Gendered context of assimilation: the female second-generation advantage among Latinos

Transcription

1 Journal of Ethnic and Migration Studies ISSN: X (Print) (Online) Journal homepage: Gendered context of assimilation: the female second-generation advantage among Latinos Nicol M. Valdez & Van C. Tran To cite this article: Nicol M. Valdez & Van C. Tran (2019): Gendered context of assimilation: the female second-generation advantage among Latinos, Journal of Ethnic and Migration Studies To link to this article: Published online: 24 Jan Submit your article to this journal View Crossmark data Full Terms & Conditions of access and use can be found at

2 JOURNAL OF ETHNIC AND MIGRATION STUDIES Gendered context of assimilation: the female secondgeneration advantage among Latinos Nicol M. Valdez and Van C. Tran Department of Sociology, Columbia University, New York, NY, USA ABSTRACT Recent macro-level trends have created a gendered context of assimilation for second-generation Latinos, pointing to favourable condition among Latina women and potential vulnerability among Latino men. This article examines second-generation assimilation in young adulthood for the top ten Latino groups using gendered reference groups to benchmark their integration into American society. We find a significant female advantage on individual-level indicators, but no clear gender differences on household-level indicators among Latinos. We also document significant heterogeneity in achievement among the top ten Latino ethnic groups, with Mexicans being the most disadvantaged group. In contrast, Cubans and South Americans are the most advantaged and have achieved parity with native whites. These findings point to the emerging significance of gender as a key dimension of stratification in the assimilation of Latinos into American life. ARTICLE HISTORY Received 3 January 2018 Accepted 28 December 2018 KEYWORDS Hispanics/Latinos; gender; second generation; assimilation; immigrant incorporation Introduction Over the last two decades, the second generation in the U.S. has come of age in sizable numbers. In 2010, the second-generation population included 20 million adults and 16 million children (Pew Research Center 2012). By 2050, the second generation is projected to make up 18.4% of the total U.S. population (Pew Research Center 2015). To place this in a historical context, the current share of the second generation is approaching proportions at the turn of the twentieth century, when the European second generation accounted for one in five of the total U.S. population (Hirschman 2005). This growing segment of the population is very diverse, with Latinos and Asians making up more than half of second-generation adults. A robust literature has examined how the second generation will assimilate into American society (Alba and Nee 2003; Kasinitz et al. 2008; Lee and Zhou 2015). As the nation s largest racial minority group, Latinos accounted for 18% of the U.S. population in 2016 and find themselves at the core of this debate. In this paper, we build on recent efforts that focused on gender as a central dimension to the assimilation experience (Feliciano 2012; Park, Nawyn, and Benetsky 2015). This is especially the case among Latinos because recent macro-level trends have created an emerging gendered context of assimilation that is slightly more advantageous towards Latina CONTACT Nicol M. Valdez nmv2116@columbia.edu 2019 Informa UK Limited, trading as Taylor & Francis Group

3 2 N. M. VALDEZ AND V. C. TRAN women. We address two questions. First, how do second-generation Latinos compare to their counterparts of the same gender from the native-born groups in terms of socioeconomic integration? Second, how do men and women within the same Latino ethnic group compare to each other? The first question addresses gender differences across groups whereas the second question examines gender differences within each ethnic group. We focus on ten Latino groups using gendered reference groups to benchmark their integration into American society because prior work has shown that the choice of reference group can shape conclusions on second-generation progress (Park, Myers, and Jiménez 2014; Tran and Valdez 2017; Kasinitz et al. 2008). While Park, Nawyn, and Benetsky (2015) have addressed gender differences among the second generation, one limitation in their analysis is their reliance on panethnic categories such as Latino and Asian which lumped together many ethnic groups with different backgrounds and experiences. Aware that Latinos and Asians are far from monolithic, they suggest that replication of these analyses for specific ethnic groups will build our knowledge (1622). Our paper builds directly on Park, Nawyn, and Benetsky (2015), while also advancing this research in four ways: by using the most recent nationally representative data, by examining intra-racial differences among Latinos, by examining gendered differences within each ethnic group, and by incorporating the different starting points in human, economic and financial resources among Latino ethnic groups. On ethnic groups starting points, we hypothesise that groups with higher levels of human capital and lower levels of legal and institutional barriers are more likely to achieve parity with the native mainstream. In addition, we make a distinction between individual-level and household-level indicators of socio-economic attainment. We expect the female advantage to be larger for the former indicators, whereas these differences should be more muted for the latter indicators because of joint households and shared resources between both genders. How gender matters for theories of assimilation Contemporary theoretical accounts on immigrant assimilation have remained mostly silent as to how gender matters for the assimilation process (Donato et al. 2006; Hondagneu-Sotelo 2013). The theory of segmented assimilation highlights the macro-structural sources of vulnerability that might lead to downward mobility among second-generation Latinos, including racial minority status, segregated urban schools, concentration in disadvantaged neighbourhoods as well as the increasingly bifurcated economy with fewer good jobs. Proponents of new assimilation theory argue that the Latino second generation will most likely follow the time-honored path among European immigrants and their descendants who achieved parity with the white American mainstream over the course of three generations (Alba and Nee 2003). The barriers confronting Latinos are significant, but they are not insurmountable. These competing theoretical frames reflect variations of straight-line assimilation, which originated from the experiences of European immigrant men at the turn of the twentieth century, and yet, made claims that represent the entire immigrant population (Donato et al. 2006; Hondagneu-Sotelo 2013). This made sense historically because most immigrants were indeed men; although even this widely held belief is not entirely accurate. During this earlier period, there were substantial numbers of female immigrants

4 JOURNAL OF ETHNIC AND MIGRATION STUDIES 3 from Mexico, China and Japan, but the architects of straight-line assimilation never considered their assimilation experiences (Jiménez 2010; Alba and Nee 2003). Furthermore, Donato and Gabaccia (2015) find evidence that women were a significant part of immigration history in their analysis of historical data tracing migration rates back to the fifteenth century. The main difference between then and now is more data availability and higher demand for female immigrant labour. Since the early 1980s, immigration scholars began to include immigrant women s accounts of their assimilation experiences (Hondagneu-Sotelo 2003) while also noting how women s migration patterns have increased over time (Donato et al. 2006; Garip et al. 2006; Massey 2009). Research on gender and immigration mostly adopts a qualitative approach focusing on one ethnoracial group (i.e. Smith 2006; Feliciano and Rumbaut 2005; Park, Myers, and Jiménez 2014). When gender is incorporated within quantitative models, it is often understood as a control variable and rarely contextualised within the comparison groups as an integral part of the assimilation process (Hondagneu-Sotelo 2003). More recent work has put a greater emphasis on gender and how it is a central aspect to benchmarking second-generation socioeconomic progress (Park, Nawyn, and Benetsky 2015). Given shifting migration patterns, theories of assimilation and integration should consider these changing gender dynamics. In this paper, we focus on the gendered context of assimilation, while also confirming the racialized context of assimilation from prior research. Gendered context of assimilation among Latinos Recent macro-level trends and the historical context under which the post-1965 second generation came of age created what we call a gendered context of assimilation for Latinos. Gendered achievements in education Over the last four decades, women have reversed the gender gap in educational achievement in the U.S. By 1982, women had surpassed men s educational attainment in higher education and accounted for 57% of all bachelor s degrees (Snyder and Dillow 2012). As women s educational attainment trends were on the rise, women were also joining the labour markets at faster rates, consequently lowering men s wages (Fischer and Hout 2006). Women s rising labour force participation not only expanded job opportunities that required further education, but also increased their incentive to earn college degrees (Buchmann, DiPrete, and McDaniel 2008; DiPrete and Buchmann 2013). Sociologists have argued that throughout all levels of schooling girls experience compounding advantages over their boy counterparts. Girls are less likely to be retained in kindergarten, more likely to graduate from high school, and more likely to experience successful transitions to college and to graduate with a college degree (Jacobs 1996; DiPrete and Buchmann 2013). While women s educational participation was on the rise the educational attainment for men began to stall, especially for men of colour. In 2012, the male dropout rate was 8% compared to the female dropout rate of 6%, with Latinos and Blacks reporting the largest male-female dropout rate gaps (National Center for Education Statistics 2013). Women also obtained graduate degrees more often earning 60% of masters degrees and 52% of doctoral degrees signalling a consistent growing female

5 4 N. M. VALDEZ AND V. C. TRAN advantage in higher education (Snyder and Dillow 2012). As a result, this increasing feminization of educational attainment shapes the motivations, expectations and norms to succeed in school for all women, including second-generation Latina women in ways not present for their co-ethnic male peers. Gendered experiences with the criminal justice system While the educational context for all women was becoming increasingly more positive, the overall context facing both black and Latino men became plagued by a detrimental rise in policing and mass incarceration (Pettit and Western 2004). The rates of incarceration for men of colour skyrocketed in the early 1980s at a time when the drug hysteria swept the nation. From 1980 to 2008, the number of people incarcerated in American quadrupled from 500, million (Schwartz 2011). In 2001, the incarceration rate was 472 per 100,000, nearly five times its historical average (Pettit and Western 2004). At the same time, early contact with the criminal justice system and zero-tolerance educational climates increased the likelihood of dropping out among Latino boys (Nolan 2011). High rates of incarceration and the increasing criminalisation of young Latino men affect their experiences in education, work, and family formation in adulthood (Pettit and Western 2004; Wakefield and Wildeman 2013). Over the last decade, the dramatic rise of deportations has disproportionately affected Latino families and the stigma of being undocumented became especially acute for Mexicans (Gonzales 2015; Dreby 2015). For example, deportations increased from 359,795 in ,421 in 2013, an increase of 21.9% in over five years (U.S. Department of Homeland Security 2014). The deportation experience is also highly gendered and racialized: 97% of deportees are sent to Latin America or the Caribbean and 88% are men (Golash-Boza 2015). Latino men are disproportionately impacted given the increasing racialisation of Latino men and the heightened criminalisation of undocumented immigrants. While second-generation Latinos are not directly subjected to immigrant detention and deportation, increasing rates of deportation pose significant consequences for the integration of all Latinos, both at the individual and group level. The undocumented stigma has become so consequential that it now affects not only mixed-status families, but also other Latino families with documented status. This is how and why mass deportation and mass incarceration can be detrimental to second-generation Latino men. The rhetoric that Latinos, specifically Mexicans, are taking jobs, and threatening the quintessential American life perpetuates negative stereotypes of Latinos on the group level (Chavez 2008). With approximately 10 million U.S.-citizen children report living with either an undocumented parent or family member in the household, the undocumented experience is no longer just a first-generation experience (American Immigration Council 2017). However, the impacts of undocumented status on Latino families also vary by ethnic origin and by gender. Second-generation men with parents from Mexico and Central America are more negatively impacted. Immigrants from these sending countries make up a larger proportion of the immigrant population in the U.S. and are more likely to enter without documentation (Menjívar and Abrego 2012). In contrast, South Americans and Cubans not only make up a smaller share of the Latino population in the U.S., but they are also more likely to be documented, to have higher levels of education and report higher household incomes than the overall foreign-born Latino

6 JOURNAL OF ETHNIC AND MIGRATION STUDIES 5 population in the U.S. (Zhong and Batalova 2018). In this sense, the undocumented experience on the family unit is more defining and impactful for some second-generation Latino groups than for others. Furthermore, deportation as a gendered racial removal program fundamentally changes family compositions by creating more female-headed households (Golash-Boza and Hondagneu-Sotelo 2013). Although Latino men are systematically targeted for deportations, women and children from these families are also impacted. The deportation crisis is not only exacerbating many Latino families already precarious economic situation, but it is also exacting significant emotional and psychological toll by separating children from parents across borders (Dreby 2012). Although second-generation Latinos are U.S. citizens, research has documented the detrimental effect of lack of parental legal status on their children, including early childhood development, educational attainment, and mental health outcomes (Bean, Brown, and Bachmeier 2015; Yoshikawa 2011; Suarez- Orozco et al. 2011). Legal status has become increasingly consequential for the integration of the Latino second generation and their situation has significantly worsened under the current administration s anti-immigrant agenda. Gendered expectations at home and in school Prior research has linked gendered expectations within the family and schools to gendered patterns of achievement (Feliciano and Rumbaut 2005; Feliciano 2012; Rendón 2014). Although girls tend to earn higher grades and report higher educational ambitions, they receive different treatments from both their parents and teachers (Feliciano and Rumbaut 2005; Suarez-Orozco and Qin 2006). For example, Feliciano (2012) finds that boys spend less time on homework, watch more TV, and confront more negative perceptions from school personnel and peers. Immigrant families also exert greater social control over daughters and give them more responsibilities at home (Zhou and Bankston 2001; Lopez 2003). Latina girls are also more tied to their homes than their male peers and are more likely to maintain cultural attachments such as language ability and familial duty (Lutz and Crist 2009; Feliciano and Lanuza 2017). The social and behavioural skills adopted by boys and girls at home translate into clear differences in rewards and penalties within the school system (Entwisle, Alexander, and Olson 2007; DiPrete and Jennings 2012). Girls and boys behaviours are differentially monitored, with boys reporting closer monitoring and more likely to receive suspension and punishment in schools (Rios 2011). In addition to lower educational attainment, Lopez (2003) documents experiences of racialisation and stigmatisation among Dominican boys because they might be stereotyped as hoodlums or thugs by school and security staff. Other scholars point to gendered differences in neighbourhood violence where Latino males get caught up in delinquent behaviours that are misaligned with doing well in school because their network ties often expose them to urban violence (Rendón 2014). Collectively, these racialized and gendered experiences within homes, schools, and neighbourhoods contribute to differential educational outcomes by gender which, in turn, fuels divergent trajectories for secondgeneration Latino men and women.

7 6 N. M. VALDEZ AND V. C. TRAN Gendered opportunities in the economic spheres Within the labour market, research has shown that men and women adopt different strategies to achieve social mobility. Although both Latino men and women confront a type of immigrant shadow in the labour market (i.e. the idea that Latinos are uneducated and unassimilable into the American mainstream), Latina women quickly learn to strategize to get ahead by adopting mainstream standards to successfully cross ethnic and socioeconomic lines (Agius Vallejo 2009). Similarly, Latina women rely on their network ties to advance to better positions, whereas Latino men s networks may not provide similar support for occupational mobility (Dominguez 2010). Given their lower educational achievement, Latino men might lack the credentials and qualifications to secure highpaying positions, leaving them on the lower rungs of occupational niches. This also further explains the increased levels of women s labour force participation across all levels of the occupational ladder over the last decades. However, factors influencing women s occupational opportunities are more complex. Occupational decisions are often influenced by family, background, and gendered expectations. Flores and Hondagneu-Sotelo (2014) find that college-educated Latinas are increasingly selecting teaching professions. Essentially, these dutiful Latinas see teaching as an immediate way to help alleviate the economic hardships faced by their families. Although women are making greater strides in the labour market due to their increased educational attainment, they remain at a disadvantage at the highest levels of occupational attainment. This effect is combined with the gender pay gap remaining intact, signalling the inherent belief that men s work is still seen as being more valuable (Ridgeway 2011). Although the glass ceiling is a well-documented phenomenon, second-generation Latinas are further hindered in their job opportunities because of both social class background and family obligations (Agius Vallejo 2015). Beyond education and work, gender differences in socioeconomic resources are often explained by variation in family structure, labour market inequality in earned wages, and women s tendency to be more risk aversive with investment decisions (Ruel and Hauser 2013). At the household level, several dynamics are at work that shape gender inequality in economic resources (i.e. home ownership, wealth, and poverty). On the one hand, trends in intermarriage and educational homogamy disproportionately benefit Latina women because women with higher levels of educational attainment are more likely to intermarry, allowing them to benefit from diverse forms of economic resources. For example, Alba (2016) finds that mixed families have household incomes more similar to mainstream white families than to minority families. Latino groups who are light-skinned and middle-class are also more likely to marry non-latino whites. On the other hand, men on average are more likely to own homes and are less likely to live in poverty, which counter the potential female advantage due to intermarriage. Finally, many economic resources are shared within the household, resulting in muted gendered differences. Finally, the context of assimilation for Latinos worsened with the onset of the Great Recession, but especially for Latino men. Latinos were hit the hardest by the foreclosure crisis: 14% of Latinos lost their homes, compared to 11% for blacks and 6% for whites (Bocian et al. 2012). Between 2006 and 2010, Latinos experienced the steepest increase in poverty rate, from 20.5% to 26.6% (DeNevas-Walt, Proctor, and Smith 2010). The

8 JOURNAL OF ETHNIC AND MIGRATION STUDIES 7 Figure 1. Conceptual schema for gendered context of assimilation among second-generation Latinos. Source: Authors compilation. impact of the Great Recession was highly gendered, with about 70% of all job losses occurring among men because men were more likely to be employed in highly cyclical industries such as construction and manufacturing (Hoynes, Miller, and Schaller 2012). While second-generation Latinos are more likely than the first generation to be in a professional occupation, they are also disproportionately represented in construction work (Tran and Valdez 2017). For example, second-generation Mexicans are twice as likely than secondgeneration respondents from other Latino groups to work in construction in traditional gateways (Tran and Valdez 2017). While rising deportation and declining fortunes among the Latino first generation might not differentially impact the Latino second generation by gender at the individual level, this broader context at the group level is more likely to affect Latino men than Latina women at the group level, given the gendered and racialized perceptions of achievement and under-achievement in American society. To sum up, Figure 1 outlines the macro-level factors and mechanisms that influence second-generation Latino socioeconomic attainment by gender. Overall, Latina women s context of assimilation in the U.S. underwent positive changes with increasing labour force participation whereas mass incarceration, mass deportation and the Great Recession disproportionately affected Latino men. In other words, women are positively reinforced to do well in school and in the workplace in ways that may not be as supportive for men of colour. To be sure, we also expect significant variations by national origin. We draw on this conceptual framework to situate our empirical findings and draw attention to how this gendered context of assimilation shapes gendered patterns of attainment among the Latino second generation. Starting points and heterogeneity in Latino second-generation trajectories Among Latinos, there is also significant internal diversity by social class, ethnoracial phenotype, gender identity, and legal status (Alba, Jiménez, and Marrow 2014; Agius Vallejo 2015; Tienda and Mitchell 2006). We highlight this heterogeneity in Figure 2, which summarises three key factors shaping socioeconomic attainment among the Latino second generation: immigrant selectivity or starting point in human and social capital, experiences of racialisation and discrimination, and legal status among the first generation.

9 8 N. M. VALDEZ AND V. C. TRAN Figure 2. Key factors shaping Latino second-generation attainment by national origin. Notes: 1. Immigrant selectivity and starting point are based on average socioeconomic indicators for the top Latino ethnic groups from López and Patten (2015). 2. Experiences of racialisation and discrimination are based on responses to the series of questions about discrimination by ethnic group from the 2007 National Survey of Latinos. 3. Lack of legal status is based on the proportion of individuals with undocumented status by Latino ethnic group from Rosenblum and Soto (2015). On selectivity, Feliciano (2005) finds that Mexicans, Dominicans and Salvadorians exhibit negative selection because of the first generation s lower educational and socioeconomic status upon arrival, which creates lower starting points for these ethnic groups. In contrast, South Americans often come to the U.S with higher than the average Latino educational attainment. As a result, the second-generation, especially girls, hold similar educational expectations and aspirations as their mothers in the home country (Feliciano 2012). On racialisation, Latinos are phenotypically diverse and fall along the entire spectrum of the black-white continuum. Latino groups with phenotypically lighter skin such as Cubans and South Americans are less likely to be undocumented or to report experiences of discrimination. The second generation for these groups might experience boundary shifting and join the native majority group given their higher socioeconomic markers and levels of educational attainment (Alba and Nee 2003). In contrast, Mexicans, Dominicans and Central Americans with more indigenous or African phenotypic components might report similar outcomes to native minority groups (Vaquera, Aranda, and Gonzales 2014). On legal status, Mexicans and Central Americans are more likely to be undocumented, which negatively affects second-generation achievement (Bean, Brown, and Bachmeier 2015) whereas Cubans were welcomed as refugees and offered support upon initial resettlement (Portes and Rumbaut 2001). Hypotheses Our discussion above leads to the following two sets of testable hypotheses: Hypothesis 1: Given the gendered context of assimilation, we predict a female advantage among second-generation Latinos. On individual-level indicators, second-generation Latinas should fare better than their male counterparts in education and occupational attainment. On household-level indicators, we expect the gender differences to be less pronounced because both genders live in joint households and share economic resources. Hypothesis 2: Given the heterogeneity among second-generation Latinos, ethnic groups with lower level of human capital and higher levels of institutional and legal barriers (i.e. Mexicans and Central Americans) should fare more similar to native minority groups. In contrast, those from Latino ethnic groups with higher level of human capital and lower level of legal barriers (i.e. Cubans) should achieve parity with the native majority group.

10 JOURNAL OF ETHNIC AND MIGRATION STUDIES 9 Data and methods We draw on the Annual Social and Economic Supplement of the Current Population Survey (CPS ASEC), a nationally representative survey that includes a diverse Latino sample. Our empirical analyses rely on the pooled dataset from CPS ASEC 2008, 2010, 2012, 2014 and 2016, which provides the most recent data and adequate sample size for the top ten Latino ethnic groups. The CPS ASEC is administered by the Census Bureau through both in-person and telephone interviews every month to monitor basic trends in the population. It uses a probability sample of about 75,000 occupied households from all 50 states and the District of Columbia. The survey design adopts a sampling scheme where households are included in the survey for the first 4 consecutive months and excluded for the next 8 months, before returning again for the last 4 consecutive months. Given this sampling design, the pooling of data from the samples will ensure the presence of non-overlapping individuals in the pooled dataset, because each of these surveys was collected two years apart. 1 The CPS ASEC, also known as the March Supplement, collects detailed information on the socioeconomic characteristics of every member in the household, providing the basis for national estimates of income and poverty. The CPS ASEC has three advantages. First, it is the only nationally representative source of data on second-generation adults in the U.S. Second, the pooled sample provides sample size for the top ten Latino ethnic groups (including Mexicans) in young adulthood. Third, it provides comprehensive socioeconomic indicators on education, occupation and poverty. However, the CPS ASEC also has one major limitation, the data cannot address social processes that underlie patterns of assimilation or inequality because it does include information about respondents family background, neighbourhood or school. As a result, our focus will be on attainment among second-generation Latinos, not on intergenerational mobility. This analysis is restricted to respondents between the ages of 25 and 40 for three reasons. First, previous research has focused on the new second generation in young adulthood. This will allow for effective comparisons of these findings to other regional surveys. Second, this strategy provides a robust set of comparisons across ethnoracial groups, effectively controlling for cohort effect. The oldest respondent in this sample would have been born between 1968 and 1976 and the youngest respondent between 1983 and Both the second-generation and the third-plus-generation respondents grew up in the aftermath of the Civil Rights movement, under the relative liberalisation of immigration policy and increasing influx of immigrants from all over the world. Third, this group of young adults came of age under similar macro-level economic conditions in an era of growing income inequality, expanding opportunities for higher education, and rising acceptance of multiculturalism. Overall, these factors provide a specific context of assimilation for immigrants and their offspring. Dependent variables This analysis focuses on socioeconomic attainments, 2 using six dichotomous dependent variables that capture the two ends of socioeconomic assimilation: having no high school degree, having a graduate degree or more, working in a low-level service

11 10 N. M. VALDEZ AND V. C. TRAN occupation, working in a professional or managerial occupation, living in poverty and owing a home. The first four are the individual-level measures whereas the last two are household-level measures. These six measures represent crucial aspects of socioeconomic integration in prior work (Park, Myers, and Jiménez 2014; Farley and Alba 2002). For example, owning a home is one of the biggest purchases the average person will make in a lifetime and exemplifies the achievement of the American dream for many immigrant families. In contrast, living in poverty has resounding implications for downward mobility across generations. Immigrant generations and ethnic groups Generation status is based on the individual s birthplace and that of the parents. The second generation includes those who were born in the U.S. to parents (one or both) born outside of the U.S. For purpose of classification of ethnic origin, we give priority to the mother s country of birth. Specifically, those who were born in the U.S. to a foreign-born mother and a U.S.-born father are classified based on the mother s country of origin. Those who were born in the U.S. to a U.S.-born mother and a foreign-born father are classified based on the father s country of origin. Those who were born in the U.S. to two foreign-born parents of different origins are classified based on the mother s country of origin. The third-plus generation includes those who were born in the U.S. to U.S.-born parents. However, we cannot distinguish between third and higher generations, so the term third-plus-generation includes both groups. We do not examine outcomes among the third-and-higher-generation Latinos in this paper because the sample size for each Latino ethnic group is too small, with the exception of Mexicans. Our analysis will focus on the second generation from the top ten Latino ethnic groups from the 2010 census: Mexicans, Cubans, Puerto Ricans, Dominicans, South Americans, and Central Americans. Specifically, Central Americans includes Salvadorans, Guatemalans, and Hondurans, whereas South Americans includes Colombians, Ecuadorians and Peruvians. (For details on the sample by ethnic group, see Table A1 in the Appendix.) The combination of the smaller ethnic groups into South Americans and Central Americans are substantively justified on two grounds. First, these ethnic groups have quite similar socioeconomic profiles and often live in close proximity to each other (Kasinitz et al. 2008). Second, the first generation among these ethnic groups faced similar contexts of exit and reception. Salvadorans, Guatemalans, and Hondurans not only came from countries where political and economic conditions have been linked together since precolonial times, but they also faced significant challenges in gaining legal acceptance in the U.S. and reported high deportation rates (Menjívar and Abrego 2012). CPS ASEC also allows for the identification of three reference groups: non-hispanic whites, non-hispanic blacks, and Puerto Ricans. The reference for most of the multivariate analyses will be the native majority, although comparisons will be made to the two native minority groups. Analytical methods The analyses proceed in three stages. First, bivariate analyses provide statistical profiles on the six indicators for each ethnic group by ethnoracial origin. Second, gender-specific

12 JOURNAL OF ETHNIC AND MIGRATION STUDIES 11 multivariate logistic regression analyses explore the socioeconomic attainment of Latino ethnic groups, relative to third-plus-generation comparison groups. Because the six dependent variables are dichotomous measures, we use logistic regressions with robust standard errors and report the odds ratios. The multivariate models for measuring across-group differences separately by gender and for measuring within-group gender differences are as followed: P(Y i = 1) log 1 P(Y i = 1) = b 0 + b 1 E i + b i X i (1) P(Y i = 1) log 1 P(Y i = 1) = b 0 + b 1 E i + b 2 G i + b 3 E i G i + b i X i (2) P(Y i = 1) where log denotes the log odds of the probability of respondent i attaining 1 P(Y i = 1) a particular socioeconomic outcome (Y). E i is the ethnic background for respondent i, the main variable of interest. G i is the gender for respondent i. X i is a vector of control variables for respondent i. To investigate intra-ethnic differences by gender, we introduce the interaction terms between ethnic origin and gender. The key control variables include respondents age, the quadratic term of age, region, and survey year. For the four measures of occupation, poverty and home ownership, we also controlled for respondent s education in the multivariate models, as education is a key predictor of these outcomes. Region is a categorical variable with four categories: Northeast, Midwest, West, and South. For the occupational measures, we only include those respondents that reported full-time employment in the previous year. We present our findings using odds ratios and all analyses adjust for the stratified survey design using appropriate final weights. Gendered attainments among second-generation Latinos in young adulthood Table 1 presents descriptive results by ethnic group and gender. Overall, we find a clear female advantage on educational and occupational attainment whereas the gender differences on measures of home ownership and poverty are much less pronounced, offering initial support to Hypothesis 1. On education, second-generation Mexicans, Dominicans and Central Americans report the lowest level of achievement. Among men, the high school dropout rate is 16.4% among Mexicans, 15.3% among Dominicans, and 15.1% among Central Americans. Among women, the dropout rates are 13.6% among Mexicans, 7.4% among Dominicans and 5.3% among Central Americans. In contrast, second-generation Cubans and South Americans report the highest rates of graduate degree completion. For men, these rates are 9.4% among Cubans and 12.1% among South Americans, compared to 10.1% among white men. For women, these rates are 21.2% among Cubans and 23% among South Americans, compared to only 14% among white women. There are also important gender differences in education. Compared to their co-ethnic female counterparts, Central Americans men are three times more likely to drop out of high school and Dominican men are twice more likely to do so. Compared to their co-ethnic male counterparts, the rates of graduate degree completion are two times higher for Cuban and South American women and three times higher for

13 Table 1. Socioeconomic attainment among second-generation Latinos by gender. Male respondents No High School Graduate Degree Service Occupation Professional Occupation Living in Poverty Own Home Second-generation Cuban South American Central American Dominican Mexican Third-plus generation Native white Native black Puerto Rican Female respondents No High School Graduate Service Occupation Professional Living in Poverty Own Home Degree Occupation Second-generation Cuban South American Central American Dominican Mexican Third-plus generation Native white Native black Puerto Rican Source: Pooled CPS ASEC ( ). Notes: All numbers are percentages. 12 N. M. VALDEZ AND V. C. TRAN

14 JOURNAL OF ETHNIC AND MIGRATION STUDIES 13 Dominican women. To be sure, a female advantage also exists among the native comparison groups, but the differences are less stark. On occupation, Mexican and Central American men report the highest concentration in service work whereas Cuban and South American men report the highest concentration in professional occupation. This is consistent with the educational profiles reported above, as educational attainment is a key component of occupational attainment. Among women, the differences in service work are not as pronounced by ethnic origin because women tend to be more concentrated in this line of work. On professional occupation, Cuban and South American women also report the highest rates whereas Mexican and Central American women report the lowest. Finally, Dominican women report the highest rate of service work and the second highest rate of professional occupations, suggesting a bi-modal distribution of their occupational profile. This is consistent with their educational profile, which includes the second highest high school dropout rate and third highest graduate school completion rate among all Latino groups. Across genders, the occupational differences are less stark within each ethnic group compared to the educational achievement, suggesting that the female advantage in education might not fully translate into their occupational gains. On poverty, Latina women generally report higher rates compared to Latino men, but gender differences are not significant, with the exception of poverty rate among Dominicans. Among men, Central Americans and Mexicans report the highest rates of poverty whereas South Americans and Dominicans report the lowest rates. In fact, the poverty rates among both South American and Dominican men are lower than the rate among white men. Among women, Dominicans report the highest poverty rate (21.2%), followed by Mexicans (15.1%) and Central Americans (12.6%). The most striking gender difference is among Dominicans where second-generation women are three times as likely as their male co-ethnics to live in poverty. On home ownership, men report higher rates compared to women, with the exception of Dominicans. The highest homeownership rates are among Cuban men and women, whereas the lowest rates are among Mexican and Central American women. Overall, these results underscore the bimodal nature of socioeconomic attainment among Dominican women and the relative disadvantage among Mexicans. On the one hand, Dominican women are both most likely to live in poverty and more likely to own a home, compared to both Dominican men and to other Latina women. One potential explanation for the comparatively high rate of poverty among Dominican women, despite their educational advantage, is the higher rates of female-headed households among Dominicans, compared to other Latino ethnic groups. More generally, these outcomes are also broadly consistent with the bimodal distribution of their educational profile. On the other hand, Mexicans of both genders report relatively low levels of achievement across many of these socioeconomic indicators compared to other Latino groups. We next turn to the multivariate results to examine ethnic differences in Hypothesis 2. Tables 2 and 3 present two sets of logistic regressions on the six outcomes separately for men and for women. Figures 2 and 3 present the predicted probabilities of attainment by ethnic group separately by gender, along with the confidence intervals for these regression estimates based on the multivariate analyses in Tables 2 and 3. Among second-generation men, Figure 3 shows that Mexicans are the most disadvantaged.

15 14 N. M. VALDEZ AND V. C. TRAN Table 2. Logistic regressions on socioeconomic attainment, male respondents only. Variables No High School Graduate Degree Service Occupation Professional Occupation Living in Poverty Own Home Odds ratios Odds ratios Odds ratios Odds ratios Odds ratios Odds ratios Ethnoracial origin Cuban (.220) (.400) (.358) (.312) (.445) (.272) South American (.385) (.344) (.446) (.249) (.323) (.143) Central 3.375*** *** American (.633) (.162) (.161) (.148) (.257) (.066) Dominican 3.585*** ** (1.316) (.255) (.503) (.451) (.238) (.130) Mexican 3.773***.274*** 1.225*.653*** (.331) (.044) (.120) (.051) (.086) (.064) Puerto Rican 1.872***.334*** 2.128***.696* *** (.319) (.084) (.291) (.098) (.232) (.050) Black 2.079***.399*** 1.888***.626*** 2.346***.439*** (.111) (.028) (.094) (.032) (.113) (.015) Constant.038**.000*** *** *** (.043) (.000) (1.089) (.007) (4.139) (.030) N 67,150 67,150 55,485 55,485 67,150 67,150 Source: Pooled CPS ASEC ( ). Notes: Robust standard errors for the odds ratios are in parentheses. Combined samples are limited to those between the ages of 25 and 40. All models also controlled for age, quadratic term of age, gender, region and survey year. Sample sizes for multivariate models on service and professional occupations were restricted to those who were employed full-time in the previous year. The reference group is third-plus-generation whites. ***p <.001, **p <.01, *p <.05. Compared to white men, Mexican men are significantly more likely to have no high school degree, less likely to hold a graduate degree, more likely to work in the service sector, and less likely to be in a professional occupation. Next, Dominican and Central American men are the second most disadvantaged groups. Compared to white men, they are significantly less likely to have a high school degree and to own a home. In contrast, Cuban and South American men report socioeconomic parity with the native majority group. This is remarkable because they have closed the gaps with the mainstream on key measures of attainment by the second generation whereas this process took three generations for descendants of European immigrants who arrived at the turn of the twentieth century (Alba and Nee 2003). Among second-generation women, the pattern is more mixed in Figure 4. Compared to white women, Mexican and Central American women report a disadvantage in education and occupation. However, Mexicans are also less likely to live in poverty compared to white women. Although Dominican women have achieved educational parity with white women, they are still more disadvantaged on indicators of poverty and homeownership. In contrast, Cuban and South American women have a slight advantage in both education and occupation compared to white women. Although there is a female advantage among Latinos on the individual-level indicators, there is a clear stratification across Latino ethnic groups compared to the native majority group. Among Latinos, Mexicans of both genders are the most disadvantaged, followed by Dominicans and Central Americans, whereas Cubans and South Americans are the most advantaged.

16 Table 3. Logistic regressions on socioeconomic attainment, female respondents only. Variables No High School Graduate Degree Service Occupation Professional Occupation Living in Poverty Own Home Odds ratio Odds ratio Odds ratio Odds ratio Odds ratio Odds ratio Ethnoracial origin Cuban *.469* (.308) (.406) (.149) (.179) (.249) (.145) South American.063** * (.064) (.455) (.146) (.289) (.397) (.149) Central American ** ** *** (.283) (.111) (.167) (.092) (.183) (.059) Dominican **.259*** (.859) (.293) (.203) (.273) (.556) (.054) Mexican 3.809***.330***.713***.823**.859*.994 (.295) (.038) (.059) (.061) (.065) (.056) Puerto Rican 2.670***.497*** ***.333*** (.490) (.085) (.140) (.108) (.176) (.032) Black 2.052***.533*** 1.357***.711*** 2.353***.282*** (.103) (.026) (.054) (.028) (.081) (.008) Constant *** *** *** (.517) (.000) (3.230) (.009) (.817) (.005) N 76,023 76,023 55,363 55,363 76,023 76,023 Source: Pooled CPS ASEC ( ). Notes: Robust standard errors for the odds ratios are in parentheses. Combined samples are limited to those between the ages of 25 and 40. All models also controlled for age, quadratic term of age, gender, region and survey year. Sample sizes for multivariate models on service and professional occupations were restricted to those who were employed full-time in the previous year. The reference group is third-plus-generation whites. ***p <.001, **p <.01, *p <.05. JOURNAL OF ETHNIC AND MIGRATION STUDIES 15

17 16 N. M. VALDEZ AND V. C. TRAN Figure 3. Predicted probabilities of socioeconomic attainment, male respondents only. Source: Pooled CPS ASEC ( ). Notes: Results are based on the subsample of 25-to-40-year-old respondents. Predicted probabilities are based on multivariate models which controlled for gender, age, quadratic term of age, region, and immigrant destinations (see Table 2 for full results). Sample sizes for service and occupational professions were restricted to those who were employed full-time in the previous year. Gendered differences by ethnic origin Although the focus of the paper is to better understand gender differences across groups, we also wanted to know if there are any significant gender differences within each ethnic group. In other words, we are interested in the interaction between ethnicity and gender, as well as how this combination creates ethnoracial inequality. We systematically test for gender differences by including a series of interaction terms in our multivariate models and these results are in Table 4. With the exception of home ownership, the main effect for the gender coefficient is statistically significant and show a clear female advantage. Among other covariates, educational achievement is significantly associated with occupational achievement, lower poverty rate and higher home ownership rate. Latinos in the Northeast report higher educational achievement and higher concentration of service workers whereas Latinos in the West report significantly higher level of professionals. There are no differences in poverty rates across the regions, but homeownership is more prevalent in the Midwest and the South. On survey year, respondents from later CPS years are more likely to report higher levels of attainment. This is reflective of the cumulative impact of time on achievement as the Latino second generation moves from early to middle adulthood.

18 JOURNAL OF ETHNIC AND MIGRATION STUDIES 17 Figure 4. Predicted probabilities of socioeconomic attainment, female respondents only. Source: Pooled CPS ASEC ( ). Notes: Results are based on the subsample of 25-to-40-year-old respondents. Predicted probabilities are based on multivariate models which controlled for gender, age, quadratic term of age, region, and immigrant destinations (see Table 3 for full results). Sample sizes for service and occupational professions were restricted to those who were employed full-time in the previous year. We now turn to the interaction effects. Although not all the coefficients for interaction terms between gender and ethnic origin reach statistical significance, the Chi-squares and p-values from the post-regression Wald tests for the interaction terms are significant for five of the six dependent variables with the exception of home ownership. Since this is a nonlinear model, we must take into account the values of all covariates in interpreting these interaction effects. To do so, we will start with the margins command in Stata that estimates the discrete difference in predicted marginal probabilities between male and female respondents for each ethnic group. In what follows, we visualise three of the six outcomes for which the interaction coefficients reach statistical significance for at least two of the seven interaction terms between gender and ethnic origin. These are having no high-school education, working in service occupations and owning a home. (However, not all interaction terms for these three outcomes reach statistical significance.) Specifically, we graph the post-regression predicted probabilities for each ethnic group by age, separately by gender. Because our primary focus is on intra-ethnic gender differences among Latinos, we limit the presentations of these results to only Latino ethnic groups for visual simplicity, but the underlying logistic regression models include native whites and native blacks (see Table 4). Because socioeconomic attainment increases with age, we