ECONOMICS OF THE FAMILY AND FAMILY POLICIES
Economic analysis of the family is a recent, but already well-established area in economics. Economists have developed new methods to analyze decisions such as marriage, childbearing, divorce and the use of family resources. The economic approach has become crucial for design and understanding in many policy areas of current interest in modern societies, such as family, tax, social security and gender equality policies. Economics of the Family and Family Policies bears evidence to the lively and relevant research in the area. The first part provides a clear picture of the state of the art of economics of the family as it relates to economic theory and economic modeling, examining the developments from common preference family models to the more recent cooperative or non-cooperative bargaining models. The second part explores theoretical and empirical applications: the effect of the intrafamily distribution of income on family decisions; the interaction between marriage markets and labour markets; and the factors behind the rise of single-parent families. The final part of the book focuses on family policies and analyzes tax, public childcare and parental leave policies in terms of the incentives they create for labor supply, time allocation, human capital accumulation and the choice of how to organize childcare. The empirical studies in this section are mainly drawn from the Nordic countries, renowned for their experience with family and gender equality policies. This volume will be an invaluable, up-to-date resource for economists and those involved in the social sciences and gender studies, as well as policy-makers themselves. Inga Persson is Professor of Economics at Lund University, Sweden. She holds a chair in the Economics of Gender and has published on labor market policy, unemployment, the welfare state and the economic position of women. Her publications include Generating Equality in the Welfare State: The Swedish Experience (1990). Christina Jonung is a University Lecturer in Economics at Lund University. Her research has covered the economic position of women in Sweden and gender equality policies.
ROUTLEDGE RESEARCH IN GENDER AND SOCIETY 1. ECONOMICS OF THE FAMILY AND FAMILY POLICIES Edited by Inga Persson and Christina Jonung
ECONOMICS OF THE FAMILY AND FAMILY POLICIES Edited by Inga Persson and Christina Jonung A selection of papers from the 15th Arne Ryde Symposium on “Economics of Gender and the Family,” in honor of Anna Bugge and Knut Wicksell
London and New York
First published 1997 by Routledge 11 New Fetter Lane, London EC4P 4EE This edition published in the Taylor & Francis e-Library, 2005. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 © 1997 Inga Persson and Christina Jonung All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Arne Ryde Symposium (15th:1995) Economics of the family and family policies: a selection of papers from the 15th Arne Ryde Symposium…/edited by Inga Persson and Christina Jonung p. cm. —(Routledge research in gender and society) Includes bibliographical references and index. 1. Family—Economic aspects—Congresses. 2. Family policy—Economic aspects—Congresses. 3. Family policy—Economic aspects—Scandinavia— Congresses. I. Persson, Inga. II. Jonung, Christina, 1945– III. Title. IV. Series. HQ518.A76 1997 306.85–dc21 97–10253 ISBN 0-203-44133-8 Master e-book ISBN
ISBN 0-203-74957-X (Adobe eReader Format) ISBN 0-415-14902-9 (Print Edition)
CONTENTS
List of figures
vii
List of tables
viii
List of contributors Preface
x xii
ANNA AND KNUT Christina Jonung and Inga Persson
2
INTRODUCTION Inga Perssonand and Christina Jonung
8
Part I Where are we in the economics of the family? 1
BARGAINING AND DISTRIBUTION IN MARRIAGE Shelly Lundberg and Robert A.Pollak
17
2
BATTLES OF THE SEXES: NON-COOPERATIVE GAMES IN THE THEORY OF THE FAMILY Kjell Erik Lommerud
33
Part II Marriage and family formation 3
INTRAHOUSEHOLD DISTRIBUTION OF RESOURCES AND LABOR MARKET PARTICIPATION DECISIONS Daniela Del Boca
49
4
A FAMILY WITH ONE DOMINATING SPOUSE Kristian Bolin
65
5
WOMEN’S HOURS OF WORK AND MARRIAGE MARKET IMBALANCES Shoshana Grossbard-Shechtman and Matthew Neideffer
77
6
PREMARITAL COHABITATION, CHILDBEARING AND THE CREATION OF ONE-PARENT FAMILIES John Ermisch
92
Part III Family policies and household allocation of time 7
CHILDCARE, HUMAN CAPITAL AND ECONOMIC EFFICIENCY Siv S.Gustafsson and Frank P.Stafford
107
vi
8
THE CHOICE BETWEEN FULL-TIME AND PART-TIME WORK FOR NORWEGIAN AND SWEDISH MOTHERS Marit Rønsen and Marianne Sundström
123
9
PUBLIC POLICY AND CHILDCARE CHOICE Seija Ilmakunnas
139
TAXATION AND THE MARKET FOR DOMESTIC SERVICES Anne-Marie Pålsson
153
Index
171
10
FIGURES
Anna and Knut The Bugge-Wicksell marriage contract 1.1 The Nash bargaining solution 5.1 Markets for (a) Female spousal labor; (b) Male spousal labor; (c) Female labor; (d) Male labor. Primary effects of an increase in the relative number of men 6.1 First partnership rates, 1950–62 and post-1962 cohorts 6.2 Comparison of first partnership rates for two broad cohorts 6.3 Destinations of never-married women entering cohabitation, per 1000, 1950–62 cohorts 6.4 Destinations of never-married women entering cohabitation, per 1000, post-1962 cohorts 6.5 Destinations of women having children in first cohabitations, per 1000, all cohorts 7.1 Intersibling equity tax 7.2 Child development and the distribution of well-being over children and parents 10.1 The market for domestic labor (services) 10.2 Impact on GDP by allowing tax deductions
1 4 22 82 94 94 99 99 100 114 115 163 168
TABLES
3.1 3.2 3.3a 3.3b 3.4a
Descriptive statistics Effect of total family non-labor income on labor market participation Effect of male and female non-labor income on labor market participation Tests for the equality of income effects Effect of male and female non-labor income on labor market participation. Families with children under six 3.4b Tests for the equality of income effects 3.5 Maximum likelihood estimates. Bivariate probit. Effect of male and female non-labor income on labor market participation 5.1 Means and standard deviations 5.2 Regressions of labor force participation, predicted wages and hours of work, married white women ages 25–29, US census, 1990 6.1 Observed first partnerships of women, BHPS 6.2 Observed and simulated first “destinations” of women, BHPS 6.3 Cox model parameter estimates 6.4 Observed and simulated outcomes for never-married women in first cohabiting unions, BHPS 6.5 Simulated outcomes for two broad cohorts (percentages) 6.6 Relative risks of birth in cohabiting union 7.1 Work and childcare arrangements, 1900 and 2000 7.2 Equity/efficiency trade-offs in child development 8.1 Descriptive statistics for the variables used in the analyses 8.2 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with preferences and human capital variables 8.3 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with policy variables 8.4 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with interactions with calendar period of birth 9.1 Recipients of home care allowance (HCA) 1985–93, whole country 9.2 The childcare modes and their economic characteristics 9.3 Description of the data 9.4 The mixed model 9.5 The changes in estimated choice probabilities: a “base case” analysis 9.6 The effects of the mother’s wage and home care allowance on the choice probabilities (a “base case” analysis) 9.7 The effects of home care allowance on the choice probabilities of the whole sample 10.1a Time allocation for men, 1990–1, average hours and minutes per week 10.1b Time allocation for women, 1990–1, hours and minutes per week
55 57 58 59 60 60 61 87 88 93 95 96 97 101 102 109 110 127 132 133 135 141 142 145 147 148 148 149 154 155
ix
10.2 Time allocation in household work, men and women 20–64 years, Sweden 1990, average hours and minutes per week 10.3 The time input (number of hours) for the teacher and the carpenter and the tax income generated, SEK; four different solutions 10.4 Summary of the marginal analysis 10.5 First-order conditions for optimality
155 157 160 165
CONTRIBUTORS
Kristian Bolin has a Ph.D. in Economics from Lund University, Sweden. His main research interest is family economics and he has undertaken both theoretical and empirical work in this area. He is now working on a project which, using Swedish data, examines the effects of the risk of divorce on female labor force participation and on the probability of having children. Daniela Del Boca has a Ph.D. from the University of Wisconsin, USA. She is Associate Professor of Economics and teaches at the University of Turin, Italy, as well as at New York University, USA. Her main interests are labor economics and applied econometrics. She has published several books and articles in Italian and international journals. John Ermisch is Professor of Economic Demography at the ESRC Research Centre on Micro-social Change at the University of Essex, United Kingdom. He is a Fellow of the British Academy and former president (1989) of the European Society for Population Economics. His primary research field is in the economics of the family and household. Shoshana Grossbard-Shechtman is Professor of Economics at San Diego State University, USA, and is a specialist in the economics of marriage and the family. Her publications include On the Economics of Marriage, a Theory of Marriage, Labor and Divorce. Siv S.Gustafsson is Professor of Economics at the University of Amsterdam, The Netherlands. She specializes in comparative population and gender economics. Her publications include Separate Taxes and Married Women’s Labor Supply. A Comparison of Sweden and West Germany and Labor Force Transitions in Connection with Childbirth. A Comparison of Germany, Great Britain and Sweden. Seija Ilmakunnas is Senior Researcher at the Labour Institute for Economic Research, Helsinki, Finland. Her main research interests revolve around labor supply, household production and social security systems. She is a member of the Network of Experts on the Situation of Women in the Labor Market, European Commission. Christina Jonung is a University Lecturer at the Department of Economics at Lund University, Sweden. Her research has concerned the economic position of women in Sweden and gender equality policies. Kjell Erik Lommerud is Professor of Economics at the University of Bergen, Norway. His primary field of interest is labor economics, with emphasis on the economics of family and gender and the economics of education. He has also published articles in the areas of public economics and industrial organization. Shelly Lundberg is Professor of Economics at the University of Washington, USA, and specializes in labor and family economics. Matthew Neideffer was a graduate student in economics at San Diego State University, USA, and is currently a graduate student at Texas A&M University.
xi
Anne-Marie Pålsson is Associate Professor of Economics at Lund University, Sweden. Her primary research field is the economics of the family and household, ranging from personal finance such as wealth accumulation, risk-taking attitudes, etc. to household production, time allocation and taxes. Inga Persson is Professor of Economics at Lund University, Sweden. She holds a chair in Economics of Gender and her research interests are labor market policy, unemployment and the economic position of women. Robert A.Pollak is Hernreich Professor of Economics at Washington University in St Louis, USA. He is the author of The Theory of the Cost-of-Living Index and co-author of Demand System Specification and Estimation and From Parent to Child. Marit Rønsen is a Senior Executive Officer in the Division for Social and Demographic Research at Statistics, Oslo, Norway. Her main research interests include female labor supply and fertility, wage formation and income distribution. Frank P.Stafford is Professor of Economics at the University of Michigan, USA, and director of the Michigan Panel Study. His publications deal with time use, earnings, childcare, human capital and economic growth. His publications include Time Goods and Wellbeing and Divergence, Convergence and Gains to Trade. Marianne Sundström is Associate Professor of Economics at the Demography Unit, Stockholm University, Sweden. Her main research interest lies in female labor force participation and social policy.
PREFACE
This volume contains a selection of the papers presented at the 15th Arne Ryde Symposium on “Economics of Gender and the Family”, held on August 18–19, 1995, at Rungsted in Denmark. Another selection of the papers presented at the symposium is published in an accompanying volume entitled Women’s Work and Wages. During two hot summer days, about 100 economists met to discuss about 40 papers in this rapidly expanding area of international research. We want to thank all the participants in the symposium for their contributions to the lively discussions that took place in the various sessions. We (and the authors) are particularly grateful to the appointed discussants who by their comments and insights helped improve the papers. We also want to express our gratitude to the Arne Ryde Foundation for financing the symposium. Since 1973, the Foundation, established in memory of Arne Ryde, a promising young doctoral student in Economics at Lund University who died in a car accident, has generously supported international symposia and lectures as well as other professional activities arranged by the Department of Economics in Lund. This support has proved to be of great value for the economics profession in Sweden and, in particular, for the doctoral students and economists at the department in Lund. Professor Björn Thalberg, Chairman of the Board of the Arne Ryde Foundation, initiated the symposium and also, as a member of the organizing committee, saw it through from start to finish. We owe him many thanks for having contributed his vast experience. Other crucial members of the team behind the symposium and this volume have been Kristian Bolin, Carole Gillis, Keith Persson and Ann-Charlotte Sahlin. Kristian Bolin helped us in planning the symposium and also acted as our consultant when the intricacies of game theory etc. threatened to overwhelm us. Carole Gillis worked hard at improving the English. Keith Persson spent part of what was supposed to be his summer vacation at the copying machine producing the conference volumes. Ann-Charlotte Sahlin, in her calm and efficient manner, took care of all practical arrangements and later also of getting the manuscripts into shape for publication. The Arne Ryde Symposium on “Economics of Gender and the Family” was held in honor of Anna Bugge and Knut Wicksell. The reasons for dedicating the symposium to them are explained in our short tale of “Anna and Knut” opening this volume. In their own way they were forerunners both in gender relations and in family economics. Their life story provided a source of inspiration and gave us a sense of continuity through the generations in our work with this project. Inga Persson and Christina Jonung Lund, December 1996
xiii
ACKNOWLEDGMENTS We want to thank the Journal of Economic Perspectives for permission to reprint as Chapter 1 in this book the paper “Bargaining and Distribution in Marriage” by Shelly Lundberg and Robert A.Pollak, which appeared in Journal of Economic Perspectives, vol. 10, Fall 1996.
Anna and Knut
ANNA AND KNUT Christina Jonung and Inga Persson
The symposium on “Economics of Gender and the Family” in Rungsted, Denmark was dedicated to the world-famous Swedish economist Knut Wicksell (1851–1926) and his wife Anna Bugge (1862–1928). In this short contribution we wish to explain why we find they deserve such an acknowledgment by telling you the story of “Anna and Knut.” Among economists Knut Wicksell is best known for his theoretical work in economics which established him as Sweden’s leading economist at the turn of the century. Paul Samuelson even ranks him alongside Adam Smith, Walras and Keynes. Knut Wicksell was Professor of Economics at Lund University between 1901 and 1916. However, it is not in his capacity as a prominent economist or a professor at our university that he was honored at this symposium. Knut Wicksell is honored as a forerunner and radical in the area of gender and the family and Anna Bugge is honored for being, in her own right, “a woman before her time.” It may come as a surprise to many to hear that Knut Wicksell first became known to a wider public in Sweden through a fiercely feminist poem. At the Scandinavian student festivities at Uppsala University in 1878, he gave the traditional poetic “Address to Woman.” The poem, which in content departed radically from the traditional themes of women’s beauty and attractiveness, was widely published by Swedish newspapers, where it was highly commended, or accused of manifesting the “depraved spirit of the times.” Unfortunately the poem is only available in Swedish, but we will try to convey some of its ideas. At this point in time Knut was not yet an economist: he was studying mathematics and physics at Uppsala University. Nevertheless the poem contains several themes that could be found in the sessions at Rungsted. After a few of the usual verses acclaiming Woman’s charming and gracious nature, the poet lets his first critical thesis burst upon his listeners: “Wealth is Woman’s true ornament.” So unjustly is society constructed that without that attribute, she will never win a respected position. If she inherits nothing from her parents, she will probably become just one more poverty-stricken seamstress, working hard to buy food for the morrow and ignored by all honorable, wealthy admirers (Gårdlund 1996:39). We recognize here a clear forerunner to Becker’s theory of marriage! The next verse introduces Wicksell’s theory of gender wage differentials. He argues simply that since women have smaller appetites than men, men have wisely arranged things in such a way that the fruits of women’s labors “in just proportion also should be small.” As the sessions at the symposium illustrated, some further theories of gender wage differentials have been developed through time, but the smallness is still with us. Another verse eloquently illustrates how unequal access to human capital shapes men’s and women’s future. Eventually Wicksell ends in a quieter and more hopeful tone. He looks forward to the day when the achievement of greater influence for women in society will create a different legal and moral climate; “a
ANNA AND KNUT
3
spirit as gentle as Woman herself would inform the law.” The strong would then protect and the weak be protected. This theme is resounding among today’s essentialist feminists. Marriage, which according to Wicksell at that time often meant slavery and the subordination of women, would also become a free and tender union between equal citizens. (Gårdlund 1996:39) In the following decade Knut Wicksell became one of Sweden’s leading radical figures, through his neoMalthusian ideas and his writings and public lectures against prostitution and in support of birth control, which was illegal at the time, and his proposals of freer forms of marriage and marriage at a younger age. Knut Wicksell’s ideas were considered extremely provocative by his contemporaries. One young female student commented on his lectures in the following way: “His talk has caused a sensation, the like of which has not been seen for many years. He arouses admiration, astonishment, loathing, hate. He has stirred the passions of all” (Gårdlund 1996:58). Hjalmar Branting, who was to become Sweden’s first Social Democratic Prime Minister, said of him, “Together with Strindberg, Wicksell was a harbinger of revolution bringing the first tidings of change to the youth of the 1880s” (Gårdlund 1996:58). It was this controversial “harbinger of revolution” that 25-year-old Anna Bugge was to meet at a Scandinavian feminist meeting in Copenhagen in 1888. Knut had then reached the respectable age of 36 and had felt lonely and unhappy for many years. It should perhaps be added that Knut was a generally very mildmannered, friendly and courteous revolutionary. A year before their first personal encounter, Anna had heard Knut give a lecture in Oslo in which he had criticized marriage on the grounds that under existing law, the husband had legal authority over his wife and children. He proposed instead a common-law marriage—a cohabitation where neither party had any legal claims on the other. This was in fact what he offered Anna when they met for the second time, in the summer of 1889 in Oslo. But who was this young Norwegian woman who dared to flout all accepted social conventions and become the close associate of the infamous Knut Wicksell, free thinker, subversive and apostle of immorality?1 When they met, Anna was already a public figure in her home country, despite her young age. Along with a few female friends, she had founded a private “gymnasium” and became the fifth Norwegian girl to take her “studentexam” and thereby gain entrance to the university. In her and her friends’ view, knowledge brought with it responsibility and they set up a debating society in order to learn to use their newly won knowledge in lectures and debates. In 1885 the society became part of the Norwegian Feminist Association, which was subsequently chaired by Anna Bugge. A few years later, Anna took part in the setting-up of the Norwegian Women’s Suffrage Association. According to Anna herself: “There has never been a more fortunate group of individuals than the young people of the 1880s. This was the period when the great issues were under discussion and nothing barred our way” (Wicksell Nordqvist 1985). This belief in the future and the firm conviction that everything was possible provided that one worked for its achievement seems to have remained with Anna for the rest of her life. The idea of a common-law marriage between equal partners must have been appealing to Anna. However, her radicalism was of a more quiet kind than that favored by the more outspoken Knut. Anna preferred to avoid conflicts and to work for long-term changes. To a greater extent than Knut she undoubtedly also understood the costs that a common-law marriage would impose upon themselves as well as upon their families. However, a few weeks later Anna traveled to Paris in order to meet Knut and a week later she had moved into what became their joint home. Some time later, this event was announced in the Stockholm press under
4
ANNA AND KNUT
the heading “United.” With the help of Karl Staff (Knut’s friend and a lawyer and future liberal Prime Minister), they drew up and signed a marriage contract that established their mutual financial obligations (Wicksell Nordqvist 1985:95). The content of the contract in English translation is reproduced here. It is quite a remarkable document; and not only for those interested in contract theories of marriage. It would be seven years before Anna was to visit Norway and her family again. At the start of their life together, Knut had still not become an economist. Although he had visited various European universities to study and attend lectures given by leading economists of the day, he had not yet produced any academic publications. The question may be asked whether there would have been as many, or indeed any at all, without Anna Bugge. Anna made numerous contributions to their life together. First, she provided the stimulus, peace of mind and security that Knut had lacked for so long and without which he had been unable to undertake any systematic academic research. Second, she continually tried to steer Knut’s efforts Whereas we the undersigned, N.N. (husband’s name) and N.N. (wife’s name), have freely entered into mutual union, we do hereby pledge and covenant as follows:
(1) (2)
(3) (4) (5)
(6)
Each of us hereby pledges to provide for the other, to the extent that our incomes and assets reasonably allow, such provision continuing as long as our union prevails. In the event that I, N.N. (wife’s name), should have children during our union or, should our union cease to exist, during such time when I, N.N. (husband’s name) could be the father, we jointly pledge to make provision for the proper support and upbringing of these children, in accordance with our respective incomes and assets, until such time as they are able to provide for themselves. Our union shall exist until such time as it is renounced by either of the undersigned. The obligation in clause 2 to make provision for our children will remain in force irrespective of a cessation of our union. The mutually agreed pledge in clause 1 to provide for the other will remain in force until such time as our union is dissolved as specified in clause 3. However, in the event of illness or any other reason that prevents either of the undersigned from being able to make adequate provision for him/herself and if he (she) is without his/her own assets, the other contracting party shall be bound to make provision for the first, without regard to the dissolution of the union. This obligation will remain in force as long as the need exists and will come into force whenever the need arises. This contract has been drawn up in duplicate, both parties receiving their own contract.
Paris, July 1889 (signatures) THE BUGGE-WICKSELL MARRIAGE CONTRACT
towards academic output and away from an overriding concern with contemporary political issues. She even went as far as trying to curtail his reading of fiction and mathematics in favor of articles on economics. Third, Anna also helped him in practical matters. This was not just a question of running a household and
ANNA AND KNUT
5
managing the family economy, an area in which Knut never had much success.2 She also played an important role as secretary, linguistic advisor and, not least, in planning his academic career. After a year in Paris, the Wicksell family moved to Stockholm. Anna was now expecting a child. In order to support himself and his family, Knut devoted the next few years to public lectures, political pamphlets and journalism. Anna had a share in this work as Knut’s assistant, but she also did work of her own. She held lectures and wrote polemic pamphlets on the rights of women (the legal standing of married women in the Scandinavian countries) and on world peace (efforts to establish international arbitration)— the two areas that were to preoccupy her for the rest of her life. It was not until 1900 when Knut Wicksell became an Associate Professor at the University of Lund that he secured his first regular source of income. For more than a decade, the family had often been fairly poor, living on temporary sources of income, irregular scholarships and loans from friends and relatives. Anna had continually tried to provide Knut with the opportunity to carry out the research that would enable him to gain a university position. Finally, Anna succeeded in persuading him to take the law degree required in order to be able to lecture in economics. Her concern with Knut’s career was not just a question of their perennial shortage of money. Her overriding desire to support Knut’s academic work reflected her belief that it was through such work that Knut’s radical views would reach a larger audience. Devotion, the demand for truth and his unswerving incorruptibility were undoubtedly among the qualities that Anna found most attractive in Knut. She also learned right from the start about his “all or nothing” moral code. However, living with Knut Wicksell cannot have been an easy matter. His regard for principles would seem to have been more important than the distress that he could inflict on his family and close friends as a result of his actions. He was even willing to put the family’s future at stake by refusing to sign his professorial application to the King with the words “most humbly.” The economic pressure eased once Knut had been appointed professor. But Anna had to continue to put up with what she called her “regular dose of torment per term” from Knut, as a result of, for example, the crisis in the union between Norway and Sweden, Knut’s pro-Russian views on defense and his period of imprisonment on a charge of blasphemy. The academic community in Lund turned up their noses at the unconventional ideas and lifestyle of the Wicksell family—the unmarried professorial couple, who were often seen walking hand in hand, who set up home out in the countryside, with the husband-professor doing the shopping in the market place and carrying his books around along with the vegetables in a discarded pram, who taught their children themselves at home and who demanded that the children should not receive religious instruction once they started school. It was in Lund that Anna finally found time to devote to the fulfillment of her own ambitions. Ever since her years in Norway, Anna had dreamed of becoming a lawyer. She had started legal studies during their year in Paris but the children, Knut’s career, the problems of supporting a family and illness had all intervened to delay her studies. It was not until 1911 that Anna, at the age of 48, took her final exam for her law degree. She was to specialize in international law. During the first two decades of this century, Swedish women campaigned actively for the right to vote. Anna had started this campaign 15 years earlier in Norway and was to become one of the movement’s leading figures in Sweden as well. If Knut’s strategy for change could be described as one of provocative action, Anna’s strategy was instead one of persistent, practical work. She formed associations, organized work on behalf of women’s suffrage, held lectures, wrote articles and political pamphlets, put forward legislative proposals, electioneered for the Liberals and regularly attended international conferences as a
6
ANNA AND KNUT
delegate for women’s suffrage. Her commitment to women’s right to vote was not just a question of justice or the belief that there would be a new outlook on political life once women were able to participate fully in the political process. It was also, she argued, a matter of social responsibility, which forms an essential part of each individual citizen’s development and maturity. Using the same methods as she had applied to feminist issues, Anna involved herself in the peace movement. The solution of international conflicts by means of international law was a question which had preoccupied her for a long time. During World War I, she acted as an expert advisor to the committee that was working on a Nordic proposal for an international legal system. On the formation of the League of Nations, Anna was appointed Swedish delegate and thereby became the country’s first female diplomat. From then on and until her death in 1928, she was to spend many weeks of the year in Geneva or traveling on feminist or peace business. Her travels made life lonely for Knut and he complained endlessly about a feeling of restlessness and an inability to get on with his academic work. His last major scholarly study was the second part of Lectures on Political Economy, which he began when Anna started her legal studies. The story of Anna and Knut is excellently told in the biography of Knut Wicksell written by Torsten Gårdlund, available in English (1996) as well as Swedish (1990). There, quite naturally, the focus is on Knut. Throughout their lives, Anna and Knut wrote to each other almost every day when they were apart. They also wrote frequently to friends—long, frank letters, full of ideas and thoughts about everyday life and work. This abundant and informative correspondence formed the basis for a biography of Anna Bugge by her granddaughter Liv Wicksell Nordqvist (1985) which provides unique insights into Anna’s and Knut’s life and thoughts. Most of the social reforms which Anna and Knut worked for are now in force, e.g. the right to vote for men and women alike, a more liberal matrimonial legislation, the access to birth control, increased social and economic equality, international agencies for the promotion of peace. In spite of all of these achievements, many of the changes that they had hoped to bring about still seem very far away today, such as the abolition of prostitution and the achievement of a lasting peace. Moreover, no matter which country we come from, we may also ask ourselves to what extent we have actually attained their overriding goal for a successful society—an open climate of public debate, characterized by a search for truth and reasoned argument. For example, do dissenting opinions, in this age of “political correctness,” meet with greater tolerance at our universities and in society at large than they did around the turn of the century? We hope we have now made it clear why we feel that Anna and Knut, through their life and their work, their dreams and their ideals, have been forerunners and may serve as examples in the work and the ambitions for equality between men and women. This is why this symposium was in their honor. ACKNOWLEDGMENTS We want to thank Liv Wicksell Nordqvist who most generously provided us with the photographs of her grandparents, Anna Bugge and Knut Wicksell. REFERENCES Gårdlund, T. (1990) Knut Wicksell, Rebell i det nya riket, Stockholm: SNS förlag (first published by Bonniers, 1956). —— (1996) The Life of Knut Wicksell, Cambridge: Edward Elgar (first published by Almqvist & Wiksell, 1958). Wicksell Nordqvist, L. (1985) Anna Bugge Wicksell. En kvinna före sin tid (Anna Bugge Wicksell. A Woman before Her Time), Lund: Liber.
ANNA AND KNUT
7
NOTES 1 Wicksell Nordqvist (1985) is a fascinating biography about Anna’s life. 2 Despite being a leading expert on Sweden’s system of taxation, Knut Wicksell was not able to complete his own tax returns (Gårdlund 1996:326).
INTRODUCTION Inga Persson and Christina Jonung
Economic models of the family and of family decision-making have by now become an important and wellestablished area of research within economics. The models have enabled economists to analyze a number of issues related to the family that previously were not amenable to economic analysis, such as marriage, fertility and divorce. Economic models of the family have also turned out to be crucial tools when it comes to policy design and policy evaluation in many of the areas that are of central importance in modern societies: policies concerning the family, taxes, social security and gender equity. Concomitantly, the increased availability of cross-sectional and longitudinal data about individuals and households has greatly enriched the possibilities for empirical tests and studies of the theories. Thus, economics of the family is definitely here to stay. That it is also a very lively and dynamic research area is witnessed by the contents of this book. The book is divided into three parts. Part I consists of two chapters which describe and discuss the current state of economic modeling of the family, the reasons behind the evolution of “second-generation” models of family decision-making and the differences in the results obtained from different models. Together the two contributions establish a clear picture of “the state of the art” within economics of the family. Part II contains four chapters with theoretical and/or empirical applications. The chapters analyze family labor supply and time allocation, the role of marriage markets, and family formation. The four chapters in Part III are devoted to issues related to family policies. They treat the equity/efficiency tradeoffs that exist between parents and children and the incentive effects for households that arise from tax policies and public childcare policies. PART I WHERE ARE WE IN THE ECONOMICS OF THE FAMILY? “First-generation” economic models of the family are “common preference” models which assume that the family maximizes a single utility function. Common preference models imply that family behavior is independent of who it is in the household that receives income or controls resources. The common preference model has proved to be a powerful analytical tool, but in recent years the approach has been seriously brought into question both from a theoretical and an empirical perspective. Shelly Lundberg and Robert A.Pollak take their starting-point in the common preference model and discuss how the common preference assumption has been rationalized either by a Samuelson consensus model or by a Becker altruist model. They describe the theoretical and empirical criticisms that have been leveled against common preference models and how “second-generation” game-theoretic models of the
INTRODUCTION
9
family have been developed as alternative analytical tools. The game-theoretic models permit independent agency of men and women in marriage and differ from the common preference models in their implications for distribution within marriage and for observed family behavior. These models are thus consistent with a growing amount of empirical evidence which indicates that income controlled by husbands has different effects on family behavior than that controlled by wives. The game-theoretic models of the family first developed in the literature were cooperative bargaining models. Lundberg and Pollak review the salient features of cooperative family models and the types of results such models generate. They also highlight and discuss the importance in such models of the assumptions made about what constitutes the threat point in marital bargaining. In common preference models and cooperative bargaining models of the family, family behavior is Pareto-optimal, i.e. no family member can be made better off without making another worse off. This is not necessarily the case in non-cooperative models of the family—the second strand of game-theoretic models of the family to have been developed. Non-cooperative models open up the possibility that family behavior might be inefficient, and consequently they allow economists to examine what conditions might enable families to sustain Pareto-optimal outcomes. Lundberg and Pollak show how this, in turn, tends to bring the role of family legislation, social norms and conventions, etc. into the picture. Such norms and conventions constitute a framework within which the bargaining in a particular marriage takes place and can influence the outcome of that bargaining. The authors also show the important role played by the marriage market and demonstrate that marriage market repercussions must be taken into account when evaluating the more longrun effects of family policy interventions and changes in family legislation. Kjell Erik Lommerud delves deeper into the insights that can be obtained from non-cooperative models of family decision-making. He focuses on fully non-cooperative family models and, within that category, on models where the family members have conflicting objectives. First he looks at situations where there are family-specific public goods (for example, care of children, a tidy home), and family members try to freeride on the public-good provision of others. Several interesting results can be derived from such models. One group of results concerns neutrality and crowding-out effects. For example, public provision of a close substitute to the family public good (e.g. public childcare) can be shown to crowd out private provision and thus might have no net effect on investments in children. On the other hand, in such models public provision might very well affect the distribution between the spouses in an equalizing direction. Another result from these models is that not only the comparative advantages (as in Becker-type models), but also the absolute advantages of the spouses will affect the family’s allocation of time between household work and market work. Lommerud then goes on to take a look at intergenerational games. He demonstrates the strong structural similarities that exist between various models of “battles of the generations” and private-provision-ofpublic-goods models of the nuclear family. He ends his chapter with an overall assessment of the contributions made so far and the future role and potential of game-theoretic modeling of the family. PART II MARRIAGE AND FAMILY FORMATION The traditional, common preference model of the family implies that the consumption of each family member is not dependent on which individuals in the household receive income or control economic resources but only on total household income. However, as discussed above, if family members differ in their preferences, household decision-making is more appropriately modeled as a (cooperative) bargaining game
10
INTRODUCTION
in which each member’s ability to influence the distributional outcome is related to his or her own contribution to household income, for example to the non-labor income of the spouses. This is the startingpoint for the chapter by Daniela Del Boca in which she uses data about Italian families to test the hypothesis that households in Italy may be treated as if they pool all their income. Rejection of the incomepooling hypothesis, however, does not allow one to distinguish between different bargaining models but only to reject the common preference model. Del Boca uses two samples of married couples: one of all married Italian couples in the age range 21–58 years and one subsample of married couples where a child under six years of age is present. For these samples labor supply functions are estimated which permit income effects to vary depending on the family member to whom the income is attributed. It turns out that the equality of the effects of male and female non-labor incomes is rejected for the total family sample but not for the sample of families with small children. Thus the extent to which the intrafamily distribution of income influences labor supply and household consumption patterns may be related to the composition of the household and to the stage in the life cycle. Del Boca also shows that the “traditional family” model (which treats the labor supply of married men as independent of the behavior of their wives and the husband’s behavior in turn as exogenous with respect to the wife’s work decision) does not seem to be an adequate description of the behavior of Italian families with small children. Taken together her results suggest that families with young children tend to behave and interact in a different way from other families, pooling their resources to a greater extent and allowing their individual labor supply decisions to be mutually and significantly influenced by the characteristics and behavior of the other partner. In the next chapter Kristian Bolin proposes a new, additional explanation for the time allocation of spouses within marriage, namely that one of them (usually the husband) dominates. Dominance is defined as the husband having a first-mover advantage in the process which settles the household’s time allocation, i.e. the husband is allowed to make an irreversible decision about his time allocation before his wife decides upon hers, leaving her no choice but to do the best she can given his time allocation. The decision process in families with one dominating spouse is modeled by a specific non-cooperative game, namely the Stackelberg game. Bolin shows that in such families not only comparative advantages, but also dominance, will matter for time allocation and that the husband will supply more time to the market and the wife will supply more time to the household when the husband has a first-mover advantage than when that is not the case. This means, for example, that in a family where the husband dominates, even if the husband and the wife were identical with respect to their preferences and comparative advantages, the husband would supply more time to the market than would the wife. Furthermore, the existence of dominance can be shown to improve the husband’s utility and to lower the wife’s utility. Does the existence of dominance have implications for family policy? Bolin examines this by looking at the incentives for a government to supply a substitute for the time inputs of family members into a familyspecific public good (such as childcare) under the presence, or respectively the absence, of dominance. The result of his analysis is that the existence of dominance tends to increase the incentives for public provision, so that a government which would not choose to supply a substitute in the absence of dominance might choose to do so in the presence of dominance. Marriage markets and labor markets are likely to interact in ways that are usually ignored in studies of labor supply. The chapter by Shoshana Grossbard-Shechtman and Matthew Neideffer studies these interactions, theoretically and empirically. Based on earlier theoretical work by Grossbard-Shechtman, they expand the theoretical model of married women’s labor supply by incorporating a market for “spousal
INTRODUCTION
11
labor” and marriage decisions into the analysis. Their model of labor and marriage is put into perspective by comparisons with traditional labor supply theory, Becker’s theory of marriage, and household bargaining theory. A major benefit of using their theory of labor and marriage is that it leads to the consideration of a number of factors which can influence labor supply but which have been ignored by traditional labor supply theory, one such factor being marriage market conditions. The effects of marriage market imbalances on labor supply is also the subject of their empirical study. Earlier empirical studies that have tested for the effects of marriage market imbalances on labor supply have been based on aggregate level data. The study by Grossbard-Shechtman and Neideffer presented here uses micro-level data from the 1990 US census (for married, white women aged 25–29) and is the first to look at the effects of marriage market imbalances on individual labor supply. The hypothesis to be tested is the one that in marriage markets with high male/female sex-ratios, women will be getting more income from spousal labor and therefore will supply less labor outside the household. The sex-ratio turns out to have a negative and statistically significant effect on the amount of market labor a woman will supply. This provides empirical evidence for the existence of a sex-ratio or marriage squeeze effect, as predicted from their theoretical model of marriage market and labor market interactions. In the industrialized countries changing patterns of family formation and family dissolution have combined to create what is often considered a problematic family form—the one-parent family. In particular, concern is often expressed about the rapid rise in one-parent families headed by never-married mothers. In Britain the proportion of one-parent families headed by never-married mothers increased from about one-fifth in 1981 to about one-third in 1992. This has triggered a policy backlash and has led to proposals to reduce state benefits which might encourage young women to become single mothers. However, in reality little has been known about the factors behind the rise. John Ermisch disentangles the role of various demographic factors in creating more never-married, single mothers in Britain, using new data which allow the issue to be addressed for more than one cohort and consequently to trace changes over time. Based on the life histories collected in the British Household Panel Study in 1992, Ermisch compares the demographic experiences of two broad cohorts of women, those born during the period 1950–62 and those born after 1962. He finds that the increase in the number of one-parent families headed by never-married mothers reflects a combination of factors. Of prime importance is the fact that a much larger proportion of women are spending some time before marriage in cohabitational unions. While marriage was the dominant mode of entering a first partnership in the 1950–62 cohort, cohabitation had become the dominant mode in the post-1962 cohort. Women in the later cohort are also much more likely to bear a child while cohabiting. If and when such fertile cohabitational unions break up, and in the post-1962 cohort about one-half of them do dissolve, one-parent families of never-married mothers are created. In fact, Ermisch’s estimates suggest that among more recent cohorts of women, about two-fifths of the one-parent families headed by nevermarried mothers are created in this way as compared to one-fifth among the 1950–62 cohort. But of course such families are also created by first births outside of partnerships. About 5 percent of the women in the 1950–62 cohort had a birth before any partnership. The incidence of pre-partnership births is larger in the post-1962 cohort, but is still estimated to be only about 8 percent.
12
INTRODUCTION
PART III FAMILY POLICIES AND HOUSEHOLD ALLOCATION OF TIME The family is often at the center of social policies. The goals as well as the means of family policies have shifted through time and differ between countries. The goals may be as varied as encouraging or discouraging fertility, achieving a more equal income distribution between families or between children growing up under varying circumstances, equalizing the human capital of children, fostering gender equality or simply supporting the family as an institution. The means utilized to achieve family policy goals are also manifold: tax and transfer policies, social security policy, childcare programs, educational policies and family legislation. The papers included in this part of the book analyze some of the policies directed at families in terms of the incentives they create for labor supply, time allocation, human capital accumulation and the choice of how to organize childcare. Economic analyses of childcare policies have generally focused on their effect on women’s labor force participation, and their economic efficiency has been evaluated in terms of whether the production value of women’s increased labor supply outweighs the costs. Siv S.Gustafsson and Frank P. Stafford want to put the focus instead on the role of childcare in the accumulation of human capital, and therefore on how it affects income distribution and long-term economic growth. The family is an important environment, and the parents’ time a valuable input in the building of children’s human capital. As women enter the labor market, some of this human capital production is transferred to the market. At the same time the importance of human capital as a generator of growth is increasing. According to Gustafsson and Stafford, family policy in the future must confront these issues and show a greater awareness of the efficiency and equity aspects of human capital investments involved in the choice of various policy approaches. Any family policy measure will give rise to efficiency as well as equity effects and often there are trade-offs between them. Gustafsson and Stafford provide an overview of the different equity/efficiency trade-offs that parents and policy-makers have to consider in their human capital investments, e.g. between siblings in a family, between children from families with unequal resources, between time for children and time for adults, and between husbands and wives. They then delve deeper into the problem of how to handle investments in children with different learning abilities. Another trade-off they elaborate upon is the intergenerational one between a parent and a child. As illustrated by examples of family policy approaches in The Netherlands, Sweden and the US, the conflict between the mother’s time use in market work, her time use for her own investments in human capital, and her time use for investments in the child’s human capital are approached quite differently in different countries. In view of the rising expenses for childcare programs as well as the increased importance attached to human capital and education in the process of economic growth in industrialized countries today, the equity/efficiency issues raised by the authors in this chapter are likely to receive much more attention in future debate and research. The Scandinavian countries are renowned for their extensive family policies such as generous parental leaves, childcare programs and other types of economic support to families with children. No doubt such policies have made it easier for women to combine work in the market with having children. Female employment in Scandinavia is characterized by very high employment rates for mothers of young children. For example, in 1988 the labor force participation rate among mothers with preschool children was 86 percent in Sweden and 72 percent in Norway. Another feature of Scandinavian female employment is the high prevalence of part-time work among working women. A question little studied is that of the relationship between the design of the family policies and the incidence of part-time work.
INTRODUCTION
13
The paper by Marit Rönsen and Marianne Sundström explores the impact of the parental leave programs and other factors on women’s choice between part-time and full-time work upon re-entry to the labor market after the birth of their first child. Through the use of two national data-sets that are very similar in design, they are also able to compare the policy effects between two countries, namely, Norway and Sweden, that are culturally close and have the same set of public policies, although with some country variations. Their findings indicate among other things that the right to a paid leave of absence makes women resume employment faster after childbirth and that the rates of part-time work as well as those of full-time work are affected. The impact is larger in Sweden, perhaps due to the longer entitlement period there, which makes it easier to reconcile work and motherhood and not drop out of the labor market. On the other hand, the “speed-premium” feature of the Swedish program, introduced in 1980 and enabling women to have a second child within 24 months (since 1986 within 30 months) and maintain their benefit level without returning to the labor market, seems to have delayed the return to both full-time and part-time work after giving birth. In the late 1980s the Finnish government introduced a new policy in the childcare area. The child home care allowance (HCA) is an income transfer paid to all families with children under the age of three who do not use public day-care services. It can be used to finance parental care or as a voucher for hired private childcare. The system is combined with a legal right for the mother or the father to be on leave from work until the child is three years old. Parents with small children may thus choose between a subsidized place in the public day-care system and a cash option. Seija Ilmakunnas analyzes the effects of this measure on the choice of childcare mode and women’s labor supply. In 1994 about three-quarters of the potential recipients chose the home care allowance. Out of those, only a small minority used it to finance private care. Thus parental care of small children has increased. As a result the labor force participation rate of women of childbearing age has dropped and the number of women in full-time household work has increased. Since 90 percent of those taking advantage of the allowance are women, the measure has encouraged a traditional division of labor by sex. An empirical study based on survey data about desired childcare modes illustrates how the use of the HCA varies according to family characteristics. As expected, having more and younger children encourages the use of HCA. There is a strong relationship between the mother’s opportunity wage in the market and the choice of childcare mode. As the mother’s time becomes more highly valued in the labor market, she is less likely to stay at home when her child is small. However, the level of the HCA plays an important role in the decisions. Simulation results indicate that the incidence of family care rises as the generosity of the allowance increases. Overall the study provides clear evidence that family policy and economic incentives play a distinct role in family decisions about the organization of care for small children. The paper by Anne-Marie Pålsson treats an issue that hitherto has received little attention in the literature: the market for employed domestic labor. In the Becker tradition the major focus has been on the household’s possibilities to substitute market goods for home time and on the substitution between the husband’s and the wife’s time within domestic production. Not much attention has been paid to the opportunity to substitute market time for home time and to the preconditions for a market for domestic services to develop. In a high-tax society, such a market will be virtually non-existent. By use of an example based on tax rates similar to the Swedish ones, Pålsson demonstrates that “do-it-yourself solutions” or illegal employment are always economically superior to legal market employment for domestic services in such an environment.
14
INTRODUCTION
She then makes a concrete proposal for a revised tax system in order to expand the market for domestic services. Her suggestion is to make expenditures for such services tax deductible, combining this with transferring social security payments, and the corresponding benefits, from the buyer to the seller of domestic services. Taxes, just like tariffs in international trade, serve as a deterrent to trade between households. Reducing the tax wedges in the domestic sector will encourage specialization and trade in services directed to households. In a Gronau-type model which separates between market goods, domestic output and leisure, she demonstrates that such a tax change may increase overall economic efficiency and thereby total output in society. All households stand to gain. Households with high productivity on the market can supply more of their labor to the market and hire domestic services to substitute their own labor at home. Low-productivity households will gain access to an additional labor market. Households who continue to do their own household work will gain from the improved economic efficiency and the rise in total output in the economy. Since women are the ones who carry out the major share of domestic work, an additional benefit may be improved gender equality in the sense that women, through having access to domestic services, will be able to compete with men on more equal terms in the market. WHY ECONOMICS CANNOT DO WITHOUT ECONOMICS OF THE FAMILY The family is a basic social institution. For a long time the focus of economic analysis on market activities excluded decisions and activities within the family from the economists’ realm. However, the choices made within the family about time allocation, about the distribution of income and consumption, and about human capital accumulation in parents and children have important repercussions for what happens in the market and to the development of the economy. There are also several reasons why the interactions and the repercussions between the market sector and the household sector have grown in importance over time. One reason is the dramatic increase in women’s participation in market work and the less dramatic, but still increased participation of men in household work. This changing division of labor between the sexes means that a much larger share of the adult population today uses their time in both market and household work, rather than being specialized in one or the other. As long as women devoted their time mainly to the household sector, and men to the market sector, it was less misleading to disregard the repercussions for household production when making economic analyses of the market sector. The two sectors then solved their problems of time allocation and resource allocation more or less independently of each other. Today changes in wages, relative prices, taxes, social security programs, etc. are likely to have more direct and immediate effects on the time allocation and household production of households, effects which must not be ignored when evaluating such changes. A second reason for greater interdependence is that with the increase in women’s market work, a lot of production with close substitutes in household production (as exemplified by childcare, school meals, care of the elderly) has been transferred to the market. And the interdependence between the market sector and the household sector will be larger, the closer substitutes in consumption are the goods and services produced in the two sectors, and the smaller are the productivity differences between the two sectors in the production of such goods and services. A third reason is the growth in the number and size of public policies and programs which do not treat market work and non-market work in the same way. Examples are tax systems, social security systems, subsidized market-produced childcare and other family policies. Such policies and programs will introduce “tax and/or benefit wedges” between market and non-market work and between market-produced and household-produced goods and services. These wedges will affect household decisions and might give rise
INTRODUCTION
15
to economic inefficiencies in production and consumption. Awareness and consideration of this is needed when designing public policies and programs. In order to analyze the growing interactions between the household sector and the market sector, economists need the theories, models and empirical insights provided by the economics of the family. The development of this rapidly growing field and the contributions to this volume show that family decisions are indeed amenable to the use of the basic tools contained in the economist’s tool kit.
Part I WHERE ARE WE IN THE ECONOMICS OF THE FAMILY?
1 BARGAINING AND DISTRIBUTION IN MARRIAGE Shelly Lundberg and Robert A.Pollak
INTRODUCTION In the 1970s, a proposed change in social welfare policy in the United Kingdom excited considerable debate. The universal child allowance, which had consisted primarily of a reduction in the amount withheld for taxes from the father’s paycheck, was to be replaced by a cash payment to the mother. An excerpt from the parliamentary debate in Hansard (House of Commons, May 13, 1975) expresses a popular sentiment: [F]ar from a new deal for families, it will take money out of the husband’s pocket on the Friday and put it into the wife’s purse on the following Tuesday. Far from being a child benefit scheme, it looks like being a father disbenefit scheme. Popular discussions of family policies such as the UK child benefit often concern their presumed effects on distribution within the family—on the relative well-being of husbands, wives and children. The economist armed only with traditional models of the family must view these discussions as naive. Until very recently, the standard of the profession for both theoretical and empirical analysis was a “common preference” model of the family, which assumes that family members act as though they are maximizing a single utility function. A family’s common preference ordering may be the outcome of consensus among family members or the dominance of a single family member, but all such models imply that family expenditures are independent of which individuals in the family receive income or control resources. Common preference models imply that all income is “pooled” and then allocated to maximize a single objective function, so that family demand behavior depends on total family income, and not the incomes of individual members. This pooling of resources within the family implies that a change from child allowances paid to fathers to child allowances paid to mothers should arouse neither the ire of affected fathers nor the opposition of their parliamentary representatives. The United Nations World Population Conference, held in Cairo in September 1994, witnessed a continuation of a long-standing debate about how to reduce birth rates in developing countries. Population experts divided into two camps: one favoring continued emphasis on family planning services, the other favoring policies that improve the status of women. The latter group argues that greater access to education, business loans and development projects would give women more control over reproduction and that, as a result, birth rates would fall. Economists understand a link between fertility rates and the educational and earnings opportunities of women that operates through the value of women’s time and the time price of children, but the proponents
18
SHELLY LUNDBERG AND ROBERT A.POLLAK
of women’s “empowerment” have emphasized the effect of women’s education and income on their decision-making authority within the household. Models that treat the family as a black box, with income flowing in and demands for goods, services, leisure and children flowing out, cannot deal with this argument, and so cannot address the intrafamily distribution issues that concern population and development agencies. If economists are to participate in this important debate (or at a minimum comprehend it), we must move beyond common preference models of family behavior. To this end, the theoretical challenge facing family economics is to develop models in which joint family decisions are derived from the sometimes divergent interests of husbands and wives, and in which the formation and dissolution of marriages provide a beginning and an end to the family allocation process. In recent years, a large number of game-theoretic models of marriage and the family have been developed, building on the seminal contributions of Manser and Brown (1980) and McElroy and Horney (1981). In general, these models impose fewer restrictions on observed family behavior than do common preference models, and recent theoretical contributions have been prompted, and supported, by a growing body of empirical evidence inconsistent with common preference models. The most provocative of this empirical work demonstrates a strong positive association between child well-being and the mother’s relative control over family resources, and has raised new questions about the potential effectiveness of policies “targeted” to specific family members. A current snapshot of family economics would show the traditional framework under siege on both theoretical and empirical fronts. The political potency of gender issues has given a certain urgency to the development of alternatives to common preference models. However, no new theoretical framework has gained general acceptance as a replacement for common preference models, and empirical studies have concentrated on debunking old models, rather than on discriminating among new ones. In this paper, we review a number of simple bargaining models that permit independent agency of men and women in marriage, discuss their implications for distribution within marriage and for observed family behavior, and present a sampling of the relevant empirical evidence.1 MODELS OF FAMILY BEHAVIOR Economic models of consumer demand and labor supply begin with an individual economic agent choosing actions that maximize his or her utility function subject to a budget constraint. How can we reconcile this individualistic theory of the consumer with the reality that people tend to live, eat, work and play in families?2 Application of a single-agent model to the household or family raises two distinct issues—the identity of the consumer and the identity of the decision-maker. The identity of the consumer is an issue because microdata on “consumption” usually report expenditures at the household level, seldom consumption at the individual level. The household purchases bread and refrigerators, ballet lessons and haircuts, but in general the data do not assign the consumption of these goods and services to individual household members. If the problem were fundamentally data-based, however, we could solve it by collecting better data on individual consumption, time allocation and income. There are two reasons why better data would be only a partial solution, and thus no solution at all. First, goods whose consumption is inherently joint are an important component of household consumption. With household public goods, better data cannot solve the assignment problem and, hence, cannot restore the integrity of the simple single-agent consumer model.3 Second, family members who are linked by love and duty have an interest in each other’s consumption. Even if we could assign direct consumption to individual family members, interdependent preferences would invalidate the single consumer assumption.
BARGAINING AND DISTRIBUTION IN MARRIAGE
19
Economists have dealt with the multiplicity of decision-makers in the family in two ways. The first approach, in ascendancy from the 1950s until the 1980s, was the common preference approach—treating the family as though it were a single decision-making agent, with a single pooled budget constraint and a single utility function that includes the consumption and leisure time of every family member. The second approach, pioneered by Manser-Brown and McElroy-Horney in the early 1980s, was to model family demands as the solution to a bargaining game. Most bargaining models of family behavior allow two decision-makers—the husband and the wife. Children are customarily excluded from the set of decisionmaking agents in the family, though they may be recognized as consumers of goods chosen and provided by loving or dutiful parents. The empirical implications of bargaining models of marriage depend upon their assumptions about the form of the bargaining game but, in general, these models widen the range of “rational” family behavior. COMMON PREFERENCE MODELS AND THE INCOME-POOLING ASSUMPTION Two models provide the theoretical underpinning of the common preference approach to family behavior: Samuelson’s (1956) consensus model and Backer’s (1974, 1981) altruist model. The consensus model was introduced by Samuelson to exhibit the conditions under which family behavior can be rationalized as the outcome of maximizing a single utility function. Consider a two-member family consisting of a husband and a wife. Each has an individual utility function that depends on his or her private consumption of goods but, by consensus, they agree to maximize a social welfare function of their individual utilities, subject to a joint budget constraint that pools the income received by the two family members. Then we can analyze their aggregate expenditure pattern as though the family were a single agent maximizing a utility function. This optimization problem generates family demands that depend only upon prices and total family income, and that have standard properties provided the utility functions are well-behaved.4 Thus, the comparative statics of traditional consumer demand theory apply directly to family behavior under the consensus model. Samuelson did not, however, purport to explain how the family achieves a consensus regarding the joint welfare function, or how this consensus is maintained. Becker’s altruist model (1974, 1981) addresses these questions, and also provides an account of how resources are distributed within the family. In Becker’s model, the family consists of a group of purely selfish but rational “kids” and one altruistic parent whose utility function reflects his concern for the wellbeing of other family members. Becker argues that the presence of an altruistic parent who makes positive transfers to each member of the family is sufficient to induce the selfish kids to act in an apparently unselfish way. The altruistic parent will adjust transfers so that each “rotten kid” finds it in his interest to choose actions that maximize family income. The resulting distribution is the one that maximizes the altruist’s utility function subject to the family’s resource constraint, so the implications of the altruist model for family demands coincide with those of the consensus model. Whether motivated by Samuelson’s family consensus story or Becker’s altruist story, the common preference framework is a simple, powerful mechanism for generating demand functions and establishing their comparative statics for use in applied problems. It remains the standard theoretical framework for analyzing consumption behavior and labor supply. Only serious deficiencies could justify replacing this approach with a more complicated alternative. In recent years, however, common preference models have been targets of an intense barrage of theoretical and empirical criticism.
20
SHELLY LUNDBERG AND ROBERT A.POLLAK
Dissatisfaction with common preference models on theoretical grounds has been the product of serious study, by economists, of marriage and divorce. Models of marriage and divorce require a theoretical framework in which agents compare their expected utilities inside marriage with their expected utilities outside marriage. Common preference models cannot be used to examine these decisions, because the individual utilities of husband and wife cannot be recovered from the social welfare function that generates consumption, labor supply, fertility and other behavior within marriage. If the analysis of marriage and divorce is awkward, the analysis of marital decisions in the shadow of divorce is even more so. If unilateral divorce is possible, individual rationality implies that marital decisions cannot leave either husband or wife worse off than they would be outside the marriage. This individual rationality requirement, however, alters the comparative statics of the model, and destroys the correspondence between the behavior of a single rational agent and the behavior of a family. Recent empirical evidence suggests that the restrictions imposed on demand functions by common preference models are not well-supported. Rejections of the family income-pooling assumption have been most influential in weakening economists’ attachment to common preference models.5 Income-pooling implies a restriction on family demand functions that appears simple to test: if family members pool their income and allocate the total to maximize a single objective function, then only total income will affect demands. The fraction of income received or controlled by one family member should not influence demands, conditional on total family income. A large number of recent empirical studies have rejected pooling, finding that earned and unearned income received by the husband or wife significantly affect demand patterns when total income or expenditure is held constant. Many studies find that children appear to do better when their mothers control a larger fraction of family resources. Empirical tests of pooling, using data from a variety of countries, invariably show that income controlled by the husband and wife have significant and often substantially different effects on family behavior, whether measured by expenditure on categories of goods and services, or measured by outcomes such as child health.6 For example, increases in the wife’s income relative to the husband’s income have been shown to be associated with greater expenditures on restaurant meals, childcare and women’s clothing (Phipps and Burton 1992),7 and with reduced expenditures on alcohol and tobacco (Phipps and Burton 1992; Hoddinott and Haddad 1995). Increases in child health, nutrition and survival probabilities have also been associated with mothers’ control over family resources (Thomas 1990, 1994; Hoddinott and Haddad 1995; Rose 1994). Estimated differences in the effects of mothers’ and fathers’ resources on child outcomes are often large: Thomas (1990) finds that the effects of mothers’ unearned income on child survival probabilities in Brazilian data is almost 20 times that of fathers’ income. A test of the pooling hypothesis requires a measure of husband’s and wife’s relative control over resources. Relative earnings would seem to be an attractive candidate for this measure, since labor income is by far the largest component of family income, and earnings data are readily available and reliably measured. Also, the earnings of wives relative to husbands have increased dramatically in the US and many other countries, and we would like to assess the distributional consequences, if any, of this change. The difficulty with this approach is that earnings are clearly endogenous with respect to the household’s time allocation decisions, so that households with different ratios of wife’s earnings to husband’s earnings are likely to face different prices and may have different preferences. If we think of earned income as the product of hours worked and a fixed market wage rate, then the first factor, hours worked, is a standard choice variable in models of household behavior and is determined simultaneously with the expenditure patterns the pooling test examines. The second factor, the wage rate,
BARGAINING AND DISTRIBUTION IN MARRIAGE
21
measures the price of time for the husband or wife, and enters the household’s demand functions in common preference models and in bargaining models. Thus the interpretation of the separate effects of wife’s earnings and husband’s earnings is problematic. Consider the finding of Phipps and Burton (1992) that expenditures on restaurant meals are more elastic with respect to the wife’s earnings than the husband’s earnings. A bargaining interpretation of this result is that, as the wife’s earnings rise relative to the husband’s, she gains more influence over the household’s spending patterns and that increased expenditures on restaurant meals reflect her preferences. The common preference interpretation is that restaurant expenditures depend upon the cost of substitutes, and that the wife’s wage is an important component of the cost of home-prepared meals. Thus, the Phipps-Burton result can be interpreted as a price effect rather than as evidence against income-pooling. One might try to avoid these problems by testing the pooling of unearned income rather than earnings. Unearned income is not contaminated by price effects, but most unearned income sources are not entirely exogenous with respect to past or present household behavior. Furthermore, variations in unearned income over a cross-section are likely to be correlated with other (possibly unobservable) determinants of consumption.8 For example, property income reflects, to a considerable extent, accumulated savings and is therefore correlated with past labor supply and, if those who worked a lot in the past continue to do so, current labor supply. Public and private transfers may be responsive to household distress due to unemployment or bad health, and may be related to expenditures through the events that prompted them (Schultz 1990). Unexpected transfers such as lottery winnings, unexpected gifts or unexpected bequests will affect resources controlled by individuals without affecting prices, but are likely to be sporadic and unimportant for most families. The ideal test of the pooling hypothesis would be based on an experiment in which some husbands and some wives were randomly selected to receive an income transfer. A less-than-ideal test could be based on a “natural experiment” in which some husbands or some wives received an exogenous income change. Lundberg et al. (1997) examine the effects of such a natural experiment—the policy change in the UK that transferred a substantial child allowance from husbands to wives in the late 1970s. They find strong evidence that a shift towards relatively greater expenditures on women’s goods and children’s goods coincided with this income redistribution,9 and interpret this as a rejection of the pooling hypothesis. Rejecting the pooling hypothesis has important policy implications. Policy-makers sometimes want to target transfers or programs to particular classes of individuals within families, such as women or children. Common preference models imply that such policies are ineffective, beyond their influence on total family resources, because the equilibrium intrahousehold allocation is independent of the distribution of income among family members. Bargaining models, on the other hand, suggest that the government can affect distribution within marriage, either by changing the income of divorced men and women, or by transferring control over resources within the marriage from one spouse to the other. Some empirical work supports the potential effectiveness of such policies, but much more will be required to establish the nature of family equilibrium in different cultural and institutional contexts, and the extent to which it can be shifted by government policies. COOPERATIVE BARGAINING MODELS A viable alternative to common preference models of the family must relax the pooling assumption and must recognize, in a non-trivial fashion, the involvement of two or more agents with distinct preferences in
22
SHELLY LUNDBERG AND ROBERT A.POLLAK
Figure 1.1 The Nash bargaining solution
determining family consumption. Bargaining models from cooperative game theory satisfy these conditions. A typical cooperative bargaining model of marriage begins with a family that consists of only two members: a husband and a wife. Each has a utility function that depends on his or her consumption of private goods (Uh for the husband Uw and for the wife). If agreement is not reached, then the payoff received is represented by the “threat point,” (Th, Tw)—the utilities associated with a default outcome of divorce or, alternatively, a non-cooperative equilibrium within the marriage. The Nash bargaining model provides the leading solution concept in bargaining models of marriage.10 This solution can be illustrated by a diagram in utility space, where AB is the utility-possibility frontier (see Figure 1.1). Nash (1950) shows that a set of four axioms, including Pareto-optimality—which ensures that the solution lies on the utilitypossibility frontier—uniquely characterizes the Nash bargaining solution. The utility received by husband or wife in the Nash bargaining solution depends upon the threat point; the higher one’s utility at the threat point, the higher one’s utility in the Nash bargaining solution. This dependence is the critical empirical implication of Nash bargaining models: family demands depend, not only on prices and total family income, but also on determinants of the threat point. In divorce-threat bargaining models, the threat point is the maximal level of utility attainable outside the marriage. If divorcing partners maintain ownership of income received separately within marriage, the demands emerging from marital bargaining will depend not on total family income, but on the income received by the husband and the income received by the wife. The divorce threat point is also likely to depend on environmental factors (extrahousehold environmental parameters, or EEPs in McElroy’s (1990) terminology) that do not directly affect marital utility, such as conditions in the remarriage market and the income available to divorced men and women. The family demands that result from divorce-threat marital bargaining will therefore depend upon these parameters as well. As McElroy points out, the absence of pooling and the presence of extrahousehold parameters in family demands yield a model that can be tested
BARGAINING AND DISTRIBUTION IN MARRIAGE
23
against the common preference alternative. For example, changes in the welfare payments available to divorced mothers, or in laws defining marital property and regulating its division upon divorce, should affect distribution between men and women in two-parent families through their effect on the threat point. In the separate spheres bargaining model of Lundberg and Pollak (1993), the threat point is internal to the marriage, not external as in divorce-threat bargaining models. The husband and wife settle their differences by Nash bargaining, but the alternative to agreement is an inefficient non-cooperative equilibrium within marriage. In a non-cooperative equilibrium, each spouse voluntarily provides household public goods, choosing actions that are utility-maximizing, given the actions of their partner. This non-cooperative marriage may be better for both spouses than divorce. Divorce, the argument goes, may be the ultimate threat available to marital partners in disagreement, but a non-cooperative marriage in which the spouses receive some benefits due to joint consumption of public goods may be a more plausible threat in day-today marital bargaining. The introduction of this internal threat point has important implications, because separate spheres bargaining generates family demands that, under some circumstances, depend not on who receives income after divorce, but on who receives (or controls) income within the marriage. Control over resources within marriage need not affect the equilibrium: if both the husband and the wife make positive contributions to each public good in the non-cooperative equilibrium, then household allocation will not depend upon how income is distributed between the spouses.11 In the separate spheres model, however, a non-pooling outcome arises when gender specialization in the provision of household public goods ensures that only one spouse makes a positive contribution. The model assumes that socially recognized and sanctioned gender roles assign primary responsibility for certain activities to husbands and others to wives. In the absence of cooperation, one household public good, q1, will be provided by the husband out of his own resources and the other public good, q2, by the wife out of her own resources. Lundberg and Pollak assume that this allocation of marital responsibilities reflects social norms, rather than preference or productivity differences between husband and wife in a particular marriage. In a non-cooperative marriage, the husband treats the level of public good chosen by his wife as fixed and chooses quantities of his private good and his assigned public good so as to maximize his utility subject to his budget constraint. Similarly, the wife treats the quantity of the public good supplied by her husband as fixed and chooses utility-maximizing quantities of her private and public goods subject to her budget constraint. These decisions lead to a pair of reaction functions that determine a Cournot-Nash equilibrium in which the public goods contributions are inefficiently low. An important characteristic of this non-cooperative equilibrium, which serves as the threat point in the separate spheres model, is that the husband’s utility depends upon the resources of his wife through his consumption of “her” public good, and vice versa. Since the demand functions generated by cooperative bargaining depend upon the threat point, they will also be independently influenced by husband’s income and wife’s income. In the cooperative equilibrium, the husband’s and wife’s utilities will depend not on total family income but on the incomes controlled separately by each spouse.12 As the divorce-threat and separate spheres models show, cooperative bargaining does not necessarily imply income-pooling. Bargained outcomes depend upon the threat point, and the income controlled by husband and wife will affect family behavior (and the relative well-being of men and women within marriage) if this control influences the threat point. This dependence implies that public policy (e.g. taxes and transfers) need not be neutral in their effects on distribution within the family, although how they affect distribution depends upon how the alternative to agreement is specified. A divorce-threat bargaining model predicts that policies
24
SHELLY LUNDBERG AND ROBERT A.POLLAK
improving the status of divorced women will shift resources within marriage to wives; it also predicts that policies affecting the control of income within the marriage will have no effect on distribution within marriage if they have no effect on the incomes of divorced men and women. A separate spheres bargaining model predicts that policies reallocating income within marriage will change distribution within marriage and family demands, even if they do not affect the well-being of divorced men and women. Consider, for example, a change in child allowance policy from one that pays husbands to one that pays wives, but suppose that in the event of divorce, the mother is always the custodial parent and receives the child allowance. Divorce-threat models predict that this change will have no effect on distribution in two-parent families, while the separate spheres model predicts redistribution towards the wife. PARETO-OPTIMALITY AND NON-COOPERATIVE BARGAINING MODELS Most models of the family either assume or conclude that family behavior is Pareto-optimal. Common preference models ensure Pareto-optimality by assuming a family social welfare function that is an increasing function of the utilities of all family members: no member can be made better off without making another worse off. Cooperative bargaining models characterize the equilibrium distribution by means of a set of axioms, one of which is Pareto-optimality. Distributional issues remain important: as Lommerud has stressed in his paper in this book, “efficiency” does not imply “harmony.” However, the focus on models that restrict us to the utility-possibility frontier is striking. Two recent departures have been the development of empirical models that permit tests of Pareto-optimality, and applications of non-cooperative game theory to the family that allow us to examine what conditions might enable families to sustain Paretooptimal outcomes. Pareto-optimality is the defining property of the “collective model” of Chiappori (1988, 1992). Rather than applying a particular cooperative or non-cooperative bargaining model to the household allocation process, Chiappori assumes only that equilibrium allocations are Pareto-optimal, and so his collective model contains cooperative bargaining models and common preference models as special cases. He demonstrates that, given a set of assumptions including weak separability of public goods and the private consumption of each family member, Pareto-optimality implies, and is implied by, the existence of a “sharing rule.” Under a sharing rule, the family acts as though decisions were made in two stages, with total family income first divided between public goods and the private expenditures of each individual, and then each individual allocating his or her share among private goods. The collective framework thus imposes a set of testable restrictions on the observed demands of the household. In essence, the ratio of the marginal propensities to consume any two goods must be the same for all sources of income, because the independent incomes of husband and wife affect consumption only through the sharing rule. The pattern of consumption expenditures in Canadian and French households has been found to be consistent in this sense with Pareto-optimality (Bourguignon et al. 1993; Browning et al. 1994). Nevertheless, the prevalence of destructive or wasteful phenomena such as domestic violence and child abuse, as well as the demand for marriage counseling and family therapy, suggests that we consider the possibility that family behavior is sometimes inefficient. Other researchers have pointed to gender segmentation in the management of businesses or agricultural plots in many countries as evidence of an essentially non-cooperative, and possibly inefficient, family environment. A rare fragment of empirical evidence is provided by Udry (1996), who finds that the household allocation of resources to male- and female-controlled agricultural plots in Burkina Faso is inefficient.
BARGAINING AND DISTRIBUTION IN MARRIAGE
25
Cooperative game theory motivates the assumption of Pareto-optimality by assuming that information is relatively good (or at least not asymmetric) and that the players can make binding, costlessly enforceable agreements. Since legal institutions do not provide for external enforcement of contracts regarding consumption, labor supply and allocation within marriage, the binding agreement assumption is unappealing. Non-cooperative game theory, in contrast, does not assume that binding agreements enforce intrahousehold allocations, but focuses instead on self-enforcing equilibria. Pareto-optimal outcomes are possible in non-cooperative games, but not necessary. Without binding agreements, much of the motivation for assuming Pareto-optimality vanishes. It is possible, however, for non-cooperative bargaining to yield Pareto-optimal outcomes under certain conditions. For example, if the voluntary contribution game played by husbands and wives in the separate spheres model is played only once, it yields an inefficient equilibrium in which public goods are underprovided; but if the voluntary contribution game is played repeatedly, many other equilibria are possible.13 In general, repeated non-cooperative games have multiple equilibria, and Pareto-optimal equilibria can often be sustained by the threat of punishment. In essence, each spouse realizes that the one-period gain from deviating from an agreement will be less than the loss associated with being punished by their spouse in the periods that follow. Browning et al. motivate their assumption of Pareto-optimality with the claim that the marital environment possesses characteristics that would promote efficient outcomes in a repeated non-cooperative game—a long-term relationship, relatively good information and a stable bargaining environment. We prefer a different research strategy. One of the benefits of modeling distribution within marriage as a noncooperative game is the opportunity to treat efficiency as endogenous, potentially dependent upon the institutions and social context of marriage in a particular society and upon the characteristics of the marital partners. The corresponding costs include the need to specify fully the set of possible actions and the timing of moves. The existence of multiple equilibria in repeated non-cooperative games and the need to choose among them suggest how history and culture might affect distribution within marriage. Kreps (1990) points out that, in many games, there seems to be a “self-evident way to play” that corresponds to a particular equilibrium. He emphasizes that which equilibrium corresponds to the self-evident way to play cannot, in many cases, be identified solely from the formal description of the game: in realistic social contexts, conventional modes of behavior may suggest to the players a “focal point equilibrium,” thus reducing or eliminating the need for pre-play negotiations. In the case of marriage, social conventions regarding the rights and responsibilities of husbands and wives may indeed suggest to the spouses a particular equilibrium. For example, consider a model with two-household public goods in which the husband and wife make voluntary contributions. Suppose that specialization is desirable in the sense that the household is better off if the wife supplies one good and the husband supplies the other. This game may possess two Nash equilibria analogous to those in the “Battle of the Sexes” game— one in which the wife supplies good one and the husband good two, and another in which the provider roles are reversed.14 The husband and wife may prefer to provide one good rather than the other, but both will prefer a coordinated provision of public goods to the inefficient alternative in which both supply the same good. The choice between the two equilibria is likely to be sensitive to history and culture, which may generate a “self-evident” way to play. The separate spheres bargaining model provides an obvious example: if some household public goods are regarded as within the wife’s sphere and others as within the husband’s sphere, then the focal point equilibrium may involve complete gender specialization in the provision of household public goods
26
SHELLY LUNDBERG AND ROBERT A.POLLAK
corresponding to this conventional gender assignment of responsibilities. In this non-cooperative model, distribution within marriage will depend on the individual resources of husband and wife, due to the corner solution in public goods provision. Treating distribution within marriage as the outcome of a repeated non-cooperative game, we see the issue of Pareto-optimality through a different lens. The existence of multiple equilibria, some of which are Pareto-optimal and some of which are not, suggests that we consider factors omitted from the formal model to explain the patterns of marital behavior and gender allocations that develop in any particular society. The behavior of any particular couple may be directed towards a focal point equilibrium that conforms with the behavior of those around them and is consistent with socially sanctioned gender roles. Viewed as the outcome of a repeated game in a social context, the Pareto-optimality of distribution within marriage must be investigated and analyzed, not simply assumed. Like any microanalysis that appeals to focal points or social norms, our analysis inevitably raises macro questions—how do the social norms and gender roles that constrain a particular marriage arise and how are they maintained—and directs our attention to these larger issues. If the achievement of a Pareto-optimal outcome depends upon such factors as the stability of the marital environment and the quality of information possessed by husband and wife, then we may be able to analyze the role of marital and other societal institutions in promoting efficient marriage (as well as affecting distribution between husbands and wives), at least in the short run. These institutional factors could include the role of older generations in arranging marriages and regulating marital behavior, restrictions on the economic behavior of married women, the costs of leaving a marriage, and the social and legal treatment of domestic violence. If one takes seriously the notion that institutions and practices, norms and gender roles are endogenous, then the analysis of individual behavior, individual well-being and Pareto-optimality must be recast. England and Kilbourne (1990) and Sen (1990) develop analyses that depend crucially on this endogeneity. England and Kilbourne argue that women are socialized to be less willing than men to drive hard bargains with their spouses and, hence, that wives get less than they otherwise would. Sen carries the internalization argument a step further, arguing that “socialization”—he avoids the word— may prevent a woman from recognizing her true interests. Non-economists’ critiques of economists’ analyses of distribution between men and women often use words like “power” that are foreign to the vocabulary of economics. Pollak (1994) argues that, although the language is unfamiliar, the substance of these critiques is that economic models of distribution between men and women focus on the subgame of bargaining within a particular marriage and that the real action is elsewhere—in the prior game that determines social norms and gender roles. Although individual men and women take the outcome of this earlier game as given, economists should not, for it determines the institutions and norms that affect the play in a particular marriage. THE MARRIAGE MARKET Models that analyze bargaining within existing marriages can give only an incomplete picture of the determinants of the well-being of men and women. The marriage market, as Becker has emphasized (1991: 13–15), is an important determinant of distribution between men and women. At a minimum, the marriage market determines who marries, and who marries whom. The extent to which the marriage market also determines distribution within particular marriages depends crucially on whether prospective spouses can make binding agreements in the marriage market. At one extreme, if binding, fully contingent contracts regarding marital distribution can be made prior to marriage, then there is no scope for bargaining within marriage: distribution within marriage simply implements agreements previously made in the marriage
BARGAINING AND DISTRIBUTION IN MARRIAGE
27
market.15 At the other extreme, if binding agreements cannot be made in the marriage market, then husbands and wives bargain over the surplus generated by a particular marriage. The marriage market can also generate substantial differences between the short-run and long-run effects of tax, transfer and other redistributive policies. In Lundberg and Pollak (1993), we consider a model in which prospective spouses can agree on a transfer payment from (for example) husbands to wives that is uncontingent on the realized values of income later in the marriage. If policy-makers attempt to redistribute income by transferring the ownership of a child allowance payment from husbands to wives, some redistribution is likely to occur within marriages in existence at the time of the policy change. For the next generation of marriages, however, there will be a new equilibrium with the same pairing of men and women, but with the agreed transfer to wives reduced by the amount of the child allowance. With binding marital agreements, therefore, targeted policies that have redistributive effects in existing marriages may be “undone” by subsequent generations in the marriage market—a pure Ricardian equivalence result. Even without binding agreements, however, the long-run effects of a redistributive policy are likely to differ from the short-run effects on existing marriages. Prospective spouses understand that marriage commits them to playing a particular bargaining game with a particular partner. A policy that transfers income from husbands to wives will make marriage relatively more attractive to women and less attractive to men. Such a change in transfer policy can alter the equilibrium number of marriages contracted in subsequent marriage markets, as well as the equilibrium matching and distribution of marital surpluses (Lundberg and Pollak, 1993). The scope for bargaining within marriage also depends upon the alternatives available to the marital partners. In the marriage market, if there are close substitutes for each individual, then the next best marriage is nearly as good as the proposed one, and the surplus to be divided by bargaining is small. Over time, however, a sizable surplus may develop in an ongoing marriage, perhaps because of investments in marriage-specific human capital. In this situation, the possibility of divorce (perhaps followed by remarriage) defines the scope for bargaining ex post within marriage by placing bounds on the distributions that can emerge as equilibria. These “divorce bounds” depend upon the costs of divorce, including psychic costs, the resources available to divorced individuals, and conditions in the remarriage market. Individual rationality ensures that no individual will accept less than he or she would receive in the next best alternative and implies that the divorce bounds apply to all bargaining models, both cooperative and noncooperative.16 Just as there is little scope for bargaining in the marriage market when the next best marriage is almost as good as the proposed marriage, there is little scope for bargaining within marriage when the divorce bounds are tight. Bargaining models of marriage are motivated by the assumption that, in at least some marriages, surpluses are large enough that their distribution is worth modeling. The role of marriage markets in determining distribution within marriage provides another example of the importance of social norms and institutions. When matching models have multiple equilibria, as they often do, which equilibrium is selected or realized may depend upon institutions and practices not specified in the formal model. For example, it is well-known that in a marital matching model, the equilibrium realized when men propose to women is more favorable to men and less favorable to women than the equilibrium realized when women propose to men. Pollak (1994) argues that when the selection of one equilibrium rather than another has important distributional implications, institutions and practices (e.g. courting conventions) should be explicitly modeled.
28
SHELLY LUNDBERG AND ROBERT A.POLLAK
CONTRIBUTIONS OF BARGAINING APPROACHES TO MARRIAGE Common preference models of the family have proven to be too limited a framework for the analysis of family behavior. Though a rigorous and powerful tool for analyzing family expenditure patterns and labor supply, its assumption of a single family utility function and its implication of family income-pooling are problematic. Furthermore, common preference models rule out analysis of intrafamily distribution or of the connection between marriage markets and marital behavior. Game-theoretic approaches to family behavior provide new models, yield new results and provoke new questions. Novel questions and areas of inquiry are numerous, but three that seem particularly interesting to us are the effect of control of resources by husbands and wives on the well-being of children, the effect of social norms on marital bargaining, and the relationship between marital distribution and marriage markets. Policies that empower women have been supported not only by claims that they will increase the wellbeing of women and reduce birth rates, but also by claims that they will increase the well-being of children. The belief that “kids do better” when their mothers control a larger fraction of family resources, which was presumably part of the rationale for changing the UK child benefit program in the late 1970s, has now attained the status of conventional wisdom among development agencies. This belief entails two distinct propositions, both confidently maintained in a recent World Bank (1995) monograph on gender equality. First, we must reject income-pooling in favor of some alternative in which control over resources influences distribution within the family. As the World Bank (1995:59) puts it, “policies that specifically target women or girls can address the needs of this group more efficiently and with greater cost-effectiveness than general policy measures.” Second, we must accept the additional hypothesis that “[f]emale household members tend to allocate resources more directly to children, while men tend to allocate more resources to adults” (World Bank 1995:59). The kids-do-better hypothesis is widely accepted and has received extensive empirical support: Bruce et al. (1995) and Blumberg (1991) cite and summarize many of the relevant studies. Economists, many of whom have been skeptical on theoretical grounds that the kids-do-better results were caused by mothers’ control over resources, have found the possible endogeneity of income sources discussed earlier to be sufficient econometric grounds for discounting the empirical evidence that supports it. Though the evidence on this point is not conclusive, we think that the burden of proof has shifted to those who doubt that children benefit when their mothers control a larger fraction of family resources. The notion that control over resources matters focuses attention on the difficult issue of the meaning and measurement of “control.” Does the individual family member whose name is on the check maintain control over its disbursement? Are in-kind transfers more controllable by individual recipients than cash? To what extent are own earnings “owned” by the worker rather than pooled for household use? Work by sociologists on family budgeting suggests considerable heterogeneity among families in money management practices (Pahl 1983; Treas 1993; Zelizer 1989, 1994; and the literature they cite). Economists, however, are unlikely to find money management practices especially interesting as outcome variables or appropriate as explanatory variables. A more interesting focus for economists is the relationship between control over resources and the extramarital environment, on the one hand, and outcomes such as expenditure patterns, labor supply, and observable indicators of individual well-being such as morbidity and mortality, on the other. In particular, empirical studies that examine the effects of differences in tax and transfer policies that appear to establish different claims on resources within the household are likely to improve our understanding of intrahousehold allocation.
BARGAINING AND DISTRIBUTION IN MARRIAGE
29
Bargaining models of marriage suggest a number of mechanisms through which social norms and institutions can affect distribution between men and women. Most directly, social norms may affect the preferences of marital partners. In bargaining models, social norms affect outcomes indirectly, often through their effect on the threat point. In the divorce-threat bargaining model, custody and child support standards and the social position of divorced men and women will be among the extramarital environmental parameters that determine the threat point. In the context of this model, an increase in welfare “stigma” will be associated with a decline in the relative well-being of women and children in low-income families, as will reductions in the real value of welfare payments. If the threat point reflects the possibility of domestic violence, marital distribution may depend upon the expected reaction of neighbors and the behavior of police on domestic violence calls. Non-cooperative outcomes may be influenced by social norms in a different way; in models with multiple equilibria, social conventions may suggest a focal point equilibrium and a way of coordinating behavior without explicit bargaining. Norms regarding appropriate marital or parental behavior for men and women may be powerful in their ability to channel the behavior of marital partners to one equilibrium among many—raising the question of how such norms develop and are maintained. Bargaining models place distribution within marriage in a theoretical framework that is consistent with existing analyses of marriage and of divorce: two decision-makers with well-defined preferences choosing an action or strategy from a well-specified set of alternatives. Bargaining models thus provide an opportunity for integrating the analysis of distribution within marriage with a matching or search model of the marriage market. In a unified model, marital bargaining is conditional upon the match (and perhaps contract) agreed to in the marriage market, and agents in the marriage market anticipate the bargaining environment within marriage. We can expect outcomes in these two arenas for male—female contracting to be closely related. A change in the bargaining environment within marriage (e.g. a change in tax policy, relative wages or social norms) can not only affect distribution within existing marriages but can also alter subsequent marriage market equilibria; under some circumstances, these marriage market effects can entirely undo the effects of the initial redistribution. We are far from a unified model of marriage, divorce and marital behavior, but a model of distribution within marriage that recognizes the independent agency of men and women within marriage is a prerequisite to a unified model. ACKNOWLEDGMENTS This paper was published in Journal of Economic Perspectives, Fall 1996. We thank the Journal of Economic Perspectives for permission to reprint the paper here. An earlier version of the paper was presented at the Arne Ryde Symposium. The authors wish to thank participants at the symposium, Elaina Rose, Dick Startz and the editors of Journal of Economic Perspectives for comments. REFERENCES Becker, Gary (1974) “A Theory of Social Interactions,” Journal of Political Economy 82, 6:1063–94. —— (1981; enlarged edition 1991) A Treatise on the Family, Cambridge, Mass.: Harvard University Press. Behrman, Jere R. (1996) “Intrahousehold Distribution and the Family,” in Mark R. Rosenzweig and Oded Stark (eds) Handbook of Population and Family Economics, Amsterdam: North-Holland Publishing Company. Behrman, Jere R., Pollak, Robert A. and Taubman, Paul (1995) From Parent to Child: Inequality and Immobility in the United States, Chicago: University of Chicago Press.
30
SHELLY LUNDBERG AND ROBERT A.POLLAK
Bergstrom, Theodore C. (1996) “A Survey of Theories of the Family,” in Mark R. Rosenzweig and Oded Stark (eds) Handbook of Population and Family Economics, Amsterdam: North-Holland Publishing Company. Bergstrom, Theodore C., Blume, Lawrence and Varian, Hal (1986) “On the Private Provision of Public Goods,” Journal of Public Economics 29, 1:25–49. Blumberg, Rae Lesser (ed.) (1991) “Income Under Female Versus Male Control: Hypothesis from a Theory of Gender Stratification and Data from the Third World,” in Gender, Family, and Economy: The Triple Overlap, Newbury Park: Sage. Bourguignon, Francois, Browning, Martin, Chiappori, Pierre-André and Lechene, Valerie (1993) “Intra Household Allocation of Consumption: A Model and Some Evidence from French Data,” Annales d’Economie et de Statistique 29:138–56. Browning, Martin, Bourguignon, Francois, Chiappori, Pierre-André and Lechene, Valerie (1994) “Income and Outcomes: A Structural Model of Intrahousehold Allocution,” Journal of Political Economy 102, 6:1067–96. Bruce, Judith, Lloyd, Cynthia B. and Leonard, Ann (1995) Families in Focus: New Perspectives on Mothers, Fathers, and Children, New York: Population Council. Chiappori, Pierre-André (1988) “Rational Household Labor Supply,” Econometrica 56, 1:63–89. —— (1992) “Collective Labor Supply and Welfare,” Journal of Political Economy 100, 3:437–67. England, Paula and Stanek Kilbourne, Barbara (1990) “Markets, Marriages, and Other Mates: The Problem of Power,” in Roger Friedland and A.F.Robertson (eds) Beyond the Marketplace: Rethinking Economy and Society, New York: Aldine de Gruyter. Grossbard-Shechtman, Amyra Shoshana (1993) On the Economics of Marriage—A Theory of Marriage, Labor and Divorce, Boulder: Westview Press. Haddad, Lawrence and Kanbur, Ravi (1990) “How Serious is the Neglect of Intrahousehold Inequality?,” Economic Journal 100:866–81. Hoddinott, John and Haddad, Lawrence (1995) “Does Female Income Share Influence Household Expenditure? Evidence from Cote d’Ivoire,” Oxford Bulletin of Economics and Statistics 57, 1:77–95. House of Commons (1980) Hansard January. Kreps, David M. (1990) Game Theory and Economic Modelling, Oxford: Oxford University Press. Lundberg, Shelly and Pollak, Robert A. (1993) “Separate Spheres Bargaining and the Marriage Market,” Journal of Political Economy 101, 6:988–1010. Lundberg, Shelly and Pollak, Robert A. (1994) “Noncooperative Bargaining Models of Marriage,” American Economic Review Papers and Proceedings 84, 2:132–7. Lundberg, Shelly, Pollak, Robert A. and Wales, Terence J. (1997) “Do Husbands and Wives Pool Their Resources? Evidence from the U.K.Child Benefit,” Journal of Human Resources 32, 3, Summer: 463–80. McElroy, Marjorie B. (1981) “Appendix: Empirical Results from Estimates of Joint Labor Supply Functions of Husbands and Wives,” in R.G.Ehrenberg (ed.) Research in Labor Economics 4, Greenwich, Conn.: JAI Press. —— (1990) “The Empirical Content of Nash-Bargained Household Behavior Journal of Human Resources 25, 4: 559–83. McElroy, Marjorie B. and Homey, Mary Jean (1981) “Nash Bargained Household Decisions,” International Economic Review 22, 2:333–49. Manser, Marilyn and Brown, Murray (1980) “Marriage and Household Decision Making: A Bargaining Analysis,” International Economic Review 21, 1:31–44. Nash, John F. (1950) “The Bargaining Problem,” Econometrica 18, 1:155–62. Pahl, Jan (1983) “The Allocation of Money and the Structuring of Inequality within Marriage,” Sociological Review 31: 237–62. Phipps, Shelley and Burton, Peter (1992) “What’s Mine is Yours? The Influence of Male and Female Incomes on Patterns of Household Expenditure,” Working Paper 92–12, Department of Economics, Dalhousie University. Pollak, Robert A. (1994) “Taking Power Seriously,” mimeo, University of Washington. Rose, Elaina (1994) “Consumption Smoothing and Excess Female Mortality in Rural India,” mimeo, University of Washington.
BARGAINING AND DISTRIBUTION IN MARRIAGE
31
Samuelson, Paul A. (1956) “Social Indifference Curves,” Quarterly Journal of Economics 70, 1:1–22. Schultz, T.Paul (1990) “Testing the Neoclassical Model of Family Labor Supply and Fertility,” Journal of Human Resources 25, 4:599–634. Sen, Amartya K. (1990) “Gender and Cooperative Conflicts,” in Irene Tinker (ed.) Persistent Inequalities: Women and World Development, New York: Oxford University Press. Thomas, Duncan (1990) “Intra-Household Resource Allocation: An Inferential Approach,” Journal of Human Resources 25, 4:635–64. —— (1994) “Like Father, Like Son: Like Mother, Like Daughter: Parental Resources and Child Height,” Journal of Human Resources 29, 4:950–88. Treas, Judith (1993) “Money in the Bank: Transaction Costs and the Economic Organization of Marriage,” American Sociological Review 58:723–34. Udry, Christopher (1996) “Gender, Agricultural Production and the Theory of the Household,” Journal of Political Economy 104, 5:1010–46. Warr, Peter G. (1983) “The Private Provision of a Public Good Is Independent of the Distribution of Income,” Economic Letters 13, 2:207–11. Weiss, Yoram (1994; revised) “The Formation and Dissolution of Families: Why Marry? Who Marries Whom? And What Happens Upon Divorce?,” Working Paper 15–93, Foerder Institute for Economic Research, Tel-Aviv University. World Bank (1995) Toward Gender Equality: The Role of Public Policy, Washington DC: World Bank. Zelizer, Viviana A. (1989) “The Social Meaning of Money: ‘Special Monies’,” American Journal of Sociology 95, 2: 342–77. —— (1994) The Social Meaning of Money, New York: Basic Books.
NOTES 1 Those interested in a more technical review of theories of the family should refer to Bergstrom (1996). Weiss (1994) provides an extensive review of models of marriage and divorce and Behrman (1996) of the empirical literature on intra-household distribution. 2 For our present purposes, we interpret “families” broadly to encompass all types of multi-person households, though some of our discussions of bargaining models of marriage emphasize the legal institutions surrounding marriage and divorce. 3 The allocation of time provides a set of family demands that are more readily assigned to individuals than is the consumption of goods and services. If leisure is assumed to be the only alternative to market work, we can assume that it is privately consumed, and standard cross-section and panel data sources report the relevant prices and quantities. For many years, the analysis of male labor supply proceeded on the basis of a single-agent model with researchers expressing few qualms about ignoring household interdependencies. When serious study of female labor supply began in the 1960s, however, the limitations of this approach became apparent. In a model of the labor force participation of married women, both leisure and time spent in home production—cooking, cleaning, childcare— are alternatives to market work. Since home production yields a variety of goods that are consumed by others in the household, the interest of other family members in the time allocation of married women is difficult to ignore. 4 More explicitly, suppose that the husband, h, has an individual utility function that depends on his private consumption of m goods, and the wife, w, has an individual utility function . If they agree to maximize a consensus social welfare function of the form W[Uh, Uw], then we can analyze their aggregate expenditure pattern as though the family were a single agent maximizing a utility function of the form U(x1,…, xm) where , subject to the joint budget constraint
32
SHELLY LUNDBERG AND ROBERT A.POLLAK
5
6
7 8 9 10
11
12
13 14
15 16
that pools the income received by the two family members. This optimization problem generates family demands . If the utility functions are well-behaved, these demand functions are homogeneous of degree zero in prices and total family income, and the implied Slutsky matrix of compensated cross-price effects is symmetric and negative semi-definite. Another restriction implied by the common preference model is the symmetry of compensated cross-price effects. McElroy (1981) surveys the evidence provided by demand systems estimation, and concludes that it provides little support for such symmetry. Although more recent work has not altered this conclusion, statistical rejections of symmetry do not appear to have shaken anyone’s faith in common preference models: rejections of symmetry can always be attributed to functional form misspecification or improper aggregation of goods. Expenditures on a purely private good are not required to test pooling. If husband and wife have different preferences, some bargaining models suggest that the marginal propensity to spend the husband’s income on a public good may differ from the marginal propensity to spend the wife’s income on the same good. Increases in expenditure on women’s clothing are also found by Browning et al. (1994) and by Lundberg et al. (1997). Behrman et al. (1995) express reservations about interpreting the results of Thomas (1990) or Schultz (1990) as conclusive rejections of pooling on these grounds. Annual expenditures on children’s clothing rose about £50, and expenditures on women’s clothing about £30, in an average two-child family receiving a child allowance of £400. The Nash bargaining solution is the allocation that maximizes the product of the gains to cooperation, given by the function: subject to the constraint that the family’s joint income equal joint expenditure, . The control of resources among the potential contributors to a public good in a voluntary provision model affects neither the equilibrium level of the public good nor the equilibrium utility levels of the potential contributors, provided that each makes a strictly positive contribution. This neutrality result is well-known in public finance; see Warr (1983) and Bergstrom et al. (1986). More explicitly, the husband treats, , the level of public good chosen by his wife, as fixed and chooses quantities of his private good and his assigned public good so as to maximize his utility subject to his budget constraint . Similarly, the wife treats the quantity of the public good supplied by her husband as fixed and chooses utility-maximizing quantities of her private and public good subject to her budget constraint. The separate spheres threat point is determined by the husband’s and wife’s utilities in this noncooperative equilibrium, and can be written as: . Since the demand functions generated by cooperative bargaining will depend upon the threat point, they will be of the form: . Lundberg and Pollak (1994) analyze distribution within marriage as a repeated non-cooperative game. The canonical battle of the sexes story relies heavily on gender stereotypes. Both the husband and wife want to spend the evening together, but the husband wants to go to a sporting event (e.g. a prize fight) and the wife to a cultural event (e.g. a ballet). The story is used to motivate a non-cooperative non-zero sum game in which the Pareto-optimal outcomes correspond to successful coordination (i.e. both go to the prize fight or both go to the ballet) and are Nash equilibria. The formal structure of the game provides no way to choose between them. For example, Grossbard-Shechtman (1993) analyzes marital distribution assuming that the marriage market determines a “wage” for spousal labor that is binding during the marriage. The divorce-threat bargaining model goes beyond the notion of divorce bounds to make the cooperative equilibrium depend explicitly on the value of divorce.
2 BATTLES OF THE SEXES: NON-COOPERATIVE GAMES IN THE THEORY OF THE FAMILY Kjell Erik Lommerud
INTRODUCTION Family economists commonly assume that decisions taken within a family are Pareto-efficient. For instance, in much of his seminal work on the family, Becker (1991) assumes that the family is able to reach efficient outcomes, and makes no attempt to describe the process through which this is achieved. Also Gronau’s (1973) well-known work on the time allocation in families is an efficiency model; a more recent example is Chiappori (1992). Some efficiency models of the family postulate that the family maximizes a joint welfare function over the family members’ utilities. This can be seen, though, as a trick to generate efficient outcomes, rather than as an attempt to describe actual family behavior. The efficiency assumption is sometimes referred to as an assumption of complete contracts: a sufficient condition for achieving efficiency is that there are no limitations whatsoever on the type of binding contract that can be written and enforced. One should be careful about interpreting efficiency models of the family as models of harmony. In an efficiency model the intrafamily distribution of resources may be very uneven, and this could be a source of conflict. Rather, the assumption of the efficiency models is not that there are no distributional fights within the family, but that such distributional conflicts do not prevent efficiency from being realized. Crudely put, an efficient family can be seen as one which first agrees on the division of labor that maximizes the resources available to the family, and then—perhaps—fights over the division of these resources.1 It should also be stressed that when one talks about efficient family decisions, one means that no inefficiency arises in the family decision process itself. If, for instance, wages in the outside labor market give incorrect signals about productivity, perhaps, for example, because there is discrimination against women, the division of labor decided upon within the family will be inefficient: this is a derived inefficiency, however, that does not have its primary cause inside the family. Around 1980 several authors argued that Nash bargaining models would be a way to introduce distribution concerns into economic models of the family. (Key references are Manser and Brown 1980 and McElroy and Horney 1981.) However, the difference between a Nash bargaining model and a welfare maximization model of a family is small. Nash bargaining models also assume that decisions will be efficient. The Nash maximand is a Cobb–Douglas welfare function over Von-Neumann–Morgenstern utilities. In the so-called generalized Nash bargaining solution, welfare-weight-like parameters are referred to as “bargaining power” parameters. Does this similarity between a welfare maximization model and a Nash bargaining model imply that the Nash bargaining model is empirically void? (See the exchange
34
NON-COOPERATIVE GAMES IN THE FAMILY
between Chiappori 1988, 1991 and McElroy and Horney 1990 and McElroy 1990.) One difference between a Nash bargaining model and a welfare maximization model is that under Nash bargaining the “threat points” of the parties influence the distribution of resources whereas this is not the case under welfare maximization. Here lies a possible route to distinguish the two models empirically—a point made by McElroy (1990). However, what should be treated as the threat point in a family bargain is quite ambiguous. In line with Binmore’s et al. (1986) non-cooperative gametheoretic underpinning of the Nash solution, it is often argued that the proper threat point is the utilities of the parties during a conflict when an agreement has not yet been struck. For example, in labor economics it has now become customary to use utility and profit during a strike as the threat point in models of labor disputes, rather than for example using the outside opportunity wage as the fall-back for workers. Correspondingly, when a family cannot reach agreement on a given decision, immediate divorce is not always the consequence. This casts doubt on the validity of using utilities as single as the threat point in family bargaining (as McElroy 1990 does). But what corresponds to a strike in a family? We will return to this question later in the paper. The use of efficiency models is extremely common in family economics. And since families are long-run projects and the members hopefully tend to be sympathetic towards each other, this is perhaps not a hopeless assumption. However, as a contrast to this ruling tradition, some authors have suggested that it might be worthwhile to use non-cooperative models in the study of the family: thus efficiency is not assumed from the start. As a by-product, of considerable value in itself, non-cooperative game theory also forces the modeler to specify, for example, the timing of events, the action spaces and the possibilities to enter into contracts, in a more rigorous way than has been usual in the received literature. Family economists in the Becker tradition wage war on many fronts. Some, both inside and outside economics, accuse family economists of placing too much reliance on the assumption of rationality, thus giving a too cynical picture of how families function. Others maintain that family economists tend to give a too consensus-oriented description of family life. Non-cooperative family models meet the latter criticism— these types of models can be set up, for example, to allow for women being oppressed in the family. However, models of non-cooperation arguably draw even more heavily on the assumption that individuals are rational —to be “rational” means that individuals will calculate more complicated strategies than in simpler efficiency models. Nelson (1995) calls both Becker-type models and subsequent game-theoretic formulations in family economics “something of a double-edged sword to feminists.” I agree with this statement. According to Bergmann (1987) the Becker type of family model “explains, justifies, and even glorifies role differentiation by sex.” This type of criticism is harder to apply to non-cooperative models. Broadly speaking, game-theoretic models of the family can be divided into two categories. First we have models of “incomplete contracts” or “transactions costs.” These models postulate that the family in many respects can make binding agreements—but that these possibilities are not complete. One then imposes some exogenous limit on the type of family agreement which can be entered into. One example of this approach is Lommerud (1989). There it is assumed that love and altruism bind spouses in an implicit family agreement as long as they remain married; after a divorce, non-cooperative behavior follows. A seminal contribution in transaction-cost family economics is Pollak (1985). Other references include King (1982), Cohen (1987) and Allen (1990). Incomplete contracts models of the family can be accused of being ad hoc in the sense that the assumption about which contracts can be entered into, and which not, is quite arbitrary. An alternative model strategy is to assume full non-cooperation in the family. This at least brings out the clearest contrast to the efficiency models and can also be a benchmark for future research in the incomplete contracts vein.
KJELL ERIK LOMMERUD
35
For the most part this paper will focus on fully non-cooperative models. It should be stressed that such models, at the best, picture life in families who are on or beyond the brink of divorce—but I hope that they can provide a fruitful starting-point for later research that combines cooperative and non-cooperative elements. There are a few non-cooperative family models in the literature. In this paper we will in no way try to give a full overview of these models. I should admit at once that I emphasize my own work in a quite biased way. Gametheoretic models of behavior in organizations—with the family as one example of an organization—fall broadly into two categories. First, there are models of conflicting objectives. Second, there are situations where goals do not necessarily conflict, but where the parties when operating noncooperatively have problems in coordinating their actions. In this paper we concentrate on the first class of models. First we will look at situations where there are family public goods, and family members try to free-ride on the public-good provision of others. This type of game is a simultaneous moves game, often used to study behavior in a non-cooperative nuclear family. Alternatively, one can operate with the Stackelberg assumption that players choose their actions in a prespecified order. Free-rider problems then translate into incentive problems: one player tries to steer the actions of another. This type of structure is often used to study relationships between family members of different generations. Coordination-problems models are potentially very interesting in family economics. There are large potential benefits if married couples can coordinate their career choices, but this is difficult to do. This means we can have self-fulfilling expectations—where women are expected to bear the main responsibility for the children, which discourages investments in career development both by the female herself and by employers, which in turn implies that it is rational that the female actually does bear the main responsibility for children. We will not discuss this class of models in this paper, primarily because this type of family economic theory is still not very well developed. Some starting-points can be found in Coate and Loury (1993) and Lommerud and Vagstad (1996). Since the title of this paper is “Battles of the Sexes,” we note that the original Battle-of-the-Sexes game is indeed a game of coordination. The paper is organized as follows. First we look at private-provision-of-public-goods games. These games share an assumption that family members non-cooperatively use their resources either to acquire a private good or a family-specific good. What exactly constitutes the “private good” and the “public good” will be seen to vary from model to model. The next section suggests that the type of non-cooperative model presented earlier can be used to reformulate the Nash bargaining theory of the household. Then we take a brief look at intergenerational games. These games, known under buzzwords as “the rotten kid theorem,” “the Samaritan’s dilemma” and “the strategic bequest motive,” are doubtless the best-known examples of non-cooperative game models in family theory. The emphasis will be on the close structural links between this type of intergenerational models and the private-provision-of-public-goods models. PRIVATE PROVISION OF PUBLIC GOODS—IN THE FAMILY In the field of public economics a fast-growing body of research looks at individual incentives to provide public goods. Among the seminal references are Warr (1982, 1983) and Bergstrom et al. (1986). An important point in this literature is that even though free-rider problems usually lead to the underprovision of a public good, this only means that agents’ incentives to provide a public good are weak, not non-existent, which previously was a standard assumption in public economics. Especially in small groups the voluntary contribution of a group member towards the public good might be quite substantial. This literature focuses on investigating the effects of public policies that aim at redistribution between the members of the group
36
NON-COOPERATIVE GAMES IN THE FAMILY
and at increasing the total provision of the public good. As we shall see, a typical result is that public policies tend to be powerless, since changes in private provisions to the public good will counteract the effect of policy. The family seems to be a natural example of a small group setting where private provision of public goods actually takes place even in the absence of binding contracts. A small amount of literature on noncooperative provision of public goods within the family does exist. To some extent this literature draws on findings in public economics, but it has also arrived at results that are novel relative to the public economics literature. These models of non-cooperative families share the description of family members as allocating their resources between a private good and a family public good. Definitions of what exactly constitutes the private and the public good vary considerably, however. For the most part we will concentrate on a set of models that, in one way or the other, assumes that agents can use their time either to work in the labor market to produce private benefits (for example private money) or to work at home to produce a family-specific public good (care of children, a tidy home, a beautiful garden). The presentation follows Konrad and Lommerud (1995). Other work will be commented on later. The Konrad-Lommerud model We study a model that is fully non-cooperative. Each spouse determines independently and simultaneously his or her time allocation, and there is no transfer of money between the parties. Time can be used either to work in the outside labor market or to produce a family-specific good, for example childcare. Money earned in the labor market is seen as a pure private good, the home-produced good is seen as a pure public good. In reality most goods consumed in a family are more or less impure public goods. The present “sharp” assumption is made to reach sharper conclusions. Both spouses have a utility function u(xi, G), where xi denotes quantity of the private good, and G denotes total quantity of the public good.2 Further, wi is the wage rate in the labor market and hi is the productivity at home (number of units produced per period of time), with subscript i referring either to individual a or b. Under some assumptions it can be shown that a unique Nash equilibrium in time allocations exists, and that an interior private provision equilibrium is characterized by (2.1) Subscripts G and x denote partial derivatives. These conditions simply say that any agent will choose his or her allocation of time—taking the allocation of time of the other as given—such that the marginal rate of substitution between the two goods in question equals something which can be interpreted as the relative price of the two goods. More precisely, wi/hi can be interpreted as the forgone earnings cost of producing one unit of the public good. The difference between this model and the standard voluntary contributions model of Warr (1982, 1983) and Bergstrom et al. (1986) is that in those models one assumes that the contribution productivity (here: wi/ hi) is equal for all, but that agents differ by having different levels of monetary wealth. In that framework some well-known results follow. With identical and interpersonally comparable utility functions, and when the equilibrium is and remains an interior one, the richer you are, the more you contribute to the public good. In fact, in an interior equilibrium people will contribute to the extent that their remaining private income is equalized. Public good consumption is by necessity the same. We therefore see that participation in private-provision-of-public-goods games is strongly redistributing, in spite of no actual redistribution of
KJELL ERIK LOMMERUD
37
money among individuals taking place. Transfers of money among members of the contributing group are neutral in the sense that if a gives $ 1 to b this will only lead b to contribute $ 1 more to the public good, and a will contribute $ 1 less. Both total contributions and both parties’ remaining income therefore are unchanged. We will refer to this result as Warr neutrality. A closely related result states that public provision of the public good crowds out private contribution in a 1:1 ratio if the public provision is financed by lump-sum taxes levied on the contributing parties. Let us assume that a government taxes $ 1 away from individual a in a non-distortionary manner. The proceeds from the tax are used to provide $ 1 worth of public good. If a now contributes $ 1 less, he can in fact choose to return to the situation before government intervention. His income after contribution to the public good will be unchanged, as will be the total amount of the public good. If the pre-intervention adaptation was privately optimal, it will be optimal to restore the same situation. This crowding-out result is a variant of Warr neutrality. The government can be seen as one of the contributors in a provision-of-a-public-good game, and transfers among the contributing parties, including transfers from an individual to the state, will then be neutral. Let us return to our specified family model. It differs from the standard model in that it allows contribution productivities to differ. This extension is quite natural for a family economist, since for decades a key focus in family economics has been the effect of comparative advantages on the allocation of time between household work and market work. A family’s allocation of time, as characterized by (2.1), will be determined both by comparative advantage and absolute advantage. Assume first that the male (a) is relatively more productive in the labor market, and the female (b) in household production, but in a way where the total value of their abilities is the “same.”3 There will then be specialization according to comparative advantage, as in an efficiency model, but the specialization will be less pronounced. In Nash equilibrium the female contributes more than the male to the public good, simply because provision is cheaper for her. However, specialization occurs at her own expense in the sense that when she works more at home, it is her own wage income that suffers. She is not compensated for this through transfers of income from the male. If one is willing to assume that the common utility function is interpersonally comparable, the distribution in the marriage can be seen to be uneven: the wife enjoys—by necessity —the same amount of the public good, but has lower wage income, and therefore lower utility. In a model such as this, as opposed to an efficiency model, absolute advantages also help determine the allocation of time. Assume now that the female is more able both in market work and in household production, but in such a way that wi/hi is the same for both spouses (i.e. there are no comparative advantages). Also in such a situation the female will contribute more towards the public good. This reflects the standard result that in voluntary contributions games the richer party contributes more, and one way to be rich is simply to be very productive overall. Family policy Let us now investigate how some common family policy measures work in this framework. We will look at the public provision of a perfect substitute to the family public good, at redistribution within marriage from the male to the female, and also at the effect of income taxation. Naturally, the results will be related to Warr neutrality, but the neutrality results obtained in a model with equal contribution productivities do not hold strictly when contribution productivities differ.
38
NON-COOPERATIVE GAMES IN THE FAMILY
Government provision of a close substitute to a family public good, with public provision of childcare as a prime example, is often thought of as an allocative policy, in the sense that it is meant to correct for some inefficiency. Here, however, there will be a strong tendency for such government provision free of charge, combined with lump-sum taxation of the contributing parties, to crowd out private contributions. Whether the crowding out is more than, less than or exactly equal to 100 percent, depends on how the contribution productivity of the government compares with that of the party who bears the tax. Government provision will increase the total amount of the public good only if the government has a cost advantage in providing the good, or if the arrangement means that the family is subsidized by other groups in society, such as singles and the elderly, or if government provision is taken beyond the point where private contributions are already crowded out. The driving force behind the crowding-out result is the unobservability of the contributions of family members.4 Had these contributions been observable, the government could have made its own contribution contingent on private contributions not being reduced. The most important family public good is probably childcare. Childcare has a quantity-quality aspect: under-provision of childcare can mean having too few children or spending too little time and resources on a given number of children. The unobservability assumption fits the quality dimension, but hardly the quantity aspect. If one wants to increase the number of children people have, a per child subsidy can indeed do the job. However, if a government cares both about the number of children and the amount of time spent on each child, it should be noted that a policy that encourages people along the observable dimension (they have more children) can have disadvantageous effects along the unobservable one (they spend less time on each child). It should be noted that government provision of the family public good may have beneficial distribution effects. Assume for the sake of illustration that the government has the same contribution productivity as the woman, whom we assume to be the least-cost provider within the family. But even though the female provides childcare more efficiently than her husband, we assume that she has the same outside wage opportunity as the husband. Assume further that government provision of day care is so extensive that it exactly crowds out private contributions, and that the lump-sum tax to finance the contribution is split evenly between the spouses. Distribution in marriage then becomes equal. Remember that in the equilibrium without government provision the wife, because she is more productive in providing childcare, is the one who worked more at home—and since she is not compensated for her forfeited earnings, she suffers from a distribution point of view. In this particular example, it can be shown that the total provision of the public good increases somewhat.5,6 Next, let us turn to policies explicitly aimed at changing the distribution in the family to the female’s benefit. For example, in several countries child allowances are paid to mothers, and not to families as entities. We represent such a policy by studying a lump-sum redistribution from b to a. For the case with equal contribution productivities and where the equilibrium is and remains an interior one, Warr’s neutrality result will apply. But what happens when contribution productivities are allowed to differ? In such a case, it is still true that such a transfer is a bad instrument for redistribution. The fact that the man gets poorer and the woman richer will tend to make the man contribute less and the woman more to the public good, which undoes the redistribution effect of the transfer. However, the transfer constitutes a Pareto improvement, since the provision of the public good is shifted to the individual with the higher productivity. And this time the transfer means that the woman is compensated for her increased public-good provision. In fact, our result implies that transfers should at least be taken to a point where the equilibrium ceases to be an interior one.
KJELL ERIK LOMMERUD
39
To sum up so far, we are left with the somewhat paradoxical conclusion that an “efficiency tool” such as government provision of day care seems to have its limitations as a means to correct an undersupply of family public goods, but that this policy can be very advantageous from an intrafamily distribution point of view. On the other hand, a “distribution tool” such as the transfer of money from the man to the woman does little to improve the intrafamily distribution, but does, in fact, improve efficiency. Since we work within a framework with comparative advantages, improving efficiency here means that the person with comparative advantage in household production will be led by the transfers to take over a larger share of the household production. As long as transfers are Pareto-improving, this means that a husband voluntarily and with no binding contract would be willing to transfer money to his wife, just because he knows how this will affect her equilibrium behavior as regards the provision of family public goods. However, if staying away from the labor market is a long-term decision, this means that the husband should transfer quite a substantial amount to the wife at the outset of marriage. Liquidity constraints may keep him from doing this. The state, though, is more likely to be able to commit itself to transferring a given sum from the male to the female each year. Liquidity constraints would thus play less of a part, explaining why we have concentrated on a transfer system implemented by the government. We will also mention briefly that the normal system of income taxation can be evaluated within a framework like the present one. It is a standard claim that progressive income taxation leads to inefficiencies in agents’ use of time. Here the crucial inefficiency is that people try to escape household production, with market work as an important alternative use of time. “Distortionary” income taxation can therefore improve efficiency—and it can be shown that under some assumptions a marginal increase in the degree of progressivity can lead at the same time to increased production of the public good, increased activity by the male in household production, and more female market work. This is, however, a typical second-best result: driving the husband back to family production is a Pareto improvement, but the global optimum is reached only when the female (with comparative advantage in household work) specializes in household production and is properly compensated for it. Other family models of private-provision-of-public-goods So far we have followed the presentation of Konrad and Lommerud (1995). Lundberg and Pollak’s (1993) work is a parallel contribution.7 One difference is that Lundberg and Pollak exogenously impose a “separate spheres” assumption, meaning that the male provides only one type of family public good and the female another. This seems unnecessarily restrictive, but the assumption is made because (as is well-known) Warr neutrality ceases to hold in corner solutions. Our paper also puts more emphasis on family policy. Bragstad (1991) looks at a situation where the spouses are different because they care differently about the public good. This is of course parallel to having different contribution productivities. Weiss and Willis (1985) tailor their model to fit a divorced couple.8 One of the parties has custody, the other does not. The non-custodian parent must decide whether or not to transfer money to the parent who has custody. The custodian parent decides independently the amount of resources to devote to the child. In a way, this is an extreme case of different contribution productivities, since the parent not living with the child cannot contribute directly to the child’s welfare at all. We should also mention that there are “voluntary contributions”-type models with a different specification of what constitutes the private and the public good. The seminal contribution is Leuthold (1968). Her focus is on public transfers to the poor and their effect on labor supply. Each family member
40
NON-COOPERATIVE GAMES IN THE FAMILY
chooses independently his or her own labor supply. Leisure is a private good, but money is seen as a family public good. Kooreman and Kapteyn (1990) suggest that a Leuthold-type model can be used as fall-back in household bargaining, and use this idea in empirical work. In their empirical application Kooreman and Kapteyn associate “leisure” with all time spent outside the labor market. It then becomes clear that the Leuthold model in a way is the Konrad-Lommerud model turned upside-down: here it is production at home that is the private good and market activity that produces the public good. The fact that Dutch women tend to have a loose attachment to the labor market is thus interpreted as their having a strong bargaining position. Ulph (1988) and Woolley (1988) study a family’s demand for goods when some goods are private for one family member, and other goods in varying degrees are public goods. An issue of interest also in such a framework is how earmarked transfers to one of the spouses alter the demand composition. THE BARGAINING FAMILY REVISITED Several writers on non-cooperative families claim that even if one does not believe in the fully noncooperative model as a true description of family behavior, non-cooperation is a natural threat in bargaining.9 This means that a better picture of non-cooperative behavior within the family is needed even for those who believe that a family is best depicted as an efficiently bargaining unit. The alternative to using non-cooperation as a bargaining lever is to use utilities as single, as for example in McElroy (1990). As mentioned in my introduction, this alternative has been criticized, and the utilities under non-cooperation intuitively seem better suited to represent “utilities during conflict.” Konrad and Lommerud (1996a) suggest a model with the following characteristics. The model is a twostage one. In the first stage, long-run, irreversible educational decisions are made. We assume that these choices are non-cooperative. In the second stage, the day-to-day allocation of time between market work and household production is decided upon. These decisions are assumed to be arrived at through Nash bargaining—but with non-cooperative behavior as the fall-back in bargaining. We should mention that in this model we work with a more specialized assumption about utility functions, namely that utility takes the form . The convex cost functions a(.) and b(.) are the costs of providing the public good and of acquiring education, respectively; gi is individual i’s contribution to the public good. A first result within this model, which combines cooperative and non-cooperative elements, is that the inefficiency may be even worse than in a fully non-cooperative model, where stage two decisions are also taken non-cooperatively. But the finding that we will focus on here is that using utilities under noncooperation as fall-back in bargaining is not very different from using utilities as single. For example, in both cases a key determinant of the distribution of resources within the family is the outside wage of the man and the woman, respectively. This warrants some intuitive explanation. Non-cooperative behavior, as we have described it, can be regarded as “internal divorce.” Transfers of money and coordination of time use cease. The parties are still linked, though, through the mutual concern for family public goods, such as children, but this is most likely also the case after a real divorce. The difference between “internal divorce” and a real divorce is that in the latter case the spouses stop living together —and this may change their valuation of the public goods. If living together or not does not change the value of the public good—people continue to love their children regardless of living arrangements—then at this level of abstraction it is difficult to spot any differences between non-cooperation and divorce at all. It should then come as no surprise that both under non-cooperation and divorce, one’s individual utility is higher the higher one’s outside wage—and consequently, the better one’s bargaining position is. If divorce
KJELL ERIK LOMMERUD
41
implies an even greater reduction in time spent in household production than does non-cooperation, there will be a difference in degree between the two cases: the outside wage will influence the distribution of resources even more when utilities as single rather than non-cooperative utilities are used as the threat. It can be argued that when family bargainers cannot reach agreement, the result will be partial rather than full non-cooperation. Some rudimentary contracts, perhaps about the sharing of financial resources, will still be honored. This may very well be true, but we still conjecture that the difference between noncooperation and divorce is one of degree rather than a qualitative one: thus, empirical work that investigates the impact of the outside wage on the allocation of resources within the family will continue to have a theoretical justification. INTERGENERATIONAL ISSUES Many of the most famous attempts to use game-theoretical formulations in family economics fall within the heading “intergenerational issues.” In spite of the title of this article, which suggests that we will concentrate on the economics of the nuclear family, we will briefly mention here some examples of “battles of the generations.” There are important structural similarities between game situations in the nuclear family and intergenerational games. Models of parent–child relationships often include an assumption about altruism within the family—but probably altruism should be included in all types of family models. Specific to the intergenerational setting is the frequent assumption that altruism is one-sided: the parent cares for the child, the child is selfish. Another difference between models of marriage and models of parent–child relationships is that in models of married couples it is most often assumed that actions occur simultaneously. In intergenerational games, the structure is often a Stackelberg one, where the parties make their moves in a prespecified order. The rotten kid theorem The “rotten kid theorem” comes from Becker (1974). Later research includes Bergstrom (1989) and Bruce and Waldman (1990). The term “rotten kid” refers to the mentioned assumption that altruism in the family is one-sided. Becker’s finding is that if a family has a head who “cares sufficiently about all other members to transfer general resources to them, then redistribution of income among members [of the household] would not affect the consumption of any member, as long as the head continues to contribute to all” (Becker 1974:1076). He continues, “The major, and somewhat unexpected, conclusion is that if a head exists, other members also are motivated to maximize family income and consumption, even if their welfare depends on their own consumption alone” (Becker 1974:1080). Here Becker does not start by assuming efficiency, but shows that for a given decision procedure efficiency is nevertheless achieved. Although the result in the rotten kid theorem is rather surprising, it is not difficult to understand. The altruistic head has a preference ordering of the consumption levels of all family members. He is rich enough so that on top of the income of each family member he always adds an additional grant that determines the member’s final income. So if one of the “kids” manages to get hold of additional resources, the head can always confiscate these by reducing his grant correspondingly, and then redistribute this additional income among family members as he pleases. Each kid’s consumption is uniquely determined by the amount of total resources the head has for distribution, so the best the selfish kid can do is to maximize family income. The rotten kid theorem is a neutrality result closely related to many other neutrality results. Let us start by looking at its relationship to Warr neutrality. Altruism makes the utility of the kid a public good for the kid
42
NON-COOPERATIVE GAMES IN THE FAMILY
and the head. If the kid manages to transfer $ 1 from the head to himself, the head, content with the original amount of public good (here: the kid’s utility level), reduces his contribution equally much, and nothing will be changed. Whether this story is an example of Warr neutrality or of the rotten kid theorem at play is hard to determine. The rotten kid theorem, and therefore also Warr neutrality, is very close to Barro’s (1974) concept of Ricardian equivalence. Ricardian equivalence means that a government policy of giving an altruistic family head tax relief financed by debt which in the end must be paid by the “kids” of this family head, will have no effect. The government policy is just an attempt to redistribute within the altruistic family. The head can undo the transfer by saving the tax reduction, and handing the money over to the generation which in the end will have to pay the bill. The relationship to the two above concepts should be immediate. The rotten kid theorem, if it is generally correct, is a very striking result. It means that under certain circumstances all non-altruistic “kids” will automatically behave in a way which maximizes the welfare function used by the head, without the head in any way using incentive schemes or thinking strategically. All team problems go away if the head is sufficiently altruistic and rich. We now turn to discussing the socalled Samaritan’s dilemma. The Samaritan’s dilemma also deals with how one-sided altruism affects the need for incentive schemes—and reaches a conclusion more or less totally opposite to the rotten kid theorem. A comparison of the assumptions underlying the two results implicitly provides a discussion of how general the rotten kid theorem is. The Samaritan’s dilemma The term “the Samaritan’s dilemma” was coined by Buchanan (1975). It refers to the fact that when an altruist wants to help somebody, this aid can result in negative incentive effects: the recipient places himself in a more impoverished state than he would otherwise have done, just to get more help. Poverty can result from either the lack of savings or the lack of work effort, or both. Buchanan’s result emphasizes that one-sided altruism allows the altruist to be manipulated; the best strategy for the altruist is to commit to some incentive scheme, rather than yielding to his altruistic impulses. Becker’s result, as we have seen, is that the altruist should follow his altruistic impulses; as an effect of this, the object of this altruism contributes to the common good. What differences in assumptions lie behind the dramatic differences in conclusions? Bergstrom (1989) has investigated under what assumptions the rotten kid theorem would hold. Becker operated under the assumption that there is only one good in the economy. Bergstrom shows that the rotten kid theorem generalizes to a multi-good setting only under special assumptions. The Samaritan’s dilemma illustrates why the move to a two-good economy matters. Assume there are two goods, consumption today and consumption tomorrow. Assume that the transfer can take place tomorrow only. By not saving, one will have low consumption tomorrow and high marginal utility of money. Even if the Samaritan understands what is happening, his optimal choice can still be to give a high transfer. One way of viewing this is that the Samaritan is a victim of having too few transfer instruments; he cannot transfer utility directly, only money in the second period. In the version of the model where the two goods are lack of work effort and consumption, the Samaritan’s problem arises because he is limited to transferring consumption rather than utility.
KJELL ERIK LOMMERUD
43
The strategic bequest motive Strategic bequest behavior to elicit higher filial attention (Bernheim et al. 1985) is a further example of an altruistic family head who thinks strategically about incentives. The parent is altruistic towards the children, so he or she wants to leave his or her monetary wealth to them, but he or she also wants them to exert disutility of effort and come to pay a visit. The issue in Bernheim et al.’s paper is to what extent a threat to disinherit a child, in order to elicit more filial attention, is a credible one. Their initial assumption is that with only one child present, such commitment is difficult: the parent is altruistic and loves his or her child, so giving the money to strangers does not seem a credible alternative. However, it is argued, the situation is different with two or more children: the alternative to giving the inheritance to one loved child is then to give it to another, equally loved child. The parent therefore pits the two children against each other in an “auction.” This auction is most naturally seen as an all-pay auction, where the participants all pay their bidded amount regardless of who wins the auction. The children bid for the inheritance by offering amounts of attention they are willing to spend to get it. It is well-known from auction theory that having two rather than one bidder can tilt the outcome in favor of the owner of the object to be auctioned away quite dramatically. However, as Bernheim et al. perhaps do not underline sufficiently, it is also well-known from auction theory that even with two bidders there are circumstances where the “bids” become rather low. The commitment assumptions underlying the Bernheim et al. model are strong: the parent can state how much of the wealth the children are going to inherit in total, and also present an inheritance rule that specifies how more filial attention is to be rewarded. The children, on the other hand, cannot enter binding contracts between themselves, for instance promising not to pay any attention to the parent at all, and then dividing the inheritance afterwards. Another problem is that if the children were sufficiently asymmetric, for example as regards popularity with their parent or in “contribution productivity” in filial attention, the equilibrium outcome could easily be that one child gave up and the other secured the inheritance with a minimum of effort. Finally, it is not obvious that a parent who loves his or her children equally is indifferent to the distribution of money among them: if the parent is not, threatening to give all the money to a better-behaved brother or sister may again not be credible.10 WHAT’S LOVE GOT TO DO WITH IT? There is little talk of love in the economic theory of the family. Most economic models of the family can be changed into models of labor-managed firms or partnerships by slightly changing the interpretation of the variables. Can an elusive concept like “love” be introduced in family economics? Can one understand the concept of family at all without including that of love? Economics is defined as a field by referring either to topics that are fit to be studied by an economist or to the methodology used. As family economics is concerned with topics that until recently were thought to lie outside economics, it is tempting to agree with Gary Becker that economics should be distinguished from related social sciences by its methodology. The economic methodology centers around rigorous analysis of optimizing behavior. In my view, though, it is not an integral element of the economic methodology that the maximand of the optimizing agents is one’s own consumption of material goods. One can for instance introduce altruism into the analysis and still associate oneself with economics. In recent years it has not been uncommon to introduce elements such as altruism, envy, status-seeking and the like into utility functions. The argument is sometimes raised that by “fooling around with utility functions,” one can in reality “assume” oneself into obtaining any result one wants from the analysis. There is some element of truth in this. However, I believe that assuming that individuals are not altruistic, do not
44
NON-COOPERATIVE GAMES IN THE FAMILY
care for status, and so on, is as equally ad hoc as the opposite assumptions. If it is true that any and all conclusions can be reached by varying the assumptions about the utility function, let us then systematically investigate how predictions and policy conclusions are affected by different assumptions. Even if one allows non-traditional utility functions, it is not clear how love should be introduced into the analysis. In his recent book, Stark (1995) works with two different notions of altruism. One is the standard concept that one’s utility depends positively on the utility level of others. This is the assumption of many intergenerational games, and should perhaps also be used more in models of the nuclear family. But altruism and love surely mean more than this: for example, you may care about your own actions towards the object of love per se by keeping promises regardless of the utility consequences of not doing so. This is Stark’s second concept of altruism. In both descriptions of “altruism,” or “love,” it would appear that the scope for opportunism and free-riding is reduced. Models like the voluntary contributions games would therefore exaggerate the tendency to inefficient decision-making within the family. But it is a far step to assume that love solves all free-riding problems. And as Bernheim and Stark (1988) point out, one cannot even be sure that the degree of inefficiency in the family always decreases with the degree of altruism. One point made by these authors is that an altruist can have trouble in sufficiently punishing a less altruistic group member who deviates from the terms of an implicit contract. This is, of course, a continuation of the theme from the Samaritan’s dilemma literature. However, if one wants to work with mathematical models of optimizing behavior, I am afraid that the psychological picture of homo economicus must remain quite simplistic, with or without altruism included. Also in other economic models of small organizations—like a partnership—a broader description of human motivations could be needed. This means that even with altruism included, the economic theory of the family will continue closely to resemble a model of two dentists sharing office space. But it is not only a disadvantage to operate at a level of abstraction where the similarities between apparently disparate phenomena catch the eye more easily. For example, it may be enlightening to notice that the dilemma of a divorced father who wants to give his child something but can do so only via the mother indeed has something in common with the problem facing a donor of development aid who wants to reach the poor in a given country but can do so only by transferring resources through a local government dominated by the rich. CONCLUSIONS In this paper I have tried to review some models that introduce game-theoretical modeling to the economic theory of the family. Apart from the string of literature that deals with the relationship between an altruistic parent and a not-so-altruistic child, this is still a small and little noticed area. My personal belief is that we shall see more of these types of models in the years to come. One reason is simply that for young economists game theory has almost become another word for microeconomics—so game-theoretical formulations will naturally be used to attack any social phenomenon, including the family. Family economists will then undoubtedly encounter a phenomenon that is much discussed in contemporary industrial organization theory. Industrial organization was one of the applied subfields of economics that first utilized modern game theory. Initially this led to an explosion of interest in the field, with new results and confirmations of older hypotheses pouring forth. However, after some time a sense of saturation could be sensed, and a common criticism against game-theoretical modeling became that “everything could be explained” simply by slightly changing the timing of events or altering some other assumption.
KJELL ERIK LOMMERUD
45
When game-theoretical family economics matures, the same argument that “everything can be explained” will certainly appear here as well. Of course everything can be explained by the right choice of assumptions; to me the idea that some results should be very robust over all assumptions about timing of events, commitment possibilities and so on, seems very strange. For family economists, as for researchers in industrial organization before them, the conclusion must be that the true value of a game-theoretical model will only be realized when both assumptions and predictions meet empirical scrutiny. But since empirical work has its limitations as well, developing a whole array of models that span the different possible assumptions which can be made will at least give the modeler a good understanding of how different assumptions interact to produce given results. REFERENCES Allen, D. (1990) “An Inquiry into the State’s Role in Marriage,” Journal of Economic Behavior and Organization 13: 171–90. Barro, R. (1974) “Are Government Bonds Net Wealth?,” Journal of Political Economy 82:1095–117. Becker, G. (1974) “A Theory of Social Interactions,” Journal of Political Economy 82: 1063–93. —— (1991; enlarged edition) A Treatise on the Family, Cambridge, Mass.: Harvard University Press. Bergmann, B.R. (1987) “The Task of a Feminist Economics: A More Equitable Future,” in Christie Farman (ed.) The Impact of Feminist Research in the Academy, Bloomington: Indiana University Press. Bergstrom, T.C. (1989) “A Fresh Look at the Rotten Kid Theorem,” Journal of Political Economy 97:1138–59. Bergstrom, T.C., Blume, L. and Varian, H. (1986) “On the Private Provision of Public Goods,” Journal of Public Economics 29:25–49. Bernheim, D., Shleifer, A. and Summers, L. (1985) “The Strategic Bequest Motive,” Journal of Political Economy 85: 1045–76. Bernheim, D. and Stark, O. (1988) “Altruism within the Family Reconsidered: Do Nice Guys Finish Last?,” American Economic Review 78, 1034–45. Binmore, K., Rubinstein, A. and Wolinsky, A. (1986) “The Nash Bargaining Solution in Economic Modelling,” Rand Journal of Economics 17:176–88. Bragstad, T. (1991) “Private Provision of a Public Good—The Significance of Thresholds,” manuscript, University of Oslo. Bruce, N. and Waldman, M. (1990) “The Rotten Kid Theorem Meets the Samaritan’s Dilemma,” Quarterly Journal of Economics 105:155–65. Buchanan, J.M. (1975) “The Samaritan’s Dilemma,” in Edmund S.Phelps (ed.) Altruism, Morality and Economic Theory, New York: Russel Sage Foundation. Chiappori, P.-A. (1988) “Nash-bargained Household Decisions: A Comment,” International Economic Review 29: 791–6. —— (1991) “Nash-bargained Households: A Rejoinder,” International Economic Review 32:761–2. —— (1992) “Collective Labor Supply and Welfare,” Journal of Political Economy 100:437–67. Coate, S. and Loury, G.C. (1993) “Will Affirmative-action Policies Eliminate Negative Stereotypes?,” American Economic Review 83:1220–40. Cohen, L. (1987) “Marriage, Divorce and Quasi-rents; Or ‘I Gave Him the Best Years of My Life’,” Journal of Legal Studies 16:267–303. Gronau, R. (1973) “The Intrafamily Allocation of Time,” American Economic Review 63:634–51. King, A.G. (1982) “Human Capital and the Risk of Divorce: An Asset in Search of a Property Right,” Southern Economic Journal 49:536–41. Konrad, K.A. and Lommerud, K.E. (1995) “Family Policy with Non-cooperative Families,” Scandinavian Journal of Economics 97:581–601.
46
NON-COOPERATIVE GAMES IN THE FAMILY
—— (1996a) “The Bargaining Family Revisited,” CEPR Discussion Paper No. 1312. —— (1996b) “Fertility Choices in a Non-cooperative Family,” manuscript, University of Bergen. Kooreman, P. and Kapteyn, A. (1990) “On the Empirical Implementation of Some Game Theoretic Models of Household Labor Supply,” Journal of Human Resources 25:584–98. Leuthold, J. (1968) “An Empirical Study of Formula Income Transfers and the Work Decisions of the Poor,” Journal of Human Resources 3:312–23. Lommerud, K.E. (1989) “Marital Division of Labor with Risk of Divorce: The Role of ‘Voice’ Enforcement of Contracts,” Journal of Labor Economics 7:113–27. Lommerud, K.E. and Vagstad, S. (1996) “Mommy Tracks and Public Policy,” manuscript, University of Bergen. Lundberg, S. and Pollak, R.A. (1993) “Separate Spheres Bargaining and the Marriage Market,” Journal of Political Economy 101:988–1010. —— (1994) “Noncooperative Bargaining Models of Marriage,” American Economic Review, Papers and Proceedings 84:132–7. McElroy, M.B. (1990) “The Empirical Content of Nash-bargained Household Behavior,” Journal of Human Resources 25:559–83. McElroy, M.B. and Horney, M.B. (1981) “Nash-bargained Household Decisions: Towards a Generalization of the Theory of Demand,” International Economic Review 22:333–49. —— (1990) “Nash-bargained Household Decisions: Reply,” International Economic Review 31:237–2. Manser, M. and Brown, M. (1980) “Marriage and Household Decision Making: A Bargaining Analysis,” International Economic Review 21:31–44. Nelson, J.A. (1995) “Feminism and Economics,” Journal of Economic Perspectives 9: 131–48. Pollak, R.A. (1985) “A Transaction Cost Approach to Families and Household,” Journal of Economic Literature 23: 581–608. Stark, O. (1995) Altruism and Beyond, Cambridge: Cambridge University Press. Ulph, D. (1988) “A General Noncooperative Nash Model of Household Behaviour,” manuscript, University of Bristol. Warr, P. (1982) “Pareto Optimal Redistribution and Charity,” Journal of Public Economics 19:131–8. —— (1983) “The Private Provision of a Public Good is Independent of the Distribution of Income,” Economics Letters 13:207–11. Weiss, Y. and Willis, R.J. (1985) “Children as Collective Goods and Divorce Settlements,” Journal of Labor Economics 3:268–92. —— (1993) “Transfers among Divorced Couples: Evidence and Interpretation,” Journal of Labor Economics 11: 629–78. Woolley, F. (1988) “A Non-cooperative Model of Family Decision Making,” Discussion paper no. TIDI 125, London School of Economics.
NOTES 1 This strict separation between efficiency and distribution applies only when utility is transferable between the parties. 2 Konrad and Lommerud (1995) do not limit their analysis to the case of identical utility functions. 3 For a more precise definition, see Konrad and Lommerud (1995). 4 To be precise, the real issue is whether or not these contributions can be observed in a verifiable way. 5 First, the crowding out of the male’s contribution by the state will mean that a contributor with higher contribution productivity crowds out one with a lower. Second, after the male’s contribution is fully crowded out, additional government contribution crowds out only the contribution of the female. As the cost of government provision is borne by both spouses, it is as if the female is subsidized by an outside party, so this too implies that total provision increases.
KJELL ERIK LOMMERUD
47
6 Konrad and Lommerud (1996b) stress that policy measures that tilt the distribution in marriage in favor of women can help to increase the number of children. If both parents can veto having a child, one reason for the female to exercise this right is that she fears that she will have to bear the bulk of the responsibility for rearing the children in a later non-cooperative phase in family life. Policy measures that ensure that the female does not suffer too much in such a situation can therefore make much sense. 7 See also Lundberg and Pollak (1994). 8 See also Weiss and Willis (1993). 9 See the already mentioned articles by Ulph (1988), Woolley (1988), Lundberg and Pollak (1993) and Konrad and Lommerud (1995). 10 One also wonders if attention based on the threat of disinheritance is equally valuable for the parent as attention given freely out of love.
Part II MARRIAGE AND FAMILY FORMATION
3 INTRAHOUSEHOLD DISTRIBUTION OF RESOURCES AND LABOR MARKET PARTICIPATION DECISIONS Daniela Del Boca
INTRODUCTION AND BACKGROUND Within the traditional neoclassical model of household behavior, the household is assumed to have a unified preference function concerning outcomes for each of the household members. Such a preference function is consistent with a situation where there is a dictator (Becker’s altruistic model) as well as a situation in which every member of the family has the same preferences defined over the consumption of all household members (Samuelson’s consensus model).1 This traditional neoclassical model of the family has however come into question both on empirical and theoretical grounds.2 Recently attention has been devoted to the intrahousehold distributional implications of family policies: for example, policies directed toward increasing female labor supply, especially when related to the care or custody of children. However, traditional neoclassical models of household behavior which assume a single utility function neglect the effect of the intrahousehold distribution of resources on the distribution of welfare in the household. Both Samuelson’s consensus model and Becker’s altruistic model imply that an increase in family resources would have the same effect on family welfare regardless of which family member receives this increase. Because the incomes of individual family members are pooled in a joint household budget, it does not matter which family member receives them: the effect of lump-sum payments (property income or transfers) will be the same. Recent empirical studies have shown that data fail to support the restrictions embodied in the single utility function models (Schultz 1990; Thomas 1990). Alternative models take into account the intrahousehold distribution of resources and make it possible to understand the separate effects of price and income and the distributional implications of various policies within the family (McElroy and Horney 1981; Manser and Brown 1980; Chiappori 1988). As discussed in chapters 1 and 2 of this book, game-theoretical bargaining models which allow heterogeneity in preferences provide an appropriate way to consider interactions within the family. Recently researchers have thus attempted to investigate alternative hypotheses about household decisionmaking, testing them against each other on the basis of the restrictions they may imply for the household demand. McElroy and Horney (1981) and Manser and Brown (1980) propose a cooperative Nash bargaining model of household behavior. In the Nash bargaining models each household member has a utility function and a threat point (maximal utility level if agreement is not reached). The greater the threat point, the more strongly that member’s relative valuation of goods is reflected in the household demands. An important element of these models is the existence of threat points that relate to the possibility of an individual “not agreeing.” The definition of the threat point is still an open question. In McElroy (1990) the
50
DANIELA DEL BOCA
threat point is the threat of divorce and is determined by the options outside the family and therefore by the wife’s/husband’s assets, while in the “separate spheres” model of Lundberg and Pollak (1993), the threat point is the utility associated with the non-cooperative outcome. If the members of the family do not agree, they will change to a non-cooperative type of behavior in which each spouse takes the other spouse’s strategy as given.3 The household bargaining model shifts the attention from resource-pooling to the control of resources. Resources are not an indistinguishable income, but are attributable to each spouse: therefore, the two partners may want to allocate a different share of resources to their consumption, possibly making the effect of the husband’s income different from that of the wife’s income. Equality of the income effects is a necessary condition in order for the traditional approach to be supported and is sufficient to reject the bargaining model (see Ott 1992 for a discussion). The issue of the intrahousehold allocation is very important for welfare considerations. Traditional household models are based on the assumption that only the distribution of income across households is relevant while the allocation within the household is not. Analyses which take into account intrahousehold distribution of resources may significantly modify a number of normative recommendations provided by the traditional approach. Various aspects of this issue have been examined. For example, Apps and Rees (1988) analyze the effects of the system of taxation on family decisions. More recently, Lundberg and Pollak (1993) compare the distributional effects of child allowances under alternative household models. This paper attempts to test the hypothesis that households in Italy may be treated as if they pool all their income. Labor supply functions are estimated which permit income effects to vary depending on the family member to whom the income is attributed. We test whether income accruing to the husband has the same effect on his leisure demand as that accruing to wives. Rejection of equality does not allow us to distinguish among the bargaining models discussed so far, but only to reject the income-pooling hypothesis of the single utility function model. Following McElroy we also test for the significance of the extrahousehold environmental parameters which affect the threat points or reservation utility that each member could achieve outside the household. The difference from the single decision-maker’s model is that here the opportunity cost of family membership is of importance for the intrafamily distribution of income and therefore for the household demands. This analysis also allows us to test the hypothesis of the “traditional family decision model” widely used in labor supply literature, which treats the labor supply decision of married men as independent of the behavior of their wives and the labor supply decision of married women as conditional on their husbands’ behavior. The paper is structured in the following way. First the neoclassical single utility model of family decisions is discussed and compared with models derived from the two-person cooperative game theory. The next section describes some of the most relevant results of the recent research on income-pooling. Then the data and the econometric models used in the analysis are described. The following section reports the results, and finally the conclusions are summarized. TESTING FAMILY DECISION MODELS In the single decision-maker model, the household is assumed to have a unified preference function concerning outcomes for each of the household members. Such a preference function is consistent with a situation where there is a dictator as well as a situation in which every member of the family has the same preferences defined over the consumption of all household members (Becker 1981).
RESOURCES AND LABOR MARKET DECISIONS
51
In the model characterized by a single decision-maker, the husband and the wife (h, w) maximize utility by purchasing market goods Xw, Xh and allocating leisure l, under the full income budget constraint. (3.1) where Li is the labor supply and will be a function of price P, wages Ww and Wh, and non-labor income Y, —that is, the total non-labor income of the family. Unearned income has an identical where effect on household demands and labor supplies regardless of the recipient of the non-earned income. (3.2) In models characterized by multiple decision-makers, husbands and wives are usually assumed to bargain over allocation, subject to the restriction that outcomes be Pareto-efficient. One example is the Nash bargaining model (McElroy and Horney 1981), where the partners maximize the product of their individual gains from marriage, i.e. the product of the difference between U, the utility level each partner achieves within the marriage, and V, the threat point or reservation utility each of them would achieve outside the family. (3.3) subject to the full household expenditures being equal to full household income: (3.4) The threat point in McElroy and Horney (1981) is the husband’s/wife’s maximized indirect utility outside the marriage and depends on the prices of each partner’s goods, wage rate, non-wage income and extrahousehold parameters. The greater the threat point, the more strongly that member’s relative valuation of goods will be reflected in the household demands. The vectors Ew and Eh represent non-price characteristics of the environment each partner would face outside the marriage such as conditions in the local marriage market as well as in the local labor market, family laws, welfare transfers that in case of separation or divorce would go to one of the spouses, etc. In the Nash bargaining model of household behavior, Yw and Yh are assumed to be resources that the wife or the husband could take with her/him when leaving the household. The solution to the maximization of (3. 3) subject to (3.4) is a system of demand equations where the arguments of the Nash demand system include all prices, separate measures of non-labor income for the partners (w, h) and the extrahousehold parameters. (3.5) Since the Nash bargaining model is a generalization and the neoclassical family demand is a special case, statistical tests can be used to determine whether the data on family behavior satisfy the restrictions implied by the neoclassical model. One testable restriction is the income-pooling hypothesis according to which nonlabor incomes of husbands and wives enter the labor supply functions (as well as the family demand functions) separately. The second restriction on (3.5) consists in the inclusion of the extrahousehold environmental parameters, since in the neoclassical demand system the opportunity costs of being married are not relevant. These tests are based on the fact that the neoclassical models are more restrictive and therefore the empirical evidence would not lead to acceptance of the Nash bargaining model, but only to rejection of the neoclassical restrictions.
52
DANIELA DEL BOCA
We also conduct tests of the “traditional” family model using the same econometric framework. In this conceptualization of family behavior, the husband is assumed to make his labor supply choice first; then the wife makes her labor supply choice conditional on her husband’s. One way to model this situation formally is as follows. Say there exists a unitary family utility function, U(Xw, Xh, lw). Let the husband “move” first and behave as though his wife would not work [i.e. lw=T]; let him also ignore any non-labor income his wife has when making his allocation decisions. He would select his labor supply and his and his wife’s private consumption, according to the following program:
(3.6) Denote the optimal labor supply decision associated with this problem by Lh (wh, Yh). The wife maximizes the same objective function as the husband, but she conditions on the husband’s labor supply choice and considers also her own non-labor income when making her decision. Thus her problem is given by (3.7) The wife’s labor supply function thus is given by This model is quite restrictive: the only reasonable interpretation is that men decide upon their labor supply when they are young and single and seldom change during their life time. Therefore male labor supply depends on the man’s own characteristics, but not on the characteristics of other family members. This model could be appropriate to describe a situation where the wife has no non-labor income and is not employed (a situation which has been quite common in the past). The implicit assumption is a joint household utility function, given that the labor supply of the spouses is treated as a twostep decision of the household. Some recent studies have tried to test alternative hypotheses of family allocative behavior (McElroy and Horney 1981; Bourgignon et al. 1994; Kooreman and Kapteyn 1990). Most of them have attempted to verify empirically whether the single utility model is consistent with the data. An implication of the traditional household utility model is the income-pooling hypothesis. If households maximize a single utility function under a budget constraint, then only total non-labor income should matter and the various recipients of the non-labor income should be irrelevant for the analysis of consumption behavior. EMPIRICAL EVIDENCE SUPPORTING THE INCOME-POOLING HYPOTHESIS Most empirical tests of household allocation have focused on leisure demand (or labor supply). In most household surveys, in fact, with the exception of leisure, data on consumption and non-labor income are collected at the household rather than at the individual level. As Kooreman and Kapteyn (1990:366) have recently pointed out, there is a need to collect more specific data on each of the players of the family unit. “In other words, not only the theorist should stop considering the family as an homogeneous unit, but also the data collector should do the same thing.”
RESOURCES AND LABOR MARKET DECISIONS
53
Because of the lack of appropriate data, only a few studies have tested the income hypothesis using consumption data. Thomas (1990, 1993), using survey data on family health in Brazil, finds that the distribution of income among men and women within the household significantly affects demand patterns. His results show that income controlled by a mother has a greater effect on her family’s health, human capital (household services, education) and also leisure (recreation activities) than does income under the control of a father. The strongest result emerges for child survival probabilities: a mother’s unearned income has almost 20 times the effect of a father’s unearned income. Bourgignon et al. (1994) investigate whether the income-pooling hypothesis can be accepted for consumption data from France. They consider a subsample of couples where both husband and wife work full time, so that they can assume that labor supply is not a choice variable, but is determined by labor demand. A consequence is that earnings (as the product of wages times the constrained number of hours) are exogenous. For a given amount of total income, the respective shares of each other’s earnings and nonlabor incomes can be treated as distribution factors. They find that income-pooling is rejected and that for a given level of total income, the share of husband’s and wife’s own income significantly affects the structure of consumption: the intrahousehold distribution of income seems to influence household behavior even when total income is fixed. In their study an empirical test of the cooperative hypothesis (that is, of a Paretoefficient collective decision-making) is also proposed. Their results reject the approach of income-pooling and seem consistent with the cooperative hypothesis. Most studies have focused on leisure demand, on which data for each member of the family are more likely to be collected. McElroy and Horney (1981) and Manser and Brown (1980) used National Longitudinal Survey data to analyze family labor supply decisions. McElroy and Horney’s results indicate equality of income effects only for the husbands’ labor supply equation, while Manser and Brown’s results reject the hypothesis that non-labor income received by the husband, the wife and other members of the household has the same effect on male and female labor supply. Schultz (1990) used the Socioeconomic Survey of Thailand to test the pooling hypothesis for both the female labor supply and the fertility demands of Thai families. The results reject the pooling restriction and imply that women with more bargaining power prefer to increase their own consumption of leisure or time in non-market activities and prefer to have more children. A woman’s non-labor income has a larger negative effect on the probability that she enters the wage labor force than does her husband’s unearned income. The opposite is true for men, while fertility is not influenced by the husband’s non-wage income. Alternative tests of the pooling hypothesis are based on natural experiments. Lundberg, Pollak and Wales (1995), using UK data, show that the policy change that transferred a substantial child allowance to wives has implied a shift towards larger expenditures on women’s and children’s goods. Other studies have tested the second restriction—that is, the relevance of the opportunity costs of being married. Carlin (1991) analyzed the effects of extrahousehold environmental variables on women’s spending patterns. According to his empirical results, women in states characterized by legal structures that allow generous divorce settlements spend more on investing in their human capital. Recent research on unmarried couples shows supporting results and points to the importance of considering the interactions between partners. A comparison between married and cohabiting families shows that among cohabiting couples a higher proportion of individuals do not share their income (van der Klaauw 1994). Similar tests of the effect of income composition on family consumption patterns have been produced for divorced households, using US Consumer Expenditure Survey data (Del Boca and Flinn 1994; Del Boca 1994). They test for the equality
54
DANIELA DEL BOCA
of the income of the exspouses on the demands for child-specific goods and for leisure demands by separated mothers.4 The empirical results they report concerning the income composition effect show that the income equality restriction could be rejected in the cases of both consumption of child-specific goods and leisure demands of mothers. These results appear to be at odds with the implications of the standard neoclassical utility-based models as well as with non-cooperative household models in which no role is allowed for the composition of income in determining demand allocations across goods, including leisure. THE DATA Our empirical analysis utilizes the data of the Bank of Italy’s Survey of Household Income and Wealth (1993). This survey is conducted every two years by the Central Bank of Italy and contains detailed information on the incomes and wealth of family members as well as several characteristics of the workplace (wages, hours of work) and socio-demographic characteristics of the households (age of the members of the family, number of children, marital status). Family structure in Italy has undergone substantial changes over the last two decades. The extended family which dominated in the southern regions was disrupted by the massive migration to the northern regions during the 1950s and 1960s. Women’s labor force participation rate has increased (even if it is still among the lowest in Europe), while women’s hourly wages have come closer to men’s hourly wages. Over these years, as an effect of these trends, there has been a remarkable decline in fertility rates: Italy now ranks lowest of all countries. In spite of the increase in the participation of women in the labor market, other changes have been slower. The percentage of part-time jobs is still extremely low in comparison with other countries and has not significantly increased in the last few years. Time budget studies show that men’s household labor has not been very responsive to women’s market work. Men married to working women spend about the same number of hours in household activities per week as men married to non-working women. The supply of subsidized childcare services increased during the 1970s and 1980s, but is characterized by extreme rigidity in the weekly hours as well as great variability across regions. The availability of public childcare services for children between 0 and 5 years of age is about 30 percent in the northern regions and only 1–2 percent in the southern regions (number of places available to the population 0–5 years of age). Profound differences remain in demographic as well as labor market aspects between the northern and central areas of the country and the southern regions, where female employment rates are much lower than in the northern regions and the availability of formal childcare services is very limited while the structure of the family is still characterized by very traditional relationships. Especially in these areas of the country, we would expect that a “traditional” model of household decision-making should perform well. Descriptive statistics of the variables used in the empirical analysis are presented in Table 3.1. The labor incomes measured by the survey are net of social security contributions and of the contributions paid by the employers toward personal income tax. We restrict our sample to married couples. Married couples with income from self-employment have been excluded from the sample. The exclusion of these families is justified since we do not observe hours worked for the self-employed. Age is restricted to the 21–58 year range. The sample size after the selection is 4074 households. Women work in 40 percent of the cases, while men work in 90 percent of the cases. Women work much less and therefore have on the average lower yearly incomes than men. However, when they work, their weekly hours of work and hourly wages are not
RESOURCES AND LABOR MARKET DECISIONS
55
very different from those of men (36 hours a week as opposed to 40 for men and an income of about 11,000 liras versus 12,000 liras). The distribution of hours for both men and women is highly concentrated, around Table 3.1 Descriptive statistics Variables Husbands Hours of work (positive values) Hourly wages (positive values) Non-labor income (positive values) Age No. children<6 Schooling Participation rate North Wives Hours of work (positive values) Wages (positive values) Non-labor income (positive values) Age No. children <6 Schooling Participation rate
All families
Families with children <6
Means
St. dev.
Means
St. dev.
38.09 39.7 11.76 12.8 223.0 408.0 43.7 0.33 10.05 90.01 0.39
8.0 7.3 6.0 5.6 173.0 1.0 9.8 0.5 4.4 18.0 0.48
29.0 39.2 6.9 14.86 169.0 290.10 29.0 11.9 11.0 96.0 0.35
8.00 7.31 6.0 5.68 73.0 11.1 4.8 4.4 4.7 19.5 0.47
13.2 35.9 4.3 11.7 66.1 265.0 40.1 0.33 10.01 40.0
8.7 10.68 7.7 4.96 34.0 4.99 9.6 0.5 3.7 49.0
11.5 32.9 4.3 12.7 35.0 163.8 25.6 11.9 10.7 47.2
8.7 10.68 6.0 4.96 22.9 61.99 4.6 3.7 3.0 41.0
35–40 hours a week. As we will discuss later, it seems that the important decision is whether to participate in the labor market or not, rather than how many hours to work. Given the differences in attitudes and behavior between men and women of younger and older cohorts and the different constraints of the presence of preschool children in the allocation of time of the spouses, we analyze separately the subsample of households where a child under six years of age is present. This selection criterion leaves us with a sample size of 1034 households. Table 3.1 reports the descriptive statistics for the sample of families with children under six years of age. The descriptive statistics show that lower hourly wages as well as lower non-wage income characterize women and men with younger children. They are also characterized by higher labor force participation rates, 96 percent for husbands and 47 percent for wives.
56
DANIELA DEL BOCA
THE ECONOMETRIC MODEL We first estimate the effect of non-labor income on the participation in the labor market of each spouse in order to test the income-pooling restriction. Labor supplies depend on wages and non-labor income of both spouses and on a vector E (which includes Ew and Eh). The labor supply functions are specified as: (3.8) The extrahousehold environmental parameters typically do not show directly in the preferences or constraints facing either member of the couple. They only shift the values of outside options for either partner. The null hypothesis for the income-pooling is: . Another hypothesis to test is whether the extrahousehold environmental parameters play a role, since in the neoclassical demand system . the opportunity costs of being married should be of no significance. The null hypothesis is: The null hypothesis for the independence of the husband’s labor supply decision is: Several problems have to be taken into account when testing the income-pooling hypothesis. Since the test for the equality of the effects of non-labor income is crucial in order to test the single decision-maker approach, the quality of the tests is dependent on the quality of the estimated parameters of non-wage incomes. Most data-sets however do not identify separately some sources of non-wage income for the husband or the wife (McElroy and Horney 1981). However, even when non-wage income information is available and is attributable to each member of the household, other problems in testing the hypothesis may remain. One problem pertains to the fact that the percentage of families receiving non-labor income is likely to be non-random. Using data on households both with and without non-wage income implies the assumption of the same income elasticity for the subsample of families receiving and not receiving non-wage income and that no systematic differences exist between the two groups. Some kinds of non-wage income, such as transfers from the social security system, depend on the employment status or on the amount of earned income and therefore cannot be considered exogenous. When we take into account the endogeneity problems using instrumental variables, we face the problem of arbitrariness regarding the choice of instruments: that is, the equality of income effects may depend on the choice of instruments. Other problems may come from measurement errors. Current unearned income is difficult to measure in household surveys. Therefore rejection of the equality of income effects may be due to differential measurement errors across individuals in the survey (Thomas 1993). Another problem is the choice of variables which influence the threat points. Many relevant variables are likely to influence the threat points, but also the demand function. Good examples are the number of children as well as the region of residence. Given the characteristics of the Italian labor market, we have introduced as extrahousehold environmental parameters the number of childcare places available for children aged 0–5 by region and the sex-ratio of the population by region, both of which are likely to alter the outside options in the marriage market as well as the options in the labor market. THE EMPIRICAL RESULTS In order to determine whether the distribution of income within a household affects household labor supply patterns, we will test for the equality of the impact of male non-labor income Yh and female non-labor income Yw. The reduced form coefficient estimates for unearned income are reported in Table 3.2 for husbands’ and wives’ labor market participation. We have chosen to estimate labor supply as a discrete choice, instead of number of hours worked, given that most recent research on Italian labor supply
RESOURCES AND LABOR MARKET DECISIONS
57
(Colombino and Del Boca 1990) has shown that the interesting behavioral response occurs at the extensive margin more than at the intensive one (that is, the number of hours supplied to the labor market). Table 3.2 Effect of total family non-labor income on labor market participation Variables
Wives
OLS
IV
OLS
IV
Constant
.238 (.027) −.011 (.006) .011 (.003) .009 (.004) .115 (.017) .002 (.001) .002 (.001) 3.073
.282 (.032) −.038 (.016) .006 (.003) .008 (.004) .117 (.019) .002 (.001) .002 (.003)
.962 (.123) −.042 (.025) .001 (.012) .002 (.002) .002 (.002) .001 (.001) .010 (.010) 1.818
Non-labor income Spouse wage Own wage North Childcare Sex-ratio Tests for endogeneity
Husbands .961 (.028) −.086 (.044) .011 (.006) .012 (.001) .005 (.005) .006 (.007) .010 (.011)
Notes: Heteroskedasticity-consistent standard errors in brackets. Instruments: age, age squared, schooling, region
The regressions include total family unearned income (Table 3.2) and separate unearned income of husbands and wives (Table 3.3). Other variables in the regression are the region of residence (aggregated into north-center and south), the number of children, the number of childcare places available by region, the population sex-ratio by region, and the hourly wages of husbands and wives. Since the wage rate is observed only for the working subsample, we use a selection-bias corrected regression procedure (Heckman 1979). The results of Table 3.2 can be summarized as follows. Total family non-labor income has a negative but marginally significant effect on the labor force participation of wives. The coefficient of one’s own and one’s spouse’s hourly wage is positive and significant. Total non-labor income has a negative but not significant effect on the labor force participation of husbands. Both one’s own and one’s spouse’s wages have positive effects, but are not significantly different from zero. Living in a region in the center or north of Italy increases the probability of working for wives, while for husbands the effect is not statistically significant. The availability of subsidized childcare and the sex-ratio variable have positive and significant coefficients only for wives. The standard errors reported in brackets are referred to as Eicker-White; they are consistent even when the distribution of the errors is not identical in the population, though the errors are still required to be independent. The OLS estimate of the parameters divided by the Eicker-White estimate of its standard error has an asymptotic standard normal distribution under the null hypothesis. As we have discussed above, some kinds of non-labor income (such as transfers or social security) may depend on employment status and therefore may not be exogenous. To test for income endogeneity we use a
58
DANIELA DEL BOCA
Wu-Hausman test, which is equivalent to including an exogenous predictor of non-labor incomes and testing for its statistical significance (the instruments used are age, age squared, schooling and region). At the bottom of Table 3.2 the tests for endogeneity are reported. The tests indicate endogeneity only for the wives’ equation. In Table 3.3a, the estimates of the labor market participation decision of wives show that the coefficient associated with one’s own non-labor income (a3) is negative while the coefficient associated with the husband’s non-labor income is positive (a4). The estimates of the participation decision of husbands are different: both the coefficient of one’s own non-labor income (b3) and the coefficient of one’s spouse’s nonlabor income are negative (b4). However, while the coefficients a3 and a4 are significantly different from zero in the wife’s equation, the analogous coefficients b3 and b4 are not significant. This result is also consistent with a situation in which the wives’ participation rate is very low and non-labor income is zero or close to zero. The estimates obtained for women conform qualitatively to other results reported in the literature that show the larger income elasticity for female than for male labor supply (Colombino and Del Boca 1990). The positive sign of the spouse’s non-labor income (a4) and wage (a2) in women’s labor market participation equation could be explained by the fact that a higher social status for the husband might increase the possibilities for gaining access to better information about labor market opportunities, raising the likelihood of women finding a job. The reverse does not seem to be true: the data do not allow us to reject the null hypothesis of the “traditional family” model, i.e. that the husband’s labor market decisions are independent of his wife’s behavior, since the coefficients for the wife’s non-labor income (b4) and wage (b2) are not significant in the husband’s decision. The coefficients of the extrahousehold environmental variables (number of subsidized childcare places available in the region and population sex-ratio) are both significant in the wife’s equation, but are not significant in the husband’s equation. These results indicate that in this case we can reject the null hypothesis as of no significance to the extrahousehold environmental parameters. We now consider the results of the test on the income-pooling hypothesis. The test statistics for the nonlabor income equality restriction are reported in Table 3.3b. We utilize a chi-squared statistic to test the nonlabor income Table 3.3a Effect of male and female non-labor income on labor market participation Variables Constant Spouse non-labor income Own non-labor income Spouse wage Own wage
Wives
Husbands
OLS
IV
OLS
IV
.259 (.012) .003 (.001) −.014 (.005) .019 (.004) .003 (.002)
.290 (.031) .039 (.032) .011 (.030) .006 (.002) .008 (.005)
.967 (.008) −.002 (.002) −.044 (.023) .014 (.010) .011 (.008)
.973 (.009) −.004 (.005) −.078 (.049) .005 (.003) .001 (.001)
RESOURCES AND LABOR MARKET DECISIONS
Variables North-center Childcare Sex-ratio
Wives
59
Husbands
OLS
IV
OLS
IV
.121 (.017) .006 (.003) .012 (.007)
.117 (.019) .002 (.001) .011 (.007)
−.003 (.005) .002 (.002) .016 (.021)
.004 (.008) .001 (.013) .013 (.011)
Notes: Sample size: 4074. Instrumented variables yw, yh. Instruments: age, age squared, schooling, region. Heteroskedasticity-consistent standard errors in brackets Table 3.3b Tests for the equality of income effects Under exogeneity Heteroskedastica Homoskedasticb Under endogeneity Heteroskedastica Homoskedasticb
Wives
Husbands
16.13 10.223
15.297 12.978
5.27 5.987
6.48 7.84
Notes: a Heteroskedastic tests are distributed as normal (0, 1). b Homoskedastic tests are distributed as t(n−k)
equality restriction. The empirical results allow us to reject the equality restriction on non-labor income for both husbands and wives. An increment of one’s own non-labor income (and therefore of the bargaining power within the family) and of the spouse’s non-labor income have opposite effects on the demand for leisure of wives. For the husband’s equation the signs of both his own and his spouse’s non-labor income are negative, but only his own non-labor income coefficient is significant. We are now interested in verifying whether non-labor income equality can be rejected also for families with young children (less than six years old). Table 3.4a shows that for this sample both the signs of one’s own and one’s spouse’s non-labor incomes are negative for the labor market participation of both wives and husbands. Also the effects of the wage rates are more symmetric: the coefficients associated with the spouse’s wage are positive and significant for the labor market participation decisions of both wives and husbands. The effect of sex-ratio and childcare are significant only for female participation and the signs of the coefficients are the same as for the total sample. The test statistics for equality of the spouses’ non-labor income coefficients are reported in Table 3.4b and show that equality cannot be rejected for this sample. The tests for endogeneity show that exogeneity of non-labor incomes can only be weakly rejected for wives.
60
DANIELA DEL BOCA
Table 3.4a Effect of male and female non-labor income on labor market participation. Families with children under six Variables Constant Non-labor income
Wives OLS
OLS
OLS
OLS
.186 (.065) −.017 (.008)
.200 (.063)
.998 (.088) −.018 (.004)
.947 (.009)
Spouse non-labor income
−.010 (.007) −.007
Own non-labor income Spouse wage Own wage North Childcare Sex-ratio
Husbands
(.005) .016 (.007) .007 (.009) .104 (.031) .003 (.001) .012 (.007)
.017 (.006) .006 (.008) .103 (.035) .002 (.001) .010 (.008)
.005 (.002) .003 (.001) −.005 (.008) .002 (.007) .012 (.019)
−.003 (.001) −.015 (.004) .004 (.001) .002 (.002) −.005 (.006) .001 (.001) .011 (.009)
Notes: Sample size: 1034. Heteroskedasticity-consistent standard errors in brackets Table 3.4b Tests for the equality of income effects
Under exogeneity Heteroskedastica Homoskedasticb Under endogeneity Heteroskedastica Homoskedasticb
Wives
Husbands
2.675 0.437
0.918 0.348
2.483 1.987
2.934 0.271
Notes: a Heteroskedastic tests are distributed as normal (0, 1). b Homoskedastic tests are distributed as t(n−k)
In a second stage we estimate a bivariate probit model to analyze jointly the decision of husbands and wives to work. We assume that and are random variables which capture the effect of unobservable characteristics. The distributional assumptions are:
RESOURCES AND LABOR MARKET DECISIONS
61
Table 3.5 reports the results. The signs of the coefficients are the same as in the earlier regressions. The ρ is estimated to be significantly different from zero. It is positive and larger in magnitude for younger couples, indicating complementarity among partners. The test statistics reported at the bottom of Table 3.5 allow us to reject the equality restriction on non-labor income for both husbands and wives for the whole sample, but not for the sample of families with younger children. The greater interdependence between the spouses in these families could be associated with the existence of greater liquidity constraints in younger cohorts, especially when young children are present. Therefore an increase in non-labor income of both partners would allow a reduction in labor supply in a period during which substitutes for parents’ time in childcare are expensive and/or difficult to organize. While the coefficient of the wife’s income on the husband’s participation is never significant for the total sample of families, for families with small children the coefficient is significant. An increase in the nonlabor income of both spouses has a similar negative effect on the probability of male labor market participation. These results indicate that the presence of young children has an important effect on labor supply interactions. The “traditional family” model (which treats the labor supply of married men as independent of the behavior of their wives and the husband’s behavior in turn as exogenous with respect to the wife’s work decision) does not seem to be an adequate description of the behavior of younger families. These empirical findings confirm previous results from earlier studies which show that families with young children behave and interact in a different way from other families (Lundberg 1988). In families with younger Table 3.5 Maximum likelihood estimates. Bivariate probit. Effect of male and female non-labor income on labor market participation All families Probability of wife working Constant Non-labour income
.186 (.065) −.006 (.003)
Spouse non-labor income Own non-labor income Spouse wage Own wage North Childcare Sex-ratio Probability of husband working
−.016 (.007) .152 (.071) .104 (.031) .007 (.002) .008 (.003)
Families with children under six .200 (.063)
.010 (.017) −.016 (.035) −.017 (.005) .159 (.079) .103 (.035) .008 (.003) .009 (.003)
.104 (.061) −.023 (.021)
.023 (.017) .152 (.091) .104 (.031) .013 (.007) .019 (.009)
.201 (.061)
−.012 (.006) −.011 (.004) .013 (.017) .159 (.089) .103 (.035) .012 (.007) .018 (.009)
62
DANIELA DEL BOCA
All families Constant Non-labor income
.177 (.050) −.036 (.018)
Spouse non-labor income Own non-labor income Spouse wage Own wage North Childcare Sex-ratio ρ Test statistics for equality
.031 (.023) .098 (.049) .107 (.067) .004 (.002) .002 (.002) .39 (.102) 4.086
Families with children under six .175 (.053)
−.009 (.019) −.025 (.010) .030 (.023) .090 (.048) .107 (.055) .002 (.003) .003 (.002)
.107 (.050) −.028 (.006)
.011 (.003) .072 (.009) .033 (.067) .012 (.011) .003 (.005) .47 (.109) 1.708
.109 (.056)
−.013 (.005) −.015 (.006) .010 (.003) .071 (.008) .031 (.055) .012 (.010) .003 (.005)
children, spouses “pool” their resources, and labor supply decisions are mutually and significantly influenced by the characteristics and behavior of the other partner. CONCLUSIONS Most economic models of the household assume that all members share the same preferences or that one member makes all resource allocation decisions. That assumption is tested on Italian data by asking whether income under male control has the same impact on leisure demands as income under female control. The equality of non-labor income effects is rejected for the total sample, but not for the sample of families with children of preschool age. The opportunity costs of being married are significant only for women, in both samples. The “traditional family” model in which only the wife’s labor supply depends on the characteristics of the partner is rejected instead only for the sample of households with younger children. These findings suggest that the distribution of income between husband and wife within the household does affect demand patterns. The extent to which the distribution of income influences labor supply decisions is strongly related however to the composition of the household as well as to the stage in the life cycle. If unearned income under the control of the wife has a different effect on household demand than unearned income in the hands of the husband, then policy effects on household welfare should be
RESOURCES AND LABOR MARKET DECISIONS
63
conditioned by the choice of the recipient of the transfer. Public transfers directed to women may affect household decisions in a significantly different way from transfers directed to the husband or the “family.” ACKNOWLEDGMENTS This research has been financially supported by a C.N.R. grant (Bilateral Project) which is gratefully acknowledged. I want to thank U.Colombino, C. Flinn, M.McElroy and Y.Weiss for helpful comments on a previous version. REFERENCES Apps, P. and Rees, R. (1988) “Taxation and the Household,” Journal of Public Economics 35:355–69. Backer, G. (1981) A Treatise on the Family, Cambridge, Mass.: Harvard University Press. Bourgignon, F. and Chiappori, P.A. (1992) “Collective Models of Household Behavior,” European Economic Review 36, 2/3:355–64. Bourgignon, F., Browning, M., Chiappori, P.A. and Lechene, V. (1994) “Incomes and Outcomes: A Structural Model of Intrahousehold Allocation,” Journal of Political Economy 102, 6:1066–96. Carlin, P. (1991) “Intra-family Bargaining and Time Allocation,” in T.P.Schultz (ed.) Research in Population Economics, vol. 7, Greenwich, Conn.: JAI Press. Chiappori, P.A. (1988) “Nash-Bargained Household Decisions,” International Economic Review 29:791–6. —— (1992) “Collective Labor Supply and Welfare,” Journal of Political Economy 100:437–67. Colombino, U. and Del Boca, D. (1990) “The Effect of Taxation on Labour Supply in Italy,” The Journal of Human Resources 25, 3:390–414. Del Boca, D. (1988) “Women in a Changing Work-place: the Case of Italy,” in J. Jenson, (ed.) The Feminization of the Labor Force, Oxford: Oxford University Press. —— (1994) “Post-Divorce Income Transfers, and the Welfare of Mothers and Children,” Labour 8, 2:259–77. —— (1996) “Market Rigidities and Costs of Children,” paper presented at the conference “The Costs of Children,” Bologna. Del Boca, D. and Flinn, C. (1994) “Expenditure Decisions of Divorced Mothers and Income Composition,” The Journal of Human Resources 29, 3:742–61. Eicker, F. (1967) “Limit Theorems for Regressions with Unequal and Dependent Errors,” Fifth Berkeley Symposium on Mathematical Statistics and Probability, Berkeley: University of California Press. Heckman, J. (1979) “Sample Selection Bias as a Specification Error,” Econometrica 47:153–61. Horney, M.J. and McElroy, M. (1988) “The Household Allocation Problem: Empirical Results from a Bargaining Model,” Research in Population Economics 6: 15–38. Killingsworth, M.R. (1983) Labor Supplyp, Cambridge: Cambridge University Press. Kooreman, P. and Kapteyn, A. (1990) “On the Empirical Implementation of Some Game-theoretic Models of Household Labor Supply,” Journal of Human Resources 24:584–98. Lundberg, S. (1988) “Labor Supply of Husbands and Wives: A Simultaneous Equation Approach,” The Review of Economics and Statistics 70:224–35. Lundberg, S. and Pollak, R. (1993) “Separate Sphere Bargaining and the Marriage Market,” Journal of Political Economy 6:988–1011. Lundberg, S., Pollak, R. and Wales T.J. (1995) “Do Husbands and Wives Pool their Resources? Evidence from the U.K.Child Benefit,” AEA meeting, January 1995. McElroy, M. (1990) “The Empirical Content of Nash-Bargained Household Behavior,” The Journal of Human Resources 25, 4:559–83. —— (1993) “The Policy Implications of a Nash Bargaining Approach,” mimeo, IFPRI McElroy, M. and Horney, M. (1981) “Models of Households Decisions,” International Economic Review 22:333–49.
64
DANIELA DEL BOCA
Manser, M. and Brown, M. (1980) “Marriage and Household Decision-making. A Bargaining Approach,” International Economic Review 34:334–75. Ott, N. (1992) Intrafamily Bargaining and Household Decisions, Berlin: Springer Verlag. Schultz, T.P. (1990) “Testing the Neoclassical Model of Labor Supply and Fertility,” Journal of Human Resources 25, 4:599–634. Thomas, D. (1990) “Intra-household Resource Allocation: An Inferential Approach,” Journal of Human Resources 25, 4:635–64. —— (1993) “The Distribution of Income and Expenditure within the Household,” Annales d’Économie et de Statistique 29:109–35. Van der Klaauw, W. (1994) “Cohabitation and Labor Supply,” mimeo, ESPE Annual Meeting, Tilburg University. White, H. (1980) “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity,” Econometrica 48:817–38.
NOTES 1 Other definitions of the single utility models are the “common preference” (Thomas 1990), the “joint utility” approach (Lundberg 1988), the “altruistic” model (Becker 1981), the “unitary model” (Chiappori 1988, 1992) and the “neoclassical model” (McElroy and Horney 1981). 2 See the discussion by Lundberg and Pollak in chapter 1 of this book. 3 One reason for choosing as an alternative threat point the non-cooperative Cournot Nash equilibrium within marriage is the monetary and time costs associated with divorce. 4 They have considered models in which parents are linked by externalities (such as consumption of children’s goods), but act non-cooperatively (that is, each parent determines the consumption under his/her control unilaterally, taking the decision of the other parent as given).
4 A FAMILY WITH ONE DOMINATING SPOUSE Kristian Bolin
INTRODUCTION It is a woman’s business to get married as soon as possible, and a man’s to keep unmarried as long as possible. (George Bernard Shaw, Man and Superman, 1903) It is common knowledge that one of the spouses in a family may dominate the other. Nevertheless, in the economic literature dominance is an overlooked phenomenon. We examine the role of dominance as it affects how husband and wife allocate their time between contributions to a family-specific public good and market work. What is new in our approach to the economics of the family is its explicit focus on conflicting internal interests. The prevailing economic theory of the family regards the family as a unit of cooperation— decisions are made in consensus. Gary Becker, who is the most prominent representative of this approach, models family decision-making by positing a household utility function.1 This model gives many insights but there are limitations. First, since each family member’s preferences are not explicitly expressed, the intrafamily procedure for reaching the final, visible decision concerning, for example, labor supply and savings cannot be analyzed by the model, nor can decisions which have a certain degree of individuality attached to them, such as settlements at divorce.2 Also, the relative time allocation of spouses relies solely on comparative advantages. Second, all models of this kind assume that spouses follow what they agreed upon. However, several agreements concerning the conduct of family life (for example, the allocation of time) are not enforceable. Hence, opportunistic behavior cannot always be excluded.3 The application of non-cooperative game theory to intrafamily decision-making is an alternative that overcomes the limitations mentioned here. A non-cooperative model explicitly embodies spouses’ individual preferences, and decisions made within a non-cooperative framework have the advantage of being self-enforcing.4 It is the thesis of this paper that one spouse may have a strategic advantage in the family decision-making game, and that this has implications for the division of time in market and non-market activities. The exact nature of the strategic advantage is assumed to be that the dominating party has the ability to make irreversible decisions concerning his/her time allocation before the other party makes his/her decisions. For instance, the wife may have to accept the decisions of her husband as a fait accompli when making her
66
A FAMILY WITH ONE DOMINATING SPOUSE
decisions. In this way, the husband has a tool for influencing the decisions of the wife. We say that the husband dominates. Dominance is formally modeled as a Stackelberg game, where the dominating party is the leader and the dominated party is the follower. It is shown that not only comparative advantage but also dominance is of importance for time allocation. Cooperation is modeled by an agreement on the sharing of incomes. The paper is organized as follows. First we present the model. Then we examine the effects of dominance on the time allocation. In the following two sections we examine some applications of dominance. For example, we examine government intervention to provide a substitute for the family-specific good, and the dependence of the effects of such intervention on the existence of dominance. THE MODEL The family consists of a dominating party (say, the husband) and a dominated party (the wife).5 The model has two stages. In the first stage a decision is made by the dominating spouse concerning how much to contribute to the family. In the second stage a decision is made by the dominated spouse. These decisions are irreversible. Both parties are considered as separate decision units. The gain from forming a family is due to the opportunities to generate benefits which are difficult to generate outside the family. The husband’s contribution of time to the family is denoted by ih, and the wife’s contribution by iw. Total time is T. Hence: (4.1) We want to examine the effects of dominance within marriage. The point is that the husband and the wife do not make decisions simultaneously. First one party chooses the contribution to the family, knowing the preferences of the other party, then the follower chooses the contribution given the leader’s contribution. Hence, the decision-making procedure is modeled as a Stackelberg game.6,7 Throughout the paper every function is three times continuously differentiable. The contributions, ih and w i , to the production of the family public good will generate the family-specific public good, according to a production function, f. Let each spouse’s contribution when the other spouse does not contribute, the standalone contribution, be denoted . Let each spouse’s contribution that completely crowds out the contributions of the other spouse be denoted .8 The family production function, f, is strictly increasing and strictly concave in the time supplied: (4.2) where α and β are the productivities of, respectively, the husband and the wife of supplying marriagespecific time. For convenience, we assume that the third-order effects are sufficiently small to be excluded from the analysis. Hence, we assume that: (4.3) As already implied, contributing to the family-specific good is costly. The reason is that the expected market income decreases with the amount of time supplied to the family. With the opportunity cost of contributing and , total incomes for the husband, yh, and the wife, to the family-specific good (wages), denoted by yw, are given by:9 (4.4)
KRISTIAN BOLIN
67
Income from market work is consumed privately, and gives utility according to an increasing, strictly concave function, u. As in the case with the family production function, we assume that the third-order effects are sufficiently small to be excluded from the analysis. Spouses may differ only in their opportunity costs of contributing to the family and in their effectiveness in supplying family-specific time. Therefore, preferences of the spouses over the family-specific public good and private consumption can be represented by the same utility function.10 Keeping the analysis simple, we examine the case in which the utility function is additively separable. The preferences of the spouses can, in this case, be characterized by: (4.5) The objectives of the husband and the wife are to maximize their respective utilities, Uh, Uw. Their utility functions can be written as: (4.6) (4.7) Once utilities have been defined, it is possible to define formally what we mean by dominance. Dominance by the husband means that he makes his decision assuming that the wife will wait and observe his decision, making her own based on his. In other words, the husband is the leader in the Stackelberg game. In this way the husband is able to influence the decision which his wife takes. Knowing his wife’s preferences, the husband can calculate how she will react to the decisions which he makes. The reaction function reflects to what extent the husband is able to affect the wife’s decisions. , yields: Differentiating the wife’s first-order condition, and solving for (4.8) This shows that when the husband lowers his contributions to the family-specific public good, the wife will increase her contribution. To examine the effects of dominance, it is useful to refer to the case where there is no dominance. Since dominance by the husband occurs when he is able indirectly to influence the decisions of his wife, no dominance occurs when neither party is able to influence the actions of the other. Formally, the family decision-making game without dominance is modeled by the Cournot model. In the Cournot model, decisions are made simultaneously, and both parties assume that their actions will not affect the actions of the other party.11 That is, in the Cournot model the husband’s conjectures about his wife’s reaction are that . Proposition 1 There exists a unique equilibrium to the family decision-making game both with dominance (Stackelberg), and without dominance (Cournot). If the stand-alone contributions are lower than the complete crowding-out contributions, the equilibrium is interior. Proof See the proofs, available on request.12 As regards the reference case, where each spouse has Cournot conjectures about the reactions of the other spouse, it is possible to define conditions so that there is a unique equilibrium in the interior of [0, T]. In our model it is particularly simple, since the reaction functions are linear. The reaction curves are not parallel, which may be seen by solving for the slopes of both spouses’ reaction curves, and noting that the assumption of equal slopes leads to a contradiction. It is then only a matter of scaling to assure that the reaction functions cross in the interior of [0, T].
68
A FAMILY WITH ONE DOMINATING SPOUSE
DOMINANCE AND THE ALLOCATION OF TIME Comparative advantages The traditional welfare maximizing model predicts specialization between the spouses according to comparative advantages. Since we want to analyze the effects of dominance, we examine the case without any comparative advantages. We begin by defining what we mean by comparative advantages: Definition The husband (wife) has a comparative advantage in the market if and only if: (4.9) The following proposition captures the effects of dominance on time allocation and welfare. Proposition 2 Consider the case in which there are no comparative advantages. Then, for an equilibrium in which both spouses supply time to both sectors, the following applies: 1 The total provision of the family-specific public good is inefficient both in the presence and in the absence of dominance, and is strictly lower in the presence than in the absence of dominance. That is: (4.10) 2 Male dominance will decrease the husband’s contribution, and increase the wife’s contribution to the family-specific public good. Concerning relative utilities, it is the case that dominance makes the husband’s utility strictly higher, and the wife’s utility strictly lower. That is: (4.11) Proof See the proofs, available on request. First, the proposition says that the effect of dominance is to lower the total provision of the familyspecific public good. When the husband dominates, he will lower his contributions to the family, knowing that the wife will then increase her contributions. However, the wife’s increased contributions will never totally compensate for the lower contributions made by the husband, and the effect is that dominance decreases the total provision. Also incomes from market work are affected by dominance. The change in the allocation of time to market work means that the husband’s income is increased and the wife’s income is decreased by dominance. Therefore, as stated in the second part of the proposition, utilities are affected by dominance. Since the total provision of the family-specific public good, and the wife’s income, are lowered by dominance, the utility of the wife is lower when the husband dominates than when he does not. As regards the husband, the situation is somewhat more complicated. His income from market work is increased by dominance, which will increase his utility, but at the same time the total provision of the family-specific public good is decreased by dominance, which will decrease his utility. The net effect on the husband’s utility can be obtained by noticing that the husband could choose to contribute to the family-specific public good at the “no dominance level”, which would induce the wife to choose the no-dominance level. But at this level the marginal utility for the husband of contributing to the family is negative, and hence his utility must increase when he lowers his contribution marginally. Thus the net effect of dominance on the husband’s utility must be positive. One further thing to notice is that when there is no dominance, the husband and the wife may attain the same utility level, which occurs when there are no comparative advantages. But when there is dominance by
KRISTIAN BOLIN
69
the husband, he attains a strictly higher utility level than the wife, even in the case where there are no comparative advantages. This is thus caused only by the force of dominance. It is not only comparative advantage that is of importance for the family’s allocation of time. Although the husband and the wife may be identical with respect to preferences and comparative advantages, the husband will supply more time to the market than the wife. This result is different from the result obtained from the welfare maximizing models, where only comparative advantage matters. Sharing and efficiency We have seen that in the absence of complete contracts, the model with dominance left us with an allocation of time which is not efficient. However, it should not be a controversial statement that most marriages are at least partly built on agreements between the spouses. Therefore, a more realistic model would allow for some agreements to be enforceable. One area where contracts may be possible to agree upon, to verify and thus to enforce is the division of market earnings. Allowing for sharing of market earnings, we are left with a model that is non-cooperative in its decision-making concerning the allocation of time, but allows for cooperation when it comes to the distribution of market earnings.13 Consider the following sharing scheme: incomes are pooled and shared so that the husband receives a fraction a of the joint market income. Consequently the wife receives the fraction (1−a). When there are no comparative advantages, then in the absence of dominance, a sharing scheme such that a=1/2 will result in an efficient allocation in which both spouses may supply time to both sectors.14 It turns out that when there are comparative advantages and no dominance, any sharing scheme such as 0
70
A FAMILY WITH ONE DOMINATING SPOUSE
THE EFFECTS OF DOMINANCE ON MARITAL TRANSFERS Spousal support is a transfer which will occur in the case of divorce. A marital transfer is a transfer which takes place within marriage.16 The effects of spousal support on time allocation etc. will have the same direction as the effects of marital transfers. Therefore, spousal support may be discussed, relying on the same formal analysis as in the case of marital transfers, even though spousal support has not formally been analyzed. Let S be a marital transfer from the husband to the wife, i.e. S≥0 when the husband pays his wife, and has the opposite sign in the reversed case. The expected utilities are: (4.12) and (4.13) Proposition 4 Consider the cases where the husband has a comparative advantage in the market, or where there are no comparative advantages. Then if both spouses share their time between the family and the market, it is the case, both in the presence and in the absence of dominance, that: (4.14) Proof See the proofs, available on request. This result applies both in the presence and in the absence of dominance. The importance of dominance is: Proposition 5 In cases where proposition 4 holds, a marital transfer is Pareto-improving, since it increases the provision of the family-specific public good. The significance of dominance is the following: marital sharing increases welfare more in the case of dominance than in the case of no dominance. Further, the transfer from the husband to the wife is larger in the case of no dominance. (4.15) Proof See the proofs, available on request. In the presence of dominance, marital sharing is potentially more beneficial. Hence, in the presence of dominance it is less likely that high transaction costs will prevent marital sharing from occurring. In our model, the husband will share his outside earnings at least until the wife or the husband himself no longer chooses an interior allocation of time. This means that sharing will have a strong impact on the spouses’ respective time allocations. As a result of this, it will improve efficiency. However, the distributional effects of a transfer from the husband to the wife are partly offset by the subsequent reallocation of time. The husband, who has a comparative advantage in the market, makes “a proposal that the wife cannot refuse,” and, hence, marital sharing may also be identified as reinforcing the effects of dominance and of comparative advantages on the relation between spouses’ time allocations. How much of his market earnings will the husband choose to share with his wife? As already stated, it will be profitable for the husband to share his income with the wife at least as long as the allocation is interior. In other words, as long as both spouses supply time to both sectors, the husband will propose a greater transfer. Proposition 5 states the relation between sharing in the presence of dominance and sharing in the absence of dominance. A correct intuition would be that dominance implies less sharing. To arrive at this conclusion, remember that dominance implies a more unequal time allocation than would be the case without dominance: that is, one may suspect that less sharing would be needed to make the allocation non-
KRISTIAN BOLIN
71
interior. Indeed, when the husband does not dominate, it will take more sharing to drive one (or both) spouse into specialization. Spousal support may also be examined within this framework. Formally, redefine utilities as (4.16) and (4.17) Qualitatively, this will produce exactly the same results as was the case with marital sharing. Why then is the occurrence of explicitly expressed divorce contracts that award spousal support so rare? One possible explanation is, again, high transaction costs. In this case it is generally argued by economists in the field of law and economics that the legal system should provide to the individuals the set of contracts which they would acquire themselves in the absence of transaction costs. It is not feasible to legislate transfers within marriage. But transfers in the case of divorce may be subject to legislation. In the Swedish case the question which then immediately arises is why very few divorcees in practice are awarded spousal support.17 A utilitarian government would legislate at least the same spousal support as would be contracted by the spouses.18 This is so since spouses would increase the spousal support as long as it is Pareto-improving, and if it is Pareto-improving, the utilitarian surplus must also increase: hence, the government would increase marital sharing. It is also interesting to notice that if the government is utilitarian, the lack of legal support for spousal support could not be explained by women being discriminated against by the government, in the sense of the government putting less weight on female utility. Suppose that the government maximizes only male utility. Then, since spousal support is Pareto-improving, the government would choose at least as high spousal support as the spouses themselves would choose in the absence of transaction costs: that is, the government would choose S>0. Finally, since marital transfers and spousal support have qualitatively the same effect on the time allocation decisions of the spouses, it is possible that an efficient allocation is reached using marital sharing. Assuming that contracts for marital transfers can be negotiated without any transaction costs, this suggests, in the case of a utilitarian legislator, an explanation for why marital sharing is not legally supported: because spouses have already, using marital transfers, attained the best time allocation. DOMINANCE AND FAMILY POLICY One important role for the state is to provide substitutes for family-specific public goods, such as education of children and childcare.19 We will now examine the significance of dominance for the public provision of such goods. In Sweden the public childcare system and its effects on the intrafamily time allocation have been discussed for years. It has been argued that a public childcare system is necessary if we want women to be able to compete with men in the labor market and not be forced back into their old roles as housewives. To examine this issue, suppose that the government provides a substitute for the time of the family members devoted to the provision of the family-specific public good. The substitute is paid for by the husband by fraction , and by the wife by fraction , throughlump-sum taxation. The productivity of the substitute for family-specific time is denoted by . The production of the family-specific public good is, then, with the government’s time provision denoted by ig:
72
A FAMILY WITH ONE DOMINATING SPOUSE
The government’s opportunity cost in the provision of the substitute is reflected by . Then if provides ig, then the total tax, T, is
(4.18) . If the government (4.19)
the government has a comparative disadvantage in the provision of time used to produce the family-specific , is the efficiency of the public provision of the substitute for the family-specific time: good. The ratio, a higher ratio means that the public provision of the substitute is less efficient than the private provision. To be able to examine the effects of dominance on the government’s willingness to provide the substitute for family-specific time, suppose that we have a utilitarian government: that is, the objective of the government is to maximize the sum of expected utilities of the husband and the wife. The government’s objective function, G, is thus: (4.20) Let the efficiency of the least efficient government which chooses to provide the substitute in the absence of and the efficiency of the least efficient government which chooses to provide the dominance be . Thus, if > , it is possible that a government substitute in the presence of dominance be that would not choose to supply in the absence of dominance, would choose to do so in the presence of dominance. With public provision of a substitute for the family-specific time, the expected utilities become respectively: (4.21) and (4.22) . First we will consider the case in which the tax, T, is divided equally between the spouses, i.e. Proposition 6 Consider an interior equilibrium in which there are no comparative advantages. Then, it is the case that (4.23) Proof See the proofs, available on request. The result says that, given that the husband dominates, the least efficient government which would supply when the husband dominates would not supply when the husband does not dominate. Next, let the government choose both how much to supply and the tax structure, . That is, ig and are chosen by the government so that its objective function, G, is maximized. Proposition 7 Consider an interior equilibrium in which there are no comparative advantages. In this case, the government will levy the tax on the dominating part, in the presence of dominance, but is indifferent to tax structures in the absence of dominance. The government would never supply if . Moreover, if the government supplies, the case is the following: (4.24) Proof See the proofs, available on request.
KRISTIAN BOLIN
73
The necessary condition for public provision is interesting because Konrad and Lommerud (1995) examine a case where >max . In this case a utilitarian government would not provide any substitute for the family-specific time at all. When the government chooses both the tax structure and the level of provision of the substitute and the husband dominates, it will tax only the husband, if there are no comparative advantages. In the case of no dominance, the government may choose any tax structure if there are no comparative advantages. Now, assume that the government is able to choose the tax structure but is forced to supply the substitute independently of the relationship between the government’s efficiency and the private efficiency, i.e. there may be a public provision even when .20 In this case: (4.25) If this is the case, both spouses would be worse off through the public provision.21 In addition, such a provision would have adverse effects on the intrafamily distribution of private consumption, as the wife would supply more time to the household, while the husband would supply less. When the government chooses an efficient provision of the substitute, both spouses will be better off. This provision might also have a redistributive effect on private consumption, as also the wife supplies less time to the household in this case. One alternative to a public provision of a substitute for family-specific time for policy-makers who want to change the family time allocation is to transfer money from one spouse to the other. The government might want to interfere and tax the husband and transfer money to the wife. One example is child allowances, which, at least in Sweden, are paid to the mother. Proposition 7 states that this will make a time allocation, in which the wife contributes more than the husband to the family, even more unequal. However, the total provision of the family-specific public good will increase. In other words, a policy intended to change the intrafamily income distribution makes this distribution even more unequal, but on the other hand has an efficiency-improving effect in that the total supply of the public good increases. Further, a policy intended to improve efficiency through a public provision of the family-specific public good will be partly offset by a reduction in the spouses’ own contributions: hence, the public provision will increase the income of the spouses. Since the tax is levied on the spouse with the higher income, the public provision will have a redistributive effect. This point has been noticed by Konrad and Lommerud (1995). CONCLUSION The raising of children, traditions and physical differences between the sexes all generate comparative advantages by gender, which is the prevailing explanation for the relative time allocation of wives and husbands within marriage. However, in the industrialized countries, there is a trend towards more male participation in the raising of children. This means that to a larger extent than before men are sacrificing a part of their investments in market-specific human capital in favor of their investments in family-specific human capital.22 What is interesting is that it seems that this tendency toward more male participation in the household has not been followed by a corresponding equalization of men’s and women’s time allocation outside the household: rather, as discussed in Konrad and Lommerud (1995), the increase in women’s labor supply has occurred in traditional female occupations, which has preserved occupational sex segregation. Besides, many women have chosen to work part time, still leaving them with the main responsibility for household work.
74
A FAMILY WITH ONE DOMINATING SPOUSE
Why have women to a larger extent chosen not to work full time, and avoided occupations that have traditionally been connected with a male labor supply? It is here that our hypothesis of dominance comes in. It states that a woman entering a marriage has a strategic disadvantage in that her husband has already made his choices, and she has to make her choices given what her husband has already done. We say that the husband dominates family decision-making. We have examined a strategic model of family time allocation. The predominant welfare maximizing model in family economics stresses the importance of comparative advantage for the time allocation decisions in the family. We have shown that comparative advantages may not be the sole determinant for time allocation. The timing of decisions is also of vital importance. In our setting, the husband moves first, knowing the response of the wife. The result is that the husband supplies more time to the market, and the wife supplies more time to the household, compared to a situation in which spouses move simultaneously. Most interestingly, this is so even in the case where there are no comparative advantages. If this is the case, then the inequality between the time allocations of men and women will prevail, even though the wages may become more equal, and men and women become equally efficient in the household. Although there are no comparative advantages, the wife supplies more time to the household in the presence of dominance than in the absence of dominance. The dominance by the husband is a complementary explanation for why men contribute less to the family than do women. A common explanation is that men have higher earnings and better career opportunities, and therefore their time at home is more expensive than their wives’. But our model suggests that men will contribute less to the family even if their time at home is not more expensive. The implication is that the inequality between the time that men allocate to the market and the time that women allocate to the market will prevail even though relative efficiencies may change so that men’s comparative advantage in the market will vanish in the long run. To obtain equality in time allocation, the reasons for male dominance must be recognized by women. When there are no comparative advantages, our intuition tells us that it ought to be feasible to obtain an efficient allocation of time, in which both spouses supply time to both sectors and share market earnings equally. When the husband does not dominate, this intuition is correct. When the husband dominates, however, it is not correct. In the presence of dominance, it is not possible to have equal sharing, efficiency and supply of time to both sectors by both spouses. In other words, a household in which the husband dominates, and both the husband and the wife perform household and market tasks, and share their market earnings equally, cannot be efficient. We have also discussed a policy issue. The result was that the presence of dominance might cause a government to supply a substitute for the family-specific time at a greater cost disadvantage than in the absence of dominance. It was also shown that a utilitarian government would make the dominant party pay for the provision, in a situation with no comparative advantages. REFERENCES Becker, Gary S. (1973) “A Theory of Marriage: Part 1,” Journal of Political Economy 81:813–46. Becker, Gary S. (1974) “A Theory of Marriage: Part 2,” Journal of Political Economy 81:11–26. Becker, Gary S. (1991; first edn 1981) A Treatise on the Family, Cambridge, Mass.: Harvard University Press. Bergstrom, T., Blume, L. and Varian, H. (1986) “On the Private Provision of Public Goods,” Journal of Public Economics 29:25–49. Bergstrom, Theodore and Bagnoli, Mark (1993) “Courtship as a Waiting Game,” Journal of Political Economy 101: 185–202.
KRISTIAN BOLIN
75
Konrad, Kai and Lommerud, Kjell Erik (1995) “Family Policy with Non-cooperative Families,” Scandinavian Journal of Economics 97:581–601. Landes, Elisabeth M. (1978) “Economics of Alimony,” Journal of Legal Studies 7:35–63. Leuthold, Jane (1968) “An Empirical Study of Formula Income Transfers and the Work Decisions of the Poor,” Journal of Human Resources 3:312–23. McElroy, Marjorie B. and Horney, Mary J. (1981) “Nash-bargained Household Decisions: Toward a Generalization of the Theory of Demand,” International Economic Review 22:333–49. Manser, Marilyn and Brown, Murray (1980) “Marriage and Household Decision-making: A Bargaining Analysis,” International Economic Review 21:31–44. Ott, Notburga (1992) Intra-family Bargaining and Household Decisions, Berlin: Springer Verlag. Varian, H. (1994) “Sequential Contributions to Public Goods,” Journal of Public Economics 53:165–86. Weiss, Yoram and Willis, Robert J. (1985) “Children as Collective Goods and Divorce Settlements,” Journal of Labor Economics 3:268–92. Weiss, Yoram and Willis, Robert J. (1993) “Divorce Settlements: Evidence and Interpretation,” Journal of Labor Economics 11:626–79.
NOTES 1 This model is one of several efficiency models, differing in the choice of which specific objective function is being maximized. 2 Manser and Brown (1980) apply the Nash bargaining solution to intrafamily decision-making. In their model explicit consideration to differences in spouses’ preferences is taken. The advocates of this approach (see, for example, Ott, 1992), stress the importance of cooperation in marriage. 3 Many arguments have been given for why spouses, in spite of the absence of perfect contracts, should be expected to behave in a cooperative way. For example Ott (1992) gives reputation as a reason for why spouses do not breach what was informally agreed on concerning the conduct of marriage. However, very few people have sufficiently many marriages during their lifetime for the reputation argument to be plausible. 4 Non-cooperative analyses of the family are rare, however. This may be so because most economists agree with Ott (1992), who argues that a husband and wife who behave non-cooperatively would not be able to reach any higher utility than they would reach on their own, and hence the reason for forming a common household would vanish. In a recent paper Konrad and Lommerud (1995) model the family as a non-cooperative game in which the spouses have to decide how much to supply of a family public good. Besides, this book contains an article by Lommerud in which he discusses non-cooperative approaches in more detail. Also the article by Lundberg and Pollak treats non-cooperative approaches to the family. 5 At this point it should be noted that it might well be the wife who is the dominating party. 6 The Stackelberg game is perhaps most easily understood by referring to the application to oligopoly. The situation is that two companies compete with identical products. They are tied together by the market demand function. The Stackelberg game analyzes this situation by assuming one company to be the leader (L), and to choose its output first. Then the other company, the follower (F), can observe the output of (L) and choose its output so as to maximize profits given the output of the leader. The important thing is that for each output he chooses (L) knows what output (F) will choose, and thus, how his own profit will be indirectly affected. Hence (L) has to choose his output level given his knowledge of the reactions of (F). 7 Konrad and Lommerud use the Cournot game, with simultaneous decisions, to model family decision-making. We want to stress the fact that spouses are not identical, and hence we use the leader-follower or Stackelberg game, which does not treat spouses as identical. 8 This follows Varian (1994). 9 We assume that there are only two alternative ways in which time can be used: either supplied to the market or supplied to the domestic production of the family-specific time. In other words, there is no leisure.
76
A FAMILY WITH ONE DOMINATING SPOUSE
10 The preferences of the spouses may of course differ, but we are only interested in the effects of dominance. These effects are best demonstrated in the simplest model with dominance as the only difference between the spouses. 11 The parties are said to have Cournot conjectures. 12 For reasons of space, we have chosen to exclude the formal proofs. However, they will be provided by the author upon request. E-mail: [email protected] 13 The family legislation in Sweden is even formulated in terms of each spouse being obliged to contribute financially what is needed to satisfy the personal needs of the other spouse. I am not sure whether this is also the case in other countries with a similar legal history. 14 In Konrad and Lommerud (1995), an efficient allocation of time and specialization according to comparative advantage result if incomes are pooled and shared equally. In their model, a solution in which both spouses supply strictly positive amounts to both the market and the household is not possible in the presence of both a comparative advantage for one spouse and equal sharing. However, when there are no comparative advantages, an efficient allocation results in which both spouses may share their time between the family and the market. This is the result of the absence of dominance, i.e. of both spouses having Cournot conjectures. 15 Again, the proof may be obtained by request. 16 Bergstrom et al. (1986) show that in a Cournot game, when there are no comparative advantages, the total supply of the public good will be unchanged when the income is redistributed among the contributors. Varian (1994) shows an analogous property in the Stackelberg game. Here, we also consider the situation when there are comparative advantages. 17 This is a difficult matter to discuss in an international framework because legal rules differ. 18 A utilitarian government is a government that maximizes the sum of utilities of the individuals in the society. In our case the society could be thought of as constituted by n identical families. Hence, a utilitarian government simply maximizes the sum of utilities of two representative spouses. 19 This is discussed by Konrad and Lommerud (1995). 20 The cause of such an inefficient supply might be that different groups, such as labor unions, are able to influence the decisions of the government. It is plausible that these groups do not have the same objective as the government, but instead seek to maximize the utility of their members. 21 The comparative statics are shown in an appendix together with the proof of proposition 7 (available by request). Given the signs above, it is straightforward to see that both spouses must be worse off; since the equilibrium is interior, and the aggregate provision of the public good goes down, the private consumption must also go down for both spouses for the first-order conditions to be satisfied. 22 In Sweden, a law concerning the distribution of parental leave between parents has been passed. The law specifies the shortest time the father must stay at home if the family is to be allowed to use the maximum amount of paid parental leave.
5 WOMEN’S HOURS OF WORK AND MARRIAGE MARKET IMBALANCES Shoshana Grossbard-Shechtman and Matthew Neideffer
INTRODUCTION Married women’s labor supply has been the focus of a substantial amount of research. Since the 1960s the theory forming the base for that research has been Mincer’s (1962) and Becker’s (1965) theory of allocation of time to work and home. Research in the last 30 years has dealt mostly with econometric issues regarding the proper way to estimate the labor supply function (e.g. Heckman 1993; Smith 1980). Grossbard-Shechtman (1984; 1993) has expanded the theoretical model of married women’s labor supply by incorporating marriage decisions into the analysis. This paper presents a modified version of the Grossbard-Shechtman model more compatible with standard practice in labor economics. Also, this paper puts the Grossbard-Shechtman model of labor and marriage into perspective, by offering comparisons with traditional labor supply theory, Becker’s (1973) theory of marriage and household bargaining theory. Finally, this paper contributes to the existing literature by testing the effect of marriage market imbalances on women’s labor supply. Giving marriage decisions an active role in the analysis opens the doors to new generations of econometric models, of which one is estimated and presented in this paper. We use microlevel data from the 1990 US census. THEORY Models of labor supply consider the relationship between market wage and reservation wage, i.e. the value an individual places on non-market activities. They posit that when the market wage exceeds the reservation wage a person will enter the labor market. One of the ways in which labor supply models have evolved consisted in expanding the utility functions from which reservation wages were derived. Early models such as Robbins (1930) assumed that an individual utility function was being maximized, leading to that individual’s choice of own leisure and market goods. The reservation wage was the value of individual leisure time. Jacob Mincer (1962) and Gary Backer’s (1965) theory of allocation of time introduced a switch from individual utility maximization to family utility models. Since then, most labor supply models have viewed individual labor supply as the result of a decision process whereby a couple maximizes its joint utility function subject to a combined income constraint and time constraint. Consequently, household characteristics—such as spouse’s income and number of children—have been incorporated as factors affecting an individual’s reservation wage and therefore individual labor supply. The following reservation
78
WOMEN’S HOURS OF WORK
wage function, which is commonly used in the literature, derives from a family-decision model in the Mincer-Becker tradition: (5.1) where h is hours of work, I is all sources of household income less the amount of wage income earned by the wife, P is a vector of prices for household goods, and A is a vector of individual traits which are related to the wife’s employment characteristics (Heckman 1974). Family-decision models have proven useful in bringing in considerations regarding the home-based determinants of reservation wages, but are of little use when the objective is to study the interrelationship between marriage and labor market outcomes. The “household” that inhabits the utility functions of most contemporary labor supply models has no intention to either marry or divorce. Consequently, these models do not offer a straightforward way to incorporate marriage market conditions into the modeling of labor supply decisions. While drawing on Becker’s (1965) theory of allocation of time and Becker’s (1973) theory of marriage, Grossbard-Shechtman (1984; 1993) offers the first model which formally integrates an economic analysis of marriage into the analysis of reservation wages and labor supply.1 Grossbard-Shechtman abandons a 30year-old tradition of family-decision labor supply models in favor of a 60-year-old tradition of individualdecision labor supply models. It follows from her theory of allocation of time to labor and marriage that the reservation wage depends on more factors than the factors included in equation 5.1, one of these factors being marriage market conditions.2 This integration between labor supply and marriage decisions is accomplished by introducing two new concepts: spousal labor and the quasi-wage for spousal labor. In essence, Grossbard-Shechtman analyzes the decisions to marry and supply labor as an occupational choice, one choice of occupation being spousal labor. Work is called “labor” if it benefits an employer outside of marriage and is called “spousal labor” if it benefits a spouse.3 Spousal labor is work not only in the sense that it is an activity often generating some disutility which has an opportunity cost in terms of foregone leisure. It is also work in the sense that most people expect to get compensated for engaging in spousal labor. There are no statistics on quasi-wages for spousal labor, a familiar problem in a literature which is used to deal with other concepts which are not directly measurable, such as reservation wages. Individual utility maximization leads to the derivation of two supplies of labor: labor and spousal labor. The same individual utility-maximization process also leads to the derivation of a demand for spouse’s spousal labor. In deciding how much spousal labor to supply, individuals compare the quasi-wage for spousal labor with wage offers in other lines of work. In deciding how much spousal labor to consume, individuals consider the cost of spousal labor, the cost of other goods and services, and their income. More formally, consider an individual i who derives (dis)utility from working in spousal labor (mi) and benefits from spouse’s j spousal labor mj.4 The individual utility function can be represented as (5.2) where i, j=f, m(m=male, f=female, i j), h denotes time allocated to labor, m is spousal labor, s is selforiented time (usually called leisure), and x denotes commercial goods. The budget constraint also undergoes a modification in comparison to standard labor supply models. Standard models have two sources of income: work and other (often called non-work) income. In Grossbard-Shechtman’s model income can also be obtained from a third source: spousal labor mi, and an additional type of expenditure is added: spousal labor by a spouse mj. The budget constraint thus becomes: (5.3)
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
79
where w is market wage for labor, y is quasi-wage for spousal labor, V is non-wage income, and p is a price vector for commercial goods and services. Individuals decide on how to allocate their time by maximizing the utility function (5.2) subject to the budget constraint (5.3) and the time constraint (5.4) where T is the maximum time available (e.g. 24 hours per day). It follows from the first-order conditions that individuals who work in both types of job will attempt to reach the following equilibrium: (5.5) Equation (5.5) indicates that time is allocated between two occupations so that the wage in one occupation equals the wage in the other occupation, plus the monetary equivalent of the difference in marginal utility produced by these occupations. The wage in equation (5.5) can be interpreted as a reservation wage. In order to be drawn into the labor force, the individual requires a wage equal at least to y, the quasi-wage available from spousal labor, plus the difference in non-pecuniary benefits generated by the two kinds of work. To the extent that work is less enjoyable than spousal labor, the reservation wage has to exceed y. The factors influencing the reservation wage according to the standard equation (5.1) also enter in a reservation wage function based on this analysis. It follows from equation (5.5) that hours of work matter as they influence the marginal utility of labor and its alternatives. Individual traits matter as they influence both a person’s utility and the quasi-wage for spousal labor. Household prices matter as they influence the amount of goods and services a person will buy and therefore the value of marginal utility of work and spousal labor. In the case of a married woman equation (5.1) included an income variable defined as household income less the amount of wage income earned by the wife. Income other than own income from work also enters a reservation wage function based on equation (5.5). However, Grossbard-Shechtman’s analysis leads to the derivation of four separate income effects, replacing the two income effects (income from work and non-work) found in standard labor supply models. Furthermore, her model points to a number of factors that are expected to influence the reservation wage which were not included in standard models. Four income effects Traditional models consider two sources of income: income from own work wihi and non-work income (I in equation 5.1). This latter category includes income earned by the spouse and income from sources other than work. According to Grossbard-Shechtman, an individual i decides on how much time to allocate to work h, or alternatively determines a reservation wage associated with a given amount of work, as a result of an individual maximization process described by equations (5.2) to (5.4). Own income from sources other than work (or spousal work) Vi enters the budget constraint and therefore i’s reservation wage. In addition, i’s reservation wage is also a function of yi, the compensation the individual can receive for spousal labor, the total income from this source being yimi. So far we have identified three sources of own income: income from work, income from spousal work, and non-work income. In this model decisions are made by individuals and not by couples. Before the effect of spouse’s income on own labor supply can be analyzed, marriage has to be introduced in the analysis. Following Becker (1973) Grossbard-Shechtman models marriage as a voluntary utility-maximizing transaction. In contrast to Becker’s model which compares individual production levels with and without marriage, Grossbard-Shechtman conceptualizes marriage as an exchange of spousal labor. Individuals
80
WOMEN’S HOURS OF WORK
marry when the amount of spousal labor they demand corresponds to the amount of spousal labor a spouse supplies at given quasi-wages for spousal labor.5 Spouse’s income enters the analysis as follows. If marriage entails that individual i is a net supplier of spousal labor to spouse j and the quasi-wage for spousal labor y is positive, then j has to pay an amount yimi to i to induce the supply of spousal labor.6 This monetary or material transfer from spouse j to spouse i is made in the form of a proportion of the spouse’s income Ij from sources other than j’s spousal work. In other words, (5.6) It is expected that k>0 to the extent that y>0 and that k
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
81
demands which also derive from an individual maximization process based on equations (5.2) through (5. 4), in the following way. Individuals j decide on how much spousal labor mi they want to consume at any given level of yi. The aggregate demand in the market for spousal labor Mi consists of the sum of all derived demands by individual j’s for labor by spouses i. Markets for spousal labor establish equilibrium values of yi, the quasiwage for spousal labor supplied by individuals i, and similarly for yj. If women are the i’s and they are net producers of spousal labor they are expected to receive a transfer from their husbands amounting to yimi–yjmj, which will be a proportion of the husband’s income. Not only are markets for spousal labor by women interrelated with markets for spousal labor by men, but markets for spousal labor are interrelated with markets for labor. Labor and spousal labor being two substitute forms of employment, the wage for labor influences the supply of spousal labor and the quasi-wage for spousal labor influences the supply of labor. Also, the demands for male and female labor by employers are related to conditions in marriage. The same products can be produced either by married couples or by firms, so that labor and spousal labor are also substitutes on the demand side. Consequently, markets for labor and spousal labor are tightly interrelated. Figure 5.1 presents four such markets—markets for male and female labor and markets for male and female spousal labor— based on the assumption that labor markets are segmented by gender (even though many employers may substitute between male and female employees). In a competitive equilibrium, these markets establish market clearing wages wf and wm and quasi-wages for spousal labor yf and ym.9 Let us compare a situation of equal numbers of marriage eligibles in markets for spousal labor to a situation with a marked excess of men. Initially, an excess of males is expected to lead to a high demand for women’s spousal labor (D' in Figure 5.1 (a)), a large supply of men’s spousal labor (S' in Figure 5.1(b)), and a large supply of men’s labor (S' in Figure 5.1(d)). An excess of males is equivalent to a shortage of females (keeping population size constant).10 Women are expected to be in limited supply in markets for spousal labor (panel a) and labor (panel c), and their demand for men’s spousal labor will be limited (panel b). The initial effect of an excess of males thus results in (1) an increase in women’s quasi-wage for spousal labor yf (panel a); (2) a decrease in men’s quasi-wage for spousal wage ym (panel b); (3) an increase in women’s wage wf (panel c); and (4) a decrease in men’s wage wm (panel d). It follows from equation (5.5) that the higher the quasi-wage for spousal labor, the higher an individual’s reservation wage. An excess of males in the market for female spousal labor causes an increase in women’s quasi-wage and therefore in women’s reservation wage. This predicted effect is based on the assumption that husbands do not automatically share their income with their wives. One way of incorporating this theoretical discussion into reservation wage equation (5.1') is by recognizing that the degree to which husbands share their own income with their wives, k, is a function of marriage market imbalances. For instance, if the ratio of marriageable men to women of a given age group (i.e. the sex-ratio) is used as a measure of marriage market imbalance and denoted by SR, it follows that k=k (SR), where the first derivative of k with respect to SR is positive. The higher the number of men relative to the number of women in a given marriage market, the more men are willing to give women access to their income. In calculating SR one needs to include all men and women who participate in the same marriage market, i.e. who actually enter or possibly enter or re-enter this market for spousal labor. This includes married and unmarried individuals. Individual access to spouse’s income k is not simply a function of SR, sex-ratio. Quasi-wages for spousal labor y vary not only with demand and supply in spousal labor markets but also with monopolistic elements
82
WOMEN’S HOURS OF WORK
Figure 5.1 Markets for (a) Female spousal labor; (b) Male spousal labor; (c) Female labor; (d) Male labor. Primary effects of an increase in the relative number of men
which enter into an existing marital relationship to the extent that divorce and marriage are costly (e.g. due to search costs). In that case one could find a divergence between the market y and the y obtained by a particular individual. Consequently, y can be separated into two elements: , where ye is the equilibrium value of quasi-wages for spousal labor in the market and b is a bargaining component specific to the individuals in a particular couple. Household bargaining models based on game theory can help understand factors influencing b. Overall, there is no good reason to believe that markets for spousal labor are fundamentally different from other labor markets. The degree to which a competitive market model is applicable will vary with circumstances, as is the case with any economic analysis. Sex ratio effects on k and therefore on reservation wage and labor supply are expected to be stronger the more y and k are influenced by the market equilibrium ye and the further a particular couple is from dual monopoly. Proportion k may thus be a function of SR and of marriage duration. The longer a couple has been married, the more divorce costs and monopolistic elements are expected to influence individual decision-making. Grossbard-Shechtman’s (1984, 1993) model leads to predictions regarding the effect of marriage duration and sex-ratios on k, y, and reservation wage which differ from those based on game theory. The games described in household bargaining models start at marriage. Wife’s and husband’s relative bargaining positions are a function of remarriage possibilities, and therefore sex-ratios. Marriage market imbalances faced prior to marriage are not considered in game-theory models. It follows from the Grossbard-Shechtman
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
83
model that sex-ratio effects will be strongest at the beginning of a marriage and will become less important as marriage duration increases. From a game-theory perspective, marriage market imbalances start playing a role only to the extent that divorce and remarriage become concrete possibilities, which may not occur at the very onset of marriage. We can now expand equation 5.1' to 5.1" (5.1") where SR denotes sex-ratio, and Aj are traits of spouse j valued in marriage markets. The reservation wage function defined in (5.1") thus differs from the traditional reservation wage function (5.1) in the following ways: our analysis leads to the inclusion of hours of spousal labor m in the reservation wage function; our analysis leads to separate income effects as explained above; in Grossbard-Shechtman’s analysis sex-ratio SR is expected to affect y and k and therefore reservation wage w*; and individual traits of spouse j (Aj) enter the reservation wage function. The first derivative of w* with respect to SR will be positive. The partial effect of mi on w* is expected to be negative: the more i works at spousal labor, the lower the marginal utility of m and therefore the less i has to be induced financially to go to work. Positive traits Aj which compensate for income transfers will be negatively related to w* (Grossbard-Shechtman and Neuman 1988).11 The implication of Grossbard-Shechtman’s analysis of labor and marriage that will be tested here is the sex-ratio effect on labor supply. More specifically, it is hypothesized that: ceteris paribus, women, especially married women, are less likely to work outside of marriage when there is an excess of marriageable males than when the number of marriageable men and women is balanced. Furthermore, the number of hours worked by women is expected to be a negative function of the relative excess of males over females. Vice versa, marriage markets with an excess of females are expected to be associated with higher participation of women in the labor force than balanced marriage markets. A qualification to this analysis is that a relative shortage of females may also cause women’s wages in the labor market to increase. It is expected that the increase in women’s quasi-wage y will exceed the increase in women’s wage outside of marriage w. If the reservation wage increases more than the wage, it is expected that marriage markets with an excess of males will be associated with lower participation of women in the labor force than balanced marriage markets. There are good reasons why imbalances in the numbers of men and women are expected to affect marriage market conditions more than labor market conditions. For some aspects of marriage heterosexual spouses have no substitutes. Consequently, men and women are poor substitutes when it comes to marriage. In contrast, men and women can often be excellent substitutes at the workplace, especially in a work environment such as the contemporary US which condemns discrimination according to gender. Given that the total number of workers does not change, quasi-wages for spousal labor will be affected considerably more by an excess of males than wages for male and female labor. Furthermore, lower male incomes associated with an excess of males will also cause income effects, and therefore a shift back to the left in the demand for women’s spousal labor. However, the effect of these income changes is not expected to neutralize the rightward shift in demand for women’s spousal labor caused by an excess of males. After all changes are integrated, it is expected that the quasi-wage for women’s spousal labor will grow with the relative excess of men over women, which leads to the marriage squeeze hypothesis above. A number of factors are expected to affect the relationship between marriage market imbalance (relative excess of one gender) and women’s labor force participation. Marriage market imbalances are more likely
84
WOMEN’S HOURS OF WORK
to affect married women than unmarried women (although it is also expected to apply to unmarried women who prepare themselves towards a career in marriage); women employed full time than women employed part time; women employed in low-paying occupations than women employed in high-paying occupations; and women with lower education (see Grossbard-Shechtman 1993). The prediction that marriage market conditions also affect the labor supply of unmarried women does not follow as easily from household bargaining models. From a marriage market perspective, unmarried women with better marriage offers are likely to work less while single or to choose occupations compatible with the fewer hours of work they plan to work outside the home when married. With respect to part-time ν. fulltime work, it is expected that the effect of marriage market imbalances on hours of work is not linear, the marginal effect being stronger when people already work many hours. One reason for that is that women working long hours are more likely to work for monetary reasons, and therefore variations in the monetary rewards for spousal labor associated with marriage market imbalances may have more effect on their behavior than on that of women working part time. Likewise, women with a high education and women employed in high-paying occupations are more likely to work for non-monetary reasons than their less skilled counterparts, and therefore their behavior is less likely to respond to marriage market imbalances affecting monetary rewards for spousal labor. Effects of marriage market imbalances on female labor supply have been tested at an aggregate level, using both time series and cross-sectional data. Grossbard-Shechtman and Granger (1994) analyzed time series data and used changes in cohort size to examine changes in labor supply. On average, husbands in the United States are two years older than their wives, which implies that changes in cohort size influence the w* of women. In comparison to women born during periods of stagnant or declining fertility, women born during periods of rapid population growth will have a choice of a relatively small number of slightly older men who are potential husbands. In contrast, men born during periods of rapid population growth will have a relatively large number of slightly younger women to choose from as potential wives (relative to men born during periods of stagnant or declining fertility). Women born during rapid population growth, in effect, are faced with an excess of females in marriage markets and therefore will have a lower y, a lower w* and a higher labor force participation rate. The results of the time series study did provide evidence, after controlling for several economic variables, for an increase in labor force participation among women born at the onset of the baby boom and a decrease in labor force participation among women born in times of decreasing fertility. A cross-sectional approach to testing the sex-ratio effect used aggregate data for large cities in the US in 1930 and 1980 and tested for the effects of marriage market imbalances on married female labor force participation rates (Grossbard-Shechtman 1993). The study did find evidence that cities experiencing an excess of males were associated with a lower average married female labor force participation rate. Also, it is generally estimated that Afro-American women in the US experience more of a marriage squeeze, i.e. a relative shortage of males, than Caucasian women. This could help explain why the participation rate of married women of African origin has traditionally exceeded that of married women of European origin in this country, even after control for all variables usually included in labor supply estimations (Grossbard-Shechtman 1985). The data previously used in testing for the effects of marriage market imbalances were aggregate level data. Since the theoretical effects of these imbalances are defined at the individual level it seems that using microdata is more appropriate. Our study is the first to look at the effects of sex-ratio on an individual’s labor supply decision.
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
85
EMPIRICAL ANALYSIS It follows from the marriage squeeze hypothesis presented above that in marriage markets with excess males, as indicated by a high sex-ratio, women will be getting more income from spousal labor and will therefore be less likely to work outside of marriage. We tested this hypothesis using a random sample of the 1990 census Public Use Microdata Sample (PUMS) data. Data A sample was constructed from the 1 percent PUMS of the 1990 United States census. These PUMS data consist of a 1 percent sample of all respondents who filled out the long census form. The sample was limited to married white women between the ages of 25 and 29, whose husband was present in the previous year. The study was limited to white women in light of two considerations: previous research has shown that black and white women’s labor supply are quite different, and we would have needed a larger sample and more resources to analyze black women as well. Also, only urban women were selected, and more specifically women who lived in the 85 most populous Standard Metropolitan Statistical Areas (SMSAs). In total, data on 17, 860 married couples was obtained. The 1990 census PUMS data was used for two reasons. First it allows for the sex-ratio to be calculated from the same data-set. The second reason is that the census data was readily available and easy to work with, making a preliminary study such as this possible. But while the census data has those two advantages it also comes with several disadvantages when compared to using other microdata-sets such as the National Longitudinal Survey. The most important problem is that the census data probably has a higher degree of measurement error. In other microdata studies effort is made to keep the data as accurate as possible by using such techniques as paying respondents and personally interviewing them. Also, longitudinal microdata studies the same people who fill out the forms on an annual basis, which adds to their accuracy. Despite the care taken in gathering large microdata-sets, the validity of the data collected is called into question (Juster and Stafford, 1991). With the census data the potential for incorrect data is even greater, especially when considering variables concerning labor supply. For instance, when asked about their average hours worked and the number of weeks worked in the previous year, the two variables used to compute total hours worked, people will give their normal hours worked per week. This creates a problem with regression analysis since it will reduce the variation of the dependent variable while at the same time increasing the error variance. Those two problems are expected to reduce the fit of the regression line in the hours of work equation. These problems will also affect the wage estimates, since the value for hourly wage was computed using a person’s total reported earnings from working divided by their total hours worked. If, as expected, people overstate their hours worked this will cause the wage rate to be inaccurate. We included information at both the household level and the aggregate level. For each couple we used the following variables: • Wife’s hours worked in 1989, in turn a product of average hours worked per week and number of weeks worked. In turn, this information was used to create a dummy for wife’s participation in the labor force (equals 1 if there were any hours of work in the past year). • Wife’s schooling. The census only reports educational categories, which we constructed into a continuous variable. • Wife’s age.
86
WOMEN’S HOURS OF WORK
• Number of children (fertility). • Husband’s income from all sources. We also attached the following information about the urban area of residence to which each individual household belongs: • Sex-ratio. We calculated the sex-ratio for each SMSA by dividing the number of white men aged 27 to 31 by the number of white women aged 25 to 29. A two-year age difference was used in view of the fact that this is the average difference in age at first marriage for men and women in the US. Only white men were selected given that marriage markets in the US are segmented by white/non-white status. This is a relatively simple and exogeneously determined measure of mate availability, in contrast to some other measures of mate availability that have been used in recent research on the effect of mate availability on marriage rates. Following Goldman (1977) many of these studies incorporate a variety of age preferences, based on observed marriages between men and women of given ages, into the mate availability measure. As age preference is endogeneous (it responds to mate availability) this is an undesirable procedure. Another measure of sex-ratio includes solely the number of unmarried men and women (e.g. Brien 1991). As the percent married responds to mate availability, this measure also suffers from an endogeneity problem. Following Wilson (1987) some researchers studying marriage have limited sex-ratios to the ratio of employed men to all women. A variation on that is incorporating men’s income into the mate availability measure (Wood, 1995). Again, matching between women’s characteristics and men’s characteristics (including income and employment) is endogeneous to the marriage market process, and we therefore preferred to use sex-ratio separately from employment measures. • Unemployment rate for the SMSA in July 1990 (Bureau of Labor Statistics, 1990). We used the total unemployment rate for men and women as we assume that both men’s and women’s unemployment has an impact on the marriage market. From the point of view of a market for women’s spousal labor, men’s unemployment is expected to have an impact on the availability of desirable mates (leads to shifts in demand for women’s spousal labor), while women’s unemployment has an impact on women’s desire to marry (leads to shifts in the supply of women’s spousal labor). Both of these impacts are expected to influence marriage market conditions. Given that local unemployment rates for men and women are highly correlated, we preferred to use a measure of the unemployment rate for both genders. Finally, we included information on whether the city is in the South or the West. The means and variances of all variables in the data-set are in Table 5.1. Methods The model tested is the common labor supply model presented by Heckman (1974, 1980), except that the hours of work equation, which depends on the reservation wage, includes an indicator of marriage market conditions (sex-ratios) and the aggregate unemployment rate, which affect both labor market conditions and marriage market conditions. We estimated the labor supply model in three stages, using both selection biascorrection and instrumental variable techniques. In the first stage a probit regression was estimated on the decision to supply labor so that we could estimate the inverse of the Mills ratio. The inverse Mills ratio was then used in the second stage as a regressor
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
87
Table 5.1 Means and standard deviations
Sex-ratio
All women
Supplied wage labor
Did not supply wage labor
(n=17860)
(n=14238)
(n=3622)
2.17 (.59) 2.17 (.21) 1.93 (1.00) 29915.62 (20660.91)
2.9 (1.08) 32593.18 (29087.42)
13.7 (2.23) 27.18 (1.40)
12.22 (2.71) 27.31 (1.39)
1.09 (.10)
Log of wages Predicted log of wages Fertility Husband’s income Unemployment Schooling Age West South
2.13 (1.09) 30458.63 (22649.60) 5.06 (1.21) 13.4 (2.41) 27.20 (1.40) 24% 23%
Note: Standard deviations in parentheses
in the wage equation, which was estimated using OLS. Then in the third stage, both the inverse Mills ratio and the predicted wages were included in the hours of work regression, which was also estimated using OLS. The last equation was estimated on the subsample of women who had supplied labor. Predicted wages were obtained through estimating the following equation (5.7) where ln W is the log of market wage and was computed by dividing the women’s total wage earnings by the product of the average number of hours worked per week and the number of weeks worked in 1989; Age is a proxy for labor market experience; South and West are dummy variables which equal 1 if the person lives in the South or the West. We estimated two different hours of work equations. The first was specified as (5.8) where ln W is the predicted log of wages; Hincome is the total income earned by the husband; and X is a vector of control variables including the number of children born to a woman, schooling, region, age of woman and local unemployment rate. The second hours of work equation was identical to the first, except for the addition of a sex-ratio variable defined for the local marriage market, (5.9) where SR is the sex-ratio.
88
WOMEN’S HOURS OF WORK
It follows from the marriage squeeze hypothesis presented above that in marriage markets with high sexratios women will be getting more income from spousal labor and are therefore likely to provide less labor outside the household. The coefficient of SR in equation (5.9) is therefore expected to be negative (b4<0). The relationship of the other variables to the amount of hours an individual works is well-established in the literature. Results Table 5.2 presents our results. All the coefficients of variables traditionally included in female labor supply models have the expected sign and are significant. After control for predicted wage the effect of schooling on hours Table 5.2 Regressions of labor force participation, predicted wages and hours of work, married white women ages 25– 29, US census, 1990 In workforce Sex-ratio
Predicted wages
−.27** (5.39)
Predicted wage Fertility Husband’s income Unemployment Schooling
−.41** (1407.6) −.00006** (197.1) −.06** (39.13) .10** (391.8)
Age West South Mills Ratio Constant R-square N
1.27 17580
−.078** (29.41) 9.98** (9.3) .04** (3.33) −.07** (7.82) −.20** (7.82) .27 .12 14190
Hours of work −117.74* (1.93) 649.68** (5.06) −412.84** (22.51) −.004** (12.16) −34.21** (−6.20) −2.33 (.21)
−20 (1.23) 160.59** (9.50) 1115.62** (10.60) 1141.92 .13 14190
Notes: t-statistics in parentheses; ** indicates significance at p>.01; * indicates significance at p>.05
of work is negative and insignificant, indicating that schooling and wages are very correlated. Increases in fertility, husband’s income and unemployment rate all cause a decrease in hours worked. This indicates that
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
89
unemployment is more of a reflection of women’s own job opportunities in the local labor market than of the husband’s opportunities and their effect on women’s marriageability and reservation wage. The parameter estimate for sex-ratio (SR) is significant and negative as predicted, offering support for the marriage squeeze hypothesis. The relatively low R-square of .13 in the hours of work equation is consistent with findings from other research (Heckman 1980; Cogan 1980). An F-test comparing an hours of work regression with and without sex-ratio indicated that including sex-ratio significantly improved the basic labor supply model.12 The results of this study show that the sex-ratio is negatively associated with the amount of labor a woman will supply. This may indicate a sex-ratio or marriage squeeze effect as hypothesized here. If so, marriage market conditions do affect the amount of labor married women supply. However, the finding of a negative sign of sex-ratio in labor supply regressions could possibly be spurious: unmeasured characteristics such as strong demand for female labor may possibly simultaneously cause migration of women, low sex-ratio and high participation of women in the labor force.13 CONCLUSION This study has provided evidence for a marriage squeeze effect on women’s labor supply. Some of the implications from these results are now discussed and suggestions are made for further work. Evidence for a marriage squeeze effect on the hours of work of married women offers support for a theory of labor and marriage. Marriage markets and labor markets interact in ways that most studies of labor supply ignore. An alternative explanation of a negative association between sex-ratios and married women’s labor force participation is that unobserved characteristics are explaining both sex-ratio and hours of work. For instance, some cities may have excellent job opportunities for women which would attract a lot of women compared to men (low sex-ratios) and cause women to work many hours. If this was the causality, one would expect the wage and unemployment rate effects to capture this. Further work is needed to develop better measures of aggregate labor market conditions for men and women and degree of segmentation between male and female jobs. As discussed in the theoretical part of this paper, degree of segmentation by gender will affect the direction of a sex-ratio effect and the magnitude of such an effect. The present empirical study improves on previous research. Previous studies by Grossbard-Shechtman have shown evidence for a marriage squeeze effect on women’s labor supply, based on cross-city comparisons and time series. The present analysis was based on microdata. This allowed us to reproduce methods that have widely been applied to the study of women’s labor supply. Also, microdata help us avoid some of the pitfalls involved in analyzing aggregate data. However, there is still plenty of room for improving the data analysis. As discussed in the data section, there are many flaws involved in the use of census data. We hope to pursue this work with more appropriate data-sets, such as the National Longitudinal Surveys or the Panel Study of Income Dynamics. Also, research on the United States could be expanded to include (1) blacks as well as whites, especially in view of the low sex-ratios observed for blacks in the US, and (2) more SMSAs. Finally, it is hoped that tests of the marriage squeeze hypothesis and other hypotheses relating marriage markets to labor supply will be performed using data from other countries as well. ACKNOWLEDGMENTS Financial support from San Diego State University Foundation and helpful comments by Bridget Heinemann, Oded Izraeli and Robert J.Willis are gratefully acknowledged.
90
WOMEN’S HOURS OF WORK
REFERENCES Becker, Gary S. (1965) “A Theory of the Allocation of Time,” Economic Journal 75: 493–515. —— (1973) “A Theory of Marriage: Part I,” Journal of Political Economy 81:813–46. —— (1981) A Treatise on the Family, Cambridge, Mass.: Harvard University Press. Brien, Michael J. (1991) “Economic Determinants of Family Structure: An Examination of Black and White Differences,” Ph.D. dissertation, University of Chicago. Chiappori, Pierre-André (1988) “Nash-Bargained Household Decisions: A Comment,” International Economic Review 29:791–6. —— (1991) “Nash-Bargained Household Decisions: A Rejoinder,” International Economic Review 32:761–2. Cogan, John (1980) “Married Women’s Labor Supply: A Comparison of Alternative Estimation Procedures,” in James P.Smith (ed.) Female Labor Supply, Princeton, N.J.: Princeton University Press. Goldman, Noreen (1977) “The Marriage Market: Supply and Demand of Potential Spouses in the United States,” Cambridge, Mass.: Harvard University, Center for Population Studies. Gronau, Reuben (1977) “Leisure, Home Production, and Work—The Theory of the Allocation of Time Revisited,” Journal of Political Economy 85:1099–124. Grossbard-Shechtman, Amyra (Shoshana) (1984) “A Theory of Allocation of Time in Markets for Labor and Marriage,” Economic Journal 94, 863–82. —— (1985) “Marriage Squeezes and the Marriage Market,” in Kingsley Davis and A.Grossbard-Shechtman (eds) Contemporary Marriage: Comparative Perspectives on a Changing Institution, New York: Russell Sage Publications. —— (1993) On the Economics of Marriage, Boulder, Co.: Westview Press. Grossbard-Shechtman, Shoshana and Granger, Clive W.J. (1994) “The Baby-Boom and Time Trends in Women’s Labor Force Participation,” presented at the American Economic Association meeting in Boston, January. Grossbard-Shechtman, Shoshana A. and Neuman, Shoshana (1988) “Labor Supply and Marital Choice,” Journal of Political Economy 96, 1294–302. Heckman, James (1974) “Shadow Prices, Market Wages and Labor Supply,” Econometrica 42:679–94. —— (1980) “Self Selection Bias as a Specification Error,” in James P.Smith (ed.) Female Labor Supply, Princeton, N.J.: Princeton University Press. —— (1993) “What Has Been Learned about Labor Supply in the Past Twenty Years,” The American Economic Review: 116–21. Hersch, Joni (1991) “Male-Female Differences in Hourly Wages: The Role of Human Capital, Working Conditions, and Housework,” Industrial and Labor Relations Review 44:746–59. Juster, Thomas F. and Stafford, Frank (1991) “The Allocation of Time: Empirical Findings, Behavioral Models and Problems of Measurement,” Journal of Economic Literature: 471–522. Mincer, Jacob (1962) “Labor Force Participation of Married Women: A Study of Labor Supply,” in H.Gregg Lewis (ed.) Aspects of Labor Economics, Princeton, N.J. : Princeton University Press. Robbins, Lionel (1930) “On The Elasticity of Demand for Income in Terms of Efforts,” Econometrica 10:123–9. Smith, James P. (1980) “Introduction,” in James P.Smith (ed.) Female Labor Supply, Princeton, N.J.: Princeton University Press. Wilson, William Julius (1987) The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy, Chicago: University of Chicago Press. Wood, Robert G. (1995) “Marriage Rates and Marriageable Men: A Test of the Wilson Hypothesis,” Journal of Human Resources 30:163–93.
SHOSHANA GROSSBARD-SHECHTMAN AND MATTHEW NEIDEFFER
91
NOTES 1 In contrast to most other applications of Becker’s (1973, 1981) theory of marriage —which have focused mostly on aspects of marriage and divorce—Grossbard-Shechtman’s (1984) primary focus was labor supply of married women. 2 Game-theory models also allow the integration of marriage market conditions into labor supply decisions (see the discussion in the chapter by Lundberg and Pollak in this book). Bargaining theories can, for example, be applied to analyze the effect of divorce on labor supply by married women. The major problem with bargaining models of marriage, however, is that they take the existence of a marriage as given and focus on effects of post-marriage eventualities on behavior within a marriage. This theory cannot serve as an integrative framework from which we can derive hypotheses regarding the relationship between aspects of marriage markets prior to marriage and various labor market outcomes such as labor force participation. Furthermore, Pierre-André Chiappori (1988, 1991) has effectively criticized some of the technical aspects of bargaining models of marriage. 3 In Grossbard-Shechtman’s model the household labor found in other Beckerian models (e.g. Gronau 1977 and Hersch 1991) is subdivided into activities benefiting oneself and activities benefiting one’s spouse, i.e. spousal labor. Spousal labor may also include time spent listening to a husband’s problems with an employer and time accompanying a wife to the doctor. In the following discussion spouses include cohabiting partners. 4 This discussion follows Grossbard-Shechtman (1984). Symbols were changed to make the model more compatible with standard labor supply models. 5 Fixed costs of marriage imply that small amounts of spousal labor will not translate into marriage. How quasiwages are determined is discussed below. 6 This is one of many possible spousal labor arrangements in marriage. An increasingly common arrangement is the egalitarian marriage in which there is no net transfer of spousal labor and no income from spousal labor. 7 A similar assumption can be found in game-theory models. 8 Commonly called “opposite sex.” We prefer to use terms that are less emotionally loaded. 9 One reason that the competitive market model may not be completely applicable is that people of certain classes, races or religions may be forbidden to intermarry. In such cases, separate markets need to be considered for each group of spouses. In that sense, spousal labor markets are not intrinsically different from other labor markets. 10 Gary Becker contributed this insight. 11 A spouse’s traits influence an individual reservation wage to the extent that they influence the portion k by which a spouse-employer j shares income with the spousal worker i. Compensating differentials in marriage will be observed to the extent that spousal employers can trade-off monetary compensations kIj with non-monetary compensations such as spouse j’s health, good looks and wits. Behavioral traits also matter. The more lovingly j acts to i, the more i will be willing to work at m for j at a given quasi-wage yi, or the lower the required yi for a given level of m. In other words, j’s behavior can compensate for lack of k or Ij. 12 The regression which does not include sex-ratio is available upon request. 13 Empirical studies of migration have indicated that job opportunities for married women are not a major reason why couples move.
6 PREMARITAL COHABITATION, CHILDBEARING AND THE CREATION OF ONE-PARENT FAMILIES John Ermisch
INTRODUCTION During recent years, the proportion of one-parent families headed by never-married mothers has been increasing in Britain. In 1981, 19 percent of one-parent families were headed by a never-married mother, but by 1992 that proportion had increased to one-third. This increase has helped foster a policy backlash, particularly proposals to reduce state benefits which might encourage young women to become single mothers. Little has been known, however, about the demographic factors which are primarily responsible for the rise, much less the underlying social and economic factors behind them. The present paper focuses on these demographic factors. The increase in the number of one-parent families headed by never-married mothers reflects a combination of factors. In particular, more women are spending some time before marriage in cohabitational unions, and are having children in such unions. Thus, when such unions break up, one-parent families are created. Of course, such families are also created by first births outside of partnerships. This paper investigates the relative importance of these two sources of one-parent families headed by never-married women and how it has changed. It first examines the dramatic rise in cohabitation among never-married women. The paper goes on to estimate the chances of becoming a never-married mother while cohabiting, and then examines how many one-parent families are created by the dissolution of the union. The analysis uses the only British data with which these issues can be addressed for more than one cohort: the life histories collected in the second wave of the British Household Panel Study (BHPS) during the last quarter of 1992.1 COHABITATION OR MARRIAGE? Before bringing births into the picture, the type and timing of first partnership is studied. Such analysis is of interest in itself because other sources of data indicate a rise in cohabitation before marriage, but the nature of these other data do not allow a cohort perspective on first partnership for more than one cohort. Furthermore, it allows a direct comparison with results from other countries, notably the USA. In this analysis, after reaching the age of 16, each woman faces three options in each month. She can marry, cohabit outside of marriage or remain single (i.e. without a partner). In order to examine change over time, the analysis is carried out for two sets of birth cohorts: women born during 1950–62 and women born after 1962. As the latter group reaches their 16th birthday from 1979 onwards, we shall label them, for
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
93
short, the “Thatcher cohort”, while the other group is called the “Pre-Thatcher cohort.” This is not meant to imply that political changes necessarily had anything to do with the differences between the two cohorts, but only to indicate the era in which they were making partnership and motherhood decisions. Table 6.1 provides a picture of the BHPS data that are being analyzed. Table 6.1 Observed first partnerships of women, BHPS Destination Cohabitation Marriage Still single Total
Born 1950–62
Born after 1962
No.
%
No.
%
369 795 61 1225
30.1 64.9 5.0 100.0
483 205 576 1264
38.2 16.2 45.6 100.0
Because they are younger, there are more women in the so-called Thatcher cohort who are still single. Nevertheless, already a larger percentage of them have started premarital cohabitation than the earlier cohort. In order to provide estimates which allow us to compare the two cohorts’ patterns of partnership entry, lifetable methods are used. That is, the main tool of analysis is “survival analysis” (in this case “surviving” as a single woman), which allows us to use efficiently the information on the young women in the Thatcher cohort who are still single at the time when the BHPS life histories were collected. Destination-specific “transition rates” for each cohort are estimated. In the case of cohabitation, for example, its transition rate can be interpreted as the probability of entering cohabitation in each month conditional on remaining single until that month. The transition rates for cohabitation and marriage sum to the total partnership rate, which implies the estimate of the survivor function for the state of being single. The two destination-specific transition rates are estimated for yearly intervals in a competing risks framework, treating the other event as censoring. Women remaining single at the time of the second wave of the BHPS are, of course, also censored. The estimated rates are shown in Figure 6.1 for the two broad cohorts. It is immediately apparent that there has been a major change in the type of partnership that young women first enter. In the Pre-Thatcher cohort, the largest transition rate is clearly the marriage rate, but in the Thatcher cohort, the cohabitation rate is largest. Thus, cohabitation has become the dominant mode of entering a partnership. Measurements more readily interpretable than the transition rates can be obtained by using the rates to simulate the outcomes for two synthetic cohorts of women, who differ according to the transition rates estimated for the Pre-Thatcher and Thatcher cohorts. Such simulations confirm that the primary difference between the two cohorts is that cohabitation is a much more important route into first partnership for the Thatcher cohort. By their 26th birthday, over half of the Thatcher cohort had entered cohabitation, compared with one-quarter of the earlier cohort. The calculations of outcomes by age 33 for the Thatcher cohort should be interpreted cautiously, as it has been assumed that transition rates for the earlier cohort prevail beyond the age of 28. These suggest that about 65 percent of more recent cohorts will cohabit in their first partnership. Roughly speaking, the relative proportions marrying directly and cohabiting are reversed between the two cohorts.2
94
JOHN ERMISCH
Figure 6.1 First partnership rates, 1950–62 and post-1962 cohorts
Figure 6.2 Comparison of first partnership rates for two broad cohorts
Partnerships are also being postponed. Just over one-fifth of women from the Thatcher cohort are still single at 26, compared with 15 percent of women in the Pre-Thatcher cohort, and postponement is also evident in Figure 6.2, which compares the sum of the cohabitation and marriage transition rates, which is the hazard rate for entering a partnership, for the two broad cohorts. The partnership rate over the ages 18– 23 is clearly lower for the later cohort. The median age at first partnership indeed increased by one year between the two cohorts, from 21 years 4 months to 22 years 4 months. It is difficult to make precise comparisons with other countries, because of differences in the way data are presented. Willis and Michael (1994) estimate that 24 percent of white women born in 1954 cohabited in
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
95
their first partnership, 65 percent married without any pre-marital cohabitation and one-tenth had not had a partner before the age of 32 (estimated from the 1986 Follow-Up Survey of the National Longitudinal Study of the High School Class of 1972). This distribution is similar to that for the Pre-Thatcher cohort at age 33. The National Survey of Families and Households, which interviewed people during 1987–8, provided the first opportunity to examine American data on cohabitation for a number of cohorts. Using these data, Bumpass and Sweet (1989) estimate that 37 percent of women born during 1960–64 had cohabited before the age of 25, compared with 16 percent among women in the 1950–54 birth cohort. The corresponding percentage for British women from the Pre-Thatcher cohort is 22 percent, but this had risen to 50 percent for the Thatcher cohort, which incorporates the experience of later cohorts than those analyzed by Bumpass and Sweet. PARTNERSHIP AND PRE-PARTNERSHIP BIRTHS The next analysis adds a third competing risk to the two partnership ones, namely having a birth outside of a partnership. After reaching the age of 16, each woman is now assumed to face four options in each month. She can marry, cohabit outside of marriage or have a birth outside of a partnership. Table 6.2 Observed and simulated first “destinations” of women, BHPS Observed Destination Pre-partnership birth Cohabitation Marriage Still single and childless Total Simulated (percentages) Destination Pre-partnership birth Cohabitation Marriage Still single Total
Born 1950–62
Born after 1962
No.
%
No.
%
59 343 769 53 1224
4.8 28.0 62.8 4.3 100.0
76 444 193 549 1262
6.0 35.2 15.3 43.5 100.0
At age 33 4.9 27.3 62.6 5.2 100.0
Born after 1962 At age 26 7.9 50.0 23.5 18.5 100.0
At age 33 7.9 59.6 26.2 6.3 100.0
Born 1950–62 At age 26 4.9 22.8 57.9 14.4 100.0
If not, she will remain single and childless (i.e. without a partner or a child). The first panel of Table 6.2 shows the observed “first destinations” in the BHPS. For instance, about 5 percent of those women born during 1950–62 had a birth before any partnership.3 Lifetable methods are again used to compare outcomes between the two broad cohorts. Now three destination-specific transition rates are estimated for each cohort in a competing risks framework, treating the other two events as censoring. The three transition rates sum to the total “escape rate” for leaving the state
96
JOHN ERMISCH
of being single and childless in each month. Using the estimated transition rates, we obtain the lifetable estimates shown in the second panel of Table 6.2. The incidence of pre-partnership births is slightly larger in the later cohort, but still only 8 percent of women have a birth before any partnership. By far the major change is the type of partnership that young women first enter. While just under one-quarter of the later cohort enter marriage directly before the age of 26, almost three-fifths of the Pre-Thatcher cohort did so. The median age at first transition rose between the two cohorts, from 21 years 3 months to 22 years. This postponement is also evident in the larger percentage single and childless at age 26 in the Thatcher cohort. WHO COHABITS? Here we investigate associations between the social background of women and the chances of women having a pre-partnership birth, cohabiting or going directly into marriage. As women are at risk for these transitions while teenagers there are few variables in the BHPS which could legitimately be called exogenous to the birth and partnership decisions. For instance, educational attainment is likely to be mutually dependent with these decisions, and influenced by similar underlying factors. The analysis focuses, therefore, on the social background of the woman, as measured by her father’s occupation when she was aged 14, and on trends across birth cohorts within each of our two broad cohorts. In the case of father’s occupation, the variable used is dichotomous, indicating whether he was in a non-manual job or not. About one-quarter of women have information on father’s occupation missing.4 As we do not wish to lose so many cases, an indicator of whether the father’s occupation is missing (MISSOCC) is also included in the model.5 Cox proportional hazard models are estimated in a competing risk framework for the three destination states and the two broad cohorts. The Cox model takes the following form: hd(t, X)= (t)exp(ßdX), where hd is the transition rate for destination d; t is months since the woman’s 16th birthday; X is a vector of explanatory variables; βd is a vector of parameters to be estimated; and (t) is a baseline hazard function that need not be specified. The parameter estimates are shown in Table 6.3. Table 6.3 Cox model parameter estimates Pre-partnership birth Variable
Born 1950–62
Born after 1962
Father in non-manual job Year of birth Cohabitation Variable Father in non-manual job Year of birth Marriage Variable Father in non-manual job Year of birth
−1.301** 0.022
−1.458** −0.020
Born 1950–62 0.099 0.077**
Born after 1962 −0.390** 0.043*
Born 1950–62 −0.308** −0.052**
Born after 1962 −0.366* −0.152**
Notes: **significant at 0.01 level or less; * significant at 0.05 level
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
97
Among the Pre-Thatcher cohort, having a father in a non-manual job lowers the pre-partnership birth rate and the marriage rate, but does not affect the transition rate into cohabitation. Thus, women from a middleclass background experience one of these transitions later in their life and are more likely to cohabit before marriage and childbearing.6 Within this cohort, there is an increasing cohabitation trend and a falling marriage trend. There is, therefore, a trend increase in the chances of premarital cohabitation, as we clearly would expect from Tables 6.1 and 6.2. The declining marriage trend is even more evident in the Thatcher cohort, and there continues to be an upward trend in cohabitation. The large negative trend in the marriage rate means, however, that young women are spending more of their young adult years single and childless. Thus, there appears to be a continuation of the trend already evident in the fall in the partnership rate between the two broad cohorts (Figure 6.2) and the rise in the median age of the first transition between the Pre-Thatcher and Thatcher cohorts discussed above. Being from a middle-class background lowers all three transition rates in the Thatcher cohort, which implies that middle-class young women remain single and childless longer than those whose fathers are in manual jobs. In contrast to the earlier cohort, the odds that a woman’s first transition is into cohabitation rather than marriage are no longer higher for those from a middle-class background. This indicates, as we might expect, that while the middle class were pioneers in cohabitation, the social background difference has disappeared as cohabitation in one’s first partnership has become more common. On the other hand, the social background difference in the time spent single and childless is larger in the later cohort.7 CHILDBEARING WITHIN COHABITATIONAL UNIONS As Table 6.4 indicates, there were 787 women who were observed to cohabit in their first partnership before becoming a mother. Four types of “outcome” are distinguished for each of these unions: the union was turned into a legal marriage before a birth; the union dissolved before a birth; the union continued and remained childless until the time of the survey; a birth occurred within the union and before marriage. The top panel of Table 6.4 shows the distribution of these outcomes for our two broad cohorts. About one in eight of the cohabitational first unions formed by childless women were observed to produce children before marriage, with a rise between the two broad cohorts. But with 23 percent of the unions in the Thatcher cohort continuing without a birth or marriage at the time of BHPS, this fraction is not a reliable estimate of the proportion of such unions that would produce children. To obtain such an estimate, lifetable methods are again used. The transition rates for the three events of interest—birth, marriage and dissolution—are estimated in a competing risks framework, treating the other two events as censoring. Unions continuing without a birth or marriage at the time of the second wave of the BHPS are, of course, also censored. Using the transition rates, estimated for 12-month intervals, the outcomes for a group of childless, never-married women entering their first cohabitation are simulated. Table 6.4 Observed and simulated outcomes for never-married women in first cohabiting unions, BHPS Observed Outcomes of union Child before marriage
Born 1950–62
Born after 1962
No.
%
No.
%
32
9.3
70
15.8
98
JOHN ERMISCH
Observed Outcomes of union Marriage before child Dissolved before marriage Still cohabiting, childless Total Simulated Outcomes of union Child before marriage Marriage before child Dissolved before marriage Still cohabiting, childless Total
Born 1950–62
Born after 1962
No.
%
No.
%
202 85 24 343
58.9 24.8 7.0 100.0
167 107 100 444
37.6 24.1 22.5 100.0
After 10 yrs 9.1 60.2 25.9 4.8 100.0
Born after 1962 After 5 yrs 17.9 42.6 27.0 12.4 100.0
After 10 yrs 18.2 47.7 29.4 4.8 100.0
Born 1950–62 After 5 yrs 8.9 55.2 23.5 12.4 100.0
The estimated median duration to termination of the childless cohabitational union is 19 months in both cohorts.8 Thus, “action,” in terms of marriage, birth or dissolution, occurs relatively quickly, and this is illustrated in Figures 6.3 and 6.4, which show the simulated cumulative proportions of women who had gone to each of the three destinations by months since the start of the union for each of the two broad cohorts. Within three years, three-quarters of each cohort had either had a birth, married or dissolved the union. It is clear from comparing Figures 6.3 and 6.4 that having a child and union dissolution are more common in the Thatcher cohort than in the Pre-Thatcher cohort, while marriage is a more frequent outcome of cohabitations in the earlier cohort. This is confirmed by the second panel of Table 6.4, which compares simulated outcomes five and ten years from the start of the cohabitation. It shows that about 9 percent of women in the Pre-Thatcher cohort entering a first cohabitational union had a birth before marriage in the union, but this percentage is predicted to double to 18 percent in the Thatcher cohort. Thus, childbearing in cohabitational unions has become more common. If a birth occurs, it primarily happens in the first three years of the union. Just under one-half of the women in the Thatcher cohort will convert the union into marriage while childless, compared with three-fifths in the earlier cohort. In addition to more unions producing children before marriage, a larger percentage of the unions among the Thatcher cohort will dissolve before marriage or childbearing. After ten years, only about 5 percent of both cohorts are still childless and cohabiting. Among the 1954 cohort of American white women, analyzed by Willis and Michael (1994), threequarters of women who cohabited in their first partnership married, one-fifth dissolved the union before marriage and 5 percent were still cohabiting at the age of 32. For this cohort, the median length of cohabitation was about 15 months. Bumpass and Sweet (1989) estimated a similar median duration of cohabitation among unions starting during 1975–84, and they estimate that three-fifths of these unions are likely to end in marriage. When similar computations are done for Britain, British cohabitations tend to be about five months longer on average (a median duration of 20 months), and they appear to be less likely to be converted into marriage.9 While 65 percent of the Pre-Thatcher cohort’s first cohabitations turned into marriage, this was
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
99
Figure 6.3 Destinations of never-married women entering cohabitation, per 1000, 1950–62 cohorts
Figure 6.4 Destinations of never-married women entering cohabitation, per 1000, post-1962 cohorts
true for only 56 percent of the Thatcher cohort cohabitations. Two-fifths of the later cohort’s cohabitations are expected to dissolve within ten years.
100
JOHN ERMISCH
Figure 6.5 Destinations of women having children in first cohabitations, per 1000, all cohorts
CREATION OF ONE-PARENT FAMILIES THROUGH COHABITATION DISSOLUTION The next step in our analysis is to estimate how many one-parent families are formed from this small but growing minority of women who have had a birth in their cohabitational union before marriage. This amounts to estimating how many of these fertile cohabitations will dissolve before marriage. Thus, there are two competing events: marriage or end of the partnership, with the latter creating a one-parent family (OPF). With only 102 of such fertile unions, no attempt has been made to split the sample by broad cohort; 39 of these are observed to dissolve, while marriage is observed for 32. Again, estimates of the risk of creation of one-parent families must also take account of the remainder of censored cases, and so competing risk transition rates are estimated for 12-month intervals, starting from the birth of the child. From these rates we calculate what would be expected to happen to a group of women giving birth in a cohabitational union. Figure 6.5 illustrates the simulated cumulative proportions of women who have married and who have had their union dissolve, by the number of months that have passed since their first birth. The simulation indicates that one-half of women having their first child in their first cohabitational union can expect to become a never-married lone mother through the dissolution of their current union. Two-fifths marry their partner. After ten years, about one in ten are still cohabiting with the same partner and their child (ren). One-half of the women have either married or become the head of a one-parent family within 21 months of the birth. The relatively low proportion of mothers who are observed cohabiting at any point in time is partly a consequence of these short durations. PATHWAYS INTO PARTNERSHIP AND CHILDBEARING It is interesting to put all of these processes together in order to estimate the outcomes for two cohorts: one reaching their 16th birthday in the Pre-Thatcher era, and one becoming 16 in the Thatcher era. Of particular
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
101
interest is the proportion of one-parent families (OPF) headed by a never-married mother created through childbearing within a cohabitation. These two cohorts differ by the following set of transition rates: first entry rates to partnerships and a pre-partnership birth, and transition rates within cohabitations. Both cohorts are assumed to have the same transition rates after the birth of a child within a cohabitation. The first entry rates to partnerships and a pre-partnership birth vary with age (as illustrated in Figure 6.1) but all subsequent transition rates are assumed to be constant with duration in each state (i.e. a competing risk Markov process is assumed for each of these processes).10 Table 6.5 shows the simulated outcomes by the cohort’s 33rd birthday for the two broad cohorts. It indicates that, of 1000 women reaching their 16th birthday in the Pre-Thatcher era, 61 are predicted to have become never-married lone mothers, but only one-fifth of these were created through childbearing within cohabitation. In the Thatcher era, 136 (per thousand) are predicted to become a never-married lone mother, with just over two-fifths of these arising because of childbearing within a cohabitational union. Table 6.5 Simulated outcomes for two broad cohorts (percentages) Outcome by age 33
Born 1950–62
Born after 1962
Directly to marriage Pre-partnership birth (OFF) Cohab.→Birth→Marriage Cohab.→Birth→OPF Cohab.→Marriage Cohab. dissolved childless Still cohabiting, childless Still cohabiting, with child Still single Total
62.5 4.9 1.0 1.2 16.5 6.9 1.3 0.3 5.2 100.0
26.2 7.9 4.7 5.7 28.0 18.0 1.9 1.4 6.3 100.0
We also see that the modal path in the Thatcher cohort is cohabitation followed by marriage, but only about 30 percent of the cohort follow this path. In the Thatcher era, diversity in partnership and birth transitions is the key characteristic. This contrasts with a very dominant, but by no means universal, direct entry to marriage in the Pre-Thatcher cohort. For the purposes of another benchmark comparison, an estimate of the proportion of first births occurring outside marriage is derived. To do so, we must also consider the transition to motherhood within marriage for women entering marriage directly and for women who cohabit and then marry that partner. For simplicity, each of these transition rates is assumed to be constant with marriage duration. Using these estimated transition rates in conjunction with the others used to construct Table 6.5, we can obtain a rough estimate of the proportion of first births occurring outside marriage.11 It is rough because we ignore first births in partnerships after the first one. These estimates imply that 10 percent of first births by age 33 in the Pre-Thatcher cohort were outside marriage, compared with 29 percent in the Thatcher cohort. Direct estimates for the 1958 cohort from the NCDS (Kiernan 1995) appear to be consistent with the estimates here. The fact that currently 32 percent of all births occur outside of marriage is also consistent with our simulation for the Thatcher cohort.
102
JOHN ERMISCH
Of the first births predicted to be born outside of marriage among the Thatcher cohort, three-fifths are predicted to be born in cohabitational unions.12 Only one-third of births outside of marriage in the PreThatcher cohort were born in cohabitational unions. WHO BECOMES A LONE MOTHER THROUGH COHABITATION DISSOLUTION? This section presents an exploration of the characteristics of women associated with having a child in a cohabitional union and with the dissolution of that union among mothers. To this end, Cox proportional hazard models are again estimated in a competing risks framework, but the sample has not been split by cohort. A limited set of characteristics that have been determined by the start of the union are considered: age at the start of the union (or at birth), whether the woman’s father was in a non-manual job when she was aged 14, and her educational qualifications (A-level or higher, O-level or CSE, or no qualifications). The “relative risks” (exp(β), calculated from Cox model coefficients, β) shown in the first panel of Table 6.6, indicate that more highly educated women and those who began the union when they were older are less likely to give birth before marriage in the union. For instance, women with A-level qualifications have a birth risk which is only 20 percent of that for women with no qualifications, and forming a union one year later reduces the birth risk by 10 percent. These associations suggest that women in poorer economic circumstances appear to be more likely to bear children in a cohabitational union. In addition, there is an upward trend in the chances of having a child in such a union.13 Table 6.6 Relative risks of birth in cohabiting union Variable
Relative risk
Age at start of union (months) Qualifications: A-level+ Qualifications: O-level/CSE Year of woman’s birth Relative risks of union dissolution, cohabiting mothers Variable Age at motherhood (months) Some qualifications
0.99* 0.19* 0.46* 1.065* Relative risk 0.99 2.47*
Note: *significant at 0.05 level or less
With only 102 mothers in the analysis, it is difficult to estimate associations between a mother’s characteristics and the hazard of union dissolution very precisely. Educational qualifications are, however, significantly associated with the creation of one-parent families from never-married mothers in cohabitational unions, as the second panel of Table 6.6 shows. Mothers with some qualifications are much more likely (2.5 times more) to have their union dissolve than those with none. While statistically significant at only the 0.10 percent level, mothers who were younger at the birth of their first child (or at the start of their partnership) are more likely to become a single-parent family through union dissolution. Thus, women who enter a cohabitational union when they are older are less likely to become single parents through having a child in the union, both because they are less likely to have a child in the union before marriage, and because, if they do have a child, their union is less likely to dissolve. On the other hand, the
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
103
relationship between educational qualifications and the creation of single-parent families through cohabitational childbearing is less clear. While better educated women are less likely to have a birth in the union before marriage, if they do, they are more likely to become a single mother through the dissolution of their union. The former effect dominates, however, because it substantially reduces the population of never-married mothers in cohabitational unions (i.e. those who are at risk to become single mothers). Simulation of a Markov model with the covariates in Table 6.6 suggests that the incidence of single parenthood among a cohort of 1000 never-married, childless women entering their first cohabitation is about halved if the cohort had achieved qualifications at A-level or above, relative to those with no qualifications. Analysis of the entry to cohabitation, summarized in Table 6.3, indicated that women from more affluent backgrounds, at least as indicated by their father’s occupation, are, if anything, more likely to cohabit, while Table 6.6 suggests that cohabiting women in better economic circumstances are less likely to have a child in the cohabitational union. Thus, the cross-section evidence (Kiernan and Estaugh 1993: chapter 2) that cohabiting mothers are in poorer circumstances than married mothers must reflect at least in part the higher chances of having a child among less-well-off cohabiting women. CONCLUSIONS This paper uses a unique source of data to explore the trend in cohabitation and the role of childbearing within premarital cohabitation in the creation of families headed by never-married mothers in Britain. It suggests that among more recent cohorts of women, about two-fifths of one-parent families headed by never-married mothers are created through childbearing within cohabitation followed by dissolution of the cohabitational union. There is a tendency for cohabiting women in poorer economic circumstances to be more likely to have a child within the union, and thereby more likely to become a never-married lone mother through the dissolution of their union. The main reason for the growth in childbearing within cohabitation is the dramatic increase in cohabitation before marriage. Indeed, it is now the most popular form of first partnership. While partnership is being postponed in young women’s lives, the odds of cohabitation relative to marriage are still rising among recent cohorts reaching young adulthood. Women from a middle-class background were more likely to cohabit than those whose fathers were in manual jobs, but this is no longer true for more recent cohorts. Middle-class young women remain single longer, and this social background difference appears to be widening over time. Among women who do cohabit, childbearing before marriage is also a more likely outcome for more recent cohorts. It is estimated that about three-fifths of first births outside of marriage take place in first cohabitational unions. About one-half of these fertile cohabitational unions dissolve, producing a nevermarried lone mother. Four in ten of these fertile unions are turned into marriage within ten years. ACKNOWLEDGMENTS I am grateful to Rachel Smith for research assistance and to the ESRC for financial support for this research.
104
JOHN ERMISCH
REFERENCES Bumpass, L.L. and Sweet, J.A. (1989) “National Estimates of Cohabitation,” Demography 26:615–25. Haskey, J. (1992) “Pre-marital Cohabitation and the Probability of Subsequent Divorce: Analyses using New Data from the General Household Survey,” Population Trends 68 (Summer): 10–19. —— (1995) “Trends in Marriage and Cohabitation: The Decline in Marriage and the Changing Pattern of Living in Partnerships,” Population Trends 80 (Summer): 5–15. Haskey, J. and Kiernan, K.E. (1989) “Cohabitation in Great Britain—Characteristics and Estimated Numbers of Cohabiting Partners,” Population Trends 58 (Winter): 23–32. Kiernan, K.E. (1995) “Transition to Parenthood: Young Mothers, Young Fathers— Associated Factors and Later Life Experiences,” Report to the Joseph Rowntree Foundation, York. Kiernan, K.E. and Estaugh, V. (1993) “Cohabitation: Extra-marital Childbearing and Social Policy,” Occasional Paper 17, London: Family Policy Studies Centre. Lancaster, T. (1990) The Econometric Analysis of Transition Data, Cambridge: Cambridge University Press. Willis, R.J. and Michael, R.T. (1994) “Innovation in Family Formation: Evidence on Cohabitation in the United States,” in J.Ermisch and N.Ogawa (eds) The Family, the Market, and the State in Ageing Societies, Oxford: Oxford University Press.
NOTES 1 The National Child Development Study has information similar to that from the BHPS for the 1958 birth cohort. The other British data on cohabitation come from the General Household Survey, but only two pieces of information are collected: whether respondents who were not living in a married partnership at the time of the survey were cohabiting, and whether respondents who were currently or previously married had cohabited with their spouse before their marriage. For further details, see Haskey and Kiernan (1989), Haskey (1992) and Haskey (1995). 2 A cross-check of the BHPS data is provided by a comparison with direct estimates from the 1958 cohort, surveyed in the National Child Development Study (NCDS), of the proportions whose first partnerships by age 33 were marriage and cohabitation. The BHPS percentage for the 1950–62 cohort is consistent with the NCDS one. As noted at the outset, alternative sources of information about cohabitation in other cohorts is very limited. Haskey (1992: Table 1(a)) estimates that 19 percent of never-married women born during 1950–59 cohabited with their spouse before their first marriage, while 41 percent of women born during 1960–69 who had married by 1989 did so. The latter figure is biased downward by the fact that those who marry at young ages are overrepresented in these truncated data and later age at first marriage is associated with a higher probability of cohabiting with one’s spouse before marriage. From a period perspective, seven out of ten first marriages in the early 1990s were preceded by the spouses’ cohabitation, compared with one out of ten first marriages in the early 1970s (Haskey 1995). 3 This is exactly the same percentage as was found among women of the 1958 birth cohort in the NCDS data. By age 33, 76 percent of women in the 1958 birth cohort had become parents, and 7 percent of the mothers had their first birth before a first partnership (Kiernan 1995). Thus, the proportion of women in the cohort having a birth before a first partnership is (0.76)*(0.07)=0.05. 4 In a very small proportion of cases, occupational data was missing because the father was deceased. In the PreThatcher cohort, 14 percent did not hav information on the father’s occupation, while this was the case for 35 percent of women in the Thatcher cohort. 5 The reference category for father’s occupation is being in a manual job or not working, and that for MISSOCC is not having or missing information for father’s occupation; thus, a woman whose father was in a manual occupation would have zero values on both variables.
COHABITATION, CHILDBEARING AND ONE-PARENT FAMILIES
105
6 The odds of entering cohabitation relative to marriage are , where x=1 if the father was in a non-manual occupation and zero otherwise, and βd is the coefficient associated with this variable for destination d. The parameter estimates in Table 6.3 for the Pre-Thatcher cohort indicate that βc– βm=0.407; thus, having a father in a non-manual job raises the odds of entering cohabitation over marriage. Analogously, βc–βp=1.400, which implies that being from a middle-class background also raises the odds of cohabiting relative to having a premarital birth. Bumpass and Sweet’s (1989) and Willis and Michael’s (1994) findings of positive effects of father’s education on the odds of cohabiting is consistent with this result. The finding here is also consistent with the tendency for British single women in the highest status socio-economic groups to be more likely to be cohabiting in 1986/7 than their counterparts in other socio-economic groups (Haskey and Kiernan 1989). 7 The coefficients on MISSOCC (not shown) indicate that those for whom information about father’s occupation is missing have lower pre-partnership birth, marriage and cohabitation rates. The other parameter estimates are similar when we exclude these missing cases. 8 Haskey and Kiernan (1989) and Kiernan and Estaugh (1993: Table 2.5) estimate median elapsed durations of cohabitation of 20 and 21 months respectively from samples from the “stock” of never-married cohabiting women in 1986/7 and 1989 (obtained from the General Household Survey). Haskey (1995) estimates that this median elapsed duration increased to 29 months in 1990–93. It should, however, be noted that the distribution of elapsed durations usually differs from the distribution of completed durations among entrants (e.g. see Lancaster 1990: 91–7), and it is the latter which is estimated in the current analysis. 9 These computations are based on lifetable estimates similar to those illustrated in Table 6.4, but ignoring a birth as a competing risk. 10 When the Markov assumption is applied to the transitions within cohabitation and after a birth within a cohabitation, it produces outcomes very similar to those in Figures 6.3, 6.4 and 6.5, even though these figures were compiled using processes which allowed for variation in the transition rate with duration. 11 The monthly birth rate following direct marriage is 0.01657 for the Pre-Thatcher cohort, and 0.02052 for the Thatcher cohort. The monthly birth rate in a marriage preceded by a cohabitation (with the same partner) is estimated to be 0.01647 and 0.02443 respectively for these two broad cohorts. 12 This proportion was two-fifths for the 1958 birth cohort, estimated from the NCDS. Birth registration statistics indicate that about 55 percent of births outside of marriage in 1993 were jointly registered at the same address, suggesting a birth to a cohabiting couple. 13 The other coefficients are similar when the trend variable is omitted from the model in Table 6.6. Similar models are estimated for the competing risks of childless marriage and dissolution, and these do not indicate systematic differences in the marriage and dissolution transition rates associated with women’s age at union, education or social background, nor do they show any significant trends in the transition rates.
Part III FAMILY POLICIES AND HOUSEHOLD ALLOCATION OF TIME
7 CHILDCARE, HUMAN CAPITAL AND ECONOMIC EFFICIENCY Siv S.Gustafsson and Frank P.Stafford
INTRODUCTION There has been a long-term shift toward a greater payoff to investment in human capital, which has led to a greater level of female labor force participation and migration to skill-intensive occupations, but at the same time has created a need for higher levels and new forms of out-of-family childcare. The western industrial economies have come up with a diverse array of policies and programs to help sponsor this out-of-family childcare. Some are predicated on the need to provide a type of social insurance for the parents and siblings of children with learning disabilities. Others can be thought of as relaxing the borrowing constraint implied by the fact that future income from the labor market usually cannot be used as collateral for loans. There are questions, however, about the effectiveness of these programs and the control of their costs. This is not surprising given the rising share of GDP devoted to these activities in the advanced industrial economies. The point of our paper is that matters relating to work and childcare are the basis for a great deal of social policy (as has been true for a very long time) but, just as importantly, they need to be considered as part of the larger picture of long-term economic growth and income distribution. As we approach the year 2000, we see that in many of the western countries, women represent half of the labor force participants and are key actors in the process of human capital formation. It is the strengthening of this dual role during the last 100 years which makes work and childcare so central to economic well-being. The formation of human capital is an efficiency requirement which social policy makers will be confronted with. At the same time possessing human capital is the major determining factor in the wellbeing of an individual. Equity considerations may therefore require that individuals acquire more equal amounts of human capital. However, there will be efficiency effects of such policies since individuals have unequal capacities to accumulate human capital, and the result of a given amount of resources devoted to human capital enhancement will differ between individuals. Also parents’ time may be devoted to human capital investments in children or to human capital investments in themselves. There are thus a number of conflicting equity and efficiency aspects within a generation and across generations. In this paper we first give some reasons for why human capital has grown in importance during the last 100 years. Second, we give an overview of different equity and efficiency trade-offs that families and policy-makers have to take into consideration and indicate some of the choices made in some western countries. We then elaborate further on one specific aspect, namely resource allocation to children with different learning abilities, which is one sort of equity/efficiency trade-off on the child/child dimension. The section that follows develops the parent/child dimension, and discusses some of the trade-offs which are
108
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
inherent in the mother’s time use. Finally, because only a few of the equity and efficiency trade-offs indicated in the second section of this paper have been discussed at any length, we indicate some questions which need further elaboration. The contention of our paper is that the efficiency of spending money on family policies cannot be seriously analyzed without taking its effect on human capital accumulation into account. THE GROWING IMPORTANCE OF HUMAN CAPITAL The rising role of families and related institutions in producing skills in the advanced, “industrialized” countries is suggested by the historical secular shift away from physical capital documented by Tinbergen (1975:98). From the 1860s to the 1950s the physical capital share of the GDP fell from 44 percent to 24 percent in England. For France the corresponding figures are 36 and 18 percent. From the turn of the century to the 1950s the physical capital share of GDP fell from 35 to 25 percent for Germany. For the US the corresponding figures are 28 and 21 percent. While the share of income arising from physical capital in the advanced industrialized countries has been stable or rose for the period 1970–95 (32.1 percent in 1970– 77 and 35.0 percent estimated for 1995 (OECD 1994)), interest in systems for human investment has been motivated by a belief that economic growth is dependent on human capital and new technology (Barro and Lee 1993; Eaton and Kortum 1995). In the United States the importance of human capital and what is referred to as skill-biased technical change (Bound and Johnson 1992) has come to depend increasingly on the utilization of the skills of educated women (Johnson and Stafford 1997). The more general relationship between women’s participation in economic and political life and economic well-being as we approach the year 2000 is suggested by the pattern of relationship between the real GDP per capita rank and the gender empowerment index (GEM) of the United Nations (United Nations 1995). The GEM is an amalgam of three kinds of variables: the share of earned income attributed to women; access to professional, technical, administrative and managerial jobs; and representation in national legislative bodies. Our interpretation is that the changing nature of work has bolstered women’s role in economic life so that more economically developed countries also have more gender equality. Although women’s role in the market economy has grown, they have maintained the primary role in the early cognitive and social development of young children, despite a secular trend toward more childcare and home production by men in the industrialized societies (Juster and Stafford 1991; Flood and Gråsjö 1995). The study of intergenerational mobility in the US using longer duration panel data indicates a stronger influence of family background variables on economic success than previously thought (Hill and Duncan 1987; Solon 1992), and it seems likely that early childcare will turn out to be one of the “background” variables most important in shaping the long-term achievement of individuals. Table 7.1 presents a selection of research findings on the character of work and women’s economic role in a selection of the western economies (Belgium, Denmark, France, Germany, The Netherlands, Sweden, the United Kingdom (Pott-Buter 1993) and the United States (Goldin 1990)), around 1900 and (close to the year) 2000. There are several issues of economic efficiency and growth related to the patterns observed in the table. These are questions of efficiency losses from occupational exclusion of women, particularly in 1900, and the emerging need to balance the lifetime careers of women with the early human capital formation of children. As discussed elsewhere by Johnson and Stafford (1997), one factor which can sustain “crowding” of women into certain occupations and industries is a broad system of institutional discrimination. If there are
CHILDCARE AND HUMAN CAPITAL
109
legal or established social practices which restrict women from working in “non-traditional” jobs (Goldin 1990: chapter 3), this will raise the real wage of men, and lower the real wage of women. In addition to these distributional elements, there will be a net or deadweight loss of economic welfare. The size of the deadweight loss will grow as the potential relative productivity of women rises, as seems plausible under the long-term movement toward a regime of skill-biased technology. Related issues are the productivity of out-of-family caregiving and external effects. Can people with specific training provide greater productivity (relative to cost) for some aspects of early child development? If parental care and child development are highly subsidized, will this create fertility rates which are “too high?” Are there significant external effects on other families from the lack of ability of some children to function in the modern job market? That is, if low job market success is an outcome for people who had less human capital investments when children, this might impose external costs on others, for example if the low achievers seek employment in the illegal sector. Can a child benefit to a family with low financial resources improve economic efficiency or will the funds be appropriated by the parents for other Table 7.1 Work and childcare arrangements, 1900 and 2000 Feature
1900
2000
Participation (overall)
Labor force participation rates in the range of 20–30% of total female population, except for France ( 35%) and the US ( 15%). Women’s share of labor force=20–35% Participation often concentrated in specific sectors such as agriculture, domestic services and industry. UK has largest concentration of women in the industrial sector. Sweden has the largest share in agriculture. Industrial work often in “female” occupations and industries
Participation higher, in the range of 35–55% of total female population. Women’s share of labor force=40– 50%
Participation (by industrial sector)
Participation (by age and family status)
Childcare arrangements
Participation concentrated among younger and unmarried women. Labor force participation rate for married women in the range of 5– 35% for US, Sweden and The Netherlands. France is very high at 56%. Childcare arrangements are family based, fertility and labor force participation are strongly connected
Agriculture disappears as a significant employment. Women’s participation broadens to a wider array of skilled occupations in the later part of the century. Still substantial occupational concentrations, particularly in the growing trade and service sectors, the latter including the public sector Participation of married women much higher. Labor force participation rate for married women in the range of 45– 80% in 1990 (The Netherlands 41; France 44; Sweden 79). Job tenure greater and less dependent on marital and fertility history. Childcare is extra-family based, link between fertility and labor force participation much reduced. Provision of childcare subsidized directly through public provision (Sweden) or through tax incentives (US). Public schooling expands to provide more of early training
Sources: Pott-Buter (1993) for labor force participation rates for France, The Netherlands and Sweden. Goldin (1990) for US figures
110
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
purposes? We now turn to a closer look at the equity and efficiency trade-offs associated with childcare by parents and external caregivers. CHILDCARE AND EQUITY AND EFFICIENCY TRADE-OFFS The purpose of this section is to consider equity and efficiency aspects of childcare that families and governments face in making decisions about resource allocations. Most family policy decisions have been motivated by equity considerations: more seldom are policy-makers aware of the efficiency dimensions which are an effect of family policies. Because decisions on human capital investment, both by parents in their children and by adults in themselves, are important in the efficiency of the economy, it is important that efficiency aspects of family policies receive attention. Table 7.2 below is an attempt to classify examples of efficiency effects of family policies. Decisions on resource allocation within a family, whether between siblings or between a child and its parents or between husband and wife, can be taken in order to achieve justice or in order to achieve an efficiency goal. Decisions taken for equity reasons will invariably have an effect on efficiency and the other way around. The left part of Table 7.2 shows some examples of equity considerations within the family and the likely efficiency consequences of such decisions. The right part of Table 7.2 shows similar equity decisions between families and examples of resulting efficiency effects. The increasing importance of human capital as compared to physical capital over time will change the calculation of the resulting potential efficiency gains and losses of different actions. Some of the efficiency effects can be seen only in a longer time perspective since it takes time to build up human capital. Because the building up of human capital is an investment there is not always a trade-off between equity and efficiency but rather resource transfers because of equity considerations can be shown to be also productivity enhancing. Below we will comment on the different examples of equity/efficiency trade-offs mentioned in Table 7.2, starting with the intrafamily child/child dimension. The intrafamily resource allocation to children can be simultaneously egalitarian and efficient if the siblings are equally able to augment their capacities through investments in their human capital. But suppose one has serious learning disabilities, such as a mental handicap. If parents care about equal outcomes for the children, they may invest more in the disabled child. If so, brothers and sisters of disabled children may receive less schooling and other resources than they would otherwise receive. This is an efficiency problem which can be described as an intrafamily equity tax. Another efficiency aspect relates to the sibling of a disabled child. Transfer of resources to the family via a social insurance mechanism will reduce the efficiency tax and provide more resources to the siblings of the disabled child—another case where equity can have positive efficiency effects. Table 7.2 Equity/efficiency trade-offs in child development Intra-family (1) child/child
Inter-family
Equity
Efficiency
Equity
Efficiency
More able siblings v. less able siblings
Human capital enhancement v. nonhuman-capital transfers
Poor and workingclass children v. middle-class children
Cost effective child development
CHILDCARE AND HUMAN CAPITAL
Intra-family
(2) child/parent
(3) parent/parent
111
Inter-family
Equity
Efficiency
Equity
Efficiency
Children with disabilities v. normal siblings Brothers v. sisters Parent’s time input in child development v. parent’s need for childfree time Resources for children v. consumption of adults
Efficiency costs of the intrafamily equity tax
Transfers to families of children with development problems
Sufficient resources to able children whose parents are poor
Inputs of parent’s time v. inputs of experts in child development
Transfers to singleparent families
Optimal population growth
Opportunity costs of parent’s time (work, investment in parent’s human capital) v. child development through parent’s time Gains through investment in specialized human capital
Subsidized childcare Paid parental leaves
Demand for parent’s time outside home Returns to human capital investments of parents
Working-class parents v. middle-class parents
Lost productivity from social stratification v. human capital investment not passing cost benefit test
Division of work between husband and wife
Bargaining over time use and expenditures between husband and wife
Men v. women: alimony, widows’ pensions, displaced homemakers’ insurance
From the interfamily perspective, suppose an equity goal is to offset the low well-being of the family with a disabled child. While this may be seen as purely an equity move, there are important efficiency aspects: what form should the transfer take? Money to the parents or resources in kind? If the parents are informed altruists, this choice may be unimportant. Suppose all parents are informed about how to provide inputs to a child with handicaps, but some parents are selfish. An in-kind transfer will have the advantage that altruists will behave in the same way as if they received cash, but non-altruists will be constrained to do more than they otherwise would. Parental transfers, particularly to children with learning problems, are possibly more efficient in the form of non-human investment, but many parents act and social programs are often designed to create more human wealth as the equity mechanism rather than simply transfer non-human wealth. In general, a system of families and social insurance with a very egalitarian view will deliver less resources to those more able to learn than will a less egalitarian system, and this might lead to underdevelopment of a country’s research and development potential and to human investments beyond the range passing a cost-benefit test. It is important to note that when some parents are poor, family reciprocal altruism may not be a viable mechanism to insure efficient investment across children of given ability in different families for reasons of liquidity constraints, uncertainty about the investment, and the robustness of the reverse transfer back to the
112
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
parent. We will develop this point below. However, resources from parents to children also raise efficiency concerns since adults too can invest in human capital or earn a rate of return on invested capital. One of the issues raised by feminists is the need for time away from young children, while traditional views have often emphasized the idea that the mother be available to meet the needs of the children 24 hours per day. The latter ideology dominated The Netherlands until very recently. Dutch feminists have spoken against this “duty of presence” assumption, although some Dutch feminists accept the duty of presence but want to share it with the father of the child in equal proportion (Droogleever Fortuijn 1993; Pott-Buter 1995). The debate over this involves issues of gender and child/parent equity1 but also relates to often implicit beliefs about the productivity of the mother’s time relative to other caregivers. Trained caregivers may be as effective or more effective in developing a child’s capacity. All countries provide general basic public education for all under the assumption that schoolteachers are needed for the development of children’s skills. The question that may be most relevant is what mix of the mother’s time and an expert’s time is most efficient at different ages of the child?2 Economic logic says that this also depends on the opportunity cost of the mother’s time, which in turn relates to the wage rate and the marginal investment value of on-the-job training. There will be the efficiency of the mother’s time use in the market as against her efficiency in the development of her child. If it were true that both child development and current GNP would rise by the substitution of specialist’s time for mother’s time and that there is limited scope for altruism, then there is a case for subsidized childcare. Swedish politicians have been more inclined than politicians in other western countries to value the contribution of trained specialists over a wide range of circumstances. Policies to alleviate the financial burden of families with children may also induce people to have more children and this may result in swings in cohort size causing what have been termed optimal population problems.3 Table 7.2 indicates possible equity and efficiency trade-offs between boys and girls and between men and women. We note that one question in this broad area is the potential efficiency gain from specialization between the husband and wife (Becker 1981) which could lead to market activity by only one spouse (the husband?) while the other (the wife?) specializes in childcare and other non-market work. This result is modified by numerous factors, ranging from discrimination in the labor market to marriage matches between men and women with similar preferences for shared consumption (importantly including children and their development). Time use panel research indicates some role for both division of labor (specialization) and family public goods (shared consumption). Stable marriages are more likely to be those where the husband and wife spend time in separate housework activities but more time together in leisure activities, ranging from active sports to socializing and even to watching television together (Hill and Juster 1985). Recently, a great deal of work has explored the bargaining approach to marital relations and indicates that market income and resources available to each spouse contingent upon separation shape the intrafamily division of resources between married couples (Ott 1992). Under these conditions public childcare creates a type of insurance for mothers as well as strengthening their bargaining position. This appears to be one of the reasons for more equal division of housework in Swedish families (Juster and Stafford 1991). If we are right about the increasing importance of human capital accumulation for individual well-being it follows that both equity considerations and efficiency considerations over time will require that each individual is ensured a right to accumulating human capital. Below we will develop in more detail some aspects of the equity and efficiency trade-off between children, the child/child dimension and some aspects of the child/parent dimension.
CHILDCARE AND HUMAN CAPITAL
113
Equality between children and effects on efficiency Parents spend time in discovering and developing particular talents and abilities in their individual children. One motivation for such a search is to put resources in a direction where a higher payoff is expected, i.e. an efficiency motivation. Another motivation for such a search is to realize a type of fairness: each child will be regarded as successful in some realm. But what if one child has a large overall developmental deficit compared to the siblings? In terms of economic theory one might postulate that the parents could compensate this child by transfer of resources in the form of physical capital or wealth rather than in the form of individual skill or human capital. Perhaps this choice of form of capital transfer is used to some extent and may be important in some cases. We do know that equal human capital outcome appears to be a substantial motivation in many countries. A dramatic example of this is in the case where a child has Down’s Syndrome. Time inputs to childcare by parents of preschool-age children (age 0–3) with Down’s Syndrome is of the order of 26.2 hours of primary childcare time per week compared to 10.1 hours per week for the US national average (Stafford 1991). Analysis suggests also that siblings without disabilities receive less parental time than they otherwise would. In this section we will present a highly stylized model to clarify the equity and efficiency considerations that may arise in the transfer of resources to different children. There is a substantial body of economic literature which investigates how families and other institutions operate to “produce” human capital and achieve some measure of equality in child development. It has been suggested that parents “try both to allocate resources equally between their children and to compensate, to some extent, for the handicaps of the children with lower natural endowments” (Griliches 1979). It has also been argued that parents and children behave in a mutually altruistic manner (Becker 1981). If this is so, the family is a central institution for the study of equity and efficiency. In some instances family altruism can create efficient outcome across families and in some instances family behavior aimed at achieving equity appears to be at odds with efficiency. Figure 7.1 depicts parents’ choices for allocating resources between two siblings. Consider the simple linear relations for “production” of child development for two siblings. (7.1) where total child development K=k1+k2 is produced with resources to children Z=c1+c2. In Figure 7.1 k1 and k2 represent the children’s respective developmental outcomes and c1 and c2 are the inputs into child 1 and child 2 of parental time and other developmental inputs. If one child is more able (a01>a02) and has greater ability to learn (a11>a12), then the production possibilities can be represented by the production frontier BC in Figure 7.1. Suppose parents are concerned with maximizing total child development, K, irrespective of how it is shared between the two children of Figure 7.1. Then child development indifference contours are simply linear with a slope of −1 as with line BC' in Figure 7.1, and the chosen point would be B, where all resources go to the more able sibling. The only point on the indifference curve BC' which can be reached along the production frontier BC is point B. Of course one could use a more plausible production frontier (one which was “bowed out” and did not allow such extreme specialization of resources), but here we wish to motivate thinking on the subject and simply highlight differing parental choices between efficiency and equity. Suppose in Figure 7.1 that the parents seek some form of equity for the siblings. One definition of equity is that there be equal inputs, which for the resource level Z implies an outcome at point D. Alternatively, equity can be defined as the parents’ preference for equal outcomes. Such cases can be represented by
114
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
Figure 7.1 Intersibling equity tax
indifference curves with a slope of −1 at the point where they cross the 45° line. As represented in Figure 7.1, we have only one point where equal outcomes are obtained, namely point S.4 This point requires that the indifference curve is tangent to the slope of the production opportunity set along the 45° line. This occurs only if the slope of the production opportunity set is also −1, which is only true if −a11/a12=−1. When one child has inherent learning disabilities in relation to the other, families will only partly equalize outcomes. If we think of successively smaller values of a12, the choices will trace out a locus such as SS". Only if the parents are concerned with equity on an absolute basis (with fixed proportion indifference curves) would they fully equalize outcomes (SS' along the 45° line). In each of these three representations (at D, along SS' or along SS"), we can think of the more able child as subject to an intersibling “equity tax”: the more limited one sibling’s ability, the less of a given amount of resources (Z) which will go to the other’s development. The more the solution deviates from the equal inputs point D, the larger is the equity tax incurred by the more able child. The ambitious goals of Swedish handicap policies (SOU 1991:46) aim at transferring government resources to the point where both parents of the handicapped child should be allowed to pursue their work careers. Also in some cases a substitute caregiver can be supplied in the evening or during night hours in order for parents to devote themselves to a sibling. If these ambitious goals are achieved the equity tax of Figure 7.1 would be paid by the general taxpayer and not by the sibling or the parents. In this case, defined by the point where parent resources are equalized, there would be no efficiency loss due to underinvestment in human capital of the sibling.
CHILDCARE AND HUMAN CAPITAL
115
Figure 7.2 Child development and the distribution of well-being over children and parents
The idea that the equity tax in the case of handicapped siblings should be paid by tax money is also the guiding principle in handicap policies in The Netherlands. The organization of handicap policies is that the mother of the handicapped child is generally expected to bring the child to the different specialists or training centers, and the goal that both parents should be allowed to pursue their work careers is not a guiding principle. Having a handicapped child in most cases negatively affects the working career and human capital accumulation of at least one of the parents. In the next sections we will analyze some efficiency trade-offs across generations and particularly the mother/child trade-off. Intergenerational equality and efficiency: one aspect of the child/parent dimension The second model is presented in Figure 7.2. It represents the choices the parents face in using resources for the child or for themselves in achieving different levels of their well-being. To illustrate some of the conceptual issues as a prelude to understanding differences in social policy and family behavior, we begin with the notion of reciprocal altruism (Becker 1981:198) between parents and their children.5 Reciprocal altruism implies a type of equity. In Figure 7.2 we have indifference curves between the well-being of the parent (P) and the child (C) which have a slope of −1 at the point where they cross the 45° line. Reciprocal altruists prefer that the other be at the same level of well-being and in our idealized world would not act opportunistically if getting to the 45° line (i.e. equality) required transfers between them, and transfers that extended through time. The investment of parental time and other resources in children involves a reduction in parental resources. In Figure 7.2, “poorer” (in term of resources) parents are closer to the origin, and the “richer” the parent, the
116
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
greater the distance from the origin. Along the P-axis is the level of resources available to a parent. By investing in the human capital of the child, resources to the child rise, but at a decreasing rate, as represented by the curves rising to the northwest from the horizontal axis. At some point the parental transfer of resources to the child is more efficiently accomplished by non-human investment, as represented by the dashed lines with a slope of −1, suggesting a zero interest reference case, so that simple transfers and non-human investment can be represented jointly. In a rich family of reciprocal altruists (Pr), the parents would invest in the human capital of the child with their own time or market inputs to the level He,6 and beyond this would “give” resources to the child in the form of financial/monetary investment (the point where the indifference curve is tangent to the 45° line). Average income (Pa) and poor (Pp) parents would (as drawn) invest in the human capital of a child of the same ability to the same extent as rich parents, provided that reciprocal altruism exists (and that the rich and poor parents have equal parenting skills or can hire specialists such as schoolteachers to achieve equal effectiveness). In this case they would “sacrifice” to invest to the same level, He, with the expectation that the child would return the favor with a reciprocal transfer later in life. In the case of poor parents this reverse transfer would end up being quite large, namely DE, testing the limits of the system and occurring less commonly. An illustration of this is the sports star who transfers resources to the parents who made success possible. If reciprocal altruism is sufficiently pervasive, it has two elements to sustain it. First, as noted by Becker, families practicing reciprocal altruism could be expected to have stronger survival value in terms of influence, if not numbers. Second, from a social perspective there is efficiency in the conventional economic definition. Children of equal ability receive equal investment so that marginal returns on human investment across children are equal and equal to the marginal return on financial/monetary investment. We have seen that one type of equity, reciprocal altruism in the family, can be compatible with normal concepts of efficiency or social welfare. Does this extend more generally or is there a trade-off between efficiency and equity under other conditions? Does the equity tax and efficiency loss from the previous section also reappear in some form in the child/parent consideration of this section? We argue that it does. Consider the case where some children have less developmental potential as represented by the thinner curves originating at Pr, Pa and Pp and hitting the vertical axis closer to the origin than is the case for the child development curves considered above. Efficiency requires a smaller investment in such children, and a rich parent would invest less in human capital and more in financial/monetary capital, as represented by the solid line with the slope of −1 from point C in Figure 7.2. Within a certain range this may be seen as “acceptable.” Now consider the situation of a poor parent with a child with pronounced learning problems or physical disabilities. On their own, and under reciprocal altruism, the result raises two obvious equity problems. First, if the parents and children act as efficient reciprocal altruists, the equilibrium could be a level of child well-being below some socially acceptable minimum, Hm,7 and the parents will be subject to a type of interfamily equity tax (relative to other parents) in their effort to improve their child’s well-being. Social insurance programs can be developed in an effort to boost the child’s well-being toward and beyond Hm. These programs use public funds to allow the parent to move in tandem with the child northwesterly along the 45° line. Important issues for the design of such social programs include whether parental earnings potential and child development skills are positively correlated and the extent to which specialists in child development and education can help children overcome developmental problems, particularly for children with the greatest degree of disabilities. Where experts are perceived or known to be effective, publicly
CHILDCARE AND HUMAN CAPITAL
117
provided education (such as special education programs or Head Start in the United States) (Lamb et al. 1992) provides both a more efficient production of human capital and a well-being transfer to the parents of such children. Another possibility is to provide services intended to boost the parents’ income, as with public programs for employment-related training of disadvantaged parents. In the US there has been a growing belief that publicly sponsored job training for poor parents, particularly those with children, will allow the families (both parents and children) to improve their well-being substantially. The purpose of our highly stylized models in Figures 7.1 and 7.2 is to set out some of the main elements in the intergenerational production and distribution of well-being. Clearly the approach is oversimplified, but it still characterizes some of the complex and conflicting elements in the overall picture. Some conflict between equity and efficiency seems inherent. As Tinbergen (1975:133) suggests, when maximizing social welfare subject to the constraint of equity, “every additional restriction will lead to lower social welfare.”8 In this context, the efficient altruist model above seems more likely to be the exception than the rule. Efficiency considerations in the mother’s time use: a second aspect of the child/ parent dimension In the previous section we analyzed resource allocation by parents in favor of children. The analysis was carried out assuming a composite transfer of time and goods. Numerous estimates of the costs of children hold the time costs to be by far the largest part of the resources that parents transfer to children. (See Cigno 1991: chapter 6 for some examples; a recent estimate for The Netherlands is Dankmeyer 1996.) The transfer of time by parents to the care of children, particularly that by mothers, carries an opportunity cost consisting of two parts: the direct market wages forgone and the human capital accumulation forgone. Empirical estimates of these costs inspired by the seminal paper by Mincer and Polachek (1974) on US data are now available for several countries (Gustafsson 1981 for Sweden; Schippers 1987 for The Netherlands) and invariably show that women who leave the labor market for some years to care for their children earn less upon return to the labor market than similar women or men with uninterrupted labor market careers. However, time inputs by parents are necessary for child development and affect the eventual well-being and capabilities of their children as adults. Based on a small panel sample from the Time Use Survey (US) over the period 1975/6–81/2, it was found that subsequent teacher ratings of an array of school performance measures were positively related to prior preschool time inputs by the mother, but were negatively related to market work time of the mother (Stafford 1987). While time used by the mother in the care of her child increases the resources available to children, there is an important conflicting time use in women’s labor force participation and investments enhancing their own skills. As human capital has become more important over time, the potential efficiency loss of women withdrawing from the labor market for mothering increases. In this section we briefly discuss the policy responses by some industrial countries to this conflict. (For a fuller account see Gustafsson 1994; Gustafsson and Stafford 1994, 1996.) The role of opportunity costs in the above choice is accentuated by situations of high unemployment and low demand for market labor. Under such conditions the public debate and public policy initiatives tend to idealize the mother’s parenting role (as during the Great Depression), while in situations of “labor shortage” (such as during World War II and the European postwar boom), there are proposals that more resources be made available for public childcare. In the United States during the Great Depression, public funds for childcare were made available to supply jobs for unemployed teachers, cooks, nurses and nutritionists (Getis and Vinovskis 1992). However, when employment increased in 1939–40, childcare funds from the
118
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
Works Progress Administration were withdrawn. As the United States entered World War II a new reason arose for subsidizing childcare: mothers were needed in the war economy, and the Lanahan Act succeeded in enrolling large numbers of children, despite limited funds. In comparing recent Swedish and Dutch history, the “shortage” of labor in Sweden has made it easier to argue for public day care, whereas high unemployment in The Netherlands has fostered the common belief that increased female labor force participation would only exacerbate the difficulty of finding work. High opportunity costs of children should lead some women to choose voluntary childlessness or to delay births up to an age where biological factors will limit fertility. Government policy then often centers on creating fertility incentives. Concerns over low fertility were one motivation behind the argument by Alva and Gunnar Myrdal (Myrdal and Myrdal 1934) in their famous Crisis in the Population Question to argue for transfers to families with children as a way of decreasing the costs of children. To discourage fertility, policies have been tried in developing countries and for low-wage populations in the US which provide incentives not to have children, by direct compensation (or reducing a preexisting level of child-dependent compensation such as AFDC) or by increasing the labor market opportunities of young women. Sweden has a policy in which benefits are dependent on having children and working in the market jointly, which “explains” that country’s high labor force participation rates in conjunction with high fertility rates (Sundström and Stafford 1992). The Dutch view that parenting is a full-time job and the Swedish view that both parents have equal responsibilities for care and a right to a market job are visible in the different ways the two countries have organized the public participation in the education of young children. Until recently there was virtually no extrafamily, full-time day care in The Netherlands, but public support has gone into part-time playschools. Also, primary school (which is universal from age four) often does not include the lunch hours, and school hours are determined individually by each school. The result is that if two siblings go to different schools there may be eight trips a day for the mother in order to bring children in the morning, take them home for lunch, bring them to school in the afternoon and take them home after school. To solve this “mother-as-achauffeur” problem, Sweden decided on the “continuous school day” in 1974. Practically, this means that all children go to school every weekday at around 8 o’clock and that schools have an obligation to supply activities for the children during breaks in the regular instruction. The situation regarding childcare for mothers with market work can be summarized as follows in the three countries. In Sweden subsidized, quality public childcare is virtually universally available, and mothers have jobs (Gustafsson and Stafford 1992; Gustafsson 1994, 1995). In the United States, there is a market for childcare and mothers to a large extent have jobs (Leibowitz, Klerman and Waite 1992). Until recently in The Netherlands there was neither a market for nor subsidized childcare, and mothers rarely had jobs. However the number of mothers with small children who work in the market has been increasing. By 1992, almost 43 percent of mothers of preschool children were labor force participants (Maassen van den Brink 1994). The Childcare Stimulation Act of 1991 in The Netherlands is the first government action which explicitly caters to the needs of the working mother rather than assigning priority to children’s educational needs. The 1991 act subsidizes childcare centers and family day care homes, but excludes (the part-time) playschools from subsidies because they are not meant for the working mother (Tijdens and Lieon 1993). The efficiency conflict in parents’ time use is in principle theoretically possible to solve. The marginal output of time spent in childcare should equal the marginal output in time spent in market work. The output of time spent in childcare may extend over the child’s lifetime and has effects on his or her productivity as
CHILDCARE AND HUMAN CAPITAL
119
an adult. Similarly there is a lifetime aspect of the output of time spent in market work due to on-the-job human capital investment. Therefore, both outputs must be computed over a considerable time period and discounted back to the time when the input in time use was made. However, the accumulation of individual human capital is also an equity problem between parents and children which could be analyzed as in Figure 7.2 above. Some parents who sacrifice accumulating their own human capital may end up in a disadvantaged position in comparison to their children (above the 45° line in Figure 7.2). The child, if not altruistic towards his or her parent, may not transfer back when adult. This is the case when a mother sacrifices her own development to cater to the needs of the children. The mother’s transfer of time resources to her child is also the major reason for another equity problem, that between men and women. It has been shown by numerous analyses of the male/female wage differential that an interrupted labor market career is one major determinant of lower female wages (Mincer and Polachek 1974; Schippers 1987; Gustafsson 1981). Indirectly, interrupted careers or expectations on the part of employers that women will interrupt their careers result in discrimination against women in the workplace which adds to the effect of a smaller lifetime investment in human capital caused directly by the labor force interruption. CONCLUDING REMARKS In this paper we have shown that human capital has become more important over time both as an individual asset and as a requirement for economic growth. The individual accumulation of human capital is therefore both an equity and an efficiency requirement. In Table 7.1 we listed some reasons for the growing importance of childcare in the accumulation of human capital. Childcare has a central role both in the accumulation of human capital and in equity considerations. In Table 7.2 we listed some examples of equity/ efficiency trade-offs that are affected by childcare policies. In the three following sections we analyzed in some further detail a few of the trade-offs listed in Table 7.2. Recently, in the face of government budget deficits, the Swedish family policies have been questioned on the basis of efficiency (Rosen 1995). We believe that efficiency can be improved in this area not only in Sweden, where a large share of the government budget is devoted to these purposes, but also in The Netherlands and the USA. If we are right in our hypothesis that individual human capital accumulation becomes more important over time, efficiency evaluations of family policies should include evaluations of such productivity-enhancing effects. A serious evaluation of the efficiency effects of family policies cannot, like Rosen (1995), disregard the vast literature on human capital accumulation and policy choices in the childcare area in different countries. Although this literature is large, so far it has not devoted much explicit attention to the evaluation of the efficiency aspects of family policies since these policies have mostly been motivated by equity reasons. Such evaluations could be framed in a comparative perspective, using policies and data from several countries, as has been demonstrated in this paper. REFERENCES Altonji, Joseph G., Hayashi, Fumio and Kotlikoff, Laurence J. (1992) “Is the Extended Family Altruistically Linked? Direst Tests Using Micro Data,” American Economic Review 82, 5:1177–98. Barro, Robert and Jong Wa Lee (1993) “Winners and Losers in Economic Growth,” National Bureau of Economic Research, Working Paper No. 4341. Becker, Gary S. (1981) A Treatise on the Family, Cambridge, Mass.: Harvard University Press.
120
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
Bound, John and Johnson, George E. (1992) “Changes in the Structure of Wages During the 1980s: An Evaluation of Alternative Explanations,” American Economic Review: 371–92. Cigno, Alessandro (1991) Economics of the Family, Oxford: Clarendon Press. Dankmeyer, Ben (1996) “Opportunity Costs of Children in The Netherlands,” Journal of Population Economics, 9, 3: 341–61. Droogleever Fortuijn, Joos (1993) Een Druk Bestaan. Tijdsbesteding en ruimtegebrek van tweeverdieners met kinderen (A Busy Life. Time and Space Use of Two Earner Families with Children), Amsterdam: Sociaal Wetenschappelijke Studies diss. Eaton, Jonathan and Kortum, Samuel (1995) “Engines of Growth: Domestic and Foreign Sources of Innovation,” paper presented at the Center for U.S.-Japan Business and Economic Studies, Tenth Annual Symposium, Stern School of Business, New York University. Flood, Lennart and Gråsjö, Urban (1995) “Changes in Time Spent at Work and Leisure: The Swedish Experience 1984– 1993,” paper presented at the 15th Arne Ryde Symposium, Denmark. Getis, Victoria L. and Vinovskis, Maris A. (1992) “History of Child Care in the United States. Before 1950,” in M.E.Lamb, K.J.Steinberg, C.-P.Hwang and A.G. Broberg (eds) Child Care in Context, Cross-Cultural Perspectives, Hillsdale, N.J.: Lawrence Erlbaum Associates. Goldin, Claudia (1990) Understanding the Gender Gap, Oxford: Oxford University Press. Griliches, Zvi (1979) “Sibling Models and Data in Economics: Beginnings of a Survey,” Journal of Political Economy 87, 5, part 2: S37–S64. Gustafsson, Siv (1981) “Male-Female Earnings Differentials and Labor Force History,” in G.Eliasson, B.Holmlund and F.Stafford (eds) Studies in Labor Market Behavior, Sweden and the United States, Conference Reports, Stockholm: The Industrial Institute for Economic and Social Research. —— (1994) “Childcare and Types of Welfare States,” in Diane Sainsbury (ed.) Gendering Welfare States, London: Sage Publications. —— (1995) “Single Mothers in Sweden: Why Is Poverty Less Severe?,” in K.McFate, R.Lawson and W.J.Wilson (eds) Poverty, Inequality and the Future of Social Policy. Western States in the New World Order, New York: Russel Sage Foundation. Gustafsson, Siv and Stafford, Frank (1992) “Childcare Subsidies and Labor Supply in Sweden,” Journal of Human Resources 27, 1:204–30. —— (1994) in Rebecca Blank (ed.) Social Protection versus Economic Flexibility: Is There a Trade-off?, Chicago: National Bureau of Economic Research and University of Chicago Press. —— (1996) “Equity and Efficiency Trade-Off s in Early Childhood Care and Education,” forthcoming in S.Barnett and S.Boocock (eds) Early Childhood Care and Education for Disadvantaged Children: Long Term Effects, New York: SUNY Press. Hill, Martha S. and Duncan, Greg J. (1987) “Parental Family Income and the Socioeconomic Attainment of Children,” Social Science Research 16:39–73. Hill, Martha S. and Juster, F.Thomas (1985) “Constraints and Complementarities in Time Use,” in F.T.Juster and F.P.Stafford (eds) Time, Goods and Well-Being, Ann Arbor: Institute for Social Research, University of Michigan. Johnson, George E. and Stafford, Frank P. (1997) “Alternative Approaches to Occupational Exclusion,” in I.Persson and C.Jonung (eds) Women’s Work and Wages, London: Routledge. Juster, Thomas F. and Stafford, Frank P. (1991) “The Allocation of Time: Empirical Findings, Behavioral Models, and Problems of Measurement,” Journal of Economic Literature 29:471–522. Knutsen, Oddbjørn (1991) “Offentlig barneomsorg i Norden. En komparativ studie av utviklingen i de nordiske land,” INAS report 91:2, Oslo. Lamb, Michael E., Steinberg, Kathleen J. and Knetterlinus, Robert D. (1992) “Childcare in the United States: The Modern Era,” in Michael E.Lamb, Kathleen J.Steinberg, Carl-Philip Hwang and Anders G.Broberg (eds) Child Care in Context, Cross-Cultural Perspectives, Hillesdale, N.J.: Laurence Erlbaum Associates. Leibowitz, Arleen, Klerman, Jacob Alex and Waite, Linda J. (1992) “Employment of New Mothers and Child Care Choice,” Journal of Human Resources 27, 1:127.
CHILDCARE AND HUMAN CAPITAL
121
Maassen van den Brink, Henriëtte (1994) Female Labor Supply, Childcare and Marital Conflict, Amsterdam: Amsterdam University Press. Mincer, Jacob and Polachek, Solomon (1974) “Family Investments in Human Capital, Earnings of Women,” Journal of Political Economy 81:76–108. Myrdal, Alva and Myrdal, Gunnar (1934) Kris i befolkningsfrågan, Stockholm: Bonniers. OECD (1994) Economic Growth, No. 56, Paris: OECD. Ott, Nottburga (1992) Intrafamily Bargaining and Household Decisions, Berlin: Springer Verlag. Pott-Buter, Hettie (1993) Facts and Fairy Tales about Female Labor, Family and Fertility. A Seven Country Comparison, 1850–1990, Amsterdam: Amsterdam University Press. —— (1995) Veranderingen in de levensloop van vrouwen. Ontwikkeling van vrouwenarbeid in zes landen, Amsterdam: Nationaal Vakbondsmuseum, Welboom. Rosen, Sherwin (1995) “Public Employment and the Welfare State in Sweden,” NBER/SNS Project on reforming the welfare state, SNS Occasional Paper No. 61, January. Schippers, Joop J. (1987) Beloningsvershillen tussen mannen en vrouwen—een economische analyse, Groningen: Wolters-Nordhoff. Solon, Gary (1992) “Intergenerational Income Mobility in the United States,” American Economic Review 82, 3:393–408. SOU (1991:46) “Handikapp, välfärd, rättvisa” (Handicap, Well-being, Justice), Stockholm: Allmänna Förlaget. Stafford, Frank P. (1987) “Women’s Work, Sibling Competition, and Children’s School Performance,” American Economic Review December: 972–80. —— (1991) “Early Education of Children by Families and Schools,” Department of Economics, University of Michigan. Sundström, Marianne and Stafford, Frank (1992) “Parental Leave and Female Labor Force Participation and Public Policy in Sweden,” European Journal of Population 8:199–215. Tijdens, Kea and Lieon, Saskia (eds) (1993) Kinderopvang in Nederland, Organisatie en financiering, Utrecht: Jan van Arkel. Tinbergen, Jan (1975) Income Distribution: Analysis and Policies, Amsterdam: North-Holland Publishing Company. United Nations (1995) Human Development Report, Oxford: Oxford University Press. Varian, Hal (1980) “Redistributive Taxation as Social Insurance,” Journal of Public Economics 14:49–68.
NOTES 1 An early Swedish government document (SOU 1938:20 cited by Knutsen 1991: 74) argued that one reason for supporting kindergarten was to give housewives a few hours of leisure time. 2 A pioneer in the study of stages of child development was the French psychologist Piaget. 3 Swings in population cause problems in planning for schools, universities, housing, and public and private pension systems. In the context of Figure 7.2 in this paper, population booms and busts create swings in the price of market goods across generations, changing the slope of the market transfer line and leading to losses of wellbeing. 4 Note that here the endowed difference, a01>a02, is fully offset so long as a11= a12. 5 Recent tests of altruism in the transfer of income among extended family members suggest a weak role for reciprocal altruism (Altonji, Hayashi and Kotlikoff 1992). However, parental resources for young children are a very different and more important matter in our opinion. Parents may transfer resources at a generous rate until at some point the children are “on their own.” 6 The actual effectiveness of one’s own time versus market inputs is of major significance. If a parent’s (mother’s) own time is valuable over a wide interval, then the market work by the mother would be lower. If the market wage of women rises and market-purchased or other non-family inputs are seen as productive of child
122
SIV S.GUSTAFSSON AND FRANK P.STAFFORD
development, then extensive market careers with on-the-job training of mothers become more efficient for altruistic families. 7 There may be a strong parental and social concern about the form of the child’s capital. Simply transferring nonhuman wealth may not be seen as equitable. It may be that substantial weight is placed on human capital, as illustrated by the strong effort by parents of children with Down’s Syndrome to achieve some level of personal functioning of the child. Societies appear to differ substantially in this regard, with disability programs in the US placing greater relative weight on total transfer rather than on the form of the transfer. 8 As noted in a somewhat different context, the cost of delivering equity in an insurance setting has a greater deadweight cost when outcomes depend on productive behavior and effort rather than on luck (Varian, 1980).
8 THE CHOICE BETWEEN FULL-TIME AND PART-TIME WORK FOR NORWEGIAN AND SWEDISH MOTHERS Marit Rønsen and Marianne Sundström
INTRODUCTION In the last decades employment for mothers of young children has risen in most western countries. This rise has perhaps been most pronounced in the Scandinavian countries, and comprised primarily part-time employment. For example, in 1988 the labor force participation rate among mothers of preschool children was 86 percent in Sweden and 72 percent in Norway, while it was only about 40 percent in West Germany and 47 percent in the UK (Eurostat 1992). The Scandinavian countries also have relatively high rates for part-time work among women. In 1988 Sweden had a part-time rate of 43 percent and Norway of about 50 percent, which was higher than in West Germany and France (31 and 24 percent, respectively), but in the same range as in the UK (44 percent) and lower than in The Netherlands (57 percent) (Eurostat 1995). The Scandinavian countries are renowned for their extensive social policies: policies such as separate taxation, subsidized childcare, generous parental leave and economic support to families with children are likely to have facilitated and encouraged the gainful employment of mothers of young children (see, e.g. Sundström and Stafford 1992). Less is known, however, about how these policies affect the mothers’ choice between full-time and part-time work. This paper analyzes the impact of public policies and other factors on women’s re-entry rates into fulltime and part-time work after the birth of their first child in Norway and Sweden. The two countries are very well-suited for a comparative analysis of policy effects since they are culturally very similar and have the same set of public policies, but with country-specific variations. We also used two data-sets which are extremely similar in design, namely the 1988 Norwegian Family and Occupation Survey and the 1992 Swedish Family Survey. The study is organized as follows: the next section sets out our theoretical and methodological framework, followed by a brief overview of the major family policy programs of the countries and a presentation of the data and variables used in the analysis. The subsequent section contains descriptive statistics of our samples as well as a comparison of full-time and part-time work by Norwegian and Swedish women in general. Next, we report our findings, finally ending with a summary and discussion. THEORY AND METHODS The analysis is based on a standard intertemporal labor supply model in which the fertility decisions have already been made. The mother maximizes the discounted expected family utility, and returns to work when
124
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
her full wage exceeds her reservation wage. The full wage reflects her forgone earnings from not returning and consists of the current market wage as well as the present value of the reduction in future earnings associated with depreciation and non-accumulation of human capital. The reservation wage is the lowest wage rate for which the woman is willing to work. It reflects the utility of her time spent at home and varies with individual preferences and the family situation. The birth of a child will raise the reservation wage by increasing the demand for the mother’s time in childcare, but also lowers it by increasing the demand for market inputs in home production (see e.g. Hotz and Miller 1988). Since the time component is more important when the child is small, the presence of a newborn child will raise the reservation wage, but as children grow older they become less time intensive and more goods intensive. Hence, the reservation wage is hypothesized to decrease with time since childbirth. The timing of the return to work will thus depend on changes in the reservation wage as well as in the full wage. When modeling these dynamics, the hazard rate is a useful concept. It is a representation of the (unobserved) rate with which an event occurs within a certain short time interval, given that it has not occurred before. In our case, the career break may end by a return to either full-time or part-time work. We then have a so-called “competing-risk” model, and the event-specific hazard function can be written as (8.1) Here hj(t) is the hazard rate associated with event j and Pj(.) is the probability that event j occurs in the time interval ∆t, given that no event happened before that time. Individuals experiencing an event other than j are censored at the time of the other event. Since the two events are mutually exclusive, it follows from straightforward probability calculus that the overall hazard function h(t)—the hazard of employment entry regardless of working hours—is the sum of the full-time and part-time hazards. There is little a priori knowledge of the functional form of the hazard rate of employment—be it full time or part time—after having a child. A rising full wage and a falling reservation wage imply a rising hazard rate, while a falling full wage and a rising reservation wage imply a falling hazard rate. If the full wage and the reservation wage move in the same direction, the direction of the resulting hazard is ambiguous. We chose the Cox proportional hazards model for our analyses as it makes no assumptions about the functional form of the baseline hazard. Besides depending on time, the hazard rate will vary with individual characteristics. In the competing risk case it can be written (8.2) where h0j is the unknown baseline hazard rate of event j, X is a vector of covariates which may or may not depend on time and βj is a vector of parameters associated with event j. The estimates of the hazards for full-time and part-time work are obtained by maximum likelihood estimation. It can be shown that the overall likelihood function for the event of employment can be split into a separate likelihood function for each of the full-time and part-time events (see e.g. Petersen 1995). Estimation can be done simultaneously for both events or separately for each event. As long as there are no restrictions on the parameters across the event-specific rates and no unobserved variables common to, or correlated across, the rates, these procedures yield identical results. We base our analysis on these assumptions, and estimate the model with a simultaneous maximum likelihood procedure available in the software package TDA (see Blossfeld and Rohwer 1995). The women are followed for 36 months or until they make their first transition to full-time or part-time work, which is defined as 10–34 hours per week in the Norwegian survey and as 16–34 hours in the
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
125
Swedish survey. Censoring occurs 36 months after the first child is born (our period of observation), when the second is born, at interview (Norway) or at the end of 1988 (Sweden), whichever comes first. NORWEGIAN AND SWEDISH FAMILY POLICIES Although Norway and Sweden do not have any specific pro-natalistic policies, the measures intended to foster gender equality and secure the economic well-being of children and their families have, no doubt, had a strong impact on women’s ability to combine family and work. Two of the main components of these policies are maternal and parental leave and subsidized childcare. In 1968, the first year of our study, employed mothers in both countries were entitled to maternity benefits, but the eligibility period was much longer in Sweden: six months, as compared to 12 weeks in Norway. The benefits were not taxable and income compensation was relatively low. Swedish mothers received a compensation of about 65 percent of prior earnings, while Norwegian mothers received a flat rate of NOK 4 per day plus 0.1 percent of prior earnings. In 1974 there were two major extensions of the Swedish system: fathers became entitled to share the sixmonth leave, and the replacement rate was raised to 90 percent of prior earnings. At the same time benefits were made taxable and pensionable. Between 1974 and 1988 the Swedish entitlement period was extended several times. In 1975 it was prolonged to seven months, in 1978 to nine months, and in 1980 to 12 months, the latter three at a flat rate equal for all (from 1987 SEK 60 per day). Since 1989 the leave period has been 15 months. The replacement rate was reduced to 80 percent in 1995. In Norway there were no extensions in the maternity leave program until 1977, when the benefit period was prolonged to 18 weeks and fathers became entitled to share the leave, except for the first six weeks which were still reserved for the mother. In 1978 the benefits were raised substantially to cover 100 percent of former income for most working women and were made taxable and pensionable. Thereafter, no further changes were made until 1987, when the entitlement was prolonged to 20 weeks, and in 1988, the last year of our study, another two weeks were added. Since 1993, maternity benefits are granted for a period of 42 weeks with full pay or 52 weeks with 80 percent compensation. The maternal or parental benefits are based on prior earnings in both countries. In Sweden, the amount depends on the earnings record 240 days prior to birth. Mothers with no previous earnings receive only the low, flat-rate payment, i.e. SEK 60 per day in 1988 or SEK 21, 600 for the 12 months (taxable). To be entitled to leave with job security, Swedish parents must have been employed either for a minimum of the last six months before delivery or for at least 12 of the last 24 months prior to birth. Before 1977, Norwegian mothers had to have been employed for eight of the last ten months before birth to be eligible for benefits. In 1977 this requirement was reduced to six months’ employment during the last ten months. Mothers who did not fulfill this criterion were granted a one-time, tax-free cash payment at birth which in 1988 equalled NOK 4730. Clearly, the programs give young women in both countries a strong incentive to be employed full time prior to having children and even to postpone pregnancies until earnings are sufficiently high. In both countries parents are entitled to unpaid leave subsequent to the standard parental leave—in Sweden until the child is 18 months old (from 1978) and in Norway until the child is one year old (from 1977). In addition, Swedish parents employed full time in all sectors of the economy have (since 1979) had the right to reduce their working hours to 75 percent of full-time employment until the child is eight years old (see Sundström 1991 for further details). In Norway, there is no general statutory right to reduced
126
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
working hours, except for nursing mothers who are allowed two hours off per day with full pay in the public sector and one hour off without pay in the private sector. In fact, in both countries public employees have long had more favorable working conditions in connection with childbirth and childcare than employees in the private sector, e.g. higher income compensation and more extensive rights to unpaid leave and reduced hours. A major difference between the programs in the two countries lies in the degree of flexibility. As of 1974 the Swedish benefits can be used full time or part time (one-quarter, one-half, three-quarters of full time) or saved and used any time before the child is eight years old. Parents are free to interrupt the leave any time to go back to work and resume the leave at a later date. During the period studied this was not the case in Norway: benefits could neither be used part time, nor saved for later usage. Parents had to use up all their eligibility in one go: if not, the remaining benefit days would be forfeited. For this reason we expect exit patterns to look very different in the two countries. Also unique to the Swedish system is a feature which encourages a closer spacing of children (Hoem 1993). Since 1980 the mother has been able to maintain the same benefit level as with the previous child without returning to work if the next child is born within 24 months of the last. In 1986 this limit was extended to 30 months. Before 1980 the maternity benefits with a subsequent child were based on the earnings record between births, as it is in Norway. Another important component of Norwegian and Swedish family policies related to the employment of mothers is the provision of subsidized, high-quality childcare. Public childcare in the form of day-care centers, family day care (childminders) and after-school centers are provided by the municipalities with the support of large government subsidies. Norway has always lagged behind Sweden in the provision of public childcare, however. In 1973 about 5 percent of Norwegian preschool children were provided with public day care, while the Swedish coverage rate was double, 11 percent. During the 1970s and 1980s public childcare expanded rapidly to include 25 percent of Norwegian and 38 percent of Swedish preschoolers by 1983. In 1988, our last year of observation, 32 percent of Norwegian preschool children had a place in public childcare as compared to 49 percent of Swedish children. Enrollment rates are lower for children aged 0–2 years than for those aged 3–6 in both countries (for Sweden: Statistics Sweden 1992; Gustafsson and Kjulin 1993; for Norway: Rønsen 1995a). Compulsory school starts at age seven in both Norway and Sweden. Parents’ fees, which are set by the municipalities, cover only a fraction of the running costs. Generally, the fees increase with family income and decrease with the number of siblings in care. Swedish parents usually pay lower fees, but unlike Norwegian parents, they cannot deduct any of the costs from their taxable income. Single parents pay a reduced rate. Since there is a greater demand for childcare than there are places, spaces are rationed and as a rule the waiting time reflects the age of the child. The excess demand has been met by several forms of private childcare arrangements: for example, relatives, private baby sitters, au pair girls or private day-care centers. This type of childcare has been more common in Norway. DATA AND VARIABLES Recently Statistics Norway and Statistics Sweden have carried out family surveys with almost identical designs: the Norwegian Family and Occupation Survey of 1988 and the Swedish Family Survey of 1992. Both surveys are national probability sample surveys of selected cohorts, containing complete retrospective life histories on childbearing, cohabitation and marriage, educational activities and employment. Interviews
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
127
were obtained from a total of 4019 Norwegian women born between 1945 and 1968 and 3314 Swedish women born between 1949 and 1969. Since the hazard model does not include a random term, it is important to keep the sample as homogeneous as possible. We therefore restricted our analysis to women who were married or cohabiting and at least 19 years old when giving birth for the first time. Single mothers were excluded since their budget sets differ from those living with a partner, and very young mothers were left out because, in keeping with our framework, they most probably did not have the time or opportunity to work before birth. Further, we excluded women whose first birth was a multiple birth, those whose child died shortly after birth and those who lived abroad at the time of birth or shortly afterwards, as well as some cases of inconsistent or incomplete information. In the Swedish sample, women with no registered work interruption were excluded. In the Norwegian sample this was not possible due to missing reports on maternity leave among employed women. As remaining benefits are lost in Norway if mothers go back to work before the leave expires, it is very unlikely that they would do so. We therefore set the length of the employment break equal to the statutory leave in those cases where women were eligible for leave but reported no leave or too short a leave (about 200 cases). Altogether this left us with a Norwegian sample of 1749 women and a Swedish sample of 1416 women. Difference in preferences is another important source of heterogeneity, but one which is not easily observable. One indicator that has proven useful in several studies of demographic behavior is religious activity. In recent studies from Nordic countries religiously active women are found to be more likely to marry and less likely to enter into unmarried cohabitation (Blom 1994), less likely to divorce (Hoem and Hoem 1992) and more likely to have a third child (Kravdal 1992). We expect religiously active women also to be more committed to the home and thus more disinclined to resume employment after birth, particularly full-time work. Religious activity is measured by church attendance in the year prior to interview with high being defined as attending church at least three times per year. As religious attitudes are probably fairly stable throughout one’s life, the timing of the measurement should not invalidate its usefulness for our purposes. Previous demographic research has also established a clear negative association between cohabitation and traditional family values (see e.g. Lesthaege and Moors 1995). In line with these findings we expect marital status to reflect something of a woman’s attitudes toward family and work. Since we know the partnership histories, we are able to distinguish between married women who cohabited with their husband prior to marriage and those who followed the traditional pattern of marrying directly. The marital status variable is thus divided into three categories—directly married, married after cohabitation and cohabiting (see Table 8.1). Based on previous findings (Rønsen 1995b; Table 8.1 Descriptive statistics for the variables used in the analyses
Age at first delivery, years Marital status at first birth: Cohabiting Married after cohabiting Directly married Religious activity: High Low
Sweden
Norway
24.5
24.1
22.9 64.5 12.6
18.6 33.2 48.2
10.1 89.9
21.6 78.4
128
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
Separated during observ. period Work experience before childbirth, years Home-time prior to childbirth, months Education attained by first childbirth: Compulsory schooling only Compulsory schooling plus one year 2–3 years of secondary school Post-secondary school education Reported status after birth: On maternity leave Managing the household Sector of employment before birth: Public sector Other Period of first childbirth: 1968–73 1974–6 1977–9 1980–84 1985–8 Duration: alla) (months) Full time Part time Number of transitions of which to full time part time Number of women
Sweden
Norway
3.1 4.9 1.3
2.4 4.5 3.0
23.3 38.0 11.3 27.4
15.6 41.0 22.4 21.0
80.5 19.5
68.8 32.2
55.3 44.7
40.9 59.1
18.4 15.0 14.6 26.9 25.1 15.6 12.8 14.5 1067 (75.4%) 418 (39.2%) 649 (61.8%) 1416
24.9 15.4 12.9 24.6 22.3 16.2 8.5 11.2 1115 (64.9%) 634 (56.9%) 481 (43.1%) 1749
Note: a) Including censored cases
Rönsen and Sundström 1996), we expect in particular women who married directly to have more traditional values and lower re-entry rates after birth, especially to full-time work. To account for changes in the budget set that may occur after birth, we include a time-varying covariate which equals one if the woman gets separated or divorces while she is still under risk of going back to work, and zero otherwise. It may be argued that a separation or divorce produces such a fundamental change in the budget set that women who go through it should be censored at the time of separation. We have also experimented with such models, but as the results were very similar to those obtained from models without censoring at separation, we have not adopted this procedure here. In the Swedish policy setting, a separation is likely to induce women to go back to work because of the loss of family income. By contrast, a separation or divorce may weaken the work incentives among Norwegian women since they receive a temporary and income-related economic support in case of separation which is partly or fully reduced if they work. Because of the reductions in benefits the effective marginal tax rate is normally higher on part-time earnings
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
129
than on full-time earnings, which are more likely to be above the limit where benefits are lost completely and where there are no more reductions. Part-time work may, therefore, seem less attractive for single mothers in Norway. As previously argued, we expect a negative relationship between the reservation wage and the age of the child. The mother’s reservation wage will be highest close to birth when the demand for the mother’s time at home is high and the market substitutes are few and costly, and then decline as the child grows older. This variation in the reservation wage will be picked up by the duration variable itself. Also in accordance with our theoretical model, we include the mother’s age at first delivery and other human capital variables. Age at delivery is hypothesized to have a negative effect on the entry into employment since younger women will have longer work horizons and thereby lose more from a work interruption. In the same manner, women who have made larger human capital investments in education and work experience may have more to lose by an employment break. Education is the highest level attained when the first child is born, and is divided into four categories: compulsory schooling only (9 years), compulsory schooling plus one year, 2–3 years of high school, and post-secondary school education. Work experience is constructed from the employment histories, and is the full-time equivalent number of years worked up to the time of the first delivery, accumulated from the year of the mother’s 17th birthday. In several US studies, work during pregnancy has been found to be a strong predictor of after-birth employment (see e.g. Even 1987; Joesch 1994; Shapiro and Mott 1994). Recently, this result has also been confirmed for Norway and Sweden (Rönsen and Sundström 1996). By working late into pregnancy, women not only maintain a high level of on-the-job training, but also show a higher degree of work commitment. Further, they qualify more often for paid maternity leave. Because of the strong correlation between work during pregnancy and maternity leave, especially in Norway, we use a different indicator of recent work behavior in the present study, labeled prior home-time, and this is the number of months without any employment or educational activity counting backwards from the time of birth. When modeling the return to work in countries with a statutory and universal parental leave system, the program in itself will be an important determinant. As a first approach, and to pick up the effects of increases in the generosity of the programs, we include period of first delivery, divided into 1968–73, 1974– 6, 1977–9, 1980–84 and 1985–8. Generally, the exit intensity is expected to decrease over periods due to the extensions in the leave program. However, this effect may be counteracted by the increase over time in the availability of public childcare. Further, the period factor will reflect other trends such as business cycle effects. In particular, we would expect re-entry rates to be lower during downturns in the business cycle such as the 1980–84 period and higher during upturns such as in 1985–8. In addition to this general approach, our data allow the study of the importance of maternity leave more specifically, as we have information on whether the mother had a paid maternity leave or not. For Sweden this indicator is based on the respondents’ own reports, while for Norway it has been constructed on the basis of the eligibility criteria and the pre-birth employment records. Previously we found that having maternity leave greatly speeds up the return to work in both countries, but more so in Sweden than in Norway (Rønsen and Sundström 1996). The question we wish to address now is whether full-time and parttime work are affected equally strongly, and to what extent there are country differences in this respect. Finally, we are able to examine the role of public sector employment (central or local government). Previous US studies suggest, for example, that the degree of an occupation’s or sector’s work convenience helps to retain mothers after the child is born (Desai and Waite 1991). The public sector has a long tradition of offering part-time jobs and other flexible working schedules as well as improved maternity leave
130
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
arrangements. We therefore expect that those employed in the public sector will return to work, and especially to part-time work, earlier because of the wider opportunities for combining work and motherhood there. FULL-TIME AND PART-TIME WORK During the period we are studying, 1968–88, gainful employment among mothers of young children increased strongly in both Norway and Sweden. As early as 1972, 48 percent of Swedish women with children 0–2 years old were a part of the labor force as compared to only 29 percent among Norwegian mothers, according to the countries’ Labor Force Surveys. Throughout the 1970s activity rates among mothers of young children grew continuously so that by 1980 Sweden had reached a rate of 75 percent while Norway’s was 48 percent. During the 1980s the rise in Norwegian rates continued but the rise in Swedish rates slowed down, resulting in participation rates of 83 percent among Swedish mothers and 68 percent among Norwegian mothers of children under three in 1988. In Scandinavia the rise in employment among mothers of young children mainly concerns part-time work. In Sweden the proportion of all employed mothers with children under three who worked part time exceeded 50 percent in all years of our study, and even exceeded 60 percent from the late 1970s until the mid-1980s (Swedish Labor Force Surveys). In Norway similar figures from the Labor Force Surveys have been published only since 1989, when 57 percent of Norwegian married and employed mothers with children under three worked part time, which is very similar to the Swedish rate at the time. Other Norwegian surveys indicate that the proportion of part-time women workers may have been higher in the late 1970s and early 1980s (Ellingsaeter 1989). All in all, the Norwegian and Swedish rates among this group of employed mothers may have been rather similar for most of the period studied. Let us now turn to the full-time and part-time re-entry rates after the birth of the first child that we observe in our study. In Table 8.1 we saw that the overall proportion of mothers who return to work within 36 months after delivery is higher for Sweden, 75 percent as compared to 65 percent for Norway (for an analysis of these re-entry rates, see Rønsen and Sundström 1996). Also, the proportion of re-entrants who take up part-time work is larger for Sweden, 62 percent as compared to only 43 percent for Norway. The reason for the seeming contradiction with the percentages in the previous paragraph is that our study covers only first-time mothers, while mothers with children under three also include mothers of more than one child. The latter group of mothers work part time to a greater extent than the former in both countries, but the difference is larger in Norway (Ellingsaeter and Rønsen 1996). There is, however, one more paradox to explain. As pointed out in the introduction, the proportion of part-time workers among all employed women was higher in Norway than in Sweden in 1988 (and during the whole period studied). Why then do we find higher rates of part-time entry for mothers in Sweden? One reason has already been mentioned: in Norway mothers of more than one child have higher rates of parttime work than the corresponding group of mothers in Sweden. A second reason is that while in Sweden a substantial proportion of mothmrs shift from part-time to full-time employment when the children start school (Sundström 1993), this is not the case in Norway. On the contrary, the proportion working part time is higher among working mothers of children aged 7–10 than among those of children under three. Ellingsaeter and Rønsen (1996) suggest that this is because school-days are short at that age and it is difficult to find childcare outside school hours, but it probably also reflects higher part-time rates among older cohorts. The higher part-time rates among employed mothers of children aged 11–15 years is another
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
131
indication in this direction. This cannot be established with certainty from the Labor Force Surveys, however, since they are cross-sectional. We have summarized the descriptive statistics of our model covariates for our two samples in Table 8.1. The Swedish and Norwegian women in our sample are very similar in many respects: the mean age when the first child is born is almost equal and so is the rate of second pregnancy. The mean duration of the career break is quite similar −15.6 months in Sweden and 16.2 months in Norway—but the Norwegian women who return to work do so much sooner, on the average, especially if they return to full time, than do Swedish women. On the other hand, Swedish women worked more before birth while Norwegian women spent more time at home. Consequently, a higher percentage of Swedish women had maternity leave after birth. Also, Swedish women were employed to a larger extent in the public sector. Further, the educational profile is different: the Swedish sample has a higher frequency both of women with low education (only compulsory schooling) and of those with education beyond gymnasium. It is possible that the latter difference is due to the fact that the Swedish cohorts are younger and therefore more educated when the first child is born. Lastly, we see surprisingly large differences in values and attitudes of our Norwegian and Swedish women. Norwegian women are more religious and more often follow the traditional pattern of marrying directly. Almost half of the Norwegian mothers had done so, but very few in Sweden, where two-thirds of women had cohabited with their partner prior to marriage. FINDINGS We start by estimating a model which includes covariates that reflect differences in preferences, budget sets and human capital accumulated at birth. The results show that some of the effects of the covariates on fulltime and part-time re-entry rates do indeed differ between the two countries (see Table 8.2). First, we see that while the effect of high religious activity in Sweden is to reduce women’s overall risk of returning to work, in Norway the effect is to discourage full-time work in particular. Second, in both countries mothers who cohabit without marriage are more likely to choose full-time work but the effect is more pronounced for Norway. Norwegian women who married directly without prior cohabitation are less inclined to return to work at all. Clearly, Norwegian mothers differ more in marital status than Swedish women and the differences in preferences (traditional v. non-traditional) that this covariate mirrors have a larger impact on their labor market choices. Third, women who separate from their partner have a higher rate of employment re-entry, especially to full time, in Sweden, while in Norway a separation seems to discourage full-time work and encourage part-time work. For Sweden this result is as expected since single mothers are encouraged to work by being charged lower fees and being given priority to public day care. But for Norway the result, although not significant, may seem at odds with the previously found higher risk of fulltime return and lower risk of part-time return for single mothers (Rønsen 1995b). Being single when the child is born or separating or divorcing afterwards may not be quite comparable, however. Single mothers have had time to adjust to single parenthood, probably know their rights and will receive benefits from the time of birth. Returning to a part-time job will be unattractive because of the very high marginal tax rates (tax rates plus reduction in benefits) that they normally would incur in these income intervals (see the section on Data and variables). Becoming a single parent through separation after the child is born may be quite a different experience. Certainly, these mothers also have to adjust to a smaller budget, but all the same they may be better off than
132
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
Table 8.2 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with preferences and human capital variables Sweden Religious activity: High Low Marital status at birth: Cohabiting Married after cohabit. Directly married Marital status change: Separated In the same union Age at delivery Work experience Prior home-time Education: Compulsory schooling only Compulsory schooling plus 1 year 2–3 years of secondary school Post-secondary school educ. Log likelihood No. of parameters
Norway
Full-time
Part-time
Full-time
Part-time
0.91 1.
0.84 1.
0.80** 1.
0.98 1.
1.10 1. 1.04
0.95 1. 1.03
1.34*** 1. 0.89
0.96 1. 0.89
1.62 1. 0.97 1.01 0.96**
1.22 1. 1.03 0.99 0.91***
0.53 1. 0.99 1.04 0.83***
1.62 1. 1.06** 0.96 0.89***
1.
1.
1.
1.
0.93
1.47***
1.25
1.63***
1.14
0.98
1.45**
2.10***
1.49**
1.12
2.27***
2.64***
−6916.9 20
−7584.0 20
Notes: For categorical variables risks are given relative to that of the base group, indicated by the value 1. For continuous variables a value lower than 1 indicates a negative gradient, higher than 1, a positive one. *** significant at the 1 % level, ** at 5 %, * at 10 %
mothers who have been single the whole time and may have greater possibilities for not working full time. In addition, they may not yet be confronted with high marginal tax rates since it takes some time to get acquainted with and registered in the benefit system. Therefore, a part-time job may be a sensible choice. Turning to the effects of differences in accumulated human capital, we find that mothers who were older when having children tend to choose part-time work when they return to a higher extent than those who were younger: the effect is significant only for Norway. There are no significant effects of work experience. Further, in line with previous research and prior expectations, months spent at home prior to giving birth are found to exert a strong negative influence on re-entry rates of mothers in both countries. As pointed out above, mothers who stay at home prior to birth are likely to be less work committed than those who work throughout pregnancy and, in addition, they may have greater difficulties in finding a job afterwards. Possibly Swedish mothers, including those who “happened” not to work during the whole pregnancy, have a more
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
133
uniform pattern of high work commitment and close ties with the labor market than Norwegian women have, resulting in larger negative effects of prior home-time for Norway. Moreover, there are clear differences in the impact of education on the return to full-time and part-time work between the two countries. For Norwegian women the effect of more education is to increase both full-time and part-time reentry rates: the part-time re-entry rates being higher than the full-time rates at each educational level. By contrast, the impact of higher education among Swedish mothers is mainly to speed up the return to fulltime work: part-time re-entry rates are highest for those with lower levels of education. This country difference might be due to the fact that maternity leave was considerably shorter in Norway than in Sweden during the period studied and also that highly educated women preferred part-time work as a combination strategy when the child was just a few months old. In the next step we test for possible effects of public policy (see Table 8.3). The results indicate that the public sector has been more important for continuous employment of mothers in Norway than in Sweden, where only the part-time rates are affected and not significantly. Further, there are clear country differences in the effects of calendar period of birth on full-time and part-time re-entry rates: in Sweden the full-time return rates were at their highest for women who had their child in 1968–73 and then declined continuously, while the part-time re-entry rates exhibit almost an inverse u-shape with a peak in the late 1970s. In contrast, the Norwegian full-time return rates were not as high in any period, nor did they fall as much, while the part-time re-entry rates were initially extremely low but increased dramatically over periods. The low return rates for women who had a child in 1980–84 in both countries can probably be explained by the economic downturn and high unemployment at the time. Furthermore, we find that the Table 8.3 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with policy variables Sweden Religious activity: High Low Marital status at birth: Cohabiting Married after cohabit. Directly married Marital status change: Separated In the same union Age at delivery Work experience Prior home-time Education: Compulsory schooling only Compulsory schooling plus 1 year
Norway
Full-time
Part-time
Full-time
Part-time
0.91 1.
0.84 1.
0.78** 1.
0.97 1.
1.16 1. 0.95
0.95 1. 1.08
1.34*** 1. 0.85
0.87 1. 1.10
1.51 1. 1.00 0.98 0.99
1.09 1. 1.03 0.97 0.95***
0.52 1. 1.00 1.03 0.86***
1.50 1. 1.04 0.95 0.92***
1.
1.
1.
1.
0.91
1.35***
1.22
1.53**
134
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
Sweden Full-time 2–3 years of 1.11 secondary school Post-secondary 1.40* school educ. Sector of employment: Public sector 0.99 Other 1. Period of first childbirth: 1968–73 2.12*** 1974–6 1.93*** 1977–9 1.77*** 1980–84 1. 1985–8 1.49*** Status after childbirth Maternity leave 2.70*** Housekeeping 1. Log likelihood No. of parameters
Norway Part-time
Full-time
Part-time
0.93
1.44**
1.83***
1.03
2.10***
2.36***
1.06 1.
1.19** 1.
1.24** 1.
1.27 1.75*** 2.15*** 1. 1.73***
43*** 1.60*** .28* 1. .33**
0.57*** 1.16 1.03 1. 1.43***
2.65*** 1. −6831.2 32
1.36** 1.
1.30** 1. −7550.6 32
Notes: For categorical variables risks are given relative to that of the base group, indicated by the value 1. For continuous variables a value lower than 1 indicates a negative gradient, higher than 1, a positive one. *** significant at the 1 % level, ** at 5 %, * at 10 %
effect of having maternity leave is to speed up the return to work in both countries, but especially in Sweden, and it is just as important for part-time entries as for full-time entries. Because of the many extensions of the leave program, especially in Sweden, the effect of having maternity leave may have changed over the years. In addition, the variations in the length of leave entitlement are likely to have produced very different full-time and part-time exit patterns across our calendar periods of birth. We explore these two possibilities by running interactions between calendar period of birth on the one hand and on the other hand (i) being entitled to maternity leave or not and (ii) duration. The first interaction reflects the extent to which the impact of having maternity leave varies across calendar periods, and the second picks up whether the exit rates of each calendar period are proportional as is assumed in the Cox model. If not, the proportionality assumption is violated, in which case there is a duration (or time) dependency (see, e.g. Cox 1972; Blossfeld and Rohwer 1995:224ff). To check for such interaction effects we re-estimated the model presented in Table 8.3 with the interaction terms included. The results are presented in Table 8.4 and show, first, that maternity leave has had a differential impact on re-entry rates across calendar periods only for Swedish women. This makes sense as there have been fewer changes in the Norwegian leave entitlement period than in the Swedish one. We observe that the importance of having maternity leave was greater before 1980 than after, and that the impact on both full-time and part-time return rates increased throughout the 1970s. In 1980–84 the positive effect of maternity leave decreased considerably and this trend continued also after 1985 for part-time work. Maternity leave may have been
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
135
more important in the earlier periods because fewer women were entitled to leave at that time, and those who were may have had a stronger work commitment relative to those without Table 8.4 Relative risks of entry into full-time and part-time employment after first childbirth among Swedish and Norwegian women. Model with interactions with calendar period of birth Sweden Full-time Period*Status after birth: 1968–73* On 2.62*** maternity leave 1974–6 * On 4.18*** maternity leave 1977–9* On 5.75*** maternity leave 1980–84* On 1. maternity leave 1985–8* On 1.30 maternity leave Period*Duration: 1968–73*[ln(T)–ln 0.29*** (mean dur.)] 1974–6*[ln(T)–ln 0.38*** (mean dur.)] 1977–9*[ln(T)–ln 0.62 (mean dur.)] 1980–84*[ln(T)–ln 1. (mean dur.)] 1985–8*[ln(T)–ln 0.48** (mean dur.)] Log likelihood No. of parameters
Norway Part-time
Full-time
Part-time
2.15**
0.87
0.71
1.58
1.23
0.62
3.12**
0.85
0.72
1.
1.
1.
0.73
0.91
0.63
0.19***
0.52***
0.67*
0.23***
0.59***
0.59**
0.54***
0.86
0.80
1.
1.
1.
0.38
1.25
1.03
−6772.9 48
−7527.5 48
Notes: This model also includes all the covariates in Table 8.3. Only interaction effects are reported. The risks are given relative to that of 1980–84, indicated by 1. For Period*Duration, T=process time. Since [ln(T)–ln(mean duration)] is negative for process times less than mean duration, a relative risk less than 1.0 implies that the risk relative to 1980–84 is higher for process times shorter than mean duration and lower for process times longer than mean duration. *** significant at the 1 % level, ** at 5 %, * at 10 %
entitlement than in the later periods when most mothers were entitled. However, the reduced importance of maternity leave after 1980 may also reflect to some extent the particular Swedish rule, introduced in the 1980s (see Norwegian and Swedish family policies), according to which women who bore their next child within 24 or 30 months of the earlier one did not have to return to work to establish eligibility for benefits for leave with that child, reducing the positive impact of leave on re-entry rates. Second, we see that the effects on re-entry rates of the interaction between period of birth and duration are indeed significant for most calendar periods in Sweden, but only for the two early periods in Norway. The low relative risks for Sweden
136
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
in particular in the two early calendar periods indicate that re-entry rates were much higher a shorter time after giving birth (and lower for longer durations), especially to part-time work, than they were in 1980–84. Of course, this is because the leave entitlement was considerably shorter at that time than in 1980–84 (6–7 months as compared to 12 months). Also, in the late 1980s the risk of exiting to full-time work in Norway was higher for longer durations than in the early 1970s when the benefit period was only about three months as compared to four and five months later on. SUMMARY AND CONCLUSION Female employment in Scandinavia is characterized by very high employment rates among mothers of young children and high part-time rates. At the same time the Scandinavian countries have a long tradition of extensive social policies, including generous parental leave programs and other economic support to families with children, which are likely to have facilitated the combination of work and family. Recent research confirms, for example, that the right to a paid maternity leave reduces women’s career breaks in connection with childbirth and thus encourages a continuous attachment to the labor market (Rønsen and Sundström 1996). In this paper we focus on the re-entry into full-time and part-time work after the birth of the first child, and study the impact of public policies and other factors on these processes using two data-sets with almost identical designs: the 1988 Norwegian Family and Occupation Survey and the 1992 Swedish Family Survey. Norway and Sweden are culturally quite similar, and have the same set of public policies, but with country-specific variation. Swedish women have, however, a longer tradition of high female employment. This will, since our data go back as far as 1968, be reflected in the Norwegian women in our sample having more traditional working patterns. They are also more traditional when it comes to cohabitation outside marriage and they are more religious. On the whole, Norwegian mothers seem to differ more in values and preferences than Swedish mothers do, and these differences are also mirrored in their after-birth employment behavior. For example, in Norway religiously active women are less inclined to work full time after delivery, while cohabiting mothers have higher full-time return rates than mothers who are married. In Sweden there is only weak and insignificant evidence of similar associations. Further, there is a stronger positive impact of education on the re-entry rates in Norway, where higher education speeds up the return to part-time work just as much as to full-time work. In Sweden, the part-time profile is quite different from the full-time profile, and the highest part-time risk is found among mothers with lower levels of education. The parental leave programs in the two countries also have many similarities, but there are important variations in generosity and flexibility. The Swedish program has offered a longer paid leave throughout the period studied, with more options for saving leave for later usage and to take leave full time or part time. In addition, since 1980 Swedish mothers have been able to maintain the same benefit level as after the first child without going back to work if the next child is born within a certain interval. This “speed-premium” is not a feature of the Norwegian system, which requires mothers to work up their entitlement before each birth. Nevertheless, in both countries women who have the right to a paid leave resume employment faster than non-eligible women, and the part-time rates are affected just as much as the full-time rates. However, the impact is larger in Sweden than in Norway. Possibly, the longer entitlement period in Sweden has encouraged more mothers to keep in touch with the labor market. With the short entitlement in Norway during the period studied, a few women will return more quickly, but a larger number will end up outside the labor force because it is more difficult to reconcile work and motherhood when the baby is very young.
FULL-TIME OR PART-TIME WORK IN NORWAY AND SWEDEN
137
During the two decades of our study, there have been several extensions of the Swedish leave program, suggesting that exit patterns will differ across calendar periods. Likewise the importance of maternity leave may have changed over the years. Not surprisingly, our results confirm that the length of the statutory leave has shaped the exit patterns. In the early 1970s, when the leave was shorter in both countries, there was a faster return closer to birth and a slower return toward the end of the 36-month observation interval. As expected this was more pronounced for Sweden than for Norway, where changes in the leave program have been fewer and smaller. Further, the interaction effects between calendar period of birth and maternity leave are significant for Sweden only, where the positive impact of having had maternity leave increased during the 1970s, but declined sharply after 1980. This is an interesting result, which may indicate that the introduction of the particular Swedish “speed-premium” may have delayed the return to both full-time and part-time work after birth. ACKNOWLEDGMENTS We want to thank Jan Kowalski and Ylva Schullström for programming assistance, and the Swedish Research Council for the Social Sciences and the Swedish Council for Research in the Humanities and Social Sciences for financial support for the Swedish study. REFERENCES Blom, S. (1994) “Marriage and Cohabitation in a Changing Society: Experience of Norwegian Men and Women Born in 1945 and 1960,” European Journal of Population 10:143–73. Blossfeld, H.P. and Rohwer, G. (1995) Techniques of Event History Modeling, Matwah, N.J.: Lawrence Erlbaum Associates. Cox, D.R. (1972) “Regression Models and Life-Tables (with Discussion),” Journal of the Royal Statistical Society B 34:187–220. Desai, S. and Waite, L.J. (1991) “Women’s Employment during Pregnancy and after the First Birth: Occupational Characteristics and Work Commitment,” American Sociological Review 56:551–66. Ellingsaeter, A.L. (1989) “Normalization of Part-Time Work. A Study of Women’s Employment and Working Time Patterns in the 1980s,” Social and Economic Studies 71, Oslo: Statistics Norway. Ellingsaeter, A.L. and Rönsen, M. (1996) “The Dual Strategy: Motherhood and the Work Contract in Scandinavia,” European Journal of Population, 12:239–60. Eurostat (1992) Women in the European Community, Luxembourg. —— (1995) A Statistical Eye on Europe 1983–1993. EurostatYearbook 1995, Luxembourg. Even, W.E. (1987) “Career Interruptions Following Childbirth,” Journal of Labor Economics 5:255–77. Gustafsson, B. and Kjulin, U. (1993) “Public Expenditures on Child Care and the Distribution of Economic Well-being: The Case of Sweden,” in B.L.Wolfe (ed.) On the Role of Budgetary Policy during Demographic Change (Public Finance 48 Suppl.): 98–121. Hoem, B. and Hoem, J.M. (1992) “Union Dissolution in Contemporary Sweden,” in J.Trusell et al. (eds) Demographic Application of Event History, Oxford: Clarendon Press. Hoem, J.M. (1993) “Public Policy as the Fuel of Fertility,” Acta Sociologica 36:19–31. Hotz, J.V. and Miller, R.A. (1988) “An Empirical Analysis of Life Cycle Fertility and Female Labor Supply,” Econometrica 56:91–118. Joesch, J.M. (1994) “Children and the Timing of Women’s Paid Work after Childbirth,” Journal of Marriage and the Family 56:429–40.
138
MARIT RØNSEN AND MARIANNE SUNDSTRÖM
Kravdal, Ø. (1992) “The Weak Impact of Female Labour-Force Participation on Norwegian Third-Birth Rates,” European Journal of Population 8:247–63. Lesthaege, R. and Moors, G. (1995) “Is There a New Conservatism That Will Bring Back the Old Family? Ideational Trends and the Stages of Family Formation in Germany, France, Belgium and The Netherlands, 1981–1990,” in Evolution or Revolution in European Population, Plenary Vol., European Population Conference, Milan: FrancoAngeli. Petersen, T. (1995) “Analysis of Event Histories,” in G.Arminger et al. (eds) Handbook of Statistical Modeling for the Social and Behavioral Sciences, New York: Plenum Press. Rønsen, M. (1995a) Family Policies and Maternal Employment, paper presented to the 9th ESPE-meeting, Lisbon. —— (1995b) “Maternal Employment in Norway,” Discussion Papers 142, Oslo: Statistics Norway. Rønsen, M. and Sundström, M. (1996) “Maternal Employment in Scandinavia. A Comparison of the After-Birth Employment Activity of Norwegian and Swedish Women,” Journal of Population Economics, 9:267–85. Shapiro, D. and Mott, F.L. (1994) “Long-Term Employment and Earnings of Women in Relation to Employment Behavior Surrounding the First Birth,” Journal of Human Resources 29:248–75. Statistics Sweden (1992) Barnomsorgsundersökningen 1992 (The Childcare Survey) S11 SM9201, Stockholm. Sundström, M. (1991) “Sweden: Supporting Work, Family and Gender Equality,” in S.B.Kamerman and A.J.Kahn (eds) Child Care, Parental Leave, and the Under 3s, Westport, CT: Auburn House. —— (1993) “The Growth in Full-Time Work Among Swedish Women in the 1980s,” Acta Sociologica 36:139–50. Sundström, M. and Stafford, F.P. (1992) “Female Labour Force Participation, Fertility and Public Policy in Sweden,” European Journal of Population 8:199–215.
9 PUBLIC POLICY AND CHILDCARE CHOICE Seija Ilmakunnas
INTRODUCTION This paper analyzes the determinants of the choice of childcare mode and the mother’s labor force participation decision. The study is carried out in the context of a well-developed Finnish welfare state with many public policy instruments targeted towards the care of small children. Most previous studies in this area have analyzed choices of childcare mode in quite different settings, where the role of government subsidies to the childcare sector is typically weaker, the informal modes of childcare are more important and the quality of care varies substantially. A partial list includes Blau and Robins (1988), Leibowitz et al. (1988) and Hofferth and Wissoker (1992). A study that is related to a rather similar institutional setting is Gustafsson and Stafford (1992), which analyzes the effects of childcare subsidies in Sweden. As far as government subsidies to the childcare sector are concerned, Finland clearly offers an example of a generous policy mix. Lengthening of maternity leaves with earnings-related benefits and the better availability of public day-care services have been its two main elements during the last two decades (see Ilmakunnas 1993a). But perhaps the clearest token of this generosity is the income transfer called child home care allowance (HCA). This family policy innovation was launched in the late 1980s and it is paid to all families with children under age three who do not use public day-care services, but rely instead on parental or private care. Accordingly, in the case of small children all three modes of childcare (parental, public and private) have become heavily subsidized, and in theory and practice families with small children have the right to choose between a publicly provided service and a cash benefit. These kinds of extensions in childcare subsidies naturally have many effects. As more public expenditure is devoted to the childcare sector and the private cost of each mode of childcare is lowered, the well-being of families with small children is expected to rise. In addition to these quite obvious effects, one might be interested in the effects on the division of labor between men and women. For instance, the use of the home care allowance seems to be closely linked with the traditional division of labor between men and women.1 The overall welfare effects might also be interesting as people without children must be willing to support the childcare programs (see Bergstrom and Blomquist 1996). For instance, one might argue that the willingness to finance the more traditional day-care programs is greater than the willingness to finance the subsidies to parental care at home, i.e. to household production. In this paper the childcare and labor supply decisions are analyzed with the help of discrete choice models. The strongest emphasis is on the sensitivity of the choices to the level of the child home care allowance. The microdata used in the study comes from a survey that was carried out in the late 1980s in
140
PUBLIC POLICY AND CHILDCARE CHOICE
order to study how families with small children react to the day-care reforms, including the introduction of HCA. The paper proceeds as follows. First we give a short description of the subsidies to the childcare sector, and then the models to be estimated are derived. The following sections are devoted to the data and the results. Finally, our conclusions are summarized. THE INSTITUTIONAL SETTING OF CHILDCARE POLICY Our study concentrates on the choice of childcare mode in the case of small children (the children’s age range is from 11 months to two-and-a-half-years). Children younger than that are typically taken care of by their own parents, usually the mothers. The rather long maternity leave with earnings-related compensation enables the mother to stay at home during the first year after childbirth.2 As practically all mothers make use of the whole maternity leave the analysis here deals with cases where the maternity leave period has come to an end. After maternity leave there are nowadays two alternative policy options available for all families with a child under three years old: a place in the public day-care system or home care allowance (HCA).3 Those families that use private childcare services also receive HCA (or part of it) to cover the fees they pay for their private childcare arrangements. Child home care allowance (HCA) HCA consists of three different parts: basic payment (BHCA), additional means-tested payment and sibling supplement. They are all taxable income. The basic payment is received by all families entitled to the benefit. In addition to the basic payment the additional means-tested payment is paid to families whose income falls below a specified level.4 For the means-tested part there is the further precondition that one of the parents has to take care of the child, and this part cannot be used to cover the fees for private childcare. When there are two or more children under school age the family also receives the sibling supplement. The sibling supplement is 20 percent of BHCA per additional child. Besides the statutory, nationwide home care allowance system described above, some municipalities (including all the big cities) have provided their own additional home care allowance schemes. These vary remarkably from municipality to municipality (Ilmakunnas 1993b). The attractiveness of HCA is increased by the childcare leave system. Finnish labor law states that the mother or the father or both in turn can be on childcare leave from their job until the child is three years old. Rather long periods of absence from work are thus possible without the risk of losing one’s job. In spite of the increased availability of low-cost and high-quality day-care services, the majority of families with children under three years old choose the child home care allowance (see Table 9.1). Of those who choose it, only a small minority use it to cover the fees for private childcare (Mikkola 1992). The popularity of HCA has actually implied that mothers have come to stay at home longer after the child is born. This trend can be seen also in the aggregate labor force statistics. In the case of women aged 20–39, the labor force participation rates have dropped and the number of women engaged in household duties has increased (Ilmakunnas 1995).
SEIJA ILMAKUNNAS
141
Table 9.1 Recipients of home care allowance (HCA) 1985–93, whole country Number of families % of all potential recipients Max. HCAa) (FIM) Max. HCA per female earningsb) (%) 1990 1991 1992 1993 1994
81,210 85,210 92,570 95,817 95,380
64 66 71 73 73
2498 2691 3512 3524 3524
33 33 42 41 40
Source: Statistical Yearbooks of the Social Insurance Institution and Statistical Yearbook of Finland, Statistics Finland Notes: a) Basic payment+maximum means-tested payment per month (local schemes not included) b) Max. HCA per average monthly earnings of female employees, %
Publicly provided day care The other policy instrument, the public day-care system, is organized and financed mainly by local authorities. The state takes part in the funding in the form of grants to local authorities. The modes within the public day-care system are day-care centers and family day care,5 the centers being the larger part of the system. The share of children in part-time care is small (about 10 percent of all children in public day care). This corresponds quite closely to the small share of part-time work among Finnish employed women. The standards for public day care are set nationally and the quality of public day care is thus rather homogeneous and also rather high. The average children/ staff ratio is 4.2 in day-care centers and 2.8 in family day care (Ministry of Social Affairs and Health 1994). The fees for public day care are income related and cover, on the average, 14 percent of the total costs of these services. Private day care The growing role of the public day-care system implies that the market share of private day care is rather small. At the beginning of the 1990s less than 10 percent of all children under school age were in private day care operating under the supervision of the local social services (Mikkola 1992). The corresponding figure for public day care was 47 percent. The difference can be understood in the light of the high subsidy rate in the public system where the parents pay only a small fraction of the total costs. Accordingly, private day care is a more expensive option for most families. The fees for private day care are heavily influenced by the recommended fees that are set by the association for private day-care minders. As HCA (or part of it) can be used to cover the fees for private care this option has become more competitive in the case of children under three years old. THE MODEL The aim of this study is to analyze how childcare choices are made in Finnish families with small children. Attention is especially devoted to the question of how the instruments of family policy affect these decisions. The most interesting instrument in this respect is the relatively new home care allowance system as it is a rather unique form of childcare policy, also in an international perspective. Furthermore, its labor supply incentives are basically different from the more “traditional” forms of childcare policy such as earnings-related maternity benefits and public day-care services. On the other hand, HCA (or part of the system) can be considered as a voucher scheme since the families using private childcare services are
142
PUBLIC POLICY AND CHILDCARE CHOICE
entitled to the basic amount of HCA. It can thus be considered as a subsidy both for parental and for private childcare. The level of that subsidy in most cases is different in the case of parental care and private childcare however, due to taxation and to the preconditions of the means-tested part of HCA. Our model includes three possible modes of childcare. The first mode is parental care. For the sake of simplicity we assume that it is the mother who takes care of the child/children in this option. When this mode is chosen, it also implies that she is a non-participant in the labor market. If the mother chooses this option, she loses her potential net earnings in the labor market, but she receives HCA and there are no childcare fees to be paid. The economic activity of her spouse, if there is one, is assumed not to be affected by the choice of childcare mode. The other two modes are public and private day care, and they both imply that the mother participates in the labor market. As we are analyzing children who are from one to two years old, this assumption is reasonable. The educational motivations for childcare play a minor role in the case of small children. In their case the need for childcare arises from the labor force participation of the mother. When either the public or the private day-care option is chosen, the mother is assumed to enter the labor market, earn labor income and pay taxes for that income. On the other hand, she has to pay for the public or private day-care services the family purchases in the market. The price she has to pay naturally varies from one mode to the other. In the case of private day-care services she also receives the voucher, i.e. the basic amount of HCA. The model is based on some crucial assumptions. First, we abstract from the time dimension of the care. Thus, there is no distinction between full-time and part-time care. All market care is assumed to be provided on a full-time basis. This is motivated by the actual small share of part-time care in the institutional setting in question. Analogously, the mother’s labor market participation takes place on a full-time basis. Second, we assume that there is no rationing affecting the availability of market day care. In the case of public day care this is motivated by a law that guarantees the parents of small children the right to have a public day-care place for their child. Admittedly, in the case of the private option the availability assumption is not as well motivated.6 Third, the quality aspect is not accounted for. This is motivated by the dominant role of public day care and its national standards. Also in the case of private day care the quality is relatively homogeneous due to the fact that it operates under the supervision of the local social service authorities. The options and their characteristics are summarized in Table 9.2. We assume that each mother evaluates the utility of each available childcare mode and then chooses the arrangement with the highest utility.7 The discrete choice models to be used are based on the maximization of a random utility Table 9.2 The childcare modes and their economic characteristics
Model 1 maternal care Mode 2 public day care Mode 3 private
Earned income (net of taxes)
Day-care costs
Cash benefits (net of taxes)
W1=0
C1=0
Net home care allowance B1
Net labor income W2=W
Public day care fees C2
B2=0
Net labor
Private day care
Net home care
SEIJA ILMAKUNNAS
day care
Earned income (net of taxes)
Day-care costs
Cash benefits (net of taxes)
income W3=W
fees C3
allowance B3
143
function where the utility associated with each alternative is the sum of a deterministic component and an unobserved random error specific to that alternative.8 We start by defining a model where the deterministic part of the utility is a linear expression and it depends only upon choice-specific attributes that may vary across individuals: the cost of that choice, the level of wage income and the level of HCA implied by that choice. The random components e have independent and identical type I extreme value distributions. In the following expressions i refers to the individual and n to the alternative (n=1, 2, 3). (9.1) The individual will choose the mode with the highest utility. Thus, individual i chooses mode k if (9.2) This formulation implies the conditional logit model CLGT where the choice probabilities are (see e.g. Greene 1991) (9.3) In the conditional logit model the impact of the explanatory variables on the choice probabilities derives from the difference in their values across the modes. The individual-specific (or family-specific) characteristics with no variation across the modes have no impact on choice probabilities. These characteristics are for example the age of the child, the father’s income, housing costs, etc. As it may well be the case that these family-specific characteristics do influence the childcare choices, we consider also alternative specifications for the choice probabilities. The other specifications we consider are: the multinomial logit model (MNLT), the mixed model (with elements from both CLGT and MNLT models) and the nested logit model. In the multinomial logit model all the explanatory variables (Xi) are characteristics of the individual or the family. As they are constant across the alternatives, the way they affect the choice probabilities is by having a different impact on the various alternatives. In the case of the multinomial logit model, the corresponding choice probabilities are (9.4) Regressor variables with the same value for all alternatives call for standardization of their coefficients. We choose the mode (1) (maternal care) as the standard of comparison and thus set the vector α1 equal to zero. In the MNLT model we include in Xi also the variables that we used as explanatory variables in the CLGT model: mother’s potential net wage income, day-care costs and cash benefits (HCA). Since the values of the explanatory variables in the MNLT model cannot vary across the alternatives, a choice has to be made. We have chosen the level of HCA to be the one obtained in the case of maternal care (i.e. B1 in Table 9.2). This example indicates the problems that are faced when the multinomial logit model is used in analyzing government policy that affects the alternative options in different ways (Hoffman and Duncan 1988). In the mixed model the utility of each mode n for individual i is a linear function of both the characteristics of the mode and the characteristics of the individual and her family (vector Zi,which is a subvector of Xi). This gives the following choice probabilities (9.5)
144
PUBLIC POLICY AND CHILDCARE CHOICE
Finally, the nested logit model is used to test whether the above models are proper descriptions of the choice problem in question. As an alternative we consider the model where we let public and private day care be closer substitutes for each other than for maternal care. In this case the choice probabilities become (see e.g. Amemiya 1985) (9.6) The estimated σ provides us with a specification test. When σ equals 1, the nested logit model reduces to the mixed model above. In this case all the error terms are uncorrelated and all the childcare modes are equally close substitutes for one another. The likelihood functions for all the models presented in this section are of the type (9.7)
THE DATA The microdata used in this study were obtained from a survey that was carried out in 1989 in Helsinki. The aim of the survey was to study how families with small children were reacting to the topical day-care reforms. Those reforms included the rapidly increased availability of day-care services for children under three years old9 and the new instrument, the home care allowance, which was introduced after the mid-1980s. Parents who were not using their right to the public service were compensated for arranging the care themselves. In the survey families the youngest child is under two-and-a-half years old. We restrict ourselves here to the subsample where the youngest child is older than ten months so that the choice between different childcare modes is relevant in the family. The analysis deals with the families whose maternity leave period has come to an end and for whom the problem of choosing a childcare mode must be faced. The sample size is 1135. The analysis is based on information about desired childcare modes. The sample was randomly divided into three subgroups. Each subgroup stated its desired day-care and labor supply choices conditional on a different (hypothetical) level of the home care allowance. The different levels of the monthly allowance were FIM 3000, FIM 3500 and FIM 4000 (approximately US$ 700–930). These figures include the nationwide statutory allowance (the basic amount) and also the additional allowance provided by the local authority, the city of Helsinki. When judging whether the information about desired childcare modes is likely to correspond to actual responses, the following issues are important. The decision concerning the childcare mode was of current interest in the families included in the survey. The possible options were available and respondents were given information about the costs and benefits of the three modes they could choose between. Further, labor demand was very strong and lack of work opportunities was not a constraint on labor market entry.10 The desired childcare mode was parental care in 629 cases (55 percent), public day care in 339 cases (30 percent) and private day care in 167 cases (15 percent). The survey provided information about the characteristics of the mother, the father and the children (see the list of the variables in Table 9.3). When having to make a decision, one of the crucial explanatory variables
SEIJA ILMAKUNNAS
145
is the value of the mother’s time in the labor market. The survey contains information about the labor income of working mothers (and fathers) and also self-reported potential earnings for those mothers who did not participate in the labor market. All income and home care allowance items used in the estimations are net of taxes.11 The costs of public day care are obtained with the help of the rules by which families pay for the service. The fees depend on the age of the child, the income of the family and the number of family members. If more than one child needs public day-care services, the fees are reduced. The costs for private day care are based on the recommended fees that are set by the association for private day-care minders.12 These fees do not depend on family income and there are no reductions for siblings. As all the families might not pay fees according to these recommended fees, there is some error in the price variable for private day care; regardless of this, these recommendations have set the guidelines for the price level in the private day-care market. They have been widely known as they have been published in the newspapers, among other places. RESULTS We will present here the results from the mixed model which includes both the family characteristics and the choice-specific attributes as explanatory variables. The conditional logit model, the multinomial logit model and the Table 9.3 Description of the data Variable
Mean
St. dev.
Wage income (net monthly wage income/1000) Home care allowance (net/1000) (in the case of parental care) Home care allowance (net/1000) (in the case of private care) Costs for public day care (total monthly costs/1000) Costs for private day care (total monthly costs/1000) Father’s income (net monthly wage income/1000) Housing costs (monthly costs/1000, gross) Child’s age (age of the youngest child in months/10) Mother’s age (age of the mother in years/10) Number of children (children under school age) Single parent (=1 for a lone mother, otherwise 0) Dummies for the mother’s educational level: (reference category is lower level of basic education)
4.723
1.449
3.149
0.379
1.770
0.222
0.893
0.655
2.277
0.937
5.724
2.819
3.437
3.677
1.636
0.494
3.198
0.613
1.488
0.613
0.075
146
PUBLIC POLICY AND CHILDCARE CHOICE
Variable
Mean
level 3 =upper level of basic education level 4 =lower+upper level of secondary education level 5 =higher education Dummies for the mother’s work experience: (reference category is 5 years or less) level 2 =6–10 years level 3 =11–15 years level 4 =15+
0.178
St. dev.
0.324 0.217
0.403 0.234 0.127
Note: Number of observations 1135
nested model were also estimated. Since the results did not differ substantially from those presented in Table 9.4 they are not included here.13 The estimation results also implied that the more complicated nested model specification (9.6) can be rejected in favor of the simpler mixed model (9.5). In the nested model the estimated value of σ is 0.917 and the asymptotic t-value for testing the hypothesis that σ=1 is −1.49. This means that σ is not significantly different from 1. The predictive power of the nested model was also weaker than that of the mixed model. In the conditional logit part the cost of care variable has been omitted. This is due to the problems caused by strong correlations between the cost variable and three other variables, namely the mother’s and the father’s income and the number of children. As public day-care fees are closely related to the incomes of the parents, their separate effect is difficult to detect reliably. Also the separate effect of the number of children and the cost variable is problematic as the cost variable in private care is simply the multiple of the fee-perchild variable. We also estimated a version of the mixed model where the price variable was included in the CGLT-part, but the number of children was omitted.14 The other parameter estimates were quite similar in these two variants of the mixed model, and the coefficient for the price variable was negative and statistically significant (−0.188, with standard error 0.042). The estimated coefficient for the mother’s wage income is positive and significant in Table 9.4. The same holds true for the other explanatory variable, HCA, in the conditional logit part. In the multinomial logit part the coefficients reveal the effects of the explanatory variables on the log-odds ratios ln(P2/P1) and ln (P3/P1). Most variables connected to the structure of the family have the expected signs. The age of the child has positive signs, thus indicating that maternal care is preferred to a greater extent in the case of the youngest children. The number of children has the expected negative signs. Furthermore, public day care becomes less likely compared to home care when the father’s income rises. The opposite is true in the case of private care. Maternal care becomes less likely when housing costs increase. We will study some of the effects more closely with the help of a “base case” (or average) mother and family. We have chosen to use the mean values for the continuous explanatory variables (see Table 9.3). For instance, the “base case” mother has a husband whose net monthly labor income is FIM 5720 (approximately US$ 1330) and her own labour income is FIM 4720 (US$ 1100). The corresponding gross
SEIJA ILMAKUNNAS
147
incomes are FIM 8640 and FIM 6920. The only child in this family is 16 months old. Furthermore, the educational attainment of the mother is secondary education and her work experience is 6–10 years. Tables 9.5 and 9.6 analyze the choice probabilities in this family. We let one characteristic in turn change and calculate the corresponding choice probabilities after each change. Table 9.5 indicates a strong relationship between the age of the youngest child and the desired childcare mode. When the child is 12 months old, the family is likely to prefer maternal care. When the child is two years old the mother is as likely to stay at home as to enter the labor market (and to choose one of the nonparental day-care options). The mother of a three-year-old child is clearly more likely to work outside the home than to stay at home. Table 9.4 The mixed model Mixed model (CGLT+ MNLT) Variable
at home
CLGT-part Wage income Home care allowance MNL-part Constant Child’s age Mother’s age Single parent Education level 3 level 4 level 5 Work experience level 2 level 3 level 4
public
private
0.524** (0.038)
0.524** (0.038) 0.427** (0.051)
−0.556** (0.064) 0.493** (0.064) −0.616** (0.052) 0.516** (0.053)
−2.732** (0.056) 0.125** (0.059) −0.581** (0.063) 1.668** (0.064)
−0.001 (0.087) −0.018 (0.087) 0.465** (0.061)
0.017 (0.070) 0.219** (0.061) 1.172** (0.062)
0.327** (0.059) 0.484** (0.060) 0.886** (0.062)
0.223** (0.061) 0.304** (0.062) 0.757** (0.063)
0.427** (0.051)
148
PUBLIC POLICY AND CHILDCARE CHOICE
Mixed model (CGLT+ MNLT) Variable
at home
Father’s income Housing costs Number of children (0–7 years)
public
private
−0.115** (0.025) 0.115** (0.028) −0.152** (0.056)
0.119** (0.028) 0.116** (0.029) −0.540** (0.059)
Notes: standard errors appear in parentheses. Mean log likelihood −0.853 Number of correct predictions 675 (59.5%)
Although the day-care costs were not included in the model, the number of children is also a noticeable factor in the childcare choices. When there are several children, the maternal care option is chosen to a greater extent even if the public sector greatly subsidizes the use of both public and private day care. The effect of the housing costs can be understood in the light of the important role they play in the finances of young couples in Finland. Table 9.5 The changes in estimated choice probabilities: a “base case” analysis Probabilities Maternal care
Public care
Private care
“Base case” Age of the child varies: 12 months 24 months 36 months No. of children varies: one sibling 2+siblings Housing costs vary: FIM 0 FIM 1000 FIM 2000 FIM 3000 FIM 4000 FIM 5000
0.582
0.290
0.128
0.621 0.508 0.385
0.249 0.369 0.506
0.130 0.123 0.109
0.642 0.693
0.275 0.254
0.083 0.053
0.674 0.648 0.621 0.594 0.566 0.537
0.226 0.244 0.262 0.281 0.301 0.320
0.100 0.108 0.116 0.125 0.133 0.142
Table 9.6 The effects of the mother’s wage and home care allowance on the choice probabilities (a “base case” analysis) Probabilities Maternal care
Public care
Private care
“Base case”
0.582
0.290
0.128
SEIJA ILMAKUNNAS
149
Probabilities Maternal care
Public care
Private care
varies:a
Mother’s gross wage income FIM 3000 FIM 4000 FIM 5000 FIM 6000 FIM 7000 FIM 8000 FIM 9000 FIM 10000 Home care allowance varies:b FIM 1000 FIM 2000 FIM 3000 FIM 4000 FIM 5000
0.806 0.744 0.690 0.631 0.561 0.503 0.439 0.376
0.122 0.160 0.205 0.250 0.297 0.346 0.391 0.434
0.073 0.096 0.105 0.119 0.142 0.151 0.170 0.189
0.384 0.452 0.521 0.574 0.618
0.498 0.421 0.347 0.287 0.236
0.118 0.127 0.132 0.139 0.145
Notes: a The home care allowance obtained in the case of private day care also changes as the mother’s wage income changes. This is due to the effect of taxation b These figures refer to the gross basic allowance paid in the case of maternal care. The allowance in the case of private day care is calculated so that it is in line with the amount paid in the case of maternal care
As the mother’s time becomes more highly valued in the labor market, she is less likely to stay at home when her children are small (Table 9.6). The probability of maternal care is twice as high for a mother whose earnings ability in the labor market is FIM 4000 as for a mother who earns FIM 10,000. The effect of the amount of the home care allowance also indicates that economic incentives play a distinct role in the childcare mode decisions. An increase in HCA increases the probability that the family chooses this policy instrument instead of public day care. An increase in the level of HCA is especially reflected in an increase in the probability that the families use it as a “maternal wage” (compensation for the formerly unpaid household work). An increase in HCA has only a relatively modest effect on the use of private day care. Let us finally try to see how a variation in the generosity of HCA affects the whole sample. The levels of HCA in Table 9.7 refer to the gross basic allowance paid in the case of maternal care. Additional meanstested payments and/or sibling payments are added if the mother is entitled to them. The tax rules are also applied. The allowance in the case of private day care is calculated so that it is in line with the amount paid in the case of maternal care. The figures presented in Table 9.7 are the sample means of the probabilities in question. Table 9.7 The effects of home care allowance on the choice probabilities of the whole sample Mean probabilities in the whole sample Maternal care Home care FIM 0
Public care
Private care
0.346
0.528
allowance:a 0.125
150
PUBLIC POLICY AND CHILDCARE CHOICE
Mean probabilities in the whole sample Maternal care
Public care
Private care
FIM 500 FIM 1000 FIM 1500 FIM 2000 FIM 2500 FIM 3000 FIM 3500 FIM 4000 FIM 4500
0.380 0.411 0.441 0.472 0.500 0.527 0.551 0.573 0.594
0.490 0.455 0.422 0.389 0.358 0.329 0.303 0.279 0.258
0.129 0.134 0.137 0.140 0.142 0.144 0.146 0.147 0.148
Note: aThese figures refer to the gross basic allowance paid in the case of maternal care. The allowance in the case of private day care is calculated so that it is in line with the amount paid in the case of maternal care
Our simulation results imply that the popularity of parental care increases quite linearly as the level of the child home care allowance is raised. Conversely the public day-care services are decreasingly popular, while the effects on the private day-care option are relatively small. These results can also be compared with the actual changes that have taken place since the introduction of HCA. In 1991 when the new legislation had been in force for a second year, 74 percent15 of the families had chosen child home care allowance in Helsinki (Statistical Yearbook of the City of Helsinki 1991). The level of HCA comparable to Table 9.7 was FIM 3300 in 1991. The prediction obtained from the simulation results is 69.7 percent (based on the sum of the mean probabilities for parental care and private care). Thus, the actual choices of the families are only slightly underestimated by our model. CONCLUSIONS We have analyzed the choice of childcare mode in Finland, where new instruments of childcare policy were recently introduced. As a result of the new, rather generous policy mix, families with small children (under age three) have the right to choose between public day care and cash. The cash option is the child home care allowance (HCA). Most families have chosen the cash option and parental care of small children has increased. In practice these changes have had the biggest effects on the activities of mothers, as over 90 percent of the persons who receive child home care allowance are women. This trend has also implied that the number of women (aged 20–39) engaged full time in household duties has increased. But not all the families and mothers are equally likely to choose child home care allowance. First, the results imply that there exists a rather strong relationship between the mother’s potential earnings and the choice of childcare mode. The lower the potential wage, the higher the probability that the mother stays at home when the child is small. The father’s income is positively related to the probability of private childcare, but negatively to the probability of public day care. The results are quite robust as far as the choice of the modelling technique is concerned. The results obtained here also indicate that the level of the home care allowance is an important factor behind these decisions. The simulation results imply that the popularity of parental care increases quite
SEIJA ILMAKUNNAS
151
linearly as the level of the child home care allowance is raised. The predictions obtained from our model correspond relatively well to actual choices that families have made. REFERENCES Amemiya, T. (1985) Advanced Econometrics, Oxford: Blackwell. Bergstrom, T. and Blomquist, S. (1996) “The Political Economy of Subsidized Day Care,” European Journal of Political Economy 12:443–57. Blau, D.M. and Robins, P.K. (1988) “Child-Care Costs and Family Labour Supply,” Review of Economics and Statistics XLL, 3:374–81. Greene, W.H. (1991) Econometric Analysis, New York: Macmillan. Gustafsson, S. and Stafford, F. (1992) “Child Care Subsidies and Labour Supply in Sweden,” Journal of Human Resources 27, 1:204–30. Hofferth, S.L. and Wissoker, D.A. (1992) “Price, Quality, and Income in Child Care Choice,” Journal of Human Resources 27, 1:71–111. Hoffman, S.D. and Duncan, G.J. (1988) “Multinomial and Conditional Logit Discrete-Choice Models in Demography,” Demography 25, 3:415–27. Ilmakunnas, S. (1993a) “The Public Day Care System in Transition,” in Shaping Structural Change in Finland, The Role of Women, Ministry of Social Affairs and Health, Equality Publications 2/1993. —— (1993b) The Child Home Care Allowance: Facts and Visions, Studies 1993:11, The City of Helsinki, Information Management Centre (in Finnish). —— (1995) “Women and the European Employment Rate: The Causes and Consequences of Variations in Female Activity and Employment Patterns in Finland,” Report for the European Commission, Working Paper, EC Network on the Situation of Women in the Labour Market, Manchester: UMIST. Leibowitz, A., Waite, L.J. and Witsberger (1988) “Child Care for Preschoolers: Differences by Child’s Age,” Demography 25, 2:205–19. Mikkola, M. (1992) “Finland: Supporting Parental Choice,” in S.B.Kamerman and Alfred J.Kahn (eds) Child Care, Parental Leave, and the Under 3s: Policy Innovation in Europe, New York: Auburn House. Ministry of Social Affairs and Health (1994) Social Security in Finland 1992, Social Security 1994:2B. Pudney, S. (1989) Modelling Individual Choice, Oxford: Basil Blackwell. Statistical Yearbook of Finland 1991–1995, Statistics Finland. Statistical Yearbook of the City of Helsinki 1991, The City of Helsinki, Information Management Centre. Statistical Yearbook of the Social Insurance Institution (Finland) 1990–1994. Tuominen, M. (1991) Who Chooses Child Home Care Allowance?, Studies 1991:9, The City of Helsinki, Information Management Centre (in Finnish).
NOTES 1 According to survey information and statistical data, over 90 percent of the persons who receive child home care allowances are women (Tuominen 1991 and Statistical Survey of the Social Insurance Institution). 2 Actually the leave consists of two parts, maternity leave and parental leave. The former can be used only by the mother (about four months) but the latter can be used by either parent (about six months). For mothers who do not qualify for the earnings-related compensation, there is a so-called basic maternity allowance. Both constitute taxable income. 3 The right of parents to choose between public day-care services and HCA came into force in 1990. HCA was gradually introduced in the late 1980s, aimed initially at families with several small children.
152
PUBLIC POLICY AND CHILDCARE CHOICE
4 The maximum amount of the additional means-tested payment is 80 percent of the basic payment (BHCA). It is paid to families with low income. The additional payment decreases as the income rises, and eventually goes down to zero. 5 In the case of family day care, the childminder is contracted by the local authority to look after children in addition to her own children. 6 The increased availability of public day care has shifted the demand from private to public day care. The freed capacity has also implied better availability of private day care for those families who prefer that mode. 7 The theoretical model is in some respects similar to that in Hofferth and Wissoker (1992). On the other hand, they assumed maternal labor force participation as given and concentrated on the choice of childcare arrangement among working mothers. Their analysis of the effects of the quality of the care as measured by the child/staff ratios is not relevant in this paper. 8 A description of the random utility models of this type is found for example in Pudney (1989). 9 Local authorities had to increase the day-care capacity for small children since from the beginning of 1990 parents have a legal right to obtain a day-care place for their children aged less than three years. 10 The unemployment rate was 1.2 percent in Helsinki in 1989. The figure for the whole country was 3.3 percent. Since then the situation has changed dramatically: the unemployment rate was 18.4 percent in 1994. The assumption that job opportunities are always available for mothers of small children is not a correct one in the new situation of the mid-1990s. 11 There is separate taxation of the spouses. The relatively complicated deduction system is taken into account as far as the standard deductions made by tax authorities are concerned. Lack of information on, for instance, expenses incurred on traveling to work and medical expenses has prevented a more accurate treatment of the deductions. 12 The recommended fee per child during the survey period was FIM 1530. It can be compared to the highest fee in the public day-care system which was at that time FIM 665. The competitiveness of private care is substantially increased by the home care allowance which the users of private childcare obtain. 13 The estimation results for the CLGT, MNLT and the nested model are available upon request from the author (Email:
[email protected]). 14 The results are available upon request. 15 At the level of the whole country, the share of the families that had chosen child home care allowance was somewhat smaller, 66 percent (Table 9.1). This is explained by the fact that the nationwide average level of HCA was lower than the one in Helsinki. Due to variation in the local policies in this respect, the nationwide average level of HCA is difficult to determine.
10 TAXATION AND THE MARKET FOR DOMESTIC SERVICES Anne-Marie Pålsson
INTRODUCTION The analysis of household production in the Becker tradition has focused on the choice between inputs of market goods and inputs of the personal time of household members. Commodities produced by combining market goods with personal time have been assumed to generate utility. However, in this analysis not enough attention has been paid to the possible substitution between one’s own domestic labor and employed domestic labor, or in other words, between own time and hired time, for producing these commodities. In this chapter it is suggested that reducing the tax wedges in the domestic sector may increase economic efficiency and thereby total output in the economy. This comes about since the production possibility set will be enlarged. Increased production is achieved simply by reallocating the labor input in the economy. Households with relatively high market productivity will substitute their own domestic work for employed domestic labor and as a consequence increase their supply of labor to the market. At the same time, households with relatively low market productivity will withdraw their labor from the non-domestic sector and instead increase their supply of labor to the domestic sector. There are essentially two ways to eliminate or reduce the tax wedges. The first method is that expenditures for hired labor can be subsidized and the second method is that expenditures for domestic service can be made tax deductible in combination with transferring social security benefits from the buyer of domestic services to the seller of such services. The disadvantage with the first method is that with progressive tax rates, households with relatively low wage rates will consume too much service, whereas households with relatively high wage rates will consume less than the amount that is optimal from an efficiency point of view. This problem is eliminated by using the second method instead, i.e. tax deductibility. The risk that the proposed tax system will be used for arbitrage purposes is also negligible as it is self-controlling when correctly implemented. Allowing tax deductions has many advantages. It contributes to stimulating an expansion of a sector for domestic services which could possibly decrease unemployment. The suggested change is also desirable from a distribution point of view since the households with the lowest market productivity can thereby enter a new market—the domestic one—which offers a higher wage rate than the market for non-domestic labor does. One may also expect it to be positive from a gender perspective since the vast majority of household work is still carried out by women and since domestic work most often acts as a binding restriction on female labor supply.
154
TAXATION AND DOMESTIC SERVICES
This chapter, however, focuses on the economic efficiency aspects. It is structured in the following way. First the range and structure of existing domestic work is illuminated. Thereafter the reasons why a market for domestic services has not yet been developed are discussed. Not surprisingly the attention here is focused on the role of tax wedges and how these prevent households from specialization and exchange. We then develop a simple model based on utility-maximizing households with three arguments in their utility function, market goods, domestic output and leisure. The model distinguishes between three groups of households depending on which position they take on the market for domestic services: as net buyers, as net sellers or as non-participants. We use the model to analyze the equilibrium conditions for each of these three types of households separately. Based on this model the aggregate supply and demand functions for domestic services are derived and the impact on equilibrium prices and the size of the market for domestic services of changing the tax rates are analyzed. The next section looks at ways to create the necessary legal framework to stimulate an expansion of a market for domestic services. Finally, we also discuss the potential gains of the proposed change in the tax system on employment, economic efficiency and the government budget. HOUSEHOLD TIME ALLOCATION Table 10.1a and Table l0.1b present data on the time allocation of Swedish men and women in the early 1990s. Table 10.2 provides information about the distribution of household work to different activities. One would expect that the bulk of time would be used for childcare and care for the elderly, but this is not the case. Of the 26 hours per week spent on household work, only slightly more than four hours on the average are spent on care activities. This figure underestimates the true figure, however, as joint production may be common. But even if this is the case, one must be surprised by the small amount of time actually spent on care, particularly on care of children, and by the large amount of time devoted to cleaning, shopping, preparing dinner and other meals, i.e. activities which could be substituted by market-provided services. Based on the figures in Tables 10.1 and 10.2, three observations can be Table 10.1a Time allocation for men, 1990–1, average hours and minutes per week
20–24, living with their parents 20–24, single 20–24, married or cohabiting 25–44, married or cohabiting, childless 25–44, single, childless Married and cohabiting parents (all ages) with small childrenb Single parents (all ages) with small children Married or cohabiting parents (all ages) with childrenc Single parents (all ages) with children 45–64, married or cohabiting, childless 45–64, single, childless All men 20–64
Market worka
Household work
Total work
38.26 31.31 40.24 41.58 39.11 44.39 − 46.53 − 38.31 33.40 41.06
10.56 13.39 14.13 19.55 15.39 25.53 − 20.00 − 20.55 20.05 20.09
49.22 45.10 54.37 61.53 54.50 70.32 − 66.53 − 59.26 53.45 61.15
ANNE-MARIE PÅLSSON
155
Table 10.1b Time allocation for women, 1990–1, average hours and minutes per week
20–24, living with their parents 20–24, single 20–24, married or cohabiting 25–44, married or cohabiting, childless 25–44, single, childless Married or cohabiting parents (all ages) with small childrenb Single parents (all ages) with small children Married or cohabiting parents (all ages) with childrenc Single parents (all ages) with children 45–64, married or cohabiting, childless 45–64, single, childless All women 20–64
Market worka
Household work
Total work
− 25.55 24.55 33.06 34.44 19.05 24.53 32.52 34.28 25.11 29.56 27.16
− 15.44 21.44 24.46 19.58 49.43 39.44 33.00 30.43 32.27 27.27 33.17
− 41.39 46.39 57.52 55.42 68.48 64.37 65.52 65.11 57.38 57.23 60.33
Source: “Living Conditions,” SCB (Statistics Sweden), 1990/91 Notes: a Including time spent on traveling to work b Small children are defined as children 0–7 years of age c Children are defined as children 7–1 8 years of age
made. First, domestic work is extensive. Each adult person in Swedish households devotes on the average more than 26 hours a week to such work. This corresponds, when added together, to 3.7 million full-time jobs which exceeds both the total number of full-time jobs in the private sector and the total number of fulltime jobs in the public sector and amounts to 80 percent of the total number of full-time jobs in both sectors together. Table 10.2 Time allocation in household work, men and women 20–64 years, Sweden 1990, average hours and minutes per week Activity
Men
Women
All
Cooking, laundry, cleaning, dishwashing Maintenance Care of children Care of adults Purchase of household goods and services Other household work Travel related to household work Total
6.11 5.01 1.59 0.56 2.13 0.42 2.41 19.50
18.26 2.34 5.08 0.56 3.15 0.35 3.09 33.57
12.08 3.51 3.30 0.56 2.41 0.42 2.55 26.40
Source: SCB (Statistics Sweden), 1990–1
Second, and more importantly, total work effort is enormous. When domestic work is added to market work, the average workload is close to 70 hours for all men and women with children. Hence a discussion of labor supply should not be limited to the supply of market labor as this seriously underestimates the total work effort of households.
156
TAXATION AND DOMESTIC SERVICES
Third, household work is not evenly distributed among households, between the sexes or over the life cycle. Not surprisingly it is most extensive among households with children and women still do the majority of it. They spend on the average 33 hours a week as compared to men who spend only 19 hours a week on household work. As the benefits provided by social security in Sweden are mainly related to market work, women are discriminated. By cleaning, cooking, doing the laundry, etc. which for them amounts to almost half a full-time job, they lose not only the wage they could have earned by using their time for market work but also social security benefits. As household work is extensive, it is important that household resources are used efficiently. The principles of how to do so, the “household production theory”, was developed independently by Becker (1965) and Lancaster (1966). Both of them explicitly distinguish between households as a production and as a consumption unit, in contrast with the traditional neoclassical theory, which tends to draw a clear distinction between the theory of production and the theory of consumption. Both also stressed the value of time for household choice. In short, men and women should allocate their labor to market activities and household activities according to their comparative advantages in order for the household to enjoy the greatest possible utility. Utility is assumed to be derived from consuming commodities, produced by combining market goods and personal domestic labor, and from leisure. The traditional household production theory stresses the substitution between one’s own time, i.e. domestic work, and market goods and recognizes the possible gains from specialization within households in market work and domestic or household work, respectively, provided that the single household consists of more than one person.1 The possible substitution between one’s own domestic labor and employed domestic labor has not been given enough attention, however. This kind of substitution is strongly hindered by the prevailing tax system in combination with high average tax rates. Consequently, in practice one’s own domestic labor often cannot be substituted for by hired help. This limitation is crucial and puts serious restrictions on the household production function and hinders households from allocating their time in an efficient way. This restriction is particularly severe for households where both spouses enjoy a relatively high, or a relatively low, wage rate. In the first case one of the spouses has to withdraw valuable labor supply from the market in order to produce household commodities. In the second case the spouses are confined to relatively low-paid market work instead of supplying labor to the domestic sector where it is valued more highly. TAX WEDGES AND TIME ALLOCATION: AN EXAMPLE The theory of household production is rather intuitive and based on the same principles as the theory of international trade. Just as countries should specialize their production according to their comparative advantages and trade, so should men and women specialize according to their comparative advantages and trade. Furthermore, just as tariffs make trade and specialization between countries more difficult, taxes severely hinder trade between households but, and more importantly, not trade within households, i.e. between household members.2 Hence, in general a low tax rate encourages specialization between different households just as low tariffs encourage international trade. The impact of taxes on household time allocation can be illustrated by the following example. Let us consider a person who wishes to change the front door of his house. He is confronted with four possible solutions. (1) He could legally employ a carpenter, i.e. he pays the relevant employment fees and the carpenter pays taxes on the income received. (2) He could legally buy the service from the carpenter. The
ANNE-MARIE PÅLSSON
157
sole difference between (1) and (2) is that in (2), value added tax is levied also. (3) He could pay the carpenter under the table. (4) He could do the job himself. In order to make the analysis very simple, it is assumed that the person in question works as a teacher with a monthly salary of SEK 18,000 and more importantly, that the monthly income of the carpenter also is SEK 18,000. The income in both cases is based on full-time work. It is also assumed that the carpenter is more efficient in changing the door. It takes him two hours while it takes the teacher four hours, but the final result is assumed to be the same. Hence this example is based on assumptions that allow for profitable trade and specialization between two persons who are equally productive in the non-domestic sector (they have the same market wage) but have different productivity in the domestic sector (it takes the carpenter two hours and the teacher four hours to change the door). Table 10.3 indicates the time input required to be allocated to this task in the four different choices. The number of hours in solutions (1), (2) and (3) reflects the number of hours needed for the teacher to work on the market in order to earn an income high enough to pay the carpenter. All figures are based on the assumption that the marginal tax rate is 50 percent, the employment tax (contributions to the social security system) is 40 percent, and the value added tax is 25 percent. From the figures in Table 10.3 it is obvious that a rational person, given this set of assumptions, would never choose alternatives (1) or (2), because they require (ceteris paribus) a larger work effort than alternatives (3) and (4). Table 10.3 The time input (number of hours) for the teacher and the carpenter and the tax income generated, SEK; four different solutions Solution
Teacher hours
Carpenter hours
Taxes SEK
(1) employ legally (2) buy service legally (3) buy service illegally (4) do-it-yourself
5.6 7.0 2 4
2 2 2 0
684 880 180 0
It is however important to point out that (1) or (2) could be superior to (4) but never to (3).3 This happens if either the differences in market productivity or the domestic productivity are large enough. Formally this can be expressed in the following way. Buying the service legally is preferable to “do-it-yourself” if (10.1) or (10.2) where the following abbreviations have been used: wb the buyer’s market wage rate=buyer’s non-domestic productivity= ws the seller’s market wage rate=seller’s non-domestic productivity= the seller’s productivity (number of hours per unit produced) in household work=1/ the buyer’s productivity (number of hours per unit produced) in household work=1/ T tax wedge=
158
TAXATION AND DOMESTIC SERVICES
where te employment tax tv value added tax tl marginal income tax rate Hence if the difference in productivity in household work and/or in the market is large enough, then it is possible for a household service sector to be established. The magnitude of this sector will however be limited to special tasks which require special skills. The size of a market for more common domestic services will be negligible for several reasons. First and most important, the legal solution will always he inferior in economic terms to the illegal one, provided that the character of the service is independent of the legal nature and that the per sons involved do not object to the illegal character of the exchange. Second, because the wage differential necessary to create the market is large, in this example 3.5, not many people could engage in this exchange. Swedish data show that 95 percent of all employed persons have a monthly wage rate in the range of SEK 10,000–20,000. The picture looks even worse when we also take into account the fact that most families consist of two adults. Given equal productivity of the two adults in the domestic sector, the relevant comparison is between the family member who earns the lower market wage and the wage rate for the person on the market who is to perform the task in question. Therefore, even if the wage differential could be substantial between the family member with the higher wage and the seller of the service on the market, this might not be the case for the family member with the lower wage. More formally, the following ratios should be compared: and and T where superscript bm denotes the male buyer and bf the female buyer of domestic services. Consequently, three parties are involved in this potential transaction: comparing the above ratios two by two, we can conclude that the party with the lower ratio should perform the domestic service. As women most often have the lower non-domestic market productivity of the two spouses, they will specialize in domestic production, which is confirmed by Table 10.1 and Table 10.2.4 TAX WEDGES AND TIME ALLOCATION: A MODEL Assumptions In order to analyze the impact of tax wedges on household time allocation more formally, we shall now consider the following simple model. Let the representative and rational individual be concerned with maximizing the following strictly quasi-concave utility function, (10.3) The special feature with the above specification is that the utility function is defined over three arguments, market goods, c, domestic output, z, and leisure, f, as opposed to the traditional household production model where individuals are assumed to derive utility from leisure and commodities. This specification can be seen as a reduced form of the two-stage model and is identical to the Gronau model given the assumptions that (i) work does not involve any direct utility or disutility and (ii) market goods and home-produced goods are not perfect substitutes (see Gronau 1986). In order to focus the attention on the allocation of time between households, the domestic output, z, in contrast to in the Gronau model, is assumed to be produced by using one’s own labor, h, and/or employed domestic labor, d. Thus: (10.4)
ANNE-MARIE PÅLSSON
159
where g( ) is a quasi-concave function. We can write the income and the time constraints facing the individual in the following way, assuming that the non-labor income is zero: (10.5) and (10.6) where the following abbreviations are used: c market consumption non-domestic labor one’s own domestic labor d employed domestic labor, d>0 for employed labor, d<0 if the household supplies its own labor to the domestic market. L available time f leisure z domestic output wl non-domestic wage rate (exogenous, given by the existing wage distribution) wd domestic wage rate (exogenous) tl marginal tax rate on market labor=income tax (exogenous) td marginal tax rate on purchased domestic service=employment tax and value added tax (exogenous), In order to simplify the analysis and focus on the impact of tax wedges on time allocation, we will make some assumptions. First we examine a household consisting of one individual only and we also assume that there is no difference between males and females. This is not a critical assumption if the tax schedule, as in the case of Sweden, is based on individual income and not on household income. Second, we assume that there is no black market. This may appear as a more serious shortcoming of the model since there is indeed a black market, although its extent is hard to estimate. In fact, it is reasonable to believe that the market for domestic services is dominated by illegal labor, as such help is the only one that can compete with one’s own labor, but that the range of the market as a whole is limited. In order to simplify the analysis, it is also assumed that the correlation between market productivity and household productivity is zero, and that the representative individual is characterized by a constant marginal productivity in both market production and domestic production. We also assume that the demand for nondomestic labor is perfectly elastic and we ignore non-economic aspects of labor supply. Three decisions Maximizing expression (10.3) does in fact involve three separate decisions. First is the choice between supplying labor to the market or to the domestic sector. As the marginal productivity is assumed to be constant, the household will supply labor either to the non-domestic market or to the domestic market depending on the relative wages, hence “l” and “−d” are perfect substitutes from the supply side. However in order to ensure a unique solution, it is assumed that for equal wage rates the household prefers nondomestic work. Hence if then and l>0, and if then d<0 and l=0.
160
TAXATION AND DOMESTIC SERVICES
The second decision concerns the choice of producing the domestic service by using one’s own labor or hired labor.5 Again we ignore all non-economic aspects and assume that the household has no preferences for d or h, but that h is preferred when the costs are equal. Then where and denote the units of z produced per hour, or the marginal productivity, of hired labor and one’s own labor respectively in domestic production. Consequently “ ” and are “ ” perfect substitutes from the demand side and denotes the relative productivity. The optimal use of d and h is determined by minimizing the cost of producing z* units of domestic output: (10.7) It follows from the above expression and from the assumption of constancy in marginal productivity that d>0 ; otherwise, d=0, where gh and gd are the marginal productivity of own for labor and hired labor, respectively, in domestic production and T denotes the tax wedge and is now defined as: (10.8) The third decision concerns the allocation of the potential income on market goods, domestic goods and services and leisure. And three solutions Hence three different cases can be distinguished, namely:
The first case is denoted the net buyer case, the second the “do-it-yourself” case and the third the net seller case. The optimization problem for each of them is studied separately and the results are summarized in Table 10.4 below. The objective function for each of these household types is the same and given by (10.3), the restrictions are given by (10.5) and (10.6), and the production function by (10.7). Uc Uz, and Uf represent the marginal utility with respect to consumption of market goods, domestic output and leisure, respectively, and and are the Lagrange multipliers associated with the budget constraints. Table 10.4 Summary of the marginal analysis Net buyer
Do-it-yourself
Net seller
h=0 d>0 l>0
h>0 d=0 l>0
h>0 d<0 l=0
Lagrange
One’s own domestic labor Hired domestic labor Non-domestic labor First-order conditions
ANNE-MARIE PÅLSSON
Net buyer
Net buyer
Do-it-yourself
Do-it-yourself
Net seller
161
Net seller
Taxation and household choice Two important points could be made with respect to taxes. First, taxes create wedges and distort the household choice of optimal time allocation. This follows directly when rewriting the first-order conditions in the following way: Net buyer
Do-it-yourself
Net seller
Second, the impact of taxation on household choice varies in the three different groups. Households characterized as being “net buyers” experience that all their choices are distorted and find themselves demanding more leisure and less domestic output than they would have done in the absence of taxes. The impact on the demand for market goods, however, is indeterminate. For households belonging to the “do-it-yourself group,” the distortion is limited to the choice between, on the one hand, consumption of market goods and, on the other hand, consumption of leisure and domestic output. The allocation of time between leisure and one’s own domestic production is not distorted. In general, taxes tend to increase the demand for leisure and domestic output for these households and reduce the demand for market goods. In terms of time allocation, taxes encourage households to devote too much time to non-market activities, such as domestic production. The net seller case is identical to the do-it-yourself case with the exception that net seller households supply labor for domestic production and not for production of non-domestic goods and services. However, not only is the choice of the different types of consumption distorted by taxes but also the distribution of households among the different types is affected. In general, the larger the tax wedges, the larger the do-it-yourself group and the smaller the market for domestic service. Thus, too many households, particularly those characterized by relatively high non-domestic productivity, will devote too much time to producing domestic output, thereby reducing their supply of labor for non-domestic production. And too many households with low non-domestic productivity will engage in producing non-domestic goods instead of producing domestic goods.
162
TAXATION AND DOMESTIC SERVICES
Market equilibrium From the first-order conditions, it follows that the individual demand for domestic labor, as:
, can be written
(10.9) where subscript i denotes the ith individual. Consequently the individual supply, , can be written as: (10.10) d s The aggregate demand for domestic labor, D , and the aggregate supply, D , for domestic labor can thus be expressed as: (10.11) and (10.12) where the households have been ordered consecutively after non-domestic wage rate so that, and wrl=wmaxl. Consequently the equilibrium on the market for domestic labor is: (10.13) The equilibrium, as can be seen from (10.11) and (10.12), will be determined not only by relative prices and household preferences but also by the distribution of non-domestic productivity over the households. The impact on aggregate demand for and aggregate supply of domestic services of changes in the arguments determining them can therefore be divided into two parts. The first part can be called the “group effect” and consists of the impact on the distribution of households over the different types by changes in relative prices. The second part consists of the usual price and income effects and is made up of changes in the demand and supply from households already buying and selling domestic service. More specifically, a reduction in the relative price of domestic service, Twd/wil, tends to increase the size of the net buyer group and thereby to increase the demand for domestic service. A reduction in the relative price also tends to increase the demand for domestic services among those households already buying them. Both the income and the price effect are likely to work in the same direction, as domestic services, if anything, must be considered as a luxury good. Hence, the combined effect is unambiguous and a reduction in the relative price increases the aggregate demand for domestic labor. What about the impact on aggregate supply? From (10.12) it follows that a change in the relative price of domestic services does not necessarily affect the supply of domestic labor. For this to happen the change in the relative price must be brought about by a change in the income tax rate and/or in the wage rate of domestic labor. More specifically, an increase in the relative wage rate encourages more households to supply their labor to the domestic market instead of to the non-domestic one. Hence the group effect is positive. Furthermore, if leisure is a normal good and if the income effect dominates, then an increase in the aftertax wage rate, (1−tl)wd, induces households already supplying domestic labor to decrease their supply, thus leaving the combined effect indeterminate. But if the substitution effect dominates, which is usually assumed to be the case in the low wage bracket, then the supply of domestic labor and the after-tax wage rate will be positively correlated. As the group effect is positive, the combined effect will also be positive.
ANNE-MARIE PÅLSSON
163
Figure 10.1 The market for domestic labor (services)
A stylized picture of the market for hired domestic labor (services) is presented in Figure 10.1.6 S, S' and S" depict the aggregate supply of domestic labor for three different income tax rates, tl, tl' and tl", where tl'
164
TAXATION AND DOMESTIC SERVICES
(10.14) and (10.15) If, as in the case of Sweden, tl and td are of the same magnitude (0.3–0.5), then (10.14) must be greater than (10.15). Hence a reduction in income tax rates will expand the market for domestic services more than will follow from an equal reduction in the taxes on hired domestic labor. HOW TO REMOVE THE TAX WEDGES? There are essentially three ways to reduce or remove the tax wedges: namely, by lowering the total tax burden, by making expenditures for domestic services tax deductible, and by subsidizing domestic services. Regarding the first way, a lowering of the total tax burden may be hard to implement since many countries, such as Sweden, suffer from budget deficits. Tax cuts must therefore be combined with compensating cuts in public expenditures, which may also be hard to accomplish. The extent to which public expenditures must be cut crucially depends, however, on the dynamic effects. We will refrain from analyzing this topic as it requires a full general equilibrium model, something which lies beyond the scope of this paper. The third way, subsidizing domestic services, could also remove or reduce the tax wedges. If income taxes are proportional, then a proper choice of subsidies can establish the same price for purchased domestic services that there would have been if expenditures were tax deductible.7 However, subsidizing instead of allowing deductions suffers from the disadvantage that it contributes to increase aggregate public spending and hence aggregate taxes, which in general will increase total distortions in the economy. Removing tax wedges by allowing expenditures for domestic services to be tax deductible does not suffer from the above shortcoming and is easy to achieve. In order to remove the entire tax wedge, the taxes (and benefits) for social security must also be transferred from the buyer to the seller of domestic services, however. In order to illustrate the impact on household choice of making expenditures for domestic services tax deductible, we shall now consider the same model as above, but in order to isolate the principles we will now disregard the value added tax. The objective function is the same, but the constraints facing the individual are altered and the time allocation problem can be stated as:
(10.16)
From (10.16) it follows that the employment tax (for social security) is no longer levied on gross income but on gross income minus the expenditures for domestic services. As a consequence pension rights and other benefits provided by the social security system must be reduced by a corresponding amount. As before, the household choice involves three different choices. The choice between supplying labor to the non-domestic or to the domestic sector is the same as previously and we need not repeat the arguments here. But the choice between purchasing labor or supplying one’s own domestic labor for producing
ANNE-MARIE PÅLSSON
165
domestic services is now different since the constraint is altered. The production function is now found by tracing out the lowest cost production opportunities: (10.17) From the above expression and from the assumption of constancy in marginal productivity, it follows that , otherwise d=0. As can be seen, the tax wedge is now eliminated and the allocation d>0 for of one’s own labor is solely determined by relative productivity on the domestic and the non-domestic markets. A summary of the equilibrium conditions for each group of households is provided by Table 10.5. As can be seen, when households are permitted to deduct expenditures for domestic services from their taxable income combined with a transfer of the employment tax to the supplier of domestic services, the tax wedge between the leisure and domestic output is completely removed for all three types of households. Furthermore, all households now face the same constraints and the allocation of labor is determined solely by the marginal productivities. But still, only those households with a marginal productivity in non-domestic production that exceeds their marginal productivity in domestic production will buy domestic services and thereby transfer their own time from use in their own domestic production to supply for non-domestic production. The allocation between market consumption on the one hand and leisure and domestic services on the other is still distorted, however, but it is important to note that making expenditures for domestic services tax deductible does not create a new tax wedge, but removes one out of three. Furthermore, from Table 10.5 it follows that the impact of deductions on household choice is most significant for those belonging to the net buyer group. For this group, it is obvious that their production possibility set is thereby enlarged. But households belonging to the net seller group may gain as well. This occurs when the new equilibrium wage rate emerging for domestic production exceeds the wage rate of this group in non-domestic production. Table 10.5 First-order conditions for optimality Net buyer
Do-it-yourself
Net seller
The distribution of households into different groups also changes when expenditures for domestic services are made tax deductible. The net buyer group and the net seller group will both increase whereas the do-it-yourself group will diminish: its size will depend solely on the relative productivity in domestic production between buyers and sellers. In case of equal productivity, the do-it-yourself group will completely vanish. The impact on equilibrium wages and on the size of the domestic market is depicted in Figure 10.1. D1 represents the market demand when tax wedges have completely vanished and D the demand schedule when the tax wedge is greater than 1. Then L1−L and w1–w represent the potential increment in the market size and the wage rate respectively by eliminating the tax wedges.
166
TAXATION AND DOMESTIC SERVICES
EFFECTS ON ECONOMIC ACTIVITY In this section the possible implications for economic activity of allowing tax deductions will be discussed. The focus is on economic efficiency and the impact on the government budget. Other aspects, such as justice and equality between men and women, are also important but will be examined here only briefly. Tax deductions and economic efficiency The impact on GDP is in fact crucial for the following reason. It was pointed out in the previous section that a removal of the tax wedges increased utility for households buying the service and for households providing it. We did not, however, analyze the impact on households belonging to the do-it-yourself group. These households neither bought nor provided services and did not change their behavior when deductions were implemented. If tax rates do not change due to the deductions, then one can argue that do-it-yourself households are not affected at all. Allowing for deductions must therefore be Pareto-efficient since no household would voluntarily change position if this did not improve utility. However, if tax rates change due to the implementation of deductions, then do-it-yourself households are also affected. If deductions result in an increase in economic efficiency, then they will also benefit do-ityourself households since taxes could be cut without reducing public spending or public spending could be increased without increasing taxes. Things would look different however if total economic activity is reduced. If public income is proportional to GDP, the government will then have to increase the tax rates in order to finance a given public budget. Do-it-yourself households will end up with lower utility. As a matter of fact, households belonging to the other groups could also experience a reduction in utility. This happens if the income effect from higher taxes is large enough to offset the substitution effect from a lower price on domestic services. It is therefore important to analyze the impact of making expenditures tax deductible for economic activity. In order to analyze this issue more formally, we will return to the model developed in the previous section. Let Y1 and Y2 denote aggregate GDP without and with tax deductions, respectively. Note that GDP here is measured as the value of non-domestic production only. The exclusion of domestic production is explained by the simple fact that if it was included, it would have created a serious bias. This is so because the registered value would depend on the producer. If it is produced using the producer’s own labor, it is not included; if it is produced using hired labor, then it is included. Bearing this in mind we get: (10.18) and (10.19) and w*d1 is the equilibrium wage rate on domestic service without and with tax deductions, respectively, and lb[ ] and lb[ ] is the labor supply from the do-it-yourself group and net buyer group, respectively. q(w) is the density function and ɸ the relative productivity, gh/gd, which for simplicity is assumed to be the same across households. Deductions increase economic efficiency, i.e.Y2 is greater than or equal to Y1 if
ANNE-MARIE PÅLSSON
167
(10.20)
which is identical to
(10.21)
The first term in the above expression corresponds to the increment in production when households transfer their own labor from the production of domestic goods and services to that of non-domestic ones. We can call this effect “the switching effect” or the “group effect.” It refers to the impact on non-domestic production when households move from the do-it-yourself group to the net buyer group. This term is greater or equal to zero provided that not all one’s own labor that is freed by buying domestic service is transferred to leisure. In that case it equals zero. The second term is also greater than zero and corresponds to the usual price effect. It refers to the adjustment within the net buyer group only. When the price of domestic services decreases, then the demand for it increases. As a consequence more of one’s own labor will be transferred from the production of domestic service to that of non-domestic. The last term is also positive and corresponds to the value of output foregone by transferring labor from producing non-domestic goods to producing domestic goods and services. Thus one can conclude that if the increase in the value of production from those new and existing households buying domestic services exceeds the decrease in the value of production from those leaving the non-domestic market in favor of the domestic, then aggregate GDP increases. To what extent this will be the case depends strongly, however, on the distribution of wages and on the propensity to replace one’s own domestic labor with hired domestic labor and to increase the supply of labor to the market instead of increasing leisure. In order to get an idea of the magnitude, we will assume equal productivity in domestic production: hence, =1 If the labor supply from the do-it-yourself group is set on the average to l and the labor supply from the net buyer group to , then expression (10.21) could be simplified to: (10.22) Now, just for the sake of simplicity let market wages be uniformly distributed, that is: (10.23)
168
TAXATION AND DOMESTIC SERVICES
Figure 10.2 Impact on GDP by allowing tax deductions
where q(wl), is the probability density function and M and K refer to the highest and the lowest wage rate on the market respectively. Then GDP increases if: (10.24) Figure 10.2 illustrates the increase in the labor supply for producing non-domestic goods, , necessary to make the impact on GDP positive for different values (wd2/wd1). Two values of T have been used, namely, T= 3.9 and T=2.5. These are the relevant values in Sweden in high and low income brackets respectively. As can be seen, there is more likely to be a positive effect on GDP for large tax wedges than for smaller ones. When the equilibrium wage rate on the market for domestic labor is sensitive to demand, i.e. for high values on (wd2/wd1), then buyers must increase the time allocated, to the production of non-domestic goods more than they would have to if the opposite were the case. Figure 10.2 also shows that the labor supply response need not be large in order to obtain a positive effect on GDP. This is so because it is not likely to expect a wage rate response of more than 1.4 as the wage structure is already extremely compressed. Thus, even for a very modest response (for example =0.1), removal of the tax wedges as proposed here, tends to increase GDP, as net buyers can be expected to be located in the high income bracket, where the tax wedge is 3.9. The suggested change in the tax structure is then Pareto-improving. Other aspects These calculations are quite cautious, however. For instance, labor demand is assumed to be perfectly elastic, meaning that unemployment does not exist. In reality, this is not the case. On the contrary, unemployment is troublesome and the present unemployment rate in Sweden falls between 8 and 20 percent depending on how it is defined. Most other countries in Europe also have a very high unemployment rate. The important implication of unemployment in this model is of course that no non-domestic production need be forgone by changing the tax structure. Hence the second term in (10.24) will be much smaller than assumed here. As a matter of fact it can even be zero if unemployment is large enough. Making expenditures on domestic services tax deductible will unambiguously increase GDP.
ANNE-MARIE PÅLSSON
169
Tax deductions on domestic services could also have important implications for the equality between men and women. According to Table 10.1 and Table 10.2 most domestic services are still carried out by women. As a consequence the female labor supply is restricted and women’s opportunities for earning a taxable income and thereby becoming eligible for social security are limited. If women would allocate their time between household work and market work in the same way as men, then they would increase their market work on the average by roughly 14 hours weekly. This corresponds to 32,000 hours over the entire active period 20–64 years or more than SEK 2.5 million calculated on a rather low wage rate (SEK 13,000 monthly). If we were also to include the pension rights forgone then the amount would be even larger. The observed allocation of time between spouses can however be seen as a consequence of the existing tax structure and distribution of productivity. Men and women simply behave as every rational economic man or woman would: they utilize their comparative advantages, specialize in those activities in which they are relatively good and trade within the family . Making expenditure tax deductible would however make it possible for women with relatively high market productivity in particular to use more time for producing non-domestic goods on the market and less time for producing domestic goods and services for their own consumption. According to the previous analysis and the assumption that men are relatively more productive in producing non-domestic goods and services, i.e. > then for it to be profitable in economic terms for a woman to buy domestic service and thereby to specialize in her own non-domestic work. If however households were allowed to deduct expenditure on domestic services from taxation, then it is sufficient that the wage rate of the female buyer exceeds the one of the seller of domestic services, with the proper adjustment for differences in productivity, that is, for exchange to be advantageous. Hence also women with much lower wages are now in the position to buy services and to reallocate their time towards more non-domestic production, thereby increasing their income. The following example could illustrate this potential. We assume that the marginal productivities in domestic production are equal between the seller and the female buyer and that the going wage rate for the seller of domestic services is SEK 12,000 monthly. The buyer must thus earn more than SEK 46,800 monthly to buy domestic services. Less than 0.5 percent of all persons employed in Sweden enjoy such a high salary and among those we would find very few women. With deductions however she only needs to earn more than SEK 12,000 monthly for trade to take place. This means that most women now are in the position to specialize and trade, if they so wish. Hence tax deductions contribute to enlarge the opportunity set which particularly benefits the one in the family with the lower wage rate, who most often happens to be the woman. SUMMARY The traditional formulation of the household production theory does not take into explicit account the possible substitution between one’s own domestic labor and employed (purchased) domestic labor. The prevailing tax system, based on a double taxation of labor principle, greatly hinders households from specialization and exchange by creating large tax wedges. In this paper it has been shown that removing tax wedges improves household time allocation from an efficiency point of view. As a consequence total output may also increase. Tax wedges can easily be removed by making expenditures on domestic services tax deductible combined with a transfer of social security taxes (and benefits) from the buyer to the seller. As a
170
TAXATION AND DOMESTIC SERVICES
consequence households will demand more domestic output than before with an increase in the wage rate for domestic services as a result. The proposed change in the tax structure is beneficial to households with a low productivity on the market, as the market for domestic services offers an opportunity to find work that is better paid than that on the regular market. It is also beneficial for households with high productivity on the market, as they now can supply more labor for non-domestic production and substitute purchased labor for their own labor in domestic production. But allowing tax deductions may also benefit households belonging to the “do-ityourself” group. This happens if economic efficiency, i.e. GDP, increases. Then the aggregate tax burden could be lowered without a corresponding reduction in public spending. REFERENCES Becker, G. (1965) “A Theory of the Allocation of Time,” Economic Journal 75:493– 517. Cigno, A. (1991) Economics of the Family, Oxford: Clarendon Press. Gronau, R. (1986) “Home Production—A Survey,” in C.Ashenfelter and R.Layard (eds) Handbook in Labor Economics, Amsterdam: Elsevier Science Publisher BV. Lancaster, K.J. (1966) “A New Approach to Consumer Theory,” Journal of Political Economy 74:132–57. Pålsson, A.-M. and Norrman, E. (1994) Finns det en marknad för hemarbete?, Stockholm: SNS. Rosen, S. (1995) Public Employment and the Welfare State in Sweden, Occasional Paper no. 61, SNS, Stockholm. Sandmo, A. (1990) “Tax Distortions and Household Production,” Oxford Economic Papers 42:91–104. SCB (1990/91) “Living Conditions,” report no. 80, The Swedish Time Use Survey, Stockholm: Statistics Sweden.
NOTES 1 See for instance Cigno (1991). 2 This requires a proportional tax structure. With a progressive income tax, trade within households as well is made more difficult but is not hindered. 3 Note that alternative (3) is superior to alternative (4) from a purely fiscal point of view. 4 Each domestic activity should be analyzed separately, however. Hence, for some activities the male productivity exceeds the female (e.g. changing a door). In such cases men tend to specialize in domestic production. 5 The separability follows from the specification of the model; see for instance Sandmo (1990). 6 Domestic labor and domestic services can be analyzed simultaneously due to the specification of the production function. They differ only by a constant. 7 Then
INDEX
“Address to Woman” (Wicksell) 3–4 age factor (working hours) 111, 112, 113, 166, 170 Allen, D. 46 altruism 57, 59, 145; Becker’s model 11, 26, 46, 55–6, 58, 65, 146–7, 149, 151; reciprocal 149–52 Amemiya, T. 184 Apps, P. 66 auction theory 57
Bourguignon, Francois 33, 69, 70 Bragstad, T. 52 Branting, Hjalmar 4 Brien, Michael J. 112 British Household Panel Study (1992) 14, 119–20, 123–6 Brown, Murray 24, 25, 45, 65–6, 70–1 Browning, Martin 33 Bruce, Judith 38, 55 Buchanan, J.M. 56 Bugge, Anna 3–9 Bumpass, L.L. 122, 128 Bureau of Labor Statistics 112 Burton, Peter 27, 28
Bank of Italy survey 71–2 bargaining: common preference models 10–11, 12, 23, 24, 25–9, 32, 37; contributions of bargaining approaches 37–9; cooperative models 11, 29–32, 33; household bargaining theory 13, 100, 107–9; marriage market (role) 35–7; non-cooperative models 11–12, 30–5, 44–60, 84 “bargaining power” parameters 45 Barro, R. 56, 140 basic home care allowance (BHCA) 179–80 Becker, Gary 4, 13, 17, 35, 44, 67, 84, 100, 101, 103, 194, 197; altruist model 11, 26, 46, 55–6, 58, 65, 146–7, 149, 151 Bergmann, B.R. 46 Bergstrom, T.C. 47, 49, 55, 56, 179 Bernheim, D. 57, 58–9 Binmore, K. 45 Blau, D.M. 178 Blom, S. 164 Blomquist, S. 179 Blossfeld, H.P. 161, 173 Blumberg, Rae L. 38 Bound, John 140
Carlin, P. 71 Chiappori, Pierre- André 32, 44, 45, 65–6 child benefits 23, 28–9, 36, 37, 71, 95 child-specific goods 71 childbearing: cohabitational unions 125–8; one-parent families 14–15, 122–3, 128–33; partnership/pre-partnership births 122–3; pathways into partnership and 129–30 childcare 92–3; choice (public policy) 178–91; efficiency/equity trade-offs 143–55; family policy 50–1; home care allowance 16–17, 178, 179–84, 187, 189– 91; human capital and 15–16, 139–55; labor market decisions and 72, 74–8; subsidized 163–4 Childcare Stimulation Act (1991) (The Netherlands) 154 children: benefits/allowances 23, 28–9, 36, 37, 71, 95; equality between 146–9; home care allowance 16–17, 178–87, 189–91; 171
172
INDEX
intergenerational relationships 12, 16, 54–8, 149–54; kids-do-better hypothesis 37–8; rotten kid theorem 26, 47, 55–6; see also family Cigno, Alessandro 152 Coate, S. 47 Cobb-Douglas welfare function 45 Cogan, John 115 cohabitation 4, 5, 14–15, 71, 119–33 Cohen, L. 46 “collective model” (bargaining) 32 Colombino, U. 75, 76 common preference models 10–11, 12, 23, 24, 25–9, 32, 37 comparative advantage 49, 52, 198; dominance and time allocation 13, 84, 85, 87–9, 90–1, 94, 96 “competing-risk” model 160 conditional logit model 183, 185, 187, 188 consensus model 11, 25–6, 65, 84 contracts: complete 44; incomplete 46 Consumer Expenditure Survey (US) 71 consumption 25–6, 32–3, 55, 56–7 “continuous school day” 154 cooperative bargaining models 11, 29–32, 33 coordination-problems models 47 Cournot-Nash equilibrium 31 Cournot model/conjectures 87 Cox, D.R. 173 Cox proportional hazard model 124, 131, 161 Crisis in the Population Question (Myrdal and Myrdal) 153 crowding-out effect 12, 49, 50 Dankmeyer, Ben 152 day care systems 180–91 decision making: family decision models 10, 67–70; see also bargaining Del Boca, Daniela 71, 75, 76 demographic factors: one-parent families 14, 119–30; working mums 164–74 Desai, S. 167 disabled children (equity tax) 147–9 distribution 44–5; bargaining models 29–35, 37–9;
income-pooling assumption 23, 25–9, 70–8; intrahousehold 12–13, 65–81; in marriage 11, 23–39; marriage market (role) 35–7 division of labour, sexual 17, 18, 44, 146, 178–9 divorce 26–7, 29–32, 36–9, 46, 52–4, 59, 66, 71, 84, 107, 166 domestic output 18, 195, 204 domestic services, taxation and 17–18, 194–215 dominating spouse 13, 84–96 Down’s Syndrome 147 Droogleever Fortuijn, Joos 145 Duncan,Greg J. 141, 184 duty of presence (childcare) 145 Eaton, Jonathan 140 economic activity, effects of tax deductions 17–18, 209– 14 efficiency: economic: childcare and human capital 15–16, 139–55; tax deductions and 17–18, 194–5, 210–13; effect of equality between children 146–9; equity trade-offs (childcare) 15, 143–55; intergenerational 149–52; models 44–6, 49–50; time allocation and 89–90, 152–4 Eicker-White estimate 76 Ellingsaeter, A.L. 168–9 employment: domestic services 17–18, 194–215; full-time/part-time (mothers) 16, 159–75; participation decisions 12–13, 65–81; unemployment 112, 113, 114–15, 153, 171, 195, 213; women’s hours of work 13–14, 100–16 empowerment of women 24, 37, 140–1 England, Paula 35 equality: between children (effects) 146–9; intergenerational 149–52 equity/efficiency trade-offs (childcare) 15, 143–55 equity tax 147–9, 151 escape rate (partnerships) 123 Estaugh, V. 132 Eurostat 159 Even, W.E. 166 extrahousehold environmental parameters (EEPs) 30, 67, 68, 71, 74, 77
INDEX
family: bargaining see bargaining; behavior (models) 23–4, 25–35, 37–9; decision models (testing) 67–70; economics of (models) 10–12; economics of (value) 18–19; formation, marriage and 12–15; marriage market (distributive role) 35–7; neoclassical model 65, 67, 68, 71, 197; non-cooperative games 44–60; nuclear 12, 46, 55, 58; one-parent see one-parent families; with one dominating spouse 13, 84, 96; -specific public good 11–12, 25, 31–4, 47–53, 85–8, 92–5; structure (Italy) 72–4 family policy 23, 50–2; childcare choice and 16–17, 178–91; childcare and economic efficiency 15–16, 139–55; domestic services and taxation 17–18, 194–215; dominance and 13, 92–5; full-time/part-time work (mothers) 16, 159–75; household allocation of time 15–18 fathers: parental leave 162–3, 180; sexual division of labor 17, 18, 44, 146, 178–9 fertility rates 24, 72, 109–10, 153 Finland (childcare policy) 16–17, 178–91 Flinn, C. 71 Flood, Lennart 141 focal point equilibrium 34–5, 39 Follow-Up Survey of National Longitudinal Study 122 free-rider problem 11–12, 46–7, 58 full-time work (for mothers) 16, 109, 159–75 full wage 160–1 game-theoretic models 24; cooperative bargaining 11, 29–32, 33; non-cooperative bargaining 11–13, 30–5, 44–60, 84–7 Gårdlund, Torsten 4, 8 gender 4; division of labor 17, 18, 44, 146, 178–9; empowerment index (GEM) 140–1; roles 34–5 Getis, Victoria L. 153 Goldin, Claudia 141 Goldman, Noreen 112 government: provision of family public good 50–1, 85, 92–5;
173
subsidies (childcare) 16–17, 178–84, 187, 189–91; see also family policy; taxation Granger, Clive W.J. 109 Gråsjö, Urban 141 Griliches, Zvi 147 Gronau, R. 17, 44, 200–1 Grossbard-Shechtman, S.A. 100, 101, 102, 103–4, 107–8, 109–10, 115 group effect 205, 206, 211 Gustafsson, B. 163 Gustafsson, Siv 152, 153, 154, 178 Haddad, Lawrence 27 handicapped children (equity tax) 147–9 hazard rate of employment 160–1 Heckman, J. 75, 100, 101, 112, 115 Hill, Martha S. 141, 146 Hoddinott, John 27 Hoem, B. 164 Hoem, J.M. 163, 164 Hofferth, S.L. 178 Hoffman, S.D. 184 home care allowance (Finland) 16–17, 178–87, 189–91 Horney, Mary Jean 24, 25, 45, 65–6, 68, 69, 70, 74 Hotz, J.V. 160 hours of work: full-time/part-time (mothers) 16, 109, 159–75; women’s 13–14, 100–16 household: bargaining theory 13, 100, 107–9; distribution of resources see distribution; production theory 18, 197–200; sector, market and (interdependence) 18–19; time allocation see time allocation; see also cohabitation; family; marriage human capital 4, 95; childcare and 15–16, 139–55 Ilmakunnas, S. 178, 180 incentive schemes 47, 56, 57 income: child benefits/allowances 23, 28–9, 36, 37, 71, 95; four income effects 103–4; maternity leave/benefit 161–4, 167, 169, 171–5, 178– 9, 181, 184–5; non-wage 73–80, 81, 103–4;
174
INDEX
-pooling assumption 12–13, 23, 25–9, 37–8, 67, 70–8; see also wages income tax 50, 52 individual utility maximization 102, 103, 105 intergenerational equality/efficiency 16, 149–54 intergenerational issues 12, 54–8 “internal divorce” 54 intersibling equity tax 147–9, 151 intrahousehold distribution of resources 12–13, 65–81 inverse Mills ratio 113–14 Joesch, J.M. 166 Johnson, George E. 140, 141 Juster, Thomas F. 141, 146 Kapteyn, A. 53, 69, 70 Keynes, J.M. 3 Kiernan, K.E. 130, 132 Kilbourne, Barbara 35 King, A.G. 46 Kjulin, U. 163 Klerman, Jacob A. 154 Konrad, K.A. (Konrad-Lommerud model) 48–50, 52, 53, 95 Kooreman, P. 53, 69, 70 Kortum, Samuel 140 Kravdal, Ö. 164 Kreps, David M. 34 labor market: childcare choice and 16–17, 178–91; domestic services 17–18, 194–215; marriage market and 13–14, 100–16; mothers in full-time/part-time work 16, 159–75; participation decisions 12–13, 65–81 Lamb, Michael E. 151 Lanahan Act (US) 153 Lancaster, K.J. 197 League of Nations 8 Lectures on Political Economy (Wicksell) 8 Lee, Jong Wa 140 Leibowitz, Arleen 154, 178 leisure 17, 53, 66, 70, 71, 81, 101, 195, 200, 204–5, 209 Lesthaege, R. 165 Leuthold, J. 53 Lieon, Saskia 154 Lommerud, Kjell Erik 46, 47; Konrad-Lommerud model 48–50, 52, 53, 95 Loury, G.C. 47 love 46, 57, 58–9
lump-sum taxation 49, 50, 51, 93 Lundberg, Shelly 28, 30, 31, 36, 52, 66, 71, 79 Maassen van den Brink, Henriëtte 154 McElroy, Marjorie B. 24, 25, 30, 45, 53, 65–6, 67, 68, 69, 70, 74 Manser, Marilyn 24, 25, 45, 65–6, 70–1 marginal tax rates 170–1 marital transfers (effects of dominance) 90–2 market equilibrium (domestic services) 204–7 market goods 17, 195, 204 market sector, household sector and (interdependence) 18–19 market wage 100, 160, 212 Markov model 129, 132 marriage: bargaining and distribution in 11, 23–39; cohabitation 4, 5, 14–15, 71, 119–33; dominating spouse 13, 84–96; family formation and 12–15; market (imbalances) 13–14, 100–16; market (role) 11, 35–7; squeeze hypothesis 14, 74–8, 106–10, 111–16 see also divorce “maternal wage” 190 maternity leave/benefit 161–4, 167, 169, 171–5, 178–9, 181, 184–5 Michael, R.T. 122, 128 Mikkola, M. 180, 181 Miller, R.A. 160 Mincer, Jacob 100, 101, 152, 154 Moors, G. 165 mothers: childcare choice and labor participation 16–17, 178– 91; full time/part time work (Norway/Sweden) 16, 159– 75; maternity leave/benefit 161–4, 167, 169, 171–5, 178– 9, 181, 184–5; opportunity cost of time 145–6; time use (efficiency) 152–4; see also childbearing; childcare; children; family Mott, F.L. 166 Myrdal, Alva 153 Myrdal, Gunnar 153
INDEX
Nash, John F. 29 Nash bargaining model 29–30, 34, 45, 47, 53, 66, 67, 68 Nash equilibrium 48, 49 National Child Development Study (NCDS) 130 National Longitudinal Survey 70–1, 111, 116, 122 National Survey of Families and Households 122 Nelson, J.A. 46 neoclassical model (of family) 65, 67, 68, 71, 197 Neuman, Shoshana 104, 108 neutrality 12, 49, 51, 52, 55–6 non-cooperative bargaining models 11–12, 30–5, 44–60, 84; Stackelberg game 13, 47, 85–7 non-wage income 73–80, 81, 103–4 Norway (full-time/part-time work) 16, 159–75 Norwegian Family and Occupation Survey 159, 164, 174 Norwegian Feminist Association 5 Norwegian Labor Force Survey 168, 169 Norwegian Women’s Suffrage Association 5, 8 nuclear family 12, 46, 55, 58 OECD 140 one-parent families 14–15, 119, 122–3; creation of 128–9, 130–3 opportunism 58, 84 opportunity costs 71, 74, 86, 93, 145–6, 152–3 Ott, N. 66, 146 Pahl, Jan 38 Panel Study of Income Dynamics 116 parental leave/benefits 162–3, 180 parents: intergenerational relationships 12, 16, 54–8, 149–54 see also family; fathers; mothers; one-parent families Pareto-efficient decisions 44, 68, 70, 210 Pareto-improvement 91, 92, 212 Pareto-optimality 11, 29, 32–5, 52 part-time work (for mothers) 16, 159–75 partnership rates 120–30 Petersen, T. 161 Phipps, Shelley 27, 28 Polachek, Solomon 152, 154 Pollack, Robert A. 30, 31, 35, 36, 37, 46, 52, 66, 71 Pott-Buter, Hettie 141, 143, 145 “Pre-Thatcher cohort” (partnership rates) 120–30 preference function 65, 67;
175
see also common preference models private day care 181, 182, 184–91 private provision of public goods (in family) 11–12, 47– 53 proportional hazard models 124, 131, 161 public day-care system 180–1, 182, 184–91 public goods 25, 31–4, 85–8, 92–5; private provision of (in family) 11–12, 47–53 Public Use Microdata Sample (PUMS) data 110–11 quasi-wages 101–9 passim reciprocal altruism 149–52 redistribution (male to female) 50, 51–2 Rees, R. 66 reservation utility 68 reservation wage 100–4, 106–8, 112, 115, 160–1, 166 resources: allocation 38, 140, 143–5, 147–52; see also distribution Ricardian equivalence 36, 56 Robbins, Lionel 100 Robins, P.K. 178 Rohwer, G. 161, 173 Rönsen, M. 163, 165–7, 168–9, 170, 174 Rose, Elaina 27 Rosen, Sherwin 155 rotten kid theorem 26, 47, 55–6 Samaritan’s dilemma 47, 56–7 Samuelson, Paul 3; consensus model 11, 25–6, 65, 84 Schippers, Joop J. 152, 154 Schultz, T.Paul 28, 65, 71 “second-generation” models (family decisions) 10, 11 Sen, Amartya K. 35 separate spheres bargaining model 30–2, 33, 34, 52, 66 sex-ratio (marriage squeeze effect) 14, 74–8, 106–10, 111– 16 sexual division of labor 17, 18, 44, 146, 178–9 Shapiro, D. 166 sharing (time allocation and dominance) 89–90 sharing rule 32–3 simultaneous moves game 46–7 single-agent model (of family) 23, 25–6 single utility function 23, 25, 26, 37, 65, 67, 69–70 Smith, Adam 3 Smith, James P. 100 social insurance 139, 143, 145, 151, 197
176
INDEX
social norms 35, 37, 38–9 social welfare function 26–7, 32, 152 socialization process 35 Socioeconomic Survey of Thailand 71 Solon, Gary 141 specialization 90, 91, 146, 197, 198 “speed-premium” (Sweden) 16, 162–3, 175 “spousal labor” 13, 14, 101–4, 105–9, 110, 112, 114 spousal support 90–2 spouse dominance 13, 84–96 Stackelberg game 13, 47, 85–7 Staff, Karl 5 Stafford, Frank P. 140, 141, 146, 147, 152–3, 154, 178 Standard Metropolitan Statistical Areas (SMSAs) 110, 111–12, 116 Stark, O. 58–9 Statistics Norway 164 Statistics Sweden 163, 164, 196–7 strategic bequest motive 47, 57–8 substitution effect 206 Sundström, Marianne 153, 162, 166–7, 168, 174 “survival analysis” 120 Sweden (full-time/part-time work) 16, 159–75 Swedish Family Survey 159, 164, 174 Swedish Labor Force Survey 168, 169 Sweet, J.A. 122, 128 switching effect 211 taxation 66, 94, 95, 185; domestic services and 17–18, 194–215; equity tax 147–9, 151; of income 50, 52; lump-sum 49, 50, 51, 93; marginal rates 170–1; wedges 17, 18, 19, 194–5, 198–214 “Thatcher cohort” (partnership rates) 120–30 Thomas, Duncan 27, 65, 70, 74 “threat point” 29–32, 38, 45, 66, 67, 68, 74 Tijdens, Kea 154 time allocation 15–18, 28, 100, 101; domestic service and 195–207; dominance and 13, 84–5, 87–90, 95–6; Konrad-Lommerud model 48–50, 52, 53, 95; mother’s time use (efficiency) 152–4 Time Use Survey (US) 152 Tinbergen, Jan 140, 152 traditional family decision model 67, 69, 76, 79, 81 transaction costs 46, 91–2 transition rates (partnerships) 120–1, 123–5, 128–30
Treas, Judith 38 Udry, Christopher 33 Ulph, D. 53 unemployment 112, 113, 114–15, 153, 171, 195, 213 United Nations 23–4, 140 US Consumer Expenditure Survey 71 utility functions 58, 66, 84, 86, 100–1, 182–3, 195, 200; single 23, 25–6, 37, 65, 67, 69–70 Vagstad, S. 47 van der Klaauw, W. 71 Vinovskis, Maris A. 153 voluntary contribution game 30, 33, 34, 47, 49, 50, 53, 58 Von-Neumann-Morgenstern utilities 45 wages 4; full 160–1; market 100, 160, 212; maternal 190; quasi 101–9 passim; reservation 100–4, 106–8, 112, 115, 160–1, 166 Waite, L.J. 154, 167 Waldman, M. 55 Wales, T.J. 71 Walras, Leo 3 Warr, P. 47 Warr neutrality 49, 51, 52, 55–6 Weiss, Y. 52 welfare maximization model 45, 56, 87–9, 96 Wicksell, Knut 3–9 Wicksell Nordqvist, Liv 5, 8 Willis, R.J. 52, 122, 128 Wilson, William J. 112 Wissoker, D.A. 178 women: empowerment of 24, 37, 140–1; full-time/part-time work (mothers) 16, 159–75; hours of work 13–14, 100–16; sexual division of labor 17, 18, 44, 146, 178–9; see also mothers Wood, Robert G. 112 Woolley, F. 53 Works Progress Administration (US) 153 World Bank 37–8 World Population Conference (UN) 23–4 Wu-Hausman test 76 Zelizer, Viviana A. 38