Geography, Coethnic Concentration, and Earnings Disadvantage of Asian American Men

Geography of Asian American Labor Market Outcomes

Understanding the earnings dynamics of Asian Americans requires careful consideration of detailed geographies. Leslie McCall (2001) found that relative wages of Asian Americans compared to Whites are highest in the Midwest net of individual human capital variables and metropolitan statistical area (MSA) characteristics. The question is whether the geographical variations in Asian American earnings affect their earnings relative to native-born Whites. Don Mar (1999) argued that the estimates of the earnings gap between Asian Americans and non-Hispanic Whites are inappropriate so long as regional differences are not accounted for. Early immigrant settlement patterns have shaped the geographic concentration of Asian American communities, and Asian Americans have been overrepresented in the areas with high wages and high costs of living (Hurh and Kim 1989; Sakamoto et al. 2009; Takei et al. 2012; Wang et al. 2017). The geographic choice of Asian Americans may be related to the proximity of the American West to Asia, the path dependence of the historical settlement, economic opportunities in urban areas, their strategic adaptation through education, and additional considerations. Asian households tend to spend more on housing to provide better K–12 education for children (Tian 2023). Regardless of what causes it, this spatial clustering has implications for earnings outcomes, as geographic location can influence access to employment opportunities, cost of living adjustments, and other socioeconomic factors.

The sorting of racial groups into various geographies is very much part of the overall stratification process (Massey 2020). Despite the well-known concentration of Asian Americans in a relatively small number of regions and urban areas, the geographic nuances of their earnings disadvantage relative to native-born Whites remain underexplored. Historically, Asian Americans were concentrated on the West Coast, specifically, California and Hawaii (Sakamoto et al. 2009; Tran 2024). While two-thirds of Asian Americans resided in California and Hawaii in 1960, the Asian American population dispersed to other regions in 2019, where 45 percent resided in the West, 24 percent in the South, 19 percent in the Northeast, and 12 percent in the Midwest (Tran 2024, 7). Within state boundaries, Asian Americans are overrepresented in metropolitan areas and underrepresented in rural, nonmetropolitan areas (Xie and Goyette 2004). The overrepresentation of Asian Americans in metropolitan areas has been strongly driven by suburban concentration since the 1990s (Zapatka and Tran 2023), by pioneering into previously White neighborhoods (Kye 2023; Logan and Zhang 2010).

During this period, differences in economic outcomes across detailed geographies have also increased with an overall increase in inequality (Manduca 2019; Moretti 2012; VanHeuvelen and Copas 2019). Since the 1980s, when the long-standing trend of regional economic convergence stalled, variations in economic levels by states and smaller regional units have begun to increase (Abel and Deitz 2019; Ganong and Shoag 2017; Manduca 2019). The peculiarities of the geographic distribution of Asian Americans and recent changes in geographic inequality raise the possibility that the previous estimates of the Asian American net earnings disadvantage, which did not control for detailed regions, may have overestimated their advantages.

There are a small number of studies that have explored the effect of locations on Asian Americans’ labor market outcomes. For example, when accounting for regional variation in Asian American disadvantage, Isao Takei and colleagues (2013) found that college-educated Asian Americans residing in the West tend to be positively selected in terms of earnings, while those in non-Western regions exhibit negative selection. Arthur Sakamoto and colleagues (2013) reported that the Asian American population in the South has grown more rapidly than in other regions, many of them have migrated from California, and after adjusting for the cost of living, Asian Americans in the South enjoy higher average earnings than Asian Americans elsewhere (Sakamoto et al. 2013, 176). Few studies, however, control for detailed residential or working areas. An exception is Sharron Xuanren Wang and colleagues (2017), in which the US metropolitan area cost of living index for the respondents’ current place of residence is controlled for. Wang and colleagues (2017) applied two-stage least squares models, by which the preference for living in high- and low-cost-of-living areas is estimated in the first stage, and then the relative earnings of Asian Americans compared with Whites are estimated in the second stage. Interestingly, Wang and colleagues (2017) reported that Asian Americans are not disadvantaged even after accounting for the cost of living. While adjusting for geographical variations in the cost of living can account for some impacts of detailed geographies, it is limited to capturing price differentials only. Adjusting for the cost of living cannot fully replace the fixed effects of detailed geographic areas.

Whether the residential dissimilarity between Asian Americans and Whites in detailed areas is correlated with the relative earnings of Asian Americans compared to Whites remains unknown. Given the growing income disparities between small geographic areas over the past several decades, accounting for these detailed areas has become increasingly important in studying racial wage gaps. Because Asian Americans tend to be concentrated in a relatively small number of affluent areas, we expect their earnings to appear lower compared to native-born Whites after accounting for local labor market fixed effects.

Local Labor Market Contexts and Asian American Disadvantages

Existing research provides reasons to suspect that the uneven distribution of Asian Americans across the United States may affect their relative earnings, as well as the mechanisms behind this effect. Due to data limitations, this study does not aim to identify which mechanisms are most likely at play. Instead, we highlight how previous studies have suggested that uneven geographic distribution affects the relative earnings of Asian Americans.

Higher concentrations of racial minorities tend to be associated with increased racial inequality, increased racial discrimination, and limited access to resources by racial minorities (Beggs et al. 1997; Blalock 1956; Frisbie and Neidert 1977; Huffman and Cohen 2004; McCall 2001; Tienda and Lii 1987; Tigges and Tootle 1993). For example, Matthew C. Snipp and Charles Hirschman (2004) reported that the occupational status of Asian American men residing in areas with large populations of coethnics is lower than that of Asian Americans living in other areas. Thus, the visibility hypothesis expects Asian Americans’ relative earnings to be lower in local areas where they are concentrated.

Contrary to the visibility hypothesis, the ethnic enclave hypothesis expects Asian Americans’ relative earnings to be higher in local areas where they are concentrated and a relatively higher portion of Asian Americans are self-employed (Portes and Jensen 1989). While self-employment and high concentration of coethnic groups in a CZ are not necessarily indicators of an ethnic enclave in themselves, ethnic enclaves are often closely associated with higher rates of self-employment in areas with relatively higher ethnic concentrations (Portes and Manning 1986). In the pioneering work on the ethnic economy, Ivan Light (1972) showed how Chinese and Japanese Americans could start their businesses relying on ethnic capital such as rotating credit associations in the early twentieth century. Charles Hirschman and Morrison G. Wong (1986) speculated that Asian ethnic enclaves have provided economic means to thrive for highly educated Asian Americans despite the apparent discrimination in the mainstream economic sectors. The ethnic solidarity of Korean Americans was a collective strategy to survive economically at a time of ethnic hostility from Black Americans (Min 2008). The positive effect of the self-employed ethnic concentration is not limited to Asian Americans. For example, Kenneth L. Wilson and Alejandro Portes (1980), focusing on Cuban immigrants, showed how residing within ethnic enclaves can yield heightened returns on human capital accrued from their countries of origin. A concentrated coethnic population, especially a self-employed coethnic population, in this sense, can foster economic opportunities and establish support networks that bolster immigrants’ labor market integration and outcomes.

However, not all research on ethnic enclaves finds a positive association between coethnic concentration and earnings. While the benefits of coethnic clustering help immigrants with resources and institutions of social support, it often runs the risk of saturated labor markets and exploitation by coethnics (Flippen and Farrell-Bryan 2021; Xie and Gough 2011). Especially for Asian immigrants with less education who have limited resources and English fluency, their high dependence on coethnic social networks and resources may increase their risk of exploitation by coethnics. Despite this vulnerability, their limited assimilation and continued reliance on coethnic resources restrict their geographic choices. This implies that, for the less educated, coethnic clustering is a result of structural constraints rather than preference. Yet, another possibility is that less-educated Asian Americans benefit from tangible and intangible resources that highly educated coethnic counterparts build in a community. An example is educational resources. In the latter case, lower earnings are a trade-off for these benefits.

Another context in which the relative earnings of Asian Americans can vary is the model minority stereotype of Asian Americans. The model minority image is also associated with the hyper-selectivity of Asian immigrants. Since the passage of the Immigration and Nationality Act of 1965, Asian immigrants have been subject to double positive selectivity (Lee and Zhou 2015). They are not only positively selected compared to their peers in their origin societies but also more educated than native-born Whites. Hyper-selectivity refers to this double positive selectivity. The hyper-selectivity hypothesis assumes that the upper-middle class culture of the hyper-selected Asian immigrants has a spillover effect into not-so-positively selected Asian immigrants in success frame (Lee and Zhou 2015). This hypothesis is mainly about the extraordinary educational achievements of Asian American children. But with a slight modification, we can apply the hyper-selectivity hypothesis to the labor market outcomes for less-educated Asian Americans. Hyper-selectivity of Asian immigrants brings about positive changes in the cognitive construction of Asian as a racial group (Lee and Zhou 2015; Tran et al. 2018, 2019). Such reconstruction can induce the host society to treat Asian Americans more favorably (Lee and Zhou 2015), which can lead to better outcomes for Asian Americans, including less-educated Asian Americans. Indeed, Byeongdon Oh and colleagues (2023) reported that less-educated Asian Americans are relatively less disadvantaged in labor markets compared to their Black and Hispanic counterparts.

Yet, there is another possibility regarding the effect of the model minority image, which involves what may happen when Asian Americans deviate from the expectation. The study of the model minority comes almost always with the label of myth (Chou and Feagin 2015; Kao and Thompson 2003; Shams 2020). This is because not all Asian Americans fit the image of a model minority, and often, the portrayal of Asian Americans as a model minority is inaccurate and damaging to Asian Americans who do not fit this image (Kao and Thompson 2003; Min 2003). The so-called model minority myth effect refers to “an earnings disadvantage that Asian Americans may have relative to whites that is net of productivity and is associated with an exaggerated characterization or level of performance that is required for Asian Americans” (Kim and Sakamoto 2014, 285). In this line of argument, the exaggerated expectations of performance or work characterization for Asian Americans will especially be high in the areas where more Asian Americans fit the model minority image by being highly educated. In such an environment, the wages of less-educated Asian Americans can suffer more relative to equally educated Whites, due to the heightened, racialized expectations.

In sum, existing theories suggest that a higher concentration of coethnic groups or a higher proportion of the highly educated individuals among Asian Americans in local labor markets, or both, may be associated with either higher or lower relative wages for Asian Americans. Although this study does not examine specific mechanisms, as a first step, we assess whether coethnic concentration is linked to relative wage outcomes. We pay particular attention to less-educated Asian Americans who may benefit more from coethnic resources in local labor markets while simultaneously being subject to coethnic competition and exploitation.

Data

The multiyear ACS offers enough sample sizes for Asian Americans in small geographic areas. Samples are limited to those non-Hispanic single-race identified Asian American and native-born White men in their prime working ages from 25 to 64. Only those who are not in school and report positive annual wages are included, excluding the self-employed.

To account for the differences in the detailed geographic areas between Asian Americans and Whites and the effects of local labor markets, we use commuting zones as a geographic unit. As interest in geographic inequality has grown, scholars have developed alternative ways to define meaningful geographic units beyond administrative boundaries, such as counties or MSAs. One such approach is the use of CZs (Tolbert and Sizer 1996). The initial effort to construct CZs stems from the inadequacy of counties as proper units for understanding an area’s economy (Fowler 2024; Fowler, Rhubart, and Jensen 2016). CZs map the boundaries within which people are likely to commute from residence to work, thereby providing meaningful context for local labor markets and everyday activities. CZs have become a standard measure of local economies in academic research (for example, Acemoglu and Restrepo 2020; Chetty et al. 2014; VanHeuvelen and Copas 2019). In this article, we treat CZs as proxies for local labor markets. To construct CZs, individual-level data are expanded following David H. Autor and David Dorn (2013) and Tom VanHeuvelen (2018), adjusting survey weights to obtain correct CZ-level estimates. The geographic scope of the data is the contiguous United States, consisting of 722 CZs.

Key Variables

The dependent variable for this research is the log of annual wages. Before log transformation, wages are adjusted to 2019 constant dollars. The log transformation of wages follows common practice, which facilitates dealing with the highly skewed distribution and eases of interpretation.

We classify Asian Americans into two groups: native-born Asian Americans and immigrant Asian Americans who immigrated before the age of thirteen (the 1.5+ generation, hereafter), and Asian immigrants who immigrated at or after the age of thirteen (the first generation). For simplicity in reporting results, we group native-born Asian Americans and the 1.5 generations together because the results for these two groups are largely similar. In our main analyses, each group is further divided by levels of education: high school graduates or less than a high school degree (HSG); some college (SC); bachelor’s degrees (BA/BS); and advanced degrees (GRAD). Thus, a total of eight Asian American groups (that is, four levels of education by two immigrant generations) are examined. In some models, we examine the variation across Asian ethnic groups by disaggregating Asian Americans into Chinese, Korean, Filipino, Vietnamese, Asian Indian, Southeast Asian, South Asian, and other Asian groups, including multiethnic Asians.

Analytic Strategy

Our analyses mainly test two models. First, net of covariates, we assess whether education-specific Asian American disadvantages change when we control for CZ-fixed effects. Second, we examine whether the disadvantages observed in the first model are associated with CZ characteristics such as the density of the Asian American population, the size of the self-employed Asian American population, and the proportion of highly educated Asian Americans. Ordinary least squares models with a large number of fixed effects are estimated.¹ All estimations are weighted. As CZ characteristics are our key independent variables in some models, CZ-year-clustered standard errors are reported. Equation (1) shows the model of the first set of analyses:

where Y_ijk refers to the log wage for individual i of group j in CZ k. With native-born Whites set as the reference group, A_j indicates eight Asian American groups separated by immigrant generation and education. C_k are CZ-fixed effects. The key interest is how the coefficients for being in Asian American groups, β_j, change before and after the inclusion of C_k.

The matrix of control variables, X, includes an extensive set of both individual-level demographic covariates, including levels of education, age, age squared, marital status (married with spouse present, married with spouse not present, separated, divorced, widowed, and never married single), detailed degree field (171 fixed effects), number of children, family size, English proficiency. It also includes labor market covariates such as weeks worked last year (1–13, 14–26, 27–39, 40–47, 48–49, and 50–52 weeks), usual hours worked per week, and class of worker status (wage or salary: private; wage or salary: non-profit; federal government; state government; local government; and unpaid family worker). Detailed occupations (328 fixed effects) are controlled in some models. T_t are survey year fixed effects.

In addition to equation (1), we perform Kitagawa–Oaxaca–Blinder decomposition analyses (Blinder 1973; Kitagawa 1955; Oaxaca 1973). The decomposition analysis aims to compute the degree to which CZ’s composition and coefficient effects account for the wage gap between Asian Americans and Whites and to compare them with other effects, such as education. Equation (1) and decomposition analyses test whether detailed geographies affect the relative wages of Asian American men compared to native-born White men.

Next, we explore the association between the relative earnings disadvantage of Asian American men and local labor market characteristics. Local labor market characteristics, Z_kt, include four variables: the logged Asian American population, the logged Asian American self-employed population,² the proportion of Asian Americans with BA/BS or higher, and average logged household income. These characteristics are measured at the CZ level separately by year among those aged twenty-five to sixty-four. For ease of interpretation, the proportion of BA/BS or higher among Asian Americans is multiplied by 10. Equation (2) shows the model:

Of main interest from equation (2) is the coefficient estimate of δ_j which quantifies the extent to which the relative earnings of Asian Americans grow or decrease as local labor market characteristics, Z_kt, increase by one unit (by 10 percent points in case of the proportion of BA/BS or higher among Asian Americans). Because four local labor market covariates covary, we estimate equation (2) separately for each covariate. Note that as we control for CZ- and time-fixed effects and their interactions, the main effect of Z_kt is dropped from the model due to perfect multicollinearity. Thus, the estimated δ_j is not affected by other observed and unobserved CZ-level heterogeneities and their variation across survey years. Equation (2) tests the association between detailed geographic characteristics and the relative wages of Asian Americans. If δ_j is negative, the concentration of coethnic groups or the proportion of Asian Americans with a BA/BS or higher is negatively associated with their wages. If δ_j is positive, it suggests that Asian Americans tend to benefit from coethnic concentration or hyper-selectivity.

Asian American Disadvantage in Local Labor Markets

Table 2 shows the changes in Asian American disadvantages as more detailed geographic regions are added in each model. Before controlling for region and with control for demographic, labor market covariates, levels of education and fields of study, model 1 shows that 1.5+ generation Asian American men are advantaged compared to White men, while first-generation Asian American men are disadvantaged. In models 2 and 3, the addition of 8 census regions and 48 state-fixed effects reduces the 1.5+ generation Asian American advantage to 0.022 and −0.010, respectively, and the first-generation Asian American disadvantage to −0.140 and −0.174. When 721 CZ-fixed effects are added in model 4, 1.5+ generation Asian American men turn out to be disadvantaged by 4.7 percent (= e^−0.048 − 1) compared to White men, and the disadvantage of first-generation Asian immigrants is enlarged to a whopping 19.5 percent (= e^−0.217 − 1). The Duncan dissimilarity index in nine census regions is 0.307, which means that around 31 percent of Asian Americans need to relocate to have the same distribution as Whites. In terms of CZs, the dissimilarity index is substantially larger at 0.465.

Next, we further desegregate Asian Americans by their levels of education and assess their relative earnings in table 3. Without controlling for CZ-fixed effects, 1.5+ generation Asian American men show higher relative earnings than their White counterparts regardless of levels of education. For first-generation Asian immigrants, their earnings are lower than Whites. The disadvantage for less educated Asian immigrants is more than 15 percent compared to equally educated Whites, while those with an advanced degree seem to have reached near parity. However, when CZ-fixed effects are added in model 2, all 1.5+ generation Asian American groups except graduate degree holders turn out to be disadvantaged relative to equally educated Whites. The disadvantage for first-generation Asian immigrants, including graduate degree holders, becomes substantially larger. Accounting for detailed local labor markets, the disadvantage suffered by first-generation Asian American men enlarges, and the advantage enjoyed by the 1.5+ generation turns to a disadvantage. These results suggest the importance of incorporating detailed geography in understanding labor market performance of Asian American men.

There is debate about whether adding occupation to the right-hand side of the regression equation creates an over-control problem (Sakamoto and Wang 2021; Tran et al. 2019). We do not claim that occupation should be controlled for when assessing the relative wages of Asian Americans. Instead, we treat the occupational distribution as an indicator of the strategic adaptation of Asian Americans and check whether the control of occupation accounts for Asian American disadvantages. First, we control for occupation without adding CZ-fixed effects in model 3. Accounting for detailed occupation has heterogeneous effects across levels of education. Compared to the coefficients estimated in model 1 in which occupation is not controlled for, the relative wages of Asian Americans without a bachelor’s degree either increase or remain the same in model 3 with the control of occupation, while those for the highly educated decreases. The differences in net wage disadvantage of Asian Americans between models 1 and 3 suggest that the distributional difference in occupations between Asian Americans and Whites is associated with lower wages for less-educated Asian Americans, while it is associated with higher wages for highly educated Asian Americans. Occupation explains almost all advantages for graduate-school-educated 1.5+ generation Asian American men.

These results highlight the strategic adaptation of college-educated Asian Americans. Asian Americans are likely to choose occupations that require the most education and can signal competence (Lee et al. 2024). Unlike the highly educated who can strategically choose fields of study in college that might lead to occupations in which the Asian American disadvantage is weak or nonexistent, less-educated Asian Americans might be devoid of such a strategic adaptation in choosing occupations, and thus the occupational distribution for less educated Asian Americans may be a source of disadvantage rather than advantage. The increase in relative wages for less-educated Asian Americans in model 3 compared with model 1 is consistent with this expectation. When both detailed occupations and CZ-fixed effects are added in model 4, the educational gradient of Asian American disadvantages disappears. All Asian Americans, regardless of levels of education and immigrant generation, are statistically significantly and substantially disadvantaged compared to their White counterparts.

Ethnic Diversity of Asian Americans

The ethnic diversity of Asian Americans is one of the key characteristics that define Asian Americans (Tran 2024; Vo et al. 2023). Thus, we reestimate model 2 of table 3 after desegregating Asian Americans by their ethnicities. Table 4 shows the result.

As expected, there are noticeable variations across Asian ethnic groups. In particular, the ethnic variation among highly educated 1.5+ generation Asian Americans is noteworthy. Asian Indian Americans with advanced degrees are the only group with a statistically significant advantage over Whites. Korean Americans are another group with small, though statistically insignificant, advantages over Whites. Chinese and Vietnamese Americans with advanced degrees have reached parity with Whites. Other Asian groups are disadvantaged compared to Whites. Among those with a bachelor’s degree, Asian Indians and Korean Americans are not disadvantaged, while other groups endure wage disadvantages. Among the less educated, almost all ethnic groups are disadvantaged. Less-educated other Southeast Asian Americans do not seem to be disadvantaged.

For Asian immigrants, all ethnic groups are disadvantaged compared to Whites regardless of levels of education. Although they are disadvantaged like other groups, the extent to which highly educated Asian Indian immigrants are disadvantaged is noticeably smaller than that of other Asian ethnic groups. Unlike the highly educated, less-educated Asian Indian immigrants are as disadvantaged as other Asian ethnic groups. This might be because highly educated first-generation Asian Indian immigrants are more hyper-selected than other groups (Lee et al. 2024) and that, unlike East Asians, they have a culture encouraging assertiveness (Lu et al. 2020), which is valued in leadership positions. Overall, table 4 uncovers that, with a few important exceptions, most Asian ethnic groups endure disadvantages in wages compared to Whites, once levels of education and local labor markets are controlled for. Detailed geography matters for the relative wages of most Asian ethnic groups.

Decomposition of Asian American Wage Disadvantage

Analyses so far demonstrate that 1.5+ generation Asian Americans’ wage advantage is a spurious outcome of the omitted-variable bias and that the disadvantages of first-generation Asian immigrants are underestimated due to the lack of control for local labor markets. The following question is how much overestimation of Asian American wages is associated with geographic distributions. To assess this, we conduct the Kitagawa–Oaxaca–Blinder decomposition analyses. The results are presented in table 5, where the first set of result columns shows the detailed decomposition of the wage advantage for 1.5+ generation Asian Americans, and the second set shows that for first-generation Asian immigrants. In this analysis, all covariates are grand-mean centered (Kim 2013), and thus, the constant quantifies the expected wage gap when all covariates are set at their grand means for both Asian Americans and Whites.

In the decomposition analyses, we are primarily interested in the composition and coefficient effects of CZs, which are the sums of 721 CZ-fixed effects. Out of the total gaps between 1.5+ generation Asian Americans and Whites, the composition effect of CZs contributes the most to the higher wage of Asian American men relative to White men. The total gap between 1.5+ generation Asian Americans and Whites is 12.4 percent, or 0.117 log dollars, which means Asian Americans earn 12.4 percent more than Whites. Out of this advantage, the composition effect of CZs accounts for about three-fourths (0.091 out of 0.117). That is, 1.5+ generation Asian Americans earn 9.5 percent (= e^0.091) more than Whites simply because they reside in high-wage CZs. The effect size of the composition of CZs is slightly larger than that of occupation and comparable to that of education, including fields of study (0.095 = 0.058 + 0.037). The overrepresentation of Asian Americans in high-wage CZs has a similar effect size to the overeducation of Asian Americans and their strategic adaptation in the science, technology, engineering, and mathematics majors. The coefficient effects of CZs are negligible, which indicates that 1.5+ generation Asian Americans are not getting particularly favorable treatment when residing in high-wage CZs.

For first-generation Asian immigrants, the grand advantage is 9.7 percent, or 0.093 log dollars. Out of this advantage, the composition effect of CZs (0.063) accounts for about two-thirds. The effect size of the composition of CZs is similar to that of occupation (0.064). Compared with the combined composition effects of levels of education and fields of study (0.117 = 0.045 + 0.072), that of CZs is around half. Thus, the lack of control for local labor markets yields more biased results when estimating the relative wages for 1.5+ generation Asian Americans than for first-generation Asian immigrants. Conversely, controlling for detailed fields of study is more important when estimating the relative wages for first-generation Asian immigrants than for 1.5+ generation Asian Americans.

Coethnic Concentration and Asian American Disadvantage Across Local Labor Markets

As shown in figure 1, panel B, the wage gaps between Asian Americans and Whites are not constant across local labor markets but vary substantially. In this section, we explore how the variations are systematically associated with Asian American coethnic concentration and educational selectivity. If Asian Americans benefit from the ethnic economy, the relative wages for Asian Americans are positively associated with either the population size of Asian Americans in the CZ or that of self-employed Asian Americans or with both. If there is a spillover effect of hyper-selected Asian Americans, the relative wages for Asian Americans are positively associated with the proportion of highly educated Asian Americans in the CZ. Conversely, if coethnic competition, coethnic exploitation, and the model minority image hurt Asian Americans who do not fit this image, the relative wages for less-educated Asian Americans would be negatively associated with the proportion of highly educated Asian Americans in the CZ. As noted earlier, we do not assess which mechanisms are most likely at play. Instead, we examine whether coethnic concentration and the proportion of highly educated Asian Americans across CZs are associated with their relative wages.

Figure 2 plots the predicted (dis)advantage of Asian American men relative to equally educated Whites across four CZ-level characteristics: log Asian American population, log population size of self-employed Asian Americans, the proportion of bachelor’s degree holders among Asian Americans, and average log family income by CZ. The first two variables indicate ethnic economy and the third variable indicates the educational selectivity of Asian Americans. It turns out that the higher the concentration of Asian Americans and the higher the number of self-employed Asian Americans, the lower the relative wages of Asian American men relative to Whites. The negative associations look steeper for first-generation Asian immigrants than for 1.5+ generation Asian Americans. An interesting finding is that for all levels of education, 1.5+ generation Asian Americans do not experience wage disadvantage if they live in CZs with a small Asian American population.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Education-Specific Wage Disadvantage of Asian American Men Across CZ Characteristics

Source: Authors’ tabulation based on American Community Survey 2010–2019 (Ruggles et al. 2024).

Note: Plotted estimates represent the expected net disadvantage relative to equally educated native-born White men, expressed as percentages, at P5, P10, P25, P50, P75, P90, and P95 of CZ-level variable. Shaded areas indicate confidence intervals at 95 percent. Based on equation (2). HSG = high school graduates, SC = some college, BA/BS = bachelor’s degree, GRAD = advanced degrees. CZ = commuting zone.

The associations between the proportion of highly educated Asian Americans in the CZ and the relative wages of Asian Americans vary depending on levels of education. As the proportion of highly educated Asian Americans grows in a local community, the earnings disadvantage of the less educated Asian American workers becomes worse. Contrary to this, the increase in the proportion of highly educated Asian Americans is positively associated with the net earnings of Asian American workers with a BA/BS. On the other hand, the association is negative for 1.5+ generation Asian Americans with an advanced degree. With these mixed results, it is hard to make a strong conclusion regarding the association between higher proportion of BA/BS holders among Asian Americans in local communities and the relative wages of Asian American men. At minimum, this result indicates that there is no positive wage spillover effect for less-educated Asian Americans in areas with higher proportions of highly educated Asian Americans.

We also test whether Asian Americans earn more by living in affluent CZs and find no such association. As shown in figure 2, panel D, average CZ family income is either negatively associated with the relative wages for Asian Americans or not associated. In particular, less-educated Asian Americans suffer from larger relative wage disadvantages when they live in affluent CZs. This result is consistent with the speculation that less-educated Asian Americans may invest in houses in affluent areas because affluent CZs offer good public schools despite such action hurting their wages (Tian 2023).

Robustness Check

While we have accounted for CZ-fixed effects in this study, some will argue that this is not enough to control for the heterogeneous preferences of Asian Americans. There is a possibility that Whites are more likely to move internally for better wages than Asian Americans, while Asian Americans are more likely to stay in coastal areas where the population density of Asian Americans is high. As shown in figure 2, panel A, Asian population density is negatively associated with wages for Asian Americans. To address this concern, we limit our sample to the native-born and estimate the model with control for 51 birth states and their full interaction with 722 CZs (= 35,358 fixed effects excluding empty cases). In this model, respondents’ preference for the current residential area given their birth states is fully controlled. Table 6 shows the results.

The estimated disadvantages for Asian Americans who are second generation or greater are reduced only marginally when the full interactions between birth states and current CZs are added. There are no changes in the direction and significance of the estimated coefficients of being Asian Americans, regardless of levels of education. When we conduct the formal test of the change in coefficients between model 1 and model 2, no changes are statistically significant.

As shown in table 1, 1.5+ generation Asian Americans are younger as a group than native-born Whites and first-generation Asian Americans. Limiting the respondents’ age ranges from twenty-five to fifty-four or from thirty-five to forty-four alters the estimated coefficients as expected but does not change our main conclusion. For an additional robustness check, we exclude California and New York states from our analyses and repeat the same analyses with table 3. Our conclusions are not altered either.

Our sample includes only wage workers. Given the higher likelihood of self-employment among Asian Americans, particularly first-generation immigrants (Le 2012; Tran and Poon 2011), less-educated Asian American wage workers may be negatively selected in terms of potential earnings, especially in CZs with a high concentration of highly educated coethnics and greater ethnic business opportunities. To address this concern, we repeat the same analyses with table 3 and figure 2 using a sample that includes the self-employed, with annual earnings as the dependent variable. The results are almost identical to what we report here.