American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic
Literature.
http://www.jstor.org
American Economic Association
Trends in the Well-Being of American Women, 1970-1995
Author(s): Francine D. Blau
Source: Journal of Economic Literature, Vol. 36, No. 1 (Mar., 1998), pp. 112-165
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2564953
Accessed: 18-03-2015 09:58 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Journal of Economic Literature
Vol. XXXVI (March 1998), pp. 112-165
Trends in the Well-Being of American
Women, 1970-1995
FRANCINE D. BLAU
Cornell Universityand NBER
I twouldlike to tlanttkDeborah Anderson, Marianne Ferber, Victor Fuchs, Joyce Jacobsen,
LawrenceKalhn,ElizabethPeters,Jane Walclfogel,Anne Winkler,antdanonymousrefereesfor
helpful commentsand suggestions. I aml17indebted to DeborahAnderson,SliubDepgupta, anid
Rana Sihwailfor excellenttr-esearch1assistanice.
I. Introduction
THE MAJOR GOAL of this paper is to
chart the trends in the well-being of
women in the United States over the last
quarter of a century. This interval covers
the period of the most significant shifts
in women's wage outcomes since 1940
and is long enough to form reasonable
conclusions regarding the extent and
magnitude of changes in other indicators
of well-being. The past 25 years are significant
not only as a period in which
women's labor market outcomes and
family structure changed substantially,
but also as a time of equally significant
changes in the labor market as a whole.
Chief among these has been the dramatic
widening of wage inequality
among workers. In a time of stagnating
overall real wage increases for men, this
has meant substantial declines in the real
wages of less skilled men. Moreover,
substantial decreases in the relative employment
of this group have also been
documented. The sources, dimensions,
and consequences of widening wage inequality,
particularly as they affect male
workers and family income distribution
have been the focus of considerable recent
research. Less attention has been
directed at the implications of these
trends for women's well-being, a major
focus of this paper.
Documentation of changes in
women's well-being cannot proceed
without the specification of a set of indicators
by which well-being may be
measured. A second major goal of this
paper is to propose a set of indicators
for this purpose and demonstrate their
usefulness for forming a more complete
picture of changes in women's well-being
than may be obtained elsewhere in
the literature. While there is always
likely to be some disagreement about
the usefulness of including any particular
indicator, I hope that the two basic
principles which guided me in my selection
will prove useful in future similar
endeavors.
First, it is important that a broad
range of indicators of well-being be employed,
encompassing not only labor
market outcomes like wages and occupations,
but also time available for leisure,
the level of family income, and
the share of women who are single family
heads, an important proxy for a family's
economic resources. In this re-
112
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 113
spect, this inquiry follows the example
set by Victor Fuchs (1986; 1988) and
James Smith and Michael Ward (1989).
In addition, recent work on bargaining
models of the family and feminist economics1
suggest that women's status
within the family is also of interest.
Thus, I have sought indicators with which
to measure changes in this facet of
well-being as well, including women's
relative wages among married couples
and data on the extent of domestic vio-
lence.
Second, in considering the trends,
it is important to examine both changes
in indicators of well-being for women
relative to men, and changes in absolute
levels for women over time, and to
conduct these inquiries at a disaggregated
as well as at an aggregate level.
An interest in relative outcomes of
women flows naturally from a concern
that women are in some sense in a disadvantaged
position in the labor market,
in the family, and in the larger
society. But such a perspective is
not necessary to motivate this investigation.
Even if gender differences in outcomes
were entirely due to differences
between men and women in preferences
and qualifications, it is still of
interest to know whether such factors
have resulted in widening or narrowing
differences in outcomes over time. In
addition to permitting an appraisal of
progress toward gender equality per se,
women's relative progress is also of
interest because data on males provide
a useful benchmark against which to
assess women's progress, in effect
enabling us to estimate a "period effect."
Thus, for example, we would
assess a 10 percent decline in real
wages for women difi-erently depending
on whether the real wages of comparable
males fell by 20 percent or increased
by 20 percent at the same
time.
Absolute trends in indicators of wellbeing
among women are also important.
So, for example, it instructive to know
not simply that the gender wage gap is
declining, but also the magnitude of any
real wages increases for women. Or, as
another example, we may inquire as to
whether in recent years women have
been upgrading their occupations in
some absolute sense or simply narrowing
the gender difference in occupational
distributions. The importance of
a disaggregated analysis is suggested by
the recent research on men which has
pointed to growing disparities among
them based on skill. This suggests the
potential usefulness of a comparable investigation
of how various groups of
women are faring relative to others,
both in terms of how quickly they are
narrowing the gap with men, but also in
terms of their standing on various indicators
of well-being compared with
other women. Because in most cases a
great deal more is known about the
trends for males and the trends in the
relevant gender differences than about
the absolute trends in indicators of
well-being for women and the differences
in these trends among various
groups of women, I have sought in this
review to place greater emphasis on the
latter.
One contribution of this paper is to
supplement what can be learned from
existing research by presenting new
tabulations of a variety of indicators of
women's well-being calculated primarily
from Current Population Survey
(CPS) data.2 While the broad outline ofI See Shelly Lundberg and Robert Pollak (1996)
for a review of the literature on family bargaining
models and the contributions in Marianne Ferber
and Julie Nelson (1993) for examples of issues
raised by feminist economics.
2 See the Data Appendix for details of the data
analysis.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
114 Journal of Economic Literature- Vol. XXXVI (March 1998)
many of the trends delineated here are
generally known, it is useful to present
data for the same years from a comparable
source and across a wide range of
outcomes. More important, data are
presented here at a more disaggregated
level than is generally the case so that
significant differences in trends by education
and age may be identified. Given
constraints of space, however, I was
not able to investigate these trends
separately by race in comparable detail;
this remains an important area for future
research. I also present new empirical
results for indicators which have
not been examined in the same way
before, including information on trends
in the wages of wives compared to their
husbands from the CPS and in the
time use of men and women from the
Panel Study of Income Dynamics.
Where possible, I compare my empirical
results with the broader literature
but I do not attempt to review this literature
in its entirety, which would of
necessity be a very different exercise.
Similarly, I endeavor to summarize
briefly and comment on what is known
about the causes of these trends, but,
here too, given the broad range of indicators
which I review, it is not possible
to fully consider this question in each
case. Finally, it is important to note that
when other work is considered, the primary
focus is on the economics literature.
Considerations of time and space
preclude doing justice to the voluminous
literature on these topics in other
fields.
For virtually all age and education
groups, I obtain substantial evidence of
rising gender equality in labor market
outcomes, notably labor force participation,
wages, occupational distributions,
and self-employment. I also find broad
evidence of greater gender parity within
married couple families as the housework
time of husbands has increased
relative to wives', and as the relative
wages of wives have risen compared to
their husbands'. However, parallel to
the recent evidence of the declining labor
market status of lower skilled men,
there has been a similar sharp decline
for less educated women, especially
high school dropouts, compared to
other women. This decline has occurred
across a wide variety of dimensions. For
example, while overall female participation
rates increased 23 percentage
points over the 1970 to 1995 period,
the participation rate of high school
dropouts, already below that of their
more educated peers in 1970, rose
by only 4 percentage points. For those
that were employed similar disparities
are evident with respect to the progress
of real wages, with average weekly
wages of full-time workers rising by
31.2 percent in real terms for all
women, but declining by 2.2 percent
for female high school dropouts. Finally,
while the headship rate of women
with less than a high school education
(12.1 percent) was fairly similar to that
of all women (9.4 percent) in 1970, the
incidence of single headship among
high school dropouts increased 12.2
percentage points, compared to an increase
of 6.5 percentage points among
all women.
The deteriorating relative economic
position of less educated women and
their families is a major finding of this
paper, drawn both from existing studies
and the original results presented here.
One question that these results raise,
and a question which has received remarkably
little attention in the literature,
is the possibility that such trends
reflect shifts in the composition of the
least educated category rather than
changes in their opportunities or behavior.
The share of the population with
less than 12 years of education has
fallen sharplyover the past 25 years, deThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 115
clining by over 60 percent among both
men and women. It is possible that as
this group has dwindled in size it has
become more negatively selected compared
to more highly educated Americans.
Were it possible to identify a
similarly constituted group in each
year, one might find smaller changes in
outcomes, or, indeed, no changes at all.
An additional compositional factor affecting
the least educated is that the
share who are foreign born has grown
considerably more rapidly for them
than for other education groups, and labor
market outcomes of immigrants
have been declining relative to natives,
even compared to natives with the same
measured characteristics (George Borjas
1995). On the other hand, it could
be argued that the college educated
group which has fared especially well
relative to high school dropouts, has increased
its share markedly-by nearly
two times among men and two and a
half times among women-potentially
making it a less positively selected
group. This suggests that, to the extent
that compositional factors are important,
their impact is an empirical ques-
tion.
Although these are undoubtedly complex
factors which will take considerable
effort to unravel completely, in
their analysis of trends in wage inequality,
Chinhui Juhn, Kevin Murphy, and
Brooks Pierce (1993) suggest a fairly
simple method for gaining some insight
into whether or not changes in the relative
status of the less educated reflect
true changes in outcomes and behavior
versus solely compositional shifts. One
can compare the magnitude of changes
in outcomes for a cohort over time to
changes experienced by successive cohorts
of individuals of the same age. Because
the former comparison could not
have been affected by changes in the
composition of the various education
groups, similar findings for the "within
cohort" and "across cohort" comparisons
would suggest true changes in opportunities
or behavior and not simply
compositional effects. I apply this technique
to a number of the indicators of
welfare considered below. The findings
strongly suggest that compositional
shifts do not entirely account for the
deteriorating economic position of less
educated Americans.
In the sections that follow, a broad
range of indicators of economic well-being
are considered. I first focus on
labor force participation and next turn
to an examination of majorlabor market
outcomes including wages and occupations,
and trends in self-employment. I
then consider indicators of women's
standard of living, including family
structure and income differences across
families, as well as indicators of
women's well-being within the family,
including the relative wages of husbands
and wives and housework hours
of men and women, as well as trends
in domestic violence. A more detailed
rationale for the inclusion of each
indicator and its relationship to
women's economic well-being is presented
below in conjunction with the
empirical results. The focus is on individuals
aged 25-64; and the following
four age groups are distinguished: 25-
34, 35-44, 45-54, and 55-64. Disaggregations
by education are presented for
the following four education groups:
less than 12 years, 12 years, 13-15
years, and 16 or more years.3 Throughout,
the term "married"is used to refer
to an individual who is married, spouse
present. A "single head" of a family or
subfamily is a person who is not married
spouse present, i.e., was never mar-
3 The specification of education in the 1995 CPS
differs significantly from previous years. These differences
and how they were dealt with are described
in the Data Appendix.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
116 Journal of Economic Literature, Vol. XXXVI(March 1998)
ried, or is separated, widowed, or di-
vorced.
II. Labor Force Participation: Trends
and Explanations
The labor force participation decision
is viewed as the outcome of a utility
maximizing decision by an individual,
possibly in conjunction with his or her
family, as to how much labor to supply
to the market. Thus, the relationship
between trends in participation and
women's economic well-being is not obvious.
One reason for examining participation
trends is that they underlie the
transformation in gender roles that has
occurred in recent years and thus
should be summarized, if only to provide
the background for understanding
changes in indicators more obviously
linked to well-being. However, our consideration
of labor force participation
also rests on more direct links between
female participation and women's well-
being.
First, with the increase in the incidence
of female-headed families and
single person female households, admittedly
not entirely exogenous developments,
there exists a segment of the female
population whose economic
well-being is quite obviously heavily dependent
on whether they participate in
the labor force and their earnings levels,
given participation. In addition,
within married couples, the contribution
of working wives has traditionally
been important in averting poverty and
has been found to play a role in determining
the extent of inequality in family
income (Maria Cancian, Sheldon
Danziger, and Peter Gottschalk 1993;
Lynn Karoly and Gary Burtless 1995).
Thus, the finding that, for example, the
growth in the labor force participation
of female high school dropouts has
lagged behind that of other groups is of
concern in light of their rising incidence
of single-headed families and the
well documented declines in the relative
earnings of the less educated men
that they are likely to marry.
Second, family bargaining models
suggest that, in married couple families,
women's participation in the labor force
and their level of earnings while employed
is likely to affect the distribution
of resources within marriage. These
models assume that the prefe.rences of
husbands and wives may diverge and
that outcomes for each spouse can be
modeled using game theory. In the
most widely applied of these models,
the cooperative bargaining model of
marriage, the utility received by each
spouse in the Nash bargaining solution
is positively related to the utility of
each partner at the "threatpoint." Most
commonly, threat point utility is taken
to be the utility that each would receive
if the marriage irretrievably broke
down, i.e., in the event of divorce. The
link to labor force participation is provided
by assuming that, if a divorce occurs,
each partner will maintain ownership
of income received separately
within marriage. Thus, a woman who
works outside the home will have a
higher utility at the threat point and
hence a more favorable distribution
within marriage. Studies reviewed by
Lundberg and Pollak (1996) provide
empirical support for this model in
their finding that earned or unearned
income received by the wife has a different
effect on demand patterns than
income received by the husband.4
Third, shifts in participation are also
4 A common finding of such studies is that children
appear to do better when their mothers control
a larger fraction of family resources. Additional
support for this view is provided by a
substantial literature in sociology which finds that
employed wives have a greater say in household
decision making than nonemployed wives (Paula
England and George Farkas 1986).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Wonmen 117
of importance because they influence
the average levels of labor market experience
of women, an important determinant
of the gender pay gap (Jacob
Mincer and Solomon Polachek 1974).
The relationship between rising participation
and women's average experience
is not obvious a priori (Smith and Ward
1985; Claudia Goldin 1990). On the one
hand, to the extent that rising participation
reflects increased entry of women,
female experience levels are likely to be
diluted by new entrants. On the other
hand, to the extent that participation increases
reflect more continuous attachment
over the life cycle, experience levels
will be raised. While the net effect
is an empirical question considered below,
in either case, participation trends
potentially affect gender differences in
labor market outcomes.
A. Trendsin Labor Force Participation:
GenderDifferences
An extensive literature documents
the substantial increases in the labor
force participation of women since
World War II, including Mincer's
(1962) classic study and the influential
papers in Smith (1980). Panel (A) of Table
1 illustrates these trends for our
time period, while Panel (B) presents
comparable tabulations for men. Data
are for the standard measure of labor
force participation based on survey
week status; both the employed and unemployed
are included in the labor
force.
Overall, female participation rose 23
percentage points between 1970 and
1995. While participation increased
substantially for all but the oldest age
group, the major development of these
decades was the substantial rise in
participation among younger women
due partly to postponements and reductions
in fertility and increases in the
divorce rate, but also reflecting a
substantial rise in labor force attachment
among new mothers, especially
among married women (e.g., Blau,
Ferber, and Anne Winkler 1998; Arlene
Leibowitz and Jacob Klerman 1995).
So, for example, published data indicate
that the participation rate of married
women with children under six rose
from 18.6 percent in 1960 to 30.3
percent in 1970 and 63.5 percent in
1995. Participation growth of women
slackened in.the early 1990s, most notably
in the two younger age groups.
Given the high rates that their participation
had attained by 1990, approximately
three-quarters were in the labor
force, such a slow-down is perhaps not
surprising.
As Goldin (1990) has shown, increases
in female participation patterns
have been fueled by rises in participation
both across and within cohorts.
This may be seen in Table 1 which
shows, looking across the rows, that
more recent cohorts have evinced
greater labor force attachment than
their predecessors, and, looking diagonally,
that participation has also risen
within specific cohorts as they age. An
exception is that for the oldest age category,
55-64, where retirement begins to
occur, there are some small cross-sectional
increases due to cohort effects,
but declining participation within cohorts
(aging from 45-54 to 55-64).
Male participation trends shown in Table
lB are in some respects the reverse
of those for women. As emphasized in
recent work by Juhn (1992), male participation
has been declining, even in
the prime working ages, though the
changes have been considerably less
than the shifts in female participation;
overall, the male rate fell by 6 points.
The net result of these participation
trends is that the difference in participation
rates between men and women
declined substantially between 1970
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
118 Journal of EcononmicLiterature, Vol. XXXVI(March 1998)
TABLE IA
LABOR FORCE PARTICIPATIONRATESOF WOMEN BYAGE AND EDUCATION, 1970-95
1970 1980 1990 1995
I. Total 0.490 0.596 0.689 0.715
II. By Age
Age 25-34 0.456 0.660 0.737 0.745
Age 35-44 0.513 0.660 0.763 0.771
Age 45-54 0.544 0.598 0.706 0.752
Age 55-64 0.437 0.417 0.453 0.492
III. By Age and Education
A. Education < 12 years 0.430 0.439 0.462 0.472
Age 25-34 0.403 0.490 0.505 0.505
Age 35-44 0.476 0.537 0.569 0.542
Age 45-54 0.479 0.476 0.507 0.538
Age 55-64 0.367 0.308 0.319 0.319
B. Education = 12 years 0.513 0.614 0.687 0.689
Age 25-34 0.455 0.642 0.721 0.722
Age 35-44 0.527 0.676 0.763 0.751
Age 45-54 0.578 0.625 0.716 0.731
Age 55-64 0.494 0.472 0.471 0.490
C. Education 13 to 15 years 0.509 0.665 0.759 0.773
Age 25-34 0.455 0.704 0.780 0.772
Age 35-44 0.527 0.700 0.803 0.824
Age 45-54 0.570 0.666 0.774 0.799
Age 55-64 0.506 0.491 0.555 0.586
D. Education 16 + years 0.608 0.736 0.811 0.828
Age 25-34 0.576 0.775 0.850 0.856
Age 35-44 0.576 0.749 0.824 0.835
Age 45-54 0.674 0.739 0.830 0.856
Age 55-64 0.641 0.545 0.595 0.636
Source: Author'stabulations from the March Current Population Surveys.
Notes: Sample for each year includes all adult civilian women between ages 25 and 64. Labor force participation is
measured during the surveyweek.
and 1995 from 45 to 16 percentage
points.
As discussed above, rising female labor
force participation has ambiguous
effects on the average experience levels
of employed women depending on the
relative importance of increases in entry
and reductions in exits (i.e., increasing
labor force attachment) in producing
the trends. The central finding here
too has been one of declining gender
differences over this period (e.g.,
Goldin 1990; Smith and Ward 1985).
However, this decrease did not occur
immediately even though women's agespecific
experience levels began to increase
in the 1970s. This is because the
especially large increases in participation
of younger women which occurred
over the decade resulted in a decrease
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 119
TABLE IA (Cont.)
Change
1970-80 1980-90 1990-95 1970-95
I. Total 0.107 0.093 0.025 0.225
II. By Age
Age 25-34 0.203 0.078 0.007 0.289
Age 35-44 0.147 0.103 0.008 0.258
Age 45-54 0.054 0.108 0.046 0.208
Age 55-64 -0.020 0.036 0.039 0.054
III. By Age and Education
A. Education < 12 years 0.009 0.023 0.010 0.042
Age 25-34 0.087 0.014 0.000 0.101
Age 35-44 0.061 0.033 -0.027 0.066
Age 45-54 -0.003 0.030 0.031 0.059
Age 55-64 -0.059 0.011 0.001 -0.047
B. Education = 12years 0.100 0.074 0.002 0.176
Age 25-34 0.187 0.079 0.000 0.267
Age 35-44 0.149 0.088 -0.012 0.224
Age 45-54 0.048 0.091 0.015 0.153
Age 55-64 -0.022 -0.001 0.020 -0.004
C. Education 13 to 15 years 0.156 0.094 0.014 0.264
Age 25-34 0.249 0.075 -0.007 0.317
Age 35-44 0.173 0.103 0.021 0.297
Age 45-54 0.095 0.108 0.025 0.229
Age 55-64 -0.015 0.065 0.031 0.080
D. Education 16 + years 0.127 0.076 0.016 0.219
Age 25-34 0.199 0.075 0.006 0.280
Age 35-44 0.172 0.076 0.011 0.259
Age 45-54 0.065 0.091 0.026 0.181
Age 55-64 -0.096 0.050 0.041 -0.005
in the average age of women workers
(Goldin 1990). This is ironic in that it is
precisely this growing labor force attachment
of women during the childbearing
years which was necessary to increase
women's overall experience
levels in the long run. In any case, by
the 1980s, this process had played itself
out and the "experience gap" between
men and women began to fall. For example,
Blau and Lawrence Kahn (1997)
find that, among full-time workers, the
gender difference in full-time experience
declined from 7.5 years in 1979 to
4.6 years in 1988; similar trends are reported
in June O'Neill and Polachek
(1993). The gender difference in job
tenure (i.e., length of time with an employer)
was also reduced (Alison Wellington
1993).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
120 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE lB
PARTICIPATIONRATESOF MEN BYAGE AND EDUCATION, 1970-95
1970 1980 1990 1995
I. Total 0.935 0.898 0.888 0.874
II. By Age
Age 25-34 0.965 0.950 0934 0.926
Age 35-44 0.971 0.953 0.942 0.914
Age 45-54 0.946 0.912 0.904 0.890
Age 55-64 0.834 0.726 0.671 0.661
III. By Age and Education
A. Education < 12 years 0.893 0.794 0.751 0.720
Age 25-34 0.951 0.892 0.868 0.842
Age 35-44 0.947 0.885 0.818 0.756
Age 45-54 0.916 0.853 0.797 0.733
Age 55-64 0.793 0.621 0.537 0.508
B. Education = 12 years 0.963 0.922 0.899 0.869
Age 25-34 0.982 0.967 0.945 0.928
Age 35-44 0.982 0.965 0.939 0.910
Age 45-54 0.963 0.926 0.911 0.869
Age 55-64 0.888 0.773 0.693 0.650
C. Education 13 to 15 years 0.958 0.927 0.915 0.901
Age 25-34 0.958 0.946 0.942 0.939
Age 35-44 0.988 0.968 0.960 0.933
Age 45-54 0.975 0.929 0919 0.909
Age 55-64 0.875 0.794 0.694 0.704
D. Education 16 + years 0.961 0.955 0.945 0.938
Age 25-34 0.954 0.960 0.952 0.956
Age 35-44 0.988 0.985 0.983 0.970
Age 45-54 0.975 0.972 0.968 0.964
Age 55-64 0.900 0.850 0.797 0.777
Source: Author'stabulations from the March Current Population Surveys.
Notes: Sample for each year includes all adult civilian men between ages 25 and 64. Labor force participation is
measured duiing the survey week.
B. Trends in Labor Force Participation:
Differences by Education within
Gender Groups
The results presented in Table 1 also
illustrate the well known strong positive
association between educational attainment
and labor force participation in
each year. Of more significance for
trends in well-being, however, are the
sharp differences in participation trends
by education for both men and women
which resulted in a steepening association
between education and participation
among both groups, with an especially
large increase in the participation
gap between high school dropouts and
others. The sharp declines in participation
of less educated men have received
considerable attention in recent years.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Womnen 121
TABLE lB (Cont.)
Change
1970-80 1980-90 1990-95 1970-95
I. Total -0.037 -0.010 -0.014 -0.061
II. By Age
Age 25-34 -0.015 -0.015 -0.009 -0.040
Age 35-44 -0.018 -0.011 -0.028 -0.057
Age 45-54 -0.034 -0.007 -0.014 -0.055
Age 55-64 -0.108 -0.055 -0.010 -0.173
III. By Age and Education
A. Education < 12 years -0.099 -0.043 -0.032 -0.173
Age 25-34 -0.059 -0.025 -0.025 -0.109
Age 35-44 -0.062 -0.067 -0.061 -0.190
Age 45-54 -0.063 -0.057 -0.064 -0.183
Age 55-64 -0.172 -0.084 -0.030 -0.285
B. Education = 12 years -0.042 -0.023 -0.030 -0.094
Age 25-34 -0.016 -0.022 -0.017 -0.055
Age 35-44 -0.016 -0.026 -0.029 -0.071
Age 45-54 -0.037 -0.015 -0.042 -0.095
Age 55-64 -0.115 -0.080 -0.043 -0.238
C. Education 13 to 15 years -0.031 -0.012 -0.014 -0.057
Age 25-34 -0.012 -0.004 -0.003 -0.019
Age 35-44 -0.020 -0.008 -0.027 -0.056
Age 45-54 -0.045 -0.011 -0.010 -0.066
Age 55-64 -0.081 -0.101 0.010 -0.171
D. Education 16 + years -0.006 -0.010 -0.006 -0.023
Age 25-34 0.006 -0.008 0.004 0.002
Age 35-44 -0.003 -0.001 -0.014 -0.018
Age 45-54 -0.003 -0.004 -0.005 -0.011
Age 55-64 -0.050 -0.053 -0.020 -0.123
But it has not been recognized that an
analogous pattern of falling relative participation
has occurred among women
as participation increases of high school
dropouts have lagged considerably behind
those for other groups. Participation
rates of the least educated women,
already below those of their more
highly educated counterparts in 1970,
rose by only 4 percentage points over
the 25 year period compared to increases
of 19 to 26 points for the other
education categories. By 1995, only 47
percent of women with less than a high
school education were in the labor force
compared to 83 percent of college
graduates.
The role of compositional changes in
explaining the trends in participation by
education is investigated by comparing
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
122 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 2
DIFFERENCES IN LABOR FORCE PARTICIPATION
RATES BYEDUCATION, ACROSS AND WITHIN COHORTS, 1970-90
Across Cohort Change Within Cohort Change
1970-80 1980-90 1970-80 1980-90
I. Women
A. Difference: Ed = 12 minus Ed < 12
Age 25-34 0.100 0.065 0.087 0.042
Age 35-44 0.088 0.055 0.098 0.071
Age 45-54 0.050 0.060 0.065 0.003
B. Difference: Ed = 16+ minus Ed < 12
Age 25-34 0.112 0.060 0.039 -0.030
Age 35-44 0.111 0.043 0.162 0.112
Age 45-54 0.067 0.060 0.041 0.013
II. Men
A. Difference: Ed = 12 minus Ed < 12
Age 25-34 0.043 0.003 0.049 0.047
Age 35-44 0.045 0.041 0.038 0.034
Age 45-54 0.025 0.041 0.104 0.083
B. Difference: Ed = 16+ minus Ed < 12
Age 25-34 0.064 0.017 0.096 0.098
Age 35-44 0.058 0.066 0.078 0.072
Age 45-54 0.060 0.053 0.170 0.141
Notes: Calculated from results presented in Table 1. Illustrating for the 1970-80 participation change of the 25-34
year age group, the "across"cohort change is the difference between the participation rate of the 25-34 year age
group in 1970 and 1980; the "within"cohort change is the difference between the participation rate of the 25-34
year age group in 1970 and the 35-44 year age group in 1980.
across and within cohort changes in participation
for those with less than 12
years of schooling to the changes for
high school and college graduates.
Were compositional changes the full explanation
for the weakening position of
high school dropouts, we would expect
to see considerable differences in across
cohort changes in participation by education
group but relatively little difference
in within cohort changes. Table 2
which is derived from the data in Table
1 presents these comparisons. (Note
that the oldest age group is dropped because
only the across cohort changes
would be available for them.) A positive
sign indicates that the change for high
school and college graduates is algebraically
larger than for high school dropouts,
i.e., more positive or less negative.
Looking first at the results for women
shown in Panel I of the table, we see
that, for the comparisons of high school
drop outs to high school graduates, all
the within and across cohort comparisons
indicate a growing gap in participation
between the two groups and all but
one of the within cohort changes are
quite similar in magnitude to the corresponding
across cohort changes. The exception
is that, by the 1980s, the size of
the within cohort participation reduction
as 45-54 year olds aged to 55-64
was similar for high school dropouts
and high school graduates. Similarly,
with one exception, the comparisons between
high school dropouts and college
graduates indicate a widening differThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 123
ence between the two groups over time
both across and within cohorts. The exception,
is for the 25-34 year age
group: within cohort increases in participation
were smaller for college
graduates than for high school dropouts
over the 1980s. Similarly, over the
1970s, the within cohort advantage for
college graduates was considerably
smaller than their across cohort advantage.
These results are likely due to the
tendency of female college graduates to
delay childbearing; thus, many of them
are making first transitions into parenthood
as they age from 25-34 to 35-44,
at the same time that many women with
a high school education or less are experiencing
diminished child care responsibilities
as their children get older
and more self-sufficient. It may also be
noted that the participation rate for college
graduates in this age group was already
extremely high by 1980 (77.5),
thus limiting the scope for further increases
and making their within cohort
participation rate increase of 5 points
particularly impressive. Finally, as was
the case for the comparison with high
school graduates, the participation rate
drop off as the 45 to 54 year olds aged
over the 1980s was similar for high
school dropouts and college graduates.
Taken as a whole, these results
strongly suggest that compositional factors
are not the full explanation for the
growing disparity in participation rates
between high school dropouts and more
highly educated women. The results for
males shown in the lower portion of the
table are, if anything, even stronger in
this regard, indicating that the within
cohort changes are consistently larger
than the corresponding across cohort
changes. Thus for males as well it appears
that the declining relative participation
of the least educated does not
merely reflect compositional shifts.
Our discussion has focused on the labor
force participation trends which
show strikingly similar patterns for men
and women of decreasing relative labor
force participation among the less educated.
Findings by Mary Coleman and
John Pencavel (1993a, 1993b) strongly
suggest that the patterns noted here
also prevail along other dimensions of
labor supply. They document remarkably
similar trends of rising annual
hours worked for well-educated male
and female workers since 1940 and decreasing
trends in annual hours among
the less educated. The changes in magnitude
were larger for women than for
men. Further evidence of the deteriorating
labor market position of less
skilled workers is presented by Henry
Farber (1995) who finds a decrease in
the probability of being in long duration
jobs for less educated men between
1973 and 1993, while women with at
least a high school education were more
likely to be in such jobs.
C. Explaining the Trends
In his pioneering study of the postWorld
War II increase in married
women's labor force participation,
Mincer (1962) concluded that the rise
was due to a dominance of the positive
substitution effect generated by increases
in wives' own wages over the
negative income effect associated with
increases in the real wages of their husbands.
Extrapolating to our period of
stagnating real wages for men, it would
be tempting to suggest that married
women's participation has increased in
recent years at least in part to compensate
for the disappointing wage growth
of their husbands. This is the question
posed in a recent study by Juhn and
Murphy (1997) which convincingly
demonstrates that, while such an interpretation
appears plausible at the aggregate
level, it does not match up well
with the cross-sectional evidence. They
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
124 Journal of Economic Literature, Vol. XXXVI(March 1998)
find that the declines in male employment
and earnings have been greatest
for low wage mnen,but employment and
earnings gains have been largest for
wives of middle and high wage men.5
Moreover, for married women, the positive
relationship between employment
and own wages has grown stronger over
time while the negative relationship between
employment and husband's earnings
has grown weaker. Based on these
findings, they conclude, as did Mincer
(1962), that own wage effects continue
to dominate cross effects between husband
and wife in accounting for changes
in female employment.
These findings not only help to illuminate
the aggregate participation
trends, but also shed light on the widening
differences in participation across
education groups. As we shall see below,
more highly educated women have
experienced faster real wage growth
than less educated women, and real
wage declines have occurred for high
school dropouts. Because women tend
to marry men with similar levels of education,
low wage couples tend to be
comprised of men and women with low
levels of education. Thus, the labor
force participation rates of husbands
and wives in these families has been declining
relative to the participation
rates of higher wage, more highly educated
couples who experienced faster
real wage growth for both partners.
More generally, this suggests that worsening
relative wage prospects for less
educated women underlie their declining
relative participation just as has
been found to be the case in recent
studies of male labor force participation
(Tuhn 1992: Pencavel 1997).
As discussed in detail below, another
important development during the
1970-95 period was the especially large
increase in the incidence of singleheaded
families among less educated
women and blacks. Nonetheless, Table
3 which shows participation trends by
education, headship and race suggests
that the differences in the overall participation
trends across education
groups were driven by the trends for
married women. For all races combined,
married women's participation
rates rose by 8 percentage points among
high school dropouts, but by 22 points
among high school graduates, and by 29
points among women with some college.
Differences in trends by education
were much smaller for single heads and
others. Moreover, the participation
rates of high school dropouts were fairly
similar across headship/marital status
categories in 1995, with the highest rate
being 49 percent for single heads. Thus,
a slower growth in female headship
would not have raised the participation
rate of less educated women as a whole.
Because the slow rise of participation
rates among married women appears to
be primarily responsible for the declining
relative participation of less educated
women, welfare is unlikely to
have played a major role in explaining
it. This conclusion is reinforced by
Robert Moffitt's (1992) finding based
on an extensive review of the literature
that higher AFDC guarantee levels
have only a small negative effect on the
labor supply of female family heads.
Over the past 25 years, there have
been significant declines in the participation
rates of blacks relative to whites.
Among men this has taken the form of
faster declines in participation among
blacks. As may be seen in Table 3,
among women this was due to considerably
slower increases in participation
among blacks than whites; by 1995, the
5 The rising correlation of wife's and husband's
earnings has also been noted by others, e.g., Cancian,
Danziger, and Gottschalk (1993); and Karoly
and Burtless (1995). We return to this issue below
when we consider income trends across families.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of AmnericanWonmen 125
TABLE 3
LABOR FORCE PARTICIPATIONRATESOF WOMEN BYHEADSHIP, EDUCATION, AND RACE, 1970-95
All Whites Blacks
Change Change Change
1970 1995 1970-95 1970 1995 1970-95 1970 1995 1970-95
I. Total 0.490 0.715 0.225 0.479 0.719 0.240 0.590 0.704 0.114
II. Education < 12 years
Married, Spouse Present 0.393 0.469 0.076 0.380 0.466 0.086 0.500 0.531 0.031
Single Head 0.520 0.487 -0.033 0.531 0.514 -0.018 0.504 0.449 -0.055
Other 0.549 0.464 -0.085 0.544 0.478 -0.066 0.564 0.406 -0.158
III. Education = 12 years
Married, Spouse Present 0.453 0.668 0.215 0.443 0.667 0.223 0.613 0.699 0.086
Single Head 0.762 0.722 -0.040 0.776 0.754 -0.022 0.714 0.653 -0.061
Other 0.795 0.742 -0.053 0.798 0.756 -0.041 0.767 0.686 -0.081
IV. Education> 12 years
Married, Spouse Present 0.476 0.767 0.291 0.458 0.763 0.305 0.805 0.838 0.033
Single Head 0.785 0.851 0.066 0.784 0.863 0.079 0.791 0.826 0.035
Other 0.843 0.863 0.021 0.835 0.868 0.033 0.921 0.860 -0.061
Source: Author'stabulations from the March Current Population Surveys.
Notes: Sample for each year includes all adult civilian women between ages 25 and 64. Labor force participation is
measured during the survey week.
historic participation advantage of black
women had been eliminated. Table 3
also indicates that smaller participation
increases for black women prevailed
within all education and headship categories.
Analyzing the growing overall
race gap in employment rates among
men, Juhn (1992) finds that declining
wage opportunities for the low skilled
combined with the lower education levels
of black men explain a substantial
part of the differences between blacks
and whites. However, a significant component
cannot be explained and is consistent
with an inward shift in the supply
curve of black males. A similar
analysis of the slower growth of participation
among black women and the role
of behavior shifts vs. labor market opportunities
in producing it would be ex-
tremnelyvaluable.
Rising educational attainment of
women is also a factor in their increasing
labor force participation via its impact
on wages. The causation likely runs
in the opposite direction as well: as
women expect to participate in the labor
force more continuously over the
life cycle, they will be inclined to invest
more in their human capital. While
years of education have increased at
only a slightly faster pace for women
than men, women have been increasingly
pursuing college, graduate and
professional education and entering traditionally
male fields of study (Blau,
Ferber, and Winkler 1998; Jerry Jacobs
1995). While a human capital explanation
for the educational shifts fits well
in some respects, it should also be
pointed out that, to the extent that
women's expectations of encountering
labor market discrimination have deterred
their educational investments in
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
126 Journal of Economic Literature, Vol. XXXVI(March 1998)
the past,6 expectations of reductions in
discrimination would encourage them
to invest more and to enter traditionally
male fields.
Demographic factors have also contributed
to the aggregate increases in
labor supply for women. Decreases in
fertility, declines in marriage rates, and
increases in marital breakup move
women into categories with higher participation
rates (e.g., Goldin 1990).
Moreover, there is some evidence that
the higher probability of divorce has increased
the participation rates of married
women (William Johnson and
Jonathan Skinner 1986). Of course to
some extent these demographic trends
may themselves be due to rising market
opportunities for women (Gary Becker
1991).
Of considerable importance in understanding
the reasons for the intertemporal
shifts in women's participation is
the evidence that changes in measured
variables, including wages and demographic
factors can explain only a portion
of the observed increase; frequently
only a small portion. So, for
example, focusing on the traditional
economic variables, Juhn and Murphy
(1997) find that increases in women's
real wages can explain only 6-7 percent
of the total increase in women's employment
between 1969 and 1989.
While changes in demographic factors
likely explain some of the remainder,
behavioral shifts even in the impact of
demographic factors have played a part
as well. Of particular significance, not
just for the level of women's participation,
but also for its growing consistency
over the life cycle is that young
children exert a smaller negative influence
on wives' participation than formerly
(Leibowitz and Klerman 1995).
Thus, a considerable portion of the
change over time in female participation
remains "unexplained"by variables
conventionally used in our analyses. As
we shall see, this appears to be the case
in each of the broad areas examined.
During a period of major shifts in gender
roles, it is perhaps not surprising
that a significant portion of the explanation
for the changes in participation appear
to be due to behavioral shifts of
unknown or at least unquantified origin.
Moreover, it can be informative to ascertain
what particular aspects of behavior
have changed. So, for example,
the findings of a diminished negative
effect on female participation of young
children and higher husband's income,
and of a growing positive effect of own
wages, suggest that women's economic
role within the family has changed and
that their participation is increasingly
determined by their own opportunities
and less by the demographic and economic
circumstances of their families.7
III. Labor MarketOutcomes:Wages
and OccupationalDistributions
The inclusion of wages as an indicator
of economic well-being requires little
justification because they are of obvious
fundamental importance as a major determinant
of economic welfare for employed
individuals, as well as of the potential
gain to market employment for
those not currently employed. Further
they serve as a significant input into a
myriad of decisions ranging from labor
6 Blau and Ferber (1991) find that while female
college seniors expected equal starting salaries
with men in the same field, they expected much
lower salaries later in their career, even under the
assumption of equal labor force attachment. It
should also be noted that a reduction in discrimination
in educational institutions themselves may
have also played a role; Title IX which bans such
discrimination was passed in 1972.
7 Goldin (1990) provides an excellent example of
how evidence of shifts in supply responsiveness
may be interpreted to suggest insights into the underlying
factors that caused them; see pp. 136-38.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 127
supply to marriage and fertility (Becker
1991), as well as a factor potentially influencing
bargaining power and relative
status within the family.
A major motive for considering sex
differences in occupational distributions
derives from their association
with earnings. Considerable research
suggests that predominantly female occupations
pay less, even controlling for
measured personal characteristics of
workers and a variety of characteristics
of occupations and industries.8 Occupational
differences between men and
women may reflect differences in preferences
or discrimination. It is not an
easy matter to distinguish between
these two empirically, and of course the
outcome may reflect a combination of
both. While most would allow that gender
differences in preferences play
some role, and there is considerable
evidence to support that view (e.g.,
Morley Gunderson 1989), the claim that
discrimination is also important is more
controversial. Some of the more convincing
evidence of the importance of
discrimination comes from descriptions
of institutional barriers that have historically
excluded women from particular
pursuits or impeded their progress
(e.g., Barbara Reskin and Heidi Hartmann
1986). In addition the finding
that women are less likely to be promoted,
all else equal, is also suggestive
though it suffers from the standard
problems of this type of exercise discussed
at greater length below.9 Moreover,
to the extent that less on-the-job
training occurs in female than in male
jobs, the incentives for women to participate
in the labor force continuously
are increased when women enter traditionally
male jobs. This is the case
whether the changes are due to voluntary
choice or declines in discrimination.
It might also be argued that the tie
to earnings is not the only reason for
including occupations as an indicator of
well-being. Occupational segregation itself
may have deleterious effects on
women's economic status by reinforcing
exaggerated notions of gender differences
in capabilities, preferences, and
social and economic roles. Such views
could adversely affect the labor market
outcomes even of women who enter traditionally
male pursuits. In addition to
occupation, we also consider the growth
in self-employment of women as at least
potentially indicating an expansion in
opportunities.
A. Issues in Measuring Wages
An important factor in interpreting
our results for wages and other labor
market outcomes is that they may be affected
by changes over time in the selfselection
of individuals into the labor
force This is of particular concern in a
study focusing on women because there
have been considerable increases in
their labor force participation rates over
time, as well as differences in growth
rates across education categories. Two
types of selectivity are potentially at issue.
The first relates to changes in the
degree of self-selection by measured
characteristics, as, for example, more
highly educated women experience
faster increases in participation. This
type of selection is of lesser concern because
all our results are disaggregated
by important measured characteristics,
i.e., education and age, thus in effect
adjusting for such shifts. (Although it
should be noted that age is a much
8 See, e.g., Elaine Sorensen (1990). A recent
study by David Macpherson and Barry Hirscl
(1995) using a 1973-93 panel of data from the
CPS, however, finds that the negative wage effect
of percent female in the occupation is substantial[
reduced when longitudinal wage change
mo els are estimated to control for unobserved
fixed effects.
9 For citations of such studies, see, Blau, Ferber,
and Winkler (1998).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
128 Journal of Economic Literature, Vol. XXXVI(March 1998)
poorer proxy for labor market experience
for women than it is for men.) In
addition, studies of the impact of
changes in men's and women's measured
characteristics, including actual labor
market experience, on the gender
gap over time are reviewed below.
Of considerably greater concern,
however, are changes in the degree of
selectivity of workers on the basis of
their unmeasured characteristics; this
important issue was brought to the fore
in the highly influential work of James
Heckman (1980). The expected effect
of this factor on the trends is uncertain.
It is theoretically possible for labor
force participants to be either a positively
selected group of those with especially
high wage offers, controlling for
measured characteristics, or a negatively
selected group of those with especially
low values of nonmarket time, all
else equal. In the former case, an increase
in the relative size of the labor
force group is likely to make it less positively
selected. In the latter case, an expansion
of the labor force could result
in its becoming less negatively selected.
A conventional approach to adjusting
for selectivity bias in unmeasured characteristics
is to employ a Heckman
(1980) selectivity bias correction to obtain
consistent estimates of the coefficients
in the earnings equation. Using
this approach, it would be possible to
obtain estimated mean male and female
wage offers after adjusting for selectivity
bias. Such parametric models have,
however, been criticized for their lack
of robustness: "seemingly small misspecifications
[including heteroskedasticity
and nonnormality] may generate large
biases in estimates" (Charles Manski
1989, p. 356). For this reason, I do not
employ such an approach here. If we
take as the most plausible scenario that
the female labor force has become less
positively selected over time as it has
grown, the degree of closing of the gender
wage gap is understated in the results
presented below, but, as our discussion
above suggests, the possibility
that women's relative wage growth is
overstated cannot be completely ruled
out. It may be noted, however, that
where the Heckman correction has been
used, it has uniformly been found that
women narrowed the gender wage gap
even after such an adjustment (Blau and
Andrea Beller 1988; Wellington 1993).
To study trends in wages, it is necessary
to define an earnings measure for
the CPS data which adjusts for time input.
Weekly wages, defined for the calendar
year preceding the survey, are
used here; they are computed as annual
earnings divided by annual weeks
worked for full-time workers. This measure
is employed because it is available
in comparable form for each of the sample
years; computation of a comparable
hourly wage variable is complicated by
the absence of data on usual weekly
hours in the 1970 CPS. The issues involved
are considered at greater length
in the data appendix. A final point to
note is that money wages, however
measured, are an incomplete indicator
of total compensation which would take
into account not only nonwage benefits
but also compensating differentials for
job amenities. This is far from a trivial
issue and particularlythe latter is a concern
given the likelihood of substantial
differences in occupational preferences
between men and women. Complex
issues are also raised with respect to
nonwage benefits because, in some instances,
married women may be covered
under their husbands' plans, thus
reducing their demand for these benefits.
Unfortunately, a full consideration
of these issues would take us well beyond
the scope of this paper; at the
same time the relevant data and prior
research needed for such an investigaThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
BWan-nThe Wpl-Binory of Ameprican Wnmen 129Q
TABLE 4
FEMALE-MALE WEEKLY WAGE RATIOS BYAGE AND EDUCATION, 1969-94
(FULL-TIME WORKERS)
Ratio Percent Change
1969 1979 1989 1994 1969-79 1979-89 1989-94 1969-94
I. All workers 0.562 0.583 0.682 0.717 3.87 17.00 5.00 27.61
Age 25-34 0.598 0.665 0.785 0.827 11.20 18.06 5.37 38.33
Age 35-44 0.520 0.542 0.664 0.720 4.29 22.53 8.37 38.49
Age 45-54 0.541 0.535 0.620 0.651 -1.16 16.03 5.00 20.43
Age 55-64 0.600 0.554 0.582 0.615 -7.66 5.03 5.78 2.59
II. Education < 12years 0.560 0.568 0.682 0.670 1.48 20.08 -1.79 19.67
Age 25-34 0.556 0.630 0.738 0.713 13.34 17.02 -3.32 28.22
Age 35-44 0.546 0.581 0.685 0.639 6.28 17.94 -6.74 16.89
Age 45-54 0.557 0.532 0.660 0.662 -4.48 24.00 0.36 18.87
Age 55-64 0.574 0.558 0.603 0.607 -2.84 8.09 0.77 5.82
III. Education = 12 years 0.553 0.577 0.662 0.710 4.22 14.79 7.24 28.30
Age 25-34 0.577 0.621 0.712 0.779 7.60 14.67 9.42 35.00
Age 35-44 0.526 0.539 0.657 0.700 2.53 21.95 6.42 33.06
Age 45-54 0.536 0.558 0.596 0.651 4.00 6.91 9.22 21.44
Age 55-64 0.563 0.567 0.606 0.633 0.83 6.83 4.44 12.51
IV. Education 13 to 15 years 0.568 0.608 0.680 0.715 7.06 11.89 5.11 25.91
Age 25-34 0.617 0.676 0.748 0.776 9.54 10.59 3.78 25.71
Age 35-44 0.538 0.565 0.665 0.711 4.93 17.68 6.93 32.05
Age 45-54 0.526 0.557 0.641 0.681 5.76 15.13 6.32 29.45
Age 55-64 0.569 0.534 0.612 0.663 -6.15 14.67 8.29 16.54
V. Education 16 + years 0.589 0.598 0.694 0.722 1.51 16.01 4.03 22.52
Age 25-34 0.660 0.715 0.796 0.831 8.36 11.34 4.34 25.89
Age 35-44 0.554 0.561 0.698 0.736 1.36 24.40 5.49 33.02
Age 45-54 0.540 0.526 0.645 0.685 -2.64 22.63 6.25 26.85
Age 55-64 0.542 0.570 0.617 0.647 5.17 8.34 4.80 19.40
Source: Author'stabulations from the March Current Population Surveys.
Notes: The sample for each year includes full-time workers between ages 25 and 64. The weekly wage ratio is
calculated as WAGE,WAGE,,,, where WAGE, is the mean weekly wage of women and WAGE., is the mean
weekly wage of men. See the Data Appendix for additional sample restrictions and wage definitions.
tion are considerably sparser than one
would like. Thus the focus here is on
the traditional wage.
B. Trendsin the Gender Wage Gap
Gender wage ratios for all workers
and for age and education categories
separately are shown in Table 4. Overall,
there has been a substantial reduction
in the gender gap over the past
25 years as the gender ratio rose from
56.2 percent in 1969 to 71.7 percent in
1994. For the most part, gains were
concentrated in the post-1979 period,
as has been noted in a number of previous
studies, although some progress
is discernible among younger women
in the earlier decade.10 The largest
10 This also accordswith published data summarized
in Blau, Ferber, andWinkler (1998).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
130 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 5
PERCENTAGE CHANGES IN REAL WEEKLY WAGES, BYSEX, AGE AND EDUCATION, 1969-94
(FULL-TIME WORKERS)
Women Men
1969-79 1979-89 1989-94 1969-94 1969-79 1979-89 1989-94 1969-94
I. All workers 12.07 11.98 4.51 31.15 7.90 -4.29 -0.47 2.78
Age 25-34 13.30 6.78 -0.56 20.31 1.89 -9.56 -5.62 -13.03
Age 35-44 15.73 16.41 4.48 40.76 10.96 -4.99 -3.59 1.64
Age 45-54 10.72 15.83 7.65 38.06 12.01 -0.17 2.52 14.64
Age 55-64 8.12 7.12 7.43 24.42 17.08 1.99 1.56 21.28
II. Education < 12 years 7.18 -1.79 -7.11 -2.23 5.62 -18.21 -5.42 -18.30
Age 25-34 9.37 -7.33 -7.69 -6.44 -3.50 -20.81 -4.52 -27.03
Age 35-44 8.39 0.50 -15.57 -8.03 1.99 -14.79 -9.46 -21.32
Age 45-54 6.07 2.18 -2.95 5.19 11.05 -17.59 -3.30 -11.51
Age 55-64 7.42 -2.55 0.87 5.60 10.56 -9.84 0.11 -0.21
III. Education = 12 years 7.02 1.95 -0.71 8.33 2.69 -11.18 -7.42 -15.56
Age 25-34 7.73 -2.56 -1.16 3.76 0.12 -15.02 -9.67 -23.14
Age 35-44 7.76 5.29 -4.42 8.44 5.10 -13.66 -10.19 -18.50
Age 45-54 8.20 4.28 0.63 13.54 4.04 -2.46 -7.86 -6.50
Age 55-64 6.81 -0.01 3.04 10.05 5.92 -6.40 -1.33 -2.18
IV. Education 13 to 15 years 2.53 6.71 -0.34 9.05 -4.23 -4.63 -5.18 -13.39
Age 25-34 6.72 0.51 -3.77 3.22 -2.57 -9.11 -7.28 -17.89
Age 35-44 2.40 9.70 -0.31 11.99 -2.42 -6.78 -6.77 -15.19
Age 45-54 4.08 12.25 -1.91 14.59 -1.59 -2.50 -7.74 -11.48
Age 55-64 -0.29 8.94 1.01 9.72 6.25 -5.00 -6.72 -5.85
V. Education 16 + years -2.00 16.26 5.55 20.26 -3.46 0.22 1.45 -1.84
Age 25-34 -0.18 15.52 -1.21 13.92 -7.88 3.75 -5.32 -9.51
Age 35-44 4.37 15.74 9.22 31.93 2.97 -6.97 3.53 -0.82
Age 45-54 -2.01 15.40 5.60 19.42 0.65 -5.89 -0.61 -5.86
Age 55-64 8.37 5.45 0.84 15.24 3.05 -2.66 -3.78 -3.49
Source: Author'stabulations from the March Current Population Surveys.
Notes: The sample for each year includes full-time workers between ages 25 and 64. Wages are expressed in 1990
dollars, adjusted for inflation using the Implicit Price Deflator. See the Data Appendix for additional sample
restrictions and wage definitions.
increases were during the 1980s, but
progress continued at a slower rate in
the early 1990s. Women in all age and
education groups substantially narrowed
the gap with their male counterparts;
increases were largest for women
in the younger two age groups, 25-34
and 35-44, and smallest for those over
54. Relative gains were fairly similar in
magnitude for high school graduates
and women with some college. Increases
were a bit lower among college
graduates, but they started and ended
the period with the highest gender
ratios. While relative gains were actually
a bit greater for high school dropouts
over the 1980s, progress for them
lagged over the early 1990s, and,
by 1994, they had the lowest gender ra-
tio.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 131
C. Trends in Real Wages and
Differences in Wage Gains Within
Gender Groups
The declining gender gap has also
meant that women as a group fared better
than men in terms of real wage
growth. Table 5 shows percentage
changes in real weekly wages; adjusted
using the 1990 Implicit Price Deflator.
Overall, women's real wages increased
by 31 percent between 1969 and 1994;
rising at a similar average annual rate in
the 1970s and 1980s, and at a somewhat
slower pace in the early 1990s. In contrast,
as has been widely noted, men's
real wages stagnated, rising by only 3
percent over the 25 year period. If the
Consumer Price Index had been employed
to adjust for inflation, the real
wage gains for women would have been
smaller and men's real wages would
have declined somewhat.1"
Table 5 also shows that, as has been
widely found by other researchers,
more educated men fared better than
less educated men in terms of real wage
growth in the post-1980 period and cumulatively
for the two decades (e.g.,
Juhn, Murphy, and Pierce 1993). Less
well known is that a similar pattern prevailed
among women. Real wage gains
between 1969 and 1994 were 20.3 percent
for female college graduates and 8
to 9 percent for women with high
school degrees or some college, while
real wages fell by 2.2 percent for high
school dropouts. As was the case for
men, declines in real wages among the
least educated women were especially
large for the younger two age groups,
falling by 6.4 to 8.0 percent. For each
educational group, however, the experience
of women was more favorable than
that of the corresponding male group.
The real wages of male high school
dropouts, for example, fell by 18.3 percent
between 1969 and 1994, declining
by 27.0 percent for the youngest men,
while the real wages of male college
graduates decreased by only 1.8 per-
cent.
Patterns of men's and women's real
wage growth by education differed between
the 1970s and the 1980s. These
differences were related to movements
in the return to a college education
which fell in the 1970s, for both men
and women, but has risen for both
groups since then (Lawrence Katz and
Murphy 1992). Thus, the disparity in
the experience of real wage growth
among education categories was more
pronounced in the 1980s and 1990s
than for the period as a whole. Since
1979, the real wages of female college
graduates have risen by 22.7 percent,
while the real wages of female high
school dropouts fell by 8.8 percent.
Finally, we may consider whether the
declining relative wages of the least
educated which are observed across cohorts
represent a true change in outcomes
or whether they can be entirely
explained by compositional changes. In
Table 6, I seek to shed light on this issue
by comparing across and within cohort
changes in real wages by education
group. Again, the focus is on the comparison
of the across and within cohort
changes of high school dropouts to high
school and college graduates shown in
panels A and B. For women, we see that
the within cohort changes generally exhibit
the same basic pattern of changes
for high school dropouts relative to the
others as the across cohort changes.
Moreover, the across and within cohort
changes are quite similar in magnitude
in the 1980s, the key period during
which high school dropouts lost ground.
11 Note that estimated changes in real wages will
vary depending on the wage measure used, the
data set employed, and the starting and ending
years selected. However, the findings reported
here are broadly consistent with those reported in
the literature.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
132 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 6
DIFFERENCES IN PERCENTAGE CHANGES IN REAL WAGES
BYEDUCATION ACROSS AND WITHIN COHORTS, 1969-89
Across Cohort Change Within Cohort Change
1969-79 1979-89 1969-79 1979-89
I. Women
A. Difference: Ed = 12 minus Ed < 12
Age 25-34 -1.64 4.78 -2.53 4.26
Age 35-44 -0.63 4.79 1.12 3.83
Age 45-54 2.12 2.10 1.94 2.37
B. Difference: Ed = 16+ minus Ed < 12
Age 25-34 -9.55 22.85 -0.25 27.40
Age 35-44 -4.02 15.24 -4.19 13.32
Age 45-54 -8.09 13.22 1.56 18.32
II. Men
A. Difference: Ed = 12 minus Ed < 12
Age 25-34 3.62 5.79 10.27 6.58
Age 35-44 3.12 1.13 -8.17 5.98
Age 45-54 -7.02 15.14 -3.85 5.81
B. Difference: Ed = 16+ minus Ed < 12
Age 25-34 -4.38 24.56 24.86 39.09
Age 35-44 0.98 7.82 -1.54 10.56
Age 45-54 -10.41 11.70 -3.90 12.62
Source: Author'stabulations from the March Current Population Surveys.
Notes: Illustrating for the 25-34 year age group in 1969-79, the "across"cohort change is the percentage change in
the real wage of the 25-34 year age group between 1969 and 1979; the "within"cohort change is the percentage
difference between the real wage of the 25-34 year age group in 1969 and the 35-44 year age group in 1979. See
Table 5 and the Data Appendix for additional information.
The same is true for males. As in the
case of the participation trends, there is
no evidence that the age specific trends
identified for high school dropouts are
merely due to shifts in the composition
of this group.
D. Trends in Gender Differences in
Occupations and Self-Employment
For many decades a high degree of
segregation of men and women into different
occupations appeared to be an
unchanging feature of the labor market
(Blau, Ferber, and Winkler 1998). A
substantial break with this pattern occurred
in the 1970s when occupational
segregation began to decline noticeably
(Beller 1985). This progress has continued
into the 1980s (e.g., Jacobsen 1994;
and Blau, Patricia Simpson, and Anderson
1997). Differences between women
and men in occupations across a wide
number of categories may be summarized
by a segregation index which gives
the percentage of women (or men) who
would have to change jobs for the occupational
distribution of the two groups
to be the same. Estimates based on
Census data for a comparable set of 470
detailed occupations indicate that the
index fell by 11 to 12 percent in each
decade, declining from 67.7 in 1970 to
59.3 in 1980 and 53 in 1990 (Blau,
Simpson, and Anderson 1997).
Changes in the extent of segregation
may be due to shifts in sex composition
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 133
within occupations (i.e., integration of
formerly male or female occupations) or
shifts in occupation mix (i.e., growth in
the size of integrated occupations or
decreases in the size of male or female
occupations). Changes in the sex composition
of occupations were the predominant
cause of the decrease in segregation
in both the 1970s and 1980s,
suggesting expanding opportunities for
women played a significant role, although
changing occupational mix was
of more importance in the latter period
(Blau, Simpson, and Anderson 1997). In
terms of segregation per se, female
college graduates made the fastest
progress in reducing the segregation
index over the decade and by 1990 the
index was 46 among college graduates
compared to 56-58 for less educated
women (Jacobsen, forthcoming).12
Another striking trend during our period
is a substantial increase in the selfemployment
rate of women which rose
over 60 percent from 4.1 percent in
1975 to 6.7 percent in 1990, compared
to a 20 percent increase for men to
12.4.13 Thus women have reduced the
"self-employment gap" with men, increasing
from less than one quarter to
nearly one third of self-employed workers
(Theresa Devine 1994a). Similarly, a
larger rise in female than in male selfemployment
prevailed within each of a
large number of racial and ethnic
groups included in a recent study
(Robert Fairlie and Bruce Meyer 1996).
However, while the rise in female selfemployment
was fairly broad-based,
here too the least skilled, as measured
by quartile in the distribution of potential
-wage and salary earnings, have
lagged (Devine 1994b).
An interesting question which has not
yet been addressed is the relationship
between the changes in the occupational
distribution of women and their
increased incidence of self-employment.
As women have entered traditionally
male pursuits, the opportunities for
self-employment have likely increased.
In evaluating the implications of the
growth in self-employment for women's
well-being, however, it should be noted
that, for workers of both sexes, this expansion
likely reflects, at least in part,
the increase in independent contractors
who comprise one component of the
growing "nonstandard"workforce (that
is, workers who do not have "regular"
full-time jobs).14 Nonetheless, a recent
study by Karen Lombard (1996) found
that the rising earnings potential of
women in self-employment compared to
the wage and salarysector explains most
of the upward trend in the self-employment
of married women between 1970
and 1990. This suggests that the growing
move of women into self-employment
does represent an expansion in
their opportunities.
E. Explainingthe Trendsin Wages:
The Overall Gender Gap
As we have seen, the narrowingof the
gender gap in recent years has taken
place in an environment of sharply rising
wage inequality. Wage inequality
has been increasing for women since at
least 1979, and for men since 1970. The
most widely accepted explanation for
this trend links it to a rise in the returns
to skill caused by an outward shift in
the relative demand for highly skilled
workers (e.g., Katz and Murphy 1992;
12 Some caution must be observed in such comparisons
because the 3-digit occupational categories
employed by the Census for blue collar,
manufacturing occupations distinguish an especially
large number of occupations thus making it
easier to detect segregation than for (relatively)
more aggregated white collar jobs.
13The se f-employment rate is the percentage of
nonagricultural workers who are self-employed.
14 For more information on these trends, see,
e.g., Blau, Ferber, and Winkler (1998).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
134 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 7
IMPACTOF CHANGES IN CHARACTERISTICSON CHANGES IN MEN'S AND WOMEN'S LOG REAL WAGES AND
THE GENDER GAP, 1979-88 (1983 DOLLARS)
Men'sWages Women's Wages Gender Gap
I. All Workers
Total Change -0.104 0.048 -0.152
Due to: All Characteristics -0.014 0.116 -0.124
Education Variables 0.040 0.049 -0.009
Experience Variables -0.018 0.040 -0.053
Occupation Variables 0.002 0.048 -0.046
Collective Bargaining -0.028 -0.011 -0.018
Industry Variables -0.009 -0.011 0.001
II. Low Skill Group (0-20 in 1979)
Total Change -0.268 -0.159 -0.109
Due to: All Characteristics -0.069 0.015 -0.083
Education Variables 0.030 0.019 0.011
Experience Variables -0.006 0.005 -0.012
Occupation Variables -0.009 0.032 -0.041
Collective Bargaining -0.051 -0.029 -0.022
Industry Variables -0.028 -0.016 -0.012
Source: Blau and Kahn (1997).
Notes: These results are from Blau and Kahn (1997) and include some additional calculations. Data are from the
Michigan Panel Study of Income Dynamics for 1980 and 1989. Wages are defined as annual average hourly
earnings. Individuals are divided into skill groups based on their predicted wages using estimated coefficients from
an overall male OLS regression estimated pooling 1979 and 1988 wage data which controls for education (years of
schooling and dummy variables for college and advanced degrees), actual experience, actual experience squared,
and year. In 1979, allocation to skill groups is determined by the own sex percentile ranking of the individual's
predicted wage as follows: low skid group (0-20); middle skill group (20-80); high skill group (80-100). Individuals
are allocated to skill groups in 1989 based on the 1979 cut-offs for predicted wages. Thus the size of each skill group
can change. For further details, see Blau and Kahn (1997).
Juhn, Murphy, and Pierce 1993). This
demand shift has in turn been related to
such economy-wide forces as technological
change and the impact of international
trade. Institutional factors
like the decline in unionism and the
falling real value of the minimum wage
may have also played a role. What is the
relationship between these developments
and trends in the gender wage
gap?
An insightful recent paper by Juhn,
Murphy, and Pierce (1991) suggests
how these aggregate trends may influence
the wage outcomes of particular
demographic groups. Juhn, Murphy,
and Pierce (1991) were concerned with
explaining the slowing of convergence
black and white wages among males
over the late 1970s and the 1980s. (A
similar slowing of race convergence occurred
among women; see, e.g., Blau,
Ferber, and Winkler 1998). Their approach
not only links the outcomes of
particular groups to the widening wage
inequality which has been a dominant
labor market trend of this period, but
also introduces a new factor into the
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 135
TABLE 7 (Cont.)
Men'sWages Women's Wages Gender Gap
III. Middle SkillGroup (20-80 in 1979)
Total Change -0.144 -0.037 -0.107
Due to: All Characteristics -0.067 0.003 -0.070
Education Variables 0.020 0.011 0.009
Experience Variables -0.021 0.002 -0.024
Occupation Variables -0.006 0.025 -0.031
Collective Bargaining -0.046 -0.019 -0.027
Industry Variables -0.011 -0.012 0.001
IV. High Skill Group (80-100 in 1979)
Total Change -0.038 0.080 -0.118
Due to: All Characteristics -0.033 0.065 -0.098
Education Variables 0.008 0.026 -0.018
Experience Variables -0.040 0.001 -0.040
Occupation Variables -0.025 0.035 -0.060
Collective Bargaining 0.001 0.000 0.001
Industry Variables 0.023 0.004 0.019
The entries in the columns headed "men'swages" and "women'swages" show the contribution of changes in the
means of the indicated category of variables to the 1979-88 change in the mean of real log wages for males and
females respectively. This is calculated for all workers and separately by skill groups. The contribution of the
indicated variables is:
A~~~~~~~~~~~~~~~ (XI
_
xo )01?11
where A3, is a vector of coefficients from an OLS regression estimated for males overall or in the indicated skill
group in 1988, and X1 is a vector of own means of the variables (i.e., for men or women) either overall or in the
indicated skill group in year t (1 = 1988 and 0 = 1979). The results are based on the "fullspecification" reported in
Blau and Kahn (1977) and include controls for education and experience (as described above), as well as dummy
variables for major industry and occupation, and unionism. The difference in the entries for men and women is
equal to the contribution of changes in the indicated characteristics to the change in the gender gap.
analysis of demographic wage differentials.
Traditional analyses focus on what
might be termed "group-specific" factors,
i.e., the group's relative skills and
the extent of labor market discrimination
against them. However, outcomes
for particular groups are also affected
by "wage structure" in general. I would
define wage structure to include overall
skill prices and rents received for employment
in favored sectors, although
Juhn, Murphy, and Pierce (1991) emphasize
the former. Because blacks are
on average less skilled than whites,
Juhn, Murphy, and Pierce show that
they were especially adversely affected
by recent trends in skill prices and that
this helps to explain the slowing of convergence
in the race gap.
The reasoning which Juhn, Murphy,
and Pierce (1991) apply to understanding
the trends in the race gap is
applicable to understanding trends in
gender wage differentials. Thus, while
there have been a number of useful
studies of the sources of the recent
narrowing of the gender gap, including
O'Neill and Polacheck (1993) and
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
136 Journal of Economic Literature, Vol. XXXVI(March 1998)
Wellington (1993), I focus on one by
Blau and Kahn (1997) which incorporates
these insights. As Blau and Kahn
point out, analyzing women's progress
in Juhn, Murphy, and Pierce's framework
raises something of a paradox. As
the prices of measured skills and rewards
for employment in high-paying
sectors have risen, women, who continue
to have less experience, on average,
and to be located in lower-paying
occupations and industries, should have
been increasingly disadvantaged. Yet
the gender wage gap has declined sub-
stantially.
Blau and Kahn (1997) investigate this
issue using data from the Michigan
Panel Study of Income Dynamics.15
They find that rising inequality and
higher rewards to skills did indeed retard
women's progress during the
1980s, "reclaiming" about one-third to
two-fifths of women's potential gains in
relative wages. The substantial decline
in the male-female pay gap which nonetheless
occurred is traced to "genderspecific"
factors which were more than
sufficient to counterbalance changes in
both measured and unmeasured prices
which worked against women. Specifically,
improvements in women's relative
qualifications and the "unexplained"
portion of the pay gap declined substan-
tially.
A decline in the unexplained gap is
generally viewed as reflecting either an
upgrading of women's unmeasured labor
market skills or a decline in labor
market discrimination against them.
This ambiguity is due to the fact that
empirical evidence for discrimination
relies on the existence of a residual gender
pay gap which cannot be explained
by gender differences in measured
qualifications. This accords well with
the economic 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. 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,
industry) themselves reflect the
impact of discrimination, then discrimination
will be underestimated. Blau and
Kahn report that, controlling for human
capital characteristics, including education
and actual labor market experience,
women earned 71.5 percent as
much as men in 1979 and 80.5 percent
in 1988. When the authors further control
for major occupation and industry
and for unionism, the ratio rose from
77.6 percent in 1979 to 88.2 percent in
1988.
These results are shown in more detail
in Table 7. Column (3) is drawn directly
from the Blau-Kahn study and
gives the contribution of changes in the
characteristics of men and women workers
to changes in the gender wage gap
over the 1980s. However, these findings
refer to relative changes and thus leave
open the question of whether women's
characteristics improved in some absolute
sense. To shed light on this question,
the first two columns of Table 7
provide new calculations based on the
Blau-Kahn results. They give the contribution
of changes in characteristics to
each sex group's real wage changes over
the period, both overall and separately
by skill group. The weights are 1988 regression
coefficients (including the constant
term) from male wage equations
estimated either for all workers or for
workers in the indicated skill group.
Differences in these changes for men
and women separately equal the final
15 In a companion paper, Blau and Kahn (1996)
present evidence that wage structure is very important
in explaining international differences in
the gender pay gap.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 137
column which gives the net result of the
changes in characteristics on the change
in the gender gap. (See the notes to Table
7 and Blau and Kahn (1997) for additional
details.)
Results in column (3) indicate that
improvements in women's relative experience16
and broad occupational category
were particularly important in narrowing
the gender gap; changes in
unionism also benefited women relative
to men, but played a smaller a role,
while shifts in industrial distribution
had relatively little effect. Looking at
columns (1) and (2), we see that women
did upgrade their major occupations absolutely,
that is they moved into higher
paying categories, evaluated at the male
returns, while changes in men's occupational
distribution had no effect on
men's real wage changes. Both men and
women lost ground in terms of collective
bargaining coverage, with larger
declines for men than for women.
Changes in industry distribution also
lowered the real wage growth of both
men and women, but by approximately
the same amount for each.
The role of occupational upgrading in
narrowing the gender gap raises the
question of the reason for the decline in
occupational segregation which occurred
over the 1970s and 1980s. Here
again, as in our consideration of the
educational improvements of women,
both the human capital and the discrimination
models potentially provide
viable explanations.17 On the one hand,
it may be that as women anticipated remaining
in the labor force for longer
periods it became profitable for them to
invest in the higher amount of on-thejob
training required in traditionally
male occupations. On the other hand,
women may have entered these areas in
response to declining barriers to their
participation. Unfortunately, there is no
research explicitly examining this, but it
is quite likely that both sets of factors
played a role.
There is also some question as to how
we should interpret the decline in the
unexplained gender gap over this period;
as noted above, it is unclear
whether this is due to an upgrading of
women's unmeasured labor market
skills or a decline in labor market discrimination
against them. Because
women improved their relative level of
measured characteristics, it is plausible
that they also enhanced their relative
level of unmeasured characteristics. For
example, it is possible that with increasing
labor force commitment, the quality
of women's labor market experience
may have also improved. A recent study
by Anne Royalty (1996), the first to explicitly
examine the importance of predicted
turnover probability in explaining
the gender training difference,
sheds some light on this possibility.
Consistent with this expectation, she
finds that predicted turnover probability
does explain some of the gender
training difference. 18 However, interestingly,
a major portion remains unexplained
even after this and other determinants
of training are taken into
account. This suggests that women's rising
labor force commitment likely did
reduce the gender training gap, though
even if the gender difference in com-
16 Similar results are reported by O'Neill and
Polachek (1993) and Wellington (1993).
17 England (1982) provides the strongest critique
of the human capital explanation for occupational
segregation. Some particularly interesting
recent evidence implicitly supporting the human
capital model is Macpherson and Hirsch's (1995)
finding of a substantial effect of skills in explaining
the lower pay in predominantly female jobs.
Theirs are among the higher estimates; for a review
see Sorensen (1990).
18 Considerable empirical evidence indicates
that women have traditionally received less onthe-job
training than men; see John Barron, Dan
Black, and Mark Loewenstein (1993) and the references
therein.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
138 Journal of Economic Literature, Vol. XXXVI(March 1998)
mitment had been eliminated some
training difference would remain.
There is also evidence that the marketability
of women's education improved.
As noted above, gender differences
in fields of study among college
students decreased over the 1970s and
1980s. This development, which is not
captured in the Blau and Kahn study
and other similar work, likely contributed
to the narrowingof the gender gap
because gender differences in college
major have been found to be strongly
related to the gender wage gap among
college graduates (Charles Brown and
Mary Corcoran 1997). The male-female
difference in SAT math scores has also
been declining, falling from 46 points in
1977 to 35 points in 1996 (College
Board 1996).19
On its face, the implication that discrimination
against women declined
seems less credible than that their unmeasured
characteristics improved, because
it is well known that the federal
government scaled back its anti-discrimination
enforcement effort during
this period (Jonathan Leonard 1989).
However, there are a number of ways in
which decreases in discrimination still
could have played a role in reducing the
pay gap in the 1980s. For example, it is
possible that the female gains reflected
delayed dividends to earlier anti-discrimination
efforts which encouraged
women to enhance their qualifications
and enter traditionally male fields.
Thus, decreases in discrimination may
underlie some of the progress women
made due to the enhancement of their
measured qualifications relative to
men's. And, to the extent that qualifications
are not fully controlled for in the
wage regression used to decompose the
change in the gender wage gap, this
may also help to explain the decline in
the "unexplained" gap. Further, it may
be that as women increased their commitment
to the labor force and their
other job skills, the rationale for statistical
discrimination against them diminished.
For example, controlling for
measured characteristics, employers
may discount women's wages due to
their higher expected turnover rate. A
reduction in the perceived turnover differential
between men and women
would result in a decrease in this discount
and an increase in women's relative
wages controlling for other factors
(Dennis Aigner and Glen Cain 1977).
Moreover, in the presence of feedback
effects, employers' revised views can
generate further improvements in
women's earnings by increasing returns
to their investments in job qualifications
and skills.20 To the extent that
some of these qualifications are unobserved
in our analysis, the result will be
a decrease in the unexplained gap. A final
scenario might be that discrimination
declined due to changes in social
attitudes which make such discriminatory
tastes increasingly unpalatable.
Thus, an important role for decreases in
discrimination in narrowing the gap
may not be as implausible as it first ap-
pears.
Two more points may be noted about
the finding that a decrease in the unexplained
gap played a significant role in
the narrowing of the gender gap. First,
these findings do not, of course, imply
that labor market discrimination has
disappeared entirely. As reported
above, ceteris paribus pay gaps between
19SAT scores were recentered by the College
Board in 1996; scores for the earlier year have
been converted to the recentered scale by the College
Board.
20 Lundberg and Richard Startz (1983) develop
an interesting model of statistical discrimination
based on women's productivity being less accurately
predicted than men's that includes feedback
effects. See also, Blau, Ferber, and Winkler
(1998).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 139
men and women, which are often taken
as an estimate of discrimination, continue
to exist, although their magnitude
has been diminished. Second, the finding
that a substantial portion of the
closing of the gap is essentially "unexplained"
parallels the findings in each
of the major areas examined. This may
be an indicator in this case of a decrease
in discrimination. It is my own
view that is the case at least to some
extent. However, it may also be, in this
area as in others, in part a reflection of
the limitations of our models and the
data sets at our disposal for estimating
them.
F. Explaining the Trends in Wages:
Differences by Skill Group
Blau and Kahn (1997) also investigate
women's progress separately by skill
group. Returning to Table 7, we see
that the pay gap declined at about the
same pace across all three skill groups
defined in terms of experience and education.
This is consistent with tabulations
from the CPS presented above
which showed similar gains for women
across education groups prior to the
early 1990s. However, the sources of
the closing differed. Consistent with a
greater negative effect of restructuring
on male blue collar and manufacturing
jobs, the results in column (3) of Table
7 indicate that industry and union representation
effects strongly favored
women relative to men at the bottom
and middle of the skill distribution; but
worked to increase the gender gap
slightly among high skill workers. High
skill women nonetheless advanced at a
similar pace as the other groups due to
the larger improvement in their human
capital characteristics and occupational
distribution, although relative occupational
gains were also substantial for
low and middle skill women.
In terms of absolute progress, we see
in Table 7 that women in each skill
group also improved their occupational
distribution absolutely over the period,
while men's distribution declined
slightly for low and middle skilled men
and more substantially for high skill
men. With respect to industrial distribution
and unionism, the experience of
low and middle skilled men and women
was less favorable than men and women
in the high skill group. Putting this
somewhat differently, -the decline of the
gender wage gap among low and middle
skilled workers was partly due to men
losing ground at a faster pace than
women. Taken together, changes in industry
distribution and unionism
worked to lower the real wages of low
skill men by about 8 percent and of low
skill women by about 5 percent. In the
middle skill group, these factors lowered
real wages by 6 percent for men
and 3 percent for women. In contrast,
in the high skill group, changes in industry
and unionism (principally the
former) worked to raise men's wages by
2 percent and left women's wages
roughly unchanged.
IV. Standard of Living: Differences
Across and Within Families
In this section I review trends in a
number of indicators of economic wellbeing
related to the distribution of resources
both between and within families.
I first consider trends in marital
status and family formation among
women, chiefly focusing on single headship.
I concentrate on headship because
it is well documented that families
headed by women are more likely to be
poor or to have low incomes. There are
also serious concerns about negative
consequences for children of living in
female headed families, due in part to
this economic deprivation (e.g., Sara
McLanahan and Karen Booth 1989).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
140 Journal of Economic Literature, Vol. XXXVI(March 1998)
Trends in well-being are influenced not
only by trends in family structure but
also by trends in family income both
across family types and by differences
in the trends within each type for particular
education groups. Thus, I also
explicitly consider these trends below.
In comparing income across family
types we confront a conceptual issue
raised by Fuchs (1988) which has generally
not been addressed in the income
distribution literature. Because fulltime
homemakers produce goods and
services of value to their families, comparisons
based only on money income
are likely to overstate the difference in
well-being between families in which
the wife works outside the home and
those in which she does not. In like
manner, when labor force participation
rates of married women are rising, the
increase in the well-being of married
couples compared to single-headed
families is likely to be overstated when
only money contributions to the economic
well-being of the family are
counted. However, Fuchs' (1988) careful
adjustment for this problem did not
alter his conclusion that the rise in the
share of female-headed families adversely
affected the economic well-being
of women relative to men.
Another issue to be considered which
Fuchs (1988) also brought to the fore is
the sharing rule within the family. If
the distribution of resources between
husbands and wives within families is
very unequal, we may overstate the advantage
in well-being for women of being
part of a higher income married
couple family over a lower income single-headed
family. While this is certainly
a reasonable qualification, the
differences in income between the two
family types are so large that it seems
justifiable to assume in general that
women in married couple families do
have more resources. Further, bargaining
models (discussed above) suggest
that higher relative wages of wives
should increase their bargaining power
within marriage. Thus, trends in wages
of wives relative to those of husbands
are reviewed as an indicator of bargaining
power.2' Relative wages of wives
are found to have been increasing overall
and within education groups. This
implies that, if anything, women's share
of resources within marriage should be
increasing. Similarly, while domestic
violence is a risk which married (or cohabiting)
women may face to a greater
extent than single women, the trends in
domestic violence which are also considered
below suggest that there was no
increase in the incidence of such violence
from the mid-1970s to the mid-
1980s.
Finally, trends in the gender division
of housework are considered. This is an
important indicator of well-being for a
number of reasons. First, as Fuchs
(1986; 1988) has pointed out, access to
leisure is important. Within families,
women have traditionally had the major
responsibility for housework, while men
have had the major responsibility for
market work. As market work has become
more equally shared between men
and women in the family, if the division
of housework remains very unequal, the
consequence is likely to be a reduction
in leisure for women. An additional reason
for concern over the division of
housework is because women's greater
responsibility for housework may adversely
affect their labor market outcomes
in at least two ways. First, it may
cause women to constrict their hours of
work, work schedules, and commuting
time and hence reduce their wages or
occupational choice. Second, even controlling
for hours of market work, it
21 This is preferable to looking at earnings, because
earnings are influenced by labor supply decisions
which could well change after a divorce.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 141
may result in their reducing their effective
effort per hour compared to men
who spend' fewer off-job hours on
homemaking tasks (Becker 1985). And,
indeed, empirical evidence suggests
that women who spend more time on
housework do have lower wages, all else
equal (e.g., Joni Hersch and Leslie
Stratton 1997). Of course, the causation
could run in the opposite direction:
women with lower wages face a lower
opportunity cost of time spent in housework.
However, Hersch and Stratton
(1997) still find a significant negative effect
of housework on wages when instrumental
variables estimation is used.22
A. Trends in Marital Status and
Family Formation
In this section I review trends in
marital and family status for women. As
noted above, "married" refers to women
who are married spouse present. Single
heads are women who head families or
subfamilies and are not married spouse
present; this excludes single person
households. One problem with these
definitions is that a significant and
growing proportion of individuals who
are officially defined as single heads are
actually members of cohabiting couples.
For example, in 1990, 7 percent of single
mothers were cohabiting (McLanahan
and Casper 1995). While this is certainly
an important caution, it should
also be noted that the extent of income
pooling in cohabiting families is uncertain.
Moreover, cohabitations tend to
end very quickly either in marriage or a
breakup (Larry Bumpass and James
Sweet 1989) 23
Table 8 shows the trends in marital
and family status for women and illustrates
a number of well-known demographic
shifts. Overall, the proportion
of women who are married spouse present
has declined, while the proportion
of those who are single heads has increased.
So, for example, the percentage
of women who were married and
living with their husbands fell from 78
to 65 percent between 1970 and 1995.
At the same time, the incidence of single
headship rose from 9 to 16 percent;
and the proportion of women raising
children on their own doubled from 6 to
12 percent. The fastest growth in female
headship over this period occurred
during the 1970s. A rise in headship occurred
in all age groups with the exception
of the oldest, and was particularly
pronounced for younger women.
While these increases in single headship
among women have received a
great deal of attention, the pronounced
differences across education groups in
these trends have been less often noted.
The contrast between high school dropouts
and college graduates is particularly
striking. In 1970, about threequarters
of both groups were married,
less than either high school graduates
or those with some college. But, by
1995, the proportion of women who
were married had fallen to 56 percent
among high school dropouts compared
to 69 percent among college graduates.
Of greater significance for economic
well-being, this represented principally
a postponement of marriage and family
for more educated women but an increasing
tendency toward single headship
among less educated women. Although
women with less than a high
school education were more likely to be
single heads than other groups in 1970,
the differences were fairly moderate in
absolute terms. For example, the proportion
of single heads was 12 percent
22 The Hersch and Stratton study finds these
conclusions to be quite robust to alternative specifications
of the instrument set. They do not, however,
find consistent evidence of a significant
negative effect of housework on men's wages.
23 For example, 40 percent of cohabitations do
not continue (as cohabitations) beyond one year,
and two-thirds do not continue beyond two years.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
142 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 8
MARITALAND FAMILYSTATUSOF WOMEN
Marital Status Single Head Child/Children Present
Married, Married,
Spouse Ever Never Never Spouse Single
Year Present Married Married All Married All Present Head
All Women
1970 0.777 0.161 0.062 0.094 0.008 0.545 0.484 0.061
1980 0.715 0.200 0.085 0.132 0.016 0.506 0.410 0.096
1990 0.666 0.211 0.123 0.148 0.032 0.462 0.356 0.106
1995 0.652 0.214 0.135 0.159 0.040 0.467 0.350 0.116
Age 25-34
1970 0.819 0.096 0.086 0.084 0.009 0.812 0.730 0.081
1980 0.687 0.159 0.154 0.141 0.029 0.686 0.551 0.135
1990 0.618 0.145 0.236 0.165 0.063 0.632 0.475 0.157
1995 0.589 0.144 0.266 0.183 0.081 0.620 0.446 0.174
Education < 12 Years
1970 0.736 0.218 0.046 0.121 0.009 0.474 0.397 0.077
1980 0.656 0.281 0.063 0.182 0.023 0.445 0.319 0.126
1990 0.596 0.287 0.117 0.217 0.056 0.404 0.258 0.146
1995 0.561 0.293 0.146 0.243 0.076 0.436 0.270 0.167
Education = 12 Years
1970 0.816 0.129 0.055 0.080 0.006 0.604 0.549 0.055
1980 0.754 0.180 0.065 0.125 0.015 0.534 0.441 0.092
1990 0.691 0.214 0.094 0.154 0.031 0.477 0.364 0.113
1995 0.669 0.219 0.113 0.166 0.043 0.459 0.337 0.122
Education 13-15 Years
1970 0.790 0.137 0.073 0.078 0.006 0.589 0.536 0.054
1980 0.711 0.193 0.095 0.127 0.015 0.551 0.449 0.102
1990 0.664 0.210 0.126 0.147 0.030 0.502 0.392 0.110
1995 0.647 0.233 0.121 0.172 0.037 0.501 0.368 0.133
Education 16 + Years
1970 0.754 0.103 0.144 0.060 0.009 0.523 0.489 0.034
1980 0.704 0.130 0.165 0.078 0.010 0.482 0.430 0.052
1990 0.671 0.144 0.185 0.082 0.016 0.441 0.387 0.054
1995 0.685 0.136 0.179 0.082 0.017 0.454 0.398 0.057
Source: Author'stabulations from the March Current Population Surveys.
Notes: Sample for each year includes all adult civilian women between ages 25 and 64.
among high school dropouts compared
to 6 percent among college graduates.
By 1995, the incidence of headship had
doubled to 24 percent among high
school dropouts compared to a 33 percent
increase to 8 percent among college
graduates. As many as 38 percent
of female high school dropouts with
children were raising them on their own
compared to only 13 percent of college
graduates.
As has been widely noted in the literature,
the rise in female headship has
been particularly large among blacks.
As Table 9 indicates, by 1995, 39 percent
of black women were single heads
and the race difference in the incidence
of single female headship rose from 17
percentage points in 1970 to 26 points
in 1995. A considerably higher incidence
of single headship prevailed
among blacks within all age and educaThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 143
TABLE 9
INCIDENCE OF SINGLE HEADSHIP BY RACE, AGE, AND EDUCATION
1970 1980 1990 1995 Change 1970-95
A. White Women 0.077 0.105 0.116 0.124 0.047
Age 25-34 0.063 0.106 0.121 0.140 0.078
Education < 12 years 0.091 0.199 0.223 0.253 0.162
Education= 12years 0.060 0.115 0.145 0.178 0.118
Education > 12 years 0.042 0.067 0.075 0.097 0.056
Age 35-44 0.078 0.126 0.136 0.148 0.070
Education < 12 years 0.097 0.169 0.196 0.235 0.138
Education = 12 years 0.070 0.115 0.149 0.153 0.083
Education> 12years 0.065 0.115 0.112 0.130 0.065
B. BlackWomen 0.244 0.347 0.381 0.388 0.144
Age25-34 0.260 0.398 0.445 0.440 0.181
Education< 12years 0.335 0.517 0.646 0.651 0.316
Education = 12 years 0.236 0.418 0.464 0.452 0.216
Education > 12 years 0.114 0.277 0.337 0.358 0.243
Age 35-44 0.274 0.396 0.413 0.416 0.142
Education < 12 years 0.319 0.456 0.484 0.536 0.217
Education = 12 years 0.227 0.369 0.432 0.421 0.194
Education > 12 years 0.184 0.348 0.351 0.375 0.192
Source: Author'stabulations from March Current Population Surveys.
Notes: Sample for each year includes all adult civilian women between ages 25 and 64.
tion groups. However, it is also true for
both blacks and whites that, within each
age category, high school dropouts had
the highest rates of single headship. By
1995, among women aged 25-34 with
less than 12 years of education, 65 percent
of black women and 25 percent of
white women were single family heads.
A final important aspect of the demographic
shifts related to headship is the
rise in the incidence of never married
women among female heads. This is of
concern because this group tends to be
the least well-off economically. As may
be seen in Table 8, never married single
heads account for a rising proportion of
women overall and within each education
group. However, the share of never
married heads has increased especially
rapidly among blacks and among
younger, less educated women in both
race groups. By 1995, among younger
women (aged 25-34) with less than 12
years of education, 44 percent of single
white heads and 72 percent of single
black heads were never married, up
from 7 percent of whites and 19 percent
of blacks in 1970.24
Across and within cohort changes in
headship are presented in Table 10. I
focus on the two younger age groups
(25-34 and 35-44) when across and (especially)
within cohort changes in headship
are still likely to occur. Comparisons
of the across and within cohort
rates are not as informative here as in
the case of the other variables considered.
Less educated women have children
earlier and increasingly form single-headed
families through out of
wedlock births, thus reducing the scope
for within cohort increases in headship
over time. In contrast, more educated
women are much more likely to form
24 Based on tabulations from the CPS.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
144 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 10
ACROSS AND WITHIN COHORT CHANGES IN SINGLE HEADSHIP OF WOMEN
BYAGE AND EDUCATION, 1970-90
Across Cohort Change Within Cohort Change
1970-80 1980-90 1970-80 1980-90
A. Education < 12 years
Age 25-34 12.39 3.98 8.87 -0.73
Age 35-44 8.84 2.79 4.05 -1.62
B. Education = 12 years
Age 25-34 7.48 3.95 6.50 3.31
Age 35-44 6.06 4.29 2.24 -2.63
C. Education 13 to 15 years
Age 25-34 6.09 3.14 10.88 5.04
Age 35-44 8.39 0.25 2.76 -5.21
D. Education 16 + years
Age 25-34 2.43 0.31 7.02 4.28
Age 35-44 4.31 -0.32 6.67 2.46
E. Difference: Ed = 12 minus Ed < 12 (B - A)
Age 25-34 -4.91 -0.04 -2.37 4.04
Age 35-44 -2.78 1.49 -1.80 -1.01
F. Difference: Ed = 16+ minus Ed < 12 (D - A)
Age 25-34 -9.96 -3.67 -1.84 5.00
Age 35-44 -4.52 -3.11 2.62 4.08
Notes: Illustrating for the 1970-80 change in single headship of the 25-34 year age group, the "across"cohort
change is the difference between the single headship rate of the 25-34 year age group in 1970 and 1980; the
"within"cohort change is the difference between the single headship rate of the 25-34 year age group in 1970 and
the 35-44 year age group in 1980.
single-headed families through divorce
or separation, and, especially in the
case of college graduates, tend to have
children considerably later. These factors
tend to raise within cohort increases
in headship of more educated
women relative both to their own across
cohort increases and to the within cohort
increases of less educated women.
Taking these factors into account, the
results in Table 10 are fairly supportive
of the view that true changes in behavior
and not just compositional shifts underlie
at least some of the observed
growth in headship of the least educated
women. Focusing on the 1970s,
when the largest across cohort increases
in headship occurred for all groups, we
find sizable within cohort increases in
headship among high school dropouts,
although in each case they are smaller
than the corresponding across cohort
increases. Moreover, with one exception,
the within cohort increases of high
school dropouts exceed those of high
school and college graduates (panels E
and F), although by considerably less
than is the case for the across cohort
differences. In contrast, in the 1980s, as
the rate of increase in headship slowed
and as more less educated women entered
single headship through out-ofThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 145
wedlock births, within cohort changes
in headship of less educated women became
smaller than their across cohort
changes (in fact they are negative) and
tended to be smaller than those of more
highly educated women as well. Nonetheless,
at least for the 1970s, these results
suggest that the rise in headship of
less educated women was due at least in
part to behavioral changes.
B. Explaining the Trends in Family
Formation
A variety of explanations have been
offered for the movement of women
away from marriage and toward female
headship. One explanation links these
developments to increasing labor market
opportunities for women which increase
their relative wages and thereby
reduce the gains to marriage (Becker
1991). A second explanation focuses on
the deteriorating labor market situation
of less skilled men and the resulting
scarcity of "marriageable" males (William
Wilson 1987). A final explanation
points to welfare availability as important
in encouraging the formation of female
headed families (Charles Murray
1984; Becker 1991). Changes in social
attitudes, including the greater social
acceptance of divorce and the declining
social stigma of unwed motherhood
have also been cited as supplementary
to the economic explanations (Rebecca
Blank 1995).
Somie conclusions may be reached
concerning the relative importance of
each of each of these factors based on
what has been learned regarding the
fundamental causes of headship at a
point in time as well as from a smaller
number of studies that explicitly examine
trends. On the basis of this research,
welfare appears the least plausible
explanation for a number of reasons.
First, Moffitt's (1992) review of the literature
suggests that welfare does not
have strong incentive effects on demographic
decisions. Second, the time pattern
of changes in welfare benefits does
not match the demographic trends. The
real value of monthly support levels
available from AFDC and food stamps
combined have been falling steadily
since the 1960s (Blank 1995). It is true,
as we have seen, that labor market options
of low-skilled men (and women)
have also been declining, possibly making
welfare relatively more attractive.
However, to the extent this is the case,
it is the labor market trends that are the
causal factor. Finally, as Blank (1995)
points out, international comparisons
are very persuasive. While the U.S. provides
much lower levels of government
support for single mothers than other
industrialized countries, it has one of
the highest rates of single motherhood
and the highest rate of teen pregnancy.
Greater empirical support has been
obtained for the first two explanations:
rising relative wages of women (e.g., T.
Paul Schultz 1994) and the declining
economic prospects of low skilled men
(e.g., William Darity and Samuel Meyers
1995).25 Blank (1995) suggests that
these two explanations may work together,
with higher relative wages and
labor market opportunities for women
explaining the general rise in single parenthood
across all income and skill
groups and declining labor market opportunities
of low skilled men explaining
the higher increases among the low
skilled women they would generally
marry. An additional factor that can be
noted based on the labor market trends
reviewed above is declining labor market
opportunities for low skill women.
This has been true over the 1980s in
both a relative sense, i.e., compared to
other women, and an absolute sense,
25 In the black community, relatively high homicide
and incarceration rates are also believed to
contribute to a scarcity of marriageable males.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
146 Journal of Economic Literature, Vol. XXXVI(March 1998)
i.e., in terms of declining real wages.
And the declines have been most pronounced
for younger women. A substantial
and growing proportion of single female
heads, especially among low
skilled women, is comprised of unwed
mothers, many of whom began their
childbearing as teenagers. The declining
market opportunities confronting
this group imply that their opportunity
cost of having children at an early age
has been falling relative to others.26
While arguments structured in terms
of the wage trends of men and women
appear more promising than the welfare
explanation, a review of the aggregate
trends suggests that they too are unlikely
to provide a full explanation. The
largest growth in female headship during
our period occurred over the 1970s,
yet that was a time when real wage
trends for those with a high school education
or less were actually more favorable
than the trends for college graduates
among both men and women. This
suggests that the 1970s trends were not
driven by a declining desirability of
male high school graduates or dropouts
as husbands in economic terms, or by a
falling opportunity cost of teen pregnancy.
Moreover, research findings suggest
that the decline in the supply of
"marriageable"men is not the prime explanation
for the sharp decrease in
black marriage rates over the 1970s and
1980s. A particularly comprehensive
study by Robert Wood (1995) using
SMSA level data for 1970 and 1980
finds that the decline in high earning
young black men explains only 3 or 4
percent of the decline in black marriage
rates during the 1970s. Moreover,
Wood's review of previous research suggests
that while the availability of marriageable
men, defined in terms of employment
or earnings, is often found to
be statistically significant, its quantitative
effect is small.
Rising relative wages of women also
appear unlikely to be the principal
cause of the trends, especially the large
rise in headship in the 1970s. While
there was some increase in the relative
wages of women among younger workers
in the 1970s, the largest increase
even for this group as well as for other
age groups occurred in the 1980s. Even
the point estimate of this effect is in
some doubt. A recent event history
analysis of first marriages begun between
1967 and 1983 using data from
the Panel Study of Income Dynamics by
Hoffman and Greg Duncan (1995) finds
no evidence that higher wages of wives
increase marital instability. Based on
simulations using their estimated coefficients,
they find that female wages,
male incomes, and AFDC benefit levels
did not play a large role in explaining
the trends in divorce rates over recent
decades.
A review of the trends presented
above suggests that a closer correlation
would be obtained between rising
women's labor force participation per se
and the decline in marriage than between
the relative wages of women and
the 1970s trends. This difference can
occur both because absolute increases
in women's wages can raise their participation
even if their relative wages
remain fairly constant-this appears to
have occurred in the 1970s-and because,
as we have seen, much of the increase
in women's labor force participation
cannot readily be explained by
measured economic and demographic
variables. As women and men become
26 Recent work suggests that the negative effects
of teen childbearing may have been overstated in
earlier studies due to overestimates of the alternative
prospects for these young women (Arline
Geronimus and Sanders Korenman 1992). At present
there exists a considerable range of estimates
of the magnitude of this adverse effect (Saul Hoffman,
Michael Foster, and Frank Furstenberg
1993).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 147
less specialized in homework and market
work, the scope for gains to specialization
is reduced. On the other hand, in
reconciling an explanation based on
participation with the aggregate trends,
we now face an embarrassment of
riches: the percentage point increases
in female participation were similar in
the 1970s and 1980s, but headship rose
at a much faster pace in the 1970s.27
Thus, wage explanations framed in
terms of changes either in levels of
male wages or in women's wages relative
to men's are unlikely to fully account
for the particularly rapid increases
in headship which occurred
over the 1970s, though they might provide
a more viable explanation for the
post-1980 trends which have not yet
been investigated. Considerably more
work is needed to fully understand the
importance of economic factors in explaining
the trends. The role of increases
in women's labor force participation,
as distinct from trends in the
wages of women relative to men, might
be a fruitful area for future study. In
addition, there is some evidence, particularly
among blacks, that greater
economic opportunity may reduce
out-of-wedlock births to young women
by raising its opportunity cost (Randall
Olsen and Farkas 1990). This suggests
that an investigation of the declining
prospects of less skilled women might
also be a promising route to explore
in explaining the post-1980 trends.
Given the apparent inadequacies of each
of the proposed economic explanations,
a principal role in explaining the
changes in family formation would at
this point have to be assigned to
changes in behavioral responses and
shifts in social attitudes. While it is always
difficult to know whether changes
in behavior follow changes in attitudes
or cause them, it is likely that, whatever
the initial cause, changes in both areas
reinforce each other through feedback
effects.
C. Implications of Changes in Family
Formation, Labor Force
Participation, and Wage Trends for
Income Differences Across Families
As we have seen, the rise in female
headship has been heavily concentrated
among less-educated women. At the
same time, the labor force participation
and wages of this group have also declined
relative to other women. Thus,
it is not surprising that the relative
incomes of individuals in femaleheaded
families have fallen. Equivalence
incomes for individuals are based
on family income after adjusting for the
number of family members and economies
of scale. In 1989, the mean equivalence
income of individuals in married
couple families was over double the income
in female-headed families, up
from 70 percent higher in 1969 (U.S.
Department of Labor 1995, p. 65). The
consequences of being in a female
headed family are also more negative
for less educated women, and are becoming
increasingly so. Fifty-nine percent
of individuals in families headed
by a single woman with a high school
education or less were in the bottom
quintile compared to 30 percent when
the head had more than a high school
education. And, between 1969 and
1989, the equivalence income of individuals
in families headed by single,
less-educated women fell from 60 to 57
percent of individuals in families
headed by women with more than a
high school education (U.S. Department
of Labor 1995).
There has also been a widening income
gap between individuals in families
headed by less educated couples
27The percentage increase in participation did
decline a bit, however, from 21.8 percent over the
1970s to 15.6 percent over the 1980s.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
148 Journal of Economic Literature, Vol. XXXVI(March 1998)
compared to more highly educated couples.
For example, in 1969, the average
equivalence income of individuals in
families headed by married couples who
had a high school education or less, was
59 percent of that of individuals in
families where both spouses had more
than a high school education; by 1989
the ratio was 52 percent (U.S. Department
of Labor 1995). This trend reflects
a larger rise in labor force participation
of wives of high earning husbands
and also an increase in the correlation
between the earnings of husbands and
wives when both are employed (U.S.
Department of Labor 1995; Karoly and
Burtless 1995; Cancian, Danziger, and
Gottschalk 1993). The consequences of
these developments for income inequality
across married couple families is unclear
a priori. Increases in participation
of wives lower wage dispersion across
all persons (or families) because there
are fewer families with zero earnings of
the wife. On the other hand, the increasing
tendency of participation to be
concentrated among high income families
and the increase in the correlation
of husbands' and wives' earnings works
to increase inequality (Karoly and
Burtless 1995). Cancian, Danziger, and
Gottschalk (1993) found that, on net,
wives' earnings have continued to equalize
family income among married couples.
Karoly and Burtless (1995), however,
using a different measure of
inequality and a different decomposition
technique, find that wives contributed
to a rise in inequality of equivalence
income of individuals since the
late 1970s among married couples. One
problem in reconciling these results is
that, as Gottschalk and Timothy Smeeding
(1997) point out, it is unclear how
to allocate changes in correlations
among income sources (e.g., between
husbands' and wives' earnings) in such
decompositions. They report that some
of the differences between the Karoly
and Burtless and Cancian, Danziger,
and Gottschalk results are due to differences
in the treatment of changes associated
with such correlations. In any
case, it is clear, as we have seen, that
the average equivalence income of individuals
in families headed by less educated
spouses has fallen relative to
those in families headed by the more
highly educated.
D. Trends in Intra-family Allocation:
Wage Ratios of Husbands and Wives
We have seen that, overall, women's
wages increased relative to men's, particularly
over the 1980s. In this section,
I examine how these trends have played
out within the family. The focus is on
the post-1979 period because preliminary
results indicated little change in
the wage ratio of wives to husbands in
the 1970s. Because I do not include results
from the 1970 CPS, it is possible
to compute average hourly earnings. An
advantage of this measure is that parttime
workers may be included; parttime
work is particularly prevalent
among married women. In Table 11,
data are presented for all women, and
separately by education category of the
wife; and for wives in the youngest age
group. Wage ratios are calculated as the
ratio of the mean wages of wives in the
indicated group divided by the mean
wages of husbands in that group. In addition,
data are presented on the percentage
of couples with a wage ratio of
.9 or higher, i.e., with rough wage parity
between the spouses (or, in some cases,
superiority for the wife).
Looking first at the actual femalemale
wage ratios, it may be seen that,
for the total group, there was considerable
improvement in the 1980s and
early 1990s Overall, the wage ratio rose
from 60.1 to 72.9 percent between 1979
and 1994. Also of interest is that the
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of AnmericanWomen 149
TABLE 11
FEMALE-MALE AVERAGE HOURLY EARNINGS RATIOSAMONG MARRIED COUPLES BY EDUCATION, 1979-94
Change
1979 1989 1994 1979-94
I. All Women
Female/male wage ratio (actual) 0.601 0.684 0.729 21.3%
Female/male wage ratio (fixed weight average) 0.592 0.663 0.689 16.4%
Wage ratio .9 or higher 0.228 0.316 0.368 0.14
II. Education < 12 years
All Women
Female/male wage ratio 0.579 0.647 0.643 11.2%
Wage ratio .9 or higher 0.226 0.286 0.276 0.05
Age 25-34
Female/male wage ratio 0.593 0.659 0.696 17.4%
Wage ratio .9 or higher 0.240 0.315 0.313 0.07
III. Education = 12 years
All Women
Female/male wage ratio 0.568 0.616 0.649 14.2%
Wage ratio .9 or higher 0.186 0.255 0.300 0.11
Age 25-34
Female/male wage ratio 0.610 0.671 0.711 16.5%
Wage ratio .9 or higher 0.214 0.270 0.344 0.13
VI. Education > 12 years
All Women
Female/male wage ratio 0.646 0.742 0.772 19.5%
Wage ratio .9 or higher 0.286 0.380 0.419 0.13
Age 25-34
Female/male wage ratio 0.718 0.799 0.830 15.5%
Wage ratio .9 or higher 0.330 0.424 0.440 0.11
Source: Author'stabulations from the March Current Population Surveys.
Notes: The sample in each year includes husbands and wives (matched partners) both of whom are between ages 25
and 64 and worked for at least one week in the previous year.The additional sample restrictions detailed in the Data
Appendix are applied to both partners except that respondents are not required to be full-time workers. Wages are
defined as annual earnings divided by annual hours; ratios are calculated as the mean wages of wives divided by the
mean wages of husbands. The ratio using "fixedweight averages"is calculated by fixing the distribution of husbands
and wives across education-age cells at their 1970 levels, but attributing current year wages to each cell. Four age
categories are employed: 25-34, 35-44, 45-54, and 55- 64; and three education categories: less than 12 years, 12
years, and greater than 12 years. The "proportionwith a wage ratio of .9 or higher" is the mean of a dummy variable
which takes a value of 1 if a couple's wage ratio is .9 or higher and 0 otherwise. Where averages are reported
separately, couples are classified on the basis of the wife's education and age.
percentage of couples with a wage ratio
of .9 or higher rose by 14 percentage
points, from 23 to 37 percent. Potential
market wages likely affect the behavior
and bargaining power of women who
are not currently employed, as well as
that of labor force participants. Because,
as we have seen, employment
growth has been largest for wives of
middle and high wage men who are
likely themselves to be better educated,
the figures in Table 11 may be affected
by shifts in the composition of employed
wives compared to their husThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
150 Journal of Economic Literature, Vol. XXXVI(March 1998)
bands. Thus, they may give a distorted
impression of the extent of increases in
the relative earnings of wives with given
characteristics.
To get a rough indication of whether
adjusting for changes in measured characteristics
would affect our conclusions,
I have calculated female-male wage
ratios for all workers based on "fixed
weight" averages; that is, based on
mean earnings for each sex fixing the
distribution of husbands and wives
across education-age cells at their 1970
levels, but attributing current year
wages to each cell.28 These fixedweight
averages thus adjust for any
shifts in the composition of women relative
to men across these measurable
characteristics. The results suggest that
some of female gains are due to improvements
in working wives' characteristics
relative to their husbands, because
the pace of progress is somewhat
reduced when the fixed weight
measure is used. Nonetheless, bulk of
the gains remain; the wage ratio based
on fixed weight averages increased from
59.2 to 68.9 percent between 1979 and
1994. Of course, this approach only addresses
the issue of gender differences
in the selectivity of labor force participants
in terms of their measured characteristics.
Shifts in the distribution of
unmeasured characteristics of wives
relative to husbands could still have an
impact on the observed progress (Heckman
1980).
The results for the gender ratios
based on fixed weight averages indicate
that most of the progress in narrowing
the gender ratio among married couples
occurred within age-education categories.
But this does not mean that the
rate of progress was the same for each
educational group. Table 11 indicates
that the ratio of wages of wives to husbands
increased in all education groups.
However, in each year, the highest ratio
was for wives with more than a high
school education and, moreover, relative
gains were larger for more highly
educated wives. By 1994, the wage ratio
for couples where the wife had more
than 12 years of schooling was 77 percent,
and rough wage parity existed for
42 percent of such couples. Among
wives with less than 12 years of education,
the wage ratio was 65 percent and
rough wage parity prevailed for only 30
percent of couples. In each year, the
wage ratio was higher for younger
women (25-34) and they experienced
similar wage gains across all education
groups. Nonetheless, young wives with
more than a high school education began
and ended the period with the highest
wage ratio.
E. Trends in Intra-family Allocation:
Allocation of Housework
There are two potential sources of
data on housework: time budget studies
and self-reports based on recollections
over some period of time. While the
former are likely to be more accurate,
the relevant surveys are generally not
available on a regular basis and tend to
have small samples. Given these data
problems, there have been few efforts
to examine trends in the allocation of
housework between men and women. A
notable exception is work by Fuchs
(1986; 1988). Fuchs reports that, on average,
women's total hours of work
(housework plus market work) increased
relative to men's between 1959
and 1979, and thus concludes that the
increased employment of women outside
the home had resulted in a relative
decrease in leisure for them. One problem
with these estimates, however, is
that, given the limited data available,
28 Four age categories: 25-34, 35-44, 45-54,
and 55-64; and three education categories: less
than 12 years, 12 years, and greater than 12 years
are employed.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 151
Fuchs relied on an imputation of housework
time based on coefficients obtained
from one time use study which
he then merged with large, nationally
representative data sets drawn from the
Census and CPS for a number of time
periods.29
Two other studies which use time
budget data reach somewhat different
conclusions from each other, but neither
find that total work time of women
increased relative to men's. Thomas
Juster and Frank Stafford (1991) report
that women's total work time fell from
60.9 hours to 54.4 per week between
1965 and 1991 compared to a decrease
from 63.1 to 57.8 for men, while a
United Nations (1991) report found
that women's total weekly work time remained
roughly constant at about 56.5
hours between 1965 and 1986 while
men's increased slightly from 58.3 to
59.4 hours. However, because these
studies do not provide data separately
by marital and employment status, it is
not clear what role changes in the composition
of the population play in producing
these results nor is there any direct
indication of the extent of
reallocation of tasks between partners
among married couples.
To investigate the extent of reallocation
of housework between men and
women, as well as trends in leisure, I
provide some new tabulations on time
allocation based on the Michigan Panel
Study of Income Dynamics (PSID) in
Table 12. There are some limitations to
the data used here, most important, due
to the phrasing of the housework questions,
it is possible that time spent on
child care or time spent outside of the
home on housework, e.g., shopping,
are not included, because respondents
were not explicitly prompted about
them.30 Bearing this caution in mind,
the opportunity to examine these trends
for a large, nationally representative sample
is nonetheless of interest. Tabulations
are from the 1979 and 1989 PSID
on reported average weekly housework
and market work hours in 1978 and
1988. "Not employed" is defined as
working less than 100 hours per year.
The results in Table 12 suggest a
moderate but significant reallocation of
time use between men and women.
Overall, for women, average hours of
market work increased and average
hours of housework decreased, while
hours of both market work and housework
increased for men. The changes
were larger for women. Their housework
declined by 5.4 hours per week,
on average. It fell a bit more for married
than for single women and, surprisingly,
more for nonemployed than for
employed wives. Comparisons across a
number of time use studies reported in
Blau, Ferber, and Winkler (1998) indicate
that women's housework time also
appears to have declined for both employed
and nonemployed wives between
the 1960s and the 1970s, suggesting that
the results in Table 12 for the 1980s
represent the continuation of a longterm
trend. The increase in wives' market
work of 6.3 hours are, in part, the
result of a shift of wives from nonemployed
to employed status, but, even
among employed wives, market work
hours increased. Again, similar findings
are reported by Blau, Ferber, and Winkler
(1998) for the earlier period.
In contrast to the changes for women,
the increase in men's housework was
entirely concentrated among married
men: housework hours declined by 1.1
hours for single men but increased by
29 For a critique of the Fuchs' approach, see
Joyce Manchester and David Stapleton (1991).
30 This and other data issues are evaluated in
detail by Hersch and Stratton (1997); they conclude
that, despite these data problems, the PSID
yields a useful measure of household production.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
152 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 12
TRENDS IN AVERAGEWEEKLY HOURS OF HOUSEWORK AND MARKETWORK BYMARITALSTATUSAND
EDUCATION, MEN AND WOMEN, 1978 AND 1988
1978
MarketWork Housework N
A. All individuals
Women 20.1 26.7 2008
Single 27.2 17.2 412
Married 18.3 29.1 1596
Wife not employed 0.3 37.1 607
Wife employed 29.3 24.3 989
Men 42.5 6.1 1812
Single 38.3 8.2 216
Married 43.1 5.8 1596
Wife not employed 42.5 5.0 607
Wife employed 43.5 6.4 989
B. Married Couples by Education
of Wife
Education 12 years or less
Wives 16.8 30.6 1088
Wife not employed 0.2 37.5 462
Wife employed 29.0 25.4 626
Husbands 42.0 5.7 1088
Wife not employed 41.2 4.8 462
Wife employed 42.6 6.3 626
Education 13 years or more
Wives 21.5 26.1 508
Wife not employed 0.4 35.5 145
Wife employed 30.0 22.3 363
Husbands 45.5 6.2 508
Wife not employed 46.7 5.4 145
Wife employed 45.1 6.5 363
Source: Based on author'stabulations from the Michigan Panel Study of Income Dynamics, 1979 and 1989.
Notes: Individuals and spouses are aged 25-64 in the indicated year and have valid data on both housework and
market work hours. Married individuals are included only if their spouse also satisfies the inclusion criteria.
Couples are allocated to education categories on the basis of the wife's education.
1.6 hours for married men. The increase
in married men's housework in
part represented a shift of their wives to
the employed category where, even in
the initial year, husbands' hours of
housework were greater. But, primarily,
it reflected a rise in hours of housework
of equal magnitude, 1.4 hours, for husbands
of both employed and nonemployed
wives. Trends were similar
among married couples where the wife
had a high school education or less, and
those where she had more than a high
school education.
The increase in husbands' hours of
housework was small and the allocation
of hours of housework between husbands
and wives remained quite unequal
in 1988, even when the wife was
employed. However, the significance of
this change is magnified by the fact that
housework hours decreased for wives.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 153
TABLE 12 (Cont.)
1988 Change 1978-88
Market Work Housework N Market Work Housework
A. All individuals
Women 26.4 21.3 2320 6.34 -5.35
Single 32.1 13.4 507 4.96 -3.83
Married 24.9 23.6 1813 6.58 -5.57
Wife not employed 0.3 33.0 415 -0.02 -4.01
Wife employed 32.2 20.8 1398 2.83 -3.52
Men 43.3 7.4 2183 0.71 1.27
Single 41.2 7.0 370 2.99 -1.14
Mariied 43.7 7.5 1813 0.54 1.63
Wife not employed 41.3 6.4 415 -1.19 1.40
Wife employed 44.4 7.8 1398 0.86 1.42
B. Married Couples by Education
of Wife
Education 12 years or less
Wives 23.1 24.9 941 6.36 -5.64
Wife not employed 0.1 33.3 248 -0.12 -4.26
Wife employed 31.4 21.9 693 2.39 -3.49
Husbands 41.7 7.2 941 -0.30 1.58
Wife not employed 38.6 6.0 248 -2.63 1.13
Wife employed 42.8 7.7 693 0.22 1.42
Education 13 years or more
Wives 26.7 22.1 872 5.22 -3.97
Wife not employed 0.4 32.7 167 0.07 -2.83
Wife employed 33.0 19.6 705 3.00 -2.69
Husbands 45.8 7.7 872 0.25 1.51
Wife not employed 45.4 7.0 167 -1.29 1.59
Wife employed 45.9 7.8 705 0.81 1.35
So, for example, in employed-wife families,
the ratio of husbands' to wives'
housework hours rose from 26 percent
in 1978 to 38 percent in 1988.31 Moreover,
the fact that married men's housework
hours increased when they decreased
for all other demographic
groups is also significant. This suggests
31 Using data from the 1975 Time Use Study
based on time diaries and the 1987 National Survey
of Families and Households based on direct
questions about specific tasks, Beth Shelton
(1992) also finds an increase in employed men's
housework time as a percentage of employed
women's over the 1975 to 1987 period, due in part
to a decline in women's time. Similarly, Juster and
Stafford (1991) report a decrease in women's
housework time and an increase in men's between
1965 and 1981; however, because as noted
above, they do not report results separately by
employment or marital status, trends in these factors
could account for some of the observed
changes.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
154 Journal of Economic Literature, Vol. XXXVI(March 1998)
the combined impact of demographic
trends together with changes in household
technology and in the availability
of market substitutes was to reduce
housework. Absent some reallocation
between spouses, we would have expected
married men's housework to
have declined rather than to have increased.
Finally, it may be noted that
the comparison of results of time use
studies for the 1960s and 1970s reported
in Blau, Ferber, and Winkler
(1998) finds no such trend toward increased
housework by husbands for that
period, either among those with employed
wives or among those with
nonemployed wives. This suggests that
increased housework by husbands was
an important new development of the
1980s.
Overall, these trends are consistent
with a reallocation of housework between
men and women due to the rising
relative wage of women.32 The timing
of the change also matches well with
the increase in married women's relative
wages in the 1980s. Such a reallocation
might occur in Becker's (1991)
model of specialization within the family
due to comparative advantage. As
women's comparative advantage in this
area declines, housework may be more
evenly divided.33 Alternatively, the increasing
relative wages of women could
signal an increase in their bargaining
power (e.g., Lundberg and Pollak
1996), and the reallocation of housework
between men and women could
be viewed in that context. Our finding
that housework hours increased for the
husbands of both employed and nonemployed
women seems more consistent
with the bargaining explanation, with a
shift in bargaining power between husbands
and wives in response to changing
market opportunities for women
(regardless of their current employment
status). However, calculations based on
results in Table 11 indicate a faster
growth in relative wages over the 1980s
for wives with some college than for
women with a high school education or
less, yet the housework of more highly
educated wives relative to their husbands
fell no faster than for less educated
women. This does not accord well
with either theory.34 Clearly more research
is needed to better understand
the reasons for the observed reallocation
of time use between married men
and women.
We may also consider the trends in
relative leisure time of men and women
implied by the results in Table 12, although
an important qualification is
that the measure of housework hours is
likely incomplete. Overall, based on this
measure, total work time (including
both housework and market work) was a
bit higher (2 hours) for men than for
women in 1978. Total work time increased
for both men and women, but a
bit more for men (2 hours) than for
women (1 hour). The trends were the
same for married men and women: an
increase of 2 hours in total work time
for husbands and 1 hour for wives. The
more favorable trend for women was
due to the sizable decrease in the time
spent on housework by both employed
and nonemployed wives; this was suffi-
32 See Hersch and Stratton (1994) for evidence
on the responsiveness of husband's and wife's
housework time to relative wages.
33 This conclusion would require some modifications
of the simple comparative advantage model
presented in Becker (1991). For example, there
might be a number of household activities, with
the degree of comparative advantage of each
spouse differing across activities. Then, as the
wife's relative wage increases, some activities
might be transferred from wife to husband.
34 Specifically, the ratio of husband's to wife's
housework time increased by 57 percent among
less educated wives (39 percent for nonemploye d
wives and 42 percent for employed wives); the ratio
increased by 47 percent among more highly
educated wives (41 percent for nonemployed
wives and 37 percent for employed wives).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 155
cient to outweigh most of the positive
impact on their total hours of the shift
of wives to the employed category (employed
wives work longer total hours
than nonemployed wives in each year).
Based on this data then it appears that,
for this period, gender differences in
trends in leisure time are small, and
where differences exist they slightly favor
women over men.35
F. Trendsin Intra-familyAllocation:
Domestic Violence
In recent years, considerable attention
has focused on domestic violence.
Economists have only recently begun
attempting to model the determinants
of the incidence of domestic violence,
and there have been very few studies
of trends, let alone efforts to explain
these trends. An endeavor of this type
faces enormous measurement problems
due to changes in the propensity to
recognize and report such incidents, as
well as changes in the particular questions
asked over time in various surveys.
However, as we shall see, it is possible
to reach some reasonable conclusions
about trends despite the data
problems.
One potential source of data on domestic
violence is the National Crime
Victimization Survey (NCVS) which annually
collects information about crimes
from a nationally representative sample
of U.S. households. This data set has a
number of advantages for our purposes,
but also some disadvantages. First, as a
household survey, it includes information
on crimes that are not reported to
the police; this is important because
many incidents of domestic violence are
not reported. Second, it separately reports
information on violent crimes
committed against women by "intimates,"
although who precisely is included
in that category has changed
over time.36 Finally, incidents reported
in the survey must be considered
"crimes" by the respondents. This is
both an advantage and a disadvantage.
On the one hand, it serves as an implicit
control for the seriousness of the incident.
On the other hand, while many
people see domestic violence as wrong,
they do not necessarily perceive it as
criminal (Murray Straus and Richard
Gelles 1988). Thus, data from the
NCVS is likely to underreport the incidence
of domestic violence, though perhaps
to a lesser extent over time if people
are increasingly inclined to see such
acts as criminal. The inherent problems
in using the NCVS to study trends are
exacerbated by a recent redesign of the
survey explicitly undertaken to produce
more accurate reporting of violence by
intimates and to increase the range of
incident types that are reported.
Bearing the data issues in mind, it is
nonetheless of interest to consider the
magnitude of the problem based on this
data source. In general, men in the U.S.
are considerably more likely than
women to be the victims of violence;
during the 1980s and 1990s, the overall
rate of violent crime victimization for
men averaged 1.5-1.7 times higher than
the female rate.37 In contrast, as may
be seen in Table 13, women are considerably
more likely than men to be victims
of violence by an intimate or relative,
some 3.5 to 5.4 times more likely.
The data in Table 13 show a jump in the
incidence of violence by intimates for
both men and women with the redesign
35 Based on diary reports of time specifically
spent on leisure activities, Shelton (1992) finds littie
difference between women and men in total
leisure time in 1975 or 1981.
36 An intimate is currently defined as a spouse,
ex-spouse, boy/girlfriend, or ex-boy/girlfriend. In
earlier data, figures for intimates are included in
the same category as "other relatives."
37 Calculated from U.S. Department of Justice
(1994); and Ronet Bachman and Linda Saltzman
(1995).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
156 Journal of Economic Literature, Vol. XXXVI(March 1998)
TABLE 13
AVERAGE ANNUAL RATE OF SINGLE-OFFENDER VIOLENT VICTIMIZATIONSFROM THE NATIONAL CRIME
VICTIMIZATIONSURVEY, 1979-93
Women Men
Victim-Offender
Relationship 1979-87 1987-91 1992-93 1979-87 1987-91 1992-93
Intimates and Relatives 6.3 6.5 12.1 1.8 1.2 2.6
Intimate na 5.4 9.3 na 0.5 1.4
Other Relative na 1.1 2.8 na 0.7 1.2
Acquaintance/friend 7.0 7.6 12.9 12.1 13.0 17.2
Stranger 11.4 5.4 7.4 29.4 12.2 19.0
Sources: 1979-87: Harlow (1991); 1987-91: Bachman (1994); and 1992-93: Bachman and Saltzman (1995).
Notes: Rate per 1000 individuals, 12 years of age or over. Intimate includes spouse, ex-spouse, boy/girlfriend, and
ex-boy/girlfriend. Data for 1979-87 include multiple offenders. na indicates not available. Results from the 1992-
93 NCVS are based on a redesigned survey and thus may not be comparable to previous years.
of the survey, indicating considerable
sensitivity of responses to how questions
are phrased and whether or not
additional cuing occurs. However, prior
to the redesign of the survey, the incidence
of violence by intimates and relatives
is virtually constant for women at
about 6.5 per thousand for the 1979-87
and 1987-91 periods, and fell somewhat
for men from 1.8 to 1.2.38 Given the
many biases which would be expected
to cause an upward trend in these figures
over time, it seems reasonable to
conclude that there was no increase in
domestic violence between the early
and the late 1980s.
Another source of data on domestic
violence is the 1975 and 1985 National
Family Violence Surveys (NFVS) reported
in Table 14. This nationally representative
survey which includes married
and cohabiting couples is portrayed
to respondents as a survey of "family
problems" rather than of crime. And respondents
are explicitly questioned
about the particular types of violent behavior
which are itemized in the footnotes
to the table. Two indexes are obtained,
one for overall violence and the
other for severe violence; the latter includes
actions which are believed to
have a high probability of causing
injury. As may be seen by comparing
the violence rates in Tables 13 and 14,
the rates obtained from the NFVS, even
for severe violence, are considerably
higher than those obtained by the
NCVS. For example, for 1992-93, after
the redesign, the rate of violence
against women by intimates from the
NCVS is 9.3 per thousand compared to
a rate of severe husband to wife violence
of 30 per thousand from the
NFVS for 1985. When all violence is
considered, the NFVS rate of 113 for
husbands to wives in 1985 is 12 times
the NCVS rate and indicates that at
least one incident of violent behavior of
husbands to wives occurred among 11
percent of American couples in 1985,
with serious incidents taking place
among 3 percent. Are the rates from
the NFVS, which are so much higher
than the NCVS rates, credible? The an-
38 Another discontinuity in the table is the sizable
fall in the rate of violence by strangers for
both men and women between 1979-87 and 1987-
91. This likely reflects the inclusion of multipleoffender
victimizations in the 1979-87 data, but
not in the later period.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 157
TABLE14
RATE OF VICTIMIZATIONBASED ON MARITALVIOLENCE INDEXES, 1975 AND 1985
HusbandtoWife Wifeto Husband
Year OverallViolence(1) SevereViolence(2) OverallViolence(1) SevereViolence(2)
1975 121 38 116 46
1985 113 30 121 44
%Change -6.6 -21.1 4.3 -4.3
Source: GellesandStraus(1986)basedonthe 1975 and1985 NationalFamilyViolenceSurveys.
Notes: Rateper 1000wives(husbands).Includescohabitingcouples.
(1)Includes:threwsomething;pushed,grabbed,orshoved;slapped;kicked,bitorhitwithfist;hitortriedto hit
withsomething;beatup;threatenedwithknifeorgun;usedgunorknife.
(2)Includesonly:kicked,bitorhitwithfist;hitortriedto hitwithsomething;beatup;threatenedwithgunor
knife;usedgunorknife.
swer seems to be yes. Straus and Gelles
(1988) summarize findings from a large
number of similar studies which obtain
results of roughly the same magnitude.
The difference between the NFVS and
the NCVS appears to occur because the
NFVS does not query respondents
about "crimes," but rather "problems,"
and also because it asks them about specific
types of violent behavior.
Another surprising finding from the
NFVS is that, when this methodology is
employed, the violence rate of wives to
husbands is similar to the rate of violence
of husbands to wives. However, as
Straus and Gelles and others have
pointed out, the greater average size
and strength of men and their greater
aggressiveness means that the same act
(e.g., a punch) is likely to inflict greater
amounts of pain and injury when committed
by a man. This reasoning is
supported by the much lower rates of
victimization of men by intimates reported
in the NCVS where "crimes"
are the focus. In addition, the data
on homicide indicate that 28.3 percent
of female homicide victims in 1992
were known to have been killed by an
intimate compared to only 3.6 percent
of males (Bachman and Saltzman
1995).39 Moreover, Straus and Gelles
argue that many of the assaults of
women against their husbands are acts
of retaliation or self-defense. Some evidence
in support of this is that while
these data suggest that women are violent
as frequently within the family as
men, they are very rarely found to be
violent outside the family.
Nonetheless, the high rates of violence
of wives against husbands reported
in this data raise some concerns
about the validly of the results for violence
of husbands against wives. Specifically,
if one dismisses much of the
violence of women against men as insignificant
because it is unlikely to cause
injury, some unknown portion of the
violence of husbands against wives
should also likely be dismissed for the
same reason. Putting this somewhat
differently, though many reasonably
feel that any domestic violence is wrong
and should not occur, the statistics from
the measure of violence employed by
NFVS may give an inflated picture of
the prevalence of violence that consti-
39 The relationship of the offender to the victim
was not identified in 30.9 percent of the cases with
female victims and 41.3 percent of the cases with
male victims.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
158 7ournal of Economic Literature, Vol. XXXVI(March 1998)
tutes a serious problem in terms of
harm or potential harm. On the other
hand, as Straus and Gelles point out,
violence by women may be of substantive
concern because it is dangerous to
them. It may set the stage for a more
serious immediate assault on them by
their partners who, as previously noted,
are generally bigger and stronger; or it
may provide a precedent or rationale
for violence on their partners' part at a
later time.
Whatever the conclusion about the
levels of violence reported in the two
surveys, the data in Table 14 from the
NFVS strongly support our conclusion
from the NCVS that there is no evidence
of an increase in the rate of domestic
violence against women, in this
case between the mid-1970s and mid-
1980s. Indeed, there is even some evidence
of a decrease, especially in the
incidence of severe violence. If we
again assume that, if anything, the tendency
would be to increase reporting of
such incidents, our confidence that a
decrease has occurred would be
strengthened further.
An interesting question that may be
raised about these trends is whether
economic variables played a role in
causing the apparent decrease in the incidence
of domestic violence. Specifically,
it is possible that, as women's
relative wages and employment rates
have risen, they have become less likely
to remain in an abusive situation. The
increased establishment of shelters for
battered women could also have contributed
to the decrease. However, currently,
there are no detailed studies of
the causes of these trends and, indeed,
as we have seen, there is some reason to
feel less than completely confident that
a decrease has even occurred. Perhaps
the most reasonable conclusion is that
there is no evidence that the inclusion
of domestic violence would in any way
change our finding, based on other indicators,
that women's status within the
family has, if anything, increased.
A last point to consider is what can be
learned about the incidence of domestic
violence among women by level of education
and by race. For this purpose,
the most recent NVCS figures are
particularly useful because they provide
the most accurate reporting of the
relevant data for this survey. The
findings are somewhat surprising. While
high school dropouts and black women
have higher overall violence victimization
rates than others, rates of violence
by intimates are similar for blacks and
whites, and across education categories,
with the exception of considerably
lower rates among women college
graduates. If it can be assumed that this
pattern has been fairly stable,40 here
again, the bottom line is that the conclusions
reached based on other indicators
do not have to be revised by a consideration
of domestic violence.
V. "HavingIt All:"CombiningWork
and Family
Given traditional male and female
roles, one potential source of gender
differences in economic well-being is a
difference in the ability to successfully
combine work and family responsibilities
so as to attain desired objectives in
both areas. Men tend not to face the
same dilemmas as women, or not to
face them to the same extent, because
desirability as a spouse and ability to
contribute economically to one's family
are positively correlated with labor market
success for men. While shifting gender
roles within the family may be increasing
the prevalence of family
responsibilities among men, as we have
40 Tabulations for previous years in Caroline
Harlow (1991) and Bachman (1994) suggest similar
patterns.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 159
seen, considerable gender differences
remain.
Differences in typical gender roles in
the family are also indicated by research
which widely finds a negative
simple correlation between women's
earnings and the presence of children.
In general, it has also been found that
children are negatively related to female
earnings, all else equal, while marriage
and children are positively related
to male earnings.41 The finding for
women depends in part on what other
variables are controlled for. A substantial
part of the negative effect of children
when education and potential experience
(age) are held constant
reflects the lower actual labor market
experience of women with children, but
some evidence of a negative effect has
been obtained in a number of studies
even when experience is controlled for
(e.g., Jacobsen and Laurence Levin
1995; Waldfogel 1997). This may reflect
that, in the past, the birth of a child
often meant that a woman withdrew
from the labor force entirely, thus
breaking her tie to her employer and
foregoing the returns to any firm-specific
training she might have received,
as well as any rewards for having made
an especially good job match. Anticipation
of this pattern could also deter
both women and their employers from
making large investments in firm-specific
training to begin with.
This is by no means a new dilemma
for wonmen Ihifltrthptrms -ofbteAhnit&perceived
by the bulk of women have
changed over time. In recent decades,
the desire to "haveit all," that is to successfully
pursue a career and to have a
family appears to have become an increasingly
common goal among women,
especially the college educated. While
there is no way of knowing precisely
how prevalent it is, recent work by
Goldin (1997) shows how rarely women
have been able to successfully combine
career and family, even among a relatively
recent cohort of college women.
Using National Longitudinal Survey
data, Goldin looks at attainment of career
and family by 1985-88 among
white women who graduated from college
between 1966 and 1979. These
women ranged in age from 34 to 44 in
1988. Goldin uses a variety of definitions
of career, but especially emphasizes
results where career is defined as
having hourly earnings in the selected
years exceeding that of the 25th percentile
of men with 16 or more years of
schooling in the Current Population
Survey (CPS) in the relevant year.
When a three-year definition is employed,
1985, 1987, and 1988, only 26
percent of the women qualify as having
careers. Using a two-year definition,
1987 and 1988, one third qualify, still a
relatively low proportion. A still smaller
proportion attain both career and family,
where family is defined by the presence
of children: only 13 percent using
the three-year definition and 16 percent
using the two-year definition. This reflects
a considerably lower incidence of
career among women with children.
This is a relatively stringent definition
of career, because a substantial proportion
of men would not have attained a
career by this definition either. However,
when Goldin shifts to a lower cutoff-being
in the labor force in each of
three years and generally working fulltime-the
incidence of career and fam-
41 See for example, Korenman and David Neumark
(1991), Waldfogel (1997), and Fuchs (1988).
A number of serious econometric issues are raised
in seeking to measure the effect of children on
women's wages due to unmeasured heterogeneity
among individuals and endogeneity of the decision
to have children. For an especially thorough treatment
of these issues, see Korenman and Neumark
(1992 and 1994). While Korenman and Neumark's
earlier findings were ambiguous, their more recent
and referred results do indicate a negative
effect of children on wages for white women.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
160 Journal of Economic Literature, Vol. XXXVI(March 1998)
ily among the women remains extremely
low, only 22 percent. And, under
all three definitions, approximately
one half of the women with careers had
not had a first birth.
Some qualifications may be noted
concerning these findings. We do not
know what fraction of the childless
women were disappointed in not having
children. However, Goldin (1997) does
present evidence that a considerable
fraction (half or more) had indicated
that they desired children when surveyed
in their early to mid twenties. Another
qualification is that this cohort
(aged 34-44 in 1988) has in general
postponed marriage and childbearing.
Thus, childlessness among them is
likely to be overestimated, because
some may still have children later, and
their measured career attainments may
appear especially low during the years
surveyed because of the presence of
young children among the women with
families. Yet even these qualifications
suggest that these women face the need
to make decisions and tradeoffs seldom
confronted by their male counterparts.
This concern is reinforced by Waldfogel's
(1997) finding that the wages of
women with children lag increasingly
behind those of women without children.
On a more positive note, Waldfogel
finds that access to family leave substantially
mitigates the negative effect
of children on women's wages. This suggests
that the difficulties which women
have faced in achieving career and family
may be reduced as more firms adopt
such policies.42
VI. Conclusion
The goal of this paper is to delineate
the trends in the well-being of American
women over the past quarter century.
I argue for a broad range of
indicators designed to capture changes
in women's well-being in the family
as well as in the labor market, and also
for the importance of examining
both women's progress relative to men
and in some absolute sense compared to
the initial levels of indicators for
women. Moreover, given trends in recent
years of growing disparities across
groups, the investigation must proceed
at a disaggregated as well as at an aggregated
level to be informative. I have
undertaken this investigation both
through direct analyses of data, primarily
from the Current Population Surveys,
and examination of existing research
results.
Overall, I conclude that women have
made substantial progress toward gender
equality over the past 25 years
across a number of dimensions. Gender
differences in labor force participation
have narrowed sharply and women now
remain in the labor market more consistently
over the life cycle. Differences
between men and women in occupations,
types of education, and rates of
self-employment have been greatly diminished;
and women have narrowed
the gender wage gap substantially.
Within the family, wages of wives rose
relative to their husbands' and, perhaps
as a consequence, there was a small but
notable reallocation of housework between
husbands and wives. And, while
the data are not as reliable as one would
like, there is no indication of any increase
in domestic violence between
the mid-1970s and mid-1980s which
would tend to offset these gains. Relative
gains also appear to have been
widely distributed across education
groups. These relative gains appear to
be matched by progress for women
overall in an absolute sense. Women's
real wages increased substantially over
42The Family and Medical Leave Act of 1993
requires all employers over a minimum size to offer
12 weeks of unpaid job-protected leave.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 161
the 1969-1994 period, while men's stagnated.
Similarly, women upgraded their
major occupations absolutely between
1979 and 1988, in that they moved into
higher paying occupational categories,
while, on net, men's occupational shifts
left their real wages unchanged.
This does not mean that discrimination
and other gender-related disabilities
affecting women have disappeared.
There still exists a considerable, although
reduced, gender wage gap after
controlling for measured characteristics,
which is often taken as an estimate
of discrimination. And the challenges
of combining work and family
appear to continue to pose serious obstacles
and dilemmas for women but, at
this point, do not seem to affect men in
the same way or at least to the same
extent. Moreover, trends in family
structure and, in particular, the increase
in families headed by single
women, have adversely affected the
economic well-being of women and
their dependent children. But this development,
unlike those cited above was
heavily concentrated among women
with little education (less than 12 years)
and black women.
The faster rise in female headed
families was just one of a number of developments
adversely affecting the relative
economic well-being of less educated
women. In a manner strikingly
parallel to trends in the labor market
for men, wage differentials by education
widened among women in the
1980s and early 1990s, and female high
school dropouts experienced real wage
declines. While women at all skill levels
upgraded their occupations, less skilled
and middle skilled women lost union
jobs, although at a slower pace than
men, and their representation in higher
paying industries declined, at about the
same rate as men's. Less educated
women have also increased their labor
force participation less than other education
groups, as have black women
compared to white women. Finally, not
only has female headship increased
most rapidly among less educated
women, the income of individuals in
families headed by couples with lower
educational attainment has fallen relative
to that of more highly educated
couples. Some simple analyses strongly
suggest that compositional shifts do not
entirely account for the deteriorating
economic position of less educated
Americans, but rather that real changes
in behavior and opportunities underlie
at least some part of them. My principal
concern has been with the well-being of
women. However, these findings for
less educated women also serve to underscore
the widening gap between
more and less skilled Americans of both
sexes, as well as to emphasize its broad
dimensions.
Much remains to be learned about
the details of the trends outlined here,
and even more about their fundamental
causes. In each major area where we
probed existing studies for explanations
of the trends, whether in participation,
relative wages, or single headship, a
substantial portion of the explanation
must at this point be allocated to behavioral
shifts and changes in tastes. The
sources of the growth in headship are
particularly poorly understood. It may
be that this is inescapable at a time of
rapid change in gender roles and social
attitudes toward women and family relationships.
But perhaps we should take
this as a challenge to develop and refine
economic models which can account for
a greater proportion of the changes. It
is in any case clear that more serious
empirical analyses of the reasons for
these trends are needed if we are to understand
these developments better.
Such analyses also provide an excellent
opportunity for testing the economic
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
162 Journal of Economic Literature, Vol. XXXVI(March 1998)
significance as well as the statistical significance
of our models.
Appendix: Data Description
The primary source of data for the
empirical results presented here are the
March Current Population Surveys
(CPS) for 1970, 1980, 1990, and 1995.
The CPS was used because it is a large,
nationally representative sample which
permits considerable disaggregation by
subgroups. The CPS data for 1970,
1980, and 1990 were selected because
the earnings measure which is based on
the previous calendar year relates to
years of comparable economic activity
over broad ten year intervals; the 1995
CPS was the most recent year available
at the time this research was conducted.
While the Census of Population could
have been used as an alternative data
source for 1970-1990, providing still
larger samples, the CPS provides a consistent
data source throughout, including
the most recent year.
There was a change in the coding of
the CPS education variables affecting
the 1995 CPS data. In earlier years, information
is available on years of
schooling completed, top-coded at 18.
The 1995 codes, with years of education
assigned in parentheses, are: less than
1st grade (1); 1st, 2nd, 3rd, or 4th grade
(2.5); 5th or 6th grade (5.5); 7th or 8th
grade (7.5); 9th grade (9); 10th grade
(10); 11th grade or 12th grade no diploma
(11); high school graduate (12);
some college but no degree (13); associate
degree (14); bachelor's degree (16);
master's degree; professional school degree;
and doctorate degree (18).
Wage results are generally for the
weekly wages of full-time workers (i.e.,
those usually working 35 hours or more
per week in the preceding year). While
average hourly earnings (i.e., annual
earnings divided by weeks worked multiplied
by hours per week) would have
been a preferable measure in some respects,
weekly hours information in
1970 is available only for hours last
week. This is in an error-ridden measure
of the preferred variable, usual
weekly hours, and, moreover, is only
available for a selected group of workers,
i.e., those employed last week.
However, information on whether the
individual usually worked part-time or
full-time in the preceding year is available
in each of the surveys, including
1970. I thus generally use the weekly
earnings measure which can be computed
on a comparable basis in each
year and which applies to an identifiable
group, full-time workers, rather
than an arbitrary group, those employed
in the previous week.43
Average weekly and hourly wage rates
are calculated based on the previous
calendar year and exclude individuals
who were self-employed or for whom
data on wage and salary income was
either missing or imputed. Except when
otherwise indicated, the sample was
further restricted to full-time workers
who worked at least one week in the
preceding year and who participated in
the labor force (i.e., were employed or
unemployed) for at least 27 weeks. Following
Katz and Murphy (1992), individuals
with real weekly earnings of less
than $67 or real hourly earnings of less
than $1.68 in 1982 dollars (i.e., one half
the value of the minimum wage, assuming
a 40 hour week for the weekly earnings
cutoff) were excluded; for individuals
whose annual wage and salary
income was top-coded by the CPS, annual
income was imputed as 1.45 the
top-coded value ($50,000 in 1970 and
43 The weekly wage measure has been used
quite extensively in recent studies of trends in
wages, wage inequality, and race wage differentials;
see, for example, Katz and Murphy (1992),
and Juhn, Murphy, and Pierce (1991 and 1993).
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 163
1980 and $1,999,998 in 1990 and 1995).
All tabulations employ the CPS sampling
weights.
The CPS data are supplemented with
information from the Michigan Panel
Study of Income Dynamics (PSID) on
average weekly hours of self-reported
market work and housework. The PSID
is a large, national sample which is
available for a number of years making
it attractive compared to smaller, and
less frequently available time use
studies. I employ data from the 1979
and 1989 PSID yielding information
for the preceding years. I begin with
1979 because housework data were not
consistently available before this. The
1989 PSID was the most recent publicly
available when this work was commenced.
These are two years of roughly
comparable levels of economic activity.
The poverty sample of the PSID was
deleted in the tabulations presented.
REFERENCES
AIGNER, DENNIS J. AND CAIN, GLEN G. "Statistical
Theories of Discrimination in Labor Markets,"
Ind. Lab. Relat. Rev., Jan. 1977, 30(2),
pp. 175-87.
BACHMAN, RONET. Violence against women. A national
crime victimization survey report. U.S.
Department of Justice, Jan. 1994.
BACHMAN, RONET AND SALTZMAN, LINDA E.
Violence against women: Estimates from the redesigned
survey. Washington, DC: U.S. Department
of Justice, Aug. 1995.
BARRON, JOHN M.; BLACK, DAN A. AND
LOEWENSTEIN, MARK A. "Gender Differences
in Training, Capital, and Wages," J. Human
Res., Spring 1993, 28(2), pp. 342-64.
BECKER, GARY S. "Human Capital, Effort, and
the Sexual Division of Labor," J. Lab. Econ.,
Jan. 1985, 3(1), pp. S33-58.
. A treatise on the family. Enlarged ed.
Cambridge, MA: Harvard U. Press, 1991.
BELLER, ANDREA H. "Changes in the Sex
Composition of U.S. Occupations, 1960-1981,"
J. Human Res., Spring 1985, 20(2), pp. 235-
50.
BLANK, REBECCA. "Teen Pregnancy: Government
Programs Are Not the Cause," Feminist Economics,
Summer 1995,1(2), pp. 47-58.
BLAU, FRANCINE D. AND BELLER, ANDREA H.
"Trends in Earnings Differentials by Gender:
1971-1981," Ind. Lab. Relat. Rev., July 1988,
41(4) pp. 513-29.
BLAU, FRANCINE D. AND EHRENBERG, RONALD
G., eds. Gender and family issues in the workplace.
New York: Russel Sage Foundation,
1997.
BLAU, FRANCINE D. AND FERBER, MARIANNE A.
"Career Plans and Expectations of Young
Women and Men," J. Human Res., Fall 1991,
26(4), pp. 581-607.
BLAU, FRANCINE D.; FERBER, MARIANNE A. AND
WINKLER, ANNE E. The economics of women,
men, and work. 3rd ed. Englewood Cliffs, NJ:
Prentice-Hall, 1998.
BLAU, FRANCINE D. AND KAHN, LAWRENCE M.
"Wage Structure and Gender Earnings Differentials:
An International Comparison,"
Economnica, May 1996, 63(250), suppl., pp. S29-
62.
. "Swimming Upstream: Trends in the Gender
Wage Differential in the 1980s," J. Lab.
Econ., Jan. 1997,15(1, pt. 1), pp. 1-42.
BLAU, FRANCINE D.; SIMPSON, PATRICIA AND
ANDERSON, DEBORAH. "Continuing Progress?
Trends in Occupational Segregation Over the
1970s and 1980s." Cornell U., Sept. 1997.
BORJAS, GEORGE J. "Assimilation and Changes in
Cohort Quality Revisited: What Happened to
Immigrant Earnings in the 1980s?" J. Lab.
Econ., Apr. 1995, 13(2), pp. 201-45.
BROWN, CHARLES AND CORCORAN, MARY. "SexBased
Differences in School Content and the
Male/Female Wage Gap," J. Lab. Econ., July
1997, 15(3, pt. 1), pp. 431-65.
BUMPASS, LARRY L. AND SWEET, JAMES A. "National
Estimates of Cohabitation," Demography,
Nov. 1989, 26(4), pp. 615-25.
CANCIAN, MARIA; DANZIGER, SHELDON AND
GOTTSCHALK, PETER. "Working Wives and
Family Income Inequality Among Married Couples,"
in Uneven tides: Rising inequality in
America, Eds.: SHELDON DANZIGER AND PETER
GOTTSCHALK. New York: Russell Sage
Foundation, 1993, pp. 195-221.
COLEMAN, MARY T. AND PENCAVEL, JOHN.
"Changes in Work Hours of Male Employees,
1940-1988," Ind. Lab. Relat. Rev., Jan. 1993a,
46(2), pp. 262-83.
. "Trends in Market Work Behavior of
Women Since 1940," Ind. Lab. Relat. Rev., July
1993b, 46(4), pp. 653-76.
COLLEGE BOARD. "College Board Reports Continuing
Upward Trend in Average Scores on
SAT 1." Press Release, Aug. 22, 1996.
DARITY, WILLIAM A., JR. AND MYERS, SAMUEL
L., JR. "Family Structure and the Marginalization
of Black Men: Policy Implications," in The
decline in marriage among African Americans.
Eds.: M. BELINDA TUCKER AND CLAUDIA
MITCHELL-DERNAN. New York: Russell Sage
Foundation, 1995, pp. 263-308.
DEVINE, THERESA J. "Characteristics of Self-Employed
Women in the United States," Mon.
La. Rev., Mar. 1994a, 117(3),pp. 20-34.
. "Changes in Wage-and-Salary Returns to
Skill and the Recent Rise in Female Self-EmThis
content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
164 Journal of Economic Literature, Vol. XXXVI(March 1998)
ployment," Amer. Econ. Rev., May 1994b,
84(2), pp. 108-13.
ENGLAND, PAULA. "The Failure of the Human
Capital Theory to Explain Occupational Sex
Segregation," J. Human Res., Summer 1982,
17(3), pp. 358-70.
ENGLAND, PAULA AND FARKAS, GEORGE. Households,
employment, and gender: A social, economic,
and demographic view. Hawthorne, NY:
Aldine de Gruyter, 1986.
FAIRLIE, ROBERT W. AND MEYER, BRUCE D.
"Ethnic and Racial Self-Employment Differences
and Possible Explanations," J. Human
Res., Fall 1996, 31(4), pp. 757-93.
FARBER, HENRY S. "Are Lifetime Jobs Disappearing?
Job Duration in the United States: 1973-
1993." National Bureau of Economic Research
Working Paper No. 5014, Feb. 1995.
FERBER, MARIANNE A. AND NELSON, JULIE A.,
eds. Beyond economic man. Chicago: U. of Chicago
Press, 1993.
FUCHS, VICTOR R. "His and Hers: Gender Differences
in Work and Income, 1959-1979," J.Lab.
Econ., July 1986, 4(3), pt.2, pp. S245-72.
. Women's quest for economic equality.
Cambridge, MA: Harvard U. Press, 1988.
GERONIMUS, ARLINE T. AND KORENMAN, SANDERS.
The Socioeconomic Consequences of
Teen Childbearing Reconsidered," Quart. J.
Econ., Nov. 1992,107(4), pp. 1187-1214.
GOLDIN, CLAUDIA D. Understanding the gender
gap. New York: Oxford U. Press, 1990.
. "Career and Family: College Women
Look to the Past," in FRANCINE D. BLAU
AND RONALD G. EHRENBERG, eds. 1997, pp.
20-58.
GOTTSCHALK, PETER AND SMEEDING, TIMOTHY
M. "Cross-National Comparisons of Earnings
and Income Inequality," J. Econ. Lit., June
1997, 35(2), pp. 633-87.
GUNDERSON, MORLEY. "Male-Female Wage Differentials
and Policy Responses," J. Econ. Lit.,
Mar. 1989, 27(1), pp. 46-72.
HARLOW, CAROLINE WOLF. Female victims of
violent crime. Washington, DC: U.S. Department
of Justice, Jan. 1991.
HECKMAN, JAMESJ. "Sample Selection Bias as a
Specification Error," in JAMES P. SMITH, ed.
1980, pp. 206-48.
HERSCH, JONI AND STRATTON, LESLIE S.
"Housework, Wages, and the Division of
Housework Time for Employed Spouses,"
Amer. Econ. Rev., May 1994, 84(2), pp. 120-25.
. "Housework, Fixed Effects, and Wages of
Married Workers." J. Human Res., Spring 1997,
32(2), pp. 285-307.
HOFFMAN, SAUL D. AND DUNCAN, GREG J. "The
Effects of Incomes, Wages, and AFDC Benefits
on Marital Disruption," J. Human Res., Winter
1995, 30(1), pp. 19-41.
HOFFMAN, SAUL D.; FOSTER, E. MICHAEL AND
FURSTENBERG, FRANK F., JR. "Reevaluating
the Costs of Teenage Childbearing," Demography,
Feb. 1993a, 30(1), pp. 1-13.
JACOBS, JERRY A. "Gender and Academic Specialties:
Trends among Recipients of College Degrees
in the 1980s," Sociology of Education,
Apr. 1995, 68(2), pp. 81-98.
JACOBSEN, JOYCE P. "Trends in Workforce Sex
Segregation, 1960-1990," Soc. Sci. Quart., Mar.
1994, 75(1), pp. 204-11.
The economics of gender. 2nd ed. Cambridge,
MA; Blackwell, forthcoming.
JACOBSEN, JOYCE P. AND LEVIN, LAURENCE M.
"Effects of Intermittent Labor Force Attachment
on Women's Earnings," Mon. Lab. Rev.,
Sept. 1995,118(9), pp. 14-19.
JOHNSON, WILLIAM R. AND SKINNER, JONATHAN.
"Labor Supply and Marital Separation," Amer.
Econ. Rev., June 1986, 76(3), pp. 455-69.
JUHN, CHINHUI. "Decline of Male Labor Market
Participation: The Role of Declining Market
Opportunities," Quart. J. Econ., Feb. 1992,
107(1), pp. 79-121.
JUHN, CHINHUI AND MURPHY, KEVIN M. "Wage
Inequality and Family Labor Supply," J. Lab.
Econ., Jan. 1997, 15(1) pt. 1, pp. 72-97.
JUHN, CHINHUI; MURPHY, KEVIN M. AND
PIERCE, BROOKS. "Accounting for the Slowdown
in Black-White Wage Convergence," in
MARVIN KOSTERS, ed., 1991, pp. 107-43.
"Wage Inequality and the Rise in Returns
to Skill," J.Polit. Econ., June 1993, 101(3), pp.
410-42.
JUSTER, F. THOMAS AND STAFFORD, FRANK P.
"The Allocation of Time; Empirical Findings,
Behavioral Models, and Problems of Measurement,"
J.Econ. Lit., June 1991, 29(2), pp. 471-
522.
KAROLY, LYNN A. AND BURTLESS, GARY. "Demographic
Change, Rising Earnings Inequality,
and the Distribution of Personal Well-Being,
1959-1989, Demography, Aug. 1995, 32(3), pp.
379-405.
KATZ, LAWRENCE F. AND MURPHY, KEVIN M.
"Changes in Relative Wages, 1963-87: Supply
and Demand Factors," Quart. J. Econ., Feb.
1992, 107(1), pp. 35-78.
KORENMAN, SANDERS AND NEUMARK, DAVID.
"Does Marriage Really Make Men More Productive?"
J. Human Res., Spring 1991, 26(2),
pp. 282-307.
. "Marriage, Motherhood, and Wages." J.
Human Res., Spring 1992, 27(2), pp. 233-55.
. "Sources of Bias in Women's Wage Equations,"
J. Human Res., Spring 1994, 29(2), pp.
379-405.
KOSTERS, MARVIN H., ed. Workers and their
wages. Washington, DC: AEI Press, 1991.
LEIBOWITZ, ARLEEN AND KLERMAN, JACOB
ALEX. "Explaining Changes in Married Mothers'
Employment Over Time," Demography,
Aug. 1995, 32(3), pp. 365-78.
LEONARD, JONATHAN S. "Women and Affirmative
Action," J. Econ. Perspectives, Winter 1989,
3(1), pp. 61-75.
LOMBARD, KAREN V. "Female Self-Employment
and the Demand for Flexible, Non-Standard
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions
Blau: The Well-Being of American Women 165
Work Schedules." Working Paper, U. of Miami,
Jan. 1996.
LUNDBERG, SHELLY AND POLLAK, ROBERT A.
"Bargaining and Distribution in Marriage," J.
Econ. Perspectives, Fall 1996, 10(4), pp. 139-
58.
LUNDBERG, SHELLY J. AND STARTZ, RICHARD.
"Private Discrimination and Social Intervention
in Competitive Labor Markets," Amer. Econ.
Rev., June 1983, 73(3), pp. 340-47.
MACPHERSON, DAVID A. AND HIRSCH, BARRY T.
"Wages and Gender Composition: Why Do
Women's Jobs Pay Less?" J. Lab. Econ., July
1995, 13(3), pp. 426-71.
MANCHESTER, JOYCE M. AND STAPLETON,
DAVID C. "On Measuring the Progress of
Women's Quest for Economic Equality," J.Human
Res., Summer 1991, 26(3), pp. 562-80.
MANSKI, CHARLES F. "Anatomy of the Selection
Problem." J. Human Res., Summer 1989, 24(3),
pp. 343-60.
MCLANAHAN, SARA AND BOOTH, KAREN.
"Mother-Only Families: Problems, Prospects,
and Politics," Journal of Marriage and the Family,
Aug. 1989, 51(3), pp. 557-80.
MINCER, JACOB. "Labor Force Participation of
Married Women: A Study of Labor Supply," in
Aspects of labor economics. Universities-National
Bureau Committee for Economic Research.
Princeton, NJ: Princeton U. Press, 1962,
pp. 63-97.
MINCER, JACOB AND POLACHEK, SOLOMON.
"Family Investments in Human Capital: Earnings
of Women," J. Polit. Econ., Mar./Apr. 1974,
82(2) pt. 2, pp. S76-108.
MOFFITT, ROBERT."Incentive Effects of the U.S.
Welfare System: A Review." J. Econ. Lit., Mar.
1992, 30(1), pp. 1-61.
MURRAY, CHARLES. Losing ground: American social
policy 1950-1980. New York: Basic Books,
1984.
OLSEN, RANDALL J. AND FARKAS, GEORGE. "The
Effect of Economic Opportunity and Family
Background on Adolescent Cohabitation and
Childbearing among Low-Income Blacks," J.
Lab. Econ., July 1990, 8(3), pp. 341-62.
O'NEILL, JUNE AND POLACHEK, SOLOMON. "Why
the Gender Gap in Wages Narrowed in the
1980s," J. Lab. Econ., Jan. 1993, 11(1), pt. 1,
pp. 205-28.
PENCAVEL, JOHN. "Changes in Male Work Behavior
and Wages." Stanford U., Feb. 1997.
RESKIN, BARBARA F. AND HARTMANN, HEIDI I.,
eds. Women's work, men's work: Sex segregation
on the job. Washington, D.C.: National
Academy Press, 1986.
ROYALTY, ANNE BEESON. "The Effects of Job
Turnover on the Training of Men and Women,"
Ind. Lab. Relat. Rev., Apr. 1996, 49(3), pp.
506-21
SCHULTZ, T. PAUL. "Eroding the Economic Foundations
of Marriage and Fertility in the United
States." Paper presented at the conference
on "Changes in Family Patterns in Western
Countries," Bologna, Italy, Oct. 6-8, Sept.
1994.
SHELTON, BETH ANNE. Women, men, and time:
Gender differences in paid iwork, housework
and leisure. New York: Greenwood Press, 1992.
SMITH, JAMES P., ed. Female labor stupply: Theory
and estimation. Princeton, NJ: Princeton U.
Press, 1980.
SMITH, JAMES P. AND WARD, MICHAEL P. "TimeSeries
Growth in the Female Labor Force," J.
Lab. Econ., Jan. 1985, 3(1), suppl., pp. S59-90.
. "Women in the Labor Market and in the
Family," J. Econ. Perspectives, Winter 1989,
3(1), pp. 9-23.
SORENSEN, ELAINE. "The Crowding Hypothesis
and Comparable Worth Issue," J. Human Res.,
Winter 1990, 25(1), pp. 55-89.
STRAUS, MURRAY A. AND GELLES, RICHARD J.
"Societal Change and Change in Family Violence
from 1975 to 1985 as Revealed by Two
National Surveys," Journal of Marriage and the
Family, Aug. 1986, 48(3), pp. 465-79.
. "How Violent Are American Families? Estimates
from the National Family Violence Resurvey
and Other Studies," in Family abuse and
its consequences: New directions in research.
Eds.: GERALD T. HOTALING ET AL. Newbury
Park, CA: Sage, 1988, pp. 14-36.
UNITED NATIONS. The iworld's women 1970-1990:
Trends and statistics. New York: United Nations,
1991.
U.S. DEPARTMENT OF JUSTICE. Criminal victimization
in the United States: 1973-92 Trends.
Washington, DC: U.S. Dept. Of Justice, July
1994.
U.S. DEPARTMENT OF LABOR, BUREAU OF LABOR
STATISTICS. Report on the American workforce.
Washington, D.C.: U.S. GPO, 1995.
WALDFOGEL, JANE. "Working Mothers Then and
Now: A Cross-cohort Analysis of the Effects of
Maternity Leave on Women's Pay," in
FRANCINE D. BLAU AND RONALD G. EHRENBERG,eds.
1997, pp. 92-126.
WELLINGTON, ALISON J. "Changes in the
Male/Female Wage Gap, 1976-85," J. Human
Res., Spring 1993, 28(2), pp. 383-411.
WILSON, WILLIAM J. The truly disadvantaged.
Chicago: U. of Chicago Press, 1987.
WOOD, ROBERT G. "Marriage Rates and Marriageable
Men: A Test of the Wilson Hyothesis."
J. Human Res., Winter 1995, 30(1), pp.
163-93.
This content downloaded from 147.251.185.127 on Wed, 18 Mar 2015 09:58:09 UTC
All use subject to JSTOR Terms and Conditions