Research in Personnel and Human Resources Management, Volume 23, First Edition (Research in Personnel and Human Resources Management)

CONTENTS OVERVIEW Joseph J. Martocchio

vii

FAIRNESS IN IDIOSYNCRATIC WORK ARRANGEMENTS: JUSTICE AS AN I-DEAL Jerald Greenberg, Marie-Élène Roberge, Violet T. Ho and Denise M. Rousseau SEXUAL ORIENTATION IN THE WORKPLACE: THE UNIQUE WORK AND CAREER EXPERIENCES OF GAY, LESBIAN AND BISEXUAL WORKERS Belle Rose Ragins

1

35

ORGANIZATIONAL STAFFING: A MULTILEVEL REVIEW, SYNTHESIS, AND MODEL Robert E. Ployhart

121

THE FOUR-FIFTHS RULE FOR ASSESSING ADVERSE IMPACT: AN ARITHMETIC, INTUITIVE, AND LOGICAL ANALYSIS OF THE RULE AND IMPLICATIONS FOR FUTURE RESEARCH AND PRACTICE Philip Bobko and Philip L. Roth

177

GOING PLACES: ROADS MORE AND LESS TRAVELLED IN RESEARCH ON EXPATRIATE EXPERIENCES David A. Harrison, Margaret A. Shaffer and Purnima Bhaskar-Shrinivas

199

INFERENCE PROBLEMS WITH HIERARCHICAL MULTIPLE REGRESSION-BASED TESTS OF MEDIATING EFFECTS Eugene F. Stone-Romero and Patrick J. Rosopa

249

v

vi

THE EMPLOYEE-ORGANIZATION RELATIONSHIP: A TIMELY CONCEPT IN A PERIOD OF TRANSITION Lynn M. Shore, Lois E. Tetrick, M. Susan Taylor, Jaqueline A.-M. Coyle Shapiro, Robert C. Liden, Judi McLean Parks, Elizabeth Wolfe Morrison, Lyman W. Porter, Sandra L. Robinson, Mark V. Roehling, Denise M. Rousseau, René Schalk, Anne S. Tsui and Linn Van Dyne

291

ABOUT THE AUTHORS

371

OVERVIEW I am pleased to present Volume 23 of Research in Personnel and Human Resources Management that contains seven interesting papers. These papers are interesting because each one establishes a strong case for bringing somewhat understudied topics into the research mainstream of the human resources management field, brings to our attention the need to critically examine conventional, conceptual and methodological approaches used in HRM research, or builds creative paths in our thinking about commonly studied topics. Some of you may find one or more papers to be controversial in nature, and this is by design. I encourage authors to challenge convention in order to stimulate healthy debates and discussions between HRM researchers. Greenberg, Roberge, Ho and Rousseau indicate that contemporary employers and employees voluntarily are entering into highly customized agreements regarding nonstandard employment terms. They refer to such idiosyncratic deals as “i-deals,” acknowledging that these arrangements are intended to benefit all parties. Examples of i-deals include an employee with highly coveted skills who is compensated more generously than other employees doing comparable work, and an employee who is granted atypically flexible working hours to accommodate certain personal life demands. The nonstandard nature of i-deals is likely to prompt questions about the fairness of the arrangement among three principal stakeholders – employees who receive the i-deal, managers with whom the i-deal is negotiated, and the coworkers of these employees and managers. The authors analyze issues of fairness that arise in the relationships between all three pairings of these stakeholders through the lenses of four established forms of organizational justice – distributive justice, procedural justice, interpersonal justice, and informational justice. Their discussion sheds light on previously unexplored nuances of i-deals and identifies several neglected theoretical issues of organizational justice. In addition to highlighting these conceptual advances, they also discuss methods by which the fairness of i-deals can be promoted. Ragins establishes that lesbian, gay and bisexual (LGB) employees constitute one of the largest, but least studied, minority groups in the workforce. Next, she examines what we know, and what we need to know, about the career and workplace experiences of this understudied population. The construct of sexual identity is defined, followed by a review of the research on sexual orientation vii

viii

OVERVIEW

in the workplace. Then, Ragins delivers a narrative analysis of the differences between LGB employees and other stigmatized groups. Three unique challenges facing LGB employees are identified, and conceptual models are developed that explain underlying processes. Finally, Ragins critically analyzes career theories, and she presents an identity-based longitudinal theory of LGB careers. Ployhart poses the question: Does staffing contribute to organizational effectiveness and sustained competitive advantage, or are many of staffing’s implications merely cross-level fallacies? This paper provides a critical examination of staffing research and practice, and proposes a multilevel model of staffing that ties together micro (e.g. personnel selection), meso (e.g. team staffing), and macro (e.g. organizational strategy, Human Resources practices) theory, research, and practice. The model is both integrative and prescriptive, providing a basic organizing structure for examining staffing research within and across levels. Ployhart begins with a review of multilevel theory, followed by a review and critique of the dominant staffing paradigms from a multilevel perspective. He shows that single level paradigms cannot answer many of the primary questions of interest to staffing specialists. In contrast, Ployhart maintains that a multilevel staffing model not only addresses these limitations, but also prompts a variety of new predictions that oftentimes run counter to prevailing wisdom. Evidence for adverse impact in applied, organizational settings can often depend on application of the “four-fifths rule.” Bobko and Roth analyze the arithmetic four-fifths rule, its operationalization, and related statistical tests. They note that the rule has intuitive appeal and has arithmetic directness, but also it contains many ambiguities because of the manner in which it is defined, as well as its use in practice. Bobko and Roth discuss the arithmetic and statistical facets of the definition. These authors also demonstrate where the ambiguities (and possibly unintended consequences) with the four-fifths rule might arise when numerical interpretations are invoked. Implications for future research and academic dialogues are then noted. Harrison, Shaffer and Bhaskar-Shrinivas review 25 years of research on the expatriate experience, concentrating on expatriate adjustment as a central construct, and relying on a general stressor-stress-strain framework. First, they consider who expatriates are, why their experiences differ from domestic employees, and what adjustment is. Next, conceptualizing (mal)adjustment in terms of stress, the authors review the stressors and strains associated with it. Harrison and colleagues note major patterns of effects and their implications for how human resource managers can facilitate adjustment based on the wide range of antecedents (anticipatory and in-country) that have been studied to date. Finally, to stimulate future efforts to understand the experiences of expatriates, the authors discuss the challenges and opportunities of continuing down this road of research.

Overview

ix

Stone-Romero and Rosopa point out that mediating effects are often tested using hierarchical multiple regression (HMR) procedures, and many researchers apply the very frequently cited procedure described by Baron and Kenny (1986). Stone-Romero and Rosopa explain that there are several limitations with this HMR procedure, and they demonstrate that it: (a) is of virtually no value for buttressing claims of mediating effects for data from nonexperimental research; (b) produces erroneous inferences about the existence of mediating effects for misspecified mediation models; and (c) is not able to provide credible evidence of mediating effects in a large proportion of cases, even for properly specified mediation models. The authors detail a number of important implications of their analyses. Shore and colleagues’ treatment of the employee-organization relationship (EOR) truly captures the spirit of multidisciplinary thinking. The EOR has increasingly become a focal point for researchers in the fields of organizational behavior, human resource management, and industrial relations. Literature on the EOR has developed at both the individual- (e.g. psychological contracts) and the group and organizational-levels of analysis (e.g. employment relationships). Shore and colleagues review both sets of literature, and they argue for the need to integrate these literatures as a means for improving understanding of the EOR. Mechanisms for integrating these literatures are suggested. Subsequently, they present a discussion of contextual effects on the EOR in which they encourage researchers to develop models that explicitly incorporate context. Shore and colleagues then examine a number of theoretical lenses to explain various attributes of the EOR such as the dynamism and fairness of the exchange, and new ways of understanding the exchange including positive functional relationships and integrative negotiations. Their paper concludes with a discussion of future research needed on the EOR. In summary, this set of seven papers represents an eclectic mix of perspectives on understudied and more commonly studied topics in the human resource management field. Altogether, Volume 23 reinforces the multidisciplinary character of HRM research, and reminds us of the possibilities from thinking outside the box. Joseph J. Martocchio Series Editor

FAIRNESS IN IDIOSYNCRATIC WORK ARRANGEMENTS: JUSTICE AS AN I-DEAL ´ ene Roberge, Violet T. Ho Jerald Greenberg, Marie-El` and Denise M. Rousseau ABSTRACT In response to demands and opportunities of the labor market, contemporary employers and employees voluntarily are entering into highly customized agreements regarding nonstandard employment terms. We refer to such idiosyncratic deals as “i-deals,” acknowledging that these arrangements are intended to benefit all parties. Examples of i-deals include an employee with highly coveted skills who is compensated more generously than other employees doing comparable work, and an employee who is granted atypically flexible working hours to accommodate certain personal life demands. The nonstandard nature of i-deals is likely to prompt questions about the fairness of the arrangement among three principal stakeholders – employees who receive the i-deal, managers with whom the i-deal is negotiated, and the co-workers of these employees and managers. We analyze issues of fairness that arise in the relationships among all three pairings of these stakeholders through the lenses of four established forms of organizational justice – distributive justice, procedural justice, interpersonal justice, and informational justice. Our discussion sheds light on previously unexplored nuances

Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 1–34 Copyright © 2004 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23001-8

1

2

JERALD GREENBERG ET AL.

of i-deals and identifies several neglected theoretical issues of organizational justice. In addition to highlighting these conceptual advances, we also discuss methods by which the fairness of i-deals can be promoted.

INTRODUCTION Standardization of human resource policies long has been a hallmark of managerial practice. Whether it’s identifying the compensation employees will receive for particular jobs (e.g. in the form of pay grades), specifying how job performance will be evaluated (e.g. in the form of performance appraisal guidelines), or delineating the set of tasks to be performed (e.g. as in formal job descriptions), uniform employment practices are the general rule. In fact, homogeneity of human resource practices has grown in popularity since the industrial revolution in the nineteenth century (Kornblith, 1998) and the advent of scientific management in the twentieth century (Taylor, 1911). Inspired by efforts to reduce variability in human behavior (Budd, 2004), the use of uniform human resource practices has been touted both for its contribution to efficient organizational functioning (Spender & Kijne, 1996) and for its important role in promoting interpersonal cooperation and trust (Lazear, 1981). In recent years, however, standardization has been eroded by two prominent forces in the workplace that have made non-standardization the new standard of human resource management. First, as the competition to hire, motivate, and retain the best workers has intensified, many organizations have emulated the tactics used to secure the best athletes to sports teams and actors to motion pictures – treating the most highly desired workers (whether job candidates or job incumbents) like “stars” (Lee, MacDermid & Buck, 2000). Such exceptional individuals deserve not merely standard treatment, but exceptional treatment, which, by definition, precludes standardization – the mantra being, “only standard people get standard deals.” A second force undermining the hegemony of standardization derives from heightened expectations for voice among workers (Budd, 2004). As teamwork and collaborative work arrangements have become commonplace, social norms have developed that lead employees to become more outspoken about work arrangements that satisfy their individual needs and desires (Freeman & Rogers, 1999). This is especially so with respect to policies that make it possible to pursue career and personal life opportunities simultaneously (Grover & Crocker, 1995). In this regard, workers often are guided by readily-available information about some employers’ willingness to make accommodations in response to labor market pressures (Cappelli, 2000). To the extent that these are embraced as new reference

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

3

standards, it is not surprising to find workers pressuring their own employers for similar arrangements on an ad hoc basis (at least until such time as they become institutionalized; Rousseau, Ho & Greenberg, in press). Taken together, these forces have contributed to the erosion of standardized human resources practices in favor of ones that are highly customized (Rousseau, 2000). This trend toward standardizing non-standard work arrangements represents a fundamental paradigm shift in contemporary employment relationships (Rousseau, in press).

The Nature of I-Deals Highly customized arrangements that allow workers and managers to negotiate idiosyncratic employment terms are growing in popularity (Rousseau, in press), now bringing them to the attention of human resource scholars (Rousseau, 2001, 2004). We have referred to such idiosyncratic deals as i-deals (Rousseau et al., in press), a term that serves both as an abbreviation and as an indication that such arrangements are intended to be “ideal” for both employers and employees. Specifically, we define i-deals as voluntary, personalized agreements of a nonstandard nature negotiated between employers and employees with respect to the terms of employment. The key features of i-deals are as follows: (1) I-deals are negotiated individually in a manner that reflects the employee’s market value as acknowledged by the employer (Bartol & Martin, 1989). (2) I-deals involve some degree of within-group heterogeneity relative to the rewards and benefits received by co-workers (Klein, Dansereau & Hall, 1994). (3) I-deals are intended simultaneously to serve the interests of both employers (who desire special contributions) and employees (who desire special considerations). (4) I-deals vary both in form (e.g. with respect to wages and perquisites) and scope (ranging from a single idiosyncratic element in an otherwise standard employment package to a completely customized package). (5) I-deals may be formed between either prospective employers and employees in an effort to create mutually desirable future employment terms (to which we refer as ex-ante i-deals), or between current employers and employees in an effort to revise existing employment terms (to which we refer as ex-post i-deals). The literature provides several good accounts of employees who, because they are considered “stars,” have been able to capitalize on their uniqueness by negotiating more desirable employment conditions than their non-star cohorts (e.g.

4

JERALD GREENBERG ET AL.

Frank & Cook, 1995; Rosen, 1981). Hochschild (1997), for example, recounts an incident in which an engineer who had especially valuable skills was considering resigning, freeing him to pursue his interest in underwater photography. Concerned about the long-term cost of his departure, the company negotiated an i-deal that permitted this valued employee to take an unprecedented educational leave for one year before returning to work. In another example, Rousseau (in press) tells of an American university professor who struck an i-deal with Fujitsu, the large Japanese electronics firm, to hire him and pay his moving expenses to Japan. This unusual arrangement was intended to benefit both the professor (by giving him valuable new experience in Japanese-style management techniques) and the company (by exposing its executives to this scholar’s expertise). It is important to underscore the win-win nature of these examples for the employer and his or her organization. This defining feature of i-deals distinguishes them from other person-specific employment relationships in which a particular employee is given preferential treatment. For example, as we described elsewhere (Rousseau et al., in press), i-deals differ from favoritism or cronyism. In that case, an individual arrangement benefits an employee (e.g. by being given favorable work assignments) and his or her individual manager (e.g. by receiving political favors in exchange), although it may not help the organization (Pearce, Branyiczki & Bigley, 2000), or even may cause it harm (Rousseau, 2004). Because such unofficial workarounds or acts of rule-bending do not add value to a company, they cannot be considered i-deals. The same may be said of the usually covert managerial practice of condoning employee theft – albeit unofficially, of course (Tomlinson & Greenberg, in press). By allowing employees to steal, some managers are relying on an “invisible wage structure” (Ditton, 1977) through which they derive additional opportunities to influence their subordinates (Greenberg & Tomlinson, 2004). In this case, although employees may benefit (because of the extra rewards received) along with their managers (because of the extra influence opportunities they gain), their organization will suffer (because of the lost resources), making such arrangements not i-deals. The key point is that for an idiosyncratic deal to be an i-deal, it must be ideal for all stakeholders. This, of course, often requires a careful managerial balancing act, as we will document in this chapter.

Justice without Standardization Although it largely has gone unacknowledged, standardized employment practices offer a built-in system for regulating justice in the workplace. (In keeping with common practice, we will use the terms “justice” and “fairness” interchangeably

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

5

in this paper.) After all, by its very nature, standardization implies equal treatment of workers in equal conditions, and “treating equals as equals” is a definitional criterion of fairness that can be traced back to Aristotle (Ostwald, 1962). It is not surprising, therefore, that efforts to ensure that workers receive “a fair day’s wages for a fair day’s work” and that provide safeguards against corrupt practices, such as the spoils system (Foulke, 1974), have been considered key to ensuring that justice was done (Jackall, 1989). Standardized human resource practices provide some assurance of fairness, at least nominally, by building uniformity into formal procedures, a point emphasized by Taylor (1911) in articulating scientific management. However, justice is far more illusive in the case of i-deals because they specify an exception to the rule, an island of heterogeneity in a sea of homogeneity. Add to this the fact that i-deals typically are made covertly and kept private, and it becomes clear that i-deals constitute a fertile breeding ground for potential injustice. Thus far, we have established that i-deals are growing in popularity and that their very nature predisposes them to be perceived as unfair. Juxtaposing these points with widespread evidence of problems associated with perceived unfairness in organizations (e.g. lowered job performance, both role-based and extra-role; various forms of employee withdrawal, such as turnover; and various forms of deviant behavior, such as employee theft; for reviews, see Greenberg & Colquitt, in press), it becomes clear that i-deals need to be managed in such a way that they are not perceived to be unfair. Importantly, we are not advocating that inherently unfair procedures be recast so as to disguise their true form. In fact, we do not believe that i-deals are inherently unfair at all. Rather, we assert that the ad hoc nature of i-deals, along with the way they typically are created and communicated, requires conscientious justice-management efforts that generally are unnecessary whenever standardized human resource management practices are used. We will describe these techniques in the present chapter. Toward this end, we will present a conceptual framework that identifies matters of justice bearing on the principal stakeholders of i-deals. We then will discuss these issues systematically, focusing on the various justice considerations likely to be salient for each stakeholder. Finally, we close with a discussion of various key issues, implications, and practical applications stemming from our analysis.

CONCEPTUAL FRAMEWORK Given that i-deals are such a prominent feature of the contemporary organizational landscape, we believe that it is important to understand people’s reactions to i-deals and their willingness to propose them in the first place. And, because

6

JERALD GREENBERG ET AL.

i-deals represent opportunities to exchange valued resources in the workplace, people are inclined to be sensitive to the messages they send about fairness in the workplace (Greenberg, 1996; Rousseau, 2001). After all, the perceived legitimacy and appropriateness of i-deals are likely to have profound effects on organizations and their employees (Rousseau, in press; Rousseau et al., in press). With this in mind, we will analyze i-deals with respect to different perspectives on organizational justice (for reviews, see Colquitt & Greenberg, 2003; Colquitt, Greenberg & Scott, in press). Each of these perspectives represents a different fairness lens through which i-deals may be understood. Specifically, our analysis of the fairness of i-deals is guided by a conceptual framework that results from combining considerations about each of four different varieties of organizational justice with each of three different dyadic relationships formed by pairings of key stakeholders in i-deals.

Stakeholders in I-Deals Although i-deals are formed between managers (M) and employees (E), these are not the only major stakeholders involved in i-deals. To fully understand the fairness of i-deals, we also must consider two additional stakeholders – the employee’s co-workers (C) and, of course, the organization (O) within which these arrangements are made. Figure 1 identifies the key dyadic relationships, E-M, E-C, and M-C, to which attention must be given in analyzing the fairness of i-deals. This illustration also provides an opportunity to differentiate graphically between ex-ante i-deals (in which case E is not a member of the organization) and ex-post i-deals (in which case E is a member of the organization) in terms of the employee’s relationship to the other stakeholders. Our analysis of the fairness of i-deals will be considered from the perspectives of each of these dyads. Before initiating our analysis, we offer three caveats. First, to simplify our presentation, we are assuming that M is the same individual to whom E reports and with whom E negotiated the i-deal. Of course, the reality of job specialization may mean that this is or is not the case in all i-deals (Rousseau, in press). Second, although O clearly is important, we will pay less attention to O in our analysis. This reflects the fact that E, M, and C already are subsumed under O (at least, potentially, in the case of E) and that the impact of i-deals on O as a whole are likely to be more distal in nature. Of course, insofar as i-deals stand to have some system-wide effects that extend beyond the immediate outcomes realized by E, M, and C, we also will consider O as a stakeholder in our analysis. Third, unless otherwise noted, we will focus on ex-post i-deals to the exclusion of ex-ante i-deals. This is justifiable given that relationships between E and other stakeholders have

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

7

Fig. 1. Key Stakeholders in Ex-ante i-deals and Ex-post i-deals.

not yet been formed in the case of ex-ante ideals, making consideration of other dyads moot. At the risk of over-simplification, the major distinction is temporal in nature – justifying arrangements that already have been made in the case of ex-post i-deals and those that are being considered in the case of ex-ante i-deals.

Types of Justice Our analysis of the fairness of i-deals will be based on four different types of justice that have been identified in the literature (Greenberg, 1993b) and whose

8

JERALD GREENBERG ET AL.

independent status has been verified empirically (Colquitt, 2001). These are as follows: distributive justice, procedural justice, interpersonal justice, and informational justice. Distributive Justice Distributive justice refers to perceptions of the fairness of outcomes received. This is the traditional approach to organizational justice (Greenberg, 1987) and follows from Adams’ (1965) seminal theory of inequity. Research on distributive justice has established that that people desire to maintain distributions of outcomes between themselves and comparison others that reflect their relative contributions to a joint task (Greenberg, 1982). This has been demonstrated both with respect to people’s negative responses to inequity (Greenberg, 1984) and their efforts to proactively maintain fairness in their reward distribution efforts (Leventhal, 1976). Procedural Justice Procedural justice refers to perceptions of the fairness of the procedures used to determine outcomes. This includes such criteria as granting voice in decision making (Thibaut & Walker, 1975), providing opportunities to appeal decisions, basing decisions on accurate information, and following rules in a consistent manner (Leventhal, 1980). This approach to organizational justice has received considerable attention for its implications for human resource management (Cropanzano, Rupp, Mohler & Schminke, 2001; Folger & Cropanzano, 1998). Specifically, considerable attention has been paid to employees’ perceptions of the fairness of performance appraisal procedures (e.g. Greenberg, 1986), compensation systems (e.g. Greenberg & McCarty, 1990), drug testing procedures (Cropanzano & Konovsky, 1995), and personnel selection systems (e.g. Gilliland & Hale, in press). Interpersonal Justice Interpersonal justice refers to the degree of concern and social sensitivity demonstrated about an individual. Research has shown higher perceived levels of interpersonal justice are associated with positive responses to negative outcomes (for a review, see Greenberg, 1996). For example, it has been found that when managers demonstrate high degrees of social sensitivity to the plight of their employees, this reduces the extent to which those employees steal in response to having their pay cut (Greenberg, 1990a), increases the willingness of smokers to accept a smoking ban (Greenberg, 1994), and decreases the tendency for terminated employees to bring wrongful termination lawsuits against their former employers (Lind, Greenberg, Scott & Welchans, 2000). This approach to organizational justice has received a great deal of recent conceptual attention (Bies, 2001).

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

9

Informational Justice Informational justice refers to the quality of the information used to explain how decisions are made and the thoroughness of the accounts provided about those outcomes. Specifically, a supervisor’s perceived degree of informational justice reflects the extent to which he or she is believed to communicate information candidly, in timely fashion, and a manner that is tailored to the needs of the individual (Colquitt, 2001). It also reflects the degree to which information is perceived as being explained thoroughly and reasonably (Shaw, Wild & Colquitt, 2003). Many of the same beneficial effects found to occur in response to high degrees of interpersonal justice, as noted above, also have been found to occur in response to high degrees of informational justice (Colquitt & Shaw, in press; Greenberg, 1994).

EMPLOYEE-MANAGER PERSPECTIVES ON FAIR I-DEALS Because i-deals are negotiated directly between employees and managers, these parties will have direct opportunities to formulate agreements that satisfy one another’s concerns about justice. After all, unless each party believes that an i-deal is fair, he or she will be unlikely to accept the arrangement and to continue negotiating until fairness is established. This is predicated on the assumption that the attainment of fairness is a desirable quality of the employer-employee relationship, a point that is not contended in the literature (Carr & Greenberg, in press). Still, as we now will describe, the non-standard nature of i-deals presents some special challenges in attaining justice for the parties directly involved.

Distributive Justice Among employees receiving standard employment terms, maintaining distributive justice is relatively straightforward. Within an organization, it involves providing employment conditions that match those of comparable others performing comparable jobs. Although this often is easier said than done (Ostroff & Atwater, 2003), the task is simplified by the use of within-company referents and standardized rules. Internal equity comparisons are likely to be given considerable weight in establishing fair treatment because workers are inclined to have at their disposable information about their colleagues’ outcomes (e.g. because of policies regarding open records) and inputs (based on direct observations). Moreover, the proximity of these individuals makes comparisons with them perceptually salient. A set of

10

JERALD GREENBERG ET AL.

predefined rewards and obligations is assigned each job, and maintaining justice is a matter of following established patterns with respect to distributing rewards and contributions. In the case of i-deals, at least some element of such established practices are not followed. By definition, in fact, they purposely are violated. Instead, existing referents may be characterized more accurately as starting points from which E and M negotiate the actual rewards and contributions that define the unique terms of their employment relationship. Here, claims about fairness are more likely to be framed with respect to external market conditions than internal referent standards (Rousseau et al., in press). A major impetus for an i-deal, in fact, is likely to be knowledge of external market conditions coupled, of course, with the desire to make them one’s personal reality. Consider an example. Suppose an experienced systems analyst is seeking employment in a more desirable location. She has identified a prospective new employer that can offer attractive employment terms, albeit with salaries below prevailing market levels. The firm is very interested in hiring this individual because the skills she would bring enable it to pursue new accounts. So as to keep the low salary from becoming “a deal breaker,” the candidate and an enlightened company representative negotiate an ex-ante i-deal by which she will be paid market wages. Although this will give her a higher salary than other systems analysts in the same firm, company officials may justify this practice as fair on the grounds that it allows a unique opportunity to move the company in a desired new direction. This win-win arrangement benefits both the new employee (who now works where she wants at prevailing wages) and the organization (which now has opportunities to develop new business). To E and M, it also may be considered distributively just insofar as each is convinced that it is deriving value relative to its contributions. Solely from this perspective, the arrangement is cost-neutral insofar as costs are more than adequately compensated for by the benefits stemming from the work. As we will explain later, however, unless appropriately managed, this arrangement may strain the relationships between E and C and between M and C, making it far more costly than envisioned. Now, let’s consider the corresponding example in which E already is employed by O. Suppose E is interested in remaining in the company, but learns that she is underpaid because her salary, although consistent with C, falls below prevailing market levels. To rectify the situation, E proposes an ex-post i-deal that grants her higher wages. M now has to decide how to react. Assuming that M acknowledges the validity of E’s claim about the market and cannot afford to risk E’s disaffection or departure, M will be inclined to entertain such an i-deal. Of course, for the resulting arrangement to be an i-deal instead of merely the result of succumbing to undue pressure (as in the case of “emotional blackmail”; Forward & Frasier,

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

11

1998), M must fully appreciate the options and accept an arrangement that benefits the company. As in our previous example of the ex-ante i-deal, this would be the case if M acknowledges that retaining E would make it possible for O to pursue new lines of business. In general terms, this i-deal might take the form of restructuring E’s pay and workload in a manner that enables her to make higher wages while also providing greater value to the company. Precisely how this is accomplished, however, can be tricky business. If E merely receives higher wages in exchange for doing more work, then her equity ratio (perceived outcomes to perceived inputs) remains identical, maintaining the underpayment perceptions that prompted the request for the i-deal in the first place. Presumably, the basis for her request is that external referent comparisons justify giving her outcomes that are disproportionately higher than her present inputs. Thus, the i-deal she seeks may take the form of granting a pay raise. In distributive justice terms, the basis for this request is that fairness demands paying employees “the going rate” established by the market, and that this has not been occurring. Although E may be paid more highly than a comparable C doing the same work, E’s argument in proposing the i-deal would be that this is only the case within O. If in other organizations, however, she would be more highly paid for doing the same work, then that market comparison standard is justified as a relevant basis for posing the i-deal. M’s immediate charge, therefore, is to investigate the accuracy of E’s claims about external equity and to make the adjustments required to make remuneration within O comparable to that received elsewhere. It is important to note that not all i-deals are predicated on external referent standards. Instead, many are couched in terms of individual need, such as those associated with opportunities to attend to personal life matters. For example, E may attempt to negotiate an i-deal with M that allows him unprecedented flexibility in work hours that make it possible for him to attend to the special needs of an ailing child. What makes this interesting is that in most work settings, efforts to promote distributive justice typically are framed in terms of differential rewards that are contribution-based. The equity norm of distributive justice (Deutsch, 1985) dictates that people be rewarded in proportion to their relative contributions to the work they perform (a fundamental assumption underlying equity theory; Adams, 1965). In this case, E is requesting a special consideration that necessitates a fundamental shift in the allocation rules used. As in the case of any i-deal, granting this variance requires making a case as to how the arrangement benefits both parties. In this instance, for example, arrangements may be made that require E to work less desirable extended hours, such as evenings or weekends, in exchange for the flexibility. Of course, given that the precipitating situation is temporary, so too would be the i-deal. Thus, after the child’s condition changes, E would be

12

JERALD GREENBERG ET AL.

expected to return to his previous work arrangement. The temporary nature of the i-deal would be in keeping with distributive justice insofar as it sends the message that “special deals apply only under special conditions.”

Procedural Justice An i-deal reflects a curious contradiction of procedural justice. On one hand, by negotiating an i-deal, M is granting voice to E on a matter of considerable personal interest, thereby promoting perceptions of procedural justice (Lind & Tyler, 1988). Based on correspondent inference theory (Jones & Davis, 1965), the highly unusual nature of this arrangement augments E’s willingness to attribute M’s behavior to his or her inherent fairness (Greenberg, 1990b). The very fact that M would listen to E and strike a deal that helps E through a special situation itself is likely to be taken as a clear indication of M’s commitment to fairness. On the other hand, the nonstandard nature of an i-deal itself is a violation of procedural justice insofar as it treats E inconsistently with ostensibly similar others (Leventhal, 1980). As such, E is sending the message that he or she is willing to break O’s rules. To E, who is the immediate beneficiary, this behavior is likely to be interpreted as M’s generous flexibility – at least, at first. However, with the passage of time, this same behavior may come to be interpreted differently, as E begins to question the extent to which M is committed to following O’s rules at all. In fact, unless the benefits to the organization are clear, it may be interpreted as favoritism, which is a form of bias inconsistent with procedural justice (Folger & Greenberg, 1985). This is especially likely to occur in the case of an ex-ante i-deal, where one of the first things E learns about M is that he or she is “a rule-breaker” – an attribution that although positive at first, later may come to be viewed as a source of uncertainty about his or her integrity (Trevi˜no & Weaver, 2003). Although the metaphor admittedly is imperfect, this seems akin to the reaction of individuals who subsequently marry those partners with whom they had an extra-marital affair. At the time of their affair they attributed their infidelity to the profoundly unique nature of their relationship. After marriage, however, they may look back at the same illicit behavior as a basis for raising questions about their new spouse’s capacity to demonstrate fidelity (Fincham & Bradbury, 1990). After all, the person knows first-hand that his or her spouse is capable of cheating. Just as the individual always may harbor some suspicions about his or her spouse’s fidelity in this case, so to may employees who received i-deals come to question M’s commitment to O’s rules. This may prove to be a key source of E’s uncertainty about M as a source of fair treatment (Lind & Van den Bos, 2002). Especially in view of evidence of primacy effects in perceptions of justice (e.g. Lind, 2001), questions

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

13

may be raised about the wisdom of beginning a work relationship with another by openly demonstrating one’s willingness to flaunt established organizational procedures. With this in mind, it is important to note that concerns about unfairness triggered by the use of i-deals may be avoided in organizations in which i-deals are unnecessary. This includes, for example, companies in which opportunities for flexibility already are built into existing human resource policies. This is the case, for example, at SAS, the Cary, North Carolina software firm where employees are given so much flexibility with respect to when, where, and how to work, that i-deals are all but unnecessary (Howells, 2004). Because flexibility is a standard part of the culture at SAS, the prospect of bias is eliminated, avoiding suspicions about unfairness. As a general rule, i-deals are unnecessary whenever a human resources system has built-in provisions to accommodate formally terms that otherwise would have to be created “under the table.” The act of maintaining openness about rewards provides an opportunity to enhance procedural justice (Greenberg, 1986). Therefore, managers who find themselves repeatedly creating i-deals may find themselves dealing under the table so much that they have a systemic problem (Rousseau, in press) – one that should be addressed not on a case-by-case basis by striking i-deals, but by promoting system-wide change (Rousseau et al., in press). Although an i-deal may be an effective bandage for a limited problem, too many i-deals, like too many bandages, may be an ineffective way to rectify a chronic problem in need of more systemic treatment.

Interpersonal Justice To be interpersonally just, E must perceive that M demonstrated appropriate levels of dignity and respect in the course of negotiating the i-deal. For example, M would be considered interpersonally unfair to the extent that he or she exerts inappropriately high pressure on E to accept an i-deal. Such treatment would fail to acknowledge E’s vital contributions to O, or even may be taken as an effort to undermine E by taking advantage of him or her. Acts that raise E’s suspicions about this are likely to send the message that the only thing “special” about the i-deal M is proposing is that it is really not ideal for him or her at all. As such, the arrangement would not only be considered unfair, but it may not be considered to be an i-deal at all. By contrast, proposing an i-deal provides a good opportunity for M to operationalize his or her appreciation of E’s value to the organization. In other words, E’s special contributions can be acknowledged with special treatment, a point in keeping with the group-value theory of organizational justice (Lind & Tyler, 1988).

14

JERALD GREENBERG ET AL.

According to this conceptualization, the effectiveness of voice-granting procedures is predicated on their tendency to send the message that one is accepted as a member of a desired group. In large part, this occurs insofar as such procedures demonstrate the authority figures’ respect for a subordinate’s rights and opinions (Tyler & Lind, 1992). By their very nature, i-deals do precisely this, thereby sending the message that E is valued by M. Not surprisingly, this process has resulted in positive reactions by E that are beneficial to O (for a review, see Cropanzano & Greenberg, 1997) – notably, enhanced levels of organizational commitment (McFarlin & Sweeney, 1992), reduced levels of turnover (Aquino, Griffeth, Allen & Hom, 1997), enhanced willingness to conform to organizational policies (Greenberg, 1994), and reduced tendencies to bring lawsuits (Lind et al., 2000). Interestingly, this effect is so robust that concern has been expressed about what has been dubbed “false consciousness” (MacCoun, Lind & Tyler, 1992, p. 127) or “hollow justice” (Greenberg, 1990b, p. 138) – the tendency for authority figures to give employees voice simply to enhance their appearance as fair. This type of manipulation would occur, for example, if M attempted to make an ostensible show of fairness, by offering E employment terms in the form of an i-deal despite the opportunity to satisfy E’s objectives by using already established human resource policies. Although we cannot say that this ever was done (nor might we ever be able to do so), we agree with the observation that some executives are not beyond the temptation to “use the appearance of fair procedure . . . as an inexpensive way to co-opt citizens and to distract them from outcomes that by normative criteria might be considered substantively unfair or biased” (MacCoun, 2001, p. 8644). The irony is that the inherently inexpensive – and readily controllable – nature of interpersonally fair procedures (i.e. by treating people with dignity and respect) that makes this type of fairness so easy to promote also makes it easy to manipulate by those with unprincipled motives.

Informational Justice For E and M to believe that their i-deal is fair, each must believe that the other’s claims are legitimate. So, for example, M must believe E’s need for the conditions specified and E must believe M’s capacity to create and deliver the negotiated terms. Accordingly, i-deals are predicated on each party’s trust in the other. However, to the extent that the parties openly share verifiable information bearing on their claims, then admonitions to “trust me” become less necessary as a basis for promoting informational justice. In fact, the notion that trust may serve as a proxy for valid information in making justice judgments has been well established (Lewicki, Wiethoff & Tomlinson, in press).

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

15

Tyler and Lind (1992) note that facts carry greater weight than opinions when it comes to promoting perceptions of fairness (their so-called “neutrality” criterion). Indeed, research has shown that employees perceive even negative work outcomes to be fair when clear, thorough, and believable information is shared as the basis for those outcomes (for a review, see Cropanzano & Greenberg, 1997). With this in mind, several important considerations regarding i-deals are worth noting. As we noted above, because i-deals represent departures from established policies, it is incumbent upon M to justify to E the fairness of his or her actions. This often is accomplished by sharing information to substantiate the claims upon which the i-deal is based. For example, E might share medical records justifying his or her need for the flexibility required to care for a sick child, and M might share copies of the paperwork required to notify others of the change. In so doing, each party is offering sound assurances that he or she is not taking advantage of the other. At the same time, there is also a positive by-product of doing this. The mere act of sharing information is itself useful in promoting the important impression that M has nothing to hide. Although he or she violated the rules, the reliance on sound information provides assurance that the potentially questionable action, in reality, was the result of a carefully considered judgment. It also is important to note the reputational benefit associated with promoting the fairness of one’s i-deals by making use of sound information. For example, as knowledge is shared that M treats employees fairly, and can support the fairness of his or her actions by pointing to data or some relevant information, M is likely to develop a reputation for his or her fair dealings with others (Ferris, Blass, Douglas, Kolodinsky & Treadway, 2003). And, as detailed elsewhere, these impressions of fairness can be beneficial to M in a variety of different ways (Greenberg, 1990b). In the present context, one of the benefits is that knowledge of M’s reputation for fairness might influence E’s or C’s willingness to pose ex-post i-deals. After all, people likely would be reluctant to approach M about an i-deal if he or she has a reputation for being unfair. And, to the extent that each party wins when an i-deal is created, this failure to even broach the topic represents an opportunity cost of unknown magnitude.

EMPLOYEE-CO-WORKER PERSPECTIVES ON FAIR I-DEALS Although i-deals are arranged directly between E and M, the nature of their arrangement is unlikely to remain private. In the most obvious case, because E works alongside C, the more visible aspects of the special arrangement between E and M will become known to C. For example, even without cause for close

16

JERALD GREENBERG ET AL.

scrutiny, C may observe that E gets only certain kinds of work assignments or that E works only certain hours. To the extent that such disparate treatment individuates E, questions are bound to be raised about how this came to pass, and the resulting assessments are likely to enter into judgments about fairness. As in the case of the E-M relationship, the fairness of E-C relationships also may be understood from the perspective of each of four types of organizational justice.

Distributive Justice C’s knowledge that E received an i-deal from M is bound to prompt questions about the extent to which E is over-benefited relative to C. “Is E now getting an undeservedly better deal than me?” is the question C will be inclined to ponder. In view of people’s keen sensitivity to underpayment inequity (Greenberg, 1982), C is likely to be highly attuned to this possibility. Even if E doesn’t feel overpaid relative to C (because the i-deal reflects some special characteristics that make E deserving of it), E is likely to suspect that C will be suspicious of his or her preferential treatment. Indeed, the very fact that E and C have different employment arrangements may be sufficient to trigger feelings of underpayment within C, regardless of how justifiable E may believe the arrangement is. After all, M negotiated the i-deal with E, not C, forcing C to confront this arrangement as a fait accompli. Thus, there is good reason to suspect that recipients of i-deals may feel somewhat over-benefited relative to their cohorts (at least initially, until cognitive adjustments are made; Greenberg, 1982). Even if not, they at least are likely to be suspicious that their co-workers will harbor such feelings. How might E respond to C in a manner that reflects this possible feeling of overpayment? By raising his or her work outcomes, the primary behavioral response to overpayment identified by equity theory (Adams, 1965), E risks sending the message to M that he or she feels overpaid, or at least that he or she is capable of being far more productive than has been demonstrated all along. As such, this is an undesirable option. Moreover, such a response may have minimal impact on creating distributive justice directly with C. Instead, E may elect to convince C that he or she is deserving of special treatment because he or she is an especially generous employee. One way E may do this is by demonstrating high degrees of organizational citizenship behavior toward C – that is, OCB-I (i.e. voluntary, extra-role behavior directed toward other individuals instead of toward one’s organization; Williams & Anderson, 1991). By doing such things as assisting co-workers with problems and helping them meet deadlines, for example, recipients of i-deals remediate any guilt they may feel. At the same time, they are creating cordial relations with others who might be resentful of them. Indeed, OCB-I has

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

17

been found to be effective in promoting goodwill among co-workers (Podsakoff, MacKenzie, Paine & Bachrach, 2000). The extent to which E will direct OCB-I toward C is likely to be based on the degree to which E identifies with C individually, and O as a whole within which C is involved. According to social identity theory (Tajfel & Turner, 1985), the more strongly individuals define themselves as members of a group, the more inclined they will be to invest in that group. This dynamic accounts for the considerable contributions people make to the teams in which they work (Ashforth & Mael, 1989). We believe, therefore, that E receiving an i-deal will be inclined to direct OCB-I to C whenever E and C work together as members of the same team. In so doing, E may be taking steps to cement relations with others toward whom he or she identifies and is unwilling to risk disaffection.

Procedural Justice Historically, because of its focus on people’s reactions to others’ decisions (e.g. Thibaut & Walker, 1975, concept of decision control), it is not surprising that the literature on procedural justice has been applied to relationships within which there is an acknowledged difference in formal power between parties (Folger & Greenberg, 1985; Lind & Tyler, 1988). This orientation explains the concept’s utility as a mechanism for explaining superior-subordinate relationships in hierarchical organizations (Colquitt & Greenberg, 2003) and its inattention to relationships between equal-level peers. Because the E-C relationship falls into this peer-to-peer category, therefore, it is not surprising that traditional procedural justice research has failed to provide much insight into such dyads. Cropanzano and Ambrose (2001), however, offer some insight into this dynamic in their analysis of procedural justice. Specifically, they propose that “Individuals may experience a sense of moral outrage, not because they are personally treated unfairly, but because others like them live and work under unfair procedures” (p. 140). Although evidence of this so-called “fraternal-procedural deprivation” is limited, it is clear that observers are sympathetic to violations of procedural justice suffered by those with whom they identify (e.g. Leung, Chiu & Au, 1993). Extrapolating from this, it is an intriguing, although yet untested, possibility that C will be sensitive to the fairness of the procedures used to determine E’s i-deals. This seems reasonable insofar as C also is in an exchange relationship with M, and may interpret the i-deal between E and M as an indication of M’s commitment to fair procedures. Thus, in searching for information upon which to base his or her own prospects for fair treatment from M, C is likely to look to M’s treatment of others, such as E. Given that their unique nature makes it possible to attribute

18

JERALD GREENBERG ET AL.

the fairness of i-deals internally to M instead of externally, to the system within which E and M are operating (Jones & Davis, 1965), it is reasonable for C to seek information bearing on such attributions. In this sense, C’s sensitivity to his or her beliefs about M’s treatment of E, as reflected by the apparent fairness of the i-deals between them, may be taken as a barometer of M’s commitment to implementing fair procedures. And, to the extent that C’s outcomes are closely tied to E’s outcomes – as in work teams, where outcomes are interdependent – it is reasonable to expect that C will be carefully attuned to this information.

Interpersonal Justice The nature of some i-deals is such that E is permitted to do less work or different work than typically performed. For example, consider our earlier example of an ex-post i-deal in which E is granted permission to leave the office earlier than C on a regular basis so that she can take care of her ailing child. Because E is required to make up these hours by working weekends, C benefits by not having to work as many of those undesirable shifts. At the same time, however, the arrangement also may be quite disruptive to C, who is required to pick up the slack caused by E’s absence (Rousseau et al., in press). Assuming that the missed hours cause a temporary understaffing condition, the pace of work for C is likely to be more hectic as a result (Dietzel & Coursey, 1998). Thus, the i-deal, although potentially acceptable to C over the long run, may come at an immediate cost. For this reason, C may harbor some resentment toward E for having caused this undesirable – and unfair – state of affairs. Assuming C is not reluctant to share his or her concerns about the resulting overwork, and that E is not insensitive to them, it becomes incumbent upon E to manage his or relationship with C accordingly. A key way E may do is by demonstrating appropriate interpersonal sensitivity toward C for having to bear the costs of the i-deal. At the very least, E must make a conscientious effort to openly acknowledge C’s extra workload and to thank C for taking it on. By demonstrating appreciation in this manner, E is taking an important step toward getting C to accept the situation as fair, thereby enhancing his or her willingness to bear this burden (O’Malley & Greenberg, 1983). In addition to merely explaining the situation to C in a highly sensitive manner, it may be a useful for E to engage in activities that behaviorally reinforce his or her appreciation for C. Earlier, we mentioned that E may engage in OCB-I in an effort to assuage his or her guilt about having inconvenienced C. Here, we are advocating that E engage in OCB-I intentionally, in an instrumental effort to reinforce his or her appreciation for C’s efforts. For example, staying at the office later than usual

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

19

to help a colleague in need of assistance, although a small gesture, may be a useful way for E to symbolize his or her appreciation for C. Although the OCB-I itself may be small, it may speak volumes about the actor’s positive intentions. In fact, the symbolic benefit of these actions may speak at least as loudly as the assistance provided by the words themselves.

Informational Justice Beyond promoting interpersonal justice by demonstrating social sensitivity for aspects of C’s outcomes created by E’s i-deal with M, it also is important for E to promote informational justice by thoroughly explaining to C the basis for the i-deal and to justify precisely why it is warranted. Thus, E should be forthcoming to C about all aspects of the deal itself, such as how it was made and the exact terms of the arrangement. Moreover, E should share complete information bearing on these points. To enhance informational justice, C must fully comprehend and accept as legitimate E’s need for an i-deal. This requires that E make an acceptable case to C about why the arrangement is warranted and deserved. In this regard, the sharing of valid, verifiable information is key. This is especially important insofar as C may be inclined to challenge E’s claims on the grounds that they are self-serving, but will find it harder to reject claims based on objective information (Greenberg, 1993a). In fact, in the absence of valid information, C may wonder if E has something to hide, leading him or her to suspect that E has been treated with favoritism by M. The failure to explain thoroughly one’s needs for an i-deal to one’s colleagues, especially when they are likely to be affected by the arrangement, at best may be taken as a sign of E’s callousness. At worst, it may be taken as a signal that he or she is “above the rules,” which is a sure way to risk disaffection from one’s colleagues. Thus, to help shape C’s perceptions of fairness, E should make C an ally by explaining in clear and thorough terms the nature of the situation that led to need for the i-deal (and, as noted above, do so in a way that demonstrates his or her sensitivity to the costs to be borne by C). At the same time, E has a delicate balancing act to perfect. While being completely forthcoming and thorough about the nature of his or her i-deal, E runs the risk of being perceived as bragging about his or her special qualities and gloating about the unique nature of his or her arrangement with M. Thus, by promoting informational justice E has to guard against violating interpersonal justice. Indeed, this is precisely what would happen if E’s motives for imparting information are believed to be self-serving instead of other-directed. Clearly, the desirable objective may be referred to as “informative but modest,” which may be

20

JERALD GREENBERG ET AL.

difficult to achieve in practice (Shlomo & Weiner, 2000). Unfortunately, because many of the investigations that have examined both interpersonal justice and informational justice in tandem have been experimental studies in which these forms of justice have been manipulated rather than measured, existing research provides no immediate guidance with respect to how this may be accomplished.

MANAGER-CO-WORKER PERSPECTIVES ON FAIR I-DEALS Although managers negotiate i-deals with employees that the two of them believe are fair, managers also have to justify the fairness of these arrangements to other employees in the work unit. There are both instrumental and symbolic reasons for this. Specifically, the i-deal is likely to have effects on the nature and degree of resources available for M to allocate to C. Moreover, C’s knowledge that M agreed to the arrangement sends a message to C about what is valued by O. This is reasonable insofar as M is acknowledged to be a legitimate agent of O and the values of M and O are believed to covary (Brockner & Siegel, 1996). As we now will illustrate, these underlying considerations are likely to have implications for managing different types of organizational justice in the M-C dyad.

Distributive Justice In discussing the distributive justice of i-deals in the E-C dyad we acknowledged the possibility that E may feel over-benefited relative to C and that he or she may engage in OCB-I in an effort to assuage resulting feelings of guilt. Given that both E and M are direct parties to the i-deal and that C is a third-party observer, it is tempting to consider that the dynamic between M and C may parallel that between E and C. In other words, to the extent that M feels guilty for arranging the i-deal, he or she may take steps to make amends toward C. However, given M’s higher position in O’s status hierarchy, this dynamic is likely to differ. Because of his status liability (Wiggins, Dill & Schwartz, 1965), M is likely to be unwilling to acknowledge any overly favorable treatment to E. To the extent that M has internalized the arguments underlying the i-deal and is convinced that it is beneficial to everyone (including C), then he or she is unlikely even to recognize, let alone admit, that he or she has created inequitable conditions. Thus, whereas E may come to think that he or she “got a pretty sweet deal” (over a period of time, at least), and feel guilty as a result, M, in his or role as maintainer of system-wide justice will be disinclined to perceive things similarly. Rather, M is likely to be convinced

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

21

that “everyone made out well,” which, of course, is precisely what justifies the i-deal in the first place. Thus, guilt associated with having caused overpayment inequity for E is unlikely to be felt, and if it is, then any signs are likely to be kept hidden because of their possible threats to one’s self-image and public image (Greenberg, 1990b). After all, M’s position in the status hierarchy requires him or her continually to justify as fair the nonstandard arrangement to anyone who might challenge its propriety on the grounds that it is favoritism. Thus, we believe that E may be more likely to acknowledge that he or she has benefited by an i-deal than C will be to acknowledge that he or she has permitted such a deal to be made. To some extent, the degree to which M will feel guilty of having over-benefited E is likely to depend on the extent to which C blames M for being disadvantaged by E’s arrangement with M. After all, in a zero-sum world in which resources are scarce, as typically characterizes organizations, C is likely to be sensitive to the resources received from M in large part because they have negative implications for oneself (Greenberg, 1981). This is a key point insofar as conceptualizations of fairness based on distributive justice (e.g. equity theory; Adams, 1965) focus almost exclusively on the dyad members involved directly in an exchange without consideration of third parties. However, third parties (C, in the case of i-deals) may play a key role insofar as they offer information about potentially unintended effects of the principals’ own exchanges with one another (E and M, in the case of i-deals). For example, C might complain to M that because of the i-deal he or she is now forced to work much harder than before. Not only might C share such information about the resulting inequity, but C also may be inclined to voice dissatisfaction as a result. This may take the form of C bringing pressure to bear on M for changing the arrangement with E. Alternatively, C might request some form of compensation for the inconvenience. Even if such dissatisfaction is not expressed orally, the underlying negative feelings are likely to be expressed in other potentially costly behavioral ways, such as by sabotaging work (Ambrose, Seabright & Schminke, 2002), by taking unexcused absences (De Boer, Bakker, Syroit & Schaufeli, 2002), or by resigning (Griffeth & Stefan, 2001).

Procedural Justice Just as an i-deal between M and E promotes E’s perceptions of procedural justice by giving E a voice in decision-making, an i-deal between M and C also may promote C’s perceptions of procedural justice because it would grant him or her voice. This may take the form of soliciting C’s opinion about the arrangement with E, thereby providing a conduit through which C openly may express his or her opinions about the i-deal.

22

JERALD GREENBERG ET AL.

In the M-C dyad, procedural justice stands to be promoted not only because of the granting of voice associated with fulfillment of an i-deal itself, but also because of M’s underlying consistency of treatment between E and C. Specifically, given that M made an i-deal with E, procedural justice with C may be promoted by giving C an opportunity to formulate a similar arrangement – either now, or whenever it is needed. Such assurances satisfy the underlying notion of consistency of treatment, which is one of the major determinants of procedural justice (Leventhal, 1980). After all, M may be able to deflect any charges of unfairness by offering comparable arrangement to comparable others. As obvious as this point may be, in actual practice it may prove challenging for M to satisfy C’s concerns about procedural justice by demonstrating consistency. The primary reason for this is that the inherently individual nature of i-deals makes it difficult for consistency to be defined in operational terms. After all, it is E’s unique needs or talents that prompt the i-deal, making the inherently ambiguous matter of comparability difficult to assess. In contrast to the relatively straightforward task of assessing consistency of treatment with respect to the administration of standard work arrangements, nonstandard arrangements are not directly comparable by definition. Accordingly, it is not so much consistency of treatment that matters in promoting the fairness of i-deals across different members of a work unit as it is consistency of opportunity. In other words, if E receives an i-deal from M, then applying the consistency criterion of procedural justice, C must have an equal opportunity to have a comparable i-deal. By its nature, this notion of opportunity, refers to future potential for fulfillment, which, by definition, is predicated on trust (Lewicki et al., in press). In this case, that trust involves C’s belief that M will be ready, willing, and able to negotiate a comparable i-deal should it become warranted. This is likely to be predicated on C’s belief that M is inherently fair as an individual (i.e. perceived consistency in underlying values) and that when an i-deal is called for, conditions will make it possible for M to negotiate a similar arrangement (i.e. perceived consistency of opportunities). For example, reductions in M’s discretionary power (which even may be the result of widespread negative reactions to M’s i-deal with E) may limit M’s future capacity to create i-deals. So too may changes in the available level of resources threaten the potential for comparable i-deals to be granted in the future. Additionally, not only may C have reason to be concerned about M’s future power in the organization, but more fundamentally, whether M even will be employed by the company in the months or years to come. This is a valid concern insofar as the unofficial and nonstandard nature of i-deals makes them dependent on the particular individuals involved. So, for example, should M leave and be replaced by another role incumbent, C will have no legitimate claim for a future i-deal. Generally, i-deals are neither inheritable

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

23

nor fungible. Thus, unless special agreement has been reached, arguments about i-deals with a departed manager will have no standing whatsoever with a new one. By the same token, the continuity of E’s i-deal also is in jeopardy. Depending on the nature of the conditions under which M was replaced, it may be likely that the new M will terminate E’s i-deal in an effort to “clean house.” Although the state of present research on i-deals requires us to label these suggestions as speculatory, we can be confident about a key aspect of the underlying dynamics. Specifically, the unofficial nature of i-deals makes them tenuous both for future recipients and those who hope to receive them in the future. This uncertainty is likely to affect perceptions of justice in ways that we are now just beginning to understand (Lind & Van den Bos, 2002).

Interpersonal Justice Earlier, we noted that for E to enhance C’s perceptions of interpersonal justice, E must acknowledge and express sincere appreciation to C for any costs borne by his or her i-deal with M. Likewise, for M to enhance C’s perceptions of interpersonal justice, M also must acknowledge and express sincere appreciation to C for any costs borne by his or her i-deals with E. Such ameliorative expressions are likely to be even more salient to C when they come from M than when they come from E. This is in keeping with the notion that M has greater power than E given that although either may propose an i-deal, M’s approval is required for that arrangement to be realized. Thus, to the extent that it is M and not E who is seen as ultimately responsible for the i-deal, it is M whose inherent fairness is reflected by the relationship – and therefore, who is obligated to demonstrate sensitivity to any harm done by the arrangement. A related consideration is that M is likely to be the individual from whom C expects to be receiving valued resources in the future. Thus, beliefs regarding M’s fairness are likely to be salient insofar as it is likely to have implications for the fairness of future treatment toward oneself. By acknowledging M’s fairness with E, for example, C is psychologically reassured of M’s capacity to demonstrate fairness toward himself or herself in the future (Lind, 2001). Extending this line of thought, we must note the strong message of social acceptance that also is communicated by virtue of expressions of dignity and respect (Tyler & Lind, 1992). We already made this point in conjunction with our discussion of efforts to promote interpersonal justice in the E-M dyad. Although group value theory (Lind & Tyler, 1988) addresses only the relationships between the parties directly involved in a social exchange, it is an intriguing possibility that C’s beliefs regarding M’s interpersonal justice toward E will contribute to C’s

24

JERALD GREENBERG ET AL.

impressions of M’s interpersonal justice toward himself or herself. Essentially, this would be a matter of indirect, or vicarious attribution. Thus, instead of E forming judgments about M’s fairness based on M’s treatment of E, C may form judgments of M based on his or her knowledge of M’s treatment of E. To the extent that this occurs, such as through judicious efforts to publicize the fairness of one’s actions, then M may be deriving impression management efforts that extend beyond their originally intended targets (Greenberg, 1990b). By contrast, consider what would occur if M were to be insensitive to C by failing to acknowledge – or, even flatly denying – that his or her special arrangement with E had any adverse effects on C. Clearly, adding the insult of insensitivity to the injury of overwork would risk further eroding an M-C relationship that already may be strained by C’s feelings of injustice. And, given that studies repeatedly demonstrate intense negative reactions to socially insensitive reactions to injustice (Greenberg, 1990a, 1993a, 1994; Lind et al., 2000), it is clearly unwise for managers to be insensitive to these concerns. Given that the nonstandard nature of i-deals is likely to make co-workers hyper-salient to matters of justice, managers would be well-advised not overlook the fairness of the impressions created by their i-deals. Compounded by the fact that there are likely to be multiple C’s who make social comparisons about matters of justice (Cropanzano & Ambrose, 2001, especially pp. 136–138), the potential for negative impact is exacerbated, thereby underscoring the wisdom of this advice.

Informational Justice Despite the oft-heard phrase, “I don’t need to explain myself,” when it comes to i-deals, M’s act of explaining himself or herself is likely to go a long way toward enhancing C’s perceptions of M’s fairness. In his or her dealings with C, M may be likely to reinforce images of himself or herself as being fair-minded by presenting to C arguments bearing on his or her reasons for the arrangement with E (Greenberg, 1990b). For these to come across as truly fair as opposed to efforts to rationalize or cover-up an ostensibly unfair decision, it is useful for M to couch his or her decisions in terms of relevant information that justifies it. Clearly, one of the contributing factors is the story told by the information itself. That is, the facts are likely to justify the decision – at least, when put into proper perspective by M. There also is reason to believe that another contributing factor is simply the degree to which M agreed to share the information in advance of being asked to do so. Being forthcoming by volunteering information is likely to be perceived as an indication that one has nothing to hide, enhancing the possibility that one is being fair (Greenberg, 1990b; Shaw et al., 2003). On this

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

25

basis, we believe that is not just the information itself, but the voluntary sharing of explanation by M that sends the message of his or her fairness to C. In this regard, “timing is everything,” as the saying goes. It is not just the information that matters, nor is it just the sharing of the information that matters. Rather, it is the sharing of information before being asked to do so that matters. After all, this timing reflects the possibility that the act is motivated by a genuine interest in being informative. This idea is predicated on research showing that people are more inclined to make internal attributions of one’s fairness when that behavior is volitional (allowing it to be attributed internally to the source) than when it is mandated by policy (in which case, it is likely to be attributed externally) (Greenberg, 2003). Following from existing research (e.g. Shaw et al., 2003) we earlier noted that E’s perceptions of M’s fairness is likely to be enhanced by E’s knowledge of the information upon which M based his or her decision about the i-deal. In addition, we believe that C’s impression of M’s fairness also is likely to be enhanced by M’s sharing of the same information with C, despite C’s third-party status. By “opening the book” in this manner through voluntarily sharing this information, M sends two important messages that promote his or images as fair. For one, it sends the message that although the arrangement is nonstandard, M “has nothing to hide.” Secondly, it suggests that M was serious about his or her promise to offer a comparable i-deal should comparable conditions exist in the future. After all, sharing the facts upon which the i-deal is based may be tantamount to inviting C to request a similar i-deal with M whenever appropriate conditions prevail.

JUSTICE AS AN I-DEAL: ISSUES, IMPLICATIONS, AND APPLICATIONS Our analysis has raised several fundamental issues about both organizational justice and the nature of i-deals that stand to be identified and that warrant further discussion. Likewise, several of the points we raised have provocative implications for understanding the nature of justice in organizations and for the practical management of i-deals. We now turn to these matters. Static vs. Dynamic Time Perspective In keeping with most prevailing approaches to organizational justice, our discussion of i-deals has been predicated on a static temporal perspective. However, it is important to caution that because fairness is perceptual in nature, what people judge to be fair at one point may change over time with the accumulation of additional

26

JERALD GREENBERG ET AL.

experiences. Although this possibility already has been acknowledged for over two decades (Cosier & Dalton, 1983), the notion that perceptions of justice are dynamic largely has been ignored. The phenomenon of i-deals, however, calls our attention to the importance of shifting justice perceptions given that people receive i-deals at one point in time and live with them and their effects over extended periods. Conceptual support for this possibility is provided by Lind’s (2001) fairness heuristic theory. Specifically, Lind claims that as people enter into new relationships with other individuals or organizations, they make quick judgments (“heuristics”) about what is fair based on the first information available to them and that this information has the greatest influence on their feelings about fairness. In other words, Lind (2001) claims that there is a primacy effect in judgments of justice. Supporting this, Lind, Kray and Thompson (2001) report experimental evidence showing that workers who experience unfair treatment from their supervisors early in their dealings with them perceive those supervisors to be less fair than those who receive the same amount of unfair treatment from their supervisors after knowing them even as little as 1 hour longer. On the basis of these and similar findings, Lind (2001) claims that By far the best time to make a worker a willing and cooperative member of an organization is at the beginning of the worker’s relationship with the organization. Manifestly fair treatment early on will pay great dividends later in terms of support, sacrifice, and acceptance of the organization and its authorities (p. 73).

The implications of this are clear: In the course of proposing ex-ante i-deals to prospective employees, managers may be sending strong messages about their fairness insofar as they are beginning their relationship with them by granting high levels of voice. Selectivity in the Use of I-Deals This point about using i-deals as a means for promoting the voice-granting benefits of procedural justice prompts a question. If granting a small i-deal, or just a few i-deals is good, is it better to grant more generous i-deals, or to grant i-deals more frequently? Evidence suggests that the answer is “no” and that there may be a limit to the so-called voice effect (Greenberg, 2000). In keeping with previous research on the justice-enhancing effects of granting voice, Peterson (1999) found that people are more satisfied with leaders who give them a voice in decision-making than those who do not. Importantly, however, he also found that people were dissatisfied with those same leaders when the magnitude of the voice they granted was inappropriately high. Under these circumstances, leaders were believed to

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

27

be violating appropriate procedures by relinquishing too much of their decisionmaking power. Satisfaction with leaders (which tends to be correlated positively with the perceived fairness of leaders; Colquitt, 2001) had an inverted-U-shaped relationship with the magnitude of voice. As they apply to the management of i-deals, these findings dictate caution. One cannot assume that the granting of an i-deal necessarily promotes perceptions of fairness. Whereas an i-deal whose basis clearly is warranted may be a path to promoting perceptions of fairness, the granting of an i-deal that either is unwarranted or inappropriate may backfire. Indeed, the message sent may be that M is showing callous disregard for the rules – thereby weakening, rather than strengthening, the degree to which he or she is perceived as fair. This notion underscores a point to which we alluded earlier about the management of i-deals being a careful balancing act. Using i-deals appropriately may not only enhance perceptions of justice, but also may be an effective means of tapping the potential of human resources. However, misusing i-deals may be courting problems. The Importance of Sincerity and Trust To a great extent, and in many ways identified throughout this paper, the appropriate use of i-deals is predicated on sincerity and trust. Thus, for i-deals to work as intended, M must be perceived to be sincere about his or her intentions regarding the offer. In other words, M must be trusted to deliver as promised (Brockner & Siegel, 1996; Lewicki et al., in press). After all, given that the arrangement is ad hoc in nature, E has no formal recourse if M fails to fulfill his or her obligations. Moreover, the informal nature of i-deals may require that at least some aspects of them remain unwritten, residing only in the minds of E and M. As such, their success depends on the strength the psychological contract between them (Rousseau, 1995, 2000). Research bearing on sincerity primarily has focused on the negative effects of insincerity on acceptance of negative organizational outcomes. A colleague’s i-deal that causes extra work for oneself, for example, would constitute such a negative outcome. Although we are aware of no study that directly examined i-deals in this regard, it is reasonable to extrapolate from existing research studying sincerity of explanations in other contexts. For example, Shapiro, Buttner and Barry (1994) found that for explanations to be accepted as fair, they not only have to be accurate and thorough, but they also must be perceived as genuine and sincere in intent (Shapiro, Buttner & Barry, 1994). So, for example, should C perceive that the explanations he or she is receiving about E’s i-deals are insincere – or even manipulative – the benefits of providing explanations are likely to be negated. Worse yet, to the extent that M’s manipulative intent is likely to generalize by sending a message about O (as has been suggested, Brockner & Siegel, 1996),

28

JERALD GREENBERG ET AL.

this action may run the risk of poisoning the perceived fairness of the entire organization. As Greenberg (2000) put it, “Unless employees are convinced that their superiors are ‘being straight’ with them, the use of explanations is likely to backfire” (p. 187). Based on this, our advice to managers is straightforward: Efforts to manipulate workers by being insincere about i-deals should be avoided at all costs.

The Role of Openness In this connection, openness is an important consideration. Indeed, throughout our discussion, we have emphasized the importance of openness as a means of promoting perceptions of fair i-deals. Specifically, we advised M to share the basis for his or her i-deal with both E and C, and for E also to be candid with C about his or her i-deal with M. In both cases, our rationale was that sharing this information reduces concerns that the i-deals were made “under the table” because they somehow were improper and in need of hiding. Rather, the message should be sent that the arrangement is being made “under the table” simply because no formal opportunities for comparable work arrangements to be made “on the table.” This is in keeping with an important point made by Salam (2000) regarding the role of visible behavior by leaders in the creation of trust: Leaders can only create trust in subordinates if they ensure that the subordinates have knowledge and information about their behavior. Hence, ensuring the visibility of one’s behavior is absolutely necessary for the creation of trust (p. 277).

Although this statement speaks to openness by leaders (M, in present terms), the underlying point applies as well to openness between peers. To illustrate this, consider one particular way of promoting openness in the E-C dyad as it might be applied to our example of the employee who in need of special flexibility to accommodate a personal situation. In the typical scenario, E would discuss the matter with M and propose an ex-post i-deal. After the deal is struck, the arrangement then would be presented to C as a fait accompli. As an alternative, we advocate that E assemble all C’s and candidly share his or her situation with them. At such a meeting E should explain thoroughly the nature of the arrangement sought with M before proposing it to M. Making the communication two-way, E also may solicit suggestions from C as to how to make the situation less taxing on them. This approach would be beneficial for at least three important reasons. First, and most obviously, such meetings may prove to be useful sources of information to C about either alternative avenues that may be pursued to address his or her situation or ways of minimizing disruption caused by the i-deal. Second, getting C’s “buy-in” would help E gain the acceptance of C who is being asked to bear

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

29

some of the costs of E’s i-deal. Not only does this give C useful information required to understand it (Wagner, 2000), but it also promotes their psychological acceptance of it (Latham, Winters & Locke, 1994). Third, assuming that such meetings effectively result in C’s acceptance, sharing information about C’s “buy in” with M may pave the way for E’s proposed arrangement to be accepted by M, who may have concerns about C’s reactions. After all, by not having to “spring” the arrangement on C, E is spared from having to perform a potentially awkward task that may threaten the relationship between themselves and C.

CONCLUSION As a growing reality of the contemporary workplace, managers, workers, and co-workers are likely to find themselves involved in i-deals – either directly, as participants, or indirectly, as third-parties affected by them. In fact, given that individuals generally play multiple roles in organizations, we may expect people to be both direct parties to some i-deals and third-parties to others. The resulting network of arrangements and reactions to them may have profound effects on organizational behavior. Given this complexity, it is clear that understanding the nature of i-deals is essential to managing them most effectively. As we have detailed here, research and theory on organizational justice offer useful conceptual mechanisms for furthering this understanding. Our discussion highlights the ways in which people’s reactions to i-deals may be interpreted in light of notions of organizational justice. Additionally, we have identified considerations of justice that are likely to enter into people’s considerations about how to form i-deals, and even whether i-deals should be created in the first place. However, because our knowledge about i-deals as a phenomenon is brand new (Rousseau, in press; Rousseau et al., in press) and our conceptual understanding of justice in organization is still growing (Greenberg & Colquitt, in press), our analysis admittedly is incomplete. We surely will have raised more questions than we answered. Instead of apologizing, we point to the issues we may have unearthed as the basis for challenging our colleagues to provide insight into them in their own research. If our remarks prove to inspire any future efforts to study i-deals, we indeed will consider them to have been ideal.

REFERENCES Adams, J. S. (1965). Inequity in social exchange. In: L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 2, pp. 267–299). New York: Academic Press.

30

JERALD GREENBERG ET AL.

Ambrose, M. L., Seabright, M. A., & Schminke, M. (2002). Sabotage in the workplace: The role of organizational injustice. Organizational Behavior and Human Decision Processes, 89, 947– 965. Aquino, K., Griffeth, R. W., Allen, D. G., & Hom, P. W. (1997). Integrating justice constructs into the turnover process: A test of a referent cognitions model. Academy of Management Journal, 40, 1208–1227. Ashforth, B. E., & Mael, F. (1989). Social identity theory and the organization. Academy of Management Review, 14, 20–39. Bartol, K. M., & Martin, D. C. (1989). Effects of dependence, dependence threats, and pay secrecy on managerial pay allocations. Journal of Applied Psychology, 74, 105–113. Bies, R. J. (2001). Interactional (in)justice: The sacred and the profane. In: J. Greenberg & R. Cropanzano (Eds), Advances in Organizational Justice (pp. 89–118). Stanford, CA: Stanford University Press. Brockner, J., & Siegel, P. (1996). Understanding the interaction between procedural and distributive justice: The role of trust. In: R. M. Kramer & T. R. Tyler (Eds), Trust in Organizations: Frontiers of Theory and Research (pp. 390–413). Thousand Oaks, CA: Sage. Budd, J. W. (2004). Employment with a human face: Balancing efficiency, equity, and voice. Ithaca, NY: ILR Press. Cappelli, P. (2000). A market-driven approach to retaining talent. Harvard Business Review, 78(1), 103–111. Carr, J. Z., & Greenberg, J. (in press). Is organizational justice self-interested? In: J. Greenberg & J. A. Colquitt (Eds), Handbook of Organizational Justice. Mahwah, NJ: Lawrence Erlbaum. Colquitt, J. A. (2001). On the dimensionality of organizational justice: A construct validation of a measure. Journal of Applied Psychology, 86, 386–400. Colquitt, J. A., & Greenberg, J. (2003). Organizational justice: A fair assessment of the state of the literature. In: J. Greenberg (Ed.), Organizational Behavior: The State of the Science (2nd ed., pp. 165–210). Mahwah, NJ: Lawrence Erlbaum. Colquitt, J. A., & Greenberg, J., & Scott, B. A. (in press). Organizational justice: Where do we stand? In: J. Greenberg & J. A. Colquitt (Eds), Handbook of Organizational Justice. Mahwah, NJ: Lawrence Erlbaum. Colquitt, J. A., & Shaw, J. (in press). How should organizational justice be measured? In: J. Greenberg & J. A. Colquitt (Eds), Handbook of Organizational Justice. Mahwah, NJ: Lawrence Erlbaum. Cosier, R. A., & Dalton, R. (1983). Equity theory and time: A reformulation. Academy of Management Review, 8, 311–319. Cropanzano, R., & Ambrose, M. L. (2001). Procedural and distributive justice are more similar than you think: A monistic perspective and a research agenda. In: J. Greenberg & R. Cropanzano (Eds), Advances in Organizational Justice (pp. 119–151). Stanford, CA: Stanford University Press. Cropanzano, R., & Greenberg, J. (1997). Progress in organizational justice: Tunneling through the maze. In: C. L. Cooper & I. T. Robertson (Eds), International Review of Industrial and Organizational Psychology (Vol. 2, pp. 317–372). New York: Wiley. Cropanzano, R., & Konovsky, M. A. (1995). Resolving the justice dilemma by improving the outcomes: The case of employee drug screening. Journal of Business and Psychology, 10, 221– 243. Cropanzano, R., Rupp, D. E., Mohler, C. J., & Schminke, M. (2001). Three roads to organizational justice. In: G. R. Ferris (Ed.), Research in Personnel and Human Resources Management (pp. 1–113). San Diego: Elsevier.

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

31

De Boer, E. M., Bakker, A. B., Syroit, J. E., & Schaufeli, W. B. (2002). Unfairness at work as a predictor of absenteeism. Journal of Organizational Behavior, 23, 181–197. Deutsch, M. (1985). Distributive justice. New Haven, CT: Yale University Press. Dietzel, L. C., & Coursey, R. D. (1998). Predictors of emotional exhaustion among nonresidential staff persons. Psychiatric Rehabilitation Journal, 21, 340–348. Ditton, J. (1977). Perks, pilferage, and the fiddle: The historical structure of invisible wages. Theory and Society, 4, 39–71. Ferris, G. R., Blass, F. R., Douglas, C., Kolodinsky, R. W., & Treadway, D. C. (2003). Personal reputation in organizations. In: J. Greenberg (Ed.), Organizational Behavior: The State of the Science (2nd ed., pp. 211–246). Mahwah, NJ: Lawrence Erlbaum. Fincham, F. D., & Bradbury, T. N. (1990). The psychology of marriage: Basic issues and applications. New York: Guilford Press. Folger, R., & Cropanzano, R. (1998). Organizational justice and human resource management. Thousand Oaks, CA: Sage. Folger, R., & Greenberg, J. (1985). Procedural justice: An interpretive analysis of personnel systems. In: K. M. Rowland & G. R. Ferris (Eds), Research in Personnel and Human Resources Management (Vol. 3, pp. 141–183). Greenwich, CT: JAI Press. Forward, S., & Frasier, D. (1998). Emotional blackmail. New York: Quill. Foulke, W. D. (1974). Fighting the spoilsmen. New York: Arno Press. Frank, R. H., & Cook, P. J. (1995). Winner take all: How more and more Americans compete for fewer and bigger prizes, encouraging economic waste, income inequality, and an impoverished cultural life. New York: Free Press. Freeman, R. B., & Rogers, J. (1999). What workers want. Ithaca, NY: Cornell University Press. Gilliland. S. W., & Hale, J. M. S. (in press). How do theories of organizational justice inform fair employee selection practices? In: J. Greenberg & J. A. Colquitt (Eds), Handbook of Organizational Justice. Mahwah, NJ: Lawrence Erlbaum. Greenberg, J. (1981). The justice of distributing scarce and abundant resources. In: M. J. Lerner & S. C. Lerner (Eds), The Justice Motive in Social Behavior (pp. 289–316). New York: Plenum Press. Greenberg, J. (1982). Approaching equity and avoiding inequity in groups and organizations. In: J. Greenberg & R. L. Cohen (Eds), Equity and Justice in Social Behavior (pp. 389–435). New York: Academic Press. Greenberg, J. (1984). On the apocryphal nature of inequity distress. In: R. Folger (Ed.), The Sense of Injustice: Social Psychological Perspectives (pp. 167–186). New York: Plenum Press. Greenberg, J. (1986). Determinants of perceived fairness of performance evaluations. Journal of Applied Psychology, 71, 340–342. Greenberg, J. (1987). A taxonomy of organizational justice theories. Academy of Management Review, 12, 9–22. Greenberg, J. (1990a). Employee theft as a reaction to underpayment inequity: The hidden cost of pay cuts. Journal of Applied Psychology, 75, 561–568. Greenberg, J. (1990b). Looking fair vs. being fair: Managing impressions of organizational justice. In: B. M. Staw & L. L. Cummings (Eds), Research in Organizational Behavior (Vol. 12, pp. 111–157). Greenwich, CT: JAI Press. Greenberg, J. (1993a). Stealing in the name of justice: Informational and interpersonal moderators of theft reactions to underpayment inequity. Organizational Behavior and Human Decision Processes, 54, 81–103.

32

JERALD GREENBERG ET AL.

Greenberg, J. (1993b). The social side of fairness: Interpersonal and informational classes of organizational justice. In: R. Cropanzano (Ed.), Justice in the Workplace: Approaching Fairness in Human Resource Management (pp. 79–103). Hillsdale, NJ: Lawrence Erlbaum. Greenberg, J. (1994). Using socially fair procedures to promote acceptance of a work site smoking ban. Journal of Applied Psychology, 79, 288–297. Greenberg, J. (1996). The quest for justice the job: Essays and experiments. Thousand Oaks, CA: Sage. Greenberg, J. (2000). Promote procedural justice to enhance acceptance of work outcomes. In: E. A. Locke (Ed.), A Handbook of Principles of Organizational Behavior (pp. 181–195). Oxford, England: Blackwell. Greenberg, J. (2003). Creating unfairness by mandating fair procedures: The hidden hazards of a pay-for-performance plan. Human Resource Management Review, 13, 41–57. Greenberg, J., & Colquitt, J. A. (in press). Handbook of organizational justice. Mahwah, NJ: Lawrence Erlbaum. Greenberg, J., & McCarty, C. (1990). Comparable worth: A matter of justice. In: G. R. Ferris & K. M. Rowland (Eds), Research in Personnel and Human Resources Management (Vol. 8, pp. 265–301). Greenwich, CT: JAI Press. Greenberg, J., & Tomlinson, E. C. (2004). The methodological evolution of employee theft research: The DATA Cycle. In: R. Griffin & A. O’Leary-Kelley (Eds), The Dark Side of Work (pp. 426–460). San Francisco: Jossey-Bass. Griffeth, R. W., & Stefan, G. (2001). A role for equity theory in the turnover process: An empirical test. Journal of Applied Social Psychology, 31, 1017–1037. Grover, S. L., & Crocker, K. J. (1995). Who appreciates family-responsiveness human resource policies? Journal of Applied Psychology, 48, 271–288. Hochschild, A. R. (1997). Time bind: When work becomes home and home becomes work. New York: Henry Holt. Howells, J. (2004, February). SAS teaches the art of empathy. IT Training, 14–17. Jackall, R. (1989). Moral mazes: The world of corporate managers. New York: Oxford University Press. Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In: L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 2, pp. 96–145). New York: Academic Press. Klein, K. J., Dansereau, F., & Hall, R. J. (1994). Levels issues in theory development, data collection. and analysis. Academy of Management Review, 19, 195–229. Kornblith, G. (1998). The industrial revolution in America. Boston: Houghton Mifflin. Latham, G. P., Winters, D. C., & Locke, E. A. (1994). Cognitive and motivational effects of participation: A mediator study. Journal of Organizational Behavior, 15, 49–63. Lazear, E. P. (1981). Agency, earnings profiles, productivity, and hours restrictions. American Economic Review, 71, 606–620. Lee, M. D., MacDermid, S. M., & Buck, M. L. (2000). Organizational paradigms of reduced-load work: Accommodation, elaboration, and transformation. Academy of Management Journal, 43, 1211–1226. Leung, K., Chiu, W. H., & Au, Y. F. (1993). Sympathy and support for industrial actions: A justice analysis. Journal of Applied Psychology, 78, 781–787. Leventhal, G. S. (1976). The distribution of rewards and resources in groups and organizations. In: L. Berkowitz & E. Walster (Eds), Advances in Experimental Social Psychology (Vol. 9, pp. 91–131). New York: Academic Press.

Fairness In Idiosyncratic Work Arrangements: Justice As An I-Deal

33

Leventhal, G. S. (1980). What should be done with equity theory? New approaches to the study of fairness in social relationships. In: K. Gergen, M. Greenberg & R. Willis (Eds), Social Exchange: Advances in Theory and Research (pp. 27–55). New York: Plenum Press. Lewicki, R. J., Wiethoff, C., & Tomlinson, E. C. (in press). What is the role of trust in organizational justice? In: J. Greenberg & J. A. Colquitt (Eds), Handbook of Organizational Justice. Mahwah, NJ: Lawrence Erlbaum. Lind, E. A. (2001). Fairness heuristic theory: Justice judgments as pivotal cognitions in organizational relations. In: J. Greenberg & R. Cropanzano (Eds), Advances in Organizational Justice (pp. 56–88). Stanford, CA: Stanford University Press. Lind, E. A., Greenberg, J., Scott, K. S., & Welchans, T. D. (2000). The winding road from employee to complainant: Situational and psychological determinants of wrongful termination claims. Administrative Science Quarterly, 45, 557–590. Lind, E. A., Kray, L., & Thompson, L. (2001). Primacy effects in justice judgments: Testing predictions from fairness heuristic theory. Organizational Behavior and Human Decision Processes, 85, 1–22. Lind, E. A., & Tyler, T. R. (1988). The social psychology of procedural justice. New York: Plenum Press. Lind, E. A., & Van den Bos, K. (2002). When fairness works: Toward a general theory of uncertainty management. In: B. M. Staw & R. M. Kramer (Eds), Research in Organizational Behavior (Vol. 24, pp. 181–223). Boston, MA: Elsevier. MacCoun, R. J. (2001). Legal issues: Public opinion. In: International Encyclopedia of the Social and Behavioral Sciences (pp. 8641–8646). San Diego, CA: Elsevier. MacCoun, R. J., Lind, E. A., & Tyler, T. (1992). Alternative dispute resolution in trial and appellate courts. In: D. K. Kanehiro & W. S. Laufer (Eds), The Handbook of Psychology and Law (pp. 111–144). New York: Springer-Verlag. McFarlin, D. B., & Sweeney, P. D. (1992). Distributive and procedural justice as predictors of satisfaction with personal and organizational outcomes. Academy of Management Journal, 35, 626–637. O’Malley, M. N., & Greenberg, J. (1983). Sex differences in restoring justice: The “down payment effect.” Journal of Research in Personality, 17, 174–185. Ostroff, C., & Atwater, L. E. (2003). Does whom you work with matter? Effects of referent group gender and age composition on managers’ compensation. Journal of Applied Psychology, 88, 725–740. Ostwald, M. (1962). Nicomachean ethics: Aristotle. Englewood Cliffs, NJ: Prentice-Hall. Pearce, J. L., Branyiczki, I., & Bigley, G. (2000). Insufficient bureaucracy: Trust and commitment in particularistic organizations. Organizational Science, 11, 148–162. Peterson, R. S. (1999). Can you have too much of a good thing? The limits of voice for improving satisfaction with leaders. Personality and Social Psychology Bulletin, 25, 313–324. Podsakoff, P. M., MacKenzie, S. B., Paine, J. B., & Bachrach, D. G. (2000). Organizational citizenship behaviors: A critical review of the theoretical and empirical literature and suggestions for future research. Journal of Management, 26, 513–563. Rosen, S. (1981). The economics of superstars. American Economic Review, 71, 845–858. Rousseau, D. M. (1995). Psychological contracts in organizations: Understanding written and unwritten agreements. Newbury Park, CA: Sage. Rousseau, D. M. (2000). Psychological contracts in the United States: Diversity, individualism, and associability. In: D. M. Rousseau & R. Schalk (Eds), Psychological Contracts in Employment: Cross-national Perspectives (pp. 1–18). Newbury Park, CA: Sage.

34

JERALD GREENBERG ET AL.

Rousseau, D. M. (2001). Idiosyncratic deals: Flexibility vs. fairness? Organizational Dynamics, 29, 260–271. Rousseau, D. M. (2004). Under the table deals: Preferential, unauthorized or idiosyncratic? In: R. Griffin & A. O’Leary-Kelley (Eds), The Dark Side of Organizations (pp. 262–290). San Francisco: Jossey-Bass. Rousseau, D. M. (in press). Idiosyncratic deals: When workers bargain for themselves. New York: M. E. Sharpe. Rousseau, D. M., Ho, V. T., & Greenberg, J. (in press). I-deals: Idiosyncratic terms in employment relationships. Academy of Management Review. Salam, S. (2000). Foster trust through competence, honesty, and integrity. In: E. A. Locke (Ed.), A Handbook of Principles of Organizational Behavior (pp. 274–288). Oxford, England: Blackwell. Shaw, J. C., Wild, E., & Colquitt, J. A. (2003). To justify or excuse? A meta-analytic review of the effects of explanations. Journal of Applied Psychology, 88, 444–458. Shlomo, H., & Weiner, B. (2000). Accounts for success as determinants of perceived arrogance and modesty. Motivation and Emotion, 24, 215–236. Spender, J. C., & Kijne, J. J. (1996). Scientific management: Frederick Wislow Taylor’s gift to the world? New York: Kluwer. Tajfel, H., & Turner, J. C. (1985). The social identity theory of intergroup behavior. In: S. Worchel & W. G. Austin (Eds), Psychology of Intergroup Relations (2nd ed., pp. 7–24). Chicago: Nelson-Hall. Taylor, F. W. (1998/1911). The principles of scientific management. New York: Dover. Thibaut, J., & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, NJ: Lawrence Erlbaum. Tomlinson, E. C., & Greenberg, J. (in press). Discouraging employee theft by managing social norms and promoting organizational justice. In: R. Kidwell & C. Martin (Eds), Managing Organizational Deviance: Perspectives and Cases. Newbury Park, CA: Sage. Trevi˜no, L. K., & Weaver, G. R. (2003). Managing ethics in business organizations. Stanford, CA: Stanford University Press. Tyler, T. R., & Lind, E. A. (1992). A relational model of authority in groups. In: M. P. Zanna (Ed.), Advances in Experimental Social Psychology (Vol. 25, pp. 115–191). San Diego, CA: Academic Press. Wagner, J. A., III (2000). Use participation to share information and distribute knowledge. In: E. A. Locke (Ed.), A Handbook of Principles of Organizational Behavior (pp. 304–315). Oxford, England: Blackwell. Wiggins, J. A., Dill, D., & Schwartz, R. D. (1965). On “status liability.” Sociometry, 28, 197–209. Williams, L. J., & Anderson, S. E. (1991). Job satisfaction and organizational commitment as predictors of organizational citizenship and in-role behaviors. Journal of Management, 17, 601–617.

SEXUAL ORIENTATION IN THE WORKPLACE: THE UNIQUE WORK AND CAREER EXPERIENCES OF GAY, LESBIAN AND BISEXUAL WORKERS Belle Rose Ragins ABSTRACT Lesbian, gay and bisexual (LGB) employees constitute one of the largest, but least studied, minority groups in the workforce. This article examines what we know, and what we need to know, about the career and workplace experiences of this understudied population. The construct of sexual identity is defined, followed by a review of the research on sexual orientation in the workplace. Then an analysis of the differences between LGB employees and other stigmatized groups is presented. Three unique challenges facing LGB employees are identified, and conceptual models are developed that explain underlying processes. Finally, career theories are critically analyzed, and an identity-based longitudinal theory of LGB careers is presented.

INTRODUCTION Gay men, lesbians and bisexuals constitute between 4 and 17% of the workforce, one of the largest but least studied minority groups in organizations (Gonsiorek & Weinrich, 1991). Although gay, lesbian and bisexual employees have been in Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 35–120 Copyright © 2004 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23002-X

35

36

BELLE ROSE RAGINS

the workforce as long, if not longer, than other groups, researchers are only now recognizing the need to understand the experiences of this “invisible group.” At the onset, it is clear that lesbian, gay, and bisexual (LGB) employees face unique challenges not encountered by other groups in the workplace. To start, discrimination against employees who are gay, or simply appear to be gay, is legal in most workplaces in the U.S. (Herrschaft & Mills, 2002). Without legal protection, LGB employees face extensive discrimination; it is estimated that between 25 and 66% of gay and lesbian employees experience discrimination in the workplace (see review by Croteau, 1996). However, this estimate is conservative, as most LGB employees do not fully reveal their sexual identity at work (Badgett, 1996, 2001; Ragins & Cornwell, 2001a). Unlike many other group memberships, sexual identity is not visible and the decision to disclose a gay identity in the workplace is an important and often stressful career decision for many LGB workers (ChrobotMason, Button & DiClementi, 2001; Ragins & Cornwell, 2001b). It is clear that LGB workers differ from other groups in important, yet understudied, ways. This article examines what we know, and more importantly, what we need to know about sexual orientation in the workplace. Although much has been written about other minority groups in organizations (cf. reviews by Nkomo & Cox, 1996; Ragins & Gonzalez, 2003), there has been relatively little theory or research on sexual orientation in the workplace. As a foundation for research in this area, we need to understand the differences between LGB workers and other minority groups in organizations. Accordingly, a principal objective of this article is to provide an in-depth analysis of the similarities and differences between LGB workers and other minority groups in organizations. This provides a fundamental research map that can be used to chart new and exciting programs of research on this understudied population. Towards this goal, this article is organized into four sections. The first section lays the foundation for the article by defining sexual orientation, heterosexism and homophobia. The construct of sexual orientation is complex, and a brief review of social constructivist and queer theory approaches to defining sexual orientation is therefore needed. The second section provides a review of the research on sexual orientation in the workplace. Embedded in this section are research agendas and methodological caveats for conducting research in this area. In the third section, three unique challenges facing LGB employees are identified and detailed analyses of the effects of these challenges on the careers and workplace experiences of LGB workers are presented. Conceptual models and research propositions are developed in order to clarify relationships and provide a research road map for future research and theory development. The fourth and final section builds on the preceding sections, and identifies unique processes in the careers of LGB workers. The limitations of existing theories for explaining LGB workers’ careers are discussed

Sexual Orientation In The Workplace

37

and an identity-based, longitudinal theory of LGB careers is presented. I hope that these models and theories will help set a solid foundation for future research in this emerging, important, and sometimes controversial, area of research.

DEFINING SEXUAL ORIENTATION, HETEROSEXISM AND HOMOPHOBIA At the start our journey it is important to have a clear understanding of the constructs of sexual orientation, heterosexism and homophobia. At first glance these constructs appear to be relatively straightforward. However, as we will discover, these constructs are quite complex and are infused with social and political influences. Although quite a bit has been written on the construct of sexual orientation, there has been little discussion on sexual identity within the context of the work environment. Accordingly, this section provides a review of the construct of sexual orientation, and applies psychological, social constructivist and queer theory perspectives to organizational settings. To further set our foundation, a brief definition of heterosexism and homophobia is presented at the end of this section. Emerging Definitions of Sexual Orientation: From Kinsey to Queer Theory Dimensions of Sexual Identity and Orientation Sexual orientation represents a complex interaction of psychological influences that are nested within cultural, political and historical contexts (Bohan, 1996; Garnets & Kimmel, 1993). The very definition of sexual orientation has changed considerably over the past 50 years (see review by Ragins & Wiethoff, in press). Early definitions used a simple bipolar behavioral perspective: individuals were classified as heterosexual or homosexual based on whether they engaged in sexual relations with someone of the same biological sex (Kinsey et al., 1948, 1953). This perspective has a number of limitations (cf. Garnets & Kimmel, 1993). First, a behavioral perspective assumes that identity is linked to behavior, and ignores the fact that individuals may engage in same-sex sexual behaviors without viewing themselves as gay. For example, adolescents may engage in same-sex sexual experimentation while maintaining a heterosexual identity. In addition, some cultures may not view same-sex sexual acts as displays of homosexuality (Schmitt & Sofer, 1992). Second, individuals may self-identify as gay but, like heterosexuals, may choose to be celibate and abstain from all sexual behaviors. Third, this dichotomy views sexual orientation as a single dimension with homosexuality on one end of the dimension, bisexuality in the middle, and heterosexuality on the other end of the dimension. This definition ignores the complexity of bisexuality

38

BELLE ROSE RAGINS

and excludes transgendered, transsexual and asexual individuals (Horowitz & Newcomb, 2001). Researchers recognized that more fine-grained analyses were needed in order to capture the full complexity of sexual identity (Coleman, 1987; Garnets & Kimmel, 1993; Shively & DeCecco, 1977). To start, instead of viewing sexual orientation as existing on a single bi-polar dimension, sexual orientation is now viewed as falling along two continuous scales of homosexuality and heterosexuality (Haslam, 1997; Shively & DeCecco, 1977). This two-dimensional approach allows for more accurate assessments of bisexuality (high on both dimensions) and asexuality (low on both dimensions). We also witness a more comprehensive approach to defining the construct of sexual orientation. Current views hold that sexual orientation involves individuals’ physical behaviors, their emotional attractions and fantasies, as well as their sense of identity (Garnets & Kimmel, 1993). These factors vary by individual and may interact in complex ways. For example, some individuals are congruent in being both physically and emotionally attracted to one gender, while other individuals are physically attracted to one gender but emotionally attracted to another. In addition, these attractions may be independent of the individual’s social and personal identity. One outcome scenario is that an individual may be in a same-gender sexual relationship, but self-identify as heterosexual. As applied to the workplace, this scenario may result in a sharp divergence between an employee’s sexual identity and the perception of that identity by others in the workplace; an employee may consider him or herself heterosexual, but may be viewed as gay or lesbian by co-workers. This may lead to a certain degree of confusion and even conflict in the workplace, as discussed later in this article. Researchers now recognize that there is tremendous individual variation in the development of a gay identity (see Martin & Knox, 2000). At least three factors contribute to this variation. First, the development of a gay identity varies by individual over the course of a lifespan. Some individuals self-identify as gay prior to adolescence, while others recognize their gay identity, or “come out” to themselves, much later in life, sometimes even after raising a family in a monogamous heterosexual marriage (Cass, 1984; Troiden, 1989). Second, lesbians and gay men differ in the development of their sexual identity. Specifically, lesbians are more delayed in their sexual identity development than gay men, and are more likely to deny same-sex attraction and even marry because of social pressures (cf. review by Garnets & Kimmel, 1993). In addition, lesbians are more likely to self-define as gay based on their affectional orientation and political perspectives, whereas gay men are more likely to rely on physical behaviors and fantasies in defining themselves as gay (Gonsiorek & Weinrich, 1991). Third, there is significant individual variation among bisexuals. Some bisexuals view themselves as more gay than

Sexual Orientation In The Workplace

39

heterosexual, while others view themselves as primarily heterosexual but receptive to same-sex relationships (Blumstein & Schwartz, 1977). These differences may explain varying reports on the prevalence of homosexuality and bisexuality (Gonsiorek & Weinrich, 1991). In essence, reports of sexual identity may vary based on the respondent’s gender, age, and developmental stage (Martin & Knox, 2000). This has implications for the study of sexual orientation in the workplace, as some employees may not recognize their gay identity, or come out to themselves, until late in their careers or after they worked at an organization for some period of time. This means that an employee may establish work and developmental relationships based on a heterosexual identity but may have to redefine, and may even lose these relationships, once his or her true identity is revealed. This may create tremendous work-related stress for LGB employees, as discussed later in this article. The Social Constructivist Perspective As illustrated in the preceding discussion, psychological perspectives hold that sexual orientation is multidimensional and involves complex psychosocial and perceptual processes. There is another level of analysis that should be included in definitions of sexual orientation. In contrast to psychological perspectives on sexual orientation, a social constructivist perspective questions the view that sexual orientation exists as a fixed, underlying essence or observed trait (cf. Bohan, 1996; Weinrich & Williams, 1991). Using historical and cross-cultural perspectives, social constructivists view homosexuality as a construct that varies by time and by place, and that only exists relative to societal construction (McIntosh, 1968; Plummer, 1975; Weinberg, 1983). For example, Weinrich and Williams (1991) note that it is meaningless to talk about “lesbians in ancient Greece,” as the lesbian construct did not exist until the term was coined in the modern West. Another more recent example is that psychologists, psychiatrists and social workers viewed homosexuality as a mental disorder and sexual pathology until the early 1970s (Garnets & Kimmel, 1993; Martin & Knox, 2000); the same behavior was viewed as a sign of mental illness in one decade, but a healthy expression of sexual identity in another. In fact, media analysts now marvel at the fact that in the course of a few short years, homosexuality has moved from the closet to prime time television (Mendoza, 2003). Weinrich and Williams (1991) aptly summarize that under a social constructionist perspective, sexual orientation “cannot be properly or completely understood apart from the social milieu in which it is embedded” (p. 47). As applied to organizations, homosexuality may be viewed as a healthy form of diversity in one organization, but a form of deviance that violates religious principles in another (cf. Creed, Scully & Austin, 2002). In essence, the environmental context takes a central role in defining and placing value on identity and group membership.

40

BELLE ROSE RAGINS

Another key element of the social constructivist perspective is that the formation of a sexual identity is a continuous, two-way interactive process in which the individual interacts with his or her environment in the quest to construct a sexual identity. Sexual identity is viewed as a construct that is fluid over time and experience, as well as over the course of an individual’s life (Horowitz & Newcomb, 2001). Social constructivists view the environment as a powerful influence on the construction of a gay identity, and it is therefore reasonable to expect that the organizational environment may play a significant role in the development and disclosure of a gay identity. This topic is discussed at length later in this article. The Queer Theory Perspective A recent outgrowth of the social constructivist perspective is queer theory. Queer theory challenges the assumption of a unified homosexual identity. Seidman (1996, p. 11) summarized that “Queer theorists argue that identities are always multiple or at best composites with literally an infinite number of ways in which ‘identity components’ (e.g. sexual orientation, race, class, nationality, gender, age, ableness) can intersect or combine. Any specific identity construction, moreover, is arbitrary, unstable and exclusionary. Identity constructions necessarily entail the silencing or exclusion of some experiences or forms of life.” Queer theory views heterosexuality and homosexuality as categories of knowledge rather than identities, and represents a movement from the politics of minority to the politics of knowledge and difference (Seidman, 1996). In particular, instead of framing the study of homosexuality as a study of a minority group, queer theory shifts the focus to analyzing institutional practices and discourses that create norms, organize social life and repress differences (cf. Foucault, 1978). Queer theory questions the use of the “hetero/homosexual binary” and in so doing moves the focus from an exclusive preoccupation with homosexuality to a focus on heterosexuality as a social and political organizing principle (Seidman, 1996). Stein and Plummer (1996, p. 134) explain that “Queer theorists claim that existing gay strategies, and minority group strategies in general, have tended to rely on conceptual dualisms (male/female gender models, natural/artificial ontological systems, or essentialist/constructionist intellectual frameworks) that reinforce the notion of minority as ‘other’ and create binary oppositions which leave the ‘center’ intact.” Stein and Plummer (1996) observed that “queer theory normalizes homosexuality by making heterosexuality deviant. Homosexuality ceases to be the exclusive cite of differences” (p. 135). This conceptualization of sexual orientation has important philosophical and practical implications for the study of sexual orientation in the workplace. For example, an application of queer theory to organizations would entail an examination of how heterosexuality is infused in workplace cultures, norms, policies and

Sexual Orientation In The Workplace

41

practices. Like the study of other minority groups in organizations (cf. Ely, 1995; Nkomo, 1992; Nkomo & Cox, 1996; Wharton, 1992), the study of LGB employees dichotomizes identity (majority/minority), assumes a deficit model for the minority group, and fails to examine the overriding and pervasive effects of majority norms as organizational frames of reference. As applied to sexual identity, LGB employees become the “other” and questions regarding their experiences, acceptance or even the discrimination they face does not challenge the heterosexual majority. Using a deviance model, the focus becomes the extent to which LGB employees “fit in” or are accepted, thus leaving the dominant group and its norms intact and unpacked. Lesbian, gay and bisexual employees become evaluated relative to the dominant norms of the organization, which are based on heterosexuality (cf. Creed & Scully, 2000). Even the study of workplace discrimination focuses on the impact of discrimination on LGB employees as “the minority” without examining how sexual identity affects the organizational context. For example, implicit in the assessment of “sexual orientation in organizations” is the message that only gay men and lesbians have a sexual orientation; the overriding effects of sexual identity are not examined or questioned. Now that we have reviewed the complexities of sexual orientation as construct, let us continue to build the foundation for this article by defining heterosexism and homophobia.

Defining Heterosexism and Homophobia Heterosexism and homophobia are often used interchangeably, but they represent two very unique constructs. Heterosexism is defined as “an ideological system that denies, denigrates, and stigmatizes any non-heterosexual form of behavior, relationship, or community” (Herek, 1990, p. 316). Heterosexism involves antigay attitudes, prejudice, and discriminatory behavior (Sears, 1997). Like racism or sexism, heterosexism is manifested not only through individual behaviors, but also through institutional, cultural and legal venues (Herek, 1990). In contrast, homophobia is defined as the fear and aversion associated with homosexuality (Weinberg, 1972). Homophobia involves irrational, deep-seated fears of LGB people, and a cultural belief system that supports negative stereotypes about them (Morin & Garfinkle, 1978). Herek (1986) points out that the term homophobia is problematic, as “phobia” connotes an individual illness model that does not address the political and social realities of anti-gay prejudice and discrimination. Heterosexism may or may not be accompanied by homophobia. For example, an employee may engage in discriminatory behavior towards a gay co-worker without aversion or fear of that worker. A similar analogy may be found with

42

BELLE ROSE RAGINS

racism. Individuals engage in racist behaviors for reasons other than fear; prejudice may be based on self-interests, beliefs, group norms or social institutions (Allport, 1954). Therefore, individuals may engage in heterosexist behaviors without being homophobic, or may hold homophobic feelings without displaying heterosexist behaviors (Jung & Smith, 1993). The distinction between homophobia and heterosexism has important implications for organizations. Like racism, heterosexism may be institutionalized and have little to do with an individual employee’s phobia. We see examples of institutionalized heterosexism in organizations that lack policies that prohibit sexual orientation discrimination and in companies that do not provide the same worker benefits for LGB and heterosexual employees. Now that we have examined some of the essential differences between homophobia and heterosexism, let us turn to the second section of this article in which we review the research on workplace discrimination and the decision to disclose a gay identity at work.

A REVIEW OF THE RESEARCH ON SEXUAL ORIENTATION IN THE WORKPLACE Although there has been extensive research on the general construct of sexual orientation using psychological, economic and sociological perspectives (cf. Badgett, 2001), only a relatively small body of research has examined the topic of sexual orientation in the workplace (see reviews by Croteau, 1996; Ragins & Wiethoff, in press). This section reviews the research on workplace discrimination and the decision to disclose a gay identity at work. This review will be used as a springboard for identifying areas for future research, as well as conceptual and methodological challenges that face scholars conducting research in this area.

Prevalence and Types of Workplace Discrimination Prevalence of Workplace Discrimination Reports of heterosexism in the workplace reveal that it is relatively widespread; between 25 and 66% of LGB employees report that they experienced discrimination based on their sexual orientation (see review by Croteau, 1996). A national study of 534 gay and lesbian professionals indicated that over a third had faced verbal or physical harassment in prior positions because of their sexual orientation, 37% faced discrimination because others suspected or assumed they were gay, and 12% left their last job because of discrimination (Ragins & Cornwell, 2001b).

Sexual Orientation In The Workplace

43

Workplace discrimination against LGB workers may not only be overt, but may also take more subtle forms (cf. Chung, 2001; Peel, 2001). For example, a study of job applicants who posed as gay men and lesbians found that while employers did not overtly discriminate against them by denying them a job offer, they were still the target of more subtle forms of interpersonal discrimination involving shorter job interviews, less eye contact and more negative communication. Applicants who posed as gay were also more likely to report that their interviewers were nervous, hostile, standoffish and less interested in helping them (Hebl, Foster, Mannix & Dovidio, 2002). This points to the importance of distinguishing between different types of discrimination faced by LGB workers. Types of Workplace Discrimination Similar to women and other under-represented employee groups, LGB workers may face three types of discrimination in the workplace: disparate treatment, adverse impact and social discrimination (cf. Collella & Dipboye, in press; Twomey, 1994). The first two types of discrimination, disparate treatment and adverse impact, involve different treatment on the basis of group membership (Ledvinka & Scarpello, 1991). These types of discrimination are illegal for protected groups (e.g. race, color, national origin, gender, religion, age, and disability) under federal legislation (e.g. Title VII of the Civil Rights Act of 1964, the Civil Rights Act of 1991, the Americans With Disabilities Act of 1990 and the Age Discrimination in Employment Act of 1967) (Twomey, 1994). Disparate treatment occurs when protected individuals are intentionally treated differently on the basis of their group membership, whereas adverse impact (also called disparate impact) is the case in which a seemingly neutral employment practice has an adverse effect on a protected group (Ledvinka & Scarpello, 1991). LGB workers have no legal protection at the federal level against these forms of discrimination, and it is therefore legal to terminate their employment simply on the basis of their sexual orientation in many states, counties and municipalities (Herrschaft & Mills, 2002). In addition to disparate treatment and adverse impact, LGB employees face social discrimination, which can be subtle in form but potent in effect. Social discrimination involves differential access to social networks, relationships and activities that can improve an employee’s job performance and promote their careers (Collella & Dipboye, in press). Although no federal law addresses this form of discrimination, lack of access to social resources has been found to damage the careers of women and employees of color (Cleveland, Stockdale & Murphy, 2000; Powell & Graves, 2003; Ragins, Townsend & Mattis, 1998; Thomas & Gabarro, 1999). The exclusion from valuable work relationships is a pervasive barrier faced by LGB employees (Woods, 1993). Moreover, there is a “Catch-22” in this relationship; LGB employees who have disclosed their sexual identity

44

BELLE ROSE RAGINS

may be shunned from social interactions with colleagues who are uncomfortable interacting with them because of their orientation, but LGB workers who hide their sexual orientation may need to maintain a social distance from their colleagues in order to conceal their sexual identity, thus limiting the development of these valuable social relationships. In sum, LGB employees may face social discrimination and exclusion from valuable mentoring and networking relationships irrespective of whether or not they have disclosed their sexual identity in the workplace. As discussed later in this article, these three forms of discrimination can be experienced directly by the individual, which we call direct discrimination, or can be experienced indirectly through watching the experience of others. As discussed later, witnessing discrimination against other LGB individuals can have a powerful impact on LGB employees, and their decisions to disclose their sexual identity at work. Finally, workplace discrimination does not occur in a vacuum, but reflects deeper patterns of heterosexism in our society. Gay men and lesbians are frequently targeted for hate crimes (Nelson & Krieger, 1997). In fact, LGB individuals currently rank fourth in the nation in reported hate crimes, after race, ethnicity/national origin, and religion (Federal Bureau of Investigation, 2001). Although most hate crimes are perpetrated in public settings by strangers, hate crimes are also committed by neighbors, co-workers and even relatives of gay men, lesbians and bisexuals (Herek, Cogan & Gillis, 2002). Given this context, let us now turn to examining some of the antecedents and consequences of discrimination in the workplace.

Antecedents to Workplace Discrimination Organizational Policies and Protective Legislation One of the primary predictors of heterosexism in the workplace is the presence of organizational policies and practices that provide a positive climate for LGB workers and prohibit discrimination against them. In a study of 537 LGB employees, Button (2001) found that the presence of affirming organizational policies was associated with fewer reports of sexual orientation discrimination. Similarly, in a national random study of 534 LGB employees, Ragins and Cornwell (2001a) found that LGB employees were less likely to report either experiencing or observing sexual orientation discrimination in organizations that had supportive organizational policies and practices. The key policies that were found to be related to reports of discrimination included: written policies forbidding sexual orientation discrimination, sexual orientation being included in definitions of diversity, and the offering of same-sex domestic partner benefits. A particularly interesting finding was that of all the policies and practices, inviting same-sex partners to company social events

Sexual Orientation In The Workplace

45

had the strongest relationship to reduced reports of workplace discrimination. This practice is not only a strong indicator of a supportive organizational climate, but also reflects a relatively high comfort level with social interactions involving LGB workers. Although these two studies found that organizational policies and practices were significantly associated with reduced reports of heterosexism, another study failed to find this relationship. In his study of 287 LGB workers, Waldo (1999) found that supportive organizational policies did not influence reports of discrimination, but found that organizational climate played a more direct role in predicting discrimination. In this study, discrimination was operationalized as having “direct” (e.g. anti-gay jokes) and “indirect” (e.g. assumptions of heterosexuality) components, and LGB employees were more likely to experience both these forms of discrimination when they reported their organization was tolerant of heterosexism. The contradictory findings of these studies may be due to the different instruments used to measure workplace discrimination. Instruments of workplace discrimination can tap direct and indirect experiences, as well as potential and enacted discrimination (Chung, 1991). Moreover, emerging instruments on heterosexism vary with respect to psychometric properties and construct validity. Future research needs to play close attention to the psychometric properties of these emerging measures, and we need to develop reliable and valid instruments of heterosexism in the workplace. Until then, established instruments that assess workplace discrimination against other groups (i.e. race and gender) may be adapted to assess heterosexism in the workplace (cf. James, Lovato & Cropanzano, 1994). Finally, the presence of protective legislation may play an important role in preventing workplace discrimination against LGB employees. Although there is no national legislation prohibiting sexual orientation discrimination, legislation has been passed in select states, counties and municipalities (Herrschaft & Mills, 2002). While it is estimated that only one-fifth of gay and lesbian Americans live in areas that offer this protective legislation, existing research indicates that legislation can play a central role in reducing discrimination against LGB workers. Specifically, in the national study of 534 LGB workers discussed above, workers who were employed in organizations that were covered by legislation prohibiting sexual orientation discrimination reported less discrimination than those employed in organizations lacking such coverage (Ragins & Cornwell, 2001a). However, it is interesting to note that while protective legislation was significantly related to reduced reports of discrimination, the presence of organizational policies and practices was the single most important factor predicting reduced reports of discrimination, perhaps reflecting the more proximal influence of organizational culture on experiences of workplace discrimination.

46

BELLE ROSE RAGINS

Work Group Demography and Support A principal antecedent of workplace discrimination is the demography and attitudes of co-workers. Ragins, Cornwell and Miller (2003) found that LGB employees reported significantly less workplace discrimination when they had gay supervisors and when they had a higher proportion of gay co-workers in their work group. The racial composition of the workgroup also affected reports of sexual orientation discrimination. Irrespective of their race, LGB employees in racially balanced groups reported less heterosexism than those in primarily White or non-White work groups. This may reflect the fact that racially diverse teams may be more likely than homogeneous teams to develop diversity skills and competencies that affect the climate for LGB team members. Finally, the study found that more heterosexism was reported among LGB workers with male supervisors or work teams, and this effect was more likely to be reported by lesbians than by gay men. This finding is aligned with other research indicating greater reports of heterosexism among work groups with greater ratios of male co-workers (Waldo, 1999). Although there has been a lack of research on co-worker’s attitudes towards homosexuality, general research on attitudes towards homosexuality can illuminate potential predictors of co-worker attitudes. Existing research indicates that men have more negative attitudes towards homosexuality than women (Herek, 2000; Kite, 1984) and this effect is pronounced among heterosexual men’s attitudes towards gay men (Herek, 2000). Other studies have found that African-Americans are more disapproving of homosexuality than Whites, but are also more likely than Whites to support civil rights and oppose employment discrimination against gay men and lesbians (Lewis, 2003). Finally, prior contact with LGB individuals and the method of disclosure have been found to be powerful predictors of attitudes towards homosexuality. In particular, more positive attitudes are associated with those who had more and closer relationships with lesbians, gay men and bisexuals (Bowen & Bourgeois, 2001; Herek & Capitanio, 1996). Herek and Capitanio (1996) found that in addition to contact, the method of disclosure had a strong impact on attitudes; heterosexuals held more negative attitudes towards gay men and lesbians when they discovered an individual’s gay identity indirectly (e.g. through guessing or third parties) rather than through direct disclosure. The impact of method of disclosure on target attitudes has important implications for the workplace, and Herek and Capitanio’s findings (1996) need to be replicated and extended to work settings. Specifically, we need to examine whether co-workers’ reactions to LGB colleagues change as a function of the method of discovery, and we need to explore the factors that influence this relationship. At the onset, it is reasonable to expect that indirect discovery may be more likely

Sexual Orientation In The Workplace

47

to lead to feelings of betrayal in work relationships that have existed for longer periods of time, are interpersonally close, and involve interdependent work functions. The key factor underlying heterosexual co-workers’ feelings of betrayal may be their perception of the intimacy and closeness of the work relationship. Heterosexual workers who thought they had a close relationship with a gay colleague may feel foolish or betrayed when they discover from others that their colleague has been concealing one of the most personal and important aspects of identity. These feelings may even turn to anger if the gay colleague deceived co-workers by constructing a heterosexual identity (cf. Woods, 1993). However, although indirect discovery may lead to initial feelings of shock, betrayal or even anger, this disclosure can ultimately lead to a closer and more authentic work relationship, particularly if the relationship involves a heterosexual co-worker who has positive or even neutral attitudes about homosexuality. In fact, heterosexual co-workers who are sensitive to issues faced by LGB workers may understand their colleagues’ need for secrecy and may have more positive reactions to indirect discovery than co-workers who are unaware of the challenges faced by LGB workers. It is clear that we need more research on this complex, multi-layered and provocative topic. Race and Gender of LGB Worker A few studies have examined whether discriminatory actions are affected by the LGB individual’s race or gender. In a recent laboratory study, Horvath and Ryan (2003) asked 236 undergraduates to rate resumes in which the gender and sexual orientation of the applicant was manipulated. The researchers found that lesbian and gay male applicants were rated less positively than heterosexual male applicants, but surprisingly, more positively than heterosexual women. They also found that subject’s beliefs in the controllability of sexual orientation, their religiosity, their beliefs in traditional gender roles and their previous contact with LGB people were significantly related to their attitudes towards LGB people, which in turn was related to their beliefs about the consequences of employing LGB people. In a simulation study that examined rater bias using a sample of 548 employed MBA students, Crow, Fok and Hartman (1998) manipulated race, gender and sexual orientation. In contrast to some of the findings of the Horvath and Ryan study cited above, this study found that White, female heterosexuals were the least likely to be discriminated against in simulated hiring decisions. In line with their expectations, the researchers found that Black gay men were the most likely targets of hiring discrimination. Finally, Ragins, Cornwell and Miller (2003) examined race and gender effects in their national sample of 534 gay employees, 162 of whom were gay employees of color. They developed and tested two competing models of the effects of multiple

48

BELLE ROSE RAGINS

group membership on reports of heterosexism. They found that lesbians and gay people of color reported equivalent experiences of heterosexism as their White male counterparts, and concluded that racism and sexism may not spillover into experiences of heterosexism in the workplace. While membership in high status groups gives some employees an advantage, this effect was not apparent for the White gay men in this study. The researchers concluded that the race and gender of these men apparently did not buffer them from sexual orientation discrimination in the workplace. Although this study provides a start, future research needs to examine more closely the complex relationships between racism, sexism and heterosexism in the workplace. We need to understand the similarities and differences in the processes underlying these forms of social prejudice as they manifest in the workplace. Now that we have examined the antecedents to heterosexism, let us turn to reviewing the research on the outcomes associated with this form of workplace discrimination. Outcomes of Workplace Discrimination Sexual Orientation and Earnings A number of economists have examined the general effects of sexual orientation on earnings (see review by Badgett, 2001). Overall, their research suggests that the effects of sexual orientation on earnings vary for gay men and lesbians. Studies based on national General Social Survey data indicate that gay men earn less than their heterosexual counterparts. For example, Badgett (1995) found that gay and bisexual male workers earned 11–27% less than heterosexual male workers with the same experience, occupation, education, region of residence and marital status. Other estimates using more recent GSS data also found significant pay gaps for gay men, although the extent of these gaps varied based on the controls used in the study. Gay men were found to earn 14–16% less than heterosexual men in one study (Black, Makar, Sanders & Taylor, 2003), while another study reported that gay and bisexual men have a striking 30–32% income disadvantage relative to their heterosexual peers (Blandford, 2003). Using data from the 1990 U.S. Census, Allegretto and Arthur (2001) found that after controlling for differences associated with being married, gay men earned 2.4% less than their unmarried heterosexual counterparts. Many of these same studies found that although gay men earn less than heterosexual men, lesbians and bisexual women earn as much if not significantly more than heterosexual women. Badgett (1995) generally found no significant differences in earnings among lesbians and heterosexual women. However, lesbians were found to earn 20–34% more than heterosexual women in one

Sexual Orientation In The Workplace

49

investigation (Black, Makar, Sanders & Taylor, 2003), while another study indicated a 17–23% wage premium for lesbians and bisexual women (Blandford, 2003). These researchers offered a number of explanations for these findings. Blandford (2003) observed that lesbian and bisexual women earn more than heterosexual women because they are more successful gaining employment in high-paying, male-dominated fields. Black et al. (2003) note that lesbians may be more likely than heterosexual women to choose high-paying careers because they expect to be the primary wage earner in their relationships. Blandford (2003) speculates that lesbian and bisexual women may be less likely than heterosexual women to incur the wage penalties associated with career interruptions involving children. To add another layer of complexity, there is some evidence that these wage effects may be related to living arrangements. Clain and Leppel (2001) found that men living with male partners tended to earn less than other men, but women living with female partners tend to earn more than other women. These patterns of relationships are intriguing and future research can examine the extent to which they reflect social roles, gender roles, occupational choices or socioeconomic artifacts (cf. Badgett, 2001). In fact, the effect of partner relationships on career choices and organizational outcomes is a fascinating area for future research, particularly if researchers compared LGB partner relationships with a matched sample of heterosexual partner relationships. Career and Organizational Outcomes There has been relatively little research that has examined the effect of sexual orientation on the careers and advancement of LGB and heterosexual workers. The Ragins and Cornwell (2001a) study discussed earlier examined career and organizational outcomes associated with discrimination. This study found that LGB employees who reported more discrimination on the basis of their sexual orientation reported less career satisfaction, less satisfaction with opportunities for promotion, and received significantly fewer promotions in the past 10 years than those who reported less workplace discrimination. While the relationship between reported discrimination and compensation was not significant in the Ragins and Cornwell (2001a) study, an intriguing finding was that gay employees who worked with gay co-workers earned significantly less than gay employees who either worked with heterosexual co-workers or in groups that were balanced with respect to sexual orientation. Moreover, the effect of work group composition on compensation was direct and not mediated through reported discrimination, suggesting that gay workers may be tracked or self-select into positions or work groups that provide less pay. Future research could assess the effects of sexual orientation on career paths within and between organizations, and the career and organizational outcomes

50

BELLE ROSE RAGINS

associated with these career paths. Ideally, such research would use pairs of heterosexual and LGB employees who are matched on gender, rank and position. This study would provide interesting insights, but faces unique methodological challenges. Specifically, matching gay and heterosexual employees in organizational settings is constrained by the fact that openly gay employees may work in organizations that are more supportive of their careers than employees who have not disclosed their sexual identity. As discussed later, this represents one of many methodological challenges facing researchers studying sexual orientation in the workplace. Job Attitude Outcomes It is not surprising that existing research indicates a negative relationship between experienced heterosexism in the workplace and job attitudes (Button, 2001; Ragins & Cornwell, 2001a; Waldo, 1999). Ragins and Cornwell (2001a) found that discrimination was significantly related to lower job satisfaction, organizational commitment, organizational self-esteem, and greater turnover intentions. Similarly, other research has found that discrimination had a negative relationship with job satisfaction (Button, 2001; Waldo, 1999) and organizational commitment (Button, 2001). An interesting finding in Button’s (2001) study was that worker’s attitudes about their sexual identity interacted with discrimination to affect job satisfaction. LGB workers who expressed a fascination with their gay identity and a strong interest in gay and lesbian culture had more negative reactions to workplace discrimination than those who did not have a strong gay identity. This research suggests that LGB workers may be more dissatisfied with inequitable work environments when they have a strong gay identity, an idea that is examined at greater length later in this article. Stress and Mental Health Outcomes Some research indicates that heterosexism in the workplace contributes to the stress experienced by LGB workers and may have a negative impact on their mental health. For example, Waldo (1999) found that heterosexism was associated with higher levels of psychological distress and health-related problems among his sample of 287 LGB workers. Similarly, in a recent study of 97 LGB individuals, Smith and Ingram (2004) found that those who experienced heterosexism at work were also more likely to report depression and psychological distress. LGB individuals may experience stress as a consequence of discrimination within and outside the workplace. For example, using a national sample of 73 gay and bisexual adults and 2,844 heterosexuals, May and Cochran (2001) reported that LGB individuals were more likely to report every-day discrimination than heterosexuals, and that this discrimination was positively associated with the odds

Sexual Orientation In The Workplace

51

of having psychiatric disorders and psychological distress. Meyer (2003) conducted a meta-analysis to assess the effects of stress on the mental health of the LGB population. He found that lesbians, gay men and bisexuals have a higher prevalence of mental disorders than heterosexuals, and theorized that this relationship is a function of minority stress. As applied to the LGB population, he defined minority stress as resulting from stigma, prejudice and discrimination associated with sexual orientation. He proposed that key factors contributing to minority stress for LGB individuals include: hiding and concealing a gay identity, expectations of rejection, the potential for violence and discrimination, and internalized homophobia, which is the internalized self-hatred of a gay identity (Bohan, 1996). The antecedents and outcomes of workplace discrimination are intricately connected to the disclosure of a gay identity in the workplace. As discussed below, disclosure may be both an antecedent and a consequence of discrimination. In addition, individuals may be less likely to experience direct discrimination aimed at them if no knows that they are gay. Therefore, from a methodological perspective, disclosure and discrimination need to be examined conjointly in research on sexual orientation in the workplace. Let us now turn to exploring some of the research on disclosure in the workplace. Disclosure of Sexual Identity in the Workplace Identity Management Strategies In a typology of identity management strategies, Woods (1993) identified three strategies LGB employees use to manage their sexual identity at work. First, LGB workers can use a “counterfeiting” strategy. In this case the worker actively constructs a heterosexual identity, and may even go to such lengths as bringing an opposite-gender individual to company social events in order to appear heterosexual. This “counterfeiting” strategy is also known as “passing,” and involves active deception about one’s sexual identity (Griffin, 1992). Second, the LGB worker can employ an “avoidance” strategy in which they attempt to evade the issue by self-editing, censoring, and telling half-truths. They may maintain a social distance and avoid any discussion of their personal lives. Some may even appear asexual in the attempt to ward off questions about their sexuality. Third, LGB workers may use an “integration” strategy in which they openly disclose their sexual identity to others at work. This usually involves direct verbal disclosures, but may also involve bringing same-gender partners to organizational events and displaying pictures of partners in work settings. The disclosure of a gay identity in the workplace is called being “out at work,” which is a general reference to being “out of the closet.” In his study of 70 gay male professionals, Woods (1993) found that nearly all of the workers sought to avoid discrimination by posing as a heterosexual at some

52

BELLE ROSE RAGINS

point in their careers. Woods observes that the use of avoidance and counterfeiting strategies isolate gay employees and restrict their ability to develop genuine, open and trusting relationships with subordinates, co-workers and mentors. Those who rely on avoidance strategies may be viewed as antisocial or aloof by co-workers. In addition, counterfeiting a heterosexual identity is stressful and consumes a considerable amount of psychological energy. Finally, Woods observes that highranking positions frequently involve work-related social functions that request the presence of a spouse or a date. This may create considerable stress for closeted LGB employees who do not want to disclose their identity, but risk speculations about their sexual identity if they consistently go “solo” to these required social events. Existing research supports the idea that LGB employees use these three identity management strategies in the workplace. For example, Button (2001) found support for Woods’ (1993) typology of three identity strategies in his study of 537 LGB employees. However, Chrobot-Mason et al. (2001) caution that measurement of disclosure is complex, as individuals may rely on a combination of strategies in the workplace. For example, LGB workers may use an integrative strategy with some co-workers, but use avoiding or counterfeiting strategies with other colleagues. In contrast to assessing identity management strategies, some studies directly ask LGB workers the extent to which they disclose their sexual identity in the workplace (cf. reviews by Croteau, 1996; Ragins & Wiethoff, in press). These studies indicate that most workers limit the disclosure of their sexual identity in the workplace. For example, the Ragins and Cornwell’s (2001a) national study of 534 gay men and lesbians cited earlier in this review found that 12% of respondents reported that they were out to no one at work, 37% were out to some people, 24% were out to most people, and 27% reported that they were out to everyone at work. These results were comparable to findings of other studies which used smaller or localized samples (Driscoll, Kelley & Fassinger, 1996; Levine & Leonard, 1984; Schneider, 1987). In sum, there are two methods for measuring disclosure of a gay identity in the workplace. The first asks respondents to indicate the type of behaviors they use to manage their identity (i.e. counterfeiting, avoiding or integrating). Although this strategy provides important insights into how individuals manage their identity, it does not reveal the absolute extent to which they have disclosed at work. Moreover, as mentioned earlier, different strategies may be used with different individuals (Chrobot-Mason et al., 2001). For example, some individuals may use integrating strategies with trusted co-workers but may avoid discussing their sexual identity with upper-management, customers, clients or vendors. It is therefore important to assess not only the identity management strategies used, but also the extent of disclosure and the target of disclosure in the workplace (for more discussion see: Ragins & Cornwell, 2001b; Ragins & Wiethoff, in press).

Sexual Orientation In The Workplace

53

Now that we have explored the strategies and prevalence of disclosure in the workplace, let us examine the antecedents and consequences associated with being “out at work.”

Antecedents of Disclosure of Sexual Identity in the Workplace Identity Strength and Disclosure A number of studies have found that that the presence of a strong gay identity is a key predictor of the decision to disclose in the workplace. By extending ethnic identity theory to sexual identity, Chrobot-Mason et al. (2001) tested the idea that identity affects disclosure decisions. Using a sample of 255 lesbian and gay workers, she and her colleagues found that employees with a strong sexual identity were more likely to use integrative strategies and disclose their gay identity at work than those who with a weaker sexual identity. Similarly, Button (2001) found that LGB workers who reported stronger identification with gay groups were more likely to disclose at work than those lacking such identification. Along similar lines, Rostosky and Riggle (2002) found a significant relationship between internalized homophobia, which reflects negative views of one’s gay identity, and reduced disclosure in the workplace. Along similar lines, in their study of 370 gay men and lesbians Griffith and Hebl (2002) found that centrality of a gay identity and selfacceptance predicted disclosure decisions. Of particular interest was their finding of a relationship between disclosure at work and disclosure to friends and family outside the workplace. The relationship between disclosure in home and work settings is an interesting concept that is discussed later in this article. Although these studies found that gay identity is an important predictor of disclosure in the workplace, some scholars challenge the notion that disclosure is driven primarily by an individual’s internal identity, and point to the influence of environmental factors on disclosure decisions (Cain, 1991; Fassinger, 1995; McCarn & Fassinger, 1996). For example, Cain (1991) observes that decisions regarding disclosure are not just due to the individual’s identity development, but are shaped by powerful social influences and the anticipated response of others in the social environment. In support of this idea, his interview study of 38 gay men in Montreal revealed that disclosure was guided by the anticipation of potentially negative consequences of disclosure within a social context, and that these anticipated consequences were distinct from gay identity development. He concluded that individuals who are open about their sexual orientation are not necessarily more developmentally advanced than those who conceal it, and that concealment in a non-supportive environment may be a very effective strategy for self-preservation.

54

BELLE ROSE RAGINS

Future research could examine the complex relationships between the strength of sexual identity and the environmental context. For example, as proposed later in this article, individuals who have a strong gay identity may self-select into organizations that support this identity. Alternatively, we know little about the effects of organizational environment on identity. LGB workers in organizations that allow them to disclose at work may develop stronger gay identities than those working in organizations that do not allow their disclosure. Finally, as discussed later in this section, much work needs to be done on the effects of work group composition on disclosure and the development of a gay identity. LGB workers who work in groups comprised primarily of LGB workers may develop stronger social identities that involve their sexual identity than those who are the only gay member of their group. This points the way to an examination of the effects of the organization’s environment and climate on disclosure decisions. Disclosure and Organizational Policies, Practices and Climate Similar to the relationships found for workplace discrimination, existing research indicates that organizational policies have a significant relationship with the decision to disclose a gay identity in the workplace. Chrobot-Mason et al. (2001) found that lesbian and gay employees were more likely to disclose when they perceived more affirming organizational climates, a finding that was replicated by Griffith and Hebl (2002). In investigating the specific policies associated with disclosure, Ragins and Cornwell (2001a) found that disclosure was associated with written policies forbidding sexual orientation discrimination, the inclusion of sexual orientation in organizational definitions of diversity, offering same-sex domestic partner benefits, offering gay support groups, and the practice of inviting same-sex partners to company social events. In addition, they found more frequent disclosure in organizations that were covered by legislation prohibiting sexual orientation discrimination. Some research suggests that a worker’s decision to disclose in the workplace may be affected by his or her partner’s disclosure decision. In a study of 118 lesbian and gay couples, Rostosky and Riggle (2002) found that individuals were more likely to disclose at work when their partner worked at an organization that had a non-discrimination policy. Future research needs to examine the effects of non-work relationships on the decision to disclose in the workplace, a topic that is examined more thoroughly later in this article. Work Group Predictors of Disclosure Although there has been little empirical research on this topic, work group demography and climate may play a central role in disclosure decisions. Ragins

Sexual Orientation In The Workplace

55

et al. (2003) found that LGB employees with gay supervisors or primarily gay work groups were more likely to disclose at work than those who worked in primarily heterosexual work groups. In addition, LGB employees were more likely to disclose when they had supervisors of the same race or ethnicity, regardless of the supervisor’s sexual orientation, indicating support for relational demography perspectives (Riordan, 2001). However, the demography of the work group may not be as important as the support provided to the LGB worker. In a study of 499 lesbians, Jordan and Deluty (1998) found that social support was a key predictor of disclosure of sexual orientation. Although the researchers did not examine disclosure in the workplace, it is reasonable to expect that social support is a critical predictor of workplace disclosure. In support of this idea, LGB employees with supportive co-workers and supervisors reported less fear of disclosure and were more likely to disclose their identity at work than those lacking a supportive work group, even when holding the sexual orientation of the group constant (Ragins & Cornwell, 2001b). Since this study is cross-sectional, we cannot determine the direction of the relationships. Future research could examine whether LGB employees who seek to disclose their identity self-select into demographically similar groups that support this disclosure decision.

LGB Worker Demography and Disclosure Some research has been conducted on the effects of demographic variables on disclosure decisions. For example, age is significantly related to disclosure decisions; older gay men and lesbians are less likely to be out at work than their younger counterparts (Bradford, Ryan & Rothblum, 1994; Schope, 2002). This may reflect generational experiences; for example, LGB workers who grew up in the pre-Stonewall 1950s may be less comfortable disclosing a gay identity at work than those who grew up in post-gay civil rights eras (Martin & Knox, 2000). A few studies have examined the effects of gender and race on disclosure decisions. This research has failed to find gender effects; lesbians have been found to be as likely as gay men to disclose in the workplace (Ragins et al., 2003) and use similar identity management strategies at work (Chrobot-Mason et al., 2001). On the other hand, some research indicates that race plays a role in the disclosure decision. In a study of 1,925 lesbians, Bradford, Ryan and Rothblum (1994) found that lesbians of color were less likely to be out in their lives than White lesbians. Similarly, in a sample of 534 gay employees, 162 of whom were gay employees of color, gay employees of color were less likely to disclose their sexual identity in the workplace than their White counterparts (Ragins et al., 2003).

56

BELLE ROSE RAGINS

There are a number of potential explanations for the finding of race differences in disclosure. One explanation is that gay people of color are already highly visible because of their race and numerical rarity, and this visibility may heighten their fears about disclosing a gay identity at work. They may fear that if they disclose their identity to even a few people, their visibility will fan the gossip mills that spread this information to the rest of the organization. Another explanation is that gay people of color who experience discrimination on the basis of race may want to avoid adding “fuel” to the discriminatory fire by disclosing their stigmatized sexual identity. Finally, gay people of color may not want to risk losing the valuable support of co-workers of color who are not comfortable working with gay colleagues. Future research could examine and test these explanations. Prior Experience with Discrimination and Disclosure The decision to disclose a gay identity at work may be affected by the individual’s prior experiences with workplace discrimination. In a study of 228 lesbians in New England, Schneider (1987) found that lesbians who had been fired from a prior position due to disclosure were less likely to disclose their sexual orientation in their present position than those who did not have this experience. In addition, some of the gay men interviewed in Woods’ (1993) study reported that past experiences with workplace discrimination contributed to a heightened awareness of the potential for discrimination in their current position. However, Ragins and Cornwell (2001b) found that while prior experiences with discrimination increased LGB workers’ fears about disclosing their identity in their current positions, individuals with prior experiences with discrimination were still more likely to disclose than those who did not have these experiences. This suggests that while a history of discrimination may heighten the perception of the risk associated with disclosure, once LGB workers are out of the closet, they may not be willing to relinquish their identity and retreat back into the closet in subsequent positions. It is clear that a number of factors affect disclosure decisions, and future research needs to untangle the effects of these antecedent variables. Now let us turn to a review of the research on the outcomes associated with disclosing a gay identity in the workplace. Outcomes of Disclosure Job Attitudes and Psychological Strain A fair number of studies have examined the relationship between the degree of disclosure in the workplace and work-related attitudes, outcomes and psychological strain at work. A basic premise behind these studies is that LGB employees who are open about their sexual orientation at work avoid the conflict and stress involved with the concealment and management of dual-identities.

Sexual Orientation In The Workplace

57

Although this is a reasonable view, the results of these studies yield inconsistent and even contradictory findings. Some researchers found that LGB employees who disclosed their sexual identity to more people at work had more positive job attitudes (Day & Schoenrade, 1997; Ellis & Riggle, 1995; Griffith & Hebl, 2002), while other studies found no relationship between the degree of disclosure and work attitudes (Croteau & Lark, 1995; Driscoll, Kelley & Fassinger, 1996). In fact, some studies found that LGB employees who disclosed to more people at work had lower continuance organizational commitment (Day & Schoenrade, 1997), earned less compensation (Ellis & Riggle, 1995; Schneider, 1987), and had less pay satisfaction (Ellis & Riggle, 1995) than those who concealed their identity from others in the workplace. The decision to disclose should be strongly related to work-related stress and psychological strain, but this research has also produced inconsistent findings. Some studies found a significant relationship between disclosure and reduced anxiety, depression and other negative psychological symptoms (Jordan & Deluty, 1998; Smith & Ingram, 2004), but other studies found no relationship between disclosure and work-related stress (Day & Schoenrade, 1997), psychological strain or occupational coping (Driscoll, Kelley & Fassinger, 1996). One reason for these inconsistent findings is that the actual decision to disclose at work may have less of an impact on the individual than the immediate fears that precede this decision. In essence, the fear of negative consequences of disclosure may have a greater impact on the psychological health and well being of LGB employees than the actual decision itself, which can bring a sense of relief to LGB employees (Friskopp & Silverstein, 1996; Woods, 1993). Ragins and Cornwell (2001b) tested this idea by surveying 534 LGB employees, and found that employees’ fear of negative consequences of disclosure had a greater impact on work attitudes and psychological strain than the actual disclosure decision. While the actual decision to disclose was generally unrelated to outcome variables, the fears associated with disclosure had an overwhelming negative impact on 14 of the 15 variables examined in the study. Fear of disclosure was significantly related to job satisfaction, organizational commitment, turnover intention, opportunities for promotion, organizational self-esteem, role ambiguity, role conflict, group participation and career satisfaction. Respondents who reported greater fears of disclosure also reported experiencing a host of physical stress related symptoms, such as somatic complaints, work-related depression, anxiety and irritation, and they received fewer promotions, and marginally less compensation than those who did not anticipate negative consequences to disclosure. The authors concluded that the combined costs associated with these fears, both to the individual and the organization, could be staggering. Finally, this study suggests that fear of discrimination may rest at the heart of disclosure. Let us now examine the relationship between

58

BELLE ROSE RAGINS

disclosure and discrimination, and the methodological challenges inherent in this research.

Methodological Challenges in Conducting Research on Sexual Orientation in the Workplace Understanding the Discrimination-Disclosure Relationship One of the first challenges in conducting research on sexual orientation in the workplace is untangling the relationship between discrimination and disclosure. Since existing research in this area has been cross-sectional, we are unable to determine the direction of these relationships; discrimination may represent a consequence or an antecedent to the disclosure of a gay identity in the workplace (Badgett, 1996). Most research indicates that LGB workers are less likely to disclose their sexual identity in workplaces that foster discrimination. In their early study of 174 lesbian and gay student affairs professionals, Croteau and Lark (1995) found that employees who faced discrimination were less likely to disclose in the workplace than those who did not face discrimination. Subsequent studies supported this finding; Smith and Ingram (2004) found a negative relationship between heterosexism and disclosure at work, and Button (2001) found that discrimination was positively related to counterfeiting and avoiding strategies, and negatively associated with integrating. Finally, Ragins and Cornwell (2001a) found that LGB employees were less likely to disclose their sexual identity when they reported either experiencing or observing sexual orientation discrimination at work. In contrast, in his study of 287 LGB workers, Waldo (1999) found greater reports of direct discrimination were associated with greater disclosure, suggesting that those who disclosed were more vulnerable to discrimination. Indeed in some workplaces, most notably the military, disclosure can be the cause, rather than the consequence of discrimination (cf. Herek, Jobe & Karney, 1996). Methodological Issues It is clear that we need more longitudinal research that disentangles the effects of discrimination from disclosure. However, there are a number of methodological challenges inherent in such research (cf. Ragins & Wiethoff, in press). To start, selfreports of disclosure may not reflect actual disclosure. For example, a gay employee may believe that no one knows he is gay, in spite of the fact that everyone is aware of his sexual identity. One outcome of this scenario is that LGB employees may experience workplace discrimination, but not report it because they believe that no one knows they are gay. Another scenario involves a gay employee who mistakenly

Sexual Orientation In The Workplace

59

believes that everyone is aware of his sexual identity, and therefore misattributes a negative performance review to his sexual orientation. These examples suggest that self-reports of discrimination and disclosure may over or under-estimate actual discrimination and actual disclosure in the workplace. Another challenge in conducting research in this area is obtaining a representative sample of LGB workers (cf. discussion by Ragins & Wiethoff, in press). The routine practice of surveying members of organizations and gathering performance-based data on respondents is quite problematic for LGB research. In addition to confidentiality and privacy concerns (Martin & Knox, 2000), surveys inquiring about sexual orientation that are administered by employers may be very stressful for LGB workers, particularly those who are not out at work. Another limitation with surveying members of an organization is that LGB employees who have not disclosed their sexual orientation at work may not return the survey, or may report they are heterosexual, thus jeopardizing the key variable in the study. In addition to creating error variance, this method creates response bias because LGB workers may be more likely to return a survey on sexual orientation if they have disclosed their sexual identity at work. Given these methodological challenges, studies on sexual orientation in the workplace often use members of LGB organizations or rely on snowballing sampling techniques. These techniques address some of the methodological issues discussed above but have limitations with respect to generalizability of findings. Herek and his colleagues (Herek, Kimmel, Amaro & Melton, 1991) offer a number of useful sampling strategies. For example, researchers can target settings in the community frequented by LGB individuals, such as coffeehouses, bookstores and community centers. Samples can also be obtained through exit polls at voting sites in gay neighborhoods and through advertising in LGB publications.

Summary of Literature Review In summary, existing research indicates that discrimination against LGB workers is relatively widespread, but is contained in organizations that have supportive policies, practices and cultures. The presence of protective legislation and supportive work teams may also buffer LGB workers from workplace discrimination. It is not surprising that experiences of workplace discrimination are related to a host of negative work and career attitudes and outcomes. Discrimination also takes its toll on the stress and overall mental health of LGB workers. Disclosure of a gay identity in the workplace is a key issue for LGB employees, and most workers have not fully disclosed their sexual identity at work. LGB

60

BELLE ROSE RAGINS

workers are more likely to disclose in supportive organizations that are covered by protective legislation. Although the work group’s demography affects disclosure decisions, the social support of the group has a stronger influence on the individual’s decision to disclose than the group’s demography. Disclosure is also associated with a number of individual differences variables. Individuals with a strong gay identity are more likely to disclose at work than those with a weak identity, but the direction of this relationship is open to question. Individuals with strong gay identities may be more likely to choose work environments that support their disclosure, and supportive settings may bolster a gay identity. Prior experiences with discrimination affects fears regarding disclosure, but it seems that few LGB workers will return to the closet after they have disclosed in the workplace. Research on attitudes and outcomes associated with disclosure yield inconsistent and contradictory findings. One explanation for this is that the fear associated with disclosure may be the predominant factor affecting attitudes and outcomes, and there is some evidence that fears associated with disclosure have a stronger impact on outcome variables than the actual disclosure decision, which may come as a sense of relief to some LGB workers. The impact of race, gender and age on discrimination and disclosure reveals a complex pattern of relationships. Gay employees of color and older workers are less likely to disclose than White and younger LGB workers, but lesbians are as likely to disclose as gay men. Although gender and race do not affect reports of workplace discrimination, gay men earn less than heterosexual men. In contrast, lesbians earn more than heterosexual women, perhaps reflecting differences in career paths, occupational choices and career interruptions. Finally, this review identified a number of methodological challenges relating to measurement, operationalization and sampling in LGB research. The stigma and invisibility of the sample creates unique challenges, and the sampling strategies used in gender and race research may not be easily transferred to sexual orientation research. As evident from this review, sexual orientation as a field of inquiry is still in its infancy. We have some insights into the antecedents and consequences of discrimination and disclosure, but we have just skimmed the surface in assessing these relationships. This represents a tremendous opportunity for scholars to break new ground in assessing the career and workplace experiences of the LGB workforce. As our field matures, we need theoretical models and frameworks to guide our research. This brings us to the next section of this article, in which we examine three unique challenges faced by LGB workers, and develop research propositions and models to guide future research in this area.

Sexual Orientation In The Workplace

61

UNIQUE ASPECTS OF SEXUAL ORIENATION AS A STIGMATIZED SOCIAL IDENTITY: THE THREE CHALLENGES An accurate assessment of the workplace challenges faced by lesbian, gay and bisexual employees require an understanding of some of the similarities and differences between LGB workers and other under-represented groups in organizations. Like other minority groups (see reviews by Nkomo & Cox, 1996; Ragins & Gonzalez, 2003), LGB employees face prejudice, stereotypes, restricted power, and multiple forms of employment discrimination. However there are also attributes unique to LGB employees that create special challenges for them in the workplace. Stigma theory provides a useful framework for understanding these challenges, and for analyzing some of the similarities and differences between LGB employees and other minority groups in organizations (cf. Ragins, 2004). According to stigma theory, groups that are stigmatized are seen as inferior, flawed or deviant (Goffman, 1963). Stigmatized individuals possess, or are believed to possess, an attribute or characteristic that reflects a devalued or deviant social identity (Crocker, Major and Steele, 1998; Goffman, 1963). Stigmas reflect attributions about the individual’s identity and character that result in prejudice and discrimination (Dovidio, Major & Crocker, 2000). While the process of stigmatization may affect lesbian, gay and bisexual employees in similar ways as other stigmatized groups (cf. Clair, Beatty & MacLean, 2002; Ragins, 2004), the stigma of sexual orientation creates unique challenges for LGB workers that are not faced by individuals in other stigmatized groups. Three unique challenges are identified and analyzed in this section. The first challenge stems from the invisibility of sexual orientation as a stigmatized identity. The invisibility of a gay identity leads to decisions about disclosure of the identity, possible disconnections between disclosure in work and non-work environments, and the potential for loss of control over the disclosure process. The second challenge involves the potential negative reaction of heterosexual co-workers to their LGB colleagues. Negative reactions stem from threatened emotional states and the perception that LGB workers are responsible for their stigma. The third challenge centers on the dual effects of identity development and lack of social support on LGB workers. Specifically, the longitudinal nature of sexual identity development for LGB employees, coupled with the lack of support in work and non-work settings, results in specific career and workplace challenges for the LGB workforce. This section explores the implications of these three unique challenges and presents conceptual models and frameworks that describe and explain the complex

62

BELLE ROSE RAGINS

set of relationships underlying these challenges. Research propositions are also presented with the goal of guiding future research in this area. As a boundary condition, it should be noted that many factors affect the workplace experiences of LGB workers. Because much has been written on the workplace experiences of other minority groups (see reviews by Nkomo & Cox, 1996; Ragins & Gonzalez, 2003), this article will not review the processes and experiences that are common to both LGB workers and other groups that are in the minority in organizations (e.g. stereotyping, tokenism and attribution effects, exclusion from formal and informal developmental relationships and networks, harassment and discrimination). Instead, I attempt to break new ground by focusing on the challenges that tend to make LGB workers unique from other minority groups and to examine the antecedents, outcomes and relationships among these distinctive challenges.

Challenge No. 1: Sexual Orientation as an Invisible Stigmatized Identity A key feature distinguishing LGB groups from many other groups is the invisibility of their stigma (Badgett, 1996; Herek, 1996). In this sense, sexual identity is similar to religion, class and invisible disabilities, but differs from gender, groups with visible disabilities, and in most cases, race (Ragins et al., 2003). Because of the invisibility of their stigma, a core challenge facing gay employees is the decision to disclose their sexual identity, or “come out of the closet” at work. This disclosure decision has a number of unique and important implications for the LGB workforce. This section examines some unique issues relating to the challenge of having an invisible stigmatized identity. In particular, I examine antecedents and outcomes relating to the disclosure, and non-disclosure, of a gay identity in the workplace. The Disclosure of Sexual Identity Continuum The decision to disclose a gay identity at work is not an all-or-none decision, but reflects a continuum ranging from total disclosure to everyone at work at one end, to absolute non-disclosure at the other end. For those who choose to reveal their sexual identity, the decision to disclose is usually conducted on a careful case-by-case basis with each new social interaction (Cain, 1991). Gay employees weigh the potential costs of disclosure, such as social rejection, job termination and even physical assault (D’Augelli & Grossman, 2001; Friskopp & Silverstein, 1996; Herek, Cogan & Gillis, 2002; Woods & Harbeck, 1992), with the psychological benefits associated with bringing one’s true identity to work. Concealing a stigmatized identity has been found in some studies to lead to psychological stress

Sexual Orientation In The Workplace

63

and depression (Jordan & Deluty, 1998; Lewis, Derlega, Berndt, Morris & Rose, 2001; Meyer, 2003; Smart & Wegner, 2000; Smith & Ingram, 2004), and concealment may also prevent the establishment of close and authentic relationships at work (Fassinger, 1995, 1996). Gay employees who are “in the closet” often keep a social distance from their co-workers so as to avoid situations that reveal their true identity (Woods, 1993). This social distance may restrict the development of important mentoring and networking relationships among LGB workers, and may have long-term detrimental effects on their careers (see reviews by Croteau, 1996; Lonborg & Phillips, 1996; Ragins & Wiethoff, in press). Some scholars note that although the decision to conceal a gay identity may be stressful, it is not automatically a bad decision (Cain, 1991; Fassinger, 1995; McCarn & Fassinger, 1996). The decision to remain in the closet may be an effective strategy for managing a gay identity in organizations that discriminate against gay employees or terminates them because of their sexual identity. An example frequently cited is the U.S. military policy of “Don’t Ask, Don’t Tell,” which makes employment in the military contingent upon gay personnel hiding their sexual identity (Herek, 1996; Herek et al., 1996). While the decision to remain in the closet may be a functional and adaptive reaction to a hostile work environment, it also increases the risk and stress associated with losing control over the disclosure process. Control Over Disclosure and Disclosure Disconnects for LGB Workers Since most LGB employees do not disclose their sexual identity to everyone at work (Button, 2001; Ragins & Cornwell, 2001b), a key issue they face is the loss of control over the disclosure process. For example, a gay employee may confide in one or two trusted co-workers, but gossip channels may spread this information to the rest of the organization, including those in influential positions who hold negative attitudes towards homosexuality. One consequence of this scenario is that LGB employees are often uncertain as to “who knows and who doesn’t.” This uncertainty compounds the stress related to lack of control over the disclosure process. The most disturbing and potentially stressful case of loss of control over the disclosure process involves LGB employees who are involuntarily “outed” by others at work. For example, a gay employee may choose not to disclose his sexual identity at work, but may be “outed” after he is spotted by a co-worker at a gay bookstore or at a gay event in the community. This can create an inordinate amount of stress for LGB employees who have chosen to conceal their sexual identity at work, but choose to be open about their sexual identity outside the workplace. This also suggests that disclosure at work is connected to disclosure at home, and that LGB workers who are fearful of disclosure of their gay identity at work may

64

BELLE ROSE RAGINS

feel pressured to monitor their behaviors, relationships and activities outside the workplace. Control over disclosure is an issue that spills not only from the home to the workplace, but also from the workplace to the home. Some LGB employees may be out at work, but may not disclose their identity to parents, siblings or relatives. For example, a recent study of 443 gay men found that 35% did not disclose their gay identity to their parents, and 23% hid their identity from their siblings (Schope, 2002). LGB workers who are out at work, but not to their families, risk being “outed” by well-meaning but oblivious co-workers who run into family members in the local community. This line of thinking supports the idea that LGB employees face sexual identity disclosure disconnects between work and non-work environments. Ragins (2004) analyzes the concept of disclosure disconnects for individuals with invisible stigmas, and this process should extend to LGB workers as well. Specifically, LGB workers may be out in one domain to a greater extent than the other, thus creating a disclosure disconnect between work and home domains. Proposition 1. LGB workers experience sexual identity disclosure disconnects when they have disclosed their sexual identity to different degrees in work and non-work settings.

Consequences of disclosure disconnects for LGB workers. Sexual identity disclosure disconnects may have at least two consequences for LGB workers. First, it is reasonable to expect that the greater the disparity in disclosure between work and home domains, the greater the risk of being intentionally or unintentionally “outed” by co-workers or members of the community. Sexual identity disclosure disconnects can therefore lead to feelings of lack of control over the disclosure process for many LGB workers. These feelings of lack of control may in turn increase the general stress in the lives of LGB workers (see Meyer, 2003 for review). Moreover, since the risk of being “outed” is amplified in small towns or communities, the size of the community is an important factor to consider when examining these relationships. It is reasonable to expect that the effect of disclosure disconnects on perceived lack of control will be greater in smaller communities than larger communities. Proposition 2. Sexual identity disclosure disconnects leads to perceptions of loss of control over the disclosure process for LGB workers. Loss of control also mediates the relationship between disclosure disconnects and experienced stress.

Sexual Orientation In The Workplace

65

Proposition 3. Size of community moderates the relationship between sexual identity disclosure disconnects and perceived lack of control over the disclosure process. The second consequence of sexual identity disclosure disconnects involves the impact of managing the same identity differently at work and at home on the worker’s sexual identity. Like other workers with invisible stigmas (Ragins, 2004), LGB workers should experience sexual identity conflict as a consequence of disclosure disconnects. A number of psychological theories on self and identity maintain that individuals need identity congruence across environments in order to avoid negative psychological outcomes (cf. Leary & Tangney, 2003). For example, selfverification theory (Swann, 1983, 1987; Swann et al., 2004) holds that individuals seek stable views of themselves to bolster their feelings of psychological coherence, and display identity cues in order to create social environments that reinforce coherent and stable self-views. Using a symbolic interactionist perspective, Thoits (1983) theorized that individuals can maintain multiple identities, and that the loss of an identity can lead to depression and anxiety. These theories support the idea that disclosure disconnects can lead to sexual identity conflict for LGB workers, and that this conflict can create stress in their lives. Identity conflict may have a direct effect on stress, but it can also serve as the process by which disclosure disconnects affect stress. Therefore, it is reasonable to propose that sexual identity conflict may mediate the relationship between sexual identity disclosure disconnects and stress for LGB workers. Proposition 4. Disclosure disconnects leads to sexual identity conflict for LGB workers. Sexual identity conflict also mediates the relationship between disclosure disconnects and stress. The effect of disclosure disconnects on sexual identity conflict may depend on the centrality or saliency of the sexual identity to the individual’s total identity. Similar to the effects of racial identity on psychological outcomes (Cross, 1978), sexual identity is a key factor to consider when analyzing disclosure processes and outcomes (cf. Button, 2001; Chrobot-Mason et al., 2001; Rostosky & Riggle, 2002; Walters & Simoni, 1993). Individuals have multiple identities (Ferdman, 1999) and LGB employees may differ in how they view their sexual identity relative to their other identities. For example, some gay people of color may view their racial identity as being more salient than their sexual identity, whereas others may view themselves primarily in terms of their sexual identity (Ragins et al., 2003). The salience of sexual identity may therefore influence the relationship between disclosure disconnects and identity conflict for LGB employees. Specifically, LGB workers may be less likely to experience sexual identity conflict as a

66

BELLE ROSE RAGINS

result of disclosure disconnects if their sexual identity is not central to their overall identity. Proposition 5. Identity salience will moderate the relationship between disclosure disconnects and sexual identity conflict for LGB workers. Finally, aligned with classic research on the effects of stress (see review by Beehr & Bhagat, 1985), the stress associated with sexual identity conflict and the lack of control over the disclosure process should affect a number of work and life outcomes, such as life satisfaction, job satisfaction, organizational commitment, turnover intentions, career satisfaction, advancement and compensation. These relationships are illustrated in Fig. 1. Disclosure and sexual identity. Now let us turn to a more fine-grained examination of the relationship between disclosure disconnects and sexual identity for LGB workers. Since disclosure falls along a continuum, let us examine how various combinations of disclosure at work and at home influence sexual identity processes among these workers. Based on more general theory of invisible stigmas, disclosure and identity outcomes (cf. Ragins, 2004) disclosure disconnects can have three distinct influences on LGB identity. First, as discussed earlier, sexual identity conflict occurs when the LGB worker discloses his or her sexual identity in one domain but not in another. The clearest case of sexual identity conflict results from situations involving strong disclosure disconnects in which an individual has disclosed to everyone in one domain, but no one in another domain. For example, a gay employee may have disclosed his sexual identity to everyone at home, but no one at work. Another example involves cases where gay employees not only hide their sexual identity at work, but also fabricate heterosexual identities in order to avoid questions and gossip (Button, 2001; Chrobot-Mason et al., 2001; Woods, 1993). Sexual identity conflict therefore results from situations in which an LGB worker maintains one sexual identity in the workplace, and another sexual identity at home. These dual identities are maintained through differing degrees of disclosure or through the fabrication of a heterosexual identity. Second, LGB workers may have a congruent sexual identity, which involves a consistency in the degree of sexual identity disclosure across work and home domains (Ragins, 2004). An example illustrating this identity state is the case where a gay employee has disclosed his sexual identity to everyone at work and everyone at home. There are no disclosure disconnects in this situation, thus the individual’s sexual identity across home and work domains is congruent.

Sexual Orientation In The Workplace

67

Fig. 1. Outcomes of Disclosure Disconnects for LGB Workers.

The third identity state falls between these two examples, and probably represents the most common situation. In this case, there is a moderate degree of disclosure disconnect: the gay employee has disclosed his sexual identity to some people at work and some people at home. Although this scenario does not represent as intense a case of disclosure disconnect as the first example of identity conflict, it may still create a strain on the individual’s identity. In essence, this

68

BELLE ROSE RAGINS

Fig. 2. Disclosure of Sexual Orientation and Sexual Identity States.

situation leads in a milder form of identity conflict, which is called sexual identity strain. These three forms of sexual identity states as outcomes of disclosure disconnects in work and home domains are summarized in Fig. 2. While this general framework of identity and disclosure has been applied to other groups with hidden stigmatized identities (e.g. Ragins, 2004), two pragmatic distinctions need to be made for LGB workers. First, while a gay employee may not have disclosed his sexual identity to anyone in his work environment, unless he is celibate, it is unlikely that he will report that he has hidden his sexual identity from everyone in his home environment. Accordingly, the disclosure dimension for home uses the descriptor “few or none” rather than “none” in Fig. 2. Second, although the congruent sexual identity state involves individuals who have parallel identities in both work and home domains, sexual identity congruence may have different outcomes for LGB workers who have disclosed their identity in both domains, as compared with those who have hidden their identity in both domains. As discussed earlier, one key issue facing LGB employees who hide their identity is that they will be “outed” by others in their work or home environment. Therefore, hiding one’s sexual identity in both home and work domains carries a risk, and associated stress, that may not be experienced by those who have fully

Sexual Orientation In The Workplace

69

disclosed their identity in both domains. In order to depict the differing degrees of stress associated with these situations, sexual identity states associated with greater stress are shaded in Fig. 2. As indicated in that figure, sexual identity congruence involving disclosure in both work and home is depicted as having less potential for stress than the other identity states. In sum, many gay employees feel that they do not have control over the disclosure of their invisible stigmatized identities inside or outside the workplace. Those who are not totally out of the closet face the ongoing risk of being “outed” against their will, even if this is done accidentally and without malicious intent. In addition, many LGB workers experience a disconnection in the extent to which they have disclosed their sexual identity at work and at home. These disclosure disconnects may lead to feelings of loss of control over the disclosure process and may create sexual identity conflict, which in turn leads to stress and other negative work and life attitudes and outcomes. Two moderators affect outcomes associated with disclosure disconnects: identity salience and the size of the community. Individuals who define themselves in terms of their sexual orientation should experience greater sexual identity conflict as a consequence of disclosure disconnects than those who define themselves on the basis of other identities. Size of the community is also important, as small communities increase the risk of being “outed” and therefore amplify feelings of lack of control over the disclosure process. This situation is not faced by groups with visible stigmas (race, gender, visible disability) and is rarely faced by other groups with invisible stigmas (e.g. invisible disabilities, religion and class), as these invisible stigmas are usually not hidden from family members. In fact, as discussed later, individuals with these stigmas are more likely to get support from their family members than lesbians, gay men and bisexuals. The challenge of having an invisible stigma leads to related processes and outcomes for LGB workers. Let us now delve into related issues that face LGB workers as they grapple with the decision to disclose their sexual identity at work. Assumptions of Heterosexuality, Indirect Discrimination and Disclosure Backlash According to stigma theory, one outcome of an invisible stigma is that the individual is assumed to be a member of the majority group (Goffman, 1963; Jones et al., 1984). As applied to sexual orientation, gay employees who have not disclosed their sexual identity are often assumed to be heterosexual by co-workers, supervisors and clients, thus complicating the disclosure decision and their work relationships (Friskopp & Silverstein, 1996; Waldo, 1999; Woods, 1993). Although invisibility of stigma certainly contributes to the assumption of heterosexuality, as discussed earlier in this article, this assumption also reflects a social construction about sexual identity in organizations; heterosexuality is

70

BELLE ROSE RAGINS

assumed partly because heterosexuality is constructed as the “norm,” thus leaving other forms of sexual identity generally invisible (e.g. Creed & Scully, 2000). In other words, in the absence of countervailing “evidence,” most employees are assumed to be heterosexual. A blending of stigma and social constructivist perspectives would hold that the invisibility of sexual orientation fuels the social construction that leads to prevailing assumptions of heterosexuality in the workplace. Indirect discrimination. The assumption of heterosexuality creates two unique outcomes for LGB employees. First, it increases the occurrence and impact of “indirect discrimination.” Indirect discrimination involves discriminatory actions that have a negative impact on an individual even though the actions are not aimed directly at the individual. This contrasts with direct discrimination, which involves actions aimed at the individual target. Examples of indirect discrimination range from watching negative treatment of other openly gay co-workers to overhearing heterosexist comments and jokes. For LGB employees who have not disclosed their sexual identity, the likelihood of indirect discrimination increases by the fact that the perpetrators of these actions are unaware that their colleague is gay, and thus do not censor their comments or actions. In fact, in some cases, they may expect their gay colleagues to participate in these discriminatory actions. Simply put, invisibility of stigma contributes to assumptions of heterosexuality, which in turn increases the probability and intensity of indirect discrimination experienced by closeted LGB employees. Although other groups face indirect discrimination, the form and impact of indirect discrimination is different for LGB workers. For example, heterosexual women and employees of color may experience indirect discrimination by observing members of their own group face discrimination, but they are rarely expected to participate in these actions. Moreover, when they do overhear comments and jokes, sometimes the comments are really meant to be overheard, thus reflecting a different intent and a direct form of discrimination. In contrast, when closeted LGB employees hear anti-gay jokes, they know these comments are uncensored and thus reflect the true sentiments of their colleagues. The comments are a form of indirect discrimination in that they are not directly aimed at harming the gay colleague, because that individual is assumed to be heterosexual. The comments may even be given in a co-conspiratorial way, with the assumption that the gay colleague will agree with their intent and anticipated outcomes. Indirect discrimination is therefore a unique outcome of the invisibility of stigma, the decision not to disclose a gay identity at work, and co-worker’s assumptions of heterosexuality. It is reasonable to expect that many LGB employees will experience indirect discrimination by witnessing discrimination against

Sexual Orientation In The Workplace

71

others, but those who have not disclosed their sexual identity at work may witness more indirect discrimination than those who have disclosed their identity because comments and behaviors are less likely to be censored. In addition, the impact of indirect discrimination may be more intense for closeted LGB employees as they recognize that the perpetrator’s actions reflect their uncensored feelings and intentions. Proposition 6. The assumption of heterosexuality increases the prevalence and intensity of indirect forms of discrimination. Proposition 7. Indirect forms of discrimination will be more likely to be experienced by LGB employees who have not disclosed their sexual identity than those who have disclosed. In addition, individuals who have not disclosed their sexual identity will experience stronger reactions to indirect discrimination than those who have disclosed. Finally, indirect discrimination may affect the decision to disclose a gay identity at work. Witnessing negative treatment of other openly gay colleagues may push LGB workers deeper into the closet, thus perpetuating the cycle. Proposition 8. Experiences of indirect discrimination will be associated with the decision not to disclose a gay identity in the workplace. Disclosure backlash. The second outcome associated with the assumption of heterosexuality is that this assumption may force LGB employees to make disclosure decisions before they are ready. For example, a gay employee who is new on the job may want to “test the waters” before disclosing his identity to his heterosexual co-workers. The disclosure decision becomes salient when his co-workers assume that he is heterosexual. At this point, the gay employee has two choices. First, he may say nothing and not correct the misconception that he is heterosexual. He may even listen to anti-gay jokes without declaring his true identity. This may not only have a negative effect on his self-esteem (cf. Meyer, 2003), but it may also increase the negative reactions, or “disclosure backlash” from his co-workers once his true identity is revealed. Specifically, once his true identity is revealed, his coworkers may feel ashamed, misled, threatened or manipulated, and these feelings may result in strained or dysfunctional work relationships. Disclosure backlash therefore has both emotional and behavioral components that revolve around hurt, anger, withdrawal and possible retaliation. The gay employee’s second choice is to disclose his sexual identity as soon as he realizes his co-workers’ assumptions, irrespective of whether it is an opportune time or whether he feels comfortable sharing this personal part of his identity with his co-workers. Unfortunately, this

72

BELLE ROSE RAGINS

immediate disclosure may also yield behavioral backlash from co-workers, most commonly in the form of accusations that the gay employee is “flaunting his sexual orientation” (Friskopp & Silverstein, 1996; Woods, 1993). This backlash reaction to disclosure is based in part on prevailing assumptions of heterosexuality in organizations. For example, a heterosexual co-worker may reason that since he does not talk about his sexual identity in the workplace, his gay colleague’s disclosure of his gay identity is inappropriate, unwarranted and even unprofessional. However, this line of reasoning does not recognize the effects of assumptions of heterosexuality in the workplace; the heterosexual employee may not need to declare his sexual identity because assumptions about his identity are accurate. Disclosure backlash also reflects a much deeper issue on how sexual identity is viewed in organizations. In fact, the negative reaction to the disclosure of a gay identity at work has a foundation that has no real parallel for other stigmatized groups in organizations. Herek (1996) hypothesizes that negative reactions to disclosure are a function of the restricted perception and sexualization of homosexuality. He explains that: “Heterosexual relationships are generally understood as involving many components, but same-sex relationships are widely perceived only in sexual terms, even though they are very similar to heterosexual relationships in that they are primarily about love, affection, and commitment” (1996, p. 305). Herek proposes that public roles associated with heterosexuals (e.g. husband, wife, parent) are viewed as indicators of social duties rather than sexual conduct, and are therefore desexualized. He observes that: “Because of the desexualized nature of heterosexual roles, disclosures that identify one as heterosexual are not perceived as an inappropriate communication about private sexual conduct” (1996, p. 203). Consequently, heterosexual employees often make statements and provide affirmations about their heterosexuality at work, by wearing a wedding ring, discussing their family, and displaying family photographs, but these actions are not construed in sexual terms. However, LGB employees who engage in comparable activities are viewed as “flaunting their sexual orientation,” and are advised to “keep their sexual identity to themselves” (Friskopp & Silverstein, 1996; Woods, 1993), thus implying that they should hide their sexual identity, a prescription that perpetuates the assumption of heterosexuality. Disclosure backlash therefore occurs as a consequence of both concealing, and eventually revealing a stigmatized sexual identity. Disclosure backlash is the emotional withdrawal and negative reaction of co-workers, which can lead to direct forms of workplace discrimination. Proposition 9. Assumptions of heterosexuality increase the probability of disclosure backlash, which in turn may lead to direct forms of discrimination.

Sexual Orientation In The Workplace

73

As reviewed earlier, there are a number of detrimental outcomes associated with sexual orientation discrimination in the workplace (e.g. Button, 2001; Ragins & Cornwell, 2001a; Waldo, 1999). It is therefore reasonable to expect that indirect and direct forms of discrimination should increase the general life stress experienced by LGB workers (Meyer, 2003). Direct and indirect discrimination may also have an adverse effect on an array of work, life, and career attitudes and outcomes, including life satisfaction, job satisfaction, organizational commitment, turnover intentions, career satisfaction, advancement and compensation. Proposition 10. Indirect and direct discrimination will have an adverse effect on work, life, and career attitudes and outcomes. These relationships are summarized in Fig. 3. Summary of Invisible Stigmas Effects In sum, the invisibility of sexual orientation as a stigma creates a unique challenge for LGB workers. Key issues stemming from this challenge include a lack of control over the disclosure process, disclosure disconnects that involve differing degrees of disclosure of sexual identity at home and at work, the potential for sexual identity conflict, and assumptions of heterosexuality that increase the prevalence and intensity of indirect forms of discrimination and disclosure backlash. While LGB employees share many of the challenges faced by other groups with invisible stigmas, they face additional pressures and reactions stemming from an underlying core of assumptions about sexual identity and sex in the workplace. Let us now examine how the invisibility of sexual orientation as a stigma creates and compounds the second challenge faced by LGB workers: co-workers’ negative reactions to their sexual identity.

Challenge No. 2: Negative Co-Worker Reactions The second unique challenge facing gay, lesbian and bisexual employees is their co-workers’ negative reactions to their sexual identity. These negative reactions include direct forms of workplace discrimination and the disclosure backlash effects discussed earlier. In addition to direct forms of discrimination and backlash, negative co-worker reactions also include the avoidance and social isolation of LGB colleagues; heterosexual co-workers may refuse or avoid job assignments with LGB colleagues and may avoid social and work-related interactions within and outside work settings.

74

BELLE ROSE RAGINS

Fig. 3. Invisibility of Sexual Orientation Effects.

Although these negative behaviors may be common reactions to other minority groups in organizations, what is unique to LGB employees is the source of these behaviors. In this section, two unique sources of these negative reactions are presented: co-workers’ perception of sexual orientation as a choice and their emotional reaction to homosexuality.

Sexual Orientation In The Workplace

75

Perception of Sexual Orientation as a Choice According to stigma theory, individuals who are seen as having control over or being responsible for their stigma will face more negative reactions than those who are viewed as not being responsible for their stigma (Goffman, 1963; Jones et al., 1984), and existing research supports this hypothesis (see review by Crocker, Major & Steele, 1998). Despite mounting evidence that sexual orientation may not be a choice for gay men and lesbians (see: Friedman & Downey, 1994; Pillard & Bailey, 1995; see also Baumrind, 1995 and DeCecco & Parker, 1995 for a review of the issues involved in this debate), homosexuality is often viewed as a “lifestyle choice” or “sexual preference,” and this perception may lead to negative reactions to LGB workers. While most studies documenting negative responses to controllable stigmas have examined such stigmas as obesity, drug abuse and AIDS (cf. Weiner, Perry & Magnusson, 1988), one recent study found similar effects for sexual orientation; subjects who perceived homosexuality as a controllable choice were less willing to engage in a relationship with a lesbian than those who believed that homosexuality was beyond the individual’s control (King, 2001). Other research has found that beliefs about controllability of sexual orientation significantly influenced negative attitudes towards gay men and lesbians (Horvath & Ryan, 2003). By applying existing theory and research on controllable stigmas to the area of sexual orientation in the workplace, we can expect that perceptions of controllability of sexual orientation will affect co-workers’ reactions to their LGB colleagues. Proposition 11. The belief that sexual orientation is a choice will increase coworkers’ negative reactions to LGB colleagues. Although LGB employees are similar to other groups that are perceived to have control over their stigmas, such as obese workers and workers with some forms of disability (e.g. smoking-related illnesses, drug abuse, alcoholism), there is one factor that sets LGB employees apart from even these stigmatized groups. Specifically, many religious groups condemn homosexuality as a “lifestyle choice” that is morally wrong (Jung & Smith, 1993). These religious views not only reinforce the perception of sexual orientation as a choice, but since they amplify negative attitudes towards homosexuality (Herek, 1988; Horvath & Ryan, 2003; Hunter, 1991; Kirkpatrick, 1993), they may also result in workplace discrimination. Recent public opinion polls indicate that the perception of homosexuality as a morally wrong lifestyle choice is relatively widespread. In a recent poll conducted by the New York Times and CBS News of a national sample of 1,057 adults, half of the sample viewed homosexual relations between adults as morally wrong (Seelye & Elder, 2003). It is interesting to note that respondents who personally

76

BELLE ROSE RAGINS

knew gay people were less likely to view homosexuality as morally wrong than those who believed they had never met a gay person, which is aligned with other research linking personal contact to attitudes towards homosexuality (cf. Bowen & Bourgeois, 2001; Herek & Capitanio, 1996; Horvath & Ryan, 2003). Finally, the belief that LGB employees choose their sexual identity can lead to the perception that they are responsible for the outcomes associated with this choice, such as workplace discrimination and social isolation. Individuals who hold this view may reason that gay, lesbian and bisexual employees could avoid workplace discrimination simply by choosing to be heterosexual. This perspective amplifies the negative reactions of co-workers, which in turn creates negative workplace climates that may ultimately push LGB workers deeper into the closet. Employee attitudes towards LGB workers are affected not only by the perception that LGB employees can control their sexual orientation, but also by a complex and unique set of emotional reactions to homosexuality that are grounded in fear and threat. Let us now examine these emotional reactions. Emotional Reactions to Homosexuality: Stigma Threat and Aversive Heterosexism Stigma theorists hypothesize that feelings of fear and threat are common emotional reactions to individuals with stigmas (Blascovich, Mendes, Hunter & Lickel, 2000; Stangor & Crandall, 2000). Stigmas can cause perceivers to feel anxiety, discomfort or threat in social interactions (Crocker, Major & Steele, 1998; Goffman, 1963). Some stigmas elicit feelings of uncertainty and unpredictability among perceivers, and may even lead to acute perceptions of physical or psychological danger (Blascovich et al., 2000). These negative emotional reactions to stigmas may have behavioral consequences for the perceiver. Specifically, Blascovich and his colleagues hypothesized that the effort required to suppress negative emotional reactions to stigmatized individuals can decrease a perceiver’s performance (Blascovich et al., 2000). They found support for this stigma-threat hypothesis in a series of experiments designed to assess physiological responses to individuals with visible stigmas; subjects interacting with partners with facial birthmarks had cardiovascular responses consistent with threat and poorer performance (Blacscovich, Mendes, Hunter, Lickel & Kowai-Bell, 2001). Their research points to the potentially powerful effects of stigma on threat, fear and performance in the workplace. Tangible and symbolic threat. Threat may not only be an outcome of stigma, but it may also serve as its antecedent. Stangor and Crandall (2000) theorized that as social constructions, stigmas are often derived from threat. Although they identified many different forms of threats associated with stigma, two are directly relevant to sexual orientation: (1) tangible threat, which involves a threat to the health or

Sexual Orientation In The Workplace

77

safety of the perceiver; and (2) symbolic threat, which involves a threat to the perceiver’s moral or political views. These two forms of threat can be applied directly to the workplace experiences of gay employees. Tangible threat is reflected in the experience of gay male employees, who are often viewed as HIV positive, and report that bouts with the flu are assumed to be symptoms of AIDS (Friskopp & Silverstein, 1996; Woods, 1993). One consequence of this is that heterosexual colleagues avoid contact with gay and bisexual male employees for fear of contracting AIDS. Although education has dispelled many of the myths surrounding HIV and AIDS, the fear of AIDS is intertwined with the fear of homosexuality in a way that has no real parallel for other groups (cf. Herek & Capitanio, 1999). LGB employees may also elicit symbolic threat. As discussed earlier, the very concept of homosexuality is a threat to many employees’ religious beliefs (Jung & Smith, 1993). The symbolic threat of homosexuality is also reflected within our larger society. For example, gay partnerships have been characterized as “a threat to the institution of marriage” (Adam, 2003; Wetzstein, 2003) and the Vatican categorizes them as a direct “threat to society” (“Same-sex marriages a threat,” 2003). A constitutional amendment that prohibits gay marriages is currently being considered in the U.S., and recent public opinion polls reveal that Americans oppose gay marriage by a 61–34 margin (Seelye & Elder, 2003). One consequence of this is that some heterosexual employees may experience symbolic threat when faced with LGB co-workers who are out at work and wish to bring their partners to company social events. The symbolic threat of homosexuality may increase if gay marriages are legalized in some states, and LGB workers bring their spouses to workplace events. As discussed earlier, religious beliefs may affect co-worker’s perception of sexual orientation as a choice. These beliefs may also create a state of symbolic threat in the co-workers of LGB employees. Proposition 12. Employees’ religious beliefs predict their perception of sexual orientation as a choice and their experience of symbolic threat. Personal threat. Although tangible and symbolic threats may be present with other stigmatized groups, there is one additional threat that is unique to gay, lesbian and bisexual workers: “personal threat.” As developed here, personal threat involves heterosexual co-workers’ feelings that their sexual identity is threatened by the presence of LGB colleagues. Personal threat springs in part from homophobia, which is heterosexual’s fear that they are gay, may become gay, or may be perceived as being gay by others (Herek, 1984). This fear is unique to sexual orientation and is not found with most other stigmatized groups in organizations; employees usually

78

BELLE ROSE RAGINS

are not afraid that they may become or may be viewed as another gender or race by others in the organization. In support of the construct of personal threat, Herek (1984) contends that some heterosexuals develop negative attitudes to gay men and lesbians as a defense reaction that is rooted in deep-seated insecurities and fears about their own sexual orientation. Personal threat is grounded in homophobia, but there are subtle distinctions between the constructs. Personal threat involves a threat to an individual’s sexual identity. Although it can be argued that all homophobics experience personal threat, not all individuals who experience personal threat are fearful of homosexuality. In essence, the distinction between threat and fear (cf. Stangor & Crandall, 2000), while subtle, parallels the distinction between personal threat and homophobia. Individuals who experience fear usually feel threatened, but those who experience threat do not always experience fear. Aversive heterosexism. It is reasonable to expect that personal threat may lead to aversive heterosexism. Following in the form of aversive racism (Dovidio & Gaertner, 1986), aversive heterosexism is where unconscious homophobia leads individuals to behave in heterosexist ways, even though they profess and believe that they are not heterosexist (Winegarden, 1994). Aversive heterosexism is derived from a deep-seated, and usually unconscious fear of gay men and lesbians. Aversive heterosexists may avoid interacting with LGB colleagues and hold negative beliefs and stereotypes about them even as they publicly take a “not that there’s anything wrong with it” position. In some cases, aversive heterosexism may lead to direct forms of workplace discrimination. However, even though aversive heterosexists engage in direct forms of discrimination, they disavow, to themselves and others, that the discriminatory behaviors are motivated by heterosexism. In contrast, “traditional heterosexists” may engage in direct discrimination and are fully aware that their behaviors are driven by their heterosexist beliefs and attitudes. Although aversive heterosexism is similar to aversive racism in that both processes involve aversive reactions of threat and fear, aversive heterosexism is distinct because it is partly grounded in the individual’s subconscious fear of being gay. Aversive racism involves fear and a negative reaction to people of color, but it does not stem from the fear of actually becoming a person of color. Another important distinction is that aversive racism stems from social norms prohibiting overt acts of racism; aversive racists will not admit their racism to themselves or others because of social norms prohibiting expressions of racism. However, social norms prohibiting discrimination against gay men and lesbians are not as well developed as norms prohibiting direct displays of racism. For example, although national legislation prohibiting racial discrimination in the workplace was

Sexual Orientation In The Workplace

79

passed nearly 40 years ago, comparable legislation protecting LGB employees has yet to be passed (Herrschaft & Mills, 2002). Moreover, although social norms prohibit overt racism, most Americans still believe that homosexuality is morally wrong. As noted earlier, a recent national public opinion poll by the New York Times and CBS News concluded that half of all Americans continue to believe that homosexuality is morally wrong (Seelye & Elder, 2003), a finding that is aligned with earlier national public opinion surveys (Newport, 1998). In short, while aversive heterosexism shares some common attributes with aversive racism, it is unique in its reflection of personal threat, homophobia and societal sanctions of heterosexism. In sum, co-workers’ negative reactions to their LGB colleagues may be fueled by tangible, symbolic and personal threat. In addition, some negative reactions are a function of aversive heterosexism, a largely unconscious process that is uniquely predicted by personal threat. Personal threat may therefore affect co-worker reactions through the unconscious processes involved with aversive heterosexism. However, personal threat may also have a direct effect on co-worker reactions. For example, some heterosexual individuals may be fully aware that their LGB colleagues constitute a personal threat to their sexual identity, and may choose to deal with their discomfort by deliberately avoiding interacting or working with their LGB colleagues. Other individuals may be uncomfortable working with LGB colleagues, but because they are experiencing aversive heterosexism, they may be unaware of the source of this discomfort, and may even attribute their discomfort to other factors, such as the gay co-worker’s personality, communication style, mannerisms or dress. Finally, although homophobia is separate construct from personal threat, it contributes to the experience of personal threat. This discussion leads to the following propositions: Proposition 13. Co-workers’ experience of tangible, symbolic and personal threat will increase their negative reactions to LGB colleagues. Proposition 14. Personal threat is fueled by homophobia and leads to aversive heterosexism. Aversive heterosexism is predicted primarily by personal threat and will lead to negative co-worker reactions. Courtesy Stigmas and Personal Threat Compounding these threats is the process of stigma-by-association, or courtesy stigmas. Courtesy stigmas are stigmas received by merely associating with stigmatized groups (Goffman, 1963). Courtesy stigmas are amplified by the invisibility of sexual orientation (Herek & Capitanio, 1996); heterosexual employees who associate with openly gay colleagues may be assumed to be gay by others in the

80

BELLE ROSE RAGINS

organization. This aspect of courtesy stigmas is unique to LGB employees. As mentioned earlier, employees are usually not assumed to be of another race or gender and they are rarely concerned that they will be viewed as another race or gender by mere association with different gender or racial groups. Courtesy stigmas build on and contribute to personal threat. Heterosexual employees who experience personal threat may avoid interacting with LGB colleagues for fear that others will assume that they are gay because of courtesy stigmas. This assumption, in turn, amplifies their experience of personal threat. One consequence of this scenario is that LGB employees may become very isolated in the workplace; heterosexual employees may avoid socializing and even working with them, and this avoidance can have a deleterious effect on LGB workers’ productivity and careers. To make matters worse, LGB employees who have not disclosed their sexual identity at work may fear that their identity would be revealed if they associate with open gay colleagues, and may therefore avoid contact with members of their own social identity group. In short, personal threat and courtesy stigmas amplify social isolation effects for LGB employees who have disclosed their sexual identity as well as for those who remain in the closet. Proposition 15. Courtesy stigmas are both an outcome and an antecedent of personal threat. They are fueled by homophobia and lead to an avoidance of LGB colleagues. Existing research suggests that courtesy stigmas influence perceptions of both LGB and heterosexual employees. Neuberg and his colleagues examined stigmaby-association threats for individuals associating with gay men in a series of studies conducted on male college students (Neuberg, Smith, Hoffman & Russell, 1994). The subjects (i.e. observers) were presented with videotapes of conversations between two male friends, and were asked to give their impressions of these actors. Prior to viewing the videos, the subjects were given information profiles that varied the sexual orientation of one of the actors and created a condition in which the heterosexual actor was “well-regarded”: they were portrayed as similar or dissimilar to the subjects and were given high or average social status. As expected, the researchers found a strong stigma-by-association effect; heterosexual actors were viewed more negatively when they had homosexual rather than heterosexual friends. Surprisingly, manipulations of the heterosexual actor’s social status and similarity to the observer failed to lessen these stigmatization effects, in fact heterosexual actors who had high social status received particularly negative reactions when portrayed as having homosexual friends. Another interesting finding was that the homosexual actors were not destigmatized when portrayed as having heterosexual friends, and in fact faced stronger stigmatization effects.

Sexual Orientation In The Workplace

81

Specifically, the homosexual actors were perceived as untrustworthy and dishonest when they were portrayed as having high-status heterosexual friends who were similar to the observer. The researchers speculated that these effects were due to a defense reaction in the observer. As an aside, it is certainly reasonable to expect that personal threat was the underlying process responsible for these reactions. Although this study needs to be replicated in work settings, future research could test the view that LGB employees who interact with high-status heterosexual colleagues may not have their status upgraded, but may face negative reactions that are fueled by personal threat and aversive heterosexism. Combined, these negative reactions may not only lead to aversive organizational climates for LGB employees, but they may also affect LGB workers’ decisions to disclose their sexual identity at work. LGB employees often “test the water” before coming out at work and are unlikely to disclose their sexual identity if they believe that co-workers will be threatened by their sexual orientation. Proposition 16. Negative co-worker reactions leads to aversive organizational climates for LGB employees. Aversive climates affect the decision to disclose a gay identity, and influence LGB employee’s stress, work and career attitudes and outcomes. Summary of Negative Co-Worker Reaction Effects As summarized in Fig. 4, there are a number of unique factors that create negative reactions to lesbians, gay men, and bisexuals in the workplace. Perceptions of controllability of stigma, combined with emotional responses grounded in fear and threat, may increase negative reactions to LGB employees. Tangible, symbolic and personal threat may lead to overt forms of discrimination, and aversive heterosexism may lead to subtle, but potent forms of discrimination and exclusion. Courtesy stigmas may amplify negative emotional reactions to LGB workers and may increase their isolation. LGB employees who associate with high status heterosexual employees may be viewed with suspicion and may even face increased stigmatization and social isolation. These processes combine to create workplace climates for LGB employees that range from aversive to toxic. These organizational climates have a negative impact on the work, career and life attitudes and outcomes for LGB workers, and may clearly affect their decision to disclose their sexual identity in the workplace. The decision to disclose a stigmatized gay identity is compounded by a lack of support and more general challenges associated with development of a gay identity. Let us now turn to this final challenge that makes LGB employees unique from other stigmatized groups in the workplace.

82

BELLE ROSE RAGINS

Fig. 4. Antecedents and Consequences of Co-Worker Reactions to LGB Workers.

Challenge No. 3: Lack of Social Support and Challenges in Identity Development An important distinction between LGB employees and other minority groups is that gay identities are not acknowledged or identified at birth, but are discovered later in life. Although some LGB people recognize their sexual identity relatively early in life, for others the recognition and acceptance of a gay identity occurs much later in life (Garnets & Kimmel, 1991). Unlike other demographic identities, gay identities are therefore an “achieved” rather than an “ascribed” status and reflect a long-term developmental process (Weinberg & Williams, 1974). This process creates specific challenges for gay, lesbian and bisexual employees, and sheds light on the some of the workplace challenges they face, most pressing of which is their decision to disclose their sexual identity to some or all of their colleagues. This section begins with an examination of the factors contributing to the lack of support for the development of a gay identity. I then provide a brief review

Sexual Orientation In The Workplace

83

of the theories of gay identity development. This is followed by an examination of how these unique processes affect the workplace experience of LGB workers. Lack of Support for Identity Development Gay men, lesbians and bisexuals obtain less support in the development of their identities than other stigmatized groups (cf. Ragins & Wiethoff, in press). Herek (1996) observes that most gay men and lesbians were raised with the expectation that they would be heterosexual, and had to discover and construct their identity against a backdrop of parental and societal disapproval. Gay and lesbian children learn and internalize negative societal messages about homosexuality long before they recognize their own sexual identity (Garnets & Kimmel, 1993; Meyer & Dean, 1998). Their eventual recognition of their sexual identity is sometimes associated with “internalized homophobia,” which is the internalization of negative beliefs about homosexuality (Bohan, 1996). Internalized homophobia may lead to lowself esteem, self-hatred, self-destructive behaviors and the belief that one deserves discriminatory treatment (Bohan, 1996). These negative feelings may be amplified by the concealability of sexual orientation. For example, Frable, Platt and Hoey (1998) found that individuals with concealable stigmas reported lower self-esteem and more negative affect than individuals with visible stigmas, and that only the presence of similar others lifted the self-esteem and mood of individuals with concealable stigmas. It is clear that the acceptance of a gay identity can be a difficult hurdle for many LGB individuals, and this hurdle may be even more formidable for LGB youth. Lack of family support. Since families of gay men and lesbians are typically heterosexual, they are unable to provide the coping mechanisms necessary for gay youth to deal with their emerging stigmatized identity. Heterosexual parents are often unable to provide the role modeling and guidance necessary to support the sexual identity development of their gay and lesbian children. Lack of support by parents may have long-term effects on the identity development of adults; recent research indicates that LGB adults who had less parental support and parental acceptance of their homosexuality had more negative gay identities and were less likely to disclose their sexual identity than LGB adults who had parental support and acceptance (Mohr & Fassinger, 2003). It is important to note that while the family is a source of support for many people of color, families are often disrupted or even lost when gay children disclose their sexual identity (D’Augelli & Hershberger, 1993). Further isolation occurs when close friends are alienated or lost after disclosure, thus limiting peer support as a buffer against lack of family support (Savin-Williams, 1995). This lack of family

84

BELLE ROSE RAGINS

and peer group support is one of the key developmental differences between sexual orientation and other stigmatized group identities. LGB people not only receive less support from their families than other stigmatized groups, but the family itself may become a source of stress and conflict in their lives. In fact, one of the most difficult challenges faced by many LGB people is the decision to disclose their sexual identity to their families (D’Augelli, Hershberger & Pilkington, 1998). Existing research indicates that this decision often results in dire consequences for gay youth; close to half of gay adolescents who are targeted for violence because of their sexual orientation receive that violence from their own families (see reviews by Bohan, 1996; see also: D’Augelli, 1998; D’Augelli et al., 1998; Savin-Williams & Cohen, 1996). It is interesting to note that family responses to disclosure varies by the gender of the child; boys are more likely to be evicted from their homes while girls are more likely to be physically and verbally abused in the attempt to “make them heterosexual” (see review by Bohan, 1996; see also: Savin-Williams & Cohen, 1996). A large-scale national study of 1,925 lesbians revealed that lesbian girls face challenges not encountered by their heterosexual counterparts. Specifically, 37% of the lesbians surveyed had been physically abused as a child or adult, 32% had been raped or sexually attacked and 19% had been involved in incestuous relationships growing up (Bradford, Ryan & Rothblum, 1994). In this sample, 27% had disclosed to all family members and 19% had not disclosed to any family member. The stress associated with coming out during adolescence has contributed to the relatively high rate of depression and suicide among gay and lesbian youth (Bohan, 1996; Meyer, 2003; Savin-Williams & Cohen, 1996). Bohan (1996) observes that according to some estimates, suicide may be the number one cause of death among LGB teens. Lack of institutional support. Societal, religious and educational institutions provide little support for the development of a gay identity (Garnets & Kimmel, 1993). Unlike other racial and ethnic groups, gay history is rarely taught in school and is not passed on through family traditions, thus limiting the sense of community and pride in a gay identity. Schools not only provide limited support, but they are often the setting for hate crimes and violence against gay, lesbian and bisexual youth (Herek, Cogan & Gillis, 2002; Meyer, 2003). While many churches, synagogues and mosques are the primary sources of social support in the community, these religious institutions often either shun LGB members, or seek to “convert them” to heterosexuality. Since many people of color and individuals in the religious minority rely on religious institutions for community support, the removal of this source of support for LGB people in these minority groups represents an even wider gap in their social support network.

Sexual Orientation In The Workplace

85

The importance of community. Given the lack of support from family, friends and school systems, LGB people often seek out communities that provide them with support and assist them in developing and sustaining their sexual identity (cf. Frable, Wortman & Joseph, 1997). The “LGB Community” is a means of establishing contact with other LGB individuals, gives social support, and is an expression of LGB culture (Bohan, 1996). Bohan (1996) notes that these communities provide support in establishing a gay identity, are an antidote for feelings of isolation, and provide a buffer for antigay prejudice experienced at work and at home. The nature and size of communities vary by geographic area, but they are generally more likely to be found in urban, rather than rural settings. Although there has been relatively little research on this topic, a recent study of 97 LGB individuals revealed that involvement in the gay community was associated with reduced reports of depression, anxiety, and negative psychological symptoms (Smith & Ingram, 2004), findings that highlight the importance of community as a source of support. A Holistic Perspective on Lack of Support Effects Lack of support within and outside the workplace. The lack of support from family, friends and educational institutions creates unique challenges for LGB workers and has important implications for their workplace experiences. The lack of support for a gay identity has at least three immediate implications for the workplace. First, given the importance of community for many LGB workers, they may seek employment in locations that offer a community, and may decline job offers or transfers to settings that do not support a gay community. Community may therefore be a distinctive and important non-work factor that affects the career and employment decisions of LGB workers. Proposition 17. The presence of a supportive local LGB community will affect recruitment, retention and relocation decisions of LGB workers. Second, the lack of social support outside the workplace may decrease LGB workers’ ability to cope with a lack of support inside the workplace. Organizational scholars are increasingly aware of the fact that social support and the presence of positive, sustaining relationships outside the workplace helps employees deal with corrosive relationships, stress and conflict within the workplace (Dutton, 2003). Taking this holistic perspective, the support LGB employees receive outside the workplace may spillover into the workplace by equipping them with the emotional resources necessary to cope with their work environments. Similarly, a lack of social support outside the workplace may deplete the emotional resources needed to cope with negative co-worker reactions and aversive workplace climates.

86

BELLE ROSE RAGINS

In essence, an application of the social support spillover hypothesis to LGB workers holds that LGB workers receive social support from work and non-work domains, and that the presence or lack of support in one domain can bolster or deplete the individual’s ability to cope with the demands present in another domain (e.g. Ragins, 2004). Given the demands of the work environment reviewed earlier, LGB workers need more social support than their heterosexual counterparts, but have less social support available within and outside the workplace. As an emotional resource, social support is therefore more critical for LGB employees than many other workers since they are less likely to receive support from home or at work. Receiving social support from family, friends and community can help LGB workers cope with negative reactions of co-workers, and may also provide LGB workers with the emotional strength and resiliency necessary to disclose their sexual identity at work. Although the social support received in the workplace may spillover to the home domain, this effect may not be as strong or as likely to occur as the home-to-work spillover effect. For example, one would expect that support from family would bolster a gay employee’s ability to disclose his sexual identity at work, but support in the workplace may not go as far in helping him disclose his identity to his immediate family. The social support spillover effect is therefore asymmetrical for LGB employees. This effect is illustrated in Fig. 5. Future research could test this holistic perspective of social support, and examine the proposition that the presence of support for a stigmatized gay identity in one domain supports disclosure decisions and increases the individual’s ability to cope with lack of support in another domain. Proposition 18. Social support for a gay identity in one domain will affect disclosure decisions and the individual’s ability to cope with lack of support in another domain. This relationship will be stronger from home to work than from work to home. As an aside, although this social support spillover effect may apply to other stigmatized groups (e.g. Ragins, 2004), the effect should be stronger for these groups than for LGB workers. As discussed earlier in this section, LGB workers are less likely to get social support from family, friends and educational institutions than other stigmatized groups. Therefore other stigmatized workers generally have more social support resources at home than LGB workers and may be more likely to use these resources as a buffer against aversive workplace climates. To the extent that LGB workers receive less social support for their identity in the workplace than other groups, they may also have less emotional resources from work that can be applied to home demands. In short, compared to LGB groups, other minority groups may be more likely to employ home social support resources in

Sexual Orientation In The Workplace

87

Fig. 5. Social Support Spillover Effects for LGB Workers.

the workplace, and less likely to need work social support resources to bolster their emotional support at home. Future research could examine the potential for these differential relationships across stigmatized groups. Incorporating an historical perspective. Finally, the decision to disclose a gay identity in the workplace needs to be viewed within the historical context of the individual’s life experiences. As discussed earlier, many LGB workers have experienced loss of family and friends as a consequence of disclosing their sexual identity. In addition, many have experienced hate crimes simply because others assume that they are gay (Federal Bureau of Investigation, 2001; Herek, 1989). For

88

BELLE ROSE RAGINS

example, a study of 450 LGB adults revealed that a third experienced hate crimes based on their sexual orientation, and that perpetrators included not just strangers, but also neighbors, relatives and co-workers (Herek, Cogan & Gillis, 2002). Another study of 416 LGB adults found that three quarters of the respondents had been victimized because of their sexual identity, and that those who disclosed their identity were more likely to be victimized than those who did not (D’Augelli & Grossman, 2001). Clearly, LGB workers realize that much is at stake in their decision to disclose a stigmatized sexual identity in the workplace. Even the presence of supportive co-workers and supervisors may not fully offset a lifelong history of rejection, isolation and victimization as a consequence of disclosure. It is reasonable to expect that individuals who have prior experience being victimized on the basis of their sexual identity will be more fearful of disclosing their identity than those who did not have such experiences. However, this will not automatically lead to disclosure decisions; while LGB employees with a history of discrimination may be fearful of disclosing their sexual identity in new positions, they may also be unwilling to relinquish their identity and retreat back into the closet (Ragins & Cornwell, 2001b). Proposition 19. Prior experience with discrimination on the basis of sexual orientation will increase fears associated with disclosure to disclose a gay identity in the workplace. Many heterosexual co-workers may not fully understand their LGB colleagues’ reluctance to fully disclose their sexual identity because they have not had the same experiences; they have not lost close relationships or risked physical assault because of their heterosexual identity. As a consequence, heterosexual workers may view the caution with which their LGB colleagues approach disclosure decisions as a lack of trust or even paranoia. This lack of understanding may create social distance and even conflict in some work relationships. One remedy for this situation is to provide diversity training that gives heterosexual workers insights into the varied life experiences of their LGB colleagues. This training could also address some basic differences in the development of sexual identity between heterosexual and homosexuals, as well as the workplace implications of these differences. We now turn to a brief review of the gay identity development process, followed by an examination of how this developmental process affects the careers and workplace experiences of LGB workers. Theories of Gay Identity Development Psychologists have been theorizing about gay identity development for over twenty years, and this has yielded a number of theories and models on the topic (see

Sexual Orientation In The Workplace

89

for example: Cass, 1984; Coleman, 1982; Morris, 1997; Troiden, 1989). Most theories of gay identity development propose two initial developmental stages (cf. reviews by Gonsiorek & Rudolph, 1991; Sophie, 1986). The first stage involves a general awareness of same-sex attraction, which is often accompanied by feelings of confusion, defensiveness, fear and denial. The second stage involves emotional and behavioral testing and exploration. This may involve “coming out to oneself” or the self-labeling of oneself as gay, seeking out other gay people, and establishing the first gay sexual relationship. The theories of gay identity development differ after the second stage, but usually involve third and fourth stages that involve identity tolerance, acceptance, commitment and integration. Disclosure of sexual orientation to friends, family and co-workers is an integral component of these later stages of development. It is important to note that while some theories propose discrete and linear stages, other theories view identity as a cyclical process involving backtracking to earlier stages involving fear, defensiveness and even denial of same-sex feelings (Rivers, 1997; Sophie, 1986). Individuals differ in the timing of these stages; some gay men and lesbians start their identity process early in adolescence, whereas others may not come to terms with their sexual identity until much later in life (D’Augelli & Grossman, 2001). In fact, some individuals may enter the first stage of gay identity development after they have married a heterosexual and raised a family (Garnets & Kimmel, 1991). Some of these individuals report that they denied early feelings of attraction for those of the same gender, but others report that these feelings first emerged midway through their lives (Garnets & Kimmel, 1991). In addition to individual differences in this identity process, existing research indicates that gender plays a strong in this process: in general, women take longer to self-identity as gay than men (see review by Garnets & Kimmel, 1991). Women are more likely to view their sexual identity within the context of community and relationships, whereas men are more likely to establish their sexual identity in conjunction with early sexual exploration (Garnets & Kimmel, 1991; Gonsiorek, 1995). The development of a LGB identity may differ across individuals not only in terms of timing, but also in terms of outcomes. Bohan (1996) observes that there is tremendous individual variation in the development and manifestation of gay, lesbian, and bisexual identities. She observes that theories of gay identity development may not fully capture the uniqueness of individual experiences, and illustrates the diversity of this experience by asking whose experience we are trying to capture with these theories: “Is it the gay man who has ‘always known’ he was gay, or is it the elective lesbian who consciously chose this identity at midlife? Is it the woman whom others may refer to as bisexual but who self-identifies as lesbian despite her involvement in a heterosexual relationship, or is it the man

90

BELLE ROSE RAGINS

who was happily heterosexually married and raised children before falling in love with another man? Is it the man who has frequent sexual encounters with other men but steadfastly insists that he is not gay, or is it the two women who regard themselves as a lesbian couple but whose relationship is not sexual?” (p. 93). One insight gleaned from this quote is that the diversity of sexual identity may be greater among gay men and lesbians than heterosexuals. Heterosexuals rarely have to address these issues in the course of developing their sexual identity, in part because their identities are socially prescribed and accepted as “the norm” in society. This diversity in the development and manifestation of an LGB identity creates ambiguity and complexity in perceptions of the identity. Although sexual orientation is not visible, many LGB individuals report that when they came out to close friends and family members, these individuals reported being relieved, rather than surprised; evidently they knew or suspected because of certain behaviors, actions or other “cues” (Friskopp & Silverstein, 1996). Of course there are tremendous individual differences in the degree to which LGB individuals give intentional or unintentional “cues” about their sexual identity. There is also variation in whether these cues are perceived or interpreted correctly by others. Some LGB individuals may not be perceived as gay, despite giving a symphony of cues about their sexual identity; others may not yet recognize themselves as gay, even as those around them view them as gay. Still others may give no cues and may not be perceived as gay. Finally, some heterosexuals may mistakenly be perceived as gay due to misinterpretations of these cues. This discussion illustrates the complexities involved with developing a gay identity, and has important implications for the LGB workforce. Let us now turn to a closer examination of how sexual identity development processes affect LGB employees’ workplace experiences. Identity Development and Workplace Experiences An understanding of the general process of gay identity development provides insights into the processes involved with establishing a gay identity at work. To start, there is quite a bit of individual variation in the development of a gay identity in the workplace. LGB employees may be at various stages in the development of their sexual identity over the course of their careers; they may vary on the degree to which they are out at work, out to relatives or friends at home, or even out to themselves. As illustrated by Bohan (1996), there is tremendous individual variation not only in sexual identities, but also in the degree to which sexual identities correspond to sexual behaviors. The diversity in sexual identity among lesbian, gay and bisexual workers may not be immediately understandable to heterosexual workers because it is so different from their own experience in developing a heterosexual

Sexual Orientation In The Workplace

91

identity. Using their own identity as a frame of reference, heterosexual workers may view their LGB colleagues as one undifferentiated group, and therefore may not understand why a LGB co-worker does not fit neatly into the expected roles associated with a “gay identity.” Second, these theories predict that the process of establishing a gay identity in the workplace may not be smooth or linear, but may involve backtracking and retreats to earlier phases. Gay, lesbian and bisexual employees may “come out” to themselves, but then enter a state of denial in which they deny their sexual identity and retreat back into the closet. This not only creates a significant amount of stress for LGB workers, but may also create confusion among heterosexual colleagues who do not understand the conflict, stress and struggles that some of their LGB colleagues experience. Third, these theories suggest that LGB employees who are in the early stages of their identity development may deny their sexual identity, even as some LGB and heterosexual co-workers perceive them as being gay. This paradoxical situation may have a number of ramifications for the work environment. Heterosexual coworkers may not understand why their LGB colleagues are denying their sexual orientation, and this lack of understanding may create social distance. Another potential outcome of this situation is that LGB employees may face discriminatory and heterosexist behavior but may attribute this behavior to other causes if they are unaware or have not accepted their sexual identity. This perceptual disconnect can lead to the under-reporting of sexual orientation discrimination in the workplace. Of course these misperceptions can also lead to the situation in which heterosexual workers are perceived, and treated as gay by their colleagues, while remaining oblivious to the true reasons behind their co-workers’ treatment. Finally, the longitudinal development of a gay identity indicates that some gay men, lesbians and bisexuals may choose careers that are not receptive to their sexual orientation (e.g. the military, primary school education) before they discover or come to terms with their true sexual identity (Chung, 1995; Prince, 1995). This may create a career crisis in which they are forced to either hide their sexual identity or change their occupation, a career decision not faced by their heterosexual counterparts. This example points to the broader issue of the impact of a gay identity on the careers of lesbian, gay and bisexual workers, which is considered in depth in the final section of this article. Summary of Social Support and Identity Effects Lesbian, gay and bisexual workers receive significantly less support in developing their identities than other stigmatized groups in organizations. While family and friends provide important social support to other groups, LGB individuals are not only unlikely to get support from these sources, but also may face outright rejection,

92

BELLE ROSE RAGINS

shame and scorn from these relationships. Societal and educational institutions also provide little support for LGB identity development, and too often become the setting for rejection and hate crimes. This lack of social support for identity development has three implications for the workplace. First, given the lack of support from friends, family and educational institutions, the local LGB community becomes an important source of support for LGB workers. In fact, the supportive local community may have a direct influence on employment and career decisions for many LGB employees. Second, lack of support in one area may affect the individual’s ability to cope with demands in another area. The social support spillover perspective holds that the presence or absence of social support in one domain bolsters or depletes an individual’s ability to cope with demands in another domain. As applied to LGB workers, those who do not receive social support for their sexual identity at home may have less emotional resources that can be used to cope with demands and disclosure decisions in the workplace. Similarly, but to a lesser extent, LGB workers who receive social support in the workplace may be better equipped to cope with identity-related demands that occur outside of the workplace. The third outcome associated with these processes is that the cumulative life-long historical effects of lack of support, discrimination and victimization are carried with LGB workers into their current workplaces, and may affect their fears and decisions about disclosing a gay identity at work. The final implication of the gay identity process for the workplace is that LGB workers may come to terms with their sexual identity at various stages in their lives and careers. There is tremendous individual variation in how stages of gay identity development are entered as well as the outcomes associated with these stages. These unique social support and identity processes may have a significant effect on the careers of the LGB workforce, as discussed later in this article.

Summary of the Three Unique Challenges This section reviewed three challenges faced by LGB employees that are not faced by other stigmatized groups in organizations. Each of these challenges creates processes that are distinct to the LGB workforce. The first challenge stems from the invisibility of sexual orientation as a stigmatized social identity. There are two issues associated with the invisibility of this stigma. The first is whether LGB employees choose to disclose their sexual identity, and if so, the conditions under which this disclosure takes place. Disclosure of a gay identity is not an all-or-none phenomenon, but usually occurs on a continuum that takes place in work and non-work settings. This process of

Sexual Orientation In The Workplace

93

disclosure may lead to sexual identity disclosure disconnects, which involve differing degrees of disclosure of sexual identity at work and at home. These disconnects may result in feelings of loss of control over the disclosure process, and may create sexual identity conflict as individuals attempt to manage one, or even two sexual identities that are at various degrees of disclosure at work and at home. The size of the local community and the salience of sexual orientation as an identity are key factors that may moderate these relationships. Combined, these processes create significant stress and may negatively affect the individual’s work and quality of life. The second issue emerging from the invisibility of sexual orientation is that invisibility combines with organizational norms to create prevailing assumptions of heterosexuality in organizations. Unless they have disclosed their sexual identity, LGB workers are often assumed to be heterosexual by others in the organization. This complicates the disclosure decision and increases the prevalence and intensity of indirect discrimination. Indirect discrimination involves discriminatory actions that have a negative impact on an individual, even though the actions are not aimed directly at the individual. Because they are assumed to be heterosexual, closeted LGB employees may be more likely to overhear uncensored anti-gay jokes or comments, or may even be expected to participate in these actions. Since these actions reflect uncensored feelings, their impact on LGB co-workers is particularly intense and may push LGB workers deeper into the closet. The assumption of heterosexuality may also lead to disclosure backlash, a negative co-worker reaction that occurs once the LGB worker’s true identity is revealed. Disclosure backlash can be both emotional and behavioral, and may involve reactions of hurt, anger, withdrawal and retaliation. The second challenge faced by the LGB workforce is the potential for negative reactions among heterosexual colleagues. Negative reactions spring from unique causes that include threatened emotional states among co-workers and their perception that LGB workers are responsible for their stigma, a view that is often fueled by religious beliefs. Heterosexual co-workers may experience three forms of threat in reaction to their LGB colleagues: tangible threat, which involves fears associated with the contraction of AIDS, symbolic threat, which involves threat to one’s religious beliefs, and personal threat, which is a threat to one’s own sexual identity. Personal threat is grounded in deep-rooted sexual insecurities, and is fueled by homophobia, which is the fear of homosexuality. Personal threat is reinforced and contributes to courtesy stigmas, the stigma-by-association process in which individuals are assumed to be a member of a group because they associate with the group. Tangible, symbolic and personal threats all contribute to negative co-worker reactions, which include discrimination, disclosure backlash, avoidance and social isolation. However, unlike tangible and symbolic threat, personal threat

94

BELLE ROSE RAGINS

has a unique relation to aversive heterosexism, which is the unconscious fear and aversion to homosexuality that leads to discriminatory behaviors. These processes combine to create aversive workplace climates that have a negative impact on LGB workers and decrease their likelihood of disclosing their stigmatized sexual identity in the workplace. The third unique challenge facing LGB workers centers on the development of their sexual identity and the lack of social support they receive in developing their identity. LGB workers often receive little support for their identity development from family, friends or educational institutions. Their local LGB community therefore becomes a critical source of support for them, and may influence their employment and career decisions. In addition, a social support spillover effect may occur in which the presence or absence of social support in one domain may bolster or deplete the individual’s ability to cope with demands in another domain. The individual’s history of victimization, discrimination, and lack of support, may affect the fears associated with disclosure and the decision to disclose a gay identity at work. Finally, the longitudinal nature of the gay identity processes means that LGB workers may come to terms with their sexual identity at various stages in their lives and careers. These social support and identity processes may have a significant effect on the careers of the LGB workforce, as discussed later in this article. These three challenges do not occur in isolation, but interact to affect the workplace experiences of LGB workers. The three challenges and their associated processes are summarized in Fig. 6. Implications of Challenges There are two key implications of these challenges. First, the invisibility of stigma is a pervasive element that directly or indirectly contributes to each challenge. Invisibility underlies assumptions of heterosexuality, and this assumption, coupled with homophobia and aversive heterosexism, amplifies personal threat. Personal threat involves insecurity about sexual identity, an insecurity that is amplified by the fact that sexual identity is invisible. Invisibility also heightens the impact of courtesy stigmas on personal threat; stigma-by-association may be more likely to trigger deep-seated insecurities among heterosexual co-workers because sexual stigmas are inferred, rather than seen. Finally while invisibility of stigma does not create a lack of social support, it certainly fuels this problem. LGB workers are isolated within and outside the workplace by the fact that, on the surface, their identities are invisible. Their invisibility allows them to “pass” as heterosexuals, which may lead to identity conflict and further contribute to a lack of support for their true identity. Moreover, unlike groups with visible stigmas, groups with invisible stigmas do not challenge the assumptions, misconceptions and stereotypes

Sexual Orientation In The Workplace

95

Fig. 6. Summary of the Three Challenges and Related Processes.

held by majority members in organizations. Contact with stigmatized individuals may change majority members’ attitudes (Allport, 1954), but not when the stigma is concealable (Herek & Capitanio, 1996). Heterosexual co-workers may not be forced to confront their stereotypes about homosexuality if they mistakenly believe that all of the people they know are heterosexual. The second implication is that non-work factors play a critical role in LGB workers’ ability to face the workplace challenges associated with their sexual identity. Increasingly, organizational scholars are approaching work holistically

96

BELLE ROSE RAGINS

by assessing the reciprocal relationships between work and non-work. Although this is important for all employees, given the challenges relating to disclosure and support reviewed here, non-work may have an unparalleled effect on the workplace experiences of LGB workers. Non-work factors spill over into workplace in negative and positive ways; most tangibly through the fear of being unintentionally outed by well-meaning co-workers and friends, or through having local communities buffer and bolster LGB workers as they deal with workplace discrimination and the decision to come out at work. The impact of non-work factors on the workplace has become even more salient with the recent controversial and public debates over gay civil rights and gay marriages. These debates are not likely to end soon, and threatened reactions by heterosexual co-workers may increase as these issues spill from the pressroom to the lunchroom. Now that we have examined some of the unique workplace challenges facing LGB employees, let us turn to the broader issue of their careers. As mentioned earlier, the longitudinal nature of sexual identity development, coupled with the lack of support in work and non-work settings, results in specific career challenges for the LGB workforce. In brief, LGB workers face identity and resource-based challenges in their careers that are not faced by other minority groups. The next section builds on this discussion by examining the effects of these identity and support processes on the careers of LGB workers, and the limitations of existing theories for explaining LGB careers. This analysis will lead to the development of a theory of career processes for LGB workers.

THE IMPACT OF A GAY IDENTITY ON CAREERS This section examines some of the unique aspects of the careers of lesbian, gay and bisexual workers, and the limitations of existing theories for explaining their career experiences. Following this, a theory of careers that may more fully capture the career experiences of the LGB workforce is presented. Limitations of Existing Career Theories A number of scholars observe that existing theories of career development were developed from research conducted on heterosexual samples, and therefore fall short in capturing the unique processes affecting the careers of LGB workers (see for example: Dunkle, 1996; Fassinger, 1996; Lonborg & Phillips, 1996; Mobley & Slaney, 1996). In the section below, two processes that have been omitted from career development theories are examined: the role of discrimination in environment fit and the role of sexual identity in career development.

Sexual Orientation In The Workplace

97

Extending the Concept of Environment Fit: The Influence of Discrimination and Safe Havens Although many career theories address the issue of individual-career fit (cf. reviews by Greenhaus, Callanan & Godshalk, 2000; Sullivan, 1999), LGB workers face additional issues of fit that are associated with being in the sexual minority (Fassinger, 1995; Lonborg & Phillips, 1996). Since it is legal to discriminate on the basis of sexual orientation in most workplaces (Herrschaft & Mills, 2002), the choice of careers, organizations and positions becomes constrained for LGB workers (Morrow, Gore & Campbell, 1996). Consequentially, LGB workers may use vocational choice as a strategy to avoid workplace discrimination. Chung (2001) observes that LGB employees may choose to be self-employed, or self-select into organizations, industries or occupations that are gay-friendly in order to avoid workplace discrimination, and existing research provides some support for this idea (Chung & Harmon, 1994). These decisions reflect a broader process that may underlie perceptions of career or organizational “fit” among employees who are susceptible to workplace discrimination. As developed here, this “safe haven hypothesis” holds that career decisions of LGB workers are driven by their desire to find careers, occupations and workplaces that allow them to disclose their sexual identity and protect them from workplace discrimination. In addition, once an LGB worker finds a safe haven in a career, organization or work group, he or she may be reluctant to leave the situation, even for an alternative that offers more opportunities, greater compensation, or a better fit with career aptitudes and interests. In sum, it is proposed here that safe havens play a pivotal role in LGB workers’ decisions regarding entry and retention in a career, organization and position. The premise behind safe havens has virtually been ignored in theories of career and organizational choice (e.g. Greenhaus et al., 2000), and in related theories of employee recruitment and retention (e.g. Hom & Griffeth, 1995). These theories do not consider the impact of prior or anticipated work discrimination on employees’ job, organizational, or career choices. However, safe havens may play a crucial role in job and career decisions for individuals who are in the minority, particularly those who have faced workplace discrimination in the past. Although the construct of a safe haven applies to women and other minority groups, the lack of legal protection against sexual orientation discrimination, and their resulting vulnerability to discrimination, makes safe havens particularly important for LGB workers. In addition to providing a buffer against discrimination to LGB employees, safe havens provide the inestimable benefit of allowing LGB workers to be out at work. Perhaps it is here that safe havens have their strongest effect on retention. LGB workers may not be willing to go back into the closet, even

98

BELLE ROSE RAGINS

for substantial increases in career opportunities, job challenges or compensation that are offered by other positions, organizations or occupations that do not support their identity. Safe havens therefore become even more vital for LGB workers who are in integrated stages of their sexual identity development. Proposition 20. Safe havens affect LGB workers’ decisions about their careers, occupations, organizations and positions. The presence of a safe haven influences LGB workers’ entry and retention decisions, and this effect is amplified among those with prior experiences of workplace discrimination and those who are currently out at work. Now that we have looked at the role of safe havens in perceptions of environmental fit, let us turn to a second limitation of existing career theories: their inability to address the unique identity processes faced by LGB employees. As discussed earlier, identity processes represent one of the unique challenges facing LGB workers, and play a pivotal role in their workplace experiences. Let us now turn to examining the role of identity in LGB careers and the extent to which this role has been overlooked in existing theories of career development.

The Role of Identity in Careers A striking limitation of existing career theories is that they contain assumptions about identity that simply do not apply to many LGB workers (Chung, 1995; McCarn & Fassinger, 1996). As a background, many career theories view careers as a process that is nested within the broader realm of identity development, and hold that individuals select occupations based on a relatively stable sense of adult identity (e.g. Holland, 1985; Super, 1957). One problem with this perspective is that it assumes that sexual identity is stable and consistent over the course of an individual’s life (cf. Dunkle, 1996; Mobley & Slaney, 1996). Although this assumption holds for most heterosexuals, it is generally not the case for most LGB workers. As discussed earlier, the development of a gay identity is not stable or consistent for LGB employees. Many do not self-identify as gay until later in their careers, and the disclosure of their sexual identity is often a defining life event associated with conflict and stress (Meyer, 2003). Existing theories of career development do not address these issues, but take heterosexual identity as a stable “given” in the individual’s life, an assumption that needs to be examined in future research (cf. Tomlinson & Fassinger, 2003). LGB identity transitions often lead to disrupted and non-linear career paths for LGB workers, paths that are not explained by existing theories of career development. For example, one interview study of lesbians found that the process

Sexual Orientation In The Workplace

99

of discovering and integrating a lesbian identity delayed, disrupted and even derailed their careers (Boatwright, Gilbert, Forrest & Ketzenberger, 1996). These interviewees reported that their early career decisions were reexamined or even completely changed once they recognized and integrated their lesbian identities. In addition to the loss of career opportunities and the fears related to others discovering their sexual identity, they reported that internalized homophobia weakened their self-confidence and impeded the development of their professional self-esteem. Boatwright et al. (1996) noted that since coming to grips with a socially stigmatized identity is not an easy process, lesbians spend more time resolving emotionally charged identity issues than their heterosexual counterparts. Boatright and her colleagues (1996) made the important observation that existing theories of identity and adult life development (e.g. Erikson, 1968) assume that identity issues are resolved by early adulthood, thus allowing individuals to focus on career development issues. She observes that since the evolution of a lesbian identity extends into adulthood, this developmental process is often delayed for lesbians. Although her discussion focuses on lesbians, other qualitative research on gay men indicates similar career processes, with career paths of gay male professionals becoming derailed as a consequence of sexual identity processes involving awareness and the eventual disclosure of a gay identity (Woods, 1993). This research suggests that LGB workers not only face career barriers directly associated with discrimination, but may also face more subtle barriers associated with the disjointed timing of the development of their sexual identity with their career and life development. It is clear that we need to revisit established theories of career development in order to assess their adequacy for explaining the career processes of the LGB population. Because existing career theories are based on heterosexual theories of adult identity and life development, we may need to modify them or develop new theories in order to explain the unique career challenges faced by lesbian, gay and bisexual workers. Moreover, existing theories of career development do not take into account the powerful effect of workplace discrimination on the careers of LGB workers and other minority groups in organizations. Experiences with discrimination may lead to a re-assessment of career goals, and may play a powerful role in future career decisions regarding safe haven choices of occupations, organizations and positions. The next section seeks to address these limitations by proposing an identity-based, longitudinal theory of careers for LGB workers. An Identity-Based and Career Resource Theory of LGB Careers As reviewed earlier in this article, “lack of social support and challenges in identity development” is one of the three challenges LGB people face that explains unique

100

BELLE ROSE RAGINS

outcomes associated with their careers. While our prior discussion examined the effect of this challenge on workplace experiences and made only cursory reference to career implications, let us now undertake a deeper assessment of the effects of this unique challenge on the careers of the LGB workforce. Importance of Longitudinal Perspective Sexual identity processes require that a longitudinal lens be used when viewing LGB workers’ careers. A longitudinal approach is necessary because stages of sexual identity development often unfold over the course of their careers. As discussed earlier, some workers may realize their sexual identity early in their lives, but others may not come to terms with their sexual identity until much later in their careers. This means that some gay men, lesbians and bisexuals may choose careers that are not receptive to their sexual orientation (e.g. the military, primary school education) before they self-identify as gay (Chung, 1995; Prince, 1995). This may create a career crisis in which they are forced to either hide their sexual identity or change their occupation, a career decision not faced by their heterosexual counterparts. Sexual identity processes therefore play a key role in the careers of the LGB workforce. To review our earlier discussion on identity, the sexual identity development process goes through distinct longitudinal stages (cf. review by Gonsiorek & Rudolph, 1991). Although there are a number of different theories of identity development, and there has been much debate on the particulars of these theories (cf. Horowitz & Newcomb, 2001; Rivers, 1997; Sophie, 1986), Gonsiorek and Rudolph (1991) observe that a general theme underlying most theories is that there are four stages in the development of a gay identity: awareness, testing, acceptance and integration. The “coming out” process generally parallels these stages, with greater disclosure occurring during the acceptance and integration stages, although significant individual differences occur in the experiences, timing and outcomes of these stages (Sophie, 1986). LGB workers may therefore vary in the stages of their identity development over the course of their careers. This may influence the salience of their sexual identity, their need to disclose their identity, and their need to integrate their sexual identity into their careers. In addition, as discussed earlier, LGB workers differ from other groups in the lack of support they receive in developing their gay identity. While other minority groups receive social support for their identity from family and friends starting at an early age, LGB workers receive little support and often face rejection from these groups as they transition through the stages of their sexual identity development. Moreover, the presence or absence of general social support is often contingent on the unfolding of their gay identity. LGB workers may lose friends, family and coworkers as a consequence of disclosing their sexual identity. In addition, societal

Sexual Orientation In The Workplace

101

and educational institutions provide little support in the development of their sexual identity. As a consequence, LGB people may seek out new friends and an LGB community that supports their emerging sexual identity. In brief, the presence or absence of social support for their sexual identity may vary over the course of LGB workers’ careers as a consequence of their stages of identity development and their disclosure of a gay identity. In sum, there are two primary factors that make LGB workers’ careers different from other minority groups: their internal sexual identity processes and the external lack of social support available for their identity development. An inclusive theory of career development should therefore include these factors and examine how they influence LGB workers’ careers over the course of a lifespan. Let us now turn to the particulars that need to be included in such a theory. Accumulation of Career Resources Similar to longitudinal career theories that explain gender differences in the accumulation of power and career resources over the course of a lifespan (Ragins & Sundstrom, 1989), so might a longitudinal theory of differential access to career resources explain how the accumulation of different life and career experiences results in restricted career outcomes among the LGB workforce. Existing research and theory indicates that gender affects the long-term accumulation of social and career capital (Lyness & Thompson, 1997; Ragins & Sundstrom, 1989; Tharenou, Latimer & Conroy, 1994); along parallel lines, it is reasonable to expect that sexual orientation has incremental and significant effects on early, mid and late career stages and advancement. In fact, the limited research in this area suggests that differential access to career resources begins early in life; gay and lesbian students report less support and guidance in their academic and career decision-making than their heterosexual counterparts (Nauta, Saucier & Woodard, 2001). It should be noted that these effects extend above and beyond the effects associated with other group memberships, such as gender and race. Like other groups, the LGB workforce encounters discrimination that leads to differential access to career resources and career tracking, but their identity process and their lack of social support within and outside the workplace create profound and unique cumulative effects on their careers. Career resources of “safe havens” and “identity supporters”. Broadly speaking, career resources involve not only the accumulation of social and career capital, but for LGB workers they include the development of relationships and environments that support their sexual identity and buffer them against workplace discrimination. Given the previous discussion, two types of career resources that are particularly

102

BELLE ROSE RAGINS

unique to LGB workers are the presence of “safe havens” and “identity supporters.” As discussed earlier, safe havens are occupations, organizations and work groups that provide protection against discrimination and support the development of a gay identity. Identity supporters are more specific and involve family, friends, developmental relationships and communities that support the development of a gay identity. Although safe havens may include identity supporters, safe havens represent a broader reflection of organizational, occupational and work group climate. In contrast, identity supporters are more limited in scope and refer primarily to relationships that support the development of a gay identity. Given this discussion, the following relationships are proposed: Proposition 21. Sexual orientation has an independent, incremental and cumulative effect on the development of career resources over the course of the LGB worker’s lifespan. Proposition 22. Sexual orientation affects careers through internal and external events that unfold over the course of the LGB worker’s career. Internal events involve the development of a gay sexual identity and external events involve career resources associated with safe havens and identity supporters. The importance of social support was discussed earlier in this section, but let us now examine how identity affects the development of identity supporters for LGB workers. Reciprocal Relationships of Identity and Career Resources Identity processes and career resources do not operate independently, but reciprocally influence one another in influencing the careers of LGB workers. Identity processes affect two primary career resources: developmental relationships and safe-havens. Identity and developmental relationships. The disclosure of a gay identity may directly affect LGB workers’ ability to obtain a critical career resource: developmental relationships that directly influence their career choices and outcomes. These developmental relationships include mentors, peer relationships, networks, and career counselors. Existing research indicates that developmental relationships, particularly those involving mentors, are a principal predictor of career satisfaction and success (see reviews by: Allen, Eby, Poteet, Lentz & Lima, 2004; Noe, Greenberger & Wang, 2002; Ragins, 1999; Wanberg, Welsh & Hezlett, 2003). While developmental relationships are important for all workers, they are vital for LGB workers. As stressed throughout this article, LGB workers face substantial

Sexual Orientation In The Workplace

103

and unique challenges in the workplace, and need special support navigating discriminatory organizational terrains and developing career strategies for managing their stigmatized identity. While developmental relationships are vital for LGB workers, they may encounter a number of barriers developing these career resources. LGB workers who disclose their sexual identity to heterosexual mentors, peer networks and career counselors risk eliciting the personal threat reactions discussed earlier, and may lose these important relationships. However, non-disclosure of a gay identity means that the LGB worker may not get the advice and support necessary to manage their stigmatized sexual identity. For example, heterosexual mentors who are unaware of their prot´eg´es’ gay identity may advise them to seek careers or positions that are hostile to LGB workers, and may not understand when their prot´eg´es do not take their advice. However, even the disclosure of a gay identity does not guarantee appropriate advice if the mentor is unaware of the unique workplace challenges facing LGB people. Finally, like other diversified mentoring relationships (Clutterbuck & Ragins, 2002; Ragins, 1997), relationships involving sexual orientation diversity may be limited in providing role modeling, friendship and other psychosocial support functions. Identity and safe havens. The second impact of identity on career resources involves the saliency and importance of safe havens. As mentioned earlier, safe havens are a critical career resource for LGB employees; they represent a refuge from discrimination that allows them to disclose their sexual identity. As career resources, safe havens may be found in occupations, organizations and work groups. As proposed earlier, the importance of a safe haven is amplified among LGB employees who are out at work, thus reflecting the influence of sexual identity processes on this important career resource. Career resources may also be instrumental in the development of a gay identity. As discussed earlier in this article, the presence of a supportive LGB community is a critical factor that may help LGB individuals accept and integrate their sexual identity. Although LGB workers may lose friends, colleagues and developmental relationships when they disclose their sexual identity, they may also seek out supportive allies who accept their identity and help them reach integration stages of identity development. Finally, occupations, organizations and work groups that provide safe havens are likely to have a critical mass of openly gay workers, who may provide needed role modeling and support in the development and integration of a gay identity at work and at home. Therefore, career resources and identity development not only have independent effects on the career processes of LGB workers, but also jointly influence one another as they affect these processes.

104

BELLE ROSE RAGINS

Proposition 23. Stages of identity development have a reciprocal impact on the presence of identity supporters and the importance of safe havens. Summary. In sum, an identity-based career resource theory of career development for LGB workers predicts that the effects of sexual orientation on careers is not a function of a single event, but is due to an incremental accumulation of internal and external events that unfold over the course of the individual’s career. Internal events involve the development of a gay identity and the degree to which this identity process coincides with career and adult life development stages. External factors reflect the availability of career resources. Career resources have two components. First, they involve identity supporters who support and sustain the worker’s development of a gay identity. Identity supporters may involve family, friends, developmental relationships and the overall community. The second career resource is the presence of a safe haven, which may occur in occupations, organizations or work groups. Safe havens protect LGB workers from discrimination and support the development their sexual identity through fostering positive workplace climates. Finally, these internal and external factors operate conjointly to affect the careers of LGB workers. Now let us now turn to an examination of how these internal and external factors affect the career stages of LGB workers. The Recursive Nature of LGB Careers Stages Like other employees, LGB workers can be expected to go through some basic stages in the development of their careers. An integration of theories of career development holds that individuals go through three linear career stages (cf. review by Greenhaus et al., 2000). The first stage involves the individual’s general awareness of themselves and their environment. This awareness includes individuals’ assessment of the general fit between their aptitudes, skills and interests and their perception of the requirements of various occupations and career choices. This stage is logically followed by second stage involving occupational choice and a third stage involving organizational choice. Although this general framework captures the career experiences of majority members of the workforce, it does not fully capture the recursive and non-linear career experiences of workers who face career challenges associated with stigmatized and minority group memberships. Accordingly, this three-stage framework of careers is modified here to include additional processes associated with the outcomes of organizational choices, the potential encounter of discriminatory environments, and the subsequent reassessment of career and organizational choices. These modifications may more adequately capture the non-linear and recursive career paths reported by LGB workers discussed earlier (e.g. Boatwright et al., 1996; Chung, 1995; Woods, 1993).

Sexual Orientation In The Workplace

105

Fig. 7. An Identity-Based Longitudinal Theory of LGB Careers.

By including the role of discrimination in career processes, we need to consider discrimination-related outcomes that are associated with organizational choice and the effects of these outcomes on the individual’s career. As displayed in Fig. 7, organizational choice may lead to two outcomes relating to discrimination. First, organizational choice may lead to positive career outcomes and career satisfaction. Career satisfaction may stem from choosing an organization that offers a supportive diversity climate and does not discriminate on the basis of sexual orientation.

106

BELLE ROSE RAGINS

Organizational choice may also lead to negative career experiences. In this case, the LGB worker has chosen an organization that does not support a gay identity. The organization may have a “don’t ask, don’t tell” climate that expects LGB workers to hide their sexual identity, or may openly discriminate against LGB workers (e.g. Chung, 2001; Herek, 1996; Peel, 2001). LGB workers may also experience career dissatisfaction if the organization has a “lavender ceiling.” Lavender ceilings, which are the equivalent of glass ceilings, reflect barriers that restrict the progress of openly gay workers (Baker, Strub & Henning, 1995). The lack of openly gay, lesbian or bisexual high-ranking managers and executives may indicate to LGB workers that there is no future for them in their organization. In fact, the lack of openly gay managers, administrators or executives in their occupation may send similar signals to them about their future career prospects in their selected occupation. The stage of identity development not only affects the choice of an organization, but may also moderate the relationship between organizational choice and career satisfaction. LGB workers who are at early stages of identity development, are not out to themselves or to others, or may not feel the need to disclose their sexual identity at work, may choose organizations based on criteria unrelated to their sexual identity, and may therefore experience career satisfaction even in organizations that do not support a gay identity. Along similar lines, LGB worker who are at integrated stages of identity development and want to be out in their workplace may select an organization based on that criterion, but will experience heightened career dissatisfaction if the organization fails to meet their expectations regarding support for a gay identity. In support of this idea, LGB workers have been found to express greater dissatisfaction with inequitable work environments when they had a strong gay identity than when they have a weak gay identity (Button, 2001). The career dissatisfaction experienced as a result of discriminatory environments may lead some LGB workers into a recursive state in which they reassess their needs, career goals, organizational choices and occupational choices. However, this reassessment stage may occur independent of the outcomes associated with organizational choice. In particular, the stage of identity development may have a direct effect on this reassessment; LGB workers who were satisfied in their careers may face an abrupt career shift as their sexual identity and the potential fit between their identity and their environment changes. LGB workers in identity transition may therefore experience recursive stages of self-environment fit, occupational choice and organizational choice. In essence, as stages of identity development unfold, LGB workers who were in the closet may no longer view this is an acceptable choice, and this decision may prompt a broader reassessment of their values, priorities, choices and careers. As depicted in Fig. 7, the reassessment stage involves a recursive reconsideration of the individual’s self-environment fit, their choice of occupations and their choice

Sexual Orientation In The Workplace

107

of organizations. Reassessment may also lead LGB employees to seek safe havens that support their sexual identity. Individuals who have experienced negative workplace experiences may explicitly seek occupations, organizations and work groups that provide supportive cultures. In turn, as discussed earlier, the experience of working in safe havens affects subsequent career choices and decisions. LGB workers who have experienced a work environment that supports their sexual identity may refuse to “lose ground” by making a career move that requires them to go back in the closet. For example, anecdotal reports indicate that some openly gay workers disclose their sexual identity during job interviews to assess the reaction of the interviewer (Woods, 1993), a behavior that can be viewed as a safe haven assessment. Finally, the presence of a safe haven increases the opportunity for developing identity supporters. As discussed earlier, safe havens may attract LGB workers. In addition, it is reasonable to expect that heterosexuals who work in safe havens that foster supportive diversity climates will be more accepting of LGB sexual identities, and may be more likely to serve as identity supporters, than those who work in organizations that are hostile to LGB workers. This discussion leads to the following propositions: Proposition 24. Reassessment of career and organizational choices will be prompted by sexual identity development or by negative career experiences stemming from organizational or occupational cultures that are aversive or limiting. Proposition 25. Reassessment will facilitate seeking career resources involving safe havens and identity supporters. These career resources in turn will affect occupational and organizational choices. In sum, this longitudinal identity-based theory of career development holds that occupational and organizational choices for LGB workers are affected directly by internal factors involving identity development, as well as by external career resource factors involving identity supporters and the presence of safe havens. In addition, the theory includes a reassessment stage that takes into account two primary factors that create non-linear recursive career paths for LGB workers: experiences of workplace discrimination and transitions in sexual identity development. Summary of the Effect of a Gay Identity on Careers This section reviewed the limitations of existing theories for explaining the career development of LGB employees and developed an identity-based, career resource theory of LGB careers.

108

BELLE ROSE RAGINS

A critical analysis of career development theory revealed that existing theories omit two factors that play a central role in the careers of LGB workers. First, existing theories do not take into account the effects of workplace discrimination on the careers of LGB workers and other minority groups. Anticipated or prior experiences with discrimination may lead to a re-assessment of career goals and “safe haven” career decisions. Safe havens are occupations, organizations and workgroups that support the disclosure of a gay identity and are a refuge from discrimination in the workplace. Safe havens play an important role in occupational and organizational choices, as well as retention decisions among LGB workers. LGB workers may not leave safe havens, even for alternatives that offer more opportunities and compensation, if those alternatives require them to go back into the closet. The safe haven effect is therefore amplified among those who have experienced workplace discrimination in the past and among those who are currently out at work. The second limitation of existing career theories is that they are based on heterosexual theories of adult identity and life development, and therefore include assumptions of identity that do not apply to LGB workers. These theories assume that identity is stable and consistent, and that identity issues are resolved by early adulthood, thus allowing individuals to focus on the development of their careers. However, unlike their heterosexual counterparts, stages of identity development unfold over the course of careers for LGB workers, often leading to disrupted, recursive and non-linear career paths. Existing theories of career development therefore fall short in capturing the central role of sexual identity development in LGB workers’ careers. An identity-based theory of LGB careers was developed in order to address these limitations. This theory addressed two unique factors embedded in LGB workers’ careers: their internal sexual identity processes and the external lack of social support available for their identity development. This theory incorporated a longitudinal perspective and held that the effect of sexual orientation on LGB careers is not a function of a single event, but is due to an incremental accumulation of internal and external events that unfold over the individual’s lifespan. Internal events involve the development of a gay identity and the degree to which stages of sexual identity development coincide with career and adult life development stages. External events involve career resources associated with safe havens and identity supporters. Identity supporters are individuals, groups and communities that support a gay identity. Safe havens reflect broader occupational and organizational climates that provide protection against workplace discrimination. These internal and external factors operate conjointly to affect the careers of LGB workers.

Sexual Orientation In The Workplace

109

A general framework of careers was modified to include career processes associated with discriminatory environments and the resulting reassessment of career and organizational choices. LGB workers may reassess their careers as a consequence of transitions in their sexual identity or because they encountered aversive or unsupportive organizational environments. Reassessment involves a recursive reconsideration of the individual’s self-environment fit, choice of occupations, and choice of organizations. It also leads to the selection of safe havens, which increase the probability of obtaining identity supporters. Identity supporters and safe havens in turn facilitate identity development processes and the ultimate disclosure of a gay identity. Although the present theory was developed to explain the unique career processes of LGB workers, its underlying structure can be applied to other stigmatized and minority groups in the workforce. In particular, the ideas of safe havens and the reassessment of careers as a consequence of discrimination are applicable to other minority groups in the workforce. In fact, it is hoped that this theory will provide guidance not only for research on LGB careers, but also for research and theory on other stigmatized groups in the workplace. We need new theories that capture the powerful effects of stigmatization and workplace discrimination on career processes and outcomes. As discussed earlier, existing career theories were developed on heterosexual samples and consequentially contain assumptions about sexual identity that severely limit their applicability to the LGB workforce. Similar limitations hold for other groups. Management scholars point out that since career theories were developed on white male samples, they contain assumptions relating to gender (Sullivan, 1999) and race (Thomas & Alderfer, 1989) that limit their applicability to these groups, and existing research supports this idea (see review by Sullivan, 1999). Although some inroads have been made in developing theories that examine the effects of gender on careers (Powell & Mainiero, 1992), comparable theories need to be developed for race, and other stigmatized identities involving disability, class, religion, physical appearance and obesity. Finally, it is hoped that future research will examine not only the effects of sexual identity, but also the cumulative effects of multiple stigmatized identities on career processes and workplace experiences. Individuals have multiple identities, and these identities may become more or less salient, to themselves and others, in different organizational and work group environments (e.g. Ragins et al., 2003). We need to understand the cumulative and interactive effects of race, gender, sexual orientation, visible and invisible disability, and other group memberships on work and career experiences.

110

BELLE ROSE RAGINS

SUMMARY AND CONCLUSIONS This article identified the unique challenges facing the “invisible workers” of the workplace: lesbians, gay men and bisexuals. Unlike other stigmatized identities, a gay identity is invisible. As we discovered, the invisibility of their identity leads to complex and pervasive challenges in their workplaces, homes and lives. While LGB workers have the choice not to disclose their identity, they also face the ongoing fear that others will make that choice for them. With each new relationship they face the decision of disclosure and the potential backlash associated with that decision. As a consequence, some LGB workers decide to maintain different identities at work and at home, but this decision has its costs; it leads to disclosure disconnects, fear of losing control over the disclosure process, identity conflicts and a grinding state of stress. To compound this challenge, heterosexual co-workers often assume that their LGB colleagues are also heterosexual. This leaves LGB workers in a “no-win” situation. If they go along with that assumption and remain in the closet, they are unable to bring their full identity to work and face the ongoing risk of being discovered. However, an inopportune disclosure of their true identity may lead to backlash, retaliation and social isolation. The reaction of heterosexual co-workers complicates this situation considerably. Independent of their behavior or disclosure decisions, LGB workers realize that their very existence threatens their co-workers in deep, personal and sometimes subconscious ways. Stigma threat, combined with homophobia, conservative religious beliefs, and aversive heterosexism can form a volatile mixture that ignites co-worker reactions ranging from social banishment to workplace violence. The social isolation that LGB workers face in the workplace is aggravated by the lack of social support they receive at home. Disclosure decisions have steep costs; friends, family and developmental relationships are lost, even as supportive communities and allies are found. It is encouraging that social support may spillover, in that support received in one domain may provide an emotional buffer against demands made in another. Underlying all of these issues is the core challenge of developing a gay identity. For many, the process of discovering, accepting, disclosing and integrating a gay identity is a life-challenge unlike any other. There is tremendous individual variation in how and when a gay identity is developed and the extent to which it is disclosed in work and home domains. The process is complex and can be enormously stressful, particularly when disclosure decisions can result in social isolation, rejection, job loss and hate crimes. For many, this challenge has a singular and prevailing influence on their work, career and life satisfaction. Safe havens and identity supporters play critical roles in helping LGB workers meet this challenge,

Sexual Orientation In The Workplace

111

but access to these career resources is often limited. Ultimately, it is clear that for the LGB workforce, the process of developing a gay identity leads to unique and challenging career and work experiences that simply are not faced by other groups in the workforce. The complexities faced by lesbians, gay men and bisexuals are amplified for transgendered workers. Their identity process differs from LGB individuals, and their decisions regarding disclosure are considerably more complex (cf. Namaste, 2000; Roen, 2002). For example, transgendered individuals may self-identify as heterosexual, even as others view them as gay or lesbian. In addition, they face a lack of support not only from heterosexuals, but also from many in the LGB community. Although the career and workplace challenges facing the transgendered workforce goes beyond the scope of this article, the issues they face could easily consume an entire article. In conclusion, LGB workers constitute one of the largest minority groups in our workforce, yet remain the “invisible stigmatized citizens” in our workplaces and communities. This treatment results in tremendous personal, social and economic loss. For our workforce and communities to prosper, we need to make this invisible group visible by creating safe havens that welcome, value and protect LGB workers.

REFERENCES Adam, B. D. (2003). The defense of marriage act and American exceptionalism: The “gay marriage” panic in the United States. Journal of the History of Sexuality, 12(2), 259–277. Allegretto, S. A., & Arthur, M. M. (2001). An empirical analysis of homosexual/heterosexual male earnings differentials: Unmarried and equal? Industrial and Labor Relations Review, 54, 631– 646. Allen, T. D., Eby, L. T., Poteet, M. L., Lentz, E., & Lima, L. (2004). Career benefits associated with mentoring for prot´eg´es: A meta-analysis. Journal of Applied Psychology, 89, 127–136. Allport, G. W. (1954). The nature of prejudice. Cambridge: Addison-Wesley. Badgett, L. (1995). The wage effects of sexual orientation discrimination. Industrial and Labor Relations Review, 48, 726–739. Badgett, L. (1996). Employment and sexual orientation: Disclosure and discrimination in the workplace. Journal of Gay and Lesbian Social Services, 4, 29–52. Badgett, L. (2001). Money, myths and change: The economic lives of lesbians and gay men. Chicago: University of Chicago Press. Baker, D. B., Strub, S. O., & Henning, B. (1995). Cracking the corporate closet. New York: Harper Books. Baumrind, D. (1995). Commentary on sexual orientation: Research and social policy implications. Developmental Psychology, 31, 130–136. Beehr, T. A., & Bhagat, R. S. (1985). Human stress and cognition in organizations: An integrated perspective. New York: Wiley.

112

BELLE ROSE RAGINS

Black, D., Makar, H., Sanders, S., & Taylor, L. (2003). The earnings effect of sexual orientation. Industrial and Labor Relations Review, 56, 449–469. Blascovich, J., Mendes, W. B., Hunter, S. B., & Lickel, B. (2000). Stigma, threat, and social interactions. In: T. Heatherton, R. Kleck, M. Hebl & J. Hull (Eds), The Social Psychology of Stigma (pp. 307–333). New York: Guilford Press. Blacscovich, J., Mendes, W. B., Huner, S. B., Lickel, B., & Kowai-Bell, N. (2001). Perceiver threat in social interactions with stigmatized others. Journal of Personality and Social Psychology, 80, 253–267. Blandford, J. M. (2003). The nexus of sexual orientation and gender in the determination of earnings. Industrial and Labor Relations Review, 56, 622–642. Blumstein, P., & Schwartz, P. (1977). Bisexuality: Some social psychological issues. Journal of Social Issues, 33, 30–45. Boatwright, K. J., Gilbert, M. S., Forrest, L., & Ketzenberger, K. (1996). Impact of identity development upon career trajectory: Listening to the voices of lesbian women. Journal of Vocational Behavior, 48, 210–228. Bohan, J. S. (1996). Psychology and sexual orientation: Coming to terms. New York: Routledge. Bowen, A. M., & Bourgeois, M. J. (2001). Attitudes toward lesbian, gay, and bisexual college students: The contribution of pluralistic ignorance, dynamic social impact, and contact theories. Journal of American College Health, 50, 91–96. Bradford, J., Ryan, C., & Rothblum, E. D. (1994). National lesbian health care survey: Implications for mental health care. Journal of Consulting and Clinical Psychology, 62, 228–242. Button, S. (2001). Organizational efforts to affirm sexual diversity: A cross-level examination. Journal of Applied Psychology, 86, 17–28. Cain, R. (1991). Stigma management and gay identity development. Social Work, 36, 67–73. Cass, V. C. (1984). Homosexual identity formation: Testing a theoretical model. The Journal of Sex Research, 20, 143–167. Coleman, E. (1987). Assessment of sexual orientation. Journal of Homosexuality, 14, 9–24. Chrobot-Mason, D., Button, S. B., & DiClementi, J. D. (2001). Sexual identity management strategies: An exploration of antecedents and consequences. Sex Roles, 45, 321–336. Chung, Y. B. (1995). Career decision making of lesbian, gay, and bisexual individuals. The Career Development Quarterly, 44, 178–203. Chung, Y. B. (2001). Work discrimination and coping strategies: Conceptual frameworks for counseling lesbian, gay, and bisexual clients. The Career Development Quarterly, 50, 33–44. Chung, Y. B., & Harmon, L. W. (1994). The career interests and aspirations of gay men: How sex-role orientation is related. Journal of Vocational Behavior, 45, 223–239. Clain, S. H., & Leppel, K. (2001). An investigation into sexual orientation discrimination as an explanation for wage differences. Applied Economics, 33, 37–47. Clair, J., Beatty, J., & MacLean, T. (2002, August). Out of sight but not out of mind: How people manage invisible social identities in the workplace. Presented at the Academy of Management Meeting, Denver, Colorado. Cleveland, J. N., Stockdale, M., & Murphy, K. R. (2000). Women and men in organizations: Sex and gender issues at work. Mahwah, NJ: Lawrence Erlbaum. Clutterbuck, D., & Ragins, B. R. (2002). Mentoring and diversity: An international perspective. Oxford: Butterworth-Heinemann. Coleman, E. (1987). Assessment of sexual orientation. Journal of Homosexuality, 14, 9–24. Collella, A., & Dipboye, B. (Eds) (in press). Discrimination at work: Psychological and organizational bases. Mahwah, NJ: Lawrence Erlbaum.

Sexual Orientation In The Workplace

113

Creed, W. E., & Scully, M. A. (2000). Songs of ourselves: Employee’s deployment of social identity in workplace encounters. Journal of Management Inquiry, 9, 391–413. Creed, W. E., Scully, M. A., & Austin, J. R. (2002). Clothes make the person? The tailoring of legitimating accounts and the social construction of identity. Organization Science, 13, 475– 496. Crocker, J., Major, B., & Steele, C. (1998). Social stigma. In: D. Gilbert, S. Fiske & G. Lindzey (Eds), The Handbook of Social Psychology (4th ed., pp. 504–553). Boston: McGraw-Hill. Cross, W. E. (1978). The Thomas and Cross models of psychological nigrescence. Journal of Black Psychology, 4, 13–31. Croteau, J. M. (1996). Research on the work experiences of lesbian, gay and bisexual people: An integrative review of methodology and findings. Journal of Vocational Behavior, 48, 195–209. Croteau, J. M., & Lark, J. S. (1995). On being lesbian, gay, or bisexual in student affairs: A national survey of experiences on the job. NASPA Journal, 32, 189–197. Crow, S. M., Fok, L. Y., & Hartman, S. J. (1998). Who is at greatest risk of work-related discrimination – Women, blacks or homosexuals? Employee Responsibilities and Rights Journal, 11, 15– 26. D’Augelli, A. R. (1998). Developmental implications of victimization of lesbian, gay, and bisexual youth. In: G. M. Herek (Ed.), Stigma and Sexual Orientation: Understanding Prejudice Against Lesbians, Gay Men, and Bisexuals (pp. 187–210). Thousand Oaks, CA: Sage. D’Augelli, A. R., & Grossman, A. H. (2001). Disclosure of sexual orientation, victimization, and mental health among lesbian, gay, and bisexual older adults. Journal of Interpersonal Violence, 16, 1008–1027. D’Augelli, A. R., & Hershberger, S. L. (1993). Lesbian, gay, and bisexual youth in community settings: Personal challenges and mental health problems. American Journal of Community Psychology, 21, 421–448. D’Augelli, A. R., Hershberger, S. L., & Pilkington, N. W. (1998). Lesbian, gay, and bisexual youth and their families: Disclosure of sexual orientation and its consequences. American Journal of Orthopsychiatry, 68, 361–371. Day, N. E., & Schoenrade, P. (1997). Staying in the closet vs. coming out: Relationships between communication about sexual orientation and work attitudes. Personnel Psychology, 50, 147– 163. DeCecco, J. P., & Parker, D. A. (1995). The biology of homosexuality: Sexual orientation or sexual preference? Journal of Homosexuality, 28, 1–27. Dovidio, J. F., & Gaertner, S. L. (Eds) (1986). Prejudice, discrimination and racism. San Diego: Academic Press. Dovidio, J., Major, B., & Crocker, J. (2000). Stigma: Introduction and overview. In: T. Heatherton, R. Kleck, M. Hebl & J. Hull (Eds), The Social Psychology of Stigma (pp. 1–30). New York: Guilford Press. Driscoll, J. M., Kelley, F. A., & Fassinger, R. E. (1996). Lesbian identity and disclosure in the workplace: Relation to occupational stress and satisfaction. Journal of Vocational Behavior, 48, 229– 242. Dutton, J. E. (2003). Energize your workplace: How to create and sustain high-quality connections at work. San Francisco: Jossey Bass. Dunkle, J. H. (1996). Toward an integration of gay and lesbian identity development and Super’s life-span approach. Journal of Vocational Behavior, 48, 149–159. Ellis, A. L., & Riggle, E. D. B. (1995). The relation of job satisfaction and degree of openness about one’s sexual orientation for lesbians and gay men. Journal of Homosexuality, 30, 75–85.

114

BELLE ROSE RAGINS

Ely, R. J. (1995). The role of dominant identity and experience in organizational work on diversity. In: S. E. Jackson & M. N. Ruderman (Eds), Diversity in Work Teams: Research Paradigms for a Changing Workplace (pp. 161–186). Washington, DC: American Psychological Association. Erikson, E. H. (1968). Identity, youth and crisis. New York: W. W. Norton. Fassinger, R. E. (1995). From invisibility to integration: Lesbian identity in the workplace. The Career Development Quarterly, 44, 148–167. Fassinger, R. E. (1996). Notes from the margins: Integrating lesbian experience into the vocational psychology of women. Journal of Vocational Behavior, 48, 160–175. Federal Bureau of Investigation (2001). Hate crime statistics, 2000. Washington, DC: author. Ferdman, B. M. (1999). The color and culture of gender in organizations: Attending to race and ethnicity. In: G. Powell (Ed.), Handbook of Gender & Work (pp. 17–34). Thousand Oaks: Sage. Foucault, M. (1978). The history of sexuality (R. Hurley, Trans.). New York: Pantheon Books. Frable, D. E. S., Platt, L., & Hoey, S. (1998). Concealable stigmas and positive self-perceptions: Feeling better around similar others. Journal of Personality and Social Psychology, 74, 909– 922. Frable, D. E. S., Wortman, C., & Joseph, J. (1997). Predicting self-esteem, well-being, and distress in a cohort of gay men: The importance of cultural stigma, personal visibility, community networks, and positive identity. Journal of Personality, 65, 599–624. Friskopp, A., & Silverstein, S. (1996). Straight jobs, gay lives: Gay and lesbian professionals, the Harvard Business School, and the American workplace. New York: Touchstone/Simon & Schuster. Garnets, L. D., & Kimmel, D. C. (1991). Lesbian and gay male dimensions in the psychology of human diversity. In: J. D. Goodchilds (Ed.), Psychological Perspectives on Human Diversity in America (pp. 143–189). Washington, DC: American Psychological Association. Garnets, L. D., & Kimmel, D. C. (1993). Introduction: Lesbian and gay male dimensions in the psychological study of human diversity. In: L. D. Garnets & D. C. Kimmel (Eds), Psychological Perspectives on Lesbian and Gay Male Experiences (pp. 1–51). New York: Columbia University Press. Goffman, E. (1963). Stigma: Notes on the management of spoiled identity. New York: Jason Aronson. Gonsiorek, J. C. (1995). Gay male identities: Concepts and issues. In: A. R. D’Augelli & C. J. Patterson (Eds), Lesbian, Gay, and Bisexual Identities over the Lifespan: Psychological Perspectives (pp. 24–47). New York: Oxford University Press. Gonsiorek, J. C., & Rudolph, J. R. (1991). Homosexual identity: Coming out and other developmental events. In: J. C. Gonsiorek & J. D. Weinrich (Eds), Homosexuality, Research Implications for Public Policy (pp. 161–176). Newbury Park: Sage. Gonsiorek, J. C., & Weinrich, J. D. (1991). The definition and scope of sexual orientation. In: J. C. Gonsiorek & J. D. Weinrich (Eds), Homosexuality: Research Implications for Public Policy (pp. 1–12). Newbury Park: Sage. Greenhaus, J. H., Callanan, G. A., & Godshalk, V. M. (2000). Career management (3rd ed.). Orlando, FL: Dryden Press. Griffin, P. (1992). From hiding out to coming out: Empowering lesbian and gay educators. In: K. M. Harbeck (Ed.), Coming Out of the Classroom (pp. 167–196). Binghamton, NY: Harrington Park Press. Griffith, K. H., & Hebl, M. R. (2002). The disclosure dilemma for gay men and lesbians: ‘Coming out’ at work. Journal of Applied Psychology, 87, 1191–1199. Haslam, N. (1997). Evidence that male sexual orientation is a matter of degree. Journal of Personality and Social Psychology, 73, 862–870.

Sexual Orientation In The Workplace

115

Hebl, M. R., Foster, J. B., Mannix, L. M., & Dovidio, J. F. (2002). Formal and interpersonal discrimination: A field study of bias toward homosexual applicants. Personality and Social Psychology Bulletin, 28, 815–825. Herek, G. M. (1984). Beyond “homophobia”: A social psychological perspective on attitudes towards lesbians and gay men. Journal of Homosexuality, 10, 1–21. Herek, G. M. (1986). The social psychology of homophobia: Toward a practical theory. Review of Law and Social Change, 14, 923–934. Herek, G. M. (1988). Hetereosexuals’ attitudes toward lesbians and gay men: Correlates and gender differences. Journal of Sex Research, 25, 451–477. Herek, G. M. (1989). Hate crimes against lesbians and gay men: Issues for research and policy. American Psychologist, 44, 948–955. Herek, G. M. (1990). The context of anti-gay violence. Journal of Interpersonal Violence, 5, 316–333. Herek, G. M. (1996). Why tell if you’re not asked? Self-disclosure, intergroup contact, and heterosexuals’ attitudes towards lesbians and gay men. In: G. M. Herek, J. B. Jobe & R. M. Carney (Eds), Coming Out in Force: Sexual Orientation and the Military (pp. 197–225). Chicago: University of Chicago Press. Herek, G. M. (2000). Sexual prejudice and gender: Do heterosexuals’ attitudes towards lesbians and gay men differ? Journal of Social Issues, 56, 25–266. Herek, G. M., & Capitanio, J. P. (1996). “Some of my best friends”: Intergroup contact, concealable stigma, and heterosexuals’ attitudes toward gay men and lesbians. Personality and Social Psychology Bulletin, 22, 412–424. Herek, G. M., & Capitanio, J. P. (1999). AIDS stigma and sexual prejudice. American Behavioral Scientist, 42, 1130–1147. Herek, G. M., Cogan, J. C., & Gillis, J. R. (2002). Victim experiences in hate crimes based on sexual orientation. Journal of Social Issues, 58, 319–339. Herek, G. M., Jobe, J. B., & Carney, R. M. (Eds) (1996). Coming out in force: Sexual orientation and the military. Chicago: University of Chicago Press. Herek, G. M., Kimmel, D. C., Amaro, H., & Melton, G. B. (1991). Avoiding heterosexist bias in psychological research. American Psychologist, 46, 957–963. Herrschaft, D., & Mills, K. I. (2002). The state of the workplace for lesbian, gay, bisexual and transgender Americans 2002. Washington, DC: Human Rights Campaign. Holland, J. L. (1985). Making vocational choices: A theory of vocational personalities and work environments. Englewood Cliffs, NJ: Prentice-Hall. Hom, P. W., & Griffeth, R. W. (1995). Employee turnover. Cincinnati: South-Western College Publishing. Horowitz, J. L., & Newcomb, M. D. (2001). A multidimensional approach to homosexual identity. Journal of Homosexuality, 42, 1–19. Horvath, M., & Ryan, A. (2003). Antecedents and potential moderators of the relationship between attitudes and hiring discrimination on the basis of sexual orientation. Sex Roles, 48, 115– 130. Hunter, J. D. (1991). Culture wars: The struggle to define America. New York: Basic Books/Harper Collins. James, K., Lovato, C., & Cropanzano, R. (1994). Correlational and know-group comparison validation of a workplace prejudice/discrimination inventory. Journal of Applied Social Psychology, 24, 1573–1592. Jones, E. E., Farina, A., Hastorf, A., Markus, H., Miller, D., & Scott, R. A. (1984). Social stigma: The psychology of marked relationships. New York: W. H. Freeman.

116

BELLE ROSE RAGINS

Jordan, K. M., & Deluty, R. H. (1998). Coming out for lesbian women: Its relation to anxiety, positive affectivity, self-esteem, and social support. Journal of Homosexuality, 35, 41–62. Jung, P. B., & Smith, R. F. (1993). Heterosexism: An ethical challenge. Albany: State University of New York Press. King, B. R. (2001). Ranking of stigmatization toward lesbians and their children and the influence of perceptions of controllability of homosexuality. Journal of Homosexuality, 41, 77–97. Kinsey, A. C., Pomeroy, W. B., & Martin, C. E. (1948). Sexual behavior in the human male. Philadelphia: W. B. Saunders. Kinsey, A. C., Pomeroy, W. B., Martin, C. E., & Gebhard, P. H. (1953). Sexual behavior in the human female. Philadelphia: W. B. Saunders. Kirkpatrick, L. A. (1993). Fundamentalism, Christian orthodoxy, and intrinsic religious orientation as predictors of discriminatory attitudes. Journal for the Scientific Study of Religion, 32, 256–268. Kite, M. E. (1984). Sex differences in attitudes towards homosexuals: A meta-analytic review. Journal of Homosexuality, 10, 69–81. Leary, M. R., & Tangney, J. P. (Eds) (2003). Handbook of self and identity. New York: Guilford. Ledvinka, J., & Scarpello, V. G. (1991). Federal regulation of personnel and human resource management (2nd ed.). Boston: PWS-Kent. Levine, M. P., & Leonard, R. (1984). Discrimination against lesbians in the workforce. Journal of Women in Culture and Society, 9, 700–724. Lewis, G. B. (2003). Black-white differences in attitudes toward homosexuality and gay rights. Public Opinion Quarterly, 67, 59–78. Lewis, R. J., Derlega, V. J., Berndt, A., Morris, L. M., & Rose, S. (2001). An empirical analysis of stressors for gay men and lesbians. Journal of Homosexuality, 42, 63–88. Lonborg, S. D., & Phillips, J. M. (1996). Investigating the career development of gay, lesbian, and bisexual people: Methodological considerations and recommendations. Journal of Vocational Behavior, 48, 176–194. Lyness, K. S., & Thompson, D. E. (1997). Above the glass ceiling? A comparison of matched samples of female and male executives. Journal of Applied Psychology, 82, 359–375. Martin, J. I., & Knox, J. (2000). Methodological and ethical issues in research on lesbians and gay men. Social Work Research, 24, 51–59. May, V. M., & Cochran, S. D. (2001). Mental health correlates of perceived discrimination among lesbian, gay, and bisexual adults in the United States. American Journal of Public Health, 91(11), 1869–1876. McCarn, S. R., & Fassinger, R. E. (1996). Revisioning sexual minority identification formation: A new model of lesbian identity and its implications for counseling and research. Counseling Psychologist, 24, 508–535. McIntosh, M. (1968). The homosexual role. Social Problems, 16, 182–192. Mendoza, M. (2003). In U.S., gays see growing influence: Observers say trend has been building. The Dallas Morning News (July 14). Retrieved 9/22/03 from Ebsco Host Research Data Base. Meyer, I. H. (2003). Prejudice, social stress, and mental health in lesbian, gay, and bisexual populations: Conceptual issues and research evidence. Psychological Bulletin, 129, 674–697. Meyer, L. H., & Dean, L. (1998). In: G. M. Herek (Ed.), Stigma and Sexual Orientation: Understanding Prejudice Against Lesbians, Gay Men, and Bisexuals (Vol. 4, pp. 160–186). Thousand Oaks: Sage. Mobley, M., & Slaney, R. B. (1996). Holland’s theory: It’s relevance for lesbian women and gay men. Journal of Vocational Behavior, 48, 125–135.

Sexual Orientation In The Workplace

117

Mohr, J. J., & Fassinger, R. E. (2003). Self-acceptance and self-disclosure of sexual orientation in gay, lesbian, and bisexual adults: An attachment perspective. Journal of Counseling Psychology, 50, 482–495. Morin, S. F., & Garfinkle, E. M. (1978). Male homophobia. Journal of Social Issues, 34, 29–47. Morrow, S. L., Gore, P. A., Jr., & Campbell, B. W. (1996). The application of a sociocognitive framework to the career development of lesbian women and gay men. Journal of Vocational Behavior, 48, 136–148. Namaste, V. K. (2000). Invisible lives: The erasure of transsexual and transgendered people. Chicago, IL: University of Chicago Press. Nauta, M. M., Saucier, A. M., & Woodard, L. E. (2001). Interpersonal influences on students’ academic and career decisions: The impact of sexual orientation. The Career Development Quarterly, 49, 352–362. Nelson, E. S., & Krieger, S. L. (1997). Changes in attitudes toward homosexuality in college students: Implementation of a gay men and lesbian peer panel. Journal of Homosexuality, 33, 63– 81. Neuberg, S. L., Smith, D. M., Hoffman, J. C., & Russell, F. J. (1994). When we observe stigmatized and “normal” individuals interacting: Stigma by association. Personality and Social Psychology Bulletin, 20, 196–209. Newport, F. (1998). Americans more likely to believe sexual orientation due to environment, not genetics. The Gallup Poll Monthly, July, 14–16. Nkomo, S. M. (1992). The emperor has no clothes: Rewriting race in organizations. Academy of Management Review, 17, 487–513. Nkomo, S. M., & Cox, T., Jr. (1996). Diverse identities in organizations. In: S. Clegg, C. Hardy & W. Nord (Eds), Handbook of Organization Studies (pp. 338–356). London: Sage. Noe, R. A., Greenberger, D. B., & Wang, S. (2002). Mentoring: What we know and where we might go. Research in Personnel and Human Resources Management, 21, 129–173. Peel, E. (2001). Mundane heterosexism: Understanding incidents of the everyday. Women’s Studies International Forum, 24, 541–554. Pillard, R. C., & Bailey, J. M. (1995). A biologic perspective on sexual orientation. Clinical Sexuality, 18, 71–84. Plummer, K. (1975). Sexual stigma: An interactionist account. London: Routledge & Kegan Paul. Powell, G. N., & Graves, L. M. (2003). Women and men in management (3rd ed.). Thousand Oaks, CA: Sage. Powell, G. N., & Mainiero, L. A. (1992). Cross-currents in the river of time: Conceptualizing the complexities of women’s careers. Journal of Management, 18, 215–237. Prince, J. P. (1995). Influences on the career development of gay men. The Career Development Quarterly, 44, 168–177. Ragins, B. R. (1997). Diversified mentoring relationships in organizations: A power perspective. Academy of Management Review, 22, 482–521. Ragins, B. R. (1999). Gender and mentoring in organizations: A review and research agenda for the next decade. In: G. Powell (Ed.), Handbook of Gender and Work (pp. 347–370). Thousand Oaks, CA: Sage. Ragins, B. R. (2004). Disclosure disconnects: Antecedents and consequences of disclosing invisible stigmas across work and home domains. Unpublished manuscript. Ragins, B. R., & Cornwell, J. M. (2001a). Pink triangles: Antecedents and consequences of perceived workplace discrimination against gay and lesbian employees. Journal of Applied Psychology, 86, 1244–1261.

118

BELLE ROSE RAGINS

Ragins, B. R., & Cornwell, J. M. (2001b, August). Walking the line: Fear and disclosure of sexual orientation in the workplace. Paper presented at the National Academy of Management Meeting, Washington, DC. Ragins, B. R., Cornwell, J. M., & Miller, J. S. (2003). Heterosexism in the workplace: Do race and gender matter? Group & Organization Management, 28, 45–74. Ragins, B. R., & Gonzalez, J. A. (2003). Understanding diversity in organizations: Getting a grip on a slippery construct. In: J. Greenberg (Ed.), Organizational Behavior: The State of the Science (2nd ed., pp. 125–163). Mahwah, NJ: Lawrence Erlbaum. Ragins, B. R., & Sundstrom, E. (1989). Gender and power in organizations: A longitudinal perspective. Psychological Bulletin, 105, 51–88. Ragins, B. R., Townsend, B., & Mattis, M. (1998). Gender gap in the executive suite: Female executives and CEOs report on breaking the glass ceiling. Academy of Management Executive, 12(1), 28– 42. Ragins, B., & Wiethoff, C. (in press). Understanding heterosexism at work: The straight problem. In: A. Collella & B. Dipboye (Eds), Discrimination at Work: Psychological and Organizational Bases. Lawrence Erlbaum. Riordan, C. M. (2001). Relational demography within groups: Past developments, contradictions, and new directions. Research in Personnel and Human Resource Management, 19, 131–173. Rivers, I. (1997). Lesbian, gay and bisexual development: Theory, research and social issues. Journal of Community & Applied Social Psychology, 7, 329–343. Roen, K. (2002). “Either/Or” and “Both/Neither”: Discursive tensions in transgender politics. Signs: Journal of Women in Culture and Society, 27(2), 501–523. Rostosky, S. S., & Riggle, E. D. B. (2002). “Out” at work: The relation of actor and partner workplace policy and internalized homophobia to disclosure status. Journal of Counseling Psychology, 49, 411–419. Same sex marriages a “threat to society” – Vatican (2003). Irish Times (August 1), p. 10. Retrieved 11/22/03 from ProQuest Newspapers Data Base. Savin-Williams, R. C. (1995). Lesbian, gay male, and bisexual adolescents. In: A. R. D’Augelli & C. J. Patterson (Eds), Lesbian, Gay, and Bisexual Identities over the Lifespan: Psychological Perspectives (pp. 165–189). New York: Oxford University Press. Savin-Williams, R. C., & Cohen, K. M. (1996). Psychosocial outcomes of verbal and physical abuse among lesbian, gay, and bisexual youths. In: R. C. Savin-Williams & K. M. Cohen (Eds), The Lives of Lesbians, Gays, and Bisexuals (pp. 181–194). Fort Worth, TX: Harcourt Brace & Co. Schmitt, A., & Sofer, J. (1992). Sexuality and eroticism among males in Moslem societies. New York: Haworth Press. Schneider, B. E. (1987). Coming out at work: Bridging the private/public gap. Work and Occupations, 13, 463–487. Schope, R. D. (2002). The decision to tell: Factors influencing the disclosure decision of sexual orientation by gay me. Journal of Gay & Lesbian Social Services, 14, 1–22. Sears (1997). Thinking critically/intervening effectively about homophobia and heterosexism. In: J. T. Sears & W. L. Williams (Eds), Overcoming Heterosexism and Homophobia: Strategies That Work (pp. 13–48). New York: Columbia University Press. Seelye, K. Q., & Elder, J. (2003). Strong support is found for ban on gay marriage. The New York Times (December 21), pp. A1, A26. Seidman, S. (Ed.) (1996). Queer theory/sociology. Cambridge: Blackwell. Shively, M. G., & DeCecco, J. P. (1977). Components of sexual identity. Journal of Homosexuality, 3, 31–48.

Sexual Orientation In The Workplace

119

Smart, L., & Wegner, D. M. (2000). The hidden costs of hidden stigma. In: T. Heatherton, R. Kleck, M. Hebl & J. Hull (Eds), The Social Psychology of Stigma (pp. 220–242). New York: Guilford Press. Smith, N. G., & Ingram, K. M. (2004). Workplace heterosexism and adjustment among lesbian, gay, and bisexual individuals: The role of unsupportive social interactions. Journal of Counseling Psychology, 51, 57–67. Sophie, J. (1986). A critical examination of stage theories of lesbian identity development. Journal of Homosexuality, 12, 39–51. Stangor, C., & Crandall, C. S. (2000). Threat and the social construction of stigma. In: T. Heatherton, R. Kleck, M. Hebl & J. Hull (Eds), The Social Psychology of Stigma (pp. 62–87). New York: Guilford Press. Stein, A., & Plummer, K. (1996). I can’t even think straight: “Queer” theory and the missing sexual revolution in sociology. In: S. Seidman (Ed.), Queer Theory/Sociology (pp. 129–144). Cambridge: Blackwell. Sullivan, S. E. (1999). The changing nature of careers: A review and research agenda. Journal of Management, 25, 457–484. Super, D. E. (1957). The psychology of careers: An introduction to vocational development. New York: Harper Row. Swann, W. B., Jr. (1983). Self-verification: Bringing social reality into harmony with the self. In: J. Suls & A. G. Greenwald (Eds), Social Psychological Perspectives on the Self (Vol. 2, pp. 33–66). Hillsdale, NJ: Lawrence Erlbaum. Swann, W. B., Jr. (1987). Identity negotiation: Where two roads meet. Journal of Personality and Social Psychology, 53, 1038–1051. Swann, W. B., Jr., Polzer, J. T., Seyle, D. C., & Ko, S. J. (2004). Finding value in diversity: Verification of personal and social self-views in diverse groups. Academy of Management Review, 29, 9–27. Tharenou, P., Latimer, S., & Conroy, D. (1994). How do you make it to the top? An examination of the influences on women’s and men’s managerial advancement. Academy of Management Journal, 37, 899–931. Thoits, P. A. (1983). Multiple identities and psychological well-being: A reformulation and test of the social isolation hypothesis. American Sociological Review, 48, 174–187. Thomas, D. A., & Alderfer, C. P. (1989). The influence of race on career dynamics: Theory and research on minority career experiences. In: M. Arthur, D. T. Hall & B. Lawrence (Eds), Handbook of Career Theory (pp. 133–158). Cambridge, MA: Cambridge University Press. Thomas, D. A., & Gabarro, J. J. (1999). Breaking through: The making of minority executives in corporate America. Boston, MA: Harvard Business School Press. Tomlinson, M. J., & Fassinger, R. E. (2003). Career development, lesbian identity development, and campus climate among lesbian college students. Journal of College Student Development, 44(6), 845–860. Troiden, R. R. (1989). The formation of homosexual identities. Journal of Homosexuality, 17, 43–73. Twomey, D. P. (1994). Equal employment law (3rd ed.). Cincinnati, OH: South-Western Publishing. Waldo, C. R. (1999). Working in a majority context: A structural model of heterosexism as minority stress in the workplace. Journal of Counseling Psychology, 46, 218–232. Walters, K. L., & Simoni, J. M. (1993). Lesbian and gay male group identity attitudes and self-esteem: Implications for counseling. Journal of Counseling Psychology, 40, 94–99. Wanberg, C. R., Welsh, E. T., & Hezlett, S. A. (2003). Mentoring research: A review and dynamic process model. Research in Personnel and Human Resources Management, 22, 39–124. Weinberg, G. (1972). Society and the healthy homosexual. New York: St. Martin’s.

120

BELLE ROSE RAGINS

Weinberg, M. S., & Williams, C. J. (1974). Male homosexuals: Their problems and adaptations. New York: Oxford University Press. Weinberg, T. S. (1983). Gay men, gay selves: The social construction of homosexual identities. New York: Irvington. Weiner, B., Perry, R. P., & Magnusson, J. (1988). An attributional analysis of reactions to stigma. Journal of Personality and Social Psychology, 55, 738–748. Weinrich, J. D., & Williams, W. L. (1991). Strange customs, familiar lives: Homosexuality in other cultures. In: J. Gonsiorek & J. Weinrich (Eds), Homosexuality: Research Implications for Public Policy (pp. 44–59). Newbury Park: Sage. Wetzstein, C. (2003). Groups pledge to protect marriage; Say gay threat demands action. The Washington Times (October 3), Pg. A.01. Retrieved 11/22/03 from ProQuest Newspapers Data Base. Wharton, A. (1992). The social construction of gender and race in organizations: A social identity and group mobilization perspective. Research in the Sociology of Organizations, 10, 55–84. Winegarden, B. J. (1994). Aversive heterosexism: An exploration of unconscious bias toward lesbian psychotherapy clients. Unpublished doctoral dissertation, University of Wisconsin, Milwaukee. Woods, J. D. (1993). The corporate closet: The professional lives of gay men in America. New York: Free Press. Woods, S. E., & Harbeck, K. M. (1992). Living in two worlds: The identity management strategies used by lesbian physical educators. In: K. M. Harbeck (Ed.), Coming Out of the Classroom Closet: Gay and Lesbian Students, Teachers and Curricula (pp. 141–166). Binghamton, NY: Haworth Press.

ORGANIZATIONAL STAFFING: A MULTILEVEL REVIEW, SYNTHESIS, AND MODEL Robert E. Ployhart ABSTRACT Does staffing contribute to organizational effectiveness and sustained competitive advantage, or are many of staffing’s implications merely cross-level fallacies? This article provides a critical examination of staffing research and practice, and proposes a multilevel model of staffing that ties together micro (e.g. personnel selection), meso (e.g. team staffing), and macro (e.g. organizational strategy, Human Resources practices) theory, research, and practice. The model is both integrative and prescriptive, providing a basic organizing structure for examining staffing research within and across levels. The article begins with a review of multilevel theory, followed by a review and critique of the dominant staffing paradigms from a multilevel perspective. It is shown these single level paradigms cannot answer many of the primary questions of interest to staffing specialists. In contrast, the multilevel staffing model not only addresses these limitations, but also prompts a variety of new predictions that oftentimes run counter to prevailing wisdom. Staffing specialists are challenged to show how our science and practice contribute to better functioning organizations.

Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 121–176 © 2004 by Published by Elsevier Ltd. ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23003-1

121

122

ROBERT E. PLOYHART

A CHALLENGE TO STAFFING SCHOLARS Over 80 years of research on organizational staffing may lead one to believe personnel selection represents one of the true “success stories” of applied psychology. For example, we have an impressive collection of data that supports the criterion-related validity of numerous predictor measures such as cognitive ability, conscientiousness, and structured interview methods (e.g. Barrick, Mount & Judge, 2001; Schmidt & Hunter, 1998). Additional research suggests organizations that use “high performance” staffing practices are more effective (e.g. Huselid, 1995). Given such impressive findings, one might believe our knowledge of staffing organizations is fairly comprehensive. My goal with this article is to critically examine this belief. If one more closely inspects the foundation of the staffing literature base, one finds many places where the cement has been poured thickly and many more places where there is none at all. While the extant research tells us a great deal about individual level predictorperformance relationships and an increasing amount about macro Human Resource (HR) practices and organizational effectiveness, it says almost nothing about how these practices at different levels fit with each other and contribute to performance and effectiveness across levels of analysis. Ten years ago, John Campbell (Campbell, McCloy, Oppler & Sager, 1993) wrote about the criterion problem in applied psychology, noting that, “If the dependent variable is the variable of real interest and if performance is perhaps our most important dependent variable, then more often than not we simply may not know what our real interest is” (p. 35). We have since made progress in defining performance at the individual level, but have scarcely scratched the surface of defining and predicting performance across multiple levels of analysis. Today we might modify Campbell’s statement slightly by noting that if performance is the most important variable, one might also ask at what level of performance are we most interested – an individual’s, a group’s, or an organization’s. The answer depends on who you ask: most organizational decision makers (as well as stockholders, investors, and governments) care little about individual performance relative to organizational performance, yet the staffing specialist will most often advocate the individual level criterion. As a discipline, organizational staffing research and practice have been aiming at a narrow individual-level performance target that may be of little interest to anyone outside of personnel selection. Perhaps we continue to rely on this individual level criterion because we can predict it, it is convenient, or it fulfills our immediate staffing goals. But if our science is to contribute to better functioning organizations, we must start to consider loftier goals. In this article I challenge staffing researchers and practitioners to adopt a multilevel perspective, arguing that only through such a perspective can we fully

Multilevel Staffing

123

appreciate and understand staffing’s relationship to organizational effectiveness. Toward this end, the article develops a multilevel staffing model that connects micro and macro staffing research and practice, with a particular focus on extending micro level findings to higher levels. In the sections that follow, I will: (1) discuss the basics of person-centered staffing perspectives; (2) summarize basic levels of analysis research; (3) review and critique past organizational staffing research and practice (particularly the dominant goals, methods, and assumptions); (4) propose and discuss the multilevel staffing model; and (5) describe several novel predictions directed by the model, many of which are counterintuitive to prevailing thinking. If we are to improve the science and practice of our profession, and to convince the consumers of our research that organizational staffing “matters,” we need to examine staffing from a broader perspective. We need to think of staffing from within a multilevel framework.

STAFFING ORGANIZATIONS: A PERSON-CENTERED PERSPECTIVE What does it mean to staff an organization? Schneider (1987) proposed an integrative model, known as the “Attraction-Selection-Attrition (ASA) model,” that helps answer this question. First, individuals are attracted to organizations that are perceived to possess similar personalities, values, and characteristics (e.g. Cable & Judge, 1997; Ryan, Sacco, McFarland & Kriska, 2000). Second, organizations select applicants based on their fit with the job’s requirements (e.g. knowledge, skills, abilities, and other characteristics; KSAO), and individuals select job opportunities that most match their needs, interests, and values. Finally, attrition from organizations can work in two ways; one with the individual withdrawing from the organization and the other with the organization terminating the individual’s employment (Schaubroeck, Ganster & Jones, 1998). Through these ASA processes, the model predicts the “homogeneity hypothesis”: over time organizations become staffed with people who share similar values, personalities, and goals, and these shared ways of thinking influence organizational structure, process, and culture. Empirical evidence suggests that homogeneity does occur within jobs, occupations, organizations, and industries, such that these categories can be distinguished in terms of the personalities of the people in them (e.g. Ployhart, Weekley & Baughman, 2003; Schaubroeck et al., 1998; Schneider, Smith, Taylor & Fleenor, 1998). Perhaps the most important implication of the ASA model is that staffing practices can and should produce homogeneity in terms of KSAOs. But how do micro staffing processes contribute to organizational outcomes? Staffing as a field has allowed this to remain an untested question

124

ROBERT E. PLOYHART

for far too long; its answer requires an understanding of multilevel methods and theory.

INTRODUCTION TO MULTILEVEL METHODS AND THEORY: ARCHITECTURES OF ORGANIZATIONS Multilevel theory, or levels of analysis, are general terms describing theory and methods useful for understanding how constructs and processes evolve, interact, and function within and between levels of analysis (Kozlowski & Klein, 2000). Levels of analysis research draws heavily from concepts developed in General Systems Theory (GST; Von Bertalanffy, 1968, 1972), a far reaching theory in many areas of science. In this section I summarize key implications of this research for staffing; readers interested in more detail about levels issues should consult several excellent sources (e.g. House, Rousseau & Thomas-Hunt, 1995; James & Jones, 1974; Klein, Dansereau & Hall, 1994; Rousseau, 1985), and especially a recent book by Klein and Kozlowski (2000). In the organizational sciences, the level of analysis is usually defined by the nature of the organization or applied problem, and most organizational theory and research tend to consider only a few levels. Working from the micro to the macro, these levels typically include: task, individual, dyad, job, group/team, department, organization, industry, nation, and culture. These levels may be seen graphically in Fig. 1 (the arrows are discussed shortly). As shown in the figure, the lowest level of typical interest in staffing research is the level of the task (e.g. job analysis). As one works up through the levels, each level becomes more encompassing such that each lower level is nested within each higher level. Although perhaps not obvious, this figure shows several important and fundamental implications for staffing research. Drawing from Kozlowski and Klein (2000), these include: operationalizations, features, cross-level movement, models, and fallacies. Table 1 provides an overview and summary of these issues.

Operationalizations There are at a minimum two important issues for how psychological constructs are operationalized within a multilevel framework (see Kozlowski & Klein, 2000). First, one must consider the level of theory. The level of theory specifies at what level the construct is hypothesized to exist. For example, group composition (i.e. the KSAO composition of a group of people) is inherently a group level construct, whereas a given individual’s KSAOs represent individual level

Multilevel Staffing

125

Fig. 1. Common Levels in the Organizational Sciences.

constructs. Second, one must consider the level of measurement. The level of measurement specifies at what level the measure of the construct will occur. For example, we may measure group composition by measuring individual level KSAOs and aggregating them to the group level. In this instance the level of theory is at the group level but the level of measurement is at the individual level.

126

ROBERT E. PLOYHART

Table 1. Summary of Levels of Analysis Issues and Implications for Staffing. Levels Concept

Definition and Implication

I. Operationalizations A. Level of theory B. Level of measurement

Theoretical level of interest for the construct Level where the construct is measured

II. Features of levels A. Bond strength B. Temporal scale C. Near-decomposability III. Cross-level movement A. Contextual B. Emergent

IV. Models A. Single level B. Cross-level C. Multilevel V. Fallacies A. Misspecification B. Cross-level C. Contextual

Strength of connections between levels A time asymmetry, such that lower level processes operate more quickly than higher level processes In a multilevel system, cannot separate one level from the system Top-down; higher level influences on lower levels (move quickly) Bottom-up; lower level influences on higher levels (move slowly). Two general forms: compilation (dispersion) and composition (consensus) Individual; unit-global; unit-shared; unit-mixed; unit-configural Direct effects; mixed determinants; mixed effects; moderator; frog pond Multilevel homologous Incorrectly specify the wrong level of theory to a construct Inappropriate generalization of findings across levels Neglecting contextual (cross-level) influences on lower levels

Note that this is not to say simple aggregation is always appropriate, as discussed shortly. Features Constructs and processes at different levels exist in a rough hierarchy such as that illustrated in Fig. 1. Each lower level is nested within a higher level, and each level borders other levels. This means the relationships between levels must be carefully understood through such concepts of bond strength, temporal scale, and near-decomposability. First, bond strength refers to how strongly related constructs at various levels are tied to each other (Simon, 1973). As a general rule, adjacent levels are more strongly related and intertwined than levels farther apart. As shown in Fig. 1, the relationship between the individual level and group level will be stronger than the relationship between the individual level and organizational level. Thus, regardless

Multilevel Staffing

127

of the level of interest, the farther one moves from this focal level to another level, the weaker the relationship. Stated differently, the effects of any one level on another will become weaker as the levels become farther apart. A closely related concept is known as coupling (Weick, 1976). More strongly coupled levels are more strongly interrelated. For example, when the nature of the team task makes team members highly interdependent, the individual level and team levels are closely coupled. Second, different levels operate on different time scales or metrics; such differences may be referred to as differences in temporal scale across levels (Kozlowski & Klein, 2000). This concept is not unique to organizational research. For example, Newell (1990) provides a framework in cognitive science that dictates how different levels of time correspond to different types of tasks and theories/domains of study. Lower level phenomena occur in terms of milliseconds and correspond to biological processes, while higher levels occur in terms of hours or days and represent bounded rationality (i.e. real world cognitive functioning). The same concept can be applied to the organizational sciences, such that higher levels operate on a longer temporal scale than lower levels (see Simon, 1973). Clearly the temporal pacing for individuals is much quicker than the pacing for organizations. Individuals can change their behavior quickly (e.g. change jobs), whereas organizations change much more slowly (e.g. change marketing strategies). Third, the principle of near-decomposability represents the fact that when examining a system of interrelated levels, one may only partially decompose part of the system to study within-level processes (Simon, 1973). For example, if the level of interest is only the individual level, a researcher may proceed to study that level, but must recognize that at least part of the variance at that level may be influenced by forces above or below it (Kozlowski & Klein, 2000). Thus, neardecomposability suggests one cannot completely disentangle levels, but can only partially do so. A complete description of a multilevel structure can therefore not be based on single, within level research. Cross-Level Movement The bond strength and temporal scale features of levels predict an important asymmetry when moving across levels. Contextual influences are those that move from higher to lower levels; a top-down perspective. Group processes influencing individual behavior would be an example of a contextual influence because the group provides a context for the individual’s behavior. In contrast, emergent influences are bottom-up processes. Emergent influences occur when lower-level constructs contribute to higher level constructs, such as when individual performance contributes in some manner to group performance. Of the two forms of cross-level

128

ROBERT E. PLOYHART

movement, one usually observes faster and stronger effects for contextual influences than for emergent influences (Kozlowski & Klein, 2000; Simon, 1973). That is, contextual effects operate more quickly than emergent effects. For example, when an organization changes its staffing strategy, all employees must adapt to this new strategy (i.e. an immediate contextual influence on behavior). However, it may take years before the benefits of this new staffing strategy show improvements in organizational performance – a slower emergent process. There are a variety of forms of emergence that can occur; these are discussed in detail in Kozlowski and Klein (2000). On one extreme is compilation; a form of emergence that occurs when the higher level construct is different from the lower level construct. Unit level diversity is a compilation model because the construct of unit diversity is conceptually distinct from the individual level demographics of the unit’s members (Bliese, 2000). At the other extreme is the isomorphic composition model. Isomorphism means the construct is essentially the same across levels of analysis. Thus, an isomorphic composition model occurs when individual performance and group performance represent the same fundamental construct. Chan (1998) presents a typology of five composition models useful for thinking about how composition occurs and how it should be modeled (see Kozlowski & Klein, 2000, for more detailed alternatives). An additive model is the summation or average of lower level scores (e.g. individual intelligence) into a unit level composite (e.g. team intelligence) irrespective of the within-unit variance. A direct consensus model is also the average of lower level scores, but it presupposes or evaluates consensus on the within-unit scores. A referent-shift model is similar to the direct consensus model except the focus of the lower level scores is on the higher level construct (e.g. asking questions about one’s own personality is a direct consensus model; asking questions about the average personality of one’s co-workers is a referent-shift model because the item reference is shifted to the unit level unit). In contrast to the previous models, a dispersion model focuses on the within-group variance rather than consensus as the measure of the construct (e.g. racial diversity). Finally, a process model proposes similar relations among composition variables and constructs at one level that are also present at other levels. Together, these various composition models help us think about and conduct cross level research, particularly in terms of conceptualizing the nature of aggregation.

Models There are a variety of types of multilevel models that articulate in theoretical, testable terms how constructs and processes at one level are related to constructs and processes at other levels. The nature of such models is a function of the level

Multilevel Staffing

129

of theory and the level of measurement. Kozlowski and Klein (2000) describe three major types of models: single-level, cross-level, and homologous multilevel. Figure 2 illustrates these various models with empirical examples (note: not all of the representative empirical examples in Fig. 2 are from the staffing literature or perfectly represent the type of model. This reflects the current lack of such research in staffing, but they are nonetheless useful for illustrating the various models). First, single level models are models where the level of theory occurs entirely within one level. A single-level model can occur within any level shown in Fig. 1; the important feature is that the entire level of theory is contained within the single level. When the level of theory is at a higher level of analysis, but the level of measurement is at a lower level of analysis, one must aggregate or combine the lower level scores into higher level scores (this is denoted by the dashed arrows in Fig. 2). Second, cross level models are defined by at least one of the independent and/or dependent variables existing at different levels of analysis. Contextual effects for higher level variables (culture) on lower level outcomes (performance), or emergent processes where lower level processes (individual KSAOs) contribute to higher level outcomes (team performance), are by definition cross-level models. The level of theory in such models contains two or more levels, as will the level of measurement. Figure 2 shows five general subtypes of cross-level models. The final type of model is the homologous multilevel model. The term homologous means the relationships between constructs within levels are the same across levels of analysis. The constructs of interest, and their interrelationships, are often hypothesized to be isomorphic across levels.

Fallacies Ignoring levels of analysis issues can lead to many fallacies in theorizing and data analysis. Rousseau (1985) describes three such fallacies. First, a misspecification fallacy occurs when the wrong level of theory is assigned to a construct. For example, if one were interested in understanding how group composition predicts group performance, a misspecification fallacy would occur if the research only examined individual KSAOs. The misspecification fallacy can also occur when the level of measurement is not aligned with the construct of interest. Second, a crosslevel fallacy occurs when one incorrectly generalizes across levels of analysis. Many researchers may assume that relationships found at the individual level must also occur at higher levels. An example of this could occur if one believed that organizations composed of more intelligent employees will perform better (a generalization of what is found at the individual level). In fact, without empirical proof this belief commits a cross-level fallacy! Finally, a contextual fallacy occurs

130

ROBERT E. PLOYHART

Fig. 2. Multilevel Staffing Models (Adapted from Kozlowski & Klein, 2000) and Examples. Note: Subscript “i” denotes individual level, “g” denotes group level, and “o” denotes organizational level. Dashed lines indicate types of aggregation.

Multilevel Staffing

131

Fig. 2. (Continued )

132

ROBERT E. PLOYHART

Fig. 2. (Continued )

when one has neglected or misattributed the effects of a higher level contextual influence on lower level processes. For example, suppose one finds the relationship between individual personality and individual performance disappears after one controls for group structure. In this example, ignoring the group context effect would result in a very inappropriate interpretation of the data. In summary, considerable theory and research suggests that a careful consideration of multilevel issues is important to fully understand any one level within the system. This means that to fully understand organizational staffing, it must be put into the appropriate multilevel context. By focusing on only one level, whether it be the individual or organizational, research will not fully illuminate the processes through which staffing influences organizational effectiveness (as predicted by the principle of near-decomposability). However, few staffing researchers have given sufficient attention to multilevel issues, as discussed next.

THE DOMINANCE OF SINGLE LEVEL STAFFING RESEARCH Consistent with the observations of similar reviews (e.g. Ostroff & Bowen, 2000; Schneider, Smith & Sipe, 2000; Wright & Boswell, 2002), there is little staffing research that is cross-level or multilevel in nature (indeed, for some of the models shown in Fig. 2, I could not identify even a single staffing example). Therefore, let us review the current state of staffing research from the perspectives that dominate it: two single-level perspectives (individual and organizational) and one crosslevel perspective (group/team with individual) are examined. Respectively, these roughly correspond to the micro, macro, and meso (intermediate) levels of HR. My purpose in the following sections is not to be exhaustive but rather to discuss the goals, theoretical traditions, and assumptions of each perspective because doing so

Multilevel Staffing

133

provides a means to compare, contrast, and ultimately integrate staffing research within a multilevel perspective. Table 2 provides an overview and summary of the characteristics described below.

Single-Level Staffing Research Individual Level (Micro Staffing) Major goals, questions, and problems. Micro staffing research is historically the dominant perspective on staffing. It is concerned with identifying theoretical and empirical relationships between individual level predictor constructs and measures with individual level performance constructs and measures. Key questions and problems include: What is job performance? What KSAOs contribute to effective performance? What are the most efficient and predictive ways we can recruit and select qualified applicants? How can one balance prediction with diversity, or at least reduce the amount of subgroup differences and adverse impact in a selection process? What new technologies improve predictor validity and utility? Campbell (1990) describes these issues as the “performance prediction problem,” and it is solving this problem that continues to keep the attention of most individual level staffing researchers. Theoretical tradition. The individual level staffing perspective draws heavily from theory and research in differential and individual differences psychology (see Ackerman & Humphreys, 1990). These theoretical traditions attempt to identify and classify the major cognitive, dispositional, physical, and psychomotor domains (e.g. Fleishman & Quaintance, 1984). In more recent years, personality psychology (particularly that based on the Five Factor Model; FFM) is an added influence. Finally, research on values and vocational interests (e.g. Holland, 1985) complements this individual level staffing research. Across these disciplines, the common theme is a focus on the differences that occur among individuals (although occupational psychology also considers the “environment”). Representative research. There have been many, many studies examining individual level staffing issues. Research has sought to identify the latent structure of job performance (e.g. Campbell et al., 1993), including contextual (or OCB; Borman & Motowidlo, 1997), counterproductive (Sackett, 2002), and adaptive (Pulakos, Arad, Donovan & Plamondon, 2000). Most research has focused on identifying the constructs and predictor measures most predictive of performance: cognitive ability, conscientiousness, work samples, structured interviews, biodata, and situational judgment tests (e.g. Barrick et al., 2001; Huffcutt, Weekley, Wiesner,

134

Table 2. Comparing and Contrasting Staffing Research Across Levels of Analysis. HR Orientation

Theoretical Orientation

Representative Predictor Variables

Representative Mediating Variables

Representative Dependent Variables

Representative Questions

Organizational

Macro

Organizational theory Strategy Sociology

Staffing practices HR practices HR strategy

Usually none

Organizational Effectiveness Return on investment Market share

Relations between HR practices and organizational effectiveness Which HR “bundles” contribute to effectiveness Linking HR practices to HR strategy

Group & Team

Meso

Groups and teams Social psychology Sociology

Group composition Individual level KSAO Attitudes toward teamwork

Cohesion Mental models Potency Backup behaviors Coordination processes

Individual and team Performance Turnover Satisfaction

Nature of group processes Types of teamwork KSAOs Form of group/team composition Input-Process-Output linkages

Individual

Micro

Differential psychology Individual differences Personality psychology Cognitive psychology

Cognitive ability Personality Physical ability

Job knowledge Job skills Motivation Judgment Goal choice

Job performance (overall, task, contextual, adaptive, etc.) Withdrawal Accidents Training

Relationships between predictors and criteria Reducing racial subgroup differences Test validity and utility Recruitment Decision making Validity generalization Applicant reactions

ROBERT E. PLOYHART

Level

Multilevel Staffing

135

DeGroot & Jones, 2001; McDaniel, Morgeson, Finnegan, Campion & Braverman, 2001; see Schmidt & Hunter, 1998). Utility research has attempted to translate these validities into indices of financial gain (see Boudreau & Ramstad, 2003; Cabrera & Raju, 2001). Racial and gender subgroup differences are common to varying degrees on many of these predictors, and understanding how to reduce these differences is an active area of research (Hough, Oswald & Ployhart, 2001; Sackett, Schmitt, Ellingson & Kabin, 2001). Finally, recruiting and applicant reactions research seeks to attract and retain the best applicants (e.g. Anderson, Born & Cunningham-Snell, 2001; Highhouse, Zickar, Thorsteinson, Stierwalt & Slaughter, 1999; Ryan & Ployhart, 2000). Assumptions. The individual staffing perspective makes several major assumptions. First, it assumes there are individual differences in both performance and predictor constructs. Second, these individual differences are expected to be reasonably stable. Third, it assumes these individual differences can be measured with assessment methods that will be both valid and reliable. These assumptions are largely agreed upon (e.g. Guion, 1998). Other assumptions are also agreed upon but should not be. One is that hiring better employees contributes to better organizational performance. Another is that contextual influences do not substantially affect the predictive validity of staffing practices. I’ll revisit these issues in a later section. Organizational Level (Macro Staffing) Major goals, questions, and problems. Macro staffing represents a more recent perspective and examines how staffing practices enacted at the organizational level contribute to indices of organizational effectiveness (Wright & Boswell, 2002). This research is currently generating a great deal of interest, in part because it shows strong effects on the types of criteria many organizational decision makers care the most about and understand (e.g. return on investment, profitability, and so on). For example, as much as 20% of the variance in organizational performance may stem from organizational HR practices (e.g. Gerhart, 1999; Shaw, Delery, Jenkins & Gupta, 1998). This can translate into millions of dollars and there is simply nothing in the micro HR literature that comes close to this figure and is also believed by management. This research tradition asks the following kinds of questions: How do organizational differences in staffing practices relate to organizational effectiveness? What kinds of staffing practices lead to better profitability? In what industries are particular staffing practices most important? Do more valid and reliable selection methods contribute to better organizational effectiveness? What are the “high performance” staffing practices that most contribute to organizational

136

ROBERT E. PLOYHART

profitability? How does human capital at the organizational level contribute to organizational effectiveness? Theoretical tradition. The major theoretical tradition of the macro HR perspective stems from organizational theory. For example, the important and innovative work by several researchers, including Argyris (1964), James and Jones (1976), Katz and Kahn (1978), Miles and Snow (1978), Miller (1978), Woodward (1958), and others, helped set the stage for current organizational theorizing. The purpose of this work was clearly on understanding organizations (as opposed to any given individuals within the organization), and attempted to delineate how organizations are structured, function, develop, and decline. In such research, the focus was less on staffing and more about organizational characteristics such as structure and span of control. A related theoretical tradition focuses on strategic HR (e.g. Boxall & Purcell, 2003; Schuler & Jackson, 1987). For example, Olian and Rynes (1984) theorize how organizational strategy has a direct influence on the particular manifestation of HR practices, and Snow and Snell (1993) describe three major models of the staffing process. Finally, research on human capital (Becker, 1964) suggests that it is the aggregate composition of individuals’ KSAOs that contribute to effective organizational performance, and provide a source of organizationalspecific competitive advantage (e.g. Wright & McMahan, 1992). Representative research. Macro staffing research tends to focus on how HR strategy relates to use of HR practices, and/or bundles of HR practices contribute to firm effectiveness (broadly defined). For example, research suggests organizations that use certain bundles of HR practices, such as particular kinds of selection and compensation, show higher levels of organizational financial effectiveness (e.g. Delery & Doty, 1996; Huselid, 1995; Huselid, Jackson & Schuler, 1997; Koch & McGrath, 1996; MacDuffie, 1995). Although many of these studies do not take a fine grained analysis of micro staffing practices, there are a few notable exceptions. Terpstra and Rozell (1993) demonstrate the use of different staffing practices (i.e. assessment of recruiting, use of validation studies, structured interviews, cognitive ability tests, and biodata) relate to organizational profitability. Lepak and Snell (1999, 2002) argue there is no single best “HR architecture” for managing human capital, an important idea because it suggests certain HR practices cannot be considered appropriate for all situations. Finally, a growing body of research examines how nation and culture influence staffing (Brewster & Larsen, 2000; Schuler, Dowling & DeCieri, 1993). Note, however, very few of these studies focus specifically on the question of macro staffing as opposed to the more general topic of macro HR practices (which include staffing, but also other practices such as compensation).

Multilevel Staffing

137

Assumptions. Like the micro individual staffing perspective, the macro staffing perspective makes several important assumptions. First, it assumes there are important differences between organizations in terms of both staffing practices and measures of effectiveness. Second, it assumes the relevant constructs and level of theory are primarily at the organizational/unit level; individual level processes are rarely considered (see Wright & Boswell, 2002). Third, it often assumes reliable and valid measures of organizational practices can be obtained with a single informant (or small number of informants) from within each company. These informants are typically upper level executives who respond to questions about the organization’s staffing practices. The problem with using such informants is that their responses may have questionable reliability and validity (e.g. Gerhart, Wright, McMahan & Snell, 2000; Wright, Gardner, Moynihan, Park, Gerhart & Delery, 2001; see Huselid & Becker, 2000, for a different perspective). Fourth, this research tends to assume there are little to no individual differences within organizations, at least with respect to perceptions of staffing practices (Wright & Boswell, 2002). Such within-organizational variability may cause problems for concepts such as “human capital,” defined as the organization’s composition of aggregate individual-level KSAOs. It also leaves open the question, “How and why do macro staffing practices contribute to organizational effectiveness?” Finally, it is assumed there are a variety of contextual (e.g. industry) and organizational (size; technology) factors that need to be statistically controlled (but statistically controlled using within level methods like regression). Cross-Level Research (Meso Staffing) Major goals, questions, and problems. Cross-level staffing research is concerned with examining predictive relationships within and between two or more levels of analysis. This is the youngest staffing perspective and is still relatively uncommon, with most research restricted to the study of teams/groups and individuals (e.g. Jones, Stevens & Fischer, 2000; Klimoski & Jones, 1995). However, this research is important because recent studies suggest structuring work within teams is already prevalent (Devine, Clayton, Philips, Dunford & Melner, 1999) and is only likely to continue (Stewart, Manz & Sims, 1999). Most conceptualizations of groups and teams consider them in terms of an input-process-output model (e.g. Hackman, 1987; Marks, Mathieu & Zaccaro, 2001; McGrath, 1964). As relating specifically to team staffing, inputs include individual and aggregate KSAOs (team composition), processes include interpersonal and social factors (e.g. team cohesion, team potency), and outputs include such constructs as team performance and turnover. These relationships are often conditional on the type of team task (e.g. degree of interdependence). Neither individual level nor organizational level staffing research give as much attention to

138

ROBERT E. PLOYHART

mediating processes. Thus, the major goal of team staffing is to combine individual level KSAOs in some form to best enhance team level mediating processes and ultimately performance in an input-process-output model. Important questions include: What composition of KSAOs contribute to better team processes and performance? What are the primary team mediating processes? Are the same types of KSAOs and team composition structures equally predictive under different kinds of team tasks? How do teams influence individual level performance and vice versa? How does leadership and the leader’s role influence team processes and outcomes? Theoretical tradition. Research on groups and teams is steeped in social psychology and sociology. Research on social facilitation (Zajonc, 1965), ingroup and outgroup categorizations (Tajfel, 1982), group interaction processes (McGrath, 1984), group cohesion (Festinger, 1950), group decision making (Janis, 1972), group information processing (Hinsz, Tindale & Vollrath, 1997), and group task type (Steiner, 1972), among others, comprise the theoretical base. To a lesser extent, this research is also steeped in more traditional aspects of social organizational theory (e.g. Katz & Kahn, 1978). Interestingly, however, most of the recent research attention given to teams has been conducted in the organizational literature (e.g. Ilgen, 1999). Representative research. While there is a large amount of research on teams, very little of this research is focused specifically on team staffing. The few studies conducted examine how team composition influences team performance and mediating processes, where team composition is essentially the aggregate composite of individual level KSAOs (Guzzo & Shea, 1992; Hackman, 1987; McGrath, 1964). Note, however, composition is not always the simple within-unit average of KSAOs; it is sometimes best represented by the variability within a unit (Barrick, Stewart, Neubert & Mount, 1998; Chan, 1998). Important theoretical issues with team staffing are identified in several good sources (Jones et al., 2000; Klimoski & Jones, 1995). They propose that to correctly staff teams, one must first understand the nature of the team and why it exists because different types of teams require different staffing strategies. LePine, Hanson, Borman, and Motowidlo (2000) provide a valuable extension and refinement to these models. Empirical research has shown various forms of team composition influence a variety of team mediating processes, outcomes, and performance measures (e.g. Barrick et al., 1998; Campion, Medsker & Higgins, 1993; Campion, Papper & Medsker, 1996; Neuman & Wright, 1999). More direct research on staffing has identified a set of individual level KSAOs important for working in teams (Stevens & Campion, 1994), and has linked these KSAOs to teamwork performance (Stevens & Campion, 1999). Note

Multilevel Staffing

139

that the approach adopted by Stevens and Campion (1994, 1999) has focused on individual level KSAOs and individual level performance; they have not considered how such KSAOs aggregate into team composition. Thus, this research is really more similar to individual level staffing research. Assumptions. The meso staffing perspective makes several important assumptions, and because it is cross-level in nature, holds more assumptions than within-level research. First, it assumes that team inputs (e.g. composition) are presumed to flow through team level processes to influence team performance (e.g. Hackman, 1987; Marks et al., 2001). Second, it assumes there are important differences between teams in terms of inputs, processes, and outputs. Third, while the primary level of theory is at the team level, individual level constructs may also be important (to the extent they contribute to team level processes). Fourth, the level of measurement for many team processes can occur at the individual level and be aggregated to the team level. Fifth, these constructs can be assessed with valid and reliable measures. Finally, time is an important component in theorizing about team processes, although it has been less explicit in team staffing research.

LIMITATIONS OF THE SINGLE LEVEL STAFFING MODEL: WHY STAFFING RESEARCH MUST BECOME MULTILEVEL Many will argue staffing research need not focus on multilevel issues, that the dominant single level paradigms discussed above are entirely adequate and appropriate. I agree that continued single level research is important and necessary for an understanding of multilevel staffing. I disagree that single level staffing research is by itself sufficient, and challenge the profession to think about staffing more broadly. I support this challenge by building upon some observations raised earlier, to critique staffing research from a multilevel perspective and identify the limits of the single level research paradigm. First, the principle of near-decomposability (Simon, 1973) suggests single level approaches provide an incomplete picture of organizational staffing, and our ability to generalize findings beyond a single level are limited (i.e. doing so commits a cross-level fallacy). When we hire applicants with higher levels of cognitive ability, for example, we assume and hope this produces better functioning organizations. And when certain HR practices lead to better organizational performance, we assume it is because these practices are changing individual level behavior. But levels of analysis research teaches us the dangers of making such assumptions;

140

ROBERT E. PLOYHART

indeed this is the classic cross-level fallacy (Rousseau, 1985). How long can we sustain a science this way? Second, one consequence of the near-decomposability and bond strength principles is that trying to argue one level of research is “more important” than another is pointless. Each level is to a certain extent interrelated to those above and below it. Further, one should tend to find stronger effects when the predictor and criterion are at the same level of analysis (Simon, 1973). We should not be surprised by claims that macro HR staffing practices explain 20% of the variance in organizational level effectiveness, or that the relationship between individual level KSAOs and individual level performance are not highly moderated by contextual factors. In each instance, the effects occur within the same level. Third, that levels differ in their time metrics has important implications for how one would go about interpreting past research, generalizing across levels, and conducting multilevel staffing research. For example, the time asymmetry may require one to measure contextual variables prior to individual level criteria to find a cross level effect. Perhaps the difficulties in showing how individual level staffing contributes to unit level effectiveness occur because most studies are crosssectional in nature. We have some theories of how individual level performance operates across time (e.g. Murphy, 1989), and how individual and team interactions evolve over time (Kozlowski, Gully, McHugh, Salas & Cannon-Bowers, 1996), but very few perspectives on how other levels mutually influence each other across time. Fourth, micro and macro staffing paradigms will continue to remain separate unless a multilevel perspective is adopted. Shouldn’t findings in macro HR impact micro HR, and vice versa, as the principles of near decomposability and bond strength predict? Levels of analysis research and GST tell us there are mutual, cross-level influences on each set of within-level relationships. Yet micro and macro staffing continue to operate independently, as though the implications of one have no bearing on the other (Wright & Boswell, 2002). In both literatures, the vast middle ground between micro and macro HR is a cloudy set of assumptions and expectations. Likewise, what variance within organizations enhances or inhibits individual level relationships (e.g. Wright & Boswell, 2002)? Macro staffing research assumes that within organization variability is not important, yet somehow the individual level KSAOs and performance must aggregate to contribute to effective organizations. A related question asks how macro staffing practices influence lower level processes. Such macro practices are actually contextual effects for department, group, and individual level phenomena, yet we have seen practically no application of such cross-level staffing models. It is also worth noting that although micro staffing researchers have ignored the macro context for micro staffing research, macro staffing researchers have also ignored how contextual factors such

Multilevel Staffing

141

as national or regional culture are contextual factors that influence macro relationships (Anderson, 2003; Ryan, McFarland, Baron & Page, 1999). Such contextual factors are at best “controlled” for statistically, or at worst completely ignored, resulting in contextual fallacies. By focusing attention entirely within a single level of analysis, the precise mechanisms describing how organizational staffing practices contribute to better organizational effectiveness remain unknown. Fifth, the review of meso staffing research suggests such research often neglects levels of analysis issues and is frequently single level in nature. Although individual level research (e.g. Stevens & Campion, 1999) is extremely valuable, it has the same limitations of the single level model. Crossing individual and team levels requires careful consideration of such issues as level of theory, level of measurement, aggregation, and appropriate specification of the cross-level model. Furthermore, searching for direct links between team inputs and team outputs misses the very important linkages representing team processes. Ignoring such issues contributes to misspecification and cross-level fallacies, thus providing an incomplete picture of team staffing (see Klimoski & Jones, 1995; LePine, Hollenbeck, Ilgen & Hedlund, 2000). In studies where levels issues have been given careful consideration (e.g. Barrick et al., 1998; Chen & Bliese, 2002; LePine et al., 1997), findings both similar to and different than the individual level have been observed, suggesting that one should not directly generalize individual level relationships to the team level. Clearly team level staffing research and practice must adopt cross-level and multilevel models (LePine et al., 2000). Finally, the very nature of modern work often requires staffing to become multilevel. The increased use of teams as the organizing framework for performing work is well documented (e.g. Devine et al., 1999). Organizations frequently use short-term project teams that are disbanded after the project is completed, or teams composed of individuals with distributed expertise. Many organizations operate in multiple cultures, and oftentimes face challenges adapting staffing practices to different cultures. Most of the jobs in today’s world exist in the service economy (Bureau of Labor Statistics, 2001), where service providers must work in a changing interpersonal and cultural context. Jobs contained within small businesses blur distinctions between individual, group, and organizational levels. Finally, as staffing practitioners we are often held to standards at multiple levels, such as individual level validity, equal employment compliance, staffing system return on investment, and so on. Thus, if the levels of analysis research has any implications for organizational staffing, it tells us our existing paradigms: (a) may not produce findings that generalize across levels; (b) give each level a false sense of its own importance; (c) do not fully consider the importance of time in design and analysis; (d) do not communicate and reference each other; (e) ignore important levels principles and make

142

ROBERT E. PLOYHART

potential fallacies; and (f) have difficulty fully conceptualizing many “modern” forms of work. Yes, these single level paradigms have made great advances, but these are not trivial limitations. For these reasons, a complete understanding of organizational staffing requires a multilevel framework that unifies the levels within and across the organization. Anything less than a multilevel perspective provides an incomplete picture of the science and practice of staffing organizations. The following section considers previous conceptual work on this topic, followed by a proposed integrative model.

HISTORICAL FOUNDATIONS: PREVIOUS CONCEPUTALIZATIONS OF MULTILEVEL STAFFING This article does not present the first attempt to study staffing from a multilevel perspective. However, it is among the first to formally adopt a multilevel perspective on staffing. Most previous attempts to conceptualize staffing as a system were really more focused on macro issues related to organizational staffing. Dobbins, Cardy, and Carson (1991) discuss and propose a systems approach to staffing as opposed to an individual differences perspective. Jackson and Schuler (1995) discuss HR within the context of local, national, and multinational external influences. Ostroff and Bowen (2000) provide a multilevel model linking HR systems and firm performance. This model predicts lower level attributes are composed of employee KSAOs and contribute to individual level performance, whereas organizational level attributes are composed of collective attributes/human capital and contribute to organizational performance and effectiveness. Boudreau and Ramstad (2003) discuss an interesting model integrating HR practices with HR strategy. Of particular interest is their emphasis on understanding the linkages between practices and strategy to create organizational value. Schneider et al. (2000) suggest there are both individual differences (e.g. KSAOs) and organizational differences (e.g. HR practices), with individual differences contributing to organizational differences. What is most interesting in this model is how both individual and organizational differences are predicted to contribute directly to both individual and organizational performance (respectively), but only organizational differences are expected to moderate the relationship between individual differences and individual performance. Stated differently, organizations can facilitate or inhibit effective employee performance, and enhance or reduce the potential of skilled employees. Although the frameworks proposed by these previous researchers are valuable, they tend to ignore the intermediate areas between micro and macro staffing – the area that connects these levels together. Although it was not the intention of these authors to propose such formal linkages, in practice neglecting this important

Multilevel Staffing

143

middle level creates a cross-level fallacy and can lead to ambiguous, misleading, or incorrect interpretations. Ployhart and Schneider (2002a) articulated a staffing model that was based on an explicit consideration of levels issues and was designed to be prescriptive in nature. This model delineated how staffing might be conducted across levels of analysis and used the levels of analysis literature as a conceptual base. For example, they noted how validation strategies would change within a multilevel perspective. But the model still leaves many questions unanswered (Ostroff, 2002; Ployhart & Schneider, 2002b; Schmitt, 2002), and the need for a more comprehensive model.

A MULTILEVEL MODEL OF ORGANIZATIONAL STAFFING Based on the previous sections, we are now in a position to integrate these diverse issues into a single coherent model of organizational staffing. Figure 3 shows the proposed multilevel model of staffing. This model draws heavily from Ostroff and Bowen (2000), Ployhart and Schneider (2002a), and Schneider et al. (2000). I first briefly review the elements of model (structure), then describe its assumptions (process) and implications (function).

Structure: The Basics of the Model Following Ployhart and Schneider (2002a), the model distinguishes between levels of theory and domains of study. For simplicity, I focus on three levels of theory (individual, group, and organizational) that roughly correspond to micro, meso, and macro HR. However, please note that intermediate levels (e.g. dyadic) and higher levels (e.g. regional, national, cultural) are also operating in accordance with Fig. 1. The focus of the research question will determine which levels are appropriate, but the principles described below will still apply regardless of the level. Thus, rows in the model represent single level research. Fully crossed with these three levels are three domains: predictor, performance, and effectiveness. Columns therefore represent domains of study. The predictor and performance domains are both internal to the organization and largely under its control, while the effectiveness domain is external to the organization. The predictor domain represents those antecedents of performance, including individual level cognitive ability and personality, team composition, and HR practices/human capital. The performance domain consists of those outcomes or criteria that are within the organization, such as individual and team level performance. The effectiveness

144

ROBERT E. PLOYHART

Fig. 3. Multilevel Model of Organizational Staffing.

domain represents the kinds of outcomes most organizational decision makers try to accomplish, including such financial indicators as revenue, marketshare, profitability, and return on investment. Clearly performance and effectiveness are two different things (Smith, 1976), such that effectiveness tends to be less controllable and more influenced by a variety of environmental factors (e.g. economy, organizational size). Note, however, that both performance and effectiveness must be considered because macro staffing tends to use external effectiveness measures as the important dependent variables (e.g. revenue), while group and individual

Multilevel Staffing

145

level staffing tend to focus on internal performance measures (it is also true that both types of criteria are not easily assessed at each level). These domains are related such that the predictor domain contributes to the performance domain, which contributes to the external effectiveness domain. Relationships among different domains within and across levels are represented by arrows (the broad arrows at the periphery of the figure represent assumptions, discussed next). Horizontal arrows typically represent “causal” relationships (as inferred by correlations or regression coefficients), vertical arrows typically represent forms of agreement, aggregation, composition, or compilation. Notice that a consequence of vertical arrows is a nesting, such that lower level units are nested within higher level units (in accordance with Fig. 1). Also of interest are the mediated relationships between predictor and performance domains. These mediated relationships acknowledge that the effects of predictors on performance are not necessarily direct (a point we shall return to shortly). A final point about the direction of “causality:” the arrows are double-sided to allow forward, backward, or reciprocal causation. Thus, the arrows weld the three domains with the three levels of analysis. This means the multilevel staffing model is a hierarchical model; intermediate processes, functions, and cultures may occur between levels in some form of a hierarchy (e.g. Schaubroeck et al., 1998; Schneider, Goldstein & Smith, 1995; Schneider et al., 1998). So when examining relationships between individual and organizational levels, there are multiple intermediate levels through which these endpoints connect. To help illustrate this point, Table 3 provides examples of constructs and measures within each of the nine cells. For example, when focusing on the predictor domain, the important constructs may differ across levels such that individual KSAOs are most important at the individual level, the aggregation of these KSAOs forms group composition, and further organizational aggregation leads to human

Table 3. Examples of Constructs and Measures Across Levels and Domains. Level of Theory

Domain Internal

External

Predictor

Performance

Effectiveness

Organizational Group

HR practices human capital Group KSAO composition

Individual

Cognitive ability personality

Productivity Group performance efficiency Performance ratings accident rates

Revenue market share Unit sales customer satisfaction Individual sales customer service

146

ROBERT E. PLOYHART

capital. However, not all constructs will necessarily aggregate across levels, as HR practices at the organizational level cannot be disaggregated into lower level components – its level of theory exists only at the organizational level. Process: Assumptions of the Model The multilevel staffing model is based on the theoretical underpinnings of the levels of analysis literature. Therefore, the model contains 10 fundamental assumptions (which fortunately are testable): (1) Across each level and across each domain, there are reasonably stable differences that reflect important variance across people, groups, departments, organizations, etc. (2) Reliable and valid measures of this variability can be developed. (3) Any set of within-level relationships across domains can be placed on a continuum from distal to proximal (e.g. Kanfer, 1990). Such was important when discussing the input-process-output team model (Hackman, 1987) and is becoming increasingly important at the individual level (e.g. Chen, Gully, Whiteman & Kilcullen, 1999). Distal influences tend to represent more traitlike and stable constructs (e.g. cognitive ability, personality) and are farther removed from criteria than are proximal constructs, which are more statelike and dynamic (e.g. motivation, choice, self-regulation) (George, 1992). All else being equal, distal constructs operate through proximal constructs to influence criteria; therefore proximal constructs are likely to be more strongly related to criteria. (4) The temporal scale differs within levels, such that higher level processes operate on a slower time scale than lower level processes. (5) The temporal scale differs across levels in an asymmetric manner. Emergent (bottom-up) influences operate on a slower time scale than contextual (topdown) influences (Kozlowski & Klein, 2000). (6) Levels exist within a hierarchy, such that lower levels are nested within higher levels. There is a nesting or clustering of organizational data. (7) The principle of bond strength operates in the model, such that adjacent levels are more strongly and mutually related than levels farther apart. (8) As a consequence of bond strength, relationships across two or more levels are less likely to be direct but rather mediated through intermediate levels. For example, this means it will rarely be the case the organizational level has a direct influence on the individual level, but rather operates through the department and/or group levels (see Fig. 1). It is therefore important to use the model to frame cross-level staffing questions and specify the appropriate type

Multilevel Staffing

147

of cross-level model and the linkages through which cross-level relationships occur. (9) Effect sizes will tend to be strongest within levels than across levels. However, there is an asymmetry with cross-level relationships, such that contextual effect sizes will be greater than emergent effect sizes. (10) To answer cross-level staffing questions one must adequately sample a sufficient number of units or observations from each relevant level of analysis (e.g. Schneider, Kristof, Goldstein & Smith, 1997). For example, if one is trying to examine organizational differences in individual level KSAOcriterion relationships, one must have a sufficient number of organizations and people within organizations. These 10 assumptions, or perhaps better thought of as 10 commandments, apply regardless of the number of levels examined, types of levels or domains examined, or constructs and processes examined. In conjunction with these commandments, the model in Fig. 3 represents a broadly applicable way to think about and frame staffing research. It allows a way to assess what levels and domains results should generalize to and at what magnitude, as well as potential limitations and confounds of ignoring levels. It also allows one to frame the various multilevel models noted in an earlier section (single level, cross-level, homologous multilevel).

Function: Implications of the Multilevel Staffing Model Considering the model shown in Fig. 3 (structure) with these ten assumptions (process), we can now work through some of the major implications of the model (function). First, the model is integrative of micro, meso, and macro staffing, and allows one to compare and contrast the goals, questions, assumptions, models, and methods of past staffing research. These are compared graphically in Fig. 4. Macro staffing research is primarily single-level research, focuses on predictor and effectiveness domains, and largely ignores mediating processes. At the organizational level, staffing practices have been shown to relate to organizational level revenue (e.g. Huselid, 1995). Micro staffing research is primarily single-level research, focuses on predictor and performance domains, and largely ignores mediating processes. For example, micro staffing research has focused on linking individual predictor KSAOs to individual performance (e.g. Schmidt & Hunter, 1998). Meso staffing research focuses primarily on predictor and performance domains, but gives more attention to mediating processes and cross-level relationships (as denoted by dashed lines). At the group level, research has linked individual KSAOs

148

ROBERT E. PLOYHART

Fig. 4. Comparison of Dominant Macro, Meso, and Micro Staffing Questions Within the Multilevel Model. Dashed Lines Represent Cross-Level Meso Staffing Approach.

to group composition, and group composition to group performance (Barrick et al., 1998; Neuman & Wright, 1999). It is clear that across macro, micro, and meso levels, staffing researchers have not been examining the same questions in the same manner – the goals, questions, assumptions, models, and methods differ. Without making linkages across levels and mediating processes explicit, past staffing research tells us very little about how macro, micro, and meso staffing practices are related, and how staffing broadly defined contributes to performance and effectiveness across levels. Second, the proposed staffing model makes explicit the need to consider mediating processes linking predictors and performance criteria, and places these predictor constructs on a proximal-distal dimension (e.g. Kanfer, 1990; see also Binning & Barrett, 1989). We have seen how such a distinction is critical at the group level, where mediating processes such as group cohesion and interpersonal interactions are the very mechanism through which group inputs (e.g. composition)

Multilevel Staffing

149

influence group performance (Hackman, 1987; Marks et al., 2001). Research at the individual level also supports such proximal mediating processes, as the relatively stable constructs of cognitive ability and personality operate through such proximal processes as job knowledge, goal choice, motivation, and self-regulation (e.g. Barrick, Mount & Strauss, 1993; Barrick, Stewart & Piotrowski, 2002; Hunter, 1983). Shortly we shall see how acknowledging these mediating processes helps enhance understanding of contextual effects, as well as understanding what KSAOperformance relationships exist across levels. Third, while arrows connect constructs and measures between adjacent levels, no arrows are used to cross more than one level. The reason is because there are a variety of such cross-level relationships that can occur (see Fig. 2), such as mixed effects or mixed determinant models, and these directions of influence may be bottom-up and top-down. Further, the model assumes that direct (as opposed to mediated) links across two or more levels is likely to produce weaker effects than multilevel mediated relationships. Indeed, a search for direct relationships across domains and across different levels of analysis is unlikely to be fruitful if one adopts conventional criteria for statistically significant effect sizes. Rather, the model proposes that indirect, mediated relationships are what tie individual and organizational levels together. For example, the model proposes one cannot directly link individual level KSAOs to organizational productivity. Rather, this relationship would be indirect and mediated through such links as KSAO composition at the group and organizational levels, and/or links with individual and group level performance. Note, however, this should not prohibit an assessment of such links. For example, practical constraints may make it impossible to study how organizational level predictors operate through group level processes to influence individual level criteria (Schneider et al., 1995). Therefore, researchers can and should examine such direct relationships between organizational and individual levels, so long as they acknowledge the relationship is most likely mediated (i.e. the principle of near-decomposability). All of this also means we should not be discouraged by small cross-level effects between levels of analysis, they are likely to be the norm and we should be hesitant to say they do not matter. Fourth, as the model assumes differences in the temporal scale within and across levels, staffing research conducted at different levels must make explicit the time metrics expected to be operating at the level. Differences in the temporal scale of individual and organizational levels suggest it should take much longer for individual level performance to influence organizational level performance than the other way around. Yet the majority of staffing research remains cross-sectional or closely linked in time. For example, individual level staffing researchers need to acknowledge it will probably take many years for individual level staffing to influence organizational effectiveness (although less so to influence group level

150

ROBERT E. PLOYHART

performance). Likewise, macro staffing research must acknowledge there is likely to be considerable individual level (within organization) variability that is never recognized because the macro level operates on a much longer temporal scale. For example, the ASA model (Schneider, 1987) argues that homogenization processes operate over time, and truly longitudinal research is necessary to fully test this homogeneity hypothesis. Similarly, Schneider, Hanges, Smith and Salvaggio (2003) showed longer time periods are critical for supporting causal inferences – a cross-lagged time series model showed corporate financial effectiveness predicted organizational aggregate employee attitudes, but when ignoring these time lags the data supported an “attitude causes effectiveness” model. Modeling time at each level is thus critically important and has tended to be ignored in staffing research. Given a time asymmetry exists across levels, it would be helpful for research to more formally theorize and build time into the design of the study. Fifth, the inherent nesting of organizational phenomena suggests that our interpretation of effect sizes should consider how closely coupled the levels are in the hierarchy. For example, bond strength suggests adjacent levels will be more strongly intertwined that levels farther apart. We should find stronger relationships between individual and team levels than between organizational and individual levels. It is also the case that the consequences of our interventions should be greater the more strongly connected two levels are. For example, suppose one wanted to determine the unit level effects of a staffing procedure (e.g. quasi-experimental design). We should focus our attention to the level immediately above the individual level (e.g. group, team, department) to see the strongest and most immediate evidence for an effect. Sixth, the level of the intervention will determine the level of the effect. This is something the general public believes – hiring the “right” CEO will impact an organization’s effectiveness (e.g. stock value, profitability). Or for that matter, electing the appropriate government official will solve the collective’s problems (e.g. the California recall election and governor-elect Schwarzenegger). But scientifically, how does this occur? At what level does the CEO operate – individual or organizational? As with much levels of analysis research, the model predicts the appropriate level is defined by the level of the phenomena’s antecedents and consequences. So when a CEO makes decisions that impact an organization, his or her consequences occur primarily at the organizational level and therefore staffing a CEO position is said to exist at the organizational level. This means the level of the staffing practice in large part determines the level of the effect; the level of the effect desired answers the question at which level to staff. Seventh, the model allows a mechanism for testing single-level assumptions. As noted earlier the effects of within-level research on other levels are assumed, largely untested, and poorly articulated. For example, nearly all individual level

Multilevel Staffing

151

staffing research assumes improvements in individual effectiveness contribute to organizational effectiveness. Simply put, hiring better people leads to better organizations. This hypothesis may make intuitive sense, but has received practically no empirical support (Harris, 1994). Utility research suggests a relationship is likely to exist in some form (e.g. Jayne & Rauschenberger, 2000; Schmidt & Hunter, 1998), but in fact there is little direct empirical data supporting this relationship. Considering levels of analysis issues in staffing helps illuminate why. The principle of bond strength predicts that adjacent levels are more strongly intertwined and therefore only weak effects should occur between the individual and organizational levels. Furthermore, the differences in the temporal scale of individual and organizational levels predict that it would take much longer for individual level performance to influence organizational level performance than the other way around. Our intuition tells us individual level performance should translate into organizational effectiveness; multilevel theory and methods tell us this relationship is not likely to be direct or found over short periods of time . . . but it should be found. The model prompts a heuristic map and conceptual frame for such issues.

MODEL-BASED BETTING: SOME INTERESTING PREDICTIONS From a scientific perspective, the interest in a theory is not so much whether it is true as much as whether it prompts interesting questions (Davis, 1999). That is, truth is not interesting because it is already known (or is already known in a commonsense manner), trying to find truth is interesting because it confronts the unknown. To express the value of the proposed multilevel model, I highlight what I believe to be some of its more interesting predictions. Empirical research will clearly be necessary to determine their “truth,” but in science it never hurts to place your bets early on the biggest payoffs (e.g. gravity can only be discovered once!).

Prediction 1: The Determinants of Performance and Their Mediating Mechanisms Change Across Levels This simple prediction is profound in its implications. For example, if cognitive ability and conscientiousness are among our best predictors of individual level performance (Schmidt & Hunter, 1998): (a) do these characteristics cluster and aggregate to higher levels such as teams, departments, and organizations; (b) do teams, departments, and organizations that have more intelligent and conscientious employees perform better and more effectively; and (c) are these unit level

152

ROBERT E. PLOYHART

criterion-related validities similar in magnitude to the lower level validities? Only very preliminary research has examined these questions. Research has found jobs, occupations, and organizations can be differentiated in terms of their personality (Ployhart et al., 2003; Schaubroeck et al., 1998; Schneider et al., 1998), but it has not shown the unit-level consequences of such personality homogeneity. Team research suggests findings at the individual level of analysis may not be the same in magnitude as the unit level, and different predictors may be relevant only at higher levels of analysis. For example, Barrick et al. (1998) found agreeableness and emotional stability were actually more “important” at the group level than at the individual level, partly because they influenced group mediating processes not relevant for individual level performance. Other research has shown how cognitive ability is not nearly so dominant of a predictor in teambased performance contexts (Devine & Philips, 2001; LePine et al., 1997; 2003; Neuman & Wright, 1999). A program of research by Chen (Chen & Bliese, 2002; Chen, Webber, Bliese, Mathieu, Payne, Born & Zaccaro, 2002) has shown how both the determinants and consequences of efficacy change across levels. It therefore appears that as the nature of performance changes across levels, the proximal and distal predictors must necessarily change as well. So should we expect organizations with smarter and more conscientious employees to perform better than those who do not (the organizational analog of the frequently argued individual level relationship)? Should we expect the magnitude and/or direction of this relationship to be the same, or operate through the same mediating processes? Results to date both support and refute such expectations (Chen & Bliese, 2002; Chen et al., 2002; LePine et al., 2000). We should therefore be hesitant to assume findings at the individual level directly generalize to higher levels (and vice versa) because to do so commits a cross-level fallacy. Indeed, based on the literature available, one might actually predict the impact of KSAO composition will be inversely related to level, such that as one moves up levels of analysis, more mediating processes come into play and the direct effects of KSAO composition become weaker (as Schneider et al., 2003, would term “extendedmediated models”). Therefore, the effects of KSAO composition become weaker as one moves up levels because they are farther removed from the criterion. But this thinking runs counter to much of the macro HR literature that examines – and finds – direct relationships between HR practices (or organizational attitudes) and organizational effectiveness. Here is a disconnect in need of resolution (see Schneider et al., 2003, for one possibility). Perhaps the key agenda for future research is to determine whether the findings from single level research exist across levels; this research must adopt cross-level and homologous mixed level models to answer these questions. Methods for analyzing such relationships are just starting to appear (Chen, Bliese & Mathieu, 2003;

Multilevel Staffing

153

Chen, Mathieu & Bliese, in press). Can anyone imagine the impact on staffing if it was found that departments or organizations with more cognitively able members do not perform better? The search for understanding why would be exciting and interesting!

Prediction 2: Effective Micro Level Staffing Might Not Help, and Could Sometimes Hurt, Organizational Effectiveness If the previous prediction bears some support and the determinants of performance change across levels, let us take this implication to the extreme and consider a situation where at least some of the determinants of performance are different at each level. For example, suppose we find no practically meaningful criterion-related validity for the personality trait openness to experience at the individual level (Barrick & Mount, 1991), but this trait (in aggregate) contributes to effectively working with co-workers and hence unit performance (LePine, 2003). Provided the KSAOs cluster within units (remember, an assumption of the model), excluding this trait from an individual level predictor battery will result in little improvement in higher level effectiveness. Further, what if some of the predictors that are used for selection are negatively related to unit level effectiveness? Could such an instance happen? Possibly, if for example the organization used test methods (e.g. interviews, assessment centers) that assess multiple constructs, and the test method correlated negatively with openness but positively with individual level performance. This would be an interesting example where effective individual level staffing might actually hurt higher level effectiveness.

Prediction 3: There is a Positive Relationship Between Level of Staffing Intervention and Magnitude of Organizational Impact Hiring a CEO has a greater organizational impact than hiring an entry level employee, team manager, or department manager. This is obvious to all, or at least has the appearance of being obvious. However, that scientific principles can be used to explain and predict why the level of staffing intervention positively correlates with the magnitude of organizational impact is a strong position that articulates this commonsense thinking. Further, it has a subtle and powerful implication – when thinking about how to best staff the organization, start from the top down. When thinking about how to best improve the effectiveness of a team, start from the top down. The model predicts the fastest way to change a unit is to change its management first. To those in practice and industry, no surprise there, but the model

154

ROBERT E. PLOYHART

allows a means to test this commonsense hypothesis in a manner to find support for it. Also recognize an interesting implication – the model makes no claims that these top level executives are “born leaders,” inherently better than the masses, or anything similar. Good or bad, the organizational consequences will be greater depending on who fills that higher level position (true enough, certain KSAOs are necessary to succeed in the position, but the model places the consequences with the position, not the person). It is an interesting idea to think that hiring one (or a few) key people can do more good or harm than hiring several hundred (or thousand). It is an idea many front line Allied infantry soldiers from World War I would have agreed with; early in the war their leaders showed no reservations about losing entire units in pointless attacks into machine-gun fire. Note this prediction is also consistent with the ASA model. The ASA model predicts the founder’s personality has a strong and lasting impact on the organization’s structure, processes, and culture (Schneider, 1987). The present model extends this thinking beyond merely the founder; it argues that any person in such a pivotal role (whether founder or newly hired CEO) will have this powerful impact on organizational structure, processes, and culture. However, this prediction may be counter to the model discussed by Boudreau and Ramstad (2003), which articulates several instances where lower level employees may actually have more strategic value to an organization. A competing hypothesis is a bet you can’t lose!

Prediction 4: Roles, Particularly the Leadership Role, Operate at the Level of Their Consequences Our colleagues who study leadership often bemoan the “simple” approach staffing researchers take when predicting leadership performance. Most frequently, the validation study will measure some KSAOs that are expected to predict leadership performance, and the criterion will be individual level ratings of leadership behavior or effectiveness. Given the model predicts leadership should have its primary effect at the unit level (e.g. group, department, organization), shouldn’t staffing research use such unit level indicators as criteria? The model answers this question “Yes!”, but apparently most previous research has not. A recent meta-analysis of the leader personality-leadership effectiveness literature did not identify a single study that used group level performance as the criterion (Judge, Bono, Ilies & Gerhardt, 2002). A handful of studies since then have shown relationships between transformational leadership behavior and unit performance (Bass, Avolio, Jung & Berson, 2003; Dvir, Eden, Avolio & Shamir, 2002; Lim & Ployhart, in press). Interestingly, research has even shown how chief executive officer personality relates (indirectly through top management team processes) to organizational performance (Peterson,

Multilevel Staffing

155

Smith, Martorana & Owens, 2003). These results are supportive of predictions based on the model and suggest we should expect to find leader-unit performance relationships, if that validation study adopts multilevel methods and models.

Prediction 5: Leadership “Matters More” Than Unit Composition The multilevel staffing model suggests we should evaluate leaders in terms of their impact on unit processes and unit performance; a focus on individual level leadership behavior is simply at the wrong level of theory. Consider a potentially interesting consequence of this statement by using the example of a team leader. Within the model, the leader impacts team processes, performance, and effectiveness. Because the model predicts team mediating processes are the proximal and primary determinants of team performance, and the leader impacts team processes (rather than KSAO composition, which is more fixed and distal), an interesting question is whether the leader, as opposed to team KSAO composition, has a faster and/or stronger impact on team performance (both are at the team level). Assuming team members have the minimum competencies, one safe bet is that both are important but not equally so. Leader behavior is no doubt more flexible, adaptive, variable, and perhaps more proximal than unit composition. A riskier bet asks which matters “more;” if given a minimum level of KSAO composition and competence I put my money on the leader. Of course, all such bets are off when recognizing the type of team task may moderate these relationships (Kozlowski & Bell, 2003; LePine et al., 2000).

Prediction 6: The “Diversity vs. Prediction” Problem Cannot Be Answered at the Individual Level There is much focus within the micro staffing literature on questions relating to enhancing diversity vs. maximizing prediction. The problem is that our most predictive tests (cognitive ability) produce sizable racial subgroup differences, and using such tests will result in adverse impact against minorities in most realistic selection contexts (e.g. Hough et al., 2001; Sackett & Wilk, 1994). As a result, an impressive amount of research has been directed towards understanding why such differences exist, and how the differences can be reduced (Ryan, 2001; Sackett & Roth, 1996; Whitney & Schmitt, 1997). This research is important and has contributed to strategies for reducing subgroup differences and adverse impact. However, the multilevel staffing model predicts such single level research cannot adequately conceptualize this problem because diversity is an inherently unit

156

ROBERT E. PLOYHART

Fig. 5. Multilevel Framing of the “Prediction vs. Diversity” Problem.

level phenomenon. Ethnic, racial, gender, personality, values, or any other kind of diversity cannot exist at the individual level of analysis and cannot be disaggregated to lower levels (e.g. what is diversity for a single person?). Just as the aggregation of within-unit KSAO scores (mean, variance) reflect unit level KSAO composition, so too does the aggregation of racial (or gender, etc.) makeup reflect unit level diversity. Diversity can only exist at levels above the individual level (e.g. group, department, organizational) because it represents a construct of similarity/dissimilarity to other group members. In typologies of emergence (Kozlowski & Klein, 2000), it is a compilation model (Bliese, 2000). And so individual level staffing attempts to address diversity concerns are only indirectly related to the level of interest, but the dependent variable (diversity) is at a higher level and an entirely different construct. For example, the classic statistic used to calculate diversity is Blau’s (1977) index of heterogeneity (which is simply one minus the (squared and summed) proportions of demographic minority groups). When one calculates an adverse impact ratio (i.e. [minority pass rate/majority pass rate]), one is likewise calculating an index of unit level diversity, but diversity defined in terms of pass rates. This is illustrated in Fig. 5. When we focus on criterion-related validity, we examine the individual level relationship between test scores and the criterion. When we calculate adverse impact by applying the adverse impact formula, we essentially define diversity in terms of pass rates (a compilation model). Clearly

Multilevel Staffing

157

individual level staffing procedures influence the unit adverse impact estimate; adverse impact is a unit level consequence of selection on individual level KSAOs. But when asking whether we should focus on validity vs. diversity, where do we see the consequences for the unit level diversity construct? Many of those who study diversity understand it relates to a host of outcomes at multiple levels (see Alderfer & Sims, 2003), but such unit level criteria are typically missing from the individual level staffing model (as denoted by the box with question marks). Yet many practitioners will be thinking of that often unexamined unit level criterion stemming from unit level diversity. Ployhart and Schneider (2002a) discuss these issues in some detail; here I offer a couple of additional observations. First, as Fig. 5 shows, focusing solely on an individual level performance criterion is not the best choice when framing the diversity vs. prediction problem. Race, ethnicity, and so on, should not typically have meaningful relationships with the usual kinds of individual level job performance criteria derived from professionally-developed methods (provided there is no subgroup difference in the criterion). Therefore, considering it in an individual level staffing model makes little sense. When the content of a predictor battery is influenced by diversity concerns (e.g. weighting noncognitive tests more than cognitive tests lowers validity but also reduces subgroup differences), many would argue such practices mix politics with science. And perhaps they are right; with a narrow individual performance criterion, what in the model says diversity is important? Rather, those who follow such practices realize the individual level criterion is only one criterion of interest (they are implicitly looking at that box with question marks in Fig. 5). For example, diversity (in terms of ideas, backgrounds, race, intelligence) may be a good thing for project teams who need to generate creative solutions, perhaps even more so than the composition of KSAOs (this is essentially arguing for a frog-pond model; see Fig. 2). Because diversity is an entirely different construct than the individual level KSAOs (a compilation model), it may relate to entirely different kinds of criteria irrespective of relationships to the KSAOs. The recent Supreme Court’s ruling on the University of Michigan’s admissions practices was in part based on such arguments, that a “critical mass” of diversity produces educational benefits – this critical mass is a unit level phenomenon. Interestingly, this case could be analyzed from the multilevel model to better express the level at which both proponents and detractors were operating. Those who supported the university’s decision argued points that were most often unit level or cross-level phenomena; those who opposed it argued points that were most often solely at the individual level. These two sides were not talking about the same level and outcome. Likewise, academic programs rarely hire faculty who have the same expertise as existing faculty; they hire faculty who bring unique sets of expertise.

158

ROBERT E. PLOYHART

This is not to push a diversity agenda; one could also find diversity has a negative impact on team mediating processes and hence performance. For example, Sacco and Schmitt (2003) found the match between store ethnic composition and local region ethnic composition had no effect on store profitability. The point is that if diversity is a higher level construct, we need to think of it in terms of multilevel models. Second, if diversity is a construct that only exists at a higher level, might other unit level criteria become important in how we test the diversity vs. prediction problem? What if we found unit cognitive ability was not predictive of unit performance – then the “problem” might become less important than what it would seem at the individual level. If an organization says it wants a diverse workforce because it contributes to better perceptions of the organization by the general public, then why not correlate unit diversity with aggregate public perceptions to determine whether such effects exist (similar to research linking unit service climate with aggregate customer satisfaction; Schneider, White & Paul, 1998)? If a primarily white police department operates in a predominately nonwhite district and desires to have more minority officers (to improve community relations), why not examine the relationships between racial composition of police departments with aggregate perceptions of those departments (or complaints and investigations) to support or refute the claim (e.g. Lovrich, Steel & Hood, 1986; Steel & Lovrich, 1987)? Notice that in each instance the new criterion occurs only at the unit level, and thus there is a need to consider criteria at multiple levels. The idea of diversity being a higher level construct, and staffing research needing to consider it at this level, hopefully removes the feeling that such questions are battles of “science vs. values.” Indeed, the multilevel model allows one to test questions about diversity and prediction empirically; there is no need for values to play a bigger role in this picture than it does in any other type of empirical research. Those who argue for the individual level criterion define their objectives within that level and correctly note diversity should have little place in developing predictor batteries – the only golden standard is the validity coefficient. But those who practice staffing often also define their objectives at a higher level and argue for the inclusion of diversity (perhaps at the expense of a few points of validity); they are thinking about the unit level consequences of diversity (the mystery box in Fig. 5). This “debate” cannot be settled unless both sides adopt a multilevel perspective, and it would be interesting for a debate to be held using the model as a way to frame the issue. Following the work of Sacco and Schmitt (2003), it would be especially interesting to see if adverse impact statistics correlate with unit level criteria (e.g. organizations using staffing procedures with more adverse impact show greater degrees of aggregate dissatisfaction), suggesting that the unit level consequences of individual staffing cascade through the organization.

Multilevel Staffing

159

Prediction 7: (a) The Form of KSAO Composition is a Consequence of HR Strategy; (b) HR Strategy is a Consequence of Management KSAO Composition As a contextual influence, HR strategy influences and creates a certain type of emergent process – a certain form of composition. But strategy research is almost always focused on more macro processes, structures, and practices. Perhaps when thinking about HR strategy we must ask the question, “What kind of KSAO composition do we want to build?” Both the level at which the staffing intervention takes place, and the type of strategy adopted, will influence the type of composition found. For example, Lepak and Snell (1999, 2002) argue that certain types of HR practices are most appropriate for certain types of employment modes (knowledge based vs. contract work), but the choice of employment mode is dependent on the value and uniqueness of the employees. Therefore, for employees that differ in their strategic value and uniqueness to the organization, certain kinds of HR practices will be most effective and thus we would expect different forms of composition to also differ across value and uniqueness configurations. That is, organizations that adopt different strategies should show different types of composition in the relevant KSAOs because they use different HR strategies. Synthesizing the Lepak and Snell (1999) framework with research on composition models should be an important and exciting area of integration. An interesting extension from this integration would be to study organizations that use different HR strategies and determine if composition differs in each organization (or for that matter, within an organization when various employees are treated differently). Relatedly, this prediction can be turned around such that one might ask whether HR practices are a consequence of certain types of composition. If the organization becomes staffed with people who are highly achievement oriented, would staffing strategy, practice, and procedure look different from organizations who are filled with people who are less achievement oriented? The ASA model (Schneider et al., 1995) and preliminary research by Wilk and Capelli (2003) and Ryan et al. (1999) suggest such organizational differences should be found. Likewise, Peterson et al. (2003) showed how leader personality influences organizational performance through influencing top management team processes. Clearly the KSAO composition of the team and leader personality interrelate as well. Another, less obvious implication is that if organizations adopt different staffing strategies at different levels (e.g. the staffing strategy for management seeks to hire people with diverse skills and expertise, while the staffing strategy for the front line employee seeks to hire people with similar skills and expertise), different forms of composition would be found at different levels within the same organization. This in turn prompts such questions as whether the KSAO composition of the management team relates to various outcomes differently (stronger, weaker) than

160

ROBERT E. PLOYHART

the KSAO composition of aggregate employee data. For example, might aggregate cognitive ability for managers more strongly predict unit level effectiveness than aggregate cognitive ability for front line employees, even after controlling for differences in levels? Certainly any such differences are influenced by higher level contextual factors such as industry type. This prediction also begs the study of many unanswered questions. Of primary interest is how individual differences “roll up” to create human capital and thus organizational differences: What is the appropriate type of composition or compilation model to consider for these questions? Do different types of strategy, organizational structure, or possibly even industries produce different types of composition and compilation? Are different KSAO compositions optimal for different levels, job types, or team types? Do emergent and contextual (strategy and HR practices) processes operate through different “columns” in the model – an emergent-contextual version of “shoots and ladders.” For example, might the internal performance domain show stronger composition and nesting because more HR practices shape this column (e.g. selection, compensation, training) to a greater extent than the predictor domain (e.g. selection). Prediction 8: Concurrent Validation Designs Are More Important Than Predictive Validation Designs In staffing practice, predictive designs are often thought to be “better” than concurrent designs because they most directly ask the question of interest: do test scores of applicants predict performance for incumbents. To the extent such factors as applicant reactions, motivation, and social desirability influence test scores (McFarland, in press), one might expect differences between the two designs (although empirical data suggests the differences are typically not large). But notice concurrent designs are necessary to build and test the multilevel staffing model. Unit level or multilevel homologous attempts to estimate validity are already starting to appear and producing some interesting findings (Sacco, 2003; Sacco & Schmitt, 2003). Quasi-experimental staffing designs, where the research is conducted across several units within the organization, would be ideal (Schneider et al., 1997). In many large organizations, multiple units (e.g. store locations, departments) are chosen to participate in the validation design, but how they are chosen is often not a formal part of the validation design and analysis. It is only a little more work to include them as a formal part of the design and analysis (e.g. Do test scores cluster within stores? Do such aggregate scores correlate with store indices of effectiveness?). All of this leads to a need for more concurrent designs, with the predictive design being used to ensure incumbent test scores are equivalent to applicant test scores (a quite narrow single level question).

Multilevel Staffing

161

Prediction 9: The Model Provides an Alternative to Utility Analysis While there have been many advances in utility analysis (Boudreau & Ramstad, 2003; Cabrera & Raju, 2001), it is has not had the impact on practice that many would have hoped. In fact, one could argue that when researchers use utility analysis estimates to make inferences about organizational benefits and costs, they are committing a cross-level fallacy because most utility models are on estimates entirely derived from the individual level (validity, Sdy)! What if we took multilevel staffing seriously and considered a different approach? When incumbents are sampled from multiple units (e.g. departments, workgroups), aggregate KSAOs to the unit level and link these aggregate variables with unit level measures of performance and effectiveness. Boudreau and Ramstad (2003) discuss some variations on this theme, and Terpstra and Rozell (1993) have already shown how organizations that use well-developed selection procedures are more effective. For consulting companies that implement similar staffing practices across many organizations, why not link the aggregate organizational-level KSAO scores, and aggregate performance scores, to indices of organizational effectiveness (see Sacco, 2003, for an illustration). Figure 6 shows what such a model might look like. Notice that one must find the correct emergent model to justify aggregation to the organizational level, which then would allow one to examine the relationship between organizational KSAOs and organizational performance/effectiveness. That would directly show the monetary organizational value of selecting on those KSAOs (i.e. the value of KSAO composition). Indeed, showing KSAO differences in organizations contribute to differences in corporate financial performance would be a compelling form of evidence supporting the test battery’s utility – no need for assumptions about SDy, baserates, or anything similar, just a set of empirical links built throughout the levels of the organization. It would sure be an interesting way to show the value of selection!

Prediction 10: The Organizational Direct Effect: Implications for Validity Generalization Validity generalization research, and meta-analysis more generally, is compelling in that it demonstrates criterion-related validity for many predictors is not context specific (or at least not dramatically affected by context). That is, there is not a strong contextual moderator of an individual level predictor-criterion relationship and the implication is that validity generalizes across contexts. This is consistent with the model’s predictions; within level effects are stronger than cross-level contextual effects because the within-level relationships are more tightly related

162

ROBERT E. PLOYHART

Fig. 6. An Alternative to Utility Analysis.

and operate on the same time metric. However, Schneider et al. (2000) argued that meta-analyses have not considered whether organizations produce a direct (rather than moderated) effect on individual level performance. It would be interesting for future meta-analysis to examine both moderated and direct/mediated effects of contextual factors on individual level relationships. It may very well be the case that validity is not affected by context, but the aggregate KSAOs of an organization (as determined by the staffing system) account for differences in unit performance. For example, Ployhart et al. (2003) found the individual level personality-job satisfaction relationship was not moderated by contextual factors (i.e. “validity” generalized), but aggregate personality had a direct effect on satisfaction such that units with co-workers who were more conscientious and agreeable were more satisfied and explained between-unit differences in satisfaction. The “validity” for personality generalized, but job/organizational differences in average personality existed and explained differences in job satisfaction beyond

Multilevel Staffing

163

the individual level personality variables. So unit differences were not detected with the individual level correlation coefficient. An implication is that validity generalization research does not suggest context is unimportant, but rather context may influence the intercept rather than a regression (validity) coefficient. It would also be interesting if rather than conducting a meta-analysis in the usual fashion, one instead tried to hypothesize and measure the types of unit level constructs that might moderate lower level relationships. To date, nearly all meta-analyses search for moderators based on experts’ coding of various design features (e.g. type of industry). But consider a study where such contextual factors were measured. For example, assessments of organizational service climate being modeled such that organizational service climate might moderate the relationship between individual level service orientation and service performance (and/or produce the direct effects mentioned above). If conducted in this manner, research might actually find stronger contextual moderator effects because we would no longer be going on a fishing expedition for moderators (casting a large net) but instead using a fishing rod with the right bait (theory-driven assessment of contextual moderators).

Prediction 11: Contextual Effects Primarily Influence Mediating Processes, But Mediators Increase in Number and Strength As One Moves Up Levels The previous prediction argues for an organizational direct effect on lower level performance criteria. There is another way contextual factors may influence lower level criteria – through their mediating processes. This is one reason it is so important to consider the distal-proximal dimension and mediators within even single level models. Because mediating processes are more state-like, they are more likely to be affected by context (George, 1992). And because they are more proximal for the outcome of interest, contextual influences operating on mediating processes may substantially reduce (or enhance) performance, and therefore indirectly influence criterion-related validity. If this reasoning holds, it would be interesting for research to search for unit level differences in individual level mediating variables: unit level differences in average job knowledge, skill, and other mediators. We should find the unit mixed model (Fig. 2), for example, where store differences in average job knowledge contribute to differences in unit level performance, irrespective of individual level KSAOs (which should show relatively constant relationships, as based on validity generalization studies). At the very least, we should find less validity generalization for these mediating variables. But there is a more subtle and perhaps more interesting prediction here. I reasoned in Prediction 1 how the magnitude of direct effects for KSAOs are

164

ROBERT E. PLOYHART

inversely related to level. If this is correct, an implication is that more or stronger mediating processes occur as one moves to higher levels. So does the distalproximal continuum “stretch” as one moves up levels of analysis? That is, if we equate each level on the distal-proximal continuum, do we find that KSAO composition becomes increasingly more distal as we move to higher levels because there are more mediating processes? Or is it simply the case that these mediators are just stronger at higher levels, and thus the distal-proximal continuum is unchanged?

Prediction 12: Micro Staffing Interventions Should Be Accountable for Unit-Level Effectiveness In micro staffing research, many claim individual level performance should be defined in terms of behavior rather than results or effectiveness (Campbell, 1990; Campbell et al., 1993; Motowidlo, 2003). The reason is twofold: our interventions are behavioral and so our effects are strongest for a behavioral criterion, and results and effectiveness are influenced by contextual factors unrelated to behavior. There is nothing wrong with this reasoning, but it reminds one of a common cause of the criterion problem: “The only criteria we use are those we can measure and predict.” We may rely on behavioral criteria because we can predict them and show reasonably strong effects, but the fact remains management is frequently more interested in higher level effectiveness criteria. The model allows a framework for thinking about connecting behavioral criteria to unit level effectiveness. We can show the effects of behavior on such “hard” effectiveness criteria, although the relationship will not be direct but mediated within and across levels.

Prediction 13: Criterion-Related Validity Is Not Always the Primary Coefficient This article has focused primarily on conceptual issues and excluded methodological ones (see Ployhart & Schneider, in press, for an description of such methodological issues in staffing, and how to conduct a multilevel staffing study). However, one methodological point to consider is whether in a multilevel staffing world the criterion-related validity coefficient will maintain its kingly status. Because issues relating to aggregation and the testing of composition models will often become issues relating to assessing agreement (whether through rwg, intraclass correlations, or related indices), these other coefficients may actually become more important for some questions. For example, justifying the link between individual level KSAOs to unit level KSAOs relies almost entirely on these agreement indices and demonstrating there is sufficient agreement to warrant aggregation (this

Multilevel Staffing

165

of course is just one type of emergence). Bliese (1998, 2000, 2002) has written several thoughtful and relatively non-technical articles about correctly using and interpreting these coefficients. While it is unlikely the criterion-related validity coefficient will lose its important role in staffing research, it would be interesting to read a validation study where intraclass correlations were also evaluated and used to form inferences of validity.

Prediction 14: The Model Provides a Translation for Research and Practice Anderson (in press) notes the inherent tension between staffing research and practice, and how such a tension must be managed. We have all no doubt heard practitioners lament the difficulties with applying research findings to practice. Oftentimes the purpose of the research was not to provide such direct implications, but to enhance knowledge more generally. It is also probably safe to say some of these concerns arise because they are thinking of levels higher than the purposes and implications of the research under question. There are many examples of test user manuals, technical reports, and so on, that attempt to express the value of micro staffing practice on some indicator of unit level effectiveness. So if practitioners must sell our science using outcomes that exist at higher levels of analysis, but research remains primarily individual and single level, how can the implications of such research be more clearly expressed? Perhaps through using the proposed multilevel model to articulate such linkages. To glue staffing research and practice together, we must conduct research at the fringes of our dominant single-level paradigms.

Prediction 15: Refining Micro Staffing Practices Will Not Make Much of an Organizational Difference To date, nearly all macro HR research examining the organizational consequences of micro practices (e.g. use of biodata, cognitive ability) has only considered whether organizations use these predictors. Therefore, most macro research treats within-organization (or within-HR practice) variance as error variance. Such within-group variance is often called “nuisance variance” by statisticians, but this “nuisance variance” comprises much of the field of micro HR! Surely this variance is worth exploring, but the macro HR literature has not examined whether using different and better versions of the same predictor matters in any practical sense (e.g. a generic vs. work-specific personality test). Yet micro research is dominated by such questions. An interesting question asks, “How much of a validity difference

166

ROBERT E. PLOYHART

must exist at the individual level for it to matter at higher levels?” Classical utility analysis suggests any non-zero increase in validity results in greater utility, I would be suspect of such direct linear improvements in a multilevel framework. Indeed, such simple utility models do not capture strategic value (Snell, Youndt & Wright, 1996). So a question becomes, does a validity increase of say 0.05, or even 0.10, “appear” in some fashion at higher levels of analysis (e.g. faster emergent processes, greater homogeneity of KSAOs, unit level differences)? For that matter, in what higher level processes, structure, or functions should we see the effects of stronger validity coefficients? Relatedly, are the types of predictors used in a staffing procedure related to organizational level factors? Ryan et al. (1999) demonstrated staffing practices are affected by national culture, and Wilk and Capelli (2003) found staffing practices differ across organizations due to the nature of work involved. Research should continue the exploration of such contextual factors in an increasingly theoretical manner. For example, what is the nature of the contextual influences that determine which staffing procedures are adopted by organizations? At what level of cultural generalizability are these findings based (Anderson, 2003)? It would be quite discouraging if such research led to the conclusion that the magnitude of validity differences we find at the micro level are not translated into similarly sized improvements in organizational level outcomes (whatever those outcomes might be). Would this suggest micro staffing practices don’t matter much, so long as they are developed according to good practice? Might it mean we need to find really whopping differences in validity (say, 0.30 difference) before we see changes at higher levels? Or would it suggest such improvements are not directly obvious, that they occur slowly over time through a variety of mediating processes? Micro staffing researchers work hard to achieve relatively minor improvements in validity, it seems worthwhile to occasionally examine the extent to which this effort contributes to better functioning organizations.

Coda In the previous sections I have attempted to articulate what I see as some of the most interesting bets of the model. With the exception of the single level individual and unit-global models, and the cross-level moderator model (metaanalysis), the remaining models in Fig. 2 have been almost entirely ignored from staffing research and quite possibly practice (the emphasis is on the term staffing because other domains of study (e.g. job attitudes) have adopted such models). Conducting research on nearly any one of these models could be an important contribution. Yet there are many other interesting questions prompted by the model,

Multilevel Staffing

167

and as research begins to examine these questions it will no doubt identify several contradictions. But I do not mean to imply the model answers all questions, is complete, and so on. Obviously it leaves out much of the detail and complexity that must be addressed to link levels together (see Chen et al., in press, 2003; Kozlowski & Bell, 2003; Kozlowski & Klein, 2000, for a discussion of multilevel methodological issues; see Ployhart & Schneider, in press, for a discussion of staffing specific methodological issues). However, that is not the purpose of the model. Rather, it is to provide a framework for thinking about multilevel staffing issues broadly defined, to stimulate research on such issues, to know the limitations of single level research, and to direct the mind to address those limitations. I would also like to acknowledge the challenge of the multilevel staffing model to current practice. Unit level validation studies may be difficult to conduct for many researchers, and the consequences of such research may run counter to existing legal standards which retain a focus at the individual level. But the integrated data management systems within many organizations make such research increasingly feasible. More important, if the model has merit and if its implications improve the science of staffing organizations, surely they must influence its practice as well. As scientists, we have the potential to influence public policy and research supporting the model may help change policy in a manner more consistent with better science and practice.

CONCLUSION Nearly a century of selection research has greatly contributed to our understanding of how to best predict and understand individual level performance and organizational effectiveness. However, the dominant micro and macro paradigms remain single level in their orientation and implications: smarter and more reliable employees perform better, organizations that use systematic and well-developed staffing practices are more effective, and almost everything in between these levels is assumed. The research being conducted in these single level models is very much a program of “refinement” (e.g. implementation of web-based testing, examining the best HR structure for a particular situation, etc.). I have and will continue to conduct such research myself; but I must now say let us push these single-level findings to their limits. Do smarter companies perform better? Does the individual employee work better or remain more satisfied in organizations using particular bundles of staffing practices? Can we show how hiring smarter employees leads to better organizational performance? Let us focus our research on the seams of these single level paradigms and push into the fringe of our literature base. Let the programs of our staffing research ride those uncharted waters that lie at the edge

168

ROBERT E. PLOYHART

of our understanding. Surely there must be some brave souls willing to illuminate these dark places with the fires of a multilevel staffing revolution!

ACKNOWLEDGMENTS I would like to thank the many people who have provided helpful comments, suggestions, discussions, and challenges to this article: Talya Bauer, Mark Schmit, Neal Schmitt, Ben Schneider, and members of the Personnel and Human Resources Research Group. The multiple iterations of feedback and discussion with Gilad Chen, Lynn McFarland, and Josh Sacco were invaluable. Finally, I appreciate the help of Michael Camburn for compiling and organizing many of the articles discussed in this article (often with little advance notice!). While I owe all of these individuals my sincere thanks, certainly not all have agreed with everything I’ve proposed here, and any criticisms with this article are my responsibility.

REFERENCES Ackerman, P. L., & Humphreys, L. G. (1990). Individual differences theory in industrial and organizational psychology. In: M. D. Dunnette & L. Hough (Eds), Handbook of Industrial and Organizational Psychology ??. Palo Alto, CA: Consulting Psychologists Press. Alderfer, C. P., & Sims, A. D. (2003). Diversity in organizations. In: R. J. Klimoski, W. C. Borman & D. R. Ilgen (Eds), Handbook of Psychology (Vol. 12, pp. 595–614). Hoboken, NJ: Wiley. Anderson, N. (2003). The future of selection and assessment: Toward a globalized science and practice. Keynote Address to the 5th Australian Industrial and Organizational Psychology Conference. Melbourne Australia. Anderson, N. (in press). Relationships between practice and research in personnel selection: Does the left hand know what the right is doing? In: A. Evers, O. Smit-Voskuyl & N. Anderson (Eds), The Handbook of Selection ??. Oxford: Blackwell. Anderson, N. R., Born, M., & Cunningham-Snell, N. (2001). Recruitment and selection: Applicant perspectives and outcomes. In: N. Anderson, D. S. Ones, K. Sinangil & C. Viswesvaran (Eds), Handbook of Industrial, Work, and Organizational Psychology (Vol. 1, pp. 200–218). UK: Sage. Argyris, C. (1964). Integrating the individual and the organization. New York: Wiley. Barrick, M. R., & Mount, M. K. (1991). The big five personality dimensions and job performance: A meta analysis. Personnel Psychology, 44, 1–26. Barrick, M. R., Mount, M. K., & Judge, T. A. (2001). Personality and performance at the beginning of the new millennium: What do we know and where do we go next? International Journal of Selection & Assessment, 9, 9–30. Barrick, M. R., Mount, M. K., & Strauss, J. P. (1993). Conscientiousness and performance of sales representatives: Test of the mediating effects of goal setting. Journal of Applied Psychology, 78, 715–722.

Multilevel Staffing

169

Barrick, M. R., Stewart, G. L., Neubert, M. J., & Mount, M. K. (1998). Relating member ability and personality to work-team processes and team effectiveness. Journal of Applied Psychology, 83, 377–391. Barrick, M. R., Stewart, G. L., & Piotrowski, M. (2002). Personality and job performance: Test of the mediating effects of motivation among sales representatives. Journal of Applied Psychology, 87, 43–51. Bass, B. M., Avolio, B. J., Jung, D. I., & Berson, J. (2003). Predicting unit performance by assessing transformational and transactional leadership. Journal of Applied Psychology, 88, 207–218. Becker, G. S. (1964). Human capital. New York: Columbia University Press. Binning, J. F., & Barrett, G. V. (1989). Validity of personnel decisions: A conceptual analysis of the inferential and evidential bases. Journal of Applied Psychology, 74, 478–494. Blau, P. (1977). Inequality and heterogeneity. New York: Free Press. Bliese, P. D. (1998). Group size, ICC values, and group-level correlations: A simulation. Organizational Research Methods, 1, 355–373. Bliese, P. D. (2000). Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis. In: K. Klein & S. W. J. Kozlowski (Eds), Multilevel Theory, Research, and Methods in Organizations: Foundations, Extensions, and New Directions (pp. 349–381). San Francisco, CA: Jossey-Bass. Bliese, P. D. (2002). Multilevel random coefficient modeling in organizational research: Examples using SAS and S-PLUS. In: F. Drasgow & N. Schmitt (Eds), Measuring and Analyzing Behavior in Organizations: Advances in Measurement and Data Analysis (pp. 401–445). Jossey-Bass. Boxall, P., & Purcell, J. (2003). Strategy and human resource management. New York, NY: Palgrave. Borman, W. C., & Motowidlo, S. J. (1997). Task performance and contextual performance: The meaning for personnel selection research. Human Performance, 10, 99–109. Boudreau, J. W., & Ramstad, P. M. (2003). Strategic industrial and organizational psychology and the role of utility analysis. In: R. J. Klimoski, W. C. Borman & D. R. Ilgen (Eds), Handbook of Psychology (Vol. 12, pp. 193–221). Hoboken, NJ: Wiley. Brewster, C., & Larsen, H. H. (2000). Human resource management in northern Europe: Trends, dilemmas, and strategy. Oxford: Blackwell. Bureau of Labor Statistics (Dec 2001). BLS Releases 2000–2010 Employment Projections. Cable, D., & Judge, T. (1997). Interviewers’ perceptions of person-organization fit and organizational selection decisions. Journal of Applied Psychology, 82, 546–561. Cabrera, E. F., & Raju, N. S. (2001). Utility analysis: Current trends and future directions. International Journal of Selection and Assessment, 9, 1–11. Campbell, J. P. (1990). Modeling the performance prediction problem in industrial and organizational psychology. In: M. Dunnette & L. M. Hough (Eds), Handbook of Industrial and Organizational Psychology (2nd ed., Vol. 1, pp. 687–732). Palo Alto, CA: Consulting Psychologists Press. Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1993). A theory of performance. In: N. Schmitt & W. C. Borman (Eds), Personnel Selection in Organizations (pp. 35–70). San Francisco: Jossey-Bass. Campion, M. A., Medsker, G. J., & Higgins, A. C. (1993). Relations between work group characteristics and effectiveness: Implications for designing effective work groups. Personnel Psychology, 46, 823–850. Campion, M. A., Papper, E. M., & Medsker, G. J. (1996). Relations between work team characteristics and effectiveness: A replication and extension. Personnel Psychology, 49, 429–452. Chan, D. (1998). Functional relations among constructs in the same content domain at different levels of analysis: A typology of composition models. Journal of Applied Psychology, 83, 234–246.

170

ROBERT E. PLOYHART

Chen, G., & Bliese, P. D. (2002). The role of different levels of leadership in predicting selfand collective efficacy: Evidence for discontinuity. Journal of Applied Psychology, 87, 549– 556. Chen, G., Bliese, P. D., & Mathieu, J. E. (2003). Conceptual framework and statistical procedures for delineating and testing multilevel theories of homology. Paper presented at the Academy of Management. Chen, G., Gully, S. M., Whiteman, J., & Kilcullen, R. N. (2000). Examination of relationships among trait-like individual differences, state-like individual differences, and learning performance. Journal of Applied Psychology, 85, 835–847. Chen, G., Mathieu, J. E., & Bliese, P. D. (in press). A framework for conducting multilevel construct validation. In: F. J. Dansereau & F. Yamarino (Eds), Research in Multi-Level Issues: The Many Faces of Multi-Level Issues (Vol. 3, ??). Oxford, UK: Elsevier. Chen, G., Webber, S. S., Bliese, P. D., Mathieu, J. E., Payne, S. C., Born, D. H., & Zaccaro, S. (2002). Simultaneous examination of the antecedents and consequences of efficacy beliefs at multiple levels of analysis. Human Performance, 15, 381–410. Davis, M. S. (1999). Aphorisms and clich´es: The generation and dissipation of conceptual charisma. Annual Review of Sociology, 25, 245–269. Delery, J. E., & Doty, D. H. (1996). Modes of theorizing in strategic human resource management: Tests of universalistic, contingency, and configurational performance predictions. Academy of Management Journal, 39, 802–835. Devine, D. J., Clayton, L. D., Philips, J. L., Dunford, B. B., & Melner, S. B. (1999). Teams in organizations: Prevalence, characteristics, and effectiveness. Small Group Research, 30, 678– 711. Devine, D. J., & Philips, J. L. (2001). Do smarter teams do better: A meta-analysis of cognitive ability and team performance. Small Group Research, 32, 507–532. Dobbins, G. H., Cardy, R. L., & Carson, K. P. (1991). Examining fundamental assumptions: A contrast of person and system approaches to human resource management. In: G. R. Ferris & K. M. Rowland (Eds), Research in Personnel and Human Resources Management (Vol. 9, pp. 1–38). Greenwich, CT: JAI Press. Dvir, T., Eden, D., Avolio, B. J., & Shamir, B. (2002). Impact of transformational leadership on follower development and performance: A field experiment. Academy of Management Journal, 45, 735– 744. Festinger, L. (1950). Informal social communication. Psychological Review, 57, 271–282. Fleishman, E. A., & Quaintance, M. K. (1984). Taxonomies of human performance. Orlando, FL: Academic Press. George, J. M. (1992). The role of personality in organizational life: Issues and evidence. Journal of Management, 18, 185–213. Gerhart, B. (1999). Human resource management and firm performance: Challenges in making causal inferences. In: Patrick Wright et al. (Eds), Research in Personnel and Human Resources Management (pp. 31–74). Gerhart, B. A., Wright, P. M., McMahan, G. C., & Snell, S. A. (2000). Measurement error in research on human resources and firm performance: How much error is there and how does it influence effect size estimates? Personnel Psychology, 53, 803–834. Griffin, M. A. (2001). Dispositions and work reactions: A multilevel approach. Journal of Applied Psychology, 86, 1142–1151. Guion, R. M. (1998). Assessment, measurement, and prediction for personnel decisions. Mahwah, NJ: LEA.

Multilevel Staffing

171

Guzzo, R. A., & Shea, G. P. (1992). Group performance and intergroup relations in organizations. In: M. D. Dunnette & L. M. Hough (Eds), Handbook of Industrial and Organizational Psychology (2nd ed., Vol. 3, pp. 269–313). Palo Alto, CA: Consulting Psychologists Press. Hackman, J. R. (1987). The design of work teams. In: J. Lorsch (Ed.), Handbook of Organizational Behavior ??. Englewood Cliffs, NJ: Prentice-Hall. Harris, D. H. (1994). Organizational linkages: Understanding the productivity paradox. Washington, DC: National Academy Press. Highhouse, S., Zickar, M. J., Thorsteinson, T. J., Stierwalt, S. L., & Slaughter, J. E. (1999). Assessing company employment image: An example in the fast food industry. Personnel Psychology, 52, 151–172. Hinsz, V. B., Tindale, R. S., & Vollrath, D. A. (1997). The emerging conceptualization of groups as information processors. Psychological Bulletin, 121, 43–64. Holland, J. L. (1985). Making vocational choices: A theory of vocational personalities and work environments. Englewood Cliffs, NJ: Prentice-Hall. Hough, L. M., Oswald, F. L., & Ployhart, R. E. (2001). Determinants, detection, and amelioration of adverse impact in personnel selection procedures: Issues, evidence, and lessons learned. International Journal of Selection and Assessment, 9, 152–194. House, R., Rousseau, D. M., & Thomas-Hunt, M. (1995). The meso-paradigm: A framework for integration of micro and macro organizational behavior. In: L. L. Cummings & B. Staw (Eds), Research in Organizational Behavior (Vol. 17, pp. 71–117). Greenwhich, CT: JAI Press. Hunter, J. E. (1983). A causal analysis of cognitive ability, job knowledge, job performance, and supervisory ratings. In: F. J. Landy, S. Zedeck & J. Cleveland (Eds), Performance Measurement and Theory (pp. 257–266). Hillsdale, NJ: Erlbaum. Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38, 635–672. Huselid, M. A., & Becker, B. E. (2000). Comment on measurement error in research in human resource decisions and firm performance: How much error is there and how does it influence effect size estimates? Personnel Psychology, 53, 835–854. Huselid, M. A., Jackson, S. E., & Schuler, R. S. (1997). Organizational characteristics as predictors of personnel characteristics. Personnel Psychology, 42, 727–786. Huffcutt, A. I., Weekley, J. A., Wiesner, W. H., Degroot, T. G., & Jones, C. (2001). Comparison of situational and behavior description interview questions for higher-level positions. Personnel Psychology, 54, 619–644. Ilgen, D. R. (1999). Teams embedded in organizations: Some implications. American Psychologist, 54, 129–139. Jackson, S. E., & Schuler, R. S. (1995). Understanding human resource management in the context of organizations and their environments. Annual Review of Psychology, 46, 237–264. James, L. R., & Jones, A. P. (1974). Organizational climate: A review of theory and research. Psychological Bulletin, 81, 109–1112. James, L. R., & Jones, A. P. (1976). Organizational structure: A review of structural dimensions and their conceptual relationships with individual attitudes and behavior. Organizational Behavior and Human Decision Processes, 16, 74–113. Janis, I. (1972). Victims of groupthink. Boston: Houghton Mifflin. Jayne, M. E. A., & Rauschenberger, J. M. (2000). Demonstrating the value of selection in organizations. In: J. F. Kehoe (Ed.), Managing Selection in Changing Organizations (pp. 123–157). San Francisco, CA: Jossey-Bass.

172

ROBERT E. PLOYHART

Jones, R. G., Stevens, M. J., & Fischer, D. L. (2000). Selection in team contexts. In: J. F. Kehoe (Ed.), Managing Selection in Changing Organizations (pp. 210–241). San Francisco, CA: Jossey-Bass. Judge, T. A., Bono, J. E., Ilies, R., & Gerhardt, M. W. (2002). Personality and leadership: A qualitative and quantitative review. Journal of Applied Psychology, 87, 765–780. Kanfer, R. (1990). Motivation theory and industrial and organizational psychology. In: M. D. Dunnette & L. M. Hough (Eds), Handbook of Industrial and Organizational Psychology (2nd ed., Vol. 1, pp. 75–170). Palo Alto, CA: Consulting Psychologists Press. Katz, D., & Kahn, R. L. (1978). The social psychology of organizations. New York: Wiley. Klein, K. J., Dansereau, F., & Hall, R. J. (1994). Levels issues in theory development, data collection, and analysis. Academy of Management Review, 19, 195–229. Klein, K. J., & Kozlowski, S. W. J. (2000). Multilevel theory, research, and methods in organizations: Foundations, extensions, and new directions. San Francisco, CA: Jossey-Bass. Klimoski, R., & Jones, R. G. (1995). Staffing for effective group decision making: Key issues in matching people to teams. In: R. Guzzo, E. Salas & Associates (Eds), Team Effectiveness and Decision Making in Organizations (pp. 291–332). San Francisco: Jossey-Bass. Koch, M. G., & McGrath, R. G. (1996). Improving labor productivity: Human resource management policies do matter. Strategic Management Journal, 17, 335–354. Kozlowski, S. W. J., & Bell, B. S. (2003). Work groups and teams in organizations. In: R. J. Klimoski, W. C. Borman & D. R. Ilgen (Eds), Handbook of Psychology (Vol. 12, pp. 333–375). Hoboken, NJ: Wiley. Kozlowski, S. W. J., Gully, S. M., McHugh, P. P., Salas, E., & Cannon-Bowers, J. A. (1996). A dynamic theory of leadership and team effectiveness: Developmental and task contingent leader roles. Research in Personnel and Human Resource Management, 14, 253–305. Kozlowski, S. W. J., & Klein, K. J. (2000). A multilevel approach to theory and research in organizations: Contextual, temporal, and emergent processes. In: K. J. Klein & S. W. J. Kozlowski (Eds), Multilevel Theory, Research, and Methods in Organizations: Foundations, Extensions, and New Directions (pp. 3–90). San Francisco, CA: Jossey-Bass. Lepak, D. P., & Snell, S. A. (1999). The human resource architecture: Toward a theory of human capital allocation and development. The Academy of Management Review, 24, 34–48. Lepak, D. P., & Snell, S. A. (2002). Examining the human resource architecture: The relationships among human capital, employment, and human resource configurations. Journal of Management, 28, 517–543. LePine, J. A. (2003). Team adaptation and postchange performance: Effects of team composition in terms of members’ cognitive ability and personality. Journal of Applied Psychology, 88, 27– 39. LePine, J. A., Hanson, M. A., Borman, W. C., & Motowidlo, S. J. (2000). Contextual performance and teamwork: Implications for staffing. In: G. Ferris (Ed.), Research in Personnel and Human Resources Management (Vol. 19, pp. 53–90). LePine, J. A., Hollenbeck, J. R., Ilgen, D. R., & Hedlund, J. (1997). Effects of individual differences on the performance of hierarchical decision-making teams: Much more than g. Journal of Applied Psychology, 82, 803–811. Lim, B. C., & Ployhart, R. E. (in press). Transformational leadership: Relations to the Five Factor Model and team performance in typical and maximum contexts. Journal of Applied Psychology. Lovrich, N. P., Jr., Steel, B. S., & Hood, D. (1986). Equity vs. productivity: Affirmative action and municipal police services. Public Productivity Review, 39, 61–72.

Multilevel Staffing

173

MacDuffie, J. P. (1995). Human resource bundles and manufacturing performance: Organizational logic and flexible production systems in the world of auto industry. Industrial and Labor Relations Review, 48, 197–221. Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. (2001). A temporally based framework and taxonomy of team processes. Academy of Management Review, 26, 356–376. McDaniel, M. A., Morgeson, F. P., Finnegan, E. B., Campion, M. A., & Braverman, E. P. (2001). Use of situational judgment tests to predict job performance: A clarification of the literature. Journal of Applied Psychology, 86, 730–740. McFarland, L. A. (in press). Warning against faking on a personality test: Effects on applicant reactions and personality test scores. International Journal of Selection and Assessment. McGrath, J. E. (1964). Social psychology: A brief introduction. New York: Holt, Rinehart, & Winston. McGrath, J. E. (1984). Groups: Interaction and performance. NJ: Prentice-Hall. Miles, R. E., & Snow, C. C. (1978). Organizational strategy, structure, and process. New York: McGraw-Hill. Miller, J. G. (1978). Living systems. NY: McGraw-Hill. Motowidlo, S. J. (2003). Job performance. In: R. J. Klimoski, W. C. Borman & D. R. Ilgen (Eds), Handbook of Psychology (Vol. 12, pp. 39–53). Hoboken, NJ: Wiley. Murphy, K. R. (1989). Is the relationship between cognitive ability and job performance stable over time? Human Performance, 2, 183–200. Neuman, G. A., & Wright, J. (1999). Team effectiveness: Beyond skills and cognitive ability. Journal of Applied Psychology, 84, 376–389. Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard University Press. Olian, J. D., & Rynes, S. L. (1984). Organizational staffing: Integrating practice with strategy. Industrial Relations, 23, 170–181. Ostroff, C. (2002). Leveling the selection field. In: F. J. Dansereau & F. Yamarino (Eds), Research in Multi-Level Issues Volume 1: The Many Faces of Multi-Level Issues (Vol. 1, pp. 141–154). Oxford, UK: Elsevier. Ostroff, C., & Bowen, D. (2000). Moving HR to a higher level: HR practices and organizational effectiveness. In: K. J. Klein & S. W. J. Kozlowski (Eds), Multilevel Theory, Research, and Methods in Organizations: Foundations, Extensions, and New Directions (pp. 211–266). San Francisco, CA: Jossey-Bass. Peterson, R. S., Smith, D. B., Martorana, P. V., & Owens, P. D. (2003). The impact of chief executive officer personality on top management team dynamics: One mechanism by which leadership affects organizational performance. Journal of Applied Psychology, 88, 795–808. Ployhart, R. E., & Schneider, B. (2002a). A multilevel perspective on personnel selection research and practice: Implications for selection system design, assessment, and construct validation. In: F. J. Dansereau & F. Yamarino (Eds), Research in Multi-Level Issues Volume 1: The Many Faces of Multi-Level Issues (Vol. 1, pp. 95–140). Oxford, UK: Elsevier. Ployhart, R. E., & Schneider, B. (2002b). A multilevel perspective on personnel selection: When will practice catch up? In: F. J. Dansereau & F. Yamarino (Eds), Research in Multi-Level Issues Volume 1: The Many Faces of Multi-Level Issues (Vol. 1, pp. 165–175). Oxford, UK: Elsevier. Ployhart, R. E., & Schneider, B. (in press). Multilevel selection and prediction: Theories, methods, and models. In: A. Evers, O. Smit-Voskuyl & N. Anderson (Eds), Handbook of Personnel Selection. Ployhart, R. E., Weekley, J. A., & Baughman, K. (2003). The consequences of personality for individual job satisfaction and organizational citizenship behaviors across individual, job, and organizational levels. Symposium presented at the annual conference of the Society for Industrial and Organizational Psychology, Orlando, FL.

174

ROBERT E. PLOYHART

Pulakos, E. D., Arad, S., Donovan, M. A., & Plamondon, K. E. (2000). Adaptability in the workplace: Development of a taxonomy of adaptive performance. Journal of Applied Psychology, 85, 612–624. Richard, O. C. (2000). Racial diversity, business strategy, and firm performance: A resource-based view. Academy of Management Journal, 43, 164–177. Rousseau, D. M. (1985). Issues of level in organizational research: Multi-level and cross-level perspectives. In: L. L. Cummings & B. Staw (Eds), Research in Organizational Behavior (Vol. 7). Greenwhich, CT: JAI Press. Ryan, A. M. (2001). Explaining the black/white test score gap: The role of test perceptions. Human Performance, 14, 45–75. Ryan, A. M., McFarland, L. A., Baron, H., & Page, R. (1999). An international look at selection practices: Nation and culture as explanations for variability in practice. Personnel Psychology, 52, 359–391. Ryan, A. M., & Ployhart, R. E. (2000). Applicants’ perceptions of selection procedures and decisions: A critical review and agenda for the future. Journal of Management, 26, 565– 606. Ryan, A. M., Sacco, J., McFarland, L. A., & Kriska, S. D. (2000). Applicant self-selection: Correlates of withdrawal from a multiple hurdle process. Journal of Applied Psychology, 85, 163– 179. Sacco, J. M. (2003). Validating personality and ability tests in customer service jobs at the business unit level. Unpublished manuscript. Sacco, J. M., & Schmitt, N. (2003). The relationship between demographic diversity and profitability: A longitudinal study. Paper presented at the 18th Annual Conference of the Society for Industrial and Organizational Psychology, Orlando, FL. Sackett, P. R. (2002). The structure of counterproductive work behavior: Dimensionality and relationships with facets of job performance. International Journal of Selection and Assessment, 10, 5–11. Sackett, P. R., & Roth, L. (1996). Multi-stage selection strategies: A Monte Carlo investigation of effects on performance and minority hiring. Personnel Psychology, 49, 1–18. Sackett, P. R., Schmitt, N., Ellingson, J. E., & Kabin, M. B. (2001). High stakes testing in employment, credentialing, and higher education. American Psychologist, 56, 302–318. Sackett, P. R., & Wilk, S. L. (1994). Within-group norming and other forms of score adjustment in preemployment testing. American Psychologist, 49, 929–954. Schaubroeck, J., Ganster, D. C., & Jones, J. R. (1998). Organizational and occupation influences in the attraction-selection-attrition process. Journal of Applied Psychology, 83, 869–891. Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124, 262–274. Schmitt, N. (2002). A multi-level perspective on personnel selection: Are we ready? In: F. J. Dansereau & F. Yamarino (Eds), Research in Multi-Level Issues Volume 1: The Many Faces of Multi-Level Issues (Vol. 1, pp. 155–164). Oxford, UK: Elsevier. Schneider, B. (1987). The people make the place. Personnel Psychology, 40, 437–454. Schneider, B., Goldstein, H. W., & Smith, D. B. (1995). The ASA framework: An update. Personnel Psychology, 48, 747–773. Schneider, B., Hanges, P. J., Smith, D., & Salvaggio, A. N. (2003). Which comes first: Employee attitudes or organizational financial and market performance? Journal of Applied Psychology, 88, 836–851.

Multilevel Staffing

175

Schneider, B., Kristof, A. L., Goldstein, H. W., & Smith, D. B. (1997). What is this thing called fit? In: N. Anderson & P. Herriot (Eds), Handbook of Selection and Appraisal (2nd ed., pp. 393–412). London: Wiley. Schneider, B., Smith, D. B., & Sipe, W. P. (2000). Personnel selection psychology: Multilevel considerations. In: K. J. Klein & S. W. J. Kozlowski (Eds), Multilevel Theory, Research, and Methods in Organizations: Foundations, Extensions, and New Directions (pp. 91–120). San Francisco, CA: Jossey-Bass. Schneider, B., Smith, D. B., Taylor, S., & Fleenor, J. (1998). Personality and organizations: A test of the homogeneity of personality hypothesis. Journal of Applied Psychology, 83, 462–470. Schneider, B., White, S. S., & Paul, M. C. (1998). Linking service climate to customer perceptions of service quality: Test of a causal model. Journal of Applied Psychology, 83, 150–163. Schuler, R. S., Dowling, P. J., & DeCieri, H. (1993). An integrative framework of strategic international human resource management. Journal of Management, 19, 419–459. Schuler, R. S., & Jackson, S. E. (1987). Linking competitive strategies with human resource management. Academy of Management Executive, 1, 207–219. Shaw, J. D., Delery, J. E., Jenkins, G. D., & Gupta, N. (1998). An organization-level analysis of voluntary and involuntary turnover. Academy of Management Journal, 41, 511– 525. Simon, H. A. (1973). The organization of complex systems. In: H. H. Pattee (Ed.), Hierarchy Theory (pp. 1–27). NY: Braziller. Smith, P. C. (1976). Behaviors, results, and organizational effectiveness: The problem of criteria. In: M. D. Dunnette (Ed.), Handbook of Industrial and Organizational Psychology. Chicago: Rand McNally. Snell, S. A., Youndt, M. A., & Wright, P. M. (1996). Establishing a framework for research in strategic human resource management: Merging resource theory and organizational learning. In: G. R. Ferris (Ed.), Research in Personnel and Human Resources Management (pp. 61–90). Greenwhich, CT: JAI Press. Snow, C. C., & Snell, S. A. (1993). Staffing as strategy. In: N. Schmitt & W. C. Borman (Eds), Personnel Selection in Organizations (pp. 448–478). San Francisco: Jossey-Bass. Steel, B. S., & Lovrich, N. P., Jr. (1987). Equality and efficiency trade-offs in affirmative action – Real or imagined? The case of women in policing. The Social Science Journal, 24, 53–70. Steiner, I. D. (1972). Group process and productivity. New York: Academic Press. Stevens, M. J., & Campion, M. A. (1994). The knowledge, skill, and ability requirements for teamwork: Implications for human resource management. Journal of Management, 20, 503–530. Stevens, M. J., & Campion, M. A. (1999). Staffing work teams: Development and validation of a selection test for teamwork settings. Journal of Management, 25, 207–228. Stewart, G. L., Manz, C. C., & Sims, H. P., Jr. (1999). Team work & group dynamics. New York: Wiley. Tajfel, H. (1982). Social identity and intergroup relations. Cambridge: Cambridge University Press. Terpstra, D. E., & Rozell, E. J. (1993). The relationship of staffing practices to organizational level measures of performance. Personnel Psychology, 46, 27–48. Von Bertalanffy, L. (1968). General systems theory. NY: Braziller. Von Bertalanffy, L. (1972). The history and status of general systems theory. In: G. J. Klir (Ed.), Trends in General Systems Theory (pp. 21–41). NY: Wiley. Weick, K. E. (1976). Educational organizations as loosely coupled systems. Administrative Science Quarterly, 21, 1–19. Whitney, D. J., & Schmitt, N. (1997). Relationship between culture and responses to biodata employment items. Journal of Applied Psychology, 82, 113–129.

176

ROBERT E. PLOYHART

Wilk, S. L., & Capelli, P. (2003). Understanding the determinants of employer use of selection methods. Personnel Psychology, 56, 103–1242. Woodward, J. (1958). Management and technology, problems of progress in industry. London: Her Majesty’s Stationary Office. Wright, P. M., & Boswell, W. R. (2002). Desegregating HRM: A review and synthesis of micro and macro HR research. Journal of Management, 28, 247–276. Wright, P. M., & McMahan, G. C. (1992). Theoretical perspectives for strategic human resource management. Journal of Management, 18, 295–320. Wright, P. M., Gardner, T. M., Moynihan, L. M., Park, H. J., Gerhart, B., & Delery, J. R. (2001). Measurement error in research on human resources and firm performance: Additional data and suggestions for future research. Personnel Psychology, 54, 875–901. Zajonc, R. B. (1965). Social facilitation. Science, 149, 269–274.

THE FOUR-FIFTHS RULE FOR ASSESSING ADVERSE IMPACT: AN ARITHMETIC, INTUITIVE, AND LOGICAL ANALYSIS OF THE RULE AND IMPLICATIONS FOR FUTURE RESEARCH AND PRACTICE Philip Bobko and Philip L. Roth ABSTRACT Evidence for adverse impact in applied, organizational settings can often depend on application of the “four-fifths rule.” We analyze the arithmetic four-fifths rule, its operationalization, and related statistical tests. We note that the rule has intuitive appeal and has arithmetic directness. On the other hand, the four-fifths rule contains many ambiguities because of the manner in which it is defined, as well as its use in practice. One purpose of this article is to discuss the arithmetic and statistical facets of the definition. A related purpose of this article is to demonstrate where the ambiguities (and possibly unintended consequences) with the four-fifths rule might arise when numerical interpretations are invoked. Implications for future research and academic dialogues are then noted.

Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 177–198 Copyright © 2004 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23004-3

177

178

PHILIP BOBKO AND PHILIP L. ROTH

INTRODUCTION A selection rate for any race, sex, or ethnic group which is less than four-fifths (4/5) (or 80%) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact, while a greater than four-fifths rate will generally not be regarded by Federal enforcement agencies as evidence of adverse impact. Smaller differences in selection rate may nevertheless constitute adverse impact, where they are significant in both statistical and practical terms. . . . Greater differences in selection rate may not constitute adverse impact where the differences are based on small numbers and are not statistically significant. . . . (Uniform Guidelines, 1978, p. 38297).

The above quotation is from the Uniform Guidelines On Employee Selection Procedures (herein after simply referred to as the Guidelines) – a document that was a joint effort of several government agencies (Civil Service Commission, Department of Labor, Department of Justice, Equal Employment Opportunity Commission). The purpose of these Guidelines was to assist employers and other groups in complying with Federal law prohibiting discriminatory employment practices. The Guidelines were “designed to provide a framework for determining the proper use of tests and other selection procedures” (see Guidelines, p. 38296). Among other things, the Guidelines attempt to define a variety of concepts and operational procedures (e.g. adverse impact, test fairness) as well as the manners in which tests might be “validated.” The opening quotation to this article appears in the section of the Guidelines that defines “adverse impact.” The operationalization of adverse impact (one selection rate to be within 80% of the other selection rate) has come to be known as “the fourfifths rule” or, equivalently, “the 80% rule.” As noted in the Background section of the Guidelines, this rule was adopted as a rule of thumb. It was not originally intended as a legal definition of discrimination, but as a “practical device to keep the attention of enforcement agencies on serious discrepancies in hire or promotion rates or other employment decisions” (see Federal Register, p. 38291). If disparity in selection rates is evidenced, then a prima facie case of disparate impact can be alleged. The purpose of this article is to review the logic of the four-fifths rule, its application in research and practice, and its relation to the potential adverse impact of selection systems. We are not providing a legal summary of this rule nor its status in case law, as that would be an analysis for a different outlet (e.g. a law journal). Rather, we first analyze the arithmetic and statistical characteristics of this rule and its operationalization. We note that both approaches (arithmetic and statistical) share common issues (e.g. error rate concerns). Then, we discuss a variety of additional ambiguities that arise from logical considerations of the rule. Finally, we derive a listing of future research needs from these issues.

The Four-Fifths Rule For Assessing Adverse Impact

179

Description vs. Inference Note that the first sentence in the definition of the four-fifths rule is based upon an algebraic, descriptive definition. That is, one selection rate is arithmetically compared to four-fifths of another selection rate. In contrast, the second and third sentences indicate that selection rates which are within 80% of each other may nonetheless constitute adverse impact – as a function of statistical and practical significance (see also Morris & Lobsenz, 2000). We point out that this change in focus (from descriptive to inferential) mirrors how most basic statistical texts are organized. That is, students are first taught about graphs, means, standard deviations, parameter estimation, etc. – all in a descriptive mode. Topics such as distribution theory, statistical hypothesis testing, and the role of inference then appear in the second half of statistics texts. We think it is interesting that this shift parallels the words in the above quote; i.e. the definition of adverse impact changes from a descriptive one to an inferential one. This shift has practical implications that we discuss below. Some History Meier, Sacks and Zabell (1984) provide an analysis of the four-fifths rule, including a discussion of how it was operationalized in an early Supreme Court case (Hazelwood School District v. United States, 1977). In a footnote, they shed some light on the history of the four-fifths rule. Specifically, those authors suggest that the origins of the Federal rule came from a similar rule in the State of California stating that adverse impact was indicated if both the disparity in selection rates was greater than 80% and the difference in rates was statistically significant. Meier et al. (1984, Footnote 93) also state that “compliance officers expressed difficulties in properly performing statistical significance tests in the field” and so the arithmetic 80% rule was, by itself, adopted. This appears to be related to the fact that the first sentence of the opening quote is about an arithmetic rule, yet subsequent sentences are about statistical significance. Room for Interpretation Note that the three sentences in the opening quote contain two “may”s and two “generally”s. These words provide substantial ambiguity in the assessment of adverse impact. Combined with the shift from descriptive to inferential statistics, the rule leaves room for interpretation in a variety of ways – thus providing additional motivation for this article.

180

PHILIP BOBKO AND PHILIP L. ROTH

ARITHMETIC AND STATISTICAL CONSIDERATIONS One of the key issues already noted is that one can approach an analysis of potentially disparate selection rates both arithmetically and statistically. This leads, in turn, to ambiguity in how two selection rates (e.g. p1 and p2 as representations of the minority and majority selection rates) are compared. Specifically, the arithmetic four-fifths rule proceeds by computing a ratio of selection rates (e.g. p 1 /p 2 or p minority /p majority ). If that ratio is less than 0.80, then adverse impact is indicated. [We also point out that the “standard” selection rate is, in practice, usually taken to be that for Whites or males, yet the opening quote clearly states that the standard is the highest rate of any race, sex, or ethnic group – which may not involve Whites or males.] In contrast, the usual statistical significance test proceeds by testing the null hypothesis that H0 : p 1 = p 2 or, equivalently H0 : p 1 − p 2 = 0. That is, the statistical test focuses on the difference in selection rates, not the ratio. As we shall see, this change in perspectives can lead to increased ambiguity (relative to the “may”s and “generally”s already noted) in the determination of adverse impact.

Typical Procedures (Arithmetic and Statistical) An Arithmetic Example The arithmetic four-fifths rule attempts to assist in deciding when a disparity in selection rates is large enough to renew a focus on the validity of an organization’s selection system. It is the current authors’ belief that the rule was adopted because, although it is arithmetic, it does provide some latitude for variation (e.g. sampling variance) in subgroup selection rates. The rule does not invoke statistical theory/assumptions and avoids the fact that statistical significance depends on sample sizes. However, the arithmetic nature of the four-fifths rule is based on percentages, and percentages are a function of the basis used (i.e. denominator in the ratio). For example, consider the data in Table 1 (modified from a court case in the 1970s). Table 1. Illustrative Adverse Impact Data. Majority

Minority

Passed the test Failed the test

92 4693

24 2639

Total

4736

2691

The Four-Fifths Rule For Assessing Adverse Impact

181

Assume that anyone who passed the test was hired, while those who failed would not be hired. In this data, the hiring rate for the majority subgroup is then 92/4736 = 0.019 and the hiring rates for minorities is 24/2691 = 0.009. (Looking ahead, it is interesting to note that there is a statistically significant relationship between ethnicity and selection rate.) In this situation, the 80% rule is violated (i.e. the ratio of the two selection rates is 0.009/0.019 = 0.47). Indeed, the plaintiffs argued that one selection rate was nearly twice that of the other. The defendant, on the other hand, compared the difference between the rates, rather than their ratio, and noted that the two rates were within 1% of each other (0.019–0.009). Thus, although the ratio of the minority rate to the majority rate is small (and the four-fifths rule therefore violated), it could be argued that there is little practical difference between two selection rates that are one percentage point from each other. Note that the four-fifths rule is open to interpretation even in an arithmetic way – consistent with the equivocation in the Guidelines. As Meir et al. suggested when considering ratios (as in the four-fifths rule) and differences (as in statistical hypothesis testing), “no single measure can capture completely the information” (1984, p. 164). The Statistical Procedure As noted, the opening quote suggests the possibility of conducting statistical analyses regardless of whether the selection rates are within 80% of each other. On the other hand, the Guidelines indicate that a statistical significance test is not a replacement for the four-fifths arithmetic rule (see p. 38294). Part of the justification for this position is that “the great majority of employers” do not hire enough individuals for sole reliance on significance testing. In other words, there is a recognition of the role of sample size. Nonetheless, it is our experience that tests of significance are believed to have substantial probative value, and they routinely appear in many court cases involving determination of adverse impact (see also Meier et al., 1984 for a review of cases where the courts have adopted the notion that gross statistical disparities alone could constitute prima facie proof of disparate impact). If sample sizes are sufficient, the statistical hypothesis test is based on the normal approximation to the binomial distribution (if sample sizes are small, a Fisher exact test is an option). Thus, the statistic typically reported is a “z-value” comparing the proportions from two independent samples (see any introductory statistics text). The hypothesis test considers if the two samples come from populations with equal probabilities of selection (H0 : p 1 = p 2 ) and the test statistic is: z=√

p1 − p2 p(1 − p)[(1/n1 ) + (1/n2 )]

(1)

182

PHILIP BOBKO AND PHILIP L. ROTH

Table 2. More Illustrative Adverse Impact Data.

White Black

Pass Test

Fail Test

Total

477 54

149 42

626 96

531

191

722

where p1 and p2 are the sample proportions, p is the pooled selection rate across both subgroups, and n1 and n2 are the subgroup sample sizes. In Eq. (1), the numerator is the difference in selection rates and the denominator is the sampling standard deviation associated with a difference in proportions. If the usual 0.05 and 0.01 levels of significance are used, and a two-tailed test is invoked, then the “critical values” of z are 1.96 and 2.58. In a Supreme Court decision that is often cited in this domain (Hazelwood School District v. United States, 1977) the court approximated these critical values by the phrase “two or three standard deviations.” In our experience, lawyers and experts have come to call the use of Eq. (1) a “standard deviations analysis.” That is, the value of z is reported, and disparities are often claimed to be significant if z (the number of standard deviations of difference between p1 and p2 ) is greater than 2 or 3. A statistical example. Suppose a populous state in the U.S. is interested in assessing the adverse impact of a particular selection test for public service employment, and the state’s concern is with potential differences between White and Black selection rates. The example data are presented in Table 2 (taken from another court case; identity and scope preserved). In this instance, the selection rate for Whites is p w = 477/626 = 0.762 and that for Blacks is p b = 54/96 = 0.563. Arithmetic application of the four-fifths rule indicates that the Black rate is not within 80% of the White rate (i.e. the ratio of the two selection rates is 0.563/0.762 = 0.739). The statistics test typically conducted considers if the two samples come from populations with equal proportions (H0 : p w = p b ). Letting nw and nb be the subgroup sample sizes (626 and 96, respectively, in this example), application of Eq. (1) yields a value of approximately z = 4.13, which is “statistically significant” (p < 0.05). In this instance, one says that H0 is rejected (p < 0.05) and might conclude that there is evidence of statistically significant covariation between race and passing/failing the test. Notice also the careful wording (“significant covariation”) here. The usual admonition of not implying causation from correlation applies. That is, the statistically significant result might be used as a way to determine adverse impact, but as in

The Four-Fifths Rule For Assessing Adverse Impact

183

the case of the four-fifths rule, the evidence is just prima facie. There could be a variety of reasons for the significant relationship (or the fact that one rate is not arithmetically within 80% of the other rate). For example, the selection system might be discriminatory, or the selection system might be associated with group differences that are related to group differences in job performance (i.e. the test is valid and “fair”), or the group differences in selection rate might be related to group differences in test scores which are related to group applicant differences in prior job (or educational) experience, or there may have been differential recruiting efforts such that the two applicant pools are not of equal quality, and so forth.

The Role of Sample Size: Interpretation and Power One Improper Interpretation Consistent with the fact that the hypothesis testing procedure has become to be known as a “standard deviations analysis,” some experts in this area have begun to present the result of the hypothesis test in terms of the “number of standard deviations” of difference between the two hiring rates. That is, the value of z is presented (usually to the nearest integer) and then adjectival interpretations to the value of z are presented (“significant,” “highly significant,” “overwhelmingly significant”) depending upon how large z is relative to the p = 0.05 critical value of 1.96. Of course, use of such adjectives is counter to standard statistical practice in which significance proceeds by an a priori choice of alpha level (Type I error rate) and the decision of “significant” (or not) is a dichotomous one (Austin, Boyle & Lualhati, 1998). Clearly, the magnitude of z can be used to calculate and report the a posteriori probability of obtaining a difference in selection rates as extreme, or more extreme, as that obtained (i.e. if one obtained a value of z = 4.13, at what level of significance would this result have been significant?). But, the value of z (or the associated a posteriori probability) depends upon sample size. Therefore, the “number of standard deviations” does not provide unambiguous information regarding the importance of the magnitude of the difference in selection rates (see APA Publication Manual, 2001, p. 25, for a reaffirmation of this fact). Thus, while a z of 4.13 would have been significant at the 0.0001 level, it is improper to add an adjective such as “highly significant” or “overwhelmingly significant” to this result (Harcum, 1989). The dependence of the value of z on sample size can readily be seen by noting that the values of “n” are in the denominator of the denominator of Eq. (1) above. Thus, the larger the ns, the larger the value of z (all else equal). To put this in context, we initially assumed that the state in which the original study was conducted was a

184

PHILIP BOBKO AND PHILIP L. ROTH

populous one. Suppose a similar study was conducted in a state with one-fifth the population (i.e. assume state services have one-fifth of the number of individuals). Suppose that the two passing rates for Blacks and Whites remain exactly the same as before, and that all other proportions remain the same. If all the subgroup sample sizes are reduced by a factor of 5 then, all else equal, the value of z will be reduced by a factor of the square root of 5 and the resulting z will be approximately 1.85! This value is no longer statistically significant, even though the two selection rates remained unchanged. Thus, the use of z as an index of the strength of the difference in hiring rates is incorrect because its magnitude is heavily dependent on sample size. Indeed, we suggest – in contrast to the current use of “number of standard deviations” in reporting adverse impact analyses – that effect size estimates which are not sample size dependent be adopted (for reviews of effect sizes, see for example, APA Manual, 2001; Breaugh, 2003). Use of effect size indices which are independent of sample sizes would also be consistent with current practice in meta-analyses. Sample Size and Power As quoted earlier, the Guidelines implied that many employers might not have sufficient sample size to conduct significance tests. Just what constitutes “enough” sample size is a statistical issue beyond the scope of our process-oriented review of the four-fifths rule – although it is an important question given that minority sample sizes can be small in practice (e.g. see Wing, 1982). We do note that the power of the statistical hypothesis test in Eq. (1) depends heavily on the smaller value of n1 or n2 (see Bobko, 2001, pp. 55–56). For example, assuming an overall sample size of 200 and all else equal, Bobko shows that the value of z in Eq. (1) will be 20% lower if n 1 = 40 and n 2 = 160, as compared to n 1 = n 2 = 100. This issue is particularly salient in adverse impact cases, as minority sample sizes can be quite small (the lack of power being the opposite situation of what was discussed in the prior section of this article). In fact, it can readily be shown that for a given overall sample size, the maximal power obtained for testing H0 : p 1 = p 2 is when n 1 = n 2 (which is one reason why cell sizes are equated in experimental design). More on Power (Combining Ethnic Groups) Some analysts take two ethnic minority groups (e.g. Blacks and Hispanics) and combine them (a general paradox associated with combining groups will also be discussed below). The apparent reason for combining subgroups is to increase the sample size of the minority “group” in order to increase statistical power (to detect differences in selection rates). The fact that the number of minorities is occasionally small lends some credibility to this process – in that the hiring rate

The Four-Fifths Rule For Assessing Adverse Impact

185

for the combined minority group will have a reduced standard error given the larger combined subgroup n. On the other hand, if the two (or more) minority groups being combined have different hiring rates themselves, then their combination might create confounding group-level variation within the combined group. There appear to be many drawbacks to combining subgroups. First, we are unaware of anything in the Guidelines that recommends combining ethnic groups. Section 4B (p. 38297) of those guidelines simply indicates that records are to be kept by sex and ethnic group (Black, American-Indian, Asian, Hispanic, White, and total). Second, in our experience, the combinations tend to have a post-hoc quality in that only groups are combined which are disfavorable when compared to the White hiring rate. For example, in one instance we are aware of, Hispanics and Blacks were combined as a group – apparently after looking at the data. In those groups, selection rates were lower than those of Whites. On the other hand, there was no apparent adverse impact for Asians and American-Indians, and these subgroups were not used as part of the combined group. Such a post-hoc quality seems to fly in the face of the logic of hypothesis testing and scientific endeavor, where hypotheses are set in advance based upon theory. These hypotheses are tested and then updated as logic and data permit. Third, combining groups may increase the statistical power (and we have also seen such a combination – Black and non-Black – required by regulatory agencies), but it is then unclear as to precisely what ethnic group is being considered. As summarized in Guinyard v. City of New York (1992, p. 1089), combining groups “may produce a statistically more useful figure, but it does not necessarily produce a meaningful one.” For example, if one is comparing subgroup differences in operational testing to expectations from the literature (e.g. “We used work sample exams – what would we have expected adverse impact to be based on the literature?”), then combining groups may cloud interpretation when comparisons are made to literature based, meta-analytic estimates. In general, we suggest that it is already a difficult necessity of life that human beings are placed in single demographic categories which, by definition, ignore variation in the myriad skills, abilities, and backgrounds that they have. To further mix these individuals into a larger, more amorphous or heterogeneous, group does not help our science progress in the understanding of human behavior, let alone the efficiency of selection systems. The Chi-Square Test and More Sample Size Issues The data in Tables 1 and 2 are in the form of a “two-by-two” contingency table. Thus, one could conduct a chi-square test of independence on these data. The null hypothesis is that there is no association between passing/failing the test and ethnic group membership. It is first important to note that the chi-square

186

PHILIP BOBKO AND PHILIP L. ROTH

test is arithmetically the same as the test for equal proportions using the normal approximation to the binomial. The value of the sample chi-square (which will have one degree of freedom) is exactly the square of the value of “z” in the binomial test. This can be algebraically confirmed and is consistent with statistical theory that a chi-square distribution is composed of squares of standard normal variables (see, e.g. Mood & Graybill, 1963). Using the data in Table 2 as an example, the value of chi-square is 17.02 – which is the square of 4.13, within rounding error. An important point to remember is the arithmetic equivalence of the two analyses, in that it would not make sense to conduct both tests and then claim that there was convergence of findings (as we have seen done in past practice). Also, thinking about the significance test in terms of chi-squares readily leads to additional statements about sample sizes. The evaluation of the adequacy of a chisquare test can be taken from perspectives of both Type I error and Type II error (see also the next section of this article for discussion of error rates). Regarding power, the issues depend upon the magnitude of the effect and the sample size in complex ways, and the reader is referred to texts on the subject (e.g. Cohen, 1988). Regarding Type I error, the literature on chi-square tests suggests that these tests reasonably maintain their nominal Type I error rates even when sample sizes are very small. For example, DeLucchi’s (1983) review of the literature indicates that even “occurrences of expected values below one generally do not invalidate” chi-square procedures (p. 167). And, Camilli and Hopkins (1979) concluded that chi-square tests of independence “were found to provide reasonably accurate estimates of Type I error probability for N greater than or equal to 8” (p. 1011). Error Rates Type I and Type II Error Rates of the Statistical Test of H0 : p1 = p2 In fact, we have been implicitly discussing error rates throughout this article. Type I error rates, as in most statistical hypothesis testing procedures, are usually controlled at 0.05 or 0.01 levels. This has led to the “2 or 3 standard deviations analysis” rule of thumb. Type II error rates are, of course, a function of the actual difference in selection rates (if any) as well as the two subgroup sample sizes. As we have discussed above, all else equal, the power to detect differences in selection rates (and avoid Type II errors) increases with increasing sample size – particularly as the smaller of the two subgroup n’s increases (Bobko, 2001; Hsu, 1993). Type I and Type II Error Rates of the Arithmetic Four-Fifths Rule Interestingly, some individuals have investigated to what extent the arithmetic four-fifths rule is subject to Type I and Type II errors. For example, suppose that in the population, the majority and minority selection rates are identical. Suppose

The Four-Fifths Rule For Assessing Adverse Impact

187

further that subsamples of n1 majority members and n2 minority members are considered. Because of naturally occurring sampling variation, the adverse impact ratio (p 2 /p 1 ) will not always exactly equal 1.0. The issue is – how often will the ratio drift below 0.80 as a result of sampling variation? The degree to which this occurs is Type I error of the arithmetic rule.1 Greenberg (1979) and Boardman (1979) estimated the rate of Type I errors using the four-fifths rule. Each author used some simplifying, and/or possibly unrealistic, assumptions. For example, Greenberg assumed that the majority and minority subgroup sizes were identical. And, Boardman assumed that individuals were selected “with replacement” (e.g. selecting a minority person from the applicant pool did not change the probability that any remaining minority would be selected – even though in practice the number of minorities then available would clearly go down). Nonetheless, their results are illustrative and important. For example, Greenberg found that if majority and minority population selection rates were set to the same value of 0.50, and if samples of n 1 = n 2 = 50 were taken, then the four-fifths rule would be violated 14% of the time (i.e. the sample minority selection rate would be below 0.80 of the sample majority selection rate about 14% of the time). Thus, the four-fifths rule would be violated in 14% of the situations even though there was no adverse impact in the population. Thus, the Type I error rate is 0.14 in this scenario. [A perusal of Greenberg’s results also show that, under conditions where the two selection rates are identical, the four-fifths rule is more likely to be violated as the common, equal selection rate is relatively low. We discuss this important implication in more detail in the next part of this article.] Greenberg also showed that Type II errors will occur with some frequency. For example, suppose that in the population the majority selection rate is p 1 = 0.50 and the minority selection rate is p 2 = 0.30. The adverse impact rate (0.30/0.50 = 0.60) is below the 0.80 threshold. However, even with samples of 100 majority and minority applicants (n 1 = n 2 = 100), about 5% of the situations would result in the minority rate being within 0.80 of the majority rate. The point is that the arithmetic four-fifths rule is not immune to the same types of issues that occur when conducting traditional hypothesis tests. (See also Lawshe (1987) for some data on the variability in adverse impact ratios across successive years or Meier et al.’s (1984) statement that changes in recruiting across time will cause instability in the ratio.) We call below for research and realistic simulations on this issue. More on Error Rates: Experimentwise Error As noted, each statistics test is typically conducted at the p = 0.05 level of significance using the “2 or 3 standard deviation” rule (see Guidelines, p. 38301

188

PHILIP BOBKO AND PHILIP L. ROTH

for the adoption of the value 0.05). However, it is important to realize that several comparisons are often computed in such analyses: White vs. Black, White vs. Hispanic, males vs. females, etc. Of course, this process ignores what the experimental design literature labels “experimentwise error.” That is, even if there are no differences in selection rates, as the number of tests and/or comparisons increases, the likelihood of obtaining at least one significant comparison increases substantially (see Greenberg, 1979, who also pointed this out). Thus, although it is not practiced in the analyses we have seen, one might correct statistical tests for experimentwise error rates (see Kirk, 1995, for remedies such as the Bonferroni correction and others). The possibility of correcting tests of adverse impact in this manner presents an interesting set of issues. From the vantage of the employer, the corrections appear reasonable, as the employer should not be subject to artificially inflated rates of the rejection of the null hypotheses. However, from the point of view of an applicant, she/he (or his/her class) wants to know if there is adverse impact against a single, particular group – and assessment should not depend upon how many other claims, or potential claims, there might be. We encourage future discussions of these issues in the academic and applied arena. Armed with a better understanding of the Guidelines and the intricacies of arithmetic vs. statistical assessment of disparate impact, we now turn to yet more issues about the four-fifths rule. Some of these issues provide operational clarification to organizations using the rule, while other issues lead to more ambiguity. We also discuss methods for conjoining the arithmetic and statistical procedures.

FURTHER AMBIGUITIES, CLARIFICATIONS, AND ADVANCES Implications for Selectivity The Selection Ratio Influences Outcomes Based Upon the Four-Fifths Rule Assume that the highest selection rate in an organization is for the majority group and that the selection rate is 0.70. The four-fifths rule implies that one hiring rate of interest is 80% of the majority rate; i.e. 0.70 × 0.80 = 0.56. Thus, any minority selection rate below a value of 0.56 might be interpreted as arithmetic evidence of adverse impact. Thus, the minority rate might drop 0.70 − 0.56 = 0.14 percentage points below the majority rate before adverse impact is claimed. In contrast, suppose another organization is more selective in its hiring, such that the majority selection rate is, say, 0.20. The four-fifths rule would then imply that any minority rate below 80% of 0.20 (i.e. 0.20 × 0.80 = 0.16) could be interpreted as evidence of adverse impact. There is much less room in this second organization

The Four-Fifths Rule For Assessing Adverse Impact

189

for selection ratios to “drift” before prima facie adverse impact is claimed. Thus, the disparity between rates can be as large as 0.14 for the organization that hires 70% of its applicants, yet the disparity between rates drops to 0.04 for the more selective organization. To put this in another context, if an organization successfully implements a recruiting program and increases the number of applicants (or advertises a position opening on an Internet website and gets increased numbers of applicants), all else equal, the overall (and subgroup) selection rate will be lowered. In turn, the degree of acceptable selection ratio “drift” will be reduced. Thus, selective organizations will increase their likelihood of violating the four-fifths rule as a result of focused recruiting and increases in the applicant pool. The above underlying arithmetic issue is related to the fact that the four-fifths rule is about a ratio of ratios and is not defined by a critical difference in ratios. That is, it might be the case that (p 1 − p 2 ) = (p 3 − p 4 ), yet p 1 /p 2 is different than p 3 /p 4 . This possibility should be obvious, as the values of the ratios substantially depend on what the value of the denominator is. Yet another way to phrase the current issue is to note that, all else equal, the four-fifths rule is less likely to be a concern to organizations that are less selective. Our conclusion is consistent with the findings of Sackett and Ellingson (1997), who considered the effects of standardized group differences in test scores (“d”) on majority vs. minority hiring rates. Those authors found that the effect of a given “d” on differential hiring rates is “greater as an organization becomes more selective” (p. 711). This facet of the four-fifths rule is potentially troublesome – in that utility analysts have readily demonstrated that lower selection rates are associated with greater gains in selection system utility (e.g. see Boudreau, 1991). Thus, there can be a tradeoff between organizations wanting to maximize their utility via increased selection standards (lower selection rates) and the potential for prima facie adverse impact. Selective organizations faced with potential lawsuits should recognize this issue as an explanation for increased violation of the four-fifths rule, relative to less selective organizations. Selectivity and Hypothesis Tests We have already noted that selective organizations have less room for “drift” in selection rates before the arithmetic four-fifths rule was violated. For different underlying reasons, similar statements apply to assessment of adverse impact using statistical significance testing. In Eq. (1), it can readily be seen that organizations with selection rates that depart from 0.50 (including selective organizations) are more likely to obtain significant results from a test for equal rates. Specifically, note that the denominator of the equation for z depends on the value p(1 − p). The value of p(1 − p) is maximized when p = 0.50. As p departs from 0.50, the value

190

PHILIP BOBKO AND PHILIP L. ROTH

of that expression gets smaller. Given that this expression is in the denominator, all else equal the value of z gets larger and the difference in rates is therefore more likely to be statistically significant if the organization is being very selective (or, conversely, hires almost everyone). Framing The Lack of Symmetry in Framing the Four-Fifths Rule Research in cognitive psychology has shown that the effects of framing can substantially affect decision-maker behavior (Kahneman & Tversky, 1984). In turn, it is important to realize how the Guidelines have framed the notion of adverse impact. Specifically, adverse impact is framed in terms of selection rate, rather than rejection rate. In fact, note that adverse impact defined in terms of rejection rate would not necessarily lead to the same conclusions as adverse impact defined in terms of selection rate. For example, suppose an organization selects 15% of its majority applicants and 10% of its minority applicants. In this instance, the four-fifths rule is violated in that the minority selection rate (0.10) is less than 80% of the majority selection rate (0.80 × 0.15 = 0.12). However, note that the majority and minority rejection rates are much “closer,” proportionately speaking. That is, the majority rejection rate is 0.85 while the minority rejection rate is 0.90. These two rejection rates are within 80% of one another. A different example could readily be developed in which the outcome was the exact opposite (four-fifths rule not violated for selection rates but the rejection rates not within 80% of one another). The point is not to rewrite equal opportunity employment law. Rather, our focus here is to note that regulatory agencies have made choices that involve decisions during the development of the rule that might go unrecognized. An important implication is that the four-fifths rule is not “symmetric” with respect to framing issues via selection rates or rejection rates. The Guidelines clearly call for framing with respect to selection rates. Note also that we have been implicitly discussing the first sentence of the quoted four-fifths rule (i.e. the arithmetic consideration of a ratio of selection rates). As suggested, the remainder of the opening quote concerns hypothesis testing, and there is a shift to the difference in selection rates (not the ratio of rates). Of course, the difference in selection rates is the same as the difference in rejection rates! (For example, in the hypothetical earlier in this section, the difference in rates is 0.05 whether one looks at selection or rejection rates.) Thus, when the focus turns to statistical inference there is a natural symmetry that does not occur when considering the arithmetic ratio of selection rates. The above has practical implications. The fact that selection and rejection rate frames are symmetric in hypothesis tests, but not in arithmetic considerations,

The Four-Fifths Rule For Assessing Adverse Impact

191

can lead to ambiguity in whether or not there is adverse impact. For example, one of us has seen a situation where the majority selection rate (for promotion) was 0.98 and the minority selection rate was 0.96. To the host organization, everything was “acceptable” because the 80% rule was not violated. However, the plaintiffs chose to note that the minority rejection rate (0.04) was twice the rejection rate for the majority group. While this does not reflect the arithmetic frame for “selection,” in this instance the 0.02 rate difference was indeed statistically significant and the “twice” word was invoked (after reporting the statistical significance). Levels of Analysis Aggregation and Simpson’s Paradox The fact that the four-fifths rule is based upon use of percentages also implies that analyses are subject to a phenomenon in mathematics often labeled “Simpson’s paradox” (Simpson, 1951). For example, consider the data contained in Table 3. In this instance, note that across the entire organization (i.e. in the row labeled “Total”) the selection rate for males (23%) is greater than that for females (15%). Application of the 80% rule is indicative of adverse impact (i.e. 15% is less than 0.80 of 23%). Given such a result, one might ask “What job in the organization is ‘causing’ the adverse impact?” Of course, as one peruses Jobs A through D, it becomes clear that within each job, there is no adverse impact and the female selection rate is always twice the male selection rate! We note there is a similar, converse point if the columns labeled “Males” and “Females” are switched. In that situation, the overall organization would show no adverse impact, yet there would have been adverse impact within each job. [The “obvious” arithmetic reason for this paradox is Job D. When one averages selection rates across jobs but within Table 3. Data Illustrative of Simpson’s Paradox. Male # Appl.

# Hired

Female % Hired

# Appl.

# Hired

% Hired

Job A Job B Job C Job D

150 200 100 50

30 70 15 2

20 35 15 4

40 50 50 600

16 35 15 48

40 70 30 8

Total

500

117

23

740

114

15

192

PHILIP BOBKO AND PHILIP L. ROTH

gender, the effect of the small rate for Job D is minimized for males (n = 50) yet substantial for females (n = 600).] We have seen the Simpson paradox labeled as a “composition effect” or as a “stratification” issue (Meier et al., 1984). Indeed, Meier et al. point out that such a “paradox” occurred for the selection system (of undergraduates) at the University of California at Berkeley. In particular, the male selection rate for one year was higher than the female selection rate, yet within virtually all departments the selection rate for males was lower than that for females (see Meier et al., 1984, p. 156). The point is that the answer to “Is there a difference in selection rate that suggests adverse impact?” depends upon the level of analysis chosen (see Ostroff & Harrison, 1999, for a treatment of this topic in meta-analysis). Defendants and/or plaintiffs might present analyses by organization, by jobs within the organization, by division, by levels of prior experience of the applicants, and so forth. Different levels of analysis might lead to very different conclusions. Again, the four-fifths rule, although based upon an arithmetic logic, is subject to ambiguity. We believe that this is one reason why the word “may” appears so often in the opening quote to this paper. Combining data across minority groups is a particular instance of the ambiguity created by choices about levels of analysis. As noted earlier, it is our experience that analysts will sometimes combine data from two ethnic groups (e.g. Blacks and Hispanics) and compute selection ratios for this combined group. The composite selection rate is then arithmetically compared to the selection rate for, say, Whites. The notion of Simpson’s paradox should make it clear why this is not necessarily a helpful procedure. (In fact, Simpson’s paradox is more likely to occur precisely when sample sizes are quite different and that is often the case in disparate impact situations.) Namely, it might be the case that there is no adverse impact for the combined group, yet adverse impact might be present in each subgroup. Or, the converse might occur.

Combining Arithmetic and Statistical Assessment A major point in our analysis of the four-fifths rule (in the opening quote) is a shift from an arithmetic focus to a statistical one. We have noted that this shift also implies a shift in focus from the adverse impact ratio of two selection rates to a focus on the difference in selection rates. Recent work by Fleiss (1994) and Morris and Lobsenz (2000) suggests that the field could benefit from statistical analyses (hypothesis tests and confidence intervals) based directly on the adverse impact (a.i.) ratio.

The Four-Fifths Rule For Assessing Adverse Impact

193

Testing the Ratio Directly Morris and Lobsenz (2000) suggest conducting statistical analyses on ratios (not differences) of rates using Fleiss’ (1994) results. Specifically, because the distribution theory of the a.i. ratio is extremely complex, analysis is based on the logarithm of the a.i. ratio (natural logarithm being used). Second, large sample (asymptotic) theory is used. Assuming that the two population selection rates are equal (i.e. the population a.i. ratio is 1.0), Morris and Lobsenz (citing Fleiss) indicate that the standard deviation of the logarithm of the a.i. ratio is approximately:
  1−p 1 1 S.D. [log(a.i. ratio)] = (2) + p n1 n2 where p is the overall selection ratio (presumed equal for both groups) and n1 and n2 are as before. (Note: a more general formula is available if the selection rates are not assumed equal.) Note that when using the standard deviation in Eq. (2), one must apply it to the logarithm of the a.i. ratio. This has intuitive appeal because the logarithm of a ratio is the difference in logarithms; i.e. log(p 1 /p 2 ) = log(p 1 ) − log(p 2 ). Thus, the test statistic for the ratio also reflects a difference (but a difference in the log of the rates, not the rates themselves). To see how such an equation might operate, consider a situation where p 1 = p 2 = 0.5. That is, the population a.i. ratio is assumed to be 1.0. Assume that n1 is very large (i.e. the term 1/n 1 is essentially ignorable). Assume also that n2 (minority group sample size) is 49. Equation (2) then results in S.D. [log(a.i.)] = 0.143. It is important to remember that the value 0.143 is not the variation associated with the a.i. ratio. Rather it is the variation associated with the log of the a.i. ratio. If interest is in a 95% confidence interval, then multiply 0.143 times 1.96 and −1.96. Take the anti-log of those values (i.e. e1.96×0.143 and e−1.96×0.143 , or 1.32 and 0.76). Those values are then multiplied by the assumed a.i. ratio, resulting in upper and lower bounds of 1 × 1.32 and 1 × 0.76, respectively. Thus, the 95% confidence interval for an a.i. ratio based on a minority sample size of n 2 = 49 includes the range between 0.76 and 1.32 – a substantial expected variation. The point is that asymptotic equations exist for analyzing adverse impact ratios directly, rather than depending upon a test of the difference in two selection rates. We encourage more work in this area. Testing the Four-Fifths Rule Statistically A direct focus on the a.i. ratio could lead to some other interesting possibilities. For example, the arithmetic four-fifths rule has, as an implied standard, the value

194

PHILIP BOBKO AND PHILIP L. ROTH

of 0.80. In contrast, the traditional statistical test focuses on the hypothesis that H0 : p 1 = p 2 and an implied standard of p 1 /p 2 = 1.0, not 0.80. Thus, another way to blend arithmetic and statistical perspectives together would be to test the null hypothesis H0 : p 1 = 0.8p 2 where p2 is the majority rate. This would involve algebraically deriving the standard deviation associated with (p 1 − 0.8p 2 ) rather than the standard deviation of (p 1 − p 2 ) as is traditionally done (see Eq. (1)). Such an idea was also considered in an appendix in Meier et al. (1984) but there is little discussion of this approach in the current literature.

IMPLICATIONS FOR FUTURE RESEARCH Some ambiguities, unintended consequences, and clarifications of use of the fourfifths rule have been explored – with arithmetic, logical, and statistical thinking in mind. Although the rule has been, and remains, important for organizations and regulatory agencies, we believe there is additional basic and applied research that can help inform the field. Based on the above analysis, we call for more research and academic dialogues on the following issues. The list is not exhaustive and we hope that we have motivated individuals in the field to think further about these issues and derive investigations beyond those listed here. Our suggestions follow topics in the same order as the above text. Introductory Issues
The four-fifths rule originated from the federal government’s desire for a common, uniform, set of guidelines – as different government agencies (e.g. Department of Labor, Department of Justice) had been separately developing their own internal guidelines. We call for a qualitative analysis of the historical origins of the four-fifths rule during the development of those uniform guidelines. Specifically, what was the logic and motivational intent of those who proposed and adopted the rule? Why was four-fifths (or 0.80) chosen and not, say, two-thirds (or 0.67)? Was the choice of a value other than 1.0 used as a way to incorporate sampling error in small sample situations? Did the developers realize they were confounding ratios with differences, the ambiguity associated with rates and percentages, and so forth? Arithmetic and Statistical Issues
We have noted the distinction between the ratio and the difference of selection rates. The four-fifths arithmetic rule uses ratios. However, when presented with

The Four-Fifths Rule For Assessing Adverse Impact













195

information in a ratio format, do decision-makers understand the nature of ratios? Or, do they automatically translate the ratios into differences? Also, do decisionmakers understand that ratios are unstable when the denominator of the ratio is small? And so forth. Although sampling variability was a facet of our analysis (both in regards to variation around the four-fifths rule, as well as variation in regard to statistical tests), the “bottom” line in many situations is a dichotomous one – i.e. a decision as to whether there is adverse impact or not. We noted that the statistical determination of adverse impact is a function of several factors, including sample size, and called for reporting of adverse impact via effect sizes that were not heavily influenced by sample size. For example, some individuals report values of the standardized difference in mean scores between two subgroups (the statistic commonly being labeled “d”). Or, the four-fifths rule directly compares the adverse impact ratio (a.i. ratio) to the value of 0.80. We suggest that research on the “continuum” of values of the a.i. ratio or values of d should be investigated. Do decision-makers really have an internal value function such that any a.i. ratio above 0.80 is equally acceptable and anything below 0.80 is equally unacceptable? That is, is the implicit value function a step-function? Or, do decision-makers have a more continuous function in mind when they think about values of the a.i. ratio? If so, what is the shape and nature of that function? Knowledge of this function could help predict when, and how strongly, organizations react to potential adverse impact in a selection system. Such functions might also help researchers investigate the perceived trade-off between adverse impact of a selection system and its validity. We also noted that sampling variation in selection ratios and adverse impact ratios might be due to a variety of factors (random sampling, changes in selection system, changes in recruiting efforts, and so forth). We call for research that investigates to what extent variation in adverse impact ratios can be attributed to these non-sampling error factors. The existence of Type I and Type II errors was documented, via both logic (understanding of how statistical procedures operate) as well as mathematical results using simplistic scenarios. We call for other, perhaps Monte Carlo, simulations that investigate the degree to which Type I and Type II errors occur when using the four-fifths rule (both for ratios and differences). Among other factors, such simulations could vary population differences in selection rates, sample sizes, and number of stages (most analyses to date have focused on a single stage of selection). We noted that the need to correct hypothesis tests for experimentwise error rate might depend on the perception of the person(s) conducting the analysis. For example, organizations might want to perform such corrections so as

196

PHILIP BOBKO AND PHILIP L. ROTH

not to be penalized for having many subgroups in their applicant population (and hence having to conduct many hypothesis tests). On the other hand, complainants may want their case treated as a separate issue, and corrections for other groups should not be made. We encourage academic dialogue on this value-laden issue. Further Ambiguities and Clarifications
The four-fifths rule is framed within the context of selection rates, rather than rejection rates. It would be interesting to study how this framing affects perceptions of adverse impact. If a decision-maker is told that selection rates are not within 80% of each other, is this perceived as a more severe violation of “equity” than if the decision-maker is told that rejection rates are not within 80% of each other? (Or, conversely.) Answers to this question are not only academically interesting. That is, we noted that the difference in rates did not depend on frame. So, if there is a statistically significant difference between rates, when presenting the difference should it be framed as a rejection or selection finding in order to achieve the maximum perceived effect?
We also noted that work by Morris and Lobsenz (2000) uses procedures based on the asymptotic distribution of the a.i. ratio, rather than sampling theory about the difference in rates. We encourage more work in this area. For example, how good is the asymptotic approximation in Eq. (2) in situations when subgroup sample sizes are small (which is not uncommon in practice)?
Using Simpson’s paradox, we also indicated that a finding of adverse impact might depend on the level of analysis chosen. We encourage work in this area. For example, researchers might delineate strategies for prioritizing levels of analysis and their relative importance in assessing adverse impact. Again, we hope that researchers are motivated to generate even other research topics and questions as a result of the above analysis. We look forward to the increased understanding that all of these studies and dialogues will bring.

NOTES 1. The Guidelines allude to sampling error – particularly sampling error in situations when sample sizes are small – in a series of questions and answers designed to clarify and interpret the Guidelines. Specifically, question No. 21 conceptually asks, “What if one more minority had been hired (rather than a majority group member)?” The answer provided is that enforcement action is not necessarily appropriate if this single, simple switch in data would have changed the outcome.

The Four-Fifths Rule For Assessing Adverse Impact

197

REFERENCES Austin, J., Boyle, K., & Lualhati, J. (1998). Statistical conclusion validity for organizational science researchers: A review. Organizational Research Methods, 1, 164–208. Boardman, A. E. (1979). Another analysis of the EEOC Four-Fifths Rule. Management Science, 8, 770–776. Bobko, P. (2001). Correlation and regression: Applications for industrial organizational psychology and management (2nd ed.). Thousand Oaks, CA: Sage. Boudreau, J. (1991). Utility analysis for decisions in human resource management. In: M. Dunnette & L. Hough (Eds), Handbook of Industrial and Organizational Psychology (Vol. 2, pp. 621–745). Palo Alto, CA: Consulting Psychologists Press. Breaugh, J. (2003). Effect size estimation: Factors to consider and mistakes to avoid. Journal of Management, 29, 79–97. Camilli, G., & Hopkins, K. (1979). Testing for association in 2 × 2 contingency tables with very small sample sizes. Psychological Bulletin, 86, 1011–1014. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. DeLucchi, K. (1983). The use and misuse of chi-square: Lewis and Burke revisited. Psychological Bulletin, 94, 166–176. Fleiss, J. (1994). Measures of effect size for categorical data. In: H. Cooper & L. Hedges (Eds), The Handbook of Research Synthesis (pp. 245–260). New York: Russell Sage. Greenberg, I. (1979). An analysis of the EEOC Four-Fifths Rule. Management Science, 8, 762– 769. Guinyard v. City of New York, 800 F. Supp. 1083, E. D. N.Y. (1992). Harcum, E. (1989). The highly inappropriate calibration of statistical significance. American Psychologist, 44, 964. Hazelwood School District v. United States, 433 U.S. 299 (1977). Hsu, L. (1993). Using Cohen’s tables to determine the maximum power attainable in twosample tests when one sample is limited in size. Journal of Applied Psychology, 78, 303– 305. Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American Psychologist, 39, 341– 350. Kirk, R. (1995). Experimental design: Procedures for the behavioral sciences (3rd ed.). Pacific Grove, CA: Brooks/Cole Publishing Co. Lawshe, C. (1987). Adverse impact: Is it a viable concept? Professional Psychology: Research and Practice, 18, 492–497. Meier, P., Sacks, J., & Zabell, S. (1984). What happened in Hazelwood: Statistics, employment discrimination, and the 80% rule. American Bar Foundation Research Journal, 83, 139– 187. Mood, A., & Graybill, F. (1963). Introduction to the theory of statistics (2nd ed.). New York: McGraw-Hill. Morris, S., & Lobsenz, R. (2000). Significance tests and confidence intervals for the adverse impact ratio. Personnel Psychology, 53, 89–111. Ostroff, C., & Harrison, D. (1999). Meta-analysis: Level of analysis, and best estimates of population correlations: Cautions for interpreting meta-analytic results in organizational behavior. Journal of Applied Psychology, 84, 260–270.

198

PHILIP BOBKO AND PHILIP L. ROTH

Publication manual of the American Psychological Association (5th ed.) (2001). Washington, DC: APA. Sackett, P. R., & Ellingson, J. E. (1997). The effects of forming multi-predictor composites on group differences and adverse impact. Personnel Psychology, 50, 707–721. Simpson, E. (1951). The interpretation of interaction in contingency tables. Journal of the Royal Statistical Society Series B, 13, 238–241. Uniform Guidelines on Employee Selection Procedures (1978, August). Federal Register, pp. 38290–38315. Wing, H. (1982). Statistical hazards in the determination of adverse impact with small samples. Personnel Psychology, 35, 153–162.

GOING PLACES: ROADS MORE AND LESS TRAVELED IN RESEARCH ON EXPATRIATE EXPERIENCES David A. Harrison, Margaret A. Shaffer and Purnima Bhaskar-Shrinivas ABSTRACT We review 25 years of research on expatriate experiences concentrating on expatriate adjustment as a central construct, and relying on a general stressor-stress-strain framework. First, we consider who expatriates are, why their experiences differ from domestic employees, and what adjustment is. Conceptualizing (mal)adjustment in terms of stress, we next review the stressors and strains associated with it. Consolidating the wide range of antecedents (anticipatory and in-country) that have been studied to date, we note major patterns of effects and their implications for how HR managers can facilitate adjustment. Although relatively less research has focused on the consequences of adjustment, enough evidence exists to establish a bottom-line impact of poor adjustment on performance. To stimulate future efforts to understand the experiences of expatriates, we discuss the challenges and opportunities of continuing down this road of research. How long did it take to adjust to living in the Netherlands? About 37 seconds. Unlike other people who came on assignment – we knew where we were, I already knew enough of the language to get by, we knew quite a bit about the culture, and we had good friends there, so it

Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 199–247 Copyright © 2004 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23005-5

199

200

DAVID A. HARRISON ET AL.

was a relative easy adjustment. Also, the Netherlands is very much Western-working and that is very close to what it is like here. But, there are some amazingly subtle differences and it is these that take getting used to. Americans tend to have large personal spaces. The Dutch, living in such a compact country and having so many people packed in a small space, that the concept of personal space is virtually unknown. There is no such thing as personal space. People physically are on top of each other all the time. They are on top of each other on the streets, on top of each other in the supermarkets, in line at the bank, and at the post office. Now the other side of that, is that unlike Americans who have large physical distances that tend to be relatively open and informal, the Dutch have compensated for the lack of physical space by being relatively formal and relatively closed in terms of social relationships. For example, if you were to move into a new neighborhood in Arlington, within 30 seconds all your neighbors would come around and bring you casserole and welcome you. In the Netherlands, it is just the opposite. A person’s home is their castle, when you’re ready, you invite them over. So, the difference between physical and social space is not apparent to a lot of Americans and it’s very disturbing to them. Formally, in the Netherlands, when you meet someone you give the full name – first and last – until they get to know you. They have to initiate that informality. Why did the company send me to the Netherlands? Two reasons: one is they had a need for somebody with my background in software development and office innovation and I wanted to go and they knew I wanted. Still, I did not initiate the move. They came after me. What I was doing was not new to me – the culture was new, the site was new, the people who worked for me were new: the work group in Netherlands was quite diverse, with two Americans, a Dane, a Swiss, several Dutch. But, the work was the work I had done before. You have to be flexible. When we were in the Netherlands there was a guy we used to call the ‘robber baron’ because he used to stock a lot of American goods. He would pay exorbitant amounts of money to have Cheerios or Hellmann’s Mayonnaise. ‘If I don’t have my Skippy peanut butter I can’t have lunch.’ That’s baloney. We never had problems like that because we didn’t create problems like that for ourselves. – a well-adjusted expatriate How long did it take to adjust to living in Italy? I never did. Living there, existing there, is not a problem. It’s the level of frustration that you either sublimate or put up with or opt to not put up with. It’s a frustrating existence because of the way the society works. Some of the most beautiful stuff in the world is there, including some of the most beautiful country. Everybody should go there for a month as a tourist – but you don’t want to live there. I can assure you that if you poll my colleagues you would get the same answer. Heck, if you polled my Italian colleagues, you would get the same answer. You just don’t want to live there – with the taxes, with the post office that has been on strike for three months and they can’t deliver the backlog, with the phone system that doesn’t work, with the fact that we had 3 kilowatts of electricity for our entire apartment so God forbid you turn on two appliances at the same time and run downstairs. It was a nightmare, but you have to take it with a grain of salt. I didn’t expect to be living in Keller when I moved to Rome. What I didn’t really expect was how bad it was going to be. IBM does not send you to language training beforehand, but they will give you language training on-site. We found that in Rome, many of the expats knew in advance they were going, so they took the Berlitz course – but by the time they got to Rome, they forgot what they had learned. My experience tells me that you are much better off learning there rather than beforehand. In Europe, anyway, enough people speak English so you can manage to get by. At work, the official language of the company is English. And English is by and large used.

Expatriate Experiences

201

However, the reality of the situation is that under stress people will always use their native language. In terms of diversity, Italy was either Italians or U.S. assignees. Very few of the company’s employees in Italy were not Italian. In Rome, I was a senior planner – it was a technical job and I was not managing. Working with people from other nationalities was great. It can get frustrating at times, of course. In Italy especially, it is tough to get anything done. There is one sentence in Italian that I made up which describes the way the entire society works and I will translate it: ‘My mother told me that I was best, and I don’t have to worry about anybody else.’ That’s the way they are trained – it filters down through all aspects of the society – driving, standing in line, trying to get a bill paid, getting through the supermarket. I don’t care what aspect of activity you pick, that sentence dominates peoples’ being. To survive, you have to adopt that philosophy for yourself – otherwise you’re don’t survive. I mean, a car waiting in line to make a left turn? No way. – a poorly adjusted expatriate

The thrills and the stresses of being an expatriate are richly captured in accounts such as the two given above. So is the dilemma for international HR managers. These two accounts came from the same person working for the same multinational firm doing roughly the same job, but posted to two different overseas locations (the quotes do not represent our opinions of either country; another person might have had a wonderful working experience in Italy, and a trying one in Netherlands). Rather than concentrate on functions, processes, and outcomes in the domestic work environment that make individuals better contributors to their organizations, an intense, overriding concern for HR professionals managing international employees is how those employees experience all of their foreign environment: their new jobs, their new relationships, and especially, their new culture (Shaffer, Harrison & Gilley, 1999). This dilemma is being faced by large and steadily increasing numbers of HR managers. For example, Dolainski (1997) estimated that nearly 500,000 U.S. employees and over 2,000,000 employees worldwide are working for their firms outside their home countries, nearly double some of the numbers estimated in the mid-1980s (Tung, 1986). Annual surveys by GMAC Global Relocations (e.g. GMAC, 2003; formerly Windham International), the National Foreign Trade Council (NFTC), and the Society for Human Resources Management (SHRM) suggest strong, double-digit percentage rises in the numbers of expatriates since 1997. Academic research on expatriates has paralleled this growth. Investigations have moved past signal studies in the 1960–1970s of the experiences of student sojourners, Peace Corp Volunteers and military personnel (e.g. Mischel, 1965). Work flourished in the late 1980–1990s in studies of experiences of managers and technical employees sent to complete long-term international assignments (typically > 1 year), at the behest of their employers. The purpose of this article is

202

DAVID A. HARRISON ET AL.

to review that latter road of research, separating what is known from what is yet to be known about expatriate experiences. Specifically, we review the past 25 years of trends in the increasingly coherent expatriate literature, consolidate findings on the nature, antecedents, and consequences of expatriate adjustment (the focal variable in the literature), and make suggestions for studies in what we hope will be another major thoroughfare of expatriate research over the next decades. Our review is split into four major sections. The first three correspond to the stressor-stress-paradigm that dominates current research, albeit not in that particular sequence. The first section deals with the stress of expatriate adjustment: the substance of adjustment as a psychological state and its likely course or process over time. In that section, we try to clarify who expatriates are, how they and their assignment “trajectories” might differ in systematic ways (e.g. from host to home, or home to host countries), and how their experiences can be separated into identifiable dimensions. In doing so, we try to hone in on defensible conceptual definitions, and therefore, greater clarification of the basic constructs and populations being studied. The second section deals with stressor constructs and their buffers, including the mounting evidence for sets of both anticipatory and on-assignment antecedents of adjustment proposed in the most popular (in terms of spawning empirical investigations) framework in the literature: Black, Mendenhall and Oddou’s (1991) model. Although there are many dozens of studies covering many dozens of possible stressors, some strong patterns have emerged. Through selection, training, and policy-based support, the evidence suggests several ways that international HR managers can facilitate adjustment. The third section covers strains, or the connections between adjustment and its possible consequences for assignees. In that review, we document evidence for the bottom-line impacts of (poor) adjustment, and we note the lack of integrated approaches to the expatriate criterion space, especially relative to the predictor space. Finally, the fourth section concerns what we see as major blind spots in the current literature, and we suggest some forks in the road: what new questions might deserve investigation given the changing nature of multi-national operations?

EXPERIENCING THE STRESS OF AN INTERNATIONAL ASSIGNMENT Three basic issues must be addressed before proceeding with the review of research on expatriate experiences, to provide greater conceptual clarity and facilitate integration of theoretical and empirical findings. These issues, posed as questions, include: (a) who is it that we are studying (who are expatriates)?; (b) why are their

Expatriate Experiences

203

experiences different from domestic employees (e.g. simple relocations or transfers)?; and (c) what about expatriates is the focus of study? Through theoretical and empirical consolidation, the latter question has become: what is (the stress of) adjustment? Who are Expatriates? Although seemingly a simple question, there are insights to be had from working with the possible answers. The process of international transfers or relocation to a foreign culture has been the subject of extensive theoretical and empirical scrutiny for over 50 years. As we mentioned above, the focus of early research was on “sojourners,” with several articles and books devoting considerable space to their adaptation and effectiveness (Brein & David, 1971; Cleveland, Mangone & Adams, 1960; Gullahorn & Gullahorn, 1963; Kelman, 1965). Sojourners are defined as individuals who travel to a foreign country, voluntarily, on a temporary basis and for a distinct, usually non-business purpose (e.g. tourism, refuge, work, studies, exploration, soldiering, religious missions etc.; Aycan & Kanungo, 1997; Gudykunst & Hammer, 1984). Beginning in the late 1980s, with the growing surge of economic and corporate globalization, researchers shifted their attention to business people who relocated overseas to fulfill an organizational purpose. Drawing from earlier definitions in the literature (see Aycan & Kanungo, 1997), we define expatriates as employees of business organizations, who are sent overseas on a temporary basis to complete a time-based task or accomplish an organizational goal. There are three components that are integral to this definition: type of the relocating employee, the duration or length of time spent overseas, and the purpose of the assignment. Given that each of these components has important implications for expatriate experiences, we explain them below. Type of Relocating Employee In addition to working for a parent organization, expatriates have also been classified on the basis of their national origin relative to that of the parent company (Borman, 1969; Briscoe, 1995). However there is little systematic study of these distinctions in the research literature. Parent country nationals (PCNs) are employees whose national origins are the same as that of country in which the MNC is headquartered, such as an American expatriate manager from Federal Express deployed to Hong Kong. Research on expatriate adjustment has been dominated by studies of PCNs. Inpatriates are foreign nationals who live and work in the parent country (e.g. a Korean national from the Korean subsidiary office of Federal Express deployed to the U.S.). Third country nationals (TCNs) are employees

204

DAVID A. HARRISON ET AL.

whose national origin and country of assignment are outside the parent country of the MNC (e.g. an Indian national on assignment at the subsidiary of Federal Express in Singapore). Host country nationals (HCNs) are work colleagues and citizens of the country where the expatriate is posted (e.g. Chinese employees from the Federal Express office in Hong Kong where an American expatriate is posted). Consistent with the theme of this article, our review excludes studies that focus exclusively on the experience of HCNs. On the other hand, building relationships with HCN’s is one of the elemental behavioral goals of expatriates, and studies of expatriate experiences have, unfortunately, spent very little effort on the HCN perspective (Toh & Denisi, 2003). Out of approximately 60 empirical articles, dissertations, and manuscripts that we reviewed on expatriate experiences, only a handful of studies make a clear distinction between PCNs, inpatriates and TCNs (e.g. Shaffer, Harrison & Gilley, 1999). Over 95% appear to use the term expatriates and PCNs interchangeably. Research rarely involves inpatriates or TCNs. Our review found three studies that included inpatriates among the sample of respondents (e.g. Caligiuri, 2000b) and two studies involving TCNs and PCNs (e.g. Caligiuri & Day, 2000), although these appear to be taken from the same original sample. Relocation experiences may vary across types of expatriates and hence it may be necessary to make distinctions in the literature among PCNs, inpatriates and TCNs. For example, TCNs, inpatriates and PCNs may espouse different organizational values due to differences in the intensity of socialization practices and processes between subsidiaries and the home base. Often selected to facilitate communication between the corporate and foreign offices, and being from the home or headquarters country of the firm, PCNs might undergo more rigorous indoctrination. Having limited authority and control, TCNs and inpatriates may be perceived as peripheral, command less support, and end up feeling isolated, abandoned or ostracized in a host country. In other words, TCNs and inpatriates may experience greater difficulties and challenges in expatriation when compared to their parent country counterparts, a possibility that has yet to receive systematic empirical scrutiny. Another way of thinking about the differences might be to say that TCNs and inpatriates would experience greater work-related, and perhaps relationship-related stress of adaptation, but stress associated with broader cultural issues might be the same as for PCNs (see the distinctions between these dimensions below). Yet another category of expatriates are self-initiated foreign workers (Inkson, Arthur, Pringle & Barry, 1997). Being personally motivated to live and work abroad, these individuals may even resign from their current jobs and self-sponsor their overseas move. Oftentimes, there may not be a pre-arranged job set up in the host country, and the individual may engage in menial work just to make ends

Expatriate Experiences

205

meet. Implicit in our definition is the notion that the relocating employee remains affiliated with the parent business organization during transfer. Furthermore, this category of expatriates has not elicited any attention in the empirical literature beyond the seminal ‘sojourner’ studies. Consequently, we strongly urge the inclusion of self-initiated foreign workers in future research that might span the boundaries between international HR and entrepreneurship studies. Duration or Length of Time Spent Overseas Our definition of an expatriate makes explicit the temporary nature of the relocation. The length of assignments may vary depending upon the purpose and organizational sub-goal for which the expatriate is originally sent (Furnham, 1987). Short term assignments typically range from six months to a year. Long term assignments, on the other hand, involve extended stay in the host country and range anywhere from 1+ to 5 years (DeCieri, Dowling & Taylor, 1991; Fontaine, 1996; Fukuda & Chu, 1994). The latter types of overseas stays are more common (GMAC, 2003). Over 85% of the expatriate studies we have reviewed do not report the expected duration of the expatriate’s assignment. Among studies that have reported it, there appears to be a tendency in the literature to focus on long term assignees (e.g. Black, 1990b). This contrasts with evidence suggesting that organizations are shifting the emphasis to short term assignments, to reduce costs and mitigate dual-career issues (ORC Worldwide Survey of International Assignment Policies and Practices, 1997–1998). The length or duration of assignments may have important implications for expatriate experiences. For example, to the extent that the end of the assignment is viewed as a goal, theories of goal-setting (Latham & Locke, 1991; Stock & Cervone, 1990) suggest that the assignment’s end date may serve as an effective self-regulation mechanism. Individuals who have nearer-term goals or shorter assignments may persist in their pursuit of the end date goal and see it through completion, as proximal goals tend to be more motivating than distal or long term goals (Shaffer & Harrison, 1998). From a cognitive perspective, splitting up a proximal goal into manageable units leads to an increase in the perception that the goal can be confidently executed. Per this view, shorter duration assignments should reduce uncertainty and lead to an increase in assignment-related effort, along with fewer thoughts of quitting. It would be useful to conduct a detailed investigation of these effects on the experiences of expatriates, perhaps by comparing matched pairs of long- and short-term assignees. Purpose of Assignments Oftentimes, the reason for using expatriates rather than local employees or HCNs is the scarcity of professionals in the host country with the specific knowledge

206

DAVID A. HARRISON ET AL.

or skills needed to meet task requirements (Harzing, 2001a, b). The reasons may include technical know-how, coordination and control, or management development. Hays (1974), classified assignments into four types based on their purpose or nature. When the expatriate assignment involves responsibility of building a system (e.g. marketing/finance/production) in the foreign subsidiary that is similar to the kind that he or she has already experienced or has adequate knowledge of, it is termed structure reproduction. Trouble-shooting assignments involve analysis and problem solution. Operational assignments require the expatriate to perform day-to-day functions in the foreign subsidiary and strategic or high profile assignments involve management of the entire foreign operation. Although there has been no stream of empirical investigations on assignment purpose, we believe that such aspects have strong potential and need to be tracked in future work on expatriate experiences. For example, strategic assignments may be perceived as more motivating and intrinsically challenging than operational ones, due to their high profile nature, perhaps staving off adjustment problems and psychological withdrawal. They also have higher potency for an expatriate’s career path as the organizational stake in the expatriate’s assignment is larger. Consequently, expatriates may expend more personal resources (time and energy) toward accomplishing their jobs in such positions. Another classification that has received some attention in the expatriate literature is partially based on the hierarchical level of the assignees (Shaffer et al., 1999). Technical assignments (e.g. IT or system engineers/analysts etc) do not require interactions with host national employees, and they involve expatriates in the lower to middle rungs of the organizational hierarchy. Higher-level, managerial assignments are often aimed at new market creation or development, and alliance management. The expatriate experience literature is dominated by studies that involve managerial assignees. Approximately 72% of the studies that have reported the nature of the assignments focused solely on expatriate managers (e.g. Selmer, 1999b). Even among studies that comprise a mix of technical and managerial personnel, sample make-ups tend to favor the latter (e.g. Nicholson & Imaizumi, 1993; Shaffer & Harrison, 1998). Almost by definition, managerial jobs have greater frequency, intensity and variability of interpersonal interactions with host national stakeholders, relative to technical jobs. In technical assignments, much of the on-the-job communication with host nationals is characterized by jargon that is part of a widely shared and accepted specialized vocabulary. This may reduce some of the opportunities for misunderstandings, miscommunication and misattributions. Consequently, the expatriation experience may become more difficult or challenging for managerial assignments vs. technical ones. At the least, the latter would be expected to involve a different mix of abilities, perhaps with a stronger social component.

Expatriate Experiences

207

Table 1. Comparison of International and Domestic Relocations. Salient Differences in . . .

International

Domestic

Environmental features (1) General environment (a) living conditions (b) health and welfare (c) security and safety (d) rule of law (e) religious beliefs (f) social support (g) language (2) Cultural environment (3) Interaction environment (4) Work environment

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Maybe No No No Maybe Yes No No Yes Yes

Length of transfer

Temporary

Permanent

Note: There may be exceptions, but we believe the cell entries accurately reflect the main trends.

Although our definition lays out the parameters required for qualification as an expatriate employee, it does not capture the nature or essence of the overseas experience. And, we have begged the question of why expatriates need to be, or have been, studied as singular populations. To better understand the uniqueness of the expatriation process, we ask the question below.

Why are Expatriate Experiences Different from Other Relocations? There are substantial differences in the magnitude and concentration of changes experienced in an overseas vs. a domestic relocation (refer Table 1 for a comparative summary). Domestic relocations, as with international relocations, involve a disruption in work and social support networks, job requirements, and organizational cultures. However, the broader economic, social, political, cultural and religious contexts of the former type of move are not likely to be vastly different from one’s prior setting; they remain familiar to the individual (Black et al., 1991). In contrast, the distinctive feature of expatriation is an individual’s (and usually, his or her family’s) move into a different embedding culture, where, on a number of salient dimensions, the expatriate is unlikely to have value, thought, or behavior patterns that fit comfortably with his or her situation (Bell & Harrison, 1996). Various aspects of the general environment such as living conditions (e.g. shopping, financial transactions, children’s schooling, public transport systems), health and welfare (e.g. access to medical facilities), safety and security (e.g. political

208

DAVID A. HARRISON ET AL.

instability, threat of personal harm), rule of law (e.g. norms of corruption, equal opportunity, ownership of property, code of conduct, citizenship rights), language, and religious differences may assume greater salience in a different host country setting. For example, with respect to religious beliefs, some cultures tend to be more or less tolerant or culturally “tight” than others (Chan, Gelfand, Triandis & Tzeng, 1996). To the extent that core dogmas and institutions, membership and regular participation in specific religious activities play a prominent role for an expatriate who is posted to a country with low tolerance for religions, the entire experience of relocation can become extremely stressful and tumultuous. Cultural differences may also become more pronounced in expatriation. Native country culture determines the way that individuals perceive, interpret and behave; indeed, it is foundational to how individuals define themselves (Markus & Kitayama, 1991; Torbiorn, 1982). When expatriates relocate to a foreign country, they carry with them a prevailing repertoire of perceptions, interpretations, and behaviors from their native culture, and often, with the underlying and mistaken assumption that these are preferable ways of seeing, understanding, and acting on the world (Bell & Harrison, 1996). In other words, after spending most of his or her life in a native country, it becomes difficult for an expatriate to adopt new norms and values and adopt beliefs and behaviors that might be demanded by host country work as well as non-work settings. This uncertainty is higher when the perceived differences between the host and native country are larger (Black & Gregersen, 1991a). A final point of difference concerns the length of transfer. A domestic relocation may seem permanent compared to an international transfer – at the very least, there is little expectation that one will return to his or her previous location in a domestic move. Because of the permanence of that move, the individual and his/her family may be forced to deal with the challenges head on. In contrast, if the move is temporary (almost all expatriates expect to be repatriated), there may be a tendency to seek recourse to a psychological escape route that can influence various short and long term decisions. For example, an individual may choose to rent rather than invest in a property. He or she may not make a conscious effort to build a social network of neighbors and friends, given that these ties will be severed once the assignment comes to an end. Such decisions have important consequences in that they do not bind individuals to the setting and may create an underlying resistance or avoidance to learning the norms and values of the host culture. To the extent that learning helps individuals cope with uncertainties, the psychological temporariness of expatriation makes it more challenging relative to a domestic transition. So far, we have distinguished between domestic and international relocations. The third type of relocation is repatriation: the return to one’s home country upon completion of an assignment. Repatriation can be as challenging as expatriation, if

Expatriate Experiences

209

not more, in that there are uncertainties with respect to continued employment in the home country and career progression, reverse culture shock and so on. Although the move may be more permanent when compared to expatriation, most of the social and financial investments made to minimize uncertainty in the host country may become barriers to repatriation. Research suggests that the psychological experience of repatriation may be altogether different (e.g. Gregersen & Stroh, 1997). Consequently, we restrict our review to an examination of the expatriation process. Our discussion of the nature and extent of differences between a domestic and international relocation highlights the need for wide-ranging psychological, affective and behavioral models to capture the breadth of uncertainties experienced during an international transition. With this background in mind, various scholars have attempted to understand the expatriation experience by proposing theoretical models. For example, Mendenhall and Oddou (1985) outlined the skills required for success in an overseas setting. Thereafter, Black et al. (1991) developed an integrated model of international transitions that has instigated and galvanized a large body of evidence. Although researchers have proposed some theoretical variations (see Aycan, 1997; Parker & McEvoy, 1993), all models of expatriation experiences place adjustment in a keystone position of psychological and behavioral mechanisms that involve anticipatory and on-assignment individual, job, organizational, and non-work variables. As such, adjustment has been the focus of most of the expatriate literature for the last two decades (e.g. Black, 1988; Shaffer et al., 1999; Torbiorn, 1982). In the following section, we review the literature on the meaning and centrality of adjustment for expatriate experiences. What is Expatriate Adjustment? Management theory and research on expatriate experiences shows an implicit recognition or agreement that adjustment is the principle force or backbone of the phenomena being studied (Aycan, 1997; Black et al., 1991; Kraimer, Wayne & Jaworski, 2001; Mendenhall, Kuhlmann, Stahl & Osland, 2002; Parker & McEvoy, 1993). Despite this convergence in general topic area, there is much less consensus regarding its constitutive definition and content. In this section, we distinguish adjustment from acculturation and adaptation, review previous viewpoints of adjustment from a state as well as a process perspective, and arrive at a working definition. Acculturation, Adaptation and Adjustment There is considerable confusion about, or an inadequate understanding of, acculturation, adaptation and adjustment, with the terms being used somewhat interchangeably in the literature on expatriate experiences (see Aycan, 1997;

210

DAVID A. HARRISON ET AL.

Mendenhall & Oddou, 1985). We believe that it is necessary to make conceptual distinctions between the terms to give a more precise understanding, and guide our review. Acculturation is a somewhat broad and overarching concept that is applicable to a range of social and cultural contexts. It refers to a continuous process of change whereby emotions, cognitions and behaviors of individuals from one culture are altered due to regular or frequent contact with individuals from another culture (Moyerman & Forman, 1992; Redfield, Linton & Herskovits, 1936). Clearly, acculturation is a two-way process that affects the relocating individuals as well as members of the host culture (Berry & Kim, 1988). Adaptation and adjustment can be viewed as critical elements within the process of acculturation. Drawing from the domestic (e.g. Chan, 2000) and international work literatures (e.g. Aycan & Berry, 1996) we define adaptation as a process whereby relocating individuals alter their behaviors to achieve a degree of fit with different aspects of the environment. Unlike acculturation, it is a means of attaining congruence with the environment; it is mainly a one-way process that is restricted to the relocating individuals. The processes of acculturation and adaptation are similar in that they make no distinction between the permanence and temporariness of the relocation and they have been widely applied in the literature to groups of immigrants and sojourners (although it could reasonably be argued that acculturation is seen as a longer-term process than adaptation). In contrast, adjustment is a more specific and narrow concept. It refers to a psychological state or process of experience (although, below, we favor its treatment as a state). It results from (an anticipated) temporary change in various aspects of the environment, and it can be proactive or anticipatory, as well as reactive or “in-country” (Dawis & Lofquist, 1984). Given the variability in definitions, we categorize them into state vs. process perspectives below. Adjustment as a State As a state, adjustment has been defined and measured in terms of various cognitive, affective and behavioral markers. There has been historical disagreement about its content and dimensionality, but there has also been a recent coalescing. Early definitions of expatriate adjustment were based on conceptualizations of subjective well-being (Campbell, 1981) which refers to a global state of dissatisfaction or feeling of unhappiness in expatriates’ perceptions of their circumstances (Munton & West, 1995). These definitions describe adjustment from the perspective of the individual experiencing the transition, and reference the overall or general environment of the host country (Ruben & Kealey, 1979; Torbiorn, 1982). In those definitions, adjustment is cast as a unitary phenomenon, similar to early conceptualizations of job satisfaction (Wanous & Lawler, 1972) and organizational commitment (Reichers, 1985).

Expatriate Experiences

211

Drawing from the acculturation literature (Aycan & Berry, 1996; Hawes & Kealey, 1981; Searle & Ward, 1990), some scholars (e.g. Aycan, 1997; Selmer, 1999b) have also referred to adjustment in terms of its socio-cultural characteristics and defined it as achieving effectiveness in interpersonal exchanges with host country nationals. This definition has the disadvantage of incorporating the effect or consequence of a construct as part of the construct itself (Harrison, 2002). We would counter that interpersonal effectiveness is further downstream in expatriate experiences, although it may well stem from adjustment – but it may not. That is, adjustment could be argued to be an intervening variable between the stressors of a cross-cultural assignment and effectiveness. Certainly, those who fail to adjust are unlikely to be effective in their exchanges with HCNs. In that sense, we posit that appropriate adjustment is a necessary condition for effective performance and must be treated as distinct from effectiveness. In addition to these views, attempts to understand adjustment have been based on a stressor-stress-strain formulation of the expatriation experience. This prevalent approach has been explicitly or implicitly relied upon by a number of investigators (e.g. Lance & Richardson, 1988; Parker & McEvoy, 1993, Takeuchi, Yun & Tesluk, 2002b). In the sequence, stressors are cast as demands in the foreign environment that are mismatched with an expatriate’s personal resources (Black & Gregersen, 1991b). Strains include affective, cognitive and behavioral outcomes such as job dissatisfaction, psychological withdrawal cognitions, turnover and poor performance (see Sullivan & Bhagat, 1992, for a review of stress outcomes in general). The experience of stress is represented by (mal)adjustment or the psychological (dis)comfort associated with various aspects of the international assignment: cultural, interaction and work. Cultural adjustment refers to the degree of psychological comfort associated with non-work aspects of the overseas environment, including transportation, entertainment, food, security and health care services. The (dis)comfort of interaction adjustment is with respect to interpersonal exchanges with HCNs both inside and outside of work settings. Lastly, work adjustment flows from the assigned job or tasks (Black et al., 1991). The tri-partite definition of adjustment clearly delineates the various aspects of the expatriate’s experiences, and it has been operationalized (Black & Stephens, 1989) and validated repeatedly (e.g. Shaffer et al., 1999). Of the three dimensions, interaction adjustment is often believed to be the most fundamental, as both work and general adjustment for expatriates are based on reconstituting their interpersonal interaction patterns in a new environment (Bell & Harrison, 1996). It is also believed that work adjustment may be the least difficult for the expatriate as most organizations use at least approximately uniform work policies and procedures across global subsidiaries (Black et al., 1991). Some researchers have proposed interrelationships between these three dimensions of adjustment

212

DAVID A. HARRISON ET AL.

(e.g. Aycan, 1997). Per that view, both psychological (well-being) and sociocultural (effectiveness in dealing with HCNs) adjustment are expected to affect work adjustment. Operationalization of Adjustment Nearly three quarters of the existing articles on expatriate adjustment have used the measure developed by Black and colleagues (Black, 1988; Black & Stephens, 1989). This scale assesses the three dimensions of adjustment given above, with seven items for cultural adjustment, four for interaction adjustment, and three for work adjustment. Cultural adjustment items measure expatriates’ adjustment to living conditions, climate, housing, food and so on. Interaction adjustment items examine expatriates’ adjustment to socializing with HCNs. Work adjustment items measure the expatriates’ adjustment to the requirements of the new position. For each item, respondents indicate their degree of adjustment on a seven-point scale (1 = very unadjusted to 7 = very adjusted). The psychometric properties of cultural and work adjustment vary somewhat across studies, with alphas ranging from 0.73 to 0.95 for cultural adjustment and 0.69–0.92 for work adjustment respectively. The band of alphas is appreciably narrower for interaction adjustment (0.82–0.93). Although it continues to serve the literature well, we believe that there may be issues in this instrument that need to be addressed, to improve its domain coverage and psychometric properties. First, there is fair amount of verbal redundancy in the interaction and work adjustment items that are not likewise reflected in cultural adjustment. For example, all the items that represent interaction adjustment contain a reference to “host country nationals.” It might be plausible that the items form one factor because of this wording cue or artifact (Harrison & McLaughlin, 1996; Shaffer et al., 1999 report a detectable artifact simply for the word “facilities” appearing in two items). Second, there are too few items in the interaction and work adjustment measure compared to those in cultural adjustment. If they are of commensurate importance to the stress of expatriate experiences, it may be necessary to broaden the latter measures. Finally, cultural adjustment also includes adjustment to climate which is a feature of the physical rather than social environment. Although it can differ dramatically across between nations, it can do so just as much within nations (e.g. Anchorage vs. Miami in the U.S.). The digression of this item may have contributed to some of the low reliabilities mentioned earlier. Adjustment as a Process Some scholars have maintained that adjustment is a process that describes the sequential how of expatriates experiencing their overseas environments. Researchers

Expatriate Experiences

213

in the organizational socialization and work role transitions domain (Dawis & Lofquist, 1984; Nicholson, 1984) have characterized adjustment as a mode or process of fit whereby expatriates adjust either by changing the new work role to fit their requirements (role innovation) or by changing their attitudes and behaviors to fit the work role (personal change). Expanding on these two dimensions, Nicholson (1984) suggested four modes of adjustment: replication (few changes in attitudes or behaviors and few changes in the work role), absorption (large changes in attitudes or behaviors and few changes in the work role), determination (few changes in attitudes or behaviors and large changes in the work role) and exploration (large changes in attitudes or behaviors and large changes in the work role). This conceptualization of adjustment as a mode is useful, and might lead researchers to characterize successful expatriate experiences as absorption or exploration, with unsuccessful experiences more likely to be failed attempts at replication or determination (both represent a foisting of one’s former, home country behavior pattern on the host country). On the other hand, this conceptualization tends to ignore the nonwork context for expatriates, which, alongside overlap with the concept of adaptation and difficulties in operationalization and measurement, has limited its application in expatriate research. Other researchers have attempted to explain and illustrate the sequential unfolding of adjustment over time in terms of a U-curve (Lysgaard, 1955; Torbiorn, 1982). It represents a stage-based model of the international assignee’s level of adjustment in the host culture. The four stages are “honeymoon,” “culture shock,” “adjustment,” and “mastery.” Black and Mendenhall (1991) explained the U-curve in terms of social learning theory (Bandura, 1977). The honeymoon stage is characterized by a fascination with the new host culture. Expatriates engage in trial and error learning and inappropriate behaviors may persist as negative feedback may not be forthcoming. Culture shock follows when expatriates recognize and learn to cope with the stresses of daily life. This is also the time at which expatriates experience their poorest performance and the greatest urges to withdraw. As expatriates learn to behave appropriately in the new setting, there is a gradual reduction in stress or adjustment to host country culture. During this stage, expatriates’ feelings of satisfaction and self-worth increase. The final stage is mastery which is marked by an increased ability to function in the new setting. Although the U-curve theory provides a detailed explanation of each adjustment step, it has not received much empirical attention. This might be attributable to the difficulty in tracking expatriates over time. Only two studies in our review studied adjustment longitudinally (Kealey, 1989; Nicholson & Imaizumi, 1993), speaking to a dramatic weakness in the expatriation literature. Even the few studies that have investigated the effects of time have been limited to student or ‘sojourner’

214

DAVID A. HARRISON ET AL.

samples, restricting generalizability to employees sent by parent firms on international assignments (Black & Mendenhall, 1991). From a temporal perspective, it is critical for organizations to track the periods during which expatriates are at the greatest risk for maladjustment, withdrawal and reduced performance. It may also be the time when expatriates require maximum support from organizations. Although saying “more longitudinal research is needed” sounds trite, we would say it is desperately needed with respect to mapping the pattern or stages of adjustment for expatriates.

STRESSORS IN INTERNATIONAL ASSIGNMENT EXPERIENCES Having clarified who expatriates are and what constitutes the stress of adjustment to international assignments, we now turn our attention to a review of where that stress comes from. Researchers have drawn upon several theoretical paradigms, including learning, stress-coping, developmental, and personality-based models, to understand expatriate experiences (Mendenhall et al., 2002). Based on these paradigms, three theoretical frameworks – all with adjustment as the focal construct – have been developed. The first framework, by Black and his colleagues (1991), postulated that expatriate adjustment was a function of both anticipatory and in-country factors. Although the other two frameworks (Aycan, 1997; Parker & McEvoy, 1993) extended conceptualizations of adjustment to include consequences, the majority of the empirical literature has focused on identifying its antecedents – the stressors that lead to adjustment problems implicit in the stressor-stress-strain formulation of the expatriation experience (e.g. Lance & Richardson, 1988; Parker & McEvoy, 1993; Takeuchi et al., 2002b). Stressors are cast as demands in the foreign environment that are mismatched with an expatriate’s personal resources (Black & Gregersen, 1991c). To facilitate a review of such mismatches, we categorized the various antecedents (i.e. personal, work, family/friends and environmental factors) and consequences (i.e. satisfaction, withdrawal cognitions and performance) of adjustment and noted the correlates that have been detected (see Table 2). Insofar as most studies have used the tripartite conceptualization of adjustment introduced by Black and colleagues (Black, 1988; Black & Gregersen, 1991a), and described above, we identify correlates of these three dimensions in Table 2. Studies using a unidimensional or composite measure of adjustment are subsumed under the cultural dimension heading.

Correlates Stressors 1. Personal factors a. Demographic characteristics Gender

Age (−) Education (+) Nationality (+) Assignment duration (+)

Previous international experience (+)

Cultural Adjustment

Florkowski and Fogel (1999) (F > M) Shaffer and Harrison (1998) (F > M)

Parker and McEvoy (1993) Caligiuri (2000b) Florkowski and Fogel (1999) Gregersen and Black (1990) Gregersen and Black (1992) Kraimer et al. (2001) Li (1995) Caligiuri (2000b) Kraimer et al. (2001) Selmer (2001a) Selmer (2002)

Tenure (+) Promotion (+) b. Personal traits Introversion (−) Work self-efficacy (+)

Interaction Adjustment

Harrison et al. (1996)

Work Adjustment

Expatriate Experiences

Table 2. Summary of Detected Relationships.

Florkowski and Fogel (1999) (F > M) Shaffer and Harrison (1998) (M > F) Selmer et al. (2000) Shaffer and Harrison (1998) Selmer (1998) Gregersen and Black (1990)

Florkowski and Fogel (1999) Gregersen and Black (1990) Kraimer et al. (2001)

Selmer (2001a) Selmer (2002)

Selmer (2001a) Selmer (2002)

Shaffer and Harrison (1998)

Shaffer and Harrison (1998) Parker and McEvoy (1993) Shaffer and Harrison (1998)

Parker and McEvoy (1993) Harrison et al. (1996) Shaffer et al. (1999)

Harrison et al. (1996) Shaffer et al. (1999)

215

216

Table 2. (Continued ) Correlates

Cultural Adjustment

Interaction Adjustment

Harrison et al. (1996)

Harrison et al. (1996) Shaffer et al. (1999) Harrison et al. (1996)

Social self-efficacy (+) Self-monitoring (+) Locus of control (−) Ethnocentrism (−) Positive affectivity (+) Cultural flexibility (+) Willingness to communicate (+) Psychoticism (−) c. Knowledge/skills/abilities/others Language ability (+)

Relational skills (+)

Responsible problem solving (+)

Black (1990) Florkowski and Fogel (1999) Kraimer et al. (2001) Black (1990b) Black (1990b)

Florkowski and Fogel (1999) Kraimer et al. (2001) Black (1990b) Black (1990b) Takeuchi et al. (2002b) Ditchburn (1996)

Black (1990b) Kraimer et al. (2001) Black (1990b) Black (1990b) Takeuchi et al. (2002b)

Caligiuri (2000b) Florkowski and Fogel (1999) Kraimer et al. (2001) Shaffer et al. (1999) Takeuchi et al. (2002b) Black (1988) Black (1990a) Black (1990b) Black and Gregersen (1991a) Caligiuri (2000b) Gregersen and Black (1992) Kealey (1989) Selmer (1999a) Selmer (2001b) Selmer (1999a) Selmer (2001a) Selmer (2001b) Selmer (2001b)

Nicholson and Imaizumi (1993) Shaffer et al. (1999) Takeuchi et al. (2002b)

Florkowski and Fogel (1999) Kraimer et al. (2001) Takeuchi et al. (2002b)

Black (1990a) Black (1990b) Black and Gregersen (1991a) Nicholson and Imaizumi (1993) Selmer (1999a)

Black (1990b) Black and Gregersen (1991a)

Selmer (2001a)

Selmer (1999a) Selmer (2001a)

Ditchburn (1996)

Selmer (1999a)

DAVID A. HARRISON ET AL.

Showing tolerance and patience (+)

Work Adjustment

Expatriate refuge (−) Predeparture knowledge (+) Conflict resolution (+) Met expectations (+) Expatriate career intentions (+) Meeting career goals (+) 2. Workplace factors a. Job features Role clarity (+)

Role discretion (+)

Selmer (1999a) Selmer (2001a) Selmer (2001a)

Selmer (1999a)

Black (1988) Takeuchi et al. (2002b) Black (1990b) Caligiuri et al. (2001) Selmer (1998) Selmer (1999b)

Takeuchi et al. (2002b)

Selmer (2001a)

Black (1990b)

Selmer (1999) Selmer (2001a) Black (1988) Takeuchi et al. (2002b) Black (1990b)

Selmer (1998) Selmer (1999b)

Selmer (1998) Selmer (1999b)

Black (1990) Florkowski and Fogel (1999) Gregersen and Black (1992) Lance and Richardson (1988) Shaffer et al. (1999) Takeuchi et al. (2002b)

Black (1990) Shaffer et al. (1999)

Gregersen and Black (1992) Shaffer et al. (1999)

Shaffer et al. (1999)

Aryee and Stone (1996) Black (1988) Black (1990) Black and Gregersen (1991) Florkowski and Fogel (1999) Lance and Richardson (1988) Li (1995) Shaffer et al. (1999) Taylor and Napier (1996) Takeuchi et al. (2002b) Anderz´en and Arnetz (1999) Aryee and Stone (1996) Black (1988) Black and Gregersen (1991) Shaffer et al. (1999) Taylor and Napier (1996)

Expatriate Experiences

Parent county escapism (−)

217

218

Table 2. (Continued ) Correlates

Interaction Adjustment

Work Adjustment

Role conflict (−)

Black (1990a) Gregersen and Black (1992) Lance and Richardson (1988) Shaffer et al. (1999)

Black (1990a) Shaffer et al. (1999)

Role novelty (−)

Kraimer et al. (2001) Stroh et al. (1994)

Role overload (−) Job demands (−) Scheduling clarity (+) Performance criteria clarity (+) Participation (+) Skill variety (+)

Lance and Richardson (1988)

Aryee and Stone (1996) Black (1990a) Black and Gregersen (1991a) Lance and Richardson (1988) Shaffer et al. (1999) Taylor and Napier (1996) Black (1990a) Kraimer et al. (2001) Nicholson and Imaizumi (1993) Lance and Richardson (1988) Anderz´en and Arnetz (1999) Aryee and Stone (1996) Aryee and Stone (1996) Li (1995) Li (1995) Aryee and Stone (1996) Shaffer et al. (1999) Aryee and Stone (1996) Shaffer et al. (1999) Black (1990a) Black and Gregersen (1991a) Kraimer et al. (2001) Anderz´en and Arnetz (1999)

b. Social context Supervisor social support (+)

Shaffer et al. (1999)

Shaffer et al. (1999)

Co-worker social support (+)

Shaffer et al. (1999)

Shaffer et al. (1999)

Home national social support (+)

Black (1990a)

Black (1990a)

LMX (+) Psychosocial work climate (+)

Kraimer et al. (2001)

c. Organizational context Logistical support (+) Realistic assignment preview (+)

Kraimer et al. (2001) Shaffer et al. (1999) Florkowski and Fogel (1999)

Shaffer et al. (1999) Kraimer et al. (2001)

Kraimer et al. (2001) Florkowski and Fogel (1999)

DAVID A. HARRISON ET AL.

Cultural Adjustment

3. Family/friends factors Spouse adjustment (+)

Spouse satisfaction (+) Accompanying spouse (+) Expatriate spouse (+) Family responsibility (+) Time spent with expatriates (−) 4. Environmental factors Culture novelty (−)

Parker and McEvoy (1993) Selmer (2000) Black (1988) Black and Stephens (1989) Black and Gregersen (1991a) Caligiuri (2000b) Nicholson and Imaizumi (1993) Shaffer and Harrison (1998) Shaffer et al. (1999) Stroh et al. (1994) Takeuchi et al. (2002b) Shaffer and Harrison (1998) Selmer et al. (2000)

Black and Stephens (1989) Black and Gregersen (1991a) Nicholson and Imaizumi (1993) Shaffer and Harrison (1998) Shaffer et al. (1999)

Parker and McEvoy (1993)

Parker and McEvoy (1993)

Black (1990a) Black and Gregersen (1991a) (+) Kraimer et al. (2001) Parker and McEvoy (1993) (+)

Black (1990a) Black and Gregersen (1991a) Kraimer et al. (2001) Parker and McEvoy (1993)

Expatriate Experiences

Compensation (+) Corporate career development (+)

Black and Stephens (1989) Black and Gregersen (1991a) Caligiuri et al. (1998) Nicholson and Imaizumi (1993) Shaffer and Harrison (1998) Takeuchi et al. (2002b)

Shaffer and Harrison (1998) Selmer et al. (2000) Selmer et al. (2000) Shaffer and Harrison (1998)

Black (1990a) Takeuchi et al. (2002b)

219

220

Table 2. (Continued ) Correlates

Standard of living (+) Living conditions (+) Strains 1. Satisfaction Job satisfaction (+)

Non-work satisfaction (+)

2. Withdrawal cognitions (−)

Cultural Adjustment

Interaction Adjustment

Shaffer and Harrison (1998) Shaffer et al. (1999) Selmer (2002) Florkowski and Fogel (1999) Shaffer and Harrison (1998)

Shaffer and Harrison (1998) Shaffer et al. (1999) Takeuchi et al. (2002b)

Li (1995) Shaffer and Harrison (1998) Stroh et al. (1994) Takeuchi et al. (2002a) Lance and Richardson (1988) Shaffer and Harrison (1998) Takeuchi et al. (2002a) Black and Stephens (1989) Caligiuri (1997) Caligiuri and Tung (1999) Florkowski and Fogel (1999) Gregersen and Black (1990) Shaffer and Harrison (1998) Takeuchi et al. (2002a) Takeuchi et al. (2002b)

Shaffer and Harrison (1998)

Aryee and Stone (1996) Shaffer and Harrison (1998) Takeuchi et al. (2002a)

Shaffer and Harrison (1998)

Black and Stephens (1989) Gregersen and Black (1990) Shaffer and Harrison (1998) Takeuchi et al. (2002b)

Lance and Richardson (1988) Shaffer and Harrison (1998) Takeuchi et al. (2002a) Florkowski and Fogel (1999) Shaffer and Harrison (1998) Takeuchi et al. (2002a) Takeuchi et al. (2002b)

Kraimer et al. (2001)

Kraimer et al. (2001) Parker and McEvoy (1993) Kraimer et al. (2001)

Florkowski and Fogel (1999) Shaffer and Harrison (1998)

Overall performance (+)

Caligiuri (1997) Kraimer et al. (2001) Sinangil and Ones (1997) Caligiuri (1997) Sinangil and Ones (1997)

Nicholson and Imaizumi (1993)

DAVID A. HARRISON ET AL.

3. Performance Task performance (+) Relationship performance (+)

Work Adjustment

Expatriate Experiences

221

What are the Antecedents of Adjustment? Most efforts to understand expatriate adjustment have been devoted to identifying inputs to adjustment. Reflecting the atheoretical nature of much of this literature (Mendenhall et al., 2002), a wide array of predictors have been investigated. Except for variables comprising the Black et al. (1991) theory, however, most have been tested in only one or two studies, thus precluding any definitive conclusions about their explanatory contributions. Integrating the classification schemes offered by the three main adjustment theories (Aycan, 1997; Black et al., 1991; Parker & McEvoy, 1993), we organized this kaleidoscope of predictors by categorizing them into four major factors: personal (demographic characteristics, personality traits, and KSAOs, including training to improve them), work (job, social context of work, and organizational features), family/friends, and environmental factors. Below we highlight and discuss some of the more persistent arguments and findings to answer two main questions: what demands exacerbate the stress of living and working in a different culture, and what resources ease the uncertainties associated with the international assignment? Personal Factors Perhaps because of their relevance to selection and training, personal factors have been scrutinized extensively. According to the Black et al. (1991) model, one set of personal inputs to adjustment is anticipatory. The basic premise is that accurate predeparture expectations and appropriate preparations for the upcoming assignment will facilitate actual adjustment in the new international setting. These anticipatory factors include the expatriate’s language ability (fluency in the host country language), previous overseas assignments (prior experience in living and working abroad), and predeparture knowledge about the host country. Host country language fluency. Knowledge of the host language is widely touted as a contributor to all aspects of an expatriate’s adjustment (Mendenhall & Oddou, 1985). The greater the proficiency in host country language, the easier it is to obtain necessary information pertaining to everyday work and non-work activities (Nicholson & Imaizumi, 1993). Yet, despite the general consensus that host country language skills should be a necessary prerequisite for international assignments (Jordan & Cartwright, 1998), empirical findings are mixed, with correlations that range from r = 0.03 for work adjustment (Shaffer et al., 1999) to r = 0.63 for interaction adjustment (Takeuchi, Yun & Russell, 2002a). That is, consistent significant effects have only been found for interaction adjustment (Kraimer et al., 2001; Shaffer et al., 1999; Takeuchi et al., 2002b), with contradictory results

222

DAVID A. HARRISON ET AL.

for the other two adjustment dimensions. Why such a puzzling discrepancy between what might seem like an obvious prediction, and the growing body of evidence? Perhaps it reflects the preponderance of English-speaking samples, for whom everyone else might be making linguistic accommodations at work rather than vice versa. Another possibility is that the self-reports almost universally collected as language skill measures tend to be less valid than something more objective. Still, one would expect self-reports of language skill to generate inflated relationships with other self-reported variables such as adjustment. Previous overseas experience. According to Black et al. (1991), previous international experience also facilitates all dimensions of adjustment. The authors suggest that seasoned veterans find adjustment easier as they are able to anticipate problems (e.g. they are more aware of what they will not know) more accurately, given their accumulated knowledge from past assignments on living and working overseas (see also Bell & Harrison, 1996, who make this an individual difference argument about biculturalism). Again, despite the intuitive appeal of this idea, research evidence has been surprisingly inconsistent. Some researchers (Parker & McEvoy, 1993) have found positive effects, while others have found null or even negative effects (Nicholson & Imaizumi, 1993). One reason may be that current operationalizations of previous experience (i.e. length of time spent overseas or number of past assignments) do not capture the quality or density of the previous experience. It may be useful to track various individual and context based dimensions of international experience that provide additional insights into the nature and relevance of prior experiences. Other KSAOs. Concerted efforts have been targeted at understanding the effects of knowledge, skills and abilities on expatriate adjustment. In particular, efforts have focused on relational skills. Such skills refer to a repertoire of tools and techniques that facilitate the formation of interpersonal ties; through those ties, expatriates obtain necessary information and behavior-relevant feedback in the host culture (Black et al., 1991). Black (1990a) and Black and Gregersen (1991c) detected positive relationships between relational skills and all three forms of adjustment; other researchers (Caligiuri, 2000a; Nicholson & Imaizumi, 1993) report significant relationships for only one or two dimensions (i.e. culture and/or interaction adjustment). In attempts to understand how expatriates cope with the stresses associated with living and working in a foreign culture, Selmer (1999a, 2001) identified several coping strategies: showing tolerance and patience, responsible problem-solving, parent country escapism, and expatriate refuge. Except for work by Selmer (1998a, 1999b), little has been done in the area of personal career issues and how these affect expatriate adjustment.

Expatriate Experiences

223

Relative to some of the other personal factors, specific predeparture knowledge of the host country seems to be more important than general experience in having worked abroad. Those who learned about the host environment before departing for the international assignment form more accurate expectations and adjust better (Black, 1988; Florkowski & Fogel, 1999; Takeuchi et al., 2002a). Why might these anticipatory adjustment features be helpful? Through the acquisition of knowledge (i.e. language skills or information about the host country) or through the experience of living in another international environment, expatriates develop competencies that enable them to engage in cross-cultural interactions more effectively. According to social learning theory, such predeparture preparation equips expatriates with a wider repertoire of cognitive and behavioral responses for dealing with the uncertainties and ambiguities that often arise in a foreign setting (Black & Mendenhall, 1990). Training. Given the assumed learning mechanisms involved, cross-cultural training has been hailed by many as being a anticipatory remedy for adjustment and performance problems (Black & Mendenhall, 1990; Kealey & Protheroe, 1996). Although the linkage between cross-cultural training and expatriate effectiveness seems intuitively obvious, the empirical literature is not entirely conclusive. Early studies were contradictory, with some researchers reporting negative or null effects for cross-cultural training on expatriate adjustment and performance (Black & Gregersen, 1991a) and others reporting strong positive relationships (Earley, 1994). Recent meta-analyses by Deshpande and Viswesvaran (1992) and Morris and Robie (2001) have tried to clarify these discrepancies. Their findings have been generally supportive of cross-cultural training (meta-analytic ␳’s of 0.41 and 0.26, respectively), but significant variations in original effect sizes suggests that training may not always be beneficial. In designing training programs, therefore, consideration should be given to the needs of the expatriates as well as the methods, timing and content of training (Morris & Robie, 2001). Various training methods such as environmental briefing, cultural orientation, culture assimilator, language training, sensitivity training, case method, and field experience, have been used to prepare expatriates for overseas assignments (Tung, 1982); however they vary in content and goals. For example, sensitivity training might be more appropriate to develop attitudinal flexibility, while cultural assimilators might be preferable to provide expatriates with experience in intercultural behavioral encounters. More fine-grained data that capture and compare the effectiveness of different methods and content of training are needed. The timing of the training is also an important issue. Although most research has focused on predeparture training, there is some evidence that on-site training provided by the host organization may be just as, if not more important,

224

DAVID A. HARRISON ET AL.

than predeparture training (Suutari & Burch, 2001), suggested in our opening quote. Demographic characteristics. In addition to anticipatory adjustment factors, various individual factors have been identified by expatriate adjustment theorists (Parker & McEvoy, 1993) as personal requirements for effectiveness in the overseas environment. From a human capital perspective, these personal factors besides KSAs (e.g. demographics and traits) are resources that might help the expatriate deal more effectively with the stresses of living and working in a foreign culture. For example, those in higher-level managerial positions generally have more autonomy and control, both of which are important in reducing stress (Karasek, 1979). Of the demographic characteristics (see Table 2), the general profile of a well-adjusted expatriate is one who is well-educated (Shaffer & Harrison, 1998), married (Caligiuri, Joshi & Lazarova, 1999; Florkowski & Fogel, 1999), is in a higher level management position (Caligiuri et al., 1999; Shaffer et al., 1999), and who has received a promotion along with the assignment (Parker & McEvoy, 1993; Shaffer & Harrison, 1998). For other demographic characteristics, relationships with adjustment are questionable. Although Selmer, Ebrahimi and Li (2000) and Shaffer and Harrison (1998) reported significant effects of age and tenure, respectively, others have reported null effects (Florkowski & Fogel, 1999; Parker & McEvoy, 1993). With respect to gender, research on psychological stress shows that environmental demands tend to have a more damaging impact on women than men (Davidson & Cooper, 1983). However, the reported effects of gender on expatriate adjustment are contradictory. In a study by Florkowski and Fogel (1999), female expatriates reported higher levels of both cultural and work adjustment. Shaffer and Harrison (1998) found that females were better adjusted to the general culture while males were better adjusted to work. With females accounting for only about 10% of the samples studied, however, these findings are tenuous. Findings regarding nationality are also conflicting. Selmer (1998b) reported that North Americans were more adapted to interacting with host nationals than Western Europeans, especially the French. However, U.S. respondents in Parker and McEvoy’s (1993) study were less adjusted to the culture than non-US respondents. The relationship between nationality and adjustment may have more to do with differences between the home and host country than with anything inherent about a particular expatriate nationality. Although some of the demographic characteristics are of theoretical and practical importance (e.g. managerial vs. technical positions and previous international experience), others are less so. In particular, at least in the U.S., employment laws prohibit firms from making employment decisions on the basis of gender, age, marital status and nationality. Consequently, while these characteristics are

Expatriate Experiences

225

potentially useful as control variables in research, there is not much to be gained by pursuing their effects on expatriate adjustment. A more interesting question, perhaps, has to do with the remaining personal features: Are there general traits associated with expatriate adjustment? Since the initial surge of research in the early 1960s surrounding the selection of Peace Corps volunteers (Mischel, 1965), several exhaustive lists of “essential” individual characteristics have appeared in the literature (Arthur & Bennett, 1995; Ones & Viswesvaran, 1999). Personality traits. The underlying rationale for the importance of some personality traits is that they provide mechanisms through which expatriates gain familiarity with what is acceptable or unacceptable in the host culture. Using a three-dimensional typology of expatriate acculturation developed by Mendenhall and Oddou (1985), Leiba-O’Sullivan (1999) classified various individual differences in terms of self-, others-, and perceptual-orientations. According to Mendenhall and Oddou (1985, 1986), the self-orientation necessary to be effective as an expatriate includes adaptive activities and qualities such as reinforcement substitution, stress reduction, and technical competence. These qualities enhance self-preservation, self-enjoyment, and mental hygiene. The others-orientation includes skills and attributes that assist in the development and maintenance of relationships and effective communication with host-country nationals (HCNs). The perceptual-orientation refers to expatriates’ abilities to accurately see and understand the behaviors of HCNs. Expatriates strong on this dimension tend to make more correct attributions and be less judgmental. Identification of “essential” individual characteristics and development of an ideal prototype has been hampered by a lack of compelling theory and few truly predictive (rather than concurrent) studies. With respect to personality traits, the limited evidence that does exist indicates that work self-efficacy (Harrison, Chadwick & Scales, 1996), positive affectivity (Kraimer et al., 2001), cultural flexibility (Black, 1990a), and willingness to communicate (Black, 1990a) are important inputs to all three forms of adjustment. Other traits have been linked with just one or two adjustment dimensions. For ethnocentrism, Black (1990a) reported a significant correlation of r = −0.24 with interaction adjustment. Florkowski and Fogel (1999) failed to replicate this, but did find that it had an important influence on cultural (r = −0.16) and work adjustment (−0.29). Interestingly, no one has examined the influence of all of the Big Five personality traits on expatriate adjustment. However, a few researchers have looked at these effects on expatriate performance and the results remain equivocal (Caligiuri, 2000a; Dalton & Wilson, 2000; Deller, 1997). Only one researcher has examined the effects of the Big Five on the criterion of withdrawal cognitions or retention (Caligiuri, 2000b). Initial results suggest that, as in domestic studies, conscientious

226

DAVID A. HARRISON ET AL.

is a positive predictor, but the most likely Big Five surrogate for flexibility – openness to experience – has not shown dependable prediction of expatriate criteria. Workplace Factors Expatriate assignments typically involve new operations, means of completing tasks, and increased responsibilities. They are often marred by policy and procedural conflicts with the parent company (Gregersen & Black, 1992). Under these circumstances, aspects of assignees’ jobs, the social context in which they work, and organizational policies and practices will likely influence the ability of the expatriate to adjust to the new environment. Because uncertainty arising from all of these variables is focused on the job environment, these inputs have been expected to affect work adjustment only (Black et al., 1991). However, evidence indicates that their influence can be cross-domain, readily “spilling over” to cultural and interaction adjustment as well (Kraimer et al., 2001; Lance & Richardson, 1988; Shaffer et al., 1999). The major job stressors of role clarity (unambiguous conveyance and understanding of assignment requirements), role discretion (decision-making autonomy), role conflict (incompatible cues regarding job expectations), and role novelty (differences between host and native country work roles) have been investigated extensively and evidence for their influence on work adjustment is fairly strong (the weakest evidence is for role novelty; Aryee & Stone, 1996; Black, 1988; Black & Gregersen, 1991c; Kraimer et al., 2001; Shaffer et al., 1999; Taylor & Napier, 1996). As maladjustment might reasonably be thought of as lingering uncertainty for an expatriate, the strong impact of these role stressors might not be surprising. They also represent features of the job that international HR managers have specific options for improving. For example, by designing global positions with sharply stated tasks, goals, and duties, providing realistic assignment previews and so on, expatriate adjustment can be enhanced. Variables emanating from the social context can also play important roles in easing adjustment (Guzzo, Noonan & Elron, 1994). By providing expatriates with information about culturally appropriate norms and behaviors in the work context (Cohen & Wills, 1985; Pinder & Schroeder, 1987), social support from co-workers and supervisors serves to reduce the uncertainty emanating from the expatriate’s new circumstances. Support from supervisors, co-workers and home nationals has been associated with all three dimensions of adjustment (Black, 1990b; Shaffer et al., 1999). Other social contextual variables, such as attitudes of host country nationals (Caligiuri et al., 1999), the psychosocial work climate (Anderz´en & Arnetz, 1999), and the quality of supervisor-subordinate relationships (Kraimer et al., 2001) have been shown to contribute to improved cultural and work adjustment as well. Given the importance of interpersonal relationships to the expatriate

Expatriate Experiences

227

experience, however, research in this area is lacking. In particular, little effort has been made to understand the dynamics of relationships and communication between expatriates and host country nationals (Shaw, 1990). One of the best opportunities for future investigations might be to study the social ties and informal network structures of assignees and their HCN colleagues (Au & Fukuda, 2002). Expatriates who are on the periphery of their work unit’s social network are less likely to receive the local job knowledge and affective support necessary to adjust and perform well, especially in terms of developing a set of alliances, contacts, or guanxi in their foreign markets. Relative to job features and the social context, researchers have given organizational level variables very little attention. Logistical support (i.e. assistance from the expatriate’s home office with respect to day-to-day aspects of living such as financial support, housing assistance and so on) is especially important because it assists adjustment by making critical resources available to the expatriate at times of necessity. Because it deals with issues outside of the work domain, it has been found to affect all forms of adjustment (Kraimer et al., 2001; Shaffer et al., 1999). However, research to date has not clarified which organizational benefits (e.g. financial, family-related benefits) are most vital. Except for single-study considerations of realistic previews of the assignment (Florkowski & Fogel, 1999), compensation (Parker & McEvoy, 1993), and career development programs (Selmer, 2000), other firm-level practices (e.g. career planning, mentoring, performance appraisal) have been ignored by researchers. Family/Friends Factors For married expatriates, another potential source of uncertainty is the spouse and family. If the spouse and children are unable to adapt to the overseas environment, their maladjustment may spread to affect the expatriate’s own level of work and nonwork adjustment (Staines, 1980). In a study by Arthur and Bennett (1995), expatriates rated family issues as the most important contributors to successful international assignments. The currently limited empirical evidence is strong in supporting the key role of family in facilitating all forms of adjustment (Nicholson & Imaizumi, 1993; Shaffer & Harrison, 1998; Stroh, Dennis & Cramer, 1994). Married expatriates whose spouses were also expatriates and who were on assignment with them adapted better to work and had a greater sense of psychological comfort (Selmer et al., 2000). Those with increased family responsibilities also adjusted better to work (Shaffer & Harrison, 1998). Given the general consensus in the literature that the spouse and family are central to expatriate effectiveness, it is surprising that so little attention has been given to this domain. Only a handful of studies have investigated factors that influence spouse adjustment (Black & Gregersen, 1991b; Shaffer & Harrison,

228

DAVID A. HARRISON ET AL.

2001), and even fewer have considered the effects of family on other attitudes and behaviors such as retention decisions (Shaffer & Harrison, 1998). Insofar as expatriates are especially likely to experience stress as a result of excessive demands placed upon them in both their work and family roles, it is also important to understand the interface between these two domains (Shaffer, Harrison, Gilley & Luk, 2001). Although other non-work relationships may play a less influential role in the adjustment process, research in this area once again suggests that informal social networks warrant further study. For example, Parker and McEvoy (1993) found that expatriates who spent more time with other expatriates were less well-adjusted to the culture and to interacting with host nationals. Environmental Factors A unique feature of the transition process for expatriates is that it requires adjustment to a foreign and perhaps dramatically different culture. Among the many antecedents or stressors that have been investigated, cultural novelty or the distance between native and host cultures, has consistently emerged as a potent influence on the stress of expatriate adjustment (Black, 1990a; Shaffer et al., 1999). Larger perceived discrepancies between host and native cultures create greater uncertainties and more severe adjustment problems as the requisite changes in thought processes and behavior patterns for the expatriate are more acute (Mendenhall & Oddou, 1985). Although most researchers report a negative relationship between culture novelty and adjustment, as predicted (Black, 1990a), in some studies a significant positive relationship with cultural adjustment has been found (Black & Gregersen, 1991c; Parker & McEvoy, 1993). These contradictory findings suggest a more complex consideration of culture novelty (and time; especially if these studies captured expatriates during different adjustment phases along the purported U-curve). One possibility is that the relationship between culture novelty and adjustment is non-linear. According to the psychic distance paradox (O’Grady & Lane, 1996) adjustment to a relatively similar culture may be as difficult as adjustment to a distant culture because differences are not anticipated (Boyacigiller, Kleinberg, Philips & Sackmann, 1996). More specific features of the foreign environment, such as standard of living and comparability of living conditions (relative to those in the home country), have also been reliably associated with better expatriate adjustment (Florkowski & Fogel, 1999; Shaffer & Harrison, 1998). In addition, the fit between host and home cultures, in terms of living standards, cultural values, and religious beliefs, may also contribute to the stresses of being an expatriate (Harrison, Shaffer & Bhaskar, 2002). Thus, it may be important to gauge the direction of differences between countries of origin and assignment. A move from an economically less advantaged to a more advantaged country (excess), might create fewer adjustment problems

Expatriate Experiences

229

than the reverse move (deficiency). The central assumption is that a higher standard of living in one’s host vs. native country would provide the expatriate with greater resources to deal with, and therefore be less sensitive to, assignment stressors. On the other hand, under conditions of deficiency (fewer resources than one is accustomed to), adjustment will be more sensitive to proposed stressors.

STRAINS OF INTERNATIONAL ASSIGNMENT EXPERIENCES As we have noted above, the main surge of research on expatriate experiences, which began in the late 1980s has focused on the antecedents and nature of adjustment. And, the findings we have reviewed make it clear that expatriate adjustment serves as a mainstay, if not the lynchpin, of the international assignment literature. To date, most of the theoretical frameworks in international HR and OB have attempted to explain it (e.g. Parker & McEvoy, 1993). Widely used measures have been developed to assess it (Black & Stephens, 1989). Training programs have been designed to improve it (e.g. Morris & Robie, 2001).

What are the Consequences of Adjustment? This convergence of scholarly effort begs an important question. Has the focus on adjustment been worth all the research investment? To provide a partial answer, we now turn to a fundamental assumption underlying this body of work: that adjustment has consequences which are both ubiquitous and potentially critical for expatriates, their families, and the subsidiaries to which they are assigned (Black, 1988). We evaluate that assumption below. Specifically, we examine arguments and evidence for attitudinal (job and non-work satisfaction) and conative (withdrawal, performance) outcomes of adjustment among expatriates – what might be termed “strains” in the dominant stressor-stress-strain paradigm for this research area (Harrison et al., 2002). Given the well-established bottom line implications of these outcomes in domestic HR contexts (Blumberg & Pringle, 1982; Griffeth, Hom & Gaertner, 2000; Judge, Thoresen, Bono & Patton, 2001), their practical utility for international HR contexts is also likely to be high. In addition, we review the (lack of) evidence for more distal but no less important outcomes in the home and workplace of the expatriate. As we note in the introduction, the logical cornerstone of research in this area has been the idea that maladjustment – the experience of distress – prompts negative emotions or dissatisfaction for the expatriate. Both in tandem with these

230

DAVID A. HARRISON ET AL.

emotions, and in indirect effects through them, maladjustment is presumed to lead to an expatriate’s psychological disengagement from his or her assignment. Such disengagement includes withdrawal cognitions or explicit turnover decisions that eventually result in costly early returns (Sanchez, Spector & Cooper, 2000). Negative affect and thoughts of quitting also have implications for holding back other facets of expatriate effort and performance (Harrison & Shaffer, 2001). Attitudinal Consequences Job satisfaction. Perhaps the closest consequence of expatriate adjustment, if it can truly be thought of as a consequence (rather than as a covariate, or something sharing the same definitional space; see below), is job satisfaction. Following classic definitions, we consider it to be the overall evaluation of, or the “emotional state resulting from, the appraisal of one’s job or job experiences” (Locke, 1976). As it is obviously and primarily task-related, job satisfaction is presumed to arise from successful adaptation to overseas work role requirements and from effective development and maintenance of interpersonal relationships with the host country’s workforce and customers (Shaffer & Harrison, 1998). That is, work and interaction adjustment should foster job satisfaction. The data clearly affirm this connection. Feldman and Thomas (1992), Stroh, Dennis and Cramer (1994), Aryee and Stone (1996), and others have all reported positive and fairly robust correlations (r’s range from 0.3 to 0.5) between job satisfaction and work adjustment on international assignment. There is also a clear pattern of content- or domain-matching among the studies that have examined multiple adjustment dimensions (e.g. Shaffer & Harrison, 1998). As one might expect, job satisfaction has the strongest relationship with work adjustment, followed by interaction adjustment, and a weak (or even null) relationship with cultural adjustment. Given the widespread and far-reaching behavioral consequences of job (dis)satisfaction: lateness, absence, turnover, quality and quantity of task performance, and organizational citizenship behavior or contextual performance (e.g. Newman, Harrison & Roth, 2003; Van Scotter & Motowidlo, 1996), if one dimension or another of adjustment is indeed a causal substrate of satisfaction, it is fundamentally important for international firms to facilitate it among their expatriate employees. Non-work satisfaction. An equally consistent but switched pattern is apparent for non-work satisfaction, which might be regarded as an expatriate’s overall evaluation of experiences outside his or her job. We include it here, and researchers have included it in their investigations of expatriate experiences because it is tied so strongly to the “bundle” of strains that might emanate from an overseas move. It is also paralleled by the breadth and potency of adjustment antecedents that originate

Expatriate Experiences

231

from the work environment. That is, a focus only on work outcomes would seem to miss the main, multifaceted theme of the adjustment approach to expatriate experiences. Lance and Richardson (1988), Shaffer and Harrison (1998), and Takeuchi et al. (2002a), among others, have found sizeable correlations between general (cultural) adjustment and non-work satisfaction, ranging from r = 0.40 to r = 0.65, and then declining from this level when the predictor is interaction or work adjustment. The size of these correlations far exceeds what might be expected solely from common methods variance (Harrison, McLaughlin & Coalter, 1996). Hence, there is little doubt that these similarly framed attitudinal constructs are substantively linked to the stress of adapting to an overseas assignment. Causal sequence. Despite the supportive evidence for substantive links between adjustment and attitudes, an open conceptual question remains. Are attitudes really a consequence, or are they merely another window on the adjustment construct itself? One way to answer this question about construct redundancy is to ask two additional questions. First, can an expatriate who is adjusting poorly to work, to interactions with HCN’s, or to the general host country culture, still be highly satisfied with each of those corresponding domains? Not likely. In fact, as we mentioned in the first section of this article, some treatments define adjustment in terms of satisfaction or affect. For example, Torbiorn (1982, p. 77) views adjustment as an “individual’s general satisfaction with his/her personal situation in the host country.” Selmer (1999b, p. 43) offers a similar take, defining adjustment as: a “feeling of happiness and an ability to enjoy one’s day-to-day activities and face up to one’s problems.” Although we have settled on a more constrained definition of adjustment, as stress or psychological (dis)comfort, it is difficult to envision a highly discomforting circumstance in which an assignee can still be very satisfied or feel extremely positive affect. Second, can an expatriate who is well-adjusted be highly dissatisfied? Perhaps. And this is where the distinction between adjustment and attitudes begins to have some theoretical leverage. Expatriates who are adjusting well, but who believe they are severely underpaid relative to their counterparts, would probably exhibit negative job attitudes. A health condition or housing problem that was equally as likely in the home country as the host country could still generate negative non-work attitudes while the expatriate is on assignment. Hence, psychological adjustment might be regarded as a sufficient, but not a necessary condition for job or nonwork (dis)satisfaction. The previously specified domain-matching correlations support this viewpoint. Although they are large, they are not so large as to be regarded as alternative measures of the same psychological state, especially given that the supporting data come from the same, self-report source. More support

232

DAVID A. HARRISON ET AL.

for their uniqueness comes from Shaffer and Harrison (1998), who showed that attitudes have unique predictive validity relative to adjustment. In a large sample of expatriates stationed in over 40 host nations, they demonstrated that job satisfaction (alongside organizational commitment) and non-work satisfaction fully mediated the effects of adjustment on expatriate return decisions to return home early from their assignments. These arguments and patterns of evidence prompt one final question. If adjustment and attitudes are distinct, are they separable enough features of expatriate experiences to merit continued study? We would answer with a qualified yes. It would not be especially useful to continue studying the adjustment – attitude relationship per se, but it would be useful to investigate how adjustment plays out, differentially, into some of the outcomes reviewed next. Conative Consequences Turnover behavior. With its visibility, (presumed) cost, and relative ease of study via routinely kept organizational records, thousands of studies have investigated employee turnover or turnover decisions. Researchers in those studies have tested hundreds of different hypotheses, many stemming from influential HR theories about the variables implicated in the process of quitting one’s job (e.g. Griffeth, Hom & Gaertner, 2000; Mobley, 1982). Visibility and cost to parent firms are perhaps even higher and more salient for international assignees than domestic employees (Tung, 1998). Yet, we are unaware of a single published study that has tracked the predictors of observable “quitting” behavior among expatriates. Perhaps it is because an early return from an international posting, although thought of as a failed assignment, is not truly turnover: permanent separation of an employee from his or her membership in an organization (Harrison, 2002). Instead, it is akin to host-home transfer: movement across unit boundaries, or leaving one’s job but staying within the same organization (Dalton & Todor, 1993). Archival measures of what would be regarded as “early” returns from international assignments are almost impossible to obtain, because: (a) they do not involve a mandatory (payroll) record of an employee separating from the firm; (b) expatriates are often sent overseas to complete a particular task rather than being sent for a fixed duration, making it difficult to determine if someone has returned “prematurely”; and (c) companies are often unwilling to label an assignment an early return because of the personal stigma attached to it. The upshot of these data restriction and definition problems is that, in spite of dramatic expense estimates (Shannonhouse, 1996), HR tracking and control systems for international employees appear to be generally underdeveloped. Or, if they are developed, the defining features of expatriate assignments (length of time, goals to be reached that signify assignment completion) are

Expatriate Experiences

233

not routinely or clearly shared with HR managers to serve as data for such systems. Lacking those data, expatriate researchers have turned to survey measures. The measures form a set of highly similar constructs variously referred to as turnover tendency (Bhuian & Al-Jabri, 1996; Birdseye & Hill, 1995), propensity to leave (Naumann, Widmier & Jackson, 2000), and early return intentions (Guzzo et al., 1994). In most cases, these variables have been offered as substitutes for turnover itself, although there is little evidence of a necessarily strong correspondence. Dalton, Johnson and Dailey (1999) recommend a limited set of substitution conditions, one of which is a correlation of r = 0.50 between intention and behavior; no such correlations have been forthcoming from research on expatriate experiences. Withdrawal cognitions. Taking another approach, Hom, Griffeth and Sellaro (1984) gathered the above set of turnover inclination variables together in the domestic HR literature and termed them withdrawal cognitions. The deliberate re-direction of thought processes and personal plans away from one’s current position, or the cognitive effort spent searching, evaluating alternatives, and preparing to leave, is also what Shaffer and Harrison (1998, p. 88) referred to as “psychological withdrawal.” The latter authors try to bolster the case for studying such a form of withdrawal by arguing it indicates a broader, more encompassing reduction of inputs by the expatriate from his or her assignment, because the sunk costs and exit barriers associated with actually leaving an international assignment are much steeper than in domestic jobs: “Many expatriates . . . would like to quit their assignments, but ‘hang in there’ to collect a bonus or because they fear a ‘political black eye’ that would jeopardize their careers.” Black and Gregersen (1991a) estimate that about half of these expatriates are ineffective; their assignments are costly failures for the firm in that subsidiary performance is likely to suffer. Unfortunately, these individual-level and cross-level (individual to unit) arguments have yet to be subjected to systematic empirical testing (cf. Harrison & Shaffer, 2001), as is true with much of the conjecture surrounding the criterion space for expatriates. Still, a number of studies have estimated effects of the three widely used dimensions of adjustment, as well as effects of overall adjustment, on some instantiation of withdrawal cognitions (e.g. Caligiuri & Tung, 1999; Florkowski & Fogel, 1999). Those effects are consistently negative, with rs sometimes approaching −0.40, despite an often restricted range of the criterion. The relationships appear not to be inflated by the use of common data sources. Correlations involving spouse or peer ratings of either adjustment or withdrawal cognitions yield the same range and size of values as those in which both variables are self-rated (Black & Gregersen, 1991b; Black & Stephens, 1989; Shaffer & Harrison, 1998; Takeuchi et al., 2002b). At the same time, adjustment-withdrawal relationships

234

DAVID A. HARRISON ET AL.

are less susceptible to criticisms of construct redundancy than those involving attitudes. Psychological neglect and/or quitting have a complex etiology outside adjustment problems, and they are only a subset of several possible responses to distress or negative affect (e.g. Rusbult, Farrell, Rogers & Mainous, 1988). More interestingly, relationships with withdrawal cognitions also seem to be of equal strength across adjustment dimensions. That is, contrary to what might be expected from the emphases of domestic research, work-related adjustment problems do not explain the lion’s share of variance in the psychological disengagement of expatriates, or in decisions to quit their assignments. Our own review of the evidence suggests that the explicitly nonwork dimension – cultural adjustment – is perhaps the strongest determinant of such disengagement and decisions. When an expatriate and his or her family leave an overseas assignment, they are leaving the host country culture and its members as much as they are leaving the international job itself. The possibility of a similar process operating in domestic turnover decisions has been largely overlooked. Even the most recent, ground-breaking turnover models (e.g. Lee & Mitchell, 1994) focus their conceptual force on work-related determinants that emanate from the current or available alternative jobs. Cultural (environmental) and interaction (social) adjustment variables that matter as expatriate’s reasons for returning home early, are relatively unexplored in other HR research, but might be potentially important reasons for quitting a domestic job. Performance The elemental belief driving much of the research devoted to expatriate experiences is that assignee performance suffers from maladjustment (Caligiuri, 1997). This belief is validated in the concepts of stressor-stress-strain that have been the root source of many expatriate adjustment models (Harrison et al., 2002). In the stress literature, theorists have argued, and data have supported the idea, that psychological stress can have pernicious effects on work performance. It induces mental fatigue or tiredness that deprives individuals of the energy needed to successfully perform their job duties (e.g. Cohen, 1980). Parallel arguments have been made about international assignments. For example, Shaffer et al. (2001) contend that well-adjusted expatriates will have a greater reserve of personal resources (time, effort, emotional investment) available to spend on the behaviors that facilitate their job performance. After reviewing various models of the criterion or performance space in international HR (e.g. Arthur & Bennett, 1997; Ones & Viswesvaran, 1997; Ronen, 1989; Tung, 1981), some common themes are clear. Fulfilling specific task requirements, as well as developing and maintaining relationships with host country nationals, are both purported to be core facets of expatriate performance.

Expatriate Experiences

235

These two facets are fairly consistent with the task/technical and contextual/interpersonal facilitation dimensions suggested by domestic researchers (e.g. Borman & Motowidlo, 1993), although the targets of the social relationships are clearly different for expatriates (who are unfamiliar not just with their cultural surroundings, but also with their work group members from that culture). In other words, task-based performance for expatriates refers to successful execution of duties, including attainment of specific goals or accomplishment of definable projects. Relationship-based performance is the effective development or maintenance of interpersonal connections with members of the host country workplace (Podsakoff, MacKenzie, Payne & Bacharach, 2000). These two dimensions have a long history of being distinguished in HR-related research on groups (task and interpersonal conflict), leadership (initiating structure and consideration, low vs. high LPC leaders), and stress (task stressors vs. social support buffers). Tests of this elemental assumption have come within the last several years. Although Feldman and Thomas (1992), Nicholson and Imaizumi (1993), and Parker and McEvoy (1993) reported fairly strong, positive connections between forms of adjustment and performance, the connections were based on concurrent, same-source data. More recent studies have also been concurrent, but they have employed multi-source designs involving performance ratings from co-workers or supervisors (Caligiuri, 1997; Harrison & Shaffer, 2001; Kraimer et al., 2001; Sinangil & Ones, 1997). Adjustment – performance connections in these studies have been consistently positive, but not especially potent. Correlations have ranged from r = 0 to 0.3. They don’t appear to be differentiated by task vs. relationship performance serving as an outcome. Expatriate adjustment likely contributes to (both major dimensions of) performance, but it is not an overriding influence.

SUMMARY What is the substantive essence of expatriate experiences? The past 25 years of theory and evidence has settled on (mal)adjustment or psychological stress as the principal construct. That construct can be dependably broken up into work, interaction (interpersonal), and cultural experiences, with the latter underscoring the clear difference between international assignments and domestic relocations. Although less settled, the literature has also tended to move toward a stressorstress-strain paradigm as describing the sequence of expatriate experiences; at the least it serves a useful heuristic for structuring adjustment mechanisms. What variables drive the stress of poor adjustment? Although a vast range of possible stressors or buffers has been studied, a few consistent findings emerge. Many can be quickly incorporated into international HR practices, and because

236

DAVID A. HARRISON ET AL.

of their regular connections with adjustment, many should be routinely included in future research. Predeparture training (including realistic previewing) is an obvious buffer of stress, but it could profitably be extended to the expatriate’s family (Shaffer & Harrison, 2001). Although understudied, family experiences (for assignees who bring family members along) serve as perhaps the most potent input to expatriate adjustment. Flexible relational skills and broad-mindedness also appear to be essential for diminishing distress, as are the potential role stressors identified in domestic research: heightened conflict and novelty and reduced clarity and discretion. Perhaps most surprisingly, language skills aid in interpersonal adjustment but not necessarily work adjustment, and the mitigation of stress by previous overseas assignments is seemingly nil. Although expatriates transfer, their international experiences appear not to. Maladjustment rises and falls because of unique local conditions. What does the stress of poor adjustment lead to for expatriates? From the viewpoint of psychological strains, our review indicates that poor adjustment manifests itself in job and non-work dissatisfaction, as well as psychological withdrawal or intentions to prematurely quit the assignment. Perhaps most critically, poor adjustment manifests itself in poor performance, although there is a great deal more to be learned about the latter outcome and its basic dimensions for international assignees. The latter sets of connections are robust to percept-percept inflation problems, and to a reasonably degree, validate the importance of adjustment as the focus of past research on expatriate experiences. In the next section, we turn our attention more fully to the future, noting research needs and suggesting ways that this area might move past the predominance of expatriate adjustment.

Fork in the Road? Research on expatriate experiences seems to have evolved into reliance on a comfortable (although difficult, at least in terms of population access) paradigm and well-worn road. Theories tend to revolve around the stress of adjustment, and they tend to concentrate exclusively on expatriates themselves rather than other elements of their social circumstances: family members, HCNs, or the rest of the members of the foreign operation. The modal empirical study seems to involve: (a) contact with a multi-national corporation (MNC) or expatriate social organizations to sponsor surveys or give distribution lists; (b) distribution of questionnaires to expatriates, either sent to multiple locations from the same MNC or living in the same location (e.g. Hong Kong) but being posted by different firms; and (c) the use of Black and Stephens (1989) three-dimensional adjustment scale as a multivariate but cross-sectional and same-source dependent variable, or to a lesser extent, as

Expatriate Experiences

237

an independent variable. Although we do not believe it is time to abandon adjustment as a primary variable in this domain, it may be time to push it into the conceptual background or highlight different features of it, and pave a new road of investigations. Below, we pick up some of the themes mentioned earlier and make recommendations for the next decade of expatriate research. The Who and When of Expatriate Experiences Past and current research has been dominated by samples of employees from the home countries of western or highly industrialized East Asian (especially Japanese) cultures, sent overseas on long-term assignments, mainly to cultures with the same or lower standards of living. Indeed, it would seem that some of the arguments in the predominant theory of expatriate adjustment by Black et al., 1991, might be predicated on the deficiency conditions produced by such a move. Yet, it would be both theoretically and practically useful to study different expatriate populations and “assignment vectors.” Use of in-depth, qualitative methods would be especially useful in identifying whether the substance of adjustment is similar (e.g. Caligiuri, Hyland, Joshi & Bross, 1998; Shaffer & Harrison, 2001). Set of TCNs and inpatriates might be easier to access, especially if they are stationed in the researchers’ own country. At the same time, members of these semi-captive populations are more likely to contain expatriates who are on an excess trajectory. Depending on their pay structures (they might experience stronger negative inequities relative to HCNs than their PCN counterparts posted elsewhere), their experiences might not be as trying as the expatriate literature tends to portray. The same populations might also be easier to track over time. As expatriate researchers who have repeatedly done cross-sectional studies ourselves, it is embarrassing to note that almost no longitudinal research has been conducted regarding expatriate experiences in the last 20 years. Despite compelling theoretical statements (e.g. Black & Mendenhall, 1991), we as a discipline have no clear data on the natural course of adjustment. If adjustment patterns do follow a systematic curve or even U-shape, knowing the low points and the reasons for them would allow the design of more effective assignments and support interventions. Internet-based surveys, along with e-mail prompts to expatriate in any country, provide a tremendous opportunity for longitudinal data collection. Researchers might also stagger their surveys at equal intervals, but random start points, in a way that provides a more complete mapping of the adjustment curve. For example one sampled expatriate might receive survey prompts at 1, 5, 9, and 13 months, while another might receive them at 3, 7, 11, and 15 months, and so on. Expatriates (as well as inpatriates and TCNs) serve as cross-cultural boundary spanners, and in some ways, act as the cultural ambassadors – the face and voice – of their home countries. Yet, just as globalization trends have created stronger

238

DAVID A. HARRISON ET AL.

needs for expatriate employees; technological trends might temper that need. For example, rather than assign someone to supervise a product development team in Western Europe, a firm might instead form a virtual international team, attempting to learn about member cultures and integrate it into managerial practices via telecommunications rather than physical presence. In an even more salient example of “virtual expatriation,” call centers that service overseas products have burgeoned in India, the Philippines, and even China. These employees, numbering in the millions, are expected to be cultural conversant with the American or European clients at the other end of the phone or network connection, even to the point of adopting a different name and identity. Although it is unlikely that the same adjustment mechanisms apply, it is likely that predictor constructs such as cultural novelty and ethnocentrism might be operating under these latter conditions. The How and Why of Expatriate Experiences The array of variables used to predict expatriate experiences is simultaneously impressive (given their breadth and number) and disappointing (given their often atheoretical nature: see Mendenhall et al., 2002 for a wide-ranging indictment). Although a good portion of the uncertainty that underlies international assignments has been brought to light, there is also a tendency to highlight predictors that can be measured on a questionnaire at the same time as the adjustment criterion. A greater yield of insights would come from not only allowing longitudinal tracing of adjustment and other mechanisms, but also from attending more closely to the expatriate’s social environment. The most important elements in the expatriate’s work environment are his or her colleagues and subordinates, who are likely to be HCN’s. The positions of the assignees and HCNs in the unit-level, informal social network – a way to gauge social capital – would help determine the resources the expatriate has available to foster adjustment and eventual performance. The human capital that expatriates carry with them to their postings has also seen far too little research. The large amount of unexplained variance in expatriate performance (nearly 90%) stands as an open door to new (and old) paradigms for understanding this aspect of expatriate experiences. For example, although enduring individual differences such as general mental ability (g) and the “Big Five” personality dimensions are predominant variables in much of the HR literature on domestic job performance, they have seen only limited empirical inquiry in the expatriate literature (e.g. Caligiuri, 2000; Dalton & Wilson, 2000; Shaffer, Ferzandi, Harrison, Black & Gregersen, 2003). One might expect the relative complexity of international assignments to require a fairly high level of g, and therefore be sensitivity to it. But, other than a small sample of Peace Corps volunteers (n = 41; Mischel, 1965) who were administered an intelligence test under what the authors admit were atypical conditions, there are no published

Expatriate Experiences

239

investigations of the implications of cognitive ability for performance among expatriates. In terms of the “Big Five” dimensions, openness to experience has seen the least predictive validity in domestic HR research. Results of current predictive studies are equivocal regarding the importance of the Big Five dimensions. However, most of the theoretical and standpoint papers on expatriate performance suggest that it, or constructs akin to it (e.g. flexibility), are essential to effectiveness on international assignments (Caligiuri, Jacobs & Farr, 2000; Ones & Viswesvaran, 1997). Perhaps we are putting the performance cart before the task horse It is strange that in the expatriate literature, researchers (ourselves included) have not spent much time closely inspecting the requirements of jobs, and instead have concentrated on the psychological disruption (maladjustment) accompanied by expatriation itself. This monolithic treatment belies what may be tremendous variability in international assignments. That is, we have not heeded our own HR prescriptions and done a thorough job or task analysis of positions to determine what KSAO’s are necessary for top performance among expatriates. Only then would we know how intense and frequent the interpersonal interactions with HCN’s are, or how much abstract decision-making is required, to get a stronger sense of the extent to which various dimensions of adjustment, ability, and personality are necessary to be effective. The So-What of Expatriate Experiences Effectiveness is the broad term we would use to characterize the criterion space for domestic and expatriate employees, but the latter realm of research has been concentrated on one or two corners in that space. If adjustment is only weakly related to performance, it leaves the “so-what” question nearly unanswered. One answer is that adjustment is fairly strongly related to withdrawal cognitions, but the counterpoint is that there are no studies estimating its relationships to actual turnover or a failed assignment (early return). Given that the costs of assignment failures are often the premise on which much of expatriate research is based (e.g. Shaffer & Harrison, 1998), a top priority of future studies should be to empirically connect observational or archival records of early return with other aspects of expatriate experiences. A quick review of the consequences section of our article reveals a fairly short list of outcomes of adjustment, at least in terms of those that have received systematic investigation (see Table 2). Although a subset of expatriate attitudes, cognitions, and behaviors are dependably connected to adjustment, there are a number of other vital, individual-level criteria that have seen little or no study. These include work motivation or effort regulation (Harrison & Shaffer, 2001), mental and physical strains (although there have been studies of a related construct: subjective

240

DAVID A. HARRISON ET AL.

well-being; Anderz´en & Arnetz, 1992; Feldman & Thomas, 1992; Forster, 1997; Nicholson & Imaizumi, 1993), and salary and promotion. Investigations of criteria beyond the expatriate are atypical in this research area. As we have documented earlier, researchers have shown that spouse or family adjustment can have a major impact on the adaptation of the expatriate. On the other hand, despite many anecdotal reports of the possible burdens an overseas move might put on the family (e.g. Fukuda & Chu, 1994), the impact of the expatriate’s stress on his or her spouse and children is virtually unexplored (DeCieri et al., 1991). Conceptually, inadequate adjustment can force the expatriate to take time and emotional investment from family activities and settings that might already be resource-impoverished because of the pressures of adapting to a foreign environment. For example, Harrison and Shaffer (2001), who, in matched-pair data showed that the largest contributor to spouse adjustment while on assignment was expatriate adjustment. In data from another sample of matched expatriate-spouse pairs, Shaffer et al. (2001) demonstrated that various forms of work-family conflict while on assignment were related to adjustment problems. Finally, if expatriate research is to have an impact on other areas of management, research needs to take a stronger strategic perspective. One of the most important questions to yet be answered is how much impact the experiences of expatriates have on performance of the subsidiary to which they are posted. Do assignment “failures” really translate into lost contracts, smaller market shares, and damaged customer, vendor, and supplier relationships overseas? Applications of strategic management theories about control, discretion, and agency are relevant (e.g. Toh & Denisi, 2003), as are ideas about how expatriates might serve as conduits for two-way, subsidiary-base knowledge transfers (Bonache & Brewster, 2001), and organizational behavior theories about person-group (and person-culture) fit (e.g. Harrison et al., 2002). If international assignees are as crucial for achieving a firm’s international goals as our unstated research assumptions suggest, it is imperative that the next 25 years of work go to new places, laying a firmer, time-sensitive, and cross-level pavement of findings about expatriate experiences.

REFERENCES Anderz´en, I., & Arnetz, B. B. (1999). Psychophysiological reactions to international adjustment: Results from a controlled, longitudinal study. Psychotherapy and Psychosomatics, 68(2), 67–75. Arthur, W. J., & Bennett, W. J. (1997). A comparative test of alternative models of international assignee job performance. In: Z. Aycan (Ed.), New Approaches to Employee Management, Vol. 4: Expatriate Management: Theory and Research (pp. 141–172). Stamford, CT: JAI Press.

Expatriate Experiences

241

Aryee, S., & Stone, R. J. (1996). Work experiences, work adjustment and psychological well-being of expatriate employees in Hong Kong. International Journal of Human Resource Management, 7(1), 150–163. Au, K. Y., & Fukuda, J. (2002). Boundary spanning behaviors of expatriates. Journal of World Business, 37, 285–297. Aycan, Z. (1997). Expatriate adjustment as a multifaceted phenomenon: Individual and organizational level predictors. International Journal of Human Resource Management, 8(4), 434–456. Aycan, Z., & Berry, J. W. (1996). Impact of employment-related experiences on psychological wellbeing and adaptation of immigrants. Canadian Journal of Behavioral Sciences, 28(3), 240–251. Aycan, Z., & Kanungo, R. N. (1997). Current issues and future challenges in expatriate management. In: Z. Aycan (Ed.), Expatriate Management: Theory and Research (pp. 245–260). Greenwich, CT: JAI Press. Bell, M. P., & Harrison, D. A. (1996). Using intra-national diversity for inter-national assignments: A model of bicultural life experiences and expatriate adjustment. Human Resources Management Review, 6, 47–74. Bhuian, S. N., & Al-Jabri, I. M. (1996). Expatriate turnover tendencies in Saudi Arabia: An empirical examination. International Journal of Organizational Analysis, 4(4), 393–407. Birdseye, M. G., & Hill, J. S. (1995). Individual, organizational/work and environmental influences on expatriate turnover tendencies: An empirical study. Journal of International Business Studies, 26(4), 787–813. Black, J. S. (1988). Work role transitions: A study of American expatriate managers in Japan. Journal of International Business Studies, 19(2), 277–294. Black, J. S. (1990a). Locus of control, social support, stress, and adjustment in international transfers. Asia Pacific Journal of Management, 7(1), 1–29. Black, J. S. (1990b). The relationship of personal characteristics with the adjustment of Japanese expatriate managers. Management International Review, 30(2), 119–134. Black, J. S., & Gregersen, H. B. (1991a). Antecedents to cross-cultural adjustment for expatriates in Pacific Rim assignments. Human Relations, 44(5), 497–515. Black, J. S., & Gregersen, H. B. (1991b). The other half of the picture: Antecedents of spouse crosscultural adjustment. Journal of International Business Studies, 22(3), 461–477. Black, J. S., & Gregersen, H. B. (1991c). When Yankee comes home: Factors related to expatriate and spouse repatriation adjustment. Journal of International Business Studies, 22(4), 671–694. Black, J. S., & Mendenhall, M. E. (1990). Cross-cultural training effectiveness: A review and theoretical framework for future research. Academy of Management Review, 15, 113–136. Black, J. S., & Mendenhall, M. (1991). The U-curve adjustment hypothesis revisited: A review and theoretical framework. Journal of International Business Studies, 22(2), 225–247. Black, J. S., Mendenhall, M., & Oddou, G. (1991). Toward a comprehensive model of international adjustment: An integration of multiple theoretical perspectives. Academy of Management Review, 16(2), 291–317. Black, J. S., & Stephens, G. K. (1989). The influence of the spouse on American expatriate adjustment and intent to stay in Pacific Rim overseas assignments. Journal of Management, 15(4), 529–544. Blumberg, M., & Pringle, C. D. (1982). The missing opportunity in organizational research: Some implications for a theory of work performance. Academy of Management Review, 7, 560–569. Bonache, J., & Brewster, C. (2001). Knowledge transfer and the management of expatriation. Thunderbird International Business Review, 43, 145–151. Borman, W. (1969). The problem of expatriate personnel and their selection in international business. Management International Review, 9, 37–48.

242

DAVID A. HARRISON ET AL.

Borman, W. C., & Motowidlo, S. J. (1993). Expanding the criterion domain to include elements of contextual performance. In: N. Schmitt & W. C. Borman (Eds), Personnel Selection (pp. 71–98). San Francisco: Jossey-Bass. Boyacigiller, N., Kleinberg, M. J., Philips, M., & Sackmann, S. (1996). Conceptualizing culture. In: B. J. Punnett & O. Shenkar (Eds), Handbook for International Management Research. Cambridge, MS: Blackwell. Brein, M., & David, K. H. (1971). Intercultural communication and the adjustment of the sojourner. Psychological Bulletin, 76, 215–230. Briscoe, D. (1995). International human resource management. Prentice-Hall. Caligiuri, P. M. (1997). Assessing expatriate success: Beyond just “being there.” In: Z. Aycan (Ed.), New Approaches to Employee Management, Vol. 4: Expatriate Management: Theory and Research (pp. 117–140). Stamford, CT: JAI Press. Caligiuri, P. M. (2000a). The Big Five personality characteristics as predictors of expatriate’s desire to terminate the assignment and supervisor-rated performance. Personnel Psychology, 53(1), 67–88. Caligiuri, P. M. (2000b). Selecting expatriates for personality characteristics: A moderating effect of personality on the relationship between host national contact and cross-cultural adjustment. Management International Review, 40(1), 61–80. Caligiuri, P. M., & Day, D. V. (2000). Effects of self-monitoring on technical, contextual, and assignment-specific performance: A study of cross-national work performance ratings. Group and Organization Management, 25(2), 154–175. Caligiuri, P. M., Hyland, M. M., Joshi, A., & Bross, A. S. (1998). Testing a theoretical model for examining the relationship between family adjustment and expatriates’ work adjustment. Journal of Applied Psychology, 83, 598–610. Caligiuri, P., Joshi, A., & Lazarova, M. (1999). Factors influencing the adjustment of women on global assignments. International Journal of Human Resource Management, 10(2), 163–179. Caligiuri, P., Phillips, J., Lazarova, M., Tarique, I., & Buergi, P. (2001). The theory of met expectations applied to expatriate adjustment: The role of cross-cultural training. International Journal of Human Resource Management, 12(3), 357–372. Caligiuri, P. M., & Tung, R. L. (1999). Comparing the success of male and female expatriates from a US-based multinational company. International Journal of Human Resource Management, 10(5), 763–782. Campbell, A. (1981). The sense of well-being in America. New York: McGraw Hill. Chan, D. (2000). Understanding adaptation to changes in the work environment: Integrating individual difference and learning perspectives. In: G. Ferris & J. J. Martocchio (Eds), Research in Personnel and Human Resources Management (Vol. 18, pp. 1–42). JAI Press. Chan, D. K. S., Gelfand, M. J., Triandis, H. C., & Tzeng, O. (1996). Tightness-looseness revisited: A systematic examination in Japan and the United States. International Journal of Psychology, 31(1), 1–12. Cleveland, H., Mangone, G., & Adams, J. C. (1960). The overseas Americans. New York: McGraw-Hill. Cohen, S. (1980). After effects of stress on human performance and social behavior: A review of research and theory. Psychological Bulletin, 88, 82–108. Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98, 310–357. Dalton, D. R., Johnson, J. L., & Daily, C. M. (1999). On the use of ‘intent to . . .’ variables in organizational research: An empirical and cautionary assessment. Human Relations, 52(10), 1337–1350.

Expatriate Experiences

243

Dalton, D. R., & Todor, W. D. (1993). Turnover, transfer, absenteeism: An independent perspective. Journal of Management, 19, 193–219. Dalton, M., & Wilson, M. (2000). The relationship of the Five-Factor Model of Personality to job performance for a group of Middle Eastern expatriate managers. Journal of Cross Cultural Psychology, 31(2), 250–258. Davidson, M., & Cooper, C. (1983). Stress and the woman manager. Oxford. Dawis, R. B., & Lofquist, L. H. (1984). A psychological theory of work adjustment. Minneapolis: University of Minnesota Press. DeCieri, H., Dowling, P. J., & Taylor, K. F. (1991). The psychological impact of expatriate relocation on international partners. International Journal of Human Resource Management, 2(3), 377–414. Deller, J. (1997). Expatriate selection: Possibilities and limitations of using personality scales. In: Z. Aycan (Ed.), New Approaches to Employee Management, Vol. 4: Expatriate Management: Theory and Research (pp. 93–116). Stamford, CT: JAI Press. Deshpande, S. P., & Viswesvaran, C. (1992). Is cross-cultural training of expatriate managers effective: A meta-analysis. International Journal of Intercultural Relations, 16(3), 295–310. Earley, P. C. (1994). Self or group? Cultural effects of training on self-efficacy and performance. Administrative Science Quarterly, 39, 89–117. Feldman, D. C., & Thomas, D. C. (1992). Career management issues facing expatriates. Journal of International Business Studies, 23(2), 271–293. Florkowski, G. W., & Fogel, D. S. (1999). Expatriate adjustment and commitment: The role of host-unit treatment. International Journal of Human Resource Management, 10(5), 783–807. Fontaine, G. (1996). Social support and the challenges of international assignments: Implications for training. In: D. Landis & R. S. Bhagat (Eds), Handbook of Intercultural Training (2nd ed., pp. 264–281). Thousand Oaks, CA: Sage. Forster, N. S. (1997). The persistent myth of high expatriate failure rates: A reappraisal. International Journal of Human Resource Management, 8(4), 414–433. Fukuda, K. J., & Chu, P. (1994). Wrestling with expatriate family problems. Japanese experiences in east Asia. International Studies of Management and Organizations, 24, 36–47. Furnham, A. (1987). The adjustment of sojourners. In: Y. Kim & W. Gudykunst (Eds), Cross-Cultural Adaptation: Current Approaches (??). Newbury Park, CA: Sage. GMAC Global Relocation Services (2003). Global relocation trends 2002 survey report (March). Website: http://www.gmacglobalrelocation.com. Gregersen, H., & Black, J. S. (1990). A multifaceted approach to expatriate retention in international assignments. Group and Organization Studies, 15, 461–485. Gregersen, H. B., & Stroh, L. K. (1997). Coming home to the Arctic cold: Antecedents to Finnish expatriate and spouse repatriation adjustment. Personnel Psychology, 50(3), 635–654. Griffeth, R. W., Hom, P. W., & Gaertner, S. (2000). A meta-analysis of antecedents and correlates of employee turnover: Update, moderator tests, and research implications for the next millennium. Journal of Management, 26(3), 463–488. Gudykunst, W. B., & Hammer, M. R. (1984). Dimensions of intercultural effectiveness: Culture specific or culture general? International Journal of Intercultural Relations, 8, 1–10. Gullahorn, J. T., & Gullahorn, J. E. (1963). An extension of the U-curve hypothesis. Journal of Social Issues, 14, 33–47. Guzzo, R. A., Noonan, K. A., & Elron, E. (1994). Expatriate managers and psychological contract. Journal of Applied Psychology, 79, 617–626.

244

DAVID A. HARRISON ET AL.

Harrison, D. A. (2002). Meaning and measurement of work role withdrawal: Current controversies and future fallout under changing technology. In: M. Koslowsky & M. Krausz (Eds), Voluntary Employee Withdrawal and Inattendance: A Current Perspective (pp. 95–132). Plenum. Harrison, D. A., McLaughlin, M. A., & Coalter, T. M. (1996). Context, cognition, and common method variance: Psychometric and verbal protocol evidence. Organizational Behavior and human Decision Processes, 68(3), 246–261. Harrison, D. A., & Shaffer, M. A. (2001). Mapping criterion space for expatriate success: Task and relationship based performance, effort and adaptation. Presented at the 2001 meetings of the Academy of Management, Washington, DC. Harrison, D. A., Shaffer, M. A., & Bhaskar, P. (2002). Moving outward from organizations: Personculture fit in the experiences of expatriates. Presented at the 2002 meetings of the Academy of Management, Denver, Colorado. Harrison, J. K., Chadwick, M., & Scales, M. (1996). The relationship between cross-cultural adjustment and the personality variables of self-efficacy and self-monitoring. International Journal of Intercultural Relations, 20(2), 167–188. Harzing, A. W. (2001a). Of bears, bumble-bees, and spiders: The role of expatriates in controlling foreign subsidiaries. Journal of World Business, 36(4), 366. Harzing, A. W. (2001b). An analysis of the functions of international transfer of managers in MNCs. Employee Relations, 23(6), 581–599. Hawes, F., & Kealey, D. J. (1981). An empirical study of Canadian technical assistance: Adaptation and effectiveness on overseas assignment. International Journal of Intercultural Relations, 5, 239–258. Hays, R. (1974). Expatriate selection: Insuring success and avoiding failure. Journal of International Business Studies, 5(1), 25–37. Hom, P. W., Griffeth, R. W., & Sellaro, C. L. (1984). The validity of Mobley’s (1977) model of employee turnover. Organizational Behavior and Human Performance, 34, 141–174. Inkson, K., Arthur, M. B., Pringle, J. K., & Barry, S. (1997). Expatriate assignment vs. overseas experience: Contrasting models of human resource development. Journal of World Business, 32(4), 351–368. Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job satisfaction-job performance relationship: A qualitative and quantitative review. Psychological Bulletin, 127, 376–407. Karasek, R. A. (1979). Job demands, job decision latitude and mental strain: Implications for job redesign. Administrative Science Quarterly, 24, 285–308. Kealey, D. J. (1989). A study of cross-cultural effectiveness: Theoretical issues and practical applications. International Journal of Intercultural Relations, 13, 387–428. Kealey, D. J., & Protheroe, D. R. (1996). The effectiveness of cross-cultural training for expatriates: An assessment of the literature on the issue. International Journal of Intercultural Relations, 20(2), 141–165. Kelman, H. C. (Ed.) (1965). International behavior: A social-psychological analysis. New York: Holt, Rinehart & Winston. Kraimer, M. L., Wayne, S. J., & Jaworski, R. A. (2001). Sources of support and expatriate performance: The mediating role of expatriate adjustment. Personnel Psychology, 54, 71–99. Lance, C. E., & Richardson, D. (1988). Correlates of work and non-work stress and satisfaction among American insulated sojourners. Human Relations, 41(10), 725–738. Latham, G. P., & Locke, E. A. (1991). Self regulation through goal setting. Organizational Behavior and Human Decision Processes, 50, 212–247.

Expatriate Experiences

245

Lee, T. W., & Mitchell, T. R. (1994). An alternative approach: The unfolding model of voluntary employee turnover. Academy of Management Review, 19, 51–89. Leiba-O’Sullivan, S. (1999). The distinction between stable and dynamic cross-cultural competencies: Implications for expatriate trainability. Journal of International Business Studies, 30(4), 709– 725. Locke, E. A. (1976). The nature and causes of job satisfaction. In: M. Dunnette (Ed.), Handbook of Industrial and Organizational Psychology (??). Chicago: Rand McNally. Lysgaard, S. (1955). Adjustment in a foreign country: Norwegian Fulbright grantees visiting the United States. International Social Science Bulletin, 7, 45–51. Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224–253. Mendenhall, M., & Oddou, G. (1985). The dimensions of expatriate acculturation: A review. Academy of Management Review, 10, 39–48. Mendenhall, M., & Oddou, G. (1986). Acculturation profiles of expatriate managers: Implications for cross-cultural training programs. Columbia Journal of World Business, 21(4), 73–79. Mendenhall, M., Kuhlmann, T., Stahl, G., & Osland, J. (2002). Employee development and expatriate assignments: A review of the expatriate adjustment theory literature. In: M. Gannon & K. Newman (Eds), Handbook of Cross-Cultural Management (pp. 155–183). Oxford, UK: Blackwell. Mischel, W. (1965). Predicting the success of Peace Corps volunteers in Nigeria. Journal of Personality and Social Psychology, 1, 517–520. Mobley, W. H. (1982). Some unanswered questions in turnover and withdrawal research. Academy of Management Review, 7, 111–116. Morris, M. A., & Robie, C. (2001). A meta-analysis of the effects of cross-cultural training on expatriate performance and adjustment. International Journal of Training and Development, 5, 112–125. Moyerman, D. R., & Forman, B. D. (1992). Acculturation and adjustment: A meta-analytic study. Hispanic Journal of Behavioral Sciences, 14, 163–200. Munton, A. G., & West, M. A. (1995). Innovations and personal change: Patterns of adjustment to relocation. Journal of Organizational Behavior, 16(4), 363–376. Naumann, E., Widmier, S. M., & Jackson, D. W., Jr. (2000). Examining the relationship between work attitudes and propensity to leave among expatriate salespeople. Journal of Personal Selling and Sales Management, 20(4), 227–241. Newman, D. A., Harrison, D. A., & Roth, P. L. (2003). Utility of overall job attitude for predicting individual effectiveness: Meta-analytic tests. Paper presented at the meetings of the Academy of Management, Seattle, WA. Nicholson, N. (1984). A theory of work role transitions. Administrative Science Quarterly, 29(2), 172–191. Nicholson, N., & Imaizumi, A. (1993). The adjustment of Japanese expatriates to living and working in Britain. British Journal of Management, 4(2), 119–134. O’Grady, S., & Lane, H. W. (1996). The psychic distance paradox. Journal of International Business Studies, 27, 309–333. Ones, D. S., & Viswesvaran, C. (1999). Relative importance of personality dimensions for expatriate selection: A policy capturing study. Human Performance, 12(3–4), 275–294. ORC (1997–1998). Worldwide survey of international assignment policies and practices. New York: Organization Resources Counselors. Parker, B., & McEvoy, G. (1993). Initial examination of a model of intercultural adjustment. International Journal of Intercultural Relations, 17, 355–379.

246

DAVID A. HARRISON ET AL.

Pinder, C., & Schroeder, K. (1987). Time to proficiency following transfers. Academy of Management Journal, 30, 336–353. Podsakoff, P. M., MacKenzie, S. B., Paine, J. B., & Bachrach, D. G. (2000). Organizational citizenship behaviors: A critical review of the theoretical and empirical literature and suggestions for future research. Journal of Management, 26(3), 513–563. Redfield, R., Linton, R., & Herskovitz, M. J. (1936). Memorandum for the study of acculturation. American Anthropologist, 38, 149–152. Reichers, A. (1985). A review and reconceptualization of organizational commitment. Academy of Management Review, 10, 465–476. Ronen, S. (1989). Training the international assignee. In: I. L. Goldstein (Ed.), Training and Development in Organizations (pp. 417–453). San Francisco: Jossey-Bass. Ruben, B., & Kealey, D. (1979). Behavioral assessment of communication competency and the prediction of cross cultural adaptation. International Journal of Intercultural Relations, 3, 15–47. Rusbult, C. E., Farrell, D., Rogers, G., & Mainous, A. G. (1988). Impact of exchange variables on exit voice, loyalty, and neglect: An integrative model of responses to declining job dissatisfaction. Academy of Management Journal, 31, 599–627. Sanchez, J. I., Spector, P. E., & Cooper, C. L. (2000). Adapting to a boundaryless world: A developmental expatriate model. Academy of Management Executive, 14(2), 96–106. Searle, W., & Ward, C. (1990). The prediction of psychological and sociocultural adjustment during cross-cultural transitions. International Journal of Intercultural Relations, 14, 449–464. Selmer, J. (1998a). Adjustment of western European vs North American expatriate managers in China. Personnel Review, 30, 6–21. Selmer, J. (1998b). Expatriation: Corporate policy, personal intentions and international adjustment. International Journal of Human Resource Management, 9(6), 996–1007. Selmer, J. (1999a). Career issues and international adjustment of business expatriates. Career Development International, 4(2), 77–87. Selmer, J. (1999b). Effects of coping strategies on sociocultural and psychological adjustment of Western expatriate managers in the PRC. Journal of World Business, 34(1), 41–51. Selmer, J. (2000). Usage of corporate career development activities by expatriate managers and the extent of their international adjustment. International Journal of Commerce & Management, 10(1), 1–23. Selmer, J. (2001). Psychological barriers to adjustment and how they affect coping strategies: Western business expatriates in China. International Journal of Human Resource Management, 12(2), 151–165. Selmer, J., Ebrahimi, B. P., & Li, M. (2000). Personal characteristics and adjustment of Chinese mainland business expatriates in Hong Kong. International Journal of Human Resource Management, 11(2), 237–250. Shaffer, M. A., & Harrison, D. A. (1998). Expatriates’ psychological withdrawal from international assignments: Work, non-work, and family influences. Personnel Psychology, 51(1), 87–118. Shaffer, M. A., & Harrison, D. A. (2001). Forgotten partners of international assignments: Development and test of a model of spouse adjustment. Journal of Applied Psychology, 86(2), 238–254. Shaffer, M. A., Harrison, D. A., & Gilley, K. M. (1999). Dimensions, determinants, and differences in the expatriate adjustment process. Journal of International Business Studies, 30(3), 557–581. Shaffer, M. A., Harrison, D. A., Gilley, K. M., & Luk, D. M. (2001). Struggling for balance amid turbulence on international assignments: Work-family conflict, support and commitment. Journal of Management, 27(1), 99–121.

Expatriate Experiences

247

Shaw, J. B. (1990). A cognitive categorization model for the study of intercultural management. Academy of Management Review, 15(4), 626–645. Staines, G. L. (1980). Spillover vs. compensation: A review of the literature on the relationship between work and non-work. Human Relations, 33, 111–129. Shannonhouse, R. (1996). Overseas failures. USA Today. International Edition, 8A, November 8. Sinangil, H. K., & Ones, D. S. (1997). Empirical investigations of the host country perspective in expatriate management. In: Z. Aycan (Ed.), New Approaches to Employee Management, Vol. 4: Expatriate Management: Theory and Research (pp. 173–205). Stamford, CT: JAI Press. Stock, J., & Cervone, D. (1990). Proximal goal setting and self-regulatory processes. Cognitive Therapy and Research, 14, 483–498. Stroh, L. K., Dennis, L. E., & Cramer, T. C. (1994). Predictors of expatriate adjustment. International Journal of Organizational Analysis, 2(2), 176–192. Sullivan, S. E., & Bhagat, R. S. (1992). Organizational stress, job satisfaction and job performance: Where do we go from here? Journal of Management, 18(2), 353–374. Suutari, V., & Burch, D. (2001). The role of on-site training and support in expatriation: Existing and necessary host-company practices. Career Development International, 6, 298–311. Takeuchi, R., Yun, S., & Russell, J. E. (2002a). Antecedents and consequences of the perceived adjustment of Japanese expatriates in the USA. International Journal of Human Resource Management, 13, 1224–1244. Takeuchi, R., Yun, S., & Tesluk, P. E. (2002b). An examination of crossover and spillover effects of spousal and expatriate cross-cultural adjustment on expatriate outcomes. Journal of Applied Psychology, 87(4), 655–666. Taylor, S., & Napier, N. (1996). Working in Japan: Lessons from women expatriates. Sloan Management Review, 37(3), 76–84. Toh, S. M., & Denisi, A. S. (2003). Host country national reactions to expatriate pay policies: A model and implications. Academy of Management Review, 28, 606–628. Torbiorn, I. (1982). Living abroad: Personal adjustment and personnel policy in the overseas setting. New York: Wiley. Tung, R. L. (1981). Selection and training of personnel for overseas assignments. Columbia Journal of World Business, 16(1), 68–78. Tung, R. L. (1982). Selection and training procedures of U.S., European and Japanese multinationals. California Management Review, 25(1), 51–71. Tung, R. L. (1998). A contingency framework of selection and training of expatriates revisited. Human Resource Management Review, 8(1), 23. Van Scotter, J., & Motowidlo, S. J. (1996). Interpersonal facilitation and job dedication as separate facets of contextual performance. Journal of Applied Psychology, 81, 525–531. Wanous, J. P., & Lawler, E. E. (1972). Measurement and meaning of job satisfaction. Journal of Applied Psychology, 56, 95–105.

INFERENCE PROBLEMS WITH HIERARCHICAL MULTIPLE REGRESSION-BASED TESTS OF MEDIATING EFFECTS Eugene F. Stone-Romero and Patrick J. Rosopa ABSTRACT Mediating effects are often tested using hierarchical multiple regression (HMR) procedures. Typical of the HMR-based strategies is the very frequently cited and widely used procedure described by Baron and Kenny (1986). Unfortunately, there are several important problems with it. More specifically, as we demonstrate below, it: (a) is of virtually no value for buttressing claims of mediating effects for data from non-experimental research; (b) produces erroneous inferences about the existence of mediating effects for misspecified mediation models; and (c) is incapable of providing credible evidence of such effects in a large proportion of cases, even for properly specified mediation models. We detail a number of important implications of our analyses.

Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 249–290 Copyright © 2004 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23006-7

249

250

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

INFERENCE PROBLEMS WITH HIERARCHICAL MULTIPLE REGRESSION-BASED TESTS OF MEDIATING EFFECTS A mediator is a variable that transmits the effects of an independent variable to a dependent variable. Figure 1a shows a model involving assumed mediation for three variables expressed in standard score form. In it, the mediator (e.g. Z2 ) intervenes between the independent variable (e.g. Z1 ) and the dependent variable (e.g. Z3 ). Researchers in many different disciplines (e.g. management, psychology, sociology) deal with assumed causal models that involve mediators. For example: (a) behavioral intentions are believed to mediate the relation between attitudes and behavior (Azjen & Fishbein, 1980); (b) resistance skills are assumed to mediate the relation between drug abuse prevention programs and the ability of children to refuse drug offers (Donaldson, Graham, Piccinin & Hansen, 1995); and (c) trust is assumed to mediate the relation between procedural justice and organizational citizenship behavior (Konovsky & Pugh, 1994). In most instances, tests of mediation models are based upon the analysis of data from non-experimental research. Thus, a key concern is the validity of inferences about mediation that stem from such tests. Unfortunately, as we demonstrate below, tests of mediation models that are based upon data from non-experimental studies have little or no capacity to serve as a basis for valid inferences about mediation. This is especially true for studies that test for mediation using the hierarchical multiple regression (HMR) procedure popularized by Baron and Kenny (1986). To demonstrate why this is so, this article considers the following issues: (a) examples of tests of mediation models in the field of personnel and human resource management; (b) illustrations of the ambiguity that surrounds causal models in the same field; (c) the difference between assumed mediation and actual mediation; (d) strategies that are commonly used to test mediation models using data from non-experimental research, including the HMR procedure; (e) the validity of inferences about mediation when the results of a study are consistent with an assumed model; (f) statistical estimates and significance tests associated with HMR analyses; (g) the partitioning of explained variance in multiple regression analyses; (h) the interpretation problems that stem from tests of misspecified mediation models; (i) the method, results, and implications of a simulation study that assessed the validity of inferences stemming from the use of the HMR procedure; (j) some other issues associated with testing causal models using data from non-experimental research; and (k) general conclusions that stem from the consideration of the just-noted issues.

Tests of Mediating Effects

Fig. 1. Relations Among Variables in Alternative Causal Models.

251

252

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Fig. 1.

(Continued ).

Mediators in Personnel and Human Resource Management Research In personnel and human resource management and allied fields (e.g. industrial and organizational psychology and organizational behavior), assumed causal models are frequently tested using data from non-experimental studies. There are numerous examples of such studies (e.g. Baum & Wally, 2003; Collins & Clark, 2003; Huselid, 1995; Man & Lam, 2003; Masterson, Lewis, Goldman & Taylor, 2000; Robinson, 1996; Vogus & Welbourne, 2003). Masterson et al. (2000) examined the mediating effect of leader member exchange on the relation between interactional justice and performance. Robinson (1996) assessed whether unmet expectations mediated the relation between psychological contract breach and employee contributions to organizations. Man and Lam (2003) tested for the mediating effects of group cohesiveness on the relation between job complexity and work group performance. Huselid (1995) investigated the mediating effects of employee turnover on the relation between the alignment of High Performance Work Practices (see U.S. Department of Labor, 1993) with a company’s competitive strategy and corporate financial performance. Baum and Wally (2003) assessed the mediating effect of strategic decision speed on the relation between:

Tests of Mediating Effects

253

(a) environmental and organizational characteristics; and (b) firm performance. Vogus and Welbourne (2003) tested for the mediating effects of intensity of innovation on the relation between human resource practices (e.g. use of skilled temporary employees and emphasis on training) and financial performance. Finally, Collins and Clark (2003) assessed whether human resource practices mediated the relation between top management team social networks and firm performance. As is demonstrated below, in all of these studies, inferences about actual mediation are highly suspect.

Ambiguity About Causal Flow in Organizational Research It is worth stressing that tests of many assumed causal models in the above-noted fields have yielded highly inconsistent findings regarding the direction of causal flow between and among measured variables. To illustrate this, we next provide examples of tests of inconsistent causal models and the findings of tests of such models from research on relations between: (a) job performance and job satisfaction; and (b) job satisfaction and organizational commitment. Job Performance and Job Satisfaction Interpretations of the results of non-experimental research on the causal connection between job performance and job satisfaction have been very contradictory. As a result of the human relations movement in the 1940s, it was believed that a satisfied employee would be a more productive employee; that is, that job satisfaction causes job performance (Pinder, 1998). In accordance with this model, the findings of research by Shore and Martin (1989) were used to infer that job satisfaction caused job performance. Interestingly, however, results of a path analysis by Keaveney and Nelson (1993) failed to support the same inference. Some researchers have argued that job performance causes job satisfaction. For example, Porter and Lawler’s (1968) expectancy theory model posits that job performance causes job satisfaction when rewards are positively contingent on performance. In support of this hypothesized causal sequence, Stumpf and Hartman (1984) used the findings of a non-experimental study to argue that job performance causes job satisfaction. However, the test of a path model by Birnbaum and Somers (1993) failed to reveal a statistically significant path between the same variables. Other researchers have posited that job performance and job satisfaction are related reciprocally. For example, Wanous (1974) used the results of nonexperimental research to argue that there was a reciprocal relation between job performance and job satisfaction, with the direction of the relation being dependent on the type of satisfaction. More specifically, extrinsic satisfaction was

254

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

viewed as a cause of performance, whereas, performance was regarded as a cause of intrinsic satisfaction. Overall, non-experimental tests of the relation between job performance and job satisfaction have yielded quite inconsistent conclusions about the direction of causal flow. This is not at all surprising because, as we demonstrate below, the findings of non-experimental research can virtually never provide valid inferences about causal connections between measured variables (Cook & Campbell, 1979; Shadish, Cook & Campbell, 2002; Stone-Romero, 2002). Job Satisfaction and Organizational Commitment Studies that have tested models dealing with the assumed causal link between job satisfaction and organizational commitment also have yielded inconsistent inferences. For example, using data from a non-experimental study, Williams and Hazer (1986) tested several such models. The test of one provided evidence consistent with the view that job satisfaction causes organizational commitment. However, the test of another yielded slight support for the view that organizational commitment causes job satisfaction. Some researchers have argued that organizational commitment causes job satisfaction (Bateman & Strasser, 1984; Koslowsky, 1991; Lance, 1991; Wiener & Vardi, 1980). For example, Koslowsky (1991) used structural equation modeling (SEM) procedures to analyze data from a non-experimental study of the job satisfaction-commitment relation. Results of the analysis were used to argue that organizational commitment causes job satisfaction. However, the findings of a non-experimental study by Lance (1991) were used to infer that there is an asymmetric reciprocal relation between job satisfaction and organizational commitment. More specifically, Lance argued that specific job satisfaction attributes cause organizational commitment, and that commitment has a weak effect on job satisfaction. A few studies have tested competing causal models of the relation between job satisfaction and organizational commitment (Currivan, 1999; Curry, Wakefield, Price & Mueller, 1986; Farkas & Tetrick, 1989; Vandenberg & Lance, 1992). The models tested were that: (a) job satisfaction causes organizational commitment; (b) organizational commitment causes job satisfaction; (c) job satisfaction and organizational commitment are reciprocally related; and (d) there is no causal connection between these variables. Vandenberg and Lance (1992) found evidence to support the inference that organizational commitment causes job satisfaction. However, Currivan (1999) found nearly identical fits for all four of the above-noted models, suggesting that the relation between job satisfaction and organizational commitment may be spurious. Moreover, on the basis of data from a longitudinal study, Farkas and Tetrick (1989) argued that the causal direction was unclear

Tests of Mediating Effects

255

and that organizational commitment and satisfaction may be reciprocally related. Finally, Curry et al. (1986) found no causal relation between job satisfaction and organizational commitment. In summary, research that has tested different assumed causal models of the job satisfaction-organizational commitment relation have yielded very conflicting views about the direction of causal flow. This is not at all surprising. As is detailed below, evidence from non-experimental research has virtually no capacity to provide valid inferences about causal connections among measured variables (Bock, 1975; Cook & Campbell, 1979; Dean & Voss, 1999; Holland, 1986; Ling, 1982; Rogosa, 1987; Shadish et al., 2002; Stone-Romero, 2002). In addition, the application of sophisticated statistical procedures to data from such research does little or nothing to improve the validity of causal inferences (Cook & Campbell, 1979; Hume, 1740/2000; Mill, 1884; Shadish et al., 2002; StoneRomero, 2002). Negative Consequences of Inappropriate Causal Inferences In the above-noted studies, causal inferences were based upon the results of applying such data-analytic techniques as ordinary least squares regression, path analysis, and SEM to data from non-experimental studies. Unfortunately, the same inferences can lead to a large number of other non-experimental studies aimed at showing support for incorrect causal models, and the conduct of such studies can lead to the waste of large amounts of valuable resources (e.g. all of the direct and indirect costs of research). In addition, inappropriate, and possibly damaging, recommendations for practice may flow from such studies. For example, as noted by Maruyama (1998), one potential consequence of finding support for an improperly conceived model could be “to design an intervention that, based on inferences drawn from an incorrectly specified model, actually manipulates an effect and not a cause” (p. 5).

Actual vs. Assumed Mediation It is important to distinguish between actual and assumed mediation. To illustrate this, we refer to several models in Fig. 1. In these models, straight lines are used to represent actual or assumed causal connections between variables and curved lines are used to denote non-causal relations. Actual Mediation In the case of actual mediation, changes induced in one variable (e.g. Z1 ) lead to changes in a second (e.g. Z2 ), which, in turn produce changes in a third (e.g.

256

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Z3 ). Figure 1a illustrates such a sequence. Assuming that the data used to test for mediation were from a well-designed and properly conducted experimental study, there would be virtually no question about the causal connections in the model. As such, a researcher could be highly confident about causal inferences stemming from the test of it. Assumed Mediation In non-experimental research mediation is always assumed, as opposed to actual. Consequently, all that a researcher can ever conclude legitimately from such research is that a set of covariances (e.g. correlation coefficients) among variables is consistent with an assumed causal model (e.g. that shown in Fig. 1a). However, as is demonstrated below, the same set of covariances also may be consistent with a large set of other causal models, including those in Fig. 1b–f (Cliff, 1983; Cook & Campbell, 1979; Freedman, 1997; Maruyama, 1998; Pedhazur, 1982; Shadish et al., 2002; Stone-Romero, 2002). As a result, in almost all non-experimental research it is inappropriate to argue that one model is more appropriate than another.

Strategies for Testing for Mediation in Non-experimental Research Numerous techniques have been developed for testing mediating effects in nonexperimental research. MacKinnon and his associates (MacKinnon, Lockwood, Hoffman, West & Sheets, 2002) identified 14 such methods. Among them are the HMR approach of Baron and Kenny (1986), the difference in coefficients methods (e.g. Freedman & Schatzkin, 1992; Olkin & Finn, 1995), and various path model-based approaches (e.g. Alwin & Hauser, 1975; Kenny, 1979; Sobel, 1988). Focus on the Baron and Kenny Procedure The focus of this article is on the often-cited HMR procedure of Baron and Kenny (1986) for testing mediation (referred to hereinafter as the HMR procedure). There are several reasons for focusing on it. First, a search of the Social Science Citation Index by MacKinnon et al. (2002) revealed more than 2,000 citations of the Baron and Kenny (1986) article. Second, the HMR procedure is frequently used to test for mediation. For instance, in 2001 the Journal of Applied Psychology published over 32 articles in which tests for mediation were conducted. Of these, 53% used the HMR procedure. For example, Erez and Judge (2001) used it to test for the mediating effect of task motivation on the relation between core self-evaluations and task performance. In addition, Tepper, Lockhart and Hoobler (2001) used it to

Tests of Mediating Effects

257

test for the mediating effect of employees’ role definitions on the relation between procedural justice and organizational citizenship behavior. Given the frequent use of the HMR procedure in non-experimental research in personnel and human resource management and allied fields, it is critical to assess its utility for demonstrating the existence of mediating effects. Thus, the major purpose of this article is to provide the needed assessment. The Baron and Kenny HMR Procedure Baron and Kenny (1986) recommend a three-step, HMR procedure to test an assumed mediation model using data from non-experimental research. They argue that it can provide evidence of either partial mediation or complete mediation. The distinction between these two types of mediation is considered below. Some conventions. Before describing the HMR procedure, we note several conventions that are used in this article. First, in the interest of simplicity, we use equations that involve standard (Z) scores, averting the need to deal with the intercept term in regression equations. Second, each of the estimated standard ˆ for the below-described HMR analyses is partial regression coefficients (␤s) accompanied by a subscript, in parentheses, that indicates the stage of the analysis. For example, for the first analysis, the subscript (1) appears with the ␤ˆ for Z1 , i.e. ␤ˆ 1(1) . Third, and finally, population regression coefficients are denoted by ␤s to ˆ differentiate them from sample estimates (␤s). The three-step HMR procedure. The HMR procedure requires three sequential regression analyses. We explain these in relation to the model shown in Fig. 1a. In Analysis 1, the assumed mediator (Z2 ) is regressed upon the assumed independent variable, Z1 . ˆ 2 = ␤ˆ 1(1) Z 1 Z

(1)

Baron and Kenny (1986) specify that this analysis must show that ␤ˆ 1(1) = 0. Hereinafter, we refer to this as meeting or satisfying Condition 1. It is a necessary but insufficient (NBI) condition for inferring either partial or complete mediation. In Analysis 2, the assumed dependent variable (Z3 ) is regressed upon the assumed independent variable, Z1 . ˆ 3 = ␤ˆ 1(2) Z 1 Z

(2)

This analysis must show that ␤ˆ 1(2) = 0. Hereinafter, we refer to this as meeting or satisfying Condition 2. It is also a NBI condition for inferring either partial or complete mediation.

258

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

In Analysis 3, the assumed dependent variable (Z3 ) is regressed simultaneously upon the assumed independent variable (Z1 ) and the assumed mediator (Z2 ). ˆ 3 = ␤ˆ 1(3) Z 1 + ␤ˆ 2(3) Z 2 Z

(3)

An inference of complete mediation requires that ␤ˆ 1(3) = 0. Below we refer to this as satisfying Condition 3a. It is a NBI condition for inferring complete mediation. Analysis 3 must also show that |␤ˆ 1(3) | < |␤ˆ 1(2) |. Henceforth, we refer to this as satisfying Condition 3b. It is a NBI condition for inferring either partial or complete mediation. Moreover, for inferring either partial or complete mediation, Analysis 3 must show that ␤ˆ 2(3) = 0. Hereinafter, we refer to this as meeting Condition 3c. It is a NBI condition. ˆ 2 = 0. This is a NBI condition for inferring Finally, Analysis 3 must show that R either partial or complete mediation. Note that when: (a) Conditions 3a, 3b, and 3c are satisfied simultaneously; and ˆ 2 = 0 in Analysis 3, we refer to this as satisfying Condition 3t for complete (b) R mediation. Note that 3t is NBI for inferring complete mediation. Inferences about complete vs. partial mediation. The results of HMR analyses are assumed to provide evidence on two types of mediation, complete and partial (Baron & Kenny, 1986). Inferences about complete mediation require the simultaneous satisfaction of Conditions 1, 2, and 3t. In the case of such mediation, Baron and Kenny (1986) contend that there is “strong evidence of a single, dominant mediator” (p. 1176). In contrast, partial mediation is said to exist when: (a) Conditions 1, 2, 3b and ˆ 2 = 0 in Analysis 3. For such mediation, 3c are satisfied simultaneously and (b) R ˆ ˆ it is critical that |␤1(3) | < |␤1(2) |. However, it is acceptable that ␤ˆ 1(3) = 0. Note that many researchers have used the HMR procedure to infer partial mediation (e.g. Frayne & Geringer, 2000; Korsgaard, Brodt & Whitener, 2002; McAllister & Bigley, 2002; Slaughter, Sinar & Bachiochi, 2002). Although the HMR procedure is frequently used to test for both partial and complete mediation, for reasons that are detailed below, we believe that changes ˆ do not provide valid evidence of in the absolute magnitudes of the above-noted ␤s either type of mediation. Moreover, because causality can never be demonstrated on the basis of non-experimental research (Bock, 1975; Cook & Campbell, 1979; Dean & Voss, 1999; Holland, 1986; Ling, 1982; Rogosa, 1987; Shadish et al., 2002; Stone-Romero, 2002), the results of HMR analyses of data from such research cannot be used to provide evidence on the superiority of any one assumed causal

Tests of Mediating Effects

259

model (e.g. Fig. 1a) over a host of rival models (e.g. Fig. 1b–f). We consider the reasons for this below.

Validity of Inferences About Causality This section considers two major issues. First, we detail the factors that are essential for demonstrating that one variable causes another. Second, we describe a number of problems that stem from attempts to support causal models using data from non-experimental research. Conditions Vital to Inferences About Cause Although the requirements for demonstrating that one variable causes another have been the subject of some controversy (Cook & Campbell, 1979; McKim & Turner, 1997; Shadish et al., 2002), there are many consistencies in terms of views on this issue. Hume (1740/2000), an eighteenth-century philosopher, described a set of eight proofs (conditions) that are necessary to demonstrate that a presumed cause is responsible for an observed effect. Typically, three of these are mentioned in the contemporary research methods literature (e.g. Cook & Campbell, 1979; Rosenthal & Rosnow, 1991; Shadish et al., 2002). More specifically, they are that: (a) the cause and effect are contiguous in space and time; (b) the cause precedes the effect in time (temporal precedence); and (c) there is a “constant union” between the cause and the effect. Analogous to Hume’s rules, Mill (1884), a nineteenth-century philosopher, argued that inferences about cause require that three conditions be satisfied: (a) the cause precedes the effect in time; (b) the cause and effect are related; and (c) there are no alternative causes of the effect (Cook & Campbell, 1979; Shadish et al., 2002). Even if it were possible to test mediation models with HMR using data from nonexperimental research, the legitimacy of estimates of causal effects would hinge on meeting a number of critical assumptions (Blalock, 1964; Bollen, 1989; Cohen, Cohen, West & Aiken, 2003; Kenny, 1979; Maruyama, 1998; Pedhazur, 1982), including the following: First, the model being tested would have to be properly specified in that the assumed independent variable caused the assumed mediator, the assumed mediator caused the dependent variable, there was no reciprocal causation, and the model being tested included all relevant variables (e.g. sources of spuriousness) and did not contain any irrelevant variables. Second, relations among variables in the model would have to be linear. Third, the variables in the model would have to be free of measurement error. Fourth, and finally, the measured variables would have to be free of systematic bias (e.g. that stemming from common measurement methods). As is detailed below, in the case of almost all

260

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

non-experimental research, very rarely, if ever, could a researcher meet these and other important assumptions. Model Specification Problems in Non-experimental Research In this section we consider a number of factors that threaten the validity of inferences about mediation that are based upon the analysis of data from non-experimental research (Cook & Campbell, 1979; Shadish et al., 2002). All such factors have to do with model specification. Alternative causal models for a given set of variables. In most non-experimental studies it is virtually impossible to make credible inferences about mediation because a number of causal models may be consistent with any set of zero-order correlation coefficients (rs) among a set of measured variables (Alwin & Hauser, 1975; Blalock, 1964; Bobko, 1990; Cliff, 1983; Freedman, 1987, 1997; Kenny, 1979; Ling, 1987; MacCallum, Wegener, Uchino & Fabrigar, 1993; Maruyama, 1998; Pedhazur, 1982). For example, as shown in Fig. 1a, Z2 may mediate the relation between Z1 and Z3 . In addition, as illustrated in Fig. 1b, Z1 may mediate the relation between Z2 and Z3 . What’s more, as portrayed in Fig. 1c, both Z1 and Z2 may be exogenous determinants of Z3 . Moreover, as illustrated in Fig. 1d, Z3 may be the cause of both Z1 and Z2 , making the relation between Z1 and Z2 spurious. Furthermore, as depicted in Fig. 1e, Z2 may be the cause of both Z1 and Z3 , making the relation between Z1 and Z3 spurious. Finally, as shown in Fig. 1f, the non-causal relations among Z1 , Z2 , and Z3 may be a byproduct of the fact that they share the common causes of Z4 , Z5 , and Z6 . The consideration of these models should make it clear that a demonstration that a set of rs is consistent with one of them does nothing whatsoever to discount the plausibility of a host of others. Thus, in deciding to test one model (e.g. Fig. 1a) as opposed to another (e.g. Fig. 1b), the HMR analyses will differ and, quite frequently, so will the conclusions stemming from them. Consequently, erroneous interpretations and inferences are likely in tests of mediation models that are based upon non-experimental research. We provide clear examples of this in sections of this article that follow. Just as the results of HMR tests of mediation that are based upon data from non-experimental research can lead to erroneous inferences about mediation, so can the results of a host of other data-analytic strategies (e.g. cross-lagged correlation, path analysis, SEM). The reason for this is that statistical procedures, however sophisticated they may be, have no capacity whatsoever to provide credible evidence about mediation. As is detailed below, such evidence can only come from appropriately designed and executed experimental studies (Bock, 1975; Cliff, 1983; Cook & Campbell, 1979; Dean & Voss, 1999; Freedman, 1987; Holland, 1986; Kirk, 1995; Shadish et al., 2002).

Tests of Mediating Effects

261

Temporal precedence and internal validity. A key requirement for demonstrating causal connections between variables is that changes in the assumed cause take place prior to changes in the assumed effect. However, with data from non-experimental research, there is virtually never any credible evidence about temporal precedence. Stated differently, the researcher typically does not know the true causal ordering between or among measured variables. For example: (a) Does job satisfaction cause organizational commitment? (b) Does organizational commitment cause job satisfaction? (c) Do job satisfaction and organizational commitment have reciprocal effects on one another? (d) Are job satisfaction and organizational commitment both a function of one or more unmeasured causes (e.g. perceived organizational support, distributive justice, or procedural justice)? In short, because of the fact that the true causal ordering of variables is almost never known, it is generally the case that evidence of covariance among variables from non-experimental studies cannot be used to infer that any one causal ordering is more plausible than any other. Thus, with non-experimental research, tests of mediation models almost always produce equivocal inferences about causality. Cook and Campbell (1979) and Shadish et al. (2002) argued that ambiguous temporal precedence is a major threat to the internal validity of non-experimental research. Because of this, assumptions about causality can never serve as an adequate substitute for experimentally-based inferences about cause and effect. So, with the HMR procedure, it is virtually impossible to determine which variable is the cause, which is the mediator, and which is the effect. As a result, although the true causal sequence among three variables may be that shown in Fig. 1e, a researcher might use the same procedure to test the improperly specified model illustrated in Fig. 1a. The same ill-conceived test would result in Z2 being partialed from the relationship between Z1 and Z3 . In this case the researcher would inappropriately control for the variance of the true cause (Z2 ). Reciprocal causation in non-experimental research. The legitimacy of HMR inferences about mediation is highly suspect in cases where there is reciprocal causation. For example, consider a causal sequence in which a supervisor made a hostile remark to a subordinate (independent variable), leading to poorer leader-member exchange (mediator variable) and this led to a hostile remark by the subordinate that was directed at the supervisor (dependent variable). For the next cycle, the subordinate’s hostile remark would be the independent variable that would lead to even poorer leader-member exchange (mediator variable), and a subsequent hostile remark by the supervisor (the dependent variable). At any given stage in the process, a given variable’s status might switch from dependent to independent. Thus, even if one non-experimental study provided evidence consistent with a particular causal sequence, this would not rule out the plausibility of other

262

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

such sequences. This would further complicate inferences about the direction of causality. Unmeasured confounds. In non-experimental research, it is impossible to rule out the possibility that the relations among variables in a mediation model are a function of one or more unmeasured confounds. And, to the degree that a model being tested fails to consider all such confounds (i.e. it is misspecified), estimates of relations among the variables in the model will be biased. Unfortunately, researchers are seldom, if ever, aware of all relevant confounds (Cook & Campbell, 1979; Freedman, 1987; Shadish et al., 2002). In addition, even if all were known, trying to control for their effects statistically would present a host of problems. For example, because of statistical power considerations, the larger the number of confounds in a model being tested, the more difficult it would be to find support for the model. It deserves noting that the Baron and Kenny (1986) HMR procedure makes no allowance for confounds. However, assuming that a researcher used a modified version of it that provided for the statistical control of confounds, there would still be a very serious problem. More specifically, statistical control is not an adequate substitute for control afforded by experimental research (Clogg & Haritou, 1997; Cook & Campbell, 1979; Rogosa, 1987; Shadish et al., 2002; Stone-Romero, 2002). Note, in addition, that no statistical control technique can convert data from a non-experimental study into data from an experimental study. Thus, inferences about mediation that are based upon data from non-experimental research are always more tenuous than those stemming from experimental research. Implications of testing a misspecified model. Assuming that a researcher made inferences about mediation on the basis of applying the HMR procedure to data from a non-experimental study, what would happen if his or her assumed causal model were misspecified? For example, although the correct causal model was that shown in Fig. 1c, the model tested by the researcher was that illustrated in Fig. 1a. This is a non-trivial issue, because as noted above, there are numerous instances in which researchers have tested seemingly misspecified models. In such cases, incorrect assumptions about the model being tested would have very serious implications for not only the outcomes of model testing, but also for the development of sound theory and the conduct of practice (Maruyama, 1998; Pedhazur, 1982). We provide clear evidence of this below in the section titled Interpretation Problems From Applying the HMR Procedure to a Misspecified Model. Validating assumptions in tests of mediation models. As should be obvious from the above, with non-experimental research, researchers can never be certain that

Tests of Mediating Effects

263

their assumptions about a given model are valid. Moreover, given the limitations of such research, there is no satisfactory method of validating the same assumptions. Thus, in non-experimental research, internal validity inferences are always suspect (Cook & Campbell, 1979; Shadish et al., 2002). Model specification perspectives considered in this article. To deal with the model specification issue, in the sections that follow, we consider two HMR analysis perspectives. The first is that of using HMR to test a properly specified mediation model (PSMM), whereas the second is that of an HMR test of an improperly specified mediation model (ISMM). Note that details on the model specification issue are provided in numerous works (e.g. Blalock, 1964; Bollen, 1989; Cohen et al., 2003; Maruyama, 1998; Pedhazur, 1982).

Statistical Estimates and Significance Tests in HMR Analyses ˆ and R ˆ 2 s, we Because users of HMR base conclusions about mediation on ␤s specify how (a) the pattern of rs among variables and (b) N affect these estimates and tests of their statistical significance. We deal with these issues in terms of a simple, three-variable mediation model (e.g. that shown in Fig. 1a). Regression Coefficients in Simple Mediation Models ˆ for the aboveAssuming an HMR test of the model shown in Fig. 1a, the ␤s described Analysis 3 are totally determined by the rs among Z1 , Z2 , and Z3 . In the case of the same figure, r13 is the correlation between the assumed independent and dependent variables, r23 is the correlation between the assumed mediator and dependent variables, and r12 is the correlation between the assumed independent ˆ for Z1 and Z2 are determined as follows: and mediator variables. The ␤s r 13 − r 23 r 12 ␤ˆ 1 = 1 − r 212

(4)

r 23 − r 13 r 12 ␤ˆ 2 = 1 − r 212

(5)

As is demonstrated below, variations in N and the magnitudes of r12 , r13 , and r23 can lead to markedly different conclusions about the existence of mediation, thus resulting in suspect inferences about it. In order to show this, in the simulation study that is described below, we manipulated the magnitudes of three zero-order rs (i.e. r12 , r13 , and r23 ) used in the HMR-based test of an assumed mediation ˆ needed to determine model (e.g. Fig. 1a). Using the same rs, we calculated the ␤s

264

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

if there was support for the above-described conditions for inferring partial or complete mediation. Then, we assessed the effects of N and the manipulated rs on ˆ and R ˆ 2 s stemming from HMR tests of mediating effects. the ␤s Coefficient of Multiple Determination ˆ and the rs, R ˆ 2 can be estimated by: Knowing the values of the just-described ␤s ˆ 2 = ␤ˆ 1 r 13 + ␤ˆ 1 r 23 R

(6)

Models with More than Two Predictors In instances of models with more than two assumed independent variables and/or two or more mediators, mediation models can be tested with a number of other procedures (see, for example, Cohen et al., 2003; Kenny, 1979; Maruyama, 1998; Pedhazur, 1982). A model involving three assumed exogenous variables, two mediators, and one dependent variable is depicted on pp. 70–73 of Kenny (1979). It deserves noting that in the case of tests of models involving multiple assumed ˆ and R ˆ 2 s can be determined by (Cohen et al., causes and/or multiple mediators, ␤s 2003; Pedhazur, 1982): ˆ = R −1 r xy ␤ xx R = ˆ2

r xy R −1 xx r xy

(7) (8)

ˆ is a column vector of ␤s, ˆ R −1 Here, ␤ xx is the inverse of the correlation matrix for the predictors (i.e. all assumed independent and mediator variables), rxy is a column vector with rs between the assumed dependent variable and each of the predictors, and r xy is the transpose of the same vector. Factors that Influence the Results of Significance Tests in HMR Tests of Mediation As noted above, in order to meet NBI Condition 1, Analysis 1 must show that ␤1(1) = 0. In addition, to satisfy NBI Condition 2, Analysis 2 must demonstrate that ␤1(2) = 0. In view of this, it is instructive to consider two factors that influence ˆ and R ˆ 2 s. the statistical significance of ␤s Magnitude of the regression coefficient. One factor that determines the statistical significance of a ␤ˆ is its magnitude. All else constant (e.g. N and the Type I error ˆ the greater the likelihood of it being statistically significant. rate), the greater the ␤, Magnitude of the squared multiple correlation coefficient. The statistical signifiˆ 2 , the more ˆ 2 is a function of its magnitude. Ceteris paribus, the larger R cance of R likely it is to be statistically significant.

Tests of Mediating Effects

265

ˆ and R ˆ 2 s is a function of Sample size. The statistical significance of both ␤s N. Ceteris paribus, the greater the N, the greater the likelihood of them being statistically significant. The very important implication of the foregoing is that all else constant, as N and/or effect size decrease, it will become increasingly difficult to meet NBI Conditions 1 and 2. However, ceteris paribus, as N increases, it will become increasingly difficult to satisfy NBI Condition 3a. Magnitudes of correlation coefficients. An inspection of Eq. (4) makes it clear that when the rs among Z1 , Z2 , and Z3 are non-zero and positive, it is very likely that |␤ˆ 1(3) | < |␤ˆ 1(2) |, satisfying NBI Condition 3b. The reason for this is that in the same equation, the product of r23 and r12 is subtracted from r13 . Thus, for a number of reasons, it is inappropriate to interpret the finding of |␤ˆ 1(3) | < |␤ˆ 1(2) | as evidence of mediation. For example, as is illustrated in the below-described Scenario 3, if instead of testing the PSMM of Fig. 1c, a researcher tested the ISMM of Fig. 1a, he or she would make an incorrect inference about mediation. Partitioning of Explained Variance in Regression Analyses Even if the HMR procedure were used to test a PSMM, inferences about the variance explained by the independent and mediator variables would be problematic, ˆ stemming from the HMR analysis. Darlington as would be interpretations of the ␤s (1968) provided a compelling set of arguments on this issue. More specifically, in cases where predictor variables (e.g. Z1 and Z2 ) are correlated with one another, it is impossible to determine the proportion of variance in the predicted variable (e.g. Z3 ) that is explained uniquely by each of the predictors. In addition, the stronger the rs among predictor variables, the greater the ambiguity in basing ˆ This has a very important inferences about the importance of variables on ␤s. implication for HMR tests of mediation. Such tests rely on the existence of non-zero (often moderate to large) rs among variables. Thus, for example, for the model shown in Fig. 1a, the greater the degree of correlation between Z1 and Z2 , the greater the ambiguity in interpreting the amount of variance explained by each of these variables. This is true for both: (a) the above-described HMR analyses; and (b) the analyses associated with any other covariance-based procedure for inferring the direct and indirect “effects” of variables in non-experimental research (e.g. path analysis, SEM). Unfortunately, many users of such procedures seem unaware of the profound implications of Darlington’s (1968) important arguments.

266

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Interpretation Problems from Applying the HMR Procedure to a Misspecified Model In non-experimental research, there is no assurance that a HMR test of mediation is based upon a PSMM. Instead it may be based upon a ISMM. This is a critical issue because the test of a ISMM can lead to very serious inference problems. In this section, we illustrate several such problems. More specifically, using a set of rs among three variables (i.e. Z1 , Z2 , and Z3 ), we apply the HMR procedure to the models associated with four scenarios that are described below. For the purpose of these scenarios, the relations among the variables are as follows: r 13 = 0.3, r 23 = 0.5, and r 12 = 0.4. Sample size (N) is not considered because (a) N is irrelevant ˆ and (b) our primary purpose here is to show the effects to the computation of ␤s, ˆ and inferences about mediation. of model specification on ␤s In the case of Scenarios 2–4, we illustrate the inference problems that result when the HMR procedure is applied to a ISMM (i.e. Fig. 1a). We do so by comparing the inferences that would result from HMR tests of: (a) a ISMM; and (b) a PSMM. Scenario 1: Hypothesized Model Same as True Model For Scenario 1, we assume that a researcher applied the HMR procedure to the PSMM shown in Fig. 1a. For this scenario, Analysis 1 would result in ␤ˆ 1(1) = 0.4, Analysis 2 would lead to ␤ˆ 1(2) = 0.3, and Analysis 3 would lead to ␤ˆ 1(3) = 0.119 and ␤ˆ 2(3) = 0.452. In addition, because |␤ˆ 1(3) | < |␤ˆ 1(2) |, the researcher would infer the existence of partial mediation. For the model shown in Fig. 1a, the 0.3 correlation between Z1 and Z3 results from two types of effects: (a) the direct effect of the independent variable (Z1 ) on the dependent variable, which would equal ␤ˆ 1(3) , and (b) the indirect effect of Z1 on Z3 , which would equal the product of ␤ˆ 1(1) and ␤ˆ 2(3) ; that is, 0.3 = 0.119 + (0.4 × 0.452). For details on the estimation of direct and indirect effects see, for example, Cohen et al. (2003), Kenny (1979), Maruyama (1998), and Pedhazur (1982). Scenario 2: True Model is Fig. 1b In the case of Scenario 2, the PSMM is that shown in Fig. 1b. The results of HMR analyses for this model can be compared to the results of HMR analyses for what in the case of this scenario is the ISMM shown in Fig. 1a. For the PSMM, Z1 mediates the relation between Z2 and Z3 . In this case, the application of the HMR ˆ that are similar to those for the ISMM of procedure to the PSMM would yield ␤s Fig. 1a, but their interpretation would differ markedly.

Tests of Mediating Effects

267

Let’s consider the results for tests of the PSMM and the ISMM. In Analysis 1 for the PSMM of Scenario 2, Z1 would be regressed on Z2 and would result in ␤ˆ 2(1) = 0.4. Note that this equals ␤ˆ 1(1) from the test of the ISMM. However, in the case of Scenario 2, the researcher who tested the ISMM would make a very serious inferential error. More specifically, although Z2 was actually the cause of Z1 , he or she would conclude inappropriately that Z1 caused Z2 . Analysis 2 for the PSMM of Scenario 2 would result in ␤ˆ 2(2) = 0.5. For the ISMM, Analysis 2 would involve the regression of Z3 on Z1 , leading to ␤ˆ 1(2) = 0.3. Note that this analysis would assume mistakenly that Z2 was a mediator. Analysis 3 for the PSMM of Scenario 2 would lead to ␤ˆ 1(3) = 0.119 and ˆ for Z1 and Z2 would be identical in value to those ˆ␤2(3) = 0.452. Although the ␤s produced by the test of the ISMM, their interpretation would differ radically. More specifically, for the PSMM of this scenario, the coefficient of 0.119 represents the effect of the mediator (Z1 ) on Z3 when controlling for Z2 , whereas for the ISMM 0.119 represents the unmediated (direct) effect of Z1 on Z3 . For Scenario 2, Analysis 3 for the PSMM would result in ␤ˆ 2(3) = 0.452, which would be less than ␤ˆ 2(2) = 0.5, leading to the inference that Z1 mediates the relationship between Z2 and Z3 . Clearly, this inference is inconsistent with the inference stemming from the test of the ISMM, i.e. that Z2 mediates the relation between Z1 and Z3 . Scenario 3: True Model is Fig. 1c In the case of Scenario 3, the PSMM is that of Fig. 1c. In it, Z1 and Z2 are both exogenous causes of Z3 and there is no mediation. In addition, although Z1 and Z2 are correlated, the relation is non-causal. The single analysis for the PSMM would involve the simultaneous regression of Z3 on Z1 and Z2 . It would result in ␤ˆ 1(1) = 0.119 and ␤ˆ 2(1) = 0.452; ␤ˆ 1(1) would represent the effect of Z1 on Z3 , controlling for Z2 , whereas ␤ˆ 2(1) would represent the effect of Z2 on Z3 , controlling for Z1 . Note that the regression analysis for the PSMM would not involve any test for mediation. However, a researcher who performed HMR analyses based upon the ISMM would make a number of invalid inferences, one being that Z2 mediates the relation between Z1 and Z3 . Scenario 4: True Model is Fig. 1e In the PSMM for this scenario (Fig. 1e), Z2 causes both Z1 and Z3 and there is no mediation. For this model, the regression of Z1 on Z2 would result in ␤ˆ 2(1) = 0.4, and the regression of Z3 on Z2 would lead to ␤ˆ 2(1) = 0.5. The results of testing the ISMM (Fig. 1a) and the PSMM would lead to markedly different inferences

268

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

about mediation. The test of the ISMM would lead to the incorrect inference that Z2 mediates the relation between Z1 and Z3 . However, the test of the PSMM would lead to the correct inference that Z2 is a common cause of Z1 and Z3 , and that the 0.3 correlation between Z1 and Z3 is a consequence of them sharing this common cause. Summary A comparison of the results of testing the above-described models shows that the ˆ for many specific variables would not change. However, their interpretation ␤s would differ markedly. Thus, the just-considered illustrations show quite clearly that inferences about mediation can be highly inappropriate when the model being tested by the HMR procedure differs from the PSMM. For a ISMM, specification error makes it highly likely that the same procedure will lead researchers to make very erroneous inferences about mediating effects. Unfortunately, researchers in many academic disciplines, including personnel and human resources management, appear to be unaware of this fact. In addition, as the below-described results of our simulation show, because of the effects of N on significance tests, for a PSMM the sequential comparison of ˆ typically will fail to provide credible evidence about mediation. Furthermore, ␤s because of the inability of multiple regression analyses to provide evidence on the importance of correlated predictor variables (Darlington, 1968), it is impossible ˆ stemming from HMR Analyses 1, 2, and to provide clear interpretations of the ␤s 3. This is made clear by the consideration of just-described Scenarios 2–4.

A Simulation Study to Assess the Ability of HMR to Provide Evidence of Mediation To provide additional empirical support for the view that the HMR procedure does not serve as a basis for valid inferences about mediation, we conducted the simulation study described in this section. It was concerned with tests of mediation models in terms of two perspectives, i.e. ISMM and PSMM. Two Mediation Model Perspectives To address the ISMM perspective, we interpreted the simulation’s findings as if they had been based upon analyses of data for a ISMM. More specifically, we assumed that the PSMM (i.e. Fig. 1e) underlying a set of rs among Z1 , Z2 , and Z3 was other than the model that was tested (i.e. Fig. 1a). For the ISMM perspective, the findings were considered in terms of what would happen if the HMR procedure had been applied to the ISMM of Fig. 1a.

Tests of Mediating Effects

269

The ISMM perspective is well worth considering because, as noted above, in the case of non-experimental research, it is typically the case that researchers are unaware of the true causal model underlying a set of rs. Thus, their tests of mediation are almost always based on assumed causal models, many of which are incorrect. Among the many instances of such tests are the above-noted studies that dealt with the assumed causal connections between job satisfaction and performance. Simulation Study Overview In order to demonstrate the many problems that result from HMR tests of mediation, we conducted a simulation in which four variables were manipulated concurrently: r12 , r13 , r23 , and N. If Fig. 1a were a PSMM: (a) r13 would be the correlation between the independent (Z1 ) and dependent (Z3 ) variables; (b) r23 would be the correlation between the mediator (Z2 ) and dependent (Z3 ) variables; and (c) r12 would be the correlation between the independent (Z1 ) and mediator (Z2 ) variables. Note that for a PSMM, the same rs would reflect the true causal connections among Z1 , Z2 , and Z3 . However, for a ISMM they would not. The simulation assessed the effects of the just-noted manipulations on the capacity of the HMR procedure to provide evidence of mediation. This was indexed by the degree to which the results of HMR tests satisfied the above-described conditions for either partial or complete mediation. Magnitudes of Manipulated Correlation Coefficients In the simulation, the magnitudes of r13 , r23 , and r12 were varied from 0.1 to 0.9 in increments of 0.1 to simulate a very wide range of relations. With nine possible values for each such coefficient there were 729 sets of rs. Sample Size (N) Conditions The simulation involved 13 levels of N. To simulate relatively small Ns, values of 20 and 34 were used. Six other Ns were used that were multiples of 68 (i.e. 68, 136, 204, 272, 340, and 408). The choice of 68 was based on the fact that Lent, Aurbach and Levin (1971) reported that it was the median N for 1,500 validation studies. In addition, to simulate relatively large Ns, values of 1,000, 1,500, 2,000, 2,500, and 3,000 were used. Sample sizes such as these are often found in research using national probability samples. Excluded Cases ˆ we encountered 78 problematic cases. For In estimating the above-described ␤s, ˆ these, the pattern of rs resulted in R2 s > 1. Sets of coefficients that produced this problem were omitted from the primary analyses that are described below.

270

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Analyses Ordinary least squares regression was used in primary and secondary analyses. Their nature is considered next. Primary analyses. Using the 651 just-described sets of rs, we calculated: (a) ␤ˆ 1(1) , ˆ 2 using Eq. (6). Statistical ␤ˆ 1(2) , ␤ˆ 1(3) , and ␤ˆ 2(3) using Eqs. (4) and (5); and (b) R ˆ ˆ 2 s using a Type I error significance tests were conducted for both the ␤s and the R ˆ rate of 0.05. The significance tests for the ␤s were one-tailed. There were a total of 8,463 primary analyses. More specifically, there was an analysis for each of 651 sets of rs at each of 13 levels of N. On the basis of the results of the primary analyses, we determined the proportion of times there was evidence that was consistent with the above-noted conditions for inferring either partial or complete mediation. Secondary analyses. Information from the primary analyses was used in two secondary analyses. More specifically, to determine the degree to which the manipˆ of the primary analyses, we separately regressed ulated variables affected the ␤s ˆ␤1(3) and ␤ˆ 2(3) on the values r13 , r23 , and r12 . The N for the secondary analyses was 651. The sampling distribution for r is not symmetric. Thus, we also conducted the just-described secondary analyses using Fisher’s Z transformed values of rs (Fisher, 1915). Below, we present the findings of the secondary analyses for both the original and transformed rs. ˆ for the secondary analyses are identified using a subscript of S. For The ␤s example, for the analysis in which ␤ˆ 1(3) was regressed on the values of r13 , r23 , and r12 , the ␤ˆ for r13 is denoted by s ␤ˆ 31 . Alternative interpretations of findings. Before turning to the findings of the simulation study, we repeat a very important point: As noted above, they can be interpreted in terms of what would have resulted had Fig. 1a: (a) been the PSMM; and (b) been a ISMM and the actual causal model were that of another figure (e.g. Fig. 1e). Findings of the Simulation Study Correlations Among Variables The means, standard deviations, and correlations among the manipulated and computed variables are presented in Table 1. As can be seen in it, rs among the manipulated variables were all very low in magnitude and not statistically significant.

Tests of Mediating Effects

271

Table 1. Means, Standard Deviations, and Correlations Among Variables. Variable r13 r23 r12 ␤ˆ 1(3) ␤ˆ 2(3)

M

S

r13

r23

r12

0.484 0.484 0.484 0.335 0.335

0.248 0.248 0.248 0.501 0.501

– 0.037 0.037 0.705* −0.369*

– 0.037 −0.369* 0.705*

– −0.099* −0.099*

␤ˆ 1(3)

– −0.863*

␤ˆ 2(3)

–

Note: For the mediation model shown in Fig. 1a: (a) r 13 = correlation between assumed independent and dependent variables; (b) r 23 = correlation between assumed mediator and dependent variables; (c) r 12 = correlation between assumed independent and mediator variables; (d) ␤ˆ 1(3) = standard partial regression coefficient for assumed independent variable from HMR Analysis 3; and (e) ␤ˆ 2(3) = standard partial regression coefficient for assumed mediator variable from HMR Analysis 3. ∗ p < 0.001.

Regression of Estimated Standardized Partial Regression Coefficients on the Manipulated Variables: Secondary Analyses As noted above, in the secondary analyses, the ␤ˆ 1(3) and ␤ˆ 2(3) values from the primary analyses were regressed separately on the rs. We next consider the results for the analyses involving each of these coefficients. Regression of ␤ˆ 1(3) on manipulated correlation coefficients in the secondary analyses. For the secondary analysis using the untransformed rs: (a) s ␤ˆ 31 = 0.723, t = 31.78, p < 0.001; (b) s ␤ˆ 32 = −0.392, t = −17.21, p < 0.001; (c) ˆ 12 = −0.111, t = −4.89, p < 0.001; and (d) R ˆ 2 = 0.67, F(3, 647) = 429.25, s␤ p < 0.001. For the secondary analysis using the Z-transformed rs: (a) s ␤ˆ 31 = 0.690, t = 28.17, p < 0.001; (b) s ␤ˆ 32 = −0.368, t = −15.00, p < 0.001; (c) ˆ 12 = −0.102, t = −4.16, p < 0.001; and (d) R ˆ 2 = 0.61, F(3, 647) = 339.52, s␤ p < 0.001. Note that s ␤ˆ 31 was positive and statistically significant for the secondary analyses based on both the untransformed and the transformed rs. Consistent with Eq. (4): (a) as r13 increased, so did ␤ˆ 1(3) ; (b) as r23 increased, ␤ˆ 1(3) decreased; and (c) as r12 increased, ␤ˆ 1(3) decreased. Also, for both of the secondary analyses, ˆ 32 was negative and statistically significant. Thus, in Analysis 3 for a PSMM: s␤ (a) the greater the correlation between the independent and dependent variables, the greater will be the ␤ˆ for the independent variable; and (b) the greater the degree of correlation between the mediator and dependent variables, the lower will be the ␤ˆ for the independent variable.

272

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Regression of ␤ˆ 2(3) on the manipulated correlation coefficients in the secondary analyses. For the secondary regression analysis using the original rs: (a) s ␤ˆ 31 = −0.392, t = −17.21, p < 0.001; (b) s ␤ˆ 32 = 0.723, t = 31.78, p < 0.001; (c) ˆ 12 = −0.111, t = −4.89, p < 0.001; and (d) R ˆ 2 = 0.67, F(3, 647) = 429.25, s␤ p < 0.001. And, for the secondary analysis using the Z-transformed rs: (a) s ␤ˆ 31 = −0.368, t = −15.00, p < 0.001; (b) s ␤ˆ 32 = 0.690, t = 28.17, p < 0.001; (c) ˆ 12 = −0.102, t = −4.16, p < 0.001; and (d) R ˆ 2 = 0.61, F(3, 647) = 339.52, s␤ p < 0.001. Note that for both secondary analyses, ␤ˆ 2(3) was: (a) negatively related to r13 and r12 ; and (b) positively related to r23 . Thus, in Analysis 3 for a PSMM: (a) the stronger the relations between (1) Z1 and Z3 and (2) Z1 and Z2 , the lower the value of ␤ˆ 2(3) ; and (b) the stronger the relation between Z2 and Z3 , the greater the value of ␤ˆ 2(3) . ˆ Regression coefficient for r12 in the secondary analyses. As was just noted, the ␤s ˆ for r12 were identical for the secondary analyses in which: (a) ␤1(3) was regressed upon the rs; and (b) ␤ˆ 2(3) was regressed upon the rs. More specifically: (a) for the untransformed values of the rs, the ␤ˆ for r12 was −0.111 for both ␤ˆ 1(3) and ␤ˆ 2(3) ; and (b) for the transformed values of the rs, the ␤ˆ for r12 was −0.102 for the secondary analyses for both ␤ˆ 1(3) and ␤ˆ 2(3) . These results are consistent with (and a necessary result of) Eqs. (4) and (5): Because r12 was constant across both ˆ for r12 were the same for both. of the secondary analyses, the ␤s For a PSMM the greater the correlation between the independent and mediator variables, the lower will be the ␤ˆ for the independent variable (or the mediator) when compared with the correlation between the independent variable and the dependent variable (or the mediator and the dependent variable). Further evidence of this can be seen in Table 1. Specifically, the r between ␤ˆ 1(3) and ␤ˆ 2(3) was −0.863, p < 0.001. Model specification issues. All of the above-noted findings were based upon the assumption that the model tested (Fig. 1a) was a PSMM. However, it is vital to recognize that the above-described findings would be identical had a ISMM been tested. Thus, if the PSMM were that of Fig. 1e, the test of the model shown in Fig. 1a would result in a host of inappropriate inferences about mediation. Proportions of Cases Satisfying Baron and Kenny’s (1986) Conditions Table 2 shows the proportion of cases that satisfied various conditions for inferring mediation. Each row shows proportions for a specific value of N.

N

20 34 68 136 204 272 340 408 1,000 1,500 2,000 2,500 3,000

Hierarchical Multiple Regression Support for Condition(s) 1

2

1 and 2

3a

3b

0.6590 0.7757 0.7757 0.8909 0.8909 0.8909 1.0 1.0 1.0 1.0 1.0 1.0 1.0

0.6590 0.7757 0.7757 0.8909 0.8909 0.8909 1.0 1.0 1.0 1.0 1.0 1.0 1.0

0.4424 0.6068 0.6068 0.7957 0.7957 0.7957 1.0 1.0 1.0 1.0 1.0 1.0 1.0

0.4455 0.3625 0.2673 0.1859 0.1536 0.1367 0.1214 0.1152 0.0630 0.0522 0.0461 0.0369 0.0338

0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759 0.6759

3a and 3b 0.3702 0.3118 0.2427 0.1797 0.1505 0.1367 0.1214 0.1152 0.0630 0.0522 0.0461 0.0369 0.0338

3c

3t

C

P

0.5545 0.6375 0.7327 0.8141 0.8464 0.8633 0.8786 0.8848 0.9370 0.9478 0.9539 0.9631 0.9662

0.1628 0.1767 0.1813 0.1398 0.1260 0.1152 0.1029 0.0998 0.0553 0.0476 0.0430 0.0353 0.0323

0.0768 0.1106 0.0952 0.1075 0.0845 0.0768 0.1029 0.0998 0.0553 0.0476 0.0430 0.0353 0.0323

0.0722 0.1490 0.2028 0.3349 0.3779 0.3917 0.5023 0.5115 0.5852 0.6037 0.6144 0.6252 0.6313

Tests of Mediating Effects

Table 2. Effect of Sample Size on the Proportion of Cases Providing Support for Various Mediation Conditions Using the Hierarchical Multiple Regression Procedure.

Notes: Column 1: ␤ˆ 1(1) = 0 in Analysis 1; Column 2: ␤ˆ 1(2) = 0 in Analysis 2; Column 1 and 2: ␤ˆ 1(1) = 0 in Analysis 1 and ␤ˆ 1(2) = 0 in Analysis 2; ˆ 2 = 0 in Analysis Column 3a and 3b: ␤ˆ 1(3) = 0 in Analysis 3 and |␤ˆ 1(3) | < |␤ˆ 1(2) |; Column 3c: ␤ˆ 2(3) = 0 in Analysis 3; Column 3t: 3a, 3b, 3c, and R ˆ 2 = 0 in the same analysis. 3; Column C: Conditions 1, 2, and 3t; Column P: Conditions 1, 2, 3b, 3c, ␤ˆ 1(3) = 0 in Analysis 3, and R

273

274

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Before turning to the results in the same table, we note that of the 8,463 primary analyses, there was evidence of: (a) partial mediation in 3,647 cases (43.09% of all primary analyses); and (b) complete mediation in 630 cases (7.44% of all primary analyses). This is very important because for the ISMM perspective, in all such instances of support for mediation, the model tested (Fig. 1a) was not the true causal model (Fig. 1e). Thus, from the ISMM perspective, the results in Columns C and P of Table 2, respectively, represent the frequency with which a researcher would make invalid inferences about either partial or complete mediation. Column 1 and Column 2. Conditions 1 and 2, respectively, are that ␤ˆ 1(1) = 0 and that ␤ˆ 1(2) = 0. These are both NBI for inferring partial or complete mediation. As the results in Columns 1 and 2 indicate, for N ≥ 136 the proportions are all ≥ 0.8909. Column 1 and 2. This column shows the proportion of cases for which Condition 1 and Condition 2 were satisfied simultaneously. A comparison of entries in this column with those in either Column 1 or Column 2 shows that it was more difficult to satisfy both conditions concurrently than to satisfy either alone. However, for Ns ≥ 136 the odds of satisfying both conditions were ≥ 0.7957. Column 3a. Column 3a shows the proportion of cases that satisfied the NBI requirement for inferring complete mediation that ␤ˆ 1(3) = 0. As can be seen in the same column, as N increased, the proportion decreased. Note, for example, that for Ns ≥ 136, the proportion was ≤ 0.1859 and for N = 3,000 it was a mere 0.0338. Column 3b. Column 3b shows the proportion of cases that provided support for the NBI condition for inferring either partial or complete mediation that |␤ˆ 1(3) | < |␤ˆ 1(2) |. As the results show, irrespective of N, the proportion was 0.6759. Column 3a and 3b. The entries in Column 3a and 3b show the probability of simultaneously satisfying: (a) Condition 3a that ␤ˆ 1(3) = 0; and (b) Condition 3b that |␤ˆ 1(3) | < |␤ˆ 1(2) |. Both such conditions are NBI for inferring either partial or complete mediation. As seen in the same column, as N increased, the proportion decreased. Note, in addition, that for N ≥ 68 the proportion was ≤ 0.2427, and for N = 3,000 it was a mere 0.0338. These results show that evidence of complete mediation is very difficult to obtain. Column 3c. Column 3c shows the proportion of cases that provided support for the condition that ␤ˆ 2(3) = 0, which is NBI for inferences of either partial or complete

Tests of Mediating Effects

275

mediation. As the results in the same column show, for N ≥ 68 the proportion was ≥ 0.7327. ˆ 2 = 0 in Column 3t. Column 3t shows the proportion of cases for which: (a) R Analysis 3; and (b) there was simultaneous support for Conditions 3a, 3b, and 3c. These are NBI conditions for inferences of complete mediation. As the results in the same column show, the proportion followed a non-linear pattern as N increased. However, it was always ≤ 0.1813. In addition, for N ≥ 68, as N increased, it became increasingly difficult to satisfy the same conditions. Column C. This column shows the proportion of cases for which there was support for complete mediation. To show such support requires satisfying Conditions 1, 2, and 3t. As the results in the same column reveal, as N increased, the proportion varied in a non-monotonic fashion, but was always < 0.06 for Ns ≥ 1,000. In addition, in no instance did the proportion exceed 0.1106. Moreover, it was only 0.0323 for N = 3,000. Overall, therefore, for a PSMM it is highly unlikely that HMR tests of mediation will yield results that satisfy all of the conditions for inferring complete mediation. In fact, as previously stated, of the 8,463 primary analyses, there were only 630 cases (7.44%) for which evidence of complete mediation was found. It deserves stressing that although the just-described results were for a PSMM, the same results would have been found had Fig. 1a been a ISMM. In this case, there would have been 630 incorrect inferences of complete mediation. Column P. The entries in Column P show the proportion of times there was support for an inference of partial mediation. To demonstrate such mediation: ˆ 2 = 0; and (b) Conditions 1, 2, 3b, and 3c have (a) Analysis 3 has to show that R to be satisfied simultaneously. As the results in the same column indicate, as N increased, the proportion increased. Note, however, that the results in Column 3b show that irrespective of N, the proportion of cases that satisfied the condition that |␤ˆ 1(3) | < |␤ˆ 1(2) | was 0.6759. Thus, the proportions in Column 3b set a limit on the proportions in Column P. For example, even for N = 3,000, the proportion in Column P was 0.6313. Moreover, as noted above, across the 8,463 primary analyses, evidence of partial mediation was found in only 3,647 cases (43.09% of all cases). A word is in order about the differences between the proportions in Columns P and C. The NBI complete mediation requirement 3a that ␤ˆ 1(3) = 0 is not a requirement for partial mediation. Thus, the values in Column P are always greater than those in Column C, showing that it is easier to demonstrate partial mediation than complete mediation.

276

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

It is important to recognize that the just-described results concerning complete mediation also would have resulted if Fig. 1a had been a ISMM. In this case, there would have been 3,647 incorrect inferences of partial mediation. Basis for the Results Shown in Table 2 The basis for the results in Table 2 can be deduced from the results in a Supplementary Table that shows whether evidence of complete mediation was detected for each of the 8,463 primary HMR analyses. The latter table, which is too large to present here, is available from either author upon request. Each cell in it shows the results of a HMR analysis for a particular set of rs and a specific value of N. The cell contains either a: (a) 0 if all of the requirements for complete mediation were not satisfied; or (b) 1 if all such requirements were satisfied. Results in the Supplementary Table show that for N = 20, the HMR procedure almost never provided support for inferences of complete mediation. That is, virtually all of the cell entries were 0. As a result, in Column C of Table 2, the proportion was a mere 0.0768. In addition, as can be seen in Column 1 and 2 of Table 2, as N increased from 68 to 136, even though the proportion of cases that satisfied NBI Conditions 1 and 2 for inferring complete mediation rose from 0.6068 to 0.7957, the proportion of cases that satisfied all conditions for inferring complete mediation changed very little, i.e. it rose from 0.0952 to 0.1075 (see Column C). Summary From a ISMM perspective, our findings showed that if the model tested by the HMR analyses were not the correct model, there would be numerous instances of support for partial or complete mediation. In all such instances a researcher would make highly erroneous inferences about mediating effects. However, even if one assumes the PSMM perspective, the HMR procedure provided a dearth of evidence for complete mediation and only modest evidence of partial mediation. ˆ In the case of complete mediation, for relatively small Ns, the standard errors of ␤s would be relatively large, resulting in very low odds of satisfying Conditions 1 and 2 simultaneously. However, as N increased, more cases would satisfy both Conditions 1 and 2, but simultaneously would fail to satisfy Condition 3t. Therefore, even for a PSMM, both small and large Ns would result in very small proportions of cases that satisfied all of the conditions for inferring complete mediation. Implications of the Simulation for HMR Tests of Mediation Models As should be clear from the above, very serious problems can arise when data from non-experimental studies are used to test mediation models. Below we comment on the HMR procedure in relation to both the PSMM and ISMM perspectives.

Tests of Mediating Effects

277

The PSMM Perspective Even if the above-noted problems of trying to infer causality on the basis of nonexperimental research did not exist and a researcher had tested a PSMM, our simulation showed that the HMR approach had (a) a very limited ability to produce results that supported inferences of complete mediation, and (b) only a modest ability to provide for inferences of partial mediation. As noted above, in the 8,463 primary analyses there was evidence of: (a) complete mediation in only 630 cases (7.44% of all primary analyses); and (b) partial mediation in 3,647 cases (43.09% of all primary analyses). The ISMM Perspective To deal with the ISMM perspective, our study assumed that the true causal model was that shown in Fig. 1e, but the HRM analyses were for the model shown in Fig. 1a. In this case, the findings of the same analyses would have led to many inappropriate inferences of partial mediation, and to a far lesser extent of complete mediation. The very important implication of this is that the findings of such analyses cannot provide a basis for credible inferences about mediation. In the subsections that follow, we address a number of problems with the HMR procedure that were revealed by our study. These are considered in relation to each of Baron and Kenny’s (1986) conditions for making inferences about mediation. Both the PSMM and the ISMM perspectives are addressed. Satisfying Condition 1. Satisfying Condition 1 requires that ␤ˆ 1(1) = 0. There are two issues associated with satisfying this condition. One has to do with inferences ˆ The other has to do with assumpstemming from a statistically significant r or ␤. tions about the causal ordering of the variables. We address these issues in turn. From the PSMM perspective, the HMR procedure is problematic when applied to studies having very small or large Ns. Ceteris paribus, for any value of ␤ˆ 1(1) that differs from 0, with a sufficiently large N, it will always prove possible to show support for Condition 1. In terms of the ISMM perspective, the HMR procedure is faulty because when N is large enough sufficient to satisfy Condition 1, incorrect inferences are likely. For instance, if the true causal model were that shown in Fig. 1e, but the model tested were that shown in Fig. 1a, Z1 would be erroneously viewed as a cause, when in fact it was an effect. This was demonstrated in the above-described Scenario 4. The use of the HMR procedure or related analyses (e.g. SEM) with data from non-experimental research requires assumptions about both causality and the direction of causal flow. However, as noted above, in such research these assumptions typically are unfounded (Cliff, 1983; Clogg & Haritou, 1997; Freedman, 1997; Ling, 1982; MacCallum et al., 1993; Rogosa, 1987; Shadish et al., 2002;

278

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Stone-Romero, 2002). For example, a finding that ␤ˆ 1(1) = 0 does not speak to the issue of the causal connection between Z1 and Z2 . The reason for this is that the same finding could have resulted from several of the models shown in Fig. 1, including Fig. 1e in which the relation between Z1 and Z2 is non-causal. Moreover, it is possible that a host of unmeasured variables caused scores on Z1 , Z2 , and Z3 to be correlated (see Fig. 1f). This would result in a statistically significant, but spurious value of ␤ˆ 1(1) . Regrettably, the application of the HMR procedure to data from non-experimental research would not allow for spuriousness to be ruled out. Overall, therefore, the finding that ␤ˆ 1(1) = 0 does not allow for credible inferences about causality. Satisfying Condition 2. Condition 2 is that ␤ˆ 1(2) = 0. From the ISMM perspective, for the reasons just mentioned with respect to Condition 1, evidence of a relation between Z1 and Z3 says nothing whatsoever about a causal connection between these two variables. And from the PSMM perspective, assuming that r 13 = 0, with a sufficiently large N, it will always be possible to reject the null hypothesis of ␤1(2) = 0. Thus, satisfying Condition 2 does not provide a sound basis for inferring the existence of mediation. Satisfying various conditions in Analysis 3. In the case of partial mediation, the HMR analyses must show that: (a) ␤ˆ 2(3) = 0, satisfying Condition 3c; and (b) |␤ˆ 1(3) | < |␤ˆ 1(2) | in absolute terms, satisfying Condition 3b. And, in the case of complete mediation, a further requirement is that ␤ˆ 1(3) = 0, satisfying Condition 3a. The results of our simulation are very instructive with respect to the factors that led to |␤ˆ 1(3) | < |␤ˆ 1(2) |. More specifically, from the PSMM perspective, if (a) we assume the model shown in Fig. 1a; and (b) r13 , r23 , and r12 are all positive, Eq. (4) makes it quite clear that any degree of positive correlation between Z1 and Z2 will likely lead to ␤1(3) < ␤1(2) in absolute terms. This is a consequence of the way in which ␤1(3) is computed. Thus, a finding that |␤ˆ 1(3) | < |␤ˆ 1(2) | has no necessary implications for the validity of inferences concerning mediation. The reasons for this are apparent from the findings of our simulation study. More specifically, it showed that the degree to which the ␤ˆ for the assumed independent variable decreased in Analysis 3 was a function of the degree of correlation between the assumed independent variable (e.g. Z1 ) and the assumed mediator (e.g. Z2 ). The very important implication of this is that because the satisfaction of Condition 3b is largely a function of this correlation coefficient (i.e. r12 ), the greater its value, the greater the downward change in ␤ˆ 1 when Z2 is added to the regression equation. It is critical to recognize that this will be true for either a PSMM or a ISMM. We illustrated it in the case of the above-described Scenario 4. Thus,

Tests of Mediating Effects

279

satisfying Condition 3b says nothing whatsoever about the correctness of any given mediation model, and it does not provide a sound basis for inferences about mediation. There is another critical deficiency with Condition 3-based inferences about partial mediation that is relevant to the PSMM perspective: To infer such mediation, the HMR analyses must satisfy the NBI condition that ␤1(3) < ␤1(2) in absolute terms. Thus, all else constant, any difference between these coefficients will allow for an inference of partial mediation. This is problematic for at least two reasons. First, the inference is not based upon a requirement that the change be of any specific magnitude. Thus, even a miniscule change in ␤ˆ could be used to support an inference of partial mediation. Second, there is no requirement that there be a statistically reliable downward change in the ␤ˆ for Z1 . To test for such a change one could, for example, establish a 95% confidence interval around ␤ˆ 1(2) and show that ␤ˆ 1(3) fell below that interval. However, the HMR procedure does not call for this or any other statistical test. Thus, the Condition 3b criterion for inferring partial mediation is a very poor one. Effects of N on inferences about complete mediation. Our simulation showed that as N increased, the odds of ␤ˆ 1(3) = 0 decreased (see Column 3a of Table 2). Therefore, from a PSMM perspective, with a sufficiently large N, NBI Conditions 3a and 3t for demonstrating complete mediation will virtually never be satisfied. For example, for r 13 = 0.4, r 23 = 0.7, r 12 = 0.4, and N > 136, evidence of complete mediation will never be found because Condition 3a will not be satisfied. It is enlightening to consider the basis for the findings in Column C of Table 2, which reflect the proportion of cases that provided support for all of the conditions for inferring complete mediation. As can be seen in the same column, for N = 20, the proportion was a mere 0.0768. The reason for this is that NBI Conditions 1 and 2 for inferring complete mediation were seldom met because ␤ˆ 1(1) and ␤ˆ 1(2) were not statistically significant. This was a direct result of the effect of N on tests of the null hypotheses that ␤1(1) = 0 and ␤1(2) = 0. Interestingly, the results in the Column C of Table 2 show that there is a nonmonotonic relation between N and satisfying all conditions for complete mediation. More specifically, as N increased: (a) more cases satisfied Conditions 1 and 2; but (b) it became increasingly difficult to satisfy Conditions 3a and 3t. For example, as the results in Column 3a indicate, as N increased from 20 to 3000, the proportion of cases for which ␤ˆ 1(3) = 0 dropped from 0.4455 to 0.0338. Additional evidence of the effects of N on meeting all of the conditions for inferring complete mediation is offered in the above-described Supplementary Table. Results in it show that when r 13 = 0.3, r 23 = 0.7, and r 12 = 0.3, for Ns of

280

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

34, 68, or 136, all of the conditions of the HMR procedure for inferring complete mediation were satisfied. However, contrary to NBI Condition 3a, for N ≥ 204, ␤ˆ 1(3) was always statistically significant. In terms of the PSMM perspective, the clear implication of the above is that as N increases, the odds of showing support for complete mediation decrease, approaching 0 for very large values of N. However, as N increases the chances of obtaining evidence of partial mediation increase because of the implications of N for significance tests. Tailoring sample size to find evidence of complete mediation. Assuming one could actually demonstrate complete mediation by applying the HMR procedure to data from a non-experimental study, the N of the study would have to be tailored so as to achieve two objectives: First, it would have to be sufficiently large to allow for ␤ˆ 1(1) = 0 and ␤ˆ 1(2) = 0. Second, it would have to be small enough for ␤ˆ 1(3) = 0. Interestingly, the second requirement is inconsistent with the first. The results in the Supplementary Table provide several examples of the Ns that meet all of the conditions for inferences of complete mediation. For instance, let’s ˆ and (c) r 13 = 0.3, assume: (a) a Type I error rate of 0.05; (b) one-tailed tests of ␤s; r 23 = 0.8, and r 12 = 0.3. In this case, all conditions for complete mediation would be satisfied for 20 < N < 272. Note that N > 20 would be critical to satisfying Conditions 1 and 2, whereas N < 272 would be essential to satisfying Condition 3t. Although the Supplementary Table shows a number of instances in which our study’s results supported inferences of complete mediation, the same table is populated largely by 0s, indicating that even for a PSMM, evidence of complete mediation was very difficult to find. In fact, as noted above, of the 8,463 entries in the table, there were only 630 cases (7.44% of all cases) that supported an inference of complete mediation. Thus, from the PSMM perspective, the HMR procedure has very low odds of supporting inferences of such mediation. Overall, given any pattern of rs in an assumed mediation model, in order to satisfy all conditions for inferring complete mediation, a researcher would have to base his or her study on an ideal N. However, determining it on an a priori basis would be quite difficult. Findings of Mediation from Tests of the Wrong Causal Model As noted above, for the purpose of dealing with the ISMM perspective, we assumed that the correct causal model underlying a set of rs (e.g. Fig. 1e) differed from the assumed causal model (Fig. 1a). However, we based the primary HMR analyses on the model shown in Fig. 1a. The reason for this was that a researcher who used the HMR procedure might very well test a ISMM. In addition, on the basis of the same test, he or she might make a number of inappropriate inferences.

Tests of Mediating Effects

281

Consistent with this, from the ISMM perspective, the results of our simulation showed that in the 8,463 primary analyses, the HMR procedure would lead to 3,647 inappropriate inferences of partial mediation and 630 inappropriate inferences of complete mediation. Regrettably, we believe that inferences of this nature are quite common when the same procedure is used in empirical research in personnel and human resource management and a host of other fields. Thus, it should not be used to test for mediation. Other Issues In this section we deal with several issues that are relevant to tests of assumed causal models. More specifically, we consider: (a) incorrect causal inferences that are common in the literature of personnel and human resource management and a host of other fields; (b) appropriate uses of multiple regression in tests of assumed causal models; (c) problems inherent in tests of assumed causal models that rely on techniques other than HMR; (d) views that several others have expressed about testing assumed causal models using data from non-experimental studies; (e) an appropriate strategy for testing causal models; and (f) the motivation for our critical analysis of the HMR procedure. Inappropriate Interpretations of Tests of Assumed Causal Models A review of published empirical studies in personnel and human resource management and a number of allied fields shows a very large number of articles that suffer from a serious problem: More specifically, although their authors performed tests of assumed causal models using data from non-experimental research, they interpreted the findings of such tests in causal terms. In view of the fact that the studies were non-experimental, both the titles of many such articles and the causal inferences in them are quite inappropriate. For the reasons noted above, in the case of non-experimental studies, all that can be argued legitimately is that the findings of statistical tests are consistent with an assumed causal model. However, any such argument must be accompanied by the caveat that they also may be consistent with a number of other causal models. Unfortunately, it is rare to encounter appropriately circumspect interpretations of tests of assumed causal models. Thus, we offer two important recommendations. First, we view it as essential that authors provide very cautious interpretations of the results of such tests. Second, we strongly encourage the editors of scientific journals to require that authors provide very cautious interpretations of such tests. Careless inferences about cause can have highly dysfunctional consequences for both: (a) the development of theory; and (b) the conduct of practice in any discipline, including personnel and human resource management.

282

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

The Appropriate Use of Multiple Regression in Testing Causal Models It merits noting that although we view the HMR strategy as inappropriate for testing mediation models using data from non-experimental research, multiple regression is a quite appropriate strategy for testing causal models in experimental research. Interestingly, it is typical for researchers to analyze data from such research with analysis of variance (ANOVA). However, there is no need to use ANOVA. For many reasons, multiple regression is an attractive alternative to ANOVA (Cohen et al., 2003; Pedhazur, 1982). Thus, our criticism of HMR has nothing to do with the appropriate use of multiple regression in research. Other Strategies for Testing Assumed Causal Models Just as credible evidence about partial or complete mediation cannot be adduced from applying the HMR procedure to data from non-experimental research, the same holds true for more complex data-analytic strategies that may be applied to such data, including path analysis, cross-lagged panel correlation, and SEM. As noted above, irrespective of the data-analytic strategy used, evidence of correlation between two or more variables in a non-experimental study does not justify any conclusions about causal connections between or among them (Bock, 1975; Cliff, 1983; Cook & Campbell, 1979; Everitt & Dunn, 2001; Freedman, 1987, 1997; Hays, 1994; Holland, 1986; Ling, 1982; Rogosa, 1980; Shadish et al., 2002; Stone-Romero, 2002). Just as a single r cannot provide evidence of causation in an assumed causal model involving two variables, a set of rs cannot provide such evidence in models involving three or more variables. Note, moreover, that the inability to infer causation on the basis of a single r is exacerbated when dealing with two or more rs. One important reason for this is that the larger the number of variables considered by a model, the greater the number of rival models that might be consistent with a study’s data. It deserves adding that neither the promulgation of path diagrams nor the use of HMR or any other data-analytic procedure (e.g. SEM), can alter the fact that internal validity inferences are always problematic in non-experimental research. Thus, even though a researcher might infer that a study’s results are consistent with an assumed causal chain, for the reasons noted above, the validity of this chain can never be demonstrated in a convincing manner. Some Other Views on Causal Modeling Procedures It deserves stressing that we are not alone in terms of our views on the inability of various causal modeling techniques to provide for valid inferences about causality when they are applied to data from non-experimental research (Cliff, 1983; Ling, 1982; Rogosa, 1987). As Cliff (1983) aptly noted, with data from non-experimental research “it is not possible to isolate the empirical system sufficiently so that the

Tests of Mediating Effects

283

nature of the relations among variables can be unambiguously ascertained” (p. 119). In addition, Rogosa’s (1987) paper on “casual” models provides an important perspective on the use of statistical procedures for making causal inferences on the basis of data from non-experimental research: A critical distinction in methodological work is between (a) building (and applying) statistical models for the processes that generate the social [and behavioral] science data and (b) tossing the data at available statistical methods. In my own work I strive for (a) and discourage others from settling for (b). Regrettably, expositions and applications of the popular causal modeling methods (under the various names path analysis, structural equation models, LISREL, etc.) contain much of (b) and little of (a). In fact, my favorite typographical error “casual models” . . . is enjoyable in large part because of its accidental accuracy. And an argument can be made that the methodological proselytizing for and dominance of causal models has retarded the much more useful methodological work of (a) (p. 185).

Rogosa (1987) went on to state that “Attempts to answer experimental . . . research questions with non-experimental data seem fundamentally askew” (pp. 193–194). Finally, in a review of Kenny’s (1979) popular book titled Correlation and Causality, Ling (1982) argued that “causal inference[s] from correlational data (in the absence of controlled experiments) . . . are at best a form of statistical fantasy” (p. 490). Regrettably, a review of the literature shows that it is quite common for researchers in personnel and human resource management and a host of other fields to base inferences about causality on data from non-experimental research, acting as if the same data had come from experimental research. Clearly, as noted above, such inferences are very seldom, if ever, tenable (Bock, 1975; Cook & Campbell, 1979; Dean & Voss, 1999; Freedman, 1987; Ling, 1982; Rogosa, 1987; Shadish et al., 2002). An Appropriate Strategy for Assessing Mediation The only way that one can make credible inferences about mediation is to perform two or more experiments. In the first, the cause (e.g. Z1 ) is manipulated to determine its effect on the mediator (e.g. Z2 ). In the second, the mediator (e.g. Z2 ) is manipulated to determine its effect on the dependent variable (e.g. Z3 ). Evidence from these experiments can be used to support inferences about the causal connections among Z1 , Z2 , and Z3 . Note, furthermore, that if experimental research shows that Z1 causes Z2 , and that Z2 causes Z3 , then on logical grounds it is safe to infer that Z1 causes Z3 . However, we caution that even if: (a) units (e.g. human participants) have been randomly assigned to experimental conditions; (b) independent variables have been manipulated; and (c) relevant outcomes have been measured, unequivocal causal inferences may not be warranted. The reason for this is that a number of factors can detract from the internal validity of experimental studies, including the

284

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

post-randomization attrition of participants from experimental conditions (Cook & Campbell, 1979; Shadish et al., 2002). Nevertheless, causal inferences stemming from experimental research are far more appropriate than those resulting from non-experimental research. Motivation for the Critique of the HMR Procedure A key objective in preparing this article was to highlight a number of problems that detract from the validity of HMR-based inferences about mediating effects when a study’s data are from non-experimental research. Our work follows in the tradition of that of others who sought to improve the methods used by researchers. We offer four examples of such work below. First, articles by both Cronbach (1987) and Dunlap and Kemery (1987) were critical of procedures advocated by Morris, Sherman and Mansfield (1986) for dealing with multicollinearity problems in tests of moderating effects. Morris et al. recommended the use of principal components regression to deal with such problems when testing for interactions using moderated multiple regression. However, subsequent work by Cronbach (1987) and Dunlap and Kemery (1987) demonstrated that multicollinearity problems can be eliminated by centering the elements of the product term prior to conducting a moderated multiple regression analysis. Second, Stone-Romero and Anderson (1994) used simulation methods to demonstrate problems stemming from the use of the subgrouping strategy in tests of moderating effects. Earlier, Arnold (1982) had recommended that moderators be tested by: (a) polychotomizing scores on a continuous variable that was hypothesized to be a moderator; and (b) testing for the equality of subgroup rs. However, Stone-Romero and Anderson’s simulation demonstrated that the subgrouping strategy suffers from numerous problems, one of the most important of which is lower statistical power than moderated multiple regression. Third, Rogosa (1980) demonstrated the many problems that can result from using the cross-lagged correlation strategy for assessing causality on the basis of data from non-experimental research. Pelz and Andrew (1964), Kenny (1979) and others believed that this strategy could be used to make inferences about causality. However, Rogosa (1980) showed it to be of no value whatsoever for this purpose. Fourth, Landis and Dunlap (2000) showed how researchers could make inappropriate conclusions about the existence of moderating effects, depending on the nature of the model that was tested. More specifically, if a researcher were testing for the moderating effect of Z on the relation between X and Y, two assumptions might serve as the basis for a moderated multiple regression analysis: (a) X causes Y; or (b) Y causes X. As Landis and Dunlap demonstrated, analyses based upon these two assumptions may not yield similar results. Thus, if the model being tested is misspecified, erroneous conclusions will be made about moderating effects.

Tests of Mediating Effects

285

In the spirit of the just-cited works, this article documented a number of very serious problems connected with Baron and Kenny’s (1986) widely-used HMR procedure. We hope that our arguments, coupled with the results of our simulation, lead to the eschewal of this procedure for testing mediating effects using data from non-experimental research. General Conclusions For several reasons we believe that the use of the HMR approach for testing mediating effects in non-experimental research must come to an end. We consider two of them below. Findings of Our Simulation Study One reason for calling for an end to the use of the HMR procedure is that the results of our simulation showed quite clearly that what is often viewed as evidence of ˆ and mediation is little more than a necessary consequence of the way in which ␤s 2 ˆ R are determined and their statistical significance assessed. Thus, for example, for a ISMM the same procedure will lead to very inappropriate inferences about mediation. Unfortunately, such inferences can have very serious negative implications for both theory and practice in personnel and human resource management and allied fields. With regard to theory, consider, for example, the fact that decades of nonexperimental research on the job satisfaction-performance relation have led to inconsistent, often contradictory, inferences about the causal connection between these variables. Regrettably, the resources that were devoted to this research were not well used because we are currently in no better a position of knowing the true causal connection between the same variables than we were before all of the studies were conducted. With regard to practice, guided by the specious evidence produced by an HMR test of mediation, practitioners might develop and implement inappropriate interventions, leading to a number of undesirable outcomes (e.g. wasted resources, poor worker attitudes, decreased productivity). For example, if a researcher tested a causal model in which a variable that was actually an effect was posited to be a cause, he or she might advance an inappropriate recommendation that practitioners manipulate the same variable in the hopes of producing desired outcomes. For instance, it would be quite foolhardy to manipulate worker productivity in the hopes of changing worker ability. Regrettably, this and other types of inappropriate manipulations might prove difficult (if not impossible), costly, and counterproductive. In addition, they would not deal with actual causes.

286

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Appropriate Strategy for Establishing Mediation A second reason for recommending that the use of the HMR procedure come to an end is that the research design literature supports the view that it has no potential to provide credible evidence of mediation. As noted above, the only viable strategy for providing such evidence is experimentation (Bock, 1975; Cliff, 1983; Cook & Campbell, 1979; Dean & Voss, 1999; Freedman, 1987; Holland, 1986; Kirk, 1995; Ling, 1982; Shadish et al., 2002). The dubious results stemming from HMR tests of mediation are no substitute for evidence of mediation based upon experimental research. In view of the above, we strongly urge that HMR and a host of allied procedures (e.g. path analysis, SEM) not be used to test for or support inferences concerning mediation.

A Closing Word It is vital that we develop a sound understanding of the mechanisms through which independent variables influence dependent variables. As such, a desideratum is that we learn about the variables that actually mediate relations between independent and dependent variables. However, our understanding of mediation will not advance unless and until we test for it with appropriate strategies. Whereas experimentation is an appropriate strategy for producing credible evidence about mediation, the HMR procedure advocated by Baron and Kenny (1986) and others is not. Thus, its continued use will not contribute to the advancement of knowledge or the improvement of practice in personnel and human resource management or any other field.

ACKNOWLEDGMENTS We thank Joseph J. Martocchio for very helpful comments on a previous version of this article. The order of authorship was randomly determined. The authors made equal contributions to this article.

REFERENCES Alwin, D. F., & Hauser, R. M. (1975). The decomposition of effects in path analysis. American Sociological Review, 40, 37–47. Arnold, H. J. (1982). Moderator variables: A clarification of conceptual, analytic, and psychometric issues. Organizational Behavior and Human Performance, 29, 143–174.

Tests of Mediating Effects

287

Azjen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bateman, T. S., & Strasser, S. (1984). A longitudinal analysis of the antecedents of organizational commitment. Academy of Management Journal, 27, 95–112. Baum, J. R., & Wally, S. (2003). Strategic decision speed and firm performance. Strategic Management Journal, 24, 1107–1129. Birnbaum, D., & Somers, M. J. (1993). Fitting job performance into turnover model: An examination of the job performance-turnover relationship and a path model. Journal of Management, 19, 1–11. Blalock, H. M. (1964). Causal inferences in nonexperimental research. New York: Norton. Bobko, P. (1990). Multivariate correlational analysis. In: M. D. Dunnette & L. M. Hough (Eds), Handbook of Industrial and Organizational Psychology (2nd ed., Vol. 1, pp. 637–686). Palo Alto, CA: Consulting Psychologists Press. Bock, R. D. (1975). Multivariate statistical methods in behavioral research. New York: McGraw-Hill. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Cliff, N. (1983). Some cautions concerning the application of causal modeling methods. Multivariate Behavioral Research, 18, 115–126. Clogg, C. C., & Haritou, A. (1997). The regression method of causal inference and a dilemma confronting this method. In: V. R. McKim & S. P. Turner (Eds), Causality in Crisis? Statistical Methods and the Search for Causal Knowledge in the Social Sciences (pp. 83–112). Notre Dame, IN: University of Notre Dame Press. Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Erlbaum. Collins, C. J., & Clark, K. D. (2003). Strategic human resource practices, top management team social networks, and firm performance: The role of human resource practices in creating organizational competitive advantage. Academy of Management Journal, 46, 740–751. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design and analysis issues for field settings. Boston: Houghton Mifflin. Cronbach, L. J. (1987). Statistical tests for moderator variables: Flaws in analyses recently proposed. Psychological Bulletin, 102, 414–417. Currivan, D. B. (1999). The causal order of job satisfaction and organizational commitment in models of employee turnover. Human Resource Management Review, 9, 495–524. Curry, J. P., Wakefield, D. S., Price, J. L., & Mueller, C. W. (1986). On the causal ordering of job satisfaction and organizational commitment. Academy of Management Journal, 29, 847–858. Darlington, R. B. (1968). Multiple regression in psychological research and practice. Psychological Bulletin, 69, 161–182. Dean, A., & Voss, D. (1999). Design and analysis of experiments. New York: Springer-Verlag. Donaldson, S. I., Graham, J. W., Piccinin, A. M., & & Hansen, W. B. (1995). Resistance skills training and onset of alcohol use: Evidence for beneficial and potentially harmful effects in public schools and in private Catholic schools. Health Psychology, 14, 291–300. Dunlap, W. P., & Kemery, E. R. (1987). Failure to detect moderating effects: Is multicollinearity the problem? Psychological Bulletin, 102, 418–420. Erez, A., & Judge, T. A. (2001). Relationship of core self-evaluations to goal setting, motivation, and performance. Journal of Applied Psychology, 86, 1270–1279.

288

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Everitt, B. S., & Dunn, G. (2001). Applied multivariate data analysis (2nd ed.). London: Arnold. Farkas, A. J., & Tetrick, L. E. (1989). A three-wave longitudinal analysis of the causal ordering of satisfaction and commitment on turnover decisions. Journal of Applied Psychology, 74, 855– 868. Fisher, R. A. (1915). Frequency distribution of the values of the correlation coefficient in samples from an indefinitely large population. Biometrika, 10, 507–521. Frayne, C. A., & Geringer, J. M. (2000). Self-management training for improving job performance: A field experiment involving salespeople. Journal of Applied Psychology, 85, 361–372. Freedman, D. A. (1987). As others see us: A case study in path analysis. Journal of Educational Statistics, 12, 101–128. Freedman, D. A. (1997). From association to causation via regression. In: V. R. McKim & S. P. Turner (Eds), Causality in Crisis? Statistical Methods and the Search for Causal Knowledge in the Social Sciences (pp. 113–161). Notre Dame, IN: University of Notre Dame Press. Freedman, L. S., & Schatzkin, A. (1992). Sample size for studying intermediate endpoints within intervention trials of observational studies. American Journal of Epidemiology, 136, 1148– 1159. Hays, W. L. (1994). Statistics (5th ed.). New York: Harcourt. Holland, P. W. (1986). Statistics and causal inference. Journal of the American Statistical Association, 81, 945–960. Hume, D. (2000). A treatise of human nature (D. F. Norton & M. J. Norton, Eds). Oxford, England: Oxford University Press. (Original work published 1740.) Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38, 635–672. Keaveney, S. M., & Nelson, J. E. (1993). Coping with organizational role stress: Intrinsic motivational orientation, perceived role benefits, and psychological withdrawal. Journal of the Academy of Marketing Science, 21, 113–124. Kenny, D. A. (1979). Correlation and causality. New York: Wiley. Kirk, R. E. (1995). Experimental design: Procedures for the behavioral sciences (3rd ed.). Pacific Grove, CA: Brooks/Cole. Konovsky, M. A., & Pugh, S. D. (1994). Citizenship behavior and social exchange. Academy of Management Journal, 37, 656–669. Korsgaard, M. A., Brodt, S. E., & Whitener, E. M. (2002). Trust in the face of conflict: The role of managerial trustworthy behavior and organizational context. Journal of Applied Psychology, 87, 312–319. Koslowsky, M. (1991). A longitudinal analysis of job satisfaction, commitment, and intention to leave. Applied Psychology: An International Review, 40, 405–415. Lance, C. E. (1991). Evaluation of a structural model relating job satisfaction, organizational commitment, and precursors to voluntary turnover. Multivariate Behavioral Research, 26, 137–162. Landis, R. S., & Dunlap, W. P. (2000). Moderated multiple regression tests are criterion specific. Organizational Research Methods, 3, 254–266. Lent, R. H., Aurbach, H. A., & Levin, L. S. (1971). Research design and validity assessment. Personnel Psychology, 24, 247–274. Ling, R. F. (1982). Review of correlation and causality by David A. Kenny. Journal of the American Statistical Association, 77, 489–491. MacCallum, R. C., Wegener, D. T., Uchino, B. N., & Fabrigar, L. R. (1993). The problem of equivalent models in applications of covariance structure analysis. Psychological Bulletin, 114, 185–199.

Tests of Mediating Effects

289

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83–104. Man, D. C., & Lam, S. S. K. (2003). The effects of job complexity and autonomy on cohesiveness in collectivistic and individualistic work groups: A cross-cultural analysis. Journal of Organizational Behavior, 24, 979–1001. Maruyama, G. M. (1998). Basics of structural equation modeling. Thousand Oaks, CA: Sage. Masterson, S. S., Lewis, K., Goldman, B. M., & Taylor, M. S. (2000). Integrating justice and social exchange: The differing effects of fair procedures and treatment on work relationships. Academy of Management Journal, 43, 738–748. McAllister, D. J., & Bigley, G. A. (2002). Work context and the definition of self: How organizational care influences organization-based self-esteem. Academy of Management Journal, 45, 894–904. McKim, V. R., & Turner, S. P. (Eds) (1997). Causality in crisis? Statistical methods and the search for causal knowledge in the social sciences. Notre Dame, IN: University of Notre Dame Press. Mill, J. S. (1884). A system of logic, ratiocinative and inductive: Being a connected view of the principles of evidence and the methods of scientific investigation (8th ed.). New York: Harper & Brothers. Morris, J. H., Sherman, J. D., & Mansfield, E. R. (1986). Failures to detect moderating effects with ordinary least squares-moderated multiple regression: Some reasons and a remedy. Psychological Bulletin, 99, 282–288. Pedhazur, E. J. (1982). Multiple regression in behavioral research: Explanation and prediction (2nd ed.). New York: Holt, Rinehart & Winston. Pelz, D. C., & Andrew, F. M. (1964). Detecting causal priorities in panel study data. American Sociological Review, 29, 836–848. Pinder, C. C. (1998). Work motivation in organizational behavior. Upper Saddle River, NJ: PrenticeHall. Porter, L. W., & Lawler, E. E. (1968). Managerial attitudes and performance. Homewood, IL: IrwinDorsey. Robinson, S. L. (1996). Trust and breach of the psychological contract. Administrative Science Quarterly, 41, 574–599. Rogosa, D. (1980). A critique of cross-lagged correlation. Psychological Bulletin, 88, 245–258. Rogosa, D. (1987). Casual models do not support scientific conclusions: A comment in support of Freedman. Journal of Educational Statistics, 12, 185–195. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton Mifflin. Shore, L. M., & Martin, H. J. (1989). Job satisfaction and organizational commitment in relation to work performance and turnover intentions. Human Relations, 42, 625–638. Slaughter, J. E., Sinar, E. F., & Bachiochi, P. D. (2002). Black applicants’ reactions to affirmative action plans: Effects of plan content and previous experience with discrimination. Journal of Applied Psychology, 87, 333–344. Sobel, M. E. (1988). Direct and indirect effects in linear structural equation models. In: J. S. Long (Ed.), Common Problems/Proper Solutions (pp. 46–64). Beverly Hills, CA: Sage. Stone-Romero, E. F. (2002). The relative validity and usefulness of various empirical research designs. In: S. G. Rogelberg (Ed.), Handbook of Research Methods in Industrial and Organizational Psychology (pp. 77–98). Malden, MA: Blackwell. Stone-Romero, E. F., & Anderson, L. (1994). Relative power of moderated multiple regression and the comparison of subgroup correlation coefficients for detecting moderating effects. Journal of Applied Psychology, 79, 354–359.

290

EUGENE F. STONE-ROMERO AND PATRICK J. ROSOPA

Stumpf, S. A., & Hartman, K. (1984). Individual exploration to organizational commitment or withdrawal. Academy of Management Journal, 27, 308–329. Tepper, B. J., Lockhart, D., & Hoobler, J. (2001). Justice, citizenship, and role definition effects. Journal of Applied Psychology, 86, 789–796. U.S. Department of Labor (1993). High performance work practices and firm performance. Washington, DC: U.S. Government Printing Office. Vandenberg, R. J., & Lance, C. E. (1992). Examining the causal order of job satisfaction and organizational commitment. Journal of Management, 18, 153–167. Vogus, T. J., & Welbourne, T. M. (2003). Structuring for high reliability: HR practices and mindful processes in reliability-seeking organizations. Journal of Organizational Behavior, 24, 877– 903. Wanous, J. P. (1974). A causal-correlational analysis of the job satisfaction and performance relationship. Journal of Applied Psychology, 59, 139–144. Wiener, Y., & Vardi, Y. (1980). Relationships between job, organization, and career commitments and work outcomes: An integrative approach. Organizational Behavior and Human Performance, 26, 81–96. Williams, L. J., & Hazer, J. T. (1986). Antecedents and consequences of satisfaction and commitment in turnover models: A reanalysis using latent variable structural equation methods. Journal of Applied Psychology, 71, 219–231.

THE EMPLOYEE-ORGANIZATION RELATIONSHIP: A TIMELY CONCEPT IN A PERIOD OF TRANSITION Lynn M. Shore, Lois E. Tetrick, M. Susan Taylor, Jaqueline A.-M. Coyle Shapiro, Robert C. Liden, Judi McLean Parks, Elizabeth Wolfe Morrison, Lyman W. Porter, Sandra L. Robinson, Mark V. Roehling, Denise M. Rousseau, Ren´e Schalk, Anne S. Tsui and Linn Van Dyne ABSTRACT The employee-organization relationship (EOR) has increasingly become a focal point for researchers in organizational behavior, human resource management, and industrial relations. Literature on the EOR has developed at both the individual – (e.g. psychological contracts) and the group and organizational-levels of analysis (e.g. employment relationships). Both sets of literatures are reviewed, and we argue for the need to integrate these literatures as a means for improving understanding of the EOR. Mechanisms for integrating these literatures are suggested. A subsequent discussion of contextual effects on the EOR follows in which we suggest that researchers develop models that explicitly incorporate context. We then examine a number of theoretical lenses to explain various attributes of the Research in Personnel and Human Resources Management Research in Personnel and Human Resources Management, Volume 23, 291–370 © 2004 Published by Elsevier Ltd. ISSN: 0742-7301/doi:10.1016/S0742-7301(04)23007-9

291

292

LYNN M. SHORE ET AL.

EOR such as the dynamism and fairness of the exchange, and new ways of understanding the exchange including positive functional relationships and integrative negotiations. The article concludes with a discussion of future research needed on the EOR.

INTRODUCTION The view that the employee-organization relationship (EOR) is changing has become a truism in both the popular business press and in the scholarly organizational literature. These changes have been due in large part to major transformations in the world of work. Organizations of today reflect a diversity of structures that was rare in the early part of the twentieth century. Organizations are flatter with more permeable boundaries, and often consist of employees who are loosely linked with one another. Work is increasingly structured by projects with natural lifecycles of growth and decline. Technology has increased the speed with which information is transmitted, and has changed the way in which work is conducted. Globalization has increased the complexity of the business environment, causing organizations to seek means of flexibility in a quickly changing and competitive environment. All of these changes have made the study of employee-organization relationships (EORs) increasingly critical, as organizational leaders seek ways to meet the needs of the organization and the people in them. A great deal of research had been conducted on the EOR since the early 1990s. Many concepts, both new and revitalized, have focused on the exchange relationship between the employee and employer, including psychological contracts (Rousseau, 1995), perceived organizational support (Eisenberger, Huntington, Hutchison & Sowa, 1986), leader member exchange (Graen & Uhl-Bien, 1985), and the employment relationship (Tsui, Pearce, Porter & Tripoli, 1997). It is our goal in this article to explore the meaning of the EOR, with particular focus on psychological contracts, and the employment relationship. We use the EOR as an overarching term to describe the relationship between the employee and the organization, with the view that both psychological contracts (an individual-level phenomenon) and the employment relationship (a group-level phenomenon) are more specific cases of the EOR. In this article, we chose to put the spotlight on psychological contracts and employment relationships because: (1) both focus on exchange, highlighting the importance of examining both parties (employee and organization) to the relationship; (2) psychological contracts research has focused on the individual perspective and the employment relationship literature on the group or organizational perspective; (3) both concepts have inspired a great deal of research; and (4) we deem it important to discuss similarities,

The Employee-Organization Relationship

293

differences, and ways of synthesizing these primarily separate literatures in the EOR domain.

EMPLOYEE-ORGANIZATION RELATIONSHIP In the following sections are discussions of the psychological contract literature, followed by the employment relationship literature, and then similarities and differences in these literatures.

Psychological Contracts The last 15 years has witnessed a huge growth in the psychological contracts literature, reflecting the tremendous interest in understanding EORs in the current environment of work organizations. In this section, we first discuss the evolution of the psychological contract concept, then follow with a discussion of similarities and distinctions with two other categories of concepts, including: (1) those reflecting other types of contracts, that is, social, legal, normative and implied contracts; and (2) those focusing on social exchange relationships in the work place, that is, perceived organizational support (POS), leader-member exchange (LMX), and team member exchange (TMX). Thus, we seek to describe the psychological contract domain, including the distinctiveness of this construct, while also exploring potential relationships with other related constructs.

Evolution of the Psychological Contract Concept The definition of the construct “psychological contract” has undergone an evolution of sorts since first being introduced into the management literature by Chris Argyis in his 1960 book Understanding Organizational Behavior. Argyris used the term “psychological work contract” (p. 96), to refer to the relationship between factory line employees and their foremen. He applied the term to the implicit understanding that existed between the two parties to the contract: if the foremen respected the norms of the informal employee culture, the employees would continue satisfactory production with a minimum of problems. The next two commonly cited works credited with expanding the definition of the psychological contract appeared in 1962 and 1965. The first of these is Men, Management, and Mental Health, by Levinson, Price, Munden, Mandl and Solley. Levinson et al. were influenced by Menninger (1958), in which he discussed the

294

LYNN M. SHORE ET AL.

intangible aspects of contractual relationships. Based on an analysis of interviews with 874 employees in an electric utility, Levinson et al. defined the psychological contract as “a series of mutual expectations of which the parties to the relationship may not themselves be even dimly aware but which nonetheless govern their relationship to each other” (p. 21). They argued that mutual expectations were mostly implicit, were often based on the person’s history prior to entering the organization, and change over time. Schein (1965) credited the earlier studies by both Argyis (1960) and Levinson et al. (1962) as providing the input for his conceptualization of the psychological contract. His view centers on the mutual expectations that individuals and organizations have of each other. According to Schein “These expectations not only cover how much work is to be performed for how much pay, but also involve the whole pattern of rights, privileges, and obligations between worker and organizations” (p. 11). He argues that the organization enforces its side of the psychological contract through its authority structure, to which the employee implicitly agrees by joining the organization, and the employee fulfills his side by upward influence. Failure to uphold the contract can endanger the functioning of the organization and lead to a situation where employees no longer cooperate with the authority system. Roehling (1997) provided an extensive historical analysis of the psychological contract literature. He notes that virtually all of the psychological contract studies during the 1970s and 1980s reference Schein’s (1965) book, but that a transition point occurred in 1989 with the publication of Denise Rousseau’s seminal article Psychological and Implied Contracts in Organizations. Rousseau defines the psychological contract as “an individual’s beliefs regarding the terms and conditions of a reciprocal exchange agreement between that focal person and another party. Key issues here include the belief that a promise has been made and a consideration offered in exchange for it, binding the parties to some set of reciprocal obligations” (1989, p. 123). Two important aspects of Rousseau’s definition are those of perceived promises, with the resulting perceived obligations, and perceived mutuality. Rousseau (2001) provides a delineation of actions that may be perceived to constitute a promise. She argues that promises do not necessarily have to be spoken or written in the form of a promise in order to be perceived as such, but can consist of words or actions taken in context. Thus an employee may perceive a promise as having been made without the employer ever having issued an explicit verbal statement of such an intention. Furthermore, Rousseau and McLean Parks (1993) make clear that some expectations are contractual, others are not – they argue that expectations are contractual if they are based on promises. From these perceived promises, perceived reciprocal obligations arise (Rousseau & Tijoriwala, 1998). Rousseau defines obligation as a commitment to some future

The Employee-Organization Relationship

295

action. When entering into the employment relationship the employee realizes that by doing so he or she is accepting an obligation to furnish particular services to the organization as well as follow the directives of management (Blau, 1964). However, the employee perceives that the organization is also obligated to provide certain items in exchange, salary and wages being the most basic of these. Spoken and written communication, as well as actions taken by the organization, provide the employee with input as to the content of what he or she perceives to be promises of some future action to which the organization has obligated itself. Furthermore the employee believes that the agreement is mutually understood whether or not it is so (Rousseau, 1995). There continues to be on-going debate as to the centrality of promises within the psychological contract framework (Roehling, 2004). There are proponents of the view that expectations play a key role, apart from promises, in the development of psychological contracts. There is clearly a need for research examining the potentially different, or redundant effects of promises and expectations on the development of the obligations perceived to constitute the psychological contract.

Comparison with Other Types of Contracts – Social, Legal, Normative, Implied The psychological contract construct has been both linked to, and distinguished from, a variety of other types of contracts: social, legal, normative, and implied. Comparing psychological contracts to these other contract types presents a significant challenge, for several reasons. First, different conceptualizations of the psychological contract construct compare differentially to other contract types. For example, the conceptualization of the psychological contract construct as involving at least some level of agreement or shared understanding between the parties (Levinson et al., 1962; Schein, 1970) bears much closer resemblance to traditional notions of a legal contract than the conceptualization of psychological contacts as involving individual level subjective beliefs (Rousseau, 1989). Second, just as there are multiple views of what constitutes a psychological contract, multiple views of the various contracts with which psychological contracts are compared are found in the literature. In some instances, the differing descriptions reflect an evolution of the contract type. However, there are also examples of contemporaneous competing views of what constitutes legal, social, or implied contracts (discussed below). Third, individual researchers have used terms describing different contract types with varying degrees of clarity and consistency, and in some instances, terms used to describe distinct contract types have been mixed or combined (e.g. “implied social contracts,” Goldberg, 1994). Given the

296

LYNN M. SHORE ET AL.

complexity of the contracts literature, we chose to provide a basic review of four contract types, followed by a discussion that focuses on selected, key similarities and contrasts between psychological contracts and the four contract types. Social Contracts It has been suggested that the psychological contract construct is an extension of all that has been written about social contracts (Schein, 1980). The concept of a social contract originated in the context of philosophical inquiries aimed at prescribing and justifying rights in society. Social contract theorists Hobbes (1651/1958) and Locke (1690/1965) argued that the state was created through a hypothetical agreement among individuals, a “social contract,” that defines rights and obligations in society. Social contract theorists did not explain social contracts as if they were based on actual historical events (e.g. “a meeting of the minds” forming an agreement), but as a statement of obligations logically arising from a specific kind of relationship (Kary, 2000). In contrast, the term social contract has been used more recently in the social sciences, and in the psychological contract literature in particular, to refer to an actual (versus hypothetical), measurable phenomenon: shared normative beliefs. For example, Rousseau (1995, p. 13) defines social contracts as “cultural, based on shared, collective beliefs regarding appropriate behavior in society.” A useful illustration of the two distinct uses of the term social contract can be found in Donaldson and Dunfee’s (1999) Integrated Social Contracts Theory (ISCT), a theory that incorporates both types of social contracts, philosophicalhypothetical and actual-shared normative beliefs, to provide decision makers with a framework for resolving ethical dilemmas. A key component of ISCT is a hypothetical “macro-social contract” that generates ethical “hyper-norms” that are considered morally binding in all situations (e.g. norms reflecting the value of the sanctity of life). Another key component of ISCT is the existing “micro-social contract,” the actual shared normative beliefs in the relevant community. Briefly, the general approach set forth in ISCT involves first identifying the norms, which make up the micro-social contract, and then determining the extent to which those community norms are consistent with the macro-social contract. If a community norm fails to pass this hyper-norm test, then it is not “legitimate” and the decision maker is not obligated to follow the norm. Legal Contracts At its most basic level, a legal contract is simply a contract enforceable by law. In theory and in practice, however, there is considerable variance in what is considered the essential characteristics of a legally enforceable contract. For example, classical contract theories: (1) employ formal reasoning that involve

The Employee-Organization Relationship

297

the application of rigid rules or axioms; (2) emphasize bargained for promises as the basis for contractual obligations; (3) require objective evidence of an offer, an acceptance, and the exchange of valuable consideration; and (4) generally treat contracts as static agreements with terms that were fixed at the moment of formation (Eisenberg, 2001). In contrast, modern contract law theories: (1) recognize obligations based on both promissory and non-promissory principles (e.g. a general duty of good faith and fair dealing); (2) apply a regime of principles that are often individualized and reflect the dynamic nature of exchange relations (i.e. taking into account conduct before and after the time of contract formation and other contextual factors when interpreting the terms of the contract); and (3) are sometimes subjective (e.g. sometimes inquiring into the actual mental state of one or both parties rather than applying an objective “reasonable person standard”; Eisenberg, 2001). Other, less widely discussed contract theories exist in the legal literature. Of particular relevance to the present discussion is Ian Macneil’s theory of relational contract (Macneil, 1972, 1985), a theory that has had a manifest impact on the psychological contract literature (see Rousseau, 1989). According to Macneil, contracts are simply “relations among parties to the process projecting exchange into the future . . . no more or no less” (Macneil, 1980, p. 4). Relational contracting theory explicitly rejects classical contract law’s objective formalism and its focus on promises as the basis for contractual obligations. Rather, relational contract theory posits that most modern contracting behavior involves obligations that arise from the relationship itself (e.g. based on customs, habits, expectations created by the status quo), “with only modest input from individual promise” (Macneil, 1980, p. 17). The impact of relational contracting theory on current contract law is under debate (Spiedel, 2000). Whether or not influenced by the arguments of relational contract theorists, it can be observed that in recent decades American contract law has increasingly recognized and enforced non-promissory, relational obligations among parties to a contract (e.g. obligations inferred from customary practice and usage, obligations recognized in order to avoid the violation of a good faith or fairness norm; Eisenberg, 2000). Normative Contracts The term normative contract has been used to describe circumstances where members of a social group (e.g. a group of employees working in the same organization) hold a common set of beliefs regarding the obligations owed among themselves and to relevant third parties (e.g. their employer; Rousseau, 1995). A normative contract may be viewed as a group level psychological contract, what might be referred to as a common or shared psychological contract. The extent to which normative contracts exist, the factors that influence their formation and their

298

LYNN M. SHORE ET AL.

influence on psychological contracting processes (e.g. reactions to perceived violations) are potentially important, but currently under investigated, areas of research. Implied Contracts The notion of an implied contract was imported into the psychological contract literature by Rousseau (1989), ostensibly from the law of contract. According to Rousseau, implied contracts are the attributions that people not party to contract (i.e. “third parties” or “outsiders”) make regarding its terms, acceptance, and mutuality (Rousseau, 1995, p. 11). This description appears to closely approximate the law’s recognition of implied-in-fact contracts in some situations. Implied-in-fact contracts arise from conduct and circumstances surrounding an exchange relationship that manifest a party’s intention to obligate him or her self (even though that intention was not explicitly stated). The distinction between an express contract vs. one that is implied-in-fact merely lies in the mode of manifesting assent to be bound; the former involve express assent, the latter infers assent based on circumstantial evidence (e.g. past course of dealing). In addition to implied-in-fact contracts, modern contract law also recognizes implied-in-law contracts, or “quasi contracts” (Corbin, 1993). Sometimes described as not involving true contracts, implied-in-law contracts involve obligations that are imposed upon a party out of fairness and equity considerations, regardless of the party’s intent to be bound (Corbin, 1993).

Discussion of Key Similarities and Differences While all of the discussed contract types can be related in some way to psychological contracts, there is considerable variation in the degree of that relation and the potential for the various contract types to contribute to psychological contract theory. Of least relevance to psychological contracting theory is the social contract of philosophers Hobbes and Locke, and more recently the macro social contract of ethicists Donaldson and Dunfee (1999). Unlike psychological contracts, which involve beliefs that are subject to scientific measurement (albeit indirect), the social contract of Hobbes and Locke is a philosophical construction, the result of “thought experiments.” More closely related to psychological contracts, and of much greater relevance to psychological contracting theory, are social contracts described as involving shared normative beliefs (Rousseau & McLean Parks, 1993). This type of social contract differs from psychological contracts on several dimensions. First, it exists at the community level vs. the individual level (Rousseau, 1989) or dyadic level (Levinson et al., 1962; Schein, 1980). Second, closely related to the

The Employee-Organization Relationship

299

level of conceptualization, social contracts involve broad, cultural based beliefs (Rousseau, 1995), whereas psychological contracts are more likely to also include more specific beliefs idiosyncratic to the relationship in question, and based on other factors (e.g. the course of dealing between the parties, individual differences). Third, the social contracts described in the behavioral sciences typically focus on normative expectations. In contrast, conceptualizations of the psychological contract construct are increasingly focusing on a narrower set of beliefs, obligations (e.g. King, 2000; Shore & Barksdale, 1998) or promised-based obligations (e.g. Rousseau & Tijoriwala, 1999; Turnley & Feldman, 2000). Clearly distinguishable from psychological contracts, the primary role of social contracts in psychological contracting theory would appear to be that of a potential antecedent, one of several general factors that may influence the beliefs that make up the psychological contract. Of the contract types discussed, the comparison of psychological contracts with legal contracts is the least straightforward. It has been observed that “[w]hatever a contract is, is bound up in what is legal and/or fair” (Rousseau, 1995, p. 3). The manifest connection between legal contracts and the origin of the psychological contract may explain why researchers attempting to address the fundamental issue of the construct’s discriminant validity take pains to differentiate psychological contacts from legal contracts (Rousseau, 1998). Typically, it is argued that whereas legal contracts require intentions and/or conduct that meet standards of objectively (e.g. acknowledgment by the parties to the contract or a judicial determination that an offer was made and accepted) and are enforceable in court, psychological contacts involve highly subjective beliefs that may, or may not, be enforceable (e.g. King, 2000; McLean Parks & Schmedemann, 1994; Rousseau, 1995). This differentiation best captures the distinction between psychological contacts and classical notions of contract law, which, as the earlier discussion indicated, require objective evidence that formal requirements have been meet (e.g. “an offer,” “an acceptance”) in order to conclude that contractual obligations exist. However, the objective-subjective distinction that is made between legal and psychological contracts is being eroded as contemporary contract law increasingly takes into account the subjective state of mind of the parties (in certain circumstances) and recognizes obligations arising out of exchange behaviors that are not based on bargained for promises (Eisenberg, 2001). As refinements of the psychological contract construct are contemplated with a view toward developing a stronger consensus in the literature, an important question to be addressed is the extent to which the conceptualization of the psychological contract construct should be bound by legal notions of what constitutes a contract. Further, if the construct is to be bounded by legal notions of contract, should theorizing be guided by classical theories of contract law (e.g. a focus on promises as the basis for all contractual

300

LYNN M. SHORE ET AL.

obligations, “mutuality”), or more modern theories of contract found in the legal literature, such as relational contracting theory?

Comparison with Leader-Member Exchange (LMX), Perceived Organizational Support (POS), and Team Member Exchange (TMX) Research and theory development on psychological contracts has evolved with little integration with three other key social exchange-based constructs, leadermember exchange (LMX), perceived organizational support (POS), and team member exchange (TMX). Given that psychological contracts reside within individuals as perceptions of the obligations for reciprocation by both parties of social exchange relationships (Rousseau, 1989), the overlap with LMX, POS, and TMX is readily apparent. Psychological contracts represent employee perceptions of what compensation, resources, and/or support can be expected from an exchange partner in return for meeting perceived job-related obligations. On the other hand, LMX captures perceptions of the quality of the relationship formed with the immediate leader, POS involves perceptions of resources and support actually received from the organization, and TMX represents the general quality of relationships with fellow team members. LMX, POS, and TMX represent perceptions of the actual state of the relationships with the immediate leader, the organization, and teammates respectively. As such, these constructs represent the tenor of the social exchange relationships with these three parties without explication of mutual obligations between the employee and any of these parties (employer, manager, or team members). Just as individuals tend to form multiple social exchange relationships in organizations, they may develop corresponding psychological contracts pertaining to each of these exchange partners. Thus, individuals may establish psychological contracts outlining the expected reciprocation with immediate superiors, teammates, and the organization, as well as perceptions of the actual resources and support received from each of these exchange partners. Most research on psychological contracts in organizational settings has made the assumption that employees develop only one psychological contract – with the organization as a whole. However, Rousseau’s (1989, p. 123) definition of psychological contracts as “an individual’s beliefs regarding the terms and conditions of a reciprocal exchange agreement between that focal person and another party,” opens the door to the possibility that individuals in organizations simultaneously develop multiple psychological contracts.

The Employee-Organization Relationship

301

Psychological Contracts and LMX Early work on LMX portrayed low quality LMX relationships as being based on the formal written employment contract whereas high quality LMX relationships were described as extending beyond the formal contract. For example, Liden and Graen (1980, p. 464) noted that, “members who have high quality exchange . . . have been selected by the supervisor because she/he feels that they can be trusted in dealing with responsibilities going beyond the formal employment contract. A non-written interpersonal contract seems to be negotiated between the leader and the preferred subordinate(s).” Despite this portrayal of high LMX relationships as being based on a psychological contract, LMX has been operationalized as perceptions of the quality of the relationship between supervisor and subordinate with the exchange in leader-member exchange conspicuously missing (Liden, Sparrowe & Wayne, 1997). The closest LMX research has come to actually measuring a psychological contract pertaining to the supervisor-subordinate relationship are studies by Liden, Wayne and Stilwell (1993) and Maslyn and Uhl-Bien (2001). Liden et al. (1993) measured leader and subordinate expectations for the relationship during the first two weeks of the subordinate’s employment in the organization. Items measured the extent to which leaders and subordinates expected their dyadic partners to be high performers and to be similar to themselves in terms of perspective and values. Maslyn and Uhl-Bien (2001) assessed the degree of effort that both leaders and subordinates felt that they and their dyadic partners had put into the relationship. Although such items are tangentially related to psychological contracts, it would be desirable to directly assess psychological contracts by asking subordinates and leaders for their perceptions of the reciprocation obligations that exist in their relationships. Leaders and subordinates could then be asked to describe the exchange of resources and support between leader and member as recommended by Liden et al. (1997). Contract breach could be assessed in two ways: (1) As the actual gap between expectations and perceptions of what is actually being exchanged; and (2) By directly asking leaders and members for their perceptions of the extent to which their expectations of the reciprocation obligations had been met. A final piece to the puzzle of understanding LMX relationships as they pertain to psychological contracts is to ascertain the desired psychological contract. Just as some individuals are low in growth need strength and do not desire enriched/challenging jobs (Hackman & Oldham, 1976), some individuals may not want the type of psychological contract that corresponds to high LMX status (Dienesch & Liden, 1986). They may prefer a low LMX relationship so that they will not be stressed by the pressing obligations that are part of a high LMX psychological contract.

302

LYNN M. SHORE ET AL.

An interesting paradox surfaces when considering that when individuals who desire low LMX relationships find themselves in such relationships, expected obligations on the part of themselves and their supervisors are low, and thus the chance of a contract breach are significantly lower than they are for people who desire and are part of high LMX relationships. Indeed, the higher one’s expectations, the greater the probability of contract breach. Psychological Contracts and POS A parallel may be drawn between LMX and POS. Similar to LMX, POS represents a perception of what is currently being received from the organization, or the quality of the relationship with the organization (Eisenberger et al., 1986; Rhoades & Eisenberger, 2002). A psychological contract with the organization, on the other hand deals with perceived obligations on the part of both the focal individual and the organization. Psychological contracts thus examine mutual obligations of employee and organization, while POS focuses on support expected from the organization. As with LMX, individuals may vary in their desire for high levels of organizational support, as such support comes at a price. Some individuals may prefer low POS so that they are not stressed by the need to fulfill obligations that come with high levels of POS. In addition, individuals desiring more from the organization will be more likely to perceive contract breach. Psychological Contracts and TMX Similar to LMX and POS, TMX refers to the overall quality of relationships with team members (Seers, 1989). Likewise, measures of TMX do not capture what is actually exchanged between a focal individual and team members. Individuals may perceive the existence of a psychological contract with fellow teammates, in terms of expected reciprocal obligations, and may be cognizant of what is actually being exchanged with teammates. As with LMX and POS, contract breach could be assessed in two ways: (1) As the actual gap between expectations and perceptions of what is actually being exchanged between teammates and the individual; and (2) By directly asking members for their perceptions of the extent to which their expectations of the reciprocation obligations with teammates had been met. Finally, as with LMX and POS, individuals may vary in terms of their desire for psychological contracts that correspond to high TMX relationships. Some, especially those who are highly individualistic in their orientation (Earley & Gibson, 1998), may prefer lower TMX so that they do not feel pressured to reciprocate for support received from teammates. Others may simply be reluctant to receive help from others (Bolino & Turnley, 2002).

The Employee-Organization Relationship

303

Integration of Psychological Contracts and Social Exchange Theory Most members of organizations participate in multiple social exchange relationships simultaneously. These include relationships formed with immediate supervisors, co-workers/teammates, and the organization in general. We contend that the psychological contracts corresponding to these three forms of social exchange overlap considerably with one another. For example, obligations to reciprocate to team members may overlap with obligations to reciprocate to immediate leaders for resources and support received. Research is needed that examines psychological contracts pertaining to leaders, teammates, and the organization, as well as perceptions of what is actually received from these exchange partners. It would then be possible to examine the discriminant and convergent validity of these three types of social exchange relationships and corresponding psychological contracts. In addition, antecedents of each type of social exchange relationship should be investigated, extending the work of Eisenberger, Stinglhamber, Vandenberghe, Sucharski and Rhoades (2002), Masterson, Lewis, Goldman and Taylor (2000), and Wayne, Shore and Liden (1997). One such variable may be the individual difference introduced here, degree of desire for high LMX, TMX, and POS.

Summary In sum, this section discussed the chronology of the psychological contract literature, and some differing views of its definition. The psychological contract construct was then compared with other types of contracts, as well as POS, LMX, and TMX. While there are some similarities among constructs, the psychological contract is distinguished by: (1) being an individual-level perception; (2) focusing on mutual obligations; and (3) explicit description of the exchange relationship between employee and employer. Based on this review, we suggest several areas in which the literature could use further development. First, examining how other forms of contracting (normative, social, implicit and legal) contribute to individual-level perceptions of the psychological contract would help to understand how contracts more broadly play a role in the EOR. Second, determining the influence of mutual obligations to multiple organizational agents (supervisor, co-workers, and organization) on employee attitudes and behavior would increase understanding of the role of obligations in the EOR. Finally, greater theorizing about how the social context at work and its various agents may contribute to the psychological contract with the organization is needed. Ongoing discussion about who represents the organization in the EOR highlights the interest in this last topic (Rousseau, 1995; Shore & Tetrick, 1994).

304

LYNN M. SHORE ET AL.

Employment Relationships In this section, the focus is on the employment relationship, the employer’s perspective on the exchange relationship with employees. This literature has more recently emerged, beginning with the conceptual article by Tsui, Pearce, Porter and Hite (1995) in which they discussed the employment relationship as consisting of clusters of human resource practices offered to groups of employees, and employee contributions as a result of such varying practices. In this section, we first discuss the evolution of the employment relationship concept, then follow with a discussion of more current work in the employment relationship. Thus, we seek to summarize the employment relationship domain as specified in the literature. Evolution of the Employment Relationship The historical origins of research on employment relationships can be traced to the work of Barnard (1938) who argued that organizational success was contingent upon the management of the “exchange of utilities” (p. 240). This exchange involved the provision of inducements (pay, power, social support, participation) in return for “quantity and quality of efforts” (p. 230). This explicit exchange framework underpins subsequent work and in particular, that of, March and Simon (1958) and Tsui et al. (1997). Most organization-level theories do not focus on the exchange per se. However, two bodies of research, namely, workforce governance and human resource management systems, have provided the basis for much theorizing about the employment relationship and so a brief review follows. Workforce governance. Etzioni (1961) views organizational governance in terms of compliance relationships that develop based on the type of power applied to organizational members and the resultant type of involvement from organizational members. The basic argument is that when organizations have congruency between the type of power applied and the type of involvement organizational members have (based on their personality, socialization etc.), this fit is likely to result in more effective organizational functioning. The management of employeeemployer relationship through control has received additional attention in the work of Ouchi (1980) and Walton (1985). Although Etzioni (1961) used power as the basis for his typology of compliance relationships, the work of Ouchi and Walton shifted the emphasis towards the specification of organizational characteristics necessary to achieve desired employee behavior (for a detailed review, see Tsui & Wang, 2002). Retaining the thread of power, Mahoney and Watson (1993) classify workforce governance in terms of employees’ ability to influence the nature of the employment relationship. They propose three models: authoritarian, collective bargaining and employee involvement, and examine the conditions under which the three

The Employee-Organization Relationship

305

workforce governance models are likely to be effective in achieving superior organizational functioning and performance. The authoritarian model is posited to be more effective in smaller unitary organizations where the technology is routine and employees are relatively homogenous with relatively little organizational dependence upon their skills and knowledge. The collective bargaining model will be more effective under similar conditions except that the organization is more dependent upon the collective skills and knowledge of employees. Finally, the employee involvement model is more suited to organizations operating in dynamic and changing environments where employees are heterogenous and “beyond contract” contributions are needed. Together, the workforce governance models focus on organizational characteristics, organizational and employee power and their consequences on the extent to which control or commitment underpins the employment relationship. Human resource management systems. The “control/commitment” strategies to workforce governance underpin much of the research on human resource management systems. Specifically, some HRM researchers have sought to specify the types of HR practices associated with a control and commitment approach to HRM (Arthur, 1994; Delery & Doty, 1996; Youndt, Snell, Dean & Lepak, 1996) while other researchers have focused on mapping HRM practices associated with a high commitment/performance approach to HRM (Huselid, 1995; Wood & Menezes, 1998). This research has established the relationship between HRM practices and organizational performance using universalistic, contingency and configurational perspectives (Delery & Doty, 1996). Less researched is an examination of the mechanisms by which HR practices may affect organizational performance. To this end, Moynihan, Gardner and Wright (2002) find empirical support for organizational commitment as a mediator in the relationship between HR practices and performance suggesting that organizational commitment plays an important role in achieving the desired outcomes. Thus, it would seem that investment in employees is likely to be reciprocated through enhanced commitment to the organization. Although inferences can be made from the workforce governance models and human resource management systems in terms of what employees receive in return for their contributions, this is made much more explicit and differentiated in research that draws on the inducement-contribution framework. Inducement-contribution model. March and Simon (1958) further developed Barnard’s (1938) idea of “exchange of utilities” in their inducement-contribution model. March and Simon view the employment relationship as an exchange of organizational inducements for employee contributions. They elaborate by examining balance in the relationship. From the employees’ perspective, satisfaction is

306

LYNN M. SHORE ET AL.

greater when there is a greater difference between the inducements provided by the organization and the contributions required in return. From the organization’s perspective, its continued existence depends on whether the contributions from employees are sufficient to generate the necessary inducements. Consequently, the exchange relationship is characterized as bi-directional whereby employee contributions are contingent upon organizational inducements and vice versa. Drawing on the inducement-contribution model (March & Simon, 1958) Tsui et al. (1997) outline four types of employment relationships that differ on two dimensions: the degree of balance/imbalance in each party’s contributions and whether the focus of these contributions is economic or social. Briefly, a balanced economic exchange (job-focused) occurs when the employer offers short-term purely economic inducements in return for highly specified outcomes. A balanced social exchange (organization-focused) occurs when both parties offer an open ended and long-term investment to each other. The underinvestment approach occurs when the employer expects open-ended commitment and long-term investment from employees in return for short-term economic inducements. The overinvestment approach is categorised by the employer offering long-term investment in return for highly specified employee outcomes. In their empirical study, Tsui et al. (1997) found that a mutual investment employment relationship was positively associated with employee attitudes and performance. Current Views of the Employment Relationship Construct The previous section focuses on the evolution of the concept of the employment relationship since Barnard introduced the ideas of inducements and contributions in 1938. Over the years, a variety of terms and models have been introduced into the literature to describe the nature of the relationship that an employer may have with their employees. It is interesting to see the consistency among the various models and their converging into two prototypes (i.e. organization-focused and job-focused) involving a balance between inducements and contributions, defined from the employer’s perspective (Tsui & Wang, 2002). The current section provides a brief account of the most recent views on the construct as reflected in the publications within the past five years. The literature suggests three major areas of development: the conceptualization of the construct employment relationship, the factors influencing the use of different employment relationship types by the employer, and studies of the employment relationship in an emerging economy – People’s Republic of China. Conceptualization of the construct employment relationship. With the changes in the economic and social environments during the past two decades and the trend continuing into the new century, there is a general agreement that the nature of the

The Employee-Organization Relationship

307

employment relationship between the employee and the employer has undergone dramatic changes as well. The major changes are the removal of traditional job security by the employers, the withdrawal of loyalty by the employees, and the increased flexibility in dealing with each other (Roehling, Cavanaugh, Moynihan & Boswell, 2000). The changes have resulted in the increased use of externalized work arrangements (Pfeffer & Baron, 1988). Given the deep penetration of these practices in many industries, are there corresponding changes in the scholarly views of the employment relationship construct in recent years? We found some revised or refined conceptions of the construct in the recent literature. Specifically, authors have used the terms of employment relationships, employment modes, and employment status to highlight the increased prevalence of non-standard employment arrangements, and its distinction from standard employment relationships. Building on the resource based view Lepak and Snell (1999) introduced four types of employment relationships categorized on the bases of the strategic value and uniqueness of the firm’s human capital. When both the value and uniqueness of the human capital are high, the firm would have an organization-focused relationship with the employees. When both the value and uniqueness are low, the firm would have a job-focused relationship with the employees. When value is high but uniqueness is low, the relationship would be symbiotic which is similar to the organization-focused approach with the exception that both parties would not expect a high level of commitment from each other. When the uniqueness is high but the value is low, the relationship would be a partnership in which trust is critical to encourage the pooling of unique resources and to ensure a longer-term relationship than would be the case in a job-focused relationship. This model of employment relationship types captures the two prototypes that have prevailed in the literature and introduces two new types of the symbiotic relationship and partnership. Lepak and Snell (1999) further specified that firms would move from one relationship type to another when the degree of value or uniqueness of the human resources changes. Thus, this is a dynamic model to accommodate possible changes in the firm’s strategy and the implication of these strategic changes for the value and uniqueness of the firm’s human capital. The authors further distinguished the four types of employment relationships discussed above from four types of employment modes. The latter refers to the manner with which human resources are to be acquired (i.e. internal development, contracting, external acquisition, and alliance) while the former refers to the nature of the relationship after the human resources are acquired. In a later paper, Lepak, Takeuchi and Snell (2003) used the term “knowledge-based” employment to refer to the internal development mode, and “job-based” mode to refer to external acquisition. Their empirical study showed that both the knowledge-based and job-based employment modes were positively associated with firm performance,

308

LYNN M. SHORE ET AL.

and these relationships varied across levels of technological intensity. Lepak and Snell (2002) further confirmed that the strategic value and uniqueness of human capital differ across the four employment modes. Though Lepak and Snell (1999) distinguished employment relationships from employment modes, their empirical tests reported in the 2003 papers used only the modes and not the types. It appears that there may not be the need to have these two separate sets of terms. Employment relationships may subsume employment mode since the former may include how human capital is to be acquired (entry), deployed, or disposed (exit) by the firm. Further work, both conceptual and empirical, is needed to either merge these two concepts (i.e. employment mode and relationship) or to better distinguish them. We can also distinguish the four employment modes discussed above by the employment status of the employee. Under both the knowledge-based (organizational-focused or internal development) and job-based (job-focused or external acquisition) modes, employees have a permanent employment status within the firm in that the termination date of the employment is open-ended. Under the contracting and alliance (or partnership) modes, employees are not permanent members of the organization and the relationship has a fixed termination date. Factors influencing employment relationships. The recent literature also shows continuing interest in analyzing the determinants of the different types of employment relationship. One factor appears to be the use of multi-employer forms (Rubery, Marchington, Coopke & Vincent, 2002). Organizations are changing toward less clear boundaries (Rubery et al., 2002). Many organizations enter joint ventures, alliances, and partnerships. The permeable nature of organizations and the development of multi-employer relationships introduce new problems and ambiguities into the employment relationship. In the traditional organization with a simple hierarchy, the problem exists with defining the “employer” for each employee. In this multi-employer context, the problem is compounded. Who has the right to supervise and control the employee? Who is responsible for handling grievances, conflict and loyalty? How are bargaining units defined? Contradictions, ambiguities and complexity may be particularly serious in the areas of pay, performance management, and organizational commitment. The authors concluded that such complexity might mean the need for a dual approach – increased use of both contracts and permanent status forms of the employment relationship. Romme (1999) introduced self-determination as an alternative to domination for internal control. Self-determination is particularly relevant in organizations that use a different kind of structure, e.g. the circular form of organizing found in some European companies. In such a structure, there is the absence of a single ultimate

The Employee-Organization Relationship

309

authority. Therefore, it has some resemblance to the multi-employer organizations discussed in Rubery et al. (2002). In such an organization, control is exercised through feedback rather than authority. The employment relationship derived is a partnership contract implying an equal division of authority. The distinction between employer and employee reduces or perhaps even ceases to exist, because participants share ownership. This form of control is especially useful in small groups dealing with open-ended, dynamic situations in which new problems and issues frequently emerge. Dominance based control may be more effective in large groups of people facing rather predictable and stable situations. Romme recognizes that “effective governance and control systems would therefore have to incorporate both domination and self-determination as fundamental human concerns” (p. 804). This implies the dual approach of employment relationship discussed by Rubery et al. (2002). Sherer, Rogovsky and Wright (1998) suggest that a need for behavioral control influences organizations’ choice of employment relationship. Based on agency theory, most organizations should prefer a non-employment relationship to eliminate the need for monitoring by the employer. However, Sherer et al. found in their study that taxicab companies in reality used the (permanent) employment relationship more often than leasing or ownership (i.e. externalization). The authors explained this phenomenon by using the resource-based view. The argument is that the employment relationship provides the firm with internal flexibility, enforcement of organizational routines to ensure quality service, and to foster internal cooperation. Therefore, the employers prefer to use the permanent employment relationship and are willing to incur the added cost of employee benefits and payment of employment tax. Lastly, Masters and Miles (2002) found support for the nature of the job influencing the employment relationship. They argued that firms use the permanent employment relationship more often when they need to make repeat hiring for the job, when the job requires firm-specific skills, and when performance in the job is difficult to assess. Their results are consistent with the model by Lepak and Snell (1999) that used the criteria of value (repeated hiring) and uniqueness (firm specific skills) in determining the different types of employment modes and relationships. These studies together suggest that the structure of the organization, job nature, and quality of human capital are the antecedents or determinants of employment relationships in firms. Employment relationships in China. The literature also revealed interest in the employment relationship in the emerging economy of the People’s Republic of China. Frenkel (2001) analyzed the use of codes of practices intended to improve the working condition of workers in two leading athletic shoe buyer-driven

310

LYNN M. SHORE ET AL.

chains and their contractor plants in Southern China. The codes encouraged a less authoritarian and more human resource-oriented approach. These studies suggest a tendency to move away from an “under-investment” approach to a more “job-focused” approach in the relationship between the factory owners and their workers. However, the employment relationship at the factory level remains very basic, focusing on fundamental terms of employment and working conditions. Tsui, Wang and Zhang (2002) studied the employment relationship of firms with their middle managers in a variety of firms in China. This study adopted the inducement-contribution framework and defined the employment relationship as the balance between expected contributions and offered inducements, from the employer’s perspective. The authors developed two forms of expected contributions (professional managerial duties and loyalty) and two forms of inducements (developmental rewards and materialistic rewards). They found differences in employment relationship practices between the state owned and the non-stated owned firms in China. Specifically, state owned firms expected more loyalty and offered fewer developmental rewards. In contrast, the foreign firms and the private domestic firms expected more managerial professional duties from the middle managers but only the foreign firms offered more developmental rewards. These studies in China offer an important starting point for future comparative research on employment relationships in different countries. Another study (Wang, Tsui, Zhang & Ma, 2003) employed the cluster analytic technique and replicated the typology of employment relationships by Tsui et al. (1997). The authors tested the moderating role of business strategy and the firm’s ownership type on the relationship between employment relationship type and firm performance. They found the relationship between the organizational focused (or mutual investment) approach and firm performance to be negative for firms using the prospector strategy but positive for firms pursuing the defender strategy. Further, the relationship is positive for the state owned enterprises but negative for the private domestic firms. The results suggest the meaningfulness of the employment relationship concept in this emerging economy and paved the foundation for comparative research in the future. The Employment Relationship and Psychological Contracts Contrasted In this section of the article, we discussed early and contemporary views of the employment relationship, reflecting changes from Barnard (1938) and March and Simon’s (1958) statements of organizational inducements such as wages and job security and employee contributions in skills and effort to the more recent focus on job based relationships emphasizing short term relationships and high performance vs. organizational based relationships requiring long-term relationships and commitment. Conversely, the prior section presented another major stream of research

The Employee-Organization Relationship

311

on the employment exchange, the psychological contract, and noted a change in the contract’s conceptualization from its early appearances in the management literature (Levinson et al., 1962; Schein, 1965), i.e. a set of mutual expectations regarding pay, rights, privileges and obligations that both the organization and employee hold for one another in the context of the employment exchange, to beliefs concerning the terms and conditions of the employment exchange (Rousseau, 1990, 1995). Given our article’s focus on providing a broad and integrated portrayal of the EOR over time, it is useful to consider how the psychological contract and employment relationship are in fact similar and different from one another.

Similarities Between the Two Literatures Several important similarities between the two concepts and their respective literatures are important to highlight. First, both view the employing organization, as the buyer of the employee’s labor and the party who has greatest control over procedures and outcomes in the workplace, as generally having greater power than the employee to establish the terms of the contract as the basis of the exchange. For example, Schein (1970) argues the importance of having a psychological contract where the employee retains sufficient power to exert some influence over the nature of the contract terms and Rousseau (2001) cites the preponderance of unilateral and large scale contract changes by the employing organization as one motivation for her decision to focus on the employee’s perspective of the contract. In discussing the employment relationship, Capelli (1999) acknowledges that organizations have made dramatic changes in the nature of the employment relationship in response to the emergence of a highly competitive external environment whereby the demands of the market for the organization’s products and/or services and to some extent the labor market, largely shape what is exchanged between the two parties. Similarly Kalleberg (2001) emphasizes flexibility as a major theme in the literature on work and workplaces, and while acknowledging the importance of flexibility for the individual employee for the purpose of balancing work and non-work activities, plainly cites workplace flexibility as an organizational strategy undertaken in order to adapt to changes in product and labor markets. Second, both contract theory and employment relationship conceptualizations are focused, albeit in different ways and with differing degrees of specificity, on the bases of exchange between the individual and organizational parties. In particular, both are highly concerned with the extent to which job security continues to be exchanged for loyalty and commitment and investment in training and development, for high quantity of effort and valued skills required to achieve the organization’s goals. Thus, both literatures are highly concerned with the

312

LYNN M. SHORE ET AL.

nature of the social exchange between organization and employee, with the organization’s power to unilaterally and rapidly alter the nature of the exchange and with the resulting implications for the welfare of each party.

Differences Between the Two Literatures Yet despite some obvious similarities, there are also three primary differences between these two concepts and literatures. (1) Perspective and level of analysis. One important difference that receives fairly consistent support in the literature is that the psychological contract is largely focused on the perspective of individual employees, and related to this, on outcomes at the individual level of analysis (e.g. employee job satisfaction, commitment, turnover, organizational citizenship behavior and performance; Robinson, Kratz & Rousseau, 1994; Rousseau, 1995). In contrast, the employment relationship literature largely reflects the organization’s perspective and is therefore more focused on either employee group/job or organizational level outcomes. For example, Mahoney and Watson (1993) propose authoritarian, collective bargaining, and employee involvement models of workplace governance and their consequences for organizational functioning and performance, while Tsui et al. (1997) identify which cells of their balanced, over or under investment model characterize the employment relationship of the group of employees working in a particular job through the assessments of supervisors’ reports of HR practices. They then assess attitudinal and behavioral outcomes at the individual level of analysis. (2) Content of obligations vs. basis of exchange. Although both the literature on the psychological contract and the employment relationship are concerned with the specific content of what is exchanged between the two parties, psychological contract theory and research is focused on the specific nature or content of the parties’ mutual obligations, while the employment relationship research generally emphasizes two to five relatively general components of the exchange between employee and organization. For example, in the psychological contract literature, the specific content within the set of mutual obligations of the two parties is generally the basis for determining the extent to which the contract is more relational (longer term, open-ended, involving intangibles, e.g. celebrating the employee’s birthday) vs. transactional (i.e. short-term in duration, highly tangible and well specified outcomes and terms, e.g. rate of pay, voluntary or involuntary overtime). Similarly, psychological contract theory views incongruence between the perceived mutual set of obligations held

The Employee-Organization Relationship

313

by employee and organization as an important cause of violations (Morrison & Robinson, 1997; Rousseau, 1995). Conversely, the employment relations literature is concerned, not with perceived mutual obligations, but with what is actually exchanged by the two parties especially in light of recent and fairly dramatic changes in the relationship (Capelli, 1999; Kalleberg, 2001). (3) Determinants of outcomes. Violations in contract verses balanced or unbalanced nature of the exchange. Discrepancies or violations between what the employee believes has been promised by the organization and what is actually delivered play a large role in shaping the individual’s attitudes and behaviors in psychological contract theory and research (Robinson, 1996). Conversely, the drivers of employee group and organizational level attitudes and performance within the employment relations literature is both the level of investment that parties make to one another and the degree of balance or imbalance between their respective levels of exchange (Shore & Barksdale, 1998; Tsui et al., 1997). Summary In combination, similarities and differences in the psychological contract and employment relationship concepts suggest some important directions for future research on the employment exchange. First, the tendency of each concept to focus on the perspective of a different party and outcomes at a different level of analysis identifies a need for greater conceptualization across the two areas. Such an integration would serve to provide a better synthesis of the two perspectives comprising this exchange and an examination of the within and across level linkages of relevant process variables and outcomes. Second, growing evidence of differences in the drivers or determinants of outcomes proposed by each body of research, i.e. fulfillment of obligations vs. the level of and balance in the content of exchange, suggests the need for studies comparing the two in terms of both relative power and comprehensiveness. It is certainly possible that each may provide a stronger explanation of outcomes at distinct levels of analysis but it is also possible that the level of and balance in what is actually exchanged between parties, rather than the extent to which obligations are fulfilled may be the more powerful determinant of individual level attitudes and behaviors. Finally, both conceptual research on the employment relationship and empirical research on the nature and causes of variability within the set of human resource practices used by organizations have raised questions about the tendency of researchers to assume that employers within an industry or within a firm do or should form the same kind of employment relationship with employees or different employee groups. In combination, this work has advanced arguments that such variability in employment practices and subsequently, in the employment relationship is not random but rather reflects other differences in organizational

314

LYNN M. SHORE ET AL.

core competencies (Capelli, 1999), the nature of the customer market (Sherer & Leblebici, 2001), the relative contributions of human and other types of capital, such as technology, to the production process (Lepak, Taylor, Tekleab, Marrone & Cohen, 2003) and strategic value and uniqueness of the contributions different employee groups make to their firm (Lepak & Snell, 1999). We concur that it is important for further research to explore the determinants and effects of such variability in employment relationships within and across organizations.

INTEGRATING THE ORGANIZATION’S AND EMPLOYEE’S PERSPECTIVES As the previous sections of this article indicate the literature examining the EOR can generally be described as taking two separate perspectives either looking at the relationship from the organization’s perspective or from the employee’s perspective. While there is value in increasing the amount and breadth of research on the organization’s psychological contract, given the large and diverse body of research existing on the employee’s perspective, ultimately, the greatest value appears to be in integrating the two bodies of knowledge and studying the relationship jointly from the perspective of both parties. However, several issues currently seem to constrain researchers’ ability to integrate the two perspectives on the employment exchange. One issue to be addressed in this section is the extent of agreement or congruence that is necessary to capture the EOR and understanding the mechanisms contributing to different perceptions of the employee-organization agreement. Another challenge in integrating the organization and employee perspectives is to explicitly address the multi-level aspects involved in understanding the EOR. We conclude this section of the article discussing the challenge of focusing on the actual relationship rather than the perceptions of each party.

Agreement vs. Congruence Rousseau’s definition of the psychological contract (1990, 1995) did not require that the individual employee and the organization agree as to the obligations to each other, since the focus is on the employee’s perspective exclusively. And, in fact, there is evidence that employees and their managers do not agree (Tekleab & Taylor, 2001). In the inducements-contributions model, it is assumed that the organization, and all the relevant agents, are consistent in making investments in all individuals

The Employee-Organization Relationship

315

occupying a specific job or family of jobs and communicating the expected contributions on the part of the job incumbents. Therefore, unlike the psychological contract model, there should be agreement among agents of the organization and job incumbents as to the investments and the expected contributions. However, social comparison processes may operate to create disagreement among job incumbents as to what the inducements and contributions actually are (Folger & Kass, 2000). If these two models are to be integrated, then the issue of agreement between the organization’s understanding of the EOR and the employee’s understanding of the EOR needs to be addressed. It is necessary to understand why different perceptions of the EOR arise as well as the role of differing perceptions in creating misunderstanding.

Why Different Perceptions of the Agreement Arise There are multiple reasons why employees and employers will tend to develop and possess a different understanding of the agreement they share (Morrison & Robinson, 1997, 2004). Morrison and Robinson have identified at least three sources of these discrepancies in perceptions: divergent schemata of the employee and employer; the complexity and ambiguity around employment contracts; and miscommunication between the two parties. Both employees and employers come to the employment relationship with different cognitive schemata, which shape, guide and influence what they observe, remember, and interpret in the employment relationship (Fiske & Taylor, 1984). Each person, and thus each employee and employer, will hold a somewhat unique schemata. Because schemata are formed from experiences, the differences in schemata will be enhanced when the employee and employer come from very different educational backgrounds, work experiences, industries and cultures. These differences in schemata will result in the employee and employer remembering, interpreting and perceiving their shared experiences and communication somewhat uniquely and thus they are likely to end up with very different understandings of the agreement between them. The very nature of employment agreements and obligations also contribute to the occurrence of divergent perceptions of the employment agreement. The EOR is complex, and the myriad of shared obligations between the employee and the employer is often enormous. As a result, it is likely many obligations may be overlooked, or forgotten. In addition, often the obligations with employees are implicit and ambiguous. For these reasons, employees and employers are likely to put their own spin on those obligations, and fill in missing information with their

316

LYNN M. SHORE ET AL.

own biased interpretation (Griffin & Ross, 1991). The result of this complexity and ambiguity is that the employee and employer are likely to hold quite different and unique interpretations of their agreement. Finally, amount and quality of the communication between the employee and employer can contribute to the degree of misunderstanding between them. The degree of effective and extensive communication at the outset of the relationship, when most obligations are formed, may greatly influence how much disagreement between the employee and employer arises. In addition, ongoing communication throughout the relationship will play a significant role, especially in turbulent environments when conditions and thus the terms of the employment agreement may change. However, effective communication is not always present. Status differences between the employee and employer, for example, may contribute to communication difficulty. Moreover, the “false consensus” effect (Ross, Greene & House, 1977) leads the employee and employer to assume that they share the same understanding of the agreement when in fact they may not and because of this bias, they do not discuss it or resolve the discrepancies between their perceptions.

The Role of Differing Perceptions in Creating Misunderstanding In most EORs, the employee and the employer will possess a different understanding of the obligations of each party as well as how well each has fulfilled their obligations. These divergent perceptions of the employee and employer will invariably produce misunderstanding, conflict and contract breach. Each party will only provide that which they believe they owe to the other; thus, if one party believes the other is obligated to provide a particular contribution yet the obligated party is unaware of that obligation, they will invariably fall short of delivering it. Likewise, if one party perceives the other has failed to fulfill some obligations but the other party does not share this perception, it will remain difficult to rectify this transgression and thus a sense of violation and conflict is likely to ensue.

Multiple Levels of the EOR Thus far, we have presented evidence that the EOR occurs at multiple levels and has multiple inputs from varying levels, including organization, job, and group. Clearly in organizations, no single level, be it the individual, group, or organization, can completely account for individual behavior or group and organization performance. The EOR is shaped by actors (individuals, senior executives) and entities (work groups, human resource departments), reflecting an

The Employee-Organization Relationship

317

array of hierarchical levels and social complexity. Individual workers, managers, co-workers, executives, their interactions and the organizational structures these can comprise all shape the employment experience. But studies of EORs need a starting point. EORs are studied differently depending on whether the focus is on the individual (e.g. psychological contract, Coyle-Shapiro, 2000; Rousseau, 1995), the supervisor-subordinate exchange (e.g. mutuality of agreement, Dabos & Rousseau, 2004) the work group or bargaining unit (e.g. normative contracts, Nicholson & Johns, 1985; or labor contract, Landers, Rebitzer & Taylor, 1996), or the firm (e.g. human resource strategy; Huselid, 1995; Rousseau & Wade-Benzoni, 1994). Ultimately a multilevel perspective is needed to identify the principles that enable a more integrated understanding of phenomena that unfold across levels (Kozlowski & Klein, 2000). Multi-level issues surrounding EORs include (but are not limited to) different levels of authority and organizational complexity, top/down cross-level effects, bottom-up cross-level effects, frog ponds and partsto-wholes, and aggregated global assessments of the employment relationship. The Parties to the EOR The parties to the EOR exist at different hierarchical levels. These levels impact each party’s influence upon others as a function of authority, social distance, and complexity. Like power, the EOR, is inherently multi-party, with actions and reactions among them that shape the meanings each party attaches to the relationship (Brass, 2000). In individual-level studies of EORs, one initial question is who the focal individual believes to be his or her exchange partner(s). Different types of commitments and consequences can result when an individual’s partners to the psychological contract are team members, as opposed to managers, or the employing organization generally (Rousseau, 2000). Individuals can construe their exchange to involve numerous “principals” including the firm, their immediate manager, senior executives, team members, and even the State (see Loic Cadin’s (2000) description of psychological contracts in France). These principals can play multiple roles, sometimes acting as agents for the firm, other times as principals in their own behalf (e.g. immediate managers, senior executives). As Guest and Conway (2000) note: . . . employees at lower organizational levels with only a vague understanding of the formal authority levels above them perceive middle-ranking individuals as key representatives of the organization; however, these selected individuals were unlikely to consider themselves as representing the organization (p. 11).

Workers report that their exchange partners can be as diverse as the firm itself, its top management, their immediate boss, the team, a client or customer (Rousseau,

318

LYNN M. SHORE ET AL.

2000). Little research exists examining whether having several distinct focal others impacts the nature of the psychological contract or its dynamics under conditions of change or violation. It is evident that psychological contracts are shaped by a broad array of exchanges and cues. In a longitudinal study, Coyle-Shapiro (2001) found that a manager’s own psychological contract impacted how that manager interpreted the firm’s obligations to its employees generally and how well it fulfilled those obligations. Although more work is needed on how managers, team members, and others shape the psychological contract, it is apparent that a complex web of exchanges is operating. Simultaneous consideration of different parties’ frames of reference is important to understanding employment as a relationship. In particular, joint consideration of worker and supervisor perspectives on the EOR yields valuable insights. Coyle-Shapiro and Kessler (2002) demonstrate that workers and managers have somewhat different views regarding the terms of the EOR and the extent of each party’s fulfillment of its obligations. This study finds initial empirical support for the norm of reciprocity in the employment relationship and for a causal link between obligations and fulfillment. In a similar study, Porter, Pearce, Tripoli and Lewis (1998) report that the greater the gap between executive and worker beliefs regarding the inducements the employer offered to workers the greater were levels of employee dissatisfaction. Interestingly, a substantial number of employees reported greater amounts of inducements than did the executives representing the employer. Again as in the case of Coyle-Shapiro and Kessler (2002) executives described inducements in terms of broad organizational practices rather than those specific to exchanges with individual workers. Lastly Ho, Ang and Straub (2003) report that supervisors who continued to work with information technology workers after those individuals have been outsourced, held the same psychological contract with those workers, and continued to see themselves as the workers’ boss despite the shift in the nature of the formal employment relationship. In contrast, outsourced workers no longer viewed their former boss as their current supervisor. Another study focused specifically on dyadic agreement between individual worker and the employer. Dabos and Rousseau (2004) used matched pairs of managers and workers where each reported on the commitments they had made to the other and where workers reported on their beliefs regarding their manager’s commitments to them. This dyadic study found that agreement on the managers’ obligations (mutuality) was consistently related to performance and retention, as was agreement between manager’s obligations and those of workers (reciprocity). Further, the effect of agreement was greater than each individual party’s psychological contract assessment. We need more research on the conditions under which agreement among the parties to the EOR is likely to occur. Results to date indicate that differences exist

The Employee-Organization Relationship

319

in the beliefs that workers have regarding their EOR relative to the perspective of their managers or senior executives. Although certain human resource practices such as equity and profit sharing are posited to promote agreement among parties at different hierarchical levels regarding workers’ psychological contracts (Rousseau & Shperling, 2003) little direct evidence exists of factors giving rise to agreement regarding employment relations. However, findings also suggest that where agreement exists its affects are positive and substantial. Top-down cross-level effects. Top-down processes can introduce specific obligations and contingent beliefs into the psychological contract via human resource practices (e.g. contingent pay). Similarly, in the form of work group social influence such processes can shape member’s beliefs regarding the degree of fulfillment. For example, information people derive from their organization-based social networks impact their beliefs regarding how well their employer has fulfilled their psychological contract (Ho, 2002). Top-down processes can both introduce content into the EOR as well as constrain its variation across individual workers, impacting the flexibility for individual employment relationships to differ from one worker to another (Rousseau, 2001). To date, there is little research on the direct effects of higher-level processes on individual psychological contracts. However, there are suggestive findings regarding within-unit variation in individual psychological contracts. Dabos and Rousseau (2004) found little within-unit agreement on the psychological contract dimensions in their sample of research scientists. Rather, individual research scientists within the same work unit displayed variation in their psychological contracts despite having the same manager. Nonetheless, those workers did tend to agree with their manager (the research director) regarding the terms of the psychological contract. When these two parties agreed, this evidence of mutuality was positively related to performance and retention. The top-down effects are manifest as differential employment relations between managers and workers consistent with those observed in research on leader-member exchange (e.g. Graen & Uhl-Bien, 1995). In all, evidence of agreement in psychological contract terms appears to be greater for individual workers with their manager than among workers themselves within the same unit or work group. Bottom-up cross-level effects. Bottom-up effects refer to emergent phenomena arising due to attributes and predispositions of individuals or interactions among groups, that is, from characteristics of lower-level actors and entities within an organization. The psychological contracts of workers can impact the work groups and firms in which they are embedded. In particular, social networks can shape worker beliefs regarding their EOR, which in turn can give rise to normative

320

LYNN M. SHORE ET AL.

beliefs that become organizational realities (Ho, Levesque & Rousseau, 2003; Nicholson & Johns, 1985). Such effects have been used to characterize the link between individual and team learning (Goodman, 2001) and between group processes and organizational culture (Martin, 1992). Levesque (2001) finds evidence of role making on the part of first incumbents of positions in start up firms where actions taken by initial incumbents constrain the subsequent roles that later incumbents fill. Research on emergent processes tell us that what we think of as social structure or organizing has its origins in the on-going experiences of individuals seeking to create an effective, satisfying and somewhat predictable role for themselves in organizations (Weick, 1996). Frog-pond (parts within wholes). Where individuals stand in relation to the others in the larger work group or organization and how that shapes their attitudes and behavior constitutes a “frog pond effect” (Firebaugh, 1980; Rousseau, 1985). Such effects call attention to the origins and consequences of within-group heterogeneity in the terms and conditions of the employment relationship. Research in social psychology and economics reminds us that social comparison processes underlie many aspects of exchange relationships, in particular the standing individuals hold relative to others in their work and social groups (e.g. Frank, 1985). In particular workers are likely to derive information regarding the nature of the EOR from the treatment they receive relative to co-workers and other referent others. Whether workers believe their employers have kept their commitments has been found to be related to the quality of treatment co-workers report, a finding that suggests that comparison processes are at work (Ho, 2002). In a study of idiosyncratic EORs, where individual workers report having received treatment different from their peers, Rousseau and Kim (2003) find that this special treatment gives rise to significant within-unit effects upon the quality of co-worker employment relations. Specifically, in those departments where certain individuals had negotiated idiosyncratic deals, their co-workers were more likely to report procedural and distributive injustice than were workers in units where such special deals were absent. Though limited, such findings suggest that co-workers are important sources of information to individuals about the nature of their relationship with the employer, its quality, and extent of fulfillment. Aggregated and global assessment. It is not evident (yet) that psychological contracts of individuals can be aggregated to create meaningful indicators of work unit-level employment relationships. Aggregated variables often measure a different construct than their namesake at the individual level (Firebaugh, 1978).

The Employee-Organization Relationship

321

To date, evidence points to within-unit heterogeneity in individual beliefs regarding their psychological contract with an employer and to functional outcomes of individual – manager agreement on the terms of that individual’s particular employment relationship. Nonetheless, it can be helpful to consider the conditions under which EORs might constitute meaningful unit-level or organizational constructs. The meaning and operationalization of employment can vary depending upon the unit of analysis. Whether individuals within a similar job or the same work group have comparable conditions of employment is an empirical question and depends on the characterization of employment. Many firm-level studies of employment use archival data such as company records to derive indicators (Ichinowski, Shaw & Prennushi, 1997). If information such as job titles are the source of employment information, scholars may construe that the workers with the identical titles have the same employment relationship. However, individuals may construe their employment relationship differently despite such apparent surface similarities. It is an empirical question what the functional relationship might be between individual-level psychological contracts and broader characterizations of the EOR (e.g. occupational categories, normative contracts, etc.). We can find some guidance in Kozlowski and Klein’s (2000) conceptualization of how apparently similar constructs across levels might be inter-related. They differentiate the composition condition where comparable phenomena occur across levels (e.g. work group and organization climate, individual and organizational trust) from the compilation condition where there are discontinuities across levels among phenomena from a common domain (individual and group learning). Constructs reflecting compilation are characterized by distinct underlying causes across levels. Thus individual learning is highly subject to the influences of skill and predisposition while group learning is more driven largely by relations among members. Learning in each involves fundamental changes in behavior but the factors that give rise to this at different levels are not the same. In contrast, composition conditions reflect common causal mechanisms across levels. What little research there exists to date suggests that operationalizing EORs in terms of psychological contracts generates a high degree of within unit variability. At one level, there is little support for characterizing group or organization-wide employment relations in terms of psychological contracts paralleling those of individuals. However, it is too soon to tell. The level of detail and specificity with respect to operationalizing EORs varies considerably in research on the psychological contract (Rousseau & Tijoriwala, 1998). It may be the case that at the broadest and most general level of types of EORs such as relational or transactional, within-unit agreement may be greater than at the level of discrete obligations. We need to

322

LYNN M. SHORE ET AL.

develop theory on the causal processes underlying features of the employment relationship in order to be able to effectively explore its possible multi-level nature.

Are We Examining Two Sides, or Should We Focus (More) on the Relationship? Up to this point in time it seems safe to say that an examination of the total body of scholarly literature that has been produced on EOR issues would show a preponderance of emphasis on either one or both of the two parties to the relationship rather than on the relationship per se. The issue is: going forward, will we make greater research advances by a relatively stronger focus on the relationship as such, rather than on the attitudes and behavior of the employee and/or the organization? It is not surprising that much of the research to date has centered on one or the other of the two parties to the relationship, especially the employee. If any type of relationship is to be studied, it makes good analytical sense to decompose the relatively complex phenomenon into its constituent elements and intensively examine the nature of those elements. This the field has done up to now – although some would argue that the “O” (organization) side has been somewhat neglected in comparison to the “E” (employee) side. There is still more to be learned, of course, about how each of the two parties approach and implement the EOR, but it is probably useful to consider what might be gained in the future if the relationship per se were to become a stronger object of focus. The term “relationship” involves the condition of being connected and associated together. Thus, a focus on the EOR would mean attention to the interaction of the two parties, just as a focus on marriage as a relationship would involve attention to the interaction of husband and wife, not just to what each party brings to, and obtains from, the relationship. Clearly, as has been pointed out already in this article, if a relationship involves interaction it also involves exchanges, and thus the broader literature on exchange theory becomes relevant to understanding the particulars of EORs. The emphasis then becomes one of concentrating on both the content (what is exchanged) and, especially, the process (how the exchanges take place) of the interactions between the parties. What are some of the implications of a more intensive research focus in the future on the exchange-based relationship between E and O? Two of the more important will be mentioned here and then discussed further in the next section. For one thing, the effects of context on that relationship will likely receive more attention than has been the case so far. EORs, just like marriages, are likely to be highly affected

The Employee-Organization Relationship

323

by both the micro and macro environments in which they take place. A marriage relationship coping with economic distress of one or both parties is likely to have a quite different character than one that is occurring in relatively benign economic conditions. Likewise, a marriage dealing with a difficult child is undoubtedly a different relationship than one with no children or at least no problem children. In the case of EORs, we know relatively little in a systematic way about which context factors – such as organizational stresses, group and organizational norms and culture, employee personal circumstances − are most critical. Which aspects of the context, for example, are most important for building mutually positive EORs, and which hinder their development? Equally importantly, we lack solid knowledge about how such factors impact the ongoing exchange interactions between the two parties. What are the mechanisms that cause a factor to be important or trivial? These and similar questions would appear to form a fertile area for future research activities. A second and related implication of focusing more directly on the relationship of E and O would be that the dynamic nature of the relationship across time would receive much more concentrated attention than it has so far. Clearly, the EOR, like marriage or any other social relationship, does not stay static from one time period to another. The evolving nature of the EOR through time certainly has to be one of its more interesting features for study and understanding. Yet, that aspect, like context, is under-nourished in terms of scholarly attention. Is there an E-O equivalent, for example, to a product lifecycle? Are there highly different shapes of developmental cycles of EORs across different types of organizations, across different individuals? Are relationships smoothly continuous or highly discontinuous? These are just a few of the intriguing types of issues that would benefit from more illumination if attention is directed toward the changing and developing nature of EORs. It should be acknowledged, of course, that placing a greater research focus on the relationship of E and O will not be easy to implement. Measuring “relationships,” especially interactions between E and O – vs. measuring some simple types of social interactions or measuring the attitudes and behavior of individual Es or the actions and policies of Os – involves considerable challenges. It particularly requires the necessity to make many inferences from data, much of which will be indirect. After all, for example, at least some of the interactions of an individual with another individual (e.g. a supervisor), or with a group, can be videotaped for analysis. It seems rather difficult, however, to imagine videotaping an employee interacting with her organization! Despite such obvious research methodology obstacles, however, the benefits of investigating the relationship qua relationship of employee and organization seem worth the efforts and costs of undertaking the challenges.

324

LYNN M. SHORE ET AL.

THE ROLE OF CONTEXT AS A LOCUS OF OBLIGATIONS Although numerous researchers have emphasized the importance of context as a critical factor that influences EORs (Cappelli & Sherer, 1991; Hackman, 1999; Johns, 2001; Mowday & Sutton, 1993; Rousseau & Fried, 2001), most research on the EOR ignores the role of context or treats it as peripheral rather than central. In a broad sense, context is the locus of obligations surrounding EORs. Thus, work relationships are embedded in context. Our purpose here is to highlight the increasingly critical role of context and provide an organizing framework that can guide future research that focuses explicitly on the effects of context in employment relationships, rather than including it as an addition or after-thought. First, we briefly mention current trends that accentuate the importance of context. Second, we discuss the meaning of context. Third, we integrate recent work on the changing nature of work and work contexts to develop a preliminary framework for guiding research that aims specifically to consider the role of context in employment relationships. Fourth, we discuss the trend of increasingly nontraditional work arrangements that are not reflected in models of the EOR. Finally, we discuss the impact of the legal context on the EOR. Current Trends that Accentuate the Importance of Context Recently, globalization and competitive demands have increased the diversity in types of individual-organizational linkages. For example, many people now work in non-traditional roles and have non-traditional work status (contingent, on-call, part-time, temporary, expatriate, and foreign status) or work locations (telecommuting, satellite offices, and virtual teams) (Gephart, 2002; McLean Parks & Kidder, 1994; Rousseau, 1997; Sherer, 1996). One key outcome of these trends is increased diversity in the types of contexts that are relevant to EORs. For example, although some employees still work traditional hours at one specific location, many employees perform their work at different times (e.g. evenings, weekends, rotating shifts, on call 24-7) or in different locations (e.g. at home, on the road, electronically across national borders). Thus, we argue three key points. First, it is increasingly important to acknowledge the effects of context on the EOR. Second, context is not homogeneous, but rather it can differ substantially from one individual or type of relationship to another, even for those performing the same work for the same organization. Third, future research should examine differences in context as a primary research question that can influence the locus of obligations in EORs.

The Employee-Organization Relationship

325

What is context? At the most basic level, context is the setting, or what surrounds, the locus of obligations. Context contains the terms and conditions of the relationship. Accordingly, context presents both constraints and opportunities (Johns, 2001; Mowday & Sutton, 1993) that influence the links or connections that individuals and organizations are able to make with each other and with other entities (Rousseau & Fried, 2001). Context is multi-faceted and multi-dimensional. There is no such thing as “one context” for an individual or for an organization. Instead, each EOR is embedded in multiple contexts that all can exert influence simultaneously. For example, an individual in a traditional job might be embedded in nested contexts involving the job, work group, department, division, organization, occupation, labor market, family and national culture. Within the organization, social, physical, and structural aspects of context might also influence the relationship. Alternatively, an individual might be nested simultaneously in multiple jobs in multiple organizations. For example, part-time workers might have relationships with multiple organizations. Still yet, a contingent worker might not be nested in a job or in an organization but might be embedded primarily in the occupational or family context. Finally, an expatriate or foreign worker might need to function simultaneously in multiple national cultural contexts. In sum, we argue that context is complex; it is multifaceted and multidimensional. This is important for research on the EOR because it suggests that different aspects of context will have differential relevance for different types of relationships. Multiple contexts can represent complementary interests and demands. More often, however, different contexts represent competing interests and demands that will differentially influence interpretation of obligations and fulfillment from the perspectives of both individuals and organizations. Thus, there is no one context and it is important for researchers to recognize the simultaneous influence of multiple internal and external contexts. Given this complexity in the conceptualization of context, it is important to consider which aspect or aspects of context should be considered in any particular piece of research. Here, we recommend the benefits of research questions focused specifically on context effects. In each case, it would be important to select aspects of context based on the unique salience of specific contextual factors for the particular relationships being contrasted. In other words, we recommend that researchers consider context as a primary substantive construct, based on theory, rather than as an ad hoc, add-on to other research studies. This would avoid repetition of some of the problems encountered, for example, in the early research on personality where individual differences were tacked onto other studies without a clear conceptual focus on personality as the basis of the research question (Mischel, 1977).

326

LYNN M. SHORE ET AL.

Preliminary Framework for Thinking About the Multiple Contexts of Relationships To date, most conceptualizations of context adopt a systems perspective where aspects of context are nested within each other (Katz & Kahn, 1966, 1978). As argued above, we propose that individuals are not necessarily nested in contexts and that multiple contexts co-exist, often with conflicting orientations or conflicting demands. Recently, the Department of Labor developed a comprehensive, new framework to serve as a guide to the labor exchange process (Dye & Silver, 1999). This new approach replaced the Dictionary of Occupational Titles with O∗Net and emphasized the changing world of work (e.g. flexibility, adaptability, horizontal relationships) instead of traditional, mechanistic, and hierarchical approaches to thinking about jobs and work relationships. In developing O∗Net, a large task force of researchers began by developing a series of white papers that dealt with various aspects of contemporary work relationships. Several of these papers focused on context and identified a wide variety of cross-level contextual effects both external (Arad, Hanson & Schneider, 1999) and internal to the organization (Strong, Jeanneret, McPhail, Blakley & D’Egidio, 1999). We draw on their work in specifying seven key categories of contextual effects (four external to the organization and three internal) that form the basis of Fig. 1, a preliminary framework for thinking about the multiple contexts of EORs. Examining Fig. 1 highlights several key aspects of our conceptualization. First, we do not assume that contexts are necessarily hierarchically nested. Second, we do assume that there is one or one primary EOR. Third, we do not assume that all aspects of context are relevant to all EORs. Instead, the figure illustrates contexts that may not be nested; it allows for multiple EORs; and it shows that multiple aspects of context may be salient at the same time. Contexts external to the organization include National Culture Context (Group and Individual Cultural Values), Family/Leisure Context (Relationships and Personal Activities), Labor Market Context (Health of Economy and Industry), and Occupational Context (Professional Affiliations and Labor Affiliations). Contexts internal to the organization include Structural Context (Size, Differentiation, Job Design, and Group Composition), Physical Context (Machines, Tools, Equipment, Technology and Physical Environment of Work), and Social Context (Organizational Culture, Human Resource Systems, and Relationships). Consistent with our prior arguments about the importance of context in differentiating contemporary work relationships, we suggest the benefits of future research that focuses specifically on the role of context as the primary research question. We also propose that future research should identify specific aspects of context to include in a particular study, based on conscious conceptual

The Employee-Organization Relationship 327

Fig. 1. Preliminary Framework for Thinking About the Multiple Contexts of Employee-Organization Relationships.

328

LYNN M. SHORE ET AL.

comparison of different work roles and while acknowledging the possibility of multiple individual-organizational relationships and multiple contexts. Changes in Formal Work Relationships Models of work relationships for the most part have been based on notions of a “traditional” full time worker, frequently ignoring the armies of non-“traditional” workers who report to work every day, work the same hours and schedules and do the same jobs as their colleagues. “Traditional” work arrangements are those with “either an implicit or explicit understanding that employment will be continuous or ongoing, assuming satisfactory performance by both the individual worker and the organization” (McLean Parks, Kidder & Gallagher, 1998, p. 19). Yet, even assuming such a worker truly is the traditional norm, this norm comprises a steadily smaller part of our economy. Empirical research only sporadically reports whether samples are “traditional” or non-“traditional,” unless the focus of the research is on specific differences between the “traditional” workers’ attitudes, behaviors and performance and those of a specific type or class of contingent worker (e.g. temps, floats, contract workers). Further, because of the plethora of shifting categories that have been used to label contingent work, little is known about them that can be generalized. These non-“traditional” workers come from a variety of industries and occupations, and across different levels in the organization hierarchy. Consequently, their work relationships vary significantly. In addition, they vary in terms of how long they have been at a given site and how long they expect to continue, they vary in terms of how many different companies they are working for, their expertise and asset specificity, and the degree of control they exercise over their work. These non-“traditional” workers have faded into the background and become the invisible workers of organizational research. As a result, with few exceptions, our models of organizational behavior are impoverished, only weakly able, if at all, to accommodate more complex perspectives and work arrangements. It is critical at this juncture that models of work relationships encompass all types of workers. Omitting non-“traditional” workers, implicitly or explicitly, from our theorizing and research leaves out what may soon become the majority of our workers; as a result, our ability to contribute to the understanding of work relationships potentially will be limited. This section focuses on the implications of less “traditional” workers for research in organizational behavior, and more specifically, research on exchange relationships between workers and their organizations. First, however, before turning to a discussion of the redirection that the changing “traditional” nature of work necessitates, we briefly discuss the current state of non-“traditional” work in the labor pool and provide a brief history of the nature of non-“traditional” work.

The Employee-Organization Relationship

329

Non-“traditional” Work Over the past several years, deregulation, a volatile and uncertain economy, and pressures of globalization have resulted in a movement of workers from non-contingent to contingent work. As a result, the U.S. labor force now is nearly a contingent (BLS, 1995), made up of an army of temps, part-timers, floats, seasonal workers, day laborers, consultants and independent contractors. Although we focus on statistics from the U.S., like many other exports, contingent work increasingly is becoming a part of the labor supply in Europe and Asia (Barker & Christtensen, 1998; Hartley, 1995; Matusik & Hill, 1998; McLean Parks et al., 1998; Van Dyne & Ang, 1998). Today, the workforce is increasingly non-“traditional.” Temporary employment, just one form of non-“traditional” work, accounted for one out of every thirteen new jobs created in the late 1980s (Henson, 1996). By 1993, temporary employment accounted for 15% of all new jobs (Lewin, 1994). Now that number is 20% (CPA, 2001). Use of temps has increased by 400% since 1982, and the practice is still growing (Klein, 2002). In 2001, an average day might see almost 3 million Americans working as temps (CPA, 2001). The Bureau of Labor Statistics (2001) reported that 8.6 million workers identify themselves as independent contractors, 2 million workers identify themselves as “on call” day laborers, 1.2 million identify themselves as working for temporary help agencies, over half a million identify themselves as working for firms that provide services to a client at the client’s work site, 23 million identify themselves as part time workers, and 7 million identify themselves as holding multiple jobs. What is clear from these numbers is that organizational research and models that do not incorporate the non-“traditional” worker in their empirical research and methods, as well as their theorizing, are in danger of becoming obsolete. With the movement towards a less “traditional” workforce as a backdrop, the time has come for researchers in organizational behavior to engage in a discussion concerning the extent to which the current prevailing theory and research pertaining to organizational relationships is readily applicable to the growing number of workers who are non-“traditional,” or a workforce in which a sizeable portion of the workers are non-“traditional.” McLean Parks and smith (1998), building on an earlier paper (McLean Parks, 1993), pointed out that the literature on organizational contracts makes several stringent assumptions in the treatment of exchange relationships, two that are key in the present context: (1) that the exchanges are voluntary; and (2) that power between the parties is symmetric. In their 1998 paper, McLean Parks et al. explicitly explored theorizing as it related to psychological contracts for both “traditional” and non-“traditional” workers. As a result, they argued strongly that work arrangements should not be described

330

LYNN M. SHORE ET AL.

in terms of somewhat arbitrary and ill defined categories, such as temporary workers, contract workers, part time workers, and the like, but rather should be conceptualized along the dimensions of the psychological contract. In attempting to capture the complexity of non-“traditional” work relationships, these authors found it necessary to specify additional dimensions, including multiple agency and volition (McLean Parks & smith, 1998). In a similar vein, Gallagher and McLean Parks (2001) argued that current models of worker attitudes, such as commitment, are rather sterile in their present treatment, lacking the conceptual richness necessary to accommodate non-“traditional” work relationships. We identify several factors and considerations critical to incorporating non-“traditional” work relationships into current models: (1) a multiple party context; (2) multiple agency relationships; and (3) volition and potential power asymmetries.

Implications of a Multiple Party Context McLean Parks and smith (1998) suggested that the “entire network or fabric of related exchanges in which the contracting parties are involved is important.” Worker attitudes are impacted by the attitudes of those around them (Ibarra & Andrews, 1993), and consequently, we might expect the attitudes of the non“traditional” workers – positive or negative – to diffuse through the workplace as the relative ratio of non-“traditional” workers increases. At the very least, workers may develop psychological contracts through “vicarious contracting” (Rousseau & McLean Parks, 1993) by inference from the observable patterns of exchange observed with other workers, through their own process of negotiating the deal or through contract “drift,” making the exchange relationships of other workers part and parcel of the process. The trust that greases the wheels in organizations and workgroups may diminish as the relative ratio of non-“traditional” workers increases (Pearce, 1993), due perhaps to equity considerations stemming from the presence of multiple “types” of workers in filling the same role, with potentially different treatment by the company. Consequently, understanding the impact of the make-up of one’s workforce may be critical in understanding the antecedents of worker attitudes and behaviors important to the organization, such as commitment, trust, and extra-role behaviors.

Implications of Multiple Agency Relationships McLean Parks et al. (1998), developed the notion of “multiple agency,” which was necessary to capture more unique aspects of non-“traditional” work relationships

The Employee-Organization Relationship

331

that may differentiate them, to a greater or lesser extent, from more “traditional” work relationships. In the case of some non-“traditional” work relationships, a worker’s contributions can simultaneously fulfill obligations to more than one company. For example, temporary accountants can simultaneously fulfill their obligations to the temporary firm that places them as they complete the work required for the client firm. It is this simultaneity that creates multiple agency relationships, or a nesting of the obligations and work relationships. According to McLean Parks et al. (1998, p. 19), “A multiple agency relationship exists when an act by a [worker] simultaneously fulfills obligations to two or more entities, with full knowledge and sanction from both.” From the company side of the equation, company actions can simultaneously fulfill obligations to the worker, as well to another entity to which the worker is ideologically attached. For example, Ben & Jerry’s One-Sweet-Whirled® flavor brings attention to global warming. In addition, Ben & Jerry’s has committed to reducing the carbon dioxide emissions of their manufacturing facilities by 10% in five years. Both actions simultaneously serve the campaign against global warming and the goals of workers who are ideologically attached to these issues, exchanging ideological currency with Ben & Jerry, rather than just the economic and socio-emotional currencies.

Implications of Volition and Power in the Work Relationship Much early theorizing implicitly assumes that, through purposeful intent and actions, workers not only have accepted the terms offered them, but did so willingly and without desire to modify the terms. As noted by Staber and Aldrich (1989), “purpose is deeply embedded in current models of business strategy and organizational transaction, but such veneration of purposeful behavior has partially blinded us to other forces . . . ” (p. 112). One of these forces may be the relative power between the parties. McLean Parks (1993) argued that contracts could only be assumed to be voluntary to the extent that the parties had equal power. If power was asymmetric, the relationship involved a contract “maker” (more powerful party) and a contract “taker” (less powerful party; McLean Parks & smith, 1998). Although a contract taker may voluntarily agree to an unattractive deal – a deal which they might actually desire for any number of reasons – the notion of volition is critical to understanding the work relationship in general, and the work relationship of non-“traditional” workers in particular. The degree to which the agreement reflects the wishes, preferences, and needs of the worker likely is a function of the relative power that worker has vis a vis the organization or contract maker. Although extreme, the historical example of slavery would exemplify the contract taker with no choice – not even the choice to remain in the relationship or not. More recently,

332

LYNN M. SHORE ET AL.

workers in Asia making shoes for Nike and Reebok are unable to refuse overtime, an offence punishable by dismissal (Klein, 2002). On the domestic front, American Airlines outsourced its ticket agents in 1993. Some workers who had earned $40,000 were offered their old jobs back for $16,000 (Bernstein, 1995). In each of these cases, volition clearly is low for the less powerful party. At the other end of the spectrum are “temporary” CEOs, who peddle their services among corporations, staying an average of only five years (Thompson, 1998). Along similar lines, workers in demand who have skills that are not easily substitutable and who have highly codified credentials (e.g. clearly defined knowledge and problem sets commonly associated with the professions; Tolbert, 1996) are likely to be relatively powerful vis a vis the company. Examples of such workers would be sports and movie stars, specialist physicians, or IT workers during the Silicon Valley boom (Boh, Slaughter & Ang, 2001). Such workers, who are, to all intents and purposes, “temporary” workers, are able to “cherry pick” among jobs, moving frequently from job to job. These illustrations highlight the fallacy of the implicit assumption that the contracts between workers and organizations are voluntary. More realistically, volition lies along a continuum. At the heart of the variation in volition lies power – whether power is asymmetric favoring the worker over the organization (CEO, physician, IT worker), balanced between worker and organization, or asymmetric favoring the organization over the worker (ticket agents, migrant workers). Non-“traditional” work relationships are one area where the implicit assumption of equal or nearly equal power is a critical omission by empiricists and theorists (whether by explicit delineation or benign neglect of the issue). Incorporating volition and power thus become central factors in models of organizational relationships that are rich enough to include all types of workers, not just “traditional” workers. Whether power is symmetric, or whether one is a contract taker or a contract maker has implications for a broad range of factors, including differential reactions to breach and violation for the contract maker and the contract taker, how behavior is governed or enforced, whether reactions to an ideological breach will result in exit, whistle blowing, or principled organizational dissent, and the metric of justice that may be applied (Edwards & Bunderson, in press; McLean Parks & smith, 1998). Changes in Labor Law Both current and anticipated developments in labor law contribute to the increasing importance of understanding the EOR. First, there is a growing legal risk associated with the management of employees, and a decreasing ability to manage that threat through informed legalistic responses. As a result, it is becoming more important for organizations to effectively manage their relationships with employees through non-legal means (e.g. good employee relations) in order to avoid negative legal

The Employee-Organization Relationship

333

consequences. Second, trends associated with the globalization of the workforce (e.g. multinational employers, employees working across country boundaries) are straining national systems of employment law, leading to the recognition of a growing need for transnational standards regulating employment relationships. The understanding of EORs from diverse international and multiple disciplinary perspectives (social, psychological, legal) will take on greater importance as policy makers struggle to develop workable transnational legal employment standards. Support for these two observations is provided in the sections that immediately follow.

Growing Importance of Non-legalistic Means of Legal Risk Management In contrast to the rights of employees to engage in collective bargaining and other forms of collective action, which are viewed by many as being under assault world-wide (e.g. Brefort, 2002; Corbett, 2001), in recent decades many countries have experienced a steady growth in the individual workplace rights available to employees. For example, it has been observed that since the 1960s the American workplace has evolved from a largely unregulated arena to one that is subject to legal regulation from a myriad of sources (federal, state, and local, and legislative as well as judicial), which provides workers a wide range of individual rights (e.g. protection from discrimination based on sex, race, age, disability and other personal characteristics; family related leaves; protection from retaliation; access to information; protection of personal privacy; Befort, 2002). The widening scope of employee individual rights, together with employees’ increased awareness of those rights, led to an “explosion” in the amount of employment litigation in the U.S. (Bliss, 2000; Kohler, 1999). Other countries have also expanded the scope of individual employee rights, and are experiencing an increase in the employment related claims being filed in courts or tribunals. For example, Ireland’s Employment Equality Act of 1998 prohibits employment discrimination based on a broader range of personal characteristics than is covered by U.S. Federal law, including marital status and family status. The act went into effect in 1999, and a year later, the Director of the Office of Equality Investigations reported that claims for discrimination at work had risen by 40% in the first half of 2000. The United Kingdom’s Employment Relations Act of 1999 expanded the individual rights provided employees under the Employment Rights Act of 1996. Even before the passage of the more expansive employee rights legislation in 1999, the United Kingdom had experienced a doubling of employment discrimination and unjust discharge tribunal claims in the 1990s, leading to concerns about the rising level of “American-style” employment litigation (McCurry, 1999). World

334

LYNN M. SHORE ET AL.

wide, the continued expansion of individual rights in the workplace seems likely, and among the members of the European Union, it is a virtual certainty. Based on the authority of Article 13 of the Treaty of Amsterdam, the Council of Ministers of the European Union adopted proposals which require all member states to make discrimination (direct and indirect) unlawful on the grounds of race, ethnic religion or belief, disability, age, or sexual orientation in the areas of employment. As a result of the proliferation of laws regulating various aspects of the employment relationship, managers are less able to avoid legal risk through well informed decisions driven by legalistic considerations. Even among those few managers who have the time and resources to keep fully abreast of new regulations and how they are interpreted, and reinterpreted, the growing legal complexity creates dilemmas that cannot be avoided, but only managed. Under these circumstances, effective legal risk management will need to increasingly focus on the management of the EOR so as to promote goodwill towards employers, and avoid the types of negative interactions that lead to claiming behavior by employees and/or perceived unfairness in the eyes of third parties (e.g. judges, juries). This will be particularly true for the growing number of employers with international operations that are facing simultaneous regulation of their employee-organization relationship by multiple countries. In international contexts, employment decisions that normally involve the analysis of one countries law now require multiple, and sometimes conflicting legal analysis.

Development of Workable, Transnational Legal Employment Standards A growing recognition of the need for transnational standards regulating employment relationships can be found among both workers’ rights advocates and business/employer oriented individuals and groups. From the perspective of worker advocates, in the absence of transnational legal standards, pressures emanating from global economic competition (e.g. the perceived need for workforce flexibility and low cost workers) will lead to ever increasing workplace deregulation and the erosion of worker protections through out the world, what has been referred to as “a race to the bottom” (Arthurs, 2001. p. 2; Befort, 2002). From the business/employer perspective, there is a need for transnational standards that harmonize legal principles across national boundaries in order to reduce current uncertainty and facilitate the ability of businesses to carry out their operations as efficiently as possible (Aman, 2000). Whether motivated by concern for worker rights, or concern for business efficiency, the growing need for workable transnational standards regulating the employment relationship further contributes to the importance of understanding

The Employee-Organization Relationship

335

of the EOR. To what extent are there universal norms relevant to the employment relationship that might be the basis for transnational legal standards? It has been argued that the need for “good” or “just” cause in order to discharge an employee fairly is a universal norm (Befort, 2002), an argument that recently received additional support from a review of studies involving American workers – the legal status of the employment at-will doctrine in the U.S. not withstanding (Roehling, 2002). Are there similar widely held norms relating to other issues that employees and employers continue to grapple with, such as privacy in the workplace, employee participation in workplace decisions, and the balancing of work and non-work lives? If there is less evidence of universal norms in these other areas, what is the likelihood that employment related norms may be shaped by the implementation of transnational legal standards? Will a legal scheme, which focuses more heavily on individual vs. collective employment rights, such as the United State’s, be acceptable across countries and cultures? These are but a few of the questions associated with the development of transnational legal standards regulating employment relationships, and the answers to these questions are far from clear. What is clear, however, is that the development of a deeper understanding of the EOR will enhance our ability to fashion and implement workable transnational legal standards regulating employment relationships. In sum, the changing landscape surrounding EORs suggests the need for theories and models that reflect currency in the world of work. Globalization, competition, and environmental complexity are changing EORs. There are pressures for consistency in “employee rights” as reflected in EORs across cultures and nationalities. At the same time, organizations are increasingly seeking ways to create flexibility in response to environmental pressures, as reflected in the many types of work arrangements that are now part of the domain of EORs. These trends suggest inherent conflicts between the rights of employees and the rights of organizations that will surely shape the content and varying forms of the EOR that emerge.

MECHANISMS FOR THEORIZING ABOUT THE EMPLOYEE-ORGANIZATION RELATIONSHIP In this section of our article on the EOR, we examine several theoretical lenses that might be used to explain the dynamism, fairness of the employment exchange, relationship role, and integration of employee and organizational perspectives. We begin with an examination of developmental models that reflect changes in the relationship over time and continue with an examination of the role of fairness in the EOR. Subsequently, we consider the possibility of taking a relationship metaphor, as opposed to the contractual one for the EOR and ask what the theoretical

336

LYNN M. SHORE ET AL.

landscape would look like through this lens. Finally, we consider what the relationship might look like if greater attention was given to awarding a higher priority to satisfying the interests of both parties through the collaborative negotiation process.

Developmental Models This section examines two developmental models that while primarily used by prior research to examine the employee’s perspective of the employment relationship may nevertheless assist us in integrating the organization and employee perspectives in the EOR: (1) the relationship as schema model; and (2) the change processes and consequences model. Each allows for changes in parties’ perspectives as a function of their experiences and amount of discrepancy between those experiences and the existing schema at the time. We begin with the relationship as schema model.

The Employee-Organization Relationship as Schema Model Our starting point is a theoretical lens that views the EOR as a belief structure or event schema (Fiske & Taylor, 1984; Lord & Foti, 1986). In the past, research has typically focused on the employee’s perspective, in which perceptions of mutual obligations between employer and employee (the psychological contract) is viewed as a schema. This schema includes ideas about the typical sequence of events in a given situation, and serves as a guide for the individual’ s behavior in the EOR. Psychological contract schemata develop within a dynamic environment in which the individual interacts with multiple organizational agents (Rousseau, 1995; Shore & Tetrick, 1994). Both the individual employee and the organization (represented by several agents) enter the employment relationship with a set of expectations about mutual obligations. The individual’s psychological contract then further develops as a result of the interaction with the organizational environment. In the novel situation of entering the employment relationship, employees will process information related to the EOR mainly in a systematic and controlled, i.e. conscious, way (Rousseau, 1995). For example, Shore and Tetrick (1994, p. 97) describe the development of the psychological contract as a “deliberate goal-oriented process, in which an individual attempts to establish an agreement with his or her employing organization that will address a variety of employment objectives.” Later, however, over time as new employees become more integrated within the organization, information processing becomes

The Employee-Organization Relationship

337

more automatic and requires less individual awareness and fewer cognitive resources. There are several phases in the initial process of developing the psychological contract. It starts with an anticipatory socialization phase, when an anticipatory psychological contract (Anderson & Thomas, 1996) is created based on images formed before entry into the organization. This anticipatory contract is actually a “na¨ıve and imperfect schema about a preliminary deal that identifies what the organization expects its employees to contribute and what rewards employees can expect to receive from the organization” (Anderson & Thomas, 1996). In this stage, organizational agents involved in recruitment and selection and the information they exchange with the prospective employee create anticipatory expectations about obligations. This process involves not only the discussion of formal and written agreements but also the employee’s inferences that are often derived from implicit promises without explicit communication (Shore & Tetrick, 1994). The next phase is the encounter phase, which begins with employment and continues through the employee’s initial months at work. In this phase, employees experience the real demands at work, and the benefits and rewards such as salary and recognition. In this phase a more enduring schema of the psychological contract is actually formed and tested (Nelson, Quick & Joplin, 1991). During this time, the employee’s actual experiences influence the final development of the psychological contract schema, modifying and replacing the anticipatory psychological contract. Subsequently, it is through an ongoing process of mutual accommodation between the employee and the agents of the organization that the psychological contract continues to develop further. The employee’s contract schema gradually takes a more “solid” form, eventually becoming very resistant to change. However, this process of mutual accommodation can take different forms. One possibility is the employee’s adaptation to the reality unfolding in the workplace by adjusting his or her schema. However, it is also possible that, equity or instrumentality principles may play a role in shaping the employee’s psychological contract schema. For example, Robinson et al. (1994) found that a sample of graduating MBA students reported increasing perceptions of their employer’s obligations to them one year after graduation, yet these same employees reported decreases in their own obligations to the organization during that time. Another process that may shape an employee’s contract schema during the mutual accommodation period is reciprocity. Social exchange research indicates that the interplay of giving and receiving within the EOR causes parties to experience an increase in reciprocal obligations, and the enhancement of their trust in the other party, when that party lives up to or meets his or her obligations (Robinson et al., 1994). In support of the reciprocity process, Coyle-Shapiro and Kessler (2000) found that over time the organization’s fulfillment of its promises was positively

338

LYNN M. SHORE ET AL.

related to employees’ growing perceptions of their own obligations and to their contractual behavior. Further, employees’ perceptions of their actual contributions to the organization over time led to perceptions that organizational promises had increased in number and level, while the level of inducements actually received from the organization had diminished. Thus, changes in the psychological contract schema are likely as employees become more knowledgeable about the terms of their employment relationship (Thomas & Anderson, 1998). Existing longitudinal research (De Vos, 2002) also shows that during the first year after entry, changes in promissory beliefs reflect a process of adaptation to reality and that reciprocity plays an important role in this adaptation process. De Vos concluded that psychological contracts are subject to change during the initial stages of the EOR, especially during the first months after entry.

Change Processes and Consequences Model According to Rousseau (1995) schemas develop through discontinuous processing of information. A certain mental model of the EOR will endure until there is a clear signal of break or interruption. Profound modifications in the psychological contract may be caused by organizational actions (for example, imposed new deals: Schalk & Freese, 1997) or by individual changes (for example, the entering of a new phase in life with different needs and expectations). Thus organizations wishing to develop long-term relationships with their employees should consider not only individuals’ expectations, perceptions, and values, but also potential changes in these characteristics over time. Individual changes will require that employers work to bring them into congruence with organizational expectations (Orpen, 1994). Much has been written about the ongoing process of change in the EOR as a result of changes in the economic and social environments in which organizations operate (e.g. Roehling et al., 2000; Sparrow, 2000). Many authors suggest that a “new deal” and, consequently, a new psychological contract are emerging in employeremployee relationships (e.g. Anderson & Schalk, 1998; Herriott & Pemberton, 1995; Hiltrop, 1995; Schalk & Freese, 1997; Sparrow, 1996). The core issue of the old psychological contract, that is the offer of job security and continuity in exchange for loyalty, has disappeared, leaving a more transactional relationship that emphasizes flexibility and employability in its place (Anderson & Schalk, 1998; Roehling et al., 2000; Sparrow, 2000). These changes in the content of the terms of the psychological contract have resulted in a new deal where organizations emphasize the development of the broad skills of their employees and employees display a greater willingness to be flexible, in skills, job assignments, and sometimes even in

The Employee-Organization Relationship

339

work locations (Sparrow, 2000). Under the new deal, organizational leaders expect employees to focus less on job security provided by the organization, and more on job security based on their value in the labor market (Gaspersz & Ott, 1996). Thus, employees must take the initiative by continually seeking out further opportunities to develop their competencies and skills, in order to increase their market value. Empirical studies testing the emergence of a new psychological contract or a new style employment relationship have produced mixed results (e.g. Huiskamp & Schalk, 2002). Many employer and employee obligations have proven to be timeless, present in both the old and new contracts. However, by viewing the psychological contract as a schema, researchers are able to better understand the effects of different situations. For example, what will happen to the deal if employers do not fulfill their obligations to employees or if they unilaterally alter contract terms in the context of an organizational transformation effort? Under such circumstances, employees’ cognitive models of the contract serve as a frame of reference from which it is easy to predict likely outcomes. People will try to match the ongoing events they experience with their existing schemata. In case of minor deviations, individuals are likely to seek corrective actions for the discrepancy rather than to alter their schemata of their contracts. In novel situations or situations which drastically deviate from their contract schemata, people are more likely to take corrective actions that result in a change of their contract schemata. Revision of the psychological contract schema becomes necessary when the mutual obligations change fundamentally over time. The psychological contract as a mental schema has boundaries that determine what is acceptable and what is absolutely intolerable to employees with respect to reciprocal obligations. The development of the exchange relationship between employer and employee, considering the employee’s perceptual biases regarding that development, may create significant changes in the employee’s view of the contract. Thus, organizations should consider employees’ cognitive boundaries of what is acceptable and tolerable in the area of reciprocal obligations when planning for changes and their consequences. What kind of changes in contract terms are likely to be considered as intolerable by employees? This would depend on the basic values of the individual employee, regardless of the framework of the existing contract (Schalk & Freese, 1997). For example, employees of health care organizations who were asked what would be intolerable for them in their employment relationship gave answers like “If my health was affected,” “If I were no longer able to build a relationship with clients or to provide quality care,” “If my job was not challenging or fun anymore,” “If the organization got too negative,” “If I was not treated as a person anymore,” “If I was confronted with sexual harassment.” Specific contracts include a zone of

340

LYNN M. SHORE ET AL.

acceptable variation. Changes falling outside of this zone are likely to be viewed as intolerable violating an employee’s basic values. The zone of acceptable variation or the zone of acceptance is likely to be broad when the EOR is based on trust and good faith. For example, when employees perceive their managers as credible, their zone of acceptance to change is broader then when they assign low credibility to their managers. There are two typical patterns of change in psychological contracts that influence employee attitudes and behavior. The obligations of the contract can be exceeded or they can be breached. These changes can be due to organizational or individual actions, unexpected events, or because of a gradual “contract drift” over time (Rousseau, 1995, p. 143). When gaps between perceived obligations and met obligations result, one would expect an employee to first show a temporary decline or increase in attitudes and subsequent behaviors (Kotter, 1973; Robinson & Morrison, 1995; Robinson et al., 1994; Rousseau, 1995; Sparrow, 1996), and then to reconsider the current contract. The employee would then likely undertake clarification and re-negotiation of the contract in the form of a revision (Schalk & Freese, 1997) or transformation (Rousseau, 1995, p. 143). A revision of the psychological contract can take different forms. Negative revisions, whereby contract terms become noticeably less favorable to the employee, may cause a decrease in the individual’s felt obligations towards the organization and/or the job itself. Decreases in felt obligations towards the organization, however, is a more common result than decreases in felt obligations toward the job itself. Strong and negative (from the employee’s perspective) changes in contract terms are likely to cause the contract to break down. Accordingly, one may expect strong negative reactions from employees such as negative workplace attitudes and extreme behavioral responses: open conflicts, emotional expressions, and signs of aggression and depression. Employee reports suggest that they tend initially to focus on the work itself in such situations, frequently focusing on contacts with clients or putting more effort in their work. However, quitting the job or negative but passive work-related behavior, e.g. apathy, boredom, etc. is likely to follow (Rousseau, 1990). Conversely, positive contract revisions may occur. This is the case where the terms of the contract become noticeably more favorable to the individual employee as frequently occurs when individuals enter new jobs or positions. The effect of positive revisions generally has not been studied but it is expected that positive revisions would improve existing attitudes and behaviors. The Role of Justice in the Employee-Organization Relationship Organizational justice plays several roles within the EOR, depending on whether one views it from the perspective of the psychological contract (employee) or the

The Employee-Organization Relationship

341

employment relationship (organization) literature. In the case of the psychological contract, Rousseau (1995) posits that third parties to the EOR (i.e. parties who are neither the employee nor the organization) will tend to view the psychological contract (termed implicit contract) differently than will the two parties most directly involved. This is because they will tend to rely on different information, including a greater dependency on the tenets of procedural justice (e.g. the participation of both parties in the decision making process, the use of impartial decision makers, and advanced warning if changes are to be made in the process), when deciding on the basis of the obligations underlying the psychological contract between employee and organization if a violation has occurred. In addition, both Rousseau (1995) and Morrison and Robinson (1997) propose that employee perceptions of contract violation are tempered by their perception of the interactional justice, and more specifically informational justice, surrounding a discrepancy between what the organization is obligated to provide to the employee and what it ultimately delivers. The conceptual arguments of both researchers suggest that negative (those favoring the organization) discrepancies will not tend to be viewed as violations by employees in cases where the organization provides a good explanation for the cause of the discrepancy (e.g. the dire financial condition of the organization) and in cases where the discrepancy is implemented according to the tenets of procedural justice (e.g. with participation and advance notice). Similarly, Rousseau and McLean Parks (1993) proposed that organizational justice, in this case distributive justice, may come into play after a violation has occurred, such that the organization that provides a fair remediation to its employees for the violation may serve to restore trust and a relational contract and prevent or lessen negative attitudinal and behavioral reactions on the part of the employee. Thus, according to psychological contract theory, all four forms of organizational justice (i.e. distributive, informational, interpersonal and procedural justice), play important roles not only in shaping third parties’ views and decisions about the nature of the psychological contract between employee and organization but also in preventing employees from viewing all discrepancies between the organization’s obligations and provisions as violations and in helping to “right” a relational contract between the two parties once the organization has violated the contract. Still other research views organizational justice as a central part of the social exchange relationship between employee and organization, the exchange that underlies the very essence of the EOR. For example, Masterson et al. (2000) reported strong evidence of linkages between employee perceptions of procedural justice and their reports of the favorability of their exchange with the organization. Cropanzano, Rupp, Mohler and Schminke (2001) argue that employees repay fair treatment by forming strong social exchange relationships with the employer and cite evidence of justice effects being mediated by both indices of the quality of

342

LYNN M. SHORE ET AL.

the EOR, such as POS, as well as direct measures of the psychological contract. Finally, in a recent chapter on the EOR, Folger (2004) argues that conformance to two types of moral mandates, promise keeping and the human covenant, defined as socially shared understandings about the moral considerations seen as applicable for governing interpersonal interactions and social conduct in general, establish boundaries of acceptable behavior among human beings. These moral mandates are essential to the organization’s part of the exchange, given the employee’s willingness to accept uncertainties about the nature of task assignments and inducements as well as the autocratic control of the organization over job assignments, the precise nature of benefits or inducements, and in general, the policies and procedures governing the EOR. He cites numerous studies showing the negative effects resulting when organizations disregard these mandates and subject employees to low levels of procedural and interpersonal justice. He thus argues for a somewhat different role of organizational justice in the EOR than has been proposed before. In Folger’s view, the formation of an EOR places moral mandates on the employer, to treat its employees in a manner reflective of high procedural and interpersonal justice. In summary, we have argued that organizational justice plays a distinct and generally empirically supported role in the EOR with respect to both the psychological contract and employment relationship literatures. These roles have implications, not only for the overall quality of the relationship that develops but also for its ability to withstand conflicts between parties, mistrust and even major violations, while remaining strong and favorable.

Using a “Relationship” Metaphor Instead of a “Contract” Metaphor The literature on the employment relationship and more specifically the psychological contract has historically used models of organizational contracting. McLean Parks and smith (1998) suggested that based on this, several broad generalizations could be drawn. First, few models of organizational contracting address the variety of resources that can be exchanged and how the resources exchanged may affect the nature of the contract. Secondly, most models of organizational contracting implicitly rely on the assumption of self-interest. Thirdly, most models of organizational contracting implicitly rely on notions of balanced reciprocity. In this section, we examine the possibility of using a relationship metaphor, instead of a contract one as a basis for the EOR by drawing on social exchange. Also, in this section we posit the need to consider a more positive functional relationship in addition to the past focus on failures of organizations and employees to meet their obligations.

The Employee-Organization Relationship

343

Social Exchange Foa and Foa (1976) in their conceptualization of social exchange theory suggested that there were six types of exchange resources that could be described based on two dimensions. The categories of resources were love, status, information, money, goods, and services with these dimensions reflecting two underlying dimensions of universalistic vs. particularistic and concreteness vs. symbolic. Foa and Foa suggested that exchanges would typically be made with like resources or at least similar resources as arranged on the two underlying dimensions. Thus, resources connoting “love” would not be exchanged for monetary resources, for example. Unfortunately, there have been few tests of Foa and Foa’s theory in the workplace; indeed we could only find one such application, that being Berg and Wiebe (1993). Therefore, the question remains open as to whether Foa and Foa’s resource exchange theory generalizes to the EOR. Foa and Foa’s theoretical framework draws on the norm of reciprocity although it does not explicitly incorporate the three dimensions of reciprocity posited by Sahlins (1972). According to the norm of reciprocity within social exchange theory (Gouldner, 1960; Sparrowe & Liden, 1997), there are three components that affect reciprocity. First, the goods or services exchanged need to be of equivalent value. This is the most direct link with Foa and Foa’s resource exchange theory. Secondly, there is some consideration as to how quickly the receipt of a good or service must be reciprocated. The immediacy of exchange can be linked with the psychological contract model in that it has been posited that short-term relationships such as might be found among some part-time and temporary workers have more transactional contracts and expect repayment within the short-term (Rousseau, 1995). However, research has not always supported the proposed differences in psychological contracts among full-time and part-time workers (Coyle-Shapiro & Kessler, in press; Craig & Tetrick, 1999). Lastly, the likelihood of reciprocation is also dependent on the mutual interest between the parties to the exchange. The first component suggests a need to identify specific obligations, as it would be expected that individuals reciprocate with acts or resources that are of equivalent value. However, this raises the question as to what is meant by equivalent value. From whose perspective is value determined? Turnley and Feldman (1999) suggested that not all obligations within individuals’ psychological contracts are equally important and thus by extension they would not have equal value. If value varies by individual, then is it possible to establish a priori resources or obligations of equivalent value? This would suggest that it is possible to establish value independently. This might be possible with certain obligations or goods and services but would be more difficult with other resources. Further, some would argue that in the

344

LYNN M. SHORE ET AL.

EOR it might not be possible for employer obligations and employee obligations to have equivalent value. This links back, in part, to the power asymmetry. The second and third components of the norm of reciprocity essentially reflect something about the relationship among the parties. For example, the immediacy with which one expects someone to reciprocate reflects, at least in part, the degree of trust between the parties. The greater the trust between two parties the less there will be a need for immediate repayment of an obligation, but the less the trust the greater the need for immediate repayment. There also appears to be ambiguity with respect to some specific obligations as to whether they can actually be met. To the extent that both parties are concerned with the interest of the other party then the stronger the influence of the norm of reciprocity. It is hypothesized that such a situation may also reflect mutual trust and therefore the increased likelihood that the parties will reciprocate although not necessarily immediately. While focusing on specific obligations may have value, especially for developing interventions in organizations, the complexity of defining the EOR based on specific obligations may outweigh the advantages. It may be that a more parsimonious and generalizable approach to understanding the employment relationship would be to try to capture the exchange relationship directly (Shore, Tetrick & Barksdale, 1999). Social exchange theory and much of the literature on psychological contracts supports at least two dimensions of the exchange relationship: economic exchange (or transactional) and social (or relational) exchange. The question remains as to whether these two dimensions adequately reflect the dimensions of an EOR and whether these two dimensions are actually the ends of a single continuum or instead two distinct dimensions of the employment exchange relationship. Lastly, the literature has tended to look at reciprocity and balance in mutual obligations at one point in time (Shore & Barksdale, 1998; Tsui et al., 1997). It may be what is perceived as unbalanced at a particular point in time may actually be balanced when a longer time frame is used. We do know from cognitive psychology that people probably don’t keep a tally of every exchange. Rather, they form global perceptions that can be relatively stable. Therefore, psychological contracts may not be as malleable as some have assumed. Similarly, from an organizational perspective, strategies do not fluctuate widely and frequently; therefore, the investments in certain job incumbents would not be expected to change rapidly nor would the expectations of the organization for the contributions from those individuals in whom they have invested. It is interesting to note that the bulk of the literature has focused on organizational failure to fulfill their obligations rather than on the failure of employees to fulfill their obligations although one might say that is the focus of much of the literature on job performance, attendance, and even (anti-) citizenship behavior. It seems that taking a positive

The Employee-Organization Relationship

345

perspective on the outcomes of fulfilling obligations may expand our understanding of the EOR.

Positive Functional Relationship Much of the empirical research on psychological contracts has focused on breach or violation. There is perhaps an underlying assumption that if the mutual obligations are fulfilled all is well. This may be similar to the ongoing debate on how to define health. Is health simply the absence of illness or is it something more? It may be that by formulating the employment exchange relationship using the contract metaphor we have inherently set the stage to focus on the negative. The purpose of a contract is to provide a mechanism for enforcing mutual obligations thus preventing loss through either inadvertent lack of agreement or arbitrary and capricious behavior on the part of one party to the agreement. Certainly one can argue that the attention to failures to fulfill obligations serves to protect and promote the survival of the relationship. However, the question arises whether one can develop a positive functional EOR much in the same framework as discussed within positive psychology (Seligman & Csikszentmihalyi, 2000) or positive organizational behavior (Luthans, 2002). Based on his criteria, Luthans (2002) proposed that self-efficacy, hope, optimism, happiness or subjective well-being, and emotional intelligence are positive organizational behaviors, which are similar to concepts that researchers have suggested to be indicative of positive individual health reflecting optimal human functioning. Some of the indicators of positive individual health that have been proposed are hope, wisdom, creativity, mastery, and interpersonal flourishing (see Tetrick, 2003, for a review). Therefore, by extension, positive functional employment relationships should promote these six positive organizational behaviors. However, this only addresses one side of the exchange relationship. To complete the equation, it is necessary to consider positive organizational health. Among the generally agreed upon indicators of positive organizational health are adaptability, flexibility, communication adequacy, optimal power/equalization, resource utilization, cohesiveness, innovativeness, and meeting goals (see Bennett, Cook & Pelletier, 2003; Hofmann & Tetrick, 2003; Newell, 1995 for reviews). Therefore, it is suggested that a positive functional EOR would create both positive individual health and positive organizational health. Research is clearly needed to determine how to promote employment exchange relationships that reflect positive functional behaviors on the part of both individual employees and organizations. It is hypothesized that a positive functional EOR would have the following elements:

346

LYNN M. SHORE ET AL.


Employees are viewed as social capital investments.
There is consistent implementation of the employment relationship throughout the organization.
There is a clear meta-contract (i.e. rules about the rules of the contract).
The contract must be both flexible (i.e. able to change with the market, technology, etc.) and consistent in terms of promises, expectations, and obligations. Therefore, a positive functional EOR values employees and recognizes the context in which the EOR exists. Additional elements of a positive functional EOR need to be considered as future research develops these ideas further.

Integrative Negotiations Model The positive functional EOR has several characteristics in common with another conceptual approach that may prove useful in integrating the perspectives of the employee and organization within the context of their relationship. Earlier in this article, we noted that both psychological contract and employment relationship concepts and research generally acknowledge the greater power of the organization within the context of the EOR. There are a number of reasons for the organization’s greater power in this context but certainly among them are a larger financial resource base with which to weather economic downturns, less dependency on a single source for livelihood – i.e. a multi employee workforce vs. a single primary organizational employer, a larger and more complex social network from which to recruit a replacement relationship, and greater control of the mechanisms, i.e. job design, compensation parameters, training and development opportunities, and so forth, used in managing the EOR. Organizations’ greater power in the EOR allows for the possibility that agents of the organization may ignore or provide little satisfaction for employees’ interests, needs or welfare in general. In fact, Capelli (1999) argues that this has been the case over the last two decades as slack labor markets caused by major organizational restructuring allowed organizations to push most of the costs of changes in the workplace off on employees. He acknowledges, however, that tightening labor markets leave open the question of whether any particular organizational employer will retain its power and predicts that many will not, ultimately proving unable to identify new ways of retaining employee commitment, recruit critical knowledge and skills needed for market success and retain their best talent in the face of an increasingly transactional labor market. We conclude then that one important issue in the future of the EOR is the identification of mechanisms that are suitable for forming and maintaining a new kind of relationship, one which both parties generally concede will be increasingly

The Employee-Organization Relationship

347

open-ended (i.e. of indeterminate length, rather than permanent). Toward this end, we suggest the utility of win/win, collaborative or integrative negotiation that yields what Thompson (2001) refers to as level 2 or level 3 agreements in her pyramid model of integrative agreements. Integrative negotiation differ from distributive or win/lose negotiation in both its assumptions and process. Integrative negotiation assumes that parties’ (such as the employee and the organization) interests are not mutually exclusive and that the goals of both can be satisfactorily achieved through a process ensuring that all creative opportunities are exploited and no resources are left unused or “on the table.” Each party is committed then, not only to satisfying his or her own interests or goals but also to satisfying those of the other party. Similarly, the process underlying integrative negotiation differs considerably from that of distributive bargaining. Integrative negotiation involves: (1) the open exchange of information between parties to learn one another’s interests, goals and constraints; (2) sincere attempts to understand the other party’s real underlying needs and objectives; (3) an emphasis on commonalities between parties and the minimization of differences in values, styles, goals, etc.; and (4) searching hard for solutions that satisfy the interests and goals of both parties. Perhaps not surprising, strategies that have proven to be helpful in reaching integrative agreements include the building of trust and promotion of information sharing, asking of diagnostic questions to get at underlying interests and needs, unbundling of issues to create more opportunities for satisfying parties’ interests, proposing package deals rather than single issue offers to stimulating the comparison and trade-off of issues, and the extending of multiple offers simultaneously, with some of these offers designed to capitalize on differences in parties’ expectations of the future, and risk and time preferences – again to capitalize on differences in preferences (Lewicki, Barry, Saunders & Minton, 2003; Thompson, 2001). Level 2 or level 3 integrative agreements are those that go further than the level 1 point of exceeding each parties’ reservation point or their best alternative to that particular agreement, or in this case the EOR. Level 2 agreements are better for both parties than other feasible agreements, meaning that parties find a solution that they both prefer and would have considerable difficulty in exceeding. Finally, Level 3 agreements are those that are impossible to improve for both parties, at least not without simultaneously worsening the outcome of the other party. Thus, the use of integrative negotiation to reach a level 2 or 3 agreement between the employee and organization is an effort that requires commitment to satisfying both parties’ interests/goals, and strives for a relationship that exceeds other feasible agreements from the parties’ joint perspective. This is clearly a high standard for the process and the resulting relationship. Thus, one might ask whether the EOR is an appropriate forum for the use of integrative negotiation.

348

LYNN M. SHORE ET AL.

Consideration of the characteristics that signal the potential for win/win negotiation confirms the usefulness of the integrative model for creating value in the context of the EOR. These characteristics are: (1) the negotiation includes more than one issue, e.g. wages, benefits, training, job learning and challenge, employee performance and innovation, commitment, loyalty, etc; (2) there are other issues that might be brought into the negotiation, e.g. work hours and location, deferred income, workplace governance, etc.; (3) the possibility of side deals exists (e.g. organizations or individuals also interested in a particular employment relationship, offering another a premium for freeing an employee from a prior work commitment; permitting another individual to participate in a work project as an outside consultant); and (4) parties have different strengths of preference across most of the issues of negotiation, meaning that the organization and employee may differ in their preferences for certain salary structures, benefits, work hours, performance quality, work assignment preferences, allowing for the possibility of crafting a deal that satisfies both (Thompson, 2001). Thus, integrative negotiation seems to offer the potential for achieving a level 2 or 3 integrative EOR. A more important question, however, might be why a more powerful organization, caught up in the pressures brought by a market economy, would choose to embrace integrative, rather than distributive negotiation, in forming its relationship with generally less powerful employees? This issue is particularly important given Pfeffer’s (1981) doubts about the viability of multi-constituent systems of fairness in the face of individual biases that make it difficult to see one’s self-serving nature, and the information and negotiation skill asymmetries that typically exist between stronger and weaker parties. Our answer here is a multi-faceted one. First, we suggest that integrative negotiation is more likely to be used in an organization’s negotiations with their core employees, i.e. long-term employees that bring strategic value to the firm through their possession of highly unique and valuable knowledge and skills. We suspect that other, non-core employee groups will increasingly be externalized (Delery & Shaw, 2001; Kalleberg, 2001) and these externalized individuals’ employment relationship, will then be with a third party vendor for whom their knowledge and skills may be strategic. Second, we do not expect that all organizations will choose to use the integrative negotiation model in forming EORs, even those for core employees. Rather, we suggest that organizations which are attuned to the long term benefits of employee creativity and commitment and understand that creativity and commitment may result from selective investments in human capital (Jackson, Schuler & Rivero, 1989) might increasingly use and reap the benefits of integrative negotiations in the context of the EOR. Thus we suggest integrative negotiation as another mechanism for theorizing about the EOR.

The Employee-Organization Relationship

349

CONCLUSIONS If one looks at the total mosaic of research carried out to date on the EmployeeOrganization Relationship (EOR), as described in this article, it seems clear that many different facets of the topic have been explored. A somewhat coherent body of knowledge and understanding is gradually being developed. In this article, we’ve argued for the need to consider ways to integrate the psychological contract and employment relationship literatures, and have provided a number of suggested mechanisms for ways to do so. We have also argued for the importance of context in both theory development and empirical evaluation of EOR models. Anyone doing research on this area is definitely not starting from scratch. An issue, therefore, is that given what has already been learned, where are the most important research gaps that need additional examination and exploration. It is unlikely, however, that there is – or even should be − any single, definitive answer to this question. Thus, below, we first describe what we believe has been learned so far in the EOR literature, and then conclude with future research directions.

What We Have Learned So Far Whether one examines the psychological contract or the employment relationship literature, significant changes in economic competitiveness, organizational structure, and the frequency of non-standard work have led to accompanying revisions in the nature, level and duration of the EOR over the last 20 years. While it can be argued that these changes have been dramatic only for a relatively small portion of the working population, as many employees never experienced the stereotypical life-long loyal and relational employment relationship with an organization, it is more likely that a large percentage of employees experienced at least a part of this idealized relationship. At the very least, the expectation of long-term employment security and relational contracts, or at least the hope of such a possibility given a good education, hard work and loyalty, remains an ideal for many workers. The empirical evidence suggests that even though formal work arrangements are changing, both employees and employers are seeking ways to develop and maintain strong social exchange relationships with inherent benefits, even in the face of extensive environmental challenges. Considerable empirical research now supports the importance of agreement on or balance in the contributions of each party to the EOR for positive attitudinal and behavioral outcomes at the individual and job levels of analyses. Note however that this research has generally been conducted from the perspective of one party, not both, to the relationship. There is also evidence that the type of EOR that

350

LYNN M. SHORE ET AL.

is strongly linked to organizational level performance may be moderated by the organization’s goals and competitive strategy. Thus, it is becoming increasingly clear that even though social exchange relationships appear consistently beneficial at the individual-level of analysis, when important obligations are fulfilled by both parties, the same principle may not always apply at more macro levels of analysis. Further testing and development of group, organization, and cross-level models will help to further enlarge the EOR domain so that similarities and differences across levels become clearer. At this point in time, relatively little research has examined the effects of a balanced or congruent EOR from an integrated perspective, reflecting the views of each party. However, the three to four existing studies that have examined both perspectives indicates that we may learn more about the outcomes of the EOR from studying them from a joint or integrative perspective. Thus, the question becomes how to best integrate the individual employee and organization perspective in future theorizing and empirical research. Our research for this article indicates that there may be much to be gained from further considering the relationship as the basis of the exchange, rather than a psychological contract per se. Likewise, exploring interactive methodologies that consider the interests of both parties in the relationship, such as win/win or integrative negotiations, as long as such conceptualizations are able to explain the dynamic nature of the relationship, over time and the impact of context on the relationship. Finally, building on the ideas presented in this article, we have highlighted the importance of seeking to develop models of the EOR that delineate healthy and productive modes of organizing work relationships. While breach and violation remain important and meaningful areas of study, models of healthy employment relationships are the positive and yet understudied analogue.

Needed Research – Where do We Go from Here? This article was written by many different scholars in the EOR field. It began as an idea at the annual Academy of Management meeting, where Lois Tetrick, Susan Taylor, Lynn Shore and Jackie Coyle-Shapiro decided (how simple it seemed then!) to help move the EOR literature forward through a series of forums. This initial goal then really took shape at a 2 21 day workshop in Houston in 2002 in which many of the authors of this article were in attendance. Subsequently, a session titled “Research Workshop Networks: A Methodology for Developing the Employee Organization Relationship” (National Academy of Management Meeting in Denver in 2002), an edited volume of the Journal of Organizational Behavior (August, 2003), an edited book (Oxford University Press, 2004), and

The Employee-Organization Relationship

351

finally this article became the products of much discussion and meetings of the minds. Those of our readers who have systematically read the entire article undoubtedly recognize the many research ideas embedded within each section. The purpose of this final section is to bring forth additional ideas, and to further develop some of the research ideas already mentioned in brief in prior sections of this article.

Integration of Employer and Employee Perspectives One of the most important suggestions for future research that this article has raised is the need to integrate the employee and the employer perspectives. We have discussed a number of conceptual and methodological issues that will need to be resolved for such integration. These issues include the idea of using “relationship” rather than “contract” as a metaphor, focusing on the interaction rather than perceptions of obligations, and encouraging a dynamic view so that changes in the psychological contract or employment relationship can be tracked over time. Most importantly, the discussion directs us to greater emphasis on the “context” in understanding the nature and changes in the relationship between the employee and the employer. One implication of the “relationship” metaphor is that we would analyze the phenomenon at the dyadic level. One advantage of a dyadic focus is that it would make the phenomenon more tractable. The “relationship” is not between the employee and an abstract entity called the organization but between the employee and agents of the organization. These agents could include the immediate supervisor, one or more co-workers, higher-level managers, the HR manager or representative, and even the executive leader. Future research could investigate the role of different agents in influencing the employment relationship representing the employer’s perspective and the psychological contract perceived by the employee. Further, employees do not negotiate with the “organization.” They negotiate with “agents” of the organization. Negotiation research on the EOR could analyze the effects of different agents on negotiation outcomes. In the context of multiple agents, it would be interesting to identify agent characteristics that would likely yield win-win negotiation outcomes between two parties with inherently different power and interest. Agreement in Dyads The dyadic view raises questions about the extent to which the agreement workers perceive to be in place in their EOR is indeed an agreement in fact. The extent to which workers and employers hold similar beliefs regarding the psychological

352

LYNN M. SHORE ET AL.

contract and the employment exchange generally is critical to the functioning of employment relations. Two key dyadic qualities of the EOR are mutuality and reciprocity. Mutuality refers to the extent of agreement regarding the nature of the relationship and its specific terms. Reciprocity is what the parties believe they owe each other in exchange for contributions the other has made. Agreement, defined by shared perceptions on the parties to employment, in psychological contract theory is fundamental to positive outcomes for both parties. (Note, that we are using agreement here to capture what previous psychological contract researchers labeled as “mutuality.” These earlier uses of the term did not differentiate between agreement on same obligations or reciprocal obligations.) The first issue from a dyadic view is to what extent agreement, operationalized as either mutuality or reciprocity, exists. Though their design did not match individual workers with employers, Coyle-Shapiro and Kessler (2002) have examined the extent of agreement among managers and groups of workers they employ, finding varying levels of mutuality. Dabos and Rousseau (2004) also found variation in degrees of agreement between employer and worker. A central question regarding the presence of agreement is the factors that drive convergence in worker and employer perspectives. Psychological contract theory suggests that agreement is likely to arise where there are shared social cues, common experiences between the parties, and consistency in human resource practices. At the very least, we cannot assume that agreement exists and subsequent researchers are advised to take levels of agreement into account in studying the exchange relationship. The second issue is the impact of agreement on processes and outcomes in the EOR. In a dyadic study where employers provided commitments they had made to each worker individually, Dabos and Rousseau looked at two forms of agreement. Mutuality was operationalized as the extent to which worker and employer agreed on the obligations the employer had to the worker. Reciprocity was operationalized as the extent to which the parties agreed as to the obligations the other party owes them. In this instance, agreement would be assessed by comparing whether comparable types and levels of obligations existed in what the employer owed the worker and what the worker owed the employer in return. Both mutuality and reciprocity were positively related to beneficial outcomes for both worker and employer. In particular, productivity, retention, and future career opportunities were positively related to both mutuality and reciprocity. Mutuality regarding employer obligations appeared in this study to be higher than shared beliefs regarding reciprocal obligations. Dabos and Rousseau (2004) attributed this difference to the time lags involved in reciprocating the actions of another. If indeed reciprocal agreement is more difficult to achieve than mutuality, research is needed into the impact of time lags and in interactions between employee and employer over time on successfully accomplishing reciprocity.

The Employee-Organization Relationship

353

Reciprocity is potentially more dynamic than mutuality, that is, reciprocity is subject to change over cycles of contribution and rewards, inducements and follow-through. Some kinds of commitments also may require more time to fulfill, as is likely the case with development opportunities and career advancement, as opposed to material support or benefits. Thus in understanding the dynamics of the employment exchange we need greater attention to the types of resources involved. Another important issue underlying the extent of agreement in the EOR is the power balance between worker and employer. To the extent that one party is less able to shape the nature of the exchange than is the other, the EOR may involve more agreement to as opposed to agreement with, undercutting the level of mutuality and reciprocity upon which the exchange is based. Social Networks The dyadic approach has another conceptual advantage. It can be expanded to a network analysis. A network is comprised of multiple inter-related and embedded dyadic relationships. The network concept may apply to the “relationships” dimension of the social context depicted in Fig. 1. However, the network is more than “relationships” in the social context. It is the structure of the relationships for an employee. The network structure would influence the employee’s perception of his or her psychological contract in relation to other’s psychological contracts. For example, the psychological contracts of employees in a cohesive or dense network are likely to be different from that of employees in a sparse network. In a cohesive network involving the supervisor as an agent for the organization and multiple employees, we may expect more similarity in psychological contracts among the employees performing the same job. In a spare network involving employees who have no knowledge of or at most infrequent interaction with each other, we would expect more divergence in both the psychological contract and the actual employment relationship. In other words, a network with multiple agents of the organizations may produce EORs quite different from those in networks where one single agent dominates. The implications of multiple agents vs. a single agent for psychological contracts would be a fruitful area for further conceptual and empirical work. The location of the employee in the structure also may influence the employee’s power to negotiate with the organizational agent about expectations and obligations. For example, a more central employee certainly would be in a stronger position to negotiate than another employee in a peripheral location. The employee in a “bridge” location may have more information about the psychological contracts of other employees in the same job or similar jobs, giving the bridge employee either an advantage in negotiation or a wider basis for judging fairness by the employer. In fact, research indicates that workers with informally powerful positions have

354

LYNN M. SHORE ET AL.

more munificent psychological contracts than their less powerful counterparts (Dabos & Rousseau, 2004; Ho, Levesque & Rousseau, under review). By integrating the psychological and sociological perspectives, the network approach is a powerful analytical tool for capturing the effect of the network as a “context” on the nature of psychological contracts and employment relationships for different employees. Comparison of Workers Within Units The comparative view addresses the question of what it means when workers in the same work unit hold different perspectives regarding their employment relationship. Researchers who have examined the psychological contracts of workers in the same work unit have found little evidence of within-unit agreement (Dabos & Rousseau, 2004; Rousseau & Tijoriwala, 1999). Since the psychological contract reflects an individual’s system of beliefs regarding an exchange relationship with another, the lack of within unit agreement is not a challenge to the integrity of the psychological contract. But do workers in the same setting have different psychological contracts? If so, it raises the issue of how workers in the same unit can come to hold such distinct perspectives regarding what they owe their employer and what they believe the employer owes them. The practical implications of managing a work unit wherein members hold widely different beliefs suggest that within-work unit heterogeneity in psychological contracts can be dysfunctional from the firm’s perspective. Moreover, disagreement among co-workers can impact perceptions of the organization’s practices, including the fairness of treatment they receive. Basic questions the comparative view raises include: To what extent do people in the same organization have different psychological contracts? What is the functional effect of this disagreement for the firm and for individual workers? Greater use of analyses of within-group heterogeneity is needed to explore the dynamics of differences within units in the EOR. Such approaches can help answer the question of how the psychological contract of a co-worker impacts a worker’s experiences on the job. The relative powers of the parties involved in the EOR shapes not only the terms of their agreement (and whether this is agreement with or agreement to), but also the extent of commitment less powerful parties express regarding the terms of their agreement. Social standing shapes one’s beliefs regarding obligations, particularly obligations on the part of the employer. Whether this is a result of differences in contribution high status members make, recognition given to higher performers or more valued workers, or the effects of auspicious social position itself is unclear. Research is needed not only into the social processes that affect the psychological contract but also the extent to which social status itself can shape the resources that flow to employees and in turn the obligations they come to believe they are party to.

The Employee-Organization Relationship

355

Dynamism of the EOR One of the critical dimensions of the overall EOR domain that seems relatively underexplored, as briefly discussed earlier in the article, is the dynamic nature of the relationship through time. By its very nature, the relationship of employees to their organizations is not fixed and static. We need to know more about the evolving stages of such relationships and, especially, the endogenous and exogenous causes of those changes. The former would include causes directly related to either or both of the two parties to the exchange. For example, what is the magnitude of the effects of changes in top management, which often are accompanied by changes in organizational objectives, on the relationship, when do those effects occur (if they do) and how long do they last? Are these effects on the relationship greater or lesser than changes in an employee’s immediate supervisor, and, particularly, what are the reasons why they are or are not greater? Exogenous causes would be those outside of the two parties themselves, such as changes in economic conditions or societal trends. Beyond identifying internal and external causes of changes in the relationship, it is also important to study the interactions of those causal factors in producing various kinds of effects. This implies more emphasis is needed in future research on attempts to capture multiple factors simultaneously impacting EOR changes.

The Role of Culture In an earlier section, we have pointed out differences in employment laws and employment practices in some European countries. However, we know much less about the meaning of psychological contracts or the nature of employment relationships in contexts beyond the Anglo-Saxon culture (Rousseau & Schalk, 2000). We have reported an emerging interest in the employment relationship of firms with middle managers in the People’s Republic of China (Tsui et al., 2002, 2003). The obligations of the employer (in the form of inducements) and of the employees (in the form of contributions) expressed by the Chinese middle manager sample were quite similar to those reported by U.S. samples. We suspect that the basis for the similarity may be contextual. Chinese managers are taking MBA-like courses in large numbers and many Chinese firms are imitating the practices of foreign invested firms (many are from the Anglo-Saxon countries). These developments suggest that institutional theory (Scott, 1995) may be a relevant theoretical lens for cross-cultural or cross-national comparative studies of employment relationships and psychological contracts.

356

LYNN M. SHORE ET AL.

There are a number of exciting questions that could or should be explored in future cross-cultural studies. Does the employment relationship or psychological contract carry the same meaning in different cultures? What do employers perceive to be employee obligations to the firm in different cultures? What do the employees perceive to be employer obligations in these cultures? What might be some universal and culture-specific obligations from both sides? Are there cultural differences in worker reactions to the breach of psychological contracts? Even though the norm of reciprocity is universal (Gouldner, 1960), there might be differences in the form with which reciprocity occurs. For example, the ideas of resource equivalence, mutual interests, and immediacy of exchange (Sparrowe & Liden, 1997) may vary across cultures. Some cultures may value equivalence and immediacy while other cultures may scorn both equivalence and immediacy. How do these cultural norms in reciprocity influence the nature of the employment relationship and perceptions of psychological contracts? Answers to the above questions would contribute to basic knowledge and have important practical implications for managers of multinational firms.

What is Exchanged? To varying degrees, the different perspectives reviewed in this article address the “currency” of exchange, with perhaps the most recent development given by Edwards and Bunderson (in press), in their development of ideological currency as a complement to the economic and social currencies explored in the research thus far. McLean Parks and her colleagues (McLean Parks & smith, 1998; McLean Parks et al., 1998) built on Foa and Foa (1976) to focus on the type of resource exchanged, many of which could be sub-classified as being characteristically economic, social, or ideological. With few exceptions, relatively little research has focused on the “what” of the exchange relationship beyond these broad categorizations in any great detail, in particular, the metric (or metrics) that go in to evaluating whether or not one has received their due when exchanges cannot be completed “in kind.” However, this comparison – this commensurability (McLean Parks, 1997) – has potentially important implications for perceived fairness/justice in the exchange, perceived compliance, breach or violation, as well as reactions to compliance (Organ, 1988) or beneficence (McLean Parks, 1997) such as organizational citizenship behaviors (Organ, 1988), or to breach and violation such as retribution and revenge (McLean Parks, 1997). With different forms of currency in use in an exchange relationship, the commensurability of the currency of exchange – or resources – becomes an important issue. For example, McLean Parks (1997) built on Pruitt’s (1981) idea of three

The Employee-Organization Relationship

357

general types of compensation, or in this case the “what” one receives for the resource one has given. Pruitt describes specific, homologous and substitute forms of compensation, forms of compensation that are differentiated by: (1) the need that is satisfied; and (2) the domain in which the need arises. A working parent might hire “Merry Maids”™ to clean the house, satisfying the need for cleanliness. In this case, the need satisfied (cleanliness) is identical to that lost by the parent’s absence, and it occurs in the same domain (home), providing specific compensation or recompense (McLean Parks, 1997), becoming an exchange of like for (nearly) like within the same domain. Homologous compensation or recompense is an exchange of like kind (my Longfellow for your Thesaurus; Pruitt, 1981), where the domain is quite similar (books) but the need that each satisfies is quite different (Pruitt, 1981). Substitute compensation or recompense is exchanged in different domains and satisfies different needs, for example, the practice of purchasing indulgences from the Church for transgressions committed against others. Only in the case of “like for like” exchanged in the same domain is compensation isomorphic (Sparrowe, Dirks, Bunderson & McLean Parks, 2003) in the exchange between the parties (see Fig. 2). Outside the domain of isomorphic compensation, these different forms of compensation or recompense suggest that the equivalence of the exchange may be judged quite differently depending on which form is used in the exchange, with a more straightforward matching of specific or homologous compensation than for substitute compensation (McLean Parks, 1997). The implications for the “what” of the exchange – and commensurability – have not been explored in standard models of exchange relationships. Referring back to the three types of currencies – economic, social and ideological (Edwards & Bunderson, in press), one begins to see the complexity of determining commensurability, not only in terms of the needs satisfied, but also the domain in which they occur. Economic currencies easily satisfy the same needs and may be

Fig. 2. Form of Compensation and Currencies of Exchange.

358

LYNN M. SHORE ET AL.

more transportable across domains (e.g. an exchange when traveling of Euros for dollars). Social currencies may not be transportable across domains (a prot´eg´e in the working domain vs. one’s child in the family domain), but may satisfy the same need (affiliation, mentoring) in the two domains. Ideological currencies likely do not satisfy other needs nor do they transport easily across domains, making them particularly prone to violation.

How are the Parties “Cued” to the Exchange? From the perspective of evolutionary psychology, Cosmides and Tooby (1989, 1992) suggest that natural selection has imbued individuals with the ability to recognize cues – in this case, to the nature of the exchange relationship – and match those cues to the appropriate script that provides expectations and information about possible contingencies and outcomes. These cues and their associated scripts then inform the individual in terms of how, or if, to reciprocate, how to enforce compliance (or conversely, to punish non-compliance) and possible response sequences. Yet if, as Cosmides and Tooby suggest, these scripts are domain specific, this would imply that exchanges with specific and substitute compensation are not scripted, and rather are emergent and evolving. Future research should endeavor to explain the cues or signals used in defining the nature of the exchange (and hence the script that will be used as a referent) as well as emergent exchange processes, in which there may be a constant state of disequilibrium.

Research Approaches Research on the EOR has primarily focused on the “usual cast of characters” including such dependent variables as satisfaction, commitment, performance and citizenship. An extension of prior consequences to include dependent variables or outcomes relevant to the EOR such as the extent of participatory or shared decision making in the relationship, the level of trust between parties over time, outcomes taken from positive relational psychology, such as quality of work life, physical and mental health (of individual and organization), morality, retrospective assessments of the relationship after it has ended, goal achievement, performance levels, innovation, and value creation vs. destruction. These consequences should focus on the societal, as well as individual and organizational effects. This expansion of the research domain will clearly require development of EOR concepts at multiple levels of analysis, as well as explicating the linkages across levels. We believe such expansion is do-able, though obviously challenging.

The Employee-Organization Relationship

359

A general characteristic of the overall body of research so far carried out on the EOR is that it has been by-and-large somewhat piecemeal. This has had the advantage of bringing a number of independent perspectives to studying the phenomenon and thus avoiding stifling orthodoxy in both issues being studied and in the methodologies employed. What may be useful to add to this mix in the future is to design and implement a large-scale, comprehensive, multi-investigator study of the phenomenon in a variety of organizations and even in a set of different countries. Here the example of the recent GLOBE study of leadership (House, Javidan & Dorfman, 2001), carried out in over 180 companies in more than 60 countries, but with common measuring instruments and systematic focus on key variables, could provide a model for an analogous investigation of the EOR. The implementation of such a data-gathering project might be one way to generate additional research leverage on an obviously complex subject.

Summary Research has informed our understanding of the EOR from the employee’s and from the organization’s perspective. The challenge is to integrate these findings to understand the actual relationship jointly considering both the employee and the organization. To do so will require theory development, new measurement techniques, and application of new (at least to this literature) methods and analytical procedures. Nonetheless, we consider the endeavor to be particularly important at this point in time. The challenges for organizational leaders and members in the twentyfirst century promise to continue, with global changes putting pressures on EORs that have vast implications for individuals, organizations, and nations. “It is not the strongest of the species that survive, nor the most intelligent, but the one most responsive to change.” Charles Darwin

ACKNOWLEDGMENTS This article has been a team effort from beginning to end. We wish to acknowledge all of the authors for the generosity with which they contributed to the project, and for making this a truly comprehensive and thought-provoking article. Our deepest gratitude to Joe Martocchio and Jerry Ferris for giving us this opportunity, and for believing in our ability to carry it off! The Employment Relationship Workshop, in which most of the authors participated in the spring of 2002, was the incubator for many of the ideas that appear in

360

LYNN M. SHORE ET AL.

this article. Support for the Workshop was provided by the W. T. Beebe Institute of Personnel and Employment Relations at Georgia State University; the Robert H. Smith School of Business and the Center for Human Capital, Innovation and Technology at the University of Maryland, College Park; the Department of Psychology and the C. T. Bauer School of Business at the University of Houston; and the London School of Economics and Political Science. We are very grateful to our respective institutions for their support and to the Graduate School of Management at the University of California, Irvine, where Lynn Shore was a Visiting Professor during the writing of this article.

REFERENCES Anderson, N., & Schalk, R. (1998). The psychological contract in retrospect and prospect. Journal of Organizational Behavior, 19, 637–647. Anderson, N., & Thomas, H. D.C. (1996). Work group socialization. In: M. A. West (Ed.), Handbook of Work Groups (pp. 423–450). Chichester: Wiley. Arad, S., Hanson, M. A., & Schneider, R. J. (1999). Organizational context. In: N. G. Peterson, M. D. Mumford, W. C. Borman, P. R. Jenneret & E. A. Fleishman (Eds), An Occupational Information System for the Twentyfirst Century: The Development of O∗Net (pp. 147–174). Washington, DC: American Psychological Association. Argyis, A. (1960). Understanding organizational behavior. Homewood, IL: Dorsey Press. Arthur, J. B. (1994). Effects of human resource systems on manufacturing performance and turnover. Academy of Management Journal, 37, 670–687. Arthurs, H. (2001). Reinventing labor law for the global economy. Berkeley Journal of Employment and Labor Law, 22, 271–294. Barker, K., & Christtensen, K. (1998). Contingent work: American employment relations in transition. Ithaca, NY: Cornell University Press. Barnard, C. I. (1938). The functions of the executive. Cambridge, MA: Harvard University Press. Bennett, J. B., Cook, R., & Pelletier, K. R. (2003). Toward an integrated framework for comprehensive organizational wellness: Concepts, practices, and research in workplace health promotion. In: J. C. Quick & L. E. Tetrick (Eds), Handbook of Occupational Health Psychology (pp. 69–95). Washington, DC: American Psychological Association. Berg, J. H., & Wiebe, F. A. (1993). Resource exchange in the workplace: Exchange of economic and interpersonal resources. In: U. G. Foa, J. Converse, Jr., K. Y. Toernblom & E. B. Foa (Eds), Resource Theory: Explorations and Applications (pp. 97–122). San Diego, CA: Academic Press. Bernstein, A. (1995). Outsourced – and out of luck. Business Week, July 17th, pp. 60–61. Blau, P. M. (1964). Exchange and power in social life. New York: Wiley. Bliss, W. (2000). The wheel of misfortune. HR Magazine, 45(5), 207–218. Boh, W., Slaughter, S., & Ang, S. (2001). Is information technology a “boundaryless profession”? A sequence analysis of the career histories of IT professionals from 1979–1998. Academy of Management Proceedings, A1–A7. Bolino, M. C., & Turnley, W. H. (2002, August). Help-seeking and help-refusing in organizational settings: An impression-management approach. Paper presented at the 2002 Academy of

The Employee-Organization Relationship

361

Management Meeting Symposium on “New Perspectives on Helping: Beyond Social Exchange”, Denver, CO. Brass, D. J. (2000). Networks and frog ponds: Trends in multilevel research. In: K. J. Klein & S. W. J. Koszlowski (Eds), Multilevel Theory, Research, and Methods in Organizations (pp. 557–571). San Francisco: Jossey-Bass. Brefort, S. F. (2002). Labor and employment law at the new millennium: A historical overview and critical assessment. Boston College Law Review, 43, 351–460. Bureau of Labor Statistics (1995). Contingent and alternative employment arrangements. Washington, DC: U.S. Department of Labor (Report No. 900). Bureau of Labor Statistics (2001). U.S. Department of Labor, Bureau of Labor Statistics, May and February supplements to the current population survey. Cadin, L. (2000). Does psychological contract work in France? In: D. M. Rousseau & R. Schalk (Eds), Psychological Contracts in Employment: Cross-national Perspectives (pp. 67–86). Thousand Oaks, CA: Sage. Capelli, P. (1999). The new deal at work. Cambridge, MA: Harvard Business School Press. Cappelli, P., & Sherer, P. D. (1991). The missing role of context in OB: The need for a meso-level approach. In: B. M. Staw & L. L. Cummings (Eds), Research in Organizational Behavior (Vol. 13, pp. 55–110). Greenwich, CT: JAI Press. Corbett, W. R. (2001). Waiting for the labor law of the twenty-first century: Everything old is new again. Berkeley Journal of Employment and Labor Law, 23, 259–306. Corbin, A. L. (1993). Corbin on contracts. St. Paul: West Publishing Company. Cosmides, L., & Tooby, J. (1989). Evolutionary psychology and the generation of culture: II. Case study: A computational theory of social exchange. Ethology & Sociobiology, 10, 51–97. Cosmides, L., & Tooby, J. (1992). Cognitive adaptations for social exchange. In: The Adapted Mind: Evolutionary Psychology and the Generation of Culture (pp. 163–228). London: Oxford University Press. Coyle-Shapiro, J. (2000, August). Psychological contracts: What are the measurement trade-offs and are they worth it? Paper presented at Symposium the measurement of psychological contracts: Research experiences, and recommendations at the Annual Meeting of the Academy of Management, Toronto. Coyle-Shapiro, J. (2001, August). Managers: Caught in the middle of a psychological contract muddle. Paper presented at the Annual Meeting of the Academy of Management, Washington, DC. Coyle-Shapiro, J., & Kessler, I. (2000, August). Mutuality, stability, and psychological contract breach: A longitudinal study. Paper presented at the Annual Meeting of the Academy of Management, Toronto. Coyle-Shapiro, J., & Kessler, I. (2002). Exploring reciprocity through the lens of the psychological contract: Employee and employer perspectives. European Journal of Work and Organizational Psychology, 11, 69–86. CPA (2001). Alternatives: Newsletter of the center for policy alternatives, 9(1). Craig, D. D., & Tetrick, L. E. (1999, May). Psychological contracts and temporary workers: The assumption of the transactional contract. Paper presented at the annual meeting for Society of Industrial and Organizational Psychology, Atlanta. Cropanzano, R., Rupp, D. E., Mohler, C. J., & Schminke, M. (2001). Three roads to organizational justice. In: J. Ferris (Ed.), Research in Personnel and Human Resources Management (pp. 1–113). Greenwich, CT: JAI Press. Dabos, G., & Rousseau, D. M. (2004). Mutuality and reciprocity in the psychological contracts of employees and employers. Journal of Applied Psychology, 89, 52–72.

362

LYNN M. SHORE ET AL.

Delery, J. E., & Doty, D. H. (1996). Modes of theorizing in strategic human resource management: Tests of universalistic, contingency and configurational performance predictions. Academy of Management Journal, 39, 802–835. Delery, J. E., & Shaw, J. D. (2001). The strategic management of people in work organizations: Review, synthesis, and extension. In: G. Ferris (Ed.), Research in Personnel and Human Resource Management (Vol. 20, pp. 165–197). Greenwich, CT: JAI Press. De Vos, A. (2002). The individual antecedents and the development of newcomer’s psychological contracts during the socialization process: A longitudinal study. Ph.D. Thesis, University of Gent, Gent. Dienesch, R. M., & Liden, R. C. (1986). Leader-member exchange model of leadership: A critique and further development. Academy of Management Review, 11, 618–634. Donaldson, T., & Dunfee, T. W. (1993, 1999 in text). Ties that bind: A social contracts approach to business ethics. Boston, MA: Harvard Business School Press. Dye, D., & Silver, M. (1999). The origins of O∗Net. In: N. G. Peterson, M. D. Mumford, W. C. Borman, P. R. Jenneret & E. A. Fleishman (Eds), An Occupational Information System for the Twenty-First Century: The Development of O∗Net (pp. 9–19). Washington, DC: American Psychological Association. Earley, P. C., & Gibson, C. B. (1998). Taking stock in our progress on individualism-collectivism: 100 years of solidarity and community. Journal of Management, 24, 265–304. Edwards, J., & Bunderson, J. S. (in press). Violations of principle: Ideological currency in the psychological contract. Academy of Management Review. Eisenberg, M. A. (2000). Why there is not law of relational contracts. Northwestern University Law Review, 805–821. Eisenberg, M. A. (2001). Contracts and property law: The emergence of dynamic contract law. Theoretical Inquiries in Law, 2, 1–76. Eisenberger, R., Huntington, R., Hutchison, S., & Sowa, D. (1986). Perceived organizational support. Journal of Applied Psychology, 71, 500–507. Eisenberger, R., Stinglhamber, F., Vandenberghe, C., Sucharski, I. L., & Rhoades, L. (2002). Perceived supervisor support: Contributions to perceived organizational support and employee retention. Journal of Applied Psychology, 87, 565–573. Etzioni, A. (1961). A comparative analysis of complex organizations. New York: Free Press. Firebaugh, G. (1978). A rule for inferring individual-level relationships from aggregate data. American Sociological Review, 43, 557–572. Firebaugh, G. (1980). Groups as contexts and frog ponds. In: K. H. Roberts & L. Burstein (Eds), Issues in Aggregation. New Directions for Methodology of Social Behavioral Science, No. 6. San Francisco: Jossey-Bass. Fiske, J. D., & Taylor, S. E. (1984). Social cognition. New York: Random House Blackwell. Foa, E. B., & Foa, U. G. (1976). Resource theory of social exchange. In: J. W. Thibaut, J. Spence & R. C. Carson (Eds), Contemporary Topics in Social Psychology (??). Morristown, NJ: General Learning Press. Folger, R. (2004). Justice and employment: Moral retribution as a contra-subjugation tendency. In: J. Coyle-Shapiro, L. Shore, S. Taylor & L. Tetrick (Eds), The Employment Relationship: Examining Psychological and Contextual Perspectives (pp. 29–47). Oxford, UK: Oxford University Press. Folger, R., & Kass, E. E. (2000). Social comparison and fairness: A counterfactual simulations perspective. In: J. Suls & L. Wheeler (Eds), Handbook of Social Comparison: Theory and Research (pp. 423–441). New York: Kluwer/Plenum.

The Employee-Organization Relationship

363

Frank, R. H. (1985). Choosing the right pond: Human behavior and the quest for status. New York: Oxford University Press. Frenkel, S. J. (2001). Globalization, athletic footwear commodity chains and employment relations in China. Organization Studies, 22, 531–562. Gallagher, D., & McLean Parks, J. (2001). I pledge thee my troth . . . contingently: Commitment and the contingent work relationship. Human Resources Management Review, 11(3), 181–208. Gaspersz, J., & Ott, M. (1996). Management van employability. Assen: Van Gorcum/Stichting Management Studies. Gephart, R. P., Jr. (2002). Introduction to the brave new workplace: Organizational behavior in the electronic age. Journal of Organizational Behavior, 23, 327–344. Goldberg, B. (1994). The new world of work. Executive Excellence, 11(7), 8–9. Goodman, P. S. (2001). Missing organizational linkages. Thousand Oaks, CA: Sage. Gouldner, A. W. (1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25, 161–178. Graen, G. B., & Uhl-Bien, M. (1995, 1985 in text). Relationship-based approach to leadership: Development of leader-member exchange (LMX) theory of leadership over 25 years. Leadership Quarterly, 6, 219–247. Griffin, D. W., & Ross, L. (1991). Subjective construal, social inference and human understanding. Advances in Experimental Social Psychology, 24, 319–359. Guest, D., & Conway, N. (2000, August). Can an organization have a psychological contract? A conceptual and empirical analysis. Paper presented at the Academy of Management, Toronto. Hackman, J. R. (1999). Thinking differently about context. In: E. A. Mannix, M. A. Neale & R. Wagemen (Eds), Research on Managing Groups and Teams (Vol. 2, pp. 233–247). Stanford, CT: JAI Press. Hackman, J. R., & Oldham, G. R. (1976). Motivation through the design of work: Test of a theory. Organizational Behavior and Human Performance, 16, 250–279. Hartley, J. (1995). Challenge and change in employment relations: In: L. Tetrick & J. Barling (Eds), Changing Employment Relations: Behavioral and Social Perspectives (pp. 3–30).Washington, DC: American Psychological Association. Henson, K. (1996). Just a temp. Philadephia: Temple University Press. Herriott, P., & Pemberton, C. (1995). New deals. Chichester: Wiley. Hiltrop, J. M. (1995). The changing psychological contract: The human resource challenge of the 1990s. European Management Journal, 13, 286–294. Ho, V. T. (2002). Evaluation of psychological contract fulfillment: A social network perspective. Carnegie Mellon University Dissertation, Pittsburgh, PA. Ho, V. T., Ang, S., & Straub, D. (2003). When subordinates become IT contractors: Persistent managerial expectations in IT outsourcing. Information Systems Research, 14, 66–88. Ho, V. T., Levesque, L. L., & Rousseau, D. M. (2003). Social networks and the psychological contract: Structural holes, cohesive ties, and employees’ perceived organizational obligations. Unpublished manuscript, Carnegie Mellon University. Hofmann, D. A., & Tetrick, L. E. (2003). On the etiology of health: Implications for “organizing” individual and organizational health. In: D. A. Hofmann & L. E. Tetrick (Eds), Health and Safety in Organizations: A Multilevel Perspective (pp. 1–26). Jossey-Bass. House, R., Javidan, M., & Dorfman, P. (2001). Project GLOBE: An introduction. Applied Psychology: An International Review, 50, 489–505. Huiskamp, R., & Schalk, R. (2002). Psychologische contracten in arbeidsrelaties: De stand van zaken in Nederland. Gedrag en Organisatie, 15, 370–385.

364

LYNN M. SHORE ET AL.

Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38, 635–672. Ibarra, H., & Andrews, S. (1993). Power, social influence and sense making: Effects of network centrality and proximity on employee perceptions. Administrative Science Quarterly, 38, 277–303. Ichinowski, C., Shaw, K., & Prennushi, G. (1997). The effect of human resource management practices on productivity: A study of steel finishing lines. American Economic Review, 87, 291–313. Jackson, S. E., Schuler, R. S., & Rivero, J. C. (1989). Organizational characteristics as predictors of personnel practices. Personnel Psychology, 42, 727–786. Johns, G. (2001). In praise of context. Journal of Organizational Behavior, 22, 31–42. Kalleberg, A. L. (2001). The advent of the flexible workplace: Implications for theory and research. In: D. B. Cornfield, K. E. Campbell & H. J. McCammons (Eds), Working in Restructured Workplaces (pp. 437–454). Thousand Oaks, CA: Sage. Kary, J. H. (2000). Contract law and the social contract: What legal history can teach us about the political theory of Hobbes and Locke. University of Ottawa Law Review, 31, 73–91. Katz, D., & Kahn, R. L. (1966, 1978). The social psychology of organizations. New York: Wiley. King, J. E. (2000). White-collar reactions to job insecurity and the role of the psychological contract: Implications for human resource management. Human Resource Management, 39, 79–92. Klein, N. (2002). No logo. NY: Picador USA. Kohler, T. C. (1999). The employment relationship and its ordering at century’s end: Reflection of emerging trends in the United States. Boston College Law Review, 41, 103–123. Kotter, J. P. (1973). The psychological contract: Managing the joining-up process. California Management Review, 15, 91–99. Kozlowski, S. W. G., & Klein, K. J. (2000). A multilevel approach to theory and research in organizations. In: K. J. Klein & S. W. G. Kozlowski (Eds), Multilevel Theory, Research, and Methods in Organization: Foundations, Extensions, and New Directions (pp. 3–90). San Francisco: JosseyBass. Landers, R. M., Rebitzer, J. B., & Taylor, L. J. (1996). Rat race redux: Adverse selection in the determination of work hours in law firms. American Economic Review, 329–348. Lepak, D. P., & Snell, S. A. (1999). The human resource architecture: Toward a theory of human capital allocation and development. Academy of Management Review, 24, 31–48. Lepak, D. P., & Snell, S. A. (2002). Examining the human resource architechure: The relationships among human capital, employment, and human resource configurations. Journal of Management, 28, 517–543. Lepak, D. P., Takeuchi, R., & Snell, S. A. (2003). Employment flexibility and firm performance: Examining the interaction effects of employment modes, technological intensity, and environment dynamism. Journal of Management, 5, 681–703. Lepak, D. P., Taylor, M. S., Tekleab, A., Marrone, J., & Cohen, D. (2003). Examining variability in high investment human resource system use across employee groups, firms, and industries. Manuscript under review. Levesque, L. L. (2001). Role creation processes in start-up firms. Carnegie Mellon University Dissertation, Pittsburgh, PA. Levinson, H., Price, C., Munden, K., Mandl, H., & Solley, C. (1962). Men, management, and mental health. Cambridge, MA: Harvard University Press. Lewicki, R. J., Barry, B., Saunders, D. M., & Minton, J. W. (2003). Negotiation (4th ed.). McGraw-Hill and Irwin. Lewin, T. (1994). Low pay and closed doors greet young in job market. New York Times (March 10), 1-A, 12-A.

The Employee-Organization Relationship

365

Liden, R. C., & Graen, G. (1980). Generalizability of the vertical dyad linkage model of leadership. Academy of Management Journal, 23, 451–465. Liden, R. C., Sparrowe, R. T., & Wayne, S. J. (1997). Leader-member exchange theory: The past and potential for the future. Research in Personnel and Human Resources Management, 15, 47–119. Liden, R. C., Wayne, S. J., & Stilwell, D. (1993). A longitudinal study on the early development of leader-member exchanges. Journal of Applied Psychology, 78, 662–674. Locke, J. (1965). Two treatises of government. In: P. Laslett (Ed.). New York: New American Library. Lord, R. G., & Foti, R. J. (1986). Schema theories, information processing and organizational behaviors. In: H. P. Sims, Jr. & D. A. Gioia (Eds), The Thinking Organization (pp. 20–48). San Francisco: Jossey-Bass. Luthans, F. (2002). Positive organizational behavior: Developing and managing psychological strengths. Academy of Management Executive, 16, 57–72. Macneil, I. (1972). The many futures of contracts. Southern California Law Review, 47(May), 691–816. Macneil, I. R. (1980). The new social contract: An inquiry into modern contractual relations. New Haven: Yale University Press. Macneil, I. (1985). Relational contracts: What we do and do not know. Wisconsin Law Review, 483–525. Mahoney, T. A., & Watson, M. R. (1993). Evolving modes of workforce governance: An evaluation. In: B. E. Kaufman & M. M. Kleiner (Eds), Employee Representation: Alternatives and Future Directions (pp. 135–168). Madison, WI: Industrial Relations Research Association, University of Wisconsin. March, J. G., & Simon, H. A. (1958). Organizations. New York: Wiley. Martin, J. (1992). Cultures in organizations: Three perspectives. New York: Oxford University Press. Maslyn, J. M., & Uhl-Bien, M. (2001). Leader-member exchange and its dimensions: Effects of self-effort and other’s effort on relationship quality. Journal of Applied Psychology, 86, 697–708. Masters, J. K., & Miles, G. (2002). Predicting the use of external labor arrangements: A test of the transaction cost perspective. Academy of Management Review, 45, 431–442. Masterson, S. S., Lewis, K., Goldman, B. M., & Taylor, M. S. (2000). Integrating justice and social exchange: The differing effects of fair procedures and treatment on work relationships. Academy of Management Journal, 43, 738–748. Matusik, S., & Hill, C. (1998). The utilization of contingent work, knowledge creation, and competitive advantage. Academy of Management Review, 23, 680–697. McCurry, P. M. (1999). Trial, American style. Director, 53(3), 3–4. McLean Parks, J. (1993). Organizational contracting: Rationality and power. Paper presented at the Pierce Memorial Conference on Non-rational decision making, Cornell University, Ithaca, NY. McLean Parks, J. (1997). The fourth arm of justice: The art and science of revenge. Monograph In: R. Lewicki, B. Sheppard & B. Bies (Eds), Research on Negotiation in Organization (pp. 113–144). Greenwich, CT: JAI Press. McLean Parks, J., & Kidder, D. L. (1994). “Till death us do part . . . ” Changing work relationships in the 1902. In: C. L. Cooper & D. M. Rousseau (Eds), Trends in Organizational Behavior (Vol. 1, pp. 111–136). Wiley. McLean Parks, J., Kidder, D., & Gallagher, D. (1998). Fitting square pegs into round holes: Mapping the domain of contingent work arrangements onto the psychological contract. Journal of Organizational Behavior, 19, 697–730. McLean Parks, J., & Schmedemann, D. A. (1994). When promises become contracts: Implied contracts and handbook provisions on job security. Human Resource Management, 33, 403–423.

366

LYNN M. SHORE ET AL.

McLean Parks, J., & smith, F. (1998). Organizational contracting: A “rational” exchange? In: J. F. Halpern & R. N. Stern (Eds), Debating Rationality: Nonrational Aspects of Organizational Decision Making (pp. 125–154). Ithaca, NY: ILR Press. Menninger, K. (1958). Theory of psychoanalytic technique. New York: Basic Books. Mischel, W. (1977). The interaction of person and situation. In: I. D. Magnusson & N. S. Endler (Eds), Personality at the Crossroads: Current Issues in International Psychology (pp. 333–352). Hillsdale, NJ: Erlbaum. Morrison, E. W., & Robinson, S. L. (1997). When employees feel betrayed: A model of how psychological contract violation develops. Academy of Management Review, 22, 226–256. Morrison, E. W., & Robinson, S. L. (2004). The employment relationship from two sides: Incongruence in employees’ and employers’ perceptions of obligations. In: J. Coyle-Shapiro, L. Shore, S. M. Taylor & L. E. Tetrick (Eds), The Employment Relationship: Psychological and Contextual Perspectives (pp. 161–180). Oxford: Oxford University Press. Moynihan, L. M., Gardner, T. M. & Wright, P. M. (2002). High performance HR practices and customer satisfaction: Employee process mechanisms. Working Paper, Centre for Advanced HR Studies, Cornell University. Mowday, R. T., & Sutton, R. I. (1993). Organizational behavior: Linking individuals and groups to contexts. In: L. W. Porter & M. R. Rosenzweig (Eds), Annual Review of Psychology (Vol. 44, pp. 195–229). Palo Alto, CA: Annual Reviews. Nelson, D. L., Quick, J. C., & Joplin, J. R. (1991). Psychological contracting and newcomer socialization. Journal of Social Behavior and Personality, 6, 55–72. Newell, S. (1995). The healthy organization: Fairness, ethics and effective management. London: Routledge. Nicholson, N., & Johns, G. (1985). The absence culture and the psychological contract. Academy of Management Review, 10, 397–407. Organ, D. (1988). Organizational citizenship behavior: The good soldier syndrome. Lexington, MA: Lexington Books. Orpen, C. (1994). The effects of organizational and individual career management on career success. International Journal of Manpower, 15(1), 27–37. Ouchi, W. G. (1980). Markets, bureaucracies, and clans. Administrative Science Quarterly, 25, 129– 141. Pearce, J. (1993). Toward an organizational behavior of contract workers: Their psychological involvement and effects on employee co-workers. Academy of Management Journal, 36, 1082–1096. Pfeffer, J. (1981). Power in organizations. Marshfied, MA: Pitman. Pfeffer, J., & Baron, J. N. (1988). Taking the workers back out: Recent trends in the structuring of employment. In: B. M. Staw & L. L. Cummings (Eds), Research in Organizational Behavior (Vol. 10, pp. 257–303). Greenwich, CT: JAI Press. Porter, L. W., Pearce, J. L., Tripoli, A., & Lewis, K. M. (1998). Differential perceptions of employers’ inducements: Implications for psychological contracts. Journal of Organizational Behavior, 19, 769–782. Pruitt, D. (1981). Negotiation behavior. New York, NY: Academic Press. Rhoades, L., & Eisenberger, R. (2002). Perceived organizational support: A review of the literature. Journal of Applied Psychology, 87, 698–714. Robinson, S. L. (1996). Trust and breach of the psychological contract. Administrative Science Quarterly, 41, 574–599. Robinson, S. L., Kraatz, M., & Rousseau, D. M. (1994). Changing obligations and the psychological contract: A longitudinal study. Academy of Management Journal, 37, 137–152.

The Employee-Organization Relationship

367

Robinson, S. L., & Morrison, E. W. (1995). Psychological contracts and OCB: The effect of unfulfilled obligations on civic virtue behavior. Journal of Applied Psychology, 16, 289–298. Roehling, M. V. (1997). The origins and early development of the “psychological contract” construct. Journal of Management History, 3, 204–217. Roehling, M. V. (2002). The “good cause” norm in employment relations: Empirical evidence and policy implications. Employee Responsibilities and Rights Journal, 14, 91–104. Roehling, M. V. (2004). Legal theory: Contemporary contract law perspectives and insights for employment relationship theory. In: J. Coyle-Shapiro, L. Shore, S. M. Taylor & L. E. Tetrick (Eds), The Employment Relationship: Psychological and Contextual Perspectives (pp. 65–93). Oxford: Oxford University Press. Roehling, M. V., Cavanaugh, M. A., Moynihan, L. M., & Boswell, W. R. (2000). The nature of the new employment relationship: A content analysis of the practitioner and academic literatures. Human Resource Management, 39(4), 305–320. Romme, A. G. L. (1999). Domination, self-determination and circular organizing. Organization Studies, 20, 801–831. Ross, M., Green, D., & House, P. (1977). The “false consensus effect”: An egocentric bias in social perception processes. Journal of Experimental Social Psychology, 13, 279–301. Rousseau, D. M. (1985). Issues of level in organizational research: Multi-level and cross-level perspectives. In: L. L. Cummings & B. Staw (Eds), Research in Organizational Behavior (Vol. ?, pp. 1–37). Rousseau, D. M. (1989). Psychological and implied contracts in organizations. Employee Responsibilities and Rights Journal, 2, 121–139. Rousseau, D. M. (1990). New hire perceptions of their own and their employer’s obligations: A study of psychological contracts. Journal of Organizational Behavior, 11, 389–400. Rousseau, D. M. (1995). Psychological contracts in organizations: Understanding written and unwritten agreements. California: Sage. Rousseau, D. M. (1997). Organizational behavior in the new organizational era. In: J. T. Spence, J. M. Darley & D. J. Foss (Eds), Annual Review of Psychology (Vol. 48, pp. 515–546). Palo Alto, CA: Annual Reviews. Rousseau, D. M. (1998). The “problem” of the psychological contract considered. Journal of Organizational Behavior, 19, 665–671. Rousseau, D. M. (2000). Personal communication. Rousseau, D. M. (2001). Idiosyncratic deals: Flexibility vs. fairness? Organizational Dynamics, 29, 260–273. Rousseau, D. M., & Fried, Y. (2001). Location, location, location: Contextualizing organizational research. Journal of Organizational Behavior, 22, 1–13. Rousseau, D. M., & Kim, T. G. (2003). Idiosyncratic deals and the employment relationship. Working Paper, Carnegie Mellon University, Pittsburgh, PA. Rousseau, D., & McLean Parks, J. (1993). The contracts of individuals and organizations. Monograph In: L. L. Cummings & B. Staw (Eds), Research in Organizational Behavior (Vol. 15, pp. 1–43). Greenwich, CT: JAI Press. Rousseau, D. M., & Schalk, R. (2000). Psychological contracts in employment:Cross-national perspectives. Thousand Oaks, CA: Sage. Rousseau, D. M., & Shperling, Z. (2003). Pieces of the action: Ownership and the psychological contract. Academy of Management Review, 28, 553–570. Rousseau, D. M., & Tijoriwala, S. A. (1998). Assessing psychological contracts: Issues, alternatives, and measures. Journal of Organizational Behavior, 19, 679–695.

368

LYNN M. SHORE ET AL.

Rousseau, D. M., & Tijoriwala, S. A. (1999). What’s a good reason to change? Motivated reasoning and social accounts in promoting organizational change. Journal of Applied Psychology, 84, 514–528. Rousseau, D. M., & Wade-Benzoni, K. A. (1994). Linking strategy and human resource practices: How employee and customer contracts are created. Human Resource Management, 33, 463–489. Rubery, J., Marchington, M., Coopke, F. L., & Vincent, S. (2002). Changing organizational forms and the employment relationship. Journal of Management Studies, 39, 645–672. Sahlins, M. (1972). Stone age economics. New York: Aldine De Gruyter. Schalk, R., & Freese, C. (1997). New facets of commitment in response to organizational change: Research trends and the Dutch experience. In: C. L. Cooper & D. M. Rousseau (Eds), Trends in Organizational Behaviour (pp. 107–123). New York: Wiley. Schein, E. H. (1965). Organizational psychology. Englewood Cliffs, NJ: Prentice-Hall. Schein, E. H. (1970). Organizational psychology (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall. Schein, E. H. (1980). Organizational psychology (3rd ed.). Englewood Cliffs, NJ: Prentice-Hall. Scott, W. R. (1995). Institutions and organizations. Thousand Oaks, CA: Sage. Seers, A. (1989). Team-member exchange quality: A new construct for role-making research. Organizational Behavior and Human Decision Processes, 43, 118–135. Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology: An introduction. American Psychologist, 55, 5–14. Sherer, P. D. (1996). Toward an understanding of the variety in work arrangements: The organization and labor relationships framework. In: C. L. Cooper & D. M. Rousseau (Eds), Trends in Organizational Behavior (Vol. 3, pp. 99–122). New York: Wiley. Sherer, P. D., & Leblebici, H. (2001). Bringing variety and change into strategic human source management research. In: G. Ferris (Ed.), Research in Personnel and Human Resource Management (Vol. 20, pp. 199–230). JAI Press. Sherer, P. D., Rogovsky, N., & Wright, N. (1998). What drives employment relationships in taxicab organizations? Linking agency to firm capabilities and strategic opportunities. Organization Science, 9, 34–48. Shore, L. M., & Barksdale, K. (1998). Examining degree of balance and level of obligations in the employment relationship: A social exchange approach. Journal of Organizational Behavior, 19, 731–744. Shore, L. M., & Tetrick, L. E. (1994). The psychological contract as an explanatory framework in the employment relationship. In: C. Cooper & D. Rousseau (Eds), Trends in Organizational Behavior (Vol. 1, pp. 91–109). New York: Wiley. Shore, L. M., Tetrick, L. E., & Barksdale, K. (1999). Transactional and relational exchange relationships. Paper presented at the Society for Industrial and Organizational Psychology, Atlanta. Sparrow, P. R. (1996). Transitions in the psychological contract: Some evidence from the banking sector. Human Resource Management, 6(4), 75–93. Sparrow, P. R. (2000). The new employment contract: Psychological implications for future work. In: R. J. Burke & C. L. Cooper (Eds), The Organization in Crisis: Downsizing, Restructuring and Privatization (??). Oxford: Blackwell. Sparrowe, R., Dirks, K., Bunderson, S., & McLean Parks, J. (2003). Reinventing the wheel . . . or spinning our wheels: Social exchange and discretionary attitudes and outcomes in organizations. Working Paper, Olin School of Business, Washington University at St. Louis. Sparrowe, R. T., & Liden, R. C. (1997). Process and structure in leader-member exchange. Academy of Management Review, 22, 522–552.

The Employee-Organization Relationship

369

Spiedel, R. E. (2000). The characteristics and challenges of relational contracts. Northwestern University Law Review, 94, 823–846. Staber, U., & Aldrich, H. (1989). Human resource strategies: Some ecological cracks? Industrial Relations Journal, 20, 110–118. Strong, M. H., Jeanneret, P. R., McPhail, S. M., Blakley, B. R., & D’Egidio, E. L. (1999). Work context: Taxonomy and measurement of the work environment. In: N. G. Peterson, M. D. Mumford, W. C. Borman, P. R. Jenneret & E. A. Fleishman (Eds), An Occupational: An Information System for the Twenty-first Century: The Development of O∗Net (pp. 127–145). Washington, DC: American Psychological Association Books. Tekleab, A. G., & Taylor, M. S. (2001, August). Aren’t there two parties in an employment relationship? Antecedents and consequences of organization-employee agreement on contract obligations and violations. Paper presented at the Annual Meeting of the Academy of Management, Washington, DC. Tetrick, L. E. (2003). Individual and organizational health. In: P. Perrewe & D. Ganster (Eds), Research in Occupational Stress and Well-Being (Vol. 2, pp. 117–141). Stamford, CT: JAI Press. Thomas, H. D., & Anderson, N. (1998). Changes in newcomers’ psychological contracts during organizational socialization: A study of recruits entering the British Army. Journal of Organizational Behavior, 19, 745–767. Thompson, C. (1998). The temp. This Magazine, February 22. Thompson, L. L. (2001). The mind and heart of the negotiator (2nd ed.). Upper Daddle River, NJ: Prentice-Hall. Tolbert, P. S. (1996). Occupations, organizations, and boundaryless careers. In: M. B. Arthur & D. M. Rousseau (Eds), The Boundaryless Career: A New Employment Principle for a New Organizational Era (pp. 331–349). New York, NY: Oxford University Press. Tsui, A. S., Pearce, J. L., Porter, L. W., & Hite, J. P. (1995). Choice of employee-organization relationship: Influence of external and internal organizational factors. In: G. R. Ferris (Ed.), Research in Personnel and Human Resource Management (Vol. 13, pp. 117–151). Greenwich, CT: JAI Press. Tsui, A. S., Pearce, J. L., Porter, L. W., & Tripoli, A. M. (1997). Alternative approaches to the employeeorganization relationship: Does investment pay off? Academy of Management Journal, 40, 1089–1121. Tsui, A., & Wang, D. (2002). Employment relationships from the employer’s perspective: Current research and future directions. In: C. L. Cooper & I. T. Robertson (Eds), International Review of Industrial and Organizational Psychology (Vol. 17, pp. 77–114). ???. Tsui, A. S., Wang, D., & Zhang, Y. (2002). Employment relationship with Chinese middle managers: Exploring differences between state and non state owned firms. In: A. S. Tsui & C. M. Lau (Eds), The Management of Enterprises in the People’s Republic of China (pp. 347–374). Boston: Kluwer Academic Press. Turnley, W. H., & Feldman, D. C. (1999). The impact of psychological contract violations on exit, voice, loyalty, and neglect. Human Relations, 52, 895–922. Turnley, W. H., & Feldman, D. C. (2000). Re-examining the effects of psychological contract violations: Unmet expectations and job dissatisfaction as mediators. Journal of Organizational Behavior, 21, 25–42. Van Dyne, L., & Ang, S. (1998). Organizational citizenship behavior of contingent workers in Singapore. Academy of Management Journal, 41, 692–703. Walton, R. E. (1985). From control to commitment in the workplace. Harvard Business Review, 63, 77–84.

370

LYNN M. SHORE ET AL.

Wang, D., Tsui, A. S., Zhang, Y., & Ma, L. (2003). Employment relationship and firm performance: A contingency analysis in an emerging economy. Journal of Organizational Behavior, 24, 511–536. Wayne, S. J., Shore, L. M., & Liden, R. C. (1997). Perceived organizational support and leader-member exchange: A social exchange perspective. Academy of Management Journal, 40, 82–111. Weick, K. M. (1996). Enactment and the boundaryless career: Organizing as we work. In: M. B. Arthur & D. M. Rousseau (Eds), The Boundaryless Career: A New Employment Principle for a New Organizational Era (pp. 40–57). New York: Oxford. Wood, S., & Menezes, L. (1998). High commitment management in the U.K.: Evidence from the workplace industrial relations survey, and employers’ manpower and skills practices survey. Human Relations, 51, 485–515. Youndt, M. A., Snell, S. A., Dena, J. E., & Lepak, D. P. (1996). Human resource management, manufacturing strategy, and firm performance. Academy of Management Journal, 39, 836–866.

ABOUT THE AUTHORS

Purnima Bhaskar-Shrinivas is a doctoral student at the Department of Management and Organization, Pennsylvania State University. She received an MBA in Marketing from NMIMS, Bombay and a Masters in Management from the University of Bombay, India. Her research interests include cross-cultural work role adaptation, organizational change and artificial neural network modeling in organizational behavior. Her work has been presented at various conferences in management and psychology, including Academy of Management and SIOP. She also serves as a reviewer for the Organizational Development and Change (ODC) Division of the Academy of Management. Prior to her academic career, she worked as a management consultant at Accenture (erstwhile Andersen Consulting), India. Philip Bobko is Professor of Management and Psychology at Gettysburg College. His publications are in methodology, measurement, management, and industrial/organizational psychology. Content domains include test fairness, adverse impact, moderated regression analysis, validation methods, goal setting, decision making, utility analysis, and performance standard setting. He has also published a text on correlation and regression analysis (Sage), co-authored several handbook chapters in industrial/organizational psychology, and served as editor of Journal of Applied Psychology. His Ph.D. is from Cornell University and his B.S. is from MIT. Jacqueline A.-M. Coyle-Shapiro is a reader in Organizational Behavior in the Department of Industrial Relations at the London School of Economics where she received her Ph.D. Prior to this, she was a lecturer in Management Studies at the University of Oxford. She is a consulting editor for the Journal of Organizational Behavior and the Journal of Occupational and Organizational Psychology. She has served as guest editor for the Journal of Organizational Behavior with Lynn Shore on a special issue titled Employment Relationships: Exchanges between Employees and Employers. Her current research interests include the employment relationship, psychological contracts, organizational citizenship behavior, and organizational change. Her work has appeared in such journals as the Journal of 371

372

Vocational Behavior, the Journal of Applied Behavioural Science and the Journal of Organizational Behavior. She has edited The Employment Relationship: Contextual and Psychological Perspectives published by Oxford University Press with Lynn Shore, Lois Tetrick and Susan Taylor. Jerald Greenberg is the Abramowitz Professor of Business Ethics and Professor of Organizational Behavior at the Ohio State University’s Fisher College of Business. Professor Greenberg is co-author of one of the best-selling college texts on organizational behavior, Behavior in Organizations, which is in its third decade of publication. As a researcher, Dr. Greenberg is best known for his pioneering work on organizational justice. He has published extensively on this topic, with over 140 professional journal articles and books to his credit. Acknowledging his research contributions, Professor Greenberg has received numerous professional honors, including: a Fulbright Senior Research Fellowship, and the William Owens Scholarly Contribution to Management Award. From the Organizational Behavior Division of the Academy of Management, Professor Greenberg has won the New Concept, and twice has won the Best Paper Award. Dr. Greenberg is co-author of the forthcoming volume, Organizational Justice: A Primer, and co-editor of Advances in Organizational Justice and the forthcoming Handbook of Organizational Justice. In recognition of his life-long scientific contributions, Dr. Greenberg has been inducted as a Fellow of the American Psychological Association, the American Psychological Society, and the Academy of Management. Professor Greenberg is also past-chair of the Organizational Behavior Division of the Academy of Management. David A. Harrison is a Professor of Management at the Department of Management and Organization, Pennsylvania State University. He received an M.S. in applied statistics and a Ph.D. in I-O psychology from the University of Illinois at Urbana-Champaign. His research on work role adjustment (especially absenteeism and turnover), time, executive decision making, and organizational measurement has appeared in Academy of Management Journal, Human Resource Management Review, Information Systems Research, Journal of Applied Psychology, Journal of Management, Personnel Psychology, Organizational Behavior and Human Decision Processes, Strategic Management Journal, and elsewhere. He has served on the editorial board of Journal of Management, and currently serves on boards of the Academy of Management Journal, Organizational Research Methods, and Personnel Psychology, and will be editor of Organizational Behavior and Human Decision Processes. Violet T. Ho is an assistant Professor in Nanyang Business School at Nanyang Technological University (Singapore). She earned her Ph.D. (2002) in organizational behavior and theory from Carnegie Mellon University. Her research

373

interests include social networks, psychological contracts, and the impact of employees’ cognitive structures on work performance and other outcomes. She has published in the Academy of Management Review, Journal of Vocational Behavior, and Information Systems Research, and was awarded the Best Paper Based on a Dissertation (2003) from the Organizational Behavior Division of the Academy of Management. Robert C. Liden (Ph.D., University of Cincinnati) is Professor of Management at the University of Illinois at Chicago. His research focuses on interpersonal processes as they relate to such topics as leadership, groups, career progression and employment interviews. He has over 50 publications in journals such as the Academy of Management Journal, Academy of Management Review, Journal of Applied Psychology, Journal of Management, and Personnel Psychology. In 2000 he was inducted into the Academy of Management Journals’ Hall of Fame as a charter bronze member. He won awards (with co-authors) for the best article published in the Academy of Management Journal during 2001, as well as the best article published in Human Resource Management during 2001. He has served on the editorial boards of the Journal of Management since 1994 and the Academy of Management Journal from 1994 to 1999. He was the 1999 program chair for the Academy of Management’s Organizational Behavior Division, and was division chair in 2000–2001. Judi McLean Parks is the Reuben C. and Anne Carpenter Taylor Professor of Organizational Behavior at John M. Olin School of Business at Washington University in St. Louis. She received her Ph.D. in organizational behavior from the University of Iowa. Her research focuses on conflict and conflict resolution, the “psychological contract” between employers and employees, the impact of perceived injustice as well as the effect of gender and ethnicity on perceived justice. Recently, she has begun to explore organizational identity and its relationship to conflict in organizations. She is editor of the International Journal of Conflict Management, former executive director of the International Association for Conflict Management, and former chair of the Academy of Management’s Conflict Management Division. Author of numerous articles and chapters, her research has been published in a variety of journals, including Academy of Management Journal, Journal of Applied Psychology, and Organizational Behavior and Human Decision Processes. Robert E. Ployhart is an associate Professor at George Mason University. His primary program of research focuses on understanding staffing within the context of forces shaping contemporary Human Resources (e.g. developing multi-level staffing models, enhancing the effectiveness and acceptability of recruitment and staffing procedures, identifying cultural/subgroup influences on staffing processes). His second program of research focuses on applied

374

statistical/measurement models and research methods, such as structural equation modeling, multilevel modeling, and longitudinal modeling. He is an active member of both the Society for Industrial and Organizational Psychology and the Academy of Management, and serves on several editorial boards. Lyman W. Porter is Professor of Management in the Graduate School of Management at the University of California, Irvine, and was formerly Dean of that School. Prior to joining UCI in 1967, he served on the faculty of the University of California, Berkeley, and, also, was a visiting professor at Yale University. Currently, he serves as a member of the Academic Advisory Board of the Czechoslovak Management Center, and a member of the Board of Trustees of the American University of Armenia, and was formerly an external examiner for the National University of Singapore. Professor Porter is a past president of The Academy of Management. In 1983 received that organization’s “Scholarly Contributions to Management” Award, and in 1994 its “Distinguished Management Educator” Award. He also served as President of the Society of Industrial-Organizational Psychology (SIOP), and in 1989 was the recipient of SIOP’s “Distinguished Scientific Contributions” Award. Professor Porter’s major fields of interest are organizational psychology, management, and management education. He is the author, or co-author, of 11 books and over 80 articles in these fields. His 1988 book (with Lawrence McKibbin), Management Education and Development (McGraw-Hill), reported the findings of a nation-wide study of business school education and post-degree management development. Belle Rose Ragins is a Professor of Management at the University of WisconsinMilwaukee and the Research Director of the UWM Institute for Diversity Education and Leadership. She studies diversity and mentoring in organizations, and her work has been published in Academy of Management Journal, Academy of Management Review, Academy of Management Executive, Journal of Applied Psychology and Psychological Bulletin. She is co-author of the book Mentoring and diversity: An international perspective. Dr. Ragins has received eight national research awards, including the Sage Award for Scholarly Contributions to Management, the ASTD Research Award, the APA Placek Award, and five Best Paper Awards from the National Academy of Management. She has or is currently serving on the boards of the Academy of Management Journal, Journal of Applied Psychology, Journal of Vocational Behavior, and Personnel Psychology. She is a Fellow of the Society for Industrial-Organizational Psychology, the American Psychological Society, and the American Psychological Association. ´ ene Roberge has a master’s degree in industrial/organizational psycholMarie-El` ogy from Universit´e du Qu´ebec a` Montr´eal and is currently a doctoral student in organizational behavior at the Ohio State University’s Fisher College of

375

Business. She has published several articles on various aspects of human resource management. Her research interests include organizational justice, deviant organizational behavior, and reactions to communication media in the workplace. Sandra L. Robinson (Ph.D., Northwestern University) is an Associate Professor of Organizational Behavior as well as an Associate Member of the Psychology Department at the University of British Columbia. Professor Robinson’s research focuses on trust, managing employment relationships, psychological contracts, workplace deviance. Her most research work focuses on territorial behavior in organizations. Her research has appeared in various journals, such as Administrative Science Quarterly, Academy of Management Journal, and Journal of Applied Psychology. Professor Robinson is an associate editor of the Journal of Management Inquiry and she also serves on the editorial boards of the Academy of Management Journal, Journal of Organizational Behavior, and the Journal of Engineering and Technology Management. She has received a number of awards, including the Ascendant Scholar Award from the Western Academy of Management, the Junior Research Excellence Award from the Faculty of Commerce at UBC, and the Cummings Scholar Award from the Academy of Management. Most recently, she was awarded a “Distinguished University Scholar” designation by the University of British Columbia. Mark V. Roehling is an Assistant Professor in the School of Labor and Industrial Relations, Michigan State University. He received his Ph.D. in Human Resource Management (HRM) from the Broad School of Management, Michigan State University, and his law degree from the University of Michigan. His primary research interests include interdisciplinary studies in HRM and the law, and responsibilities in the employment relationship (psychological, legal, and ethical perspectives). His work has appeared in academic journals (e.g. Personnel Psychology, Journal of Applied Psychology, Employee Responsibilities and Rights Journal, Human Resource Management, Journal of Business Ethics) and the popular press (e.g. The Wall Street Journal, New York Times). Dr. Roehling is currently serving on the editorial review boards for the Employee Rights and Responsibilities Journal and Human Resource Planning. He is a member of the Academy of Management, the Society for Industrial and Organizational Psychology, and the Academy of Legal Studies in Business. Patrick J. Rosopa is a doctoral student in Industrial and Organizational Psychology at the University of Central Florida (UCF). He earned a B.S. in Psychology from Tulane University and an M.S. in Industrial and Organizational Psychology from UCF. He has conducted research on teamwork mental models, the results of which have been presented at the meeting of the Society for Industrial and Organizational Psychology. His current research interests include:

376

(a) decision-making in personnel selection; and (b) the use of simulation methods to evaluate the utility of statistical techniques. Philip L. Roth is Professor of Management at Clemson University. Phil’s research interests are employment interviews, grade point average, and utility analysis. He is also interested in missing data, outliers/influential cases, and meta-analysis. He is a fellow of the Society for Industrial and Organizational Psychology and the American Psychological Society. His Ph.D. is from the University of Houston. Denise M. Rousseau is the H. J. Heinz II Professor of organizational behavior at Carnegie Mellon University’s Heinz School of Business. Professor Rousseau is President of the Academy of Management (2004–2005), and Editor of the Journal of Organizational Behavior. Dr. Rousseau is best known for her work on the changing psychological contract in employment, human resource strategies, and the effects of organizational culture on performance. She has published extensively on these topics and has over 100 professional journal articles to her credit. Her books include: Psychological Contracts in Employment (Sage, with Rene Schalk); Relational Wealth: The Advantage of Stability in a Changing Economy (Oxford, with Carrie Leana); and Psychological Contracts in Organizations (Sage). In 1996, her book, Boundaryless Careers: Work, Mobility, and Learning in the New Organizational Era (Oxford, with M. Arthur) won the Academy of Management’s George Terry Award for the best management book. Professor Rousseau’s additional professional honors, include the William A. Davis Award for scholarly research in educational administration and the National Institute for Health Care Management research award. In recognition of her life-long scientific contributions, Dr. Rousseau has been inducted as a Fellow of the American Psychological Association, the American Psychological Society, and the Academy of Management. Professor Ren´e Schalk holds a special chair in Policy and Aging at Tilburg University in the Netherlands and is a faculty member of the department of Organization Studies at Tilburg University. He earned his Ph.D. in Social and Organizational Psychology from Nijmegen University. His research focuses on complexity and dynamics in organizations, with a special focus on the psychological contract, international differences, and policy and aging. He is editor-in-chief of Gedrag en Organisatie, consulting editor for the Journal of Organizational Behavior, editorial board member of the Journal of Managerial Psychology, and reviewer for fourteen international journals. He is co-editor of the book Psychological Contracts in Employment: Cross-national Perspectives, and wrote books on absenteeism and older employees. His publications appear in journals such as Journal of Organizational Behavior, Leadership and Organization Development Journal, International Journal of Selection and Assessment,

377

European Journal of Work and Organizational Psychology, Journal of Social Behavior and Personality, and International Small Business Journal. Margaret A. Shaffer is an associate Professor with the Department of Management, Hong Kong Baptist University. She received a Ph.D. in organizational behavior and human resource management from the University of Texas-Arlington. Prior to joining HKBU, she taught at the Hong Kong Polytechnic University. Her research interests are in the areas of expatriate adjustment and performance and life balance. Her work has appeared in various management journals, including Journal of Applied Psychology, Personnel Psychology, Journal of Management, Journal of International Business Studies, and Journal of Vocational Behavior. One of her papers on expatriate adjustment (co-authored with David Harrison) received the first “Best International Paper” award from the Academy of Management. Lynn Shore is Visiting Professor at University of California, Irvine, and is joining the faculty at San Diego State University in fall of 2004. Her research on the employee-organization relationship focuses on the influence of social and organizational processes, and her work on diversity has examined the impact that composition of the work group and employee/supervisor dyads has on the attitudes and performance of work groups and individual employees. She has published numerous articles in such journals as Academy of Management Journal, Academy of Management Review, Journal of Applied Psychology, Personnel Psychology, Journal of Organizational Behavior, Human Relations, and Journal of Management. Dr. Shore is a Fellow of the American Psychological Association and the Society for Industrial and Organizational Psychology. She served as the Chair of the Human Resources Division of the Academy of Management. Dr. Shore is the associate editor for the Journal of Applied Psychology. Eugene F. Stone-Romero received his Ph.D. from the University of CaliforniaIrvine, and is now Professor of Psychology and Management at the University of Central Florida. He is a Fellow of the Society for Industrial and Organizational Psychology, the American Psychological Society, and the American Psychological Association. His research interests include moderator variable detection strategies, ethnic bias in personality measures, cross-cultural influences on organizational behavior, reactions to feedback, work-related values, job satisfaction, biases in performance ratings, and privacy in organizations. Professor Stone-Romero’s work has appeared in such outlets as the Journal of Applied Psychology, Organizational Behavior and Human Performance, Personnel Psychology, Organizational Research Methods, Journal of Vocational Behavior, Academy of Management Journal, Journal of Management, Educational and Psychological Measurement, Journal of Educational Psychology, International Review of Industrial and

378

Organizational Psychology, Research in Personnel and Human Resources Management, Applied Psychology: An International Review, Multivariate Behavioral Research, and the Journal of Applied Social Psychology. He is also the author of numerous chapters in books dealing with issues germane to the related fields of industrial and organizational psychology, human resources management, and organizational behavior. Finally, he is the author of a book titled Research Methods in Organizational Behavior, and the co-author of a book titled Job Satisfaction: How People Feel About Their Jobs and How It Affects Their Performance. M. Susan Taylor is Dean’s Professor of Human Resources, 2003 University Distinguished Scholar Teacher and Director, of the Center For Human Capital, Innovation and Technology (HCIT) at the Robert H. Smith School of Business, University of Maryland College Park. She received her Ph.D. in Industrial/ Organizational psychology from Purdue University and has been a visiting faculty member at the Amos Tuck School, Dartmouth College, Bocconi University in Milan Italy, the University of Washington, Seattle, London Business School and Wuhan University, in China. Taylor is currently a member of the Academy of Management Board of Governors, incoming senior editor for Organization Science, and Human Resource editor for Sage Publications Foundations of Organizational Science Series, and serves on the editorial boards of the Journals of Applied Psychology and Organizational Behavior. She is also a SIOP Fellow. Taylor’s research interests include the employment relationship, organizational justice, executive career mobility, and organizational innovation and dynamic capabilities. Lois Tetrick is the Director of the Industrial and Organizational Psychology Program at George Mason. Professor Tetrick has served as associate editor of the Journal of Applied Psychology and is currently an associate editor of Journal of Occupational Health Psychology. She also serves on the editorial board of Journal of Organizational Behavior. Dr. Tetrick’s research has focused primarily on individuals’ perceptions of the employment relationship and their reactions to these perceptions including issues of occupational health and safety, occupational stress, and organizational/union commitment. She is active in the Society for Industrial and Organizational Psychology (SIOP) and was recently elected to represent SIOP on the American Psychological Association Council of Representatives. She also is active in the Academy of Management and has served as Chair of the Human Resources Division. Dr. Tetrick is a fellow of the Society for Industrial and Organizational Psychology, the American Psychological Association, and the American Psychological Society. Anne S. Tsui is Motorola Professor of International Management at Arizona State University, Professor of the Hong Kong University of Science and Technology

379

and Distinguished Visiting Professor at Peking University. She was the 14th editor of the Academy of Management Journal, a Fellow of the Academy, and Founding President of the International Association for Chinese Management Research (www.iacmr.org). Her recent research interests include guanxi relationship of managers, employment relationships, executive leadership and organizational culture, especially in the Chinese context. She has received the Outstanding Publication in Organizational Behavior Award (1993), the Administrative Science Quarterly Scholarly Contribution Award (1998), the Best Paper in the Academy of Management Journal Award (1998), and the Scholarly Achievement Award in Human Resource Management (1998). She has held faculty appointments previously at Duke University and the University of California, Irvine. She received her Ph.D. from the University of California, Los Angeles. Linn Van Dyne is Associate Professor, Department of Management at the Broad Graduate School of Business, Michigan State University, USA. She received her Ph.D. from the University of Minnesota in Strategic Management and Organizations. Her research focuses on proactive employee behaviors (such as helping, voice, and minority influence), international organizational behavior, and the effects of work context, roles, and groups on employee attitudes and behaviors. Her work has been published in Academy of Management Journal, Academy of Management Review, Journal of Applied Psychology, Journal of Organizational Behavior, Organizational Behavior and Human Decision Processes, Research in Organizational Behavior, and other outlets. Elizabeth Wolfe Morrison (Ph.D. Northwestern University) is a Professor of Management at the Stern School of Business, New York University, and Chair of the Management and Organizations Department. She has won several research awards, including the Cummings Scholar Award from the OB Division of the Academy of Management. Professor Morrison’s research focuses on proactive behaviors by employees (information seeking, networking), how employees adjust to new jobs, the experience of psychological contract violation, and determinants and effects of employee voice and silence. She is interested with how people make sense of, cope with, and impact their work environments. Professor Morrison has published articles in a range of journals, including Academy of Management Journal, Academy of Management Review, Journal of Applied Psychology, Journal of Organizational Behavior, and Organizational Behavior and Human Decision Processes. She is on the editorial board of the Journal of Organizational Behavior and the Journal of Management.