Generation, education, and intermarriage of Asian Americans

Abstract

The influx of immigrants from Asia to the United States (U.S.) has expanded the pool of coethnic marriageable partners, strengthened racial identity, and contributed to the decline in interracial marriage with whites among Asian Americans. Yet, retreat from interracial marriage with whites may well vary by immigrant generation, an important factor in marital assimilation. Using data from the March Current Population Survey (1994 to 2015), we examine generational differences in intergenerational marriage and interracial marriage with whites among Asian Americans. The results reveal that over time third-plus-generation Asians show no significant change in interracial marriage with whites but declines in intergenerational marriage with first-or second-generation Asians. Second-generation Asians, on the other hand, have become more likely to marry first-generation Asians and less likely to marry whites. In addition, education provides different opportunities for intermarriage, with highly-educated Asian Americans more likely than their less-educated counterparts to marry whites and less likely to marry other Asians. Notably, highly-educated second-generation Asians tend to marry third-plus-generation Asians and whites while their less-educated counterparts marry first-generation Asians. These findings highlight the importance of generation and education in integration of Asian Americans.

Asians in the U.S. are fast growing. The share of the Asian population increased from less than 1% in 1965, the year when Congress passed the Immigration and Nationality Act, to nearly 6% in the early 2010s (Pew Research Center 2013). Continuous waves of immigrants from Asia are the engine that drives this growth. According to the Pew Report (2013), new arrivals from Asia have outnumbered those from Latin America since 2009; in 2010, about 36% of new immigrants were Asian and 31% were Hispanic; and today, three quarters of Asians in the U.S. are immigrants. The large influx of immigrants from Asia has contributed to the decline in intermarriage with whites among Asian Americans (Qian and Lichter 2007). Asian immigrants, diverse in migration history, socioeconomic status, culture, language, and religion, are less likely to intermarry than their U.S.-born counterparts. At the same time, Asian immigrants provide a large pool of marriageable partners, which may increase marriages between U.S.-born and foreign-born Asians.

Asian Americans exhibited high levels of interracial marriage with whites in the past. In 1990, nearly half of U.S.-born Asians married whites (Qian, Blair and Ruf 2001). The high level of intermarriage was related to population size. Population size influences social interactions between groups (Blau 1977). When a population group is small, contact opportunities are low within the group and high outside the group. Thus, Asian Americans had limited opportunities to interact among themselves in schools, workplaces, or residential neighbourhoods. In recent decades, a growing Asian immigrant population has not only replenished the marriage pool for their native-born counterparts, but also increased cultural awareness, promoted ethnic solidarity, and developed in-group contact opportunities (Massey 1995, Okamoto 2014).

Indeed, the past few decades brought significant declines in intermarriage with whites among Asian Americans (including Asian immigrants, thereafter) (Qian and Lichter 2007, 2011). Yet, it is unclear to what extent foreign-born Asians marry their U.S.-born counterparts; whether each generation, especially the second generation—children of many immigrants who arrived after the congressional passage of the Immigration Reform and Control Act in 1986—married whites or other Asians; and whether educational attainment played a role in marriage with whites or other Asians. Our paper examines patterns of intergenerational marriage (i.e., marriage across first, second, and third-plus generations) and interracial marriage with whites among Asian Americans, by pooling annual data from the March Annual Demographic Supplements of the Current Population Survey (1994–2015). Specifically, we explore: 1) whether there has been an increase in intergenerational marriage and a decline in marriage with whites among Asians over time; 2) whether educational attainment plays a role in intergenerational marriage or interracial marriage; and 3) whether second-generation Asians are more likely to marry first-generation Asians or third-plus-generation Asians or whites. Our results reveal the role generation plays in Asian Americans’ integration in the U.S. and shed light on how differences in marriage with co-ethnics and whites by educational attainment or generation shape the future of American society. This article contributes to our understanding of how the role of marriage and assortative mating plays in assimilation among Asian immigrants and Asian Americans (Yeung and Mu 2019).

Assimilation and intermarriage

Gordon (1964) formulated classical assimilation theory based on the experiences of European immigrants who came to the U.S. around the turn of the twentieth century. The immigrants were diverse in language, religion, socioeconomic status, and national origin at the time of arrival. Yet it did not take long that they and their descendants became culturally integrated and achieved parity with their native-born counterparts in education and socioeconomic status. Cultural and structural assimilation fostered marital assimilation as intermarriage with U.S.-born whites became commonplace.

Classical assimilation theory views marital assimilation of immigrants into middle class white America as an inevitable outcome after cultural and structural assimilation. This theory explains well the experiences of European immigrants and their descendants at the turn of the twentieth century (Gordon 1964). Asian immigrants today are different. Immigrants from Asia are positively selected and Asian Americans in general have achieved high levels of educational attainment and socioeconomic status (Lee 2015, Pew Research Center 2013). Yet, as one of the racial minority groups, Asian Americans face low returns to education and high glass ceilings (Sakamoto, Goyette and Kim 2009). Asian Americans, regardless of whether they are first, second, or third-plus generation, are perceived forever foreign and experience routine prejudice and discrimination (Okamoto 2014). As a result, their racial minority status and their U.S. experiences suggest that they are unlikely to follow a single path of marital assimilation stipulated by classical assimilation theory.

Alba and Nee (2003) reformulate the assimilation theory, taking into account that American society is racially diverse and immigration continues in the U.S. This new assimilation theory does not assume a universal outcome and posits that assimilation may be segmented and take diverse paths. Asian Americans who have achieved success in education, employment, and residential location may be more likely than other Asians to marry whites. For them, cultural and ethnic differences pale in comparison to their social and economic standing. Yet, other Asian Americans, especially first-generation immigrants, are at the other end of the spectrum. They have low levels of education, do not speak English well, work in segregated workplaces, and reside in ethnic enclaves (Zhou 1992). They may seek to marry their U.S.-born counterparts as a way of getting connected with the communities and becoming eligible for naturalisation (Bean and Stevens 2003, Stevens, Ishizawa and Escandell 2012).

Clearly the growing number of Asian Americans, to say the least, increases the number of potential partners of the same race. Yet, opportunities to meet and marry an Asian are unequal and depend on an individual Asian American’s position in American society. The generation mix of Asians, that is, the shares of the Asian American population that comprise the first, second, and third-plus generations, may play an important role.

Generation and education

Asian Americans have had a long history in the U.S., dating back to the 1850s when Chinese immigrants came to work in gold mines and on railroads (Hirschman and Wong 1981). For a long time, Asian Americans, mostly Chinese and Japanese Americans, suffered from exclusion, discrimination, prejudice, and even internment in the case of Japanese Americans during World War Two. They were considered ‘unassimilable,’ banned from intermarriage, and unfit for citizenship until 1952 with the passage of the McCarren-Walter Immigration and Naturalization Act (Lee 2015). The third-plus generations of Asian Americans are descendants of earlier Asian immigrants who lived and worked in ethnic enclaves with limited outside contact.

The 1965 passage of the Immigration and Nationality Act finally opened doors for immigrants from Asia. Asian immigrants who arrived after 1965 were more socioeconomically selective than those who came in the nineteenth-and early-twentieth-century. They were more ethnically diverse than in the past, originating from China, the Philippines, Korea, India, and Southeast Asia (Hirschman and Wong 1986, Xie and Goyette 2004). Although ethnic enclaves still exist, immigrant selectivity puts many in integrated or predominantly white neighbourhoods (White, Biddlecom and Guo 1993). These immigrants’ children have come of age and become today’s second generations.

With the continuous influx of immigrants from Asia, we would expect that intermarriage with whites is least likely among new immigrants. First, this is due to their socioeconomic position. Despite their selectivity in socioeconomic status relative to the peers in their countries of origin, they do not speak English well, are less familiar with American culture, and have lower educational attainment than their U.S.-born counterparts (Lee and Edmonston 2005). The second reason is the population size. The structural theory of intergroup relationships argues that in-marriage necessarily increases as the relative population size increases (Blau, Blum and Schwartz 1982). These two factors indicate that the changing generational mix of Asian Americans, especially given that today three quarters of Asians are immigrants, will give demographic impetus to a retreat from intermarriage with whites among Asian Americans, also seen among Hispanics (Qian, Lichter and Tumin 2018).

Previous studies on Asian Americans’ intermarriage with whites focus on a comparison between foreign-born and U.S.-born Asians (Qian and Lichter 2011). The retreat from intermarriage with whites among the U.S. born does not distinguish between the second and third-plus generations. While third-plus-generation Asians may not be well versed in ethnic culture, second-generation Asians grew up in immigrant families. Arguably, the second generation is the cornerstone in the assimilation process. Second generations often juggle through both worlds, the one of foreign-born parents, their cultures, and social networks versus the one where they interact with their peers, schools, work places, and community organisations (Zhou 2009). It is conceivable that marriage patterns of second-generation Asians, whether they are more likely to marry whites or other Asians (especially first-generation Asians), help understand the assimilation process of Asians in the U.S.

Studying generational differences in intermarriage among Hispanics, Lichter, Carmalt, and Qian (2011) highlight the role of second-generation Hispanics. Second generations span boundaries that often separate the foreign born and the U.S. born. Intermarriage patterns of second-generation Hispanics indicate how quickly or slowly Hispanics integrate in American society (Lichter, Carmalt and Qian 2011). They find that second-generation Hispanics are more likely to marry first-rather than third-plus-generation Hispanics or whites, which suggests a retreat from intermarriage with whites among second-generation Hispanics. They offer two hypotheses that are highly relevant and can also be tested for Asian Americans. One is that enormous growth of immigration reinforces racial identity and increases demographic opportunities for their U.S.-born peers. As a result, U.S.-born Asians, especially the second generation, are increasingly more likely to marry their immigrant counterparts and less likely to marry whites over time. The second hypothesis states that rather than retreating from interracial marriage with whites, second-generation and third-plus-generation Asians may continue to marry whites at the same level as in the past and do not show much increase in marriage with their immigrant peers. Under this scenario, they move culturally and economically toward the mainstream society, and social distance between U.S.-born and foreign-born Asians increases. Third-plus-generation Asians, in particular, are at least one generation removed from ethnic cultures and identities and may share little in common with contemporary Asian immigrants because the countries their ancestors came from have changed dramatically or no longer sent many immigrants to the U.S. (Japan for example). As a result, social distance between first and second generations may be smaller than that between second and third-plus generations. We expect second-generation Asians to have higher rates of marriage with first-generation Asians than with third-plus-generation Asians.

How generation influences intermarriage may also be associated with educational attainment. Although Asians have the highest percent among all racial groups to have earned a college degree, disproportionately more have less than a high school degree (Pew Research Center 2013, Sakamoto, Goyette and Kim 2009). For Asian Americans, educational attainment is positively associated with intermarriage with whites (Qian and Lichter 2011). College-educated Asians, for example, have greater opportunities of meeting whites on campus, in the workplace, and in residential areas. These opportunities separate college-educated Asians culturally and socially away from their less-educated counterparts and reduce social distance with whites. In contrast, U.S.-born Asians with low levels of education lack such opportunities and tend to live and work in segregated ethnic enclaves (Zhou and Logan 1991). They may instead have more opportunities to meet and marry co-ethnic newcomers from Asia in ethnic enclaves. Consequently, second-generation Asians with low levels of education are more likely to marry first-generation Asians and less likely to marry whites than their highly-educated counterparts. It is likely that less-educated second-generation Asians may marry highly-educated first-generation Asians as an exchange for higher educational attainment with ‘higher’ generation status (Merton 1941).

Current study

Selectivity of first-generation Asians with high socioeconomic status increases the likelihood of success in socioeconomic integration and Asian Americans are considered at the vanguard of rising intermarriage in the U.S. (Lee 2015). Previous studies of Asian intermarriage have focused on marriages with non-Hispanic whites (Qian and Lichter 2007, 2011). Little is known about how generational mix of Asian Americans shapes intergenerational marriage and interracial marriage with whites. This paper fills the void. In this paper, we first examine whether a larger marriage pool of Asian immigrants reinforces cultural identities and strengthens intergroup boundaries. We hypothesise that over time there is an increase in intergenerational marriage among Asians and a decline in interracial marriage with whites.

Second, we examine how educational attainment plays a role in intergenerational marriage and interracial marriage. We hypothesise that less-educated Asian Americans are more likely to marry across generations and less likely to marry whites than their highly-educated peers. Third, we examine whether there is retreat from intermarriage among Asian Americans by focusing on the second generation. Given that the second-generation is tied to their immigrant parents and cultures, we hypothesise that second-generation Asians are more likely to marry first-generation Asians than third-plus-generation Asians.

Data and methods

Data for this study come from the Integrated Public Use Microdata Series March Current Population Survey (IPUMS-CPS) for the years 1994–2015 (https://cps.ipums.org/cps/). In 1994, the IPUMS-CPS began including information about each household member’s birthplace and the birthplace of the household members’ parents. This information, coupled with information about household members’ race and ethnicity, allows us to examine detailed racial and ethnic patterns of intermarriage among the first, second, and third-plus generations of Asians over the 1994–2015 period.

Before 2003, the CPS did not distinguish Asians from Hawaiians/Pacific Islanders, whereas from 2003 onwards, Asian only, Hawaiian/Pacific Islander only, and Asian-Hawaiian/Pacific Islander biracial individuals are coded into three different categories. To maintain consistency through the survey years, Asians in this study refer to Asians or Hawaiians/Pacific Islanders. We believe, however, including Hawaiians/Pacific Islanders and Asian-Hawaiian/Pacific Islander biracial individuals does not change our results because they only comprise about 6% of the Asian population based on data from 2003–2015.

We consider the comparative intermarriage patterns of: (1) foreign-born Asians (firstgeneration Asians) who arrived in the U.S. roughly prior to age 20.¹ This will increase the chance that they had not married before immigration and had been exposed to the U.S. marriage market conditions before marriage; (2) native-born Asians with at least one foreign-born parent (second-generation Asians); (3) native-born Asians with two native-born parents (third-plus generation Asians); and (4) non-Hispanic whites (of any generation). Intermarriages between Asians and all other races are excluded due to extremely small sample sizes. To reduce zero cells in contingency tables and highlight the central role of college education in shaping marriage patterns (Cherlin 2010), we dichotomise educational attainment into two categories: no bachelor’s degree and at least a bachelor’s degree.

Our data consist of information on married heads of households who are then linked to similar information on their co-resident spouses. Following Lichter and his colleagues (2011), we restrict the sample to household heads and their spouses who were aged 18–34 at the time of interview. Marriages among young adults are more likely to have been formed recently and thus, this age restriction helps minimise potential selection bias caused by marital disruption and remarriage (Qian and Lichter 2007). Our sample consists of 61,993 heterosexual married couples. To examine changes over time and ensure enough cases for each time period, we divide the data into two time periods: 1994–2004 and 2005–2015.

To examine changes in assortative mating between 1994–2004 and 2005–2015, we fit log-linear models that identify the associations between spouses’ race/generation and educational attainment independent of the marginal distributions of these characteristics. When studying race/generation assortative mating, we employ crossings models (e.g., Mare 1991, Qian and Qian 2014, Schwartz and Mare 2005).

Table 1 presents the crossings parameters in detail. Supposing that intermarriage is a process of crossing barriers of different levels, under crossings models, each barrier is determined by which two adjacent levels it separates. For instance, the barrier between first-and second-generation Asians is νl, the barrier between second-and third-plus-generation Asians is ν2, and the barrier between third-plus-generation Asians and whites is ν3 (Hout 1983). Thus, crossings models can reveal the barrier to intermarriage between two race/generation groups (Mare 1991). Parameters in Table 1 indicate the log odds of intermarriage across two adjacent race/generation groups relative to the log odds of endogamy, controlling for marginal distributions of husband’s and wife’s race/generation. Prospective spouses with a greater distance in race/generation must cross more barriers to get married. In other words, the log odds of marriage for couples across several race/generation boundaries are the sum of the crossings parameters separating husbands’ and wives’ race/generation (Schwartz and Mare 2005).

Table 1.

Parameters for crossings effects on race-generation assortative marriage

	Wives’ race/generation
Husbands’ race/generation	First-generation Asian	Second-generation Asian	Third-plus-generation Asian	White
First-generation Asian	1	ν1	ν1 + ν2	ν1 + ν2 + ν3
Second-generation Asian	ν1	1	ν2	ν2 + ν3 ν3
Third-plus-generation Asian	ν1 + ν2	ν2	1	ν3
White	ν1 + ν2 + ν3	ν2 + ν3	ν3	1

Note: Table is adapted from Schwartz and Mare (2005).

In order to examine gender differences in marriage patterns, we add a race/generation hypergamy parameter in the model to explore gender asymmetries (Mare 1991, Qian 2016). Specifically, we constrain the cells in which Asian husbands are in later generations than their Asian wives or white husbands are married to Asian wives into one parameter (i.e., the bottom half below the main diagonal of the four-by-four contingency table of husbands’ and wives’ race/generation). To measure the tendency for college graduates to marry each other (Schwartz and Mare 2005), we create a variable which indicates that both spouses are college graduates (=1; otherwise = 0). When exploring how race/generation assortative mating varies by spouses’ education, we add interactions between the crossings parameters and the variable indicating that both spouses are college graduates. Furthermore, we add three-way interactions of the crossings parameters, the variable indicating that both spouses are college graduates, and the time period indicator to examine change over time.

Results

We first present descriptive results of the spousal distributions by race/generation and time period for men and women, respectively, in Table 2. In 1994–2004, generational endogamy (i.e., marriage in which two spouses belong to the same generational status group) was most common among first-generation Asian men: 70% of them married first-generation Asian women, followed by third-plus-generation Asians (52%); second-generation Asian men were spread out in their marriage patterns with nearly one quarter married to first-generation Asian women, one third married to second-generation Asian women, and 9% married to third-plus-generation Asian women. Notably, generational endogamy became stronger between 1994–2004 and 2005–2015, from 32% to 49% among second-generation men and from 52% to 69% among third-plus-generation men. Among Asian men over the 1994–2004 period, 23% of the first generation, 35% of the second generation, and 28% of the third-plus generation married whites. Over time, percent marrying whites declined for all three generations, to 19%, 22%, and 22%, respectively. Meanwhile, there was an increase among first-generation Asian men marrying second-generation Asian women, from 5% in 1994–2004 to 15% in 2005–2015, and there was a decline among third-plus-generation Asian men marrying first-generation Asian women, from 13% in 1994–2004 to 3% in 2005–2015.

Table 2.

Percentage distribution of marriages by spouses’ race-generation, by men’s and women’s generation and time period

	Spouses
	First-generation Asian	Second-generation Asian	Third-plus-generation Asian	White	N
Men
1994–2004
First-generation Asian	70.2	5.0	1.9	22.9	423
Second-generation Asian	23.4	32.4	9.0	35.1	111
Third-plus-generation Asian	12.9	7.5	51.7	27.9	147
2005–2015
First-generation Asian	65.6	14.9	0.8	18.7	369
Second-generation Asian	22.8	48.5	7.1	21.6	268
Third-plus-generation Asian	3.4	5.2	69.0	22.4	116
Women
1994–2004
First-generation Asian	62.9	5.5	4.0	27.5	472
Second-generation Asian	14.5	24.8	7.6	53.1	145
Third-plus-generation Asian	5.4	6.8	51.4	36.5	148
2005–2015
First-generation Asian	54.4	13.7	0.9	31.0	445
Second-generation Asian	18.3	43.3	2.0	36.3	300
Third-plus-generation Asian	2.1	13.2	55.6	29.2	144

Now we examine intermarriage patterns from women’s perspectives, as shown in the lower panel of Table 2. As expected, there was a noticeable decline in marriage with whites from 53% in 1994–2004 to 36% in 2005–2015 among second-generation Asian women and from 37% to 29% among third-plus-generation Asian women. One exception is that percent marrying whites actually increased somewhat for first-generation Asian women, from 28% to 31%, possibly due to the improved socioeconomic profile of more recent immigrants. Meanwhile, from women’s perspectives, there was an increase in percent of marriages between first-and second-generation Asians (6% and 14% of first-generation Asian women were married to second-generation men in 1994–2004 and 2005–2015 respectively, and 15% and 18% of second-generation Asian women were married to first-generation men, respectively). There was also an increase in marriages between third-plus-generation Asian women and second-generation men, from 7% in 1994–2004 to 13% in 2005–2015.

The descriptive results reveal a decline in interracial marriage with whites and an increase in intergenerational marriage among both Asian men and women. Yet, these results are confounded by men’s and women’s generational differences and shifts in marginal distributions. We now introduce log-linear models to control for changes and differences in marginal distributions. In addition, we include educational attainment of men and women in the models to examine how couples’ generational statuses interact with their educational pairings. Table 3 reports the goodness-of-fit statistics—the deviance and the Bayesian information criterion (BIC) statistics—for the log-linear models examined in this study. The goal of log-linear modeling is to reveal the association among the variables in consideration of finding a parsimonious model with acceptable goodness of fit, using the Likelihood Ratio Test (L²) and the BIC (Hout 1980). The BIC statistic is equal to L² - (df) log(N), which adjusts L² based on degrees of freedom (df) and sample size (N). A smaller value of BIC indicates a better-fitting model (Raftery 1986). We mainly focus on the BIC statistic due to our large sample size (Gullickson and Torche 2014).

Table 3.

Fit statistics for log-linear models of race-generation and educational assortative marriage

Model		df	Deviance	BIC
1	Marginals (HR + HE + WR + WE + P) + HR×HE + WR×WE	112	29,597.95	28,362.06
2	Model 1 + C	109	21,397.36	20,194.57
3	Model 2 + RHyper	108	21,397.36	20,205.60
4	Model 2 + EduCol	108	1,362.63	170.87
5	Model 4 + HR×P + WR×P	102	1,175.88	50.33
6	Model 5 + C×P	99	1,164.61	72.17
7	Model 5 + HE×P + WE×P	100	328.55	−774.93
8	Model 7 + EduCol×P	99	325.10	−767.34
9	Model 7 + C×EduCol	97	250.70	−819.68
10	Model 9 + C×EduCol×P	90	225.60	−767.53

Notes: N = 61,993; cells = 128. df = degrees of freedom.

HR = husbands’ race/generation; HE = husbands’ education; WR = wives’ race/generation; WE = wives’ education; C = race/generation crossings parameters; RHyper = race/generation hypergamy (marriages in which Asian husbands are in later generations than their Asian wives or white husbands marry Asian wives = 1, otherwise = 0); EduCol = parameter indicating both spouses are college graduates (both spouses are college graduates = 1, otherwise = 0); P = time period (2005–2015 = 1; 1994–2004 = 0).

Model 1 includes only the time period indicator, marginal distributions of men’s and women’s race/generation and education, and associations between race/generation and education of both husbands and wives. In other words, Model 1 assumes no association between husband’s and wife’s attributes. Not surprisingly, the BIC for Model 1 is much larger than zero, indicating a poor model fit.

From Models 2 through 4, we investigate time-invariant patterns of assortative mating on race/generation and education. In Model 2, we examine barriers to race/generation intermarriage by adding the crossings parameters. The great reduction in the BIC statistic, relative to Model 1, indicates that assortative mating on race/generation is far from random. Instead, there are severe barriers to cross before individuals marry spouses from other generational or racial groups. In Model 3, adding a uniform race/generation hypergamy parameter does not further decrease the BIC relative to Model 2, indicating that once we control for marginal distributions of spouses’ race/generation, race/generation assortative mating patterns are symmetric with respect to gender. In other words, the probability of a marriage between two persons from different race/generation groups is unaffected by whether the husband or the wife is white or later-generation Asian. Building on Model 2 (a better-fitting model than Model 3), we examine the tendency for college graduates to marry each other in Model 4. After we add this term, the BIC decreases considerably relative to Model 2, indicating a strong tendency for college graduates to marry each other.

In models 5 through 8, we examine how marginal distributions of spouses’ race/generation and educational attainment as well as assortative mating on race/generation and education have changed over the two time periods. In Model 5, we add interaction terms to investigate whether or not the marginal distributions of husbands’ and wives’ race/generation vary by time period (1994–2004 or 2005–2015). Based on the BIC, Model 5 fits more closely to the data than Model 4, which is not surprising due to the large, sustained influx of Asian immigrants in recent decades. Building on Model 5, further adding the interactions between the crossings parameters and time period (Model 6) does not significantly improve the model fit, as the BIC statistic becomes more positive with the loss of 3 degrees of freedom. This result suggests that percentage changes in marriage we see in Table 2 were largely due to shifts in marginal distributions of spouses’ race/generation; once shifts in marginal distributions are controlled for, the likelihood of forming intermarriage between two race/generation groups relative to forming racial/generation endogamy did not change much between 1994–2004 and 2005–2015. Building on Model 5 (the best-fitting model thus far), we add the interaction terms between spouses’ educational attainment and time period in Model 7. Model 7’s BIC drops substantially relative to Model 5’s BIC, indicating that husbands’ and wives’ educational distributions have changed significantly over the two time periods. Model 8 includes the interaction term between the parameter indicating both couples are college graduates and the parameter for time period. Model 8 does not improve relative to Model 7, indicating that among couples under examination, the tendency for college graduates to marry each other did not change over time after shifts in spouses’ educational attainment are taken into account.

Do patterns of race/generation assortative mating differ by spouses’ educational attainment? Building on Model 7 (the best-fitting model thus far), we include the interactions between the crossings parameters and the parameter indicating both couples are college graduates in Model 9. According to the BIC statistics, Model 9 fit more closely to the data than Model 7, suggesting that education indeed shapes race/generation assortative mating. Has the interaction between educational and race/generation assortative mating changed over time? We include three-way interactions of the crossings parameters, the variable indicating both spouses are college graduates, and the time period indicator in Model 10, but the model is not the most parsimonious. In the end, we focus on the parameter estimates based on the three better-fitting models (Models 7, 9 and 10, all with BICs much smaller than zero) to examine intergenerational marriage and interracial marriage by couples’ educational attainment and time period.

Table 4 presents the parameter estimates based on the three better-fitting models for married couples involving Asians Americans of different generations and whites. Model 7 includes the crossings parameter estimates. All three parameters were negative, suggesting strong endogamy (Asian Americans married within their own generational group and whites married whites). Boundaries were the easiest to cross between first-and second-generation Asians (β = −1.33,p < .001) and hardest to cross between third-plus-generation Asians and whites (β = −3.31, p < .001).

Table 4.

Select parameters from Models 7, 9, and 10, married sample

Select Parameters	Model 7			Model 9			Model 10
	β		Std. Err.	β		Std. Err.	β		Std. Err.
Race/generation crossings
(1)First- / second-generation Asian	−1.33	***	0.09	−1.21	***	0.11	−1.67	***	0.19
(2)Second- / third-plus-generation Asian	−2 .35	***	0.12	−2 .74	***	0.17	−2 .55	***	0.21
(3)Third-plus-generation Asian / white	−3.31	***	0.04	−3.51	***	0.05	−3.46	***	0.07
Both spouses are college graduates	2.67	***	0.02	2.66	***	0.02	2.69	***	0.03
Crossings × Both spouses are college graduates
Crossings(1) × Both spouses are college graduates				−0.33	+	0.19	0.13		0.32
Crossings(2) × Both spouses are college graduates				1.00	***	0.24	1.17	***	0.30
Crossings(3) × Both spouses are college graduates				0.59	***	0.08	0.44	***	0.11
Crossings × Period
Crossings(1) × 2005–2015							0.68	**	0.23
Crossings(2) × 2005–2015							−0.46		0.34
Crossings(3) × 2005–2015							−0.12		0.10
Both spouses are college graduates × Period
Both spouses are college graduates × 2005–2015–0.08							−0.08	+	0.04
Crossings × Both spouses are college graduates × Period
Crossings(1) × Both spouses are college graduates × 2005–2015							−0.70	+	0.37
Crossings(2) × Both spouses are college graduates × 2005–2015							−0.27		0.46
Crossings(3) × Both spouses are college graduates × 2005–2015							0.29	*	0.14

Notes: Std. Err. = Standard Errors.

^***p < 0.001;

^**p < 0.01;

^*p < 0.05;

⁺p < 0.1

In addition, barriers to intergenerational and interracial marriage vary by couples’ educational pairing. Model 9, which includes the interactions between the crossings parameters and the educational pairing parameter, reveals that compared with their less-educated counterparts, Asian American couples who were both college graduates were less likely to form intergenerational marriages between first and second generations, more likely to form intergenerational marriages between second and third-plus generations, and to a less extent, more likely to form marriages between third-plus-generation Asians and whites.

Model 10 includes the three-way interactions between time period, the crossings parameters, and the educational pairing of spouses. For ease of presentation, we graph the predicted odds ratio in Figure 1 based on the parameter estimates of Model 10. Among couples who had both completed college education, the odds of marriage between second-and third-plus-generation Asians declined from .25 in 1994–2004 to .12 in 2005–2015 (p = .03). This decline may be due to increasing opportunities of generationally endogamous marriages among the U.S. born. Meanwhile, there was a small but insignificant increase in interracial marriage with whites (from .05 to .06, p = .14) among the college-educated couples. In contrast, among couples at least one of whom had not completed college education, the odds of intergenerational marriage between first and second generations doubled, from .19 in 1994–2004 to .37 in 2005–2015 (p = .003). This is accompanied by a small and insignificant decline in intergenerational marriage between second and third-plus generations, from .08 to .05 (p = .18). The results suggest that second-generation Asians move closer to their first-generation Asians in partnering, possibly due to increasing contact opportunities between first-and second-generation Asians. Yet, the beneficiaries were selective of second-generation Asians without bachelor’s degrees. Second-generation Asians’ integration in American society very much depends on their educational attainment. Highly-educated individuals are far more likely to work in integrated workplaces, live in integrated neighbourhoods, and have greater opportunities for intergroup contact than their less-educated counterparts (Zhou and Logan 1991).

Figure 1.

Odds of intergenerational marriage or intermarriage with whites, relative to odds of race-generation endogamy, by time period and educational pairing

The results discussed above are limited to married couples. An emerging living arrangement is cohabitation. Cohabitation is short lived, but it has played an important role in the decline of marriage rates (Bumpass, Sweet, and Cherlin 1991). Cohabitation is more likely than marriage to involve couples who are interracial or exogamous (Blackwell and Lichter 2000). To examine how cohabitation may influence intergenerational or interracial marriages among Asian Americans, we replicate the analyses by including cohabiting couples in the models. Table 5 presents the results for the expanded sample (married and cohabiting couples). For Model 7, the crossings parameter between the first-and second-generation Asians was −1.28, slightly less negative than −1.33 as in Table 4 when only the married sample was considered. The other two crossings parameter estimates moved to similar positions (−2.25 versus −2.35 and −3.21 versus −3.31, respectively). These results along with those from the other two models confirm that cohabitation was more likely than marriage to be intergenerational or interracial. Overall, the conclusion remains the same when we consider both marital and cohabiting unions.

Table 5.

Select parameters from Models 7, 9, and 10, including all unions (marriage and cohabitation)

Select Parameters	Model 7			Model 9			Model 10
	β		Std. Err.	β		Std. Err.	β		Std. Err.
Race/generation crossings
(1)First- / second-generation Asian	−1.28	***	0.08	−1.19	***	0.09	−1.56	***	0.17
(2)Second- / third-plus-generation Asian	−2.25	***	0.10	−2.57	***	0.14	−2.45	***	0.19
(3)Third-plus-generation Asian / white	−3.21	***	0.03	−3.39	***	0.04	−3.33	***	0.06
Both spouses are college graduates	2.69	***	0.02	2.68	***	0.02	2.70	***	0.03
Crossings × Both spouses are college graduates
Crossings(1) × Both spouses are college graduates				−0.29	+	0.17	0.05		0.29
Crossings(2) × Both spouses are college graduates				0.93	***	0.21	1.13	***	0.28
Crossings(3) × Both spouses are college graduates				0.56	***	0.07	0.39	***	0.10
Crossings × Period
Crossings(1) × 2005–2015							0.68	**	0.23
Crossings(2) × 2005–2015							−0.23		0.27
Crossings(3) × 2005–2015							−0.12		0.08
Both spouses are college graduates × Period
Both spouses are college graduates × 2005–2015–0.08							−0.04		0.04
Crossings × Both spouses are college graduates × Period
Crossings(1) × Both spouses are college graduates × 2005–2015							−0.50		0.34
Crossings(2) × Both spouses are college graduates × 2005–2015							−0.35		0.38
Crossings(3) × Both spouses are college graduates × 2005–2015							0.30	*	0.12

Notes: Std. Err. = Standard Errors.

^***p < 0.001;

^**p < 0.01;

^*p < 0.05;

⁺p < 0.1

Discussion and conclusion

For Asian Americans, intermarriage rates with whites have declined in recent decades (Qian and Lichter 2007). This decline is in large part due to the continuous influx of immigrants originating from Asia. The rapid growth of the Asian population in the U.S. has increased the number of Asian ethnic neighbourhoods, stores, restaurants, and religious institutions, which has strengthened and heightened Asian ethnic identities. The growing population and reinforced cultural boundaries may redefine marriage markets where Asian Americans have more opportunities to meet, date, and marry among themselves. This raises the questions about how Asian Americans integrate in the U.S. and how increases of first-generation Asians influence intermarriage patterns of second-and third-plus-generation Asians.

Applying new assimilation theory (Alba and Nee 2003), we posit that Asian Americans’ integration does not follow a single path. Specifically, growing Asian immigrants and changing generation mix of Asian Americans may lead to different intermarriage patterns among Asian Americans, depending on generation and socioeconomic position. Following Lichter, Carmalt, and Qian (2011) and using data from the March Current Population Survey for the years 1994–2015, we focus on generational differences in intergenerational marriage among Asian Americans and interracial marriage with whites. Descriptive statistics shows that marriage rates with whites declined for Asians of all generations (with the exception of first-generation Asian women) and marriage rates between first-and second-generation Asians increased between 1994–2004 and 2005–2015 (with the exception of second-generation Asian men who married first-generation Asian women).

Our statistical analyses yield several important findings. First, generational endogamy and white endogamy are much stronger than intergenerational and interracial marriages. Barriers between second-and third-plus-generation Asians are lower among couples both with college education than among couples with at least one non-college educated spouse. This suggests divergent paths of integration between highly-educated and less-educated second-generation Asians. Highly-educated second-and third-plus-generation Asians may have greater contact opportunities and thus a higher likelihood of intergenerational marriage than their less educated counterparts.

Second, over time, there is no strong retreat from intermarriage with whites among third-plus-generation Asians. This suggests that third-plus-generation Asians are assimilating, via intermarriage with whites, albeit not at a faster pace. They appear to have moved away from their cultural and immigrant roots due to their residential and workplace integration in the U.S. (Hirschman and Wong 1981, Hirschman and Wong 1986, Xie and Goyette 2004).

Third, differences in intergenerational and interracial marriages between second-and third-plus-generation Asians have grown stronger over time. The retreat from intermarriage with whites for the second generation is evident, as odds of intergenerational marriage between first-and second-generation Asians have increased especially among couples with at least one non-college graduated spouse. The population growth of first-generation Asians and their positive selectivity in terms of socioeconomic status have increased contact opportunities and narrowed social distance between the two generations (Lee 2015). The shared immigrant roots and national origins may have reinforced ethnic identities and fostered marriages across first and second generations (Qian, Blair and Ruf 2001, Shinagawa and Pang 1996).

The annual March demographic supplements of the Current Population Survey provide a great source of data for understanding generational differences in intergenerational marriage among Asians and interracial marriage with whites. These data, however, are not without limitations. We have pooled twenty-two years of CPS supplements from 1994 to 2015, yet the sample size of Asians is still small. A more severe limitation is that we are unable to further classify Asians into different ethnic or national origin groups (Chinese, Filipinos, or Asian Indians, for example). This limitation could confound findings as residential patterns may vary by national origin. Furthermore, ethnic or national origin compositions may vary by generation due to different immigration histories over time. For example, first-generation Asians consist of more national origin groups than ancestors of third-plus-generation Asians. The compositional differences in national origin may play a role in lowering levels of intergenerational marriage between the two generations.

Intermarriage helps understand paths of integration in American society. Patterns of intergenerational marriage and generational differences in interracial marriage among Asian Americans shed light on how generational and racial boundaries are shaped and reinforced. Third-plus-generation Asians have been integrated or assimilated as measured by interracial marriage with whites, despite a rapid increase in the share of the Asian population. Second-generation Asians on the other hand experienced a pause and had much greater likelihoods of marrying their first-generation counterparts, especially among those with less education. In the end, education has become a strong divider and leads to diverse paths of integration.