Chapter 48 RACE AND GENDER IN THE LABOR M A R K E T JOSEPH G. ALTONJI* Institute Jbr Policy Research and Department of Economics, Northwestern University and NBER REBECCA M. BLANK* School of Public Policy, University of Michigan and NBER Contents Abstract 3144 JEL codes 3144 1 Introduction 3144 2 A n overview of facts about race and gender in the labor market 3146 2.1 Trends and differences in labor market outcomes and background characteristics 3146 2.2 Methodologies for decomposing wage changes between groups 3153 2.3 Estimating simple models of wage determination 3156 2.4 Estimating simple models of labor force participation 3161 3 Theories o f race and gender differences in labor market o u t c o m e s 3164 3.1 The impact of group differences in preferences and skills 3165 3.2 An introduction to theories of discrimination 3168 3.3 Taste-based discrimination 3170 3.4 Discrimination and occupational exclusion 3176 3.5 Statistical discrimination, worker incentives, and the consequences of affirmative action 3180 4 Direct e v i d e n c e on discrimination in the labor market 3191 4.1 Audit studies and sex blind hiring 3192 4.2 Discrimination in professional sports 3195 4.3 Directly estimating marginal product or profitability 3196 4.4 Testing for statistical discrimination 3198 5 Pre-market h u m a n capital differences: education and family b a c k g r o u n d 3201 5.1 Race differences in pre-market human capital 3201 5.2 Gender differences in pre-market human capital 3204 6 Experience, seniority, training and labor market search 3207 6.1 Race differences in experience, seniority, training and :nobility 3208 6.2 Gender differences in experience, seniority, training and mobility 3213 * We are grateful to the Russell Sage Foundation and Institute for Policy Research for research support, and to Rachel Dunifon, Todd Elder, Raymond Kang, Joshua Pinkston, and James Sullivan for excellent research assistance. We also thank Orley AshepXelter and David Card for their patience and encouragement and partici pants in the Handbook pre-conference for helpful suggestions. All errors and omissions are our responsibility. Handbook of Labor Economics, Volume 3, Edited by O. Ashenfeher and D. Card �� 1999 Elsevier Science B.V. All rights reserved. 3143

3144 J. G. Altonji and R. M. Blank 7 Job characteristics, taste differentials, and the gender wage gap 3220 7.1 Overview 3220 7.2 The occupational feminization of wages 3221 7.3 The impact of other job characteristics 3223 8 Beyond wages: gender differentials in fringe benefits 3224 9 Trends in race and gender differentials 3225 9.1 Methodologies for decomposing wage changes between groups over time 3225 9.2 Accounting for trends in the black/white wage differential 3234 9.3 Accounting for trends in the male/female wage differential 3240 9.4 The overlap between race and gender 3244 10 Policy issues relating to race and gender in the labor market 3244 10.1 The impact of anti-discrimination policy 3245 10.2 The role of policies that particularly affect women in the labor market 3247 11 Conclusion and comments on a research agenda 3249 References 3251 Abstract This chapter summarizes recent research in economics that investigates differentials by race and gender in the labor market. We start with a statistical overview of the trends in labor market outcomes by race, gender and Hispanic origin, including some simple regressions on the determi- nants of wages and employment. This is followed in Section 3 by an extended review of current theories about discrimination in the labor market, including recent extensions of taste-based theories, theories of occupational exclusion, and theories of statistical discrimination. Section 4 discusses empirical research that provides direct evidence of discrimination in the labor market, beyond "unexplained gaps" in wage or employment regressions. The remainder of the chapter reviews the evidence on race and gender gaps, particularly wage gaps. Section 5 reviews research on the impact of pre-market human capital differences in education and family background that differ by race and gender. Section 6 reviews the impact of differences in both the levels and the returns to experience and seniority, with discussion of the role of training and labor market search and turnover on race and gender differentials. Section 7 reviews the role of job characteristics (particularly occupational characteristics) in the gender wage gap. Section 8 reviews the smaller literature on differences in fringe benefits by gender. Section 9 is an extensive discussion of the empirical work that accounts for changes in the trends in race and gender differentials over time. Of particular interest is the new research literature that investigates the impact of widening wage inequality on race and gender wage gaps. Section 10 reviews research that relates policy changes to race and gender differentials, including anti-discrimination policy. The chapter concludes with comments about a,future research agenda. �� 1999 Elsevier Science B.V. All rights reserved. JEL codes: J7 J15 J16 1. Introduction Race and gender differentials in the labor market remain stubbornly persistent. Although the black/white wage gap appeared to be converging rapidly during the 1960s and early

Ch. 48: Race and Gender in the Labor Market 3145 1970s, black/white male wages have now stagnated for almost two decades. The black/ white female wage gap has actually risen over the past 15 years. The Hispanic/white wage gap has risen among both males and females in recent years. In contrast, the gender wage gap showed no change in the 1960s and 1970s. Not until the late 1970s did it begin to converge steadily (although a significant gender gap still exists). Of course, these wage gaps are only the most visible form of differences in labor market outcomes by race and gender. Substantial differences in labor force participation, unemployment rates, occupa- tional location, non-wage compensation, job characteristics and job mobility all exist by both race and sex. This chapter is designed to provide an introduction into the literature that analyzes these differences. As we shall show, there are significant differences in the discussion of race versus gender. Where appropriate, we deal with both issues simultaneously, but in many sections we deal with race and gender differences sequentially, both because the literature on the two is quite distinct and because the conceptual models behind race and gender differences are often dissimilar. It is important to note that our use of the term "race" in this chapter is extremely limited. With only a few exceptions, we discuss black/white differences in labor market outcomes throughout this chapter. This reflects a major lack in the research literature. There is remarkably little empirical work on Hispanic/non-Hispanic white differences or on Hispanic/black differences in labor market outcomes. There is even less empirical work looking at other racial groups, such as Asian Americans or American Indians. In part, this reflects a lack of data on these groups. However, the widespread availability of Census data and an increase in the race/ethnic categories in a host of datasets makes this excuse increasingly inadequate. We strongly hope that future research will remedy this gap, investigating many of the issues that we discuss here for other labor market groups. The chapter attempts to summarize some of the most important research areas relating to race and gender in the labor market. Of necessity, there are topics which we will cover inadequately or not at all. In Section 2 we provide a statistical overview of the differentials by race and gender in the labor market. Section 3 discusses theories about how race and gender differences in the labor market arise, with particular attention to new theoretical developments integrating costly search into models of discrimination. In Section 4 we begin our review of the empirical literature by considering recent studies that provide what we consider to be direct evidence on the role of discrimination, a literature that is remarkably small. In Section 5 we examine the role of differences in human capital accumulation prior to labor force entry, touching on the recent literature on the role of race differences in basic skills, and the literature on the role of differences in the type of education that women receive on the gender gap in wages and occupational location. Section 6 considers the contribution of experience, seniority, training, and labor market search to race and gender differentials. In Section 7 we consider the consequences of different job characteristics for the gender wage gap, including the effects of occupational location, the "feminization" of occupa- tions, and the impact of part-time and temporary jobs. This research is closely related to

3146 .L G. Altor~]i and R. M. Blank the extended and controversial discussion about the extent to which these differences are related to taste differentials versus constraints in the types of jobs available to men and women. While most of the chapter focusses on wage differentials, and to a lesser degree, employment rate differentials, in Section 8 we discuss the much smaller literature on the race and gender differentials in fringe benefits. Perhaps more high quality research has been devoted to the analysis of changes over time in race and gender differentials than any other topic in this chapter. This has been a very active area over the past 10 years, and the work has been closely connected to more general analyses of changes in wage structure and the rise in inequality. Section 9 begins with a presentation of the standard methodology for decomposing wage changes between groups and then turns to research on the effects of changes in the prices of observed and unobserved skills. Our emphasis is on recent methodological developments. In Section 10 we consider the effect of labor market policy on labor market outcomes. We summarize the research evaluating the impact of anti-discrimination legislation, and also briefly review two areas where policy has had large impacts on female workers, namely, the impact of maternity leave benefits and the impact of comparable worth legislation. We close with a few comments on a future research agenda in Section 11. 2. An overview of facts about race and gender in the labor market 2.1. Trends and differences in labor market outcomes and background characteristics Race and gender differentials in the labor market have been persistent over time, although the nature and magnitude of those differences have changed, as this section discusses. We begin with a basic set of facts about gender, race, and Hispanic/white differences in labor market outcomes and in personal characteristics (such as human capital measures) that are likely to be related to labor market outcomes. We then provide some simple estimates of how differences in wages and employment are related to differences in characteristics and differences in labor market treatment given characteristics. One purpose of this analysis is to illustrate with the most recent data the basic regression techniques that have been used in hundreds of labor market studies of race and gender differences. We particularly discuss the difficulties that arise in differentiating between the effects of labor market discrimina- tion and the effects of race and gender differences in preferences and human capital. Table 1 shows a current set of key labor market outcomes for all workers, for white, black, and Hispal~c male workers, and for white, black, and Hispanic female workers. It is based on tabulations of the C~iarrent Population Survey (CPS) data from March 1996. Row 2 of Table 1 indicates that black and Hispanic men as well as white women earn about two-thirds of that earned by white male workers on an hourly basis. Black and Hispanic women earn even less than minority men, only slightly over half of what white males earn. Figs. 1 and 2 show median weekly earnings among full-time male and female

Ch. 48." & Race and Gender in the Labor Market -4 C~ -d ~ ~ . ~ o ~ o ~ o'3 ,~- u'] ',o r-- ~ c q c~ o ~5~5 c5 u'] o o oo c'-I o o~ ~5~5 cq.o. o o c3 oo & ~ k [ 3 ��� 0 o ~5 ~5 3147 Z C~ ..= �� % E b~ "5 5", b c~

3148 800 J. G. Altonjiand R.M. Blank 700 6O0 o "o 500 I'..- (33 o) T - - 40O 300 1965 . . " , - . . . . . . . - " . . . . . . . , . - " . . . . , - I " ' " " ". White ................. . . . . . . . . . ��� . . . . . . . . . . . . . . . . . . . , ' - , . . . . . . . . . . . - . . . . . . . . . . . . . . . . " " " . . . . . . . . . Hispanic ......................... i i i i I i i i i I i r i r F i ~ ~ i t r ~ i i I i I i i I i i 1970 1975 1980 1985 1990 1995 Fig. 1. Median weekly earnings of full-time male workers. Source: Bureau of Labor Statistics. 800 700 600 o~ ' o 500 ob 4OO 300 ... White ....... , . - . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - . . . . . . . . . . . . m " ' " ~ . . . . . . . . . . . . . _ . . . . . . - - - - ....... ~ i s a ~ " ................................................. T 1965 1970 1975 1980 1985 1990 1995 Fig. 2. Median weekly earnings of full-time female workers. Source: Bureau of Labor Statistics.

Ch. 48: Race and Gender in the Labor Market 3149 workers from 1967 to the present fbr whites and blacks and from 1986 to the present for Hispanics. 1 The wage trends in these two figures reveal that women, particularly white women, have experienced an increase in their earnings relative to men. But after declining in the 1960s, wage gaps have widened among racial/ethnic groups for both men and women. Although black men's wages rose faster than white men's in the 1960s and early 1970s, there has been little relative improvement (and even some deterioration) in the 25 years since then. Both white and black men show declines in their median weekly earnings over the last decade. Hispanic men show the strongest recent wage declines, but some of this is due to immigration, which has brought an increasing population of less-skilled Hispanic men into the workforce. Among women, white women's wages have risen steadily since 1980, as Fig. 2 indi- cates. Black women's wages almost reached parity with white women in the 1970s, but have diverged again in the last 15 years, as black women have experienced little wage growth. Hispanic women, like Hispanic men, are doing relatively worse over the past decade, in part because of shifts in labor force composition due to immigration. Annual earnings (shown in row 3 of Table 1) show an even larger differential than hourly wages, suggesting that weeks and hours worked are lower among minorities and females. Indeed, rows 4 and 5 confirm that white men not only earn more per hour, they also work more weeks per year and more hours per week. These differences are less among full-time/full-year workers as rows 8 and 9 indicate, but they are still substantial. Row 6 shows that women are particularly likely to be working part-time. Consistent with the weeks and hours data, rows 10-13 indicate that white men are more likely to ever be employed over the past year and to be employed at any point in time. Unemployment among white women has been as low or lower than among white men since the early 1980s. Blacks have about twice the unemployment rates of whites. Figs. 3 and 4 graph unemployment rates from 1955 to the present among men and women and between whites, blacks and Hispanics. Unemployment rates are quite cyclical among all groups of men, although black male unemployment is more cyclical than white male unemployment. The differential between black, white and Hispanic male unemployment rates is remarkably constant over much of this time period. W o m e n ' s unemployment has been less cyclical than men's. As has occurred with their wages, the gap between black and Hispanic women's unemployment rates and white women's unemployment rates is higher over the 1980s and early 1990s than it was in the early 1970s. Wages and unemployment rates are often affected by overall labor force participation rates, which have changed dramatically over time. Labor force participation rates by race and gender are shown in Fig. 5 from 1955 to the present. This chart clearly depicts the convergence in labor force participation among all groups. Men have experienced a steady i Data for Figs. 1-5 are from the Bureau of Labor Statistics, tabulated from the Current Population Survey. Prior to 1972, the data for blacks includes all non-whites. Beginning in 1979, the data in Figs. 1 and 2 are for workers ages 25 and over.

3150 25 J. G. Altonji and R. M. Blank . . o 1 5 c- (11 P 11) 10 20 ",. "t.\ "t '.. .-, " i '"', .."" : ........ '-. ,-"" '",. 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 Fig. 3. Male unemployment rates (annual averages). Source: Bureau of Labor Statistics. decline in their labor force involvement, with the largest declines among black men. Women have shown dramatic increases in labor force participation over these years. 25 2 0 .~ 15 c 1 0 1 9 5 0 1 9 5 5 1 9 6 0 1 9 6 5 1 9 7 0 1 9 7 5 1 9 8 0 1985 1990 1995 Fig. 4. Female unemployment rates (almual averages). Source: Bureau of Labor Statistics.

Ch. 48: Race and Gender in the Labor Market lOO 3151 o d.) 13_ 9O 8 0 7 0 6O 5O 4 0 0 1 9 5 0 ...................................................................................... wh_!t_e =============================================================================================== ..J'" White w o m e n . , ............ . . - - i F r i I i i r i r q i i i P i i i t I i i i i I r i i i F i i i i I , i i i I i i i i I i , 1955 1960 1965 1970 1975 1980 1985 1990 1995 Fig. 5. Labor force participation rates, 25 54-year-olds. Source: Bureau of Labor Statistics. White women have entered the labor market at a particularly high rate. While their rates of labor force participation used to be far lower than those of black women, they are now at parity. Hispanic women's labor force participation, although rising steadily, is still far below that of black and white women. In delineating the causes of these labor market differences, labor economists look first at the substantial differences in the attributes that different workers bring with them to the workplace. Table 2 shows a set of key personal characteristics among all persons in 1996, and among the same six race/gender groups observed in Table 1.2 Educational differences among these groups are large, with race and ethnicity mattering much more than gender. Both male and female Hispanics have particularly low education levels. White women's educational levels are quite similar to white males (this was not true in earlier periods), while blacks have less education than whites but more than Hispanics. These differential investments in education may reflect different preferences and choices, and/or they may reflect "pre-market" discrimination. For instance, there is substantial evidence that blacks have been consistently denied access to suburban housing and crowded into inner city residential neighborhoods with substandard schools. Under these circumstances, blacks will receive a poorer public education and may leave school earlier. Row 7 of Table 2 shows a "potential experience" calculation, based on calculating (age - years of education - 5) for each individual. This calculation assumes that people are working during all their adult years when they are not in school. Although this variable 2 The results in Table 2 would not be very different if the tabulations included all workers rather than all persons.

3152 " a �� C9 �� ..~ZZ d d ~ d d ~ d d ~ d ~ d ~5 d d ~ d d ~ d d o, o o ~ d d ~ ~ d o o ~ d d ~ d L d v ~ , ~ _ ~ ~ ~ .~ . ~ , ~ J. G. Altonji and R. M. Blank ~ Q O Q ~ O 0 0 0 d d & d d d d d d ~ o N N~ N �� cq

Ch. 48: Race and Gender in the Labor Market 3153 is commonly used because many datasets lack information on actual experience, it is a particularly poor proxy for experience among women, who are more likely to leave the labor market during their child-bearing years. We return to this point below when we look at alternative data with information on actual experience. Rows 8-10 of Table 2 indicate that the family and personal commitments of different workers also vary substantially. Whites are much more likely to be manied Hispanics have more children to care for and black females have greater child care responsibilities than black males. To the extent that family responsibilities influence labor market choices and create labor market constraints, these differences may be important in explaining differences in labor market outcomes. Rows 11-20 of Table 2 indicate substantial variation in the geographic location of different groups. Blacks are more likely to be in the southern regions and Hispanics are more likely to be in the western regions. Minorities are also far more likely to be in major urban areas (a relatively recent shift for black Americans, who were traditionally more likely to be located in rural areas.) As Bound and Freeman (1992) and Bound and Holzer (1993, 1996) emphasize, to the extent that local labor markets differ and that labor is largely immobile in the short-run, 3 these differences in regional location will also shape labor market outcomes. Table 3 looks at occupation and industry differences by race and gender. As others have observed, these differences are large. Black and Hispanic men are more likely to be in less skilled jobs. Women are generally more likely to be in clerical and service occupations or in professional services (which includes education). White women and Hispanic men are more likely to be in retail trade blacks are more likely to be in public administration. A key question is whether occupational and industry differences represent preferential choices or constraints. If one believes that firms discriminate in their propensity to hire into certain occupations, then occupational location is an outcome of discrimination rather than a choice-based characteristic. We discuss the research literature on this issue below. In the regressions reported in this chapter, we follow standard procedure and report regressions with and without controls for occupation, industry and job characteristics (public sector location or part-time work.) Regressions that do not control for these variables in any way probably underestimate the importance of background and choice-based characteristics on labor market outcomes. Regressions that fully control for these variables probably under- estimate the effect of labor market constraints. We allow readers to look at both outcomes. 2.2. Methodologies f o r decomposing wage changes between groups One way to explore the wage differential between groups is to decompose it into "explained" and "unexplained" components. Assume that wages for individual i in group 1 at time t can be written as Wji: = tgl:Xli: + tzli: (2.1) 3 Indeed, the more mobile is labor, the less local labor markets will differ.

3154 7Z ��.~ g 3 o .= o 3. G. Altonji and R. M. Blank q q q q ~ q q ~ q q o o o o o ~ o o o o o o d d ~ 0 0 ~ 0 0 ~ 0 0 0 0 0 ~ ~ ~ ~ ~. ~ 0 0 ~ 0 0 0 0 0 0 0 0 d d d d d d d d d d d d d d _~ ~ ~ ~ .o ~ ~ ~ o o 0 d o ~ d . ~ ~

Ch. 48: Race and Gender in the Labor Market 3155 o o o o o o o o o o o o o 0 O O O 0 0 0 0 0 0 0 0 0 0 o ~ o o o o o o o ~ o d d d d d 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ~ 0 0 0 0 ~ 0 d O 0 d d d d d o d o d o d d o d d d q q q q q q o ~ o o o ~ o o o o o o o d d d d d d d d b ~ ~' ~ ~ ~ 0= c~ �� "a r j b o=

3156 J. G. Altonji and R. M. Blank and wages for individual j in group 2 at time t can be written as w2j~ = ,~2tx2j~ + ~2j,, (2.2) where/31t and/32t are defined so that E(ujit [ Xjit) = 0 and E(u2j r I X 2 j t ) = O. The difference in mean wages for year t can be written as 4 Wit - W2~ = (Xl, - X2t)/31t + (/31, -/32t)X2t, (2.3) where Wut and Xut represent the mean wages and control characteristics for all individuals in group g in year t. The first term in this decomposition represents the "explained" component, that due to average differences in background characteristics (such as educa- tion or experience) of workers from groups 1 and 2. It is the predicted gap between groups 1 and 2 using group 1 - typically white men - as the norm. The second term is the "unexplained" component, and represents differences in the estimated coefficients, i.e., differences in the returns to similar characteristics between groups 1 and 2. The share of the total wage differential due to the second component is often referred to as the "share due to discrimination." This is misleading terminology, however, because if any important control variables are omitted that are correlated with the included Xs, then the/3 coeffi- cients will be affected. The second component therefore captures both the effects of discrimination and unobserved group differences in productivity and tastes. It is also misleading to label only this second component as the result of discrimination, since discriminatory barriers in the labor market and elsewhere in the economy can affect the Xs, the characteristics of individuals in the labor market. 2.3. Estimating simple models o f wage determinatio,~ In this section we explore race and gender gaps in wages through a set of simple models of wage determination. Table 4 shows the differences in race and gender coefficients over time, across specifications and between all workers and full-time/full-year workers. Columns (1) and (4) report regressions of log hourly wages in 1979 and 1995 respectively on dummy variables for black, Hispanic and female, without including any further control variables. Columns (2) and (5) include controls for education, experience and regional location, a minimal set of personal characteristics that an individual brings to a job. Colurmas (3) and (6) add further controls for occupation, industry and job characteristics. Part A of Table 4 focuses on all workers. As control variables are added to the model the negativ6 effect of race or gender on hourly wages becomes less significant. In 1995, black males received 21% lower hourly wages than white males if no control variables were included they received 12% less once education, experience and region were controlled for, and they received 9% less when a full set of control variables were included. Among white women, there is only a small effect of adding controls for education and experience 4 Alternatively, the average wage difference can be decomposed as Eq. (2.3~): W~t - W2t = (Xtt - X 2 t ) ~ 2 t +(/31~ -/32~)X~t. This 'alternative decomposition can produce quite different results from the first. Many authors report both results, or (occasionally) the average of the two.

Ch. 48: Race and Gender in the Labor Market Table 4 Coefficients on race and gender in wage regressions ~ 3157 1979 1995 Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Part (A) all workers (1) Black -0.143 -0.107 -0.061 0.207 -0.119 -0.089 (0.010) (0.010) (0.010) (0.012) (0.011) (0.011) (2) Hispanic -0.152 -0.053 0.040 0.379 -0.131 -0.102 (0.010) (0.010) (0.010) (0.010) (0.010) (0.009) (3) Female -0.436 -0.421 -0.348 -0.279 -0.272 0.221 (0.006) (0.005) (0.006) (0.007) (0.006) (0.007) Controls" (4) Education, No Yes Yes No Yes Yes experience, and region (5) Occupation, No No Yes No No Yes industry and job characteristics b Part (B) full-time-full year workers (6) Black -0.139 -0.115 -0.064 0.148 -0.102 -0.067 (0.012) (0.011) (0.011) (0.012) (0.011) (0.010) (7) Hispanic -0.184 -0.093 -0.076 -0.344 -0.139 -0.101 (0.012) (0.012) (0.011) (0.010) (0.010) (0.010) (8) Female -0.421 -0.399 -0.360 0.265 0.266 -0.241 (0.006) (0.006) (0.007) (0.007) (0.006) (0.007) Controls (9) Education, No Yes Yes No Yes Yes experience, and region (10) Occupation, No No Yes No No Yes industry and job characteristics b ~' Source: Authors' regressions using tile Current Population Survey, March 1980 and March 1996. Standard errors are in parentheses. b Job characteristics include public sector and part-time status. ( s u g g e s t i n g that t h e s e c h a r a c t e r i s t i c s a m o n g w h i t e w o m e n a n d w h i t e m e n are quite s i m i l a r as T a b l e 2 i n d i c a t e s ) , b u t c o n t r o l l i n g f o r o c c u p a t i o n a n d i n d u s t r y r e s u l t s in s u b s t a n t i a l l y s m a l l e r n e g a t i v e effects. P a r t B o f T a b l e 4 l o o k s o n l y at f u l l - t i m e / f u l l - y e a r w o r k e r s . 5 T h e r e s u l t s are s u r p r i s i n g l y