Article Analysis Essay

CATCHING UPt

The Gender Gap in Wages, circa 2000

The transition of women into the U.S. labor market was surely one of the most profound economic and social changes of the 20th cen- tury.' In 1900 about 20 percent of women were in thi labor force. This percentage rose to about 34 in 1950 and reached 61 percent in 2000; not far below the 75-percent participation rate of men. A key element in this change was the dramatic rise in market work among mamed women with children under the age of 18, whose labor-force participation increased from a rate of 18 percent in 1950 to 7 1 percent in 2000.

However, for much of the last 50 years the rise in women's labor-force activity and its growing convergence with that of men, did not appear to be matched by a narrowing of the gender gap in pay. Between 1955 and 1980, the most com- monly cited measure of that gap, the female- to-male ratio of median annual earnings of full-time year-round workers, hovered around 60 percent. But using the same measure, the ratio began to rise after 1980, reaching 69 per- cent in 1989 and 74 percent in the mid 1990's, after which it leveled off. Based on a more appropriate measure, average hourly wage rates (available since 1979), the gender gap is smaller, but the pattern of change is similar, and the ratio rises from 66 Dercent in 1979 to 80 percent in 1993 and then' stabilizes (Fig. 1).

Through the years the gender gap in wages frequently has been a source of public concern and a puzzle to researchers. In this paper I

tDiscussants: Donald Deere, Texas A&M University; Haniett Duleep, Urban Institute; Jeffrey Grogger, Univer- sity of California-Los Angeles.

* Department of Economics and Center for the Study of Business and Government, Baruch College, City University of New York, 17 Lexington Avenue, New York, NY 10010. I thank Mei Liao and Wenhui Li for excellent research assistance, as well as Alex Cavallo of Lexecon, Inc.

See Jacob Mincer's (1962) pioneering work on the determinants of women's labor force participation.

examine evidence from the Current Population Survey (CPS) and the National Longitudinal Survey of Youth (NLSY79) on recent trends and current sources of the gender gap.

I. Unique Factors Underlying Gender Differences in Skills

In comparing the earnings of different demo- graphic groups it is usually important to exam- ine the effect of productivity differences between the groups that might account for any earnings differential. In the case of differences in earnings between racial and ethnic groups of the same sex, productivity differences most of- ten stem from differences in the quantity and quality of education and other human capital acquired at home as well as in school. Differ- ences in productivity between men and women, however, are not likely to be due to differences in educational background. Sisters and brothers are exposed to the same parental background and attend schools of the same quality. Their current educational attainment and their cogni- - tive skills, as measured by achievement test scores, are similar.

Instead, the main source of productivity dif- ferences between women and men stems from the lesser amount of time and energy that many women can commit to labor-market careers as a result of the division of labor within the family.* Even though women's home responsibilities have fallen dramatically over the past 50 years, they are nonetheless, still significant. donse- quently, women are less likely than men to work continuously after leaving school and therefore are less likely to gain experience that

In an extension of his work on the economics of the family Gary Becker (1985) has developed a model of the allocation of energy which shows how the energy demands of childcare and housework reduce the energy available for market work.

80

55

310 AEA PAPERS AND PROCEEDINGS MAY 2003

Hourly wage rates, wage and

L L Median annual earnings,

u -tlrne year-rou wo ers

can only be acquired on the job. In addition, anticipation of child-related work intemptions and the need to coordinate home responsibilities with market work are likely to influence choice of occupation and type of firm.

One can argue whether the source of these gender role differences is a form of discrirnina- tion rather than an outcome of biological and other deeply rooted psychological and cultural factors. However, by the time they are old enough to make choices, many women make different choices than men regarding the extent of career attachment.

Current data continue to show the strong ef- fect of the presence of children, particularly young children, on work participation and on hours of work among those who do work. In March 2001, at ages 25-44, the prime period for career development, 34 percent of women with children under the age of six were out of the labor force, compared to 16 percent of women without children. Thirty percent of em- ployed mothers worked part-time, compared to 11 percent of women with no children. Among men, however, the presence of children is asso- ciated with an increase in work involvement. Only 4 percent of men with chldren under the age of six are out of the labor force, and among employed fathers only 2 percent work part-time.

The expectation of withdrawals from the la- bor force and the need to work fewer hours during the week are likely to influence the type of occupations that women train for and ulti- mately pursue. More subtle factors such as the

level of stress at work and the ability to take unplanned time off for family emergencies are also likely to influence the choice of occupation and work place. Thus, certain characteristics of jobs may affect women's occupational choices because they are particularly compatible or in- compatible with women's dual homelmarket roles. These adaptive occupational choices will tend to lower the market earnings of women relative to men.

For example, some occupations require lengthy investment in skills with applicability only to highly specific market activities (e.g., aerospace engineer, surgeon, top management in large complex organizations). The payoff to such investments is obviously reduced when years in the labor force are reduced. Moreover, skills depreciate during periods of withdrawal from work (Jacob Mincer and Haim Ofek, 1982), and the rate of depreciation is likely to vary depending on the rate of technological change and obsolescence of the slulls acquired. Fields such as physics, where knowledge depre- ciates rapidly have disproportionately fewer women. Other types of schooling and training are more general in their applicability to differ- ent situations and impart skills that are less prone to depreciate. For example, nursing and teaching skills are valuable to mothers and can be practiced widely in different settings with relatively little additional firm-specific training.

Certain characteristics of the workplace are more compatible with women's home responsi- bilities than others. The depreciation in skills and earnings related to complete withdrawal from the labor force may be ameliorated by work situations that accommodate the need for less demanding work while raising a family. Part-time work is the most obvious manifesta- tion of this adjustment. Even if a woman does not always work part-time she may be more likely to choose an occupation or job setting that provides a shorter or more flexible work week in the event it may be needed, or a more infor- mal work setting where time off for unpredict- able events is acceptable.

Both work attachment and the choice of oc- cupation are expected to be important determi- nants of women's earnings and important factors underlying the gender wage gap. In the analysis discussed below I incorporate mea- sures and proxies for these factors. I examine

311 VOL. 93 NO. 2 CATCHING UP

the factors associated with changes in the wage gap over the past two decades using data from the CPS and then examine sources of the current differential for a cohort of workers using the more comprehensive and detailed variables of the NLSY.

LI. Findings from the Current Population Survey: 1W9-2001

The CPS analysis is based on data from the CPS outgoing-rotation-group files (CPS ORG) merged with data on occupational characteris- tics from the Department of Labor's Fourth Dictionary of Occupational Titles (DOT), 1991 revision. The analysis includes part-time and full-time wage and salary workers, ages 20-60.

The major changes that have occurred during the 1979-2001 period in the gender differential in earnings-related characteristics are as fol- lows. Women continue to be much more likely than men to work part-time (19 percent versus 5 percent in 2001), although that difference nar- rowed. With respect to education, women gained relative to men at the college level. By 2001 they were somewhat more likely than men to be college graduates and were almost as likely to receive a hlgher degree. Women also have been entering occupations requiring more job-specific slulls, as measured by SVP (spe- cific vocational preparation), the time required to attain the average level of proficiency in an occupation-a DOT variable. The gender gap in SVP declined by almost half between 1984 and 1994 and has since declined further, but at a slower rate.

However, despite these changes, women and men remain in occupations that are dispropor- tionately female or male. In 2001 women, on average, worked in occupations in which the percentage of female employees was close to 68 percent; men worked in occupations that were only 30-percent female. The percentage female in an occupation is one simple way of measur- ing the characteristics of an occupation that are conducive to women's particular needs. How- ever, in the CPS analysis I have taken the more direct path of including specific characteristics of occupations as individual variables.

Returns to "Potential Experience."-As a number of studies have shown, there is evidence

TABLEI - W E AND FEMALERETURNS TO POTENTIAL EXPERIENCEAT 15 YEARS(LOG WAGEGAIN)(CPS)

Difference Year Men Women (M - W)

1979 0.447 0.207 0.240 1981 0.439 0.223 0.216 1983 0.497 0.263 0.234 1985 0.497 0.288 0.208 1987 0.47 1 0.305 0.165 1989 0.482 0.307 0.175 1991 0.477 0.312 0.166 1993 0.463 0.334 0.128 1995 0.477 0.358 0.119 1997 0.452 0.350 0.102 1999 0.425 0.306 0.119 2001 0.398 0.309 0.089

that the years of work experience of employed women increased during the 1980's (M. Anne Hill and O'Neill, 1992). In fact, the narrowing of the work-experience gap was a key factor causing the gender wage gap to narrow during the 1980's (O'Neill and Solomon Polachek, 1993; Francine Blau and Lawrence Kahn, 1997). Nonetheless, longitudinal data show that a significant experience gap remains. The CPS, however, contains no direct measure of years of work experience. The standard way of infemng past work experience in the CPS is to construct "potential experiencew-essentially the number of years since leaving school (or since age 17, if the person left school at a younger age). Actual experience is reasonably close to potential ex- perience for men. For women that is not the case. The return to potential experience is typ- ically lower for women than for men, and the fact that the difference between actual and po- tential experience is larger for women than for men likely accounts for at least part of the difference in returns. Therefore, if women's ac- tual experience has been catching up to their potential experience, one would expect that the effect of potential experience on the female wage rate would increase over time for women, and more so than for men, if the return to experience generally was rising for other rea- sons. As shown in Table l , that is in fact what has happened. I have conducted a series of annual cross-sectional regressions, for the years 1979-2001, separately by sex, in which the log wage is regressed on potential experience (qua- dratic specification), schooling, whether the

312 AEA PAPERS AND PROCEEDINGS MAY 2003

TABLE~--CHARACTERIST~CSOF NLSY MEN AND WOMEN M2: adds occupational characteristics to M I IN 2000 (AGES 35-43)

Difference 'fl ----._I

<- - y Characteristic Men Women (M - W) - - - / . . . . _ _ _ . _ _

M I (adj. for schooling, PT wprk .. - . _ _ _ . . ana cbmmT5j --+ .- .-

_ _ . - '/ Actual FIM wage ratio

individual worked part-time, and basic demo- graphic controls. The results indicate that, eval- uated at 15 years of potential experience, the return for both women and men increased from 1979 until about 1995 after which it declined somewhat. However, women's retums to poten- tial experience increased much more rapidly than men's, and the difference between men and women narrowed sharply. This suggests that the relative quantity andor quality of women's ac- cumulated work experience probably continued to rise through 2001.

The "Adjusted" Wage Gap Using Available CPS Variables.-To discern the effect of gen- der differences in characteristics on the wage gap and how the relation may have changed, I have conducted standard decompositions of the CPS annual regressions using, alternatively, male and female coefficients and different model specifications. The results, using male coefficients, are shown in Figure 2.3 Model 1 (MI), which adjusts for potential experience, schooling, whether the individual worked part- time, and basic demographic controls, raises the wage ratio by about 5 percentage points in the early years and by a little less than 4 percentage points in the later years. Model 2 (M2) adds a series of variables measuring a variety of occu- pational characteristics including SVP and other

Regression specifications and complete results, both for the CPS and NLSY regression analyses, are available from the author upon request.

AFQT 42.29 39.61 2.68 High-school dropout 25.05 19.54 5.51 College graduate 12.82 12.26 0.56 Higher degree 9.55 10.00 -0.45 Full-year equivalents 16.14 14.20 1.93

worked since age 18 (weeksl52)

Percentage of full-year 8.66 equivalents worked part-time

Tenure (years) 6.49 Years out of labor force 2.5 1 Percentage worked for 4.66

nonprofit Percentage with one child 75.31

or more Percentage age at first birth 13.85 r 30

Percentage childcare cited 12.85 as reason for out of labor force

Characteristics of Occupation:

Percentage female in occupation

SVP (months) Percentage use computer

for database Percentage use computer

for programming Percentage left labor force

in a year Percentage exposed to:

Work hazards Fumes High noise Medium strength

requirement Outdoor conditions

variables that are proxies for aspects of working conditions. (Table 2 lists these variables for the NLSY regressions; the CPS analysis uses the same variables, except for computer usage.)

Occupational characteristics explain a sub- stantial portion of the wage gap. (I have not added these variables prior to 1983 because of the major change in occupational codes.) The femaldmale wage ratio, adjusted for all model-2 variables, increased from 84 percent in 1983 to 90 percent in 2001; the unadjusted ratio rose from 70 percent to 80 percent over the same

313 VOL. 93 NO. 2 CATCHING UP

TABLE 3-EXPLAINING THE WAGE GAP IN 2000 BETWEEN NLSY WOMEN AND MEN,AGES 35-43

(USING MALE COEFFICIENTS)

Measure Model 1 Model 2 Model 3

Log Wage Gap Attributable to D~Terencesin Characteristics:

Demographic 0.0086 0.0067 0.0067 Education -0.0120 -0.0071 -0.0072 AFQT 0.0100 0.0063 0.0063 Work experience since 0.1472 0.1037 0.1030

age 18 Workplace characteristics - 0.0231 0.0222 Child-related factors - 0.0167 0.0173 SVP - 0.0129 0.0119 Other occupational - 0.0335 0.0173

characteristics Percentage female in - 0.0436

occupation

Unadjusted log wage gap 0.2463 0.2463 0.2463 Gap explained by model 0.1563 0.1957 0.2210 Unexplained gap 0.0900 0.0506 0.0253

Observed F M wage ratios: 78.2 78.2 78.2

Adjusted F M wage ratios: 91.4 95.1 97.5

period. When female coefficients are used, the effects are smaller, largely because women's earnings are less negatively affected by working in occupations that provide part-time work and allow for labor-force turnover.

111. Findings from the NLSY

Analysis of data from the NLSY permits a more complete assessment of the extent to which important differences in human capital and job and occupational characteristics can explain the gender gap in wages. The analysis uses the 2000 NLSY when the cohort has reached ages 35-43.

The variables included in the analysis and the differences in the characteristics of the NLSY men and women are given in Table 2. Table 3 shows the proportion of the wage gap "ex- plained" by sets of variables in three model specifications, using male coefficients. Here are the highlights:

(i) The gross log wage differential in 2000 was 0.246, corresponding to a wage ratio of 78.2 percent. Years of schooling and

scores on the AFQT (Armed Forces Qual- ification Test) explain hardly any of the differential because women and men differ little in these characteristics.

(ii) Actual work experience accounts for much of the gap. Using model 1, a basic human- capital specification, work experience, ac- counts for 0.1472 of the unadjusted log wage gap, which is almost all of the ex- plained portion of the gap using male co- efficients. Although it is reduced, the contribution of work experience remains large when other, intercorrelated variables are added as in models 2 and 3.

(iii) The addition of occupational and work- place characteristics in model 2 reduces the unexplained portion of the gap from 0.0900 (model 1) to 0.0506. Model 3 adds the variable FEM (percentage female in occupation). For men, FEM has a strong negative effect. It accounts for 0.0436 of the gap and reduces the effect of the occu- pational characteristics with which it is obviously correlated. (But note that in the separate regressions for women, the effect of FEM is weak but positive.14 Including all of the variables in model 3 reduces the unexplained gap to 0.0253, a wage ratio of 97.5 percent. (The comparable ratio using female coefficients is 91.3 percent.)

I have conducted additional analysis of the NLSY cohort separately by schooling level. Gender differences in work experience are much greater at the high-school level than they are for college graduates. Consequently, work experience accounts for a particularly large share of the gap. At the high-school level the wage gap falls to 3 percent using model 1; it is eliminated when occupational characteristics and FEM are added.

The unadjusted wage gap is larger for college graduates than it is at the high-school level. The field of college major, a harbinger of occupational choice, accounts for a significant amount of the

In an extensive analysis of the effect of occupation on the gender wage gap, David Macpherson and Bany Hirsch (1995) find that the effect of FEM on the wage rate is sensitive to model specifications and is negative and signif- icant in female as well as in male regressions under certain specifications.

314 AEA PAPERS AND PROCEEDINGS MAY 2003

gap, a result consistent with that of Charles Brown and Mary Corcoran (1997). At the col- lege-graduate level FEM does not have a sig- nificant effect on the outcome. The results are similar whether the male or female coefficients are used. The unexplained gap is about 6 per- centage points when both field of college major and occupational characteristics are included, and that is the case using either the male or female coefficients.

IV. Concluding Comments

Understanding the gender gap in pay is im- portant because even in the absence of any labor-market discrimination it is unlikely that the wage rates of women and men would be equal. As I have shown in this paper, the unad- justed gender gap can be explained to a large extent by nondiscriminatory factors. Those fac- tors are unllkely to change radically in the near future unless the roles of women and men in the home become more nearly identical. Thus an unadjusted gender gap may be with us for quite a while.

REFERENCES

Becker, Gary S. "Human Capital, Effort and the Sexual Division of Labor." Journal of Labor

Economics, January 1985, 3(1), Part 2, pp. S33-58.

Blau, Francine D. and Kahn, Lawrence M. "Swirn- ming Upstream: Trends in the Gender Wage Differential in the 1980s." Journal of Labor Economics, January 1997,15(1), Part 1, pp. 1-42.

Brown, Charles and Corcoran, Mary. "Sex Based Differences in School Content and the Male- Female Wage Gap." Journal of Labor Eco- nomics, July 1997, 15(3), Part 1, pp. 431-65.

Hill, Anne M. and O'Neill, June. "An Inter-cohort Analysis of Women's Work Patterns and Earnings," in R. Ehrenberg, ed., Research in labor economics, Vol. 13. Greenwich, CT: JAI Press, 1992, pp. 215-86.

Macpherson, David A. and Hirsch, Barry T. "Wages and Gender Composition: Why Do Women's Jobs Pay Less?" Journal of Labor Economics, July 1995, 13(3), pp. 426-71.

Mincer, Jacob. "Labor Force Participation of Married Women," in Gregg Lewis, ed., As- pects of labor economics, Universities- National Bureau Conference Series, No. 14. Princeton, NJ: Amo, 1962, pp. 63-97.

Mincer, Jacob and Ofek, H. "Interrupted Work Careers: Depreciation and Restoration of Hu- man Capital." Journal of Human Resources, Winter 1982, 17(1), pp. 3-24.

O'Neill, June and Polachek, Solomon. "Why the Gender Gap in Wages Narrowed in the 1980s." Journal of Labor Economics, Janu- ary 1993, 11(1), Part 1, pp. 205-28.