Narrowing the Gender Wage Gap Among Computer Science Professionals: What Differentiates the Earnings of Asian and White Workers?

Human Capital Explanations for the Gender Wage Gap

The differential in wages between men and women has been extensively covered in the economics literature (Blau and Kahn 2017). This research focuses on disparities in human capital accumulation, such as the choice of college major and work histories (Becker 1985), with gendered variation in fields of study, educational attainment, hours worked, experience, and job sector used to explain wage differences. Some of these factors—such as educational attainment—have declined in explanatory power as the average education levels among women have surpassed those of men (Blau and Kahn 2017), and the proportion of women with degrees in STEM fields has grown (Michelmore and Sassler 2016). Studies focused on the college-educated population note, for example, that the college wage premium is similar for men and women (England et al. 2020).

Yet other markers of human capital investment—hours worked, and compensation for working long hours, along with selection into job sectors with high returns—continue to differentiate the earnings of men and women, resulting in the persistent gender wage gap (Goldin 2014). Men are significantly more likely than women to engage in long work hours (more than fifty hours a week), which perpetuates the gender wage gap, especially among professionals (Cha and Weeden 2014; Weeden et al. 2016). Furthermore, occupational segregation remains high, and the importance of industry and occupation now accounts for a larger share of the gender wage gap than in the latter years of the twentieth century. In 1980, gender gaps in industry and occupation accounted for one-fifth of pay disparities between women and men, but by 2010 they accounted for 51 percent of the gender gap— despite the decline over time in that gap (Blau and Kahn 2017). Even when women work in occupations stereotypically considered more “masculine,” such as STEM jobs, the greatest economic rewards continue to flow to workers engaged in jobs with the lowest shares of women, such as programming-intensive occupations (Cheng et al. 2019; Michelmore and Sassler 2016). Among those working in CS, for example, occupation accounts for a considerable share of the (unexplained) gender wage gap, largely because women are less likely than men to work in the highest-paid jobs in CS, such as software developers (Sassler and Meyerhofer 2023).

Several recent explorations of gender disparities in occupational attainment and pay have highlighted the need to pay closer attention to differential returns to men’s and women’s attributes— suggesting that despite legislation that prohibits discrimination, bias against women persists (Fernandez and Campero 2017; Galperin 2021; Moss-Racusin et al. 2012). Occupational sorting plays an important role in perpetuating gendered wage disparities among highly educated workers (Quadlin et al. 2023; Smith-Doerr et al. 2019; Zheng and Weeden 2023), and emerges early in the professional life-course— well before family formation begins. Furthermore, gender wage disparities persist within occupations, even among men and women with observably identical attributes (Sassler and Meyerhofer 2023). Certain occupations also impose heavy penalties on employees who may desire more flexible employment for caregiving responsibilities (Goldin 2014), with those who work long hours benefiting the most in terms of earnings (Weeden et al. 2016).

Family Explanations for the Gender Wage Gap

Sociological explanations for the persistent gender wage gap have focused on gendered norms that assign primacy to women’s obligations as wives and mothers—factors that are difficult to integrate into the human capital theoretical framework, due to concerns over endogeneity.¹ The literature often refers to earnings advantages and disadvantages—premiums and penalties—that are associated with those who have the intersecting identities of employees and parents. Whereas working mothers often experience penalties such as wage reductions, diminished opportunities for career advancement, and lower perceived professional competence relative to childless women and both childless men and fathers, fathers are often viewed as more committed and responsible workers and rewarded accordingly (Correll et al. 2007). Recent research has pointed out that family factors, namely marital and parental status, play a sizable role in differentiating men and women’s earnings, especially when focusing on the highly educated (Beutel and Schleifer 2022; Cha et al. 2023; Sassler and Meyerhofer 2023). Such research has moved beyond asking whether there is a motherhood penalty, a fatherhood premium, or a marriage bonus, to instead asking whether the gender wage gap is really a family wage gap (Cha et al. 2023; Sassler and Meyerhofer 2023).

The evidence of gender disparities among men and women with at least a college degree has shifted in recent years, as the returns to education rise, marriage is increasingly delayed, and parental roles change due to delayed fertility, childlessness, and the spread of blended families (Sassler and Lichter 2020). Although married men earn more, on average, than unmarried men, the association between marriage and wages among women has changed across cohorts (Killewald and Gough 2013). Michelle J. Budig and Misun Lim (2016) found that both baby boomer and millennial men and women received marriage premiums, though the advantages were smaller for millennial men and larger for millennial women. Katherine Michelmore and Sharon Sassler (2016) found that married women with college degrees who worked in STEM occupations also earned more than their unmarried counterparts (see also Sassler and Meyerhofer 2023). Selection into marriage in part explains this finding (Budig and Lim 2016); along with the rise in educational homogamy, and the increase in dual-career couples, returns to remaining in the paid labor force have grown, resulting in higher levels of female employment among college-educated women (Percheski 2008). We therefore expect to find a marriage premium among CS professionals.

But marital status alone does not capture the importance of couple-level characteristics in shaping the gender wage gap. Jennifer Glass and colleagues (2013), looking at baby boomer women with degrees in a STEM field, found that those whose spouse worked in STEM were significantly more likely to remain in STEM jobs than those whose spouse worked in another field, but that as a spouse’s work hours increased the likelihood of women exiting the labor force rose. Youngjoo Cha and Kim Weeden (2014) suggest that men’s greater likelihood of engaging in overwork (defined as working fifty or more hours per week), in conjunction with the rising hourly wage returns to overwork, contributes to the persistent gender wage gap, particularly among those for whom at least one partner works in a professional and managerial occupation.² In other work, Cha and colleagues (2023) find that the gender wage gap was substantially larger among married White respondents compared with their married Black or Hispanic counterparts. Yet few studies explore the importance of couple-level attributes in determining the size of the gender wage gap, further reifying gendered assumptions about marriage and parenthood.

In fact, the bulk of research on the family wage gap has focused on rewards or penalties to unpaid care work, such as looking after children or aging parents, rather than marriage (Lim and Jeong 2026). While much of this research focuses on the general population, studies concentrated on those with a college degree or more or those engaged in demanding professions challenge the accepted wisdom that mothers are penalized relative to childless women in terms of their earnings. Among college-educated men, fathers earn more than their childless counterparts (Buchmann and McDaniel 2016; Glauber 2008). The association between motherhood and earnings is less clear for highly educated mothers. A growing body of evidence shows that among professionals, the size of the negative wage differential for motherhood has decreased when childless women are the referent (Beutel and Schleifer 2022; Buchmann and McDaniel 2016). In fact, in traditionally male-dominated professions, such as STEM, medicine, and law, women with children earn more than their childless counterparts (Buchmann and McDaniel 2016; Michelmore and Sassler 2016; Sassler and Meyerhofer 2023). Women in female-dominated professions continue to experience a motherhood penalty (Buchmann and McDaniel 2016), and the gender wage gap widens at the top of the earnings distribution among professionals (Quadlin et al. 2023; Zheng and Weeden 2023). Findings regarding the motherhood wage penalty, then, depend on women’s educational attainment and job sector, as well as on marital status and race. Given our sample, we anticipate a parenthood premium for both women and men.

Further disaggregation of the impacts of motherhood focuses on whether children of a particular age are more detrimental to earnings for women, with some exploring the impact of preschool-aged children or various ages of children (Percheski 2008; Sassler and Meyerhofer 2023), and others examining the number of children (Buchmann and McDaniel 2016). Buchmann and McDaniel (2016) find that among professional women employed in medicine, STEM, and law, mothers earn more than childless women, regardless of number of children, though there is variation across professions; among women working in STEM professions, those with one child or three or more children earn similar wages to childless women (Buchmann and McDaniel 2016, note 13; see also Sassler and Meyerhofer 2023). One recent study of college graduates working in CS found that both men and women with parenting responsibilities earned more than childless CS professionals; this held regardless of the age of children, but the premium was greatest when children were the youngest (Sassler and Meyerhofer 2023). While men with only preschool-aged children earned seven cents more than did childless men, women with preschool-aged children only earned 2.1 cents more per hour than their childless counterparts (Sassler and Meyerhofer 2023). The wage premium was somewhat smaller when there were both young and school-aged children, at least for men, and was smallest for those with only school-aged children. What matters here is the referent: while mothers earned more than childless women, they continued to earn significantly less than fathers, thereby widening the gender wage gap—even though women were significantly less likely to be parents than were men. We therefore anticipate that both mothers and fathers will experience a parenthood premium, but it is not clear ex ante how this may vary by age and number of children and ethnicity.

Given changes in union formation patterns, little attention has been paid to the importance of maternal age at birth as a factor shaping wage penalties. One such study, using data from the NLSY79, found that for college graduates, motherhood penalties were larger when first births occurred at earlier ages, but declined with older ages at first birth; women who delayed fertility until their mid-30s received a motherhood premium (Doren 2019). Since baby boomer women came of age, however, the median age at first marriage, especially for college-educated women, has increased, and college-educated women mostly bear children within marital unions (Sassler and Lichter 2020). In fact, the majority of college-educated women’s first births were at ages older than thirty (Bui and Cain Miller 2018), suggesting additional factors associated with returns to motherhood that require study. Heterogeneity in family formation and caretaking behaviors may further exacerbate or minimize the gender wage gap, as older first-time mothers accrue additional job tenure and work experience prior to parenthood.

Racial and Ethnic Variation in Work Behaviors Contributing to the Gender Wage Gap

Asians are frequently overlooked in research, particularly in studies that explore the factors contributing to the gender wage gap (Lu 2024). Despite being overrepresented in professional occupations—particularly in STEM jobs—Asians accounted for only 6 percent of the US population in 2020 (Jones et al. 2021).³ While Asians are the fastest-growing racial group in the United States, given their overall population size, longitudinal surveys frequently do not contain enough Asian respondents to assess wage gaps across occupations or differentiate between ethnic groups. Furthermore, the work-family literature that explores gender variation in earnings generally overlooks racial and ethnic heterogeneity (Perry-Jenkins and Gerstel 2020). Research that has assessed racial variation in the gender wage gap tends to contrast Black and Hispanic workers with White workers (Budig et al. 2021; Cha et al. 2023). Such studies have found that the marriage premium is smaller for Black men than for White men, while the motherhood penalty is smaller for Black and Hispanic women (England et al. 2016; Glauber 2007), especially among the highly educated.

The few studies that have examined gender wage gaps among Asians have found that the gender wage gap is smaller among Asians than among Whites working in STEM professions (Greenman 2011; Greenman and Xie 2008; Michelmore and Sassler 2016). Utilizing data from the 2000 Census, Emily Greenman and Yu Xie (2008) showed that there were smaller gender earnings gaps among various Asian groups than among Whites, though they limited their analysis to native-born workers. Other analyses that included foreign-born Asians found that the gender earnings gap was larger among immigrants than among native-born workers (Goyette and Xie 1999; Xie and Shauman 2005); while not testing it directly, these articles suggested that this disparity arose because foreign-born wives were trailing spouses (Goyette and Xie 1999; Xie and Shauman 2005). Others have noted more of an earnings premium for parenthood among some groups of women; for example, Asian mothers (along with their White and Black counterparts) earned more than their childless counterparts (see Michelmore and Sassler 2016), but Hispanic women did not. Greenman and Xie (2008) also found less of an earnings penalty for marriage and parenthood among their sample of native-born minorities (Asians, Blacks, and Hispanics) relative to Whites, leading them to suggest the possibility of greater traditional role specialization among White couples. Taken as a whole, these results suggest the possibility that Asian families utilize somewhat different care work strategies than other race groups in the United States, with important ramifications for economic well-being.

Other, causal work has sought to unpack the mechanisms behind the smaller wage gaps and higher returns to family status among Asian workers. Such studies, relying on various data sources, have found that Asian women were less likely than White women to reduce their labor supply on becoming mothers. Using nationally representative data from SIPP, Yao Lu and colleagues (2017) found that Asian women who were employed before childbirth demonstrated greater labor market continuation than their White counterparts (see also Greenman and Xie 2008). Lu and colleagues (2017) also found that immigrant women who had been in the US for longer durations demonstrated greater labor force attachment. This also appears to be the case among Asian women employed in STEM fields. Using data from SESTAT, Emily Greenman (2011) found that Asian women were less likely than White women to reduce their labor force supply on becoming mothers. This enabled them to accumulate more work experience. Other family behaviors—the timing of marriage, for example, or the age at which women become mothers and men fathers—may also contribute to wage premiums or penalties (Doren 2019).

We advance the research into wage gaps and ethnic attainment by exploring how family attributes are associated with the gender wage gap among different Asian ethnic groups compared with Whites. If patterns found for other racial minorities (England et al. 2016; Glauber 2007) hold, the marriage premium may be smaller for Chinese and Indian women than it is for White women. But the evidence also suggests that parenthood penalties might be greater for White women than for Chinese and Indian women (Greenman and Xie 2008; Michelmore and Sassler 2016). These studies, then, do not provide a clear direction for how the impact of motherhood may differ by age of children.

Other Explanatory Factors

A key factor distinguishing Asian workers from their White counterparts—even among those working in a similar occupation—is the large proportion who are foreign-born, as well as their disproportionate representation in STEM occupations. Asians are overrepresented among those working in all STEM occupations in general, and in computer occupations in particular (Landivar 2013). Furthermore, many of those working in STEM jobs are foreign-born. As of 2021, over one in four (26 percent) foreign-born workers employed in the United States worked in STEM occupations (National Science Board 2024), and among Asians this proportion is even greater.⁴

Therefore, any reports about the size of the gender wage gap among Asians must account for the share who are foreign-born, and the factors differentiating earnings among them (such as naturalization and education). One recent exploration of the gender wage gap among CS workers, for example, found that Asian workers earned significantly more than their White counterparts, though the Asian earnings advantage was entirely driven by foreign-born workers (Sassler and Meyerhofer 2023). But various factors differentiate the earnings trajectories of the foreign-born in ways that may shape the gender wage gap. Where one’s education was obtained, the prestige of schools attended, English-speaking ability, and work experience in the US all have been shown to influence Asians’ earnings (Lee et al. 2024; Zeng and Xie 2004). Looking at natives and immigrants in the early 1990s, Zhen Zeng and Yu Xie (2004) found that foreign-educated immigrants’ earnings were substantially lower than those of immigrants and natives who obtained their schooling in the United States; they concluded that foreign education was less valuable in the US job market. Jennifer Lee and colleagues (2024) report that Asians in the United States engage in strategic adaptation to a discriminatory marketplace by pursuing credentials at prestigious universities; such strategies, however, often still fail to eliminate bias in job returns.

Others argue for the need to differentiate culturally between East Asians, (for example, ethnic Chinese) and South Asians (for example, ethnic Indians). Jackson G. Lu (2024) asserts that whereas Chinese professionals in the business world tend to experience a “bamboo ceiling,” Indian professionals do not. Examining those who attain MBAs or work in consulting, Lu (2024) documents differential treatment in starting salaries (Lu 2023) and leadership attainment (Lu 2022), with Chinese professionals experiencing more barriers than their Indian counterparts. Explanations for these differences focus on social-cognitive disparities, such as assertiveness and perceived creativity (Lu et al. 2022). These studies examine professionals working in law, business, and consulting; it is unclear whether such factors shape work in CS to the same extent. Furthermore, these arguments are focused on the successes of Indian men and give short shrift to the fact that Indian women do not fare as well (Lu 2024). We therefore differentiate between Indians and Chinese CS professionals, contrasting them with their White counterparts and with each other, to better understand the cultural as well as structural dynamics shaping the gender wage gap.

Variables

Our key dependent variable is the natural log of pretax wage and salary income for individuals working in CS. Our wage measure includes all pretax income respondents receive as an employee, which includes bonuses that can make up a large portion of all income in CS occupations (Dice 2025). All wages are converted into 2018 dollars using the Consumer Price Index. The ACS top codes the income data;⁷ the maximum salary in our sample is $780,000.

Our analysis examines how both gender and race influence earnings. Gender in our sample is a binary variable (male or female). We focus on two race groups: non-Hispanic White and non-Hispanic Asian. We further disaggregate Asians into the two largest subgroups working in CS, Indians and Chinese, to disentangle ethnic-specific strategies for balancing work and family. Sample size precludes us from examining additional Asian ethnic groups.

Our measures of family characteristics include respondents’ marital and parental status, and whether they reside in a three-generation family. Those who are currently married serve as the reference group; dummy variables capture those who are previously married (divorced, separated, or widowed) and never married. Parental status is measured with two ACS questions on the number of children in the household (number of children under age eighteen in the household, and number of children under age five in the household) to capture differential demands children impose on parents’ ability to work at different stages of child development.⁸ We include a continuous variable for the number of children a respondent has, before further differentiating the age groups of respondents’ children into four mutually exclusive groups: those with no children, those with only children under age five, those with only children five and older, and those with children who are both under five and over five. We also include an indicator of whether the respondent lives in a multigenerational household, where at least three generations are present, which is more prevalent among Asian families (Lim and Jeong 2026).

In additional sensitivity analyses, we limit the sample to those who are married to explore the impact of occupational homogamy on the gender wage gap. This allows us to more thoroughly assess a previously understudied aspect of the family wage gap. The ACS enables us to link married respondents to their spouses if they reside in the same household. We then determine whether both spouses were employed in the same field and then the same occupation, first exploring the broad field of STEM, before narrowing the analysis to determine if both spouses worked in a CS occupation.

Our measures of human capital accumulation include variables relating to experience, degree field and educational attainment, and occupation. While the ACS does not have a variable that captures work experience or time in the labor force, we proxy this (imperfectly) with age and its square, to account for its nonlinear impact over the work life course. We include the college major of individuals, comparing those with CS majors to other majors prevalent among those working in CS occupations—engineering, other STEM majors, business, and other non-STEM majors. We also include an indicator for educational attainment, comparing having a bachelor’s degree with having a professional or master’s degree, and with a doctorate.⁹ Our final indicator of human capital differentiates between twelve CS occupations. In some model specifications we also include an indicator for overwork, noting whether an individual reports that their usual hours of work exceed fifty hours a week.

Additionally, we include controls for citizenship status, given that sizable proportions of our Asian respondents were born abroad, and research suggests that immigrants who become American citizens fare better in the workforce than those who do not (National Academies of Sciences, Engineering, and Medicine 2017).¹⁰ We compare native US citizens with those who are non-US born and who have or have not naturalized to be US citizens.¹¹ In some analyses, we include fixed effects for all fifty states, as well as year fixed effects for all years.

Descriptive Results

Despite numerous efforts designed to increase the representation of women in CS occupations, only 24 percent of our sample of CS workers are female. Table 1 shows basic summary statistics for our full analysis sample and split by gender. Female CS professionals are more diverse than male CS professionals; 67 percent of all female CS professionals are White, compared to 72 percent of all male CS professionals. Chinese women are overrepresented among female CS professionals compared to Chinese men, making up 9 percent of all female CS professionals, compared to Chinese men who are just 6 percent of all male CS professionals. Finally, women employed in CS jobs earn significantly less than their male counterparts, regardless of race or ethnic group. The gender wage gap for the entire sample is $19,500.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1.

Descriptive Statistics for Analysis Variables Whole Sample and by Gender

Because the overall means for our sample mask considerable variation across race and ethnic groups, we review descriptive statistics for our variables of interest for White, Indian, and Chinese groups (see table 2). While the gender wage gap is evident across race groups, Asians earn more than their White counterparts (see the top column in table 2). But Asians are a heterogeneous group; Chinese men and women earn substantially more than Indian men and women, who earn more than White men and women.

View this table:

• View inline
• View popup
• Download powerpoint

Table 2.

Descriptive Statistics for Analysis Variables by Race and Gender Subcategory

If, as the literature suggests, there is increasingly a family premium among professional workers—both in terms of financial returns to marriage and parenthood—then variation in the family attributes of those working in CS suggests one possible contribution to the earnings disparities observed earlier. Even though most men and women who work in CS occupations are currently married, White CS professionals are significantly less likely to be married than their Asian counterparts, and the gap is the largest between Whites and Indians. Only 58 percent of White women and 66 percent of White men who work in CS occupations are married, compared with 86 percent of Indian women and 83 percent of Indian men. White CS workers are also significantly more likely to be previously married, a status that may not have the same returns as marriage. Furthermore, over half of White women and men working in CS (60 percent of women, and 53 percent of men) have no minor children, compared with only about a third of Indian men and women, and half of Chinese men and women.

The research literature also suggests the importance of other factors shaping career investments, including partner attributes (such as their occupation or hours worked), as well as marriage timing and the age of first parenthood (see table 3). Focusing only on married couples, we explored the proportion of CS professionals whose spouse works in any STEM field, before refining this to assess occupationally homogamous couples—where both spouses work in CS jobs. This subanalysis reveals that across all groups, women who worked in CS are significantly more likely to be married to a spouse who worked either in a STEM field or more specifically in CS than their male counterparts. Marriage to another STEM professional or a spouse whose job was in CS was least likely for White men in our sample of CS professionals (9 percent and 6 percent, respectively). Indian and Chinese men were considerably more likely than White men to have a spouse who also worked in a STEM field. But occupational homogamy is far more prevalent among women, and Indian women stand out for their incredibly high level of occupational marital homogamy. Over two-thirds of Indian women’s spouses are also employed in a STEM field, with a surprising 59 percent of Indian women who are CS professionals married to a spouse who also works in CS. While still considerable, the proportions are much lower for Chinese women working in CS (40 percent of whose spouses also worked in a CS job), and only 18 percent of White women working in CS have spouses also working in CS.

View this table:

• View inline
• View popup
• Download powerpoint

Table 3.

Descriptive Statistics of Spouses for Married Members in Analysis Sample

There are also other ways that the family behaviors of our Asian groups differ from Whites that may shape earnings and contribute to the gender wage gap. Additional analyses reveal that Indian women marry earlier, become mothers earlier, and have more children on average than either their White or Chinese counterparts. Figure 1 shows these trends across gender and racial groups in our sample. The age gap at first marriage is largest among Indians, and smallest among Whites. Indian women also have their first child earlier, on average, than White and Chinese women, as shown in figure 1, panel B. Indian women are just over the age of twenty-nine when their first child is born. While this is older than the population in general, it is younger than other groups in our specific sample of highly educated women (Bui and Cain Miller 2018). The average age at first childbirth for White women is thirty-one, while for Chinese women the average age is 32.5. The average age at first parenthood for men is later than for their female counterparts, at ages in the early thirties. There is important heterogeneity in the family behaviors of Chinese and Indian women that is overlooked when Asian women are analyzed as a monolith.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 1.

Family Differences by Race and Gender

Source: Authors’ calculations using American Community Survey one-year estimates from 2009–2022.

We also observe important differences in measures of human capital, by gender as well as race and ethnicity (see table A.1). White CS workers are significantly older than their Asian counterparts, suggesting that Asian CS workers are more recent graduates, who may have had less time to experience wage compression. Furthermore, White CS professionals are far less likely than their Indian and Chinese counterparts to have obtained their degrees in CS or engineering, or to have post-graduate degrees. Finally, the descriptive results reveal stark differences in occupational concentration across ethnic groups, depicted in Figure 2. Chinese and Indian individuals are much more likely to be employed in software development—the second highest paid occupation, on average, in our sample—than their White counterparts; over half of all Indian and Chinese individuals work as software developers. Indian women are significantly more likely to work in this high-paying occupation than Chinese women; however, female Indian software developers make about $20,000 less per year than their Chinese counterparts, indicating factors beyond occupational segregation that drive wage gaps across and between ethnic groups. Both Indian and Chinese women are significantly more likely than White women to work in high-paying occupations.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 2.

Occupation Median Salary by Occupation Share Across Race and Gender Groups

Source: Authors’ calculations using American Community Survey one-year estimates from 2009–2022.

Note: Size of marker is determined by the raw number of individuals within each occupation.

Ethnic Variation in the Contribution of Family Factors

To isolate the effects of family factors—assessing the extent to which the gender wage gap is really a family wage gap—we explore the impact of our family variables from our second set of regressions (table 4, column 2), presenting these as a coefficient plot in figure 3. Shown here are the coefficient estimates and 95 percent confidence intervals for regressions run individually for each race and gender group. The graph shows the impact of marriage and parenthood on earnings, across women and men from different ethnic groups with similar levels of human capital. Specifically, within this set of regressions, we compare unmarried women and men to their married and divorced counterparts and childless women and men to those with children in varying age brackets within each racial group. To capture the full effect of parenthood, we also run a specification including only a parent indicator.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 3.

Impact of Family Status on Wages by Race and Gender

Source: Authors’ calculations using American Community Survey one-year estimates from 2009–2022.

Note: Models are run separately for each race by gender pair. Each marker displays the coefficient estimate for the coefficient labeled on the y-axis. We display 95 percent confidence intervals. Controls include: children indicators, including children 0–5, children 0–18, children 5–18 (omitted = Childless) [except in the parenthood indicator set of models where these are excluded]. Marital status indicators include married and previously married (omitted = never married). Citizenship indicators include naturalized citizen and foreign-born noncitizen (omitted = citizen from birth). Additional controls include age and age squared. Field-of-degree indicators include engineering, other STEM, business, other non-STEM (omitted = computer science). Highest degree indicators include master’s, professional degree, PhD (omitted = bachelor’s). Computer science occupations include CIS manager, computer scientist, computer systems analyst, information security analyst, computer programmer, web developer, support specialist, database administrator, network administrator, network architect (omitted = software development).

Figure 3 reveals considerably greater divergence in returns to marriage and parenthood among women than among men. Men of all groups examined here, White, Indian, and Chinese, experience a marriage premium, though this is greater for those who are currently married than the previously married. Married White and Chinese men experience an earnings boost of around 10 percent compared to their never married counterparts, while married Indian men earn 8.4 percent more; these differences are all statistically different from zero. White men experience a significantly larger marriage premium than Indian men.

In contrast to men, and to expectations drawn from the literature on professional working women, only White women experience a marriage premium that is significantly different from zero, earning 5 percent more than their never married counterparts. While married Chinese women do not experience an earnings premium, they also do not experience a marriage penalty; though imprecisely estimated, the point estimate is very close to zero. But Indian women experience large marriage penalties, which are significantly different from the returns to marriage experienced by White and Chinese women. Being married decreases Indian women’s earnings by 14 percent relative to their never married counterparts; they also experience a marriage penalty relative to married Chinese and White women. A similar pattern with regard to returns to parenthood can also be observed. Though returns to fatherhood vary somewhat across different stages of parenthood, most fathers receive a small premium for being parents, relative to their childless counterparts. There is some slight variation across ethnic groups, though these differences are not significant. White men who are fathers of preschoolers, for example, experience a small 4.8 percent earnings premium over their childless counterparts. But the earnings of Indian and Chinese fathers of preschool-aged children do not differ significantly from their childless counterparts. On the other hand, Indian fathers experience a slight earnings premium (5.4 percent) over their childless counterparts when they have school-aged children, though this difference is not significantly different from the bonus White fathers of older children experience. Chinese fathers experience no earnings premium, regardless of the age of their child or children, over childless Chinese men.

Among women, there is considerably more heterogeneity in the effect of parenthood on earnings, both across ages of children and ethnic groups. Indian women have the largest estimated parenthood penalty, of around 13 percent, statistically larger than that of White women, who have a very small positive effect of parenthood that is not different from zero. Chinese women have a parenthood penalty roughly in the middle of the two groups, around 5 percent. Whereas research on professional women has found that those with preschool-aged children experienced a wage premium relative to childless women, disaggregating motherhood penalties by ethnicity and age of children reveals a more nuanced picture. The impact of having children of different age groups on wages also shows that the effect sizes are generally larger for Indian and Chinese women than for White women, but they are more noisily estimated due to smaller sample sizes. Consistent with the results on marriage, there are larger penalties for Indian mothers compared to Indian childless women. But having older children—whether all school-aged, or in combination with preschool-aged children—has the largest estimated earnings penalty for Indian women, though we cannot say that this is significantly different from the penalty for having younger children. Indian women with both preschool- and school-aged children earn 13.5 percent less than childless women, while those with only older children (aged five to eighteen) earn 15 percent less. Chinese women also experience a mother penalty when they have only school-aged children, though the point estimates are smaller.

Figure 4 reveals how the effects of marriage and parenthood differ within racial groups across genders, highlighting across-sex variation in returns to family roles. To compare the effects of marriage and parenthood across men and women within the same racial or ethnic group, we use a set of interaction terms. Relying on the same set of controls described previously, we include an interaction between the female indicator and the associated variables for each set of relevant characteristics. We run three regressions for each racial group: the first includes interactions between females and both married and previously married (relative to never married); the second includes the interaction of female and a parenthood indicator; the third includes the interaction between female and each child age group. The coefficients on the interaction term for each of these regressions are shown in figure 4.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 4.

Differential Impact of Family Status for Females Compared to Males by Race

Source: Authors’ calculations using American Community Survey one-year estimates from 2009–2022.

Note: We run models separately for each racial group. Each marker is the coefficient for an interaction term between female and the variable displayed on the y-axis. We display 95 percent confidence intervals. Controls include: female, children indicators, including children 0–5, children 0–18, children 5–18 (omitted = childless) [except in the parenthood regression where these controls are omitted]. Marital status indicators include married and previously married (omitted = never married). Citizenship indicators include naturalized citizen and foreign-born noncitizen (omitted = citizen from birth). Additional controls include age and age squared. Field-of-degree indicators include engineering, other STEM, business, other non-STEM (omitted = computer science). Highest degree indicators include master’s, professional degree, PhD (omitted = bachelor’s). Computer science occupations include CIS manager, computer scientist, computer systems analyst, information security analyst, computer programmer, web developer, support specialist, database administrator, network administrator, network architect (omitted = software development).

The vertical dashed line crossing through zero shown in figure 4 reflects the balance between returns for men and women of the characteristic being interacted. The coefficients can be interpreted as the difference between females and males in the effect of the variable on income, not the overall effect of the characteristic on income. Returns to family attributes are universally lower for women than for men. In other words, even when women receive a marriage premium, it is significantly lower than that received by their male counterparts. White women experience the smallest penalties relative to their male counterparts, while Indian women generally experience the largest penalties relative to Indian men. Compared to White men who are married, married White women earn 11 percent less per year, on average, which is significant at the 1 percent level. But the marriage penalty between married Indian women and men is substantially larger, and the difference from Whites is statistically significant; married Indian women earn 21 percent less per year than married Indian men. The marriage premium for Chinese men relative to women is 15 percent, which is not statistically different from the estimates for White or Indian women.

A similar pattern is observed with regard to the parenthood penalty. White women also face significantly smaller parenthood penalties versus white men when compared to the parenthood penalties Indian mothers face relative to Indian fathers. In our specification with just a parenthood indicator, White women have about an 8 percent parenthood penalty compared to White men (significant at the 1 percent level). On the other hand, this penalty for Indian women is 15 percent (p < .001) and this estimate is also significantly more negative than the parenthood gap for White men and women. The parenthood gap between Indian women and men with older children is likely what is driving the larger parenthood gap for this group.

Given these large ethnic differences in returns to family attributes—particularly the wide disparities in returns to marriage—we further explore the role of occupational homogamy and whether those whose spouse is working in the same field or occupation are disadvantaged in terms of relative earnings. We do this by focusing on married couples and running regressions for each racial and gender group separately; the outcome is the log of wage and salary income, consistent with our previous analysis of the entire sample. We first include only an indicator for whether a respondent’s spouse was in a STEM occupation, before further limiting it to those where both partners worked in CS. Then we include the controls described earlier, along with this indicator. The coefficients on the spousal occupation indicator are shown in figure 5.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 5.

Effect of Spouse in Computer Science (STEM) Occupation on Wages by Race and Gender

Source: Authors’ calculations using American Community Survey one-year estimates from 2009–2022.

Note: Models are run separately for each race by gender pair. Each marker displays the coefficient estimate for the coefficient labeled on the y-axis. We display 95 percent confidence intervals. Controls include: female, children indicators, including children 0–5, children 0–18, children 5–18 (omitted = childless) [except in the parenthood regression where these controls are omitted]. Marital status indicators include married and previously married (omitted = never married). Citizenship indicators include naturalized citizen and foreign-born noncitizen (omitted = citizen from birth). Additional controls include age and age squared. Field-of-degree indicators include engineering, other STEM, business, other non-STEM (omitted = computer science). Highest degree indicators include master’s, professional degree, PhD (omitted = bachelor’s). Computer science occupations include CIS manager, computer scientist, computer systems analyst, information security analyst, computer programmer, web developer, support specialist, database administrator, network administrator, network architect (omitted = software development). CS = computer science. STEM = science, technology, engineering, and mathematics. CIS = Computer and Information Systems.

The results of these models demonstrate that for men working in CS, having a spouse employed in either a STEM occupation more broadly, or in CS more specifically, is associated with roughly a 5 percent increase in income. The one outlier here is Chinese men, who appear to particularly benefit from having a spouse in CS, which is associated with income gains of around 10 percent, at statistically significant levels (though the confidence intervals around this estimate are fairly wide). Notably, for Chinese and White women, having a spouse in CS (STEM) is also associated with income gains, with these gains also being around 5 percent in the naive model, a difference that shrinks to around 3 percent when controls are included. Chinese women are again an exception to this. Like their male counterparts, they appear to particularly benefit from having a spouse in CS; having a husband who works in CS is associated with income gains of around 10 percent for Chinese women, at statistically significant levels. For Indian women who work in CS, being married to a spouse who also is employed in CS is associated with a 5 percent decline in income, significant at conventional levels. This is striking, as nearly 60 percent of Indian women in our sample have spouses who work in CS.