## Abstract

This study investigates gender-specific changes in the total financial return to education among persons of prime working ages (35–44 years) using U.S. Census data from 1990 and 2000, and the 2009–2011 American Community Survey. We define the total financial return to education as the family standard of living as measured by family income adjusted for family size. Our results indicate that women experienced significant progress in educational attainment and labor market outcomes over this time period. Ironically, married women’s progress in education and personal earnings has led to greater improvement in the family standard of living for married men than for women themselves. Gender-specific changes in assortative mating are mostly responsible for this paradoxical trend. Because the number of highly educated women exceeds the number of highly educated men in the marriage market, the likelihood of educational marrying up has substantially increased for men over time while women’s likelihood has decreased. Sensitivity analyses show that the greater improvement in the family standard of living for men than for women is not limited to prime working-age persons but is also evident in the general population. Consequently, women’s return to education through marriage declined while men’s financial gain through marriage increased considerably.

## Introduction

In the mid-twentieth century, men were much more likely than women to earn a bachelor’s degree in the United States (Goldin et al. 2006). Since the 1980s, however, women have overtaken men in the rate of college completion (DiPrete and Buchmann 2013; Goldin et al. 2006). As of 2010, women in their late 20s have an approximate 8 percentage point advantage over their male counterparts in acquiring a bachelor’s degree (DiPrete and Buchmann 2013:2). The advantage of women in higher education now extends well beyond bachelor’s degrees to postbaccalaureate education (DiPrete and Buchmann 2013:36–38). How has this rise in the female advantage in higher education contributed to their economic well-being? Have women begun to enjoy a higher level (or at least a faster improvement) in their economic well-being compared with men because of the female educational advantage?

Previous studies have examined the return to education mainly in terms of personal earnings or wages, reporting that women’s earnings grew faster than men since the early 1970s (Morris and Western 1999). As a result, the earnings gap between men and women narrowed substantially (Leicht 2008). Although compensation in labor markets is intrinsically important, the economic well-being of women and the financial return to their education has traditionally been significantly mediated by the marriage market (Goldin 2006). To more fully account for the total financial return to women’s higher education, the economic rewards garnered in both the labor and marriage markets need to be considered simultaneously.

It might seem intuitively logical to assume that women’s economic well-being is lower than men’s because women earn less in labor markets. However, DiPrete and Buchmann (2006) demonstrated that in regard to the family standard of living (as measured by family income after adjusting for family size), women are advantaged not only in terms of the relative return to education (i.e., compared with less-educated women) but also in terms of the absolute return to education (i.e., compared with equally educated men albeit the extent of the latter advantage is small). Using data from 1964–2002, DiPrete and Buchmann (2006:12) showed that the family-level returns to higher education were trending upward faster for women than for men after 1980 when they limit their sample to 30- to 34-year-old whites. DiPrete and Buchmann suggested that the faster rise in the return to higher education over time for women compared with men is at least partially responsible for the female-friendly change in college completion.

However, the assumption that the female educational advantage directly translates into a higher standard of living for women is premature. Economic well-being reflects the incomes of all members of a family. Total family income is thus affected by patterns of assortative mating, or the educational association between spouses, which could reverse the positive trend in women’s personal socioeconomic attainments. The relative stagnation of men’s education and earnings implies that women may receive lower returns to their education in the marriage market compared with the 1980s. The fact that husbands are still the main breadwinners for most households heightens this possibility (McKinnish 2008). In this article, we investigate how the gender difference in the total financial return to education in terms of the standard of living has changed between 1990 and 2009–2011 in the United States.

## Literature Review

The labor market return to college education has increased significantly for both men and women (Kim and Sakamoto 2008). Over the last several decades, college-educated workers enjoy higher financial return in labor markets than before while less-educated workers are suffering from stagnating or even declining annual earnings (Long 2010). Several studies have shown that the college wage premium is higher for women than for men (e.g., Card and DiNardo 2002; Charles and Luoh 2003). One of the most prominent stylized facts with regard to the gender difference in the return to education is that the relative college premium is higher for women than for men (Dougherty 2005).1 Studies have also shown that the return to college education has increased faster for women than for men (DiPrete and Buchmann 2006; Grogger and Eide 1995; McCall 2000; Murphy and Welch 1992).

Although the return to education has been rising, so too has the proportion of women who have completed college (Buchmann and DiPrete 2006; DiPrete and Buchmann 2006). Women are currently more likely than men to go to college and to earn a bachelor’s or higher degree. In 1970, females accounted for 41 % of students in all degree-granting institutions, but that share increased to 57 % in 2005 (Snyder et al. 2008). The rise of women’s educational advantage does not stop at the level of the bachelor’s degree. In 2009 and 2010, more than one-half of all graduate degrees were granted to women (DiPrete and Buchmann 2013). Progress has also been seen across fields of study: gender segregation in fields of study has reduced significantly in the 1970s and 1980s (England 2010).

Since the 1990s, there have been some concerns about the stalled progress in achieving gender equality with regard to wages, occupational segregation, and segregation in fields of study (Blau and Kahn 2006; England 2010; Tomaskovic-Devey et al. 2006). Despite these issues, one area that does not show stalled progress is personal earnings. Personal weekly and annual earnings for full-time working women grew faster than for men during the 1990s and the 2000s so that the gender gap in annual earnings has continuously narrowed (Goldin 2006; Goldin et al. 2006). In regard to the standard of living (at least as conventionally measured), annual earnings is much more consequential than wage rates. Therefore, even if the gender gaps in wages or occupational segregation were to remain unchanged, one might expect that women’s standard of living would continue to grow faster than for men well into the twenty-first century. The double progress of women—in both education and annual earnings—might be presumed to accelerate a rapid improvement of women’s standard of living.

As for the total return to education, the return via the marriage market cannot be dismissed. Family formation influences and constrains the financial return to education for both genders, and likely continues to have a particularly salient impact on women’s return to education. Goldin (1997) demonstrated that for women who graduated from college between 1945 and 1960, almost one-half of the return to higher education comes in the form of a spouse with higher earnings.

This phenomenon may well have changed, however. The last few decades have seen several demographic and institutional changes in family formation (Cherlin 2004), which may have had a strong effect on the financial return to education through marriage. Differential returns by educational level may also be becoming more significant.

In general, the introduction of oral contraceptives and the reduction in fertility increased the number of women in the paid labor force and raised their annual hours worked (Bailey 2006; Lehrer and Nerlove 1986). At the same time, the rise of women in higher education has further decreased fertility (Brand and Davis 2011). The total reduction in fertility over the life course is greater for more highly educated women than for the less-educated (Musick et al. 2009). One would expect these changes to have increased women’s standard of living.

Another major and significant change relating to the total return to education is the shift in assortative mating. Role-specialization perspectives argue that a person’s economic well-being can be maximized by a gendered division of labor within the family (Becker 1991). The sharp dualism of the husband as the main breadwinner and the wife as the main homemaker has, however, waned over the last several decades (Buss et al. 2001; Cherlin 2005; Oppenheimer 1994, 1997; Sweeney 2002). With the deinstitutionalization of American marriage (Cherlin 2004), educational homogamy has risen. A well-known stylized fact is that assortative mating by education has increased over the last several decades (Schwartz and Mare 2005).

The deinstitutionalization of marriage and the accompanying change of marriage from educational hypergamy to educational homogamy suggest that women’s return to education in the marriage market will also rise. Assortative mating patterns are important causes and consequences of marriage behavior and related processes of social stratification (Mare 1991; Schwartz 2013). As assortative mating rises, the incidence of two highly educated partners earning two high incomes increases, thereby increasing their standard of living (Schwartz 2013).

On the other hand, it is possible that women’s standard of living might not improve as much as expected. Over the last several decades, men’s earnings have stagnated except among higher-income earners (Leicht 2008; Morris and Western 1999). The relative stagnation of men’s earnings for most of the distribution implies that most women may garner less return to their education in the marriage market than in earlier decades.

Another important but rather neglected factor regarding women’s standard of living is the effect of the change in assortative mating. Women’s rising educational attainment could have lowered their return to education in the marriage market. As women’s educational attainment rises, women are less likely to marry up. Partly as a consequence of this trend, educational homogamy has continuously increased in America since the 1960s (Schwartz and Mare 2005).

Recent years have seen yet another significant aspect of the change in assortative mating. The Pew Research Center reported that the share of couples in which the husband’s education exceeds his wife’s increased between 1960 and 1990, but had fallen by 20 % as of 2012 (Wang 2014). Now a record share of women live with a less-educated spouse. Educational marrying-down may reduce husbands’ contributions to the family standard of living. This rise in the rate at which men marry down could have further eroded women’s return to higher education in the marriage market in the most recent period.

These changes suggest the possibility that the rise in women’s education and the accompanying increase in their labor earnings may not be sufficient to offset the decline of the return to education in marriage market. A blurred yet still persistent gender division between housework and paid labor enhances this possibility. In spite of the notable progress in women’s earnings, men are still the main breadwinner in the majority of families (McKinnish 2008). The gender gap in earnings remains stubbornly present in both the public and private sectors (Mandel and Semyonov 2014). When husbands bring less income into a family over time, the growth in wife’s earnings might not be sufficient to raise a family’s economic well-being. The paradoxical consequence is that all the progress experienced by women over the last several decades can result in the deterioration of their overall economic well-being.

This trend conversely implies that married men’s return to education in terms of the standard of living has grown faster than for women. Men are now more likely to marry women who are more educated than themselves (Schwartz and Han 2014). These more-educated wives bring in more income into family than in previous decades. Indeed, in the marriage market, the importance of women’s earnings potential has increased over time (Sweeney and Cancian 2004). For the same level of education, men’s return to education may have increased faster than women’s.

Whether the rise in women’s education and their gain in labor earnings increases their economic well-being or paradoxically deteriorates it (because of the reduction in the return to education in the marriage market) depends on the relative size of these two effects. Rather than being some inevitable process (e.g., due to modernization), the trend in women’s standard of living is an empirical outcome reflecting demographic changes associated with the returns to education in the labor and marriage markets.

In sum, the flip side of the “rise of women” perspective—which predicts that the standard of living for women has improved faster than for men—is the contrasting trend that might be thought of as its paradox. This female advantage should be associated with rising personal income, increased assortative mating, and the deinstitutionalization of marriage. Contrary to this view, the paradox envisions that the standard of living for women is deteriorating compared with men despite greater female educational attainment. This deterioration would be related to the reduction in the return to education in marriage markets for women.

## Analytic Strategy

### Data and Target Population

We use the 1990 and 2000 Integrated Public Use Microdata Series (IPUMS) 5 % sample and the IPUMS-American Community Survey (ACS) 2009–2011 three-year combined data (Ruggles et al. 2015). Our samples are limited to persons aged 35–44, which is often considered to be the prime working-age population. We focus on this age group because by that time in the life course, most persons have completed their schooling, have been employed in their major occupational career, and have had sufficient time to become married and to become engaged in the process of family formation. In other words, the age group of 35–44 provides a more accurate cross-sectional assessment of a person’s returns to education both in the labor and marriage markets.

Some studies have investigated a younger group, such as those aged 25–34 (e.g., DiPrete and Buchmann 2006; Goldin 2006). Although informative to consider, the return to education in the marriage market may not be realized among many persons aged 25–34. The proportion of persons who marry by age 30 has steadily declined in the United States (Parker et al. 2015). In the early 2000s, 56 % of men and 39 % of women were never married by age 25 (Aughinbaugh et al. 2013). The percentage decreases rapidly, however, between ages 25 and 35. By age 35, the percentage is reduced to 22.5 % for men and 15.8 % for women. The further reduction of the never married beyond age 45 is small. The advancing age at first marriage in recent decades thus suggests that the age group of 35–44 provides a more accurate assessment of a person’s return to education in the marriage market.

Furthermore, as educational attainments have increased in recent decades, many persons may still be engaged in graduate education at ages 25–34. The returns to a graduate degree cannot be accurately assessed among persons who are still enrolled in their graduate studies. Such persons have yet to obtain their major occupational employment associated with those studies. Many persons enrolled in graduate studies may marry only after completing them. Especially for persons with some graduate training, the age group of 35–44 is preferable to the age group of 25–34 in providing a more accurate assessment of a person’s returns to education in both the labor and marriage markets. Nonetheless, following the discussion of our major findings, we provide a sensitivity analysis of them by considering other age groups as well.

An additional rationale for focusing on ages 35–44 is because this age group most clearly evidences recent decades of change in gender-specific educational attainment and assortative mating patterns. Our analysis includes three cohorts of persons aged 35–44. In the 1990 data, persons aged 35–44 were born between 1946 and 1955; in the 2000 data, persons aged 35–44 were born between 1956 and 1965; and in the 2009–2011 data, persons aged 35–44 were born between 1965 and 1976 (or more specifically, 1965–1974 for the 2009 ACS; 1966–1975 for the 2010 ACS; and 1967–1976 for the 2011 ACS).2 Because these three cohorts are all at the same age range in our data, our study investigates their economic well-being while they are at approximately the same stage in the life course.

The three cohort groups of this study represent the changing gender dynamics in education and assortative mating. Men are still more educated than women for the first cohort (born in 1946–1955). Women catch up with men and have an equal level of educational attainment in the second cohort (born in 1956–1965). Women overtake men in education in the third cohort (born in 1965–1976).

As for assortative mating, women of the first cohort are more likely to marry up rather than to marry down, whereas men are more likely to marry down. However, by the time of the third cohort, the pattern is reversed: women are more likely to marry down rather than to marry up, whereas men are more likely to marry up rather than to marry down. Thus, the changes between the first and the third cohorts indicate the changes in the gender-specific return to education when women begin to outpace men in education (see also Goldin et al. 2006:135).

To avert problematic measurement error, our samples are limited to persons with positive family income (i.e., we deleted those with negative incomes). To avoid the complexity of resource sharing among members residing in extended or nontraditional families,3 we delete multigenerational families and all families in which relatives other than parents and their children live together.4 Divorced, separated, and widowed persons are deleted as well.

In sum, our target population is limited to persons aged 35–44 who are never-married singles or who are residing in nuclear families (i.e., married heterosexual parents and their children). Using this approach, we can reasonably assume that families mostly share their resources; consumption of family members is a function of total family income and family size; and investment in education has been largely realized in both the labor market and marriage markets.

### Main Dependent Variables

Our main dependent variable is equivalized income, which is measured by total family income divided by square root of family size. As a measure of economic well-being, equivalized income has been studied and advocated by many scholars, including the recent report by the Commission on the Measurement of Economic Performance and Social Progress (Stiglitz et al. 2009); a study on the return to education (DiPrete and Buchmann 2006); and a study on how assortative mating is associated with earnings inequality (Breen and Salazar 2011). In the following, we use the terms “equivalized income” and “standard of living” interchangeably. Inflation is adjusted by scaling all income figures to 2010 constant dollars. To compare the difference between the return to education in equivalized income and that in personal earnings, we also investigate annual personal earnings. In all multivariate analyses (except the decomposition analyses), incomes are log-transformed.

### Statistical Model

Using the log-transformed equivalized income as the dependent variable, we estimate the following model:
$yT=αT+∑j=1βjTEdujT+∑j=0γjTEdujT×SingleT+∑j=0δjTEdujT×SingleT×FemaleT+∑j=0θjTEdujT×MarriedT×FemaleT+∑k=1πjTXk+eT,$
1
where y refers to log-transformed equivalized income at time T. Eduj is a vector of eight levels of education: less than high school (LTHS, the reference group), high school graduate (HSG), some college (SC), associate degree (AA); bachelor’s degree (BA), master’s degree (MA), professional degree (Prof), and doctoral degree (PhD). Edu × Single is a vector of the interactions between education and never-married single. All eight levels of education are added in Edu × Single, and a female dummy variable is not controlled for in Eq. (1). Thus, the γj refer to the penalty in the standard of living for being a never-married single at each level of education compared with the married. Edu × Single × Female is a vector of the three-way interactions among education, being single, and female. The reference group for δj is equally educated, never-married single men. A statistically significant δj indicates that single women are significantly (dis)advantaged compared with equally educated single men when the other variables are held constant.

Edu × Married × Female is a vector of the three-way interactions among education, married, and female. The reference group of this set of variables is equally educated, currently married men. The estimate of θj shows the extent to which married women are (dis)advantaged in the family standard of living compared with equally educated married men. The vector of control variables, X, includes age, age squared, race (white, black, Hispanic, Asian, and other race), region (nine census regions), and migration status (natives vs. immigrants).

Equation 1 is estimated for each year. The change in the estimated coefficients between years represents the change in the relative effect of education. For example, $θBA2009–2011−θBA1990$ quantifies the change in the relative (dis)advantage of BA-educated married women compared with equally educated married men between 1990 and 2009–2011. If women’s progress makes them richer than men, then $θBA2009–2011−θBA1990$ will be positive. If men and women are equally benefitted, then it will be statistically 0. If men gain more, the difference will be negative.

In addition to the regression analyses, we decompose the change in equivalized income over time. Our decomposition yields three components, including the change in respondents’ personal earnings, the change in spouse’s earnings, and other remaining changes. We then decompose the change in equivalized income into distributional changes and coefficient changes by applying Oaxaca-Blinder decomposition methods.

## Empirical Findings

Table 1 shows the descriptive statistics. Women’s annual personal income grew about three times faster than men’s between 1990 and 2009–2011, although men still earned 69 % more than women as of 2009–2011. Substantial personal income growth is experienced by both highly educated and less-educated women. In the case of men, only highly educated workers earned more income in 2009–2011 than in 1990, while less-educated men suffered stagnant (or slightly declining) income. In terms of education, women overtook men in acquiring a degree in higher education. The proportion with a BA or higher increased by 39 % between 1990 and 2009–2011 for women, while that for men grew only by 10 %.

As we suggested earlier, an interesting phenomenon that is likely related to women’s greater gains in education is the change in assortative mating. In 1990, women were more likely to educationally marry up than to marry down. In 2010, however, women were more likely to marry down rather than marry up. Men’s assortative mating contrasts with the pattern for women. Men aged 35–44 were more likely to marry up rather than marry down in 2009–2011.

### Change in the Return to Education: Personal Income Versus Equivalized Income

Table 2 shows the estimated coefficient for being female on personal income and equivalized income after controlling for education and other demographic covariates. Over time, being female became less negative in regard to personal annual income. The female disadvantage dropped by 22 % (i.e., exp(–.710 – (–.910)) – 1 = .221) between 1990 and 2009–2011. This reduction in the gender gap in annual personal income is consistent with findings from previous research (e.g., Blau and Kahn 2006).

The relative (dis)advantage of being female is quite different for equivalized income than for personal income. In 1990, women enjoyed a higher standard of living compared with equally educated men. Surprisingly, however, women’s advantage completely dissipated in 2000 and actually reversed into a statistically significant disadvantage in 2009–2011. The deterioration in the relative standard of living for women is a clear trend in these results and is unlikely to be simply attributable to the Great Recession. In fact, the reduction in the relative standard of living for women was slightly larger in the late twentieth century rather than in the early twenty-first century. Overall, the net advantage of being female decreased by 0.122 log dollars (i.e., approximately 13 %) between 1990 and 2009–2011.

To assess whether this change varies by educational level and marital status, we estimated the full model as given by Eq. (1). Table 3 shows the regression results, which indicate the changes in relative income compared with the reference group. The bottom two sets of rows show the relative (dis)advantage of women compared with equally educated men with the same marital status after controlling for demographic covariates. Because the reference point of each coefficient in Table 3 varies, it is not straightforward to ascertain how much income changes over time for each group. To more clearly illustrate the change in the net female (dis)advantage over time for each educational group, we show the estimated net effect of being female in personal income and equivalized income in Fig. 1.

In addition, using the regression results of Table 3, we estimate the predicted income by gender, marital status, and level of education. We then compute the expected income growth between 1990 and 2009–2011. Because all incomes are log-transformed, the expected income change in Fig. 2 can be interpreted as the income growth rates between 1990 and 2009–2011.

As shown in Fig. 1, the (dis)advantage of being female has changed differently by educational level and marital status. The pattern observed in Table 2—the reduction of gender gap in personal income and the reversal of gender gap in equalized income—does not apply to all female groups. A stark contrast is evident between never-married singles and married persons regarding the change in personal income. Although the female disadvantage in personal income was substantially large at all three time points, the gap was significantly reduced for married persons at most levels of education (except MA and PhD). For example, high school–educated married women earned 64 % (i.e., e–1.016 – 1= –.638) lower personal income than high school–educated married men in 1990. The gap narrowed to 57 % (i.e., e–0.848 – 1 = –.572) in 2000, decreasing further to 52 % (i.e., e–0.734 – 1 = –.520) in 2009–2011—a 12 percentage point drop over a 20-year period. In contrast with married women, the gender gap in personal income increased for single women at most educational levels (except LTHS, Prof, and PhD) when compared with equally educated single men.

A major research concern for our study is whether the change in the relative (dis)advantage of women’s personal income is consistent with the change in their equivalized income. Both single and married women experienced a reduction in their equivalized income. The pattern of change in the standard of living for single women shown in panel c of Fig. 1 aligns well with their change in personal income shown in panel a. In the case of married women, however, our results demonstrate that the pattern of changes in personal income is not consistent with that in equivalized income. As illustrated in panel d of Fig. 1, married women appear to have enjoyed a higher standard of living than equally educated men (except for professional degree holders) in 1990. This disadvantage, however, declined steadily over the 20-year period. In 2009–2011, the advantage almost completely disappeared except for PhD-educated married women. This decline occurred despite women’s personal earnings outpacing that of men.

Unique patterns for female groups at the highest end of the educational achievement are noteworthy. As of 2009–2011, equivalized income for PhD-educated married women were still advantaged by 18 % compared with equally educated men. However, compared with 1990, their advantage in 2009–2011 declined by 13 percentage points. The only female group with a rising relative standard of living is those with a professional degree. Unlike other women, women with a professional degree encountered substantial disadvantage in equivalized income compared with men with the same degree in 1990, but this disadvantage entirely vanished in 2009–2011.

As noted earlier, the change in female’s relative (dis)advantages over time varies by marital status. The reason behind this variation is not that men’s earnings growth rates differ by marital status but rather that single women’s income growth rates are lower than those of their male counterparts. Average income for single men grew by the same rate as for married men. The further deterioration in being female for single women over time is caused by low, sometimes negative, income growth for single women. Panel a of Fig. 2 shows this pattern clearly. In contrast with single women, the income growth rates for married women are all positive (except for LTHS) and exceed those for married men as shown in panel b, which indicates that they are statistically significant at most educational levels.

Overall, these findings indicate that for most educational groups, women’s relative standard of living has deteriorated compared with their male counterparts. Why this trend has occurred is not obvious. In the following remaining analyses, we seek to shed light on this issue.

### Why Have Women’s Relative Advantages in the Standard of Living Declined?

In the case of single women, lowered personal income can account for the change in their standard of living.5 The puzzle that we focus on is the paradoxical change among married women. The female disadvantage in personal income has been reduced over time for the married women, but this female-favorable progress ironically does not improve their standard of living.

This counterintuitive outcome is not because the standard of living for women at all educational levels has deteriorated. Highly educated women have experienced a rise in equivalized income. Their relative advantage compared with men has disappeared, however, because it has been overtaken by the equivalized income of their male counterparts. In the case of women with an AA education or less, not only is the change relative, but their absolute level of equivalized income also declined between 1990 and 2009–2011. That is, the standard of living for less-educated women was lower in 2009–2011 than in 1990. Contrary to women, less-educated married men’s equivalized income increased (except for LTHS men).

The comparison of Figs. 1 and 2 makes clear that gender-specific change in the labor market cannot explain this apparently contradictory result. Married women’s relative standard of living worsened even though their personal income increased. The sources of this change are therefore likely to be associated with the return to education in the marriage market.

In regard to the return to education in the marriage market, we consider the impacts of three marriage-related behavioral patterns: assortative mating, the number of children, and double incomes. As shown in Table 1, the incidence of educationally marrying down increased for women, but the incidence of educationally marrying up decreased. The trend for men is in contrast with that for women. This changing assortative mating pattern can lower equivalized income for women. The average number of children for married women aged 35–44 slightly increased from 1.64 in 1990 to 1.75 in 2009–2011, but it did not change for men. The proportion of families with two earners decreased for both genders, but the extent of decline is slightly larger for married women than for married men. All these changes can help to explain the declining equivalized income for married women.

To understand these factors more fully, we focus on married persons in the next part of our analysis. Because the changes between 1990 and 2000 and between 2000 and 2009–2011 are consistent, we consider only 1990 and 2009–2011 for simplicity.6 We estimate the following new model:
$y=α+∑j=1βjEduj+∑j=0γjEduj×F+δD+θC+∑lρlMl+∑j=0βjTEduj×T+∑j=0γjTEduj×F×T+δTD×T+θTC×T+∑lρlTMl×T+∑kπjXk+∑kπkTXk×T+e,$
2
where y is log-transformed equivalized income. F is a dummy variable indicating female. Because we dropped the variable for singles, the estimates of Edu × F quantify married women’s (dis)advantage compared with equally educated married men. In addition to all control variables in Eq. (1), we now add three new variables: D is a dummy variable indicating that both wife and husband are working; C refers to the number of children; and M refers to a set of dummy variables indicating educational marrying up and marrying down (with the reference group being persons whose spouse has the same educational level). Educational marrying up occurs when the spouse’s level of education is higher than respondent’s in terms of the eight educational levels. T is a dummy variable for the year 2009–2011. All variables are interacted with T. Thus, the effects of Edu × F × T (i.e., $γjT$) measure the change in married women’s (dis)advantage between 1990 and 2009–2011 compared with equally educated married men after the other variables are controlled for.

We first estimate Eq. (2) without these three new control variables (i.e., D, C, and M). Then we add each variable sequentially into the model in order to ascertain how much change in the relative (dis)advantage of women in equivalized income between 1990 and 2009–2011 is accounted for by these new control variables (that is, how much $γjT$ is reduced). Table 4 shows the results.

Of the three additional covariates, the assortative mating pattern and their changing effects over time explain the largest portion of the change in women’s relative (dis)advantage for HSG, SC, AA, BA, and MA. For example, after the marital pattern is controlled for (Model 4), the entire reduction in the BA-educated women’s advantage over time is explained. That is, the decline in equivalized income is closely associated with the reduction in the proportion of women who marry up and the increase in the proportion of women who marry down. The changes in assortative mating patterns account for 17 % of the reduction in the relative advantage in equivalized income for LTHS women; 66 % for HSG women; 69 % for SC women; 81 % for AA women; 94 % for BA women; and 42 % for MA women.

Interestingly, the number of children is associated with changes at the highest educational level (i.e., PhD), whereas its effects for the other educational groups are relatively small. The effect of two incomes is significant, and its impact is increased over time. The effects of a dual income, however, do not affect the change in the standard of living for married women.

As shown in Model 5, the combined effects of assortative mating patterns, number of children, and double incomes account for most of the changes in women’s relative (dis)advantage in their standard of living for LTHS, HSG, SC, AA, and BA. The same variables explain a substantial portion of the changes for MA and PhD. However, more than one-half of the changes in MA and PhD remain unexplained. In case of professional degree holders, these additional covariates are not associated with the change in equivalized income. The rise in personal income is probably the key factor that explains the improvement in the standard of living for women with a professional degree.

These results suggest that women tend to receive a relatively smaller return to education in the marriage market than in 1990. Conversely, men tend to receive a relatively larger return to education in the marriage market than in 1990. Overall, women are no longer advantaged in terms of standard of living compared with men.

### Decomposition of the Change in Equivalized Income

Our findings raise the issue about the extent to which the change in equivalized income reflects the respondent’s personal income versus spouse’s income. Here, we investigate this issue more specifically. We decompose the sources of the change by applying Oaxaca-Blinder decomposition methods.

First, we decompose the change in the equivalized income into three income sources: (a) respondent’s own personal income; (b) spouse’s personal income; and (c) other income. This decomposition is derived as follows:
$Family IncomeNumber Family Members=Respondent Income+Spouse Income+Other IncomeNumber Family Members,$
3
where the three income sources are divided by the square root of the number of family members as shown in Eq. (3). The addition of these three income sources is equal to the equivalized income. This decomposition quantifies the change over time due to the contribution of respondent’s, spouse’s, and other income on the standard of living. Because the three log-transformed income sources do not add up to the log-transformed value of equivalized income, we use actual dollar values for this decomposition.7
Next, we calculate an Oaxaca-Blinder decomposition for respondents’ own income and spouse’s income, respectively. The change in the contribution of the respondent’s own personal income between 1990 and 2009–2011 is decomposed into (a1) compositional changes and (a2) coefficient changes as follows:
$Y¯EduR,09–11−Y¯EduR,90=∑ωEduR,09–11−ωEduR,90Z¯EdyR,90+∑Z¯EduR,09–11−Z¯EduR,90ωEduR,90,$
4
where $Y¯EduR,09–11−Y¯EduR,90$ measures the change in the mean value of the respondent’s own contribution to equivalized income between 1990 and 2009-–2011; $ωEduR,09–11−ωEduR,90$ measures the change in the coefficients; and $Z¯EduR,09–11−Z¯EduR,90$ measures the change in the mean values of the control variables. The control variables include age, age squared, race, residential region, and being immigrants. Equation (4) is repeated for each educational level. Thus, (a1) quantifies the component associated with the changes in the control variables within each educational level, and (a2) quantifies the component associated with the changes in the coefficients within each educational level. The differences in the earnings growth rate over time by educational level are therefore reflected in (a2).

The change in the contribution of spouse’s personal income is also decomposed into compositional change versus coefficient change. To measure the net impact of changing assortative mating patterns, we further decompose the compositional change into (b1), the component associated with the spouse’s educational distribution, and (b2), the component associated with other compositional change. The expected change in equivalized income due to the change in assortative mating pattern is measured by (b1). Even if the assortative mating pattern remains constant over time, the increase in the educational premium can alter the contribution of spouse’s income to equivalized income. In our decomposition, (b3) quantifies the expected change in equivalized income due to the increase in the educational premium assuming that assortative mating remains constant. The control variables in the regression models for spouse’s income include spouse’s education, spouse’s race, spouse’s age and age squared, residential region, and spouse’s migration status. Table 5 presents the decomposition results.

For women, (a) the contribution of their own personal income in accounting for the change in equivalized income (i.e., the column, Δ) is positive for all levels of education except LTHS. The increased contribution of women’s own income is mainly due to the rise in their income as measured by (a2) rather than distributional changes (i.e., (a1)). In contrast to their own contribution, spouse’s contribution is negative for women whose education is LTHS, HSG, SC, AA, or PhD. Even when spouse’s contribution is positive, the extent of spouses’ contribution is lower than women’s own contribution.

The reasons why husband’s contribution is negative or less positive are two-fold. One, women aged 35–44 in 2009–2011 tend to have married less-educated men compared with women in the same age range in 1990. Second, income for husbands of less-educated women has declined between 1990 and 2009–2011. Husband’s net income change, (b3), is positive for BA or higher but negative for LTHS, HSG, SC, and AA.

For men, the equivalized income increased over this period for all levels of education. For the less-educated, the equivalized income grew despite the decline in men’s own contribution. The increase in their wife’s contribution was large enough to offset husband’s negative contribution. Highly educated men’s equivalized income grew because both their own personal income and their wife’s income have increased.

When wife’s contribution is decomposed into (b1), the change in educational distribution, and (b3), the change in earnings rate for women, the extent to which (b1) and (b3) contribute is roughly the same. For example, the equivalized income for BA-educated married men increased by $11,768. Approximately one-half of this increase ($5,665) is due to the rise in wife’s contribution. Of the total contribution from wives, 42 % ($2,355) is accounted for by the compositional change in wives’ education, and 57 % ($3,216) is explained by the rise in wives’ earnings rate.

In sum, standard of living improved more for married men than for married women between 1990 and 2009–2011, which partly arises because men increasingly married more-educated women. The other source of this result is that even after the distribution of wives’ educational attainment is held constant, wives’ earnings increased over this period.

### Sensitivity Analysis

Because our findings refer to those aged 35–44, the issue naturally arises as to whether our basic conclusions are more broadly applicable to the general population. To address this concern, we assess the sensitivity of our findings using various age and racial groups. First, we extend the age range to 25–54 and measure the relative (dis)advantage of being female in equivalized income. The model specification in Fig. 3 is identical to that in Table 2. The changing pattern for ages 25–54 is the same as that for ages 35–44. Next, we remove the family type restrictions by including multigenerational families; adult-adult two-generation households; and those who live with partners, friends, and visitors. We again find the same trend. The cost of being female became more negative over time.

Second, we check whether the reduction in the relative standard of living for women is evident across different age groups. As shown in panel a of Fig. 3, all three groups exhibit a deterioration in the relative (dis)advantage of being female over time. Although the magnitude of change and the extent of female (dis)advantages in a given year differ across groups, the extent of decline over three periods for those aged 25–34 is almost identical to that for respondents aged 35–44.

Finally, we examine whether the changes in the net female (dis)advantage vary by racial/ethnic groups. As shown in panel b of Fig. 3, all five racial/ethnic groups experience the same general trend. These analyses corroborate that our major findings are robust and that the changes reported in this article for our primary target population are also representative of overall changes in the broader American population.

## Conclusions

This study investigates gender-specific changes in the total financial return to education between 1990, 2000, and 2009–2011. In contrast to prior research that mostly examined the return to education in terms of personal labor market earnings, we study the return to education in regard to the family standard of living as measured by equivalized income. The latter accounts for educational returns in both the labor market and the marriage market.

Our analysis yields several notable findings. Even though their personal earnings are lower than men’s, married women enjoyed a higher standard of living than their male counterparts regardless of education in 1990. Consistent with the “rise of women” argument (DiPrete and Buchmann 2006), women’s personal earnings grew faster than men’s between 1990 and 2009–2011. This progress was mostly driven by married women. Paradoxically, however, married women’s progress led to a faster improvement in the family standard of living for men than for women themselves. As a result, the previously observed married women’s advantage in the return to education in terms of the family standard of living almost completely disappeared by 2009–2011.

We explore the mechanisms behind these changes, finding that the shift in assortative mating pattern is associated with this paradoxical result. The likelihood of educational marrying up increased substantially for men over time, and women’s likelihood decreased. Consequently, men’s financial gain through marriage increased between 1990 and 2009–2011, and women’s gain declined. No educational groups for married men experienced a drop in their standard of living. Married men at all levels of education benefited from a rising standard of living despite the decline in personal earnings among men without a bachelor’s degree. By contrast, women without a degree in higher education endured a dwindling standard of living over this period regardless of whether they are married. As far as the standard of living is concerned, our results show that married men have been the main beneficiary of women’s progress. However, we caution that this association should not be interpreted as being causal.

Overall, these results highlight the importance of taking into account both the labor market and the marriage market in studying the gender-specific return to education. Marriage markets used to be an important mechanism that mediated women’s return to education (Goldin et al. 2006). Although women benefit more than ever from their investment in education in labor markets, the importance of the return in marriage markets has not disappeared. Instead, our results suggest that the return in marriage markets is becoming increasingly important not only for women’s economic well-being but also for men’s. The rise of women in educational attainment (DiPrete and Buchmann 2013) and the changes in assortative mating (Schwartz and Mare 2005) are altering the economic landscape for family demography in the United States.

The gap in equivalized income between wives and husbands has converged over time. This convergence is consistent with the argument that marriage is becoming increasingly egalitarian (Sullivan 2006). In the United States, marriage has become less institutionalized and more personalized (Cherlin 2005; Sweeney 2002). The norm of marriage has shifted from being an economically functional institution to personalized companionship (Cherlin 2004) so that assortative mating has increased. The increase in women’s financial contributions to household finances will tilt the power relations at home toward more equal settings (Agarwal 1997; England and Farkas 1986). We argue that the gender convergence in the family standard of living is associated with this shift in the norm of marriage.

Our results also have an important implication regarding the relation between assortative mating and income inequality. Paradoxically, the growing inequality in men’s earnings across educational levels hurts married women with less than a bachelor’s degree more than less-educated married men themselves in regard to the standard of living. The decline in personal income for less-educated men is offset by the rise in income from their wives. For less-educated women, however, the contribution of their husbands has been substantially reduced so that their standard of living has diminished even though their personal earnings have grown. The rise in educational hypogamy aggravates the negative effect of rising inequality (Fernandez et al. 2005).

This study provides new insights on the links between changing patterns of assortative mating and economic well-being. At a minimum, our results demonstrate the importance of the demographic links among the gender gap in education, assortative mating patterns, and labor market outcomes in regard to rising inequality in economic well-being (Schwartz 2013). Future research needs to monitor how family demography still shapes and directly underlies inequality even as family relations continue to evolve.

## Acknowledgments

We thank the Editor and the anonymous reviewers of Demography for helpful comments. Thanks also to Kimberly Goyette, Young-mi Kim, and Yool Choi for their comments. Earlier versions of this article were presented at the 2015 RC28 summer meeting and at Yonsei University in Seoul, Korea. ChangHwan Kim received financial support for this study from the University of Kansas (GRF #2301065).

## Notes

1

However, Hubbard (2011) argued that no gender difference exists in the college wage premium after correcting for bias associated with top-coding.

2

Because we use the three-year combined ACS, we considered the sensitivity of our results by changing the sample restriction to those who were born in 1966–1975, finding that the results are almost identical. Using the ACS 2010 one-year sample instead of a three-year combined sample does not alter our conclusion, either.

3

Even within nuclear families, resource allocation can be skewed depending on power relations between couples and related transaction cost considerations (Agarwal 1997; Bergstrom 1996). How well this assumption holds is very difficult to test because the decision to participate in the labor force is not exogenous with respect to the allocation of resources within the family (Lundberg et al. 1997).

4

The proportion living with partners, friends, or visitors is only 4.2 % among persons aged 35–44 in 2009–2011.

5

Why single women’s incomes grew more slowly than equally educated single men is another question beyond the scope of this study and should be investigated in future research.

6

The analyses of the change between 1990 and 2000 and the change between 2000 and 2009–2011 yield the same conclusion that we present here.

7

Mathematically, the log of a sum does not equal the sum of the logs.

## References

Agarwal, B. (
1997
).
“Bargaining” and gender relations: Within and beyond household
.
Feminist Economics
,
3
(
1
),
1
51
. 10.1080/135457097338799
Aughinbaugh, A., Robles, O., & Sun, H. (
2013
).
Marriage and divorce: Patterns by gender, race, and educational attainment
.
Monthly Labor Review, October
,
1
16
.
Bailey, M. J. (
2006
).
More power to the pill: The impact of contraceptive freedom on women’s life cycle labor supply
.
Quarterly Journal of Economics
,
121
,
289
320
.
Becker, G. S. (
1991
).
A treatise on the family
.
Cambridge, MA
:
Harvard University Press
.
Bergstrom, T. C. (
1996
).
Economics in a family way
.
Journal of Economic Literature
,
34
,
1903
1934
.
Blau, F. D., & Kahn, L. M. (
2006
).
The U.S. gender pay gap in the 1990s: Slowing convergence
.
Industrial and Labor Relations Review
,
60
,
45
66
. 10.1177/001979390606000103
Brand, J. E., & Davis, D. (
2011
).
The impact of college education on fertility: Evidence for heterogeneous effects
.
Demography
,
48
,
863
887
. 10.1007/s13524-011-0034-3
Breen, R., & Salazar, L. (
2011
).
Educational assortative mating and earnings inequality in the United States
.
American Journal of Sociology
,
117
,
808
843
. 10.1086/661778
Buchmann, C., & DiPrete, T. A. (
2006
).
The growing female advantage in college completion: The role of family background and academic achievement
.
American Sociological Review
,
71
,
515
541
. 10.1177/000312240607100401
Buss, D. M., Shackelford, T. K., Kirkpatrick, L. A., & Larsen, R. J. (
2001
).
A half century of mate preferences: The cultural evolution of values
.
Journal of Marriage and Family
,
63
,
491
503
. 10.1111/j.1741-3737.2001.00491.x
Card, D., & DiNardo, J. E. (
2002
).
Skill–biased technological change and rising wage inequality: Some problems and puzzles
.
Journal of Labor Economics
,
20
,
733
783
. 10.1086/342055
Charles, K. K., & Luoh, M. C. (
2003
).
Gender differences in completed schooling
.
Review of Economics and Statistics
,
85
,
559
577
.
Cherlin, A. J. (
2004
).
The deinstitutionalization of American marriage
.
Journal of Marriage and Family
,
66
,
848
861
. 10.1111/j.0022-2445.2004.00058.x
Cherlin, A. J. (
2005
).
American marriage in the early twenty-first century
.
Future of Children
,
15
(
2
),
33
55
. 10.1353/foc.2005.0015
DiPrete, T. A., & Buchmann, C. (
2006
).
Gender-specific trends in the value of education and the emerging gender gap in college completion
.
Demography
,
43
,
1
24
. 10.1353/dem.2006.0003
DiPrete, T. A., & Buchmann, C. (
2013
).
The rise of women: The growing gender gap in eduation and what it means for American schools
.
New York, NY
:
Russell Sage Foundation
.
Dougherty, C. (
2005
).
Why are the returns to schooling higher for women than for men?
.
Journal of Human Resources
,
XL
,
969
988
. 10.3368/jhr.XL.4.969
England, P. (
2010
).
The gender revolution: Uneven and stalled
.
Gender and Society
,
24
,
149
166
. 10.1177/0891243210361475
England, P., & Farkas, G. (
1986
).
Households, employment, and gender: A social, economic, and demographic view
.
New York, NY
:
Aldine Pub. Co.
.
Fernandez, R., Guner, N., & Knowles, J. (
2005
).
Love and money: A theoretical and empirical analysis of household sorting and inequality
.
Quarterly Journal of Economics
,
120
,
273
344
.
Goldin, C. (
1997
).
Career and family: College women look to the past
. In Blau, F. D., & Ehrenberg, R. G. (Eds.),
Gender and family issues in the workplace
(pp.
20
58
).
New York, NY
:
Russell Sage Foundation Press
.
Goldin, C. (
2006
).
The quiet revolution that transformed women’s employment, education, and family
.
American Economic Review
,
96
,
1
21
. 10.1257/000282806777212350
Goldin, C., Katz, L. F., & Kuziemko, I. (
2006
).
The homecoming of American college women: The reversal of the college gender gap
.
Journal of Economic Perspectives
,
20
(
4
),
133
156
. 10.1257/jep.20.4.133
Grogger, J., & Eide, E. (
1995
).
Changes in college skills and the rise in the college wage premium
.
Journal of Human Resources
,
30
,
280
310
. 10.2307/146120
Hubbard, W. H. (
2011
).
The phantom gender difference in the college wage premium
.
Journal of Human Resources
,
46
,
568
586
. 10.1353/jhr.2011.0030
Kim, C., & Sakamoto, A. (
2008
).
The rise of intra-occupational wage inequality in the United States, 1983 to 2002
.
American Sociological Review
,
73
,
129
157
. 10.1177/000312240807300107
Lehrer, E., & Nerlove, M. (
1986
).
Female labor force behavior and fertility in the United States
.
Annual Review of Sociology
,
12
,
181
204
. 10.1146/annurev.so.12.080186.001145
Leicht, K. T. (
2008
).
Broken down by race and gender? Sociological explanations of new sources of earnings inequality
.
Annual Review of Sociology
,
34
,
237
255
. 10.1146/annurev.soc.34.040507.134627
Long, M. C. (
2010
).
Changes in the returns to education and college quality
.
Economics of Education Review
,
29
,
338
347
. 10.1016/j.econedurev.2009.10.005
Lundberg, S. J., Pollak, R. A., & Wales, T. J. (
1997
).
Do husbands and wives pool their resources? Evidence from the United Kingdom Child Benefit
.
Journal of Human Resources
,
32
,
463
480
. 10.2307/146179
Mandel, H., & Semyonov, M. (
2014
).
Gender pay gap and employment sector: Sources of earnings disparities in the United States, 1970–2010
.
Demography
,
51
,
1597
1618
. 10.1007/s13524-014-0320-y
Mare, R. D. (
1991
).
Five decades of educational assortative mating
.
American Sociological Review
,
56
,
15
32
. 10.2307/2095670
McCall, L. (
2000
).
Explaining levels of within-group wage inequality in U.S. labor markets
.
Demography
,
37
,
415
430
. 10.1353/dem.2000.0008
McKinnish, T. (
2008
).
Spousal mobility and earnings
.
Demography
,
45
,
829
849
. 10.1353/dem.0.0028
Morris, M., & Western, B. (
1999
).
Inequality in earnings at the close of the twentieth century
.
Annual Review of Sociology
,
25
,
623
657
. 10.1146/annurev.soc.25.1.623
Murphy, K. M., & Welch, F. (
1992
).
The structure of wages
.
Quarterly Journal of Economics
,
107
,
285
326
. 10.2307/2118330
Musick, K., England, P., Edgington, S., & Kangas, N. (
2009
).
Education differences in intended and unintended fertility
.
Social Forces
,
88
,
543
572
. 10.1353/sof.0.0278
Oppenheimer, V. K. (
1994
).
Women’s rising employment and the future of the family in industrial societies
.
Population and Development Review
,
20
,
293
342
. 10.2307/2137521
Oppenheimer, V. K. (
1997
).
Women’s employment and the gain to marriage: The specialization and trading model
.
Annual Review of Sociology
,
23
,
431
453
. 10.1146/annurev.soc.23.1.431
Parker, K., Horowitz, J. M., & Rohal, M. (
2015
).
Parenting in America: Outlook, worries, aspirations are strongly linked to financial situation
Ruggles, S., Genadek, K., Goeken, R., Grover, J., & Sobek, M. (
2015
).
Integrated Public Use Microdata Series: Version 6.0
[Data set].
Minneapolis
:
University of Minnesota
.
Schwartz, C. R. (
2013
).
Trends and variation in assortative mating: Causes and consequences
.
Annual Review of Sociology
,
39
,
451
470
. 10.1146/annurev-soc-071312-145544
Schwartz, C. R., & Han, H. (
2014
).
The reversal of the gender gap in education and trends in marital dissolution
.
American Sociological Review
,
79
,
605
629
. 10.1177/0003122414539682
Schwartz, C. R., & Mare, R. D. (
2005
).
Trends in educational assortative marriage from 1940 to 2003
.
Demography
,
42
,
621
646
. 10.1353/dem.2005.0036
Snyder, T. D., Dillow, S. A., & Hoffman, C. M. (
2008
).
Digest of education statistics, 2007
(NCES Report No. 2008-022).
Washington, DC
:
National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education
.
Stiglitz, J., Sen, A., & Fitoussi, J-P (
2009
).
The measurement of economic performance and social progress revisited: Reflections and overview
(OFCE Report No. 2009-33).
Paris, France
:
Centre de Recherche en Économie de Sciences Po
.
Sullivan, O. (
2006
).
Changing gender relations, changing families: Tracing the pace of change over time
.
Lanham, MD
:
Rowman & Littlefield Publishers
.
Sweeney, M. M. (
2002
).
Two decades of family change: The shifting economic foundations of marriage
.
American Sociological Review
,
67
,
132
147
. 10.2307/3088937
Sweeney, M. M., & Cancian, M. (
2004
).
The changing importance of white women’s economic prospects for assortative mating
.
Journal of Marriage and Family
,
66
,
1015
1028
. 10.1111/j.0022-2445.2004.00073.x
Tomaskovic-Devey, D., Zimmer, C., Stainback, K., Robinson, C., Taylor, T., & McTague, T. (
2006
).
Documenting desegregation: Segregation in American workplaces by race, ethnicity, and sex 1966–2000
.
American Sociological Review
,
71
,
565
588
. 10.1177/000312240607100403
Wang, W. (
2014
).
Record share of wives are more educated than their husbands
.
Pew Research Center Fact Tank