Introduction

The population of Asian Americans has been growing in recent decades both in terms of absolute numbers as well as a percentage of the total U.S. population (Ren et al., 2022). As of 2020, the Asian American population was over 24 million or about 7.2% of the total U.S. population (Jones et al., 2022). By comparison, the African American population was about 14.2% of the total U.S. population in 2020 (Jones et al., 2022) so that Asian Americans are now about half of the population size of the African Americans. Yet relative to the abundance of research on other minority groups that is published every year, the Asian American population has been relatively understudied. Its essential demographic characteristics are poorly understood despite Asian Americans having become a widely recognized and highly visible group throughout the contemporary U.S. (Ren et al., 2022).

Prior social scientific research on the socioeconomic characteristics of South Asian Americans is particularly limited (Tiagi, 2013). Numbering over 5 million (SAALT 2019), South Asian Americans have become highly visible in leadership positions such as the vice president of the U.S. (i.e., Kamala Harris) and recent CEO’s of many influential corporations (e.g., Adobe, Citibank, Google, IBM, Microsoft, PepsiCo, Twitter) as well as in other areas in American society including in academics, arts, entertainment, media, medicine, military, science, technology, and sports.Footnote 1 Although descriptive average incomes are sometimes noted (Marger, 2014; Chakravorty et al., 2017; Iceland, 2017; Budiman & Ruiz, 2021), multivariate analyses for South Asian Americans are lacking despite being essential for understanding the net differentials that are more informative about racial discrimination (Hirschman & Wong, 1984; Iceland, 2019; Wang et al., 2017). The dearth of multivariate analyses of socioeconomic data for such an obvious group as South Asian Americans is a curious lacuna in the literature especially given the great emphasis on being “inclusive” in regard to the study of racial inequality (Bhopal, 2017).

Our study investigates education, wages, household income, poverty, and affluence among South Asians residing in the U.S. Adequate sample sizes by generational status permit us to separately consider Bangladeshis, Indians, Nepalese, Pakistanis, and Sri Lankans.Footnote 2 Although for context some descriptive statistics are provided for first-generation immigrant South Asians, our focus is on second-generation (and the closely related 1.5-generation) South Asian Americans. Their socioeconomic attainments primarily reflect their opportunities in the U.S. in contrast first-generation immigrants whose stratification outcomes are heavily influenced by their countries of origin (Zeng and Xie, 2004; Sakamoto et al., 2012). Our investigation includes not only Indians who are by far the largest demographic group, but also the first ever multivariate analyses of socioeconomic data for Bangladeshis, Nepalese, Pakistanis, and Sri Lankans.

Although our data identify nativity status, we cannot formally distinguish between second-generation and third-and-higher generation persons. However, when studying adult South Asian Americans (i.e., persons at least 25 years of age) using data from 2014 to 2018, we can reasonably assume the great majority of the native-born are second-generation. The South Asian immigration is primarily post-1965 with most immigrants arriving in the U.S. since the 1990’s (e.g., Chakravorty et al., 2017:34). In addition, our statistics are limited to single-race persons which excludes many (if not most) third-generation persons who have at least one grandparent who was born in India (Duncan & Trejo, 2017).

Immigrants from India did come to the U.S. before the restrictive Immigration Act of 1924, but they were only about 8,000 (Min, 2006). They were mostly male, low-skilled laborers and small-scale farmers in the west coast who furthermore struggled with miscegenation laws as well as other discriminatory anti-Asian legislation that was more widespread in California during that era (Kitano & Daniels, 1995). After immigration from Asia was terminated by the Immigration Act of 1924, and after the Thind decision in 1923 ruled that the alien land laws applied to Indians, this early South Asian community basically ceased to socially reproduce itself (Kitano & Daniels, 1995; Rahma & Paik, 2017).

Regarding the sociohistorical context of these migration streams, all of these South Asian countries (with the exception of Nepal) remained under British colonial rule until about 1947. Moreover, Pakistan, India and Bangladesh were a single country until 1947 while Pakistan and Bangladesh remained a single country until 1971. Pakistan and Bangladesh are Muslim dominated countries, whereas in India and Nepal, the primary religion is Hinduism. In Sri Lanka 70% of the population are Buddhists. Nonetheless, the U.S. Census of 1930 and 1940 referred to South Asians as the “Hindu” race in order to differentiate them from Native Americans (Lee, 1993) although the majority of these early immigrants to the U.S. were actually Sikhs from Punjab (Kitano & Daniels, 1995).

Are South Asian Americans Penalized by “Whiteness Privilege” in the Labor Market?

An enduring theme in studies of Asian Americans is the over-education hypothesis which states that “Asian Americans approach socioeconomic parity with whites because of their overachievement in educational attainment” (Hirschman & Wong, 1984:584). Hirschman and Wong argued that the average earnings of native-born Asian Americans did not differ very much from those of non-Hispanic Whites due to Asian Americans having a higher average level of educational attainment. The labor market was therefore discriminating against Asian Americans in that they needed to make a higher investment in human capital in order to obtain the same earnings as non-Hispanic Whites. As stated by Hirschman and Wong (1984:602), “The apparent equality between Asians and whites is largely a function of educational overachievement by Asians. If Asians experienced the same process of stratification as whites, their educational credentials would shift their (Asians’) occupational and earnings levels substantially above those of the majority population.”

As more recently reiterated by Tran et al. (2019a:2253) in their study of occupational attainment, “despite Asian Americans’ out-performance of white Americans in education, Asian Americans have been less successful in translating these gains into the labor market. For example, researchers have found that Asian Americans were systematically overeducated relative to their labor market attainment, implying that some degree of ethnic or racial discrimination still persists….” Lee and Kye (2016:256) conclude that “significant earnings deficits” and “significant earnings disparities” continue to penalize many native-born Asian American groups after taking into account educational attainment. The over-education hypothesis underscores the need for multivariate analysis―as was noted above―when assessing discrimination against minorities in the labor market (Tran et al., 2019a).

Tran et al. (2019a:2262) furthermore state that, “despite their remarkable educational gains, there is no discernible advantage in the labor market for the non-Chinese second-generation Asian groups compared to whites. Put differently, Asian ethnic groups are over-credentialed with regards to education to achieve parity with whites in the labor market. This is consistent with qualitative research that documents a ‘bamboo ceiling’ among the Asian second-generation in the workplace….” Tran, Lee and HuanTran et al. (2019b:2273) conclude that the empirical veracity of the over-education view is considered “non-controversial.”

However, the over-education hypothesis needs to be more specifically investigated for South Asian Americans rather than considering Asian Americans as an undifferentiated racial category. In contrast to the older study by Hirschman and Wong (1984), the more recent study by Tran et al. (2019a) does identify second-generation Indian Americans. Tran et al. (2019a) interpret their results as being definitively supportive of the over-education hypothesis for second-generation Indian Americans. Summarizing their empirical findings succinctly for a more general audience, Jennifer Lee (who is one of the authors of Tran et al., 2019a]) states that “we found that Indian-Americans are no more likely to attain a professional job than white Americans, despite being eight times as likely to graduate from college….” (Craig, 2019).

Why are South Asian Americans Ignored in the Debate about “Colorism”?

Consistent with the supposedly “non-controversial” over-education view in studies of Asian Americans (Tran et al., 2019b:2273), a burgeoning literature argues for the causal effects of skin tone in social stratification (e.g., Bonilla-Silva, 2004; Hersch, 2011; Monk, 2015; McDonald & Thompson, 2016; Saenz & Morales, 2019; Abascal & Garcia, 2022; DeAngelis et al., 2022). Echoing prior conclusions by Bonilla-Silva (2004:944), Monk (2021:86) states that “skin tone stratification appears to be quite pervasive in the United States, so much so that it seems quite fair to label it a pigmentocracy….” Monk (2021) states that not only African Americans and Hispanics, but also Asian Americans “are all significantly stratified by skin tone” (Monk, 2021:86). This conventional view about “White privilege” assumes that all non-White minority groups face labor market discrimination which in the case of Asian Americans most clearly takes the form of over-education (Tran et al., 2019a).

A review of demographic studies of racial and ethnic inequality concludes that skin tone and “Whiteness privilege” are important sources of social stratification because “White persons gain privileges because of structural arrangements benefitting them” (Saenz & Morales, 2019:169). Similarly, according to DeAngelis et al. (2022:1796), “Colorism [i.e., discrimination by skin tone] also established and perpetuated White dominance by linking dark skin with danger, savagery, and incompetence—deeming dark-skin persons uncivil and unworthy of freedom—while linking light skin with moral virtue, civility, beauty, and intelligence—deeming light-skin persons as innately entitled to socioeconomic resources and privileges.” In a recent presidential address to the Population Association of America, systemic racism is identified as the key factor underlying “the large racial and ethnic disparities in sickness, hospitalization, and death” as emblematically illustrated by “the 2020 murder of George Floyd” (Hummer, 2023:633).

However, most of the above-mentioned studies focus on third-and-higher generation African Americans (e.g., Monk, 2015), and none of them consider any multivariate analyses of Asian Americans. While engaging in the analysis of skin tone stratification is beyond the scope of our study (e.g., we do not have any individual-level data on skin tone), this area of research should consider South Asian Americans as an additional case. While skin tone certainly varies across South Asian individuals, on average it tends to be darker than among persons of European ancestry as well as perhaps some other Asian groups (e.g., Stokowski et al., 2007; Jonnalagadda et al., 2019). Conclusions about “colorism” and how “skin tone stratification appears to be quite pervasive in the United States” would be more generalizable if they were not characterized by an inherent multicollinearity problem that derives from focusing on third-and-higher generation African Americans (e.g., DeAngelis et al., 2022) which is an increasingly specialized minority group in 21st century America. This literature cannot clearly ascertain the extent to which skin tone has significant effects for other groups including even second-generation African American groups some of whom have higher socioeconomic attainments than non-Hispanic Whites (Sakamoto et al., 2021). Although Tran et al. (2019a) do not explicitly raise the issue of skin tone per se, “White privilege” as implicitly construed by the over-education view needs further investigation for South Asian Americans who may be especially underpaid if they are subject to discrimination by both Asian minority status as well as skin tone.

Data and Methods

Data

We investigate data from the 2014–2018 American Community Survey (ACS). This five-year sample provides enough cases for some of the demographically smaller South Asian American groups. To our knowledge, the socioeconomic data for some of these Asian American groups―such as Bangladeshi, Nepalese, Pakistani, and Sri Lankans―have never before been studied. We restrict our analysis to single-race persons within the poverty universe of the ACS. We therefore exclude multi-racial persons and unrelated individuals under the age of 15 as well as people in institutionalized group quarters.

A limitation of the ACS is that it does not provide any information about parental place of birth. The ACS does include a variable about the respondent’s place of birth so we can clearly distinguish the native-born versus the foreign-born. Without parental place of birth, however, we cannot differentiate the third-and-higher generation from the second-generation among the native-born. Nonetheless, as noted above, based on the immigration histories of South Asian Americans, the majority of single-race South Asian Americans are likely second-generation in contrast to Whites and African Americans (Min, 2006).

Dependent Variables

Socioeconomic status is a broad analytical concept that is multi-dimensional. It cannot be fully measured by any one variable. We therefore investigate several different outcomes that are well-known due to their intrinsic importance. They have been widely considered in prior research in demographic and other social scientific literature. We extend the demographic literature by investigating these socioeconomic outcome variables for South Asian Americans.

First of all, highest educational level completed is widely recognized not only for its intrinsic value (Kingston et al., 2003), but also for its extrinsic rewards including lifetime earnings (Tamborini et al., 2015) which might even be becoming more larger in recent years (Kim et al., 2018). Following common practice, we use five dichotomous variables including: less than high school; high school; associate’s degree; and bachelor’s or graduate degree. In analyses using education, we limit the data to persons who are at least 25 years of age by which time most individuals with a bachelor’s degree will have completed it.

Poverty status is another dichotomous variable that is provided in the ACS data. This is based on the official poverty measure as defined by the U.S. Bureau of the Census. A person is considered poor if she lives in a household or family that below the official poverty line as established by the U.S. Census Bureau. The poverty line is characterized by different income thresholds that vary by family size and composition which differentiate the basic economic needs of a household.

Another dichotomous variable that we investigate is affluence which indicates whether the person lives in a household or family that is at least five times above her poverty line (i.e., based on family size and composition). This measure was originally proposed by Farley (1996). Although affluence has not been as widely studied as poverty, recent research on racial inequality suggests its rising importance as household income inequality continues to increase in recent years (Iceland, 2019).

Household income is another commonly studied measure. It is still a substantively significant indicator of socioeconomic circumstances even though it does not control for household size or composition. This continuous variable is defined for each individual as the total annual income of the household in which she resides. Additionally, since we are using data from 5 years of the ACS from 2014 to 2018, we have adjusted household income for inflation using the Consumer Price Index (CPI); all incomes have been converted to the 2018 price level. When used as a dependent variable in a multiple regression, household income is logged to reduce its positive skew. In this case, the tiny proportion of household incomes with values of zero or less (possibly due to self-employment business losses) were reassigned a value of 1 in order to permit logging (Sakamoto et al., 2021).

The last dependent variable of interest in the hourly wage rate which is determined by dividing the total annual earnings by the total hours worked last year. The total hours worked last week is estimated by multiplying “usual hours worked per week” by the number of weeks worked last year by the individual. For the small proportion of cases where the hourly wage is less than $1.00, we recoded them to be equal to $1 which is thus the minimum possible score for the hourly wage. Due to the inherent measurement involved in estimating total hours worked last year as well as the possibility of low earnings due to unincorporated self-employment, retaining low values of the hourly wage is preferable to truncation on the dependent variable which can lead to bias (Sakamoto et al., 2021).

When used in regression analysis, the hourly wage is logged in order to account for the positive skew in its distribution. Those regression analyses are limited to persons between the ages of 25 to 64 who were not enrolled in school and were employed in the labor force in the previous year at least 25 h per week for at least 27 weeks (i.e., at least 675 h). These stipulations remove from the sample persons who do not have a clear attachment to the labor force or who may be retiring from it.

Independent Variables

Various control variables are used in the regression models. Given our research objective, our study certainly requires indicators of race and ethnicity. Using the ACS data, we restricted the investigation to non-Hispanic, single-race persons, and created dichotomous variables to identify the following Asian American groups: Bangladeshi; Indian; Nepalese; Pakistani; Sri Lankan; East Asian; and Southeast Asian. For comparative purposes, East Asian refers to the category including Chinese, Japanese, Korean, Mongolian, and Taiwanese while Southeast Asian refers to the category including Burmese, Cambodian, Filipino, Hmong, Indonesian, Laotian, Malaysian, Thai, Vietnamese, and Other Asians not elsewhere classified. An additional reference group that we consider is non-Hispanic Whites as is customary in the literature.

Generational statuses are additional dichotomous variables. They are determined on the basis of the person’s place of birth and age at immigration if foreign born. In keeping with prior research, the 1.5-generation is defined as individuals who were not born in the United States and who were 12 years of age or younger when they immigrated to the U.S. (Rumbaut, 2004). The first-generation is defined as individuals who were not born in the United States and who were older than 12 years of age when they immigrated to the U.S. Native-born refers to persons who were born in the U.S.

Age and age-squared are also used as control variables in the analysis. Two household level variables that are utilized include family size and the household type. Family sizes refer to the number of persons living in the household. Household type consists of three categories: married couple; female headed household (with or without children); and other households. Regional control variables refer to the 9 U.S. Census divisions (i.e., New England, Middle Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain and Pacific division).

Statistical Models

Being measured as an ordinal variable, educational attainment as an outcome is investigated with ordered logistic regression. Being measured as dichotomous variables, poverty and affluence are studied using logistic regression for binary outcomes. Household income and the hourly wage are continuous variables so OLS regression models are used to estimate net racial and ethnic effects on those outcomes. Sample survey weights were utilized in our statistical analyses although their usage does not affect any of our findings substantively.

Empirical Results

Descriptive statistics are shown in Table 1. Although the first-generation is not used in the multivariate analyses, descriptive statistics for that group are nonetheless shown in Table 1 for descriptive purposes. Most of the sample sizes for the South Asian groups seem adequate for the estimation of the main effect of a dichotomous independent variable in multivariate analysis. For example, among native-born Indians, native-born Bangladeshi, and native-born Pakistani, Table 1 shows that the respective sample sizes are 10,785, 203, and 1,268 for persons aged 25 and older. The two groups with somewhat limited sample sizes are native-born Nepalese (i.e., 77) and native-born Sri Lankans (i.e., 112).

Table 1 Socioeconomic outcomes by race, ethnicity, and generation, American community survey 2014–2018
Full size table

In regard to educational attainment, Table 1 indicates that most of the 1.5-generation and native-born Asian groups are more likely to have a bachelor’s degree or a graduate degree than native-born non-Hispanic Whites (hereafter simply “Whites”). For example, among native-born Indians, the percentage with a bachelor’s degree is 34 while for Whites the percentage is 22. Although compared to Whites, the point estimate for a bachelor’s degree is substantively higher for 1.5-generation Nepalese, this differential is not statistically significant at the 0.05 level for this group (which as mentioned above, has a smaller sample size).

The higher educational attainment of native-born South Asian Americans is perhaps not too surprising because, for each of these groups, their first-generation ethnic counterparts are all more likely to have a college degree. While Table 1 does not utilize the “immigrant generation cohort method” per se (Park & Myers, 2010), and while many Asians from less educated families are able to attend college in the U.S. (Sakamoto & Wang, 2021), the descriptive statistics for first-generation do suggest that a substantial component of recent Asian immigrants are highly educated as has been emphasized elsewhere (Zhou & Lee, 2017). Even in the case of first-generation Nepalese, their percentage with a college degree is slightly higher than for Whites.

However, the simple version of the so-called “hyper-selectivity” argument of Zhou and Lee (2017) is problematic for various reasons (Sakamoto & Wang, 2021; Kim & Kim, 2023). For example, as shown in Table 1, first-generation Asian groups are more likely to have never finished high school than Whites especially for Bangladeshi, Nepalese, Pakistani, East Asians, and Southeast Asians. Rather than viewing Asian immigrants as being uniformly very highly educated, the results in Table 1 indicate that most of these first-generation Asian groups are more bifurcated in their educational distributions compared to Whites.

Probably for this reason, Table 1 shows that most first-generation Asian groups (Bangladeshi, Nepalese, Pakistani, and East Asians) have higher poverty rates than Whites. Prior research suggests that poverty among recent Asian immigrants tends to decline after having been in the U.S. for a few years (Lee, 2014; Takei & Sakamoto, 2011) with significant earnings growth in the American labor market for two decades after arrival (Villarreal & Tamborini, 2018). Previous studies furthermore show that Asian immigrant groups with a high level of poverty are generally characterized by significantly reduced poverty rates among their native-born (Takei & Sakamoto, 2011; Sakamoto et al., 2022) as is also evident in Table 1 for the case of South Asian Americans.

Table 2 shows the results for ordered logistic regression models predicting highest level educational completed among 1.5-generation and native-born persons. Model 2 controls for age and gender. Given the results from Tables 1 and 2 also not surprisingly reports positive effects for all of these Asian groups relative to Whites. These effects are statistically significant except for 1.5-generation and native-born Nepalese.

Table 2 Ordered logistic regression models of educational attainment, American community survey 2014–2018
Full size table

Table 3 shows the results for four nested models of logistic regression predicting poverty status among 1.5-generation and native-born persons. The independent variables have been used in prior research on poverty (e.g., Iceland, 2019). Model 1 controls for ethnicity and gender. Model 2 adds age and educational attainment. Model 3 adds family size and type. Model 4 adds region of residence. While not justified in terms of any well-articulated theory, these descriptive models are motivated by the view that the simpler specifications include the progressively “most exogenous” independent variables relative to annual earnings (Sakamoto et al., 2021). This approach provides the readers with a fuller range of relevant empirical evidence.

Table 3 Logistic regression models of poverty status, American community survey 2014–2018
Full size table

Some of the net effects are slightly negative and statistically significant across the models for 1.5-generation and native-born Southeast Asians in Table 3, but substantively these effects are rather small. Statistically significant, net effects for native-born Indians are also discernable but again these effects are not substantively significant. The only two ethnic groups that are notable are 1.5-generation and native-born Bangladeshi and Pakistani. These groups have higher likelihoods of poverty but mostly due to the family size and composition variables. These two ethnic groups have high levels of poverty among the first-generation as is evident in Table 1. We speculate that the higher likelihood of poverty among 1.5-generation and native-born Bangladeshi and Pakistani may be associated with their caring for lower-income elderly parents and relatives.

Table 4 shows the results for these same four nested models of logistic regression among 1.5-generation and native-born persons, but in this case the dependent variable is affluence. The groups that are consistently most likely to be affluent across the models are Indians and Sri Lankans. Although the effects are not quite as large, East Asians, Southeast Asians, and 1.5-generation Whites also are more likely to be affluent than Whites. Contrasting the results for Model 2 versus Model 3 in Table 4, family size and type do not seem to play as important a role in affecting affluence as they do for poverty (i.e., in comparison to the contrast between Model 2 and Model 4 in Table 3).

Table 4 Logistic regression models of affluence, American community survey 2014–2018
Full size table

Table 5 shows the results for these same four nested models for 1.5-generation and native-born persons, but in this case the dependent variable is household income, and the equations are multiple regressions estimated by OLS. After controlling age, education and gender in Model 2 (i.e., those independent variables that are generally prior to annual household income), none of the ethnic groups have lower household incomes compared to Whites. To the contrary, the results for Model 2 in Table 5 show household income advantages to 1.5-generation and native-born Indians, 1.5-generation and native-born Pakistani, 1.5-generation and native-born East Asians, 1.5-generation and native-born Southeast Asians, and 1.5-generation Whites. After controlling for family size and type in Model 3, some ethnic disadvantages do become evident for 1.5-generation Bangladeshi, and 1.5-generation and native-born Pakistani, but slight household income advantages persist for 1.5-generation and native-born Indians, 1.5-generation and native-born East Asians, 1.5-generation and native-born Southeast Asians, and 1.5-generation Whites. Family size and type seem to have bigger impacts on household income in the case of 1.5-generation and native-born Bangladeshi and Pakistani as was mentioned above in regard to poverty.

Table 5 OLS regression models of household income, American community survey 2014–2018
Full size table

These same four nested models were also estimated for 1.5-generation and native-born persons but using the logged hourly wage as the dependent variable. As is commonly done in labor market research, the models are estimated separately for men and women due to significant gender interactions (e.g., Xie & Goyette, 2004; Wang et al., 2017). Table 6 shows the results for men while Table 7 shows the results for women. The only change in the independent variables is that gender is of course no longer included in these models.

Table 6 OLS regression models of hourly wage among men, American community survey 2014–2018
Full size table
Table 7 OLS regression models of hourly wage among women, American community survey 2014–2018
Full size table

After controlling for age and education in Model 2, none of the ethnic coefficients are negative and statistically significant for men in Table 6. In contrast to the predictions of the over-education view discussed above, these results do not indicate systemic labor market disadvantage against any of these Asian ethnic groups due to “White privilege.” To the contrary, they are actually more likely to be slightly advantaged over Whites with the same age and education. That is, the results for Model 2 in Table 6 indicate that 1.5-generation and native-born Indian men earn about a 17% higher wage while 1.5-generation and native-born East Asian men earn about a 8% higher wage compared to White men with the same age and education. 1.5- generation Pakistani men are advantaged by about 8% while native-born Pakistani men are advantaged by about 7% compared to White men with the same age and education. These slight positive effects may derive from Asian American men being more likely than White men to have degrees in science, technology, engineering and math (i.e., STEM) which has been shown in prior research to be relevant for Asian American wages (Kim & Sakamoto, 2010; Wang et al., 2017).

Most of these estimates of the ethnic effects for men remain about the same in Model 3 after controlling for family size and type as is evident in Table 6. The major change is that, in contrast to Model 2, the positive effects for 1.5-generation and native-born Pakistanis are no longer statistically significant at the 0.05 level in Model 3. Nonetheless, none of these estimates are negative and statistically significant, and thus these results do not alter the basic generalization mentioned above that there is no evidence for systemic labor market disadvantage against any of these Asian ethnic groups. The estimates for Model 3 do show a highly negative effect of being in a female-headed household, but this coefficient likely pertains to a selective group of younger men who continue to reside in their mother’s residence.

Model 4 in Table 6 adds in the controls for region of residence. The positive advantages for Indian men continue to be statistically significant. For the other South Asian groups, none of the coefficients are statistically significant at any conventional level. These findings for Model 4 are consistent with the conclusion that the over-education view does not apply to the wages of 1.5-generation and native-born South Asian men.

The estimates for Model 4 in Table 6 do indicate a negative effect of 8% that is statistically significant at the 0.05 level for both 1.5-generation and native-born Southeast Asian men. This net disadvantage is much less evident in Model 2 and Model 3 and is only evident after controlling for region. The latter variable, however, may possibly be endogenous since many of these ethnic groups are more likely to reside in the Pacific (Sakamoto et al., 2009; Kim & Sakamoto, 2010). State-level differences in the cost of living are of course associated with regional differences in the cost of living (Takei & Sakamoto, 2012). Takei et al. (2012) find evidence for the endogeneity of living in the Pacific for 1.5-generation and native-born Asian men. Consistent with that view, Wang et al. (2017) conclude that state-level differences in the cost of living are endogenous for native-born Asian men. Model 4 is shown for descriptive purposes in order to provide readers with the full range of relevant empirical findings. Nonetheless, our own opinion based on prior research is that the estimates of Model 2 and Model 3 are more accurate because region of residence is unlikely to be an exogenous independent variable for 1.5-generation and native-born Southeast Asian men.

In contrast to Southeast Asian men, the estimates for South Asian men do not substantively change between Model 3 and Model 4. The estimates in Model 4 do not have any negative and statistically significant coefficients for any of the South Asian groups. To the contrary, 1.5-generation and native-born Indian men continue to be advantaged over White men in Model 4. In regard to testing the over-education view, including region as a control variable does not fundamentally alter the results about the net effects for any of the South Asian groups of men.

The slight negative effect for Southeast Asian men is not evident for South Asian men perhaps because the latter groups are much less likely than the former group to reside in the Pacific region. In our data, for example, 56% of East Asian and Southeast Asians live in the Pacific compared to 25% for Indians, 9% for Bangladeshi, 17% for Pakistani, and 11% for Whites. A more systematic analysis of state-level differences in the cost of living finds that they are lower for second-generation Indian men compared to second-generation Chinese, Filipino, Korean and Japanese men (Takei & Sakamoto, 2012). These results are consistent with the interpretation that the endogeneity of living in the Pacific is less of an issue for 1.5-generation and native-born South Asian men than for East Asian and Southeast Asian men.

Table 7 shows the estimates for these same four nested regressions of the logged hourly wage for women. None of the results in Table 7 support the over-education view for women because none of the estimates of the group effects are negative and statistically significant. To the contrary, the results for Model 3 in Table 7 show that 1.5-generation and native-born Indian women are advantaged over White women by about 26%.Footnote 3 1.5-generation Pakistani women are advantaged by about 13%. Native-born Pakistani women are advantaged by about 10%. 1.5-generation and native-born East Asian women are advantaged by about 23% while 1.5-generation and native-born Southeast Asian women are advantaged over White women by about 17%.

For the South Asian groups, these estimates are only slightly reduced after controlling for region in Model 4 in Table 7. For East Asian and Southeast Asian women, controlling for region in Model 4 reduces the coefficients somewhat more but even in these cases, their advantages over White women are still statistically significant. These results in Table 7 are consistent with the discussion of Sakamoto et al. (2012) who argued that the evidence does not support the over-education view for native-born Asian women.

Discussion and Conclusions

In percentage terms, Asian Americans are the fastest growing racial category (Rama, 2012; Budiman, 2020). Yet prior demographic research has unfortunately not adequately investigated this large, expanding, and diverse minority group. Socioeconomic characteristics are of critical importance for understanding racial and ethnic inequalities which have been widely studied for other minorities, but previous sociological research in this area has somehow largely neglected South Asian Americans.

Our results indicate that, relative to native-born Whites, all of these Asian groups (i.e., South Asians, East Asians, and Southeast Asians) have higher proportions of first-generation adults who do not have a high school degree. That is, the educational attainments of first-generation Asians are more bifurcated in comparison to Whites especially for first-generation Bangladeshi, Nepalese, Pakistani, East Asians, and Southeast Asians. For these latter groups, their proportions without a high school degree is two or three times greater than for native-born Whites. These results caution against over-emphasizing the so-called “hyper-selectivity” narrative that basically views all Asian immigrants as being college educated (Lee & Zhou, 2015). Even in the cases of first-generation Indians and Sri Lankans, rather than being “hyper-selective,” many college-educated Asian immigrants actually obtained their degree in the U.S. and do not have college-educated parents (Sakamoto & Wang, 2021) in contrast to the “hyper-selectivity” assumption than Asian immigrants are uniformly from the upper-classes (Lee & Zhou, 2015).

Due to their lower educational levels, first-generation Bangladeshi, Nepalese, Pakistani, and East Asians have above-average poverty rates. However, with the exception of the Nepalese, all of the native-born South Asian groups obtain higher levels of education than Whites. Poverty rates among Bangladeshi, Nepalese, Pakistani, and East Asians all significantly decline in the second-generation compared to the first-generation although second-generation Bangladeshi and Pakistani have somewhat higher than expected poverty due to family size and composition. The poverty rate for second-generation East Asians actually declines to a level below that of Whites, and the poverty rates for second-generation Indians and Sri Lankans are also lower than for Whites. Second-generation Indians and Sri Lankans have higher household incomes, and are much more likely to be affluent than Whites. To a somewhat lesser degree, Bangladeshi, Pakistani, and East Asians are also more likely to be affluent than Whites.

The results confirm the well-known socioeconomic significance of educational attainment, but they furthermore caution against the common if often implicit assumption that Confucianism underlies the high educational attainments of second-generation Asian Americans (Zhou & Wang, 2016). With the exception of the Nepalese, all of the second-generation South Asian groups have higher levels of educational attainment than native-born Whites, but no South Asian groups have a Confucian cultural tradition. In the cases of second-generation Indians and Sri Lankans, their educational levels even notably exceed second-generation East Asians who are mostly from Confucian societies (i.e., China, Hong Kong, Japan, Korea, Singapore, and Taiwan). Specific religious and sociohistorical cultural orientations per se may be less important than collectivist family child-rearing patterns and concerns for the educational attainment of offspring as markers of family status (Sakamoto & Kim, 2018). Qualitative studies of Indian Americans suggest a strong familial emphasis on the education and academic achievement of children (Bhattacharya, 2000; Dhingra, 2020; Warikoo, 2022) which is consistent with what has been labeled as the “success frame” for East Asian Americans (Lee & Zhou, 2015).

As noted earlier, Bangladeshi and Pakistani are Muslim societies. Globally from a cross-national perspective, predominately Muslim societies tend to have lower levels of educational development (Pew Research Center, 2016). However, in the U.S. context, any broad generalization about a Muslim influence depressing education seems premature. Rather than being educationally disadvantaged, second-generation Bangladeshi and Pakistani have higher levels of educational attainment than native-born Whites despite the fact that first-generation Bangladeshi and Pakistani immigrants are over twice as likely as native-born Whites to lack a high school degree. While U.S. Census data never include information about religion, future research on Asian Americans may be able to investigate other data sources to discern religious effects on socioeconomic outcomes.

Regarding hourly wages, prior studies on the estimation of the net effects of Asian American ethnicity are a bit complicated by the high preponderance and probable endogeneity of East Asians residing in the Pacific where the cost of living is higher (Kim & Sakamoto, 2010; Wang et al., 2017). Our results for East Asians are compatible with that interpretation. However, we find that South Asian groups are substantially less likely to reside in the Pacific than East Asians. Estimating the net effect of ethnicity on wages is likely not complicated by the issue of controlling for region such as the Pacific in the case of South Asians. Whether or not the model controls for region, none of the groups of second-generation South Asian men face any wage penalty. Second-generation Indian men are actually significantly advantaged over White men with the same demographic characteristics in all of the model specifications. The second-generation groups of South Asian women also do not face any wage disadvantage relative to White women. Second-generation Indian, Pakistani, and East Asian women have significantly higher wages that White women with the same age, education, and family characteristics.

These findings on hourly wages do not support the over-education view that second-generation South Asians are under-rewarded in the labor market (Craig, 2019; Tran et al., 2019a) and that there is a “bamboo ceiling” in American firms (Lee & Tran, 2019). None of our results are consistent with that claim for any of the 1.5-generation or native-born South Asian groups. To the contrary, the reverse is more likely to be true since the wages of 1.5-generation and native-born Indian men are advantaged over White men while the wages of 1.5-generation and native-born Indian, Pakistani, and East Asian women are advantaged over White women. Tran et al. (2019b:2273) conclude that the correctness of the over-education view is considered “non-controversial” but that assessment is inconsistent with our findings.

While our data do not directly link generations at the individual level, overall these empirical results nonetheless suggest that the socioeconomic characteristics of the first-generation immigrant groups influence the socioeconomic characteristics of their second-generations which is consistent with prior research on intergenerational mobility (e.g., Corak, 2013). Except for Indians who have a low poverty rate among their first-generation, poverty rates decline among second-generation South Asian, East Asian, and Southeast Asian groups. Except for Nepalese (whose patterns are somewhat less discernable due to a smaller sample size), educational attainment increases among second-generation South Asian, East Asian, and Southeast Asian groups. Because Indian and Sri Lankan immigrants already have high educational levels, their second-generations have very low poverty rates, high household incomes, and high levels of affluence. In regard to explaining average socioeconomic differences across groups, our findings are consistent with the view that parental differences in education play a significant role in the assimilation of the second-generation (Borjas, 1995) although some remaining ethnic differentials are still apparent (Borjas, 1992; Sowell, 1996).

The upward educational mobility of second-generation Bangladeshi, Pakistani, East Asians, and Southeast Asians implies that they do not simply replicate the occupational and labor market positions of their parental generation. The ethnic enclave literature for first-generation South Asian immigrants describes self-employment and small family-run businesses (e.g., motels, convenience stores, taxis, restaurants) as an adaptation to the lack of other opportunities in the U.S. labor market (e.g., Kibria, 2006). However, given high levels of American educational credentials obtained by second-generation South Asians, their rates of self-employment (and unpaid family-business employment) are not only lower than their immigrant parental generations but even lower than native-born and 1.5-generation Whites.Footnote 4 In other words, second-generation South Asian Americans tend to quickly leave ethnic enclave in favor of professional and managerial jobs.

The upward educational mobility of second-generation Bangladeshi, Pakistani, East Asian, and Southeast Asians are also not obviously consistent with so-called segmented assimilation theory. The latter is actually rather vague due to a lack of conceptual clarity, but in general, less assimilated youth do not seem to have higher levels of educational attainment (Xie & Greenman, 2011). According to Zhou (1997:72), “especially dark-skinned immigrants can be trapped in permanent poverty in the era of stagnant economic growth….” but the high upward educational mobility of second-generation Bangladeshi, Pakistani, East Asians, and Southeast Asians do not appear to provide strong support for that hypothesis to the extent that these groups may be considered to be “dark-skinned.” While second-generation Bangladeshi and Pakistani do have slightly higher levels of poverty than Whites, this slight disadvantage seems to be associated more with family size and composition rather than discrimination in the labor market or the educational system.

Broadly speaking, our findings do not provide much support for the “Whiteness privilege” perspective on stratification in the U.S. at least in regard to South Asian Americans. “Skin tone stratification” and “pigmentocracy” may be popular topics in discussions of third-and-higher generation African Americans (Monk, 2021:86), but Indian Americans―who are by far the largest demographic group among South Asian Americans―are greatly advantaged over Whites in terms of education, poverty, household incomes, and affluence. Most second-generation South Asian groups obtain more education than Whites, and their rewards in the labor market are at least commensurate with their degrees if not sometimes higher.

Our data do not include information on skin tone, but in general, it most likely tends to be darker for persons who are indigenous to areas that are closer to equator (Jablonski & Chaplin, 2002). Among the groups investigated in this study, Sri Lanka is the most southern and is closest to the equator. Yet our results do not indicate that Sri Lankans are disadvantaged relative to Whites or even relative to other South Asian groups.

Rather than revealing a highly stratified “pigmentocracy,” we interpret our results for South Asian Americans as being broadly consistent with contemporary assimilation theory (Alba & Nee, 2012). To be sure, the labor market is now characterized by a higher level of overall inequality than was encountered by earlier generations of immigrants (Portes & Zhou, 1993). However, a higher level of overall inequality does not necessarily imply increased relative disadvantage for minority groups or rising racial discrimination (Sakamoto and Wang 2015). Our foregoing findings suggest that by obtaining higher levels of education, most 1.5-generation and native-born South Asian Americans are quickly integrating into the mainstream of American society in contrast to Zhou’s (1997:72) prediction that “especially dark-skinned immigrants can be trapped in permanent poverty….”

Globally, the generally successful socioeconomic attainments of second-generation South Asians in the U.S. may not necessarily be as evident elsewhere in the world. According to some suggestive evidence provided by Raza and Erfani (2015), second-generation Sri Lankans in Canada may perhaps face earnings penalties in the labor market. South Asians in Hong Kong, including their second generations, are generally described in terms of “disadvantage” and “underachievement” (Tsung & Gao, 2012). Racial discrimination including its most conspicuous and brutal forms is also well-known in the recent histories of South Asians in Kenya, South Africa, and Uganda (Sowell, 1996). In the current era, South Asians (among many other migrant workers) are barred from permanently settling in Kuwait, Saudi Arabia, Qatar, and United Arab Emirates despite the considerable economic contributions that these immigrants make towards the development of those countries (Weeraratne, 2020). Further research on the socioeconomic outcomes of South Asians in the U.S. and elsewhere would greatly enhance our understanding of racial and ethnic inequalities in the contemporary world.