Cross-Era Gender Differences in Educational Attainment Among Second-Generation Immigrants

Abstract

Starting in the 1990s, the United States experienced a gender revolution in education whereby later-born cohorts of women surpassed men in rates of higher education completion. However, little research has explored how gender differences in education for second generation immigrants compare to the children of U.S.-born Whites over historical and contemporary time periods. Immigrants arrive with varying levels of socioeconomic status and may come from countries with paternalistic ideologies that reinforce traditional gender norms. However, immigrants also experience assimilation over time and may begin to mirror the U.S.-born in their educational outcomes by the second generation. While national trends show that women have surpassed men in years of education, we question whether immigrants will experience similar trends, or whether their outcomes will vary by national origin. We analyze newly obtainable linked census data collected from 1940 to the present, a timeline where linked data were previously unavailable, to test these ideas. These data offer insight into gendered trends in education by family background and socioeconomic status using a broader timeline than studies before. We find that while there is variation in the degree to which gender differences in education occur by ethnic origin, overall trends for immigrants are similar to those for U.S.-born Whites regardless of socioeconomic status in childhood. In the Industrial Era, men generally attained more years of schooling than women. However, there is greater gender equality and often a female advantage occurring for most groups in the post-Industrial Era. Educational trends for Blacks are an anomaly, whereby women have attained more years of schooling than men in both Eras.

Introduction

Immigrant integration is a gendered process (Anastasiadou et al. 2023; Boyd 2002; Curran et al. 2006; Donato et al. 2006; Itzigsohn and Giorguli-Saucedo 2005). One particularly striking example is that second generation daughters of immigrants have better educational outcomes than sons of immigrants even though many immigrant families originate from countries where men have advantages in schooling and the labor market (Feliciano and Rumbaut 2005; Suárez-Orozco and Qin 2006). However, it is unclear whether the female educational advantages among children of immigrants are amplified among immigrant families or if they mirror broader patterns and trends occurring in the U.S. population.

On the one hand, female advantages in schooling have been attributed to factors that stem from immigrants’ family strategies and experiences with discrimination, such as greater monitoring of daughters than sons, which is associated with more attention to homework, and the “criminalization” of boys from immigrant families by school teachers and administrators, which is associated with negative school experiences (Feliciano 2012). These patterns are especially pronounced among families with low socioeconomic status (Feliciano 2012), which may stem in part from familistic cultural patterns often observed among poor labor migrants (Toyokawa and Toyokawa 2019; Marsiglia, Parsai, and Kulis 2009).

On the other hand, research on the female advantage among children of immigrants has largely been conducted among contemporary immigrant groups, after American women started attaining higher levels of education than men (Diprete and Buchmann 2006; Jones 2009; Parker 2021). What might appear to be unique to immigrant families may instead reflect the assimilation of American gender norms or adjustment to the ways in which gender constrains and shapes opportunities in American society for all groups. Some support for this idea comes from research showing that even though gendered patterns of educational attainment and employment among first generation immigrants in the United States mirror those from immigrants’ source countries, such associations fade over time and across generations (Blau 2015).

In this paper, we use linked U.S. census data to examine gender differences in educational attainment among the children of immigrants and children of U.S.-born parentage across the twentieth century, from those born in the 1920s (i.e., the children of Industrial Era immigrants) to those born in the 1970s (i.e., the children born of the post-1965 immigration waves, or the “post-Industrial Era”). Our data linkages attach parental nativity and socioeconomic status, measured in childhood, to the adult records of children of immgrants for both contemporary and historical cohorts. These novel data enable us to advance beyond research on how gender and early life socioeconomic status shape educational attainment among children of immigrants, which has been conducted on contemporary samples (see Park, Nawyn, and Benetsky 2015; Feliciano 2012), with historical comparisons. Prior data that seek to answer questions of intergenerational integration have been limited to either the historical time period (e.g., the late 1800s and early 1900s; see Catron 2019) or the contemporary time period (post-1965) and do not include both eras simultaneously, which these data do. The historical sweep of our data enables a valuable test of the distinctiveness of gender inequality among immigrant families, as we can compare gender differences in adult educational attainment between children of immigrants and children of U.S.-born parents over time across social contexts. If we see female advantages or greater equalitarianism among Industrial Era children of immigrants, even before the gender crossover occurred among American adults, this would provide stronger evidence of unique gender dynamics among immigrant families. Alternatively, if we see gender patterns for children of immigrants that evolve over time in the same manner as occurs among children of U.S.-born parents, this would suggest that immigrant families conform to U.S. forms of gender stratification.

One of the challenges of comparative historical research is that immigrants and their circumstances have changed dramatically over time. Just as gender roles in American society have changed, so have immigrants. Most Industrial Era immigrants originated from Europe and arrived with very low levels of education (Van Hook and Bachmeier 2024), while the majority of contemporary immigrants come from Latin America and Asia and arrive with a mixture of high and low educational attainments (Foner 2013). We therefore account for the socioeconomic backgrounds of second generation adults (measured in childhood), and we focus on three national origin groups that we are able to track in sufficient numbers across the twentieth century: Mexicans, Italians, and Northwestern Europeans. This enables us to compare the same groups in both the Industrial and contemporary periods. The inclusion of Mexicans and Italians in the study also allowed us to test theoretically-derived expectations about how source country gender norms may constrain young women’s opportunities for some groups, and how racial barriers may be detrimental to young men’s educational opportunities among other groups. Northwest Europeans were selected as an immigrant reference group due to their similarity to U.S.-born Whites in their socioeconomic integration and mobility (see Lowrey et al. 2021). Overall, the results shed light on how the intersection of gender, national origin, and socioeconomic status shapes the integration of the children of immigrants, a group that makes up about 30 percent of young adults today.

Background

The Gender Revolution in the United States

Before the 1970s, American women were less likely than men to acquire higher education (Goldin 2006). Women who did go to college often came from families with high socioeconomic status, whereas women from poorer class backgrounds were more likely to stay home or work to help take care of the family (Sassler and White 1997). However, changes in the social and economic landscape began to shift as early as the 1950s. Feminist movements, the introduction and practice of the use of contraceptives, and the changing policy sphere with the introduction of antidiscrimination measures such as the Equal Pay Act of 1963, and the Civil Rights Act of 1964 all led to increased levels of educational attainment (Jones 2009, p.249). In these years, women of all skill levels became less likely to leave work for marriage and instead began to prefer careers before and sometimes instead of family life (Becker 1981; Cherlin 2010). As women began to anticipate the labor market, they also invested in education to obtain high-skilled, higher paying jobs (Goldin 2006).

The structural changes during the 1980s from a manufacturing- to a knowledge- and service-based economy may have also advantaged women more than men. As opportunities for white collar jobs in office and service settings expanded, women tended to move from “pink-collar” clerical and service work to higher-skilled occupations. Conversely, as well-paid manufacturing work moved overseas or was mechanized, men tended to shift toward lower-skilled occupations (Autor 2010). Men of course still hold many white-collar positions. Yet it is important to keep in mind that roughly 70 percent of the labor force is blue-collar where a higher education may be optional (e.g., trade jobs that may receive on-the-job training). For this sizable group, men from lower class backgrounds may not have strong incentives to pursue higher education.

Together, these cultural and economic shifts have contributed to a gender crossover in educational attainment. By the 1980s, women began to have similar levels of education as men, and by the end of the 1990s, they began to outcompete men in college completion (Goldin 2006). Today, White men have lower rates of college completion than women, and that gap appears to be continually growing over time (Hurst 2024). Importantly, the female advantage in education appears to be most pronounced among those with lower class backgrounds (Diprete and Buchmann 2006), perhaps a reflection of the shrinking opportunities noted above for working class men.

Just as gender differences in educational attainment vary by class, they also vary dramatically by race. There is no evidence of a gender crossover among African Americans. Rather, Black women have had higher levels of educational attainment than Black men since at least 1940 and the gender gap has grown in recent decades. Some scholars have attributed the persistent male disadvantage among African Americans to mass incarceration, but the empirical evidence for this explanation is weak (McDaniel et al. 2011). Rather, low returns to schooling may disincentivize school attendance. For example, highly educated African American men have historically been shut out of high-status occupations such as engineering, and thus have faced lower rates of employment and higher rates of underemployment in occupations for which they are overqualified (McDaniel et al. 2011; Rury 2020). At the same time, African American women have historically had higher employment rates than White women (perhaps in response to lower employment among African American men), and have had greater returns to their education in high-skilled public-sector occupations, such as teaching (McDaniel et al. 2011). This research does not fully explain why Black men are rewarded less for their achievements, or why black women fare better in the public-sector occupations than Black men. However, these patterns of discrimination may be milder forms of other instances of historical racism (e.g., lynching) that together suggest that black men posed threats to white masculinity, leading them to be more often criminalized and violently victimized in retaliation for crossing racial boundaries.

Expectations for Immigrant Families

How might we expect gender to shape educational attainment among the second generation children of immigrants both historically and today? Although research on the gender crossover among children of immigrants is sparce (Donato et al. 2006; Park, Nawyn, and Benetsky 2015), research on gender differences in educational attainment among contemporary children of immigrants provides a basis for developing research expectations. We outline three possibilities.

One is that we will see a “male advantage” among the children of immigrants, namely a persistent male advantage in educational attainment over time relative to White children of U.S.-born parents. This could be due to the retention of traditional gender roles from source countries. As the gender revolution progressed in the US, parts of Europe, and other Westernized countries, traditional gender ideologies persisted in places such as Mexico and Italy, where women’s roles were constrained to that of a homemaker, caregiver, and mother and men’s roles extended outside the home and into the workforce (Anxo et al. 2011; Lam, McHale, and Updegraff 2012; Landale and Oropesa 2007). Patriarchal gender norms may take form in the strategies immigrants employ to gain a foothold following migration. For example, the combination of patriarchal gender norms and limited resources may have led immigrant parents to invest more in boys’ education than girls’ (Fuligni, Tseng, and Lam 1999), especially during the earlier years of the twentieth century when women’s opportunities in the U.S. labor market were more limited (Mickelson 1989).

Indeed, prior research suggests that cultural norms surrounding gender equality and family obligations may be maintained or even reinforced after migration to the U.S. (Parrado and Flippen 2005; Zhou 1992). For example, one study on Mexican women found that those residing in the U.S. did not experience “liberation” but rather exhibited higher compliance with traditional gender norms than even their counterparts back in Mexico (Parrado and Flippen 2005). Much of the reasoning for this was due to an increased dependence on men and Mexican women’s limited autonomy in the workforce. Similar trends occurred for early arriving Italian immigrants (years 1915-1925). Economic need coupled with cultural norms led Italian parents to send their children (especially girls) to work rather than school, leading to lower rates of school enrollment among girls (Perlmann 1988; Rury 2020). These practices date back to immigrant’s origin communities in Italy, where education was undervalued, especially for girls who could learn what they needed by continuing to work within the home or through apprenticeship (Covello 1967). Italian American women lagged behind their brothers and their ethnic American counterparts in education until about 1960 (Alba 1985). This leads us to expect daughters of immigrants from more patriarchal origin societies to have lower educational attainment than their male counterparts, and to have experienced a stall in the gender crossover in education as their families may invest more in sons’ than daughters’ education, even in the face of growing opportunities for women in the U.S. labor market.

A second possibility is that the opposite occurs. Rather than a heightened and persistent male advantage, we might see more gender equality and earlier signs of a “female advantage” among immigrant families than children of U.S.-born parents, especially among immigrant families that are engaged in labor migration or who face an acute economic need for boys to work outside of the household. This expectation derives from research on contemporary immigrant families. This work shows that immigrant families often exert pressure on young men and women to help support and consider the needs and wishes of the family (Fuligni 2001). Yet this pressure is highly gendered, leading boys and young men to work outside the home, which in turn could interfere with their schoolwork (Telzer and Fuligni 2009). For example, in traditional Mexican households where a father is absent, boys and young men are often the first to leave the home, sometimes dropping out of school, to secure financial stability for the family (Antman 2011). Conversely, gender role expectations in these families may be concordant with girls investing time in schooling. Girls are often used as language brokers by their non-English-speaking parents, which is positively associated with educational attainment (Blair and Cobas 2006; Lopez et al. 2002). Additionally, Mexican parents often grant less autonomy to daughters than to sons (Raffaelli and Ontai 2004), especially for children with higher associations with deviant peers (Updegraff et al. 2010), and thus girls may spend more time in school or at home focusing on schoolwork. For example, girls with greater parental monitoring have higher educational achievement, while higher levels of familism among immigrant boys may help to explain their lower levels of education (Feliciano 2012; Lopez 2003).

For some immigrant groups, male disadvantages in educational attainment may also stem from structural racism. Boys from Latino and West Indian immigrant families are often treated more punitively by school teachers and administrators, leading to them having more negative school experiences (Lopez 2003; Feliciano 2012). For similar reasons, male disadvantage may have been evident among Latinos in the past. In Texas and other parts of the South, where many Mexican immigrants settled during the early twentieth century, race-segregated schools and punitive Jim Crow laws (e.g., vagrancy laws) operated to secure and control a pool of low-skilled, inexpensive labor. Low-paying jobs were often filled by Latino immigrant men, and explicit efforts were made to funnel workers (often boys and young men) away from schools and into the labor force at early ages. For example, Montejano’s history of Mexicans in Texas describes how there was little interest in providing educational opportunities to Mexican-American children in Texas because schooling kept older children out of the labor force and could lead them to seek middle-class occupations. As Montejano noted, “The key element in Mexican school policy was the concern of farmers in securing and controlling farm labor. If farmers were to keep this labor reservoir, Mexicans had to be kept ignorant” (p. 191). Although Mexican-origin women also felt the negative effects of structural racism, Mexican-origin men may have been especially likely to drop out of school at an early age given the pressure for men to fulfill their breadwinner obligations and the demand for their labor by employers.

Italian men may have also felt the negative impacts of racism given their low status as labor migrants and their “non-white” racialized status within the U.S. that stereotyped them as criminally inclined and ill-suited for high-skilled labor (see Alba and Nee 2003). However, these factors may have been offset by the educational opportunities available to them (Van Hook and Bachmeier 2024). In the early twentieth century, Italian immigrants predominantly settled in the Northeast, a region of the country where compulsory school attendance laws were more likely to be enforced than in the South, even against the wishes of immigrant families who would have preferred their sons to work rather than attend school (Sassler 1996). These considerations led us to expect early signs of equalitarianism (or even female advantage) and an accelerated gender crossover in education among the children of immigrants like that observed for African Americans, especially for immigrant groups involved in labor migration and who were subjugated under Jim Crow systems of structural racism (e.g., Latinos).

A third possibility is “U.S. conformity”, whereby immigrants followed the same trends as the broader American population. As theories of selective assimilation suggest, immigrants may initially preserve aspects of their origin country’s values and culture while also integrating socioeconomically (Portes and Zhou 1993). Yet, by the second generation, children of immigrants may adopt “Americanized” gender roles and attitudes. Additionally, the structural changes in the U.S. labor market and schools that led to the gender crossover among American families may have impacted children of immigrants in the same way. To be clear, this seems most likely among immigrants who originate from places with egalitarian gender ideologies (e.g., Northwestern Europe) but may occur even among immigrants who originate from more patriarchal cultures (see Hondagneu-Sotelo 1992).

To summarize we expect to see the “male advantage” pattern among immigrants from cultures with gender ideologies that limit girls’ opportunities in schooling and careers, such as Italians. This pattern may be especially prominent for lower-SES families, given that families may prioritize sons’ over daughters’ schooling when funds are limited. In contrast, we expect to see a “female advantage” pattern among nonwhite labor migrants, such as Mexicans, as well as among African Americans. Mexican boys and young men may feel pressure to work rather than attend school to contribute to family expenses, and discriminatory practices may have made it difficult for them to attend school. Again, this pattern may be especially pronounced for lower-SES families because they are more likely to engage in labor migration. Finally, we expect to see the “U.S. conformity” pattern among Northwestern Europeans, who are culturally most like non-Hispanic White Americans and who experience little labor market discrimination.

Data and Methods

Data

To examine gender differences among second-generation immigrants over time and by family socioeconomic status, we need to account for parental place of birth and family background in childhood. Unfortunately, data limitations have made this difficult to achieve. Much of the research on second-generation attainment in the U.S. comes from the Current Population Survey (CPS) and/or decennial census data, such as Park et al. (2015), but the CPS did not consistently collect data on parental place of birth until 1994, and it has never collected data on family background in childhood. Other U.S.-based research has relied on the Children of Immigrants Longitudinal Study (CILS) (e.g., Feliciano and Rumbaut 2005; Feliciano 2012), but those data focus only on children of the post-1965 immigrant waves and do not permit comparisons with earlier waves of immigrants. These two studies are the only two (to our knowledge) that use longitudinal data to assess how family background is related to gender differences in education, and these studies rely on contemporary data (CILS). The historical nature of our linked dataset allows us to overcome prior shortcomings in data availability.

Specifically, we use several U.S. Census Bureau surveys to create “linked-generation” samples of first- and second-generation individuals. These surveys provide nationally representative population-level estimates that help determine demographic and socioeconomic characteristics of individuals who reside in the U.S. In an initial census survey, we identify first-generation immigrant parents and their co-residential second-generation children, and then find and link the second-generation individuals to their adult records in a subsequent survey. We attach the same ethnic origin identification of the first generation to the second generation. Likewise, we attach family background information (e.g., household income, parental education, and family size and structure) from childhood to adults’ records. We recognize that this methodology does not allow for changes in self-identification for the second generation, but it does help us to follow the same national origin groups over time. Spanning several decades, our data permit us to examine changes in educational attainment across birth cohorts for the adult children of immigrants from multiple national origin groups. Our linked data sample is novel because it is the first to span historical and contemporary periods in two distinct immigration eras (mass European immigration in the early 1900s and an era of increased diversity of immigrant inflows beginning in 1965). We use the linked data to confirm already-observed trends in gender gaps in education and to delve deeper into our understanding of the socioeconomic characteristics that shape educational trajectories by gender. For example, our linked data approach allows us to interact several variables such as childhood socioeconomic status, gender, origin group, and birth cohort to reveal more detailed trends in educational attainment than studies before.

Sample

The data for this study are drawn from U.S. census data collected over the last eight decades: the full-count 1940 census, years 1973, 1979, and 1981-1990 of the Current Population Survey (CPS), the 2000 census long form, and years 2001-2018 of the American Community Survey (ACS). We use these datasets to link family members across surveys to culminate several two-generation family files (or “vintages”). The U.S. Census Bureau has attached protected identification keys (PIKs) to individuals in these data files as part of its CLIP project (the Census Longitudinal Infrastructure Project—a collaborative effort to create new datasets that link individual census records across surveys), which allows us to link individuals across the data sources. For the first vintage, parents with co-residential children were identified in the 1940 census (the “first family generation”). Then, the children of those parents were identified and linked to their later adult records in the 2000 census and/or the ACS (the “second family generation”). The second vintage follows a similar methodology, yet includes individuals found in the 1940 census who are then linked to the CPS. The third vintage links the CPS to the 2000 census and/or the ACS, and the fourth links the 2000 census to the ACS. The four vintages together encompass our final dataset, representing birth cohorts from 1920 to 1970. Individuals with duplicate records were limited to their first entry based on vintage order (see Figure 1 for a depiction of the vintages). Because the CPS’s collected before 1994 do not include information on place of birth, we attached Social Security Administration (SSA) records to the CPS to find whether individuals were born outside of the U.S. The data linkages allow us to attach parental and childhood characteristics (e.g., parental place of birth and parental education) to our second-generation adult sample.

Figure 1.

Description of Vintages by U.S. Census Survey and Generation

The final analytic sample is limited to individuals born between 1920 and 1970 who are the second generation children of immigrants, and who were 25 or older at the time they were followed up as adults (N = 680,000). We supplemented the sample with a similarly selected comparison group of children of U.S.-born non-Hispanic Whites and African Americans (N = 649,000). Our final sample N is 1,329,000¹.

Measures

Our two main outcomes are related to the educational attainment of the second generation: Years of schooling and college completion (0/1). Years of schooling is a continuous measure of the highest grade of education attained for adults aged 25 and older.

Ethnic origin is a family-level measure based on the place of birth and racial-ethnic identification of the parents in each family. Ethnic origin is treated as fixed across generations even if the second generation no longer self-identifies with the ethnic group. This is an important strength of this study (Alba 2020; Duncan and Trejo 2018). We distinguish among families in which at least one parent was born in Mexico, Italy, and Northern and Western (N/W) Europe. The countries we include in the N/W European group are Finland, Iceland, Sweden, Norway, Scotland, Belgium, Denmark, Switzerland, France, Ireland, the United Kingdom, Monaco, Luxembourg, Liechtenstein, Lapland, and the Netherlands. In 1940, N/W Europeans comprised about 2.5 percent of the total U.S. population (just over 3.2 million people) and made up 31 percent of Europeans living in the U.S. In contrast, Mexicans comprised 0.3 percent of the total population and Italians comprised 1.3 percent (and 16.3 percent of the total U.S. European population) (author’s calculations from the public-use 1940 census data).

We also include families in which both parents were U.S.-born non-Hispanic Black, and the reference group includes families in which both parents were U.S.-born non-Hispanic White. For simplicity, we refer to these two groups as “Blacks” and “Whites” respectively throughout the remainder of the article. All groups were represented in sufficiently large numbers across both the earlier and later born cohorts in our sample. Additionally, Italians and Mexicans were selected because of the similarities between historical era Italians and contemporary Mexicans in status and racialized positions (e.g., Perlmann 2005). In the contemporary era, Black and Mexican adults are often compared due to their low levels of socioeconomic status and the levels of racialization and discrimination that they experience (Brown, Jones, and Becker 2018; Portes, Fernández-Kelly, and Haller 2005; Portes and Rumbaut 2001; Portes and Zhou 1993). We include North/West Europeans as a reference group because they are a higher-resourced migrant group often viewed as having experienced upward mobility and socioeconomic integration with Whites.

We examine educational differences for each ethnic origin group by birth cohort and sex (0/1). Birth cohorts are reported in ten-year intervals between 1920-1970 and are further grouped by Industrial Era (1920-1940s) and post-Industrial Era (1950s-1970s). In regression analyses, we account for several demographic factors: family size in childhood (continuous), whether the person lived with two parents in childhood (0/1), whether the family was above or below the median income in childhood (0/1), parental education (averaged across both parents), employment status in adulthood (0/1), and survey year (continuous) to account for any secular increases in educational attainment that may have occurred during the year covered in the CPS or ACS samples. The means for all analytic variables are presented in Table 1.

Table 1.

Sample Descriptives for Second Generation by Gender

	Men		Women
	Mean/Proportion	Std. Dev.	Mean/Proportion	Std. Dev.
Origin
Mexican Children of Immigrants	0.0424	0.2015	0.0417	0.1998
Italian Children of Immigrants	0.3159	0.4649	0.3117	0.4632
N/W European Children of Immigrants	0.1638	0.3701	0.1497	0.3567
Black Children of U.S.-born Parents	0.3754	0.4842	0.4050	0.4909
White Children of U.S.-born Parents	0.1025	0.3033	0.0919	0.2889
Birth cohort
1920	0.4950	0.5000	0.5094	0.4999
1930	0.4238	0.4942	0.4081	0.4915
1940	0.0347	0.1829	0.0323	0.1768
1950	0.0084	0.0911	0.0090	0.0947
1960	0.0145	0.1195	0.0157	0.1242
1970	0.0236	0.1518	0.0255	0.1577
Survey year	1943	12.23	1943	12.66
Family size	6.06	2.17	6.12	2.16
Two parents in childhood	0.8488	0.3582	0.8342	0.3719
Years of schooling
Own education	12.11	3.57	11.97	3.00
Parent's education	6.45	3.52	6.35	3.53
Educational credentials
High school completion	0.6988	0.4588	0.7300	0.4439
Post secondary education	0.3396	0.4736	0.2640	0.4408
College completion	0.1980	0.3985	0.1277	0.3338
Above median income	0.4736	0.4993	0.4709	0.4992
N	600000		729000

Note: Median income adjusted to 1982/1984 dollars. Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Analytic Strategy

To assess whether and when the gender crossover occurred for the immigrant second generation, we first provide descriptive patterns in years of schooling and the percentage with a college degree across ethnic origin groups (Table 2). We then estimate regression models (OLS linear models for years of schooling and logistic regression for college completion) that include the demographic controls listed in the previous sections. To assess how gender differences changed across birth cohorts and ethnic origin groups, we tested three-way interactions among these three factors, and then graphed predicted values generated from these models in Figures 2 and 3 (see Appendix A for regression results). In models predicting years of education, we also assessed whether these gender differences were more or less pronounced for those from low-SES families by testing four-way interactions: gender x birth cohort x ethnic origin group x above/below median income. We were unable to estimate these models for college completion due to sparce cell sizes. As before, we generated predicted probabilities while holding all other characteristics at their mean levels, and plotted the predicted values in Figure 4. While we provide an example of our models in Appendix A, we do not include the regression results in the main text due to the complexities associated with interpreting three and four-way interactions. Coefficients from regression analyses are available upon request from the authors. All results have been approved for disclosure under review numbers CBDRB-FY24-0437 and CBDRB-FY22-P2357-R9503.

Table 2.

Educational Outcomes by Ethnic Origin Group, Gender, and Birth Cohort

		Years of Schooling		Percentage with a College Degree
		Industrial Era Birth Cohorts	Post-Industrial Era Birth Cohorts	Industrial Era Birth Cohorts	Post-Industrial Era Birth Cohorts
Children of Mexican Immigrants	Women	10.03	13.18	6.70	22.73
	Men	10.55	12.83	12.16	17.68
	Difference	0.52 * b	−0.34 *	5.46 * a	−5.05 * b
Children of Italian Immigrants	Women	12.07	14.53	10.03	43.47
	Men	12.69	14.88	22.22	51.70
	Difference	0.62 * b	0.35 b	12.19 * b	8.24 * b
Children of North/West European Immigrants	Women	12.66	14.65	15.11	46.86
	Men	12.96	14.38	26.42	41.98
	Difference	0.31 * a	−0.27	11.31 *	−4.88 * b
Children of Black U.S.-born Parents	Women	11.44	13.35	12.06	23.78
	Men	10 95	12.49	13.37	12.88
	Difference	−0.49 * a	−0.86 * a	1.30 * a	−10.90 *
Children of White U.S.-born Parents	Women	12.78	14.16	17.00	42.11
	Men	13.16	13.68	28.34	32.94
	Difference	0.38 *	−0.48 *	11.34 *	−9.17 *

Industrial Era birth cohorts = 1920s-1940s; post-Industrial Era birth cohorts = 1950s–1970s "Any views expressed are those of the authors and not those of the U.S. Census Bureau. The Census Bureau has reviewed this data product to ensure appropriate access, use, and disclosure avoidance protection of the confidential source data used to produce this product. This research was performed at a Federal Statistical Research Data Center under FSRDC Project Number 2357. (CBDRB-FY24-0437)"

^{a =}gender difference is significantly different from that observed among Whites and is either more equalitarian or shows a greater female advantage (statistical significance is based on Clogg tests)

^{b =}gender difference is significantly different from that observed among Whites and shows a greater male advantage (statistical significance is based on Clogg tests)

Figure 2. Predicted Gender Differences in Years of Schooling by Ethnic Origin and Birth Cohort.

N = 1,329,000; error bars depict 95% confidence intervals; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Figure 3. Predicted Gender Differences in College Attainment by Ethnic Origin and Birth Cohort.

N = 1,329,000; error bars depict 95% confidence intervals; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Figure 4. Predicted Differences in Years of Schooling by Ethnic Origin Group and Childhood Income Level (Full Controls).

Note: N = 1,329,000; standard error bars are included; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Results

Descriptive Patterns in Education

Estimates of average years of schooling and the percentage that completed college are shown in Table 2 by gender, birth cohort, and ethnic origin. In the table, we show the size and significance of gender gaps for each birth cohort and ethnic origin group. Where the difference between men and women’s education is positive, men have an educational advantage. Asterisks denote statistically significant gender differences. We also tested whether the gender differences for each ethnic origin group differed significantly from those observed for children of White parents. Gender differences that are more equalitarian (i.e., a smaller male advantage) or show a greater female advantage than among Whites are marked with an “a”; gender differences that show a greater male advantage are marked with a “b”.

First, we tend to see similar gender patterns and trends among second generation origin groups as for Whites. For Industrial Era cohorts, men attained more education than women across all groups. Blacks were the sole exception, whereby Black women attained more years of schooling on average than Black men (although Black men were slightly more likely to have completed college). Among post-Industrial cohorts, we see female advantages in educational attainment across most groups. The only exceptions are the children of Italian immigrants, among whom we continue to see a significant male advantage in college completion and no significant gender difference in years of schooling.

Second, although the second generation experiences distinctive gender patterns in some cases, such distinctions are more likely to favor men than women. For Industrial Era cohorts, male advantages among the children of immigrants tend to be no different or significantly more pronounced than among Whites, and for post-Industrial Era cohorts, female advantages tend to be no different or significantly smaller than among Whites. The only exceptions are that the male advantages in (1) college completion for Mexicans and (2) in years of schooling for N/W Europeans, are significantly smaller than among Whites during the Industrial era. In any case, there are no circumstances in which female advantages among post-Industrial cohorts are significantly greater among children of immigrants than among Whites.

Circling back to our research expectations, the descriptive results are consistent with the “male advantage” model in the case of children of Italian immigrants. Italians are the only group that did not experience a gender crossover. To some degree, children of Mexican immigrants also mostly follow a “male advantage” pattern given their greater male advantage for Industrial Era cohorts, and smaller female advantage for post-Industrial cohorts than among Whites. Finally, results for children of N/W European immigrants are consistent with the “U.S. conformity” hypothesis as we see fewer (and smaller) significant differences from Whites. In any case, we do not find strong support for the “female advantage” model among any group of children of immigrants. Any female advantages observed for them tend to be smaller than among Whites. African Americans were the only group to show significant female advantages that are larger than among Whites among both Industrial Era and post-Industrial Era cohorts.

Multivariate Analyses: Gendered Patterns of Educational Attainment

To what extent do the descriptive findings hold after adjusting for childhood family background? To answer this question, we estimated OLS models of predicting years of education as a function of gender, ethnic origin, and birth cohort while adjusting for demographic characteristics and socioeconomic status. The three-way interactions were statistically significant. Predicted values derived from this model are displayed in Figure 2, expressed as predicted gender differences in years of schooling across ethnic origin groups and birth cohorts. The bars extending above the x-axis denote male advantages in education, while bars dipping below the x-axis indicate female advantages. The error bars depict 95% confidence intervals. The bars on the left in the dark grey shading are for Industrial Era cohorts, and those on the right in the lighter grey are the post-Industrial Era cohorts.

We focus first on Industrial Era cohorts. After adjusting for family background, men of all ethnic origins except Blacks attained more years of schooling than women. Similar to the descriptive findings, the male advantages are significantly larger for Mexicans and Italians than among Whites, and the male advantage for NW/Europeans is smaller than among Whites, although this contrast is substantively small (about a tenth of a year of schooling). Among post-Industrial Era cohorts, women attained significantly more years of schooling than men for all groups except Italians. Consistent with the “U.S. conformity” hypothesis, however, the female advantage among children of Mexican and N/W European immigrants is not significantly different than that among Whites (this departs from the descriptive findings, which yielded a significantly smaller female advantage among these two immigrant groups).

Similar patterns emerged when we examined college completion in logistic regression models that adjust for family background. Predicted values generated from these models again show male advantages among Industrial Era cohorts and female advantages among post-Industrial cohorts. Moreover, the male advantages are significantly larger for Mexicans and Italians among the earlier-born cohorts (consistent with the “male advantage” hypothesis) (Figure 3). Among post-Industrial Era cohorts, however, we see few significant differences from Whites in the female advantage (consistent with the “U.S. conformity” hypothesis), with the one exception being that we see no significant female advantage among the children of Italian immigrants.

We conducted supplementary analyses in which we replaced the dichotomized birth cohort indicator with ten-year birth cohort categories in our models. Consistent with the results shown in Figure 2, all groups except Blacks show a male advantage in years of schooling and college completion among earlier-born cohorts, but then started to experience a gender convergence for cohorts born in the mid 20^th century (see Appendix Figure 1). For children of Mexican immigrants, women appear to catch up to men slightly earlier, while Italian-origin women appear to take slightly longer (not until around 1960) and never show a significant female advantage. Finally, African Americans stand apart; Black women consistently have more years of schooling than Black men in all birth cohorts. Results portray similar patterns for college completion (Appendix Figure 2).

Overall, the results that adjust for family background point to similar conclusions as the descriptive results. They are consistent with the “male advantage” model in the case of children of Italian immigrants and “U.S. conformity” among the children of N/W European immigrants. Results concerning children of Mexican immigrants are mixed in that they are consistent with the “male advantage” model among earlier-born cohorts, but the “U.S. conformity” model for later-born cohorts.

Variations by Class

Before concluding that the gender patterns among children of immigrants do not align with the “female advantage” model, it is important to account for family SES. Prior research has shown much stronger evidence of a “female advantage” pattern among children of immigrants with low-SES family backgrounds (Feliciano 2012). To investigate this possibility in our data, we tested whether gender differences in years of schooling differed by childhood income level, specifically distinguishing between those with household incomes falling above versus below the median income level (measured in childhood).² Figure 4 shows the gender difference in years of schooling by ethnic origin group and SES background among Industrial Era cohorts. We do not show results for post-Industrial Era cohorts because the magnitude of the gender differences did not differ significantly by childhood income level. Given the large sample size required for testing four-way interactions, we caution that the lack of significance could stem from limited statistical power and should not be interpreted as evidence that class differences do not exist among post-Industrial cohorts.

Consistent with the research just cited (Feliciano 2012), we find male advantages in schooling to be smaller (or the female advantage to be greater) for those from low-SES families (Figure 4). For Italians and Mexicans of the Industrial Era, we see a male advantage regardless of income level, but outcomes were more equalitarian among low-SES families. Similarly, N/W Europeans experienced a significant female advantage only among low-SES families, and the female advantage among Blacks was much more pronounced among low-SES families than high-SES families.

However, the gender equalitarianism seen among low-SES children of immigrants was not particularly distinctive in that we see the same pattern among children of U.S.-born Whites. Moreover, schooling outcomes among low-SES children of Mexican and Italian immigrants were less equalitarian (i.e., more favorable for men) than among their White counterparts. Low-SES children of N/W European immigrants and Blacks were exceptions in that both groups exhibited significant female advantages in schooling.

Discussion

It is well documented that immigrant integration is a gendered process, particularly in the arena of schooling where women from immigrant families tend to outperform men (Feliciano 2012; Feliciano and Rumbaut 2005). But does this gender difference represent a distinctive “immigrant” pattern, perhaps emerging from discrimination directed toward boys and vigilant parental monitoring of daughters? Or is it a common pattern – perhaps a response to the incentive structures and norms that shape gender inequality across most groups in American society? To answer this question, we compared gender differences in educational attainment among the children of immigrants to those observed for the children of U.S.-born parents. Importantly, we examined both contemporary birth cohorts and those born during the first half of the twentieth century (i.e., Industrial Era cohorts), before the well-known gender crossover in schooling occurred in the United States. As argued here, studies of contemporary birth cohorts may not reveal large differences between immigrant and U.S.-born families because female advantages in schooling are now widespread in American society, but we may see a greater contrast among earlier-born cohorts when female advantages in education were less common.

One of our research expectations was that we would find “female advantage” among the children of immigrants, particularly among Mexicans and those raised in low-SES families, but this expectation was not borne out in the results. Consistent with prior research on contemporary children of immigrants, our analysis of post-Industrial Era cohorts revealed that second generation women attained more years of schooling and were more likely to complete college than men, especially among Mexicans and N/W Europeans. However, this pattern was not unique to immigrant families. We also see it for children of White and Black U.S.-born parents. Moreover, the female advantage among post-Industrial Era cohorts was significantly weaker for children of immigrants than for children of White U.S.-born parents, although this difference was not significant in multivariate models. Additionally, we did not find significant female advantages among pre-Industrial cohorts for any group except low-SES N/W Europeans and African Americans.

Rather than a “female advantage”, the results were more consistent with the “U.S. conformity” and “male advantage” hypotheses. First, children of immigrants tended to exhibit similar gender patterns and trends as children of White U.S.-born parents (“U.S. conformity”). For example, we see similar gender crossovers (both in their occurrence and timing) between the children of U.S.-born Whites and children of immigrants. Additionally, we find no significant differences between these two groups among contemporary cohorts. Where we do see distinctive immigrant patterns, they tend to favor men over women. For example, there is evidence of an especially large male advantage among pre-Industrial era Mexican and Italian birth cohorts, and an especially persistent male advantage (even among contemporary cohorts) among children of Italian immigrants. As we theorized earlier, this may owe to patriarchal gender norms that lead immigrant parents to prioritize boy’s over girl’s schooling in the face of resource constraints. The fact that we find especially large male advantages only for earlier-born cohorts may reflect a waning of patriarchal gender norms and growth of educational and labor market opportunities for women in immigrants’ source countries.

Finally, it is vital to underscore the contrast between children of immigrants and African Americans. One of our research expectations was that the gender differences in schooling for Mexicans (and to some degree, Italians) would mirror those seen for African Americans due to common experiences with racism in schools and the labor market. Yet children of U.S.-born Blacks stood apart from Mexicans and the other groups in that they alone experienced a significant and persistent female advantage (or perhaps more aptly, a male disadvantage) both for Industrial Era and contemporary cohorts. This may reflect the unequivocably disadvantaged position that African Americans have long occupied as racialized minorities in the American social stratification system (Lee and Bean 2010). Precisely how this translates into schooling disadvantages for Black men remains unclear and worthy of further research, but as we suggested earlier, it may stem from the threats black men posed to white masculinity, especially when they were perceived as crossing over into “white” spaces such as schools. In contrast, the racialization of Mexican Americans has been less consistent, varying across regions, time periods, and class (Fox and Guglielmo 2012).

This study has important limitations. One is that we are not directly able to observe the factors that underlie the gender differences we find (e.g., cultural norms about gender, gendered parenting practices, or discrimination in schools and the labor market). Even though we tend to observe similar gender differences and trends in education between children of immigrants and U.S.-born Whites, these similarities may emerge from different social processes. We are also unable to investigate birth cohorts before 1920 or after 1970 given limitations in the data and small cell sizes for those birth cohorts. Finally, our analysis is limited to immigrant groups that we were able to compare across time periods: Mexicans, Italians, and N/W Europeans. It is possible that we would find different results for other groups. Despite these limitations, this investigation adds new information that helps us better contextualize the educational advantages observed for daughters of contemporary immigrants: (1) these patterns are new, having emerged among cohorts born during the latter half of the twentieth century, and (2) they largely mirror the patterns and trends seen among U.S.-born White Americans.

Appendix

Appendix Figure 1. Predicted Years of Schooling by Ethnic Origin Group, Birth Cohort, and Gender (Full Controls).

Note: N = 1,329,000; standard error bars are included; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Appendix Figure 2. Predicted College Attainment (Percentage) by Ethnic Origin Group, Birth Cohort, and Gender (Full Controls).

Note: N = 1,329,000; standard error bars are included; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437

Appendix Table 1.

Regression Results for 2^nd Generation Educational Outcomes (Children of Immigrants and U.S.-born)

	Years of Education		College Completion
	Model 1A	Model 2A	Model 1B	Model 2B
Origin (Ref = USB whites)
Mexican	−1.214***	−1.570***	−0.668***	−0.722***
Asian	1.120***	1.278***	0.462***	0.437***
Italian	0.506***	0.499***	−0.126***	−0.150***
N/W European	0.012	−0.132***	−0.088***	−0.145***
Black USB	−0.736***	−1.449***	−0.400***	−0.613***
Survey year	−0.033***	−0.035***	−0.027***	−0.027***
Family size in childhood	−0.163***	−0.162***	−0.144***	−0.145***
Two parents in childhood	0.325***	0.320***	0.313***	0.318***
Sex
Female	−0.073***	−0.252***	−0.532***	−0.722***
Parent's education	0.256***	0.254***	0.750***	0.746***
Employed		0.872***		0.581***
Year of birth (1920 = reference)
1930	0.624***	0.386***	0.317***	0.135***
1940	0.846***	0.478***	0.339***	0.144***
1950	1.928***	1.004***	1.051***	0.356
1940	2.003***	0.851***	1.042***	0.08
1970	2.395***	1.266***	1.345***	0.430***
Sex * Origin interaction
Female * Mexican		−0.149**		0.022
Female * Asian		−0.531***		−0.289***
Female * Italian		−0.262***		−0.312***
Female * N/W European		0.073*		−0.060*
Female * Black USB		0.888***		0.585***
Year of birth * Sex
1930 * Female		−0.171***		0.053
1940 * Female		−0.151*		0.139***
1950 * Female		0.237		0.719***
1960 * Female		0.531***		1.113***
1970 * Female		0.998***		1.404***
Year of birth * Origin (Ref=1920 & USB white)
1930 * Mexican		0.677***		0.074
1930 * Asian		0.089		0.253***
1930 * Italian		0.149***		0.229***
1930 * N/W European		0.189***		0.138***
1930 * Black USB		0.447***		−0.025
1940 * Mexican		1.005***		−0.015
1940 * Asian		−0.082		0.132
1940 * Italian		0.113		0.324***
1940 * N/W European		0.153*		0.072
1940 * Black USB		0.650***		−0.153***
1950 * Mexican		0.842*		−0.325
1950 * Asian		−0.321		−0.274
1950 * Italian		0.601		0.743*
1950 * N/W European		0.26		0.323
1950 * Black USB		0.723***		−0.172
1960 * Mexican		1.671***		0.556*
1960 * Asian		0.056		0.714***
1960 * Italian		0.501		0.870***
1960 * N/W European		0.27		0.216
1960 * Black USB		0.792***		−0.115
1970 * Mexican		1.612***		0387*
1970 * Asian		−1.104***		−0.242
1970 * Italian		0.199		1.018***
1970 * N/W European		−0.058		−0.022
1970 * Black USB		0.616***		−0.243***
Year of birth * Sex * Origin (Ref = 1940, male, usb white)
1930 * Female * Mexican		0.034		0.004
1930 * Female * Asian		0.081		0.06
1930 * Female * Italian		0.077		0.137***
1930 * Female * N/W European		−0.093*		0.058
1930 * Female * Black USB		−0.006		0.003
1940 * Female * Mexican		−0.005		0.098
1940 * Female * Asian		0.061		0.165
1940 * Female * Italian		0.06		0.204**
1940 * Female * N/W European		0.009		0.210**
1940 * Female * Black USB		−0.131		0
1950 * Female * Mexican		1.550*		0.739
1950 * Female * Asian		0.308		0.48
1950 * Female * Italian		−0.323		−0.067
1950 * Female * N/W European		−0.074		0.255
1950 * Female * Black USB		−0.154		0.058
1960 * Female * Mexican		0.347		−0.253
1960 * Female * Asian		−0.076		−0.555
1960 * Female * Italian		−0.066		−0.05
1960 * Female * N/W European		0.161		0.091
1960 * Female * Black USB		−0.345*		−0.105
1970 * Female * Mexican		−0.23		−0.13
1970 * Female * Asian		0.49		0.18
1970 * Female * Italian		−0.459		−0.665
1970 * Female * N/W European		−0.393		−0.413
1970 * Female * Black USB		−0.715***		−0.367***
Constant	74.37***	78.38***	51.28***	51.02***
BIC	6850000	6833000	968000	957000

N = 1,329,000; error bars depict 95% confidence intervals; Data: Vintages 1-4 (1940 census; 1973, 1979, 1981-1990 Current Population Survey; 2000 census; 2001-2018 American Community Survey). Disclosure review number: CBDRB-FY24-0437