Earnings Quality

Earnings Quality

Earnings Quality Jennifer Francis Duke University Durham, NC USA [email protected]

Per Olsson Duke University Durham, NC USA [email protected]

Katherine Schipper Duke University Durham, NC USA [email protected]

Boston – Delft

R Foundations and Trends in Accounting

Published, sold and distributed by: now Publishers Inc. PO Box 1024 Hanover, MA 02339 USA Tel. +1-781-985-4510 www.nowpublishers.com [email protected] Outside North America: now Publishers Inc. PO Box 179 2600 AD Delft The Netherlands Tel. +31-6-51115274 The preferred citation for this publication is J. Francis, P. Olsson and K. Schipper, R Earnings Quality, Foundation and Trends in Accounting, vol 1, no 4, pp 259–340, 2006 ISBN: 978-1-60198-114-1 c 2008 J. Francis, P. Olsson and K. Schipper

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, mechanical, photocopying, recording or otherwise, without prior written permission of the publishers. Photocopying. In the USA: This journal is registered at the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923. Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients, is granted by now Publishers Inc for users registered with the Copyright Clearance Center (CCC). The ‘services’ for users can be found on the internet at: www.copyright.com For those organizations that have been granted a photocopy license, a separate system of payment has been arranged. Authorization does not extend to other kinds of copying, such as that for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. In the rest of the world: Permission to photocopy must be obtained from the copyright owner. Please apply to now Publishers Inc., PO Box 1024, Hanover, MA 02339, USA; Tel. +1-781-871-0245; www.nowpublishers.com; [email protected] now Publishers Inc. has an exclusive license to publish this material worldwide. Permission to use this content must be obtained from the copyright license holder. Please apply to now Publishers, PO Box 179, 2600 AD Delft, The Netherlands, www.nowpublishers.com; e-mail: [email protected]

R Foundations and Trends in Accounting Volume 1 Issue 4, 2006 Editorial Board

Editor-in-Chief: Stefan J. Reichelstein Graduate School of Business Stanford University Stanford, CA 94305 USA reichelstein [email protected] Editors Ronald Dye, Northwestern University David Larcker, Stanford University Stephen Penman, Columbia University Stefan Reichelstein, Stanford University (Managing Editor)

Editorial Scope R Foundations and Trends in Accounting will publish survey and tutorial articles in the following topics:

• Auditing

• Financial Reporting

• Corporate Governance • Cost Management

• Financial Statement Analysis and Equity Valuation

• Disclosure

• Management Control

• Event Studies/Market Efficiency Studies

• Performance Measurement • Taxation

• Executive Compensation

Information for Librarians R Foundations and Trends in Accounting, 2006, Volume 1, 4 issues. ISSN paper version 1554-0642. ISSN online version 1554-0650. Also available as a combined paper and online subscription.

R Foundations and Trends in Accounting Vol. 1, No. 4 (2006) 259–340 c 2008 J. Francis, P. Olsson and K. Schipper

DOI: 10.1561/1400000004

Earnings Quality Jennifer Francis1 , Per Olsson2 and Katherine Schipper3 1

2

3

Fuqua School of Business, Duke University, Durham, NC 27708, USA, [email protected] Fuqua School of Business, Duke University, Durham, NC 27708, USA, [email protected] Fuqua School of Business, Duke University, Durham, NC 27708, USA, [email protected]

Abstract This review lays out a research perspective on earnings quality. We provide an overview of alternative definitions and measures of earnings quality and a discussion of research design choices encountered in earnings quality research. Throughout, we focus on a capital markets setting, as opposed, for example, to a contracting or stewardship setting. Our reason for this choice stems from the view that the capital market uses of accounting information are fundamental, in the sense of providing a basis for other uses, such as stewardship. Because resource allocations are ex ante decisions while contracting/stewardship assessments are ex post evaluations of outcomes, evidence on whether, how and to what degree earnings quality influences capital market resource allocation decisions is fundamental to understanding why and how accounting matters to investors and others, including those charged with stewardship responsibilities. Demonstrating a link between earnings quality and, for example, the costs of equity and debt capital implies a basic

economic role in capital allocation decisions for accounting information; this role has only recently been documented in the accounting literature. We focus on how the precision of financial information in capturing one or more underlying valuation-relevant constructs affects the assessment and use of that information by capital market participants. We emphasize that the choice of constructs to be measured is typically contextual. Our main focus is on the precision of earnings, which we view as a summary indicator of the overall quality of financial reporting. Our intent in discussing research that evaluates the capital market effects of earnings quality is both to stimulate further research in this area and to encourage research on related topics, including, for example, the role of earnings quality in contracting and stewardship.

Contents

1 Introduction

1

2 Component of Information Quality in Capital Markets

7

2.1 2.2 2.3

Quality as an Attribute of Information in the Capital Markets Financial Reporting Quality Summary Indicator of Financial Reporting Quality

8 10 13

3 Determinants of Earnings Quality

17

3.1 3.2

19

Separating Innate and Discretionary Sources Reporting (Discretionary) Determinants of Earnings Quality

24

4 Measures of Earnings Quality

37

5 Earnings Quality and Market Outcomes

53

5.1 5.2 5.3 5.4

53 56 65 68

Market Outcomes Earnings Quality and Expected Returns Earnings Quality and Abnormal Return Earnings Quality and Other Market Outcomes ix

6 Concluding Remarks

71

Acknowledgments

73

References

75

1 Introduction

This discussion review aims to provide insights on earnings quality. We describe and analyze several determinants and measures of earnings quality, and we discuss both research which uses those determinants and measures and research which provides evidence on the capital markets consequences of earnings quality. In discussing earnings quality research, we focus both on the question addressed and the researcher’s design choices. In our view, an examination of earnings quality and its determinants or consequences must address crucial research design issues, including the definition and measurement of both earnings quality itself and the construction of tests for its capital market effects (such as the cost of capital). Our perspective, which is described in more detail in Section 2, can be summarized as follows. First, we associate earnings quality with precision, in the sense that higher quality earnings are more precise with respect to an underlying valuation-relevant construct that earnings is intended to describe. Differences in the construct chosen as the benchmark for earnings quality imply differences in the research question posed, and therefore also imply differences in how earnings quality is measured. Researchers have proposed distinct constructs against 1

2

Introduction

which earnings quality is measured,1 as well as different approaches to measurement. Second, we take a capital allocation view of earnings quality, as opposed to a contracting or stewardship view, and therefore we are concerned with the capital market consequences of earnings quality. Third, we view earnings quality as comprising both an innate, relatively stable component that is driven by factors intrinsic to business models and operating environments and a relatively more discretionary and fluctuating component that is driven or influenced by management’s financial reporting decisions. We do not view earnings quality as only, or even primarily, the outcome of a period-by-period management choice; rather, we view the precision of earnings as arising from both reporting decisions and long-run strategic decisions about, for example, what should be the firm’s lines of business. Earnings quality is of considerable interest to participants in the financial reporting process, including standard setters, preparers, auditors, regulators, analysts, and financial press commentators. It is also of interest to accounting educators and accounting researchers. As evidence of this interest, we point to several analyses of earnings quality, including Dechow and Schrand (2004), Schipper and Vincent (2003), and the 2002 special issue of The Accounting Review devoted to research on earnings quality. These analyses vary in the benchmark construct that earnings is presumed to capture or describe, and in the evidence on earnings quality that is presented or discussed. Our review is intended to complement and extend the insights available from these analyses. Dechow and Schrand (2004) analyze earnings quality from a financial analysis perspective. They take the view that earnings are of high quality if they “accurately annuitize the intrinsic value of the firm” (p. 5). They identify this annuitizing-of-value attribute with reporting a normalized, sustainable or representative earnings number that corresponds to permanent earnings and they describe such an earnings 1 For

example, research has used, as measures of earnings quality, the value relevance of earnings (as reflected by the mapping of earnings into stock returns), timeliness, conservatism, accruals quality, persistence, and predictability, among other measures. As discussed in Section 4, fundamental differences among these measures likely make them differentially effective at capturing the multi-dimensional construct of earnings quality.

3 number as being of high quality because it has three attributes: it accurately reflects current performance; it indicates future performance; and it is a useful summary for assessing firm value. Our perspective complements that taken by Dechow and Schrand in that we associate earnings quality with precise (that is, low variance) information about a construct that earnings is intended to describe; in the context of Dechow and Schrand’s discussion, this construct would be permanent earnings. However, as discussed in Section 4, researchers have also posited other constructs. Our analysis of alternative measures of earnings quality distinguishes between accounting-based constructs (such as cash flows) and market-based constructs (such as the information summarized in returns) that earnings might be intended to capture. In contrast, Schipper and Vincent (2003) consider earnings quality both from a decision usefulness perspective, following the Financial Accounting Standards Board’s (FASB) conceptual framework, and from a Hicksian income perspective, following the idea that accounting earnings should faithfully represent changes in wealth. Applying our perspective to their paper, the construct that is captured with precision by high quality earnings would be wealth changes (which would probably not correspond to the normalized or sustainable earnings number proposed by Dechow and Schrand). While Schipper and Vincent discuss some of the same earnings quality measures that we analyze in Section 4, they do not consider the capital market consequences of earnings quality, one of the main purposes of our discussion review. Finally, the American Accounting Association-sponsored conference on Quality of Earnings resulted in a special issue of The Accounting Review (2002) that contains six papers and related discussions that consider earnings quality from a balance sheet perspective, from a measurement perspective, from an auditor independence perspective, from an international perspective, from an analyst expertise perspective, and from an earnings management perspective. Our discussion review complements these papers by relating some of them to the earnings quality measures, determinants, and consequences that we consider. We note that the reviews in this special issue were not intended to encompass all, or even a particular subset of, the possible approaches to research pertaining to earnings quality.

4

Introduction

Those with interests in earnings quality also tend to be interested in a related concept, earnings management. We do not attempt to describe the voluminous literature on earnings management.2 This research tends not to consider the capital market consequences of that activity, particularly for broad samples of firms, while one of our main objectives is to discuss how research can provide evidence on the extent to which differences in earnings quality are reflected in capital market outcomes, such as the cost of capital. Our precision-based perspective on earnings quality is related to earnings management because management’s financial reporting decisions are one of several determinants of earnings quality. In this regard, we note that the effect of earnings management on the precision of earnings as a descriptor of an underlying construct is likely to be highly context specific. For example, incentives to increase idiosyncratic returns volatility would be expected to encourage earnings management that decreases precision.3 The rest of this review proceeds as follows. In Section 2, we describe our perspective on earnings quality and place earnings quality in the context of overall capital market information quality. In Section 3, we discuss the determinants of earnings quality, broadly separated into intrinsic determinants that derive from business models and operating environments and reporting determinants that derive from the management’s implementation decisions in financial reporting process per se. In Section 4, we describe and compare 12 measures of earnings quality, provide examples of research which uses these measures, and, for some measures, offer some views about which measures are preferable in a given context. Section 5 discusses research design issues, and research findings, pertaining to the capital market consequences of earnings quality, in particular, its association with expected returns (or the cost of capital) and unexpected (abnormal) returns. Section 6 concludes the review. 2 Surveys

of earnings management research include, for example, Healy and Wahlen (1999) and Dechow and Skinner (2000). In addition, Dechow and Schrand (2004) discuss aspects of earnings management, and related research, as these pertain to their perspective on earnings quality. 3 These incentives could arise if call options comprise a significant portion of management’s compensation because, other things equal, the value of a call option is increasing in total volatility of the underlying shares; for additional discussion see Francis et al. (2007b).

5 Throughout the review, we use examples of research to illustrate earnings quality concepts, theories, measures, and results. These examples are not intended to provide a survey of published and unpublished research on earnings quality. Because research on earnings quality continues to evolve, we do not attempt to make our examples exhaustive, nor do we view our discussions or summaries of existing research work as definitive.

2 Earnings Quality as a Component of Information Quality in the Capital Markets

The objective of this section is to provide a perspective on the detailed discussions of several earnings quality measures and determinants that follow in Sections 3 and 4. Because earnings quality, as a construct, implies nothing about its measurement (that is, earnings quality measurements are context-specific), this section does not describe specific earnings quality measures. We begin our perspective broadly, with the construct of information quality, making particular reference to the types of information used by capital market participants, and describing what is implied by the idea of “quality” as an information attribute. We then narrow the discussion, first to focus on financial reporting quality as a special case of information quality and then to consider earnings quality specifically as a summary indicator of financial reporting quality. Although the discussion in this section ends, and the rest of the review continues, with a focus on earnings quality, we acknowledge the likely existence of statistical associations and even causal links between earnings quality and other components of overall capital market information quality. Our intent in providing this early, broader discussion is to provoke interest in investigating unanswered questions about these associations. 7

8

Component of Information Quality in Capital Markets

2.1

Quality as an Attribute of Information in the Capital Markets

The quality of information in the capital markets is of interest to those who participate in any step in the process whereby information is created, disseminated, and used for purposes of capital allocation. We broadly construe capital allocation to include any ex ante resource allocation (investment decision making). Information quality is also of interest for ex post evaluations of outcomes (contracting purposes, including compensation linked to outcomes). We view the ex ante use of information as providing a foundation for the ex post use and, because the purpose of capital markets is to allocate capital, we focus on quality as it enhances the usefulness of information for capital allocation decisions. We identify “quality” of information in the capital markets with a statistical notion, specifically, the precision of a measure with respect to a valuation relevant construct. For a given construct, higher quality information is more precise (contains less uncertainty) with respect to that construct. Quality defined as precision is broadly applicable, in the sense that precision can be an attribute of a wide variety of information items that have the potential to affect a capital market decision. For example, precise information about inflationary expectations in various jurisdictions would be expected to affect capital allocations across those jurisdictions, as would detailed (precise) information about political risk (e.g., the risk of expropriation, the risk of violence), weather, commodity prices, or the outcomes of elections. As another example, with regard to using information about outcomes for evaluating managers, precise information that can be used to separate state-of-theworld effects on outcomes (e.g., weather) from managerial skill/effort effects would in principle be pertinent. Having chosen to focus on precision, we acknowledge that other attributes of information increase its value to capital market participants — that is, other attributes also contribute to the quality of information. For example, information is valuable to capital market participants to the extent it is timely (information that arrives too late to be used in reaching a decision not valuable to decision makers). As another example, information is valuable (useful) to the extent it is

2.1 Quality as an Attribute of Information in the Capital Markets

9

unbiased (the mean of the information variable is the same as the mean of the construct described by the information). The Financial Accounting Standards Board has created a conceptual framework that provides a set of qualitative characteristics of financial reporting that contribute to decision usefulness — presumably these would all be viewed as quality attributes. The elements of the conceptual framework include timeliness (as a component of relevance), neutrality (as a component of reliability) as well as comparability.1 We note that if quality encompasses any attribute that makes information more valuable to capital market participants, then a full characterization of information quality is surely empirically intractable. Our focus on precision as the construct that underlies information quality is based on the (previously discussed) applicability of precision to many types of information and on its wide use in the accounting literature. As an alternative to assessing information quality by reference to its precision, which is an intrinsic attribute of the information, some research has tended to focus on summary indicators that might be interpreted as the outcomes of differing quality. Two prominent examples of such summary indicators are bid-ask spreads and probability-ofinformed trading (PIN) scores. In particular, Huang and Stoll (1997) report that about 10% of the bid-ask spread of large firms is linked to information, in the form of adverse selection/information asymmetry. Easley et al. (2002) present PIN scores as indicators of the probability that a trade originates from an investor with private information. A third and more recent example of a summary indicator which might be interpreted in the context of overall capital market information quality is returns synchronicity, which has been presented as a summary indicator of the amount of firm-specific information that is reflected in stock returns (e.g., by Morck et al. (2000) and Morck and Yeung (2002), among others). We return to this idea of outcome indicators 1 Accounting

researchers have not been able to devise a single measure that captures the qualitative characteristics of the FASB’s conceptual framework. Researchers have devised empirical measures that combine relevance and reliability (as discussed, for example, in Barth et al. (2001)) but this measure does not encompass comparability. In addition, researchers (e.g., Holthausen and Watts (2001)) differ with regard to the significance of combined relevance-reliability analyses for one key component of the overall process that generates capital market information, namely, standard setting.

10

Component of Information Quality in Capital Markets

of differing quality in our discussion of capital market consequences of earnings quality, in Section 5.

2.2

Financial Reporting Quality

In this section, we narrow our discussion to focus on financial reporting quality as a special case of information quality, and consider the process by which financial reporting information is created, disseminated, and used for purposes of capital allocation. We describe certain non-earnings-based indicators of financial reporting quality used by researchers and pose some questions about financial reporting quality and its determinants. Our aims are to lay the groundwork for later discussions which view earnings quality as a key component of financial reporting quality, and to offer some observations on research that considers financial reporting quality generally, including research on participants in the financial reporting process. Those participants include accounting standard setters, preparers of financial reports, governing boards that provide oversight to financial statement preparation (e.g., audit committees), auditors, enforcement agencies, other regulators, credit rating agencies, investors in both debt and equity instruments and their advisors (e.g., analysts) and commentators in the financial press. The process includes the implementation of reporting standards to produce financial reports, assurance (auditing), and dissemination of the reports. Our discussion considers the financial reporting process and the information that results more or less in isolation from other information-generating processes in the capital markets and their resulting information. That is, we take those other processes and information components as given, and do not consider the existence and nature of interactions between financial reporting and other types of information used in capital markets. Accounting researchers have, for the most part, taken the same approach, although there are exceptions.2 2 For

example Francis et al. (2002) investigate whether analyst reports compete with (substitute for) earnings announcements or complement them. Their results suggest that the relation is generally complementary — that is, market responses to earnings announcements are positively associated with market responses to analyst reports. Furthermore, they report an over-time aspect of this relation in the form of a positive association between earnings announcements and both preceding and following analyst reports. As another

2.2 Financial Reporting Quality

11

However, interactions among components of capital market information undoubtedly exist and are potentially important for understanding financial reporting quality. In particular, statements that financial reporting information accounts for a decreasing portion of the total information used in the capital markets are really statements about (usually unspecified) substitutive relations that are posited to exist and to have been increasing over time. Those statements suggest any number of open questions. For example, whether non-financial-reporting capital market information in general substitutes for or complements financial reporting information; if substitute or complement relations exist, which components are substitutes and which are complements, and why; whether existing substitute/complement relations shift over time and why; whether higher quality in some components of the information used in the capital markets compensates for lower quality in other components; the extent to which financial reporting quality is largely determined by economic conditions that are predetermined at a given reporting date (e.g., volatility of factor input prices and output markets) as opposed to being largely determined by financial reporting decisions made at that reporting date; and whether capital market information quality as a whole is affected by the same forces that affect financial reporting quality, so that financial reporting quality (both level and change) is a good indicator of overall capital market information quality. Turning now to financial reporting as a process, with differing participants at every step, we note that a capital market participant who receives financial reporting information is typically using that information, along with other information, to make a judgment (e.g., about default risk, or about the divergence between an outcome and a prediction) and/or a decision (e.g., about where and how much to invest). Therefore, we take as given that the primary purpose of

example, Francis et al. (2008b) examine the relation between earnings quality and voluntary disclosure (of financial information), where the latter is proxied by a self-constructed disclosure index, similar to Botosan (1997), applied to voluntary financial disclosures in annual reports. They find that firms with good (poor) earnings quality have more (less) extensive voluntary disclosures, suggesting a complementary association between earnings quality and voluntary disclosure.

12

Component of Information Quality in Capital Markets

financial reporting information in the capital markets is to support certain judgments and decisions. Financial reporting quality is of interest, then, primarily because of the view that high quality information leads to higher quality judgments and decisions; that is, high quality financial reporting information is more decision useful than low quality information. We follow the accounting literature in distinguishing between the judgments and decisions that pertain to capital allocations (debt and equity investments) and those that derive from contracts (e.g., a decision to foreclose on a loan). We believe this difference in perspective has implications for empirical measures of financial reporting quality; our discussions focus on the capital allocation use, as opposed to the contracting use. We also suggest that a secondary purpose of financial reporting quality is to provide a signal to participants in the information creation/ dissemination portions of the process as to how well they are doing their assigned tasks. This use of information quality might be highly context specific and indirect. For example, an auditor might use the quality of the financial reports as a whole as an indicator of the quality of the audit and a standard setter such as the FASB might view the quality of the financial reports as an indirect indicator of the quality of financial reporting standards. These uses require an assumption, namely, that the activity being evaluated — for example, auditing or standard setting — has a first-order effect on the quality of the financial reports. Put another way, the importance to overall financial reporting quality of each step in the financial reporting process is an empirical question on which evidence is relatively sparse and indirect; we view this as an issue that is worthy of further study. Although they are not typically described this way, design choices can be useful in analyzing researchers’ assumptions about what activities most affect overall financial reporting outcomes, including reporting quality. For example, using an indicator variable for large versus small audit firms in research on financial reporting outcomes requires an assumption that auditing has a first-order effect on those outcomes (otherwise, the audit firm size indicator would not meaningfully separate the financial reporting outcomes). Similarly, research which posits that certain non-US firms voluntarily adopt US GAAP in order

2.3 Summary Indicator of Financial Reporting Quality

13

to commit to high quality financial reporting requires an assumption that financial reporting standards have a first-order effect on reporting quality.3 On the other hand, certain summary indicators of reporting quality are not useful for evaluating the effectiveness of a given step in the financial reporting process because they cannot be clearly associated with a step in that process. For example, analyst ratings of reporting quality (as analyzed by, for example, Lang and Lundholm (1996)) capture (presumably expert) external assessments of financial reporting quality but they say nothing at all about how that quality arises and what components of the financial reporting process are most important in determining the outcome. Still other indicators seem ambiguous, in terms of what they capture. For example, analyst coverage might be viewed like an outcome indicator similar to analyst rating or analyst disclosure score (more analysts follow firms with high reporting quality) or as a causal factor similar to auditing (analyst coverage means closer scrutiny that in turn induces greater efforts to increase reporting quality by those who create the information being scrutinized). Overall, we believe that indicators of financial reporting quality used in accounting research differ qualitatively in what they assume and what they are intended to capture. Some indicators are valid only if the activity they capture in fact plays a key role in determining reporting quality while others represent overall assessments of the quality of reporting outcomes, regardless of the factors that cause those outcomes.

2.3

Earnings Quality as the Summary Indicator of Financial Reporting Quality

As previously indicated, we focus on earnings quality as a summary indicator of financial reporting quality. This approach follows the stream of accounting research that views earnings as either a summary indicator or a premier component of the financial reporting package. 3 However,

some research on earnings management could be interpreted as indicating a view that incentive-driven implementation decisions are the most important determinant of financial reporting quality.

14

Component of Information Quality in Capital Markets

This view may reflect a pragmatic compromise with difficulties in identifying and measuring overall indicators of financial reporting quality; that is, earnings quality may be used to summarize financial reporting quality simply because it is easier to observe the former than the latter. Alternatively, some researchers may view earnings quality as a sufficient statistic for overall financial reporting quality; or, even if earnings is not sufficient, statements by some researchers (e.g., Lev (1989)) suggest a belief that earnings is the most important outcome indicator of the financial reporting process. In either case, a focus on earnings is natural and appropriate. A focus on earnings quality can also be justified by the ubiquity of earnings and earnings-based metrics. These measures appear in contracts for compensation and borrowing, so payouts to employees and creditors are based on earnings (defective earnings can cause unintended payouts to employees and can hide deteriorating credit quality from lenders). Earnings also appears in many valuation models and investment decision models. Defective earnings can therefore result in capital misallocations (in the worst cases of fraud, earnings cause investors to provide resources that are immediately dissipated because there are no actual projects in place or contemplated to create returns). However, the idea that earnings quality is a key component or summary indicator of financial reporting quality does not imply that research on financial reporting quality should focus exclusively on earnings. The earnings number aggregates numerous line items whose properties, including their importance for capital market decisions, would be expected to vary. Focusing on earnings in total masks these differences. In addition, balance sheets contain information that by definition is not captured in earnings, and one would expect that earnings and balance sheets would be complements in terms of their decision properties and other properties. Taking this perspective, several of the questions posed about the relation between overall information quality and its reporting quality component apply to the earnings quality component of financial reporting quality as well. For example, is earnings in fact the best overall indicator of financial reporting quality, or is there a component of earnings that dominates; would a higher quality non-earnings component of financial reporting quality compensate for lower quality

2.3 Summary Indicator of Financial Reporting Quality

15

earnings and if so what is that component; do earnings components differ in quality and if so why, and with what consequences? To summarize the discussion of Section 2, we focus on precision as the attribute that measures information quality, and, more specifically, earnings quality. That is, we consider earnings to be of high quality if they are precise with respect to an underlying construct that pertains to capital market decisions. We recognize that earnings is but one component of an overall financial reporting package — albeit a very important one — and we have offered some comments on research that pertains to the financial reporting process, broadly construed.

3 Determinants of Earnings Quality

In this section, we distinguish between two sources (or determinants) of earnings quality. We discuss each source and describe related research, and we offer some comments on research questions pertaining to the determinants of earnings quality. Figure 3.1 schematically represents our perspective on the sources and capital market outcomes of earnings quality. We view earnings quality as influenced by two types of factors: those that reflect innate features of business models and operating environments, and those that reflect the financial reporting process per se. Both sources of earnings quality may have direct (that is, not operating through earnings quality) effects on market outcomes, as indicated by the dashed lines in Figure 3.1 connecting market outcomes directly to both innate and reporting determinants of earnings quality. Research on earnings quality has been primarily concerned with the link between innate and reporting determinants of earnings quality, and we focus our discussions there as well.1 1 It

is an empirical question whether direct effects of innate sources or reporting sources on market outcomes would be captured by asset pricing model risk factors that are unrelated to earnings quality, such as the market factor, size factor or book to market factor in the 3-factor model.

17

18

Determinants of Earnings Quality

Fig. 3.1 Schematic overview of the sources of earnings quality, their relation to the innate and discretionary components of earnings quality, and the overall links to market outcomes.

Innate sources of earnings quality arise from business models and operating environments; reporting sources arise from the financial reporting process, including (1) management’s financial reporting implementation decisions, including judgments and estimates; (2) the quality of the information systems used to support financial reporting; (3) monitoring activities, including internal and external audits, (4) governance activities, including actions taken by boards of directors, compensation arrangements, and ownership structures; (5) regulatory scrutiny (including the extent and nature of securities laws and the enforcement of those laws); (6) reporting standards, such as generally accepted accounting principles (GAAP). While some aspects of the financial reporting process (e.g., reporting standards) are slow to change, management’s implementation decisions reflect period-byperiod decisions.

3.1 Separating Innate and Discretionary Sources

19

The dashed line in Figure 3.1 connecting innate and reporting sources of earnings quality indicates an association between these determinants, perhaps in some cases even a causal link. For example, as we discuss in Section 3.2, governance structures such as board composition are likely to be influenced by innate factors such as the firm’s business model. As another example, most reporting choices are restricted by securities laws, state corporation laws and authoritative accounting guidance, and many of these restrictions may exist precisely because of innate factors. At the same time, accounting guidance requires managers to make numerous implementation decisions which affect the reporting component of earnings quality. We begin in Section 3.1 by discussing how accounting research has identified innate determinants of earnings quality and used those determinants to estimate an innate component and a discretionary (that is, a financial reporting) component of earnings quality.2 In Section 3.2, we discuss the reporting sources of earnings quality in more detail. By organizing the discussion this way, we do not mean to imply that the total amount of earnings quality is unimportant or that it is always necessary to identify the determinants of earnings quality. That is, some research considers the total amount of earnings quality, without being concerned about its components or determinants. For example, Francis et al. (2005a) analyze the cost of capital effects of total earnings quality and Francis et al. (2008b) analyze the link between total earnings quality and voluntary disclosure.

3.1

Separating Innate and Discretionary (Reporting) Sources of Earnings Quality

This discussion follows Francis et al. (2004, 2005a) who distinguish between innate (that is, operating and environmental) and discretionary (that is, reporting) determinants of earnings quality. Innate determinants derive from business models and operating environments, whereas discretionary determinants are associated with accounting choices, implementation decisions, managerial error, auditing, gover2 Our

use of the word “discretionary” follows the existing literature and is not meant to imply that this component of earnings quality is the outcome of free choice.

20

Determinants of Earnings Quality

nance, and enforcement. We view innate factors as being slow to change relative to factors that influence discretionary earnings quality; that is, we view them as predetermined at any given reporting date, but susceptible to modification over time. We do not identify the time period over which the business model and operating environment can change, because we recognize that some changes can be abrupt (e.g., securitizing receivables, exiting a line of business) while others can be gradual (e.g., building a brand, entering a line of business by means other than acquisition of an existing business). Empirical research designs that separate innate from discretionary earnings quality take advantage of the presumed relatively slow-tochange nature of innate determinants, taking one of two approaches. Under the first approach, the researcher regresses a measure of total earnings quality (TotalEQ equal to the combined innate and discretionary earnings quality) on innate factors that the researcher believes describe the firm’s business model and its operating environment. For example, Francis et al. (2005a) implement this approach using Dechow and Dichev’s (2002) accruals quality measure of earnings quality and innate factors: firm size, variability of operating cash flows, variability of sales, operating cycle, and incidence of loss.3 T otalEQj,t = λ0 + λ1 Sizej,t + λ2 σ(CF O)j,t + λ3 σ(Sales)j,t + λ4 OperCyclej,t + λ5 N egEarnj,t + µj,t , where

(3.1)

T otalEQj,t = the researcher’s measure of firm j’s total earnings quality (Francis et al. (2005a) use Dechow and Dichev’s (2002) measure of accruals quality) where larger values indicate poorer earnings quality, Sizej,t = log of firm j’s total assets in year t,

3 Expression

(3.1) is used to illustrate approach 1 for separating earnings quality into components. The researcher determines both the measure of total earnings quality and the innate factors.

3.1 Separating Innate and Discretionary Sources

21

σ(CFO j,t ) = the standard deviation of firm j’s cash flow from operations,calculated over some interval (Francis et al. (2005a) use the past 10 years), σ(Sales)j,t = the standard deviation of firm j’s sales, calculated over some interval (e.g., the past 10 years), OperCyclej,t = the log of firm j’s operating cycle, N egEarnj,t = the number of years over some interval (e.g., past 10 years), where firm j reported negative net income before extraordinary items. The prediction error from (3.1) is the estimate of the discretionary component of firm j’s earnings quality, DiscEQj,t = µ ˆj,t . The predicted value from (3.1) is an estimate of the innate portion of firm j’s earnings quality: ˆ0 + λ ˆ 1 Sizej,t + λ ˆ 2 σ(CF O)j,t + λ ˆ 3 σ(Sales)j,t InnateEQj,t = λ ˆ 4 OperCyclej,t + λ ˆ 5 N egEarnj,t . +λ

(3.2)

This approach yields distinct estimates for each of the two components of earnings quality; in Figure 3.1, these estimates correspond to the ellipses under the total earnings quality measure. The precision of the estimates of these two components depends on the researcher’s ability to identify and measure the factors that describe the firm’s business model and its operating environment, and thereby determine the innate portion of total earnings quality. The second approach controls for innate factors affecting earnings quality by including the factors as independent variables in a regression whose dependent variable is total earnings quality. The coefficient on the test variable in these augmented regressions captures the portion of earnings quality that is both associated with the test variable and incremental to the effects captured by the innate factors — that is the discretionary portion. For example, suppose a researcher is interested in the question of whether the choice of depreciation methods is associated with poorer (or better) earnings quality. Assume further that the researcher tests this relation by estimating Equation (3.3): T otalEQj = γ0 + γ1 DepnM ethodj + λk Innate(k )j + εj ,

(3.3)

22

Determinants of Earnings Quality

where DepnM ethodj = 1 if firm j uses straight line depreciation and 0 otherwise, Innate(k )j = vector of innate factors for firm j. A result that γ1 > 0 (γ1 < 0) would indicate that straight line depreciation is associated with worse (better) earnings quality, after controlling for the innate portion of earnings quality. Assuming that all innate factors have been included in Innate(k )j , a significant value for γ1 provides evidence about the influence of the choice of depreciation method on the discretionary portion of earnings quality. Research that adopts the second approach includes Francis et al. (2008a) who analyze the influence of CEO reputation on earnings quality, and Francis et al. (2007b) who examine the relation between earnings quality and the use of call options for financing purposes and for compensation. A variant of the second approach would replace TotalEQ with a dependent variable that captures an outcome of interest to the researcher, for example, the cost of equity. In that case, the independent variables would include TotalEQ and the innate factors, and the estimated coefficient on TotalEQ captures the effect of discretionary earnings quality on the dependent variable. The first and second approaches to separating total earnings quality into innate and discretionary components are not substitutes. Only the first approach permits direct comparisons of the effects of innate versus discretionary earnings quality, because it provides distinct estimates of these components. The second approach does not provide a distinct estimate of the discretionary component of earnings quality; rather, it implies the discretionary effect by controlling for the innate effect. The two approaches also differ in their sensitivity to omitted innate factors. Under the first approach, omitted innate factors lead to model misspecification in the regression of total earnings quality on innate factors, and more noise in the error term. Other things equal, this noise would bias the coefficient estimate relating the dependent variable of interest to DiscEQ toward zero. Under the second approach, if a dependent variable of interest is regressed on TotalEQ and an incomplete set of innate factors, the result would be an upwardly biased coefficient estimate on TotalEQ (that is, an upwardly biased estimate of the effect

3.1 Separating Innate and Discretionary Sources

23

of discretionary earnings quality) to the extent that the omitted innate factors are positively correlated with InnateEQ. These considerations suggest that researchers might prefer to use both approaches, where practicable, because the first approach would be expected to yield a lower bound inference about the effects of discretionary earnings quality and the second approach would be expected to yield an upper bound inference. For example, Francis et al. (2005a) employ and contrast both approaches in their investigation of whether innate earnings quality is associated with predictably larger cost of capital effects than discretionary earnings quality. They predict and find that discretionary earnings quality (measured under either of the two approaches) has a smaller cost of capital effect than does the innate portion of earnings quality. Under either approach to separating innate and discretionary effects of earnings quality, the researcher must specify innate factors that describe business models and operating environments. Dechow and Dichev (2002) put forward the following innate factors: firm size (proxied by total assets or sales revenues), cash flow variability (calculated as the standard deviation of the firm’s cash flows from operations, scaled by total assets), sales variability (measured as the standard deviation of the firm’s sales revenues, scaled by total assets), operating cycle (equal to the log of the sum of the firm’s days accounts receivable and days inventory) and the firm’s incidence of negative earnings realizations (measured as the firm’s proportion of losses over some prior period). Building on Dechow and Dichev, Francis et al. (2004) add measures of intangibles intensity (proxied by the sum of reported research and development expenses and advertising expenses as a proportion of sales revenues) and capital intensity (proxied by the ratio of the net book value of property, plant, and equipment to total assets). More recently, Francis et al. (2008a) add industry and year dummy variables. Research examining the ability of innate factor to explain earnings quality typically finds that innate factors account for between 50% and 70% of the variation in the earnings quality metric. As noted previously, the precision of the separation between innate and discretionary earnings quality derives from the researcher’s ability to describe or specify the firm’s business model and its operating

24

Determinants of Earnings Quality

environment using accounting variables and other factors. While the approaches we have described appear to work reasonably well, as demonstrated by the relatively high explained variation of the models, there has been no systematic and rigorous demonstration of the construct validity of the innate and discretionary earnings quality measures, independent of the specific research contexts in which they have been used. We believe future research could address the adequacy of the innate factors identified so far for describing the firm’s business model and its operating environment, to determine whether one or more variables should be added. We also suggest that future research could (re)consider whether earnings quality is a primitive construct with innate (that is, operating) components and discretionary (that is, reporting) components, each having separately identifiable capital market consequences, or whether both earnings quality and its (apparent) capital market consequences are in fact determined by innate factors.

3.2

Reporting (Discretionary) Determinants of Earnings Quality

In the introduction to Section 3, we identify six determinants of earnings quality that involve financial reporting decisions, and therefore affect what we have termed the discretionary portion of earnings quality. We contrast these reporting determinants with innate determinants that involve decisions about the operations of the business model, and therefore affect what we have termed the innate portion of earnings quality.4 The reporting factors we identify include management’s period-by-period reporting decisions, the quality of the firm’s information systems, the extent and quality of auditing, the quality of 4 This

dichotomy is imperfect, in that many determinants of reporting quality have both discretionary and innate characteristics. For example, the selection of Big 8/6/4 audit firm versus a non-Big 8/6/4 audit firm has been used by researchers as a measure of managerial discretionary choice that affects the credibility of financial reporting. However, for a large multinational firm, the choice of a Big 8/6/4 audit firm is arguably not discretionary, but in fact necessitated by innate factors such as the firm’s size and complex operating environment. As another example (discussed in detail later in this section), governance decisions influence, and are influenced by, both the discretionary component and the innate component of earnings quality.

3.2 Reporting (Discretionary) Determinants of Earnings Quality

25

governance systems, regulatory scrutiny and enforcement, and accounting standards themselves. In this subsection, we discuss research that considers these factors. (a) Management decisions. A large stream of accounting research has analyzed management’s financial reporting decisions from an earnings management perspective, examining both incentives for earnings management and its consequences. Earnings management activities are multifaceted (they include implementation decisions, choices among accounting treatments, and errors) and they can have both positive and negative earnings quality consequences. In particular, Guay et al’s (1996) discussion of the exercise of managerial discretion over accruals suggests that the discretionary (that is, the reporting) component of earnings quality contains three subcomponents. The performance subcomponent reflects management’s attempts to improve the ability of earnings to reflect performance in a reliable and timely way; this subcomponent would increase information precision and, therefore, contribute to better earnings quality. In contrast, the opportunism and noise sub-components would reduce information precision, and therefore lead to poorer earnings quality. We note, however, that the magnitudes of the earnings quality effects of these subcomponents would be expected to differ, implying that the net earnings quality effect of management’s implementation decisions could be positive, negative, or neutral. Arguments advanced by Guay et al. suggest that the performance and opportunism subcomponents dominate the noise component (that is, the discretionary component of accruals is not mostly noise). However, their empirical results do not point to either the performance subcomponent or the opportunistic subcomponent as being empirically stronger for the broad sample they consider. They note that managerial discretion over accruals has existed for many years, and it is hard to see how such a pervasive and sustained practice could survive if it systematically harmed earnings quality.5 Moreover, both 5 Under

this view, the net earnings quality effect of management’s reporting decisions is expected to be positive on average, indicating that the performance subcomponent dominates the opportunism subcomponent. Empirical support for this view is provided by Subramanyam (1996) and Xie (2001) who show that in large samples, total accruals and discretionary accruals improve earnings as a signal of performance, as measured by the contemporaneous returns-earnings relation.

26

Determinants of Earnings Quality

Guay et al. and Healy (1996) note that management in a given firm may engage in behaviors differentially affecting the magnitudes of the performance and the opportunistic subcomponents at different points in time. Further, in a broad cross-section of firms, management of one firm can report opportunistically and management of another can report unbiasedly. As a consequence of these intra-firm, across-firm, and over-time influences, the overall effect of management’s financial reporting decisions observed for a given sample will be a weighted average of the separate effects. This netting is precisely why most studies of opportunistic managerial exercise of discretion (i.e., earnings management studies) focus on carefully selected, nonrandom samples where incentives for opportunistic behaviors are strong and, therefore, where the researcher is most likely to find an effect, if one exists. In contrast, much of the research examining the effects of managerial discretion on earnings quality has not focused on nonrandom samples, including those characterized by strong financial reporting incentives. There are, of course, exceptions. In tests designed to validate their e-loading measure as a proxy for earnings quality (this measure is discussed in Section 4), Ecker (2006) identify three nonrandom samples of firms where earnings quality is arguably poor: firms that restated their earnings, firms that went bankrupt, and firms that were sued for allegedly defective financial reporting. For all three samples they find significantly higher e-loadings (indicating poorer earnings quality) than observed for matched control samples of firms. Another exception is Krishnan et al.’s (2008) investigation of the link between earnings quality and the choice of LIFO versus FIFO for inventory valuation. Krishnan et al. demonstrate analytically that, all else equal, use of the LIFO method will result in better earnings quality (relative to the FIFO method) in the specific case when earnings quality is defined as the accruals quality measure put forward by Dechow and Dichev (2002). That is, LIFO firms will have smaller variability of residuals from a regression of current accruals on past, current, and future cash flows. They demonstrate that this analytical result holds empirically by providing evidence that LIFO firms have better accruals quality than FIFO firms.

3.2 Reporting (Discretionary) Determinants of Earnings Quality

27

(b) Information systems. We have not identified empirical research which directly studies the quality of corporate information systems as a determinant of earnings quality.6 Instead, existing research tends to focus on earnings or the accrual component of earnings as a summary output signal, making the (implicit) assumption that information systems quality has minor effects, or effects with little or no cross-sectional or over time variability. This may be because it is not practicable to identify a proxy for the quality of firms’ information systems that is separate from a measure of the quality or precision of the outputs from those systems. However, this practical (empirical) difficulty does not in our view diminish the importance of the issue. Indeed, as financial reporting increasingly requires sophisticated measurements (such as fair value measurements based on estimates as opposed to observed prices) the role of information systems in enhancing (or detracting from) earnings quality should increase. (c) Auditing. In contrast to the sparse or nonexistent literature on how information systems affect earnings quality, a sizable literature investigates the extent to which auditing affects both earnings quality and influences investors’ perceptions of earnings quality. For example, Teoh and Wong (1993) study how the choice of auditor affects investors’ perceptions of the credibility of earnings. Their measure of earnings credibility (which we associate with quality, because more precise earnings should be more believable) is the slope coefficient relating the short term market reaction to the unexpected earnings news in quarterly earnings announcements (the earnings response coefficient, or ERC; this measure of earnings quality is discussed in Section 4). They

6 Verrecchia

(1990) models the relation between the precision of the information signal observed by the firm’s manager (which can be thought of as a proxy for the quality of the information system) and the manager’s voluntary disclosure decision. He finds that the probability of disclosure increases as the precision of a manager’s private information increases. The intuition for this complementary association is that because the precision of the manager’s private information is known, rational market participants will discount a high quality firm that withholds information from the market. This force causes the high quality firm’s disclosure threshold to decrease, resulting in more (less) disclosures for firms with good (poor) information quality. Empirical tests of the association between earnings quality and voluntary disclosure show results that are consistent with this predicted complementary association (Francis et al. (2008b)).

28

Determinants of Earnings Quality

find that the ERC is positively related to auditor quality as proxied by whether the auditor is a Big 8 firm. Other research in this area includes Becker et al. (1998) who examine whether clients of Big 6 audit firms report larger or smaller abnormal accruals than clients of non-Big 6 audit firms. Researchers have also investigated the extent to which characteristics of the auditor (other than the Big 8/6/4 designation) influence earnings quality. The key auditor characteristics studied are expertise and independence. Concerning expertise, empirical research draws on behavioral auditing research demonstrating that experience enhances performance (Bonner and Lewis (1990); Bedard and Biggs (1991); Wright and Wright (1997)) to examine the link between auditor experience (in the same industry as the client) and the quality of the client’s financial reporting (as proxied by the absolute value of abnormal accruals, a measure of earnings quality discussed in Section 4). In testing for a link between auditor industry experience and absolute abnormal accruals, Krishnan (2003) finds that client-firms audited by specialist auditors (i.e., auditors with high concentrations of clients in the same industry as the client-firm) have smaller absolute abnormal accruals than those audited by non-specialist audit firms. He concludes that specialist auditors use their expertise to constrain clients’ reporting behavior; this conclusion suggests that auditor expertise (as proxied by experience) affects earnings quality. Research on the relation between auditor independence and earnings quality has focused on the extent of nonaudit services supplied by the audit firm (pre-Sarbanes-Oxley) and on auditor tenure, defined as the number of years the same audit firm or the same audit partner has performed the audit of a given client firm. Both nonaudit fees and auditor tenure are predicted to be associated with diminished earnings quality, to the extent they impair independence. That is, auditor independence is the (unobservable) construct that is predicted to influence earnings quality, and nonaudit fees and auditor tenure are the (observable) indicia of lack of independence. Concerning the relation between nonaudit fees and earnings quality, research generally shows weak evidence or no evidence that larger

3.2 Reporting (Discretionary) Determinants of Earnings Quality

29

nonaudit fees are associated with worse earnings quality.7 Concerning the alleged deleterious effects of auditor tenure (i.e., the over-time dependency of having the same auditor audit the firm year after year), Myers et al. (2003) and Johnson et al. (2002) show that longer tenure is associated with better, not worse, earnings quality. Overall, the inference that has been drawn from this research is that neither the provision of nonaudit services by audit firms nor long auditor tenure appears to impair the auditor’s ability to perform audit services that result in good earnings quality. Whether there is no effect or a beneficial effect remains an open issue. (d) Governance structures (including ownership and compensation). Research has examined the relation between governance structures and both earnings quality and outcomes related to earnings quality. We believe the distinction between innate and discretionary quality is crucial for this type of research, because whether governance structures drive quality differentials or react to quality differentials affects the sign of the predicted relation. Stated differently, a view that earnings quality is entirely or mostly innate (and therefore intrinsic to the business model and operating environment) generally leads to different predictions about the association between earnings quality and governance variables (such as board composition and stock ownership by board members) than does a view that earnings quality is entirely or mostly discretionary.8 To illustrate, Bushman et al. (2004) assume that earnings timeliness proxies for the inherent limitations of accounting measures to capture value relevant information (i.e., earnings timeliness proxies for innate earnings quality). They investigate whether firms with low earnings timeliness measures respond with countervailing governance structures, such as increased external monitoring. In contrast, other papers take governance structures as given and investigate whether deficiencies in these structures (such as low external monitoring) facilitate earnings management, which they associate with poorer earnings quality (e.g., 7 An

exception is Frankel et al. (2002) who report a positive coefficient relating the magnitude of nonaudit fees to the absolute value of abnormal accruals. 8 However, both perspectives would predict (and research has found) that poor earnings quality is associated with less independent boards.

30

Determinants of Earnings Quality

Klein (2002), Peasnell et al. (2005), Larcker and Richardson (2004), Dechow et al. (1996), and Beasley (1996)). Prior research generally finds opposite relations between corporate governance characteristics and earnings quality depending on whether the researcher views earnings quality as primarily innate — that is, governance structures respond to earnings quality — or primarily discretionary — that is, earnings quality responds to governance structures. When earnings quality is viewed as innate, as in Bushman et al., results are broadly consistent with the view that firms with poor quality structures have boards characterized by more specialized knowledge (but less independence), greater stock ownership, and greater external monitoring. When earnings quality is viewed as discretionary, prior work generally shows that firms with greater board independence, greater ownership of stock and greater external monitoring have better earnings quality.9 We suggest that this research could be extended by relaxing the restriction that earnings quality is either innate or discretionary. Specifically, distinct measures of the innate and discretionary components of earnings quality (developed using the first approach described in Section 3.1 or another approach) could be used to analyze the relations between governance structures and each of the innate and discretionary components of earnings quality, separately. Such an investigation might help to reconcile the sometimes disparate results in the literature. We conjecture that these disparate results may occur because some measures of earnings quality likely capture more of the innate component of earnings quality, while other measures of earnings quality are more likely to capture the discretionary component. For example, Bushman et al. treat their composite earnings timeliness measure as primarily innate, whereas studies that treat earnings quality as discretionary tend to use measures of earnings quality based on abnor9 These

generalizations should be viewed cautiously. It is very difficult to characterize the governance literature because of differences across studies in terms of: (1) samples (this research often uses relatively small samples for relatively short time periods, so results may not be generalizable to broader populations or longer time periods); (2) the choice and construction of governance and test variables; and (3) the inclusion (or not) of certain control variables which can and do affect the relation between governance constructs and the test variable(s).

3.2 Reporting (Discretionary) Determinants of Earnings Quality

31

mal accruals. Each type of study may be correct in their assumptions about whether their measure of earnings quality is primarily innate or discretionary, and if so the results are not conflicting. However, the question remains as to whether a given earnings quality measure contains both innate and discretionary components and if so which one dominates.10 In addition to board composition as a governance structure, several studies point to the effect of ownership structures on the quality of financial reporting. This research begins with the presumption that factors related to ownership structures influence the credibility of accounting reports, which in turn affects the informativeness or quality of earnings. For example, Warfield et al. (1995) investigate the link between managerial share ownership and earnings quality/informativeness; Fan and Wong (2002) analyze earnings informativeness when concentrated managerial ownership is accompanied by the separation of cash flow rights from voting rights (as it is in several East Asian countries); and Francis et al. (2005b) examine earnings informativeness in dual class and single class firms in the United States. All three studies document a link between ownership structure and their measure of earnings quality, and posit that the causality runs from ownership to quality. Specifically, Warfield et al. (1995) find that earnings quality (as proxied by the slope coefficient from a regression of returns on earnings, and the correlation between returns and earnings) is generally increasing in the level of managerial ownership.11 They explain this result in terms of a decreased demand for accounting-based constraints when managers own more stock, because stock ownership aligns management’s interests with those of shareholders. 10 This

question is complicated by the fact that empirical studies using abnormal accrualstype measures find that innate factors explain 50%–70% of the cross-sectional variation in abnormal accruals, that is, innate factors have high explanatory power for accruals measures that are supposedly purged of their innate component. We discuss this issue more in Section 4. 11 The relation between managerial ownership and earnings informativeness is not linear. It has been shown by several studies to be increasing in small to medium levels of ownership and decreasing in high levels of ownership (Warfield et al. (1995); Gul and Wah (2002); Yeo et al. (2002)).

32

Determinants of Earnings Quality

For a large sample of East Asian firms, Fan and Wong (2002) posit that when voting rights and cash flow rights diverge, concentrated ownership is associated with lower quality earnings because of an entrenchment effect (the controlling shareholder has the ability to report self-interestedly) and an information effect (there are incentives to disclose as little proprietary information as possible). However, high levels of management ownership create an incentive effect, which acts to align managers’ interests with those of shareholders, producing higher quality earnings. Fan and Wong’s results suggest that entrenchment and information effects dominate incentive effects. They find that the earnings informativeness of their sample of concentrated ownership firms is negatively related to the divergence of cash flow rights from voting rights (consistent with the entrenchment effect) and negatively related to the percent of total votes held by the largest ultimate owner (consistent with the information effect). Francis et al. (2005b) extend Fan and Wong’s analysis to the United States setting where dual class ownership structures separate cash flow rights from voting rights, similar to the separations created by pyramids and crossholding structures in Asia. In addition, Francis et al. examine whether the lower predicted earnings quality of dual class firms leads investors to place greater reliance on dividend signals relative to earnings signals for these firms. Their results are consistent with all predictions: relative to single class firms, dual class firms have lower earnings quality (as measured by the slope coefficients relating returns to earnings and earnings changes), and investors in these firms place greater weight on dividends relative to earnings than do investors in single class firms. A related stream of research has probed the relation between compensation and earnings quality, in particular, the consequences of specific forms of compensation (such as employee stock options, or ESOs) for managerial decisions that would be expected to affect earnings quality. This research builds on prior findings that ESOs contain both riskbased incentives and share-price-based incentives. Risk-based incentives arise to the extent that ESOs induce desirable increases in volatility because they motivate risk-averse managers to select riskier projects with positive net present values. Both direct and indirect tests of the

3.2 Reporting (Discretionary) Determinants of Earnings Quality

33

risk-based incentive feature of ESOs support this perspective (see, for example, Rajgopal and Shevlin (2002), Tufano (1996), Guay (1999)), and Hanlon et al. (2003). Share-price-based incentives arise to the extent that ESOs provide incentives to increase stock price (or income, a determinant of stock price). For example, Hanlon et al. (2003) report that ESOs are associated with increased operating income. A variant of this research takes the view that ESOs create undesirable incentives to manage earnings in an effort to manipulate share prices upward (e.g., Bartov and Mohanram (2004) and Cohen et al. (2005)). Building on this research, Francis et al. (2007b) posit that managers’ financial reporting decisions provide a mechanism for increasing returns volatility via poorer earnings quality, and for increasing stock price via better earnings quality (through a reduction in the cost of capital). That is, ESOs create incentives both to decrease earnings quality and thereby increase returns volatility and to increase earnings quality and thereby increase stock price.12 Consistent with these predictions, they find that greater ESO risk-based incentives (as proxied by the sensitivity of the firm’s ESOs to returns volatility) are associated with poorer earnings quality (as proxied by the Dechow and Dichev (2002) accruals quality measure), and that greater ESO share-price-based incentives (as proxied by the sensitivity of the firm’s ESOs to stock price) are associated with better earnings quality. Further tests show that the risk-based effects dominate the share price effects, resulting in net poorer earnings quality for firms with greater ESO usage. To summarize, research has documented an association between several measures of earnings quality and governance structures, defined to include board composition, share ownership by managers and board members, ownership structures in which cash flow rights diverge from voting rights, and compensation in the form of employee stock options. In some cases, the researcher posits a causal link from governance to earnings quality; that is, governance arrangements are a determinant of earnings quality. In other cases, the researcher posits the opposite, 12 Francis

et al. (2007b) posit a causal link from ESOs to earnings quality. Their tests also allow for an endogenous relation, such that boards of directors (aware of the link from ESOs to earnings quality), can alter both the extent of use of ESOs and the risk-based and share-price-based incentive features of ESOs in compensation contracts.

34

Determinants of Earnings Quality

namely, a governance response to (innate or intrinsic) earnings quality. As noted previously, we believe that a fruitful avenue for future research is to analyze the relation between governance arrangements and earnings quality components, based on an explicit separation between innate earnings quality and discretionary or reporting earnings quality. We conjecture that the innate component should be determined largely by the business model and operating environment and therefore slow to change, suggesting that governance arrangements would respond to this component. In contrast, discretionary or reporting earnings quality can shift period by period and would be expected to be determined by governance, ownership and compensation arrangements and other elements of the financial reporting process. (e) Regulatory scrutiny and financial reporting standards. By regulatory scrutiny, we mean the elements of the securities laws that pertain to financial reporting plus the related enforcement mechanisms. By financial reporting standards, we mean the applicable authoritative guidance (e.g., United States generally accepted accounting principles or GAAP). Because these two determinants of earnings quality share two important characteristics, we discuss them in the same section. First, they vary by jurisdiction, not by firm; in comparison, many other determinants of earnings quality are firm-specific (e.g., business model, management’s reporting decisions, governance structures). Second, they vary within jurisdiction over time, but slowly and sporadically. For example, US GAAP changes only after the Financial Accounting Standards Board (FASB) completes extensive due process procedures which often take years and the US securities laws change only with Congressional action. Research suggests that both regulatory scrutiny and financial reporting standards can have discernible economic effects, at least in extreme conditions. For example, entities that restate their earnings to correct a GAAP violation, often but not always identified by enforcement bodies, experience statistically and economically meaningful share price decreases (e.g., Palmrose et al. (2004)). However, research has not systematically demonstrated an overall effect on earnings quality for either financial reporting standards or regulatory scrutiny for more general (non-pathological) samples. In addi-

3.2 Reporting (Discretionary) Determinants of Earnings Quality

35

tion, research designs have not been able to disentangle the effects of accounting standards from the effects of regulatory scrutiny. This is likely because of difficulties both in measuring the quality of accounting standards and regulatory scrutiny and in identifying settings with sufficient variation in the quality of both constructs so as to enable the effects of each to be separately extracted. We believe that these difficulties are most likely to be overcome in cross jurisdictional or over time settings.

4 Measures of Earnings Quality

Earnings quality is a multi-dimensional concept; therefore, the choice of an earnings quality measure will depend on the research question posed (which dimension of earnings quality is implied by the research question) and the availability of data and estimation models (which measures can be estimated). Some research questions call for a measure of earnings quality that is linked to investors’ perceptions of earnings. For example, research that examines the value relevance of earnings presumes earnings is useful to a particular class of market participants (namely investors) whose aggregate judgments and decisions are summarized by share prices and returns. In contrast, other research questions focus on direct measures of earnings quality constructed using accounting data alone (i.e., without reference to stock prices or returns). Yet another dimension that is important for some research questions is the distinction among total, innate and discretionary earnings quality, as discussed in Section 3.1 1 Every

earnings quality measure discussed in this section can be thought of as having both an innate component and a discretionary component. Indeed, as we note in the context of the often-used abnormal accruals measure, even metrics which are viewed as having

37

38

Measures of Earnings Quality

Francis et al. (2004) identify seven measures of earnings quality (which they refer to as earnings attributes) that have been widely used in accounting research. They characterize the seven earnings attributes as either “accounting-based” or “market-based” depending on the underlying assumptions about the function of financial reporting, and they note that these assumptions will, in turn, influence the way the attributes are measured. The accounting-based earnings attributes are accruals quality, persistence, predictability, and smoothness. These attributes take cash or earnings (or other measures that can be derived from these, such as accruals) as the reference construct, and are estimated using accounting data (not market data). The marketbased attributes are value relevance, timeliness, and conservatism. These attributes take returns or prices as the reference construct and rely on both accounting data and returns data for their estimation. As Francis et al. note, the differences in reference constructs are based on implicit or explict assumptions about the intended function of earnings. Specifically, accounting-based earnings quality measures assume that the function of earnings is to allocate cash flows to reporting periods via accruals, while market-based earnings quality measures assume that the function of earnings is to reflect economic income as represented by stock returns. The remainder of this section describes several measures of earnings quality that have been used in accounting research. (a) Accruals quality. Accruals quality as a measure of earnings quality is based on the view that earnings that map more closely into cash flows are of better quality. Dechow and Dichev (2002) measure of earnings quality captures the mapping of working capital accruals into last-period, current-period, and next-period cash flows from operations. A typical Dechow–Dichev accruals quality measure begins with a model that relates current accruals to lagged, current, and future cash

been purged of the effects of factors believed to contribute to normal fluctuations in the metrics appear to show considerable correlation with factors viewed as innate with respect to earnings quality.

39 flows from operations2 : CF Oj,t−1 CF Oj,t T CAj,t = ϕ0,j + ϕ1,j + ϕ2,j Assetsj,t Assetsj,t Assetsj,t + ϕ3,j

CF Oj,t+1 + νj,t Assetsj,t

(4.1)

where T CAj,t = firm j’s total current accruals in year t = (∆CAj,t − ∆CLj,t − ∆Cashj,t + ∆ST DEBTj,t ); Assetsj,t = firmj’s average total assets in year t and t − 1; CF Oj,t , = cash flow from operations in year t, is calculated as net income before extraordinary items (NIBE, Compustat #18) less total accruals (TA), where (TAj,t = ∆CAj,t − ∆CLj,t − ∆Cashj,t + ∆ST DEBTj,t − DEP Nj,t , and ∆CAj,t = firm j’s change in current assets (Compustat #4) between year t − 1 and year t, ∆CLj,t = firm j’s change in current liabilities (Compustat #5) between year t − 1 and year t, ∆Cashj,t = firm j’s change in cash (Compustat #1) between year t − 1 and year t, ∆ST DEBTj,t = firm j’s change in debt in current liabilities (Compustat #34) between year t − 1 and year t, DEP Nj,t = firm j’s depreciation and amortization expense (Compustat #14) in year t. To obtain a firm-specific, time-series measure of accruals quality, Equation (4.1) is estimated over some interval (e.g., rolling 10-year windows in Francis et al. (2004)), with each estimation yielding firmTime-series _ and year-specific residuals, ν j,t . To obtain a firm-specific, crosssectional measure of accruals quality, Equation (4.1) is estimated each year at the industry level, with each estimation yielding firm- and Cross-sectional _ year-specific residuals, ν j,t . For both time-series and industry cross-section estimations, the regression residuals are used to calculate the accrual quality metric, AccrualQualityj,t = σ(ˆ νj,t ), equal to the standard deviation of firm j’s estimated residuals. 2 McNichols

(2002) discussion of Dechow and Dichev suggests adding variables capturing the change in revenues and fixed assets in place: T CAj,T = φ0,j + φ1,j CF Oj,T −1 + φ2,j CF Oj,T + φ3,j CF Oj,T +1 + φ4,j ∆Revj,T + φ5,j P P Ej,T + υj,T .

40

Measures of Earnings Quality

The first idea behind AccrualQuality as a measure of earnings quality is that cash from operations is the reference construct; large (small) values of the measure correspond to poor (good) accruals quality because there is less (more) precision about the mapping of current accruals into current, last-period and next-period cash flows. The second idea behind the AccrualQuality measure is that the variability of the residuals from Equation (4.1) and not their magnitudes drives the measure. This idea captures the notion that systematically large (or small) residuals in a regression of accruals on cash flows do not create an inference problem for investors, in terms of being able to predict future earnings, because the systematic component of the residual can be identified and adjusted. The standard deviation of a series of systematically large positive residuals may very well be low, indicating that there is little inference problem. AccrualsQuality therefore is consistent with the view that high-mean, low-variance firms have good (not poor) earnings quality. One limitation of the AccrualQuality measure is that it uses a relatively modest portion of total accruals; specifically, it focuses on the mapping of current accruals into lagged, current and one-year-ahead cash flows from operations. The measure does not capture the effects of larger, more numerous and arguably more complicated accruals, such as pensions, depreciation, asset retirement obligations, and deferred tax assets and liabilities. For a sample of firms with a long time series of accounting data (21 years), Ecker et al. (2005) estimate that mean ratio of current accruals to total assets is 1.6%, as compared to a mean ratio of total accruals to total assets of 5.1%. Therefore, for their sample, the non-current portion of accruals is over three times the size of the current portion. Ecker et al. (2005) use a variety of statistical methods to estimate the relation between total accruals (that is, both current and non-current accruals) and prior, current and future period free cash flows. Their analysis is complicated by the fact that, unlike current accruals which should reverse within one year, the reversal periods of non-current accruals vary by the type of accrual as well as by accounting implementation choices. Allowing for firm-specific reversal periods as long as 21 years, their study finds that estimates of noncurrent accruals quality are significantly positively correlated with estimates of current

41 accruals quality. This result suggests that the easier-to-compute and less data-intensive current accruals measure may be a reasonable proxy for the more difficult-to-operationalize but conceptually preferred measure of total accruals quality that encompasses both current accruals and noncurrent accruals. (b) Abnormal accruals. Abnormal accruals as a measure of earnings quality is based on the view that accruals which are not well explained by accounting fundamentals (fixed assets and revenues) are an inverse measure of earnings quality. The abnormal accruals measure is typically estimated using some version of the Jones (1991) approach. Similar to the Dechow–Dichev model, estimations of abnormal accruals use either a firm-specific, time-series approach, or a cross-sectional estimation-by-industry-year approach, to obtain residuals from the following regression: TAj,t ∆Revj,t PPE j,t 1 = κ1 + κ2 + κ3 + εj,t . Assetj,t−1 Assetj,t−1 Assetj,t−1 Assetj,t−1 (4.2) Regardless of the estimation procedure (time-series or cross-sectional), the parameter estimates from Equation (4.2) are used to estimate firmspecific normal accruals (NA) as a percent of lagged total assets: N Aj,t = κ ˆ1

(∆Revj,t − ∆ARj,t ) PPE j,t 1 +κ ˆ2 +κ ˆ3 , Assetj,t−1 Assetj,t−1 Assetj,t−1

where ∆ARj,t = firm j’s change in accounts receivable (Compustat #2) between year t − 1 and year t, and to calculate abnormal accruals (AA) T Aj,t in year t, AAj,t = Assetj,t−1 − N Aj,t .3 Abnormal accruals are traditionally viewed as capturing management’s financial reporting decisions (or discretion). When studying earnings quality questions, researchers tend to focus on the absolute value of abnormal accruals (|AAj,t |) because the research question typically does not impose a directional sign on management’s expected 3 The

original Jones (1991) model has been extended in two ways. The first is the inclusion of the change in accounts receivable in the estimation of normal accruals. The second is an adjustment of the abnormal accruals by performance-matching (Kothari et al. (2005); McNichols (2000)). Based on Francis et al. (2005a) finding that variations of the |AAj,t | measure from these extensions yield similar results, we do not discuss these extensions or adjustments in more detail.

42

Measures of Earnings Quality

financial reporting decisions. In contrast, research questions related to earnings management typically impose a directional prediction, so the focus is on a variable which reflects that directional prediction, such as the signed value of abnormal accruals (AAj,t ). We distinguish earnings management research (which posits a specific reporting context and incentive structure and focuses on whether signed abnormal accruals are consistent with predictions) from earnings quality research (which may or may not examine a specific setting, but typically does not generate predictions about the sign of abnormal accruals). We think of the absolute value of abnormal accruals as an accruals-based proxy for earnings quality, with larger values of |AAj,t | indicating poorer accruals quality. The abnormal accruals measure of earnings quality (|AAj,t |) differs from the Dechow–Dichev accruals based measure (AccrualQuality) in that the former is intended to reflect the portion of accruals that is not driven by accounting fundamentals (or what we term innate factors), whereas the latter measure includes accruals that arise from both innate and discretionary sources. The fundamental or innate factors driving accruals in the Jones model are changes in revenues and fixed assets; the ability of these factors to explain the ratio of total accruals to total assets determines the magnitude of normal accruals, from which abnormal accruals are determined by subtracting estimated normal accruals from total accruals. Abnormal accruals are intended to reflect reporting influences on earnings quality, not innate influences. Given this distinction, it is natural to ask whether these two accruals-based measures of earnings quality (AccrualsQuality and |AAj,t |) capture similar constructs, given that one is intended to measure total accruals quality while the other is intended to measure discretionary accruals quality. The answer to this question lies in the completeness of the innate factors used to capture the determinants of normal accruals. Specifically, should this set include other fundamentals, beyond the change in revenues and fixed assets? Summary data reported in Francis et al. (2008b, Table 2, Panel C) indicate that several innate factors proposed by Dechow and Dichev (2002) and others (the standard deviation of cash flows, the standard deviation of sales revenues, firm size as measured by the log of total assets, operating

43 cycle, incidence of loss observations, intangibles intensity, and capital intensity) explain about 52% of the variation in the Dechow–Dichev measure of accruals quality and over 65% of the variation in |AAj,t |. Assuming that the set of innate factors explains normal fluctuations in accruals arising from the firm’s business model and operating environment, Francis et al.’s (2008b) finding implies that residuals obtained from estimations of modified Jones type models will contain sizable amounts of accruals that are not unusual; that is, they would be expected given the business model and operating environment. Stated differently, the finding that innate factors explain 65% of an abnormal accruals measure raises questions about whether the measure is, in fact, abnormal, in the sense of capturing effects of management’s reporting decisions, as opposed to capturing the financial reporting effects of business fundamentals. It is not our intent to comment on the appropriateness of the fundamental variables used in Jones-type estimations. Rather, our point is that, in the context of earnings quality research where distinctions between innate and discretionary sources are important, we believe it is not appropriate to treat the abnormal accruals measure (as conventionally estimated) as capturing only discretionary accrual behaviors. Rather, we believe |AAj,t | should be viewed as a measure of total accruals quality, distinct from the Dechow–Dichev measure. (c) Persistence. Persistence as a measure of earnings quality is predicated on the view that more sustainable earnings are of higher quality. In its simplest form, earnings persistence is measured as the slope coefficient estimate, φ1,j , from an autoregressive model of order one (AR1) for annual split-adjusted earnings per share (Xj,t , measured as firm j’s net income before extraordinary items in year t divided by the weighted average number of outstanding shares during year t): Xj,t = φ0,j + φ1,j Xj,t−1 + υj,t .

(4.3)

Equation (4.3) is typically estimated in time-series, firm-by-firm, using maximum likelihood estimation. The resulting estimate of φ1,j captures firm j’s persistence of earnings. Values of φ1,j close to one imply highly persistent (i.e., high quality) earnings, while values of φ1,j close to zero imply highly transitory (i.e., low quality) earnings.

44

Measures of Earnings Quality

(d) Predictability. Predictability, defined as the ability of earnings to predict itself (Lipe, 1990), as a measure of earnings quality is based on the view that an earnings number that tends to repeat itself is of high quality. This view is not dissimilar to the view, implied by Dechow and Schrand (2004), that a high quality earnings number is representative, that is, a good predictor, of future earnings. One common measure of earnings predictability is derived from the same firm-specific models used to estimate earnings persistence; this measure is the square p root of the error variance from Equation (4.3), P redictability = σ 2 (ˆ υj ). Large (small) values of Predictability imply lower (higher) quality earnings. A second common measure of earnings predictability is based on analysts’ forecast errors; this measure is the average absolute forecast error of analysts’ annual earnings forecasts. Forecast error is typically measured as the analyst’s forecast of EPS less reported EPS, scaled by share price 10 days before the forecast date. Relative to the time-series based measure of predictability, the analyst-based measure contains two additional behavioral influences that have no bearing on earnings quality. The first influence reflects analyst biases that affect their earnings forecasts (self-selection, cognitive, or incentive-based) and the second reflects management’s earnings guidance (e.g., Bartov et al., 2002). Both influences affect forecast errors, but are not related the quality of the earnings number itself. (e) Smoothness. The smoothness of earnings is typically measured relative to some measure of cash flows. This measurement uses cash flow as the reference construct for unsmoothed earnings, and therefore assumes that cash flows are unmanaged. As an indicator of earnings quality, smoothness reflects the idea that managers use their private information about future income to smooth out transitory fluctuations and thereby achieve a more representative (i.e., normalized) reported earnings number. To the extent that current earnings which are more representative of future earnings are of higher quality, smoother earnings indicate higher quality earnings. Not all researchers accept the premise that managers use their private information about the future to manipulate accruals with the goal of achieving a more representative earnings number. An alternative

45 view, expressed by, for example, Leuz et al. (2003), is that smoothness reflects the extent to which accounting standards (as well as other influences on earnings such as regulatory oversight mechanisms and legal regimes) allow managers to artificially reduce variability in earnings, presumably to obtain some capital market benefits associated with a smooth earnings stream. Under this view, smoother earnings would indicate poorer earnings quality. Results reported by Francis et al. (2004) suggest that capital market participants reward smoother earnings streams with reduced costs of equity and debt capital. This finding is consistent with a view that earnings smoothness is desirable (at least in the eyes of investors) because it reflects higher quality financial reporting decisions. However, results in Francis et al. cannot rule out the possibility that investors reward earnings smoothness for reasons unrelated to earnings quality. Smoothness has been measured in several ways, all of which are likely to be highly correlated: (1) the ratio of firm j’s standard deviation of net income before extraordinary items divided by beginning total assets, to its standard deviation of cash flows from operations divided by beginning total assets (Francis et al., 2004); (2) the ratio of the firm’s standard deviation of operating income scaled by assets, to the standard deviation of cash flows from operations scaled by assets (Leuz et al., 2003); and (3) the ratio of the standard deviation of non-discretionary net income (equal to operating cash flows plus nondiscretionary accruals) to the standard deviation of cash flows from operations (Hunt et al., 2000). (f) Earnings variability. Earnings variability, typically measured as the standard deviation of (scaled) earnings, is statistically and conceptually related to both smoothness and accrual quality. Therefore, the presumptions that would make earnings variability a measure of earnings quality are similar to the presumptions needed to support smoothness and accrual quality as indicators of earnings quality. As documented by Dechow and Dichev (2002, Table 4) earnings variability is the strongest instrument for their accrual quality measure. Earnings variability is also an instrument for smoothness, since the two measures differ only in the presence or absence of standardization by cash

46

Measures of Earnings Quality

flow variability. Francis et al. (2004) document, for example, Pearson (Spearman) correlations between EarnVar and AccrualQuality of 0.77 (0.84) and between EarnVar and Smoothness of 0.57 (0.66), respectively. Measures of earnings variability are typically estimated using a firm-specific time-series of scaled earnings. For example, Dechow and Dichev (2002) measure earnings variability as the rolling standard deviation of firm j’s net income before extraordinary items, scaled by beginning of year total assets, EarnV arj,t = σ(N IBEj,t ). Larger (smaller) values of EarnVar indicate more (less) volatile earnings. (g) Value relevance. Value relevance as a measure of earnings quality is based on the idea that accounting numbers should explain the information that is impounded in returns. Therefore, value relevance is the ability of one or more accounting numbers to explain variation in stock returns. Earnings with greater explanatory power are viewed as more desirable; that is, earnings that explain greater variation in returns are of higher quality. Typically, in accounting research (e.g., Francis and Schipper (1999); Collins et al. (1997); Bushman et al. (2004)), value relevance is based on the explanatory power (the adjusted R2 ) of the following regression of returns on the level and change in earnings: RETj,t = δ0,j + δ1,j EARNj,t + δ2,j ∆EARNj,t + ζj,t .

(4.4)

Small (large) values of the adjusted R2 imply less (more) value relevant earnings. In estimating regressions like expression (4.4) the researcher chooses an earnings metric and estimation periods. Earnings periodicity can be annual or quarterly. For example, RETj,t = firm j’s 15-month return ending 3 months after the end of fiscal year t; EARNj,t = firm j’s income before extraordinary items in year t (NIBE ), scaled by market value at the end of year t − 1; ∆EARNj,t = change in firm j’s NIBE in year t, scaled by market value at the end of year t − 1. Expressions like (4.4) can be estimated in time-series, in cross-section, and in pooled time-series cross-sections. To obtain a firm-specific estimate of value

47 relevance, the sample is limited to firms with a sufficient time-series of data to estimate (4.4) at the firm-level. (h) Earnings informativeness (or earnings response coefficient). Earnings informativeness (or the earnings response coefficient) is measured as the estimated slope coefficient on the level or change in earnings, or some aggregation of the estimated slope coefficients on both the level and change in earnings, from expressions like (4.4). The dependent variable could be a long-term measure (e.g., annual returns) or an indicator of a short term market reaction to an event, such as a three-day cumulative abnormal return surrounding an earnings announcement. The use of the slope coefficient on earnings (level or change) as an indicator of earnings quality is based on Teoh and Wong’s (1993) analytical model of the relation between share price responses to earnings and the precision of the earnings signal.4 Teoh and Wong show a positive relation between the credibility of accounting information and the coefficient relating returns to earnings. Because many factors affect the slope coefficient relating returns to earnings,5 most prior research tends to control for these other factors in estimating earnings informativeness. Stated differently, earnings informativeness likely proxies for earnings quality as well as other factors. Thus, to understand the portion of earnings informativeness that is associated with earnings quality, researchers typically posit an intervening variable that causes earnings quality to vary (e.g., Teoh and Wong (1993) posit audits by a Big 8 auditor) and test for an incrementally positive slope coefficient for firms with better earnings quality based on that intervening variable (e.g., firms with Big 8 auditors are posited to have better earnings quality and therefore greater earnings informativeness). Under this approach, researchers typically do not calculate a firm-specific estimate of earnings informativeness; rather, they

4 Their

model draws on Holthausen and Verrecchia (1988) model of the determinants of the magnitude of the price response to an information release. Specifically, the share price response to earnings (that is, earnings informativeness) is an increasing function of the amount of prior uncertainty about firm value (ν) and a decreasing function of the noise (i.e., the lack of credibility) of the earnings signal (η). 5 Prior research (Easton and Zmijewski (1989); Collins and Kothari (1989)) shows that the slope coefficient varies with earnings persistence, firm size, and interest rates.

48

Measures of Earnings Quality

infer lower or higher earnings informativeness for subsamples of firms, separated by variation in the intervening variable.6 (i) Earnings opacity. Bhattacharya et al. (2003) define earnings opacity as the extent to which the distribution of reported earnings fails to accurately reflect the true distribution of (unobservable) economic earnings.7 They note that earnings opacity at a point in time should be inversely related to the average level of earnings informativeness at that same point in time. Bhattacharya et al.’s proxy for earnings opacity is an index that combines earnings aggressiveness, loss avoidance, and earnings smoothing. Each of these constructs, as well as the opacity index, could be viewed as a measure of earnings quality; therefore, the reasoning to support earnings opacity as a measure of earnings quality parallels previously presented discussions. Earnings aggressiveness is measured as total accruals scaled by lagged total assets. Larger (smaller) fractions of accruals are viewed as being more (less) aggressive. Loss avoidance is measured as the ratio of the percentage of firms reporting small positive earnings (i.e., ratios of net income to total assets between 0% and 1%) to the percentage of firms reporting small negative earnings (i.e., ratios of net income to total assets between −1% and 0%). The higher is this ratio (that is, the more likely are small positive earnings relative to small negative earnings), the greater is loss avoidance. Finally, earnings smoothing is measured as the correlation between accruals and cash flows, both scaled by lagged assets. (j) Timeliness. Timeliness is similar to value-relevance, in that the reference construct for this measure is stock returns and the measure itself is based on explanatory power. Timeliness captures the ability of earnings to reflect good news and bad news that is impounded in returns, and is measured as the explanatory power of a reverse regression of earnings on returns. The use of timeliness as a measure 6 Nothing

precludes estimation of the slope coefficients at the firm-specific level using timeseries data. However, prior research suggests that such estimates would not be likely to reflect solely (or perhaps even mostly) the effects of earnings quality. 7 Bhattacharya et al.’s focus is on the pricing effects of earnings opacity world wide. Therefore, they measure earnings opacity at the country level not the firm level. Of the individual measures used to form their earnings opacity index, only loss avoidance is not readily amenable to calculation at the firm level.

49 of earnings quality is based on the same presumptions that support value relevance as an earnings quality measure. Following Ball et al. (2000) and Bushman et al., the measure of timeliness is the adjusted R2 from Equation (4.5). Smaller values of Timeliness imply less timely (i.e., lower quality) earnings. Equation (4.5) is typically estimated on a firm-specific basis in time series. EARNj,t = α0,j + α1,j N EGj,t + β1,j RETj,t + β2,j N EGj,t · RETj,t + ςj,t ,

(4.5)

where N EGj,t = 1 if RETj,t < 0 and 0 otherwise; all other variables are as previously defined. (k) Conservatism: Watts (2003) argues that conservatism is a desirable attribute of earnings because conservative reporting constrains overpayments to stakeholders. This approach is based more on an ex post contracting/stewardship approach to earnings quality than the ex ante capital allocation approach we have adopted in this discussion review. Taking a market-based approach, Ball et al. (2000) define conservatism as the differential ability of accounting earnings to reflect economic losses (measured as negative stock returns) versus economic gains (measured as positive stock returns). Following Basu (1997), Ball et al. measure conservatism as the ratio of the slope coefficients on negative returns to the slope coefficients on positive returns in a reverse regression of earnings on returns, as, for example, in Equation (4.5).8 Accounting-based measures of conservatism include Penman and Zhang’s (2002) C-Score, equal to the sum of the LIFO reserve, estimated R&D assets and estimated advertising assets, scaled by net operating assets. C-Score is a balance sheet measure of conservatism which captures the notion that firms with larger unrecorded assets (larger C-Scores) are more likely to have conservatively reported financial statements than firms with smaller C-Scores. Penman and Zhang also calculate a Q-score which captures the effect of conservatism on the income statement. 8

The combination of timeliness and conservatism is sometimes termed “transparency,” viewed by some researchers as a desirable attribute of earnings (see, for example, Ball et al. (2000)).

50

Measures of Earnings Quality

(l) e-loadings. Ecker et al. (2006) put forward a returns-based measure of earnings quality. The e-loading is the slope coefficient from a regression of excess returns on a factor mimicking portfolio that captures earnings quality (AQfactor ), controlling for other factors known to affect returns (market risk premium, size, and book-to-market ratio). The calculation of AQfactor follows the Fama–French procedures for calculating factor-mimicking portfolios; the exact procedures are explained in Ecker et al. (2006).9 Because AQfactor is time-specific, not firm-specific, it can be correlated with the returns of any firm to determine that firm’s exposure to poor earnings quality. This use of AQfactor is analogous to correlating a firm’s excess returns with the market risk premium and interpreting the resulting firm-specific beta estimate as a measure of investors’ perception of that firm’s exposure to market risk. The specific correlation measure Ecker et al. use, that is, the e-loading is the coefficient estimate on AQfactor obtained from 1-factor (superscript 1f ) and 3-factor (superscript 3f ) asset pricing regressions which include AQfactor as an independent variable: 1-factor :

1f 1f Rj,t − RF,t = αj,T + βj,T (RM,t − RF,t ) 1f + e1f j,T AQf actort + εj,t

3-factor :

(4.6)

3f 3f (RM,t − RF,t ) + s3f + βj,T Rj,t − RF,t = αj,T j,T SMB t 3f 3f + h3f j,T HMLt + ej,T AQfactor t + εj,t

(4.7)

Expressions (4.6) and (4.7) can be measured using daily, weekly or monthly returns, over intervals that are a research design choice. Assuming use of daily returns, t = index for a trading day in year T ; Rj,t = firm j’s return on day t; RF,t = the risk free rate on day t; RM,t = the market return on day t; SMB t = Fama–French small-minusbig factor on day t; HMLt = Fama–French high-minus-low book-tomarket factor on day t. 9 Ecker

et al. (2006) calculate a factor-mimicking portfolio based on the Dechow–Dichev (2002) accruals quality measure. However, the same procedures could be used to create a factor-mimicking portfolio based on alternative earnings quality measures.

51 3f For the 1-factor (3-factor) specification, e1f j,T (ej,T ) is the estimate of firm j’s sensitivity to poor earnings quality in year T and is therefore a returns-based representation of earnings quality. The other slope 1f 3f 3f coefficients, βj,T (or βj,T ), s3f j,T and hj,T , capture the firm’s exposure to returns-based representations of market risk, size, and book-to-market, respectively, in year T . Larger values of the loadings (slope coefficients) imply greater sensitivity to the factor; thus, larger e-loadings imply a greater sensitivity to poor earnings quality. Ecker et al. estimate 1-factor and 3-factor e-loadings for each firmyear, using daily data for year t = 1970–2003, for all firms with at least 100 trading returns in year t. They perform several construct validity tests to determine whether e-loadings behave as one would expect if they indeed capture earnings quality. High e-loading firms, for example, have smaller earnings response coefficients and greater dispersion and lower accuracy of analyst forecasts. Higher e-loadings also characterize firms in financial trouble, as indicated by bankruptcy filings, restatements or shareholder lawsuits over financial reporting. Chen et al. (2007b) also validate the e-loading as a measure of earnings quality. They find that dividend-decreasing firms show a predictable increase in their e-loadings following the announcement of the dividend decrease. They interpret this result as evidence that the dividend decrease signals to investors a deteriorating assessment of future performance, thus creating greater uncertainty about the sustainability of current earnings. Ecker et al. discuss several advantages of the e-loading, relative to other measures of earnings quality. These advantages relate to sample size, sample selection, and flexibility. The sample size and sample selection advantages arise because the data requirements for estimating a firm’s e-loading are modest, relative to the data requirements for estimating other earnings quality proxies. An e-loading can be estimated for any firm with sufficient daily returns in a given year to estimate a CAPM (or 3-factor) asset pricing regression. Sample sizes will therefore be larger and exhibit less selection bias than if, for example, 10 years of firm-specific accounting data were required to estimate an earnings quality proxy. The flexibility advantage of e-loadings relative to other

52

Measures of Earnings Quality

earnings quality measures arises because e-loadings can be calculated with daily returns data, they need not be aligned with reporting periods and they can be estimated over intervals as short as 45 days. This flexibility means that e-loadings can be used to examine shifts in earnings quality over short intervals and around events. In comparison, other measures of earnings quality are linked to annual or quarterly reporting periods, cannot be applied to short intervals, and cannot be specific to a given financial statement date because they are constructed from current and prior data.

5 Earnings Quality and Market Outcomes

In Sections 1–4, we have discussed concepts and measurements of earnings quality, with reference to market outcomes. In this section, we begin (Section 5.1) with a discussion of market outcome indicators that have been considered by researchers, and describe some empirical research design issues that we believe should be considered in constructing tests of the relation between earnings quality and market outcomes and in interpreting the results of those tests. Sections 5.2 and 5.3 discuss analytical and empirical evidence on the link between earnings quality and expected (normal) returns and unexpected (abnormal) returns, respectively. Section 5.4 concludes with a discussion of the relations between earnings quality and other market outcomes, such as bid-ask spreads and idiosyncratic volatility.

5.1

Market Outcomes

For purposes of this discussion review, we identify two broad categories of market outcomes: (1) the cost of capital, which we view as a summary indicator of investors’ resource allocation decisions and (2) other market outcomes, viewed by some researchers as intermediary outcomes, 53

54

Earnings Quality and Market Outcomes

including, for example, properties of analysts’ earnings forecasts such as accuracy and dispersion. We describe each of these broad categories in turn. Cost of capital.1 Conceptually, the cost of equity capital is the ex ante return demanded by suppliers of equity capital, or, equivalently, the discount rate which, when applied to expected cash flows, yields the current stock price. Empirically, the cost of equity capital can be estimated by reference to analysts’ forecasts (of earnings, price, and dividends) or by reference to market data such as price-earnings ratios or returns. Analyst-forecast-based measures of the cost of equity are ex ante, in the sense that they reflect investors’ (as proxied by analysts) expectations or forecasts of outcomes such as earnings and dividends. A number of these measures, including their derivations from a dividend discount formula for share price, the assumptions required and the outcomes of construct validity assessments, are described in Botosan and Plumlee (2005). Because analyst-based estimates of the cost of equity capital are all intended to capture the same construct, differences among actual estimates would be attributable to differences in assumptions, demands on the data and samples used in the estimation. For this reason, researchers who use the cost of equity capital as a market outcome measure tend to test the sensitivity of their results to differences in cost of capital estimates. When the cost of equity capital is analyzed by reference to returns, the researcher bases estimates on a time-series of realized returns, where a firm’s realized return is defined as the price-scaled change in its stock price between two dates, adjusted for dividends. The realized return has two components: the expected (normal) return and the unexpected (abnormal) return. That is, the realized return can be thought of as containing both an ex ante expected component (the normal or expected return) and an ex post outcome component (the abnormal or unexpected return). The normal or expected return is the returns-based analog of the analyst-forecast-based measures of the cost of equity.

1 This

discussion focuses on the cost of equity capital. Proxies for the cost of debt capital include bond ratings (an ex ante measure) and realized costs of debt (an ex post measure); see, for example, Francis et al. (2005a).

5.1 Market Outcomes

55

The distinction between the normal and abnormal return components of realized returns is pertinent for analyses of the link between earnings quality and market outcomes as captured by realized returns, because earnings quality has the potential to affect both components, but in different ways. Specifically, earnings quality will affect the expected return component of realized return if earnings quality is perceived by investors as proxying for information risk (or information precision), and information risk is a priced factor. This information precision effect, to the extent it exists, is expected to be linear: poorer (better) earnings quality is associated with higher (lower) expected returns. However, the effect of earnings quality on the abnormal return component of realized return is expected to be nonlinear, because when earnings quality is poor there is a higher incidence of mispricing, which can result in either positive abnormal returns or negative abnormal returns. As a consequence, the relation between abnormal returns and earnings quality is U-shaped. These different effects present research design issues for empirical investigations of the association between earnings quality and realized returns, because those returns contain both the expected returns (linear) piece and the abnormal returns (nonlinear) piece. Other market outcomes. In the context of earnings quality analyses, researchers have identified a number of alternative (to the cost of capital) indicators of market outcomes. In this section, we describe three of these indicators, two of which have been used in research that is discussed in Sections 5.2–5.4. The bid-ask spread (or one of its components) has been used as a measure of liquidity (e.g., Amihud and Mendelson, 1986) or of information asymmetry (e.g., Huang and Stoll, 1996). Regardless of the interpretation of this market outcome measure, research which has linked it with earnings quality has taken the position that higher quality earnings should reduce bid-ask spreads. A direct measure of information asymmetry, put forward by, for example, Easley et al. (1996); Easley et al. (2002), is the probability-of-informed-trading score or PIN score. As discussed in more detail in Section 5.2, information asymmetry is viewed in some models as a mediating link between information quality and the cost of capital.

56

Earnings Quality and Market Outcomes

The quality of analysts’ earnings forecasts, as measured by accuracy and dispersion, has also been viewed, either implicitly or explicitly, as a market outcome indicator in analyses of information quality. For example, Dechow and Scharand’s (2004) discussion of earnings quality indicates that high quality earnings would, among other things, be a normalized or representative performance indicator; we view this characteristic as manifesting itself as both accuracy and consensus (low dispersion) of analysts’ earnings forecasts. As another example, Ecker et al. (2006) use analyst forecast accuracy and dispersion in construct validity tests of a returns-based measure of earnings quality. To summarize, research has used a variety of indicators of market outcomes in investigations of the link between earnings quality capital market effects. One class of indicators defines market outcomes such as cost of capital, abnormal returns, and returns volatility. Other indicators focus on information asymmetry, liquidity and the behaviors of information intermediaries.

5.2

Earnings Quality and Expected Returns

Analytical models Traditional models of market equilibrium, such as the capital asset pricing model (CAPM) generally assume homogeneous beliefs among investors, and that all information is incorporated in share price at all times (e.g., Sharpe (1964), Lintner (1965), Black (1972)). Under these assumptions, the quality of information has no bearing on asset prices. However, models that relax the homogeneous beliefs assumption — that is, models that allow for information asymmetry — do allow for the possibility that information quality matters. For example, Merton (1987) develops a model in which some investors are less informed than others in the sense that they are unaware of certain firms; information that increases uninformed investors’ awareness of these firms leads to a lower cost of capital for those firms. As another example, Diamond and Verrecchia (1991) model the link from the information provided by the firm to liquidity. They show that policies that reduce the information asymmetry between informed and uninformed investors can reduce a security’s liquidity risk, which in turn can affect expected return; Amihud and Mendelson (1986) provide an empirical

5.2 Earnings Quality and Expected Returns

57

demonstration of this analytical result. These studies argue (and show) that, under certain conditions of information asymmetry, the quality of information affects asset prices. Easley and O’Hara (2004) take the information asymmetry effect of information quality to the level of an asset pricing model. In their model, information asymmetry between informed and uninformed investors creates unavoidable information risk, which is priced in equilibrium. Taking a different approach, Hughes et al. (2007) conclude that while information asymmetry affects factor risk premia, it does not lead to cross-sectional differences in expected returns in their model. However, Hughes et al. note that while an asymmetric information factor does not arise endogenously in their model, their model is silent on whether a systematic information factor exists. Lambert et al. (2007b) also present an equilibrium model that allows for a information asymmetry/liquidity effect on cost of capital under conditions of imperfect competition among investors, along the lines of Diamond and Verrecchia (1991). However, their main result is not based on information asymmetry; rather, it is based on information imprecision. They dispute Easley and O’Hara’s attribution of the expected returns effect of information quality to information asymmetry. They argue that the expected returns effect stems from the average level of information imprecision, regardless of how information is distributed across investors — that is, the expected returns effect is due to information uncertainty per se, not to the presence of differences in information uncertainty across investors. They further show that the information precision effect on expected returns is not diversifiable when there are many firms whose cash flows covary. There are also theoretical studies that investigate the effect of earnings quality in a CAPM-consistent world. Yee (2006) links earnings quality to the equity premium in an infinite-horizon consumption CAPM economy. He splits risk into two components: fundamental risk and earnings quality risk. Fundamental risk (uncertainty about future dividend payments) is driven by factors such as business models and organizational structures. Earnings quality risk is generated by noisy earnings reports, caused by a weak association between accounting

58

Earnings Quality and Market Outcomes

earnings and economic earnings, or manipulated accounting numbers. In this model, fundamental risk must exist if there is to be an earnings quality effect on expected return. When fundamental risk increases, the effect of earnings quality on the cost of capital increases; that is, risk effects due to earnings quality magnify the effects of fundamental risk. Lambert et al. (2007a) also build a CAPM-consistent model in which accounting quality can influence expected returns both directly and indirectly. The direct effect occurs because higher quality disclosures affect the firm’s assessed cash flow covariance with other firms’ cash flows, and this effect is non-diversifiable. The indirect effect occurs because higher quality disclosures affect real decisions, which can change the firm’s ratio of expected future cash flows to the covariance of these cash flows with the sum of all the cash flows in the market. Lambert et al. (2007a) point out that there is no separate role for earnings quality as a risk factor in a CAPMconsistent world; the earnings quality effect on expected return occurs because earnings quality is one determinant of the unobservable forward-looking beta. In his discussion of Lambert et al., Indjejikian (2007) notes that the last point follows as a matter of course, because a single-factor model such as the CAPM cannot, by definition, generate predictions about other risk factors. More generally, Indjejikian notes that since the CAPM fails to explain the size and value premia in the cross-section of expected returns, and fails to explain the variation in the equity premium over time, the empirical relevance of predictions that rely on the CAPM may be limited. Empirical evidence. Empirical research is generally supportive of an earnings quality effect on expected returns. For example, Bhattacharya et al. (2003) find an association between country-level measures of the average cost of equity and earnings opacity, measured as a combination of earnings aggressiveness, loss avoidance, and earnings smoothing, as described in Section 4. At the firm level, Francis et al. (2004) calibrate the expected returns effects of accruals quality, persistence, predictability, smoothness, value relevance, timeliness, and conservatism on a sample of Value-Line-followed firms. Using an implied cost of equity metric

5.2 Earnings Quality and Expected Returns

59

derived from Value Line’s target price and dividend forecasts,2 they find that firms with the least favorable values of each attribute, i.e., poor earnings quality firms, generally have higher expected returns. Three accounting-based attributes (accruals quality, persistence, and smoothness) have the largest expected returns effects in terms of both economic magnitude and statistical significance: up to 41 basis points expected returns differential per ranked attribute decile.3 The expected returns effects of the market-based attributes they examine (value relevance, timeliness, and conservatism) are more muted, never larger than 13 basis points per ranked attribute decile. Francis et al. (2005a) provide a detailed investigation of the expected returns effects of two proxies for earnings quality: the Dechow–Dichev (2002) measure of accruals quality and absolute abnormal accruals (estimated using various implementations of the modified Jones (1991) model). For both proxies, they confirm the expected returns effect as well as a cost of debt effect of earnings quality. Specifically, they show that industry-adjusted (as well as unadjusted) earnings-to-price ratios are significantly higher for poor earnings quality firms than for good earnings quality firms; that poor earnings quality firms have significantly higher equity betas, higher realized costs of debt, and worse debt ratings; and that an earnings quality factor mimicking portfolio loads significantly in asset pricing regressions, where it both adds significant incremental explanatory power (over and above the CAPM and the Fama–French (1993) 3-factor model) and significantly lowers the magnitude of the size and market loadings. Francis et al. also investigate whether the expected returns effect is attributable to the innate portion of earnings quality, the discretionary portion, or both. Their approaches to identifying the innate and discretionary components of earnings quality are described in Section 3. Consistent with their predictions, they find that the expected 2 Francis

et al. (2004) use an implied cost of equity measure both to avoid the substantial methodological concerns associated with using realized returns, especially for short time periods, and because their focus is on expected returns, not the combination of expected and abnormal returns that are embedded in an analysis of realized returns. 3 Another way of stating this difference is to say that the 10% of the sample firms with the worst earnings quality have a cost of capital that is about 369 basis points higher than the 10% of sample firms with the best earnings quality (9 steps of 41 bp each).

60

Earnings Quality and Market Outcomes

returns effects of innate earnings quality are substantially larger (up to two to five times the effect) than the effects of discretionary earnings quality. We offer two observations on Francis et al.’s finding that innate earnings quality has larger expected returns effects than does discretionary earnings quality. First, the finding is consistent with the idea that the information uncertainty which results from operating a business model in an uncertain world is fundamental, in the sense of dominating reporting uncertainty, which can easily shift from period to period and is largely the result of relatively short term management decisions. This finding is also consistent with Guay et al.’s (1996) view (discussed in Section 3.2) that in broad samples and over long time periods, accruals will contain a mixture of quality-increasing and quality-decreasing accruals. Second, in the context of models in which information quality has the potential to have capital market effects, discretionary earnings quality more readily lends itself to an information asymmetry interpretation than does innate quality. In particular, a manager making a financial reporting decision as to how to account for an event or transaction would know the underlying economics of the event or transaction being reported on. (It is possible that some or much of this information is also known to well informed analysts and investors.) The manager can reduce the information disadvantage of uninformed investors (i.e., he can reduce information uncertainty) by selecting a more revealing accounting implementation. In contrast, the innate portion of earnings quality reflects economic fundamentals — volatility emanating from business models and operating environments — about which the manager, and potentially well informed investors, have a much smaller information advantage relative to market participants in general. As a result, it seems more (less) difficult to conceptualize the innate (discretionary) component of earnings quality as capturing an information asymmetry type notion. Viewed in this way, the differential capital market effects of innate and discretionary earnings quality can potentially inform on the debate about whether information asymmetry drives expected returns effects (Easley and O’Hara (2004)) or whether the average level of information precision drives the observed expected returns effects

5.2 Earnings Quality and Expected Returns

61

(Lambert et al. (2007b)), with information asymmetry effects being largely diversifiable (Hughes et al. (2007)). Several studies report effects similar to those in Francis et al. (2005a) regarding a substantially larger cost of capital effect of the innate portion of earnings quality compared to the discretionary portion. For example, Gray et al. (2007) use realized cost of debt, industryadjusted price earnings ratios and asset pricing regressions to test for cost of capital effects of accruals quality in the Australian market. When they split accruals quality into an innate portion and a discretionary portion, they find strong evidence that innate accruals quality is priced, whereas the discretionary portion has small and insignificant pricing effects. Gray et al. attribute the finding that discretionary earnings quality has an insignificant cost of capital effect to Australian institutional and regulatory factors that limits disclosure discretion. Similarly, Cohen (2006) finds that total earnings quality (he uses four proxies for earnings quality) is significantly priced. When he uses an instrumental variable technique to account for the choice portion of earnings quality (using predicted values from a regression of earnings quality on variables likely to influence managerial choice), the pricing effects for this choice portion are insignificant. This finding is consistent with the smaller cost of capital effects documented by Francis et al. for the discretionary component of earnings quality. Also similar to Francis et al. (2005a), Liu and Wysocki (2007) find that when one controls for innate determinants of earnings quality as measured by accruals quality, the remaining pricing effect is modest. In fact, they find that the pricing effect of accruals quality is insignificant when one includes all innate determinants and all the other earnings quality and information asymmetry measures in the same regression, and restricts the sample to NYSE firms. Two empirical cost of capital studies explicitly link earnings quality and fundamental risk. First, Chen et al. (2007a) test the empirical prediction from Yee’s (2006) model, described earlier. Chen et al. confirm that the expected returns effect of earnings quality as measured by accruals quality is an increasing function of underlying fundamental risk. They use several proxies for expected return: asset pricing

62

Earnings Quality and Market Outcomes

regressions, price-earnings ratios, and the average implied cost of capital estimate from four models (Gebhardt et al. (2001), Claus and Thomas (2001), Gode and Mohanram (2003), and Easton (2004)). Second, Nichols (2006) shows that an accruals quality factor mimicking portfolio has a substantial realized returns effect (an annualized factor premium of 10.8%)4 and that the effect is associated with cash flow shocks, i.e., it is associated with fundamental risk. Researchers have also explored the link between cost of capital and other measures of earnings quality. Barone (2003) finds that metrics based on Lev and Thiagarajan (1993) scores derived from fundamental financial ratios, as well as a self-constructed financial reporting quality score, are significantly related to expected return, measured as the implied cost of capital from a modified PEG ratio (Easton (2004)) and, separately, as the implied cost of capital from the Ohlson and J¨ uttnerNauroth (2005) model. Barth et al. (2006) find that the extent to which earnings and change in earnings covary contemporaneously with stock returns, which they label financial statement transparency, is significantly associated with expected return, as measured by the Fama and French (1993) 3-factor model. Berger et al. (2006) construct and validate a returns-based information quality metric, the ratio of firm-specific return variation to firm-specific cash-flow variation, and show that their measure is significantly associated with expected return. Berger et al. use three proxies for expected return: the implied cost of capital from the Gebhardt et al. (2001) model, the implied cost of capital from the Ohlson and J¨ uttnerNauroth (2005) model as implemented in Easton (2004), and cost of capital estimates from the Fama and French (1993) 3-factor model. Chaney et al. (2007) perform a latent cluster on the seven earnings attributes analyzed by Francis et al. (2004) and show that firms can be classified into different levels of reporting quality based on multiple earnings quality attributes. Moreover, they find that their empirical categorization adds incremental explanatory power to the earnings quality — expected returns link, over and above the effects of the indi4

Nichols also argues that the empirical expected returns effect of earnings quality is too large to be attributable solely to earnings quality.

5.2 Earnings Quality and Expected Returns

63

vidual attributes documented in Francis et al. (2004).5 Chaney et al. use expected returns estimates from the Fama and French (1993) threefactor model. Motivated by theoretical studies described earlier in this section that predict a direct link from earnings quality to cost of equity or an indirect link through information asymmetry, Bhattacharya et al. (2007b) analyze the relative importance of the different paths that potentially link earnings quality and cost of equity. They find that the direct link from earnings quality to cost of equity dominates the indirect link (through information asymmetry) in all settings they investigate: innate versus discretionary earnings quality, CAPM versus 3-factor model controls for fundamental risk, different degrees of market competitiveness. They further show that the degree of dominance varies predictably when the setting varies. Bhattacharya et al. use PIN scores and the adverse selection component of the bid-ask spread (Huang and Stoll (1996)) as proxies for information asymmetry, and a targetprice-based implied cost of equity estimate (similar to the one used Brav et al. (2005)) as well as cost of capital estimates from a 4-factor model. In contrast to these studies which report evidence of an expected return effect of earnings quality, Core et al. (2007) conclude that the expected returns effect of earnings quality is too small to be significant. They argue that for an asset pricing factor to be empirically valid, its realized returns premium should be significantly positive. Using a twostage approach, Core et al. report a significant positive risk premium for the Fama–French HML factor, an insignificant risk premium for accruals quality as well as for the Fama–French SMB factor, and an insignificant or negative market risk premium.6 Their results, which provide support only for HML as a positively priced factor, are broadly 5 Chaney

et al. also find that those categories with the best overall quality of earnings attributes contain firms that use Big 5/4 auditors with the greatest frequency and that the highest quality clusters are associated with longer auditor tenure. 6 In the first stage, Core et al. regress excess returns on asset pricing factor mimicking portfolios (including the market risk premium, SMB, HML and an accruals quality factor). In the second stage, they regress mean excess returns on the coefficients from the first stage regressions. The coefficients from the second stage thus capture the estimated risk premia associated with each factor, conditional on the first stage estimates.

64

Earnings Quality and Market Outcomes

consistent with Fama and French (1997), who show that realized risk premia in both CAPM and 3-factor asset pricing models generally are insignificant.7 We interpret Core et al.’s findings to mean that research designs that fail to show significance for other asset pricing models, such as the CAPM or the 3-factor model, also fail to show significance for an earnings quality factor. Building on the definition of realized return as a composite of true expected return, cash flow news and expected return news (Campbell and Shiller (1988)), Ogneva (2007) notes that, as a proxy for expected return, realized returns will be contaminated by information shocks. She documents that poor accrual quality firms experience negative cash flow shocks in the future, which results in negative returns that offset the higher expected returns for such firms. Using (among others) the same tests as in Core et al., Ogneva finds a significant negative association between realized returns and accrual quality after controlling for cash flow shocks (either by including proxies for future cash flow shocks in the asset pricing regressions or by using an accrual quality measure that is less correlated with future cash flow shocks). Ogneva thus provides additional evidence of the importance of innate factors for earnings quality; her results, in particular, indicate the importance of this issue for research designs that use realized returns. In summary, most empirical studies conclude that earnings quality is significantly associated with expected return. The earnings quality effect is not stronger than the effect of other expected returns determinants in asset pricing models (nor would we expect it to be). The weight of the existing empirical evidence also suggests that the bulk of the expected returns effect of earnings quality appears to be associated with the portion of earnings quality that is associated with innate determinants; the earnings quality effects associated with the discretionary or reporting determinants discussed in Section 4.2 are less important.

7 Similarly,

Aboody et al. (2005), discussed in Section 5.3, conclude that while the realized return factor premium for earnings quality is large is magnitude (about 1% per month for their sample), it is also very noisy, with the result that its significance is only marginal.

5.3 Earnings Quality and Abnormal Return

5.3

65

Earnings Quality and Abnormal Return

As discussed in Section 5.1, the link between earnings quality and expected returns is linear (the poorer the earnings quality, the higher the expected return), while the link between earnings quality and abnormal return is U-shaped (poor earnings quality is associated with mispricing that yields both negative abnormal returns and positive abnormal returns). This distinction affects research designs used to investigate potential mispricing attributable to earnings quality, as well research designs that use realized returns to proxy for expected return. In the case of the latter, the abnormal return component of total return will act as a confounding variable, with the magnitude of the confound increasing with deteriorating earnings quality. Thus, when investigating the expected returns consequences of poor earnings quality, realized returns will be at best a noisy proxy variable for expected returns. And, depending on the sample, realized returns can also be a biased proxy depending on the relation between earnings quality and the abnormal returns component of the realized return. Francis et al. (2007a) investigate earnings quality and abnormal returns to post earnings announcement drift strategies, and, as an extension, earnings quality and abnormal returns to other accountingbased trading signals. Francis et al. begin with the assumption that investors are rational Bayesian information processors: when confronted with new information, such as earnings announcements, investors revise their beliefs more (less) strongly for more (less) precise earnings signals. Their results show that investors’ reaction to earnings announcements, as indicated by earnings response coefficients to earnings surprises, is more muted the poorer is earnings quality, measured either as accruals quality or as absolute abnormal accruals from a modified Jones (1991) model. Francis et al. also find that post earnings announcement drift portfolios that yield abnormal returns over the six months following the earnings announcement (i.e., the portfolios based on large positive or negative earnings surprises) are characterized by significantly poorer earnings quality, on average, than the intermediate portfolios which do not yield abnormal returns. This finding exists for both positive and negative earnings surprises (followed by positive and negative

66

Earnings Quality and Market Outcomes

abnormal returns, respectively). Further, within the extreme post earnings announcement drift portfolios (i.e., those with extreme earnings surprises) the portfolio abnormal return is driven by firms with poor earnings quality, defined as the poorest quintile in the ranked distribution of firms. Francis et al. interpret these results as consistent with the view that investors update their prior beliefs less when the initial earnings signal is noisy (i.e., the signal emanates from a firm with poor earnings quality). Subsequent abnormal returns are generated as the share price adjusts gradually as more information becomes available. The same basic results hold for other trading signals that involve accounting numbers, such as the so-called value-glamor strategies that build on accounting fundamentals’ relation to stock price and the total accruals strategy (Sloan (1996)). Specifically, regardless of the initial trading signal, firms with subsequent positive or negative abnormal returns have significantly poorer earnings quality than firms that do not exhibit abnormal returns. Further, Francis et al. find that it does not matter much for the measurement of abnormal returns whether one uses the CAPM, or the 3-factor model, or whether one adds an earnings quality mimicking factor to control for the expected returns effects of earnings quality. We believe that Francis et al.’s results have practical implications for the development of trading strategies. The magnitude of the abnormal returns effects documented for earnings quality indicates that applying an earnings quality screen plus the initial trading signal can substantially increase the profitability of accounting based trading strategies, such as post earnings announcement drift. Brav and Heaton (2002) point out that it is difficult, if not impossible, to disentangle a Bayesian updating theory explaining abnormal returns from a theory built on irrational investor behavior, because the empirical predictions from both theories are similar. Francis et al. acknowledge this fact, but attempt to shed some light on the issue by investigating the link between earnings quality and idiosyncratic stock return volatility,8 which has been used as a proxy variable for “limits to 8 As

described in more detail in Section 5.3, Rajgopal and Venkatachalam (2006) find that earnings quality is a partial determinant of idiosyncratic returns volatility.

5.3 Earnings Quality and Abnormal Return

67

arbitrage” (e.g., Pontiff (1996, 2005), Wurgler and Zhuravskaya (2002), Ali et al. (2003), Mendenhall (2004), Mashruwala et al. (2006)). Limits to arbitrage is the common explanation in the behavioral finance literature for why irrational investor behavior can cause long periods of abnormal return (Shleifer and Vishny (1997)). Francis et al. confirm that idiosyncratic volatility has explanatory power for abnormal return, but the effect is substantially smaller than that of earnings quality. When then idiosyncratic volatility is orthogonalized with respect to earnings quality (i.e., when shared variation with earnings quality is removed), idiosyncratic volatility no longer has any effect on abnormal return. In contrast, when earnings quality is orthogonalized with respect to idiosyncratic volatility, the earnings quality effect on abnormal return remains significant and economically substantial, although diminished in magnitude. On balance, we believe these results suggest that earnings quality is more important than idiosyncratic returns volatility for explaining abnormal returns, which would point more toward a Bayesian explanation than an irrational investor behavior explanation for the documented abnormal returns to accounting-based trading signals. Two other studies that relate earnings quality to abnormal trading profitability are Aboody et al. (2005) and Ecker (2006). Aboody et al. argue that earnings quality can proxy for asymmetric information, and poor earnings quality should thus be exploitable by insiders. They find that insiders trade more profitably in firms that have high exposure to an earnings quality factor mimicking portfolio. As a separate but related issue, they conclude that the realized return factor premium for earnings quality is large (around 1% per month over their sample period) but very noisy, resulting in only marginal statistical significance. This result is consistent with conclusions of other studies: poor earnings quality causes abnormal returns which can be profitably exploited. At the same time, earnings quality creates substantial noise in realized returns, so when realized returns are used to proxy for expected returns effects in a factor-mimicking portfolio, this portfolio exhibits volatile realized returns. Ecker (2006) focuses on long-term abnormal returns following initial public offerings (IPOs), extending Loughran and Ritter (1995). Ecker

68

Earnings Quality and Market Outcomes

extends, to the earnings quality setting, Kim and Ritter’s (1999) finding that IPO firms are generally initially valued on industry-matched earnings multiples such as P/E ratios. Ecker argues that initially relying on the industry average P/E ratio to price IPO firms makes sense, because these firms lack the historical data that would allow investors to estimate firm-specific parameters that affect valuations, especially information risk parameters. Using the industry average as the expected information risk parameter for an IPO firm, Ecker shows that the cross-section of post-IPO long-run abnormal returns is explained by the difference between realized and expected information risk. Initial underestimation (overestimation) of information risk results in a negative (positive) abnormal return over the period investors update their beliefs. Ecker further finds that positive abnormal returns on betterinformation firms are less persistent than negative abnormal returns on worse-information firms, consistent with Bayesian investors being able to gradually assess the information risk of newly listed firms: the better the information quality, the quicker is the adjustment. Two other studies (Jiang et al. (2005) and Zhang (2006)) also espouse the view that variables related to information quality can be used as conditioning factors in trading strategies, although neither study focuses directly on earnings quality directly or on accountingbased trading signals. Both studies investigate how the profitability of trading strategies based on analyst forecast revisions and returns momentum are influenced by variables such as firm size and age, returns volatility and analyst forecast dispersion. Given the documented association between these variables and information uncertainty, these studies speak to an indirect relation between earnings quality and abnormal returns. Both studies find effects; however, in contrast to Francis et al. (2007a) and Ecker (2006), the authors interpret their findings as consistent with behavioral biases among investors.

5.4

Earnings Quality and Other Market Outcomes

Research has also investigated the relations between earnings quality and market outcomes other than expected and abnormal returns. Rajgopal and Venkatachalam (2006) find that earnings quality, measured

5.4 Earnings Quality and Other Market Outcomes

69

as accruals quality or absolute abnormal accruals, has deteriorated over the last 40 years, and that idiosyncratic return volatility has increased over the same period. They conduct both cross-sectional and time-series tests, and conclude that there is a strong correlation between earnings quality and idiosyncratic returns volatility. Other studies report a relation between earnings quality and market microstructure measures related to information asymmetry, as part of the descriptive data. Ecker et al. (2006) report that the correlation between a returns-based measure of earnings quality, the “e-loadings” (described in Section 4), and probability of informed trading (PIN, Easley et al. (1996)) scores is statistically significant but modest in magnitude, 0.13 (Pearson) and 0.16 (Spearman). They further report that the Pearson (Spearman) correlation between e-loadings and average bid-ask spreads is equally modest, 0.14 (0.12). Aslan et al. (2006) report that the Spearman correlation between absolute abnormal accruals (Jones, 1991) and PIN scores is 0.19, but this association is not significant in multivariate tests. Bhattacharya et al. (2007a) investigate in detail how earnings quality, measured as accruals quality or abnormal accruals, affects information asymmetry measures. They use the effective spreads and the price impact of trades as measures of adverse selection risk of less informed investors (similar to Huang and Stoll (1996)), and find that these measures increase disproportionately for firms with poor earnings quality, especially in the days surrounding the earnings announcements. The modest associations between earnings quality and long-term average measures of information asymmetry (annual PIN scores and annual average bid-ask spreads), coupled with the fact that adverse selection effects of earnings quality spike around earnings announcements, indicate that the returns effects of earnings quality discussed in Sections 5.2 and 5.3 may be only moderately related to the longterm returns effects associated with PIN scores (Easley et al. (2002), Aslan et al. (2006)). It appears, however, that earnings quality has an effect around earnings announcements (Ecker et al. (2006), Francis et al. (2007a)), which is associated with information asymmetry, and with its (partial) resolution when firms release financial (and other) information at earnings announcements dates.

6 Concluding Remarks

Our aim in this review is to describe and expand upon research that examines the capital market implications of earnings quality. We begin by describing why we characterize earnings quality by reference to the precision of information. The idea that the precision of financial reporting information might affect capital market participants’ perceptions of that information has a long history in the accounting literature. For example, researchers have studied whether investors react to the credibility of reported earnings (where credibility is proxied by auditor reputation), and whether investors distinguish the credibility or quality of reported earnings based on differences in governance or ownership structures. Relative to this literature, more recent research on earnings quality uses direct estimates of earnings quality. This research takes the view that factors examined in prior studies as intervening variables (e.g., auditor size, independence of the board of directors, or an ownership structure that does or does not separate cash flow rights from voting rights) might cause, or be responses to, earnings quality. As indicated in our discussion, we believe more research is needed to provide evidence on the factors that influence earnings quality. We have also noted that 71

72

Concluding Remarks

it is an open issue as to whether factors that influence earnings quality have their own direct capital market consequences as well. We discuss several measures used by accounting researchers as indicators of earnings quality. This discussion is by no means exhaustive. Indeed, it is hard to imagine how any single variable could adequately capture the many dimensions of earnings quality. Moreover, we discuss only measures of earnings quality derived from financial information; we do not, for example, address measures of disclosure quality (such as might be captured by analyst ratings or researcher-constructed disclosure scores). Finally, we describe research that addresses whether earnings quality has diversifiable or undiversifiable pricing effects, and related research design issues. Although this question continues to be controversial, we believe that the weight of the empirical evidence points to a statistically significant and economically meaningful association between earnings quality and both expected returns and abnormal returns. We interpret this empirical evidence as supporting the perspective that the precision of financial reporting information (as captured by various measures of earnings quality) affects capital allocation decisions.

Acknowledgments

This research was supported by the Fuqua School of Business, Duke University. We thank Stephen Penman and an anonymous referee for helpful comments.

73

References

Aboody, D., J. Hughes, and J. Liu (2005), ‘Earnings quality, insider trading, and cost of capital’. Journal of Accounting Research 43, 651–673. Ali, A., L.-S. Hwang, and M. Trombley (2003), ‘Arbitrage risk and the book-to-market anomaly’. Journal of Financial Economics 69, 355–373. Amihud, Y. and H. Mendelson (1986), ‘Asset pricing and the bid-ask spread’. Journal of Financial Economics 17, 233–249. Aslan, H., D. Easley, S. Hvidkjær, and M. O’Hara (2006), ‘The determinants of informed trading: Implications for asset pricing’. Working paper, University of Houston, Cornell University and University of Maryland. Ball, R., S. P. Kothari, and A. Robin (2000), ‘The effect of international institutional factors on properties of accounting earnings’. Journal of Accounting and Economics 29, 1–51. Barone, G. (2003), ‘Perceptions of earnings quality and their association with the cost of equity capital’. Working paper, University of Texas.

75

76

References

Barth, M., W. Beaver, and W. Landsman (2001), ‘The relevance of the value relevance literature for financial accounting standard setting: Another view’. Journal of Accounting and Economics 31, 77–104. Barth, M., Y. Konchitchki, and W. Landsman (2006), ‘Cost of capital and financial statement transparency’. Working paper, Stanford University, University of Southern California and University of North Carolina. Bartov, E., D. Givoly, and C. Hayn (2002), ‘The rewards to meeting or beating earnings expectations’. Journal of Accounting and Economics 33, 173–204. Bartov, E. and P. Mohanram (2004), ‘Private information, earnings manipulations, and executive stock-option exercise’. The Accounting Review 79, 889–920. Basu, S. (1997), ‘The conservatism principle and the asymmetric timeliness of earnings’. Journal of Accounting and Economics 24, 3–37. Beasley, M. S. (1996), ‘An empirical analysis of the relation between the board of director composition and financial statement fraud’. The Accounting Review 71, 443–465. Becker, C., M. DeFond, J. Jiambalvo, and K. R. Subramanyam (1998), ‘The effects of audit quality on earnings management’. Auditing: A Journal of Practice and Theory 15, 1–24. Bedard, J. and S. F. Biggs (1991), ‘Pattern recognition, hypotheses generation, and auditor performance in an analytical task’. The Accounting Review 66, 622–642. Berger, P., H. Chen, and F. Li (2006), ‘Firm specific information and the cost of equity capital’. Working paper, University of Chicago, University of British Columbia and University of Michigan. Bhattacharya, N., H. Desai, and K. Venkataraman (2007a), ‘Earnings quality and information asymmetry: Evidence from trading costs’. Working paper, Southern Methodist University. Bhattacharya, N., F. Ecker, P. Olsson, and K. Schipper (2007b), ‘Direct and mediated associations among earnings quality, information asymmetry and the cost of equity’. Working paper, Southern Methodist University and Duke University. Bhattacharya, U., H. Daouk, and M. Welker (2003), ‘The world pricing of earnings opacity’. The Accounting Review 78, 641–678.

References

77

Black, F. (1972), ‘Capital market equilibrium with restricted borrowing’. Journal of Business 45, 444–455. Bonner, S. and B. Lewis (1990), ‘Determinants of auditor expertise’. Journal of Accounting Research 28(Supplement), 1–2. Botosan, C. (1997), ‘Disclosure level and the cost of equity capital’. The Accounting Review 72, 323–349. Botosan, C. and M. Plumlee (2005), ‘Assessing alternative proxies for the expected risk premium’. The Accounting Review 81, 21–53. Brav, A. and J. B. Heaton (2002), ‘Competing theories of financial anomalies’. Review of Financial Studies 15, 575–606. Brav, A., R. Lehavy, and R. Michaely (2005), ‘Using expectations to test asset pricing models’. Financial Management 34, 31–64. Bushman, R., Q. Chen, E. Engel, and A. Smith (2004), ‘Financial accounting information, organizational complexity and corporate governance systems’. Journal of Accounting and Economics 37, 167–201. Campbell, J. and R. Shiller (1988), ‘The dividend-price ratio and expectations of future dividends and discount factors’. Review of Financial Studies 1, 195–228. Chaney, P., B. Cooil, and D. Jeter (2007), ‘A latent class model of earnings attributes’. Working paper, Vanderbilt University. Chen, L., D. Dhaliwal, and M. Trombley (2007a), ‘The effect of fundamental risk on the market pricing of accruals quality’. Working paper, University of Arizona. Chen, S., T. Shevlin, and Y. Tong (2007b), ‘What is the information content of dividend changes? A new investigation of an old puzzle’. Journal of Accounting Research 45, 1–40. Claus, J. and J. Thomas (2001), ‘Equity premia as low as three percent? Evidence from analysts’ earnings forecasts for domestic and international stock markets’. Journal of Finance 56, 1629–1666. Cohen, D. (2006), ‘Does information risk really matter? An analysis of the determinants and economic consequences of financial reporting quality’. Working paper, New York University. Cohen, D., A. Dey, and T. Lys (2005), ‘The Sarbanes Oxley Act of 2002: Implications for compensation structure and risk-taking incentives of

78

References

CEOs’. Working paper, New York University, University of Chicago and Northwestern University. Collins, D. and S. P. Kothari (1989), ‘An analysis of intertemporal and cross-sectional determinants of earnings response coefficients’. Journal of Accounting and Economics 11, 143–181. Collins, D., E. Maydew, and I. Weiss (1997), ‘Changes in the valuerelevance of earnings and book values over the past forty years’. Journal of Accounting and Economics 24, 39–67. Core, J., W. Guay, and R. Verdi (2007), ‘Is accruals quality a priced risk factor?’. Journal of Accounting and Economics (forthcoming). Dechow, P. and I. Dichev (2002), ‘The quality of accruals and earnings: The role of accrual estimation errors’. The Accounting Review 77(Supplement), 35–59. Dechow, P. and C. Schrand (2004), Earnings quality. CFA Institute. Dechow, P. and D. Skinner (2000), ‘Earnings management: Reconciling the views of accounting academics, practitioners, and regulators’. Accounting Horizons 14, 235–250. Dechow, P., R. Sloan, and A. Sweeney (1996), ‘Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC’. Contemporary Accounting Research 13, 1–36. Diamond, D. and R. Verrecchia (1991), ‘Disclosure, liquidity, and the cost of capital’. Journal of Finance 46, 1325–1359. Easley, D., S. Hvidkjær, and M. O’Hara (2002), ‘Is information risk a determinant of asset returns?’. Journal of Finance 57, 2185–2221. Easley, D., N. Kiefer, and M. O’Hara (1996), ‘Cream-skimming or profit-sharing? The curious role of purchased order flow’. Journal of Finance 51, 811–833. Easley, D. and M. O’Hara (2004), ‘Information and the cost of capital’. Journal of Finance 59, 1553–1583. Easton, P. (2004), ‘PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital’. The Accounting Review 79, 73–95. Easton, P. and M. Zmijewski (1989), ‘Cross-sectional variation in the stock market response to accounting earnings announcements’. Journal of Accounting and Economics 11, 117–141.

References

79

Ecker, F. (2006), ‘Information risk and long-run performance of initial public offerings’. Working paper, Duke University. Ecker, F., J. Francis, I. Kim, P. Olsson, and K. Schipper (2006), ‘A returns-based representation of earnings quality’. The Accounting Review 81, 749–780. Ecker, F., J. Francis, P. Olsson, and K. Schipper (2005), ‘Comparing total and current accruals quality’. Working paper, Duke University. Fama, E. and K. French (1993), ‘Common risk factors in the returns on stocks and bonds’. Journal of Financial Economics 33, 3–56. Fama, E. and K. French (1997), ‘Industry costs of equity’. Journal of Financial Economics 93, 153–194. Fan, J. and T. J. Wong (2002), ‘Corporate ownership structure and the informativeness of accounting earnings in East Asia’. Journal of Accounting and Economics 33, 401–425. Francis, J., A. Huang, S. Rajgopal, and A. Zang (2008a), ‘CEO reputation and reporting quality’. Contemporary Accounting Research (forthcoming). Francis, J., R. LaFond, P. Olsson, and K. Schipper (2004), ‘Costs of equity and earnings attributes’. The Accounting Review 79, 967– 1010. Francis, J., R. LaFond, P. Olsson, and K. Schipper (2005a), ‘The market pricing of accruals quality’. Journal of Accounting and Economics 39, 295–327. Francis, J., R. LaFond, P. Olsson, and K. Schipper (2007a), ‘Information uncertainty and post-earnings-announcement-drift’. Journal of Business, Finance and Accounting 34, 403–433. Francis, J., D. J. Nanda, and P. Olsson (2008b), ‘Voluntary disclosure, earnings quality, and cost of capital’. Journal of Accounting Research 46, 53–99. Francis, J., P. Olsson, and K. Schipper (2007b), ‘Call options and accruals quality’. Working paper, Duke University. Francis, J. and K. Schipper (1999), ‘Have financial statements lost their relevance?’. Journal of Accounting Research 37, 319–352. Francis, J., K. Schipper, and L. Vincent (2002), ‘Earnings announcements and competing information’. Journal of Accounting and Economics 33, 313–342.

80

References

Francis, J., K. Schipper, and L. Vincent (2005b), ‘Earnings and dividend informativeness when cash flow rights are separated from voting rights’. Journal of Accounting and Economics 39, 329–360. Frankel, R., M. Johnson, and K. Nelson (2002), ‘The relation between auditors’ fees for nonaudit services and earnings management’. The Accounting Review 77(Supplement), 71–105. Gebhardt, W., C. Lee, and B. Swaminathan (2001), ‘Toward an implied cost of capital’. Journal of Accounting Research 39, 135–176. Gode, D. and P. Mohanram (2003), ‘Inferring the cost of equity using the Ohlson-J¨ uttner model’. Review of Accounting Studies 8, 399–431. Gray, P., P.-S. Koh, and Y. Tong (2007), ‘Accruals quality, information risk and cost of capital: Evidence from Australia’. Working paper, University of Queensland, Hong Kong University of Science and Technology and Nanyang Technological University. Guay, W. (1999), ‘An empirical analysis of the convexity between stock price and CEO’s wealth’. Journal of Financial Economics 53, 43–71. Guay, W., S. P. Kothari, and R. Watts (1996), ‘A market-based evaluation of discretionary accrual models’. Journal of Accounting Research 34(Supplement), 83–105. Gul, F. and L. K. Wah (2002), ‘Insider entrenchment, board leadership structure and informativeness of earnings’. Working paper, City University of Hong Kong. Hanlon, M., T. Shevlin, and S. Rajgopal (2003), ‘Are executive stock options associated with future earnings’. Journal of Accounting and Economics 36, 3–43. Healy, P. (1996), ‘Discussion of “A market-based evaluation of discretionary accrual models” ’. Journal of Accounting Research 34(Supplement), 107–115. Healy, P. and J. Wahlen (1999), ‘A review of the earnings management literature and its implications for standard setting’. Accounting Horizons 13, 365–383. Holthausen, R. and R. Verrecchia (1988), ‘The effects of sequential information releases on the variance of price changes in an inter-temporal asst multi-asset market’. Journal of Accounting Research 26, 82–106.

References

81

Holthausen, R. and R. Watts (2001), ‘The relevance of the value relevance literature for financial accounting standard setting’. Journal of Accounting and Economics 31, 3–75. Huang, R. and H. Stoll (1996), ‘Dealer versus auction markets: A paired comparison of execution costs on NASDAQ and NYSE’. Journal of Financial Economics 41, 313–357. Huang, R. and H. Stoll (1997), ‘The components of the bid-ask spread: A general approach’. Review of Financial Studies 10, 995–1034. Hughes, J., J. Liu, and J. Liu (2007), ‘Information asymmetry, diversification and cost of capital’. The Accounting Review 82, 705–729. Hunt, A., S. Moyer, and T. Shevlin (2000), ‘Earnings volatility, earnings management and equity value’. Working paper, University of Washington. Indjejikian, R. (2007), ‘Discussion of “Accounting information, disclosure, and the cost of capital” ’. Journal of Accounting Research 45(Supplement), 421–426. Jiang, G., C. Lee, and Y. Zhang (2005), ‘Information uncertainty and expected returns’. Review of Accounting Studies 10, 185–221. Johnson, V., I. Khurana, and J. Reynolds (2002), ‘Audit tenure and the quality of financial reports’. Contemporary Accounting Research 19, 637–660. Jones, J. (1991), ‘Earnings management during import relief investigations’. Journal of Accounting Research 29, 193–228. Kim, M. and J. Ritter (1999), ‘Valuing IPOs’. Journal of Financial Economics 53, 409–437. Klein, A. (2002), ‘Audit committee, board of director characteristics, and earnings management’. Journal of Accounting and Economics 33, 375–400. Kothari, S. P., A. Leone, and C. Wasley (2005), ‘Performance matched discretionary accrual measures’. Journal of Accounting and Economics 39, 163–197. Krishnan, G. (2003), ‘Does Big 6 auditor industry expertise constrain earnings management?’. Accounting Horizons 17(Supplement), 1–16. Krishnan, G., B. Srinidhi, and L. Su (2008), ‘Inventory policy, accruals quality and information risk’. Review of Accounting Studies (forthcoming).

82

References

Lambert, R., C. Leuz, and R. Verrecchia (2007a), ‘Accounting information, disclosure, and the cost of capital’. Journal of Accounting Research 45(Supplement), 385–420. Lambert, R., C. Leuz, and R. Verrecchia (2007b), ‘Information asymmetry, information precision, and the cost of capital’. Working paper, University of Pennsylvania and University of Chicago. Lang, M. and R. Lundholm (1996), ‘Corporate disclosure policy and analyst behavior’. The Accounting Review 71, 467–492. Larcker, D. and S. Richardson (2004), ‘Fees paid to audit firms, accrual choices and corporate governance’. Journal of Accounting Research 42, 625–658. Leuz, C., D. J. Nanda, and P. Wysocki (2003), ‘Earnings management and investor protection: An international comparison’. Journal of Financial Economics 69, 505–527. Lev, B. (1989), ‘On the usefulness of earnings and earnings research: Lessons and directions from two decades of empirical research’. Journal of Accounting Research 27(Supplement), 153–192. Lev, B. and R. Thiagarajan (1993), ‘Fundamental information analysis’. Journal of Accounting Research 31, 190–215. Lintner, J. (1965), ‘The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets’. Review of Economics and Statistics 47, 13–37. Lipe, R. (1990), ‘The relation between stock returns and accounting earnings given alternative information’. The Accounting Review 65, 49–71. Liu, M. and P. Wysocki (2007), ‘Cross-sectional determinants of information quality proxies and cost of capital measures’. Working paper, Penn State and MIT (version: July 2007). Loughran, T. and J. Ritter (1995), ‘The new issues puzzle’. Journal of Finance 50, 23–51. Mashruwala, C., S. Rajgopal, and T. Shevlin (2006). ‘Why is the accrual anomaly not arbitraged away?’ Journal of Accounting and Economics 42, 3–33. McNichols, M. (2000), ‘Research design issues in earnings management studies’. Journal of Accounting and Public Policy 19, 313–345.

References

83

McNichols, M. (2002), ‘Discussion of ‘The quality of accruals and earnings: The role of accrual estimation errors”. The Accounting Review 77(Supplement), 61–69. Mendenhall, R. (2004), ‘Arbitrage risk and post-earnings announcement drift’. Journal of Business 77, 875–894. Merton, R. (1987), ‘A simple model of capital market equilibrium with incomplete information’. Journal of Finance 42, 483–510. Morck, R. and B. Yeung (2002), ‘The puzzle of the harmonious stock prices’. World Economics Journal 3, 105–119. Morck, R., B. Yeung, and W. Yu (2000). ‘The information content of stock markets: Why do emerging markets have synchronous stock price movements?’ Journal of Financial Economics 58, 215–260. Myers, J., L. Myers, and T. Omer (2003). ‘Exploring the term of the auditor-client relationship and the quality of earnings: A case for mandatory auditor rotation?’ The Accounting Review 78, 779–799. Nichols, C. (2006), ‘Fundamental or information risk? An analysis of the residual accrual volatility factor’. Working paper, Cornell University. Ogneva, M. (2007), ‘Accrual quality and expected returns: The importance of controlling for cash flow shocks’. Working paper, University of Southern California. Ohlson, J. and B. J¨ uttner-Nauroth (2005), ‘Expected EPS and EPS growth as determinants of value’. Review of Accounting Studies 10, 349–365. Palmrose, Z., V. J. Richardson, and S. Scholz (2004), ‘Determinants of market reactions to restatement announcements’. Journal Accounting and Economics 37, 59–89. Peasnell, K., P. Pope, and S. Young (2005). ‘Board monitoring and earnings management: Do outside directors influence abnormal accruals?’ Journal of Business Finance and Accounting 32, 1311– 1346. Penman, S. and X.-J. Zhang (2002), ‘Accounting conservatism, the quality of earnings and stock returns’. The Accounting Review 77, 237–264. Pontiff, J. (1996), ‘Costly arbitrage: Evidence from closed-end funds’. The Quarterly Journal of Economics 111, 1135–1151.

84

References

Pontiff, J. (2005), ‘Costly arbitrage and the myth of idiosyncratic risk’. Journal of Accounting and Economics 42, 35–52. Rajgopal, S. and T. Shevlin (2002), ‘Empirical evidence on the relation between stock option compensation and risk taking’. Journal of Accounting and Economics 33, 145–171. Rajgopal, S. and M. Venkatachalam (2006), ‘Information risk and idiosyncratic return volatility over the last four decades’. Working paper, University of Washington and Duke University. Schipper, K. and L. Vincent (2003), ‘Earnings quality’. Accounting Horizons 17(Supplement), 97–110. Sharpe, W. (1964), ‘Capital asset prices: A theory of market equilibrium under conditions of risk’. Journal of Finance 19, 425–442. Shleifer, A. and R. Vishny (1997), ‘The limits of arbitrage’. Journal of Finance 52, 35–55. Sloan, R. (1996). ‘Do stock prices fully reflect information in accruals and cash flows about future earnings?’ The Accounting Review 71, 289–315. Subramanyam, K. R. (1996), ‘The pricing of discretionary accruals’. Journal of Accounting and Economics 22, 249–281. Teoh, S. and T. J. Wong (1993), ‘Perceived auditor quality and the earnings response coefficient’. The Accounting Review 68, 346–366. Tufano, P. (1996), ‘Who manages risk? An empirical examination of risk management practices in the gold mining industry’. Journal of Finance 51, 1097–1137. Verrecchia, R. (1990), ‘Information quality and discretionary disclosure’. Journal of Accounting and Economics 12, 365–380. Warfield, T., J. Wild, and K. Wild (1995), ‘Managerial ownership, accounting choices and informativeness of earnings’. Journal of Accounting and Economics 20, 61–91. Watts, R. (2003), ‘Conservatism in accounting, Part 1: Explanations and implications’. Accounting Horizons 17, 207–221. Wright, A. and S. Wright (1997), ‘An examination of factors affecting the decision to waive audit adjustments’. Journal of Accounting, Auditing and Finance 12, 15–36. Wurgler, J. and E. V. Zhuravskaya (2002). ‘Does arbitrage flatten demand curves for stocks?’ Journal of Business 75, 583–608.

References

85

Xie, H. (2001), ‘The mispricing of abnormal accruals’. The Accounting Review 76, 357–373. Yee, K. (2006), ‘Earnings quality and the equity risk premium: A benchmark model’. Contemporary Accounting Research 23, 833–877. Yeo, G., P. Tan, K. Ho, and S. Chen (2002), ‘Corporate ownership structure and the informativeness of earnings’. Journal of Business, Finance and Accounting 29, 1023–1046. Zhang, F. (2006), ‘Information uncertainty and stock returns’. Journal of Finance 61, 105–136.