Contextual Boundaries: Skin Tone Stratification and Skill Transferability Among Mexicans in the Age of Mass Migration

Abstract

An extensive literature has focused on the association between human, social, and economic capital and better immigrant economic attainment, and how these characteristics contribute to stratification among members of the same group. However, few studies have explored how racialization processes contribute to these within-group differences. We examine the role of intragroup differences in skin tone in stratifying outcomes among Mexican immigrants in the early twentieth century. We create a new dataset of 1910–1940 Mexican border-crossing records that we then link to the U.S. 1940 census. We use characteristics at entry to predict income in 1940 and find that—in line with dominant assimilation theories—standard measures of capital are associated with within-group attainment differences. However, we also find skin tone to be a source of within-group stratification: being perceived as having darker skin is associated with lower subsequent economic attainment than being perceived as having lighter skin. Furthermore, whereas human and social capital transcended context to allow migrants to transfer those skills anywhere, the effect of skin tone was significant only in Texas and not in other major receiving places like California. We argue that although standard measures of assimilation typically predict later outcomes, the stratifying effect of skin tone has long been a feature of Mexican immigration.

Introduction

In the last century, Mexican immigrants have been the largest and most consistent immigrant group arriving in the United States. How these immigrants fare after arrival has become integral to theoretical understandings of assimilation and integration (Alba and Nee 2003; Portes and Rumbaut 2001; Waters and Pineau 2015). Central to these debates are labor market success and how individual- and contextual-level processes are associated with income differences between Mexican immigrants and other groups (Alba and Nee 2003; Portes and Rumbaut 2001). Research has long addressed between-group differences, but researchers have begun to analyze how background, demographic, and contextual traits vary across members within the same immigrant group (e.g., see Catron 2019; Connor 2019; Fox and Bloemraad 2015). In this article, we seek to understand how individual-level characteristics produce differences within the Mexican immigrant population and whether they vary across contexts.

Traditional accounts of within-group differences often point to background differences in human capital that produce differing trajectories after settlement. For example, immigrants with more education can often transfer their skills across contexts, enhancing their employment prospects over time (Ichou 2014; Potochnick and Hall 2021; Villarreal and Tamborini 2018). However, other research suggests that differences might also stem from variations in how Mexicans are racialized given their ambiguous position vis-à-vis the dominant Black–White racial structure (Alba 2005; Catron 2023; Escamilla-Guerrero et al. 2021; Perlmann 2005). Specifically, individuals from the same sending country are often perceived to have different racialized characteristics (Selod 2015) that shape how the dominant group treats them (Marrow et al. 2022). Thus, individuals perceived to have a lighter skin complexion often have better employment prospects net of their individual human capital (Han 2020).

However, the availability of longitudinal data allowing researchers to assess the impact of premigration characteristics and skin complexion on Mexican immigrants’ outcomes after settlement is limited. Further, the available longitudinal data cover modern contexts, where legal status will likely override background characteristics. In this article, we turn to the first half of the twentieth century—a time with fewer legal barriers to entry for Mexican immigrants—to assess the association between migrants’ characteristics at entry and their later economic attainment. Specifically, we create a new dataset of border-crossing records of Mexican individuals who arrived in the United States in 1910–1940 that we then link to their 1940 census records. U.S. border-crossing records provide a wealth of information not included in U.S. censuses, such as individuals’ physical description, occupation before arrival, and other demographic and economic characteristics. Importantly, these records include the person's skin complexion as perceived by a border inspector. We use this information to predict income in 1940. This dataset is the first to connect Mexican immigrants’ at-migration characteristics and physical description, including skin color, to their after-settlement characteristics.

Our findings suggest that social and human capital are important for explaining within-group differences in attainment. Consistent with traditional accounts of within-group differences, we find that immigrants could translate literacy skills, education, and social capital into better labor market outcomes after settlement. We also find that being perceived as having a light skin tone is associated with greater subsequent economic attainment.

These patterns, however, vary by context. We demonstrate that perceived skin complexion had a greater effect on attainment for those who settled in Texas than those who settled in California. Using a Oaxaca–Blinder decomposition, we find that the distributional differences between Texas and California are not explained by observable differences in migrants’ characteristics but rather by unmeasured social processes. Drawing from previous work on ethnic boundaries during the historical period, we argue that the presence or absence of skin tone stratification is likely due to the ethnic boundaries established in specific local labor markets. We argue that although standard measures of assimilation typically predict later outcomes, the stratifying effect of skin tone has long been a feature of Mexican immigrant history in the United States. These findings contribute to recent work that highlights infracategorical differences as an important but understudied organizing feature of inequality (Monk 2022).

Mexican Immigration in the First Half of the Twentieth Century

Mexican immigration is one of the most mature migration flows to the United States, and its dynamics have evolved over time as various sociopolitical and demographic phenomena unfolded on both sides of the border. Soon after the seizure of Mexico's northern territories following the Mexican–American War of 1846–1848, Mexicans began crossing the newly formed territorial border for seasonal work or permanent stay. Apart from 1910–1920, when many Mexicans arrived as refugees fleeing the Mexican Revolution, the Mexican flow has been characterized as economic. Mexicans seeking work in agriculture, railroads, and mining settled throughout the U.S. Southwest, where wages were six times higher than wages for similar work in Mexico (Pedraza-Bailey 1985). Despite its involvement in U.S. immigration history, however, the Mexican flow did not compose a significant part of the immigrant share until the early twentieth century. Between 1900 and 1930, the share of Mexicans among the total U.S. immigrant stock rose from less than 1% to 4.5% (author's calculations of U.S. Census data). On the U.S. side, the rise in Mexican immigration was fueled by strong demand for workers in the U.S. Southwest and Midwest (Martinez 1978). On the Mexican side, emigration was facilitated by the spread of railroads that connected the interior of Mexico to the United States, as well as the economic devastation of the 1910–1920 revolution and other political conflicts in later decades (Verduzco 1995).¹

In the early twentieth century, Mexicans primarily settled in southwestern states, where large-scale irrigation projects had expanded farming and required a cheap workforce. Because the 1882 Chinese Exclusion Act and the 1907 Japanese Gentleman's Agreement limiting the number of workers who traditionally held these positions had created labor shortages, employers in the Southwest began recruiting Mexicans to fill the demand for labor. In addition, meatpacking houses and manufacturing plants in the Midwest sought a Mexican workforce to provide cheap labor and serve as strikebreakers. Following U.S. entry into World War I, the demand for Mexican workers increased in agricultural, railroad, construction and maintenance, and mining jobs throughout the United States (Cardoso 1980).

How Mexicans fared in the labor market after settlement was associated with several individual- and contextual-level processes, which largely kept Mexican immigrants in low-paid occupations during this time (Perlmann 2005). However, Mexicans arrived with varying levels of human and social capital. In addition, the Mexican flow was heterogeneous across class and ethnic and Indigenous profiles, contributing to differences in their postmigration racialization experiences.² The flow included individuals of European descent, Mestizos, Indigenous groups, and Afro-Mexicans. Although the flow of Indigenous and Afro-Mexicans was small during this period (Fox and Rivera-Salgado 2004), a small flow of Indigenous men from northern Mexico (Mayos, Yaquis, and Pápagos) worked in the Arizona mines beginning in the late nineteenth century (Oehmichen-Bazán 2015; Weber 2008).³

Skill Transferability and Skin Tone Stratification

Dominant explanations of within-group economic differences focus on the role of capital in immigrants’ socioeconomic attainment (Feliciano and Lanuza 2017; Ichou 2014; Portes and Rumbaut 2001; Potochnick and Hall 2021). Individuals with more capital often fare better in the labor market than individuals with less capital. For example, immigrants with higher levels of education experience accelerated assimilation, with strong wage growth in the first years after arrival (Villarreal and Tamborini 2018). This pattern, however, depends on immigrants’ ability to transfer the skills and credentials they obtained in the sending country to the destination country's labor market. Immigrants might be unable to successfully transfer their skills to new contexts where those skills hold less labor market value or where discriminatory ideology relegates immigrant groups to a class of “unskilled” workers (Iskander 2021). Thus, immigrants’ labor market success also depends on their placement in the racial hierarchy they enter (Fox and Guglielmo 2012).

Racialization processes, therefore, also condition immigrants’ postarrival economic outcomes, net of their human and social capital (Han 2020; Louie 2012; Villarreal and Tamborini 2018). Racialized meanings are applied to newcomers’ cultural and physical characteristics (Selod 2015), impacting their integration, adaptation, and attainment trajectories. A growing literature has shown associations between skin tone (i.e., having or being perceived to have a lighter or darker complexion) and a range of outcomes, including health (Monk 2015), socioeconomic status (Hunter 2005), educational attainment (Hughes and Hertel 1990), and policing and punishment (Monk 2019). Skin tone stratification emerges either through discrimination according to skin color or through the inheritance of (dis)advantage within families, which is itself the result of discrimination experienced by previous generations (Abascal and Garcia 2022; Campos-Vazquez and Medina-Cortina 2018; Monroy-Gómez-Franco et al. 2022). Discrimination may occur from native-born individuals or members of the same coethnic community, who often have a greater preference for lighter skin (Monk 2015). Among immigrants, dark skin tone is associated with greater downward mobility upon integration and slower subsequent upward mobility (Han 2020).

Skin tone might be associated with economic trajectories because dark complexion penalties can be large (Han 2020). Observers use skin tone to identify individuals within a racial group, conditioning how observers treat those individuals. In the labor market, employers might use these physical characteristics to discriminate in hiring and pay, regardless of the individual's skills and talents.

Context Effects on Skill Transferability and Skin Tone Stratification

Skill transferability and skin tone stratification might vary depending on where immigrants settle. That is, immigrants do not choose the racial and ethnic hierarchy into which they enter nor how their skills may be utilized within local labor markets, but they are nevertheless subjected to those contextual forces that ultimately determine their life chances. One of the defining features of how these processes unfold is determined by the ethnic boundary within a given area.

As Alba (2005) argued, ethnic boundaries can be bright, with individuals falling on either side of a divide. Alternatively, ethnic boundaries can be blurry, with individuals falling in an indeterminant location and viewed as members of either side or as having situation-dependent membership. When boundaries are bright, barriers to better socioeconomic attainment are large (Alba 2005; Zolberg and Woon 1999). When boundaries are blurry, ethnic divides are more easily crossed, and individuals might experience better socioeconomic attainment. We do not seek to describe the conditions under which boundaries are created and maintained (for more discussion, see Wimmer 2008). However, we note that such boundaries can override processes of skill transfers and produce skin tone stratification in only some contexts because some ethnic boundaries are associated with high levels of discrimination and exclusion, whereas others do not matter in everyday aspects of life. In areas with a high degree of social closure, immigrants might have little chance of transferring their skills or succeeding because of the ethnic category imposed on them by employers and the dominant group within an area.

In the first half of the twentieth century, Mexicans were famously affected by ethnic boundaries in a given area (Alba 2005; Fox and Guglielmo 2012). These boundaries had both formal and informal components, depending on the geographic location, that made them blurred or blurrable in some places but bright in others. For instance, in some places, Mexicans were systematically excluded from economic and civic life and relegated to “Mexican” schools, jobs, neighborhoods, and public accommodations (Fox 2012; Fox and Guglielmo 2012). In others, Mexicans were welcomed into the mainstream and held public office and citizenship (Fox 2012). For instance, in New Mexico, newcomers were welcomed by immigration-friendly policies and native-born individuals, many of whom were of Mexican descent, which allowed for better economic attainment (Jacobson et al. 2018). In places such as Arizona, however, Mexicans experienced a harsher environment that severely depressed their chances of leaving unskilled positions (Jacobson et al. 2018). Harsher social contexts produce a form of social control that can shape individual and group behaviors in an area (De Genova 2002).

The defining difference for Mexican immigrants in the first half of the twentieth century was whether they settled in Texas or California. These states held nearly 80% of the U.S.-based Mexican population during this period, yet each state had a different political regime and economy that ultimately shaped ethnoracial boundaries, whether bright or blurry, in the first half of the century. In Texas, where ethnic boundaries tended to be bright, informal and formal practices shaped Mexican experiences. Texas began as a Confederate state, producing a rigid Black–White racial binary that gave Whiteness a higher social premium (Ballinas and Bachmeier 2020). As a result, Mexicans entered a racialized context in which their treatment was based (often explicitly) on skin color (Fox and Guglielmo 2012). In this period, informal boundaries were often instituted with a preference for lighter skin. For instance, Texas farmers explicitly noted their preference for lighter-skinned workers to do higher-paying jobs (Menchaca 2011; Montejano 1987). However, boundaries in Texas were also formalized. For example, judges in Texas would deny naturalization claims for individuals deemed “White by law” but whose skin was considered too dark (Fox 2012). Similarly, Texas unions (particularly the American Federation of Labor) enforced a color bar for higher-paying positions within cities (Montejano 1987), and Texas had no public relief system, greatly affecting the Mexican population (Fox 2012). Texas also had a strong history of racialized violence by the native-born and Texas Rangers through lynchings and other systematic campaigns against Mexican communities, which have been linked to worse economic outcomes for Mexican individuals (Escamilla-Guerrero et al. 2021).

California, on the other hand, was incorporated into the United States as a nonslaveholding state. The arrival of many different ethnoracial groups to this state meant that the White majority was less unified around an Anglo identity, blurring many social boundaries. For instance, Mexicans were often classified with the Black population in Texas but were often treated as separate from the Black population in California (Fox 2012). This treatment as separate from the Black population entitled Mexicans to access to social welfare spending, citizenship, and voting. Mexicans in California could also intermarry with native-born Whites, and native-born individuals often thought of Mexicans as White (Almaguer 1994). California's wage structure and great need for skilled and unskilled workers also contributed to Mexicans’ socioeconomic attainment. Demand for labor meant that employers could not be discerning on factors such as skin color, which greatly benefited the Mexican population overall (Laslett 2014). Because boundaries within the state were blurrier than in Texas, the salience of skin color likely mattered less during this era. The association between skills, skin tone, and Mexican socioeconomic attainment and their variation by context during this period, however, have rarely been studied.

Data and Methods

U.S. Border-Crossing Records

In assessing the conditions of Mexican economic success after settlement, we used U.S. border-crossing records to create a unique data source of Mexican immigrants who arrived between 1910 and 1940. Since its founding, the United States has kept detailed records of individuals stepping onto U.S. soil. However, most laws governing what information was recorded corresponded to those arriving via ships from Europe since 1813 (Catron 2020). Land border agents, especially at the southern border, did not systematically keep information on individuals who entered the United States until the early 1900s, when the Immigration Act of 1903 called for rules governing entry and inspection. Initially, inspectors at the Mexican border were tasked with enforcement of the Chinese exclusion law. However, their duties expanded in 1905 to include completing a standardized form, Form 548, for each alien applying for admission at an official port of entry. Form 548 collected much of the same information found on ship manifest lists, which recorded information on European immigrants who entered at that time—for example, immigrants’ physical descriptions and demographic and geographic characteristics, depending on the year of entry.

For illustration, we reproduce a border-crossing record from 1920 in Figure 1. These manifests include a physical description, occupational and economic characteristics, where the immigrant intended to reside initially after crossing the border, and motivations for migration. However, the information recorded in the manifests changed over time. Before mid-1917, manifests included only basic demographic information (e.g., age, marital status, and birthplace), a physical description, and socioeconomic characteristics (e.g., how much money the person carried, their occupation, and whether the person could read or write). After 1917, however, more information about the individual was recorded, including the purpose for entry, the length of time they intended to remain in the United States, their intention to become a citizen, and whether they had ever been deported. The change in information collected by border inspectors corresponds with the 1917 Literacy Act, which explicitly increased the amount of documented information required for entrants at all ports of entry. The information collected after 1917 remained largely the same until these records stopped being generated in the 1950s.

As of this writing, the information from Form 548 records has not been transcribed.⁴ We collected information from all ports of entry in Arizona (Douglas, Naco, and Nogales), California (Calexico and San Ysidro), New Mexico (Columbus), and Texas (Brownsville, Del Rio, Eagle Pass, El Paso, Hidalgo, Laredo, and Rio Grande City). The records can be found at the National Archives and Records Administration (NARA), and we accessed the file images through Ancestry.com.

We conducted two data collections to create a stratified random sample of individuals who entered between 1910 and 1940. First, we used data from an earlier project on Mexican refugee integration to collect information on individuals who entered between January 1, 1910, and December 31, 1920 (Catron and Vignau Loria 2021). Second, we used data on all entrants from 1921 to 1953. In both cases, the records in the NARA files are sorted by entry date. Therefore, we created a stratified random sample of individuals for each entry port. Within each entry port, we partitioned consecutive strata of 1,000 records, randomly selected 20 records from each strata, and then hand-coded those records.⁵ At least a portion of records for some ports (San Ysidro, Columbus, El Paso, Naco, Calexico, Douglas, and Nogales) were sorted alphabetically. In these files, we collected a simple random sample of individuals, obtaining 28,454 records of men, women, and children.⁶Figure 2 presents the number of individuals sampled by year alongside U.S. official counts. Our data track the official flow counts reasonably well.

This dataset is representative of the Mexican flow through official border-crossing stations but is not necessarily representative of the entire Mexican flow. Many crossed the unmarked U.S.–Mexico border without presenting themselves to a border inspector.⁷ No reliable data exist for the entire (documented and undocumented) flow of the Mexican population during this time (Kosack and Ward 2020). Estimates of the flow during this period based on various data sources vary dramatically (Cardoso 1980). Our results, therefore, apply to documented migrants who remained in the United States but not necessarily to all Mexican migrants.

Matching Border Records to the 1940 Census

The sample who crossed the Mexican border includes Mexican-born individuals and other immigrant groups, including European, Chinese, and Japanese individuals. We limited our sample to Mexican-born men who arrived at ages 14–55 between 1910 and 1940, yielding a sample of 11,768 individuals. We tracked these individuals to their 1940 U.S. Census record using algorithms developed for historical record linkage (Abramitzky et al. 2021; Ruggles et al. 2017). We focused on individuals 65 or younger in 1940 to target the working population. We restricted the sample to men because of naming practices of individuals and our interest in economic outcomes. First, we relied on given names and surnames to follow individuals over time. Women often change their names at marriage, precluding us from following them across records.⁸ Second, only one third of women in the 1940 census had a recorded occupation (usually “housewife”), and even fewer had a recorded income, preventing us from examining their economic outcomes.

We used a linking procedure that follows standard algorithms to match individuals using their name, age, and birthplace (Abramitzky et al. 2021; Catron 2019, 2020). This technique linked individuals from their border-crossing record to their 1940 census record by first standardizing given names and surnames by correcting for nicknames and then using New York State Identification and Intelligence System phonetic coding to account for alternate spellings and misspellings of names. Observations were first matched using the exact criteria already mentioned. If one unique match was identified, the procedure stopped, and the individual was considered matched. If no match was identified, we tried matching within a one-year age band (older and younger) and then within a two-year age band; if these attempts yielded one unique match, the individual was included in the final sample. However, if the procedure identified multiple matches or no match, the observation was discarded as unmatched.⁹

The matching procedure produced a final sample of 2,832 individuals from the border records. Our match rate is approximately 24%, consistent with prior research using similar data sources (Catron 2019). Table 1 reports differences between the full and matched samples. As with all matched datasets, our matched sample differs from the full border records. Our matched sample individuals were more likely to be younger, of dark complexion, and unskilled at arrival; they were also more likely to have had previous U.S. trips and to have arrived alone. Reasons for not matching include a common name (i.e., José Pérez), name changes, mortality, age heaping, and transcription error. During this period, Mexican immigrants also held one of the highest (voluntary and forced) return migration rates, and some of the flow was characterized by short stays (Sánchez 1993). Short stays were especially prominent among white-collar migrants who would enter the United States to conduct banking or pursue other business interests and then return a few days later. Thus, match rates were lower for this group than for others. This article, therefore, is about Mexican immigrants who settled in the United States long term.

U.S. Border Record Variables

We use individuals’ at-entry characteristics from the border records to predict their later economic attainment. We control for standard measures associated with within-group differences, including dummy variables for whether the migrant had previously traveled to the United States, at-entry literacy, and educational attainment (in the United States or Mexico). Prior experience in the United States is also an important indicator, signaling the accumulation of U.S.-specific knowledge and networks that may yield positive associations with economic outcomes after settlement. To measure premigration occupation, we code the detailed occupational strings using the Historical International Social Class Scheme (HISCLASS) developed by van Leeuwen and Maas (2005) and then further group these codes into five categories: white collar, skilled blue collar, farmers, semiskilled blue collar, and unskilled.¹⁰ These categorizations allow us to analyze broad occupational classifications and their association with later success (Catron 2020).

We also include information on individuals’ perceived skin complexion, which was included in historical immigration records beginning in the late 1890s, largely in response to the influx of southern and eastern Europeans (Perlmann 2018). The U.S. Immigration Bureau recognized that official nationality appearing on passenger lists and identifying passengers as nationals of the multiethnic Austro-Hungarian, Russian, German, and Ottoman empires obscured evaluations of who could be considered “desirable” or “undesirable.” A new tool was therefore implemented in immigration records to deconstruct the White racial category to differentiate race, nationality, and complexion. Given that Mexicans were included in the White racial category because of a treaty agreement following the Mexican–American War, documents on Mexican nationals also included this additional information.

Although our skin complexion variable is an indicator of perceived race, measurement of complexion can take many forms. Objective measures include an individual's skin complexion as a predefined color of lightness and darkness. Subjective measures, which typically rely on interviewers’ perceptions of skin tone, are considered important because how people are treated might not correspond to exact measures of skin pigmentation but rather to skin tone with other racial cues. Debates on the best measure of complexion, whether objective or subjective, have been widespread in previous research (e.g., Abascal and Garcia 2022; Monk 2015; Villarreal 2014). Although we do not weigh in on this debate, we rely on subjective measures because of the nature of the information collected on our sample. However, using subjective measures might attenuate some of our results if different observers attend to different features. For instance, an individual perceived to be light by a border inspector but darker by an employer might have been penalized in the labor market. Because we do not have multiple measures of complexion for these individuals over time, our results might be the lower-bound estimates.

Border inspectors were tasked with filling in an open-ended line on the border records, with subjective answers including light, medium, dark, olive, ruddy, and brown. We code individuals into light, medium, or dark categories on the basis of the inspector's answer. Individuals considered light are those whom the inspector described as light, olive, or fair. Individuals considered dark are those whom the inspector described as dark. Other individuals fall into a catch-all category that includes both those who were medium and those whose classifications were unclassifiable (e.g., brown, sallow). Tables A2 and A3 (online appendix) present results from multinomial logit models and an ordinary least-squares (OLS) regression predicting complexion.¹¹ Correlates with complexion include premigration occupation, arrival decade, and port of entry. Figures A2 and A3 (online appendix) also provide maps of origins for individuals with various complexions. Given that we use a subjective measure, we interpret our findings as evidence of perception of race rather than any specific outcomes of objective skin pigmentation.

In our analyses, we also control for a person's height to rule out differences in immigrant selectivity when assessing at-entry characteristics. Height is correlated with higher earnings, health, physical strength, and other features associated with socioeconomic attainment over time (Steckel 2009). The Mexican flow during this period was positively selected: immigrants were taller than the average Mexican population (Kosack and Ward 2020) and were more likely to be literate and have higher educational attainment (Feliciano 2001). In addition, Mexicans worked in labor-intensive industries, where better physical strength and health would be particularly useful. We also control for port of entry because the distribution of complexion categories varies greatly across ports.¹² Although this variation could be due to underlying differences in the population passing through these ports, it could also reflect port-specific norms used by employees when assessing and categorizing migrants’ skin color (see Figures A4–A6 in the online appendix for comparisons of birth subregion and skin color classifications by port of entry).¹³ Ultimately, however, migrants were mobile after their U.S. entry. Table A4 (online appendix) displays the cross-tabulation for state of settlement and port of entry, showing that patterns of migrant flows through ports were substantially similar regardless of the place of ultimate settlement. In Table A6 (online appendix), we also present descriptive results by arrival decade to show shifts in the flow over the study period.

Our other control variables include additional information taken from the border records: age at arrival, whether the migrant had ever been to the United States, whether the immigrant arrived alone or with family, whether they could read or write, marital status, occupation before arrival, place of birth, and decade of arrival. Because some border records showed no information written in some fields and some writing was illegible, we include dummy variables for blank or illegible fields in all our categories to avoid listwise deletion. In Table A6, we also include a year fixed effect (as opposed to arrival decade) to control indirectly for many factors associated with time-base variation in the migration flow, and we find similar results.

Construction of Income Scores

Using the matched border-crossing–to–census samples, we seek to understand whether border-crossing characteristics were associated with later economic attainment. We leverage the extra information from border records to predict income in 1940. The 1940 census was the first to include information on individual earnings. However, because income was not reported for individuals who were self-employed, including farmers, using only income reported in the 1940 census would omit successful self-employed individuals, potentially overestimating differences between groups. For instance, individuals in our matched sample who were classified as having a dark complexion were more likely to be self-employed and therefore missing reported income, although the difference in self-employment is not statistically significant. Omitting potentially successful self-employed individuals could amplify differences between these two groups. Therefore, we create a predicted (rather than actual) income variable for the individuals observed in the 1940 census (Abramitzky et al. 2021; Escamilla-Guerrero et al. 2021).

To construct our income measure, we use a statistical model to predict income from a rich set of covariates for Mexican men aged 14–65 in the 1940 complete-count census. We regress log income in 1940 on several fixed effects and a complete set of interaction terms using three-digit occupation, age, and 1940 state of residence as our explanatory variables. Following Abramitzky et al. (2021), we account for farmers’ incomes by drawing on farm laborers’ reported income in 1940. We multiply farm laborers’ 1940 income with the ratio of earnings for farmers versus farm laborers in the 1960 census—the first census with a large sample reporting wage, business, and farm income—by region. This measure is similar to the IPUMS occupational income score (OCCSCORE), which is the median income of each occupation, used in prior research (Catron 2019, 2020, 2023). Unlike the OCCSCORE, our measure considers birthplace, age, and place of residence.¹⁴ Results using actual (rather than predicted) income in the 1940 census, shown in Table A7 (online appendix), are similar to those in the main tables.

Estimation Strategy

Our analyses rest on a series of multiple regression models. Each observation is an immigrant from the Mexican border-crossing sample that we linked to the 1940 census.

To consider the relationship between at-entry characteristics and later attainment, we estimate models of the following form:

where the dependent variable $Yi$ represents the logarithm of the income score described earlier. Our main right-side variables are indicators of capital, including literacy, prior U.S. trips, premigration occupation (⁠$OCCUPATIONi$⁠), and skin complexion (⁠$COMPLEXIONi$⁠). The regression also includes a vector of controls (X) that include arrival age and arrival age squared, arrival decade, whether they arrived alone or with others, marital status at arrival, port of entry, and birth region. In some specifications, we also control for height. Our regressions include controls for 1940 variables, such as urban status, education, and whether the person lived in the Southwest. In all regressions, we cluster our standard errors at the county level in 1940.

We complement these analyses by estimating a series of twofold Oaxaca–Blinder decompositions that look deeper into differences in economic attainment between migrants who settled in Texas versus California (Blinder 1973; Oaxaca 1973; for the first presentation of methods for decomposing differences of two social groups on the same outcome, see also Kitagawa 1955). Our decomposition estimates how much of the difference in income score between migrants who settled in California and those who settled in Texas can be explained by group differences in migrants’ observed premigration and postmigration characteristics (the endowment effect) and how much remains unexplained. We include all premigration and postmigration characteristics described in our earlier models as the covariates that capture the endowment effect, although we acknowledge that the unexplained portion could represent potential effects of differences in unobserved or unmeasurable variables. Because the size of these effects depends on which group's coefficients are used as the benchmark to weigh the other group, our main twofold decomposition approach uses the coefficients from a pooled model for both groups as the reference coefficients (Jann 2008; Montenovo et al. 2022). Thus, the difference in the income score between those with a light versus dark skin tone can be expressed as follows:

where $Y¯C$ and $Y¯T$ represent the mean income scores for Mexican immigrants who settled in California and Texas, respectively. $[(X¯C−X¯T)β*]$ is the portion of the difference in income score that is explained by our observed covariates (the endowment effect), where $X¯C$ and $X¯T$ are the means of all covariates for each group, and $β*$ represents the variable coefficients from the pooled model. $[X¯C(βC−β*)+X¯T(β*−βT)]$ is the portion of the gap left unexplained by our covariates, and $(αC−αT)$ is the difference in the intercepts between the two groups.

We estimate additional twofold decompositions, using each group's coefficients as the benchmark to weigh the other group. These decompositions estimate the difference in income score if the immigrants who settled in California had the same characteristics as those who settled in Texas, and vice versa. Overall, the results of both methods are similar to those of the pooled approach. Finally, because decomposition results can vary by the omitted category among categorical variables, we follow recommended procedures of normalizing dummy variable effects (Yun 2005). We conducted these analyses using the oaxaca Stata command (Jann 2008).

Results

We begin by describing the association between at-entry characteristics and later economic attainment. We use the border-crossing characteristics to predict 1940 income scores in the matched sample. We find that having light skin, employment in skilled blue-collar professions, being literate, and previous U.S. experience are associated with better attainment after arrival.

Table 2 presents results from basic regressions controlling for various individual characteristics. The first model controls for skin complexion, arrival age, arrival decade, and port of entry. The results show that having a perceived light complexion is associated with higher earnings vis-à-vis having a perceived dark complexion. As noted earlier, we believe our perceived measures represent an individual's ethnoracial cues and other forms of privilege in education and social class that might yield later (dis)advantages.

The second model controls for premigration occupation, literacy, previous U.S. trips, arrival age, arrival decade, and port of entry. The results show that traditional measures of within-group differences, such as literacy and previous U.S. trips, are positively associated with better outcomes. Literacy is an important proxy for education, and moving back and forth between the United States and Mexico might allow for more labor market experience and more social connections and networks that could facilitate better occupational outcomes. Thus, these first two models are consistent with research on modern immigrants showing that both capital and perceived complexion are associated with long-term outcomes after settlement.

The second model of Table 2 also shows differences between premigration class positions. For instance, skilled blue-collar workers have higher earnings vis-à-vis unskilled migrants. However, white-collar migrants show no statistically discernible difference in earnings from unskilled migrants. The contrasting outcomes of skilled workers and white-collar workers are due to the historical structure of the local labor market and discrimination within it. Although Mexicans earned considerably less than other groups in the Southwest, skilled Mexican workers enjoyed an occupational premium. As the Dillingham Commission noted (S. Rep. No. 633, 1911), skilled Mexican immigrants were able to enter skilled positions “where white men [were] difficult to obtain on account of social and climatic conditions” (S. Rep. No. 633, 1911: pt. I, p. 44). Such opportunities provided higher income for these workers relative to unskilled Mexican immigrants. However, Mexicans were excluded from professional and managerial occupations, where companies actively discriminated against them (García 1982; Laslett 2014). Thus, whereas skilled workers could more easily transfer their skills to the U.S. labor market, white-collar migrants had a more difficult time. In contrast to prior research showing associations between more skill at arrival and better outcomes (Catron 2020), our results suggest that there are limits to skill transferability at the top for some immigrant groups.¹⁵

Model 3 of Table 2 combines Models 1 and 2 and includes other at-entry characteristics from the border records, including birth region, whether the individual arrived with others, and marital status. Results for this model continue to show that perceived light complexion was associated with higher earnings than perceived dark complexion. Those in skilled blue-collar occupations, those who are literate, and those who had made previous U.S. trips also exhibited higher earnings.

One concern with the differences shown in Models 1–3, however, is that individuals coded as light or skilled might have moved to places with a better wage structure. For instance, in their analysis of the 1930 Mexican racial category, Fox and Bloemraad (2015) found that census enumerators were more likely to code individuals as White if they lived outside the Southwest. Additionally, living outside the Southwest confers a stronger wage structure that might explain some differences between groups. Furthermore, individuals living in rural areas generally earn less owing to low wages in agriculture and mining. Thus, Model 4 of Table 2 adds control variables for the individual's urban status in 1940, educational attainment, and whether they lived in the Southwest (Arizona, California, New Mexico, or Texas). Model 5 limits the sample to those living in the Southwest. After adding these controls and sample limits, we continue to see the positive effects of light complexion, literacy, prior U.S. trips (p < .10), and skilled blue-collar occupation.

Finally, immigrant selection might contribute to differences between our categories. In particular, perceived skin complexion variables are susceptible to endogenous factors associated with the individual that may “Whiten” observers’ perceptions of some individuals regardless of their actual ethnoracial cues (Flores and Telles 2012). For instance, observers might evaluate individuals with more money or experiencing upward mobility as Whiter than those with less favorable socioeconomic characteristics (Abascal and Garcia 2022). Border inspectors had access to information about how much money the individual carried, their name, their occupation before arrival, and several other cues that could have changed their perceptions of complexion. Although we already control for occupation and several other socioeconomic and demographic characteristics, we add the person's height in Model 6. We limit this model to men who arrived after age 18 because younger individuals are less likely to have reached their full height. Our results continue to show differences between groups, although not all differences reach traditional levels of statistical significance. Although our data do not allow us to draw a causal link between skin tone and later economic outcomes, the results suggest that skin tone stratification is a feature of the Mexican immigrant experience contemporarily and in the past.

Effects of racialization processes and skill transferability, however, might operate differently at the subnational level. Beyond our findings of effects in the Southwest (Table 2), differences could exist within the Southwest, particularly between Texas and California. Nearly 80% of all Mexicans lived in these two states during the study period. We split the sample into those who settled in Texas and those who settled in California—43.1% and 35.8% of our sample, respectively. We control for all variables from the border records as well as urban status in 1940 from Models 3, 4, and 6 in Table 2, but we report results for only our key variables of interest.

As shown in Figure 3, the effects of complexion are stronger in Texas, where light individuals earned more than dark individuals. Similarly, those in a skilled blue-collar position at arrival had higher earnings vis-à-vis unskilled workers in Texas. By contrast, no statistically substantive differences by complexion are evident for California. Literacy, however, yields positive effects in both California and Texas (although literacy is not statistically significant in Texas when the controls in Models 4 and 6 are added), suggesting that Mexican migrants were able to transfer their skills across labor markets.

The presence of a complexion–income association in Texas but not in California suggests that the racialization of the Mexican population varied by social and political contexts. The ex-Confederacy state of Texas discriminated more, penalizing dark-skinned individuals but rewarding (or at least not penalizing) light-skinned individuals. California entered the United States under a different regime and had a massive demand for labor that not only shifted the complexion hierarchy but flattened it. Comparisons of the mean income scores between these two states suggest that it was better to be dark in California than light in Texas. The unlogged median income scores were significantly lower in Texas (dark = 361.21, light = 444.10, and other = 372.37) than in California (dark = 588.85, light = 566.49, and other = 614.15).

In our final analysis, we decompose the income score differences between immigrants settling in California and those settling in Texas. The pooled model uses the coefficients from a pooled regression in the decomposition. The twofold models apply the coefficients from the opposing state, as is common in discrimination research. All three models allow us to assess how much of the difference is due to observable differences in personal characteristics and how much is due to unexplained social processes. The disaggregated endowment effects for each of our endowment variables are shown in Table A8 (online appendix).

As shown in Table 3, few income differences are explained by observed at-migration characteristics. Roughly 5% of the difference is due to endowment effects, of which roughly half is due to educational differences. No other endowment coefficients significantly contribute to the explained differences. By contrast, nearly all the differences between the two states are unexplained by our observed covariates. As noted earlier, we expect the differences between Texas and California to be more about the contextual configuration rather than individual differences. This interpretation is consistent with an extensive literature suggesting that context can override personal characteristics or at least make them less salient (Portes and Rumbaut 2001). From our findings overall, we argue that the ethnic boundaries were bright in Texas. Thus, skin tone is a more salient feature in Texas than in California, where they were blurrier. These findings are consistent with prior qualitative work.

Discussion and Conclusion

Researchers have often argued that immigrants’ capital at arrival conditions their economic trajectories and ultimate attainment. Immigrants arriving with more skills often fare better in the labor market than immigrants with fewer skills, producing within-group differences. However, premigration characteristics are not the only source of within-group differences; rather, they operate in specific intranational contexts with diverse social, economic, and political characteristics (Alba et al. 2015). The racialized context immigrants enter is one of the key features that can block their upward mobility. Immigrants do not choose the racial and ethnic hierarchy into which they enter, but they are nevertheless subjected to its forces and the mechanisms by which it impacts their economic trajectories. For example, racialization processes often condition immigrants’ skill transferability and attainment of white-collar and high-status jobs (Feliciano et al. 2011; Han 2020; Louie 2012). Similarly, disadvantaged racial and ethnic immigrant groups might experience blocked job mobility if their skills are not recognized (Louie 2012). Indeed, many immigrants arrive with different skin complexions from the native-born, which strongly influences the positions they eventually occupy in the U.S. racial hierarchy. Thus, whether racial and ethnic boundaries in an area are bright or blurry can ultimately shape immigrants’ life chances and opportunity structures beyond the traditional associations between human capital and economic attainment.

In this study, we created a novel dataset of U.S. border-crossing records that we linked to the 1940 complete-count census. The addition of administrative data to census data furnishes information on aspects of applicants not available from the census, such as skin tone, detailed birthplace, intention for entering the United States, and social network information. It also allows us to develop a panel dataset of individuals linking their characteristics at arrival to their characteristics after settlement. To our knowledge, this is the first scholarly effort to collect detailed information on individuals who arrived at the Mexican border during the first decades of the twentieth century and to track them in the United States over time.

In the Age of Mass Migration, Mexicans entered many labor markets that greatly impacted their attainment. These labor markets were part of racialized contexts in which ethnic boundaries and the stratifying effect of skin tone varied across geographies. In places where boundaries were more easily crossed, such as California, immigrants could achieve better socioeconomic attainment, depending on their individual skill levels. Where ethnic boundaries were brighter, such as in Texas, crossing was more difficult for some. In these cases, ascribed characteristics were associated with later attainment. During the Age of Mass Migration, racial and ethnic barriers were formally and informally institutionalized in some places to create a dark-skin penalty. This penalty limited the life chances of those with darker skin, leading to worse outcomes than they otherwise would have experienced. Because the ethnic boundaries varied across time and place for Mexicans during this period, skin tone stratified the population differently across contexts.

We show that dark-skin penalties were large in some areas and not in others. This variation was partly related to the historical development of places and the salience of Whiteness in an area. As noted, California entered the United States as a free state and developed an economy requiring a large labor force that overrode the importance of light skin tones among the Mexican population. Texas, however, entered as an ex-Confederate state and established a regime that placed a burden on individuals with darker skin. We found that the role of skin complexion in the labor market differed between California and Texas. These differences were large enough that individuals with darker skin in California fared better in the labor market than individuals with lighter skin in Texas. We argue that context should be considered in studies of skin tone stratification because it might influence whether ascribed or achieved characteristics are more rewarded.

One issue with these conclusions, however, is our reliance on the 1940 census. A large proportion of the Mexican population returned to Mexico before the 1940 census. The declining economy and native-born individuals’ increased hostility toward Mexicans triggered repatriations, potentially skewing our results if it altered the composition of those who stayed versus left. Historians have argued, however, that “most repatriates were dark complected, and that only the lighter-skinned braceros remained in the United States for extended periods of time . . . the prietos (the dark ones) often had problems because of their skin coloring” (Cardoso 1980: chap. 5). Further, “the first wave of repatriates [in early 1930s were] comprised of Mexicans of modest means who decided to leave before things got worse” (Ngai 2004:73). Because repatriates tended to be darker and less successful, our estimates are likely lower than they would have been in earlier periods, when these individuals still lived in the United States.

Another potential concern with our results is the undocumented immigrant flow during this period. As noted earlier, our results are generalizable only to the legally entering population. The undocumented flow might have had different characteristics than the flow we analyzed. In particular, the main deterrent effect for entering at an official port of entry was an entry fee ($8). Those with fewer resources likely entered without paying this fee. To the extent that skin tone, premigration skills, and mode of entry are correlated, we expect our findings would have been even larger if we had included individuals with fewer resources in our study.

Our findings underscore the importance of analyzing skin tone in myriad ways. The conceptualization of skin tone and its stratifying effect among some populations is rooted in structural and historical terms. Researchers often argue that skin tone stratification among the Mexican population has its origins in colonialism that subordinated people of Indigenous and African origins (Hunter 2005). Research has similarly shown that skin tone is one of the main sources of variation in the incorporation patterns of Mexican immigrants (Telles and Murguia 1990). However, until the present study, researchers analyzing the impact of skin tone have focused only on modern immigrants. Yet, if these structural and historical factors impact immigrants today, we should also expect that these forces shaped immigrants’ experiences in the past. In line with the contemporary literature, we found that dark-skinned Mexicans were sometimes subjected to economic penalties. We argue that skin tone stratification has been a long-term feature of Mexican immigration in the United States.

Acknowledgments

Support for this research came from the University of Washington Center for Studies in Demography and Ecology, whose funders include the Eunice Kennedy Shriver National Institute of Child Health and Human Development (research infrastructure grant P2C HD042828) and multiple units at the University of Washington: the College of Arts & Sciences, Office of Research, eScience Institute, Evans School of Public Policy & Governance, College of Built Environments, School of Public Health, Foster School of Business, and School of Social Work. Other support came from the University of Wisconsin–Madison's Center for Demography and Ecology (NICHD P2C HD047873) and Office of the Vice Chancellor for Research and Graduate Education, with funding from the Wisconsin Alumni Research Foundation. The content is solely the authors’ responsibility and does not necessarily represent the official views of the National Institutes of Health or other funding sources.

Notes

References