You are on page 1of 30

1212464

research-article2023
ASRXXX10.1177/00031224231212464American Sociological ReviewCha et al.

American Sociological Review

Is the Gender Wage Gap 2023, Vol. 88(6) 972­–1001


© American Sociological
Association 2023
Really a Family Wage Gap DOI: 10.1177/00031224231212464
https://doi.org/10.1177/00031224231212464
journals.sagepub.com/home/asr

in Disguise?

Youngjoo Cha,a,b Kim A. Weeden,c


and Landon Schnabelc

Abstract
Despite large literatures on gender and family wage gaps (e.g., the motherhood wage penalty,
fatherhood wage premium, and the marriage premium) and widespread recognition that the
two gaps are intertwined, the extent and pattern of their relationships are underexplored. Using
data from the 2018 Survey of Income and Program Participation, we show that family wage
gaps are strongly associated with the gender wage gap, as long assumed in the literature, but
with important caveats. The gender-differentiated wage returns to parenthood contribute 29
percent of the gender wage gap. One third of this is associated with occupation, but very little
with other worker and job attributes. The gender-differentiated returns to marriage contribute
another 33 percent, two thirds of which is associated with worker and job attributes but very
little with occupation. However, 36 percent of the gender wage gap is unrelated to these
family wage gaps, and the gender wage gap among childless workers remains substantial.
Moreover, for Black and Hispanic workers, the pattern of association is more complex and
generally weaker than for White workers. These results caution against focusing solely on the
wage gap between “mothers and others” and suggest new directions for research.

Keywords
gender wage gap, family wage gap, motherhood wage penalty, fatherhood wage premium,
gender inequality

In the contemporary United States, as in other “fatherhood wage premium”), although the
advanced industrialized economies, wage dif- latter finding is not as robust (Budig and Eng-
ferentials between men and women, parents land 2001; England et al. 2016; Glauber
and childless adults, and married and unmar- 2008; Jee, Misra, and Murray-Close 2019;
ried workers remain large and consequential Killewald 2013; Lundberg and Rose 2000;
(for a review, see Misra and Murray-Close
2014). On average, employed women earn a
Indiana University
approximately 83 cents for every dollar that b
Yonsei University
men earn, representing an increase of approx- c
Cornell University
imately 5 cents on the dollar over the past 30
years (see England, Levine, and Mishel 2020: Corresponding Author:
Youngjoo Cha, Department of Sociology,
Figure 9). Mothers earn less than childless Indiana University, 771 Ballantine Hall, 1020 E.
women (the “motherhood wage penalty”) and Kirkwood Ave, Bloomington, IN 47405, USA
fathers earn more than childless men (the Email: cha5@indiana.edu
Cha et al. 973

Pal and Waldfogel 2016; Waldfogel 1998). will elaborate, this difference in substantive
Married men typically earn higher wages than focus is manifest in the datasets and modeling
unmarried men, and wage disparities between strategies each research tradition favors.
married and unmarried women vary by The largely separate strands of research on
cohort, race and ethnicity, and life-course gender and family wage gaps leave a critical
stage (Avellar and Smock 2003; Cheng 2016; hole in our knowledge of the relationship
Cohen 2002; Glauber 2007; Juhn and McCue between the two. Estimates of family wage
2017; Killewald and Gough 2013; Korenman gaps based on data from a single cohort or
and Neumark 1991). gender cannot be extrapolated to estimate the
Parental, marital, and gender wage gaps overall contribution of family wage gaps to
are not independent of each other. Logically, the gender wage gap. Inferring this contri-
if mothers are paid less than childless women bution is even more problematic for Black
and fathers are paid more than childless men, and Hispanic workers, given family wage
a gender wage gap will emerge even if child- gaps and the gender wage gap are smaller
less women and childless men are paid equiv- for Black and Hispanic workers than for
alently. Indeed, much of the public discourse non-Hispanic White workers (Cheng 2016;
around the gender wage gap focuses on the Glauber 2007, 2008; Mandel and Semyonov
disparate effects of parenthood, sometimes 2016; Van Winkle and Fasang 2020).
with an explicit claim that “the gender wage The separate strands of research on gender
gap is mostly a penalty for bearing children” and family wage gaps also shift attention
(e.g., Kliff 2018). In the academic literature, away from gender wage disparities among
examinations of the “family wage gap” (i.e., childless or never-married adults. In fam-
wage differentials by marital and parental ily wage gap research, childless women and
status) using single-gender samples are often never-married women are often used as the
justified by their presumed effect on the gen- reference group for within-gender compari-
der wage gap. Similarly, much of the con- sons, and their wages relative to childless or
temporary scholarship on gender wage gaps never-married men are masked. In the gender
focuses on the conflicting demands of work wage gap literature, family demography is
and family, sometimes with the explicit claim rarely incorporated in wage gap models, or
that this conflict is the last major barrier to it is estimated with imperfect measures of
gender parity in wages, particularly among family structure. This leaves a key question
highly educated workers (Goldin 2021). unanswered: Have childless women indeed
Despite widespread recognition that fam- reached wage parity with childless men?
ily wage gaps are consequential for the Although some studies of particular occupa-
gender wage gap, the two phenomena are tions suggest this may be the case (see, e.g.,
typically analyzed separately and with very Correll, Benard, and Paik [2007] on entry-
different goals and analytic strategies. At the level management consultants, or Bertrand,
risk of oversimplification, most research on Goldin, and Katz [2010] on young MBAs
the gender wage gap focuses on identifying from top U.S. business schools), neither fam-
its individual and structural correlates and ily wage gap nor gender wage gap research
how these correlates vary across time and offers a clear answer for the labor force as a
place (e.g., Blau and Kahn 2017; Cha and whole.
Weeden 2014; Goldin 2014; Mandel and Rot- The main contribution of this article is to
man 2021). Most research on family wage describe the empirical relationship between
gaps, by contrast, focuses on identifying the family wage gaps and the aggregate gen-
causal effects of marital or parental status on der gap in wages for the full labor market
wages among workers of the same gender and, where sample sizes allow, for Black,
(e.g., Budig and England 2001; Cheng 2016; Hispanic, and White workers separately. We
England et al. 2016; Killewald 2013). As we first document gender wage differentials
974 American Sociological Review 88(6)

within eight demographic groups defined by for a relatively modest share of the gender
marital and parental statuses (“family status wage gap. This is unsurprising given that
groups”). We then apply a standard wage average education levels among women now
decomposition to estimate the contribution exceed men’s, the college wage premium is
of gender-differentiated “returns” to parental similar for women and men, and job tenure
and marital statuses—that is, the motherhood and continuous labor force experience are
wage penalty, the fatherhood wage premium, converging (Blau and Kahn 2017; England
and men’s marriage premium—to the gender et al. 2020; Mandel and Semyonov 2014). By
wage gap, and describe how much of these contrast, gender differences in work hours,
gender-differentiated returns to parenthood when combined with differential returns to
and marriage are associated with human capi- work hours, widen the gender wage gap (Cha
tal or other pre-market attributes, job attrib- and Weeden 2014; Goldin 2014; Mandel
utes, and occupation. and Semyonov 2014). Similarly, occupational
All analyses use data from the 2018 Sur- segregation and the higher average pay of
vey of Income and Program Participation “men’s” occupations account for a large and
(SIPP), a household survey that is repre- growing share of the gender wage gap, with
sentative of the U.S. labor market. As we will recent estimates ranging from 10 percent in
discuss, the SIPP has important advantages the private sector and 25 percent in the public
in both measurement and generalizability sector (Mandel and Semyonov 2014) to 33
over the single-cohort longitudinal surveys percent overall (Blau and Kahn 2017).
favored in the family wage gap literature and Analyses of the predictors of the gender
the cross-sectional labor force surveys that wage gap rarely explore in depth the con-
dominate the gender wage gap literature. Our tributions of parental or marital status. One
goal is not to provide or test a causal model reason may be that parental and marital sta-
of wages or wage disparities, but to provide a tus do not fit neatly into the human capital
rich description of the empirical relationship theoretical framework that underpins much of
between family wage gaps and the gender the work on gender wage differentials, given
wage gap. Our analyses, although explicitly endogeneity between the two: human capi-
descriptive, bring together the gender and tal (e.g., education, work experience) medi-
family wage gap literatures and provide an ates the association between family status
empirical foundation for new questions into and wages, and family status also mediates
persistent and consequential gender inequali- the association between human capital and
ties in labor market outcomes. wages. Another may be that the household
survey data (e.g., Current Population Survey
[CPS], ACS/Census) on which many analy-
The Gender Wage Gap ses of the gender wage gap in the United
And Family Status States rest do not include complete informa-
In accounting for the gender wage gap, tion about parental status. The upshot is that
researchers often focus on gender differ- the gender wage gap literature offers few
ences in the distribution of and returns to conclusive answers about the strength of the
human capital (as measured by education or contribution of family wage gaps.
labor force experience), occupation, job and
labor force attributes, and other structural
predictors of the gender pay gap (Blau and Family Wage Gaps
Kahn 1997, 2006, 2017; Cha and Weeden And Gender
2014; Cortés and Pan 2019; Goldin and Katz An extensive and largely independent body
2007; Mandel and Rottman 2021; Mandel of research identifies the strength and pattern
and Semyonov 2014). In the United States, of the association between family status and
gender differences in human capital account wage differentials, mostly using within-gender
Cha et al. 975

comparisons (Anderson, Binder, and Krause mothers and by Black or Hispanic mothers
2003; Avellar and Smock 2003; Budig and (Glauber 2007; Pal and Waldfogel 2016).
England 2001; Cheng 2016; England et al. Although there is some evidence that fathers
2016; Glauber 2007; Jee et al. 2019; Kille- earn a wage premium relative to childless
wald 2013; Killewald and Gough 2013; Pal men, this finding is less robust across studies
and Waldfogel 2016; Van Winkle and Fasang than the motherhood wage penalty, and it may
2020; Waldfogel 1996). Studies that focus on be restricted to younger, married, White, and
the association between marriage and wages highly educated fathers of residential chil-
generally show that married men earn higher dren (Glauber 2008; Hodges and Budig 2010;
wages, by 7 to 11 percent, than unmarried Killewald 2013; Killewald and Gough 2013;
men after adjusting for education, work expe- Van Winkle and Fasang 2020).
rience, occupation, and other indicators of A central goal of the family wage gap
labor market position (Budig and Lim 2016; literature is to identify whether these within-
Chun and Lee 2001; Killewald and Gough gender associations between parenthood, mar-
2013; Korenman and Neumark 1991; but on riage, and wages are due to a causal effect of
the role of selection, see Killewald and Lun- marriage or parenthood, to reverse causality
dberg 2017; Ludwig and Brüderl 2018). For wherein higher wages increase the likelihood
women, by contrast, marriage does not have of marriage or parenthood, to wage-relevant
a consistent positive or negative association factors that influence the selection of workers
with wages, with estimates ranging from into marriage or parenthood, or to a spurious
negative or null effects (Anderson et al. 2003; correlation between family status and wages.
Avellar and Smock 2003; Hersch and Stratton This concern with causality is evident in the
2000; Loughran and Zissimopoulos 2009) questions that motivate some of the field’s
to an adjusted marriage premium of 3 to 7 most prominent papers: Is there a motherhood
percent (Budig and England 2001; Budig and wage penalty? (Budig and England 2001;
Lim 2016; Glauber 2007; Juhn and McCue Correll et al. 2007; Waldfogel 1996). Is there
2017; Killewald and Gough 2013; Waldfogel a fatherhood wage premium? (Glauber 2008;
1997). A fair but cautious summary is that the Hodges and Budig 2010; Killewald 2011).
“marriage premium” for women is lower than And what explains the marriage premium?
for men, if it exists at all. (Budig and Lim 2016; Killewald and Gough
A related stream of research focuses on 2013; Ludwig and Brüderl 2018).
the relationship between parental status and Consistent with this focus on causal
wages, again within gender. Mothers earn effects, most contemporary research on the
5 to 7 percent less per hour than childless family wage gap relies on longitudinal studies
women with similar education, experience, of a single cohort for which there are com-
and jobs (Budig and England 2001; Budig prehensive fertility and labor force histories,
and Hodges 2010, 2014; England et al. 2016; and applies models to these data that, under
Glauber 2007; Jee et al. 2019; Pal and Wald- specific assumptions, can plausibly support
fogel 2016; Waldfogel 1997). The size of this causal inference. One common strategy is to
“motherhood wage penalty” varies by mari- apply fixed-effect models to longitudinal data
tal status (Budig and England 2001; Budig to “net out” observed and unobserved time-
and Hodges 2010; Glauber 2007; Pal and invariant differences across workers, includ-
Waldfogel 2016), race (Anderson et al. 2003; ing their human capital (e.g., England et al.
Glauber 2007; Van Winkle and Fasang 2020), 2016; Killewald and Gough 2013). Other
and education (Doren 2019; England et al. strategies include relying on unique samples,
2016). Married, White, and highly educated such of monozygotic twins (Antonovics and
women experience the largest motherhood Town 2004), or focusing on particular groups,
penalty, although it has declined more rapidly such as older grooms or men in “shotgun”
than the penalty experienced by less educated marriages followed immediately by the birth
976 American Sociological Review 88(6)

of a child (Ginther and Zavodny 2001; Kille- al. 2016). And, of course, single-cohort stud-
wald and Lundberg 2017). ies may not generalize to the full labor force.
These efforts show that selection into Audit studies likewise suffer from challenges
marriage of earners with higher wages or to generalizability and to the problem of
higher wage growth potential accounts for “scaling up” evidence of differences in call-
at least some, if not all, of men’s marriage back rates or recommended wages to the total
wage premium (Antonovics and Town 2004; gender wage gap.
Budig and Lim 2016; Cheng 2016; Ginther A related concern is that a focus on the
and Zavodny 2001; Killewald and Lundberg causal effects of marriage or parenthood, at
2017; Korenman and Neumark 1991; Ludwig least as it is estimated in much of this litera-
and Brüderl 2018). The unadjusted father- ture, renders invisible the wage disadvantages
hood wage premium for men also seems to experienced by childless or unmarried adults.
be partly driven by selection (Killewald 2013; Part of the issue is that in within-gender
Van Winkle and Fasang 2020). For women, analyses, childless or unmarried workers are
differential selection appears to be less con- typically treated as the baseline group, both
sequential for the motherhood and marriage by convention and by theory, and accord-
penalties. Instead, the birth or adoption of a ingly tend to “disappear” in discussions of
child leads to reductions in work experience results. More importantly, within-gender
or in work hours and, in turn, lower wages analyses preclude comparing unmarried men
(Budig and England 2001; Cheng 2016; Eng- and unmarried women to each other, or child-
land et al. 2016; Killewald and Gough 2013). less men and childless women to each other.
This is broadly consistent with experimental In short, the family wage gap literature
studies that document differences in job call- provides insight into whether and why moth-
back rates and in subjects’ perceptions of job ers are paid less or fathers are paid more than
candidates depending on whether a fictive childless men and women, but not the extent
job-seeker’s application materials include to which family wage gaps contribute to the
signals of motherhood (e.g., Correll et al. gender wage gap. There are a few notable
2007; Ishizuka 2021). exceptions. Waldfogel (1998) used NLSY79
As informative as they are, efforts to parse data to show that at age 30, 56 percent of the
causal from selection effects of family status total gender gap in the log of hourly wages is
come at the cost of understanding patterns “explained by” the gender-differentiated effects
in the full labor force. In the standard fixed- of marital and parental status on wages. More
effect approach, workers who remain unmar- recently, Juhn and McCue (2017) found that
ried or childless throughout the observation the association of marriage and parenthood
period or who are missing all but one obser- with the gender gap in wages among workers
vation do not contribute to estimates of mar- age 30 and 40 increased between 1968 and
riage or parental status effects. In the National 2014. Another study of Swedish administra-
Longitudinal Survey of Youth (NLSY) 1979, tive data shows that the gender gap in wages
a dataset commonly used in the family wage within couples increases by 10 percentage
gap literature, these exclusions eliminate points within 15 years after the birth of the
about 33 percent of workers (Killewald and first child (Angelov, Johansson, and Lindahl
Lundberg 2017). Because of various data 2016). In each of these studies, restrictions
limitations and decisions to obtain cleaner in the analytic sample (e.g., a single birth
causal estimates, analysts may also exclude cohort, the exclusion of childless adults) or
workers who had a child before entering the in the quantities estimated (e.g., cohort- and
panel (Cheng 2016; Doren 2019; Killewald age-specific effects) offer cleaner causal esti-
and Gough 2013), who remain unmarried mates of marital and parental effects on gen-
until age 45 (Cheng 2016), or who are not der wage gaps, but at the cost of limiting
White (Budig and England 2001; England et coverage to a small slice of the labor force.
Cha et al. 977

Our analysis complements these studies, trad- affecting the denominator in estimates of the
ing away any pretense of causal inference in share of the gender wage gap associated with
exchange for the ability to describe patterns family wage gaps. Racial and ethnic dispari-
for the full labor force. ties in the observed covariates of wages (e.g.,
human capital, job and labor force attrib-
utes, occupation) and their association with
Racial DIFFERENCES In wages may also differ across racial and ethnic
The Gender And Family groups. Rather than gloss over these varia-
Wage Gap Relationship tions in the relationship between family wage
gaps and the gender wage gap, we repeat our
The extant research literature offers even analysis of the entire labor force with analy-
less systematic evidence on the association ses of Black, Hispanic, and White workers
between family wage gaps and the gender alone, where sample sizes allow.
wage gap among workers from racialized
minority backgrounds. We see this as a criti-
cal gap, given that pervasive labor market dis- Data, Variables,
crimination and other structural inequalities And Methods
and adaptations to them fundamentally affect Data
the meanings and experiences of marriage
and parenthood (see Dow 2019), further call- Our analyses are based on data from the
ing into question the validity of generalizing 2018 SIPP, the most recent SIPP panel in
the patterns from the full sample to patterns which data quality was unaffected by budget
for Black and Hispanic workers (see Cheng cuts (as in the 2019 SIPP) or the pandemic.
2016). The SIPP uses an event history calendar to
The few available “family wage gap” stud- collect information about the relationship
ies of Black and Hispanic workers suggest among household members, demography, and
the motherhood wage penalty is smaller for employment and wage data for each month.
Black and Hispanic women than for White To minimize recall bias, we use event data
women (Budig and England 1991; Eng- from the month preceding the first wave of
land et al. 2016; Glauber 2007; Waldfogel the SIPP panel, meaning the reference period
1997), and the fatherhood wage premium is is December 2017.1
less robust for Black and Hispanic fathers For our goals, the SIPP offers important
(Glauber 2008; Hodges and Budig 2010; Van benefits over more commonly used datasets
Winkle and Fasang 2020). The wage boost in the gender and family wage gap literatures.
associated with marriage is smaller for Black First, it allows researchers to identify parents
men than for White men (Kilbourne et al. of grown or nonresidential children. This
1994; Korenman and Neumark 1991), and the avoids bias in estimates of the within-group
wage penalty associated with marriage may wage gap from including these parents in the
be exclusive to White women (Cheng 2016). “childless adult” category. Second, it includes
Given the family wage gaps among Black a direct measure of work experience, although
and Hispanic workers are smaller than among in the 2018 SIPP panel this measure is limited
White workers, we might also expect fam- to tenure with the current employer.2 Third,
ily wage gaps to have a weaker association it is representative of the entire labor force
with (race-specific) gender wage gaps among and includes fertility histories that cover all
Black and Hispanic workers. However, this children, instead of just those born or entering
straightforward story is complicated by the the household after the respondent enters the
fact that the gender wage gap is also smaller longitudinal panel, as in most family wage
in these racial and ethnic groups (Budig, Lim, gap research that relies on fixed-effects mod-
and Hodges 2021; Hegewisch and DuMon- els applied to panel data. Finally, the SIPP
thier 2016; Mandel and Semyonov 2016), sample is large enough to support models
978 American Sociological Review 88(6)

that adjust for detailed occupations and to fit wages fall below $2 or above $250, under the
the more parsimonious models to the Black, assumption that these reflect errors in reports
Hispanic, and White subsamples. The full of wages or of hours worked (Blau et al.
sample includes Asian and “other race” work- 2021).4 With top-coding and this trimming,
ers, but there are too few respondents from women in our analytic sample earn an aver-
these groups to obtain reliable estimates in age of $24.31 per hour and men $29.68 per
race-stratified samples. hour (see Table 1). This yields an unadjusted
Our analytic sample includes wage and gender wage ratio of approximately 82 cents
salary workers age 18 to 67 who have non- on the dollar, consistent with estimates from
zero earnings. This age range captures most other data sources.
of the working population and acknowledges The key variables in our analyses are gen-
that the wage effects of marriage and mother- der, marital status, and parental status. Gender
hood can linger well beyond the departure is indicated with a binary variable in which
of adult children from the home (Abendroth, the reference category is men. Marital status
Huffman, and Treas 2014; Cheng 2016; Van is coded into four categories: currently mar-
Winkle and Fasang 2020). Note 8 provides ried; separated, divorced, or widowed but not
additional results after restricting the sample currently married (“previously married”); cur-
to adults of prime working age (i.e., age 25 rently cohabiting; and never married, the ref-
to 54). The final analytic sample includes erence category. Cohabitation is indicated by
23,765 workers. We weight the data by the SIPP’s household relationship question, from
final person weight provided by the SIPP. which we identified respondents who have
unmarried romantic partners currently living in
the same household. By necessity, respondents
Variables
who previously cohabited but are currently
Our measure of the gender wage gap is based neither cohabiting nor married are included
on men’s and women’s hourly rate of pre-tax in the “never married” category. In the SIPP,
wage and salary income, excluding income a greater proportion of men (.55) than women
from self-employment or investments.3 In (.50) were married, a greater proportion of
SIPP 2018, respondents were allowed to women (.14) than men (.08) were previously
report wages as they prefer: hourly, bi-weekly, married but not currently married, and approx-
weekly, monthly, bi-monthly, annually, or as imately equal proportions of women and men
the gross amount from W-2 forms. Based were cohabiting (.11) (see Table 1).
on this information, the SIPP data provid- The proportions of workers in the four mari-
ers calculated weekly earnings. For workers tal status groups differ by race (see Table 2).
who did not report hourly wages directly, we Black workers and, to a lesser extent, His-
calculate hourly wages from weekly earnings panic workers are less likely to be married
and the number of usual hours worked per than White workers. Gender differences in
week at the main job. For a small percentage the distribution of marital status are also
of respondents whose usual work hours are larger among Black and Hispanic workers
missing, we use the number of hours worked than among White workers. The proportion of
in the week preceding the survey. married individuals is .42 for Black men and
To preserve confidentiality, the SIPP top- .32 for Black women, and .51 for Hispanic
codes wage and income data, meaning earn- men and .44 for Hispanic women, compared
ings above a threshold are replaced with to .57 for White men and .55 for White
the mean (for hourly wages) or median (for women. The proportion of never-married
all other earnings) among the demographic individuals is .34 for Black men and .42 for
group to which high-earning respondents Black women, about .29 for Hispanic men
belong. We exclude 463 workers, or 1.9 per- and Hispanic women, and .24 for White men
cent of the analytic sample, whose hourly and .21 for White women.
Cha et al. 979

Table 1. Means and Standard Errors of


Parental status is indicated by a binary
Variables Used in the Analysis by Gender:
All Workers variable that identifies respondents who have
children of any age, relationship (biological,
Variable Men Women step, or adoptive), and residential status.5 We
Hourly wages (logged) 3.101 2.925 construct the parental status measure from
(.007) (.006) three sources in the SIPP: a parent screener
Hourly wages (2017 US$) 29.681 24.305 variable; the household roster data, which link
(.302) (.205) parents and all residential children including
Parent (reference: “never had .603 .653 non-biological children; and fertility history
child”)
data, which identify all biological children
Marital status
and their birth year regardless of their resi-
Never married (reference) .265 .255
Married .548 .500
dential status.6
Previously married .075 .137 Differences in parental status across the
Cohabiting .112 .108 Black, Hispanic, and White samples are more
Age 40.538 40.793 modest than differences in marital status (see
(.090) (.085) Table 2). The gender gap in parental status
Race is also similar across all groups of workers,
White (reference) .621 .616 with the proportion of women who are par-
Black .109 .136 ents exceeding the proportion of men who
Hispanic .184 .162
are parents by 4 to 6 percentage points. The
Asian .062 .058
weighted percentages of the parental status
Other .024 .028
Education
groups and their unweighted counts are pro-
Less than high school .085 .054 vided in Table S1 in the online supplement.
High school graduate .279 .223 In some analyses, we adjust for common
(reference) covariates of wages: age and its square, educa-
Some college .276 .303 tion (less than high school, high school gradu-
College graduate .231 .256 ate, some college, college graduate, advanced
Advanced degree .129 .164 degree), years of job tenure and its square,
Job tenure (years) 7.499 7.102 weekly work hours (under 35, 35 to 49, 50 or
(.070) (.073)
more), sector (public or private), region (East,
Weekly work hours
Less than 35 hours .127 .257
Midwest, South, and West), metropolitan sta-
35 to 49 hours (reference) .660 .642 tus, and union membership or contract cover-
50 or more .214 .101 age. In analyses based on the full sample, we
Public sector .137 .187 also adjust for race (White, Black, Hispanic,
Region Asian, and other race). Additional models fit
East (reference) .181 .185 a full set of dummy variables for 510 occupa-
Midwest .216 .223 tions, coded into the 2010 Census Occupation
South .361 .368 Classification scheme. Descriptive statistics
West .242 .224 for these covariates, with the exception of
Metropolitan .874 .872
occupation, are provided in Table 1 for all
Union .146 .133
workers and Table S2 in the online supple-
Occupations 510 Occupations
N 12,079 11,686
ment for Black, Hispanic, and White workers.

Source: SIPP 2018 Panel.


Note: All data are weighted using the final Methods
weight provided by the SIPP. Standard errors are
estimated using Fay’s modified balanced repeated We first estimate unadjusted and adjusted gen-
replication (BRR) method, using replicate der gaps in wages using estimates from OLS
weights provided by the SIPP. regressions in which gender, marital status,
980 American Sociological Review 88(6)

Table 2. Means and Standard Errors of Hourly Wages and Parental and Marital Status
Variables for Black, Hispanic, and White Workers

Black Hispanic White

Variable Men Women Men Women Men Women


Hourly wages (logged) 2.903 2.832 2.867 2.729 3.187 2.982
(.021) (.017) (.015) (.016) (.009) (.008)
Hourly wages (2017 US$) 24.395 22.603 22.653 19.377 31.889 25.401
(.855) (.613) (.582) (.479) (.367) (.264)
Parent (reference: “never had .622 .679 .631 .690 .596 .640
child”)
Marital status
Never married (reference) .344 .419 .296 .293 .240 .209
Married .420 .318 .513 .440 .571 .547
Previously married .095 .169 .062 .162 .081 .130
Cohabiting .141 .095 .128 .105 .108 .114
N 1,193 1,498 2,261 2,082 7,583 7,125

Source: SIPP 2018 Panel.


Note: All data are weighted using the final weight provided by the SIPP. Standard errors are in
parentheses and estimated using Fay’s modified balanced repeated replication (BRR) method, using
replicate weights provided by the SIPP.

and parental status fully interact.7 Model 1 (i.e., occupational segregation) contributes a
does not include any adjustment variables. It large and growing share of the gender wage
allows us to estimate “raw” gender wage gaps gap (e.g., Blau and Kahn 2017; Mandel and
across the family status groups—including Semyonov 2014). Comparison of Models 2
never-married adults and childless adults— and 3 highlights if, and in what patterns,
and the total contribution of family wage gaps occupational segregation by family status
to the gender wage gap. Model 2 adjusts for contributes to gender wage gaps. Second,
worker and job attributes that are commonly and more practically, the Black and Hispanic
included in family and gender wage gap mod- subsamples are too sparse to be confident in
els (age, race, education, job tenure, work results from occupation-adjusted models. To
hours, sector, region, metropolitan status, be able to compare identically specified mod-
and union membership). We combine these els across the race-stratified and full samples,
attributes into one model to keep the volume we need to separate occupation from the other
of results more manageable, maintain focus covariates in the models fit to the full sample.
on the family wage coefficients, and avoid In estimating Models 2 and 3, we do not
the need to make tacit assumptions about the mean to imply that the adjusted models get
causal order among covariates (see note 9 for at the “true” association between gender or
the results of an interim model that includes family status and wages. After all, many of
only “pre-market” factors). the covariates in Models 2 and 3 are not clear
Model 3 adds indicators of detailed occu- antecedents of marital and parental status and
pation. We separate occupation from the are better understood as mediators of the rela-
other covariates in the adjustment models tionship between pre-market factors, family
for two reasons. First, although occupation status, and wages (Killewald 2013; Lundberg
does not feature especially prominently in the and Rose 2002; Petersen and Morgan 1995).
family wage literature, recent gender wage Instead, we fit these models to put the magni-
gap research shows that the uneven distribu- tude of family status contributions to gender
tion of men and women across occupations wage gaps in context, and to reveal how
Cha et al. 981

commonly estimated individual and struc- a result, the endowment effects are more
tural covariates of wages in gender wage gap substantial, and we present them in addition
research (e.g., Blau and Kahn 2017; Cha and to the coefficient effects.
Weeden 2014) are embedded in family demo- The main decomposition results use the
graphic processes in which they confound or men’s equation as the baseline. The coef-
mediate the association between marriage and ficient effects thus represent the expected
parenthood and wages, or both. change in women’s mean of log hourly wages
Our second analysis decomposes the gen- if women had men’s coefficients, and the
der wage gap, estimated from separate wage endowment effects represent the expected
equations for men and women, into the por- change in women’s mean of log hourly wages
tion of the mean gender wage gap associ- if women had men’s mean values on the
ated with gender differences in the means predictors. We also estimated decomposition
of the predictors (“endowment effects,” in results using the women’s equation as the
common decomposition terminology), the baseline. This yields substantively identical
portion associated with gender differences results for all workers and for Hispanic work-
in coefficients of predictors (“coefficient ers, but a slightly larger endowment effect of
effects”), the portion associated with simulta- marriage for Black workers.
neous differences in endowments and coeffi-
cients (“interaction effects”), and the residual
(Blinder 1973; Jann 2008; Oaxaca 1973; see Results
also Kitagawa 1955). These wage equations How Do Gender Wage Gaps
only fit the main effects of marital and paren- Differ across the Family
tal status, given most coefficients for the Status Groups?
marriage-parental interaction effects in the
decomposition model are not significant. In SIPP 2018, not surprisingly, the gender
We are primarily interested in the coef- gap in unadjusted (logged) hourly wages
ficient effects of marital and parental status, among all workers is larger among parents
which show the portion of the gender wage than among childless adults. Specifically, the
gap associated with gender-differentiated ratio of women’s mean wages to men’s is .77
wage returns to family status. The coefficient among parents, corresponding to a 23 percent
effects for categorical variables are sensitive wage gap, but only .93 among childless adults,
to the choice of the reference group. One solu- corresponding to a 7 percent wage gap (not
tion is to transform the variables to express shown). The story becomes more complex
them as deviation from the grand mean (Jann when we examine the unadjusted gender wage
2008; Yun 2005). Compared to the results gaps by eight family status groups defined by
using untransformed variables, this solution the cross-classification of family and marital
generates substantively similar patterns (see status, as shown in Figure 1a. Among child-
Table S3 in the online supplement). However, less workers (see left panel of Figure 1a), the
we can no longer interpret the coefficients in wage gaps between men and women who are
relation to a marriage premium, a motherhood cohabiting or who were previously married are
wage penalty, or a fatherhood wage premium. close to zero, and the standard errors are too
Given the similarities of the results, we pre- large to reject the implicit null of no gender
sent results using untransformed variables. wage gap. Among never-married childless
In the full sample and White subsample, adults, the gender wage gap is estimated at
similar family status distributions by gender 6 percent, or an earnings ratio of .94 (e (2.701
translate into very small endowment effects – 2.768)
= .94), which is statistically significant
of marital and parental status on gender wage at conventional thresholds. Among married
gaps. However, in the Black and Hispanic childless adults, the gender wage gap is more
subsamples, men and women have quite dif- than twice as large at 15 percent, or a gender
ferent distributions across marital status. As wage ratio of .85.
982 American Sociological Review 88(6)

(a) Model 1a childless adults parents


2.768 2.738

never married 2.701 2.669

3.216 3.320

married 3.059 3.065

men
3.019 3.080 women

previously married 3.033 2.899

2.977 2.895

cohabiting 2.895 2.779

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

(b) Model 2b
childless adults parents
2.955 2.965

never married 2.852 2.864

3.124 3.215

married 2.940 2.958

men
3.022 3.091 women

previously married 2.943 2.883

3.046 3.070

cohabiting 2.914 2.897

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

(c) Model 3c
childless adults parents
2.966 2.969

never married 2.902 2.921

3.078 3.164

married 2.946 2.976

men
3.012 3.069 women

previously married 2.977 2.925

3.022 3.042

cohabiting 2.955 2.934

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

Figure 1. Log of Hourly Wages by Gender and Family Status, Full Sample
Source: SIPP 2018 Panel.
Note: N = 23,765. Figures were created using Stata graphic schemes developed by Jann (2014, 2018).
a
Model 1 does not include any covariates.
b
Model 2 adjusts for age and its square, race, education, years of job tenure and its square, work hours,
sector, region, metropolitan status, and union status.
c
Model 3 adjusts for detailed occupations in addition to covariates included in Model 2.
Cha et al. 983

Gender wage gaps among parents also one quarter of the adjusted gender wage gap
vary by parents’ marital status, as shown in among married parents is associated with
the right-hand panel of Figure 1a. Married occupational segregation by gender. With the
parents show the largest gap, at 23 percent exception of cohabitors, all parent groups
(i.e., a women-to-men wage ratio of .77). Pre- show a larger decline in the net gender wage
viously married and cohabiting parents also gap from adjusting for occupation than do the
show sizable gender wage gaps of 17 and 11 childless adult groups with the same marital
percent, respectively. Never-married parents, status.
by contrast, show no significant gender wage Notably, the residual gender wage gap
gap. The gender wage gap among cohabiting within family status groups remains substan-
parents (11 percent) is not significantly differ- tial after adjusting for occupations: 17 percent
ent from—and, potentially, smaller than—the among married parents, 13 percent among
gender wage gap among married childless previously married parents, 10 percent among
adults (15 percent). cohabiting parents, and 12 percent among
Adjusting for demographic and human married childless adults. In all eight family
capital factors (age, race, education, job ten- status groups, over 70 percent of the total
ure) and job attributes (work hours, public gender wage gap remains “unexplained” in
or private sector, region, and union) reduces the occupation adjusted models. This residual
the observed differences in average wages gender wage gap could, of course, be tied to
across family status groups, as shown by the unobserved covariates (e.g., firm size, worker
contrast of Figure 1a to Figure 1b. However, commitment) or to unobserved processes
it does little to reduce gender wage gaps (e.g., employer discrimination) that are asso-
within the family status groups (see also Blau ciated with gender, family status, and wages
and Kahn 2017). Indeed, for most family sta- (see, e.g., Correll et al. 2007; Stainback and
tus groups, the estimated gender wage gaps Tomaskovic-Devey 2012).
increase. For example, among never-married
childless adults, the unadjusted model shows
Decomposition Results
a 6 percent gender wage gap (see Figure 1a),
but adjusting for worker and job attributes How do the gender-differentiated marriage
expands this to a 10 percent gap (see Figure premium, the motherhood wage penalty, and
1b). Among never-married parents, the gap the fatherhood wage premium contribute to
increases from a non-significant 7 percent the overall gender wage gap? Table 3 shows
gap to a significant 10 percent gap. For most decomposition results derived from an unad-
other family status groups, the increase is 2 to justed model (Model 1); a model that adjusts
5 percentage points. The larger observed gaps for worker education, experience, work hours,
in the adjusted model are attributable to wom- and job and labor market attributes (Model 2);
en’s greater likelihood of earning at least a and a fully adjusted model that adds occupa-
baccalaureate degree, and the higher wages of tion dummies (Model 3). The regression coef-
degree-holders relative to non-degree-holders ficients used to produce these decomposition
(see also note 9). results are shown in Appendix Table A1.
Adjusting for occupation, by contrast, Model 1 in Table 3 shows that gender dif-
reduces observed gender wage gaps for all ferences in the wages of parents relative to
family status groups relative to the human those of childless adults account for .051 log
capital and job attribute model (compare points, or about 29 percent, of the total gap
Figures 1b and 1c). The greatest effect is of .176 log points observed in the unadjusted
seen among married parents, where adjusting model. Put differently, if mothers’ wages rela-
for occupation reduces the estimated gender tive to those of childless women were the
wage gap by 6 percentage points, from 23 same as fathers’ wages relative to childless
to 17 percent. In other words, approximately men, the total gender gap in wages would be
984 American Sociological Review 88(6)

Table 3. Wage Decomposition: Estimated Contributions of Family Wage Gaps to the Gender
Wage Gap for All Workers

Model 1a Model 2b Model 3c

Percent of Percent of Percent of


Contribution Total Gap Contribution Total Gap Contribution Total Gap
Total gap = .176
Parent .051** 29.1 .052** 29.5 .034* 19.3
(.015) (.014) (.014)
Married .058** 32.9 .033** 18.6 .031* 17.6
(.013) (.012) (.012)
Previously .004 2.2 .003 1.6 .002 .9
married (.005) (.005) (.005)
Cohabiting <.001 .1 .001 .3 –.001 –.4
(.004) (.004) (.003)

Source: SIPP 2018 Panel.


Note: N = 23,765. The “contribution” columns show the contribution to the gender gap in mean wages
of gender differences in coefficients associated with marital and parental status. See Appendix Table A1
for regression coefficients and Table S4 in the online supplement for the full decomposition results.
a
Model 1 fits only parental status and marital status without any other covariates.
b
Model 2 adjusts for all covariates except for occupations (see Table 1).
c
Model 3 adds dummy variables for 510 detailed occupations to Model 2.
*p < .05; **p < .01 (two-tailed test).

29 percent smaller. The higher wage returns the total marriage premium effect is associ-
to marriage for men than for women account ated with gender differences in human capital
for approximately 33 percent of the total and job attributes between married and never-
gender wage gap. Gender differences in the married workers (i.e., [.058 – .33]/.058 = .43).
relative wages of previously married work- The large attenuation likely reflects selectivity
ers and cohabiting workers (both relative to into marriage based on factors associated with
never-married workers) are negligible and wages (see Killewald and Lundberg 2017;
noisy, together contributing only about 2 per- Ludwig and Brüderl 2018). By contrast,
cent of the total gender wage gap. adjusting for these covariates has virtually
Taken together, these results suggest that no effect on the percent of the gender wage
approximately 62 percent of the total gen- gap associated with gender-specific returns
der wage gap is associated with gender- to parenthood. This result is anticipated by
differentiated wage “returns” to marriage or recent gender wage gap research that shows
parenthood (29.1 + 32.9 = 62 percent). only a modest share of the gender wage
Conversely, approximately 36 percent of the gap is associated with gender differences in
gender wage gap is unrelated to gender-dif- education and other measures of human capi-
ferentiated wage returns to any type of family tal (e.g., Blau and Kahn 2017; Mandel and
status (i.e., 100 – [29.1 + 32.9 + 2.2 + .1] = Semyonov 2014).9
35.7 percent).8 Adjusting for occupation (Model 3) has a
Adjusting for human capital and job attrib- more substantial effect on the contribution of
utes (but not occupation) decreases the contri- the gender-specific wage returns to parent-
bution of gender-differentiated wage returns hood, but not marital status. Specifically,
to marriage to 19 percent of the total gender the contribution of parental status declines
wage gap, compared to 33 in the unadjusted from 30 percent in Model 2 (human capital
model. In other words, a little under half of and job attributes) to 19 percent in Model
Cha et al. 985

3 (occupation; see Table 3). The contribu- wage gaps among Black (Figure 2) and His-
tion of marriage, however, remains relatively panic (Figure 3) workers are smaller than
stable at around 18 percent. Put differently, those among White workers (Figure 4) or all
approximately one third of the total parental workers (Figure 1). Similarly, the unadjusted
status effect is associated with occupational gender wage gaps are smaller among Black
segregation by gender and parental status, but parents (Figure 2a) and Hispanic parents (Fig-
virtually none of the marital status effect. We ure 3a) than among White parents (Figure 4a;
identify possible reasons behind the different compare also Model 1 for the three groups in
patterns for parenthood and marriage in the Table 4; see also Glauber 2007, 2008). These
Discussion section. racial differences are especially pronounced
To put the magnitude of the family wage for married parents, where a wage gap of 28
gap contributions in context, it is helpful to percent (a wage ratio of .72) among married
compare them to other covariates of wages. In White parents is reduced to 7 percent (a wage
Model 2, the endowment effect of education ratio of .93) with overlapping 95 percent
(i.e., the overrepresentation of women among confidence intervals among married Black
college graduates) is –.044 log points, or 25 parents, and to 16 percent (a wage ratio of
percent of the total wage gap (see Table S4 in .84) among married Hispanic parents. Child-
the online supplement). This is smaller than less Black workers (of any marital status)
the wage gap-exacerbating effect of parental also show smaller gender wage gaps than do
status (.052 log points), meaning the gender- childless White workers and Black parents.
differentiated returns to parenthood entirely The SIPP data also reveal racial hetero-
offset women’s educational advantages. The geneity in the gender wage gaps by marital
effect of gender-differentiated wage returns status. Never-married and childless Black
to marriage (.033) is about 75 percent of workers have a 4 percent gender wage gap
the size of the education endowment effect with overlapping confidence intervals, com-
(–.044). In Model 3, the coefficient effects pared to a 5 percent wage gap observed for
of parental and marital status are the largest never-married and childless White workers
observable sources of the gender wage gap (see Table 4, also compare Figures 2a and
except occupation. 4a). Married and childless Black workers
have a 3 percent gender wage gap, compared
to a 14 percent wage gap observed for White
Racial Differences in the Family and
workers. However, childless Hispanic work-
the Gender Wage Gap Relationship
ers have similar gender wage gaps as child-
In Figures 2, 3, and 4, we present results from less White workers.
models fit to the subsamples of non-Hispanic As was the case for all workers (Figure 1b),
Black (Figure 2), Hispanic (Figure 3), and adjusting for worker and job attributes com-
White (Figure 4) workers. To ease compari- presses wage gaps between the marital and
sons, Table 4 presents the gender wage ratio parental status groups for Black (Figure 2b),
within family status groups for each race sub- Hispanic (Figure 3b), and White (Figure 4b)
sample, where this ratio is calculated from the workers, but it does not reduce the gender
same coefficients as in the figures. Because wage gaps within family status groups. In
the samples of Black and Hispanic workers are fact, for many family status groups, covari-
too small to estimate Model 3, which includes ate adjustment increases the estimated gen-
detailed occupation effects, we do not show der wage gaps. This is most pronounced in
these results. Smaller sample sizes may also the Hispanic subsample, where adjustment
account for the larger standard errors in models results in a 5 percentage-point increase (from
fit to the Black and Hispanic samples. 16 to 21 percent) in the wage gap among mar-
What can we learn from the race-specific ried parents and a 6 percentage-point increase
models? As prior research has shown, gender (from 22 to 28 percent) among previously
986 American Sociological Review 88(6)

(a) Model 1a
childless adults parents
2.698 2.599

never married 2.663 2.646

3.065 3.118

married 3.038 3.045

men
3.057 2.927 women

previously married 3.124 2.847

2.898 2.777

cohabiting 2.930 2.732

2.62.72.82.93.03.13.23.33.4 2.62.72.82.93.03.13.23.33.4

(b) Model 2 b

childless adults parents


2.846 2.700

never married 2.766 2.740

2.989 3.008

married 2.907 2.923

men
3.113 2.906 women

previously married 2.918 2.784

2.953 2.862

cohabiting 2.952 2.766

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

Figure 2. Log of Hourly Wage by Family Status for Black Men and Women
Source: SIPP 2018 Panel.
Note: N = 2,691. Figures were created using Stata graphic schemes developed by Jann (2014, 2018).
a
Model 1 does not include any covariates.
b
Model 2 adjusts for age and its square, race, education, years of job tenure and its square, work hours,
sector, region, metropolitan status, and union status.

married childless adults. Among Black work- the gender wage gap, the unadjusted coef-
ers, adjusting for covariates also increases the ficient effects of parenthood and marriage are
estimated gender wage gaps, but by a smaller smaller for Black and Hispanic workers than
amount. In all family status groups, the con- for White workers or all workers (shown in
fidence intervals for Black men and women Table 3). Specifically, gender differences in
overlap in the covariate-adjusted models. the wage returns to parenthood in the unad-
Table 5 shows substantial variations justed model (Model 1) range from .005 log
among Black, Hispanic, and White work- points, or about 7 percent of the total gender
ers in the contributions of family wage gaps wage gap (.071 log points), among Black
to the (race-specific) gender wage gaps. In workers and .014 log points, or about 10 per-
both absolute terms and as a percentage of cent of the gender wage gap (.138 log points),
Cha et al. 987

(a) Model 1a
childless adults parents
2.693 2.872

never married 2.594 2.628

3.004 2.960

married 2.852 2.783

men
3.076 2.962 women

previously married 2.822 2.763

2.843 2.735

cohabiting 2.784 2.725

2.62.72.82.93.03.13.23.33.4 2.62.72.82.93.03.13.23.33.4

(b) Model 2 b
childless adults parents
2.802 3.007

never married 2.666 2.680

2.926 2.928

married 2.744 2.695

men
3.049 2.948 women

previously married 2.717 2.703

2.886 2.842

cohabiting 2.775 2.757

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

Figure 3. Log of Hourly Wages by Family Status for Hispanic Men and Women
Source: SIPP 2018 Panel.
Note: N = 4,343. Figures were created using Stata graphic schemes developed by Jann (2014, 2018).
a
Model 1 does not include any covariates.
b
Model 2 adjusts for age and its square, race, education, years of job tenure and its square, work hours,
sector, region, metropolitan status, and union status.

among Hispanic workers, to an astonishing the gender differences in the returns to being
.098 log points, or half of the gender wage previously married (instead of never mar-
gap (.205 log points), for White workers. The ried) may be slightly larger among Black and
estimates for Black and Hispanic workers are Hispanic workers than among White workers,
not statistically significant. Similarly, gender- but again these estimates are not significant.
differentiated returns to marriage contribute The general pattern is that family wage gaps,
25.4 percent of the gender wage gap for Black and in particular the motherhood wage pen-
workers and 16.2 percent for Hispanic work- alty and fatherhood wage premium, contrib-
ers compared to 34.3 percent for White work- ute smaller shares to gender wage gaps for
ers, but the former two percentages are based Black and Hispanic workers than for White
on imprecise estimates. The coefficients for workers.
988 American Sociological Review 88(6)

(a) Model 1a
childless adults parents
2.774 2.809

never married 2.722 2.736

3.239 3.440

married 3.087 3.113

men
2.983 3.151 women

previously married 3.035 2.959

2.994 3.018

cohabiting 2.897 2.819

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

(b) Model 2b
childless adults parents
2.992 3.026

never married 2.914 2.914

3.176 3.320

married 3.015 3.022

men
2.995 3.155 women

previously married 2.982 2.924

3.085 3.162

cohabiting 2.955 2.948

2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4

Figure 4. Log of Hourly Wages by Family Status for White Men and Women
Source: SIPP 2018 Panel.
Note: N = 14,708. Figures were created using Stata graphic schemes developed by Jann (2014, 2018).
a
Model 1 does not include any covariates.
b
Model 2 adjusts for age and its square, race, education, years of job tenure and its square, work hours,
sector, region, metropolitan status, and union status.

At the same time, Black and Hispanic capital and job attributes reduces this endow-
workers reap more substantial endowment ment effect of marriage by more than half, to
effects of family status on the gender wage .018 log points or 25 percent of the gender
gaps than do White workers (see Table 5). wage gap (see Panel A in Model 2 of Table 5).
In the unadjusted model, the higher rate of We see a similar pattern for Hispanic work-
marriage among Black men (42 percent) than ers, for whom gender gaps in marital status
Black women (32 percent) is associated with are larger than for White workers but smaller
.042 log points, or 60 percent of the (race- than for Black workers (see Table 2): among
specific) gender wage gap (see Panel A in Hispanic workers, the endowment effect of
Model 1 of Table 5). Adjusting for human marriage contributes to .015 log points, or
Cha et al. 989

Table 4. Ratio of Women’s to Men’s Hourly Wages for Black, Hispanic, White,
and All Workers

Black Hispanic White

Model 1a Model 2b Model 1 Model 2 Model 1 Model 2


Childless adults
Never married .96 .92 .91 .87 .95 .92
Married .97 .92 .86 .83 .86 .85
Previously married 1.07 .82 .78 .72 1.05 .99
Cohabiting 1.03 1.00 .94 .89 .91 .88
Parents
Never married 1.05 1.04 .78 .72 .93 .89
Married .93 .92 .84 .79 .72 .74
Previously married .92 .88 .82 .78 .83 .79
Cohabiting .96 .91 .99 .92 .82 .81

Source: SIPP 2018 Panel.


Note: N = 2,691 Black workers, 4,343 Hispanic workers, and 14,708 White workers. The ratios are
calculated by exponentiating the gender gaps in log of hourly earnings.
a
Model 1 does not include any covariates.
b
Model 2 adjusts for age and its square, education, years of job tenure and its squared term, work hours,
sector, region, metropolitan status, and union status.

about 11 percent of the total gender wage cohabitation. However, systematic evidence
gap in the unadjusted model, but becomes a of the contribution of family wage gaps to the
trivial (and non-significant) .001 log points, gender wage gap is largely lacking, especially
or .9 percent, in the adjusted model. The for the full labor force. Our core contribution
endowment effect of parental status is small is to provide this evidence.
and imprecise in both models for Black and The top-line result is that approximately
Hispanic workers, as anticipated by the small 64 percent of the overall gender wage gap
gender gaps in parenthood for both groups is associated with gender-specific wage dif-
(see Table 2). For White workers, the endow- ferences between family status groups, with
ment effects of both parental and marital most of this driven by gender-specific wage
status are negligible in all models. As we will differences between married and never-mar-
discuss, the implication is that the relation- ried workers (33 percent) and between parents
ship between family status and the gender gap and childless adults (29 percent). In models
in wages among Black and Hispanic workers that adjust for a host of worker demographic
is weaker than it is for White workers, and and human capital attributes, job attributes,
driven more by the gender-differentiated dis- and occupation, family wage gaps are the
tribution of marriage than gender-differenti- second-largest contributors to gender wage
ated wage returns to parenthood or marriage. gaps, behind occupation, and contribute 37
percent of the total gender wage gap.
The decomposition analysis shows that
Discussion approximately 40 percent of the association
It has become something of a truism in aca- between family wage gaps and the gender
demic and popular press reports of the gender wage gap is correlated with gender differ-
gap in wages that “work is greedy, and fami- ences in human capital, work hours, sector,
lies are needy” (Folbre 2022), a discourse that occupation, and other labor market and job
locates remaining wage disparities between attributes. However, the pattern of these
men and women in the gendered effects of associations differs between marital and
becoming a parent or entering marriage or parental wage effects: the contribution of
990 American Sociological Review 88(6)

Table 5. Wage Decomposition: Estimated Contributions of Family Status to the Gender Wage
Gap for Black, Hispanic, and White Workers

Model 1a Model 2b

Percent of Percent of
Contribution Total Gap Contribution Total Gap
A. Black Workers (total gap = .071)
Coefficient effects
  Parent .005 7.3 –.013 –18.4
(.047) (.046)
  Married .018 25.4 .025 34.9
(.023) (.022)
  Previously married .008 11.3 .020 28.3
(.016) (.016)
  Cohabiting .002 2.3 .006 8.0
(.009) (.009)
Endowment effects
  Parent .004 5.1 .003 4.9
(.003) (.003)
  Married .042** 59.7 .018** 25.3
(.007) (.005)
  Previously married –.019** –27.0 –.004 –6.2
(.005) (.003)
  Cohabiting .007* 10.0 .003 4.8
(.003) (.003)
B. Hispanic Workers (total gap = .138)
Coefficient effects
  Parent .014 9.9 .044 32.1
(.047) (.044)
  Married .022 16.2 .034 24.4
(.028) (.028)
  Previously married .015 10.5 .018 12.7
(.015) (.014)
  Cohabiting –.010 –7.3 –.008 –6.0
(.010) (.009)
Endowment effects
  Parent .001 1.0 .001 1.0
(.002) (.002)
  Married .015** 10.7 .001 .9
(.003) (.003)
  Previously married –.018** –12.8 –.002 –1.8
(.005) (.005)
  Cohabiting .003 2.5 .002 1.5
(.002) (.002)
C. White Workers (total gap = .205)
Coefficient effects
  Parent .098** 47.7 .079** 38.8
(.018) (.017)
  Married .070** 34.3 .040* 19.7
(.019) (.018)
  Previously married –.002 1.2 –.001 –.4
(.006) (.006)
  Cohabiting .004 1.8 .002 1.1
(.005) (.004)

(continued)
Cha et al. 991

Table 5. (continued)

Model 1a Model 2b

Percent of Percent of
Contribution Total Gap Contribution Total Gap
Endowment effects
  Parent <.001 .1 <.001 .1
(.001) (.001)
  Married .009** 4.5 .003** 1.4
(.002) (.001)
  Previously married –.012** –6.0 –.002 –.8
(.002) (.002)
  Cohabiting –.001 –.4 <.001 –.1
(.001) (<.001)

Source: SIPP 2018 Panel.


Note: N = 2,691 Black workers, 4,343 Hispanic workers, and 14,708 White workers. The “contribution”
columns show the contribution to the gender gap in mean log wages of gender differences in
coefficients associated with marital and parental status (“coefficient effects”) and gender differences
in compositions of marital and parental status (“endowment effects”). See Appendix Tables A2, A3,
and A4 for regression coefficients and Tables S5, S6, and S7 in the online supplement for the full
decomposition results.
a
Model 1 fits only parental status and marital status without any other covariates.
b
Model 2 adjusts for all covariates except for occupations (see the full list in Table 1).
*p < .05; **p < .01 (two-tailed test).

gender-differentiated wage returns to mar- marital wage gaps and occupation implies that
riage declines from 33 to 19 percent after “vertical” occupational segregation—that is,
adjusting for demographic, human capital, and segregation across occupations that system-
job attributes, but it remains unchanged after atically differ in pay—between married and
further adjusting for occupation, whereas the never-married men and women is no more
contribution of gender-differentiated wage pronounced than segregation between all men
returns to parenthood is unaffected by adjust- and women. This could reflect the wither-
ing for demographic, human capital, and job ing away of the discrimination, and in some
attributes but declines by one third, approxi- sectors the legally proscribed discrimination,
mately from 30 to 20 percent of the gen- against married women that characterized
der wage gap, after adjusting for occupation. labor markets in the post-World War II era,
Although some previous research has noted a near irrelevance of marriage to contempo-
that (within-gender) marriage premiums are rary workers’ choice of occupations, or both.
driven more by positive selection into mar- Conversely, the strong association between
riage on the basis of pre-market factors and parental wage gaps and occupation implies
other job attributes than by occupation (Kille- that fathers and mothers, even more than
wald and Gough 2013), to our knowledge, the childless men and childless women, continue
differential sensitivity of marital and parental to be unevenly distributed across occupations
status effects on the gender gap in wages has that differ in pay. This segregation could occur
not been previously identified in either the if discrimination against mothers is stronger
family wage or the gender wage literatures. in some types of occupations (but see Ishi-
How can we understand the differential zuka 2021), if mothers and fathers are espe-
sensitivity of marital and parental wage gaps cially likely to hold gender essentialist beliefs
to occupation? The weak association between that make traditionally female- or male-typed
992 American Sociological Review 88(6)

occupations attractive (see Thébaud and Tay- contribute far less to the gender wage gaps
lor 2021), or if occupation-specific norms and among Black and Hispanic workers than
expectations about work hours create stronger among White workers in both absolute and
conflicts with dependent-care responsibilities relative terms. The contributions of gender-
in some occupations than in others and trig- differentiated returns to marriage are also
ger gender-specific occupational mobility (see smaller among Black and Hispanic workers
Cha 2013; Goldin 2014). These are mere in absolute terms, but comparable or even
speculations, though, and we hope our results larger at times as a share of the total gender
will encourage future work on the patterns, wage gap among these groups. Moreover,
sources, and consequences of occupational unlike White workers, Black and Hispanic
segregation by family status and its interaction workers also show strong endowment effects
with gender. of marriage (but not parental status) on gen-
The general theme of our results is that der wage gaps.
the strong relationship between family wage At the most basic level, these results imply
gaps and the gender wage gap, long assumed that even if one accepts the “family wage
in the family and gender wage gap literatures, gap in disguise” framing of the gender wage
is on the mark. At the same time, our analysis gap for White workers, it does not obtain for
reveals two important caveats to a “family Black and Hispanic workers. Our results thus
wage gap in disguise” story. reinforce recent calls to pay more attention
First, our results caution against a single- to racial heterogeneity in family wage gap
minded focus on the gendered effect of work- research. A more subtle implication is that
family conflict or structural barriers that where the goal is to understand gender wage
generate family wage gaps. Indeed, if 64 gaps among Black and Hispanic workers, the
percent of the gender wage gap is associated payoff to identifying the sources of gender
with family wage gaps, 36 percent is not. differences in the distribution of family status
Similarly, the SIPP data show a gender wage may be greater than identifying the sources of
gap of 6 to 10 percent, depending on adjust- gender differences in the payoff (or penalty)
ment, among never-married childless work- to marriage or parenthood.
ers. Although this is less than half of the wage The generally weaker association between
gap among married parents (17 to 23 percent, family wage gaps (and especially parental
depending on adjustment), it is far from neg- wage gaps) and the gender wage gap among
ligible. These gender differences in wages Black and Hispanic workers is in line with
among childless and unmarried workers call qualitative scholarship that argues that Black
into question the common interpretation of and Hispanic experiences of parenthood,
the wages of these “baseline groups,” as if work, and work-family conflict fundamen-
they are the wage that would obtain in the tally differ from those of White women. Con-
absence of discrimination or of gender dif- temporary notions of work-family conflict
ferences in response to work-family conflict. presume a norm of separate spheres and the
Second, the strong association between “cult of domesticity,” wherein good mother-
the family wage differentials and the gender ing implies a devotion to family that is hard
wage gap is limited to White workers (see to reconcile with incentives and expectations
Table 5). Consistent with other research, we at work, such as the outsized wage returns
find smaller gender wage gaps among Black to working long hours (Cha and Weeden
and Hispanic workers than among White 2014; Goldin 2014). Dow (2019) argues that
workers or the all-worker sample, so much this is primarily a White, middle-class norm,
so that in the fully adjusted model, these and that Black women’s standards of wom-
gender wage gaps are no longer statistically anhood place greater value on paid work
significant. Our more novel finding is that and financial independence and allow for
gender-differentiated returns to parenthood greater reliance on extended kin to support
Cha et al. 993

mothers’ paid work. Differences in Black and inequality research. Why is occupation so
White women’s relationship to paid work are consequential for the association between the
further exacerbated by differences in eco- gender wage gap and the gender-differentiated
nomic resources, historical legacies of Black wage returns to parenthood, but not mar-
enslavement and institutional discrimination, riage? What accounts for the wage differen-
and Black men’s structural position within the tials between childless men and women, and
labor market. In this social and cultural con- between never-married men and women? To
text, we might expect not only smaller gender what extent do racial differences in the social
wage gaps among Black workers, but also meaning of motherhood (see Dow 2019)
weaker family status effects on the gender account for the observed racial differences
wage gaps. We are not wedded to this inter- in the relationship between family wage gaps
pretation of our findings, although we think and gender wage gaps? What explains the
it holds promise, but to the simpler point that gender-differentiated access to marriage and
racial heterogeneity undermines sweeping its wage consequences among Black and His-
claims about the relationship between family panic workers? Do family wage gaps make
wages and the gender wage gap. similar contributions to the gender wage gap
The patterns we document here—the main at all points of the wage distribution as at the
finding of a strong association between fam- mean, as we studied here, or do these con-
ily wage and gender wage gaps as well as tributions differ between high- and low-paid
the two caveats—shed new light on how workers? At the policy level, our empirical
wage-setting processes are embedded in fam- results suggest that efforts to address family
ily demography processes. Our results also wage gaps can complement, but not supplant,
open up new questions for gender and family efforts to address gender wage gaps.

Appendix
Table A1. Regression Coefficients Used for the Decomposition Analysis for All Workers
(Table 3)

Model 1 Model 2 Model 3

Men Women Men Women Men Women


Parent .040* –.039* .070** –.010 .057** .005
(.018) (.016) (.016) (.015) (.016) (.014)

Marital status (“never married” omitted)


Married .509** .393** .183** .117** .134** .071**
(.019) (.018) (.018) (.017) (.017) (.016)
Previously .276** .248** .067** .047* .043 .031
married (.027) (.024) (.025) (.022) (.025) (.021)
Cohabiting .153** .152** .069** .064** .042 .049*
(.025) (.025) (.022) (.024) (.022) (.022)
Age .040** .040** .029** .029**
(.004) (.003) (.003) (.003)
Age squared –.000** –.000** –.000** –.000**
(<.001) (<.001) (<.001) (<.001)
Race (“White” omitted)
Black –.171** –.014 –.105** .024
(.021) (.019) (.021) (.018)

(continued)
994 American Sociological Review 88(6)

Table A1. (continued)

Model 1 Model 2 Model 3

Men Women Men Women Men Women


Hispanic –.153** –.095** –.105** –.056**
(.018) (.016) (.017) (.016)
Other race –.074** –.010 –.094** –.022
(.028) (.031) (.026) (.030)
Education (“high school graduate” omitted)
Less than high school –.126** –.153** –.083** –.076**
(.022) (.023) (.022) (.022)
Some college .163** .147** .100** .072**
(.014) (.014) (.014) (.015)
College graduate .481** .449** .258** .261**
(.020) (.019) (.020) (.020)
Advanced degree .754** .691** .475** .470**
(.023) (.021) (.027) (.026)
Tenure .009** .009** .007** .006**
(.002) (.002) (.002) (.002)
Tenure squared <.001 <.001 <.001 <.001
(<.001) (<.001) (<.001) (<.001)
Weekly work hours (“35 to 49” omitted)
Less than 35 –.116** –.109** –.013 –.008
(.023) (.015) (.022) (.015)
50 or more .030* –.020 –.016 –.070**
(.015) (.020) (.015) (.019)
Public sector –.094** –.064** –.057** –.005
(.018) (.016) (.020) (.017)
Region (“East” omitted)
Midwest –.074** –.081** –.084** –.074**
(.021) (.020) (.020) (.019)
South –.053** –.103** –.072** –.098**
(.018) (.018) (.017) (.017)
West .049* .040* .029 .042*
(.020) (.020) (.018) (.019)
Metropolitan .142** .174** .114** .142**
(.016) (.016) (.016) (.014)
Union .119** .027 .175** .101**
(.018) (.018) (.018) (.018)
Occupations Not included Not included Included
Constant 2.760** 2.703** 1.726** 1.607** 2.865** 2.583**
(.013) (.015) (.071) (.066) (.100) (.131)
R-sq .104 .051 .323 .282 .435 .396
N 12,079 11,686 12,079 11,686 12,079 11,686

Source: SIPP 2018 Panel.


Note: All data are weighted using the final weight provided by the SIPP. Standard errors are in
parentheses and estimated using Fay’s modified balanced repeated replication (BRR) method, using
replicate weights provided by the SIPP.
*p < .05; **p < .01 (two-tailed test).
Cha et al. 995

Table A2. Regression Coefficients Used for the Decomposition Analysis for Black Workers
(Table 5, Panel A)

Model 1 Model 2

Men Women Men Women


Parent –.056 –.063 –.080 –.061
(.053) (.043) (.053) (.042)

Marital status (“never married” omitted)


Married .470** .414** .253** .176**
(.057) (.044) (.054) (.041)
Previously married .307** .260** .179* .060
(.077) (.047) (.077) (.044)
Cohabiting .170* .153* .134* .074
(.068) (.070) (.062) (.066)
Age .024* .047**
(.011) (.009)
Age squared –.000* –.000**
(<.001) (<.001)
Education (“high school graduate” omitted)
Less than high school –.140* –.116
(.055) (.073)
Some college .150** .095*
(.042) (.038)
College graduate .425** .446**
(.063) (.055)
Advanced degree .763** .698**
(.093) (.061)
Tenure .022** –.001
(.007) (.006)
Tenure squared –.000 <.001
(<.001) (<.001)
Weekly work hours (“35 to 49” omitted)
Less than 35 –.049 –.015
(.069) (.045)
50 or more –.090 –.047
(.050) (.062)
Public sector –.011 .061
(.051) (.046)
Region (“East” omitted)
Midwest –.116 –.171**
(.064) (.065)
South –.115* –.234**
(.052) (.050)
West .021 .004
(.076) (.076)
Metropolitan .139* .165**
(.066) (.045)
Union .062 –.014
(.044) (.048)
Constant 2.686** 2.685** 1.951** 1.564**
(.042) (.039) (.211) (.191)
R-sq .075 .060 .265 .289
N 1,193 1,498 1,193 1,498

Source: SIPP 2018 Panel.


Note: All data are weighted using the final weight provided by the SIPP. Standard errors are in
parentheses and estimated using Fay’s modified balanced repeated replication (BRR) method, using
replicate weights provided by the SIPP.
*p < .05; **p < .01 (two-tailed test).
996 American Sociological Review 88(6)

Table A3. Regression Coefficients Used for the Decomposition Analysis for Hispanic
Workers (Table 5, Panel B)

Model 1 Model 2

Men Women Men Women


Parent –.004 –.023 .041 –.023
(.054) (.040) (.052) (.041)

Marital status (“never married” omitted)


Married .252** .201** .094 .017
(.056) (.034) (.051) (.037)
Previously married .267** .177** .133 .024
(.078) (.048) (.068) (.050)
Cohabiting .053 .149* .012 .090
(.069) (.059) (.066) (.056)
Age .037** .040**
(.009) (.008)
Age squared –.000** –.000**
(<.001) (<.001)
Education (“high school graduate” omitted)
Less than high school –.090* –.186**
(.035) (.036)
Some college .192** .124**
(.042) (.031)
College graduate .418** .433**
(.057) (.040)
Advanced degree .528** .673**
(.082) (.069)
Tenure <.001 –.001
(.006) (.005)
Tenure squared <.001 <.001*
(<.001) (<.001)
Weekly work hours (“35 to 49” omitted)
Less than 35 –.012 –.055
(.060) (.034)
50 or more –.040 –.057
(.039) (.047)
Public sector .046 –.007
(.058) (.036)
Region (“East” omitted)
Midwest –.089 –.097
(.076) (.054)
South –.146** –.130**
(.055) (.038)
West –.100 –.040
(.052) (.041)
Metropolitan .147** .200**
(.047) (.042)
Union .179** .133**
(.047) (.044)
Constant 2.716** 2.612** 1.855** 1.594**
(.033) (.029) (.163) (.140)
R-sq .032 .017 .162 .255
N 2,261 2,082 2,261 2,082

Source: SIPP 2018 Panel.


Note: All data are weighted using the final weight provided by the SIPP. Standard errors are in
parentheses and estimated using Fay’s modified balanced repeated replication (BRR) method, using
replicate weights provided by the SIPP.
*p < .05; **p < .01 (two-tailed test).
Cha et al. 997

Table A4. Regression Coefficients Used for the Decomposition Analysis for White Workers
(Table 5, Panel C)

Model 1 Model 2

Men Women Men Women


Parent .147** –.006** .120** –.004
(.021) (.021) (.020) (.021)

Marital status (“never married” omitted)


Married .517** .388** .194** .120**
(.025) (.027) (.025) (.025)
Previously married .232** .251** .027 .033
(.035) (.035) (.032) (.032)
Cohabiting .172** .139** .068** .048
(.028) (.033) (.024) (.029)
Age .045** .040**
(.004) (.005)
Age squared –.000** –.000**
(<.001) (<.001)
Education (“high school graduate” omitted)
Less than high school –.080* –.078*
(.033) (.037)
Some college .151** .170**
(.016) (.018)
College graduate .482** .452**
(.021) (.022)
Advanced degree .711** .680**
(.026) (.026)
Tenure .009** .013**
(.003) (.003)
Tenure squared <.001 –.000
(<.001) (<.001)
Weekly work hours (“35 to 49” omitted)
Less than 35 –.179** –.161**
(.027) (.020)
50 or more .061** –.018
(.018) (.026)
Public sector –.136** –.108**
(.021) (.021)
Region (“East” omitted)
Midwest –.060* –.057*
(.024) (.024)
South –.021 –.071**
(.023) (.024)
West .088** .071**
(.026) (.026)
Metropolitan .150** .175**
(.017) (.020)
Union .128** .014
(.021) (.023)
Constant 2.767** 2.725** 1.555** 1.582**
(.017) (.020) (.086) (.092)
R-sq .143 .051 .353 .276
N 7,583 7,125 7,583 7,125

Source: SIPP 2018 Panel.


Note: All data are weighted using the final weight provided by the SIPP. Standard errors are in
parentheses and estimated using Fay’s modified balanced repeated replication (BRR) method, using
replicate weights provided by the SIPP.
*p < .05; **p < .01 (two-tailed test).
998 American Sociological Review 88(6)

Editors’ Note because some respondents erroneously reported


To avoid any possible conflict of interest, the ASR Editors earnings using a different periodicity than they had
were not involved in the evaluation of this paper. The indicated in a lead-in question. Trimming the sam-
entire review process was handled by a Deputy Editor ple of respondents with wages in excess of $250/
who is not affiliated with Indiana University. hour reduces the effect of these likely errors on our
estimates. See https://www.census.gov/programs-
surveys/sipp/tech-documentation/user-notes/2018-
Acknowledgments usernotes/2018-high-wages.html.
We thank Paula England, Misun Lim, Mary C. Noonan, 5. Although the SIPP allows analysts to differentiate
and participants of the Gender Inequality Seminar at Har- parental status by the number, age, relationship, and
vard University, Center for Demography and Ecology at residential status of children, we leave these poten-
the University of Wisconsin-Madison, and Broom Center tial sources of heterogeneity to a follow-up paper
for Demography at the University of California-Santa to meet space constraints and to focus on the core
Barbara, WZB Talk, and WZB AAM/NEPS/HIS Col- research questions.
loquium for their helpful comments on previous versions 6. The referent time point for the parent screener ques-
of this article. tion is the date of the interview in 2018; the refer-
ent time point for most other variables is December
2017. For consistency, we used the household ros-
ORCID iDs ter and fertility data to identify respondents who
became first-time parents between December 2017
Youngjoo Cha https://orcid.org/0000-0003-
and their interview in 2018, and then recoded the
1482-648X
parent screener variable accordingly. Also, 156 cases
Kim A. Weeden https://orcid.org/0000-0002-
show discrepancies between the household roster
9975-8457
and fertility history data, which could occur because
Landon Schnabel https://orcid.org/0000-0002-
deceased children are in the fertility history but not
2674-3019
the household roster data, or because SIPP top-codes
the number of children in the fertility history data.
For these cases we used the household roster data.
Data Note
7. ln Y i = βXi + γGi + δMit + πPi + τGMi + υGPi +
Replication materials are available with the article’s τMPi + υGMPi + εi., where i indexes an individual;
online supplementary materials and at https://osf.io/ujds5/. Y is a column vector for the log of hourly wages; X
is the covariate matrix; G, M, and P are column vec-
tors for gender, marital status, and parental status;
Notes and β, γ, δ, π, τ, υ, τ, and υ are column vectors of
1. The analysis using all available months shows sub- coefficients.
stantively similar results. 8. When we restricted the sample to workers age 25
2. SIPP panels from 2008 and earlier collected informa- to 54, we found that nearly 50 percent of the gen-
tion on the total years of work experience in the paid der wage gap is associated with the parenthood
labor force as well as job tenure. In a supplementary effect. The higher contribution of parental status
analysis, we used the SIPP 2008 and refit our models in the age-restricted sample than in the full sample
with and without the total years of work experience, is mainly due to the exclusion of younger workers
specified as a quadratic. This analysis shows that the age 18 to 24. However, the parental status effect in
main effect of total work experience is significant but the adjusted models and the marriage effects across
smaller than the effect of job tenure. More impor- all models are quite similar across the full and age-
tantly, the inclusion of total work experience has no restricted samples.
effect on the estimated gender wage gaps by family 9. An interim model that only adjusts for demographic
status. Assuming no dramatic shifts in wage-setting factors and education but not job attributes shows
processes between 2008 and 2018, this suggests our that roughly half of the reduction in the contribu-
conclusions are not biased by the lack of a total work tion of marital wage gaps shown between Models
experience measure in SIPP 2018. 1 and 2 in Table 3 is associated with these “pre-
3. We also fit models using a wage measure that market” factors. The contribution of parental status
includes income from commissions, overtime, tips, wage gaps increased by 14 percent in this pre-mar-
and bonuses. The gender wage gap using this alter- ket model, from .051 to .058. The decline of the
native measure is slightly larger than the gap with coefficient in Model 2 (.052) from the pre-market
the more restricted measure, but family wage gaps model implies that additional covariates (e.g., job
and their contribution to the gender wage gap are tenure, work hours, other labor market attributes)
very similar. are associated with 10 percent of the contribution of
4. User notes to the SIPP report that a small number parental status wage gaps shown in the pre-market
of cases have unusually high hourly earnings, likely model.
Cha et al. 999

References Gender-Neutral Household Specialization Effects.”


Journal of Marriage and Family 78(5):1352–70.
Abendroth, Anja-Kristin, Matt L. Huffman, and Judith Budig, Michelle J., Misun Lim, and Melissa J.
Treas. 2014. “The Parity Penalty in Life Course Per- Hodges. 2021. “Racial and Gender Pay Dis-
spective: Motherhood and Occupational Status in 13 parities: The Role of Education.” Social Science
European Countries.” American Sociological Review Research 98:102580 (https://doi.org/10.1016/j.
79(5):993–1014. ssresearch.2021.102580).
Anderson, Deborah J., Melissa Binder, and Kate Krause. Cha, Youngjoo. 2013. “Overwork and the Persistence of
2003. “The Motherhood Wage Penalty Revisited: Gender Segregation in Occupations.” Gender & Soci-
Experience, Heterogeneity, Work Effort, and Work- ety 27(2):158–84.
Schedule Flexibility.” ILR Review 56(2):273–94. Cha, Youngjoo, and Kim A. Weeden. 2014. “Overwork
Angelov, Nikolay, Per Johansson, and Erica Lindahl. and the Slow Convergence in the Gender Gap in
2016. “Parenthood and the Gender Gap in Pay.” Jour- Wages.” American Sociological Review 79(3):457–
nal of Labor Economics 34(3):545–79. 84.
Antonovics, Kate, and Robert Town. 2004. “Are All Cheng, Siwei. 2016. “The Accumulation of (Dis)advan-
the Good Men Married? Uncovering the Sources of tage: The Intersection of Gender and Race in the
the Marital Wage Premium.” American Economic Long-Term Wage Effect of Marriage.” American
Review 94(2):317–21. Sociological Review 81(1):29–56.
Avellar, Sarah, and Pamela J. Smock. 2003. “Has the Price Chun, Hyunbae, and Injae Lee. 2001. “Why Do Married
of Motherhood Declined over Time? A Cross-Cohort Men Earn More: Productivity or Marriage Selec-
Comparison of the Motherhood Wage Penalty.” Jour- tion?” Economic Inquiry 39(2):307–19.
nal of Marriage and Family 65(3):597–607. Cohen, Philip N. 2002. “Cohabitation and the Declining
Bertrand, Marianne, Claudia Goldin, and Lawrence F. Marriage Premium for Men.” Work and Occupations
Katz. 2010. “Dynamics of the Gender Gap for Young 29(3):346–63.
Professionals in the Financial and Corporate Sec- Correll, Shelley J., Stephen Benard, and In Paik. 2007.
tors.” American Economic Journal: Applied Econom- “Getting a Job: Is There a Motherhood Penalty?”
ics 2(3):228–55. American Journal of Sociology 112(5):1297–338.
Blau, Francine D., and Lawrence M. Kahn. 1997. “Swim- Cortés, Patricia, and Jessica Pan. 2019. “When Time
ming Upstream: Trends in the Gender Wage Differ- Binds: Substitutes for Household Production, Returns
ential in the 1980s.” Journal of Labor Economics to Working Long Hours, and the Skilled Gender Wage
15(1):1–42. Gap.” Journal of Labor Economics 37(2):351–98.
Blau, Francine D., and Lawrence M. Kahn. 2006. “The Doren, Catherine. 2019. “Which Mothers Pay a Higher
US Gender Pay Gap in the 1990s: Slowing Conver- Price? Education Differences in Motherhood Wage
gence.” ILR Review 60(1):45–66. Penalties by Parity and Fertility Timing.” Sociologi-
Blau, Francine D., and Lawrence M. Kahn. 2017. “The cal Science 6:684–709.
Gender Wage Gap: Extent, Trends, and Explanations.” Dow, Dawn. 2019. Mothering while Black: Boundaries
Journal of Economic Literature 55(3):789–865. and Burdens of Middle-Class Parenthood. Berkeley:
Blau, Francine D., Lawrence M. Kahn, Nikolai Bob- University of California Press.
oshko, and Matthew L. Comey. 2021. “The Impact England, Paula, Jonathan Bearak, Michelle J. Budig, and
of Selection into the Labor Force on the Gender Wage Melissa J. Hodges. 2016. “Do Highly Paid, Highly
Gap.” No. w28855, National Bureau of Economic Skilled Women Experience the Largest Mother-
Research. hood Penalty?” American Sociological Review
Blinder, Alan S. 1973. “Wage Discrimination: Reduced 81(6):1161–89.
Form and Structural Estimates.” Journal of Human England, Paula, Andrew Levine, and Emma Mishel.
Resources 8(4):436–55. 2020. “Progress toward Gender Equality in the
Budig, Michelle J., and Paula England. 2001. “The Wage United States Has Slowed or Stalled.” Proceedings of
Penalty for Motherhood.” American Sociological the National Academy of Sciences 117(13):6990–97.
Review 66(2):204–25. Folbre, Nancy. 2022. “Review of Career and Family:
Budig, Michelle J., and Melissa J. Hodges. 2010. “Differ- Women’s Century-Long Journey toward Equity.” EH.
ences in Disadvantage: Variation in the Motherhood net (https://eh.net/book_reviews/career-and-family-
Penalty across White Women’s Earnings Distribu- womens-century-long-journey-toward-equity/).
tion.” American Sociological Review 75(5):705–28. Ginther, Donna K., and Madeline Zavodny. 2001. “Is the
Budig, Michelle J., and Melissa J. Hodges. 2014. “Sta- Male Marriage Premium Due to Selection? The Effect
tistical Models and Empirical Evidence for Differ- of Shotgun Weddings on the Return to Marriage.”
ences in the Motherhood Penalty across the Earnings Journal of Population Economics 14(2):313–28.
Distribution.” American Sociological Review 79(2): Glauber, Rebecca. 2007. “Marriage and the Motherhood
358–64. Wage Penalty among African Americans, Hispanics,
Budig, Michelle J., and Misun Lim. 2016. “Cohort Dif- and Whites.” Journal of Marriage and Family
ferences and the Marriage Premium: Emergence of 69(4):951–61.
1000 American Sociological Review 88(6)

Glauber, Rebecca. 2008. “Race and Gender in Families Killewald, Alexandra, and Ian Lundberg. 2017. “New
and at Work: The Fatherhood Wage Premium.” Gen- Evidence against a Causal Marriage Wage Premium.”
der & Society 22(1):8–30. Demography 54(3):1007–28.
Goldin, Claudia. 2014. “A Grand Gender Convergence: Kitagawa, Evelyn M. 1955. “Components of a Difference
Its Last Chapter.” American Economic Review between Two Rates.” Journal of the American Statis-
104(4):1091–119. tical Association 50(272):1168–94.
Goldin, Claudia. 2021. Career and Family: Women’s Kliff, Sarah. 2018. “A Stunning Chart Shows the True
Century-Long Journey toward Equity. Princeton, NJ: Cause of the Gender Wage Gap.” Vox, February 19
Princeton University Press. (https://www.vox.com/2018/2/19/17018380/gender-
Goldin, Claudia, and Lawrence F. Katz. 2007. “Long- wage-gap-childcare-penalty).
Run Changes in the US Wage Structure: Narrowing, Korenman, Sanders, and David Neumark. 1991. “Does
Widening, Polarizing.” Brookings Papers on Eco- Marriage Really Make Men More Productive?” Jour-
nomic Activity 38(2):135–68. nal of Human Resources 26(2):282–307.
Hegewisch, Ariane, and Asha DuMonthier. 2016. “The Loughran, David S., and Julie M. Zissimopoulos. 2009.
Gender Wage Gap: 2015: Annual Earnings Differ- “Why Wait? The Effect of Marriage and Childbear-
ences by Gender, Race, and Ethnicity.” Institution for ing on the Wages of Men and Women.” Journal of
Women’s Policy Research. Human Resources 44(2):326–49.
Hersch, Joni, and Leslie S. Stratton. 2000. “Household Ludwig, Volker, and Josef Brüderl. 2018. “Is There a Male
Specialization and the Male Marriage Wage Pre- Marital Wage Premium? New Evidence from the United
mium.” ILR Review 54(1):78–94. States.” American Sociological Review 83(4):744–70.
Hodges, Melissa J., and Michelle J. Budig. 2010. “Who Lundberg, Shelly, and Elaina Rose. 2000. “Parenthood
Gets the Daddy Bonus? Organizational Hegemonic and the Earnings of Married Men and Women.”
Masculinity and the Impact of Fatherhood on Earn- Labour Economics 7(6):689–710.
ings.” Gender & Society 24(6):717–45. Lundberg, Shelly, and Elaina Rose. 2002. “The Effects
Ishizuka, Patrick. 2021. “The Motherhood Penalty in of Sons and Daughters on Men’s Labor Supply
Context: Assessing Discrimination in a Polarized and Wages.” Review of Economics and Statistics
Labor Market.” Demography 58(4):1275–300. 84(2):251–68.
Jann, Ben. 2008. “The Blinder–Oaxaca Decomposition Mandel, Hadas, and Assaf Rotman. 2021. “Revealing the
for Linear Regression Models.” The Stata Journal Concealed Effect of Top Earnings on the Gender Gap in
8(4):453–79. the Economic Value of Higher Education in the United
Jann, Ben. 2014. “Plotting Regression Coefficients and States, 1980–2017.” Demography 58(2):551–70.
Other Estimates.” The Stata Journal 14(4):708–37. Mandel, Hadas, and Moshe Semyonov. 2014. “Gender
Jann, Ben. 2018. “Customizing Stata Graphs Made Easy Pay Gap and Employment Sector: Sources of Earn-
(Part 2).” The Stata Journal 18(4):786–802. ings Disparities in the United States, 1970–2010.”
Jee, Eunjung, Joya Misra, and Marta Murray-Close. Demography 51(5):1597–618.
2019. “Motherhood Penalties in the US, 1986–2014.” Mandel, Hadas, and Moshe Semyonov. 2016. “Going
Journal of Marriage and Family 81(2):434–49. Back in Time? Gender Differences in Trends and
Juhn, Chinhui, and Kristin McCue. 2017. “Specialization Sources of the Racial Pay Gap, 1970 to 2010.” Ameri-
Then and Now: Marriage, Children, and the Gender can Sociological Review 81(5):1039–68.
Earnings Gap across Cohorts.” Journal of Economic Misra, Joya, and Marta Murray-Close. 2014. “The Gen-
Perspectives 31(1):183–204. der Wage Gap in the United States and Cross Nation-
Kilbourne, Barbara Stanek, Paula England, George Far- ally.” Sociology Compass 8(11):1281–95.
kas, Kurt Beron, and Dorothea Weir. 1994. “Returns Oaxaca, Ronald. 1973. “Male-Female Wage Differentials
to Skill, Compensating Differentials, and Gender in Urban Labor Markets.” International Economic
Bias: Effects of Occupational Characteristics on the Review 14(3):693–709.
Wages of White Women and Men.” American Jour- Pal, Ipshita, and Jane Waldfogel. 2016. “The Family Gap
nal of Sociology 100(3):689–719. in Pay: New Evidence for 1967 to 2013.” RSF: The
Killewald, Alexandra. 2011. “Opting Out and Buying Russell Sage Foundation Journal of the Social Sci-
Out: Wives’ Earnings and Housework Time.” Journal ences 2(4):104–27.
of Marriage and Family 73(2):459–71. Petersen, Trond, and Laurie A. Morgan. 1995. “Separate
Killewald, Alexandra. 2013. “A Reconsideration of the and Unequal: Occupation-Establishment Sex Segre-
Fatherhood Premium: Marriage, Coresidence, Biol- gation and the Gender Wage Gap.” American Journal
ogy, and Fathers’ Wages.” American Sociological of Sociology 101(2):329–65.
Review 78(1):96–116. Stainback, Kevin, and Donald Tomaskovic-Devey.
Killewald, Alexandra, and Margaret Gough. 2013. 2012. Documenting Desegregation: Racial and
“Does Specialization Explain Marriage Penalties Gender Segregation in Private Sector Employment
and Premiums?” American Sociological Review since the Civil Rights Act. New York: Russell Sage
78(3):477–502. Foundation.
Cha et al. 1001

Thébaud, Sarah, and Catherine J. Taylor. 2021. “The reinforce gender inequality and what organizational and
Specter of Motherhood: Culture and the Production institutional conditions help to change this trend; the
of Gendered Career Aspirations in Science and Engi- sources of occupational variation in the gender pay gap;
neering.” Gender & Society 35(3):395–421. the role of stereotypes for the labor market outcomes of
Van Winkle, Zachary, and Anette Eva Fasang. 2020. Asian-origin individuals in the United States; and how
“Parenthood Wage Gaps across the Life Course: A the pandemic has changed employees’ and employers’
Comparison by Gender and Race.” Journal of Mar- perspectives on work.
riage and Family 82(5):1515–33.
Waldfogel, Jane. 1997. “The Effect of Children on Women’s Kim A. Weeden is the Jan Rock Zubrow ’77 Professor
Wages.” American Sociological Review 62(2):209–17. of Sociology, Director of the Center for the Study of
Waldfogel, Jane. 1998. “Understanding the ‘Family Gap’ Inequality, and Stephen H. Weiss Presidential Fellow
in Pay for Women with Children.” Journal of Eco- at Cornell University. Her current projects explore the
nomic Perspectives 12(1):137–56. relationship between the social organization of work and
Waldfogel, Joel. 1996. “The Deadweight Loss of Christ- gender inequalities in wages in the United States and
mas: Reply.” The American Economic Review abroad; social change in the structure of inequality; race,
86(5):1306–08. gender, and class-based sorting into occupations, college
Yun, Myeong-Su. 2005. “A Simple Solution to the Identi- majors, and college courses; intergenerational occupa-
fication Problem in Detailed Wage Decompositions.” tional mobility; and differences in young adults’ occu-
Economic Inquiry 43(4):766–72. pational aspirations and educational decisions, including
their entry and persistence in STEM fields.
Youngjoo Cha is Associate Professor of Sociology at
Indiana University-Bloomington and Yonsei University. Landon Schnabel is the Rosenthal Assistant Professor
Her current research investigates how the trend toward of Sociology at Cornell University. His research exam-
long work hours and the normative forces behind them ines inequality, gender, religion, and politics.

You might also like