New Evidence against a Causal Marriage Wage Premium
Alexandra Killewald*, Harvard University
and
Ian Lundberg, Princeton University
July 2015
Word count: 7,244
Keywords: marriage, divorce, wages, transition to adulthood, panel data models
*We are grateful to Javier Garcia-Manglano, Florencia Torche, and Christopher Winship for
helpful comments and to Siwei Cheng for both comments and sharing her in-progress
manuscript. An earlier version of this manuscript was presented at the annual meeting of the
Population Association of America, May 2014, Boston, MA. Direct all correspondence to
Alexandra Killewald, 436 William James Hall, Harvard University, 33 Kirkland St, Cambridge,
MA 02138; killewald@fas.harvard.edu; 617-495-3818 (phone); 617-496-5794 (fax).
1
Abstract. Marriage is associated with increases in men’s wages. Recent research claims the
long-term wage benefits of marriage for men are as high as 20 percent and begin prior to
marriage, as men anticipate marriage or experience wage benefits of unmarried partnership. We
argue instead that marriage has no causal effect on men’s wages in either the short or long term
and that research on the marriage wage premium has overlooked literature in other subfields
suggesting that marriage occurs when wages are already rising unusually rapidly. A vast
literature documents that entrance into marriage depends on economic circumstances, suggesting
that effects may flow from wages to marriage, rather than the reverse. Furthermore, the
demographic literature on the transition to adulthood suggests that emerging adulthood is a time
of both union formation and unusually rapid improvements in work outcomes. Using data from
the NLSY79, we evaluate these perspectives, considering both the effects of getting married and
remaining married. We conclude that the observed wage patterns are most consistent with men
marrying at a time that their wages are already rising more rapidly than expected and divorcing
when their wages are already falling, with no additional causal effect of marriage on wages.
2
Prior research suggests that marital partnerships have substantial effects on men’s time
use, well-being, and priorities (Nock 1998, Waite and Gallagher 2000). In particular, the benefits
of marriage for men’s wages are of long-standing interest to both economists and sociologists
(Cheng Forthcoming, Chun and Lee 2001, Cohen 2002, Cornwell and Rupert 1997, Dougherty
2006, Ginther and Zavodny 2001, Gray 1997, Hersch and Stratton 2000, Killewald and Gough
2013, Korenman and Neumark 1991, Loh 1996). Traditional fixed-effects models typically
estimate that marriage increases men’s wages by about 5 to 7 percent (Cornwell and Rupert
1997, Dougherty 2006, Killewald 2013, Killewald and Gough 2013).
More recently, Dougherty (2006) and Cheng (Forthcoming) tracked men’s wages in the
years leading up to marriage and then at subsequent marital durations. Both find that men’s
wages are already rising more rapidly than expected several years prior to marriage and argue
that, including this premarital increase as part of the marriage premium, the long-term wage
benefits to marriage are about 20 percent, much higher than estimated by traditional fixed-effects
models. If correct, the effect of marriage on men’s wages is enormous – about half as large as the
effect of a college degree (Brand and Xie 2010).
We argue that caution is warranted before accepting these conclusions. Estimating the
causal effect of marriage on wages is challenging. Reverse causality poses a threat to causal
inference, as substantial evidence shows that men’s employment and earnings affect the
likelihood of marriage (Oppenheimer, Kalmijn and Lim 1997, Oppenheimer 2003, Sweeney
2002, Xie et al. 2003), and couples report delaying marriage until achieving certain financial
goals (Edin and Kefalas 2005, Smock, Manning and Porter 2005). Furthermore, marriage
typically takes place in the midst of a more general transition to adulthood. The transition to
adulthood occurs at different ages for different individuals and involves rapid maturation in
3
multiple domains, including work and family formation (Arnett 2015, Hogan and Astone 1986,
Rindfuss 1991, Shanahan 2000). We argue that the premarital rise in wages observed by
Dougherty and Cheng is due to these time-varying selection processes, not a causal effect of
marriage that predates the marriage itself. Thus, we argue that there is no causal effect of
marriage on wages, either before or after the marriage date.
Using data from the 1979-2012 waves of the NLSY79, we begin by replicating the
findings of Dougherty and Cheng, which show rapidly rising wages prior to marriage and
through the first several years of marriage. Next, we explore whether the premarital rise in wages
is likely due to a causal effect of marriage that predates the wedding, through anticipation
effects. We show that men who marry after age 26 experience no premarital increase in wages
and no marriage premium following marriage. This pattern of results is consistent with the
hypothesis that marriage has no causal effect on men’s wages, and that instead the premarital
wage gains observed for young husbands are due to spurious rapid growth in wages in early
adulthood, while men who marry later have largely completed this period well prior to their
marriage and therefore experience no premarital wage increase.
We also compare the experiences of men whose marriage is followed by a birth within
seven months – shotgun marriages – to those who do not have a birth shortly following their
marriage. If the premarital rise in wages is due to real effects of anticipating marriage, men
entering shotgun marriages should not experience this rise, as their marriage is less anticipated.
Instead, we find similar patterns for both groups.
We also estimate the wage benefits of staying married. Prior research suggests that some
or all of the marriage premium dissipates following divorce (Cohen 2002, Gray 1997, Killewald
2013, Killewald and Gough 2013, Korenman and Neumark 1991, Loh 1996). However, when we
4
examine trends in men’s wages prior to and following divorce, we find that wage declines
predate the divorce. These results are consistent with reverse causality, as prior research suggests
that declines in men’s economic circumstances are associated with marital disruption (Charles
and Stephens 2004).
Our results indicate that common theories of the ways marriage benefits men’s wages,
such as household specialization or employer discrimination, need to be revised. Minimally,
there is no evidence for a premium that begins at marriage or cohabitation. We argue that our
results are more consistent with the hypothesis that marriage, including its anticipation, has no
effect on men’s wages. These findings call for situating marriage within the more general life
stage of the transition to adulthood and considering more deeply male breadwinning as a
requirement for both marriage entry and stability.
Background
Married men’s wage advantage compared to both never-married and divorced men is
well-documented (Cheng Forthcoming, Chun and Lee 2001, Cohen 2002, Cornwell and Rupert
1997, Dougherty 2006, Ginther and Zavodny 2001, Gray 1997, Hersch and Stratton 2000,
Killewald and Gough 2013, Korenman and Neumark 1991, Loh 1996). Furthermore, both
marriage’s wage benefits and divorce’s wage costs have sometimes been found to increase with
years spent in the marital status (Cheng Forthcoming, Ginther and Zavodny 2001, Gray 1997,
Korenman and Neumark 1991, Loh 1996).
In this section, we first describe why marriage might lead to improvements in men’s
wages and discuss what this means for likely effects of staying married compared to divorce and
for changes in wages in advance of the marriage date. We then discuss the possibility that the
5
association between marriage and wages is driven by selection rather than a causal effect of
marriage.
Why might marriage affect men’s wages?
Specialization is the most prominent theory to explain a causal male marriage wage
premium. It posits that wives’ contributions to unpaid labor allow husbands to increase effort in
paid labor, increasing husbands’ wages (Becker 1991). Consistent with this perspective, marriage
is associated with less time in household labor for men (Gupta 1999). Specialization suggests
that cohabiting men might also experience wage gains compared to men living singly, but for
men living singly no premarital increase in wages is predicted. Since divorce is associated with
increases in men’s housework time compared to remaining married (Gupta 1999), specialization
also predicts a decline in men’s wages following divorce.
Employer discrimination may also cause higher wages for married men, especially if
marriage generates biased perceptions by employers of men’s stability or competence.
Unfortunately, we know of no studies evaluating this possibility. Nonetheless, this perspective
predicts no premarital marriage premium and a likely erosion of the premium following divorce.
Cheng (Forthcoming) and Dougherty (2006) document that men’s wages begin to
increase more rapidly than expected several years before marriage and that the marriage date
itself leads to neither an immediate increase in wages nor a trajectory of more rapid wage
growth. These findings cast doubt on the specialization and discrimination perspectives for the
male marriage premium, since neither can explain the premarital wage increase and both suggest
that the formation of the marital union itself should have wage consequences. Dougherty (2006)
uses his results to argue against attributing the male marriage premium to specialization, and
6
other research has similarly rejected the specialization explanation (Hersch and Stratton 2000,
Killewald and Gough 2013, Loh 1996, Pollmann-Schult 2011).
However, the male marriage premium may arise for other reasons. Because wage-earning
is normative for husbands (Nock 1998), marriage may motivate men to invest more in wageearning. Consistent with this perspective, pay is ranked a more important job trait by married
than unmarried men (Gorman 2000), married men are more likely to experience job changes
associated with wage gains (Gorman 1999, Pollmann-Schult 2011), marriage is associated with
increases in occupational prestige (Nock 1998), and a portion of the male marriage premium is
explained by occupation, industry, sector, and job tenure controls (Killewald and Gough 2013).
This perspective, unlike specialization or discrimination, allows that men’s wages might increase
prior to marriage, “as grooms-to-be shrug off bachelor habits and begin to assume the outlook
and priorities of married men” (Waite and Gallagher 2000, p. 100). When men exit the husband
role, following divorce, we expect wages to decline. In the remainder of the paper, we focus on
this causal explanation for the marriage premium, because it is more consistent than
specialization or discrimination with the observed premarital increase in wages.
Selection into marriage
Fixed-effects or first-difference models address the concern that there are time-invariant
but unmeasured characteristics of men associated with both their wages and their marital status.
Another possibility is that men with unusually rapid wage growth, regardless of marital status,
tend to marry early, upwardly biasing estimates of the marriage premium from conventional
fixed-effects models (Loughran and Zissimopoulos 2009). To address this source of selection
7
bias, we allow the age-wage profile to vary by age at marriage, education, academic
achievement, and race.
Neither of these techniques accounts for selection into marriage on the basis of timevarying characteristics associated with wages. A large literature shows that men’s economic
circumstances predict marriage (Oppenheimer, Kalmijn and Lim 1997, Oppenheimer 2003,
Sweeney 2002, Xie et al. 2003). If this selection into marriage occurs only on the basis of timeinvariant traits, fixed effects will address the selection problem. However, prior research
suggests that the financial circumstances shaping marriage decisions are time-varying.
Unmarried couples discuss postponing marriage until they have achieved financial security (Edin
and Kefalas 2005, Smock, Manning and Porter 2005). A premarital increase in men’s wages is
therefore equally consistent with men’s wages affecting the transition to marriage as with an
anticipatory effect of marriage on men’s wages. A similar problem may arise for divorce, since
declines in men’s labor market outcomes are associated with heightened risk of divorce (Charles
and Stephens 2004). Thus, reverse causality poses a substantial threat to estimates of the effect of
marriage on wages.
A spurious association between marriage and wages may also arise if marriage typically
occurs during young or emerging adulthood, a demographically dense time of high rates of both
union formation and work instability and growth (Arnett 2015, Rindfuss 1991, Shanahan 2000).
Men may experience several years of rapid changes in both work and family formation due to a
spurious general maturation. Furthermore, the transition to adulthood does not happen all at
once; it is an ongoing process (Hogan and Astone 1986), and transitions to marriage and work
are especially likely to be “blurred” (Rindfuss 1991). These transitions happen in different orders
and at different ages for different individuals (Arnett 2015, Hogan and Astone 1986, Rindfuss
8
1991, Shanahan 2000), as some individuals experience greater challenges navigating the
transition to adulthood than others, particularly in employment (Oppenheimer, Kalmijn and Lim
1997). If blurred transitions to marriage and work stability overlap in the transition to adulthood,
men’s wages will begin to rise more rapidly than usual in the years prior to marriage, but not
because of a causal effect of marriage anticipation on wages. Because the transition to adulthood
is individual-specific in timing and duration, even flexible controls for age-wage associations
will not fully net out this pattern.
Although there is no clear age at which the transition to adulthood ends, 30 is often used
as an approximation (Arnett 2015, Rindfuss 1991). Since in our analytic sample men are age 30,
on average, at divorce, compared to age 25 at marriage, we expect that the association between
divorce and wages will not suffer as much from this spurious association, as most men
approaching divorce have already completed the transition to adulthood.
Plan for analyses
How can we separate reverse causality or a spurious association between marriage and
wages from a causal effect of marriage that partly predates the marriage itself? We lack a source
of random variation to separate men who marry from those who do not. We rely instead on a
variety of analyses that together provide evidence against a causal effect of marriage or marriage
anticipation on men’s wages. Table 1 summarizes these empirical tests and the three theories
described previously: a causal effect of marriage on wages due to normative expectations of
husband breadwinning, reverse causality due to selection of men into marriage following wage
gains, and a spurious association due to a shared maturation process affecting both wages and
marriage.
9
[Table 1 about here]
We argue that anticipation effects of marriage are likely to be confined to a year or two
before marriage. Likewise, if marriage follows wage gains, we expect to see premarital wage
gains concentrated in the years immediately prior to marriage. If wages start to rise unexpectedly
quickly further in advance of marriage, we believe the transition to adulthood explanation is
more plausible.
However, it is possible that men adjust their behavior in anticipation of marriage three or
more years in advance. We perform two analyses to gain additional insight. First, we examine
trends in wages prior to marriage for men marrying at different ages. If the premarital ramp-up in
wages is due to anticipation effects, we expect that they will be shared by all marrying men,
regardless of their age at marriage. On the other hand, if the pre-marriage trends are due to the
transition to adulthood, we hypothesize that the ramp-up will be greater for men marrying earlier,
while men marrying later will largely have already completed their period of rapid wage gains,
experiencing no additional gains just prior to marriage. The predictions under reverse causality
are not as clear, but it is possible that delaying marriage until achieving a particular financial
threshold is more important for younger men, who have lower wages, on average. Thus, seeing
no premarital wage gains for men marrying at older ages provides evidence against a causal
effect of anticipation of marriage on men’s wages, but provides less evidence to distinguish
between the transition to adulthood and reverse causality perspectives.
Second, we take advantage of a group of marriages that follow a different selection
process: shotgun marriages (Ginther and Zavodny 2001). We argue that marriages following a
premarital conception are less anticipated than other marriages. If the premarital rise in wages is
due to either a causal effect of marriage anticipation on wages or selection into marriage
10
following wage gains, the premarital wage increase for men forming shotgun marriages should
be smaller than for other husbands. Similar wage gains for both groups are more consistent with
a spurious association. Thus, this analysis helps to distinguish the transition to adulthood
perspective from both the causal marriage premium and reverse causality perspectives.
Separately, we examine men’s wages in the years surrounding divorce. We expect that, if
marriage has a causal effect on wages, it will fade following divorce, leading to declining wages.
If divorce is prompted by declines in economic resources, the reverse causality perspective, we
expect wage declines that predate divorce rather than follow it. Finally, since men are older at
divorce than marriage, the transition to adulthood perspective suggests stability in wages in the
years surrounding divorce, as most men are no longer experiencing rapid wage growth.
Data and Methods
Our analyses use the 1979-2012 waves of the 1979 cohort of the National Longitudinal
Survey of Youth (NLSY79) (U.S. Department of Labor Bureau of Labor Statistics 2014). The
NLSY79 is a nationally-representative panel study that follows individuals from 1979, when
they were age 14-22, through the present. We include observations from men’s entry into the
adult labor force through the end of data collection at age 47-56. We define labor force entry as
the first year the respondent is not enrolled in full-time school, provided he remains out of fulltime school in the next interview. We exclude observations prior to labor force entry,1 prior to
age 18, and when the respondent is active-duty military. We exclude person-year observations
1
This restriction excludes 78 men who never enter the labor force.
11
from those who are currently self-employed, since for this population wages are an incomplete
measure of labor income.
As described in more detail below, we use fixed-effects models, which rely on withinperson variation. Therefore, we exclude from the models of marriage entrance men who are
followed through 2012 but never marry (8.1 percent) or attrit from the sample prior to 2012 and
have not married when they attrit (11.5 percent). We also exclude observations that lack wage
values because men are not employed (7.5 percent). Finally, we exclude 0.3 percent of the
remaining observations because they do not have at least two observations for the fixed-effects
model. The analytic sample for marriage entry includes 50,516 observations on 4,424 men.
Similarly, we exclude from the models of exit from marriage (divorce) those who are
followed through 2012 but never divorce (30.6 percent of those who marry), attrit from the
sample prior to 2012 and without having divorced (24.8 percent of those who marry), or end
their first marriage due to widowhood (1.8 percent of those whose first marriage ends). We
exclude 8.4 percent of observations for nonemployment and 0.6 percent of remaining
observations for providing only one wage observation. The analytic sample for divorce includes
15,458 observations on 1,841 men.
Our dependent variable is the log of hourly wages in the respondent’s current job. We
use the Bureau of Labor Statistics Consumer Price Index to adjust to 2012 dollars, and we topcode at an hourly wage of $100 to reduce the effect of outliers.
To adjust for age differences between married and unmarried individuals, we control for
the log of potential experience, defined as the respondent’s age minus the age at which he
12
entered the labor market, plus one (so that all values are positive and can be logged).2 As
discussed previously, we allow the background potential experience-wage profile to vary by
educational attainment, academic achievement, race, and age at marriage (under 23, 23-26, and
over 26). We measure education with four categories: no high school degree, a high school
degree but no postsecondary education, some college, and four or more years of college.
Academic achievement is measured with percentile on the Armed Forces Qualification Test
(AFQT). We measure race as classified by the NLSY79 during household screening: nonHispanic white or other, non-Hispanic black, and Hispanic. For the 159 observations missing
education data, we use surrounding years to impute values. For respondents with missing
percentiles on the AFQT, we include an indicator for missing.
We include controls for other family formation changes in order to isolate the effects of
marriage. We measure cohabitation with a dummy variable set to one if the individual is
currently unmarried and reports a partner in the household roster. We measure parenthood with
respondents’ numbers of coresidential children. Lastly, we measure period or business cycle
effects with year indicators.
2
The log specification fit the data better than linear, linear spline, and quadratic specifications,
and the results are robust across functional forms.
13
Models
To examine wage patterns in the years leading up to and following marriage or divorce,
we follow Dougherty (2006) and estimate distributed fixed-effects models.3 Like conventional
fixed-effects models, distributed fixed-effects models net out time-invariant, unobserved
characteristics of individuals associated with both their wages and the independent variables.
While conventional fixed-effects models measure marriage with a dummy variable for current
marital status, or current status plus years spent in that status, our key independent variables are a
series of indicators marking the number of years until or since marriage (or divorce). In the
models analyzing wage changes relative to entrance into marriage, the reference category is
those who will marry in five or more years. These models include a separate dummy variable for
each year from four years before to nine years after marriage, and an additional variable for ten
or more years after marriage. Thus, coefficients indicate wages in the current year relative to five
or more years prior to marriage. Respondents are censored after the end of their first marriage.
In the models analyzing wage changes around divorce, the reference category is those
who will divorce in three or more years. These models include separate indicator variables for
each year from two years before to four years after divorce, and an additional indicator for five
or more years after divorce. In these models, respondents are censored prior to the start of their
3
While our approach is similar to Dougherty (2006), our quantity of interest is different.
Dougherty conditions on work experience and current job tenure, attempting to identify the
effect of marriage on wages net of these mediators. We are interested in the total effect of
marriage on men’s wages, some of which may operate through increased work experience and
job tenure, so we do not condition on these variables.
14
first marriage and after the start of their second marriage, if one exists. Thus, coefficients
indicate wages in the current year relative to while married and three or more years prior to
divorce. All models are weighted, although unweighted results are similar. Standard errors are
clustered.
As previously described, to distinguish between the causal effect of marriage on wages,
reverse causality of wages on marriage, and transition to adulthood perspectives, we estimate
two additional models. First, we examine wages in the years surrounding marriage separately by
tercile of age at marriage, to test whether men marrying later experience a smaller premarital
increase in wages than men marrying younger. Second, we test whether men whose marriage is
followed by a first birth within 7 months experience a similar premarital wage increase to other
grooms. In this analysis, men with a premarital birth are excluded. In our sample, about 44
percent of men with premarital conceptions enter a shotgun marriage.
Table 1 reviews how these analyses relate to the theories being tested. Although these
tests do not definitively distinguish among the three theories, they may provide evidence more
consistent with some theories than others.
Results
Table 2 provides weighted descriptive statistics for the samples used to evaluate the
marriage wage premium at entry to and exit from a first marriage. Our weighted sample is about
12 percent black, 6 percent Hispanic, and 83 percent white or other. Because our sample was
entering young adulthood primarily in the 1980s, cohabitation is still relatively rare: men report a
cohabiting partner in only 5 percent of pre-marriage observations and 21 percent of post-divorce
unmarried observations. Compared to when never-married, married men are older (33 versus 23)
15
and have higher average wages ($23.43 versus $15.82). Men who are divorced or will divorce
are less likely than the general population of married men to hold a college degree and have
lower wages, suggesting some negative selection into divorce on the basis of wage-relevant
traits.
[Table 2 about here]
Distributed fixed effects of marriage entry and exit on wages
Figure 1 shows the results from the distributed fixed-effects models for wages in the
years surrounding the entry to and exit from a first marriage. Tables used to generate all figures
are available in the online supplement. Because the outcome is the log of hourly wages, we
exponentiate the coefficients and subtract one to yield the predicted percent difference in wages
for each year, relative to 5 or more years prior to marriage or 3 or more year prior to divorce. The
dotted lines show the 95 percent confidence intervals for the estimated effects.
[Figure 1 about here]
The left panel in Figure 1 shows men’s changing wages in the years prior to and
following entry into marriage. For comparison, we estimated traditional fixed-effects models
with an indicator variable for marital status and find a 4.0 percent marriage wage premium,
similar to prior results that use this method. However, our distributed fixed-effects models reveal
that hourly wages are already increasing more rapidly than expected five years before marriage
and continue increasing through the first few years of marriage. By three years after marriage
wages are 7.6 percent higher than five years before marriage, net of changes in covariates. After
the first three years of marriage, wages grow at expected rates, leading to a flat period in the
graph. This pattern is similar to that found by Dougherty (2006) and Cheng (Forthcoming) and
16
confirms that conventional fixed-effects models provide an incomplete picture of wage changes
at marriage.
The second panel in Figure 1 shows a decline in wages beginning prior to divorce,
consistent with prior evidence that declines in men’s economic circumstances can prompt
divorce (Charles and Stephens 2004). Our comparison conventional fixed-effects model
estimates a divorce penalty compared to remaining married of 4.7 percent. Similar to prior
research, conventional fixed-effects models suggest that the marriage premium evaporates
following divorce. Our models, however, reveal that divorce appears to be the consequence
rather than cause of falling wages.
Marriage premium, reverse causality, or transition to adulthood?
The pattern of wages shown in Figure 1 confirms that, if marriage has any causal effect
on wages, gains stabilize by about three years after marriage, are no faster following marriage
than in the years just prior to marriage, and do not dissipate as a consequence of divorce. Since
unusually rapid wage growth is already taking place five years before marriage, attributing this
growth to an anticipatory effect of marriage requires believing that young men are preparing for
their roles as husbands quite far in advance of the marriage itself.
The wage pattern in the years surrounding marriage is also consistent with young men
experiencing rapid wage growth for several years during the transition to adulthood, when
marriages are also frequently happening, and then a return to typical wage growth levels after the
transition to adulthood is complete. While Dougherty and Cheng challenge that conventional
fixed-effects models have underestimated the marriage premium by failing to include premarital
wage gains, we argue the reverse: conventional fixed-effects models have overestimated the
17
wage benefits of marriage by failing to consider that the rapid wage growth observed in the first
few years after marriage may be only the continuation of an unrelated maturation process that
began several years earlier.
In the first panel of Figure 2, we separated wage patterns by age at marriage. We find that
men who marry after age 26 (the top third of the distribution) experience no unusual wage gains
prior to or during marriage: their wages continue to grow at the expected rate throughout this
period. By contrast, men who marry at younger ages experience faster-than-expected wage
growth leading up to marriage and through the first several years of their marriages. For these
men, the years immediately prior to marriage are the late teens and early 20s, a time when we
might expect the most rapid wage changes. By contrast, men who marry at age 27 or later are
more likely to be approaching marriage after their improvements in wages have stabilized and
are well-captured by our controls. These results support the claim that marriage is associated
with wage gains simply because the timing of marriage is correlated with a transition to
adulthood. It may also be consistent with delay of marriage until financial thresholds are met,
which may especially affect younger men, who have lower average wages.
[Figure 2 about here]
In the second panel of Figure 2, we test whether the premarital increase in wages is
shared by men who enter shotgun marriages, which are likely to be less deliberately timed and
anticipated. We find that men entering shotgun marriages experience premarital wage gains
similar to other grooms. Again, the results cast doubt on the claim that premarital wage gains are
due to anticipation of entering marriage. Since premarital conceptions may also lead couples to
marry under less than ideal financial circumstances, the results also suggest that the premarital
increase in wages is not entirely due to men delaying marriage until financial thresholds are met.
18
Instead, the pattern is most consistent with the argument that the premarital wage gains shared by
husbands with anticipated and less anticipated marriages is due to marriage often occurring in the
middle the transition to adulthood, when wages are rapidly growing.
Potential effects on those who never marry
The possible benefits of marriage, including for those not currently married, has attracted
both policy and scholarly interest (for an overview, see Nock 2005). Even if marriage has no
effect on the wages of men who currently marry, men who did not marry might have benefited if
they had. To shed light on this possibility, we estimated wage patterns surrounding marriage for
men who were unlikely to marry, yet married nonetheless. If those who do not marry would
benefit by marrying, we might also expect greater wage growth at marriage among those who
were unlikely to marry than among other married men.
We created a subsample of respondents who were not already married in 1979 and were
observed past age 40. We estimated a logistic regression model predicting whether these
respondents ever married based on their marital expectations in 1979, educational expectations in
1979, race, AFQT score, mother’s education, and whether the respondent lived with both parents
at age 14 (for more detail, see the online supplement). We stratified respondents by the predicted
probability of marriage: the first quartile (less than 0.74), the middle 50 percent (0.74 to 0.88),
and the fourth quartile (over 0.88). Figure 3 displays results from distributed fixed-effects wage
models for each stratum. There is no evidence of a larger wage premium for men less likely to
marry, casting doubt on the hypothesis that men who do not marry would experience wage gains
from marriage.
[Figure 3 about here]
19
Supplementary models
To test the robustness of our results, we estimated a variety of supplementary models.
The results are shown in the online supplement.
Our main models excluded men missing wage data because they were not employed
since the last interview. We estimated alternative models that imputed missing wage values with
(1) the most recent reported wage and (2) the wage as predicted by the respondent’s age
interacted with race, education, AFQT score, and age at marriage, along with indicators for
survey year, plus the 5th percentile of residuals to simulate an unusually low wage draw. Results
were similar to those in the main text.
Our decision to estimate year-specific wage patterns through 10 years after marriage
follows Dougherty (2006). We also do so for reasons of attrition: the sample size is 2,857 1 year
after marriage, by 10 years after marriage has declined by about one half, and by 15 years after
marriage has declined by almost 75 percent. However, Cheng (Forthcoming) argues that part of
the marriage wage premium develops 15-20 years after marriage and that this pattern provides
evidence that the observed association between marriage and wages is at least partly causal. To
replicate those findings, we estimated a model with a time window of five years before to 20
years after marriage. We, too, observe an increase in wages at 15-20 years of marriage, but the
estimates are imprecise due to small sample sizes and we cannot reject the null hypothesis of no
unusual wage growth during this period. Therefore, the observed wage increases in this period do
not provide strong evidence of a causal marriage premium, especially since it is unclear why
marriage would lead to additional wage gains only in this period, after a decade of relative
stability. Similarly, we estimated models examining wages from five years before to 20 years
20
after divorce. The results are consistent with a decline in wages predating divorce, and the
estimates for later years are very noisy.
We considered that marriage may have a causal effect on wages for some men, perhaps
those who exhibit the most traditional family arrangements or have more traditional attitudes.
However, we found no evidence that men with nonemployed or part-time employed wives or
men who agreed that “It is much better for everyone concerned if the man is the achiever outside
the home and the woman takes care of the home and family” experienced greater wage gains in
the years surrounding marriage.
In 1982, unmarried respondents were asked whether they expected to marry within a
year. We refer to men who expected to marry within the year, yet did not marry through 1984, as
being “left at the altar.” If men who expected marriage but did not marry fare just as well as men
who married as expected, this suggests that being married does not benefit men’s wages. We
matched those who were left at the altar with those who married as expected on race, education,
age, and AFQT score (restricting the sample to those with valid AFQT scores). Inference is
tentative given the small sample sizes (82 men per group), but the results suggest similar
trajectories regardless of whether one actually marries. Using a similar approach, we examined
the wage trajectories of men who married in 1982-1984, comparing the outcomes of those who
did and did not report in 1982 that they expected to marry within the year. Again, the sample size
is small (152 men per group), but we find no difference in wage patterns. These results further
support our conclusion that anticipation of marriage is not driving the premarital increase in
wages, since wage gains in the years surrounding marriage are shared regardless of whether
marriage was expected, and that being married has no additional effect on wages, since men left
at the altar experience similar wage outcomes as men who actually marry.
21
In all models we control for cohabitation. We also estimated models evaluating wage
patterns relative to the start of a first coresidential partnership (either marriage or cohabitation)
and found extremely similar results. We also examined transitions to second marriages. The
estimates are less precise because of the smaller sample size, but, if anything, suggest a drift
upward in wages that predates the marriage itself, similar to first marriages. This may suggest
that divorced men time remarriage after they have recovered from the wage decline that predated
their divorce, but more research, both qualitative and quantitative, is needed to explore whether
remarriage patterns follow similar logics as first marriages in terms of the role of economic
circumstances.
Because the marriage models censor individuals after divorce and the divorce models
censor individuals prior to marriage and after remarriage, it is possible that wage trends are due
to changing sample composition. In an alternative specification, we dropped the censoring
restriction. The results were similar to those in the main text.
We also estimated separate models by race and education subgroups to test the robustness
of our results. Although the subgroup-specific results are of course less precise, we found that
the rate of unexpected wage growth prior to marriage and through the first few years is greater
for Hispanic men, but similar between blacks and whites. Men with less than a high school
degree experience no rapid wage increases in the years surrounding marriage, but patterns for all
other education levels are similar. It is possible that those with less than a high school degree do
not experience wage growth through the transition to adulthood because they work low-paying
jobs with little opportunity for advancement. The subgroup-specific patterns may be read as
inconsistent with the reverse causality perspective: we might have expected larger premarital
gains for less economically advantaged groups of men, who are more likely to be hampered by
22
not meeting financial standards deemed necessary for marriage. On the other hand, it is possible
that these standards vary by subgroup. Especially given the imprecision of our estimates, we
leave further exploration of subgroup variation to future research, but we find no evidence of a
causal effect of marriage on wages for any subgroup.
Conclusion
In this paper, we bring together three related but largely independent lines of research:
the study of the effect of marriage on men’s wages, the literature on the economic determinants
of marriage, and the literature on the increasingly long and heterogeneous transition to
adulthood. Consistent with recent research, we find that men’s wages begin to increase more
rapidly than expected beginning at least five years prior to marriage (Cheng Forthcoming,
Dougherty 2006).
This pattern contrasts with the notion that marriage leads to an immediate, one-time
increase in men’s wages, as is implicitly assumed by conventional fixed-effects models. Like
Dougherty (2006), we find that the rate of wage growth returns to expected levels by several
years after marriage. Thus, our results are also inconsistent with claims that marriage leads to an
enduring acceleration in men’s wage growth (Cheng Forthcoming, Ginther and Zavodny 2001,
Korenman and Neumark 1991). Instead, our results reveal that the years immediately
surrounding marriage are a time of unusually rapid improvements in wages. Minimally, these
patterns are inconsistent with the wage patterns hypothesized by household specialization or
employer discrimination, neither of which predicts a premarital increase in men’s wages.
Cheng (Forthcoming) and Dougherty (2006) interpret the premarital increase in wages as
revealing a greater marriage premium than estimated by prior research. However, research on the
23
effect of men’s economic circumstances on marriage and on the rapid changes in both work and
family life men experience during the transition to adulthood suggest possible confounding
processes that may lead to a positive association between marriage and wages in the absence of a
causal effect. Our analyses seek to adjudicate between these three possible processes: a causal
effect of marriage on men’s wages, reverse causality of men’s wages on marriage, and a spurious
association driven by the transition to adulthood.
For divorce, we show that wage losses estimated by conventional fixed-effects models in
fact predate divorce, suggesting that prior evidence of a divorce wage penalty is likely due to
failure to account for this reverse causality. For marriage, we first reiterate that, if the premarital
wage gains are due to men’s anticipation of entering marriage, this anticipation begins at least
five years prior to marriage.
Second, we show that men who marry after age 26 do not experience a premarital wage
increase. It is possible that anticipation of marriage acts to “settle down” young men, while by
the late 20s men experience this transition in work regardless of marital status. In this case,
marriage might shift men’s period of wage growth earlier, while their peers catch up later.
However, we believe that the observed variation by age at marriage is more consistent with the
transition to adulthood argument, which suggests that the spurious association between marriage
and rapid wage growth is particularly likely for men marrying in the late teens and early 20s.
Men who marry later are more likely to have already experienced that period of rapid wage gain
prior to their approach to marriage.
Third, this interpretation is reinforced by our finding that the premarital increase in wages
is shared by men who enter shotgun marriages and men whose marriages are not as rapidly
followed by a birth. If the five-year period of rapid wage increases before marriage is due to
24
anticipation, we would not expect that men whose marriage is likely to be less anticipated would
share fully in these anticipatory effects. Furthermore, we would not expect that men entering
shotgun marriages have as much discretion as other men about delaying marriage until financial
thresholds are met. Our results thus cast doubt on the explanation that the premarital increase in
wages is due to marriage being timed to follow economic improvements. This does not mean
men’s economic circumstances do not matter for marriage, only that we do not believe this effect
is responsible for the premarital increase in wages. Likewise, our results by no means rule out the
possibility that marriage affects other measures of financial well-being, such as wealth or
household income.
Of course, our analyses are not without limitations. Due to sample size, we are limited in
our ability to test the effects of rarer transitions (such as to second marriages) and to estimate
longer-term trends (such as more than a decade after marriage). We recognize that it is
impossible to establish causal effects with complete certainty in the absence of random
assignment. If we allow that marriage’s effects may predate the marriage, this challenge is
heightened, as it is unclear what the relevant comparisons are to estimate a causal effect.
However, we provide a variety of analyses that consistently support the argument that men
simply tend to marry at times when wages are already increasingly particularly rapidly. For each
result, it is possible to tell an alternative story of how the observed pattern could still be
consistent with a causal effect of marriage, although that effect may be limited to particular years
or particular subgroups. Taking our results together, however, shows how complicated and
limited such a causal story would have to be to be consistent with our results.
Because our main focus is questioning whether marriage affects wages, we are less
concerned with adjudicating between the reverse causality and transition to adulthood
25
perspectives. We believe that our results favor the transition to adulthood interpretation for
marriage and the reverse causality perspective for divorce, but our core claim – that there is no
causal effect of anticipating, getting, or staying married on men’s wages – does not depend on
which alternative process is at work.
Our results are limited to a particular cohort of men. Given changes in age at marriage,
premarital cohabitation, nonmarital childbearing, and the timing of the transition to adulthood
(Cherlin 2010), different patterns may appear for more recent cohorts. The birth cohorts of 19801984, which are followed by the NLSY97, will soon be old enough to update the findings
presented here. We anticipate that the premarital increase in wages may be diminished for this
group, given increasing marital delay.
Our results are also informed by context because, while we find no individual-level effect
of marriage on men’s wages, we are unable to evaluate the effect of “marriage culture” on all
men’s wages, regardless of whether they personally marry. For example, if wage-earning is
normative for husbands, it may also become a more general expectation of adult masculinity,
particularly if most men expect to marry and therefore perceive themselves as potential future
husbands, regardless of whether they eventually marry. Because 96 percent of the men in our
sample expect to marry, we cannot separately identify the effects of marital expectations on
men’s wages, or what men’s wages would be if the marriage rate were substantially lower.
Our results have implications for research on the economic determinants of marriage. If
wages rise prior to marriage for spurious reasons, positive associations between men’s current
economic circumstances and marriage in the next year may be biased just as the wage premium
is. Alternatively, if men do time marriage in response to changes in financial circumstances,
more research is needed to understand which aspects of financial change and achievement, or
26
which financial thresholds, are particularly relevant to enabling marriage, and whether these
thresholds vary by subgroup.
Although we cannot provide definitive evidence that marriage does not benefit men’s
wages, our analyses show that the empirical patterns of men’s wage changes in the years
surrounding marriage are entirely consistent with an alternative explanation: men marry during a
transition to adulthood when their wages are already improving, and they tend to divorce after
their wages are already declining. Given that premarital wage gains are inconsistent with
explanations for the marriage premium grounded in specialization or discrimination, the finding
in prior research that men’s anticipation of marriage and desire to take on the breadwinner role
leads to a cumulative, long-term gain of 20 percent in wages suggests an incredible ability for
men to alter their financial circumstances through changes in their incentives and motivation.
Instead, we argue that there may be no causal effect of anticipating, getting, or staying married
on men’s wages. Marriage often occurs during a time of rapid transformation in men’s wages,
but we find no reason to believe that marriage itself is the transformative force.
27
References
Arnett, Jeffrey Jensen. 2015. Emerging Adulthood: The Winding Road from the Late Teens
through the Twenties, 2nd ed. New York: Oxford University Press.
Becker, Gary S. 1991. A Treatise on the Family, Enl. ed. Cambridge, MA: Harvard University
Press.
Brand, Jennie E and Yu Xie. 2010. "Who Benefits Most from College? Evidence for Negative
Selection in Heterogeneous Economic Returns to Higher Education." American
Sociological Review 75(2):273-302.
Charles, Kerwin Kofi and Melvin Stephens, Jr. 2004. "Job Displacement, Disability, and
Divorce." Journal of Labor Economics 22(2):489-522.
Cheng, Siwei. Forthcoming. "The Accumulation of Dis(Advantage): The Intersection of Gender
and Race in the Long-Term Effect of Marriage." American Sociological Review.
Cherlin, Andrew J. 2010. "Demographic Trends in the United States: A Review of Research in
the 2000s." Journal of Marriage and Family 72(3):403-19.
Chun, H. and I. Lee. 2001. "Why Do Married Men Earn More: Productivity or Marriage
Selection?". Economic Inquiry 39(2):307-19.
Cohen, Philip N. 2002. "Cohabitation and the Declining Marriage Premium for Men." Work and
Occupations 29(3):346-63.
Cornwell, Christopher and Peter Rupert. 1997. "Unobservable Individual Effects, Marriage and
the Earnings of Young Men." Economic Inquiry 35(2):285-94.
Dougherty, Christopher. 2006. "The Marriage Earnings Premium as a Distributed Fixed Effect."
Journal of Human Resources 41(2):433-43.
28
Edin, Kathryn and Maria Kefalas. 2005. Promises I Can Keep: Why Poor Women Put
Motherhood before Marriage. Berkeley, CA: University of California Press.
Ginther, Donna K and Madeline Zavodny. 2001. "Is the Male Marriage Premium Due to
Selection? The Effect of Shotgun Weddings on the Return to Marriage." Journal of
Population Economics 14(2):313-28.
Gorman, Elizabeth H. 1999. "Bringing Home the Bacon: Marital Allocation of Income-Earning
Responsibility, Job Shifts, and Men’s Wages." Journal of Marriage and the Family
61(1):110-22.
Gorman, Elizabeth H. 2000. "Marriage and Money: The Effect of Marital Status on Attitudes
toward Pay and Finances." Work and Occupations 27(1):64–88.
Gray, Jeffrey S. 1997. "The Fall in Men's Return to Marriage: Declining Productivity Effects or
Changing Selection?". Journal of Human Resources 32(3):481-504.
Gupta, Sanjiv. 1999. "The Effects of Transitions in Marital Status on Men’s Performance of
Housework." Journal of Marriage and the Family 61(3):700–11.
Hersch, Joni and Leslie S Stratton. 2000. "Household Specialization and the Male Marriage
Wage Premium." Industrial and Labor Relations Review 54(1):78-94.
Hogan, Dennis P and Nan Marie Astone. 1986. "The Transition to Adulthood." Annual Review of
Sociology 12:109-30.
Killewald, Alexandra. 2013. "A Reconsideration of the Fatherhood Premium: Marriage,
Coresidence, Biology, and Fathers’ Wages." American Sociological Review 78(1):96116.
Killewald, Alexandra and Margaret Gough. 2013. "Does Specialization Explain Marriage
Penalties and Premiums?". American Sociological Review 78(3):477-502.
29
Korenman, Sanders and David Neumark. 1991. "Does Marriage Really Make Men More
Productive?". Journal of Human Resources 26(2):282-307.
Loh, Eng Seng. 1996. "Productivity Differences and the Marriage Wage Premium for White
Males." Journal of Human Resources 31(3):566-89.
Loughran, David S. and Julie M. Zissimopoulos. 2009. "Why Wait?: The Effect of Marriage and
Childbearing on the Wages of Men and Women." Journal of Human Resources
44(2):326-49.
Nock, Steven L. 1998. Marriage in Men's Lives. New York: Oxford University Press.
Nock, Steven L. 2005. "Marriage as a Public Issue." The Future of Children 15(2):13-32.
Oppenheimer, Valerie Kincade, Matthijs Kalmijn and Nelson Lim. 1997. "Men’s Career
Development and Marriage Timing During a Period of Rising Inequality." Demography
34(3):311-30.
Oppenheimer, Valerie Kincade. 2003. "Cohabiting and Marriage During Young Men’s CareerDevelopment Process." Demography 40(1):127-49.
Pollmann-Schult, Matthias. 2011. "Marriage and Earnings: Why Do Married Men Earn More
Than Single Men?". European Sociological Review 27(2):147-63.
Rindfuss, Ronald R. 1991. "The Young Adult Years: Diversity, Structural Change, and Fertility."
Demography 28(4):493-512.
Shanahan, Michael J. 2000. "Pathways to Adulthood in Changing Societies: Variability and
Mechanisms in Life Course Perspective." Annual Review of Sociology 26:667-92.
Smock, Pamela J., Wendy D. Manning and Meredith Porter. 2005. "'Everything’s There except
Money’: How Money Shapes Decisions to Marry among Cohabitors." Journal of
Marriage and Family 67(3):680–96.
30
Sweeney, Megan M. 2002. "Two Decades of Family Change: The Shifting Economic
Foundations of Marriage." American Sociological Review 67(1):132–47.
U.S. Department of Labor Bureau of Labor Statistics. 2014. "National Longitudinal Survey of
Youth 1979 Cohort, 1979–2012." Columbus, OH: Center for Human Resource Research,
The Ohio State University.
Waite, Linda J. and Maggie Gallagher. 2000. The Case for Marriage: Why Married People Are
Happier, Healthier, and Better Off Financially. New York: Broadway Books.
Xie, Yu, James M. Raymo, Kimberly Goyette and Arland Thornton. 2003. "Economic Potential
and Entry into Marriage and Cohabitation." Demography 40(2):351-67.
31
Table 1. Theories relating marriage to wages
Theory
(1)
Causal marriage premium:
Anticipation
(2)
Reverse causality:
Wages affect marriage
Marriage
Marriage
(3)
Spurious association:
Transition to adulthood
Marriage
Transition to
adulthood
Wages
Wages
Wages
Men’s work behavior changes due to
marriage, including in anticipation of
taking on the husband role.
Men delay marriage until financial
stability is achieved.
During the transition to adulthood, men
mature rapidly, often marry, and
experience wage increases.
Cheng (in press), Dougherty (2006),
Pollman-Schult (2011), Waite and
Gallagher (2000)
Edin and Kefalas (2005), Oppenheimer
(2003), Oppenheimer, Kalmijn and Lim
(1997), Smock, Manning and Porter
(2005), Sweeney (2002), Xie et al. (2003)
Arnett (2004), Hogan and Astone (1986),
Shanahan (2000)
Wages increase 1-2 years prior to
marriage.
Wages increase 1-2 years prior to
marriage.
Wages increase over 3+ years prior to
marriage.
Marriage exit
Wages decline after divorce.
Wages decline 1-2 years before divorce.
Wages do not decline at divorce.
Heterogeneity by
timing
Wages increase prior to marriage
regardless of marriage timing.
Wages increase most among those who
marry young.
Wages increase most among those who
marry young.
Shotgun
marriage
Wages do not increase prior to a shotgun
marriage.
Wages do not increase prior to a shotgun
marriage.
Wages increase prior to a shotgun
marriage.
Relevant literature
Empirical
predictions
Marriage entry
Table 2. Descriptive statistics of sample, by marital status
Marriage premium
Divorce penalty
Never married
Married
Married, will divorce
Divorced
Cohabiting
0.05
0.21
Number of children
0.03
0.98
0.72
0.26
(0.23)
(1.04)
(0.94)
(0.66)
Age
23.17
32.65
28.53
36.66
(3.97)
(8.43)
(6.19)
(8.64)
Potential experience
2.95
12.09
8.58
16.62
(3.49)
(8.15)
(5.94)
(8.43)
Education
Less than high school
0.15
0.13
0.19
0.10
Exactly high school
0.46
0.42
0.50
0.53
Some college
0.18
0.19
0.18
0.19
4+ years of college
0.21
0.26
0.13
0.14
Race
Hispanic
0.06
0.06
0.07
0.06
Black
0.12
0.11
0.13
0.13
White/other
0.83
0.83
0.80
0.79
Hourly wage (2012 $)
15.82
23.43
18.68
20.44
(9.40)
(15.74)
(11.01)
(13.97)
Total N in analytic sample
17,977
32,352
9,909
5,440
Note: Samples are restricted to observations on employed respondents who are included in the models in Figure 1. Observations are
weighted using sample weights to be representative of the national population within the age range of the cohort. The married
columns refer to individuals in a first marriage. The divorced column refers to those whose first marriage has ended and who have
not yet remarried.
Figure 1. Distributed fixed effect of marriage and divorce on wages
Divorce
(N = 1,841)
20%
5%
15%
0%
% change in wages
% change in wages
Marriage
(N = 4,424)
10%
5%
0%
-5%
-10%
-15%
-20%
-5%
-5
0
5
Years since marriage
Distributed FE
10
95% CI
-3 -2 -1 0 1 2 3 4 5
Years since divorce
Distributed FE
95% CI
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. Figures show
the predicted percent change in wages, relative to the value 5 or more years prior to marriage or 3 or more years
prior to divorce. The marriage model excludes respondents who never marry and observations after the end of a first
marriage. The divorce model excludes respondents who never divorce, observations prior to the start of a first
marriage, and observations after the start of a second marriage. Conventional fixed-effects models with dummy
variables for current marital status show a 4.0% marriage premium and a 4.7% divorce penalty. Full results are
available in the online supplement, Tables A1 and A2.
Figure 2. Heterogeneous wage patterns: Anticipation of marriage or a transition to adulthood?
By age at marriage
By shotgun marriage
40%
16%
35%
14%
12%
% change in wages
% change in wages
30%
25%
20%
15%
10%
5%
10%
8%
6%
4%
0%
2%
-5%
-5
0
5
Years since marriage
10
0%
-5
0
5
10
Years since marriage
Marries at age 22 or under (N = 1,846)
Marries at age 23-26 (N = 1,352)
Non-shotgun marriage (N = 3,017)
Marries at age 27+ (N = 1,390)
Shotgun marriage (N = 586)
Note: Models are distributed fixed-effects models analogous to those in Figure 1. The analysis by age at marriage
begins at 3 years before marriage because those who marry at age 22 or younger do not have valid wage
observations more years prior to marriage. Models are estimated separately within subgroups defined by terciles of
the age at first marriage distribution (left figure) and by whether the marriage was a shotgun marriage, defined as a
marriage followed by a first birth within 7 months (right figure). Full results are available in the online supplement,
Table A3.
Figure 3. Heterogeneity in the distributed effect of marriage by the probability of marriage
By probability of marriage
14%
% change in wages
12%
10%
8%
6%
4%
2%
0%
-2%
-4%
-5
0
5
10
Years since marriage
1st quartile of P(marriage)
(N = 632)
Middle 50% of P(marriage)
(N =1,599)
4th quartile of P(marriage)
(N = 897)
Note: Models are distributed fixed-effects models analogous to those in Figure 1. Models are estimated separately
within subgroups defined by the propensity of marriage, and respondents who were already married in 1979, were
not observed past age 40, or were missing AFQT percentile were excluded. The first quartile is all propensity scores
below 0.74, and the third quartile is all propensity scores above 0.88. Full results from the logistic regression model
used to predict marriage are available in the online supplement, Table A4, and full results from the wage model
stratified by the predicted probability of marriage are available in the online supplement, Table A5.
Online Supplement
Figure A1. Distributed fixed effect of marriage and divorce on wages, imputed with prior wage
Divorce
(N = 1,848)
20%
5%
15%
0%
% change in wages
% change in wages
Marriage
(N = 4,424)
10%
5%
0%
-5%
-10%
-15%
-20%
-5%
-5
0
5
Years since marriage
Distributed FE
10
95% CI
-3 -2 -1 0 1 2 3 4 5
Years since divorce
Distributed FE
95% CI
Note: Samples include non-employed respondents with wages imputed at their most recent values within the past three years. Models are linear regression
models with person fixed-effects estimated with sampling weights. Figures show the predicted percent change in wages, relative to the value 5 or more years
prior to marriage or 3 or more years prior to divorce. The marriage model excludes respondents who never marry and censors observations after the end of a first
marriage. The divorce model excludes respondents who never divorce, observations prior to the start of a first marriage, and observations after the start of a
second marriage. Conventional fixed-effects models with dummy variables for current marital status show a 4.2% marriage premium and a 5.6% divorce penalty.
Full results are available in Tables A1 and A2.
Figure A2. Distributed fixed effect of marriage and divorce on wages, missing wages imputed with predicted wage from regression
Divorce
(N = 1,861)
14%
12%
10%
8%
6%
4%
2%
0%
-2%
-4%
% change in wages
% change in wages
Marriage
(N = 4,463)
-5
0
5
Years since marriage
Distributed FE
10
95% CI
2%
0%
-2%
-4%
-6%
-8%
-10%
-12%
-14%
-16%
-18%
-3 -2 -1 0 1 2 3 4 5
Years since divorce
Distributed FE
95% CI
Note: Samples include non-employed respondents with wages imputed at their predicted values from a linear regression model conditioning on age interacted
with race, education, AFQT percentile, and age at marriage, as well as indictors for survey year and individual fixed effects. Models are linear regression models
with person fixed-effects estimated with sampling weights. Figures show the predicted percent change in wages, relative to the value 5 or more years prior to
marriage or 3 or more years prior to divorce. The marriage model excludes respondents who never marry and censors observations after divorce. The divorce
model excludes respondents who never divorce, censors observations prior to the start of a first marriage, and censors observations after the start of a second
marriage. Conventional fixed-effects models with dummy variables for current marital status show a 3.6% marriage premium and a 4.4% divorce penalty. Full
results are available in Tables A1 and A2.
Figure A3. Distributed fixed effect of marriage on wages, with longer time window
Marriage, longer window
Sample size by years since
marriage
45%
3,000
40%
35%
25%
% change in wages
% change in wages
2,500
30%
20%
15%
10%
5%
2,000
1,500
1,000
0%
500
-5%
-5
0
5
10
15
Years since marriage
Distributed FE
95% CI
20
0
-5
0
5
10
15
20
Years since marriage
Note: Model is a distributed fixed-effects models analogous to that in Figure 1. Although the point estimates rise 15-20 years after marriage, we cannot reject the
null hypothesis that the coefficients representing those years are all equal. Full results are available in Table A6.
Figure A4. Distributed fixed effect of divorce on wages, with longer time window
Divorce, longer window
Sample size by years since
divorce
10%
1,200
5%
0%
1,000
-10%
% change in wages
% change in wages
-5%
-15%
-20%
-25%
-30%
-35%
-40%
800
600
400
200
-45%
-5
0
5
10
15
Years since divorce
Distributed FE
95% CI
20
0
-5
0
5
10
15
20
Years since divorce
Note: Model is a distributed fixed-effects models analogous to that in Figure 1. Although the graph suggests the possibility of a divorce penalty that develops 510 years after divorce, we cannot reject the null hypothesis that the coefficients representing those years are all equal. Full results are available in Table A6.
Figure A5. Distributed fixed effect of marriage on wages, by gender role attitudes and wife’s employment
By gender role attitudes
By wife's employment
10%
% change in wages
% change in wages
8%
6%
4%
2%
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
-0.02
-0.04
-5
0%
-2%
-5
0
5
Years since marriage
Egalitarian
(N = 2,134)
Traditional
(N = 2,239)
10
0
5
Years since marriage
10
Wife not employed
(N = 960)
Wife employed part-time
(N = 729)
Wife employed full-time
(N = 2,682)
Note: Model is a linear regression model with person fixed-effects estimated with sampling weights, analogous to that in Figure 1. Respondents were classified
as traditional or egalitarian on the basis of agreement or disagreement with the statement, “It is much better for everyone concerned if the man is the achiever
outside the home and the woman takes care of the home and family.” Wife’s employment is measured in the first married observation on each respondent and
filled in with the same value for surrounding years to stratify the sample in three time-invariant groups. Full-time employment is defined as 35 or more hours per
week. Full results are available in Table A7.
Figure A6. Distributed fixed effect of marriage on wages, by expectations in 1982
Among those who expect
marriage, does following
through matter?
Among those who marry in
1982-1984, does expecting it
matter?
Exp(mean log hourly wage)
Exp(mean log hourly wage)
$18.00
$17.00
$16.00
$15.00
$14.00
$13.00
$12.00
$11.00
$20.00
$19.00
$18.00
$17.00
$16.00
$15.00
$14.00
$13.00
$12.00
$11.00
$10.00
1975
$10.00
1975
1980
1985
1990
1995
1980
1985
1990
1995
Survey year
Survey year
Marries in 1982-1984
(N = 82)
Left at altar
(N = 82)
Does not expect marriage
(N = 152)
Expects marriage
(N = 152)
Note: Both samples are restricted to respondents who were interviewed in 1982, and for whom AFQT percentile was not missing. In the left panel, the sample is
further restricted to respondents who expected in 1982 that they would be married in a year, and these respondents are stratified by whether they actually married
by the end of 1984. In the right panel, the sample is restricted to respondents who married in 1982-1984, and it is stratified by whether respondents expected in
1982 that they would be married within a year. In each figure, the two groups are matched on AFQT percentile with a caliper of 20 percentile points, within strata
defined by age, race, and education. Results presented are the mean of log hourly wages in each year, exponentiated to ease interpretation. Results showing the
balance achieved on the covariates are available in Table A8.
Figure A7. Distributed fixed effect of second marriage and coresidence on wages
2nd Marriage
(N = 1,070)
Coresidence
(N = 5,176)
25%
25%
20%
% change in wages
% change in wages
20%
15%
10%
5%
0%
-5%
15%
10%
5%
0%
-10%
-15%
-5%
-3 -2 -1 0 1 2 3 4 5
Years since 2nd marriage
Distributed FE
95% CI
-5
0
5
10
Years since coresidence
Distributed FE
95% CI
Note: The first year of coresidence is defined to be the first year in which the respondent reports a coresidential partner, regardless of whether the couple is
married. Models are distributed fixed-effects models analogous to those in Figure 1, except that the coresidence model does not condition on an indicator for
cohabitation since this is built into the years since coresidence term. In the 2nd marriage model, observations prior to 1st divorce and after 2nd divorce are
censored. Full results are available in Table A9.
Figure A8. Distributed fixed effect of marriage and divorce on wages, without censoring
Marriage
(N = 4,609)
Divorce
(N = 2,045)
20%
4%
2%
15%
0%
-2%
10%
-4%
5%
-6%
-8%
0%
-10%
-12%
-5%
-5
0
5
Years since marriage
Distributed FE
10
95% CI
-3 -2 -1 0 1 2 3 4 5
Years since divorce
Distributed FE
95% CI
Note: Models are distributed fixed-effects models analogous to those in Figure 1, except that the marriage model does not censor respondents at divorce, and the
divorce model does not censor respondents prior to their first marriage or after remarriage. Full results are available in Tables A1 and A2.
Figure A9. Heterogeneity in the distributed effect of marriage on wages by education and race
By education
By race
15%
14%
12%
10%
% change in wages
% change in wages
10%
5%
0%
-5%
-10%
8%
6%
4%
2%
0%
-2%
-15%
-5
0
5
Years since marriage
10
-4%
-5
0
5
10
Years since marriage
Less than high school (N = 914)
High school (N = 1,995)
Hispanic (N = 743)
Some college (N = 772)
Black (N = 962)
College (N = 743)
White (N = 2,719)
Note: Models are distributed fixed-effects models analogous to those in Figure 1, estimated separately by education category prior to marriage or by race. Full
results are available in the appendix, Table A10 and A11.
Table A1. Distributed fixed effect of marriage on wages, with alternate imputation methods and censoring restrictions
(1)
(2)
(3)
Log wages,
Log wages, missing
Log wages, missing
no imputation
values imputed with
values imputed with
prior report
predicted values
Years since first marriage
(reference category is 5+ years before marriage)
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 years before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
0.006
(0.017)
= 1.006
0.008
(0.019)
= 1.008
0.022
(0.019)
= 1.022
0.028
(0.020)
= 1.029
0.043*
(0.021)
= 1.044
0.059*
(0.023)
= 1.061
0.063*
(0.025)
= 1.066
0.073**
(0.025)
= 1.076
0.068*
(0.029)
= 1.071
0.064*
(0.030)
= 1.066
0.073*
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.011
(0.016)
= 1.011
0.001
(0.020)
= 1.001
0.008
(0.021)
= 1.008
0.016
(0.022)
= 1.016
0.032
(0.021)
= 1.032
0.049*
(0.023)
= 1.050
0.053*
(0.025)
= 1.054
0.068**
(0.025)
= 1.070
0.063*
(0.029)
= 1.065
0.057
(0.030)
= 1.059
0.064*
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.004
(0.016)
= 1.004
0.005
(0.018)
= 1.005
0.023
(0.018)
= 1.023
0.028
(0.019)
= 1.029
0.044*
(0.020)
= 1.045
0.059**
(0.022)
= 1.061
0.059*
(0.023)
= 1.060
0.064**
(0.024)
= 1.066
0.061*
(0.027)
= 1.063
0.056*
(0.028)
= 1.057
0.064*
(4)
Log wages, no
imputation, without
censoring
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.010
(0.017)
= 1.010
0.017
(0.019)
= 1.017
0.032
(0.018)
= 1.033
0.041*
(0.019)
= 1.042
0.060**
(0.020)
= 1.062
0.075***
(0.022)
= 1.078
0.076***
(0.023)
= 1.079
0.088***
(0.023)
= 1.092
0.085***
(0.025)
= 1.089
0.078**
(0.026)
= 1.082
0.088**
7 years after marriage
8 years after marriage
9 years after marriage
10+ years after marriage
Cohabiting
Number of children
(0.031)
𝑒 𝛽 = 1.075
0.071*
(0.032)
𝑒 𝛽 = 1.073
0.068*
(0.034)
𝑒 𝛽 = 1.070
0.074*
(0.036)
𝑒 𝛽 = 1.076
0.064
(0.039)
𝑒 𝛽 = 1.066
0.032
(0.017)
𝑒 𝛽 = 1.033
0.026***
(0.006)
𝑒 𝛽 = 1.026
(0.031)
𝑒 𝛽 = 1.066
0.065*
(0.033)
𝑒 𝛽 = 1.067
0.058
(0.034)
𝑒 𝛽 = 1.060
0.064
(0.036)
𝑒 𝛽 = 1.067
0.054
(0.040)
𝑒 𝛽 = 1.055
0.029
(0.016)
𝑒 𝛽 = 1.030
0.026***
(0.006)
𝑒 𝛽 = 1.027
(0.029)
𝑒 𝛽 = 1.066
0.054
(0.031)
𝑒 𝛽 = 1.056
0.057
(0.032)
𝑒 𝛽 = 1.059
0.055
(0.034)
𝑒 𝛽 = 1.056
0.049
(0.037)
𝑒 𝛽 = 1.051
0.035*
(0.016)
𝑒 𝛽 = 1.036
0.028***
(0.006)
𝑒 𝛽 = 1.029
(0.027)
𝑒 𝛽 = 1.092
0.089**
(0.028)
𝑒 𝛽 = 1.093
0.083**
(0.029)
𝑒 𝛽 = 1.087
0.090**
(0.030)
𝑒 𝛽 = 1.094
0.071*
(0.033)
𝑒 𝛽 = 1.074
0.032**
(0.011)
𝑒 𝛽 = 1.033
0.033***
(0.005)
𝑒 𝛽 = 1.034
0.082
(0.047)
= 1.085
-0.052
(0.044)
= 0.949
0.118
(0.070)
= 1.125
0.146***
(0.026)
= 1.158
-0.018
(0.013)
= 0.982
0.048**
(0.015)
= 1.049
0.094*
(0.048)
= 1.099
-0.052
(0.044)
= 0.950
0.127
(0.070)
= 1.135
0.148***
(0.027)
= 1.159
-0.021
(0.014)
= 0.979
0.049**
(0.015)
= 1.050
0.046
(0.041)
= 1.047
-0.013
(0.040)
= 0.987
0.188**
(0.064)
= 1.207
0.156***
(0.024)
= 1.169
-0.007
(0.012)
= 0.993
0.036**
(0.014)
= 1.036
0.108**
(0.038)
= 1.114
-0.031
(0.040)
= 0.970
0.102
(0.063)
= 1.107
0.149***
(0.023)
= 1.161
-0.007
(0.012)
= 0.993
0.035**
(0.013)
= 1.036
Education
Less than high school
𝑒𝛽
Some college
𝑒𝛽
College
𝑒𝛽
Potential experience (log)
𝑒𝛽
x less than high school
𝑒𝛽
x some college
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
x college
𝑒𝛽
x AFQT percentile
𝑒𝛽
x missing AFQT
𝑒𝛽
x Hispanic
𝑒𝛽
x black
𝑒𝛽
x marries at age 23-26 (reference is <23)
𝑒𝛽
x marries at age 27+
𝑒𝛽
0.041*
(0.020)
= 1.042
0.001***
(0.000)
= 1.001
-0.002
(0.033)
= 0.998
0.007
(0.013)
= 1.007
-0.041***
(0.012)
= 0.960
0.020
(0.015)
= 1.021
-0.004
(0.015)
= 0.996
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.042*
(0.020)
= 1.042
0.001***
(0.000)
= 1.001
0.003
(0.031)
= 1.003
0.008
(0.013)
= 1.008
-0.041**
(0.013)
= 0.960
0.020
(0.015)
= 1.020
-0.003
(0.015)
= 0.997
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.030
(0.018)
= 1.030
0.001***
(0.000)
= 1.001
-0.001
(0.028)
= 0.999
0.014
(0.012)
= 1.014
-0.034**
(0.011)
= 0.966
0.023
(0.013)
= 1.023
0.006
(0.013)
= 1.006
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.048**
(0.017)
= 1.050
0.001***
(0.000)
= 1.001
0.003
(0.029)
= 1.003
-0.003
(0.011)
= 0.997
-0.037***
(0.011)
= 0.964
0.014
(0.012)
= 1.014
-0.005
(0.013)
= 0.995
Observations
50,516
52,906
55,409
65,649
R-squared
0.612
0.607
0.634
0.582
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. Models additionally condition on a series of indicators
for survey year. To improve interpretability for the log wage outcome, we include exponentiated coefficients which show the factor by which wages are
predicted to change with a unit change in the covariate, relative to 5 or more years prior to marriage. Models (1)-(3) exclude respondents who never marry and
censor observations after the end of a first marriage, and Model (4) excludes those who never marry but does not censor at divorce. Model (1) includes only
employed men, model (2) imputes missing wages with the most recent previous wage report within three years, and model (3) imputes missing wages with
predicted values from a linear regression model conditioning on age interacted with race, education, AFQT percentile, and age at marriage, as well as indictors
for survey year and individual fixed effects.
Table A2. Distributed fixed effect of divorce on wages, with alternate imputation methods and censoring restrictions
(1)
(2)
(3)
Log wages,
Log wages, missing
Log wages, missing
no imputation
values imputed with
values imputed with
prior report
predicted values
Years since first divorce
(reference category is 3+ years before divorce)
2 years before divorce
𝑒𝛽
1 years before divorce
𝑒𝛽
Year of divorce
𝑒𝛽
1 year after divorce
𝑒𝛽
2 years after divorce
𝑒𝛽
3 years after divorce
𝑒𝛽
4 years after divorce
𝑒𝛽
5+ years after divorce
𝑒𝛽
Cohabiting
𝑒𝛽
Number of children
𝑒𝛽
Education
-0.054
(0.031)
= 0.947
-0.040
(0.023)
= 0.961
-0.062*
(0.024)
= 0.940
-0.074*
(0.032)
= 0.928
-0.102**
(0.035)
= 0.903
-0.077*
(0.031)
= 0.926
-0.116**
(0.038)
= 0.890
-0.117**
(0.042)
= 0.889
0.083***
(0.023)
= 1.087
0.015
(0.009)
= 1.016
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.051
(0.029)
= 0.950
-0.034
(0.023)
= 0.967
-0.058*
(0.023)
= 0.944
-0.093**
(0.035)
= 0.911
-0.100**
(0.034)
= 0.904
-0.077*
(0.030)
= 0.926
-0.106**
(0.036)
= 0.899
-0.120**
(0.042)
= 0.887
0.090***
(0.024)
= 1.095
0.017
(0.009)
= 1.017
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.054
(0.028)
= 0.948
-0.042
(0.022)
= 0.959
-0.058**
(0.022)
= 0.943
-0.066*
(0.030)
= 0.936
-0.091**
(0.032)
= 0.913
-0.086**
(0.029)
= 0.918
-0.111**
(0.035)
= 0.895
-0.110**
(0.039)
= 0.895
0.085***
(0.022)
= 1.089
0.016
(0.008)
= 1.016
(4)
Log wages, no
imputation, without
censoring
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.040
(0.024)
= 0.961
-0.026
(0.019)
= 0.974
-0.024
(0.018)
= 0.976
-0.035
(0.022)
= 0.965
-0.028
(0.023)
= 0.973
-0.035
(0.024)
= 0.966
-0.034
(0.023)
= 0.967
-0.054*
(0.024)
= 0.947
0.039**
(0.013)
= 1.039
0.036***
(0.007)
= 1.036
Less than high school
𝑒𝛽
Some college
𝑒𝛽
College
𝑒𝛽
Potential experience (log)
𝑒𝛽
x less than high school
𝑒𝛽
x some college
𝑒𝛽
x college
𝑒𝛽
x AFQT percentile
𝑒𝛽
x missing AFQT
𝑒𝛽
x Hispanic
𝑒𝛽
x black
𝑒𝛽
x marries at age 23-26 (reference is <23)
𝑒𝛽
x marries at age 27+
𝑒𝛽
0.159
(0.101)
= 1.173
0.060
(0.116)
= 1.062
0.024
(0.175)
= 1.025
0.115
(0.068)
= 1.122
0.005
(0.036)
= 1.005
0.006
(0.035)
= 1.006
0.034
(0.055)
= 1.034
0.002*
(0.001)
= 1.002
0.036
(0.048)
= 1.036
-0.007
(0.028)
= 0.993
-0.039
(0.030)
= 0.961
-0.001
(0.037)
= 0.999
-0.042
(0.064)
= 0.959
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.145
(0.098)
= 1.156
0.083
(0.123)
= 1.087
0.006
(0.192)
= 1.006
0.127
(0.072)
= 1.135
0.005
(0.036)
= 1.005
0.001
(0.036)
= 1.001
0.017
(0.060)
= 1.017
0.002*
(0.001)
= 1.002
0.042
(0.050)
= 1.042
-0.007
(0.028)
= 0.993
-0.037
(0.031)
= 0.963
-0.002
(0.042)
= 0.998
-0.029
(0.068)
= 0.971
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.153
(0.091)
= 1.166
0.028
(0.104)
= 1.029
0.035
(0.163)
= 1.036
0.131*
(0.064)
= 1.141
-0.004
(0.032)
= 0.996
0.010
(0.033)
= 1.011
0.038
(0.052)
= 1.039
0.002**
(0.001)
= 1.002
0.027
(0.046)
= 1.028
-0.000
(0.025)
= 1.000
-0.053
(0.027)
= 0.949
0.010
(0.033)
= 1.010
-0.033
(0.058)
= 0.968
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.109*
(0.049)
= 1.116
0.034
(0.058)
= 1.035
0.188
(0.106)
= 1.207
0.215***
(0.032)
= 1.240
0.003
(0.016)
= 1.003
0.005
(0.017)
= 1.005
0.012
(0.028)
= 1.012
0.001**
(0.000)
= 1.001
0.017
(0.036)
= 1.017
-0.020
(0.015)
= 0.980
-0.033*
(0.015)
= 0.968
0.024
(0.015)
= 1.024
0.018
(0.016)
= 1.019
Observations
15,458
16,337
17,090
31,186
R-squared
0.580
0.579
0.610
0.531
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. Models additionally condition on a series of indicators
for survey year. To improve interpretability for the log wage outcome, we include exponentiated coefficients which show the factor by which wages are
predicted to change with a unit change in the covariate, relative to 5 or more years prior to marriage. Models (1)-(3) exclude respondents who never divorce
and censors observations before the start of a first marriage or after the start of a second marriage, and Model (4) excludes those who never divorce but does
not censor prior to the start of a first marriage or after the start of a second marriage. Model (1) includes only employed men, model (2) imputes missing wages
with the most recent previous wage report within three years, and model (3) imputes missing wages with predicted values from a linear regression model
conditioning on age interacted with race, education, AFQT percentile, and age at marriage, as well as indictors for survey year and individual fixed effects.
Table A3. Distributed fixed effect of marriage on wages, by age at marriage and shotgun marriage
(1)
(2)
(3)
Marries at age 22 or
Marries at age 23-26
Marries at age 27+
under
(4)
Non-shotgun marriage
Years since first marriage
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 year before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
𝑒𝛽
0.031
(0.046)
= 1.031
0.089
(0.056)
= 1.094
0.183**
(0.061)
= 1.201
0.230**
(0.071)
= 1.258
0.263***
(0.077)
= 1.301
0.290***
(0.085)
= 1.336
0.312***
(0.093)
= 1.366
0.323**
(0.100)
= 1.382
0.335**
(0.107)
= 1.397
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.065*
(0.025)
= 1.067
0.088**
(0.032)
= 1.092
0.100*
(0.040)
= 1.105
0.133**
(0.047)
= 1.142
0.142*
(0.057)
= 1.153
0.170**
(0.060)
= 1.185
0.164*
(0.069)
= 1.178
0.132
(0.079)
= 1.141
0.164*
(0.083)
= 1.178
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.018
(0.018)
= 1.018
0.022
(0.021)
= 1.023
0.025
(0.024)
= 1.026
0.024
(0.024)
= 1.024
0.020
(0.027)
= 1.020
0.016
(0.032)
= 1.016
0.003
(0.043)
= 1.003
0.039
(0.035)
= 1.039
0.038
(0.037)
= 1.038
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.003
(0.020)
= 1.003
-0.001
(0.024)
= 0.999
0.017
(0.023)
= 1.017
0.022
(0.024)
= 1.022
0.037
(0.027)
= 1.037
0.059*
(0.028)
= 1.061
0.065*
(0.030)
= 1.067
0.078*
(0.031)
= 1.081
0.069*
(0.035)
= 1.072
0.063
(0.037)
= 1.065
0.073*
(0.037)
= 1.076
(5)
Shotgun marriage
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.074
(0.060)
= 1.076
0.066
(0.059)
= 1.068
0.074
(0.060)
= 1.077
0.119
(0.064)
= 1.126
0.102
(0.066)
= 1.108
0.086
(0.074)
= 1.090
0.109
(0.072)
= 1.115
0.132
(0.079)
= 1.141
0.151
(0.086)
= 1.164
0.164
(0.089)
= 1.178
0.194*
(0.095)
= 1.214
7 years after marriage
𝑒𝛽
8 years after marriage
𝑒𝛽
9 years after marriage
𝑒𝛽
10+ years after marriage
𝑒𝛽
Cohabiting
𝑒𝛽
Number of children
𝑒𝛽
0.334**
(0.114)
= 1.397
0.348**
(0.119)
= 1.417
0.361**
(0.124)
= 1.434
0.384**
(0.134)
= 1.468
0.053
(0.053)
= 1.054
0.017
(0.009)
= 1.017
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.182*
(0.089)
= 1.200
0.152
(0.095)
= 1.165
0.196
(0.104)
= 1.216
0.147
(0.122)
= 1.158
0.005
(0.030)
= 1.005
0.025*
(0.012)
= 1.025
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.025
(0.043)
= 1.026
0.052
(0.042)
= 1.053
0.017
(0.051)
= 1.018
0.023
(0.051)
= 1.023
0.027
(0.020)
= 1.028
0.028*
(0.011)
= 1.028
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.089*
(0.039)
= 1.093
0.071
(0.041)
= 1.074
0.085
(0.044)
= 1.088
0.069
(0.048)
= 1.072
0.027
(0.023)
= 1.027
0.021**
(0.007)
= 1.022
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.143
(0.098)
= 1.153
0.155
(0.105)
= 1.167
0.156
(0.103)
= 1.168
0.149
(0.118)
= 1.161
-0.034
(0.061)
= 0.967
0.022
(0.013)
= 1.023
Observations
15,970
14,955
19,591
33,662
5,706
R-squared
0.598
0.645
0.597
0.618
0.644
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. The reference category is 3 or more years before
marriage in models (1)-(3) and 5 or more years before marriage in models (4) and (5). We use a shorter time window in the analysis by age because those who
marry at age 22 or younger do not have valid wage observations more years prior to marriage. Age categories are defined by the terciles of the age at marriage
distribution. A marriage is defined as a shotgun marriage if it is followed by a first birth within 7 months. Models exclude respondents who never marry and
censor observations after divorce. As in Table A1, models additionally condition on log potential experience interacted with education, AFQT percentile, race,
and age at marriage, as well as a series of indicators for survey year. To improve interpretability, we include exponentiated coefficients which show the factor
by which wages are predicted to change with a unit change in the covariate.
Table A4. Logistic regression model predicting marriage
(1)
Marriage
Age respondent expects to marry
< 20
20-24
25-29
30+
Never
Race (other omitted)
Hispanic
Black
Linear spline for AFQT percentile
(0, 25)
(25, 50)
(50, 75)
(75, 100)
Missing AFQT score
Mother’s education (high school omitted)
Less than high school
Some college
College
No mother
0.727*
(0.315)
OR = 2.068
0.713***
(0.202)
OR = 2.039
0.341
(0.200)
OR = 1.407
0.216
(0.218)
OR = 1.241
-0.149
(0.405)
OR = 0.862
-0.408**
(0.125)
OR = 0.665
-0.917***
(0.111)
OR = 0.400
0.021**
(0.007)
OR = 1.021
0.004
(0.008)
OR = 1.004
-0.004
(0.009)
OR = 0.997
-0.003
(0.012)
OR = 0.997
0.250
(0.203)
OR = 1.284
-0.096
(0.103)
OR = 0.908
-0.027
(0.168)
OR = 0.973
-0.070
(0.184)
OR = 0.933
-0.291
(0.730)
OR = 0.747
Missing data
Highest grade respondent expects to complete
Less than high school
Some college
College
Missing
Lived with both parents at age 14
Missing data
Constant
-0.198
(0.163)
OR = 0.820
-0.151
(0.154)
OR = 0.860
0.216
(0.133)
OR = 1.241
0.163
(0.111)
OR = 1.178
0.167
(0.428)
OR = 1.182
0.149
(0.089)
OR = 1.160
-0.374
(0.884)
OR = 0.688
0.876***
(0.248)
OR = 2.402
Observations
4,020
Note: Models are logistic regression models estimated with sampling weights. Each respondent provides 1
observation: whether that respondent is ever observed to marry. Model excludes respondents who were already
married in 1979, were not observed past age 40, or were missing AFQT percentile were excluded. To improve
interpretability, we include exponentiated coefficients (odds ratios) which show the factor by which the odds of
marriage are predicted to change with a unit change in the covariate.
Table A5. Distributed fixed effect of marriage on log wages, by propensity score strata
(1)
(2)
1st quartile of
Middle 50% of
P(marriage)
P(marriage)
Years since first marriage
(reference category is 5+ years
before marriage)
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 years before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
𝑒𝛽
7 years after marriage
𝑒𝛽
8 years after marriage
𝑒𝛽
9 years after marriage
𝑒𝛽
10+ years after marriage
𝑒𝛽
Cohabiting
0.040
(0.033)
= 1.040
0.038
(0.033)
= 1.038
0.017
(0.037)
= 1.017
0.029
(0.043)
= 1.029
0.062
(0.042)
= 1.064
0.066
(0.045)
= 1.068
0.057
(0.047)
= 1.058
0.061
(0.048)
= 1.063
0.003
(0.062)
= 1.003
0.037
(0.054)
= 1.038
0.063
(0.059)
= 1.065
0.051
(0.062)
= 1.052
0.030
(0.067)
= 1.031
0.029
(0.065)
= 1.029
0.083
(0.072)
= 1.086
0.070*
(0.028)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.004
(0.026)
= 0.996
-0.020
(0.033)
= 0.980
0.000
(0.028)
= 1.000
0.013
(0.029)
= 1.013
0.038
(0.032)
= 1.039
0.042
(0.034)
= 1.043
0.045
(0.037)
= 1.046
0.055
(0.039)
= 1.056
0.040
(0.045)
= 1.041
0.052
(0.043)
= 1.053
0.009
(0.045)
= 1.009
0.040
(0.049)
= 1.041
0.040
(0.050)
= 1.041
0.039
(0.055)
= 1.039
-0.001
(0.059)
= 0.999
0.037
(0.025)
(3)
4th quartile of
P(marriage)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.005
(0.027)
= 1.005
0.029
(0.031)
= 1.030
0.037
(0.034)
= 1.038
0.032
(0.037)
= 1.032
0.031
(0.039)
= 1.032
0.065
(0.043)
= 1.067
0.081
(0.045)
= 1.084
0.090
(0.046)
= 1.094
0.099
(0.051)
= 1.104
0.086
(0.053)
= 1.090
0.120*
(0.056)
= 1.127
0.114
(0.059)
= 1.121
0.086
(0.063)
= 1.090
0.106
(0.065)
= 1.112
0.119
(0.075)
= 1.126
0.006
(0.033)
Number of children
𝑒 𝛽 = 1.073
0.015
(0.011)
𝑒 𝛽 = 1.015
𝑒 𝛽 = 1.038
0.042***
(0.011)
𝑒 𝛽 = 1.043
𝑒 𝛽 = 1.006
0.017
(0.010)
𝑒 𝛽 = 1.017
Observations
7,957
21,127
12,337
R-squared
0.544
0.607
0.595
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. See Table
A4 for the logistic regression model used to predict the probability of marriage and define the propensity score
strata. Models are estimated separately within subgroups defined by the propensity of marriage, and respondents
who were already married in 1979, were not observed past age 40, were missing AFQT percentile, or never
married were excluded. Observations after divorce are also censored. As in Table A1, models additionally
condition on log potential experience interacted with education, AFQT percentile, race, and age at marriage, as
well as a series of indicators for survey year. To improve interpretability, we include exponentiated coefficients
which show the factor by which wages are predicted to change with a unit change in the covariate.
Table A6. Distributed fixed effect of marriage and divorce on log wages, with a longer time window
(1)
(2)
Years since marriage
Years since divorce
Years since first event
(reference category is 5+ years before
event)
4 years before event
𝑒𝛽
3 years before event
𝑒𝛽
2 years before event
𝑒𝛽
1 years before event
𝑒𝛽
Year of event
𝑒𝛽
1 year after event
𝑒𝛽
2 years after event
𝑒𝛽
3 years after event
𝑒𝛽
4 years after event
𝑒𝛽
5 years after event
𝑒𝛽
6 years after event
𝑒𝛽
7 years after event
𝑒𝛽
8 years after event
𝑒𝛽
9 years after event
𝑒𝛽
10 years after event
𝑒𝛽
11 year after event
𝑒𝛽
0.017
(0.017)
= 1.018
0.022
(0.020)
= 1.022
0.039*
(0.019)
= 1.040
0.048*
(0.020)
= 1.050
0.067**
(0.022)
= 1.069
0.087***
(0.024)
= 1.090
0.095***
(0.027)
= 1.099
0.108***
(0.028)
= 1.114
0.107***
(0.032)
= 1.113
0.107**
(0.033)
= 1.112
0.119***
(0.035)
= 1.127
0.120**
(0.037)
= 1.128
0.123**
(0.039)
= 1.130
0.131**
(0.042)
= 1.140
0.120**
(0.046)
= 1.128
0.123**
(0.046)
= 1.131
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.024
(0.024)
= 0.976
-0.036
(0.023)
= 0.965
-0.085*
(0.036)
= 0.918
-0.074*
(0.030)
= 0.929
-0.101**
(0.033)
= 0.904
-0.115**
(0.041)
= 0.891
-0.149***
(0.044)
= 0.862
-0.124**
(0.043)
= 0.883
-0.171***
(0.049)
= 0.843
-0.136**
(0.050)
= 0.873
-0.171**
(0.056)
= 0.843
-0.119
(0.076)
= 0.888
-0.184**
(0.065)
= 0.832
-0.135*
(0.067)
= 0.874
-0.265***
(0.079)
= 0.767
-0.190*
(0.080)
= 0.827
12 years after event
𝑒𝛽
13 years after event
𝑒𝛽
14 years after event
𝑒𝛽
15 years after event
𝑒𝛽
16 years after event
𝑒𝛽
17 years after event
𝑒𝛽
18 years after event
𝑒𝛽
19 years after event
𝑒𝛽
20+ years after event
𝑒𝛽
Cohabiting
𝑒𝛽
Number of children
𝑒𝛽
0.145**
(0.048)
= 1.156
0.091
(0.052)
= 1.095
0.154**
(0.055)
= 1.167
0.136*
(0.056)
= 1.145
0.160**
(0.058)
= 1.174
0.140*
(0.062)
= 1.150
0.187**
(0.065)
= 1.205
0.187**
(0.065)
= 1.206
0.204**
(0.070)
= 1.227
0.034*
(0.017)
= 1.034
0.026***
(0.006)
= 1.027
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.312**
(0.121)
= 0.732
-0.308**
(0.106)
= 0.735
-0.281**
(0.100)
= 0.755
-0.286**
(0.093)
= 0.751
-0.213
(0.116)
= 0.808
-0.250**
(0.093)
= 0.779
-0.243
(0.138)
= 0.784
-0.260*
(0.102)
= 0.771
-0.287*
(0.117)
= 0.750
0.085***
(0.023)
= 1.088
0.015
(0.009)
= 1.015
Observations
50,516
15,458
R-squared
0.613
0.582
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. The
reference category is 5 or more years before marriage or divorce. As in Table A1, models additionally condition
on log potential experience interacted with education, AFQT percentile, race, and age at marriage, as well as a
series of indicators for survey year. The marriage model excludes respondents who never marry and censors
observations after the end of a first marriage. The divorce model excludes respondents who never divorce, censors
observations prior to the start of a first marriage, and censors observations after the start of a second marriage. To
improve interpretability, we include exponentiated coefficients which show the factor by which wages are
predicted to change with a unit change in the covariate.
Table A7. Distributed fixed effect of marriage on wages, by gender role attitudes and wife’s employment
(1)
(2)
(3)
Egalitarian gender
Traditional gender
Wife not employed in
role attitudes
role attitudes
first married
observation
Years since first marriage
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 year before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
0.019
(0.025)
= 1.020
0.002
(0.031)
= 1.002
0.042
(0.026)
= 1.043
0.043
(0.029)
= 1.044
0.068*
(0.031)
= 1.071
0.064
(0.033)
= 1.066
0.080*
(0.034)
= 1.084
0.090*
(0.036)
= 1.095
0.074
(0.042)
= 1.077
0.081
(0.042)
= 1.084
0.089*
(0.042)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.009
(0.023)
= 0.991
0.013
(0.023)
= 1.014
0.003
(0.027)
= 1.003
0.014
(0.027)
= 1.014
0.020
(0.029)
= 1.020
0.056
(0.031)
= 1.058
0.047
(0.035)
= 1.048
0.054
(0.036)
= 1.056
0.061
(0.039)
= 1.063
0.044
(0.041)
= 1.045
0.052
(0.043)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.033
(0.039)
= 1.034
0.006
(0.069)
= 1.006
0.042
(0.041)
= 1.043
0.047
(0.048)
= 1.048
0.089
(0.051)
= 1.093
0.095
(0.056)
= 1.099
0.153*
(0.060)
= 1.166
0.106
(0.064)
= 1.111
0.107
(0.067)
= 1.113
0.096
(0.073)
= 1.100
0.125
(0.077)
(4)
Wife employed parttime in first married
observation
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.002
(0.053)
= 0.998
0.010
(0.054)
= 1.010
-0.001
(0.057)
= 0.999
-0.004
(0.053)
= 0.996
-0.006
(0.060)
= 0.994
0.039
(0.064)
= 1.040
0.012
(0.069)
= 1.012
-0.007
(0.073)
= 0.993
-0.011
(0.081)
= 0.989
0.005
(0.083)
= 1.005
-0.009
(0.087)
(5)
Wife employed fulltime in first married
observation
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.001
(0.020)
= 0.999
-0.002
(0.021)
= 0.998
0.013
(0.023)
= 1.013
0.024
(0.025)
= 1.024
0.035
(0.026)
= 1.036
0.047
(0.028)
= 1.048
0.048
(0.030)
= 1.049
0.081**
(0.030)
= 1.084
0.073*
(0.035)
= 1.076
0.064
(0.035)
= 1.066
0.076*
(0.036)
7 years after marriage
8 years after marriage
9 years after marriage
10+ years after marriage
Cohabiting
Number of children
𝑒 𝛽 = 1.094
0.071
(0.046)
𝑒 𝛽 = 1.074
0.077
(0.048)
𝑒 𝛽 = 1.080
0.078
(0.051)
𝑒 𝛽 = 1.081
0.081
(0.057)
𝑒 𝛽 = 1.084
0.017
(0.026)
𝑒 𝛽 = 1.017
0.029**
(0.009)
𝑒 𝛽 = 1.029
𝑒 𝛽 = 1.054
0.068
(0.046)
𝑒 𝛽 = 1.071
0.056
(0.048)
𝑒 𝛽 = 1.058
0.064
(0.049)
𝑒 𝛽 = 1.066
0.040
(0.053)
𝑒 𝛽 = 1.041
0.053**
(0.020)
𝑒 𝛽 = 1.054
0.024**
(0.008)
𝑒 𝛽 = 1.025
𝑒 𝛽 = 1.133
0.085
(0.085)
𝑒 𝛽 = 1.088
0.119
(0.082)
𝑒 𝛽 = 1.126
0.135
(0.085)
𝑒 𝛽 = 1.145
0.104
(0.094)
𝑒 𝛽 = 1.110
0.076*
(0.031)
𝑒 𝛽 = 1.079
0.034**
(0.012)
𝑒 𝛽 = 1.035
𝑒 𝛽 = 0.991
0.007
(0.092)
𝑒 𝛽 = 1.007
-0.011
(0.095)
𝑒 𝛽 = 0.989
0.000
(0.106)
𝑒 𝛽 = 1.000
-0.010
(0.108)
𝑒 𝛽 = 0.990
-0.025
(0.034)
𝑒 𝛽 = 0.975
0.023
(0.017)
𝑒 𝛽 = 1.024
𝑒 𝛽 = 1.079
0.076*
(0.037)
𝑒 𝛽 = 1.079
0.070
(0.040)
𝑒 𝛽 = 1.072
0.072
(0.041)
𝑒 𝛽 = 1.075
0.066
(0.047)
𝑒 𝛽 = 1.068
0.028
(0.023)
𝑒 𝛽 = 1.028
0.022**
(0.008)
𝑒 𝛽 = 1.023
Observations
24,534
25,372
10,437
9,122
29,418
R-squared
0.587
0.642
0.570
0.618
0.614
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. The reference category is 5 or more years before
marriage. Models exclude respondents who never marry and censors observations after divorce. Models (3)-(5) additionally exclude respondents who are
married but missing data on wife’s employment hours. As in Table A1, models additionally condition on log potential experience interacted with education,
AFQT percentile, race, and age at marriage, as well as a series of indicators for survey year. Respondents were classified as egalitarian or traditional on the
basis of agreement or disagreement with the statement, “It is much better for everyone concerned if the man is the achiever outside the home and the woman
takes care of the home and family.” Wife’s employment is measured in the first married observation on each respondent, and filled in with the same value for
surrounding years to stratify the sample in three time-invariant groups. Full-time employment is defined as 35 or more hours per week. To improve
interpretability, we include exponentiated coefficients which show the factor by which wages are predicted to change with a unit change in the covariate.
Table A8. Matching results based on expectations of marriage within a year, reported in 1982
Given expected marriage within a year
Given married in 1982-1984, by
in 1982, by whether was left at the altar
whether expected marriage within a
year in 1982
Married in 1982Left at the altar
Did not expect
Expected
1984
marriage
marriage
Race (matched exactly)
Hispanic
Black
White/other
Education in 1982 (matched
exactly)
Less than high school
High school
Some college
College
Age in 1982 (matched exactly)
AFQT percentile (matched to
nearest neighbor)
0.21
0.23
0.56
0.21
0.23
0.56
0.16
0.10
0.74
0.16
0.10
0.74
0.38
0.48
0.12
0.02
21.04
(1.91)
0.38
0.48
0.12
0.02
21.04
(1.91)
0.28
0.53
0.13
0.06
21.19
(1.89)
0.28
0.53
0.13
0.06
21.19
(1.89)
27.59
(22.73)
23.04
(22.31)
39.50
(28.11)
36.27
(27.32)
Respondents
82
82
152
152
Note: Both samples are restricted to respondents who were interviewed in 1982, and for whom AFQT percentile
was not missing. In the left two columns, the sample is further restricted to respondents who expected in 1982 that
they would be married in a year, and these respondents are stratified by whether they actually married by the end
of 1984. In the right two columns, the sample is restricted to respondents who married in 1982-1984, and it is
stratified by whether respondents expected in 1982 that they would be married within a year. In both analyses, the
two groups are matched on AFQT percentile with a caliper of 20 percentile points, within strata defined by age,
race, and education.
Table A9. Distributed fixed effect of 2nd marriage and 1st coresidence on wages
(1)
Wages on years since 2nd
marriage
(2)
Wages on years since 1st
coresidence
Years since event
4 years before event
𝑒𝛽
3 years before event
𝑒𝛽
2 years before event
𝑒𝛽
1 year before event
𝑒𝛽
Year of event
𝑒𝛽
1 year after event
𝑒𝛽
2 years after event
𝑒𝛽
3 years after event
𝑒𝛽
4 years after event
𝑒𝛽
5 years after event (5+ years in Model 1)
𝑒𝛽
-0.005
(0.048)
= 0.995
0.045
(0.035)
= 1.046
0.049
(0.036)
= 1.051
0.057
(0.044)
= 1.059
0.082
(0.042)
= 1.086
0.093*
(0.046)
= 1.097
0.072
(0.049)
= 1.075
0.057
(0.054)
= 1.059
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
6 years after event
𝑒𝛽
7 years after event
𝑒𝛽
8 years after event
𝑒𝛽
9 years after event
𝑒𝛽
10+ years after event
𝑒𝛽
Cohabiting
Number of children
0.035
(0.027)
𝑒 𝛽 = 1.035
0.011
0.010
(0.018)
= 1.010
0.025
(0.019)
= 1.025
0.053**
(0.020)
= 1.054
0.072***
(0.020)
= 1.074
0.088***
(0.021)
= 1.092
0.108***
(0.022)
= 1.114
0.102***
(0.024)
= 1.107
0.127***
(0.025)
= 1.136
0.125***
(0.026)
= 1.133
0.118***
(0.028)
= 1.125
0.127***
(0.029)
= 1.135
0.133***
(0.030)
= 1.142
0.117***
(0.031)
= 1.124
0.139***
(0.032)
= 1.150
0.117***
(0.035)
= 1.124
0.032***
(0.015)
𝑒 𝛽 = 1.011
(0.005)
𝑒 𝛽 = 1.032
Observations
7,724
72,403
R-squared
0.600
0.582
Note: The first year of coresidence is defined to be the first year in which the respondent reports a coresidential
partner, regardless of whether the couple is married. The second marriage model excludes respondents who never
re-marry, censors observations prior to first divorce, and censors observations after second divorce. The
coresidence model excludes those who never report a coresidential partner. The coresidence model does not
condition on an indicator for cohabitation since this is built into the years since coresidence term. Models are
linear regression models with person fixed-effects estimated with sampling weights. The reference category is 3 or
more years before remarriage in model (1) and 5 or more years before coresidence in model (2). As in Table A1,
models additionally condition on log potential experience interacted with education, AFQT percentile, race, and
age at marriage, as well as a series of indicators for survey year. To improve interpretability, we include
exponentiated coefficients which show the factor by which wages are predicted to change with a unit change in the
covariate.
Table A10. Distributed fixed effect of marriage on wages, by education level prior to marriage
(1)
(2)
(3)
Less than high
High school
Some college
school
Years since first marriage
(reference category is 5+ years before
marriage)
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 years before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
𝑒𝛽
7 years after marriage
𝑒𝛽
8 years after marriage
𝑒𝛽
9 years after marriage
𝑒𝛽
10+ years after marriage
𝑒𝛽
Cohabiting
-0.008
(0.043)
= 0.992
-0.034
(0.048)
= 0.966
-0.063
(0.048)
= 0.939
-0.029
(0.054)
= 0.972
-0.027
(0.057)
= 0.973
-0.023
(0.063)
= 0.977
-0.048
(0.069)
= 0.953
-0.046
(0.070)
= 0.955
-0.064
(0.077)
= 0.938
-0.079
(0.078)
= 0.924
-0.091
(0.089)
= 0.913
-0.111
(0.093)
= 0.895
-0.106
(0.097)
= 0.899
-0.103
(0.093)
= 0.902
-0.141
(0.101)
= 0.869
0.030
(0.040)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.001
(0.025)
= 0.999
0.015
(0.025)
= 1.015
0.037
(0.026)
= 1.038
0.020
(0.028)
= 1.020
0.055
(0.029)
= 1.057
0.077*
(0.031)
= 1.080
0.085*
(0.034)
= 1.089
0.105**
(0.034)
= 1.111
0.114**
(0.036)
= 1.121
0.112**
(0.039)
= 1.118
0.127**
(0.041)
= 1.136
0.107*
(0.043)
= 1.113
0.121**
(0.045)
= 1.128
0.119**
(0.046)
= 1.127
0.117*
(0.050)
= 1.124
0.052**
(0.020)
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
-0.018
(0.038)
= 0.982
0.007
(0.043)
= 1.007
-0.017
(0.042)
= 0.983
0.024
(0.049)
= 1.024
0.038
(0.050)
= 1.039
0.081
(0.055)
= 1.084
0.055
(0.058)
= 1.057
0.055
(0.062)
= 1.056
0.086
(0.071)
= 1.089
0.039
(0.080)
= 1.040
0.051
(0.076)
= 1.052
0.085
(0.079)
= 1.088
0.044
(0.086)
= 1.045
0.078
(0.087)
= 1.081
0.068
(0.096)
= 1.070
0.017
(0.027)
(4)
College
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.049
(0.039)
= 1.051
0.009
(0.053)
= 1.009
0.059
(0.047)
= 1.061
0.066
(0.047)
= 1.069
0.056
(0.056)
= 1.057
0.044
(0.058)
= 1.045
0.085
(0.061)
= 1.089
0.089
(0.066)
= 1.093
0.046
(0.078)
= 1.047
0.076
(0.074)
= 1.079
0.088
(0.076)
= 1.092
0.101
(0.082)
= 1.106
0.089
(0.084)
= 1.093
0.091
(0.096)
= 1.096
0.096
(0.108)
= 1.101
-0.009
(0.066)
Number of children
𝑒 𝛽 = 1.030
0.024
(0.013)
𝑒 𝛽 = 1.025
𝑒 𝛽 = 1.053
0.014*
(0.007)
𝑒 𝛽 = 1.015
𝑒 𝛽 = 1.017
0.035
(0.018)
𝑒 𝛽 = 1.036
𝑒 𝛽 = 0.991
0.032*
(0.016)
𝑒 𝛽 = 1.033
Observations
9,362
22,581
9,245
9,328
R-squared
0.494
0.602
0.548
0.547
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. The
reference category is 5 or more years before marriage. Models exclude respondents who never marry and censor
observations after divorce. As in Table A1, models additionally condition on log potential experience interacted
with education, AFQT percentile, race, and age at marriage, as well as a series of indicators for survey year.
Education prior to marriage is defined as the highest reported education level in a survey prior to the year of the
marriage. To improve interpretability, we include exponentiated coefficients which show the factor by which
wages are predicted to change with a unit change in the covariate.
Table A11. Distributed fixed effect of marriage on wages, by race
Years since first marriage
(reference category is 5+ years before marriage)
4 years before marriage
𝑒𝛽
3 years before marriage
𝑒𝛽
2 years before marriage
𝑒𝛽
1 years before marriage
𝑒𝛽
Year of marriage
𝑒𝛽
1 year after marriage
𝑒𝛽
2 years after marriage
𝑒𝛽
3 years after marriage
𝑒𝛽
4 years after marriage
𝑒𝛽
5 years after marriage
𝑒𝛽
6 years after marriage
𝑒𝛽
7 years after marriage
𝑒𝛽
8 years after marriage
𝑒𝛽
9 years after marriage
𝑒𝛽
10+ years after marriage
𝑒𝛽
Cohabiting
𝑒𝛽
Number of children
(1)
Hispanic
(2)
Black
0.046
(0.039)
= 1.047
0.080
(0.042)
= 1.083
0.062
(0.045)
= 1.064
0.053
(0.055)
= 1.055
0.065
(0.051)
= 1.067
0.105
(0.056)
= 1.111
0.124*
(0.058)
= 1.132
0.122*
(0.059)
= 1.129
0.112
(0.062)
= 1.118
0.112
(0.068)
= 1.119
0.119
(0.068)
= 1.126
0.095
(0.075)
= 1.100
0.105
(0.080)
= 1.111
0.117
(0.079)
= 1.124
0.092
(0.086)
= 1.096
0.012
(0.031)
= 1.012
0.020
0.023
(0.027)
= 1.023
0.021
(0.029)
= 1.021
-0.013
(0.034)
= 0.987
0.033
(0.035)
= 1.034
0.038
(0.036)
= 1.038
0.048
(0.038)
= 1.049
0.051
(0.041)
= 1.052
0.044
(0.043)
= 1.045
0.006
(0.050)
= 1.006
0.051
(0.048)
= 1.052
0.067
(0.051)
= 1.069
0.016
(0.056)
= 1.016
0.021
(0.057)
= 1.021
0.026
(0.057)
= 1.026
0.042
(0.069)
= 1.043
0.055*
(0.025)
= 1.057
0.016
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
(3)
Other
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
𝑒𝛽
0.003
(0.020)
= 1.003
0.003
(0.024)
= 1.003
0.028
(0.023)
= 1.028
0.030
(0.024)
= 1.030
0.048
(0.026)
= 1.049
0.062*
(0.028)
= 1.064
0.066*
(0.030)
= 1.068
0.079*
(0.031)
= 1.082
0.078*
(0.035)
= 1.082
0.069
(0.036)
= 1.071
0.076*
(0.037)
= 1.079
0.081*
(0.040)
= 1.085
0.077
(0.041)
= 1.080
0.082
(0.044)
= 1.086
0.071
(0.048)
= 1.073
0.028
(0.021)
= 1.028
0.028***
(0.014)
1.020
(0.009)
1.016
(0.008)
1.028
Observations
9,479
11,541
29,496
R-squared
0.581
0.563
0.612
Note: Models are linear regression models with person fixed-effects estimated with sampling weights. The
reference category is 5 or more years before marriage. Models exclude respondents who never marry and censor
observations after divorce. As in Table A1, models additionally condition on log potential experience interacted
with education, AFQT percentile, race, and age at marriage, as well as a series of indicators for survey year. To
improve interpretability, we include exponentiated coefficients which show the factor by which wages are
predicted to change with a unit change in the covariate.