Content uploaded by Lawrence Kahn
Author content
All content in this area was uploaded by Lawrence Kahn
Content may be subject to copyright.
NBER WORKING PAPER SERIES
WHERE ARE WE IN THE ECONOMICS
OF GENDER? THE GENDER PAY GAP
Francine D. Blau
NBER Working Paper 5664
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
July 1996
An earlier version of this paper was presented at the 15th Arne Ryde Symposium of Lund University
on “The Economics of Gender and the Family,” Rungstedgaard, Denmark, August 18-19, 1995. This
paper draws heavily on joint work with Lawrence Kahn. Iam also indebted to him for helpful
comments. This paper is part of NBER’s research program in Labor Studies. Any opinions
expressed are those of the author and not those of the National Bureau of Economic Research.
O1996 by Francine D. Blau. All rights reserved. Short sections of text, not to exceed two
paragraphs, may be quoted without explicit permission provided that full credit, including@ notice,
is given to the source.
NBER Working Paper 5664
July 1996
WHERE ARE WE IN THE ECONOMICS
OF GENDER? THE GENDER PAY GAP
ABSTRA CT
Empirical research on gender pay gaps has traditionally focused on the role of gender-specific
factors, particulmly gender differences in qualifications and differences in the treatment of otherwise
equally qualified male and female workers (i.e., labor market discrimination). This paper explores
the determinants of the gender pay gap and argues for the importance of an additional factor, wage
structure, the array of prices set for labor market skills and the rewards received for employment in
favored sectors. Drawing on joint work with Lawrence Kahn, Iillustrate the impact of wage
structure by presenting empirical results analyzing its effect on international differences in the gender
gap and trends over time in the gender differential in the U.S.
Francine D. Blau
School of Industrial and Labor Relations
Cornell University
265 Ives Hall
Ithaca, NY 14853-3901
and NBER
While the gender pay gap has been an important focus of modem economists concerned
with the economics of gender, it was not necessarily the primary concern of early commentators
on gender inequities. For example, in the 19th century, with few married women employed
outside the home, observers like American feminist, Charlotte Perkins Gillman (1898), and
Marxist, FriedrichEngels(1884), focused on the gender division of labor itself and espoused the
emancipating effects of women’s participation in market work. Among economists as well as
the general public, interest in gender issues including the gender pay gap has proceeded hand-in-
hand with the gro~h in women’s labor force participation. As women have come to comprise a
larger share of the paid labor force and as market work has loomed larger in the typical woman’s
life, interest in the determinants of gender differences in labor market outcomes has also grown.
Of these labor market outcomes, the wage is of fundamental importance as amaj or
determinant of economic welfare for employed individuals, as well as of the potential gain to
market employment for those not currently employed. Further it serves as asignificant input into
amyriad of decisions ranging from labor supply to marriage and fertility, as well as afactor
influencing bargaining power and relative status within the family. Thus, Ifocus here on wages
in cotidence that Iam examining aquestion of considerable interest to economists and of
considerable importance to women’s economic well-being. However, Ireadily acknowledge that
wages are by no means the whole story even as ameasure of economic well-being.
Research on the gender pay gap has traditionall yfocused on the role of what might be
termed, gender-specific factors, particularly gender differences in qualifications and differences
in the treatment of otherwise equally qualified male and female workers (i.e., labor market
discrimination). An innovative feature of recent research on gender and race differentials has
been to integrate the analysis of the gender pay gap as well as other demographic differentials
LUND96.DOC 6/24/96 1:34 PM
2
into the stud yof wage structure in general. ]Wage structure describes the array of prices set for
various labor market skills (measured and unmeasured) and rents received for employment in
particular sectors of the economy.
Wage structure is potentially of considerable importance in determining the relative
earnings of groups like women who tend on average to have lower skills or to be located in lower
paying sectors of the economy. In this paper Iwill first consider the determinants of gender
differentials, highlighting the role of wage structure. 1will then illustrate the impact of wage
structure by summarizing some of my recent work with Lawrence Kahn on international
differences in male-female wage differentials (Blau and Kahn 1992a, Blau and Kahn 1995, and
Blau and Kahn 1996), and on trends over time in gender differentials in the U.S. (Blau and Kahn
1994, Blau and Kahn forthcoming). Iwill then offer some concluding thoughts and suggest
some implications for public policy.
DETERMINANTS OF THE GENDER PAY GAP: GENDER SPECIFIC FACTORS
An initial early impetus to the study of wage differentials was provided by the British
experience in World War I. Pursuant to the war effort, there was some substitution of women
into traditionally male civilian jobs, although not nearly to the degree that there would be during
World War II. Questions of the appropriate pay for women under these circumstances arose and
stimulated anumber of economic analyses of the sources of the gender pay differential—all of
which gave aprominent causal role to occupational segregation (e.g., Edgeworth 1922, Fawcett
1918, Webb 1919). Modem efforts to understand the gender pay gap have generally rested on
3
WO strong pillars: the human capital explanation and models of labor market discrimination.
These are gender sDecific explanations in that they focus on gender differences in qualifications
or treatment as the cause of the pay gap.
Human capital explanations developed by Mincer andPolachek(1974), Polachek (1981 )
and others explain gender differences in economic outcomes on the basis of productivity
differences between the sexes. This explanation is based on the gender division of labor within
the family which, as we have seen, was the focus of the 19th century commentators, and traces
the impact of this division on the wages and occupations of men and women. Anticipating
shorter and more discontinuous work lives as aconsequence of their role within the family,
women will have less incentive to invest in market-oriented formal education and on-the-job
training than men. Their resulting smaller human capital investments will lower their earnings
relative to men’s. Similar considerations are also expected to produce gender differences in
occupations, as women choose occupations where such investments are less important and where
the wage penalties for work force interruptions are smaller. In the absence of parental leave
policies, women will especially avoid jobs requiring large investments in firm-specific skills
because the returns to such investments are reaped only as long as one remains with the firm.
Since the costs of firm-specific training are shared by employers and employees, employers are
reluctant to hire women for these jobs due to their shorter expected tenure on average. The
difficulty of distinguishing more career-oriented women from less career-oriented women means
that the former may be the victims of such “statistical discrimination” as well (see below).
Thus, the human capital model provides alogically consistent explanation for gender
differences in economic outcomes based on the traditional division of labor in the family. Not
4
only will women earn less, but they will tend to be located in different occupations. Gender
differences in industrial distribution could also occur if industries vary in their skill requirements,
Thus the human capital model provides arationale for the pay gap based on the voluntary
decisions of women and men. Working in asimilar direction is Becker’s (1985) model in which
the longer hours women spend on housework lowers the effort they put into their market jobs
compared to men’s and hence reduces their wages. But these models may also be viewed as
shedding light on how the traditional division of labor in the family disadvantages women in the
labor market. Thus, in this sense, they provide some support for the claim of the 19th century
observers that the traditional division of labor is of fundamental importance in determining
women’s status within the larger society. To the extent that gender differences in outcomes are
not fully accounted for by productivity differences derived from these and other sources, models
of labor market discrimination offer an explanation.
Theoretical work on discrimination was initiated by Becker’s (1957) examination of race
discrimination. Becker conceptualized discrimination as ataste or personal prejudice against
members of aparticular group. Models of statistical discrimination were later developed, in part
to explain the persistence of discrimination in the long run in the face of competitive labor
markets (e.g., Phelps 1972, Aigner and Cain 1977, and Lundberg and Startz 1983). Such models
assume aworld of uncertainty and imperfect information and focus on differences between
groups in the expected value of productivity or in the reliability with which productivity may be
predicted. Since the real or perceived average gender differences that underlie statistical
discrimination against women in the labor market tend to stem from the traditional division of
labor in the family, this constitutes another route by which traditional gender roles within the
5
family adversely effect women’s labor market outcomes. Another aspect of interest is the
relationship between occupational segregation and adiscriminatory wage gap formulated in
Bergmam’s (1974) overcrowding model. Discriminatory exclusion of women from “male” jobs
results in an excess supply of labor in “female” occupations, depressing wages there for
otherwise equally productive workers,
These two explanations, gender differences in qualifications and differences in treatment
of otherwise similar men and women, do not necessarily constitute mutually exclusive sources of
gender wage differentials. Both may play arole and empirical studies based on cross-sectional
data within countries provide considerable empirical support for each, One problem here is that
evidence for discrimination relies on the existence of aresidual gender pay gap which cannot be
explained by gender differences in measured qualifications. This accords well with the definition
of labor market discrimination, i,e., pay differences between groups that are not explained by
productivity differences, but may also reflect group differences in unmeasured qualifications or
compensating differentials. If men are more highly endowed with respect to these omitted
variables then we would overestimate discrimination. Alternatively, if some of the factors
controlled for (e.g., occupation, tenure with the employer) themselves reflect the impact of
discrimination, then discrimination will be underestimated.
Another challenge to empirically decomposing the gender pay gap into its constituent
parts is the existence of feedback effects. The traditional division of labor in the family may
influence women’s market outcomes through its effects on their acquisition of human capital and
on rationales for employer discrimination against them. But it is also the case, that by lowering
the market rewards to women’s human capital investments and labor force attachment,
6
discrimination may reinforce the traditional division of labor in the family (e.g., Blau 1984, Blau
and Ferber 1992, Weiss and Gronau 1981, and Lundberg and Startz 1983). Even small initial
discriminatory differences in wages may cumulate to large ones as men and women make human
capital investment and time allocation decisions on the basis of them. Another 19th century
observer, JohrI Stuart Mill, touched on this very relationship over 100 years ago when he
advocated women’s “admissibility to all the finctions and occupations hitherto retained as the
monopoly of the sti-onger sex,” claiming that “their disabilities elsewhere are only clung to in
order to maintain their subordination in domestic life” (1878: 94).
DETERMINANTS OF THE GENDER PAY GAP: THE ROLE OF WAGE STRUCTURE
Thus, we see that the clear determination of the impact of qualifications VS,
discrimination in the gender pay gap is difficult for both empirical and conceptual reasons.
However, both explanations share acommon focus of being gender specific explanations of the
pay gap. Analyses of trends over time in the gender differential within countries as well as
intercountry comparisons of gender earnings ratios have traditionally tended to emphasize these
types of gender-specific factors. The last 15 to 20 years have been atime of ferment in the labor
market with rapid changes in skill differentials and thus wage inequality in much of the
industrialized world. Nowhere have these changes been more dramatic than in the U.S. It has
been anatural extension of the study of these types of realignments to examine their
consequences for various demographic groups. Moreover, upon further reflection it is clear that
the traditional gender specific factors imply an important role for wage structure.
7
The human capital model suggests that men and women tend to have different levels of
labor market skills (especially work experience) and to be employed in different occupations and
perhaps in different industries. Discrimination models too suggest that women may be
segregated in different sectors of the labor market. This implies apotentially important role for
wage structure in determining the pay gap. All else equal, the larger the returns to skills and the
larger the rents received by individuals in favored sectors, the larger will be the gender gap.
Similarly, labor market discrimination and/or actual female deficits in unmeasured skills result in
employers treating women as if they have lower unmeasured as well as measured skills, Thus,
the higher the rewards to unmeasured skills, the larger will be the gender gap, other things being
equal.
The notion of a“high” or a“low” return is intrinsically arelative concept, Thus, the
framework provided by wage stracture requires some frame of reference and is particularly
useful in analyzing changes over time in gender differentials or differences across countries in
gender gaps. Such intertemporal and cross-country comparisons enable us to measure the effects
of wage structure comua rativelv with reference to the situation that existed at an earlier point in
time or that prevails in another country.
Consider the following examples. Suppose that in two countries, women have lower
levels of labor market experience than men but that the gender difference in experience is the
same in the two countries. If the return to experience is higher in one country, then that nation
will have alarger gender pay gap. Similarly, an increase in the return to experience within a
country will, all else equal, raise the gender gap. Or, as another example, suppose that the extent
of occupational segregation by sex is the same in two countries but that the wage premium
8
associated with employment in “male” jobs is higher in one country, Then, again, that nation
will have ahigher pay gap. In like manner, an increase over time in the wage premium for
“male” jobs will increase the gender gap, ceteris paribus. This second example suggests that a
clear-cut distinction between gender-specific factors and wage structure may be difficult to
achieve. Agender-specific policy like comparable worth which mandates pay adjustments
across male and female jobs to provide for equal pay for work of equal value within the firm, can
obviously affect wage structure. Nonetheless, as Ihope to show below, the notion of wage
structure is quite usefil and can shed considerable light on international differences in the gender
gap as well as trends over time within countries.
Wage structure itself is determined by avariety of factors, including relative supplies of
labor of various skill levels, technology, the composition of demand, and wage setting
institutions. In recent years, there has been an increase in wage inequality within most of the
industrialized countries (Gottschalk and Joyce 1995). Juhn, Murphy and Pierce (1993), in their
work on the U.S. trends, make astrong case that this trend reflects arising return to skills, both
measured and unmeasured. We do not have fill consensus regarding the reasons for this increase
in the return to skill, but technological change and the impact of international trade are two of the
chief candidates in the United States (e.g., Katz and Murphy 1992, Bound and Johnson 1992,
Borjas and Ramey 1995). In addition, institutional factors, including declining union density and
afalling value of the minimum wage, appear to have also contributed to rising inequality
(Freeman 1994, Card forthcoming).
With respect to international comparisons, Kahn and Ihave emphasized in our work that
systems of centrally-determined pay are likely to entail less wage inequality and smaller gender
9
wage differentials for anumber of reasons. First, in the U. S., asignificant portion of the male-
female pay gap has been found to be associated with interindust~ or interfirm wage differentials
(Blau 1977, Johnson and Solon 1986, Sorensen 1990, and Groshen 1991). The relatively large
pay variation across industries and firms in the U.S. is to some extent an outgrowth of our
relatively decentralized pay-setting institutions. Therefore, centralized systems which reduce the
extent of wage variation across industries and firms are likely to lower the gender differential, all
else equal. Second, since in all countries the female wage distribution lies below the male
distribution, centralized systems that raise minimum pay levels regardless of gender will also
tend to lower male-female wage differentials. In Blau and Kahn (forthcoming a), we find
considerable evidence consistent with the view that, compared to the U. S., the more centralized
wage setting institutions of other industrialized countries not only reduce overall wage
inequality, but that this reduction is primarily due greater compression at the bottom of the wage
distribution in these countries rather than at the top. This tendency to bring up the bottom of the
wage distribution in turn reflects not only the impact of conscious government and union policies
in some countries, but, more generally, wage setting institutions in both the union and nonunion
sector which lead to greater wage compression in each sector compared to the U.S. Of particular
interest, is the greater prevalence in other countries of contract extension and other informal
mechanisms which extend union-detemined wages (and thus the more compressed union wage
structure) to the nonunion sector.
10
INTERNATIONAL DIFFERENCES IN THE GENDER GAP: AU.S.-SWEDEN
COMPARISON
puzzle
My work with Lawrence Kahn on international differences in the gender gap addresses a
While the relative qualifications of American women are high compared to women in
other countries and the U.S. has had alonger and ofien stronger commitment to anti-
discrimination laws than most industrialized nations, the U,S. has traditionally been among the
countries with the largest gender gaps. Our results based on comparisons of the U.S. to nine
other industrialized nations (Blau and Kahn 1995 and Blau and Kahn 1996) suggests that the
resolution of this paradox
the lower ranking of U.S.
lies in the enormous importance of overall wage structure in explaining
women. That is, the gender gap in the U,S. is relatively high, not
because of the traditional gender-specific factors but rather due to the relatively large penalty that
the U.S. wage structure imposes on groups that have below average skills (measured and
unmeasured) or are located in less favored sectors. We find that the U.S. gap would be similar
to that in countries like Sweden and Australia (the countries with the smallest differentials) if the
U.S. had their level of wage inequality.
Iillustrate the role of wage structure in influencing international differences in the gender
pay gap in more detail by presenting some of our findings for the U.S.-Sweden comparison (Blau
and Kahn 1996). This comparison is of interest because the U.S. and Sweden represent cases at
the extremes of an international ranking of gender ratios with Sweden having among the highest
gender ratios of the advanced industrialized countries, and the U. S. having among the lowest.
This was particularly the case for the year from which we draw our data, 1984, since the gender
gap has been narrowing in the U.S. and widening in Sweden. An additional reason why our
11
results for these two countries are especially interesting is that our data sources, the Michigan
Panel Study of Income Dynamics (PSID) for the U.S. and the Household Market and Nonmarket
Activities Survey (HUS) for Sweden, contain information on actual labor market experience and
thus permit us to control for this important variable in our wage regressions and corresponding
decompositions.
Sweden Differences in the Gerider GarI
The extent of the U.S ,-Sweden difference in gender ratios maybe seen in Figure 1which
shows unadjusted and adjusted gender log wage ratios for each country in 1984,2 The unadjusted
ratios of 66,9 percent for the U.S. and 82.7 percent for Sweden, indicate that the ratio was nearly
16 percentage points higher in Sweden—a considerable difference. Swedish women also fare
better afier adjusting for all variables, including education and experience, as well as major
industry and occupation: the adjusted wage ratio3 was 82.2 percent for the U.S. and 90,9 percent
for Sweden. Thus, while adjusting for measured characteristics raises the ratio in each country
and reduces the gap between the two countries, asubstantial differential in gender ratios of
almost 9percentage points relmains.
Interpreting these findings in terms of the conventional gender-specific explanations
would lead us to view the smaller unexplained gender gap in Sweden as indicating that,
compared to U.S. women, Swedish women encounter less discrimination or have more favorable
levels of unmeasured characteristics compared to men or both. The reduction in the U. S.-
Sweden difference in the gender gap when controls are added for measured characteristics would
imply that Swedish women, on net, also have more favorable levels of measured characteristics
12
compared to their male counterparts than do U.S. women, These conclusions make intuitive
sense in some respects. With regard to measured characteristics, Sweden’s considerably more
generous family leave policy may result in Swedish women being more firmly attached to their
employers and to the labor market than U.S. women.4 With respect to discrimination, the results
are somewhat surprising in that, as noted above, the U.S. has aconsiderableylonger commitment
to anti-discrimination laws than Sweden. However, it could be that Sweden’s long term
commitment to attacking traditional gender roles through avariety of policies is responsible for
the smaller unexplained pay gap in Sweden.
The Role of Waqe Structure
Despite the apparent reasonableness of the conclusions based on the conventional
approach, our examination of the role of wage structure suggests that they are incorrect. It is not
differences in women’s qualifications or labor market treatment that are responsible for the larger
U.S. gender gap, rather it is differences in overall labor market prices in the two countries. To
illuminate the role of wage structure, Kahn and Iadapted aframework developed by Juhn,
Murphy and Pierce (199 1) to analyze trends over time in race differentials in the U.S. Gender-
specific factors, including differences in qualifications and the impact of labor market
discrimination, are regarded as determining the percentile ranking of women in the male wage
distribution, while the overall wage structure (as measured by the magnitude of male wage
inequality) determines the wage penalty or reward associated with this position in the wage
distribution.
The basic premise here is that males at the same percentile ranking as women may be
viewed as comparable in the eyes of employers. Thus, the same set of factors will determine the
13
relative rewards of women and of these comparable males. The portion of the gender differential
associated with women’s lower ranking in the distribution in country Aas compared to country B
is ascribed to differences between the two countries in gender-s~e cific factors (i.e., qualifications
and treatment), while the portion that is due to the wage penalty associated with that position
(i.e., greater wage inequality) in country Athan in country Bis attributed to wage structure.
Some indirect evidence for the assumption that the same factors determine the relative rewards of
women and of these male comparable, is that wage inequality is higher in the U.S. than in the
other countries among ~men and women (Blau and Kahn 1995 and Blau and Kahn 1996).s
Similarly, wage inequality in the U,S. has been increasing among both men and women (Katz
and Murphy 1992, Blau and Kahn forthcoming b), This suggests that the same sets of factors--
the prices of measured and unmeasured skills and wage-setting institutions--affect the wages of
both men and women in asimilar way.
It should be noted, however, that the possibility of discrimination complicates this
division into gender-specific factors and wage structure because what we have labeled as the
impact of wage structure may also include acomponent which is due to the interaction between
wage structure and discrimination. That is, discrimination pushes women down in the
distribution of male wages, while wage structure determines how large the penalty is for that
lower position in the distribution.
Emp irical Findinp sfor the Effect of Wag eStructure
Figure 2presents the mean percentile rankings of women in each country’s overall male
wage distribution and residual male wage distribution.7 Our reasoning suggests that the female
percentile rankings may be taken as overall indicators of gender-specific factors, that is the
14
relative qualifications and treatment of women in each country. The placement of women in the
overall male wage distribution represents the combined effects of gender differences in
qualifications and treatment (or unmeasured characteristics). The ranking of women in the male
residual wage distribution represents the effect of gender differences in treatment (or unmeasured
characteristics) only, Differences between the rankings of the two countries in Figures 1and 2
represent the role of wage structure or the wage penalty associated with being below average in
the distribution in each country.
Looking first at the findings for the unadjusted gender wage difference, we see that,
despite the large U.S.-Sweden difference in the unadjusted gender gap, the mean percentile
rankings of women in the male wage distribution in Sweden and the U.S, are virtually identical.
On average in each country women rank at about the 30th percentile in the male wage
distribution. This implies that the large difference in the gender gap between the two countries is
entirely due to differences in wage structure, i.e., the larger wage penalty placed on women’s
lower position in the male wage distribution in the U.S. This means that at the same mean
percentile ranking, the resulting U.S. gender ratio is much lower than in Sweden.
As discussed above, the overall mean female percentile rankings in Figure 2show the
combined effect of both sets of gender-specific factors: qualifications and labor market
discrimination. It would be interesting to compare the U.S, and Sweden with respect to the
latter, that is their treatment of otherwise equally qualified men and women. We have seen that
the traditional estimate which involves computing adjusted earnings ratios, indicates that
Swedish women fare considerably better. As noted above, such estimates are subject to bias; and
15
the adjusted ratios presented in Figure 1are no exception. On the one hand, we may lack data
some factors which influence wages, although our inclusion of actual labor market experience
least surmounts this particular problem. On the other hand, we control for broadly defined
industry and occupation even though these variables may reflect the impact of labor market
discrimination. What is primarily of interest here, however, is the contrast between the
traditional measure, the adjusted ratio shown in Figure 1, and the female residual percentile
on
at
ranking shown in Figure 2.
For each country, we drop each woman’s residual from the male wage regression into the
distribution of male wage residuals and find the female mean of the resulting percentiles. This is
an indicator of the relative wages of women in each country &controlling for gender
differences in personal characteristics, industry and occupation. Unlike the traditional measure,
it is not contaminated by differences in wage structure between the two countries, The caveat of
course remains that, as with any analysis of this type, differences in the female rankings may also
represent cross-country differences in the unmeasured characteristics of women relative to men,
Indeed, in the absence of discrimination, one way to think of the rewards @enalties) associated
with ahigher (lower) position in the residual wage distribution would be as areturn to
unmeasured skills or characteristics.
Looking at Figure 2, we again find that while there are simble U.S.-Sweden differences
in the adjusted gender ratios, the mean percentile rankings of women in the residual wage
distribution are vitiually identical; in each country women rank at about the 37th percentile, on
average. This implies that the observed differences in the adjusted gender ratios in Figure 1are
entire] ydue to differences in wage structure between the two countries, i.e., the larger wage
16
penalty placed on women’s lower position in the male residual wage distribution in the U.S.
Putting this somewhat differently, the extent of labor market discrimination against women (or
unmeasured productivity differences between men and women) appears to be no greater in the
U.S. than in Sweden.
Our detailed decomposition of the U.S.-Sweden gender log wage differentials (Blau and
Kahn 1996) sheds additional light on the specific factors underlying these general results. Our
finding that gender differences in observed characteristics do not contribute to the U.S.-Sweden
difference in the gender gap reflects off-setting effects. On the one hand, there is asomewhat
smaller gender difference in actual experience in Sweden (5 years) than in the U.S. (6 years) and
aslightly more favorable relative occupational distribution of Swedish than of U. S. women. On
the other hand, this is offset by smaller gender differences in the U.S, in educational attainment
and industrial distribution. At acommon set of prices (or rewards for measured characteristics)
for both countries, the gender differences observed characteristics contribute about the same
amount to the male-female pay gap in each country. As we saw in our discussion of Figure 1,
however, using the conventional approach, differences characteristics contribute to alarger
gender differential in the U. S. than in Sweden. This is because the conventional approach uses
own-country prices; that is Swedish prices for Sweden and U.S. prices for the U.S. In general,
the prices of skills are higher in the U.S. and this means that the female deficits (i.e., the Iow-er
qualifications of women compared to men) are more heavily penalized in the U.S.
Breaking this price factor out separately, we find that the impact of differences in
observed prices between the U.S. and Sweden strongly favors Swedish women. The Swedish-
U.S. differences in relative rewards to employment by industry is the most important factor,s
17
although less favorable (for women) prices of education and experience in the U.S. also play a
role, Overall, the effect of wage structure, including the impact of prices of both measured and
unmeasured characteristics, is more than sufficient to account for the considerably larger gender
gap in the U.S.
Finally, Figure 3presents the female percentile comparisons of the U.S. and Sweden in
greater detail. It shows the female cumulative distribution finctions that results from placing
women in male wage deciles on the basis of the male log wage cutoffs. 9So, for example, in
the U. S., approximately 20 percent of women fall in the first decile of the male distribution of
log wages; almost 44 percent in or below the second male decile; etc. The results indicate that
our conclusions based on acomparison of the mean female percentiles in the male distribution
are full ysupported by the more detailed comparison, Specificallyy, the U.S. female cumulative
distribution finction is quite similar to that of Sweden. Particularly notable is the larger
proportion of women in the lowest male wage decile in Sweden (29 percent) than in the U.S. (20
percent). This suggests an important role for labor market institutions which tend to “bring up
the bottom” in reducing the gender pay gap in Sweden relative to the U.S. That is, although the
percentage of women who fall in the bottom male decile in Sweden generally exceeds that in the
U. S,, the gender pay gap is smaller in Sweden. This suggests that formal or de facto wage floors
in Sweden lessen the wage penalty for those in the bottom male decile.
Iwould argue that the rankings in Figure 2and the distributions in Figure 3provide an
informative basis for comparing the economic status of women in the two countries. The fact
that, on average, women in both the U.S. and Sweden “out-earn” about 30 percent of male
18
workers tells us agreat deal about the relative qualifications and treatment of women compared
to men in each country. What makes the rankings so informative in this respect, is that they are
not affected by differences in wage structure between the two countries. Moreover, our finding
that U.S. and Swedish women have similar rankings, on average, in the male wage distribution
is quite surprising given the large disparities in gender earnings ratios shown in Figure 1.
However, what this similar ranking in the male wage distribution in each of the two countries
buys in terms of relative wage levels is of course also extremely important. The wage is an
important indicator of economic well-being in and of itself and also, as noted above, asignificant
input into decision-making. So Iwould suggest that both measures, conventional earnings ratios
and percentile rankings, are of interest.
SWIMMING AGAINST THE TIDE: TRENDS IN THE GENDER GAP IN THE U.S.
While these findings on international differences in the gender gap help to resolve one
paradox, they generate another. Wage inequality has been increasing in the U.S. Our analysis
implies that in the face of rising wage inequality, American women are essentially swimming
upstream in alabor market growing increasingly unfavorable for workers with below average
skills. In the face of rising rewards to labor market skills (measured and unmeasured) as appears
to have occurred over the last 15 to 20 years, women’s relative skills and labor market treatment
have to improve merely for the pay gap to remain constant; still larger gains are necessary for it
to be reduced. Yet the gender pay gap has actually been falling in the U.S. since the late 1970s,
How can we explain this apparent contradiction?
19
Kahn and Ihave examined this issue (Blau and Kahn 1994, Blau and Kahn, forthcoming
b) and found that women were able to swim against the current during the 1970s and 1980s and
narrow the gender gap because gains in gender specific factors were large enough to
counterbalance the negative effect of the trends in wage structure on their relative earnings.
Women gained from improvements in their relative qualifications, particularly their relative
10
experience and occupational distribution. Alarger negative impact of de-unionization on male
than female workers also contributed to anarrowing of the differential. Women also benefited
from asubstantial decline in the “unexplained” portion of the gender gap, particularly when the
adverse effects of widening residual inequality are netted out. These reductions in the
unexplained gap may reflect improvements in unmeasured characteristics or reductions in
discrimination. Both explanations are credible for this period. Since women improved the
relative level of their measured characteristics, it is plausible that they also enhanced the relative
level of their unmeasured characteristics. Further, as women increased their commitment to the
labor market and their other job skills, it is possible that the rationale for statistical discrimination
against them diminished, Moreover, while government efforts to enforce the anti-discrimination
laws appear to have been reduced during the 1980s (Leonard 1989), it is possible that women’s
relative wage gains indirectly reflect the impact of government enforcement efforts in earlier
years which had the effect of encouraging women to train for and enter traditionally male fields.
Another insight that may be derived from afocus on wage structure relates to the
possibility that shifts in skill prices may have impacted men and women differently. In particular
men and women appear to be viewed by employers as imperfect substitutes in the labor market.
This is suggested by the considerable differences in the occupations and industries in which they
20
work, as well as the substantial pay differences that exist for men and women with the same
measured characteristics (e.g., Blau and Ferber 1992), Thus, while rising skill prices may be
expected to widen the gender pay gap, such changes need not affect men and women in precisely
the same way. We examined this issue in Blau and Kahn (forthcoming b) and found that, over
the 1980s, shifis in demand for output across industry-occupation cells favored women over men
for low and medium skilled workers, but men over women among high skilled workers. The
growth in the supply of women was also considerably larger at high skill levels.
Such a“gender twist” in the net supply of skill may have affected the relative gains of
women within skill groups. We find evidence that this is indeed the case, ~ile the unadjusted
pay gap closed slightly faster for high skill women over the 1980s, industry and union
representation effects strongly favored women at the bottom and middle of the skill distribution
relative to those at the top. High skill women nonetheless advanced at aroughly similar pace as
the other groups due to the large improvement in their human capital characteristics and
occupational distribution. Further, wage gains for low skilled women would have been greater
had the minimum wage not declined in real value over this period. The progress of high skilled
women during the 1980s is particularly impressive given the relatively unfavorable demand and
supply shifis they faced.
CONCLUSION
As wage inequality has been increasing in recent years in many of the industrialized
countries, labor economists have increasingly turned their attention to understanding its
21
determinants andthereasons forchanges overtime. Inthistalk Ihaveendeavored tohighlight
the role of wage structure in determining the size of the gender pay gap both across countries and
within aparticular country, in this case the United States, where wage inequality is especially
large, In addition to bringing auseful new construct to bear on analyses of male-female pay
gaps, such afocus serves to integrate analyses of demographic differentials into the study of
wage structure in general. This in itself is an interesting new development in labor economics.
From the perspective of our consideration of the determinants of the gender gap, the
analysis of the trends in the pay gap over time in the U.S. provides an interesting comparison to
our consideration of international differences in the gender gap. With respect to the puzzle of the
relative] yhigh U.S. gender pay gap compared to other countries, wage structure provides the
whole story—the traditional gender-specific factors do not appear to play arole. In contrast,
with respect to the narrowing of the gender gap over time in the U.S., the traditional gender
specific factors, i.e., improvements in women’s relative qualifications and declines in labor
market discrimination as conventionally measured, are an extremely important part of the story,
The insight which wage structure contributes is nonetheless also important; that is, the notion
that women were indeed swimming against the tide. In the absence of substantial improvements
in the gender-specific factors, the gender gap would have widened not narrowed. Acomparison
of these findings also serves to illustrate the more general point that the relative importance of
gender-specific factors vs. wage structure in any particular situation is an empirical question.
With hindsight it is perhaps not surprising that wage structure proved to be amore significant
pti of the story in the international comparison than in the intertemporal one, Differences across
countries in wage inequality, particularly between the U.S. and other industrialized nations, are
22
of considerably greater magnitude than the changes in the level of inequality that have occurred
in the U.S. over time, as significant as those changes have been.
The points made here about the potential importance of wage structure for the gender pay
gap can readily be expanded to encompass the relative wages of other demographic groups, such
as racial or ethnic minorities or immigrants, which have below average skills andor face
discrimination in the labor market. Thus, the well-being of particular demographic groups
depends not only on group-specific factors like relative qualifications and the extent of
discrimination against them, but also on the market factors and institutional arrangements which
determine the return to skills in general and the relative rewards to employment in the particular
sectors of the economy. This insight has anumber of policy implications. While Iwill
particularly address policies aimed at women, these point can readily be generalized to other
demographic groups.
First, in evaluating the effectiveness of “gender-specific” policies, that is polices which
are specifically designed to impact economic outcomes for women, it is important to net out the
effects of wage structure. On the one hand, policies may be erroneously deemed ineffective, or
their impact may be underestimated, because the positive effect of the policies are disguised by
adverse shifts in wage structure. So, for example, cross-national comparisons of the gender gap
could lead one to conclude that gender-specific policies in the U.S. have been relativel y
ineffective in comparison to those in other countries. On the contrary, my work with Kahn
implies that U.S. gender specific policies have been quite successful. U.S. women have lagged
behind those in other countries because of the high level of wage inequality in the U.S. which
heavily penalizes workers with below average wages, regardless of gender. On the other hand,
23
gender-specific policies could be incorrectly judged successful, or the extent of their success
exaggerated, if they happen to be accompanied by changes in wage stmcture which benefit
women as agroup.
Asecond potential policy implication that follows from our work is that outcomes for
women are affected not only by policies specifically targeted at them, but also by wage structure
in general. This means that policies designed to alter wage structure, such as the promotion of
more centralized wage determination or the establishment of relatively high minimum wages,
constitute an alternative approach to improving wage outcomes for women. While this
constitutes apotential benefit of such policies, it is important to bear in mind that they also have
costs which need to be balanced against this benefit. These costs may be substantial.
First there is the concern that minimum wages, particularly the imposition of relatively
high wage floors, may create unemployment. 1] Second, centralized wage setting may allow
firms too little flexibility to respond to differences in market conditions across industries or at the
local level. Moreover, the compression of wage premia for skills may dampen workers
incentives to acquire appropriate training. 12 Finally, overly ambitious attempts to regulate the
labor market may result in the growth of an uncovered sector, as appears to be the case, for
example, in Italy.
The substantial potential costs to direct government intervention in wage setting suggest
caution in using this approach to attack gender (or other demographic) differentials. An
additional issue is that developments in the 1980s and 1990s have led to the decentralization of
bargaining in virtually every industrialized country (see, Katz 1993, Edin and Holmlund 1995,
and Edin and Topel 1994). As the protection of centralized wage structures fall away, women
24
who continue to have less human capital on average and to encounter labor market
discrimination are Iefi exposed to downward pressures on their relative wages. Thus, the
fundamental answer to the gender pay gap may well with gender-specific policies designed to
increase women’s human capital and reduce discrimination against them. In away, this
conclusion is not entirely surprising in that it is these gender-s~ecific differences in qualifications
and treatment that constitute the basic cause of women’s lower labor market outcomes. Were
their no such differences, men and women would be similarly affected by the overall wage
structure and by changes in it,
25
REFERENCES
Aigner, Dennis and Cain, Glen (1977) “Statistical Theories of Discrimination in Labor Markets,”
Industrial and Labor Relations Review 30, 2: 175--87.
Becker, Gary S. (1957) The Economics ofDiscrimination, Chicago: University of Chicago
Press,
Becker, Gary S. (1985) “Hunlan Capital, Effort, and the Sexual Division of Labor,” Journal of
Labor Eco nomics 3, l(S): S33--S58.
Bergmann, Barbara (1974) “Occupational Segregation, Wages and Profits When Employers
Discriminate by Race or Sex,” Eastern Economic Journal 1, 1--2: 103--10.
Bemhardt, Amette, Morris, Martina and Handcock, Mark (forthcoming) “Women’s Gains or
Men’s Losses? ACloser Look at the Shrinking Gender Gap in Earnings, ”~merican
Jou nal of Sociology.
r
Blau, Francine D. (1977) Equal Pay in the Office, Lexington, MA: DC Heath.
Blau, Francine D. (1984) “Discrimination Against Women: Theory and Evidence,” in William
A. Darity (cd.) Labor Economics: Modem Views, Boston: Kluwer-Nijhoff Publishing.
Blau, Francine D. and Ferber, Marianne A. (1992) The Econo mics of Women. Men. and Work.
2nd ed,, Englewood Cliffs, NJ: Prentice Hall.
Blau, Francine D. and Kahn, Lawrence M. (1992a) “The Gender Earnings Gap: Learning from
International Comparisons,” ~merican Economic Review 82,2: 533--8.
Blau, Francine D, and Kahn, Lawrence M, (1992b) “Race and Gender Pay Differentials,” in D.
Lewin, O. Mitchell, and P. Sherer (eds.) Research Frontiers in Industrial Relations and
Human Resources. Madison, WI.: Industrial Relations Research Association.
Blau, Francine D. and Kahn, Lawrence M. (1994) “Rising Wage Inequality and the U.S.
Gender Gap,” American Economic Review 84,2: 23--8.
Blau, Francine D. and Kahn, Lawrence M. (1995) “The Gender Earnings Gap: Some
International Evidence,” in Richard Freeman and Lawrence Katz (eds,,) Differences and
Ch n~es
ain Wage Structures. Chicago, IL: University of Chicago Press.
Blau, Francine D. and Kahn, Lawrence M. (1996) “Wage Structure and Gender Earnings
Differentials: an International Comparison,” ~co nomica 63. 250 (S): S29--S62.
Blau, Francine D. and Kahn, Lawrence M. (forthcoming a) “International Differences in Male
Wage Inequality: Institutions Versus Market Forces,” Journal of Political Economv.
26
Blau, Francine D. and Kahn, Lawrence M. (forthcoming b) “Swimming Upstream: Trends in
the Gender Wage Differential in the 1980s,” Journal of Labor Economics.
Borjas, George and Ramey, Valerie (1995) “Foreign Competition, Market Power, and Wage
Inequality: Theory and Evidence,” Quarterly Joumal of Economics 110,4: 1075--1110,
Bound, John and Jokson, George (1992) “Changes in the Structure of Wages in the 1980s: An
Evaluation of Alternative Explanations,” American Economic Review 82, 3: 371--92.
Card, David (fotihcoming) “The Effect of Unions on the Distribution of Wages,” Econometrics,
Card, David and Krueger, Alan B, (1995) Mvth and Measurement: The New Economics of the
~inimum Waee, Princeton, NJ: Princeton University Press.
Edgewofih, F. Y., “Equal Pay to Men and Women for Equal Work,” (1922) Economic Journal
32, 128: 431--57.
Edin, Per-Anders and Bertil Holmlund (1995) “The Swedish Wage Structure: The Rise and Fall
of Solidarity Wage Policy,” in Richard Freeman and Lawrence Katz (eds.) Differences
and Changes in Wage Structures, Chicago, IL: University of Chicago Press.
Edin, Per-Anders and Zetterberg, Johnny (1992) “Interindustry Wage Differentials: Evidence
from Sweden and aComparison with the United States,” ~merican Economic Review 82,
5: 341--49.
Edin, Per-Anders and Topel, Robert(1994)’’Wage Policy and Restructuring: The Swedish Labor
Market Since 1960,” unpublished, University of Chicago.
Engels,Friedrich(1972, orig. pub, 1884) The Ori~in of the Family, Private Property and the
State New York: International Publishers.
—,
Fawcett, Millicent G. (1918) “Equal Pay for Equal Work,” Economic Journal 28, 109: 1-6.
Freeman, Richard (1994) “How Much has De-Unionization Contributed to the Rise in Male
Earnings Inequality?” in Sheldon Danziger and Peter Gottschalk (eds,) Uneven Tides:
Risin~ Ineaualitv in Ameri ca, New York: Russell Sage Foundation.
Gillman, Charlotte Perkins (1966; orig. pub. 1898) Women and Economics: AStudv of the
Economic Relation Between Men and Wome nas aFactor of Social Evolution, New
York: Harper &Row.
Gottschalk, Peter and Joyce, Mary (1995) “The Impact of Technological Change, De-
industnalization and Internationalization of Trade on Earnings Inequality: An
International Perspective,” in K. McFate, R. Lawson and W.J. Wilson (eds.) Povertv,
Ineaualitv. andthe Future of Social Policy ,New York: Russell Sage Foundation.
27
Groshen, Erica L. (199 1)“The Structure of the Female/Male Wage Differential: Is it Who You
Are, What You Do, or Where You Work?” Joumal of Human Resources 26, 3: 457--72.
Industrial and Labo rRelations Review, (1995) “Review Symposium: Mvth and Measurement:
The New Econom ics of the Minimum Wae_e, by David Card and Alan B. Krueger,” 48,
4: 827--49,
Johnson, George and Solon, Gary (1986) “Estimates of the Direct Effects of Comparable Worth
Policy,” American Econom ic Review 76, 5: 1117--25.
Jti, Chinhui, Murphy, Kevin M. and Pierce, Brooks (1991) “Accounting for the Slowdown in
Black-White Wage Convergence,” in M, Kosters, cd., Workers and Their Waees,
Washington, DC: AEI Press.
Juhn, Chinhui, Murphy, Kevin M. and Pierce, Brooks (1993) “Wage Inequality and the Rise in
Returns to Skill,” Journal of Political Economv-, 101,3: 410--42.
Katz, Harry C. (1993) “The Decentralization of Collective Bargaining: ALiterature Review and
Comparative Analysis,” Jndus trial and Labor Relations Review 47, 1: 3--22.
Katz, Lawrence F., Loveman, Gary W. and Blanchflower, David (1995) “A Comparison of
Changes in the Structure of Wages in Four OECD Countries,” in Richard Freeman and
Lawrence Katz (eds.) Differences and Chanees in Wage Strut tures, Chicago, IL:
University of Chicago Press.
Katz, Lawrence F. and Murphy, Kevin M. “Changes in Relative Wages, 1963-87: Supply and
Demand Factors,” Ouarterly Joum al of Econom its, 107, 1: 35--78,
LaLonde, Robert J. and Topel, Robert H, (1992) “The Assimilation of Immigrants in the U.S
Labor Market,” in G. Borjas and R. Freeman, eds., Immigration and the Work Force,
Chicago:, 111.:University of Chicago Press.
Leonard, Jonathan (1989) “Women and Affirmative Action,” Journal of Economic Perspectives
3, 1: 61--75.
Lundberg, Shelly and Startz, Richard (1983) “Private Discrimination and Social Intervention in
Competitive Labor Markets,” American Economic Review 73, 3: 340--7.
Mill, John Stuart (1878; orig. pub. 1869) On the Subjection of Women, 4th cd., London:
Longmans, Green, Reader& Dyer,
Mincer, Jacob and Polachek, Solomon W. (1974) “Family Investments in Human Capital:
Earnings of Women,” Journal of Poht]
. .cal Econom y82,2, pt. 2: S76--S108.
O’Neill, June and Polachek, Solomon (1993) “Why the Gender Gap in Wages Narrowed in the
1980s,” Jouma 1of Labor Econo mics 11, 1: 205--28.
28
Polachek, Solomon (198 1) “Occupational Self-Selection: AHuman Capital Approach to Sex
Differences in Occupational Structure,” Review of Economics and Statistics 63, 1: 60--9.
Sorensen, Elaine (1990) “The Crowding Hypothesis and Comparable Worth Issue,” Journal of
Human Resources 25, 1: 55--89.
Sorensen, Elaine (1991 )Exploring the Reasons Behind the Narrowing Gender GaD in Eamin~s,
Washington, DC: The Urban Institute Press.
Webb, Beatrice (191 9) The Wa~es ofMen and Women: Should They Be Equal, London:
Fabian Bookshop.
Weiss, Yorem and Gronau, Reuben (198 1) “Expected Interruptions in Labour Force Participation
and Sex-Related Differences in Earnings Growth,” Review of Economic Studies 48, 4:
607--19.
Wellington, Alison J. (1993) “Changes in the Male/Female Wage Gap, 1976 -85,” Journal of
Human Resources 28,2: 383--411,
29
FOOTNOTES
]Work engender isdescribed below. Wage structure hasalso been found toplayarolein U.S.
trends in black-white andimigrmt-native wage differentials (Juh, Mu~hy and Pierce 1991,
LaLonde and Topel 1992).
2Wages are equal to average hourly earnings.
3That is, for each country j, we estimate amale wage equation:
(1) Yim =Xim~m +eim
where Yi is the log of wages; Xi is avector of explanatory variables including education,
experience and its square, and major industry and occupation, ~is avector of coefficients and ei
is aresidual. The adjusted wage ratio is:
(2) Ra =exp {-(em -ef)} -exp (ef) ❑~@f)/(X@m)
where em and ef are the mean residuals from the male wage regression for men (m) and women
(f), Xf is avector of means of the explanatory variables for women, and fim and J3fare vectors of
estimated coefficients from wage regressions estimated for men and women separately.
4The expected impact of family leave (disproportionately taken by women even when it is
available to men) is unclear apriori. On the one hand, it is possible that such policies raise the
relative earnings of women by encouraging the presemation of their ties to particular firms and
hence increasing the incentives of employers and women to invest in firm-specific training. On
the other hand, the existence of such policies could increase the incidence and/or duration of
temporary labor force withdrawals among women, raising the gender gap, Fufiher, the
incremental costs associated with mandated leave policies may increase the incentives of
employers to discriminate against women.
5Similarly, across our fill sample of countries, male and female wage and residual wage
variation are highly correlated (Blau and Kahn 1995 and Blau and Kahn 1996).
bWe assign each woman in country japercentile ranking in country j’s male wage distribution.
The female mean of these percentiles by country is presented in Figure 2as the “female
percentile.”
7We find the percentile ranking of each woman’s wage residual from the male wage regression
(eif) in the distribution of male wage residuals from the male wage regression (eim), The mean
female percentile for each country is presented in Figure 2as the “female residual percentile.”
30
8This is not surprising given Edin and Zetterberg’s (1992) finding that interindustry wage
differentials are much smaller in Sweden than in the U.S.
9Befiardt, Morris and Handcock (forthcoming) use asimilar methodology to examine the
distribution of women’s wages relative to the male distribution over time in the U.S.
10O’Neill and Polachek (1993) and Wellington (1993) also provide evidence of the importance
of women’s gains in relative experience in narrowing the gender gap.
1]For example, Katz, Loveman andBlanchflower(1995) report that in France, where the
minimum wage increased from 45.7 to 53.3 percent of median earnings from 1967 to 1987, the
problem of youth unemployment has been more severe and the duration of unemployment has
tended to be longer than in other OECD countries. And Edin and Topel (1994) find that the
solidarity wage policy followed in Sweden in the 1960s and 1970s disproportionately raised pay
and lowered relative employment in low wage industries. On the other hand there is evidence for
the U.S. that suggests that relatively small increases in the minimum wage do not have adverse
employment effects, See Card and Krueger (1995); for responses to their research see, ILRR
(1995).
12Both complaints have been voiced about Sweden’s “solidarity” wage policy by employers, and
that count~’s generous student stipends and subsidized loans for higher education may be
viewed in part as ameans of off-setting the distortions caused by wage compression (Edin and
Holmlund 1995).
LUND.XLS Chartl 6/20/96
85
75
70
Figure 1: The Gender Wage Ratio in the US and Sweden, 1984
66.9
65
60
55
Unadjusted
82.2
90.9
82,7
Adjusted
OUSA
8Sweden
Source: F. D. Blau and L. M. Kahn, “Wage Structure and Gender Earnings Differentials: An International
Comparison,” ECONOMICA, 1996.
LUND.XLS Chatt2 6/20/96
Figure 2: The Mean Female Percentile in the Male Distribution in the
U.S. and Sweden
40
38
36
34
32
30
28
26
37.4
36.6
29.9
LY.D
Female Female
Percentile Residual
Percentile
❑USA
-Sweden
Source: F. D. Blau and L. M. Kahn, “Wage Structure and Gender Earnings Differentials: An International
Comparison,” ECONOMICA, 1996.
LUND.XLS Chati3 6/20/96
Figure 3: Cumulative Distribution Function, Female Wages Relative
to the Male Wage Distribution, U.S. and Sweden
100.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
12 3 45678
Source: F. D.
Male Wage Decile
Blau and L. M. Kahn, “Wage Structure and Gender Earnings
910
‘—us
—-Sweden
Differentials: An International
Comparison,” ECONOMICA, 1996,
