Age and Inequality: Diverse Pathways Through Later Life

Citation preview

Age and Inequality

Social Inequality Series Marta Tienda and David B. Crusky, Series Editors

Age and Inequality: Diverse Pathways Through Later Life, Angela M. O'Rand and John C. Henretta Children, Schools, and Inequality, Doris R. Entwisle and Karl Len Alexander Rational Choice Theory and Large-Scale Data Analysis, edited by Hans-Peter Blossfeld and Gerald Prein Generating Social Stratification: Toward a New Research Agenda, edited by Alan C. Kerckhoff Social Stratification: Class, Race, and Gender in Sociological Perspective, edited by David B. Grusky

Age and Inequality Diverse Pathways Through Later Life

Angela M. O'Rand John C. Henretta

"---»--" A Member ot the Perseus Books Group

Sock! Inequality Series

All rights reserved. Printed in the United States of America. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Copyright © 1999 by Westview Press, A Member of the Perseus Books Group Published in 1999 in the United States of America by Westview Press, 5500 Central Avenue, Boulder, Colorado 80301-2877, and in the United Kingdom by Westview Press, 12 Hid's Copse Road, Cumnor Hill, Oxford OX2 9JJ Find us on the World Wide Web at www.westviewpress.com Library of Congress Cataloging-in-Publication Data O'Rand, Angela M. Age and Inequality ; diverse pathways through later life / Angela M. O'Rand, John C. Henretta. p. cm. — (Social inequality series) Includes bibliographical references and index. ISBN 0-8133-1906-4 (he)—0-8133-9812-6 (pb) 1. Aged—United States—Economic conditions. 2. Equality—United States. I. Henretta, John C. II. Title. III. Series. HQ1064.U50665 1999 305.26'0973—-dc21

99-30973 CIP

The paper used in this publication meets the requirements of the American National Standard for Permanence of Paper for Printed Library Materials Z39.48-1984.

PERSEUS

POD

ON DEMAND

10

9

8

7

6

5

4

3

2

Contents List of Tables and Figures

1

vii

Introduction

1

Cohorts, Inequality, and Social Change

5

Inequality and Aging: Stratification over the Life Course, 8 Theories of Intracohort Heterogeneity and Inequality, 9 Economic Inequality Across Countries and Between Age-Groups, 11 Aging Cohorts and Economic Inequality, 16 Employment Institutions and Inequality in Old Age, 27 State Policies, Inequality, and Aging, 29 Conclusion, 31 2

Pathways to Inequality: Intracohort Differentiation over the Life Course

33

Aging and Social Theory: Linkages Between Work and Retirement, 34 Research on Linkages Across the Life Course, 36 An Overview of Income Sources, 42 Cohort Trends in Women's Lives and Intracohort Differentiation, 55 Implications for Intracohort Variation in the Life Course, 62 Poverty, 64 Conclusion, 69 3

Asynchronous Lives: The Normal Life Course and Its Variations

71

The Age Integration of Lives, 71 The Resilience of Gender Structure, 74 Family Pathways: The Breadwinner and Role-Sharing Models, 79 v

Contents

VI

Asynchronous Lives, 85 Studies of Asynchronous Lives and Inequality, 87 Family-Work Pathways to Retirement, 91 Conclusion, 96 4

Pathways to Retirement: The Timing of Retirement

99

Earlier and More Universal Exit, 105 Institutional Structure and Segmentation of Work Exit, 117 Alternate Institutional Pathways Producing Early Exit, 120 The Intersection of Social Structure and Individual Trajectories, 123 Conclusion, 129 5

Labor Markets and Occupational Welfare in the United States

131

From Wage Inequality to Pension Inequality, 133 Employee Benefits as Decentralized Occupational Welfare, 137 Defined Benefit and Defined Contribution Pension Plans, 146 The Private-Public Linkages in Occupational Welfare, 154 Conclusion, 157 6

U.S. Labor Force Participation Trends in Comparative Perspective

158

Changing Patterns of Late-Life Labor Force Participation, 159 Changing Patterns of Employment, 165 Social Policy and Early Exit, 169 Future Changes, 176 The Shifting Life Course, 181 Conclusion, 184 7

Aging in the Welfare State: Strategic Cross-National Comparisons of Life Course Variability and Inequality

186

The Origins of Welfare States, 187 Comparative Pathways to Variability and Inequality, 195 Aged Inequality, 200 Conclusion, 205 8

Conclusion: The Future of the Age-Structured Life Course

207

The Deinstitutionalization of the Life Course? 208 Aging Populations and Social Policies, 212 References Index

219 243

Tables and Figures Tables

1.1

Trends in earnings and income inequality

13

2.1 2.2 2.3

Aggregate income from each income source Median net worth and home ownership after age 70 Poverty rates for persons 65 and over

43 53 67

3.1

Benefits payable to couples with identical earnings

82

7.1 7.2

Comparative pathways to inequality Within decile aggregate income composition of the aged in three countries

188 201

Figures

1.1 1.2 1.3 1.4

Synthetic cohorts: Women's labor force participation Synthetic cohorts: Men's labor force participation Female/Male weekly hours Female/Male pension income ratios

19 20 24 26

2.1

Percentage with pension participation on current job, by gender Percentage with pension participation on current job, by race Monthly pension income by marital status Poverty status of individuals by age

48 49 52 66

2.2 2.3 2.4 3.1 3.2 3.3

4.1 4.2

Types of social structure Percentage of women aged 60+ with non-Social Security survivors' benefits Primary source of non-Social Security survivors' benefits for women aged 60+ Percentage of men in the labor force by age Percentage of women in the labor force by age

73 97 98 107 108 VII

VIII

4.3 4.4 4.5

5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 6.1 6.2 6.3 6.4

Tables nnd Figures

Women's labor force participation by birth cohort Percentage evaluating health as excellent or very good, by education Percentage evaluating health as excellent or very good, by race and ethnicity Male/Female decile earnings and pension distribution cutpoints Female/Male earnings and pension income decile outpoint ratios Industry pension participation rates Occupational rates of women's pension participation Industry rates of health insurance coverage Establishment size and benefit coverage rates Distributions of defined benefit and defined contribution plans Distribution of private pension plans by industry Men's labor force participation at ages 60-64 in the United States and four European Union countries Men's and women's labor force participation rates at ages 65+ in the United States and four European Union couiitries Women's labor force participation at ages 60-64 in the United States and four European Union countries Percentage of population aged 65 and over: 1994 estimates and 2020 projections

110 127 128

134 136 141 143 144 145 149 152

160 162 164 178

/ returned, and saw under the sun, that the race is not to tlie swift, nor the battle to the strong, neither yet bread to the wise, nor yet riches to men of understanding, nor yet favor to men of skill; but time and chance happeneth to them all. Ecclesiastes 9:11

If thou hast gathered nothing in thy youth, how canst thou find any thing in thine age? Apocrypha Ecclesiasticus 25:3

This page intentionally left blank

Age and Inequality

This page intentionally left blank

Introduction

This volume examines the structural and individual processes that differentiate the life course of men and women and yield patterns of inequality in adulthood and old age. Demographic, economic, and political trends have led to the increased differentiation of the life course of men and women. Processes of age structuring and gender structuring of the life course have produced mixed patterns of uniformity and diversity in life transition schedules. Meanwhile age structures and gender structures are themselves being transformed by global processes associated with population aging, economic restructuring, and welfare state reorganization. Master trends in the United States and other industrial countries have developed, and they include the following: (1) an increase in inequality within and across age-groups; (2) the shortening of men's work lives; (3) the lengthening of women's work lives; (4) diverse family arrangements and schedules; (5) earlier retirement of some segments of the workforce; (6) the synchronization of joint retirement among older couples; and (7) the feminization of poverty into late life. These patterns are observable across societies, with some systematic variations. Our primary focus is on changes in life course patterns over recent decades in the United States. However, strategic comparisons of the United States with other Western countries are also presented. Chapter 1 introduces these master trends. Aging and inequality are interrelated in complex ways. On the one hand, age is a relatively enduring principle of stratification in developed countries. Age serves to allocate differential government resources to the young and the old. It is also an efficient mechanism for setting timetables in the workplace. Access to employee benefits that include pensions, disability insurance, and some kinds of health insurance are frequently tied to age. Age is also highly correlated with entry into and exit from major life statuses related to schooling, work, family formation, and dissolution, empty nest, and retirement. On the other hand, recent decades are notable for the decreasing importance of age for the conduct of more and more social roles. The age at which marriage, full-time work, childbearing, and retirement begin, and J

2

Introduction

at which schooling, the work career, marriage, childbearing, and family care giving end, have become more variable. Historical circumstances have introduced succeeding cohorts to changing life conditions and new uncertainties, leading to a loosening of the association between age and social roles. Increased variability in the life course is also associated with increased economic inequality. Diverse life pathways have different economic implications for later life. Some studies suggest that inequality is an inevitable outcome of social aging and the differentiation of cohorts over time. What is being debated is whether, and under what circumstances, individuals in the same cohort become more unequal. The United States exhibits among the highest levels of inequality across the age span, including perhaps the highest relative inequality among the elderly when compared to other nations. In addition, an apparent surge in economic inequality worldwide has attracted the recent attention of many scholars who are concerned with how inequality is produced over the life course over cycles of prosperity, stagnation, and decline. Chapter 2 examines pathways to inequality within aging cohorts. The interplay of employment, family, and state welfare institutions is central to the production of variability and inequality in the structure of the life course and in late-life outcomes. Earnings over the work career are the strongest predictors of economic status in retirement, both for retired workers and for their spouses and survivors. Social Security arid pension income account for over four-fifths of the median income of retired households. Accordingly, labor force behavior and workplace institutions are critical components of aging and inequality. In addition, public sector institutions such as old age insurance and monetary and in-kind transfers serve to offset inequalities stemming from the workplace for workers and their families. The product of the interaction among these institutional spheres includes complex patterns of inequality and processes of status maintenance, status leveling, and cumulative advantage. Chapter 3 examines specific processes that link family and work in the production of inequality over the life course. Families and work careers are not rigid structures, but dynamic processes. Families are changing configurations of interdependency that require the coordination of different life schedules and the management of changing individual and family needs. Diverse patterns of family formation and breakup within and between cohorts produce additional complexity in the scheduling of mid-life and late-life transitions such as work and retirement, respectively. Together, these factors create complex family pathways that covary with employment pathways to produce variability and inequality. Public policies associated with welfare benefits and income taxation differentially support or penalize family arrangements and work career pat-

Introduction

3

terns in ways that facilitate inequality. Traditional "breadwinner" families receive more support in employment and welfare policies, although changing social roles in the family are challenging these institutions. The implications of the interdependence among family, workplace, and state are especially significant for women, who are more likely to face a longer time in old age alone with higher risks for poverty. Chapter 4 closely examines complex pathways through later life, including the heterogeneity of pathways to retirement. Retirement has become a common feature of life in industrial societies. Traditional pathways through work and pension acquisition have contributed to its universality. Yet alternative pathways to retirement are also institutionalized and contribute to the heterogeneity of the retirement transition. The intersection of individual biographies with alternative institutional arrangements that include unemployment, disability, and family status— as well as pensions derived from employment—allocates individuals within cohorts along different pathways from work to retirement and from dependence on earnings to dependence on different transfer programs. The scheduling of these transitions increases variability and inequality in old age. Chapter 5 focuses on the U.S. occupational welfare system. The U.S. welfare system is market centered, highly decentralized, and increasingly individualized. New forms of occupational welfare are placing more responsibilities on workers and fewer on employers arid public sector programs. The individualization of pensions, health insurance, and other life course support programs is introducing new risks to income maintenance and new sources of inequality in old age. The final chapters examine inequality and aging in a comparative perspective. Important similarities are apparent among Western countries in the shifting pattern of the life course, especially in the trend toward early retirement. Chapter 6 compares the labor force participation patterns of older men and women in the United States over the past three decades with those in other countries, particularly the United Kingdom, France, Germany, and Sweden. Changing population structures and employment opportunities have created similar schedules of early labor exit, including retirement for men and women alike. But these schedules are regulated by different institutional arrangements among employment and state institutions. Chapter 7 presents a general framework for comparing the bases of variability and inequality of United States with other countries. The interplay of employment, family, arid state welfare institutions is central to the production of variability and inequality in the structure of the life course and in late-life outcomes. The interconnectedness of educational, employment, family, and welfare systems constrains the level of variability and

4

Introduction

degree of inequality across countries. The United States is distinguished by the loose coupling of these institutional sectors and by a relatively weak welfare institution. Sweden and Germany, on the other hand, have stronger welfare systems and lower levels of life course variability and inequality, although they differ in the interrelationship between family (and gender) structure and other institutions for reinforcement of the coherence of the life course. These three countries provide a strategic comparison of the institutional sources of life course organization and latelife inequality. Chapter 8 concludes with a consideration of the future of the age-structured life course. Aging and inequality in the twenty-first century will be marked by the strong imprint of the baby boom generation in the United States. Proposals to extend the working life by delaying eligibility for full retirement benefits until age seventy and to move to more individualized retirement savings accounts tied to the market have countervailing implications. Changing institutional arrangements are responsive to the increased longevity and physical and mental viability of aging populations. However, global processes are introducing more obstacles to economic well-being over the lifetime and in old age for some segments of the population.

1 Cohorts, Inequality, and Social Change

Historical circumstances have exposed succeeding cohorts over this century to changing life conditions, new uncertainties, chances, and turning points. These historical changes have produced cohort-based variations in life course patterns and outcomes. Increased opportunities for higher education, improved economic prospects for wider segments of the population, and continuously extended life expectancy in the United States and other advanced industrial societies have proceeded throughout this century. Yet succeeding cohorts have not benefited equally by these trends, within and across national boundaries. The complex interplay of structural and individual factors has produced variable experiences under conditions of historical change that have yielded uneven consequences and divergent fortunes. This situation is dramatically illustrated by the case of the so-called baby boom cohort of the United States, a demographic category that is often, and erroneously, treated as a homogeneous population. The U.S. baby boom cohort, born between 1946 and 1960, has begun "turning 50" in the late 1990s. Its members arrive at this transition after having begun their lives during a time of unprecedented general prosperity and educational opportunity in the first decade and a half following World War II and then maturing through periods that have vacillated between relative prosperity and economic uncertainty. The 1990s have presented them with more of the same: recession followed by economic boom. On average, the baby boom cohort as a whole is relatively prosperous in historical terms. Its relative prosperity is inherited from a nearly century-long trend of rising economic well-being, albeit with a few interruptions precipitated by depressions, recessions, and wars. Accordingly, 5

6

Cohorts, Inequality, and Social Change

this cohort has participated in a master trend of this century. Its experiences have been both ordinary and exceptional. The size of the baby boom cohort relative to earlier and later cohorts has given the age structure of the United States a distinctive though changing shape over the years. As the baby boomers have aged, they have been numerically and institutionally significant for American society. As children in the 1950s they required expanded residential, health, educational, and cultural institutions. In the 1960s their numbers empowered them to resist traditional sources of authority associated with the family and the state and to participate with some effect in the new "status politics" of that era. In the 1970s and 1980s they crowded labor markets and courtrooms. As they turn 50 in the 1990s, they are inspiring extensive projections and debate on their prospective impact on the stock market and on the public retirement system that was designed a decade before their birth for a different population. The cohort concept, as well as the baby boom cohort as a historical case, provokes commonsense expectations of similarity, common experience, and shared fate. Yet as members of the baby boom cohort turn 50, they are sociologically interesting not only because of their similarities but also because of their differences. First, those turning 50 at the end of the twentieth century are the vanguard of this cohort. As such, they have experienced the post-World War II period differently than their younger counterparts. Those born before the mid-1950s compose the Vietnam War generation, whereas those born closer to 1960 have been far less directly exposed to war. Alternatively, those born earlier in the cohort entered the labor market before globally driven changes in the workplace began to introduce new uncertainties into the employment relationship; they completed their education and entered their work career ahead of restructuring, downsizing, arid economic reorganization. Meanwhile the later born women and men of the baby boom initiated their work careers in the midst of a turbulent labor market. In effect, the intersections of historical changes with the early life transitions of sequential segments of the baby boom cohort have made their respective aging experiences different. Second, cohort members turning 50 in the late 1990s reflect considerable diversity among themselves. The accumulation of life experiences, achievements, and disappointments has moved individuals within baby boom segments along different pathways that present them with a mix of expectations of the future. One aspect of within-cohort diversity is economic inequality. Variations in educational achievement, labor force participation, labor market opportunities, family formation and dissolution patterns, pension and asset accumulation, and health trajectories—all of which are cumulative and interrelated over time—produce diverse life chances with unequal outcomes in middle and late life. Economic in-

Cohorts, Inequality, and Social Change

7

equality is pervasive among baby boomers and, in the 1990s, appears to have increased (Gottschalk and Smeeding 1997). The cohort concept and the baby boom case also provoke a strong image of "agency," implying that cohorts are somehow exogenous to the institutional and historical contexts that actually produce and condition them. However, social institutions define the partitions and the boundaries of the "normal life course" within cohorts. Youth, adulthood and old age are socially constructed categories. Their boundaries are drawn and redrawn by changing institutional arrangements. Age-related roles are temporally organized by educational schedules, employment rules and timetables, family life cycles, and state welfare regimes. Cohorts and the socially diverse subgroups that constitute them—gender groups, racial and ethnic groups, and economic status groups—are constructed and reconstructed by history and social structure and cannot be reduced to a set of fixed demographic traits (Elder 1994). On the other hand, socially produced cohorts do make history. As succeeding cohorts face changing historical conditions their aggregate responses exert pressures for institutional change (Riley 1998; Riley, Kahn and Foner 1994). Riley (1987) aptly illustrates this point in the case of women's rising labor force participation patterns over this century. Women began entering the labor force in large numbers in periods during which this behavior did not conform to gender role expectations and was not encouraged by workplace structures nor by the norms regulating the family division of labor. Nevertheless, participation rates continued to climb. Today, normative expectations have changed with regard to gender roles and the reasons why women do not work have become perhaps more intriguing and problematic than in previous times (Bianchi 1995). Accordingly, historical changes and chance events interact with the institutions that shape the normal life course through individuals' day-today decisions and behaviors over their lives. This interaction produces diversity and variability; subgroups deviate from the normal life course in the scheduling of their lives in education, work, and family domains. Hence, at any time in the aging of a cohort, people are in different places in life. The cumulative patterns of life send people down different pathways. Chance events or constraints can interrupt or deflect them from previous paths. Passages between significant statuses, across temporal or chronological boundaries, and through historical periods bring into focus how lives are shaped and reshaped at social interfaces (Elder 1994). Biography produces diversity within the boundaries of institutional arrangements (Kohli 1988). The example of the baby boom cohort illuminates the theoretical and empirical value of examining variability and inequality between and

8

Cohorts, Inequality, and Social Change

within cohorts. Several theories have emerged in American sociology to explain these patterns. These theories are introduced in the next section of this chapter and are integrated with analyses in later chapters. Those of particular interest to this book explain the processes by which variability and inequality develop within a cohort over time (i.e., with age) and produce diverse pathways through later life. The rest of this chapter introduces major cohort trends and historical processes that underlie the relationship between inequality and aging and that are considered in greater detail in later chapters. First, the level of economic inequality in the United States will be compared with other developed countries. Economic inequality, as it is reflected in earnings, income, and pensions, will be emphasized because these sources of inequality stem from the market, which is the major source of inequality in modern societies. Second, intercohort trends in gender inequality will be reviewed. The long-term upward trend of women's labor force participation has reorganized the life course of women and men alike and has contributed to increased within-cohort heterogeneity with implications for inequality within and across cohorts. Third, the organization of labor markets and employment relationships directly influences the distribution of economic rewards. In recent years, significant changes in these institutions have provided the conditions for increased inequality by deinstitutionalizing the long-term employment contract and individualizing pension savings. Finally, the role of public policies and state structures in producing economic equality will be introduced. The remaining chapters of the book will examine these major processes and trends in more detail, especially as they bear upon intracohort patterns of variability and inequality. Inequality and Aging: Stratification over the Life Course

The baby boom cohort provides a useful introduction to the study of aging and inequality because it provokes overblown images of homogeneity and agency. It has also inspired a body of work that tests the Easterlin effect (Easterlin 1961; 1987), the paradigmatically appealing hypothesis that has been critically reviewed by Pampel and Peters (1995). The Easterlin effect posits that cohort size is the major determinant of potential income. Larger cohorts face higher competitive pressures for limited resources and rewards. This basic inverse relationship influences cohorts' expectations regarding their future standard of living and thus motivates age-specific behaviors, including lower fertility and higher labor force participation, that affect the normal life course. Accordingly, the demographic characteristics of cohorts, especially relative cohort size, shape the life chances of individuals over and above other

Cohorts, Inequality, and Social Change

9

social forces including family, educational, workplace, and government institutions. The Easterlin effect has been tested on a wide array of age-specific behaviors ranging from educational attainment to consumption behavior, divorce, and crime rates. The evidence has been mixed—except for a robust relationship between cohort size and fertility observed between 1960 and 1980. The mixed results stem from the countering effects of two interrelated factors: social and historical contexts that attenuate or otherwise condition the influence of cohort characteristics and diverse pathways that emerge within cohorts and differentiate lives over time (Pampel and Peters 1995). Inequalities within cohorts more often exceed those between cohorts and stem from the interaction among structural conditions, individual agency, and chance. Structural conditions include dominant institutional factors such as the distribution of educational and employment opportunities and family roles and historical circumstances related to government policy changes or economic cycles. Agency stems from the day-to-day choices made by cohort members as they face normal life transitions such as marriage, educational completion, job entry, or retirement. Structural conditions and agency also affect the patterns by which cohort members encounter and manage probabilistic life events and life course risks such as illness, divorce, or unemployment. Status characteristics such as social class, gender, and race also serve to stratify chances and fortunes even further within cohorts, net of their relative sizes, since these characteristics interact with institutional structures to produce inequality over the life span, from birth to death.

Theories of Intracohort Heterogeneity and Inequality

The idea of intracohort differentiation conditioned by social structure and time (Dannefer 1987; 1988a; 1988b; Maddox 1987) has generated three major alternative hypotheses of intracohort heterogeneity that have become as paradigmatically compelling as the Easterlin effect. The status maintenance hypothesis posits that over time status effects are preserved across social transitions and status episodes in the life course. Resources and rewards obtained early in the life course, particularly in the transition from education to work, have persistent effects over time and serve to maintain individuals' relative status within cohorts. Educational attainment influences placement in the labor market, which in turn anchors future earnings trajectories and asset accumulation. Consequently, economic status in later life, such as at the time of retirement, is a linear function of earlier status achievement. In this vein, Henretta and Campbell (1976) find that the effects of education and occupation on income remain the same before and after the retirement transition among men.

10

Cohorts, Inequality, and Social Change

Educational and occupational status also influence other cumulative patterns of the life course such as marriage, which, in turn, has subsequent economic consequences for couples and individuals alike. Assortative mating in the marriage market is influenced by the joint employment and occupational status of women and men (Oppenheimer 1988; 1994). Educational homophily adds to the socioeconomic effects of individual educational attainment. Over time, economic inequality between married and unmarried individuals is maintained, if not accentuated (Oppenheimer 1994; Blau 1998). This pattern has been observed across industrial societies (Blossfeld 1995a). The hypothesis of cumulative advantage predicts increasing inequality within cohorts on the basis of initial advantage or disadvantage (Crystal and Shea 1990; Dannefer and Sell 1988; Dannefer 1991). Contrary to predictions from status maintenance theory that inequality among individuals within a cohort remains relatively constant over the life course, the cumulative advantage hypothesis predicts that economic transitions, such as the one from work to retirement, are critical temporal thresholds which increase relative inequality (Crystal and Shea 1990). The majority of studies focused on this hypothesis take cross-sectional approaches and report strong trends in inequality by age. Using three waves of the 1984 panel of the U.S. Census Bureau's Survey of Income and Plan Participation, Crystal and Shea (1990) report that income inequality increases with age. After age 44, income dispersion increases steadily. Inequality after age 64 is the highest for all ages. The Gini coefficient (which represents the percentage of the age-group that would have to change for income to be equal) for the age-group over 64 (Gini = .41) is 21 percent higher than that for the age-group between 35 and 44 (Gini = .34). Crystal and Shea (1990) look closely at the income distribution of the oldest age-group in their study. The lowest quintile (lowest 20 percent) of the income distribution is composed of women (71 percent), the widowed (51 percent), those living alone (58 percent), those with an elementary education (53 percent), and those in poor health (34 percent). They conclude that early advantage or disadvantage and early educational achievement patterns initiate pathways leading to late-life status. These pathways diverge in the employment system, in which cumulative advantage processes predominate. Workers with more resources (education, experience, skills) gain higher rewards (wages, benefits, higher skills); workers with fewer resources experience diminished value and status in the employment system. In a comparative cross-sectional study, Easterlin, Mucanovich, and Crimmins (1993) track income inequality across five-year age cohorts using two Current Population Surveys—1964 and 1985. They report similar patterns of increased income inequality with age. Gini coefficients for the oldest age-groups in 1964 and 1985 (aged 55 to 59) are higher than all other groups. This cross-sectional approach is limited in determining the

Cohorts, Inequality, and Social Change

11

mechanisms that produce inequality, but it nevertheless lends partial evidence for the cumulative disadvantage hypothesis. In Chapter 2, these limitations are explored further. Pampel and Hardy (1994a) find support for both the status maintenance and the cumulative advantage hypotheses in their study of a panel of older men followed between 1966 and 1981 (National Longitudinal Survey of Older Men). The effects of background status variables (race, education, and occupational status) remain significant in predicting market, transfer, and total family income among these men. But increases in inequality are nevertheless observed. Although status variables persist in their effects, market and private pension income distributions within educational groups become more unequal with age. Since this study ends with 1981, the effects of market changes in the 1980s cannot be compared. The status leveling or redistribution hypothesis predicts a reduction in relative inequality following the receipt of public transfers at retirement. In most industrial countries, inequality among the elderly is less than among the younger and adult populations largely because of the leveling influences of state policies. The status leveling hypothesis assumes that the stratification effects of the market are attenuated by the transition to income from other sources than earnings. Since retirement is defined by the change in income source from earnings to annuities from Social Security and accumulated assets and since the progressive redistributive structure of Social Security benefits serves to offset prior market inequalities, the leveling of income inequality among older age-groups is predicted (Crystal and Shea 1990). Although these three hypotheses offer predictions about the relative emergence and persistence of economic inequality as an outcome within cohorts, they also specify the intracohort process producing these outcomes. The increased differentiation in the characteristics of lives within cohorts is a cumulative process in which structure, agency, and chance interact over time. Early life transitions and status attainments influence—but do not completely determine—later ones. Timing and circumstances interact over time to move individuals along different pathways at different tempos. At any one time in the aging of a cohort, variations in status and economic inequalities are results of the interplay of social structures and individual biographies. Chapter 2 will examine the relative utility of these competing hypotheses in a more detailed review of the mechanisms producing inequality with age. Economic Inequality Across Countries and Between Age-Groups

The status maintenance, cumulative advantage, and redistributive hypotheses in life course research will reappear in later chapters of this book to identify inequality as a critical element of the aging process. But

12

Cohorts, Inequality, and Social Change

like the concept of cohort, the concept of inequality lends itself to alternative though equally legitimate definitions and measurements. Recent economic trends have stimulated considerable research on the arguable rise of inequality in the United States and elsewhere. Striking declines in wage equality in the United States and the United Kingdom and growing levels of sustained unemployment in continental Europe during the 1980s have generated much scholarship and debate on the origins and implications of these phenomena (Levy and Murnane 1992). Skill-biased technological change and a changing industrial mix (specifically the shift away from manufacturing and towards service employment in advanced countries) appear to have produced wage stagnation or unemployment or both in relative proportions across countries. In states with more active labor market policies and extensive welfare regimes, a mix of wage controls, unemployment benefits, and mandated job and family benefits constrains the level of income inequality. But in the United States and the United Kingdom, less active labor market intervention and less extensive provisions for the unemployed result in lower rates of unemployment but higher income inequality. However, in spite of differences of opinion on specific issues of measurement and the substitutability of wage inequality and unemployment in these comparisons, the evidence is sufficiently weighty, consistent, and persuasive on two matters: (1) economic inequality varies systematically across agegroups in cross-section and in cohorts over time and (2) income inequality has increased over recent years, especially in less regulated market economies like the United States. The post-World War II period has been divided by Frank Levy (1987; Levy and Murnane 1992) into two distinct economic periods. The first was the twenty-seven-year period from the end of the war until 1973 during which men's inflation-adjusted wages grew between 2.5 arid 3.0 percent per year. The second, between 1973 and the 1990-1991 recession, brought an average stagnation but also a widening distribution of wages. Since the recession and during a period applauded as one of unprecedented productivity, wage inequality has continued to grow. Levy (1998) attributes these recent trends to an acceleration in the U.S. economy's demand for skills and to a shift in power "away from the average worker" and "toward the shareholder." Until 1980, economists repeatedly identified a strikingly stable U.S. distribution in family income. Gini coefficients of family income—the average income difference between all pairs of families divided by mean income which yields a range from 0.0 (perfect equality) to 1.0 (perfect inequality)—changed little between 1949 and 1979. However, beginning in 1975 and sharply increasing after 1979, a reversal in this stable or slightly fluctuating trend occurred in the direction of higher inequality, measured

Cohorts, Inequality, and Social Change

13

as a jump in the Gini coefficient from a low of .348 in 1969 to .401 hi 1989 (Danziger and Gottschalk 1993, 7). A substantive interpretation of these measurements is that the distribution of family income reversed direction from a more widely shared base in 1969 to a narrower one in 1989. In 1989, families at the twentieth percentile received a 5 percent lower real income than in 1969, whereas families at the eightieth percentile were receiving a 19 percent higher real income. The trend toward increased economic inequality has been observed across advanced societies, but the increase in the dispersion of both individual earnings and disposable household income in the United States appears larger than almost all other countries. Table 1.1 is taken from Gottschalk, Gustafsson, and Palmer (1997, table 1.1) and summarizes the results from several different national studies of levels and trends in earnings and income inequality. Low, moderate, and high levels of inequality are based on Gini coefficients in the ranges of .20-.25, .26-.30, and .30 and over, respectively. Generally, earnings and income inequality have increased the most in the major market economies of Japan, the United Kingdom, and the United States, where public social protections are weakest. Countries with more centralized labor markets (Finland,

TABLE 1.1 Countries

Trends in Earnings and Income Inequality Among Selected Earnings Inequality

Australia Canada Czech Republic Finland France West Germany East Germany Greece Hungary Ireland Israel

Japan Netherlands Sweden U.K. U.S.

Income inequality

Level

Trend, 1983-1990

Level

Trend, 1983-199

High .Moderate Very low Low Moderate Moderate n.a. n.a. Very low High High n.a. Low Low High High

(+) Moderate (+) Moderate (+) Very high (+) Moderate (+) Slight (+) Slight n.a. n.a. (+) Very high No change (+) Very high n.a. (+) Slight (+) Low (+) High (+) High

High Moderate n.a. Low Moderate Moderate Low n.a. Low High High High Moderate Low Moderate High

(+) Moderate (+) Low n.a. No change (+) Slight (+) Slight (+) Very high (+) Moderate (+) Very high (-) Slight (+) Slight (+) Moderate Moderate (+) Moderate (+) High (+) High

SOURCE: Table taken from Gottschalk, Gustafsson, and Palmer 1997, table 1.1.

14

Cohorts, Inequality, and Social Change

Germany, the Netherlands, and Sweden) or more extensive income protection policies (France and Canada) have lower levels of and trends in inequality. Finally, Eastern European countries appeared to have had the lowest levels but greatest change in inequality in the 1980s. However, the political and economic transformations occurring by the end of 1980s have rendered more current comparisons of this kind difficult. Important distinctions between earnings arid income inequality have been observed across many studies replicating the general findings summarized in Table 1.1. Earnings are market outcomes: more centralized markets lead to higher equality of the wage distribution and less centralized markets produce greater wage dispersion. Of course, more centralized markets often benefit from supportive governmental policies related to taxation and transfer payments that indirectly influence wage distributions; in these instances state policies are influenced by strong labor representation and influence in the government as well as in the market. Income, on the other hand, is derived directly from multiple sources, including the market, the state, and the family or household. In a recent review of these studies, Gottschalk and Smeeding (1997) summarize recurrent findings across countries in patterns of earnings and income inequality. First, almost all industrialized countries experienced increased earnings and income inequality during the 1980s. Wage inequality among prime aged males increased in all countries except Germany and Italy. Household income inequality did not change as dramatically as earnings inequality across countries, but the inequality in men's earnings largely accounted for changes in income inequality. Second, the largest increases in earnings and income inequality occurred in the United States and the United Kingdom and the smallest in the Nordic countries. Third, although women's labor force participation, hours, and wages increased in almost all countries in the 1980s, wide differences between men's and women's earnings existed across societies at any given time. Modest increases in the association between husbands' and wives' earnings across countries contributed further to income inequality. And fourth, differences in the demand for and supply of skilled workers across countries accounted in large measure for differences in wage returns to education and experience. The rise of unemployment rates among the least skilled in some countries with more centralized wage constraints limited wage inequality in those countries. The relative position of the United States in the level of inequality among the population aged 65 years and older matches its rank in general earnings and income inequality. Findings from the Luxembourg Income Study (LIS), a database established in 1983 that contains social and economic data from household surveys and administrative records drawn from different countries in Europe and North America, indicate

Cohorts, Inequality, and Social Change

15

that the aged income and wealth distributions of the United States are among the most unequal of all. In addition, although the United States has among the wealthiest groups of aged married couples, it also has among the most impoverished older single women in developed countries. Finally, although in most countries inequality among the aged is significantly less than that among the nonaged, in the United States this is not the case, especially in recent years (see Atkinson, Rainwater, and Smeeding 1995; Smeeding 1997). Smeeding, Torrey, and Rainwater (1993) provide documentation for aged inequality from the LIS. TTiey compare eight countries (Australia, Canada, France, Germany, Netherlands, Sweden, the United Kingdom, and the United States) between 1979 and 1987 (with some variation in dates of measurement across countries). They examine several indicators of economic inequality, including age-related family income distributions, poverty levels, and wealth distributions. In nearly all comparisons, the United States is an "outlier," insofar as it persistently tends to exhibit higher levels of inequality. For example, Gini coefficients of household disposable personal income (adjusted for differences in family size and structure) reveal the highest values across years in the United States for all age-groups between 25 and 75+ and the largest increases across agegroups between 1979 and 1986. Gini coefficients for those aged 25-54 increase from .30 to .34 over the period, and for those aged 65-74 increase from .34 to .36 over the period. Other countries in the comparison (witli the exception of France among those aged 55-64 over the period) have lower Gini coefficients for all age-groups at each observation and either decrease or increase less in the levels of inequality (Smeeding, Torrey, and Rainwater 1993, table 3). Comparisons of poverty among nonaged and aged households also serve to distinguish the United States from other countries. Although tlie U.S. poverty rates tend to decline across groups over the period, they are nevertheless higher across period and across groups: over 13 percent of nonaged U.S. households are at 40 percent of median income across tlie period, a level more than twice that of all other countries with the exception of Canada. Elderly U.S. couples fall over the period from 8 to 6 percent poor using the same 40 percent of median income threshold, but they remain between two and six times more likely to be poor than their counterparts in other countries. Finally, single elderly women in the United States drop from 21.5 percent to 17.6 percent poor; by the late 1980s single elderly women in the other seven countries have less than a 4 percent chance of being poor (Smeeding, Torrey, and Rainwater 1993, table 6). Finally, economic inequality in the United States is distinctive when financial wealth is compared across countries. Wealth usually refers to

76

Cohorts, Inequality, and Social Change

assets minus debts or similar measures of potential income. Smeeding, Torrey, and Rainwater (1993) calculate the ratio of liquid wealth (i.e., property income from stocks, bonds, savings, and rent divided by .05) to adjusted disposable personal income across these eight countries in the late 1980s. They find the highest overall ratio of liquid wealth to income among the aged in the United States, but the ratio for U.S. elderly who do not exceed 50 percent of median family income falls sigiiificantly below that found in most other countries. In effect, the United States is a study in extremes: the elderly share higher average wealth, but greater income and wealth inequality and higher risks for poverty than most of their counterparts in other advanced industrial countries. All in all, the explanations for national variations in aged inequality are complex and varied. These variations ultimately reflect differences in labor market structures, welfare regimes, and family systems that regulate life-course processes of social mobility, status maintenance, and cumulative advantage. They also represent intercohort differences resulting from changing opportunity structures for earnings from employment, savings for retirement, and the changing composition of the labor force. Suffice it to say, distributional differences across these countries are readily observed and raise important questions about what these outcomes mean and how they have come about. The remainder of this chapter addresses these general questions about the processes producing aged inequality in the United States and elsewhere and how important differences in market and state policies bear upon education, work, and family pathways over the life course. Aging Cohorts and Economic Inequality

Age stratification theory suggests that age, cohort, and period have interacted in complex ways to produce relatively higher income coupled with relatively higher income inequality in the United States, when compared to earlier years in the post-World War II period. Different U.S. cohorts have shared the two periods characterized by Levy (1987) earlier, and they have been affected differently by them as a result of encountering these times at different stages of their lives and with different educational and experiential resources, Strong period effects have produced growing inequality. But other studies reveal age-, cohort-, and gender-specific patterns embedded in these general distributions. A cross-sectional study of income inequality in 1964 and 1987 using the Current Population Surveys for these two years and examining five-year birth cohorts born between 1915 and 1964 reports two cohort patterns identifiable during Levy's two periods (Easterlin, Mucanovich, and Crimmins 1993). First, age-group specific Gini co-

Cohorts, Inequality, and Social Change

/-

efficients of mean adult incomes show that inequality is higher in 1987 than in 1964 for all age-groups below age 55-59, although inequality below age 45 is generally higher than after age 45. Second, the cohort analyses demonstrate that whereas succeeding cohorts are initially more economically advantaged than their predecessors, postwar birth cohorts' income trajectories improve at lower rates as they age than earlier cohorts' trajectories. The slopes of baby boom trajectories reveal steeper increases in income through the mid-1970s and flatter trajectories afterward that coincide with the economically more depressed and uncertain 1980s. A related study using the 1989 Current Population Survey replicated some of these results, pointing to an average relative income advantage among baby boomers when compared to their age counterparts twentyfive years earlier (those 35-44 in 1964) (Easterlin, Schaeffer, and Macunovich 1993). This study reports that baby boomers aged 35-44 in 1989 had 73 percent higher incomes than their counterparts in the mid-1960s. But this study identified only the earlier portion of the baby boom cohort (born between 1946 and 1955 according to the study), and other scholars suggest that the fortunes of later baby boomers have not been the same (Radner 1995 1998). A more recent study by Crystal and Johnson (1998) replicates the pattern of higher relative income of baby boomers when compared to earlier cohorts at similar ages but supports the argument that the baby boom's fortunes are highly differentiated and unequal. This study employs the National Longitudinal Studies of Mature and Young Women, conducted between 1967 and 1992, and the Current Population Surveys (March files) between 1980 and 1991 to compare cohort differences in family income and pension coverage. The longitudinal studies follow two cohorts of women: one cohort aged 30-44 in 1967 and the second, 14-24 in 1968. Among its major findings is that within-cohort differences in formal education stratify baby boom earnings and pension savings outcomes. Only the most highly educated group of "leading edge" baby boomers (those born by 1954) can be expected with any certainty to fare better than preboom cohorts. Other portions of this cohort, including those less educated and those born after 1955, have more uncertain futures that are contingent on events such as business cycles, wage growth or decline, and labor market shifts that can deflect or enhance their work careers and their wage and retirement savings trajectories. Individual circumstances in their lives, such as illness, disability, or marital dissolution can differentiate them further. The relative contributions of women's labor participation and earnings to family income before and after retirement are also subject to within-cohort variation, according to this study (Crystal and Johnson 1998).

16

Cohorts, Inequality, and Social Change

Among "leading edge" boomers, couples will be more likely to receive retirement benefits based solely on husbands' earnings because wives in this cohort earned less than half of their husbands' earnings on average, and even less during their childbearing years. Those in later boomer groups have been confronted by the two-pronged phenomenon during the 1980s of wage stagnation in husbands' earnings and increased pension coverage of women versus men. But the latter trend does not compensate for the former, since women's lower average earnings translate directly into lower pension benefits.

Age and Gender Inequality

The history of labor force participation by women in the U.S. economy is not a simple upward slope of women's integration into the economy earlier in their lives. Rather it is a complex sequence of cohort compositional shifts moving through a changing opportunity structure in the workplace. The fundamental and persistent paradox is that whereas women's labor force participation rates have accelerated, their integration into the economy—if we can measure the latter in terms of wage parity, occupational integration, and pension coverage based on their work—has been slow. Women's average annual participation rates reached over 70 percent in 1995, after having increased 10 percentage points every decade since 1950 (Goldin 1990). Perhaps the single most important component of this change was a compositional one: married women's as opposed to single women's increased labor force rates have been the stronger driving force in the postwar period. Not surprisingly, one consequence of this pattern is greater gender parity within married couple families (Blau 1998), a pattern of inequality that appears to carry over into retirement populations (see table 1.2 in Smeeding 1997). Figures 1.1 and 1.2 present the labor force participation rates of selected synthetic cohorts of women and men, respectively, observed between 1965 and 1995 using the matched files of the March Current Population Surveys. The 25-29-year-old cohorts are selected in five-year intervals and followed every five years beginning in 1965 and ending in 1990. Thus, for example, the labor force participation rate of women aged 25-29 in 1965 is 40 percent; by the time they reach ages 40-44 their rate is 68 percent and remains at this level until ages 55-69, when it drops to 60 percent. Those 25-29 in 1975 begin at a rate of 56 percent that increases to nearly 80 percent by ages 40-44 and begins to drop. Similarly, men and women in the cohorts aged 70 and over are selected in the years between 1970 and 1995 and their participation rates are graphed. Thus, for example, women aged 70 or more in 1995 have labor force participation rates at ages 40-44 (in 1965) of 57 percent. Women aged 70 or over in 1985 have

^5

£

Cohorts, Inequality, and Social Change

21

a labor force participation rate of 50 percent when they were 45-49. Synthetic cohort participation rates over thirty years are represented. These cohorts are selected to provide several comparisons. First, the difference between women's and men's cohort-specific labor force histories is readily obvious. Men's intercohort patterns displayed in Figure 1.2 are more similar across cohorts, with sharper drops in labor force participation evident after age 55, especially following 1980. Women's intercohort changes are substantial and reveal several patterns. First, younger women's (ages 25-29) initial cohort rates increase significantly every five years from 40 percent for the 1965 cohort to between 75 and 80 percent for the 1965 cohort to between 75 and 80 percent for the 1985 and 1990 cohorts. When followed into middle age ranges, successive younger cohorts continue to increase their participation rates over time until they reach their 50s, when rates begin to drop, although when compared to older cohorts depicted to the right in the graph their rates at ages 55-59 are higher (at 60 percent). Second, older women cohorts begin steep declines in their rates after age 55, although successive cohorts reveal higher relative rates between ages 55 and 69. These data corroborate the already well-documented countervailing trends in older women's labor participation: increasing rates across the age span, on the one hand, and declines in these rates after age 55, on the other. The data also show increased variability in the second pattern across cohorts as more women in recent cohorts stay in (or perhaps return to) the labor force beyond ages 55 and 60. We will explore these trends and their structural and life-course bases in greater detail in Chapters 2, 4, and 6. Historically, women have tended on average to retire early. The primary underlying explanation stems from gender earnings inequality. Earnings inequality operates to affect women's retirement timing and economic status in several ways. First, they tend to retire with their (historically older) husbands, whose lifetime earnings have yielded spousal benefits (50 percent of husband's retiree benefits) from Social Security that have exceeded those based on many women's own earnings records even after achieving more than minimum participation histories. As mentioned in the earlier reference to the Crystal arid Johnson study (1997), this pattern persists among early baby boomer cohorts. Second, women's lower preretirement earnings result in relatively higher replacement ratios for them; the ratio of retirement benefits to preretirement wages is higher for women and minorities, and it encourages labor exit. Both the wage gap and the pension gap between men and women are highly associated with occupational segregation (Blau 1998). A substantial literature exists on the conceptualization, measurement, and estimation of

22

Cohorts, Inequality, and Social Change

pay inequity across gender, racial, and ethnic groups (e.g., Oaxaca and Ransom 1994; Blau and Kahn 1996). Generally, pay inequity is examined by comparing the differential returns in earnings/wages to a set of observable covariates measured across status groups (gender and race groups); the residual differences in returns reflect the "unexplained" component of wage inequality that may result from discrimination. Individual-level or human capital variables and job-level or labor market variables, on average, yield lower returns in earnings/wages to women and nonwhite workers when compared to men and white workers. These differences are typically revealed by the direct comparisons of coefficients across groups, including comparisons using simulations of crossgroup coefficients. The human capital variables that produce wage inequality in general include education, experience or work history, part-time/full-time work, and job tenure. These variables interact with status group membership (gender, race) to produce different returns in earnings and wages across groups. Childbearing history also differentiates gender-based returns to work; each child costs women more than men in earnings and pension income as a result of gendered care-giving roles (Bianchi 1995; O'Rand and Landerman 1984). The labor market variables that differentiate earnings and wage returns for all groups include unionization, firm size, and industrial sector. Union coverage, large firm size, and location in core or monopolistic industrial sectors are associated with higher earnings and pension coverage. Labor market variables that specifically differentiate gender and racial groups tend to be occupation-, industry-, and firm-specific. Occupational segregation by gender and race across firms and industries has been determined across many studies to be the major structural basis of pay inequality (Jacobsen 1994; Jacobs and Steinberg 1990; Sorensen 1989). Hartmann and Treimann (1981) have estimated that occupational segregation specifically accounts for 35-40 percent of the wage gap between men and women. Other occupational characteristics with wage outcomes include occupation-contingent employment rates and part-time /partyear worker rates, and job content. The Dictionary of Occupational Titles has been used over two decades to measure the intrinsic features of jobs, such as skill requirements, cognitive complexity, responsibility, authority/autonomy, and physical hazards, that translate into wage structures. In general, skill/cognitive complexity and authority/autonomy attached to the job tend to be associated with higher workplace rewards. Recently, Paula England has identified another characteristic that she labels as "nurturance," or the extent to which the delivery of personal care to clients or customers is a prominent feature of a job; she finds that nur-

Cohorts, Inequality, and Social Change

23

turance negatively affects the wages of both women and men and thus influences wage inequality (England et al. 1994). Pension inequity has been studied far less than pay inequity, although recent approaches adopt similar decompositional methods as found in the pay inequity literature to estimate pension income gaps and their origins. This literature finds that differences in cash earnings are the primary source of male-female pension inequality (Lazear and Rosen 1987) and that labor market characteristics (such as unionization, firm size, occupational segregation, and industry) are more important than individual-level characteristics for producing gender-based pension inequality (e.g., Even and Macpherson 1990). From Wage Cap to Pension Cap

The earnings-to-pension income trajectory is a critical mechanism of variability in middle- to late-life labor force participation and in economic inequality. It is widely known that the ratio of female to male median earnings among full-time workers remained relatively constant for most of the postwar period at ratios between 0.56 and 0.60 until 1980, when the gap began to narrow rapidly (Goldin 1990). By 1994, the overall ratio had reached 0.72 (Blau 1998). However, the apparent trend toward equality reflected in these average changes actually masks multiple underlying sources of variability or dispersion. First, men's wages stagnated after 1980 with obvious effects on this ratio. Second, the overall ratio was heavily affected by age-specific wage-ratio changes, especially after 1980; younger workers' wages were more equal than older workers'. And third, inequalities among women workers within and across age-groups grew, a trend attributed by Blau (1998) to an acceleration in the stratifying effects of variable education levels, which have also increased inequality among male workers, particularly across racial groups (Farley 1988). Figure 1.3 exhibits age-specific female to male weekly wage ratios drawn from the matched files of the Current Population Survey (CPS) between 1970 and 1995. In 1970, all four age-groups are positioned within range of the constant wage gap observed since 1950—somewhere between 0.50 and 0.60. The youngest and the oldest age-groups displayhigher relative parity, probably reflecting the effects of more equal levels of work experience within these groups. After 1980, the wage ratios of the youngest age-groups march strongly upward achieving a level above 0.80 by 1995. Also after 1980, although all age-specific wage ratios increase across observation years, a stronger inverse relationship develops between wage ratios and age. The age-group situated at the median

£

Cohorts, Inequality, and Social Change

25

overall female/male wage ratio of 0.72 in 1995 consists of women aged 35-44 in that year, born between 1951 and 1960. The "leading edge" of the baby boomer cohort falls among those in the adjacent age-group (ages 45-54) in 1995 (born between 1941 and 1950). They fall below the median wage ratio in 1995. Figure 1.4 presents a similar female to male comparison of pension income ratios drawn from the same data series. Pension income is measured as cash income from two annuities (employee pensions and Social Security). The general trend is toward a convergence of all age-groups in 1995 at ratios below 0.60. Wage ratios for 1995 (reported in Figure 1.3) for women and men aged 45-64 hover at 0.6. Pension income ratios for similar age categories in 1995 fall slightly below wage ratios. The relative transferability of wage ratios to pension ratios suggests a process of status maintenance—one in which wage inequality converts directly to pension income inequality. The 70+ age-group's relatively greater equality in 1980 is reduced by more than 20 percent over the period as it falls below the ceiling in 1995 along with other age-groups. Higher relative equality in the oldest agegroup in 1980 probably reflects two conditions characteristic of the period before the 1980s. First, in the early 1970s major adjustments to the Social Security system were enacted that had the eventual effect of cutting poverty rates among the elderly to less than half the levels observed in the 1960s. These changes introduced the expansion of old-age welfare and medical assistance to the poor and cost-of-living adjustments for all beneficiaries. Supplemental Security Income (SSI), a means-tested safetynet program to provide minimum benefits to elderly persons who did not meet service requirements, fell outside of the Social Security system during the employment years, or else received incomes significantly below poverty standards, began paying benefits in 1974. By 1980 older cohorts born in the first decade or two of this century were at far lower risk of poverty. Since these oldest age-groups are dominated by women, whose limited earnings histories placed them at risk of poverty, SSI and related legislation facilitated some equalization of income, including improved gender pension income ratios. Second, imtil the 1980s the proportion of the retiring workforce with pensions from their jobs was relatively small, with fewer than 40 percent of male workers in the late 1970s holding pensions and about half as many women workers. Those actually receiving benefits from employer pensions were below these coverage rates. Hence those receiving occupational pensions in earlier years are a highly selected group. Men's pension benefit rates and levels are still higher than women's today, although women's pension coverage rates on their jobs have increased while men's have decreased during the past decade (Woods 1989).

£

Cohorts, Inequality, and Social Change

27

These patterns have been accompanied by a change in the types of pensions available to workers. The new pensions are much more individualized instruments for retirement saving that depend more on workers' volition and contributions than in the past. Thus by the 1990s men and women who retire with private pensions receive highly variable annuity levels. This variability produces more inequality. The variability in pathways to inequality in retirement through workplace structures are discussed in greater detail in Chapter 2 and Chapter 4. Changes in pensions are discussed further in Chapter 5. Figures 1.3 and 1.4 demonstrate how, in the aggregate, earnings inequality translates directly into pension income inequality when age-specific female/male ratios are compared over time. In this examination, gender inequality is maintained by way of the transferability of ratios of inequality between earnings and retirement income. However, among the oldest age-groups (70+), who tend to depend relatively more on transfer income (i.e., Social Security) than any other retirement group, the level of gender inequality appears to be increasing. Should this trend continue, the prospects for moderating inequality through Social Security among the oldest old may be at risk. As Chapter 4 and Chapter 5 will emphasize, it is important to remember that women's participation in pensions has increased dramatically over the past two decades from quite low levels in the 1970s. Pension inequality is, in large part, historically contingent on women's recent access to and participation in pensions and their shorter pension accumulation histories. Furthermore, the coincidence of women's shorter history of pension participation with a change in the types of pensions offered to them may be restricted only to the United States (and perhaps the United Kingdom). Women's entrance into the employment sector has contributed significantly to the variability in U.S. labor force participation patterns and retirement outcomes. Employment Institutions and Inequality in Old Age

The centrality of the marketplace and the employment institutions associated with it for the production of intercohort and intracohort inequality is well established. Market work constitutes the productive core of industrial societies. Institutional arrangements in the workplace stratify work roles and reward them differentially. F,ven in countries with centralized markets and highly redistributive welfare policies, inclusion in the core employment sector is the principal source of economic and social advantage (Kohli 1988; Kohli, Rein, Guillemard, and van Gunsteren 1991). Workers excluded from the formal market sector or those situated precariously at its margins experience cumulative disadvantage in the

28

Cohorts, Inequality, and Social Change

acquisition of income and wealth over a lifetime (Farley 1988; Crystal and Shea 1990). In the most conservative welfare states they may derive fewer benefits as a result of their lower market contributions. Workers with limited labor force participation over a lifetime are not only represented in lower earnings or wage categories, but they are also excluded from income protection structures that take diverse forms in the workplace, including seniority, tenure track, promotional ladders, union-protected job security, and major employee benefits such as pensions and disability and health insurance. However, as noted in the previous section on gender inequality, even full-time workers in formal employment sectors may have unequal access to these income maintenance protections over their work life depending on their industrial, organizational, and occupational locations (O'Rand 1986). The modern workplace is highly segmented and has undergone nearly continuous transformation in recent decades. Finally, even the most privileged sectors in the workplace have been shaken over the past two decades as globalization, plant closings, organizational restructuring, downsizing, mergers, and contracting out, as well as employers' growing preference for contingent workers, have become less extraordinary events and more akin to business as usual (Belous 1989; Doeringer 1991; Pfeffer and Baron 1988). The employment contract between worker and employer that was associated historically with manufacturing model of industrial relations and more recently with bureaucratized workplaces (Jacoby 1985) is being renegotiated or dismantled (Kalleberg 1996). The implications of these changes for income security and retirement savings are far-reaching. The loosening of workplace contracts has introduced more variability in the patterns of work careers and earnings acquisition, including increased job shifts and employer changes, early and flexible retirement arrangements, and increasing rates of self-employment (e.g. Doeringer 1991; Belous 1989; Henretta 1992). Chapter 2 and Chapter 4 examine the variability in employment pathways that these structural changes are producing and their implications for increasing individualization in the work career and patterns of aged inequality. New pension structures were among the first indications of this changing contract. The Employee Retirement Income Security Act of 1974, referred to as ERISA, was enacted to protect workers' employment pensions from mismanagement and malfeasance by employers. Employers were made liable for their pension promises. But this legislation coincided with new market uncertainties emanating from international oil crises, the competitive threat raised by Japanese and European enterprises, especially in the automobile industry, and the new realities of globalization. Among these new realities were cheaper offshore labor

Cohorts, Inequality, and Social Change

29

forces, just-in-time management systems that discouraged the maintenance of large inventories and related physical plants, and the greater integration of information technologies into the production, coordination, and distribution processes (Doeringer 1991). As Chapter 5 will show, these events encouraged the development of new pension instruments that relieved employers of pension liabilities by shifting more responsibly for pension saving to workers (Gustmann, Mitchell, and Steinmeier 1994). The major change in the retirement savings structure has been the rapid growth of individualized pension instruments that are largely represented by what are called defined contribution plans. These plans often supplement the traditional defined benefit plans in core industrial sectors, but they have become the exclusive plans provided by employers in the most rapidly growing sendee and communications industries and small firm sectors, in which much job creation has taken place over this period (Birch 1987). Worker-centered savings and investment plans are portable tax shelters that offer cash surrender a n d / o r loan options. They give the worker more control of retirement savings and investments, but they also require more worker knowledge and responsibility. These options make the defined contribution plans attractive for their availability to handle shortterm major expenses such as hospitalization, new home purchases, and college costs, but problematic as certain sources of retirement annuities when workers need them. Their portability encourages job mobility. Their accessibility as cash discourages long-term saving. The responsibility placed on workers to manage these plans is another source of their growing individualization. State Policies, Inequality, and Aging Welfare states are systems of statutory social protection that are more or less developed around market versus citizenship rights. A number of typologies and theories of the welfare state have developed over the past half century to describe the alternative forms taken by public and private structures for the allocation of social goods such as education, health care, and pension benefits and for the protection of vulnerable or "naturally" dependent populations. The earliest theories (Titmuss 1958) were based on the "logic of industrialism" thesis or the related "logic of capitalism" thesis, which generally posits that welfare structures are inevitable outgrowths of the market and hence are residual structures developed to preserve and reproduce market relations. Later theories have either revised or refuted earlier ones and have adopted a diverse array of institutional arguments for the development of these structures, including class coalitions (Esping-Andersen 1990), relative class power (Myles

w

Cohorts, Inequality, and Social Change

1989), the social organization of production (Quadagno 1984), "structured polity" (Skocpol 1992), and patriarchy (Orloff 1993; Sainsbury 1996). Esping-Andersen's "three worlds of capitalism" and more recently developed feminist (patriarchy) theories (Sainsbury 1996; Orloff 1993) provide complementary conceptual frameworks in Chapter 7 for comparing basic differences between the United States and other countries in the level of special expenditures and the provision of welfare benefits. Esping-Andersen has empirically developed three ideal types of welfare structure: the liberal model based on market efficiency found in the United States; the conservative corporatist model based on status allegiances developed in Germany; and the social democratic (or citizenship) model of the Nordic countries. The relative emphasis of U.S. welfare policies on market-based welfare rights, means testing, particularistic needs assessment, and minimal levels of social insurance results in higher levels of economic inequality among older populations, including higher rates of poverty. Women's benefits stem largely from a "breadwinner" model of entitlements (Sainsbury 1996) that assigns them to derived benefits. The Social Security system's benefit structure is biased toward married couples with traditional gender roles (Burkhauser and Smeeding 1994). Dual career couples receive relatively less social protection than traditional breadwinner couples or dual-earner couples in which wives are secondary market workers (Gilbert 1994). Alternatively, conservative corporatist states like Germany allocate more generous non-market-based benefits to abate risks of falling into poverty. Gender-sensitive policies in Germany recognize family statuses such as motherhood as legitimate bases for social protection. Consequently, childbearing and care giving are less likely to present risks for falling into poverty over the life course in Germany. Finally, social democratic states like Sweden allocate benefits universally on the basis of citizenship and actively participate in the wage-setting process, reserved strictly for the private sector in the United States. Strong life-transition benefit systems, generous transfers to children, guarantees of employment, and the absence of tax and benefit penalties for part-time work make its gender-sensitive policies the most liberal in the world. All three systems are confronted in the 1990s with fiscal, economic, and political problems challenging current policies. Population aging and persistently high rates of unemployment provide the major sources of strain for European systems (Pampel 1998; Bosworth and Burtless 1998). In the United States, concerns are more prospective and related primarily to population aging (with concerns regarding the long-term care of the oldest old, particularly single aged women), the personal savings pat-

Cohorts, Inequality, and Social Change

31

terns of workers, and the looming exit from the labor force of the baby boom cohorts (Rappaport and Scheiber 1993). Meanwhile, global forces are driving the increase in economic inequality, especially in the United States. Conclusion

This chapter began with the simple idea that aging cohorts are not homogeneous. The sheer size and social impact of the baby boom cohort in the United States over the years has led to its popularization as a monolithic, homogeneous, and exogenous force on American society. But this characterization contradicts what life course and stratification theories have revealed about the process of social aging. Although cohort members may encounter similar institutional conditions and historical events as they age, they do not carry away the same experiences with these conditions and events. They do not share the same outcomes as a result of their experiences. They all do not end up in the same place. Over time cohort members lives' in industrial societies become differentiated. Hie process begins with socioeconomic origins and early educational attainment. The middle passages of life include the differential engagement of cohort members with family, state, and work institutions. Later life is the outcome of these cumulative processes. Along the way, the interplay of lives with social institutions produces variability and inequality. Three theories capture the distinctive but countervailing elements of these multilevel processes in industrial societies: status maintenance, cumulative advantage, and leveling. The first refers to patterns of stability of inequality within cohorts over time. Early inequalities persist on average in relatively stable ways into old age. Educational attainment is converted into wages. Wages are converted into assets and pension wealth. Thus gender and class inequality over the life course reflect strong status maintenance patterns. Cumulative advantage characterizes patterns of divergence or increased inequality over time. In market-centered societies advantages and disadvantages can be compounded over time, particularly at critical life transitions like retirement. Gender and class groups are differentially vulnerable to the loss of status or income over the life course resulting in variable risks for poverty. The sources of differential vulnerability are structural and biographical. Educational systems, workplace institutions, and state protective and redistributive policies stratify opportunities and constrain choices over individuals' lives, with cumulative effects. State institutions with welfare functions operate to offset (or level) these two patterns by redistributing resources from the more advantaged to the less advantaged, although the degree to which

12

Cohorts, Inequality, and Social Change

the state intervenes to ameliorate market-based and gender-based inequalities varies considerably across welfare regimes. The complex operation of all these processes are observable across societies, but they differ in their pervasiveness. The United States has among the wealthiest but also the most socially differentiated and economically unequal populations when compared to Europe and other North American coiuitries. Presently, as a surge in inequality appears to have spread to all market economies, U.S. cohorts—including the baby boom cohorts in the middle passage of life and their parents, who largely make up the retirement population—are among the most unequal. Labor market institutions, family patterns, and welfare policies in the United States interact in ways to increasingly differentiate and stratify cohorts along life pathways. The force of differentiation produces the increased individualization of the life course, a process whereby relatively universal experiences such as retirement occur under highly variable circumstances and on different schedules for different segments of the population. The liberal welfare regime in the United States demarcates it from other welfare regimes because it intervenes minimally in the marketplace. In the United States state policies are often asynchronous with market policies, with the effect of sorting lives into different tracks. Other countries have developed welfare systems that, for historical reasons, redistribute resources in different ways to offset potential and real inequalities that stem from the market and from dislocations and derailing life events that can deflect the life course at the individual level and contribute to increased relative inequality. Social democratic states like Sweden are much more equal as a result of stronger interventions in the market. Conservative corporatist states like Germany fall between the United States and Sweden in the level of inequality across the life course into old age. The remainder of this book will develop these themes further. Intracohort differentiation and inequality, segmented pathways, life course variation and asynchrony, and individualization are elements of social aging that achieve distinctive forms in the United States. Yet the United States and other Western countries face common challenges to the preservation of their respective systems. Population aging is a worldwide phenomenon pressing upon public policies that developed under very different demographic and historical circumstances. Globalized market forces are exerting pressures for individualization and privatization that directly oppose welfare structures and ideologies. All of these patterns are converging across countries to produce interesting times.

2 Pathways to Inequality: Intracohort Differentiation over the Life Course

Men and women wrho begin life similarly situated may end it in quite diverse circumstances. There is great human drama as individuals learn their destinies, particularly when life brings a major reversal. Stable trajectories in work, health, or family life can alter course in a moment. On a smaller scale, time and chance do indeed happen to us all—raising the question of continuity in individuals' lives as they age. The uncertainty of life is an inexhaustible topic for reflection, but a sociological approach to changing life course trajectories focuses instead on the role of social institutions in producing continuity or discontinuity over the life course. We pose the issue in this chapter as one of age-based intracohort differentiation produced by social institutions: What are the implications of a cohort's shift from a life stage in which the regime of daily life is dictated by work institutions to a stage dominated by retirement institutions? Does this shift produce continuity in individuals' lives or discontinuity and shifting patterns of inequality? By retirement institutions, we mean public and firm policies affecting income and other entitlements of the older population. Retirement institutions vary in the extent to which they use age or employment status as a trigger for entry; thus older people who are employed may be affected by some retirement institutions—for example, the age-based Medicare program—even while they are still employed. Both employment and retirement institutions vary over time, producing differences between cohorts in the experience of aging (see Chapter 1). Change that occurs between cohorts, making the experiences of each cohort distinct from earlier or later ones, is the standard topic of cohort 33

34

Pathways to Inequality: Intracohort Differentiation over the Life Course

analysis. There is equal conceptual interest in the pattern of within-cohort differentiation with aging. The importance of intracohort differentiation derives from its focus on the role of social structure—that is, institutional rules and process—in producing increasingly more or less equal outcomes as a cohort ages (Dannefer 1987; 1988b). Of course, changing employment and retirement institutions imply that intracohort change will itself vary among cohorts. In examining the implications of a cohort's shift from work institutions to retirement institutions, this chapter poses several questions: Are there identifiable processes associated with this shift that create divergent trajectories among cohort members and either accentuate or reduce inequality with aging? What patterns of inequality do we observe among the elderly? Do these patterns represent continuity from earlier life, accentuation of earlier inequality, or a late-life leveling? We begin by discussing the nature of the linkages across this life course division. Aging and Social Theory: Linkages Between Work and Retirement

Individual pathways become problematic across the work arid retirement phases of the life course because retirement means exit from employment and entry into retirement institutions. That is, the retirement transition involves major change in the primary institutional structure within which individuals lead their lives, and it is not immediately clear what might produce continuity in individual lives across these two structures. Just as sociologists have examined the linkages between the preparation and work phases of the life course that also involve a shift in institutional structures between school and work (e.g., the links between parental resources, individual investments in education, and important socioeconomic outcomes such as occupation and income), it is equally important to examine the strength of socioeconomic linkages between the work and retirement phases of life. Martin Kohli, in an important paper entitled "Aging as a Challenge for Social Theory" (1986), analyzes the issue in a particularly useful way. Though the boundary between work and retirement may have become less clear in recent years (see Chapter 4), over the longer term the development of retirement institutions has created retirement as a distinct stage of life. In this modern life course division into separate age-segregated education, work, and retirement stages, each life segment occurs within a distinct set of social institutions (Riley 1998). The two nonwork periods of the typical life course, education and retirement, have expanded. The retirement period, particularly, has expanded at both ends because of declining retirement age and declining mortality at advanced

Pathways to Inequality: Intracohort Differentiation over the Life Course

35

ages (Wise 1997). Yet beginning with nineteenth-century attempts to understand the nature of modern society, sociological theory has generally held that the most important institutional structures affecting individuals' lives arise from work institutions. The challenge to social theory is to conceptualize the role of work in modern social life, given the reduction in lifetime hours spent in employment. Is work still central in defining social position and meaning in life? For the study of aging, meeting this challenge requires conceptualizing the meaning and structure of retirement since retirement exists outside the sphere of production and employment and, unlike education, is not preparation for work. Kohli's solution is to argue that Western industrial societies remain "work societies" because, despite the nonwork character of retirement, work remains the central factor determining retirement economic and social status. Although the governing institutional structures may shift, the meaning and significance of retirement is created by the work phase of life, and a central task is understanding the intraindividual mechanisms linking the phases of the life course. Kohli suggests the linkage across the divisions of the life course may be found in the "biographical conception of class," a time-based view that focuses on linkages within individuals' lives and connects the work and retirement phases of life despite their separate institutional contexts. "Class" refers to much more than the individual's economic and social position. It implies an individual's self-concept and subjective understanding. The significance of this idea is that it focuses on intraindividual linkages across time as a central mechanism producing a type of continuity in individual lives between work and retirement phases. The result is not a perfect reproduction of the preretirement phase but continuity in individual lives. For example, two workers with similar current occupations (and incomes) may differ in their pension entitlements. After retirement, they will not have equal incomes (barring higher savings by the worker without a pension, a topic considered later). A woman's early pattern of employment and childbearing will affect both retirement timing and economic status after retirement. The result is not simple continuity of status because the sources and institutional rules governing income, as well as daily activities, change. But there are clear intraindividual links between work and retirement. At first glance, the idea that individuals are constrained by past events appears to be a commonsense view lacking profound theoretical significance. Yet Kohli argues that this approach is quite distinct from some traditional gerontological approaches which assume that old age is a break with the past—a new stage of life in which the distinction between work and nonwork determines interests and values. The implicit view is that retirement means abandonment of the commitment arid meaning created

36

Pathways to Inequality: Intracohort Differentiation over the Life Course

during the work phase of life because it means that a different set of institutional structures are immediately relevant. Kohli discusses two varieties of these older ideas. One focuses on economic interests, arguing that nonwork gives the elderly common economic interests in public programs. A second alternative approach focuses on cultural values instead of economic interests, arguing work versus nonwork is the central distinction shaping values and attitude. The view that shared leisure is an important element creating a "class" was pioneered early in the history of social gerontology by Arnold Rose (1960), who argued that older people in industrial societies had become increasingly separated from the institutions of employment and family that linked them to the social world of younger people. This isolation provides the opportunity to become more aware of their common interests, including shared economic interests, and perhaps to develop political movements based on them. Kohli concludes that public transfer programs and shared leisure do create some shared interests among the elderly, but these shared interests are understood and experienced by individuals in different ways; there are important links between past and present in the lives of individuals that shape the nature of aging. Hence, the emphasis on continuity and linkage across phases of life, despite the new institutional structure, defines a distinctive view of the nature of retirement. The concepts of continuity and linkage in individual lives forms the core critique of both traditional views that define old age as a new phase of life.

Research on Linkages Across the Life Course Status Maintenance, Status Leveling, or Cumulative Advantage?

The empirical literature addressing age-based changes in economic status also reflects the same distinction between views that focus on continuity and discontinuity across phases of the life course. The contrasting patterns of continuity and change between work and retirement phases are conceptualized by Henretta and Campbell (1976; 1978; Campbell and Henretta 1980) through the processes of status maintenance and status leveling. The view that there are strong links in the individual determinants of status between work and retirement is captured by the concept of relative status maintenance. Occupation and industry placement produce careers with higher income, greater pension wealth, and increased accumulation of other types of economic assets. Both pensions and accumulation of assets are important linking mechanisms that tie work and retirement together. That is, the factors that predict market success are equally strong

Pathways to Inequality: Intracohort Differentiation over the Life Course

37

in predicting retirement income. An alternative hypothesis, status leveling, holds that there are relatively weaker links between occupational attainments and income after retirement than before because public policy attenuates the link between life phases. The structure of public programs, such as Social Security retirement benefits and Medicare, minimizes the effects of past earning history on benefits. Hence, to the extent that these sources of income are important, resulting retirement income will bear less relationship to education and occupation than preretirement income. The leveling view parallels the description of older gerontological theory presented by Kohli (1986) in its contention that retirement is a new stage of life subject to its own unique rules. Status leveling does not argue there will be no link between early attainments and retirement income, only that it will be severely attenuated in old age. More recently. Crystal and Shea (1990) have presented a third conceptual possibility—"cumulative advantage." They argue that the shift from employment to retirement institutions means an increase in the relative importance of those income sources that are most unequal. Hence, they argue, the retirement transition is associated with increasing cohort differentiation. Varying conceptualizations of the work-retirement link create an important agenda for research: what are the processes and mechanisms that produce linkages between events across time, despite change in work status? Does the change in work status lead to status leveling, status maintenance, or cumulative advantage? Henretta and Campbell (1976) compare income attainment path models for the same cohort before and after most of its members have retired. Their results provide strong support for the relative status maintenance hypothesis. In examining repeated cross-sections of one cohort, they find that the link between education, occupation, cuid income is very similar before retirement and after. Hence, education and occupational attainments are as useful in predicting income during both periods. They conclude that the linkage across time between statuses is not attenuated by a shift from work to retirement institutions and suggest that the linkages may be found in Social Security, firm pensions, and individual savings (Henretta and Campbell 1976; 1978; Campbell and Henretta 1980). Pampel and Hardy (1994a), using more recent panel data, have confirmed the general similarity of education and occupation effects on income before and after retirement. Although the research on continuity of status effects generally indicates status maintenance, researchers who have examined inequality of income in retirement have generally concluded that inequality increases with age (in cross-sectional data). Hurd (1990) reviews estimates of inequality among the elderly and nonelderly for various years between 1973 and 1984 and shows greater inequality among the elderly, measured

38

Pathways to Inequality: Intracohort Differentiation over the Life Course

by the Gini index. This difference is stable over various adjustments in income. Crystal and Shea (1990) similarly argue that there is increasing inequality with age; hence there is a pattern of "cumulative advantage." They find that Supplemental Security Income (SSI) and Social Security tend to reduce inequality but this effect is not adequate to counterbalance the inequality produced by great disparities in asset income. These findings have been elaborated and partially challenged recently by an analysis that shows complex and shifting patterns of inequality. Radner (1995) examines inequality among the aged and nonaged between 1967 and 1992. He finds that inequality in the under-65 population, as measured by the Gini index, increased monotonically during the period; this increase in inequality is also generally reflected in the shares of income going to the bottom and top 20 percent of the population. Among elderly households, the overall effect of a more complex pattern of change is that there is less inequality among the elderly in 1992 than in 1967. But there is more inequality in 1992 than in 1979, findings also reflected in examining income shares. Until 1992, there is clearly more inequality among elderly households than among younger households. In 1992, an analysis of the Gini index suggests more inequality among the elderly, but an income share analysis does not yield a clear conclusion. Radner finds that elderly and nonelderly inequality levels appear to be converging for two reasons—the increase in inequality among nonaged households and the decline in aged inequality before 1979. Pampel and Hardy (1994b) address the seemingly conflicting findings that inequality is usually greater among the elderly and the relationship of income to background status characteristics is not affected by retirement. They find that inequality in total family income increases in panel data as the cohort moves from preretirement to postretirement sources of income. However, most of the increase in inequality occurs within status categories (i.e., within levels of education and occupation) and relatively little occurs between status categories. The within-group increase in inequality may be due to random, unexpected evexits, or it may result from variations in social structure. For example, firm pensions are correlated with education and ocaapation, thereby producing stability in the effects of these attainments on income. But pension variation within education or occupational groups might well produce more diverse within-group outcomes in postretirement income than in preretirement earnings. The overall conclusion is that the role of status background characteristics in differentiating the population is relatively stable across the work and retirement phases of the life course. This finding is partly due to the correlation of opportunities to accumulate resources for retirement through pensions and savings with education and occupation. However, it is equally important to consider the role of individual agency in creating a link between work and retirement.

Pathways to Inequality: Intracohort Differentiation over the Life Course

39

Individual Agency as a Link Between Work and Retirement

In economic theory, an important link between status in the work and retirement phases is active planning by the individual to spread lifetime income across years of work and nonwork. The standard life cycle framework widely used in studies of saving and consumption assumes that individuals consciously plan for the future in an attempt to spread their income over their expected lifetimes and keep the marginal utility of expenditure equal over time (Hurd 1990; Browning and Lusardi 1996). Saving in one period allows higher consumption later, though the resulting predicted behavior is quite complex. First, an attempt to maintain equal utility is consistent with different levels of expenditures at different times. Higher expenditures when children are young, for example, may be required to produce the same utility as lower expenditures when household size is smaller (Browing and Lusardi 1996). Consumption may also vary because it is not always possible to borrow against future income (Hurd 1990), and future consumption generally doesn't have as much present value as current consumption. Second, individuals may not wish to spend all their income because they wish to leave an inheritance or guard against unexpected events. Third, individuals with equal savings and lifetime income may consume at different rates because they expect to live for different periods of time. Fourth, unexpected events may foil the individual's plan. Despite these conditions, the economic view is that individuals actively plan and execute, and their general goal is to maintain their status over time. Critics of the model have argued that it ignores the limits to rationality inherent in human behavior. Individuals can't plan in the way the theory suggests; and, even if they could, they lack the self-control to execute the plan (Thaler 1990; 1994). Bernheim and Scholz (1993), for example, argue there are socially patterned differences in the learned skills required to make the complex long-term planning decisions. They find that the standard life cycle framework describes the behavior of college graduate households but not those with less schooling. Sociologists are likely to find this critique congenial. To the degree that individuals can't or don't plan, a sociological model in which exogenous factors—such as firm-based pension policies or parental status that affects the individual's education—become more important. In other words, if individuals have limited ability to plan forward, what happens to them—exogenous factors—not what they make happen, becomes more important. Still, sociological approaches often underplay the role of individual agency in creating or shaping outcomes. Attention to the life cycle model is a useful corrective. The applicability of the life cycle framework to the behavior of the elderly is a continuing topic of research. Hurd's (1990) review finds a range

40

Pathways to Inequality: Intracohort Differentiation over the Life Course

of behavior that is generally consistent with the model, but other research is inconsistent. For example, Bernheim, Skinner, and Weinberg (1997) find a sudden drop in consumption at the time of retirement and pre- and postretirement consumption patterns that they argue are not consistent with the range of factors usually considered in the life cycle framework. Despite the unsettled research issues, the life cycle framework has important implications for examining the link between status before and after retirement. First, it implies that researchers should define income in a way that combines current income and the potential lifetime income from assets (including human capital)—instead of examining realized income in any one year as has generally been the procedure in sociological research. There are a number of compelling reasons for developing a comprehensive measure of well-being instead of using current year's income only. A central mechanism through which individuals spread income across their lifetime is through the accumulation of assets. The current income realized from investments is not a good proxy at later ages because the annuity value of an asset is generally greater than realized income, particularly at later ages when assets need finance consumption over fewer years. Even if assets do not provide current income, they still add to well-being since they provide a source of potential income and, hence, material and psychological security. The limitation of the comprehensive measure combining income and the full annuity value of assets measure is that asset classes may have different social and personal meanings to their holders. For example, housing equity arising from home ownership has distinctive meanings separate from its monetary value (Henretta 1987). Either for social or economic reasons, older people are less likely to use housing equity to finance daily consumption (Hurd 1990; Venti and Wise 1990; 1991). Although there remain good reasons for using comprehensive income measures, they raise complex issues of which assets should be included, how they should be valued, the period over which assets should be apportioned, and the relation between the comprehensive measure and actual behavior (Radner 1990). The emphasis on spreading lifetime income, which for most persons is produced by earnings, over the entire life course very strongly implies a status maintenance outcome. Research confirms this expectation; for example, Burkhauser, Butler, and Wilkinson (1985) find a .70 correlation between pre- and postretirement comprehensive income. Yet there are also some findings in the life cycle framework that are consistent with what one might expect in either the leveling or cumulative advantage approaches. For example, Kotlikoff, Spivak, and Summers (1982) find that Social Security benefits raise the standard of living of the aged population above the level they could have financed through their lifetime economic resources. Since public programs benefit low earners more, a topic

Pathways to Inequality: Intracohort Differentiation over the Life Course

41

discussed later, this finding suggests the possibility of leveling. Hurd and Shoven (1985) use a comprehensive income measure that adds the implied insurance value of Medicare and Medicaid to examine comprehensive income distributions over time (but not within individuals). Overall, median and mean real incomes declined; however, Hurd and Shoven's results also show that the real income of those at the very bottom of the income distribution actually increased because of the very strong effect of Social Security, Supplemental Security Income (SSI), and Medicare benefits for persons at low income levels. For very low income persons, Social Security or SSI constitutes the most stable income source of their lifetime. As a result, aggregate declines in income were slightly greater for those with higher incomes, since Social Security makes a higher proportion of the wealth of the low income population. In examining changes between 1969 and 1979, they find a small trend toward equalization (for a similar result, see Burkhauser, Butler, and Wilkinson 1985). On the other hand, there are some results suggesting the possibility of cumulative advantage. Bernheim and Scholz's (1993) results, discussed above, indicate that more educated households accumulate more assets iii relation to income than do less educated households. This finding is consistent with earlier research showing that, controlling for long-term earnings, those with greater schooling have higher levels of assets (Henretta and Campbell 1978; Campbell and Henretta 1980; Kotlikoff, Spivak, £ind Summers 1982). That higher asset position may be produced by more actual saving out of current income (implying lower consumption while working) or greater investment skill. In the former case, cumulative advantage would have been financed by lower consumption earlier-—thereby tending to equalize total lifetime consumption between less and more educated households at the same income level. Still, these results for education suggest different pre- and postretirement consumption trajectories that conform to the cumulative advantage argument. Summary

Over 20 years of research on the linkage between the work and retirement life course phases allows some robust conclusions on inequality. First, there is a relatively strong correlation between rank ordering of workers before and after retirement, indicating that there is strong carryover of economic well-being over the phases of the life course. Income is highly correlated across the two periods, and the relation of income to status background characteristics such as education and occupation remains stable. The mechanisms producing this continuity involve either the effects of an exogenous social structure which provide different opportunities to those differently situated (as indexed by education and

42

Pathways to Inequality: Intracohort Differentiation over the Life Course

occupation) or individual agency through the mechanisms captured in the life cycle saving framework. Exogenous social structure and individual agency may also provide a combined mechanism. The argument that college graduates have a social environment more conducive to learning effective saving and investment skills provides one example. Second, there is good evidence for an increase in intracohort inequality after retirement, though the difference may be less now than in the past, and some research shows a leveling produced by public programs. Different results may stem from the sensitivity of the various methodologies used to particular aspects of the income distribution. The literature addressing inequality nearly always makes the point that public policy attenuates the linkage across life course phases while market mechanisms accentuate it. Government transfers (e.g., Social Security) tend to reduce inequality. Other types of wealth—pension wealth and financial or real assets—tend to either perpetuate or increase inequality. To examine this issue, we now turn to an analysis of income sources. An Overview of Income Sources Examining retirement income sources and the links those sources create between the work and retirement phases of the life course provides one way to examine status shifts. Social Security benefits are the most common income source for elderly (ages 65+) households—over 90 percent receive Social Security income, followed by 67 percent receiving income from assets, 30 percent with pension income from private employers and 14 percent with public employee pensions, and 22 percent with income from earnings. The distribution of these income sources varies by income level. The lowest 20 percent of money income recipient households rely on Social Security and, secondarily, assets and public assistance; upper income quintiles are more likely to receive pensions, asset income, and earnings (Grad 1996). Means-tested public assistance benefits such as Supplemental Security Income are received by 6 percent of elderly income units (Grad 1996; Bureau of the Census 1996); they are particularly important in the bottom quintile of income recipients, where they are received by 20 percent of households. Table 2.1 (derived from Grad 1996) presents the proportion of aggregate income from the various sources by marital status and income quintile. The table indicates great heterogeneity across marital status and income level. For both the married and unmarried, Social Security benefits are relatively less important at higher income levels, whereas pensions, assets, and earnings are more important aggregate income sources. Clearly, higher income elderly households have that position because of their asset holdings, pensions, and continued earnings.

TABLE 2.1 A g g r e g a t e I n c o m e from Each I n c o m e Source of Income U n i t s A g e d 65 and O v er, 1990, b y Marital Stat us a n d Income Quintile Married Couples

Total

Retirement i n c o m e : Social Security Railroad retirement Gov't employee pension Firm p e n s i o n s / a n n u i t i e s Asset Income: Earnings Public assistance Other

37.6 .6 8.7 10.6 16.6 22.9 .4 2.6

Income

Nonmai ried persons

Quintile

Total

1

2

3

4

5

Men

Women

82.1 .3 1.7 2.9 4.2 2.9 3.9 1.9

70.9 .6 4.1 7.9 7.2 7.1 .6 1.6

53.4 1.0 7.5 12.5 11.8 11.4 .3 2.0

36.5 1.4 11.2 14.1 15.8 18.7 .1 2.3

18.7 .1 10.0 10.0 22.7 35.3 0.0 3.3

41.7 .7 8.7 11.7 19.6 13.8 .9 2.9

52.4 .6 7.6 6.9 18.9 8.8 2.0 2.8

Income

Quintile

1

2

3

4

5

78.3 .6 .8 1.2 2.6 .5 14.3 1.8

85.1 .8 1.2 1.9 3.4 .8 5.2 1.7

78.3 1.3 3.2 4.9 6.6 2.8 1.1 1.9

56.6 .8 7.4 11.0 13.8 6.3 A 3.6

26.2 .4 11.9 10.4 30.3 17.4 .2 3.2

Quintile limits for married coi .iples: SI!5,040; 2:1,366; 29,!586; 44,801. For u nmarriei 1 persons: $6,239; 8,819; 12,324; 19 ,094. SOURCE: Derived from tables \ TL3 and Vn.5 in Susan G rad. 19% . Income of the Population 55 or Older, 1994. Was tungton, D.C.: Sjcial Seen rity Adir linistrat:ion.

^5

44

Pathways to Inequality: Intracohort Differentiation over the Life Course

lit addition to these general patterns, there are important differences across marital status. Unmarried households tend to be older than the married and are disproportionately female and widowed. They depend more on Social Security benefits arid means-tested public assistance and less on earnings compared to the married. Unmarried women receive less of their income from pensions. The incomes of unmarried households are generally very low. The bottom 60 percent of the unmarried have incomes that would fall in the lowest fifth of the married group. Overall, 80 percent of unmarried households have income lower than the bottom 40 percent of married households. It is not surprising, therefore, that the bottom three quintiles among the unmarried have aggregate income sources that are roughly similar to the bottom 20 percent of the married. The relative importance of income sources has changed somewhat over time. Earnings provide a smaller part of total income, and assets and pensions provide more. The picture is complex, however. The role of assets in aggregate income has fluctuated over time, and Reno (1993) suggests the fluctuation in the asset share of income is affected by interest rates that were very high in the mid-1980s. In addition she suggests that the long-term increasing role of assets may result from increasing numbers of lump sum distributions from retirement plans. Whereas Social Security income is widely distributed among the elderly, pension and asset income as well as earnings are somewhat more narrowly distributed among higher income elderly—indeed, their presence results in higher income. As noted, there are differences in income source by marital status and gender. To elaborate on the role of the main retirement income sources in maintenance of inequality, we discuss the specific characteristics of the different income sources in the following sections. Despite its importance, we do not discuss the role of earnings in producing inequality because it represents the outcome of work institutions, not retirement structures. Variation in retirement age is discussed in Chapter 4.

Public Transfers: Social Security, SSI, and Medicare

The structure of public transfers to the elderly reduces inequalities that would otherwise exist. Public transfers also provide incentives for retirement at certain ages, a topic discussed in Chapter 4. Here we discuss selected characteristics of a very complex system, focusing on those aspects that reduce inequality among older age-groups. The Social Security Bulletin Animal Statistical Supplement, which is published annually, is an excellent source for more extensive discussion of Social Security details. Our discussion focuses on the Social Security system today, but its role in reducing inequality among the elderly has probably increased com-

Pathways to Inequality: Intracohort Differentiation over the Life Course

45

pared to twenty-five years ago because of the more than 50 percent rise (Ippolito 1991) in Social Security benefits in the early 1970s. We discuss the effects of this increase on a reduction in elderly poverty levels in a later section of this chapter. Social Security provides retirement benefits to almost all retired workers. A worker today must have earnings covered by Social Security for ten years in order to qualify for benefits; in 1997, a worker received a year's credit for earnings of $2,680 during the year (Social Security Administration 1997, 28). The individual worker's benefit level is based on average monthly earnings over the worker's lifetime (in addition to age of retirement). Prior to computing average monthly earnings, earnings in earlier years are adjusted to allow for the general rise of earnings over time. For most retiring workers today, the benefit computation is based on the highest thirty-five years of earnings, regardless of when they occurred in the worker's life. Average adjusted monthly earnings are usee! to calculate the worker's "primary insurance amount" (PIA). The algorithm for this calculation is primarily responsible for the leveling effect of Social Security. For workers reaching age 62 in 1997, the PIA is the sum of A. 90 percent of the first $455 of average monthly earnings B. 32 percent of the next $2,286 of earnings C. 15 percent of monthly earnings above $2,741. The PIA is the worker's benefit that would be payable at age 65. Retirement before age 65 leads to a reduction in benefits—to 80 percent of PIA at age 62 (Social Security Administration 1997, 39, 60). To see the effect of this benefit formula on inequality, consider the following three cases. A worker whose average indexed monthly earnings equaled the average earnings in 1996 ($2,058) compared to workers who earned 75 percent and 150 percent of the average a m o u n t . The average worker retiring at 65 would receive $913 while the other two would receive $750 and $1,153 respectively (Social Security Administration 1997, 56). Although the high-income worker had twice the earnings (and paid twice the taxes) of the low-income worker, the resulting PIA is only 54 percent higher. Similarly, the higher income worker's earnings were 50 percent higher than the average worker's but the resulting PIA is 26 percent higher. In addition, there is also a special PIA for those workers with very low earnings but long labor force attachment, though the examples listed above are far above the level that would qualify for the low earnings adjustment. Medicare, which pays for hospital and physician bills, has an even greater equalizing effect. Hospital insurance is provided at age 65 to all

46

Pathways to Inequality: Intracohort Differentiation over the Life Course

those eligible for Social Security or Railroad Retirement benefits plus some other groups. Since all elderly receive the same coverage regardless of their previous earnings, the program produces greater equality. Coverage for physician services is optional, and those choosing it pay a monthly fee. Yet this program is heavily subsidized by general revenues (Social Security Administration 1997, 95-97). The overall effect of the Social Security retirement benefit system and Medicare is to dull the effect of earlier lifetime economic attainments. In addition to Social Security retirement benefits, Supplemental Security Income raises the relative income of very poor elderly. It replaced Old Age Assistance in 1974. It is a means-tested program; for an individual over age 65, living alone, and with no other income and no countable assets, SSI provided $484 per month in 1997 (Social Security Administration 1997, 78). States have the option of supplementing SSI payments. The research of Burkhauser, Butler, and Wilkinson (1985), Hurd and Shoven (1985), and Crystal and Shea (1990) all show the clear equalizing effect of these government programs. Their benefit structures blunt the effect of earlier careers; by design, they limit the extent of carryover of inequality across the work and retirement phases of the life course. If they provided the only sources of income for the elderly, the story would be a simple one of status leveling produced by conscious social policy. They do provide a floor on income and, as Hurd and Shoven (1985) suggest, may raise the well-being of those at the bottom of the income scale after retirement. Since leveling does not characterize the changes in overall inequality with retirement, we now turn to a discussion of income sources producing inequality. Pensions and Earlier Socioeconomic

Attainments

If government transfer programs have an important leveling effect, inequality arising from other sources must increase with age to produce the overall patterns discussed earlier. Pensions provide a particularly important instance of the effect of employment structures and individual careers on the continuation of inequality across the work and retirement phases of the life course. Discussion of pensions is more complex than Social Security, however. There are many firm- or industry-specific pension plans, each with different rules. In addition, there are very strong individual career effects on pensions. Finally, both the social structure of pensions and patterns of individual careers vary by cohort. Individuals pass through different firms' pension structures at different points in their lives and for different amounts of time; this intersection of diverse pension structures and individual lives results in a very complex set of outcomes. (Pension structures are discussed extensively in Chapter 5.)

Pathways to Inequality: Intracohort Differentiation over the Life Course

47

Social Security is also affected by cohort differences—for example, the large unexpected increase in benefits experienced by retirees in the early 1970s (Ippolito 1990)—and individual earnings histories affect benefit amounts. But with the exception of these increases, pensions have probably changed more in recent years. Figure 2.1 provides information on one important link between retirement status and earlier attainments: the relation between pension participation and education level. Pension participation indicates current inclusion in any type of employer-provided pension plan, including both defined contribution plans such as 401K plans and thrift plans as well as traditional defined benefit plans. Data are from the 1992 wave of the Health and Retirement Study (see Juster and Suzman 1995 for a description of the study) and include respondents aged 51-55 who are currently employed by public or private employers. The self-employed and those not currently working are omitted. The figure shows a very strong relationship between pension participation on the current job and education. Men with a college education or more have a participation rate of 85.6 percent compared to 73 percent for high school educated men; the comparable figures for women indicate a greater education gradient: 79.4 versus 60.5 percent. Older workers are more likely to participate than younger workers, and those working at ages 51-55 are a selected sample. Hence these results cannot be generalized to the entire U.S. adult population. Participation rates are lower but the education effect is similar, however, when all respondents, including the self-employed, the retired, and unemployed, are considered and pension coverage is defined more broadly to include pensions expected or being received from past jobs. Differences in pension coverage while working are later reflected in pension amount received. Even among those receiving a pension, college graduates have pension incomes 1.8 times higher than high school graduates (Short and Nelson 1991), possibly reflecting longer years of coverage, coverage on more jobs, and higher earnings that result in higher pensions. Figure 2.2 presents results from a comparison of current job pension participation by race from the same data. The two highest educational attainment categories, thirteen to fifteen years and sixteen years or more, are combined in this figure, which also combines men and women (providing a minimum of 144 observations in the smallest education by race category). Overall, blacks have lower pension coverage rates—65.3 percent compared to 70.4 percent for whites. There are two reasons for this lower coverage rate. As the figure indicates, whites have a higher level of coverage among high school graduates. In addition, a higher proportion of blacks (68 percent) than whites (54 percent) are located in the lower two education categories.

£

^5

50

Pathways to Inequality: Intracohort Differentiation over the Life Course

The analysis presented in this section has ignored the cohort issue. The data on pension amounts for current pension recipients and pension participation for current workers refer to different birth cohorts. Changes in pension coverage over time, discussed in Chapter 5, imply that pension coverage and vesting have changed for successive cohorts. There is some evidence that the education difference in pension coverage has increased over time (Reno 1993). Even in the face of declines in coverage, the socioeconomic differentia] in coverage has been maintained and, perhaps, increased. In sum, pension coverage and benefit levels favor those with higher levels of education. Pensions are a central mechanism linking earlier socioeconomic differences to retirement income level, thus creating important links across the retirement transition. Gender Differences in Pension Coverage and Income

Among current workers, women are less likely to be covered by pensions, as shown in Figure 2.1. One important reason, discussed earlier, is that pension coverage is correlated with earnings levels, and women are still in lower-wage jobs, even among women who work full-time (Korczyk 1993). Women are also more likely to be working part-time (Tilly 1991). The pension gender gap in Figure 2.1 may be unique to this cohort and earlier ones. As discussed below and in Chapter 5, evidence suggests the gap may not occur in younger cohorts. The gap partly reflects past patterns of early labor force participation which are quite different from more recent cohorts (see Chapter 1). There are intriguing age differences in the pension coverage gap between men and women. There are no differences at ages 25-29 in the 1987 Survey of Income and Program Participation (SIPP) data, but there are difference at later ages (Short and Nelson 1991). Reno's (1993) comparison of 1979 and 1988 Current Population Survey (CPS) data indicates a similar pattern. The lack of gender differences at younger ages may indicate a cohort pattern of change as the labor force participation of employed women becomes more similar to men's. Retirement income sources among today's older population reflect the large pension differences between men and women found in the past. Thirty-six percent of married men and 29 percent of unmarried men receive pensions from private employers; 14 and 12 percent, respectively, receive public employee pensions. Twelve percent of married women and 21 percent of unmarried women receive private employer pensions, and 7 and 9 percent, respectively, receive government employee pensions (Grad 1996). Among currently married persons, men are much more likely to receive pension income. Unmarried women, a group that in-

Pathways to Inequality: Intracohort Differentiation over the Life Course

51

eludes never married and divorced women who are more likely to have earned pensions than other women and widows who may be receiving a survivor's benefit, are also less likely than unmarried men to have pension income. Although married men are more likely to have pension income than unmarried men (see also Campbell and Henretta 1980), the opposite is true among women. Among men, either lower status characteristics associated with being unmarried lead to placement in jobs without pension coverage or men who are unmarried are less concerned with pension in choosing a job because they generally have fewer family responsibilities. Unmarried women are more likely than married women to receive pension income because they have always been more dependent on their own employment or are receiving a widow's benefit. In both cases, marital status has important effects on sources of income. Figure 2.3 (derived from Short and Nelson 1991) shows men's and women's pension amounts, by marital status, for those who are receiving a pension. Older widows have particularly low pension income, reflecting either a reduced survivor's pension or a loss of a pension with a husband's death. Changes in the law governing survivors' pensions are discussed later in this chapter. They occurred too late to affect many of the widows in these 1987 data. Women who never married receive higher pension income than other women, reflecting their higher level of lifetime labor force participation. It is likely that future cohorts of women retirees will have higher levels of pension receipt and income because of their increasing levels of coverage. (Yet this trend must also be viewed in light of the pension changes discussed in Chapter 5.) One important reason that today's working women will have higher levels of pension income when they retire is that they have higher lifetime levels of labor force participation. At each preretirement age they are more likely to be in the labor force and less likely to interrupt their work careers for child rearing. Although more women in recent cohorts have long-term labor force attachment, some women still follow the intermittent employment pattern that was more common in previous cohorts. This pattern creates a link between family events, pension attainment, and retirement that is unique to women.

Wealth

Accumulation

Wealth accumulation is a second central process linking work and retirement phases of the life course. Income from assets is broadly important to the economic well-being of the elderly. During the mid-1980s, the real nonhome assets of the elderly increased, though the increase was not broadly shared (Ryscavage 1992). During the early 1990s, however, the net worth of the elderly declined, even controlling for age within the

£

Pathways to Inequality: Intracohort Differentiation over the Life Course

53

elderly group (Eller and Fraser 1995). Assets are very unevenly distributed among the elderly population. Table 2.1 demonstrates the small role played by asset income among lower-income elderly and its important role at higher income levels. Some asset income may reflect lump sum settlement from pensions (Reno 1993). Bernheim and Scholz (1993) find a strong relationship between education and both saving and assets. College graduates had much higher levels of saving and nonhome assets in relation to income. These findings are consistent with earlier research showing lower assets in late life among those with less education, holding long-term income constant (Henretta and Campbell 1978; Campbell and Henretta 1980; Kotlikoff, Spivak, and Summers 1982). The strong relationship between education and median net worth among the current elderly is shown in Table 2.2, which presents data from the 1993-1994 Asset and Health Dynamics Among the Oldest Old Study (for a description of the study, see Soldo, Hurd, Rodgers, and Wallace 1997). Because the respondents to this study are older and have been

TABLE 2.2

Median Net Worth and Home Ownership After Age 70 Net Worth

Home Value

Nonhome Assets

Percentage Owners

By Marital Status and Age Married 70-79 80+ Single Man 70-79 80+ Single Woman 70-79 80+

152,000 110,500

70,000 55,000

60,000 35,800

89.5 82.7

88,100 47,000

38,000 8,000

25,000 16,000

66.9 53.7

65,999 40,500

35,000 12,000

11,200 6,700

65.1 54.3

5,000 31,000 47,000 118,000

62.9 74.4 79.8 82.0

30,000 1,000 255

74.4 56.1 55.4

By Education Less than high school High school Some college College or more

40,000 98,866 132,000 231,000

25,000 50,000 65,000 85,000

By Race and Ethnicity White Nonwhite Hispanic

100,000 25,000 20,000

50,000 9,000 10,000

SOURCE: Asset and Health Dynamics Among the Oldest Old Study, 1993-1994.

54

Pathways to Inequality: Intracohort Differentiation over the Life Course

affected by events that may lead to a spending of some assets, their asset levels reflect a complex mixture of processes. Still, there is a strong relationship between education and assets. The second panel of the table shows the median net worth of college graduates is 2.3 times that of high school graduates. Since there is a great deal of evidence that the elderly do not use their home equity to finance retirement, a topic discussed in the next section, the third data column presents median nonhome assets—total net worth minus home equity. The education differences are greater here; the nonhome assets of college graduates are 3.8 times those of high school graduates. The bottom panel of the table presents asset differences by race and ethnicity, factors also associated with opportunities to accumulate wealth dtiring the working years. Whites have much higher asset levels than blacks or Hispanics. The median nonhome assets of minority elderly are not substantively different from zero. The top panel of the table provides results by the cross classification of marital status and age. The highest asset positions belong to younger, married households. Single men and women have much lower asset positions. Those over age 80 also have fewer assets. These results reflect a number of complex processes. Survival as a married couple may be associated with characteristics that allowed greater asset accumulation. Those over age 80 belong to earlier cohorts that had lower income from which to save. The results for singles also reflect the effects of widowhood. More than eight out of ten single women in the sample are widowed, as are seven out of ten single men. Health care costs of the deceased spouse might be responsible for part of the marital status difference. About 8 percent of the single women and 17 percent of the single men are divorced, a factor associated with lower total assets net of income, particularly lower home equity and financial assets (Henretta and Campbell 1978; Campbell and Henretta 1980). The remainder, 10 percent of men and 6 percent of women, are never married persons. Home

Ownership

Home ownership is a particularly interesting component of assets for two reasons. First, home ownership is linked to events early in the life course. Marriage and the presence of children (Henretta 1987), as well as being white, results in early home purchase. Differences in home purchase rates at young ages (Henretta 1984) appear to be responsible for a portion of the large difference in net worth position between older blacks and whites (Henretta 1979). The trajectory may be disturbed by later divorce (Campbell and Henretta 1980), which reduces home equity, at least among men. Second, there is considerable evidence that the elderly do not use their home equity to finance consumption (Venti and Wise 1990;

Pathways to Inequality: Intracohort Differentiation over the Life Course

55

1991). There might be many reasons for this reluctance—wanting to leave something to children, not wanting to move, or not wanting to reduce housing quality. Third, home ownership is a very important contributor to the net worth of the elderly. The important role of home equity is illustrated in Table 2.2. In each row of the table, and particularly for single women, those with little education, and minorities, median nonhome assets are considerably lower than total net worth. Also, home ownership is broadly distributed among the elderly as shown in the last column of the table. The importance of housing in the wealth picture of most older persons has spurred attempts to develop ways for older persons to use some of this wealth during their lifetime. Venti and Wise note that even if the elderly wish to use their home equity for consumption, conversion would not provide much help to the elderly most in need. Consider, for example, minorities and single women over the age of 80. Almost all their assets are tied up in their homes. But the relatively low value of their homes (which reflects both home value and the proportion who own homes) limits the degree to which conversion of home equity would help. It is possible to convert partial home equity to cash using a reverse annuity mortgage, a financial instrument allowing the older homeowner to live in the house until death. The Federal National Mortgage Association (popularly known as Fannie Mae) began purchasing reverse mortgages in 1996 (Wall Street Journal 1995). This development is important because it creates a secondary market for reverse mortgages and creates a standard product. These two factors will make reverse mortgages more widely available. Results over the next several years will provide a better test of interest in home equity conversion than has occurred heretofore. Cohort Trends in Women's Lives and Intracohort Differentiation

During the last half century, more women in successive cohorts have been affected by employment-related social structures as lifetime labor force attachment has increased and employment has become less tied to marriage and fertility. The trends of increasing women's labor force participation and earlier retirement among men and women have led some observers (e.g., Paukert 1984) to argue that age, not gender, is becoming the primary factor now organizing the life course. In related developments, age-based family roles have become more diverse. Later age at marriage and decline in proportion who ever marry, the rise in divorce, and changing fertility patterns have created greater variety in life course family attachments. These family changes have proceeded at different rates. In some cases, such as age at marriage arid fertility levels, the

56

Pathways to Inequality: Intracohort Differentiation over the Life Course

direction of trends has reversed. The pace of change has been so rapid that women's cohort patterns are more distinct than men's. As a result, women's lives have changed in two conceptually important ways. First, they are more affected by employment institutions and less derivative of family institutions. Second, they have become more diverse. While we often overestimate the uniformity of family life in the past (Uhlenberg 1969; 1978), there is increasing evidence of an invertedU shaped pattern of heterogeneity in family characteristics for female cohorts bom in the first half of the twentieth century. Recent cohorts and those born early in the century have more diverse family lives than 1930s birth cohorts. Changing employment levels and increasing variability in family patterns have helped shape the pattern of intracohort differences between men and women. In the next section we examine ways in which the average pattern of women's lives has changed (or not changed) in successive cohorts through major transformation in tine twin domains of work and family, hi addition to examining changes in average pattern, we also address the issue of shifting intracohort heterogeneity. That is, have women's lives become more homogeneous or less? Finally, we trace the effects of the extensive change in the first half of women's lives— changes in early and midlife employment and childbearing patterns—on intracohort heterogeneity in later socioeconomic inequality.

Cohort Differences in Early and Midlife Labor Force Participation

Women's

Figure 1.1, discussed in the previous chapter, uses partial life course data on a number of different cohorts to trace cohort differences in women's labor force participation since 1965. In this figure, each successive young cohort exhibits a higher level of labor force participation at each age. In addition, the shape of the labor force trajectory suggests a changing relationship of employment with fertility. The earliest young cohort, women born between 1936 and 1940 and aged 25-29 in 1965, still exhibit some traces of the classic M-shaped pattern of labor force participation identified by Masnick and Bane (1980) for earlier cohorts. In cohorts earlier man the ones shown in Figure 1.1, labor force participation reached a first peak at age 20-24 and then declined in the childbearing and rearing years; it increased again in middle age and then declined with retirement. In the 1936-1940 cohort shown in Figure 1.1, the drop in labor force participation during childbearing has disappeared, though participation is low during the child rearing years. Participation levels increase steadily throughout the portion of the childbearing and rearing years shown and reach their peak in middle age. This attenuated M-shaped pattern disappears in successive cohorts so that there is no evidence of

Pathways to Inequality: Intracohort Differentiation over the Life Course

57

lower labor force participation when children are young. This cohort difference can also be described as a difference in historical periods—the most rapid growth in women's labor force participation in recent years has occurred among women with children under 6 and under 18 (DaVanzo and Rahman 1993). The cohort perspective is particularly valuable, however, because it demonstrates the different life course patterns of more recent women that will have major consequences for their incomes and preparation for retirement Although women's employment remains more family related than men's, the role of early family events has clearly declined over time. The large changes in labor force participation mean that women's labor force participation age profiles are becoming more similar to men's (Peracchi and Welch 1994). A comparison between Figures 1.1 and 1.2 shows the increasing similarity, albeit women's participation levels are lower than men's. Women and the Changing

Workplace

A number of other changes in employment have made men's and women's job patterns more similar as men's work has become less stable while women's employment has become more stable. These trends are currently small, but their effect has been to make men's and women's work more similar. First, between 1983 arid 1997, men's median current job tenure has declined while women's tenure has increased (Bureau of Labor Statistics 1997). It is possible that men's tenure has declined because men are moving to better jobs when unemployment is low; however, the years examined have included some with high as well as low unemployment, and the trend has been monotonic across the four measurements over these years. Second, between 1969 and 1989, the proportion of adult men working part-time increased slightly while adult women's part-time work remained approximately stable (Tilly 1991). Third, between 1972 and 1988 the proportion of pension-covered workers who were vested increased, but pension coverage for men declined while women's coverage increased (Reno 1993). The central reason for these changes is that the prevalence of internal labor markets (Osterman 1994)—also called age-structured employment (Henretta 1994b)—has peaked and may have declined slightly. These work arrangements are ones in which there are workplace rules producing nontransferable seniority-based entitlements such as pension rights, higher wages, and job security (Spilerman 1977). Two large economic changes have made employment less stable. First, the rapid pace of economic change and increased competition have reduced job security. Second, technological change and higher educational job requirements

58

Pathways to Inequality: Intracohort Differentiation over the Life Course

(which also means that the worker is more likely to pay the costs of training) both reduce the long-term relationship found in age-structured arrangements (Osterman 1994). In addition, changing pension regulation and an increase in small employers has meant that pension coverage has lagged. The change affects men more than women because men have traditionally been more likely to hold protected jobs. It is important not to overstate the extent of these trends. Most midlife men are still in protected employment (Osterman 1994) and, as shown earlier, a large percentage of midlife men participate in pensions on their current jobs. Finally, there has been a shift from manufacturing to service employment. Between 1975 and 1994, manufacturing employment in the United States remained stable at about 18 million while employment in service jobs increased from 53 to 91 million. One component of services, retail trade, increased from 13 to 21 million (Economic Report of the President 1995). Hence, as a proportion of employment, manufacturing has declined while services have increased. There is a great variety in service industries (Meisenheimer 1998), and there are many good jobs in service fields such as health care or engineering services. Yet the shift has probably reduced the quality of jobs available to the unskilled (Wilson 1996), and it has affected the difference between men and women. For men, the change in industrial distribution has likely meant a decline in job security and less pension coverage because manufacturing was more likely to offer both (particularly as it was organized in the past). Women have been overrepresented in services and hence have less to lose from industrial distribution changes. Because women are entering jobs previously held by men, their opportunities have remained stable or have improved despite change in employment opportunities. Continuing Employment Differences Between Men and Women

Although the changes just described are important, important differences remain between the work histories of men and women. Employed women are not a homogeneous group; there are differences in experience, continuity of work, and labor force attachment. This variation is greater than among men. Though career pattern differences may be declining, it is premature and oversimplified to conclude that men's and women's careers are rapidly becoming indistinguishable. First, men and women are segregated in different occupations. Data from the 1984 Survey of Income and Plan Participation indicate that among high school graduates men were in occupations that were 21 percent female and women were in occupations that were 68 percent female. Results for those with more and less education were roughly similar (Bureau of the Census 1987). Though occupational segregation has been declining, even

Pathways to Inequality: Intracohort Differentiation over the Life Course

59

in 1992 one-third of employed women were in six occupations (U.S. Department of Labor 1993). Second, women are more likely to experience interruptions in their work than men. Equally important are the reasons for the interruptions—men's periods of nonemployment tended to result from difficulty in finding work, whereas women's work careers were more likely to be interrupted by family considerations (Bureau of the Census 1987). Third, women's earnings are less than men's, as discussed in Chapter 1. Ironically, increasing women's labor force participation rates tend to slow increases in two other measures of labor force commitment—experience and continuity. Labor force experience and continuity of employment among employed women does not increase at the same rate as women's labor force participation because women newly drawn into the labor force include many who have not been employed previously. Hence, the process of increasing labor force participation implies a slow increase in the average labor force experience of employed women (Smith and Ward 1984). Even in cohorts who are in late middle age today, there will be considerable heterogeneity in lifetime labor force participation because of the large numbers of women entering the labor force in their 30s. In the four most recent cohorts shown in Figure 1.1, this variability is progressively declining. The data currently available suggest both higher levels of labor force participation and a smaller proportion of the cohort entering the labor force in their thirties. Although each successive cohort has higher and less variable lifetime labor force participation, it will be about thirty years before the cohorts of women entering the retirement years show markedly lower variability in lifetime labor force participation compared to today's retirees. And, even in recent cohorts, women's employment levels remain below men's, indicating continuing variability.

Reasons for Increasing Women's Labor Force Participation

Research has identified a number of causes for the increase in women's labor force participation. The most important are the grcwth in the demand for female labor and the long-term decline in fertility. Smith and Ward (1984) point out that women's labor force participation rates have been increasing since early in this century. They attribute this increase primarily to rising women's wages created by growth of clerical jobs. Growth of these jobs after 1910 created a new role for employed women, who previously had been limited to domestic service. Similar changes occurred in Canada (Jones, Marsden, and Tepperman 1990). Recent changes in employment structure have also added to the increasing demand for women in the labor force. As noted earlier, manufacturing employment has remained stable while service employment has grown. Since women

60

Pathways to Inequality: Intracohort Differentiation over the Life Course

are underrepresented in manufacturing and overrepresented in the growing service sector, the demand for women's labor continues to increase (Smith and Ward 1989). Therefore a major contributor to the increase in women's employment levels has been a set of important structural changes in the economy. A second contributor to rising labor force participation rates has been the secular decline in fertility. Fertility of cohorts of women has been declining since the nineteenth century (Rogers and O'Connell 1984) with the exception of the postwar baby boom. In recent years, total fertility—an estimate of the average number of children born to a women, derived from birth rates in a particular year—has hovered around 2.0 during the 1990s (Ventura, Martin, Curtin, and Matthews 1997). This decline in fertility has been accompanied by an increasing variability in childbearing patterns, a topic discussed later. Yet there is only a loose link between fertility and aggregate labor force participation, since even among cohorts who were mothers of the baby boom, labor force participation continued to increase but at a slower rate than before or after (Smith and Ward 1984). Researchers have pointed to a number of additional factors to account for increasing labor force participation among women. These include the declining wages of men, rising education (Smith and Ward 1984; Jones, Marsden, and Tepperman 1990), and a growing urban population (Smith and Ward 1984). Changing Family Rttterns and the Declining Centrality of Marriage

In addition to changes in employment, marital patterns have changed greatly. These changes are most usefully summarized by examining cohort marital status patterns. The proportion of surviving women who ever marry rose through this century, peaked in the 1938-1942 birth cohort, and has since declined. At the same time, the proportion of marriages ending in divorce has increased steadily for successive cohorts in this century, though some evidence suggests divorce peaked during the 1980s (Schoen and Weinick 1993). Increasing divorce balances a decline in marriages that end in early widowhood or death. Because of the changes in marriage probabilities and divorce, the 1948-1952 birth cohort—currently about 50 years old—will spend less of its lifetime in marriage, compared to cohorts born in the 1920s and 1930s (Schoen, Urton, Woodrow, and Baj 1985). The data on marriage indicate increasing diversity of patterns. That is, age tells us less about an individual's lifetime marital status in recent cohorts than in 1930s birth cohorts. Projection of lifetime experience for younger cohorts requires certain assumptions. Schoen et al. (1985) assume that rates in effect in 1980 will continue for the remainder of the cohort's life. More recent life tables for 1988 (Schoen

Pathways to Inequality: Intracohort Differentiation over the Life Course

61

and Weinek 1993) indicate that this assumption has been reasonable so far, except that the proportions ever marrying and the proportion marrying after a divorce have declined since 1980. If projections were being done with these more recent data, they would show higher levels of heterogeneity across the life course in recent cohorts. Viewing these patterns forward and backward from the position of women retiring in the early to middle 1990s—roughly the 1928-1932 birth cohort—women who generally retired in the early 1990s differ from earlier and later cohorts of women in that they have spent more of their lives married. Despite having higher divorce rates than earlier cohorts, the higher proportion marrying and earlier age of marriage increase time spent married. Today's younger cohorts are slightly less likely to marry and much more likely to divorce. And they are less likely to remarry after a divorce. Therefore, women retiring today have had relatively uniform marital patterns compared to greater diversity in cohorts before and after them.

Changing Fertility

There have been two important changes in fertility patterns. First, fertility peaked during the postwar baby boom and has declined since then. Examining the issue in cohort perspective, women born in the 1930s had the highest cohort fertility during this century. The 1930-1934 cohort had total fertility of nearly 3.1 children per woman (Rogers and O'Connell 1984). Earlier and later cohorts had lower fertility and, as noted, total fertility has been fairly stable at 2.0 during the first half of the 1990s. The age at childbearing and the proportion childless has also changed. In recent years, more births occur to women over age 30 and an increasing proportion of women aged 40-44 are childless (DaVanzo and Rahman 1993). These changes make recent cohorts similar to pre-baby boom cohorts. Cohorts born in the mid-1930s were more likely to have their first child by age 30 than earlier or later cohorts (O'Connell 1991), indicating greater diversity in family patterns before and after the baby boom. Equally important, patterns of childlessness have changed. Women born in the 1930-1934 cohort were very likely to have had children. About 10.4 percent of women in this cohort were childless (Rogers and O'Connell 1984), and childlessness declined further in the 1935-1939 cohort (O'Connell 1991). Women born in the 1930s were in their childbearing years during the postwar baby boom, and both earlier and later cohorts have had higher levels of childlessness. For example, among women in the 1946-1950 birth cohort, 16 percent were childless at ages 40^44; among the 1951-1955 birth cohort, 17.5 percent were childless (Bureau of the Census 1997).

62

Pathways to Inequality: Intracohort Differentiation over the Life Course Implications for Intracohort Variation in the Life Course

The overall effect of changes in marriage, divorce, and fertility is that cohorts born during the 1930s, the first of whom reached retirement age in the 1990s, have had the most uniform and regular family life course of women born this century, even though all the indicators of family characteristics have not changed at the same rate or in the same direction. The 1930s cohorts have spent the largest proportion of their lives married, even though nearly one-third of them have divorced. In addition, they have had the lowest rates of childlessness, the highest cohort fertility, and the highest proportions having a birth before age 30. With the exception of divorce, the indicators discussed show a U-shaped pattern of change during this century centered on the 1930s cohorts. The 1930s cohort employment patterns are midway between those of earlier and later cohorts. They show the M-shape that has attenuated in later cohorts and an agespecific labor force participation rate that is between earlier and later cohorts. Thus the overall picture of women's lives in successive cohorts is made complex by variation in the historical patterns of change in employment and family structure. Although some changes follow a monotonically increasing pattern, other changes are characterized by a U-shape that differentiates 1930s cohorts from earlier and later ones. Today, women entering retirement are more likely to be divorced or in second marriages than in the past. They have had highly variable lifetime labor force participation patterns, and most have had children. In the future, cohorts will progressively have higher levels of labor force participation, though as discussed earlier, new retiree cohorts will show high lifetime variability in labor force participation for many years. Increasing numbers will be divorced or in second marriages. More will be childless; among those with children, there will be more variation in mother's age at child's birth. Variation in these patterns constitute an important aspect of intracohort variability by themselves, but they also have implications for intracohort variation in socioeconomic status. There are no comprehensive data allowing examination of family change implications for women's socioeconomic status across the broad sweep of cohorts discussed in this section, but research has examined the effects of lifetime employment patterns, marriage and divorce, and childbearing on socioeconomic variation

Employment Patterns and Late-Life

Heterogeneity

As a result of greater lifetime labor force participation, increasing numbers of women now receive Social Security retirement benefits based at least partially on their own work records. This proportion has increased from 50.6 percent in 1970 to 62.6 percent in 1996. However, as women's

Pathways to Inequality: Intracohort Differentiation over the Life Course

63

labor force participation has increased, the pool of women who are eligible for benefits based on their own work records has become more heterogeneous. In this group, the proportion of women receiving benefits as workers only has declined while the proportion "dually entitled" has increased. Dual entitlement means that a woman could receive benefits based on her previous work, but her benefit as a wife (half her husband's benefit, ignoring early retirement reductions) or a widow (100 percent of her husband's benefit ignoring early receipt reductions) is greater (Ferron 1997). Although more women have worked in covered employment for the required ten years to qualify for some benefits based on their work, the patterns of their employment, and therefore their Social Security benefits, are highly variable. Smith and Ward (1984) show that the increase in women's labor force participation has meant a large change in lifetime experience among all women but a very small change among women in the labor force. The reason is the same as the Social Security retirement benefit finding. A cohort pattern of increasing labor force participation means mixing of women with extensive experience with women who have had much less. Paradoxically, the increasing levels of labor force participation to this point have increased variability in women's own earned benefits. Women's retirement is affected by the same workplace factors that affect men's retirement, such as pensions. Yet women differ from men in the effects of their fertility histories on life course employment and retirement income (O'Rand and Landerman 1984). Early events such as intermittent employment, part-time employment, and limited job choice are strongly related to lifetime home and childbearing responsibilities. The heterogeneity in women's careers carries over into their retirement patterns. Women with extensive early labor force participation retire earlier than women who enter employment after their childbearing years. This pattern affects both currently married and unmarried women (O'Rand and Henretta 1982a; O'Rand, Henretta, and Krecker 1992; Henretta, O'Rand, and Chan 1993). Early employment is also related to a higher standard of living for women in late life (Hofferth 1984). The pattern of early absence from the labor force for family reasons will persist for a considerable number of women in cohorts reaching retirement age over the next two decades, though the increase in childlessness and growing labor force participation among today's childbearing cohorts will eventually reduce the prevalence of this pattern. Family Formation and Breakup

Marital status plays a particularly important role in producing variation in the experience of women. One route of marital breakup, widowhood, is examined in the next section's discussion of poverty. Divorce in midlife

64

Pathways to Inequality: Intracohort Differentiation over the Life Course

also has important effects on economic well-being. Divorce is associated with a decline in the income to needs ratio—the ratio of income to the poverty level for the household of the size a woman lives in each year. However, divorce has different effects depending on other transitions. Women in empty nest households experience more decline in the income to needs ratio than do women with children still at home (Smith and Moen 1988). A central social policy issue is how retirement policy can accommodate women whose family commitments led to lower early labor force participation, thereby affecting their individual retirement income, but later become divorced. The rising divorce rate and the early labor force participation patterns of women now in their 40s and 50s suggests this issue will remain for at least the next two decades. Men retiring today receive higher pensions than women (Short and Nelson 1991), and the higher lifetime probability of divorce among older cohorts reduces the probability that a woman will reach retirement in a first marriage. Social policy has come part way in acknowledging the implications for retirement of the changed family status of older women. Since 1983, federal law has provided for splitting of accrued pension rights in the case of divorce (Goodfellow and Scheiber 1993). Social Security law has made fewer changes. Spouses are eligible for the spouse benefit at the regular retirement age if they were married ten years. Changes in 1977 reduced the length of the marriage needed to qualify for this treatment from twenty to ten years, and 1983 changes allowed women to receive benefits at Social Security retirement ages even if their former husbands were not retired (Social Security Administration 1997). Social Security has not provided for a splitting of accrued entitlements in the manner of occupational pensions.

Poverty

Until the early 1970s, poverty was widespread among the elderly because payment levels in Social Security were very low. With the large benefit increases that occurred in the early 1970s, poverty is now no more common among the old than among younger persons. As with other aspects of aging discussed in this chapter, there is significant intracohort variation. In this section, we discuss the general issue of poverty among the elderly and then focus on the amount and sources of intracohort variation. James Schulz (1985) makes the important point that the proportion of the elderly living in poverty is no longer an appropriate measure of the adequacy of their income. As retirement has become a normal and expected part of the life course, individuals expect to maintain their standard of living, not experience a decline to near poverty. Indeed, it is likely that

Pathways to Inequality: Intracohort Differentiation over the Life Course

65

the rise in retirement income has played a major part in making retirement part of the normal life course (see Chapter 4). In addition, a poverty measure throws away information about the well-being of the elderly by treating a quantitative concept as a dichotomy. Though the poverty measure is arbitrary and wastes information, it does provide a summary description of the state of the elderly popvilation and provides insight in the ways that marital status, gender, and race define important components of intracohort variation that persist and may increase with age. Historical

Overview

Figure 2.4 presents an overview of poverty rates in the U.S from 1966 to 1996 by age-group (based on Lamison-White 1997) and shows some wellknown results concerning poverty. First, poverty among children under age 18 and adults 18-64 declined to their low points in 1973 and have generally trended upward since then, particularly during the 1980s. Among those 65 and over, poverty declined sharply until 1975 and has continued to trend downward since then. In 1996, 10.8 percent of the elderly were below the poverty line compared to 11.4 percent of those 18-64 and 20.5 percent of children. Yet the elderly were more likely to be "near poor" (between 100 and 125 percent of the poverty threshold) than were younger people (Lamison-White 1997). Increases in poverty at ages under age 65 are usually explained either by changing jobs and unemployment levels or by the increasing proportion of female-headed families in the U.S. population. (See Sandefur and Tienda 1988 for a discussion of these different approaches.) In 1966, poverty rates were highest in the over-65 population. Poverty among the elderly dropped dramatically between 1970 and 1975 for reasons specific to them. Social Security benefits were increased about 50 percent in real terms over this period (Ippolito 1990), a change that delinked the elderly's trend from that of the remaining population. In addition, the establishment of need-based Supplemental Security Income for the elderly also reduced poverty (Burkhauser, Holden, and Feaster 1988). Since that time, elderly poverty rates have tended to decline further. The growing adequacy of firm pensions and individual savings probably account for declining poverty among the elderly since 1975. Age-based public policy did work to reduce elderly poverty; hence, the elderly have lower poverty rates today than the general population. Intracohort Variability in Poverty

The overall success in reducing poverty hides large amounts of heterogeneity in the elderly population as shown in Table 2.3, which examines

£

Pathways to Inequality: Intracohort Differentiation over the Life Course

67

TABLE 2.3 Poverty Rates for Persons 65 and Over in Common Household Arrangements, by Sex White Total Married Unrelated Individuals

Black

Hi:>panic

Men

Women

Men

Women

Men

Women

5.7 4.0 11.0

12.1 3.9 21.0

18.1 10.6 25.4

29.8 10.3 46.9

19.9 14.6 38.9

27.7 15.0 46.7

SOURCE: Leatha Lamison-White. 1997. Poverty in the United States: 1996. Current Population Reports, series P60-198. Washington: Government Printing Office. Table 2.

poverty rates by gender and race and ethnicity for two common household arrangements. Married couple households (which may have children living with them) are one type of family. Unrelated individuals are persons living alone or with non-relatives, though most live alone. Married couple families and unrelated persons include most, but not all, the total population; for example, an elderly person living with a child fits neither of these categories. The table shows major differences by living arrangements that affect women more strongly then men. Within each racial or ethnic group, those who live in married couple families have the lowest poverty rates and generally show small gender differences. Unrelated individuals have high poverty rates, particularly among women arid minority groups. Race and ethnicity also have strong effects; for each gender and living arrangement, minority groups have poverty rates more than twice those of whites. The very strong effect of household arrangements for older women has led analysts to focus on the effects of widowhood on women. In recent years, research has begun to answer the question of why older widows have high poverty rates First, both men and women who become widowed had lower income and wealth several years prior to widowhood. Second, the event of widowhood results in further loss, though the loss is less for men than for women (Hurd and Wise 1989; Zick and Smith 1991). The lower income and wealth of widows several years before the event may result from lower earnings (Zick and Smith 1991), or may result from poorer health's effect on the ability to save from a certain level of earnings (Hurd and Wise 1989). Some of the income and wealth loss that occurs at the time of widowhood probably results from expenses associated with the illness and death of the spouse, but the effects of widowhood are primarily determined by the institutional rides governing retirement benefits. While nonhousing bequeathable wealth—savings, for example—declines about 15 percent with widowhood, the largest declines are in Social Security wealth (39 percent) and pension wealth (63

68

Pathways to Inequality: Intracohort Differentiation over the Life Course

percent) (Hurd 1990). Declines in Social Security income reflect the loss of the spouse's benefit for widows. Couples with one worker and a nonworking spouse are eligible for a spouse's benefit that is equal to 50 percent of the individual's benefit. After widowhood, the spouse's benefit is lost, reducing income by one-third. Social Security benefits the married more than standard household budgets suggest is required by two persons compared to one, and there have been a number of proposals to shift benefits from the time a couple is married to the time of widowhood. This issue is particularly important because reducing a couple's benefit by one dollar would allow a $1.45 increase in benefits to widows (Sandell and lams 1997). Pension income also declines substantially for widows (Hurd and Wise 1989). One reason is the loss of the pension if there is no survivor's option, as was often the case in the past. This situation may have changed because of changes in social policy. Prior to passage of the Employee Retirement Income Security Act in 1974, which required offering a survivor option, not all firm pensions offered survivor benefits (Schulz 1985). In addition, many workers declined such options (Schulz 1985), but legislation in 1984 required that both husband and wife decline the benefit (Hurd 1990). The effects of widowhood may depend on the measure of economic welfare used. Burkhauser, Holden, and Feaster (1988) found that there was little difference between widows whose husbands had pensions that ended with the husband's death or provided a survivor option. The explanation for this finding may lie in the economic status of women before widowhood—widowhood may precipitate poverty among women who had few economic resources. But among those whose husbands had pensions it may have little effect because of the higher level of resources (Hurd 1990; Choudhury and Leonesio 1997). In other words, the dichotomous nature of the poverty measure means that events that reduce economic status will only lead to poverty among those whose economic resources were limited already. In sum, successive cohorts of older persons are less likely to be poor, but within cohorts there is considerable variation. Intracohort differentiation comes from two sources: different lifetime careers and middle- or late-life events and the interaction between them (Choudhury and Leonesio 1997). The most important class of late-life events derives from family breakup—either through divorce or widowhood. There are clearly others, however—poor health (Choudhury and Leonesio 1997) or an unexpectedly early end to a job. Public policy also plays a role. The most important contributor to the declining poverty among the elderly was the 1970s rise in Social Security benefits. Yet Social Security is also an important contributor to intracohort variation because the structure of benefits advantages individuals when they are married at the cost of lower benefits during the time they are widowed. Hence poverty, as with most

Pathways to Inequality: Intracohort Differentiation over the Life Course

69

aspects of inequality among the aged, is produced by a complex interaction of early- and middle-life events, late-life events, and public policy. Conclusion

Kohli's conceptualization of the links between the work and retirement phases of life presented at the beginning of this chapter argues that there are important links between work and retirement phases of the life course in a broad range of areas, from self-concept to economic status. This chapter has traced the linkages in economic status between work and retirement. We defined three possibilities: the leveling of inequality, its maintenance, or its accentuation. In the end, these possibilities characterize components of economic status more clearly than they do the overall result. There is clearly a great deal of status maintenance as midlife attainments transmit their effects across the retirement transition. However, the components of income have quite different effects. Social Security and Medicare reduce inequality; through design they have relatively little variance in the benefits they provide regardless of previous earnings. Pensions and savings, however, tend to magnify the effects of earlier status. The reason is that they are more unequally distributed than earnings, which begin to drop in importance as an income source. These processes have two important results. Those at the bottom of the elderly's income distribution are probably better off absolutely than those at the bottom of the middle-aged income distribution. At the same time, there is more inequality as the income sources with the greatest inequality become more important. Thus, although status maintenance describes the overall result quite well, there are important aspects of inequality that would better be described as leveling and accentuation. With the exception of the leveling effect of Social Security, the resulting process indicates very important links between work and retirement phases of life. Intracohort differences among women and between women and men are strongly affected by early- and middle-life family commitments. Extensive change in lifetime marital, fertility, and employment patterns have made women's lives more diverse and will produce highly diverse outcomes for women in the future. Women's late-life status is affected by a complex combination of early- and middle-life socioeconomic and family attainments as well as late-life contingencies. Poverty provides one important example. The late-life event of widowhood produces poverty among women with limited resources. Although increasing variability in the lives of women produces variation in outcomes as they age, it does not alter the link between early and later events in individuals' lives.

70

Pathways to Inequality: Intracohort Differentiation over the Life Course

The important point for this discussion is that there are two central elements to the production of intracohort variation. Life course work and family careers and public policy are the primary players. The vicissitudes of life—poor health, a lost job, widowhood, or divorce—are important and may have devastating effects. Yet their effects play out within the context of long-term careers and public programs.

3 Asynchronous Lives: The Normal Life Course and Its Variations

Long-term, cross-national trends—including population aging, higher educational attainment, and women's increased labor force participation over their lifetime—have generated speculation regarding the relative decline of the gender structuring and increase of the age structuring of the life course (Paukert 1984). The age-structuring hypothesis proposes that the life course of men and women is becoming increasingly uniform. Educational and workplace institutions are superceding family institutions in the timing and organization of life transitions, such as the transition into and out of the workplace. Increased rates of labor force participation by succeeding cohorts of women are associated with lower fertility rates, delayed marriage, and increased marital dissolution patterns. These trends represent a decrease in time over the life course spent in the family roles of childbearing and child rearing and an increase in time spent in the workplace. As such, men's and women's life courses are now more commonly anchored by age-graded educational schedules that regulate the transition to adulthood in the life course and by employment and public welfare institutions that shape work and family careers and final labor exits through the implementation of age-based rule structures and time-based role participation and tenure criteria. The Age Integration of Lives

Educational and market-based institutions have contributed to the triphasic organization of the life course. Matilda White Riley (Riley 1993a; 1993b; Riley and Riley 1994) terms this the age differentiated or 71

72

Asynchronous Lives: The Normal Lite Course and Its Variations

age-segregated model of the life course (which is depicted in the left panel of Figure 3.1). The age-structuring hypothesis emanates from this model and argues that the normal life course is organized more along these lines rather than along separate, gendered pathways wherein women diverge as a result of their transitions into and out of reproductive and family roles. The first secular trend informing the age-structure hypothesis is population aging. The aging of the population in the United States and other advanced industrial countries (see Chapter 6) reflects general declines in fertility and increases in longevity (Bosworth and Burtless 1998; Uhlenberg 1992). The extension of life for men beyond the "productive" work years and for women beyond the "reproductive" childbearing/rearing years has made age an increasingly salient issue for public and private sector policies pertaining to health and welfare over people's lifetimes, especially those that pertain to income maintenance. The extension of adulthood beyond the productive-reproductive years has been paralleled by the extension of the adolescent or preadulthood phase, as work roles have become delayed due to educational regimes or limited occupational opportunities for young men and women (Kohli 1986; Sheppard 1991). As such, age is a fundamental criterion for allocating public and private rights and resources. Second, the relative expansion of educational opportunities to men and women across countries has been an independent basis for the integration of their life courses (Shavit and Blossfeld 1993). Education has affected gendered life course schedules in at least two ways. One effect has been to delay the transitions to more gendered roles, such as those associated with marriage, fertility, and care giving (Bianchi 1995). Educational regimes that extend secondary, postsecondary, and tertiary (vocational) programs to more of the population serve to standardize adolescent and young adult role schedules. Age ranges in late adolescence and young adulthood historically associated with the onset of fertility and family roles among women have served to separate them from—or to significantly delay their movement into—the marketplace. Expanded educational opportunities have delayed these schedules. The second effect has been to transfer cultural resources in the forms of general knowledge, specific skills, and tastes for learning to a wider segment of the population. This transfer does more than delay the transition to other roles; it establishes a foundation of cultural capital with cumulative effects on subsequent educational and occupational attainments (Blossfeld and Shavit 1993). The structure of educational systems varies widely across countries in the extent to which the education-to-work transition is tightly coupled via such mechanisms as apprenticeship and vocational education programs (see Kerckhoff 1996), but the increasingly

Asynchronous Lives: The Normal Life Course and Its Variations

AGE DIFFERENTIATED

73

AGE INTEGRATED

AGE

OLD

LEISURE

o I—.

H < U D Q

MIDDLE

YOUNG

c

—*

EDUCATION

FIGURE 3.1 Types of Social Structure SOURCE: Courtesy of Matilda White Riley. universal access to many forms of education has standardized early life transitions. Third, as noted in earlier chapters, the feminization of labor worldwide has proceeded as a relatively steady increase of women's labor force participation in all age-groups (Easterlin et ai. 1983). Goldin's (1990) economic history of gender-based wage inequality in the United States documents that in the first half of this century, women were clearly divided into two groups: those who worked relatively consistently over their lives, and those who never worked. Over the second half of this century, women's patterns have been more heterogeneous. Part-time and more diverse discontinuous work patterns developed after World War II. Changing family patterns among cohorts of women born in the two decades before the war contributed to these trends. As mentioned in Chapter 2, these cohorts of women spent more of tlieir lives

74

Asynchronous Lives: The Normal Lite Course and Its Variations

married. The cohorts that have followed have resumed the consistent pattern of work Goldin (1990) identifies earlier in the century. Women have been going to work early in adult life at increasing rates and have sustained high levels of labor force participation patterns over their lifetime (Paukert 1984). The demand for female labor, stemming from the rise in clerical occupations and the growth of the service industries, attracted primarily younger women (under 25) in the United States until World War II, Then older women's (over 35) labor force participation rates grew more rapidly until the 1960s. The subsequent period has been dominated again by the labor force patterns of younger women, especially mothers of young children. Easterlin et al. (1983) argue that the relative "substitutability" of younger and older women in the service industries has permitted the shifting dominance of age-groups. In any case, across European countries and the United States between 60 and 85 percent of women work in the service sectors. Today over three-fourths of U.S. women aged 20-24 are in the labor force, reflecting an average annual growth rate in labor force participation between 1970 and 1989 of just under 1 percent. Older women also have sustained relatively high participation rates: whereas six out of ten women aged 45-54 have been in the labor force annually for nearly two decades, approximately four out of ten women aged 55-64 have participated as well (Goldin 1990; see Figure 1.1 in chapter 1). The growth of young and middle-aged women's labor force activity after the U.S. baby boom period of the 1950s stems from the conjunction of sectoral demand (Easterlin et al. 1983), the declining male wage (or "family wage" paid to male workers) (Kamerman and Kahn 1989), and changing family roles (Oppenheimer 1994). Families have become smaller and less traditional in the United States and throughout the wrorld (Goldscheider and Waite 1991; O'Rand and Agree 1993). These are outcomes of growing patterns of delayed marriage or nonmarriage, divorce, lower fertility, and dual-earner marriages (Fuchs 1988; Blau 1998)—all of which are highly correlated with labor force participation across nations and welfare regimes. The Resilience of Gender Structure

One ironic consequence of these long-term patterns is that whereas women's work rates have increased dramatically in recent decades, their relative well-being in the United States has declined. Fuchs (1988) has determined that women's wages have risen as a result of increased labor participation, but their increased responsibility for the economic and social maintenance of families, particularly children, has led to a loss of

Asynchronous Lives: The Normal Lite Course and Its Variations

75

leisure and access to goods and services relative to men. Hochschild's (1997) depiction of the "third shift" in women's lives is a telegraphic characterization of women's new time budgets: work time, household time, arid care-giving time inside and outside of the household preempt leisure time. As such, the increased importance of women's earnings for family income has been more compensatory for the economic maintenance of the family, and less supplementary for surplus or discretionary family income. Blau's (1998) most recent examination of trends in women's economic well-being further confirms these views. Gender and the Educational System

In spite of the effects of wider educational opportunities for standardizing life schedules, gender structure has been preserved and reproduced in the educational system and the workplace via processes of allocation, segregation, selection, and socialization. In the United States the wide availability of general education at the secondary and postsecondary levels has not resulted in the uniform acquisition of skills that translate into integrated occupational locations. Hout, Raftery, and Bell (1993) studied educational stratification in the United States between 1925 and 1989 and found relative female advantage at primary and secondary levels of education completion (young men from economically disadvantaged backgrounds drop out at higher rates than young women), and near equity in postsecondary patterns of completion. Yet educational tracking systems via public-private sectors and curricular differentiation (vocational vs. general education preparation) stratify by class and gender. Other studies identify consequential patterns of differentiation by gender via pathways of selection into academic majors and gendered vocational tracks, with long-term implications for job placement and occupational and earnings attainment. In secondary, postsecondary, and tertiary programs, gender segregation is prominent. Distributions of academic majors are bimodal by gender with science and engineering tracks predominantly male and humanities/social science tracks predominantly female (Wilson and Boldizar 1990). Vocational tracks are even more segregated, channeling trainees into gender-segregated trades like automobile maintenance and electrical and plumbing repair for men and nursing and cosmetology for women (U.S. Department of Education 1994). Persistent inequality in educational attainment extends across all industrialized societies (see Shavit and Blossfeld 1993). Selected comparisons of these differences are provided in Chapter 7. Economic models dominate the educational and occupational selfselection literature. Becker (1985) provides a starting point for this

76

Asynchronous Lives: The Normal Lite Course and Its Variations

perspective with his argument that it is rational for parents to invest more in their sons' than their daughters' education, since the life cycle payoff in earnings among sons is higher. Jacobs (1996) and other researchers criticize this view as, at best, cohort-centric—that is, as relevant to earlier cohorts in the middle of the century following more traditional views gender roles. One of the distinctive features of educational trends in the United States is the predominance of women in the secondary and postsecondary sectors. In 1992, women represented 53.1 percent of all enrolled college students; nearly two-thirds of women who graduated from high school in 1992 enrolled in college the following fall, compared to only 60 percent of men (Jacobs 1996). More recently, parity has been achieved in the progression to graduate (postbaccalaureate) and professional training (Goldin 1995). However, gender segregation is readily observed in educational institutions. For example, a persistent gender gap in education exists with respect to women's presence in top-tier universities. Hearn (1990) and Jacobs (1996) propose a number of contributing factors: the costs of attending top-tier institutions, the dominance of some of these institutions by large engineering programs, and the greater tendency of women to attend postsecondary institutions as part-time students. Their analyses suggest a pattern of cumulative disadvantage in college (subject) preparation and economic resources, including financial aid support for women. In short, educational structures differentially and cumulatively allocate gender groups along different tracks. Selection mechanisms also have operated to differentiate gender experiences and outcomes from education. Based on her study of college cohorts in the 1950s, Goldin (1995) argues that women's self-selection into college has been more motivated by marriage prospects than by human capital investment. Women's investments in college are investments in the marriage market, in which prospective gains are multiplied through marriage. Jacobs (1996) outlines the shortcoming of this argument. Social historians have determined that earlier generations of women college graduates in fact tended to delay marriage while in college. Similarly, the so-called Mrs degree may have had some empirical basis before the 1960s, but after that time considerable survey evidence supports the opposite argument: women's career motives and aspirations have grown more and more specific and predominant. But women's upward trends in educational attainment reflect much more than investment in the marriage market. Modell (1989) argues that succeeding cohorts of men and women alike over the twentieth century have entered adulthood with higher levels of material and cultural resources. Every decade of this century has ushered in fundamental changes in the transition to adulthood. The transition to adulthood

Asynchronous Lives: The Normal Lite Course and Its Variations

77

(Hogan and Astone 1986) is the general umbrella concept that refers to the timing and sequencing of a set of events in late adolescence that includes the onset of sexual behavior, leaving high school, going to college, going to work, getting married, becoming a parent. Modell illustrates that these separate events have become more loosely coupled and reordered over time. These normative changes have come about as more young people at succeeding periods have wider cultural exposure outside of the constraints of family, including within the educational domain. Cultural exposure extends normative options and provides a broader range of life choices. Increasing volition in early life thus produces increased differentiation and individualization later. In short, intercohort trends link educational opportunities for wider sectors of the population to an increased deviation from gender-based role transitions in adolescence. The rise in women's labor force participation and the recent improvement in wage ratios among younger subgroups of women are economic outcomes of these general trends. Byproducts of women's attachment to the market include the rising material and cultural affluence for the U.S. middle class which is now anchored by the dual-earner family structure. In addition, Jacobs (1996) argues that education has become a general consumption item as much as a direct investment in human capital. As such, it is becoming a lifelong factor in the construction and reconstruction of the life course and is no longer restricted to the early years.

Gender Segregation in the Workplace

Gender segregation in the workplace follows from patterns of educational segregation and disadvantage, though recent trends point to improvements in occupational integration (Sorensen 1989; Jacobsen 1994; Jacobs and Steinberg 1990). The Gini coefficient of occupational segregation by gender in 1990 was .68, meaning that about two-thirds of women would have to change jobs in order for there to be occupational integration (Bianchi 1995). Over three-fourths of clerical and related administrative support positions were held by women in that year, although women's share of management positions doubled over the previous two decades (Bianchi 1995). Moreover, gender segregation persists across (and can be amplified by) other structural factors, including firm size (Goldin 1990), level of unionization (Krecker and O'Rand 1991), and industrial rates of part-time work (e.g., Drobnic and Wittig 1997). Part-time work itself is a gendered institution. Hakim (1997) has identified three features of part-time work patterns observable across countries over the past two decades. First, part-time work is growing faster than (and in some sectors is replacing) full-time work worldwide. The

78

Asynchronous Lives: The Normal Lite Course and Its Variations

increase in these work schedules over the past twenty years is evident across regulated and unregulated labor markets, associated with women and concentrated in lower wage (service) jobs (Rosenfeld and Birkelund 1995). Second, part-time work is one component of a broader trend toward nonstandardized and atypical work relationships associated with globalization processes and the related decline in the long-term employment contract discussed elsewhere in this book. Work schedules are becoming more diverse. Nonstandard work schedules are moving away from the (standard) fixed five-day/forty-hour hour per week schedule to work schedules that span day and night to accommodate a twenty-four-hour economy (Presser 1998). In 1991, only 32 percent of American workers followed me traditional schedule; dropping the forty-hour schedule criterion causes the percentage of workers with daytime schedules (working less than thirty-five hours or more than forty hours) to increase to a bare majority of 55 percent. When weekend and night schedules are taken into account, the diversity of work scheduling is unambiguous. Part-time work is a critical element of this diversifying schedule. Part-time jobs are almost exclusively filled by women worldwide, although in the United States they are less female concentrated than in Europe. In the United States about 25 percent of women are employed parttime at any given time, whereas in Europe almost half of all working women are in part-time or related nonstandard work. Country-specific percentages range from a low of 8 percent in Greece to much higher levels of 66 percent in the Netherlands, 44 percent in the United Kingdom, 41 percent in Sweden, and 33 percent in Germany (Blossfeld and Hakim 1997a; 1997b). As in the case of education, self-selection, labor market segmentation based on sex typing, and employer discrimination all operate to preserve the resilience of gender structure in the labor market. Self-selection (choice) results from lifelong socialization to gender roles and from immediate life circumstances or events that impinge upon educational arid work decisions. Family experiences become compounded by school experiences. These, in turn, influence gender-related occupational aspiration and preparation (England 1982; 1992). On average, women and men are as segregated in their educational preparation and career aspirations after they complete their education as before (Jacobs 1996).

Workplace Segregation and the Unequal Division of Household Labor

Cumulative socialization experiences also condition women and men to divide household and care-giving roles unequally, even when both

Asynchronous Lives: The Normal Lite Course and Its Variations

spouses wrork full-time (Kalleberg and Rosenfeld 1990). This division has consequences for the continuity, trajectories, and reward levels of their work careers. Although each child "costs" women in wages and pensions, children often signal higher rewards for men. Shelton and John (1996) suggest that the persistence of unequal household labor stems from a number of interrelated factors. Some are specific to marriages and Others stem from societal institutions that influence marriage generally. Among the marriage-specific factors are the relative economic dependence of women on their husbands, the relative marital power among spouses, and the gender ideology of couples that defines the norms regulating household roles. Wider institutional constraints on the division of household labor include such things as tax exemptions to support child care, employer-based benefits such as family leave, flextime, and worksite day care arrangements, and the availability of part-time work, which has countervailing effects. The effects of part-time work on the division of household labor are complex; they serve to penalize or to reward women by constraining them to secondary market roles contingent on household responsibilities or by permitting them to maintain a joint foothold in family maintenance and market participation. The following section of this chapter develops this argument more specifically. Labor market segmentation has its historical roots in the emergence of factory work and office work and in the structuration of internalized versus externalized labor pools (Jacobs 1990). Women and minorities in the private sector are more likely to fall outside or on the margins of protected (internalized) labor market structures with promotional ladders, income mainteniince benefits like pensions, health insurance, arid family leave, and continuing training opportunities inside and outside the employing organization (Kalleberg 1996; O'Rand 1986). Discrimination stems from the residual influence of stereotypes of worker skills and preferences for work held by employers (England 1982; 1992). The structure of these labor markets and their implications for women's late-life chances are discussed further in Chapter 4 and Chapter 5, which focus on the employment system and occupational welfare in the United States Family Pathways; The Breadwinner and Role-Sharing Models

The resilience of gender structure is maintained in the family or household as well as in the marketplace. The traditional family structure that has developed in tandem with market and state institutions in industrial societies has been labeled the male breadwinner mode] (Sainsbury 1996). This model represents the idealized family division of labor in which husbands are the primary wage earners and pension beneficiaries, and

79

80

Asynchronous Lives: The Normal Lite Course and Its Variations

wives are the primary unpaid caregivers who either do not participate in the labor market or participate only as intermittent, secondary, or marginal workers. As such, wives are dependent on their husbands for economic maintenance throughout the marriage—and even after the marriage in the case of widowhood. This structure has been sustained through the interplay of market, household, and state institutions represented, respectively, in the market wage system, the division of household work, and government policies that treat family members differently. The market wage system has its roots in the so-called family wage, which goes back to the late nineteenth century and early union efforts to increase worker wages and employers' efforts to stabilize the workforce and stave off unionization (May 1982; Jacoby 1985). Tine family wage was predicated on the assumption of the male breadwinner role. It is well illustrated by Henry Ford's introduction of the "five dollar day" in 1914, which was a conscious effort to improve production and achieve stability through profit sharing and family policy. He offered this bonus to married men who "should live with and take good care of their families" (Ford 1925, 129). Along with the wage offer, he also institutionalized "welfare work" within the firm by creating a personnel division ("Sociological Department") to look after the welfare of workers' families, including providing legal services for credit and mortgage assistance. Welfare work was recognized as a widespread practice among employers by a government report in 1916 which defined it as "anything for the comfort and improvement, intellectual and social, of the employees, over and above wages paid, which is not a necessity of the industry or required by law" (see Jacoby 1985, 49-52). The family wage and welfare work have persisted in various forms over most of the twentieth century in the United States. They have evolved as compensation packages to workers providing wages, benefits, and forms of job security (seniority, tenure) developed through management-union accords and in response to government taxation and social insurance policies (Cornfield 1990). These packages are unevenly distributed across labor markets and are highly associated with industrial and occupational sectors that are male dominated. Employee benefits evolved from "welfare work" (Jacoby 1985) before World War II into "hidden payrolls" during and after World War II that were developed in the context of wage controls. Federally imposed limitations on wages encouraged employers during the 1940s to develop "fringe benefits" for their favored workers that allowed them to skirt wage controls. By the mid-1950s, in an environment of growing union-management accords, tax legislation emerged to provide incentives to employers to offer selected fringe benefits, including pensions, to workers. These offers were most likely to be made to union-covered workers and to protected labor

Asynchronous Lives: The Normal Lite Course and Its Variations

81

markets-—usually referred to as internal and occupational labor markets—composed of employees with long-term employment contracts and access to career ladders (Kalleberg 1996). The overwhelming majority of workers covered by protective fringe benefits were male (O'Rand 1986). By the 1990s long-term employment contracts and career ladders are disappearing for large segments of the labor force, just as the proportion of women in the labor force and across occupational sectors approach parity with men (Kalleberg 1996). Long-term arrangements are being replaced by contingent workforces, which consist of part-time or contract workers who are not usually eligible for extensive fringe benefits. The contingent workforce is heterogeneous. High-level contract workers, like accountants or computer software specialists, often negotiate generous arrangements that permit them to provide for their own pension and health care needs. Low-level workers are typically part-time service or clerical workers without market power and without benefits; the latter are more likely to be women (Drobnic and Wittig 1997). This segmentation of part-time work is gendered and congenial with both the traditional breadwinner model of the family and the occupational segregation system of the workplace. State and public structures also reward or otherwise reinforce these traditional family arrangements through taxation, social insurance, and public policies. Skocpol's (1992) historical analysis of the origins of the U.S. welfare system distinguishes collaterally developing "maternalist" and "paternalist" policies that have produced gendered public support structures; "mothers" and "soldiers" typify the categorization of sets of public policies earmarked for different groups based on their family or work roles. Achenbaum's (1986) history of the U.S. Social Security System from its establishment in the 1930s through the early 1980s depicts a sequence of amendments and extensions of the system that reinforces its contributory market base yet incrementally extends benefits to formally defined vulnerable populations outside of the market system (particularly dependents and survivors of workers, the infirm and disabled, and the poor). Social Security was enacted to protect retired workers; later extensions of the act were clearly designated for dependents and survivors of workers first, and then for indigent and marginalized persons who fell outside of the employment system. Hence the system has both reproduced the breadwinner model and provided an ameliorative or equalizing balance to the market's capacity to deliver "comfort and improvement" to workers and their families. In effect, incremental changes over the history of U.S. welfare policies have reinforced gender categories. Orloff (1996) further characterizes the U.S. welfare system as a two-track system of welfare in the United States—a contributory social insurance

Asynchronous Lives: The Normal Lite Course and Its Variations

82

track for workers and a family-related, means-tested track for nonworkers and dependents—that is highly correlated with gender. Income and social security tax structures also reproduce this model, by penalizing nontraditional family arrangements and rewarding traditional structures (Sainsbury 1996). The adoption of the logic of community property after World War 11 changed the unit of taxation from the individual to the family. Following this logic, tax schedules were developed that privileged marriage in general, and single-earner or primary-secondary earner couples specifically; dual-worker (career) couples with more equal earnings were penalized. This structure persists. This income tax structure has been accompanied by similar privileging in the Social Security benefit system that favors spouses in traditional marriage relationships over single individuals and dual-career couples. Table 3.1 is taken from Burkhauser and Smeeding's (1994) calculations from the Social Security Administration Office of the Actuary, June 1994, benefits database. The table assumes that husbands and wives in both couples are age 65 in 1995 when they retire and that the workers began their earnings history at age 22. The one-earner couple example assumes that the husband has earned the taxable maximum over his career and that two-earner spouses have each earned one-half the taxable maximum. The latter assumption is a stringent one when considering the pervasive wage inequality in the market. However, the two ideal types of couples define the extremes of the benefit system and exemplify the breadwinner role ideology. They both pay the same amount in taxes, but (1) the one-earner couple receives higher benefits as retirees and (2) the survivor of the one-earner couple receives a higher benefit as widow.

TABLE 3.1 1994

Couple One earner Husband Wife Total Two earner Husband Wife Total

Benefits Payable to Couples with Identical Total Earnings Through

Average Lifetime Earnings

Social Seen rity Benefits Couple

Survivor

$60,600 — 60,600

$14,400 7,200 21,600

14,400

30,300 30,300 60,600

9,636 9,636 19,272

9,636

Survivor Benefit/ Couple Suririvor

2/3

1/2

SOURCE: Data from the Social Security Administration Office of the Actuary, June 1994, with calculations reported in Burkhauser and Smeeding 1994.

Asynchronous Lives: The Normal Lite Course and Its Variations

83

Finally, the division of household work both defines and reproduces the breadwinner model of the family. Neoclassical theories of home economics argue that tine traditional family structure is functional in the division of labor, since it assigns complementary roles to the wage earner and caregiver (Becker 1985). The caregiver role consists of unpaid time allocated to household, family, and kin care. Research on changing allocations of market and home-work time reveals the persistence of the breadwinner model even in households with full-time worker couples. Among dualworker couples, women assume between two-thirds and fourth-fifths of all caregiving and housekeeping tasks (Shelton and John 1996). The level of household task sharing appears to vary according to the couples' earnings gap, level of education, and egalitarian attitudes (see Presser 1994). Smaller differences in earnings (as well as relatively equal higher earnings) are associated with more household task sharing. Similarly, a husband's higher education is related to more household work and a wife's higher education is related to less household work. Egalitarian attitudes toward joint role sharing among couples, which appear to be highly associated with educational level, also appear to influence the division of household labor above and beyond the effects of education and earnings. Finally, the most recent studies suggest that the spread of nonstandard work schedules may also affect the division of household work, above and beyond the effects of education, earnings, and gender role ideology. Presser's (1998) analysis of work and household task schedules reported in the 1991 Current Population Survey of 55,000 households finds that couples with nonstandard work schedules are constrained to share household tasks, including child care, more than those with standard schedules. These schedules confront problems with day care and night care that are not anticipated by traditional family or workplace institutions. In sum, the life courses of men and women are organized similarly by educational and labor force schedules. Resilient institutions attached to gender and family roles serve to differentiate men's arid women's experiences, although recent global trends are eroding these institutions. Succeeding cohorts across countries are confronted with new challenges and choices in the construction of their life course. These choices are constrained by educational and workplace opportunity structures. The breadwinner model of family structure, which continues to be reinforced by institutional arrangements emanating from the workplace and the state, is being challenged by the press of globalization processes and demographic changes related to population aging, educational attainment, and women's labor force participation. Alternative family models have spread in recent decades. Dual-earner couples, single-parent families, single persons, and cohabiting house-

84

Asynchronous Lives: The Normal Lite Course and Its Variations

holds are growing faster than the traditional breadwinner unit. Unmarried-couple households tripled between 1960 and 1990; by 1990 twothirds of these households included children (Goldscheider and Waite 1991). Approximately six out of every ten married-couple families in the United States are dual earners—a 33 percent rise in only two decades. Together, dual-earner couples and unmarried couple households account for most households with children in the 1990s (Bianchi 1995). These alternative forms of family arrangements condition men's and women's patterns of market and household labor. Cross-national comparisons of family formation in industrial societies provide compelling evidence for the spread of the role-sharing model. Blossfeld (1995a) has compiled a number of studies from continental European countries, the United Kingdom, and the United States on changing gender roles. His principal hypothesis is that the increased educational attainment of women across countries has had a "liberating effect" on the timing of their entry into marriage and motherhood, on the division of labor in the household, and on their labor force attachment. He does not argue that the breadwinner model has been wholly superceded by rolesharing, but he notes the strong upward trend in the direction of the latter. The case of Sweden sets the standard of "liberation" by historically demonstrating the earliest trend away from traditional marriage immediately preceding and following World War II. Declines in marriage rates and fertility were earliest and steepest in Sweden. Blossfeld (1995a; 1995b) argues that economic factors were less important at this time and changing cultural values and religious traditions more important. Cultural values that emphasized egalitarianism and individual freedom in public life and secularization in private life encouraged processes of individualization that eroded gender structure. Over the following decades, state institutions have largely "sponsored" this liberation through generous day care and family leave policies, including the largest system of public day care facilities in the world (Sainsbury 1996). Other countries fall well below Sweden in this respect and reflect piecemeal approaches to supporting role sharing with integrative work and family policies (Kalleberg and Rosenfeld 1990). Germany lies at the opposite pole from Sweden. It adheres more closely to the breadwinner model, and its female labor force participation patterns, which are very low, reflect state policies that operate to keep women within the household. The United States is located between these two systems. These patterns receive more attention in Chapter 6 and Chapter 7. However, across countries a significant countervailing trend is notable—and highly relevant to the purposes of this study. An observation from Blossfeld's collection of cross-national studies (1995a; 1995b) is that although educational opportunities have expanded for men and women

Asynchronous Lives: The Normal Lite Course and Its Variations

85

alike and have presented them with more individualized options for work and family formation, educational homogamy has also increased. Assortative marriage patterns pervade industrialized countries resulting in persistent social and economic inequality. Blau's (1998) recent study of trends in women's well-being over the past three decades (cited in Chapter 1) confirms these observations for the United States. She reports that differential educational advantages stratify women's fortunes in the marketplace and through homogamous marriage. Men's fortunes are also stratified. According to Oppenheimer (1994), the trends in women's labor force participation and marriage patterns operate to the disadvantage of less educated, underemployed, and unemployed men. Declining economic opportunities for subpopulations of men marginalize them in the workplace and constrain their marriageability. This is perhaps most starkly revealed among the urban poor in the United States (Wilson 1996). Thus men's and women's commensurate role investments in the workplace and the household are increasing and challenging the male-breadwinner model with an alternative role-sharing model. The role-sharing model indicates a decline in the gender structuring of the life course as women and men follow new pathways through work and family. It may also reflect the rise of the individualization of the life course, wherein men and women increasingly share universal experiences in education, work, and family, but not necessarily at the same time in their lives and with the same resources and rewards. Asynchronous Lives

Men's and women's lives are challenging the age-segregated paradigm of the life course based on age-graded, sequential roles over the life span in spite of the resilience of gender structure in the family and the workplace in the United States and other countries. In Figure 3.1, the age-integrated model of the life course in the second panel presents an alternative model. This model explicitly suggests that events and transitions within the domains of education, work, and leisure can occur independently of age; implicit in this model is the notion that family events are also independent of age and can occur and recur across the life span in a similar fashion (Riley and Riley 1993; 1994; Riley 1998). The model also proposes that as a result of age integration, roles across domains are more interdependent over time. Market time and domestic time are becoming more integrated and simultaneous over the life course, since men and women increasingly allocate equal shares of their lifetime to co-occurring, as opposed to sequentially arrayed, roles as workers, spouses, parents, students, and children. Market time and domestic time are thus less gender segregated and decreasingly age segregated.

86

Asynchronous Lives: The Normal Lite Course and Its Variations

Nowhere is this more evident than in the labor force behavior of married women over the past four decades (Blau 1998; Goldin 1990) and, more recently, in the behavior of women with young children. Mothers with children under the age of 6 increased their labor force participation patterns between 1950 and 1995 by a factor of five (from 12 percent in 1950 to 62 percent in 1995) (Hayghe 1997). Bianchi (1995) argues that the latter shift reflects nothing short of a normative transformation: a change so fundamental that the right of young mothers to remain at home is no longer taken for granted. Rather, the pressures are to maintain these roles simultaneously. She reports that this change is spreading to other developed countries as well. Participation in multiple roles over time, simultaneously a n d / o r sequentially, is a fundamental source of variability and inequality in cohorts (O'Rand and Krecker 1990). The variability in the timing of transitions across these activities produces diversity in life scheduling and in social and economic outcomes over the life course. The timing and outcomes of events in one domain (say, education) influence transitions in other domains (family and work). With age, the possible array of sequences of timing and influence defies easy calculation and classification. Longitudinal studies of individuals moving through critical life transitions, such as from school to work, school to marriage, marriage to work, work to retirement, and so on, have uncovered such diversity in sequences that they appear to be "disordered" (Rindfuss, Swicegood, and Rosenfeld 1987). But closer examination of life course sequences is revealing that the variability observed is not so much "disordered" as it is asynchronous. Role domains are regulated by different timetables based less on age than on institutionalized status sequences and durations. Timetables introduce temporal contingencies that regulate the conduct of specific roles and that may be asynchronous with other role schedules and result in role conflicts or change. Instances of asynchronous timetables in the family and work domains abound, especially among women who have exhibited more heterogeneity in their management of multiple roles (Moen 1994). For example, since some women tend to enter their "careers" (or longest jobs) later in their lives than men as a result of earlier caregiving demands, their pension participation schedules tend to be delayed when compared to men's, leading some women to retire later from these jobs rather than earlier. Other women who enter their work career earlier and do not interrupt it for family obligations retire earlier (Ruhm 1990; O'Rand and Henretta 1982b; Henretta, Chan, and O'Rand 1993). The asynchronies involved in this example stem from multiple "clocks" that regulate life course transitions in work and family domains. Han and Moen (1998) study retirement in the same vein as "clocking

Asynchronous Lives: The Normal Lite Course and Its Variations

87

out," that is, as a multiplex process governed by multiple institutional schedules and by the diverse "pacing" of individual biographies, which leads to variability in the timing of retirement. They argue that individual biographies intersect with institutional timetables and yield diversity in retirement. In Chapters 4 and Chapter 6, the institutional timetables embodied in public policies and in the employment sector will be explored in detail as they relate to retirement schedules and labor force participation patterns worldwide. Studies of Asynchronous Lives and Inequality

In the remainder of this chapter, we will review three patterns of variability in role transitions in middle and late life that illustrate age integration and the mutual contingency of roles across work and family domains. The areas include patterns of midlife reentry into education, marital dissolution associated with education, labor force participation, and family life cycle, and interdependent family-work pathways to retirement. Besides illustrating variability and asynchrony, these patterns also demonstrate how inequality emerges from the interaction of institutional and biographical clocks.

Education in Midlife

The first domain of interest is education. Two major assumptions in the stratification and life course literature are that in the United States, major transitions in education and training end early in adulthood and that postsecondary or postcollegial skills training is employer provided. But recent trends in aggregate patterns of job mobility in midlife appear to be correlated with midlife reentry into non-employer-provided formal education and vocational training for some groups (Elman and O'Rand 1998a, 1998b). Increases in job exit and job shift patterns among middleaged workers appear to be motivating more workers in this age-group to return to formal education or to some form of vocational training. Economic restructuring and technological changes calling for the upgrading of skill levels are among the structural conditions underlying these trends, which are also implicated in labor force participation patterns generally and in retirement scheduling specifically (see Chapters 6-7). Women have always constituted a substantial fraction of adults returning to (or continuing in) school. In 1991 over one-third of college students were over age 24. Women represented nearly two-thirds of those enrolled. But even older men and women were enrolled in substantial numbers that year. Elman and O'Rand (1998a) use adjacent October Supplements of the Current Population Surveys in 1990-1991 merged with the

88

Asynchronous Lives: The Normal Lite Course and Its Variations

Dictionary of Occupational Titles and selected aggregate labor statistics to study midlife job shifts and educational reentry patterns among persons aged 45-61. The study finds that about 2 percent of these middle-aged and older workers participate in vocational training. The national estimate of the on-the-job training ever received across all age-groups is 10 percent; therefore the 2 percent figure is substantial. Women are more likely to reenter into both educational settings. Elman and O'Rand (1998a) observe significant relationships among gender, educational background, and entry into vocational training. Reentry increases with educational background up to grades thirteen to sixteen for both men and women and with women's reentry rates higher. However, men with backgrounds of grades seventeen or more are about twice as likely as women to return for training. These findings suggest a status maintenance process in which earlier educational attainment translates into continued education and training and later life. Multivariate logistic analyses controlling for occupational status, work-related skills (computer use), and industrial location provide support in this study for both the status maintenance and cumulative advantage processes. After controlling for the effects of earlier educational attainment, more highly skilled and advantaged workers are likely to reenter vocational training in midlife, thus increasing their market power relatively with age. Finally, besides gender, widowed status also increases the likelihood of retraining (Elman and O'Rand 1998a). Changes in family roles are linked to changes in work and educational roles over the life course and not only in early adulthood. Asynchronies develop over time and, in the case of educational reentry in midlife, demonstrate variations from the normal life course defined by the age-differentiated model in Figure 3.1. These results reveal the multiple contingencies of this exceptional midlife pattern. A change in multiple role participation precipitates educational reentry. In the case of women between the ages of 45 and 61, who are at a life stage usually associated with decreased household role demands, the onset of widowed status may similarly produce role changes in other domains—in this case a shift from work to further schooling. Educational reentry at midlife is an age-integrated role transition that, according to some sources (Office of Technology Assessment 1986), is expected to rise over the next few decades. Organizational restructuring and changing skill demands in the workplace provide incentives for later education, assuming local opportunity structures for retraining. Also, given rising levels of early educational achievement across successive cohorts and the triggering effects of changes in family roles related to marital dissolution through widowhood or divorce, this pattern may become more prevalent.

Asynchronous Lives: The Normal Lite Course and Its Variations

Marital Dissolution in Middle and Late Life

The case of marital dissolution in midlife provides a second example of life course variability, age integration, and relational and temporal contingency and asynehrony. In Chapter 2, trends in family patterns—especially the declining centrality of marriage over the life course—were reviewed iii the context of intercohort trends. Separation and divorce are more widespread events than midlife educational reentry. Between onehalf and two-thirds of all first marriages are likely to end in divorce or separation (Castro-Martin and Bumpass 1989). Although most divorces occur in the first few years after marriage, their temporal incidence varies considerably. The dissolution of marriages of longer duration among middle-aged populations provides the opportunity to examine several topics of interest to us here. First, the causes and consequences of divorce have implications for patterns of inequality. The major hypotheses predicting marital disruption center on the countervailing effects of two economic characteristics. The economic independence hypothesis predicts that, all else equal, women's relative economic independence, usually measured as wages arid market participation rates, will increase the risk of divorce. The economic status hypothesis, on the other hand, predicts that couples with higher levels of income and wealth are likely to lose more by divorce; therefore higher economic status measured as family income and net worth or homeownership will decrease the risk of divorce. Women tend to experience greater drops in relative economic status following divorce than men, especially in marriages of longer duration. Second, the examination of divorce in marriages of longer duration provides the opportunity to observe a nonnormative transition regulated by relational and temporal contingencies. Family development theory views the family as a changing configuration of interdependence among family members over time that is organized by processes of investment, commitment, and solidarity (Mattesich and Hill 1987; Aldous 1978). Longer marriages have survived earlier periods of higher risks for divorce and reflect partners' growing investment in children, property, and each other over time. One of the risks in the survival of marriages is maturing children. A U-curve of marital satisfaction over the marital career depicts a trajectory of marital satisfaction that begins at a positive level, declines with the appearance and maturing of children, and finally improves with the onset of empty nest. Accordingly, the survival of a marriage is, at least partially, a function of the interaction between the duration of a marriage and its configuration over time. Hiedemann, Suhomlinova, and O'Rand (1998) use the National Longitudinal Surveys of Mature Women between 1967 and 1989 to apply a

89

90

Asynchronous Lives: The Normal Lite Course and Its Variations

family development perspective to midlife patterns of marital disruption. They use hazards models to examine three hypotheses. Two are summarized above as the economic independence and the economic status hypotheses; the first predicts an increase in the rate of divorce and the second predicts a decrease in the rate of divorce. The third hypothesis they examine predicts an increase in the rate of divorce with the onset of empty nest in marriages of shorter duration. The logic of the third hypothesis is drawn from family development theory, which argues that marriage is a duration-dependent process that is driven by the changing interdependence of marital partners over time. Hiedemann and colleagues find support for all three hypotheses. In the case of empty nest, the pattern that emerges can be summarized as follows. For couples married twenty years, the last child's departure from the household triples the risk of marital disruption. They also find that in marriages of this duration the youngest child's eighteenth birthday increases the risk of marital dissolution by 56 percent. However, for marriages that have lasted over thirty years, the youngest child's departure or eighteenth birthday decreases the risk of marital disruption. Divorce has especially deleterious effects on older women's economic security, if they spent limited time in the workforce or worked in jobs not covered by pensions and other income maintaining benefits. Access to former spouses' Social Security benefits legislated under the 1983 Amendments to the Social Security Act has not solved problems of income security for this group. The divorced wife's Social Security benefit is one-third of the Social Security income of the retired couple that includes her former spouse and his new spouse. Also, because most women gain access to health insurance through their husband's group insurance plans, they are doubly vulnerable to poverty or near poverty. Divorce £ind widowhood result in losses of income, assets, health insurance coverage, and quality of life. The risk of falling into poverty after divorce has been the subject of considerable research and controversy. Estimates of women's income loss after divorce range from 21 percent to over 70 percent. Recent conservative comparisons of income loss among divorced women who were followed in the Panel Study of Income Dynamics, who were married ten years or less and eighteen years or more respectively, suggest that women from longer marriages lose relatively more income. Across predivorce annual family income categories, divorcees from intermediate to high predivorce income categories experienced net losses from divorce, whereas those from the lowest income category (