A tale of two contexts: U.S. migration and the labor force trajectories of Mexican women

Abstract

Even though women have long participated in Mexico-U.S. migration studies assessing the labor market implications of international mobility for women are rare. Especially lacking are studies that follow a life-course approach and compare employment trajectories across contexts and in connection with other transitions. Using life-history data collected in Mexico and the United States, we explore the impact of migration on women’s employment, focusing on how the determinants of employment vary across contexts. We show that U.S. residence eliminates or even reverses the employment returns to education found in Mexico, and that the constraints imposed on women’s work by marriage are actually stronger in the U.S. context. We also explicitly connect migration to other life-course events, documenting how the impact of context varies not only by marital status but also by where women’s unions were formed.

The sizeable increase in immigration from Mexico to the United States in recent decades (U.S. Census Bureau 2010) has touched off a robust debate about the prospects for the social and economic integration of new arrivals. Labor market outcomes among the immigrant generation are central to this debate, and yet research on employment patterns among immigrant men and women remains uneven. Mexico-U.S. migration has long been disproportionately male, and the presumption that most women were secondary migrants, joining husbands or parents already established in the United States, contributed to their relative neglect in early research on the subject. While several studies have subsequently addressed the initial dearth of labor market analyses of immigrant women (Donato, Wagner, and Patterson 2008; England, Garcia-Beaulieu, and Ross 2004; Flippen 2013; Greenlees and Saenz 1999; Kahn and Whittington 1996; Myers and Cranford 1998; Stier and Tienda 1992), a number of critical gaps remain. Given that roughly 45 percent of adult foreign born Mexicans in the United States are women (Donato, 2010), and the growing importance of women’s earnings to the economic survival of low-income families, greater attention to the forces shaping labor market outcomes among this group is in order.

Theoretical perspectives based primarily on the male experience assert that migration is a form of investment, undertaken to enhance employment opportunities and the economic returns to human capital (Sjaastad 1962). In a related vein, empirical research has shown that for men migration is intimately tied to other life-course events such as marriage and labor market entry (Massey 1987; Parrado 2004; Smith and Thomas 1998). However, research on immigrant Latinas remains more limited in scope. Prior studies document important constraints imposed on employment by immigrant Latinas’ family responsibilities and limited human capital. However, studies of immigrant women that follow a life course approach and connect migration, family, and employment transitions are extremely rare. As a result, our understanding of how immigrant women’s labor market trajectories are shaped by migration, and how the various constraints on market work vary across contexts, has not been adequately explored.

A major impediment to a deeper understanding of the determinants of immigrant Latina’s employment outcomes stems from the lack of longitudinal information that can track labor market experiences before and after migration. Multi-sited data, and the ability to compare migrants and non-migrants, is essential for disentangling the effects of structural and family constraints on market work for immigrant women, and yet is seldom available. In addition, detailed information on the timing of migration as it relates to other life-course events, including marriage, childbearing, and labor market entry, is necessary to provide a comprehensive account of how women’s careers are shaped by migration. This paper links job search and life-cycle perspectives to investigate the role of migration in shaping the labor force trajectories of Mexican women. We focus on two main research objectives. First, we investigate the extent to which women’s labor force participation is affected by context. This entails both assessing the degree to which U.S. residence increases employment opportunities and investigating whether and how the determinants of women’s work, particularly human capital endowments and family characteristics, vary between Mexico and the United States. Second, we take a longitudinal approach to delve into the complex interplay between marriage, migration, and employment decisions. Specifically, we compare the labor market careers of single and married women in Mexico and the United States, emphasizing the extent to which human capital endowments and prior employment experiences differentially affect female employment across contexts.

Theoretical background

Classical economic and sociological approaches view migration as an investment in human capital aimed toward maximizing economic well-being by enhancing employment prospects, wages, or both (Blau and Duncan 1967; Sjaastad 1962). According to this view migration is a life course decision undertaken as part of a job search process to improve labor market returns. In the case of Mexico-U.S. migration the connection between migration, job search processes, and labor market careers has typically been conceptualized from the male perspective. For instance, research has shown that men’s migration tends to coincide with the completion of schooling and search for opportunities at the time of labor market entry (Massey 1987; Lindstrom and Lauster 2001; Parrado 2004). The close connection with job search processes is further demonstrated by the fact that greater attachment to the Mexican labor market lowers the likelihood of U.S. migration. Moreover, these economic considerations also interact with men’s other life-course decisions, as both marriage and childbearing in many cases are the impetus behind men seeking better economic fortunes abroad.

Far less is known about the connection between migration and other life-course events for women. In Mexico, migration has long been a gendered phenomenon (Cohen, Rodriguez, and Fox 2008; Hondagneu-Sotelo 1994), with men frequently migrating in order to provide for their families financially (Broughton 2008) and women more often portrayed as migrating to join husbands or parents already established in the United States (Cerrutti and Massey 2001). Although there is growing recognition of the importance of the migration of young, single women which challenges this characterization (Donato et al. 2008; Donato 2010; Dreby 2010; Morokvasic 1984; Poggio and Woo 2000; Woo and Morales Mena 1999), research on the impact of U.S. migration on the labor market trajectories of Mexican women remains under-developed. Regardless of why women migrate, there is ample reason to expect considerable differences from the male experience in terms of migration outcomes. Participation in the labor force is nearly universal among Mexican men, regardless of whether they live in Mexico or migrate to the United States. Thus, most studies connecting migration and labor market outcomes among men tend to focus on unemployment, occupational attainment, or wages. For Mexican women, on the other hand, labor force participation is far lower and more variable. Thus a first step in assessing how migration shapes economic rewards is to examine the decision of whether or not women work in the first place.²

The salience of labor force participation for understanding the impact of migration on women’s life-course trajectories extends beyond its economic implications. Research in both the United States and Latin America has highlighted that employment is a central factor in gender relations. In Latin America, studies have recognized that increases in female labor force participation throughout the region have re-organized domestic life and enhanced women’s power and authority within the household (Chant 2002). While some have gone so far as to argue that the growth in female employment portends the demise of “traditional” male-dominated family arrangements (Cerrutti and Zenteno 2000), others have pointed out that labor force participation does not immediately translate into greater female autonomy and that much of the increases in women’s paid employment stems from deteriorating economic conditions and the adoption of household economic strategies that do not necessarily correspond with female independence (Chant 2002). Research on immigrant Latinas in the United States strikes a similar tone. On the one hand, the greater opportunities for employment, enhanced protection against domestic violence, and exposure to a more egalitarian gender regime help immigrant women gain on some dimensions of interpersonal power and authority. On the other hand, the highly marginalized position of immigrant Latinos in the U.S. economy and society, particularly when undocumented, poses significant impediments to immigrant women’s attempts to negotiate more egalitarian gender relations (Hondagneu-Sotelo 1994, 2003; Parrado and Flippen 2005; Parrado, Flippen, and McQuiston 2005; Pessar 2005).

To understand how migration shapes women’s work trajectories, it is first necessary to consider the determinants of women’s employment in Mexico. Female labor force participation in Mexico has increased considerably in recent decades from 34 percent in 1990 to 44 percent in 2010. A main factor driving this growth has been improvements in the educational system, as human capital is a strong determinant of women’s work in Mexico. While much of the growth in female employment in recent years has occurred among less educated women, there remains a strong, nearly linear relationship between education and labor force participation. Women with more than a high school education are particularly likely to work outside the home, as are those with prior work experience. Research also demonstrates the considerable constraints on market work imposed by family responsibilities. Both the total number of children and presence of young children, in particular, reduce female labor force participation. The same is true of marriage, over and above its association with childbearing. In fact, research on Mexico has documented that marriage itself is for many women a stronger constraint on female employment than childbearing, suggesting the importance of patriarchy and husbands’ resistance to wives’ work (Cerrutti and Binstock, 2009; Cerrutti and Zenteno, 2000; Garcia and Oliveira, 1994; Parrado 2006; Rojas 2010). Among married women, employment rates are lower among those with employed and higher income husbands (Cerrutti and Zenteno 2000).

Studies of Mexican immigrant women in the United States have documented similar patterns. Mexican immigrant women’s work is responsive to human capital endowments such as education and work experience (Blau and Kahn 2007; Greenlees and Saenz 1999; Stier and Tienda 1992), and those who are married or have children are also significantly less likely to work than their unmarried or childless peers (Flippen 2013 Kahn and Whittington 1996; Read and Cohen 2007). Studies have also shown that the employment of Mexican immigrants increases as they gain U.S. specific human capital through longer durations of U.S. residence and English language ability (Greenlees and Saenz 1999; Kassoudji and Ranney 1984; Livingston 2006).

Comparing employment patterns across contexts clearly shows that U.S. residence is associated with greater female employment. For 2010, Census data show that the labor force participation rate of singe women aged 20 to 65 was 52 percent in Mexico, compared to 66 percent in the United States. The disparities are even larger among married women, among whom 26 and 51 percent were employed in Mexico and the United States, respectively. While a substantial body of work has examined the issue of Mexican women’s employment behavior on one side of the border or the other, research that links the two locales and moves beyond cross-sectional analyses to examine the impact of migration on women’s work trajectories over time remains rare. This is limiting because migration holds the potential to shape Mexican women’s labor market position not only by offering them greater employment opportunities overall (i.e., raising the levels of employment), but also by transforming the structural constraints on employment (i.e., shaping the determinants of employment).

There are several reasons to expect that human capital might have a differential effect on women’s work across Mexico and the United States. Skills are often not directly transferrable across contexts, and while educational attainment has been rising steadily in Mexico, it remains lower, on average, than in the United States. Thus in relative terms even women who are highly educated in Mexico tend to be far further down the educational distribution in the United States. Moreover, the structure of the U.S. labor market also has a profound effect on the returns to education for immigrant women. The U.S. labor market has grown increasingly bifurcated and segmented in recent decades, and nativity and legal status have become increasingly important determinants of the consignment of workers across segments (Kalleberg 2011; Hudson 2007; Massey and Gelat 2010; Sassen 1984). Immigrant Mexican women tend to be overwhelmingly concentrated in cleaning, domestic, and food service jobs (Flippen 2013) where the returns to education may be muted. A main implication is that while the U.S. context expands employment opportunities for poorly educated women it might not provide the higher quality jobs that educated workers seek.

In a related vein, the constraints imposed on market labor by family responsibilities could also vary across contexts. Studies of U.S. immigrant women’s labor supply generally note that they work at lower rates than non-Hispanic white and black women (Read and Cohen 2007). Moreover, the negative effect of marriage on women’s work is generally not present for native-born women, suggesting that cultural factors, particularly the greater adherence to patriarchy, may be an important factor suppressing employment among immigrant women (Kahn and Whittington 1996). However, without examining the effect of family structure on employment across settings it is difficult to draw conclusions. On the one hand, the greater employment opportunities in the U.S. context could offer women more leverage in their interpersonal relationships, weakening the impact of marriage and childbearing on employment. On the other hand, immigrant women are cut off in many ways from social support and are in a far more precarious position than their counterparts in Mexico (Parrado and Flippen 2005), which could potentially make it more difficult for them to balance work and family in the United States than in their home communities.

Finally, the life course approach suggests that women’s various social positions intersect with one another and that previous life-course events shape the impact of subsequent events. This requires greater attention to work trajectories over time as opposed to in a given cross-section. For instance, a comprehensive account of the employment trajectories of immigrant and non-immigrant women should investigate not only differences between single and married women, but also take into consideration whether transitions occurred in Mexico or the United States. For all these dimensions, multi-sited longitudinal data that can reconstruct the life-course of Mexican migrants and non-migrants is necessary. Our aim is to sequentially investigate different dimensions of Mexican women’s labor market trajectories to construct a more comprehensive account of the connection between international migration and female employment.

Cross-national data

We investigate these issues with data from the Gender, Migration, and Health among Hispanics study.³ The project applied multi-sited data collection to explore the impact of migration on gender roles, health behaviors, and employment patterns, among other factors. By comparing the experiences of Mexican immigrants in a newly emerging area of destination in the United States with those of non-migrants in communities of origin, we are able to gain insight into how changes in context affect social behavior. In particular, the comparison of residents of sending and receiving areas helps disentangle the role of personal resources from structural factors and enhances our understanding of how the constraints on female employment are altered by migration.

The U.S. setting: Durham, North Carolina

The data for immigrants in the United States was collected in an emerging immigrant receiving context, namely the Durham, Chapel Hill, and Carrboro metropolitan area of North Carolina (for the sake of expediency referred to as “Durham,” where the majority of respondents lived). Durham offers a valuable vantage point from which to explore the economic implications of migration. Growth of the high-tech sector during the 1990s spurred a boom in business and residential building, heightening demand for construction and other semi-skilled laborers, as well as for domestic work and other service employment for the growing class of professionals in the area. The result of these forces was dramatic. In 1990, less than 2,000 foreign-born Latinos were residing in the area, but by 2010 the number had reached close to 40,000, or 12 percent of the total population (Flippen and Parrado 2012).

The highly marginalized position of immigrants in contemporary U.S. society presented unique challenges to collecting a locally representative sample of the Latino immigrant population of Durham. The study relied on Community Based Participatory Research (CBPR) to help overcome these difficulties. CBPR is a participatory approach to research that incorporates members of the target community in all phases of the research process. In our case, a group of 14 community members assisted in the planning phase of the study, survey construction and revision, and devising strategies to boost response rates and data quality. In addition, CBPR members were trained in research methods and conducted all surveys. Finally, through ongoing collaborative meetings, they were also influential in the interpretation of survey results and giving culturally grounded meaning to the findings.

In addition, the relatively recent nature of the Hispanic community in Durham rendered simple random sampling inefficient. We therefore employed targeted random sampling techniques. Based on extensive fieldwork and CBPR input, we identified 49 apartment complexes and blocks that house large numbers of immigrant Hispanics. We then collected a census of all the apartments in these areas and randomly selected individual units to be visited by interviewers. All foreign-born Hispanics between the ages of 18 and 45 were eligible for interview. Although our survey may have been less likely to capture established immigrants, this method was far superior to nonrandom methods of recruitment such as snowball or convenience sampling. To evaluate potential bias arising from targeted random sampling, we compared our sample with data from the 2000 Census. The results show no statistically significant differences in main socio-demographic characteristics such as age, education, employment status, wages, and time in the United States.

The Mexico setting: Immigrant sending municipalities

In order to construct a transnational data set the project collected parallel surveys in sending municipalities in Mexico. To do so, our first task was to obtain a sample of communities in Mexico representative of the communities of origin of the Mexican population in Durham. A list of migrants’ places of origin showed that 20 communities provided approximately one-third of the Mexican population of Durham. From these 20 sending places and based on information obtained from the 2000 Mexican Census, eight communities were purposively selected to reflect different population sizes, economic activities, level of education, and geographic areas of Mexico. The selected communities include: Celaya, Guanajuato; Ciudad Altamirano, Guerrero; Cutzamala de Piñon, Guerrero; Pachuca, Hidalgo; Maravatío, Michoacán; Tacámbaro, Michoacán; Pahuatlán, Puebla; and Playa Vicente, Veracruz. While our sample of communities does not capture the full universe of migrant sending communities in Mexico, they constitute a reasonable representation of the typical places sending migrants to Durham. We also evaluated the characteristics of men and women in our communities against data from the Mexican Migration Project that contains a much larger sample of Mexican communities. The results show no statistically significant differences in main socio-demographic characteristics such as age, education, and employment status (self-identifying reference).

Analytic strategy and model specification

In both Durham and Mexico data was collected in two phases, the first in 2002-03 and the second in 2006-07. The surveys were independently drawn and the samples pooled for the statistical analysis. While the larger project collected data from men and women, and from Latin American immigrants of any national origin, the current analysis is restricted to the 681 women of Mexican origin in Durham and the 800 women interviewed in Mexican sending communities.

Both the Durham and Mexican surveys collected retrospective yearly information for central life-course transitions, including union formation and dissolution, fertility, employment, and migration. We also collected basic demographic and socioeconomic information about respondents and their spouses. In situations where multiple events occurred in the same year we were especially careful to register the order of events and whether they occurred in the United States or Mexico, which allows us to precisely specify the temporal sequence and location of events. Based on the retrospective information we construct a person-year, longitudinal data file that tracks women as they age in Mexico and the United States. By pooling samples, we obtain person-years for all women for the time they spent in Mexico, with additional person-years in the United States for migrant women. This approach confers a number of advantages. First, we are able to directly examine the intersection of different life-source domains by dating migration, employment and family-related transitions. Second, for migrant women we can also construct their labor market trajectories in Mexico before migrating, enhancing our understanding of the impact of migration on work trajectories as well as the comparison of migrant and non-migrant experiences. Since the U.S. sample can be considered as an oversample of the migrant population from sending communities, the descriptive analysis reports results separately for the data collected in Mexico and the United States. Following survey research methods for dealing with oversampling, the statistical models control for this design effect by including a dummy variable indexing whether the data was collected in the U.S. or Mexico (Winship and Radbill 1994).

Our statistical analysis applies the methods developed for categorical longitudinal data analysis of binary outcomes (Liang and Zeger 1986; Pettit et al. 2002) to disentangle the interplay between context and family constraints in structuring Mexican women’s employment decisions. The analysis is divided into two main parts. The first part focuses on differences in labor force participation rates in the United States and Mexico, as well as how the determinants of employment vary across contexts. We thus first estimate a binary logistic regression model predicting the likelihood of women working in a given year, based on self-reports of employment which includes both part- and full-time work. The model is estimated using generalized estimating equations (GEE) to account for the correlation among repeated observations obtained for the same individual typical of longitudinal data. The GEE approach provides population-average, sometimes called marginal effects, of the explanatory variables on the likelihood of, in our case, employment. For the purposes of interpretation, coefficient estimates in GEE models describe changes in the population mean given changes in covariates while accounting for the lack of independence among observations obtained for the same individual. This is particularly appropriate in our case since we are primarily interested in the role of context in affecting female employment (Hu et al. 1998; Hubbard et al. 2010; Neuhaus 1991).⁴ Several correlation structures can be assumed in the GEE specification. Parameter estimates and covariance matrices are consistently estimated even if the correlation structure is miss-specified (Allison 1995). For the purposes of our analysis we assume an exchangeable correlation structure that expects every observation within an individual to be equally correlated with every other observation from that individual.

In addition to age, the model includes as predictors educational attainment measured as a set of dummy variables that follow the Mexican educational system (0–5, 6–8, 9–11, and 12 or more years of education) to capture the effect of human capital endowments on employment. In addition, we assess the role of family constraints by incorporating an indicator of whether the woman is married and the number of children in the prior year. Finally, a dummy variable indicating U.S. residence captures the effect of context on female employment. To investigate the interaction between context and human capital and family determinants we also estimate the models separately for person-years spent residing in Mexico and Durham. The comparison of coefficients across settings allows us to elaborate on the role of context in modifying the determinants of women’s labor supply.

Since there are a number of conditions that shape Mexican women’s employment patterns that are unique to the U.S. labor market, we also estimate a discrete time event history model that follows women after arrival to the United States. Thus, instead of age it is years in the United States that is the time variable affecting work. In addition to the prior indicators of educational attainment and family constraints, the model incorporates controls for women’s age at arrival and pre-migration statuses, namely if a woman was married or worked in Mexico before moving to the United States. We expect these prior experiences to affect the role of U.S. context on employment.

The second part of the analysis delves more in depth into the interplay between marriage, migration, and employment. The impact of migration, and the change in context it represents, is likely to vary between single and married women. Likewise, the constraints imposed on market work could be different for women who married prior to migration than for those who migrated single and subsequently married. To better illuminate the intersection of these multiple life-course domains, we approach the issue from two directions – from the perspective of migration and from the perspective of marriage. Thus, we track employment patterns in the years before and after migration, separately for married and single women. We do so relying on the information collected for migrant women and constructing their pre- and post-migration labor market trajectories. We then track employment patterns in the years before and after marriage, separately for women who married in Mexico and those who married in the United States. Finally, we examine these issues together in a multivariate framework by modeling the determinants of employment during the years that women were single, and during the years that they were married. That is, we first estimate a population-averaged logistic regression GEE model predicting single women’s labor force participation according to human capital endowment and U.S. residence. We then estimate separate models for person years spent single and in Mexico and the United States and compare results. We then investigate women’s labor market trajectories in the years after marriage. In addition to educational attainment, number of children, and U.S. residence, the models include measures of time in union, age at marriage, and whether the respondent worked before marriage. As in the prior analysis the models are estimated separately for person-years in the United States and Mexico.

Results

Assessing the impact of context on Mexican women’s employment

As a starting point for the analysis, Figure 1 plots the share of women working by age in the United States and Mexico. The salience of context for women’s work is clearly evident. At all ages Mexican immigrants in the United States are considerably more likely to work than non-migrant women in Mexico. In fact, immigrant women in Durham average 20 percentage points higher employment rates than women in Mexico. Moreover, while just under half of women worked outside the home in Mexico with little variation across adult ages, women in Durham show a more pronounced age pattern to employment; in Durham, employment rates rise from 62 percent at age 20 to nearly 68 by 24, show a dip between ages 25 to 28 (most likely due to childbearing), and then rise again after age 28.

Figure 1.

Percent of women working across contexts by age

Of course, migrant and non-migrant women are likely to differ in important ways that could potentially shape the differentials evident in Figure 1. Likewise, context could shape not only the absolute level of employment among Mexican women, but also influence how human capital and family characteristics structure employment decisions as well. Accordingly, drawing on our longitudinal data, Table 1 presents estimates from population-averaged logistic regression GEE models predicting the likelihood of women working in a particular year. The first column reports estimates from the pooled sample that includes women in Mexico and Durham, while columns 2 and 3 report estimates for person years in Mexico and Durham, respectively. Finally, column 4 expands the estimates obtained for Durham by adding migration-specific characteristics not applicable to women in Mexico.

Table 1.

Coefficients from population-averaged logistic regression GEE model predicting likelihood of Mexican women working in a given year across contexts (s.e. in parenthesis)

	All	Mexico	U.S.
Intercept	−1.469** (0.355)	−1.832** (0.358)	2.313 ** (0.895)	1.211 ** (0.509)
Age/Years in U.S.1	0.113** (0.027)	0.131 ** (0.027)	−0.014 (0.054)	−0.044 (0.026)
Age/Years in U.S. Squared	−0.002 ** (0.001)	−0.002 ** (0.001)	0.000 (0.001)	0.002 (0.001)
Educational attainment (reference= 0–5 years)
6–8	0.091 (0.129)	0.117 (0.145)	−0.005 (0.206)	0.100 (0.201)
9–11	0.107 (0.129)	0.168 (0.143)	−0.020 (0.217)	0.119 (0.213)
12+	0.456 ** (0.149)	0.653 ** (0.159)	−0.320 * (0.223)	−0.242 * (0.122)
Statuses in prior year
Married	−0.743 ** (0.081)	−0.608 ** (0.091)	−0.920 ** (0.123)	−0.874 ** (0.123)
Number of children	−0.080 ** (0.037)	−0.080 ** (0.041)	−0.126 ** (0.053)	−0.087 * (0.056)
Residence status in a given year
In the U.S.	1.251 ** (0.114)
Migration-specfic characteristics
Age at arrival				0.039 ** (0.014)
Pre-migration statuses
Married in Mexico				−0.305 * (0.178)
Worked in Mexico				0.324 ** (0.143)
Sampling design control
Data collected in the U.S.	−0.313 ** (0.107)	−0.266 ** (0.098)	−0.673 ** (0.383)	−0.663 * (0.389)
Quasi-Akaike Information Criteria (QIC)	29831.8	23658.8	6203.3	6141.0
Person-years	22,785	17,764	5,021	5,021

^*p<.10

^**p<.05

¹The duration variable is age for the pooled and Mexican samples and years in the U.S. for the U.S. sample

Results from the pooled model confirm the dramatic effect of U.S. residence on female labor force participation. Overall, Mexican women residing in the United States are 3.5 times (exp(1.251)=3.49)) more likely to be employed than if they were residing in Mexico. As in previous studies, the likelihood of women working is strongly connected to human capital and family characteristics. Female labor force participation is an age-graded phenomenon; it increases at younger ages, peaks at middle age, and then declines. Human capital is also an important determinant of employment, with better educated women more likely to work for pay than their less educated peers. Family, on the other hand, is a clear constraint on women’s work, with both marriage and childbearing negatively associated with employment.

When we examine the same models run separately according to residence in Mexico or the U,S., however, we see important differences in the impact of human capital and family characteristics across contexts. Specifically, both the age and education patterns clearly evident in the pooled sample hold for Mexico but not the United States. In Mexico, for instance, women with more than 12 years of education are on average twice (exp(.653)=1.9)) as likely to work as women with less than five years of education. In the United States, on the other hand, the effect of high levels of educational attainment actually reverses. Contrary to Mexico, results show that women with 12 or more years of education are less likely to work (−0.320) than their peers with 0–5 years of schooling.

There are two ways to interpret these differences in the effect of age and education across contexts. The optimistic interpretation would be that employment opportunities for women are so much better in the United States that even those whose age and low level of education would have undermined their work effort in Mexico are able to find employment. A perhaps more realistic explanation, on the other hand, would be that financial insecurity (including the need to provide for family members in sending communities) encourages U.S. employment at all ages, while the difficulties that immigrant women face in transferring their educational qualifications to the U.S. context mute the otherwise greater incentive that more educated women have to work outside the home. Immigrant Mexican women tend to be highly concentrated in a handful of low-skilled occupational niches where the returns to education are slight (Flippen 2013). These jobs are likely to represent considerable downward mobility for more educated women, discouraging their employment relative to their less educated peers.

Estimates also show interesting differences across contexts in the constraints on employment posed by family responsibilities. Specifically, while the presence of young children in the household depresses female labor force participation in both Mexico and the United States, the negative effect of marital status on women’s work is significantly stronger in the U.S. context relative to Mexico. Married women in Mexico are on average only 0.54 times (exp(−0.608)) as likely to work as their single counterparts. In the United States they are only 0.40 times (exp(−0.920)) as likely to work. The finding that the negative association between marriage and labor force participation is stronger in the U.S. setting than in Mexico, after accounting for context and other personal characteristics, could be a reflection of the weakened position that married Mexican women face after migration, and the lack of institutional and family support that could allow them to reduce family constraints (Parrado and Flippen 2005).

Finally, U.S. labor force participation is affected by a number of factors that are not applicable to women residing in Mexico. We thus also estimate a model predicting the likelihood of women working in a particular year after U.S. arrival (column 4). Results show that older ages at migration are associated with greater labor force participation. Women who migrate at young ages often do so as a result of decisions taken by other family members, and they may be less motivated to work outside the home than their counterparts who made a more purposive decision to migrate at an older age. Duration of U.S. residence, on the other hand, does not significantly predict employment. The labor market position of immigrant women is also affected by events taking place prior to migration; women who married in Mexico and migrated either with or following their husbands are on average only 0.74 times (exp(−0.305)) as likely to work as women who married in the United States. This suggests that there are important underlying differences between women who migrate single and those who migrate married that endure even after the former marry in the United States. At the same time, those with work experience in Mexico are 1.4 times (exp(0.324)) more likely to work than Mexican immigrants who did not work in Mexico. Thus, prior orientations towards the labor market also affect the labor force experience of women after they migrate.

Context and women’s labor market experience by marital status

While the dramatic differences in employment rates across contexts are informative, they fail to capture the complex interplay between marriage, migration, and employment decisions. To illustrate, we first examine the relationship from the perspective of migration, demonstrating how employment trajectories are altered by the act of migration, separately for married and single women. Figure 2 draws from the experience of migrant women (and is thus restricted to the Durham sample, as few women surveyed in Mexico had prior migration experience) and reports their labor force participation rates during the years before and after migration, irrespective of age. Results directly link changing context to employment opportunities. Among all immigrant women, labor force participation increases from 40 to almost 70 percent after migration from Mexico to the United States. However, the change is even more pronounced for women who migrated single. While marital status at migration did not distinguish employment rates prior to migration, immediately after arrival women who migrated single worked at much higher rate than those who migrated married (79 vs. 65 percent). Moreover, labor force participation rates continued to climb steadily in the years after arrival for single migrants (reaching 85 percent only 5 years after migration), while they fell steadily for married migrants (dropping to 55 percent after 7 years in the United States).

Figure 2.

Percent of women working before and after migration

Thus while context is clearly important to employment opportunities for Mexican women, the impact of context is at least partly dependent on family constraints, especially marriage. Marriage tends to disrupt women’s labor market trajectories but the extent of the disruption may differ between Mexico and the United States. To illustrate these patterns, and look at the interplay between marriage, migration, and employment from the perspective of marriage, Figure 3 pools the Durham and Mexico samples and plots employment rates before and after marriage according to whether the union was formed in Mexico or the United States. Among women who formed a union in Mexico results show that marriage represents a considerable, though largely temporary, setback to employment. That is, the years leading up to marriage employment rates rise substantially, from 40 to nearly 50 percent. Employment rates begin to fall in the year immediately prior to marriage, and fall even more sharply in the year after marriage, to 38 percent. After two years of marriage, however, women’s labor force participation rates begin to rebound, though even after 10 years of marriage they do not quite reach the levels attained prior to union formation.

Figure 3.

Percent of women working before and after marriage across contexts

A very different pattern is evident among women who migrated single and married in the United States. Consistent with the previous figures, employment rates among single Mexican women in the United States are quite high, exceeding 80 percent in the years prior to marriage. However, even in the U.S. context marriage represents a clear disruption to women’s market work. Employment rates fall gradually in the years leading up to marriage, and then fall sharply, by a full 18 percentage points (from 68 to 50 percent) during the five years following marriage. It is not until after five years of marriage that employment rates begin to rise again among these women, and like their counterparts married in Mexico, they do not attain their pre-marriage employment levels even 10 years after union formation. It is worth noting that while women in the United States are consistently more likely to work than those in Mexico, once again the disruptive effect of marriage on employment is actually larger in the U.S. context.

While these figures are informative and begin to illuminate the complexity of the interplay between marriage and migration timing, on the hand, and employment outcomes, on the other, they are potentially driven by demographic and human capital differences across groups. We therefore next examine the factors that predict work before and after marriage for all women in a multivariate setting. Specifically, we pool women in the Mexican and Durham samples and estimate the impact of demographic and human capital characteristics first during all person-years that women spent single (presented in Table 2) and then for all person-years that women spent married (presented in Table 3). Thus, Table 2 reports results from population-averaged logistic regression GEE models predicting the likelihood of labor force participation among women who were single in a given year. The Mexican sample is restricted to the person-years occurring before both marriage and migration to reflect the experience of single women in Mexico, while the Durham sample is restricted to person-years occurring before marriage, to reflect the experience of single women in the U.S. The comparison captures differences in the experiences of single women across contexts.

Table 2.

Coefficients from population-averaged logistic regression GEE model predicting the likelihood of single Mexican women working in a given year across contexts (s.e. in parenthesis)

	Both	Mexico	U.S
Intercept	−4.671 ** (0.581)	−4.599 ** (0.587)	−0.679 (1.390)	1.290 (0.803)
Age/Years in U.S.1	0.339 ** (0.049)	0.336 ** (0.050)	0.105 (0.140)	0.109 (0.138)
Age/Yrs. In U.S. Squared	−0.005 ** (0.001)	−0.005 ** (0.001)	−0.005 (0.014)	−0.005 (0.014)
Educational attainment (reference= 0–5 years)
6–8	0.168 (0.191)	0.110 (0.175)	−0.547 (0.610)	−0.507 (0.610)
9–11	0.152 (0.179)	0.071 (0.175)	0.112 (0.633)	0.092 (0.637)
12+	0.045 (0.204)	0.033 (0.200)	−0.684 (0.611)	−0.598 (0.602)
Residence status in a given year
In the U.S.	1.476 ** (0.277)
Age at migration			0.106** (0.050)
Worked in Mexico				0.599 * (0.384)
Sampling design control
Data collected in the U.S.	−0.248 ** (0.126)	−0.279 ** (0.112)	0.282 (0.533)	0.129 (0.527)
QIC	8610.1	8072.3	547.5	537.3
Person- years	6,851	6,250	601	601

^*p<.10

^**p<.05

¹The duration variable is age for the pooled and Mexican samples and years in the U.S. for the U.S. sample

Sample restricted to person-years before marriage

The U.S. sample restricted to person-years after migration

Table 3.

Coefficents from population-averaged logistic regression GEE model predicting the likelihood of post-marital work in a given year among Mexican women across contexts (s.e. in parenthesis)

	Both	Mexico	U.S.
Intercept	0.235 (0.349)	0.201 (0.425)	1.452** (0.732)
Years in Union	0.014 (0.020)	0.044** (0.021)	−0.034 (0.033)
Years in Union Squared	0.000 (0.001)	−0.001 * (0.001)	0.001 * (0.001)
Age at Marriage	−0.079 ** (0.018)	−0.097 ** (0.022)	−0.070 ** (0.025)
Educational attainment (reference= 0–5 years)
6–8	0.097 (0.191)	0.385 ** (0.189)	0.073 (0.225)
9–11	0.214 (0.186)	0.483 ** (0.189)	0.094 (0.235)
12+	0.817 ** (0.223)	1.346** (0.211)	−0.119 (0.251)
Number of children	−0.016 (0.047)	−0.008 (0.053)	−0.119** (0.065)
Worked before marriage	1.102** (0.129)	1.157** (0.131)	0.713 ** (0.161)
Residence status in a given year
In the U.S.	1.305** (0.149)
Age at migration			0.057 ** (0.019)
Sampling design control
Data collected in the U.S.	−0.030 (0.165)	−0.032 (0.129)	−1.040* (0.591)
Quasi-Akaike Information Criteria (QIC)	18893.2	13648.0	4670.5
Person- years	14,714	10,631	3,612

^*p<.10

^**p<.05

U.S. sample restricted to person-years after marriage among women who formed unions in the U.S.

Results from the pooled sample show that U.S. residence dramatically increases single women’s labor force participation. The likelihood of an average single woman being in the labor force is 4.4 times (exp(1.476)) higher in the United States than in Mexico. The effect is so strong that it results in nearly universal employment among single women in the United States. Educational characteristics do not explain variation in single women’s labor force participation in either the United States or Mexico. Later ages at migration, on the other hand, are associated with higher levels of U.S. labor force participation. Again, this finding likely reflects differences in the impetus behind migration between single and married women. In addition, having worked in Mexico before migrating also is positively associated with U.S. employment. Unlike educational attainment, prior work experience appears to be a dimension of labor market attachment that does transfer to the U.S. context.

Table 3 presents results from a similar model for the years after marriage. The Mexican sample is restricted to person-years after marriage and before migration, while the Durham sample is restricted to person-years after marriage for women who formed a union in the United States. This model demonstrates how the impact of demographic and human capital characteristics on employment differ for married women across contexts. As with single women the likelihood of employment is dramatically higher in the United States than in Mexico; an average married woman is 3.7 times (exp(1.102)) more likely to work in the United States than in Mexico. Nevertheless, the labor force gains arising from U.S. residence are smaller than those observed for single women. Results also show important contextual differences in the likelihood of working according to marital duration. Overall, married women’s employment increases during the early years of marriage and then declines. These results are mirrored in the Mexican sample. In the United States, in contrast, the employment of married women decreases slightly during the first year of marriage but then increases steadily, which is consistent with the image of a more monotonic attachment to the labor market.

In all samples, labor force participation is affected by age at marriage with women who marry at older ages being less likely to work than women who marry younger. The negative effect, though, is statistically larger in Mexico than in the United States. Arguably the most salient result is once again the differential return to educational attainment across contexts. Educational attainment is a strong predictor of married women’s employment in Mexico; on average married women with 12 or more years of education in Mexico are 4 times more likely to work than married women with 0 to 5 years of education. In the U.S. context, in contrast, the employment returns to education are completely absent.

Results again support the notion that net of differences in the level of labor force participation family constraints are stronger in the U.S. than Mexico. While on average the number of children decreases married women’s employment in the U.S. (−0.119) the association is not present in Mexico. Results also show that labor market attachment is affected by pre-marital experiences. Women who worked before marriage are more likely to work after marriage. The effect is actually stronger in Mexico than in the United States, which is again consistent with the image of less variability in labor force participation among women in the U.S. setting relative to Mexico.

Conclusion

Two major themes in migration research link immigration to both other life-course events and to the search for employment opportunities. In the case of Mexico-U.S. migration the study of these connections has been dominated by the male experience. This is particularly the case for labor market outcomes where the construction of Mexican women as secondary migrants has contributed to their relative invisibility as workers and job search agents. While many studies have investigated labor force processes among Mexican women on one side of the border or the other, multi-sited research capable of connecting life-course trajectories at origin and destination are rare. A central limitation is that the implications of Mexican women’s migration for job search outcomes and their relation to human capital and family structures are not well documented. In this paper we draw on multi-sited retrospective longitudinal data to connect migration to the employment trajectories of Mexican women in Mexico and the United States. We address two central concerns. The first is the role of context in shaping Mexican women’s employment and how the effect of human capital and family characteristics vary between Mexico and the United States. The second elaborates on an event-oriented approach that intersects marital status and migration to investigate how the labor market trajectories of single and married women vary by location.

While the view of Mexican women as secondary migrants continues to be contested and debated, our results document that, irrespective of their motivation to migrate, Mexican women’s labor force participation is markedly higher in the United States than in Mexico. For all types of women and modes of entry into the United States, Mexican women’s employment trajectories are powerfully shaped by migration. This is not to say that women are impacted by migration uniformly. The increase in employment is particularly pronounced among single women, whose employment is nearly universal in the U.S. setting. Among married women the effect is less pronounced, although their labor force participation rate increases from 40 to 65 percent after migration.

A central finding of the multi-sited comparison is that the employment gains resulting from migration vary systematically according to women’s educational endowments. The female labor market in Mexico is highly stratified by education, and better educated women are more likely to work than their less educated peers. U.S. migration, however, eliminates educational differentials in employment and in some instances even reverses them; women with the highest level of educational attainment are actually less likely than those with very low levels to work for pay. The interpretation of this change relates to the demand for immigrant labor in more developed countries. On the optimistic side, results could be interpreted as showing that the U.S. labor market offers women employment opportunities irrespective of their educational credentials. More realistically, it is likely that the difference reflects immigrant Latinas’ concentration in the low-skill service sector, where there are few barriers to entry for less educated workers but also few opportunities for advancement. Thus women with relatively high levels of education must either accept jobs that are of lower standing than what they could have obtained in Mexico or eschew employment entirely.

The role of family structure, especially marriage, in constraining female employment also shows considerable variation across contexts, contradicting the common view of the impact of migration on gender relations as liberating. Our results show that after accounting for differences in the level of women’s labor force participation across contexts, being married is a stronger constraint on female employment in the United States than in Mexico. While the effect is conditional on the increase in work opportunities resulting from migration, it is nevertheless sizeable. Drawing on the coefficients from Table 1 and taking single women as the reference and assuming that they work at the same rate, namely 70 percent, in Mexico and the United States, after accounting for human capital characteristics being married would reduce women’s employment to 52 percent in Mexico and further still, to 40 percent, in United States. Of course since the overall level of employment is so much higher in the United States than in Mexico, married women in the United States are still more likely to work than their Mexican counterparts (65 vs. 46 percent). But because the boost in employment is even greater among single women (87 vs. 65 percent) the gap between married and single women in greater in the United States than in Mexico. Again, we interpret the difference in terms of the characteristics of the U.S. labor market. In the context of a highly segmented labor market, where Mexican immigrants are concentrated in low-skilled occupations, being employed is often less a liberalizing experience than a response to acute economic need. These patterns also undermine the argument that culture is the primary barrier to female labor force participation among immigrant women in the United States. In tandem with research that shows that married women’s employment is more sensitive to legal status than that of single women (Flippen 2013), these findings highlight the structural underpinnings of family behavior and its impact on women’s work.

Results also document a complex relationship between marriage and migration timing and employment outcomes. Women who follow the stereotypical pattern and migrate after marriage work in greater numbers than their peers who remained in Mexico, but clearly get a smaller employment boost from migration than women who migrated single and subsequently married in the United States. Indeed, the retrospective data show that the impact of marriage on women’s work trajectories plays out very differently across contexts. In Mexico, labor force participation during women’s years spent single is largely determined by age and education. In the United States, in contrast, employment during women’s unmarried years is nearly universal, and does not vary significantly with either age or education. The only dimension increasing U.S. labor force participation is having worked in Mexico prior to migration. For the years spent married, the patterns are very different. For married women, employment trajectories vary widely across contexts. As with single women, U.S. residence significantly increases the employment of married women, although the gains are 40 percent lower than among single women. Married women in Mexico experience considerable disruption to their work careers. However, their employment increases with education and marital duration. In the United States, in contrast, neither educational attainment nor marital duration affects the work experience of married women. Likewise, having worked before marriage also increases labor force participation after marriage but the effect is weaker in the United States than in Mexico. These patterns reinforce prior findings on the lower employment returns to human capital in the United States compared to Mexico.

Overall, the findings highlight the dramatic transformation in the work trajectories of women and families arising from migration. In migrating to the United States, Mexican women become important income-generating agents. This has implications for their own personal well-being and independence but also for the social organization of immigrant families. Dual earner families diversify risks and are better able to cope with short term disruptions to employment of either spouse. Given concerns over the economic incorporation of Mexican immigrants in the United States more attention to the transformations in income generating activities and the life-course trajectories of immigrant women is in order. Multi-sited approaches that can relate and compare experiences before and after migration, as well as with non-migrants at origin hold the potential to expand our understanding of these transformations.

However, multi-sited approaches, especially in the case of longitudinal designs also present challenges and limitations. The almost inevitable reliance on retrospective information, given the difficulties of cross-sited prospective designs, implies that more subtle dimensions of individual behaviors cannot be captured. For instance, it is well documented that living arrangements, social support for childcare, and local employment conditions all affect female labor force participation. These dimensions are difficult to capture retrospectively, limiting the analysis to how major statuses, such as educational attainment or marriage, shape employment. Likewise, variation in work effort, or differences between part- and full-time work, are difficult to capture retrospectively. Finally, meaningful cross-sited understandings of behaviors should recognize that the factors affecting outcomes might differ across locales. In our case, major characteristics affecting employment at destination, such as time in the U.S. or whether women worked before migration, have no meaning in the Mexican context. The implication is that a comprehensive analysis of women’s employment at destination will not always allow for a comparative understanding of its predictors. From a broader research stand point though, the main implication is that, rather than a narrow either/or position, it is important to recognize that different approaches complement each other and that a comprehensive account of immigrants’ experiences requires integrating different designs.