Between Two Worlds: Source-Country Gender Roles and Gender Differences in Educational Attainment among Immigrant Children*

Between Two Worlds: Source-Country Gender Roles and Gender Differences in Educational Attainment among Immigrant Children* Teresa Abada, University of Western Ontario Kristyn Frank and Feng Hou, Statistics Canada Recent scholarship has focused on the relationship between source-country characteristics such as female labor force participation, fertility, level of economic development, gender role attitudes, and immigrants labor market assimilation. These studies refer to national-level factors when accounting for the vast differences in home-country groups in labor market outcomes. This study asks to what extent these source-country characteristics affect immigrant children s educational outcomes. Using data from the 2006 Canadian Census and World Values Survey, this article examines the extent to which the gender gap in educational attainment among immigrant children is associated with source-country factors. Female child immigrants who come from countries with high female labor force participation and high levels of GDP have an advantage over their male counterparts in university education. Source-country gender role ideology played a role in university completion rates for immigrant parents, but not for child immigrants. Introduction Increases in educational attainment have become widespread across industrialized societies. The advancement of women s educational attainment has been particularly notable, with women surpassing men in all aspects of their educational career (Diprete and Buchmann 2013). The rapid increase in female educational attainment has resulted in women having higher university degree completion rates than men across all racial groups. An interesting and important question is whether similar progress in women s educational attainment has been achieved by the children of immigrants from less egalitarian countries (Fenella and Kristen 2014). Immigration studies have increasingly focused on variations in the educational attainment of immigrant children as a means of assessing their acculturation to host-country norms (Buchmann, Di Prete, and McDaniel 2008; Pekkarinen 2012). The question we examine is: Are immigrants source-country gender role attitudes passed on to their children in the receiving country? This is examined by measuring group differences in the gender inequality of university educational attainment among immigrant Sociological Inquiry, Vol. 88, No. 1, February 2018, 131 154 2017 Alpha Kappa Delta: The International Sociology Honor Society DOI: 10.1111/soin.12183

132 TERESA ABADA ET AL. children. Canada s increasingly racially and ethnically diverse immigrant population provides a useful context for examining this question (Fong and Chan 2008). It allows for an examination of whether immigrant children who come from countries with traditional norms and values assimilate to their host country s gender differences in university attainment. The theoretical frameworks proposed in examining the educational attainment of immigrant children have mainly focused on destination-country effects including individual-level family and postmigration contextual factors (Alba and Nee 1997; Perreira, Harris, and Dohoon 2006; Portes and Zhou 1993; Zhou 1997). However, they have not been able to fully explain these differing educational pathways (Feliciano 2006). Examining the extent to which sourcecountry characteristics remain salient among immigrants and their children may provide further insight into gender differences in the educational outcomes of childhood immigrants. Recent scholarship has focused on the relationship between origin effects or source-country characteristics such as female labor force participation (FLFP), fertility, level of economic development, gender role attitudes, male/female university enrollment ratios, and immigrants labor market assimilation (Blau et al. 2013; Fernandez and Fogli 2009; Frank and Hou 2016; Kaida 2013) and the division of labor within families (Frank and Hou 2015). These national-level factors are used as proxies for gender ideology when accounting for the differences in labor market outcomes across homecountry groups (Antecol 2000:412). We extend this line of inquiry to assess the saliency of immigrant parents gender attitudes among their children. Some source-country characteristics, such as fertility behavior and female labor force participation, have been found to be transmitted intergenerationally (Fernandez and Fogli 2009:147). However, the strength of such traits may be more prominent among immigrants than among their children, as the latter become acculturated to their host country s norms regarding work and family. The degree to which women from countries with traditional gender role attitudes acculturate to the high levels of women s educational attainment in the host country is indicative of the strength of the transmission of gender-related cultural norms for this generation. Employing measures of source-country characteristics provides an advantage over using country of ancestry alone because we can be explicit as to why source-country matters for immigrant groups socioeconomic outcomes and whether sourcecountry norms are a determining factor. This article explores the correlates of gender differences in educational attainment among immigrant children. The focus is on the extent to which ethnic differences in the gender gap in educational attainment among childhood immigrants those who came to the host country before age 18 is associated with source-country factors including female/male ratio in labor force

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 133 participation rate, female/male ratio in tertiary education access, gender role attitudes, and GDP per capita. More specifically, this article seeks to address the following question: Do source-country characteristics remain intact among childhood immigrants despite having been educated in Canada and being exposed to Canadian norms and values? Theoretical Background Scholarship on the socioeconomic adaptation of child immigrants has primarily relied on the segmented assimilation framework in accounting for differences in educational and labor market outcomes (Portes and Zhou 1993). The interaction between individual-level factors including language proficiency, place of birth, and age at arrival and structural background factors such as racial status, family socioeconomic background, and place of residence determines to what segment of the host society specific immigrants will assimilate (Zhou 1997:984). This framework focuses on how modes of incorporation such as the context of reception for immigrant groups, discrimination that specific groups may encounter in the destination country, and characteristics of the ethnic community may determine the differing pathways of integration. While these modes of incorporation consider postmigration attributes, this study examines the premigration group-specific characteristics of immigrants origin countries. This study builds on the notion that in addition to individual and family background factors, group-specific social structures that are influenced by source-country socioeconomic environments also play a role in the educational outcomes of immigrants children. These premigration characteristics include differences in gender ideology that can influence one s perception of what is conceivable in the host country. A growing number of studies employ measures of source-country characteristics such as female labor force participation rate and level of economic development to investigate the link between source-country norms and immigrants integration, especially immigrant women s labor force activities (Frank and Hou 2015; Blau Kahn and Papps 2011). These studies underscore the importance of origin-country characteristics as indicative of attitudes and beliefs regarding family structure and women s role in market work versus housework (Antecol 2000). Notably, source-country female labor force participation level has been widely regarded as an important proxy for a society s gender ideology as it reflects the market work decisions of women as conditioned by the economic and institutional environment as well as their individual characteristics (Fernandez 2011). The preference to work (or not) underscores the perception of women s roles in the household and the workplace, whether children benefit from a mother working, and her society s views on her working (Fernandez and Fogli 2009:148). The ratio of female to male labor force activity and the

134 TERESA ABADA ET AL. ratio of female to male education represent the macrolevel gender roles in the immigrants source countries, and these cultural factors prevail in the host country after migration. Source-country female labor force participation rates in particular influence the earnings of immigrant women (Frank and Hou 2016), division of labor in immigrant families (Frank and Hou 2015), and the labor force participation of immigrant women in the host country (Antecol 2000; Blau, Kahn, and Papps 2011). The authors argue that these portable cultural factors prevail in the labor market activities of immigrants within the host country (Antecol 2000:419). These studies are based on the premise that individuals encounter with situations that are in line with feminist ideals, such as women s workforce participation and educational attainment, will lead to more progressive attitudes toward women s place in society (Bolzendahl and Myers 2004). Being raised in an environment where women are working and pursuing higher education enhances one s consciousness regarding financial independence, the ability to work and have a family, and general acceptance toward gender equity issues (Banaszak and Plutzer 1993; Bolzendahl and Myers 2004). As mothers are the primary role models for their children, those who have high levels of education and participate in the labor force lean more toward gender equality and will likely pass this ideology to their children (Bolzendahl and Myers 2004). Immigrant children who come from countries with high female labor force participation and high female educational attainment may be socialized early on in an environment that encourages the pursuit of higher education, especially among girls. It is possible that source-country female labor force participation and educational attainment reflects more than just gender role attitudes; it may also reflect the degree to which labor market institutions provide the qualifications and the expertise needed to compete in the host-country labor market. Frank and Hou (2016) find that immigrant women who come from countries with high FLFP obtain higher paying employment in the host country. This is largely explained by their occupational allocation, indicating that higher FLFP may contribute to greater labor market knowledge and familiarity with effective job search methods, which facilitate the location of better job opportunities. Immigrant women who know how to navigate labor market institutions will be in a better position to realize their children s educational aspirations because they have the knowledge to obtain information regarding the best schools, financial assistance, and requirements pertaining to the admission process. They also would be in a stronger position to encourage their daughters to pursue higher education because of greater awareness of women s opportunities in the labor market. A source-country s level of economic development may similarly prepare immigrants for competing in the host-country labor market. Origin countries

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 135 with high levels of economic development have more resources for education and the technologies for job training, hence providing immigrants from these nations with a higher quality of human capital (Coulombe, Grenier, and Nadeau 2012). Immigrants who come from countries with similar economic structures as the host country may be impelled to participate in the labor market (Blau et al. 2013). This market orientation among immigrants may translate to high expectations in their children s educational attainment. The higher quality human capital of some immigrant groups may also translate to social capital in the destination country in the form of role models, behavioral norms, academic expectations, and interpersonal networks. These positive role models can provide the motivation for immigrant children to pursue higher education (Bredtmann and Otten 2013; Zhou and Kim 2006). However, previous research in the United States has found that despite increased educational attainment, many immigrant groups do not reach parity with the native-born. In line with the segmented assimilation framework, the disadvantage of these groups is primarily attributed to structural factors (Barringer and Kassebaum 1989; Barringer, Takeuchi, and Xenos 1990). Increasing returns to education in terms of rising opportunities for women in the labor market have been noted as contributors to the reversal in the gender gap in education (Diprete and Buchmann 2013). Rising opportunities in the labor market for women in traditional female occupations such as teaching and nursing, and high-status managerial positions, the declining gender gap in wages, and a rising college wage premium particularly for female graduates have provided the incentives for further educational investments among women (DiPrete and Buchmann 2006). The gradual decline of discrimination in the labor market for women also provided the impetus for women to enter higher education (Vincent-Lancrin 2008). Educational attainment is less productive in countries with lower female labor force participation, whereas in more developed countries, opportunities for social mobility abound for highly educated women (Fenella and Kristen 2014). Immigrant women who come from countries with high female labor force participation and higher levels of economic development would have encountered greater returns to their education prior to moving to the host country. Because of their premigration awareness of the benefits to higher education, they likely possess the incentive to invest in their daughter s education to ensure that they too will have the same opportunities for advancement in the labor market. On the other hand, immigrants who come from countries with low female labor force participation may not see the advantage of educational mobility for their daughters and instead place greater value on men s education over women.

136 TERESA ABADA ET AL. Will Source-Country Gender Norms Prevail Among Immigrant Children? Qualitative research reveals the intricacies of the negotiation of gender roles among new immigrants. As traditional gender roles are regarded as central to the ethnic identity of many immigrant groups, recent immigrants may reinforce these values in an effort to ensure the successful transfer of culturally significant behaviors to their children (Dasgupta 1998: 954). Therefore, while many immigrant women regard gender equality in their host countries positively, there is evidence that they are careful to maintain traditional gender roles within their families as a means of retaining ethnic identity and traditions (Dasgupta 1998; Kibria 1990). Strong ethnic communities in the host country often play an important role in maintaining the norms and values of the source country. However, it is also clear that the negotiation of gender roles is dynamic and affected by migration. While the gender ideology espoused in the source country may continue to be highly influential within the family structure in the host country, changes also arise from external social and economic forces within the host country (Foner 1997:961). Increased exposure to these forces may lead to an alteration of gender attitudes in some realms. Childhood immigrants, or the 1.5 generation, are likely socialized into the norms and values of the host society and are exposed to the same educational opportunities and constraints as the children of native-born parents (Alba and Nee 1997). The degree to which the 1.5 generation adapt to the norms of their host country may be tempered by distinct differences reflected in their parents attitudes and beliefs, and their own place of birth (Rumbaut 2012). Sourcecountry gender role attitudes can exert a separate influence on immigrant children s educational attainment, independent of the origin country s female labor force participation and tertiary educational access. Immigrants exposed to a lifelong socialization of traditional gender attitudes may perpetuate these traditional gendered identities to their children, consequently affecting their educational and labor-related choices. On the other hand, immigrant women who come from countries with more traditional attitudes toward gender roles may find that adoption of more egalitarian attitudes in the host countries serves their interests well because it provides the means to overcome barriers and challenge systems of inequality that they encountered in their countries of origin (Bolzendahl and Myers 2004). Challenging traditional gender ideologies will likely manifest in immigrants investing equally in their children s education regardless of gender. Immigrant children may view traditional gender norms as an inherent part of their culture. However, they also enter an educational system early on where traditional values regarding work and family are questioned and even challenged. R oder and M uhlau (2014) find that immigrants adapt their gender

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 137 ideology to the host-country values and that the context of the origin country does not exert an influence on the gender attitudes of immigrant children. Therefore, the 1.5 generation may exhibit more optimistic attitudes toward their educational goals, despite coming from countries with low FLFP and low tertiary education, as well as traditional gender ideologies regarding work and family. An examination of the educational attainment of the 1.5 generation allows for an assessment of the transmission of gender attitudes among children of immigrants whose source-country norms differ significantly from those of the host country. Data, Measures, and Methods Data The Canadian 2006 census 20 percent sample microdata file is the source of individual-level variables on educational attainment and other socio-demographic characteristics. The study focuses on 1.5-generation immigrant youth aged 25 34. Immigrants in the 1.5 generation are defined as those who immigrated to Canada before age 18. For the purpose of comparison, university completion rates by gender are also presented for the second-generation, the 2.5-generation, and the third-and-higher-generation youth in 2006, as well as for their parents generation in 1996 when the selected youth groups were in or approaching the ages of university education. The second generation is defined as those who were born in Canada to two immigrant parents. The 2.5 generation is defined as those who were born in Canada to one immigrant parent and one Canadian-born parent. The third-and-higher generation is defined as those who were born to two Canadian-born parents. For youth aged 25 34, their parents generation is defined as adults aged 40 49 in 1996 when this selected cohort of youth were 15 24 years old. To examine the effect of source-country attributes on the gender difference in university completion rates among the 1.5-generation groups, the 1.5-generation immigrant youth selected from the 2006 census microdata file are matched with the database on source-country attributes based on detailed country of birth and year of arrival. Put differently, childhood immigrants in the census file are assigned a set of his/her source-country attributes measured at the year in which they came to Canada. The data for immigrant source-country attributes are compiled from various sources (see the Measures section). Two criteria are used to select 1.5-generation immigrants aged 25 34 to study the effect of source-country attributes. First, immigrants have to come from a source country that has at least 20 observations in the 2006 census microdata file. This is to ensure that the gender difference in educational attainment at the group level can be reliably estimated. Second, immigrants have to

138 TERESA ABADA ET AL. come from a source country for which data on all the selected source-country attributes are available (see the Measures section for details). There are a total of 73 immigrant groups that meet these two criteria. The final sample size at the individual level is 44,106. About 14,000 1.5-generation immigrants aged 25 34 are excluded from the analysis because source-country attributes are not available or the group sample size is smaller than 20. Measures The outcome variable is university completion which is defined as obtaining a university degree. At the group level, the outcome of interest is the gender difference in university completion rates. The focal explanatory variables are three indicators that have been used to reflect the degree of traditional gender roles in immigrants source countries (e.g., Blau et al. 2013; Frank and Hou 2016). First, the female-to-male labor activity ratio in the source country (hereafter referred to as female/male labor activity) is the ratio of women s labor force participation rate to men s labor force participation rate. These rates are calculated for individuals aged 15 and over. 1 The ratio of female to male labor force participation rates reflects the relative differences between women and men in labor force participation and thus captures the gender division of labor explicitly (Blau, Kahn, and Papps 2011). This relative indicator can mitigate the problems associated with cross-country differences in the definition and measurement of labor force participation because such problems may affect women s and men s labor force participation rates similarly (Antecol 2000; Blau, Kahn, and Papps 2011). The second indicator is the ratio of the female enrollment rate in tertiary education to the male enrollment rate in tertiary education (hereafter referred to as female/male education). 2 This ratio reflects the gender difference in accessing tertiary education within a given education system. The third focal source-country attribute is an explicit measure of gender role attitudes derived from the World Values Survey (WVS). The WVS collects measures of people s beliefs, values, and attitudes from nationally representative samples. The questions related to gender role attitudes are available for 83 countries that participated in at least one of waves 3 5 of the WVS over the 1994-to-2007 period. 3 Three questions are used to derive the scale representing gender role attitudes: (1) Do you agree or disagree that when jobs are scarce, men should have more right to a job than women? (2) Do you agree strongly, agree, disagree, or disagree strongly that a university education is more important for a boy than for a girl? (3) Do you agree strongly, agree, disagree, or disagree strongly that on the whole, men make better political leaders than women do? These statements reflect beliefs about whether women should have equal access to jobs and higher education and whether they have the ability to take on leadership roles. Each item was coded from 1 = agree strongly to

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 139 4 = disagree strongly. 4 The final scale was constructed from the mean of the three items. This scale has good reliability, with a Cronbach alpha of.65. The values of the scale range from 1.83 to 3.59, with a mean of 2.70. A high score of the scale indicates a more favorable attitude toward women s roles in the society. It should be noted that not all source countries have lower gender attitude scores than the host country. While the score for Canada is quite high at 3.28, there are six countries that have higher gender attitude scores: the Netherlands, Norway, Sweden, Finland, France, and Ethiopia. However, the vast majority of countries studied (92%) have a lower gender attitude score than Canada. Source-country GDP per capita (in 2005 U.S. dollars) is also included as an indicator of a source-country s overall level of economic development. 5 Immigrants from more developed countries may be more likely to support gender equality and women s education. Therefore, this control variable ensures that any observed association between the source-country gender role indicator and gender differences in university completion among 1.5-generation youth is independent of source-country economic development. The gender differences in university completion among the 1.5 generation s parents generation is another control variable used in the multivariate analysis. This variable is measured as the female/male ratio in university completion among adult immigrants aged 40 49 in 1996 by source country. This control variable serves two purposes. First, it will show whether any observed effects of source-country gender roles on gender differences in university education among the 1.5 generation work through gender differences in university completion within the parents generation. If parents generation gender differences in university completion play an intermediate role, its inclusion would reduce or account for the observed association between source-country gender roles and gender differences in university completion among the 1.5 generation. Second, it serves as a control for immigration selection at the group level. Immigrants tend to be different from their source-country population in socioeconomic status, behaviors, and attitudes. This selectivity may tend to render the association between source-country attributes. When source-country attributes are controlled for, immigrant groups with a higher female/male ratio in university completion have a higher degree of selection in gender role attitudes. At the individual level, the following explanatory variables are included as controls: age at immigration, years since immigration, racial minority groups (blacks, Asians, Latinos, and others, with whites as the common reference group), mother tongue (English or French=1; others=0), and geographic region of residence (living in metropolitan areas=1; others=0). At the group level, four variables are included as controls: the share of mothers with a university degree, the share of mothers speaking an official language, the share of lone mothers, and the population share of own immigrant group in the municipality where

140 TERESA ABADA ET AL. immigrant parents lived. These group-level variables are derived from adult immigrants aged 40 49 in 1996 by source country. They are used to capture parental human capital (education and official language ability), parental family structure, and the context of reception. These measures are also derived from fathers, and their inclusion in multivariate models produces similar results. Methods The analysis starts with summary statistics on university completion rates for youth and their parents generation by gender and immigration status. It is followed by a detailed look at gender differences in university completion rates for 1.5-generation youth and their parents generation by source country. Simple correlations are then calculated to examine whether the gender differences in university completion among the 1.5 generation and their parents generation are significantly associated with source-country attributes. Multivariate logistic regression analysis is conducted to examine whether each of the three source-country gender role variables is significantly associated with the gender differences in university completion among 1.5-generation youth when source-country economic development, basic demographic factors, group-level parental characteristics, and the gender difference in university completion among their parents generation are controlled for. The dependent variable is a dichotomous variable coded as 1 if a 1.5-generation youth obtained a university degree and 0 otherwise. Individual-level explanatory variables include women (coded 1 for women and 0 for men), as well as the demographic variables defined in the Measures section. Explanatory variables at the source region level include the three gender role indicators, GDP per capita, gender differences in university completion for the parents generation, parental education, language ability, family structure, and ethnic enclave. Thus, the outcome variable is predicted by both individual- and grouplevel variables. As individuals from the same source country have the same value on the group-level variables, there is a problem of dependence for observations drawn from the same source country. To address this issue, we use multilevel logistic models (Raudenbush et al. 2000). The two-level logistic model includes both random effects for the intercept and for the slope for the variable representing gender (women=1). Specifically, the intercept is allowed to vary across source countries. The intercept for each source country represents the overall university completion rate for 1.5-generation youth from that source country when group differences in individual-level demographic factors are taken into account. The variation of the intercepts across source countries is predicted by the selected source-country-level explanatory variables. Furthermore, the coefficient of the variable women is also allowed to vary across source countries. The coefficient of this variable for

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 141 each country represents the gender difference (the advantage of women over men) in university completion among 1.5-generation youth from that country. The variation of these coefficients is predicted by the selected source-countrylevel explanatory variables. Results Table 1 presents a broad picture of the gender difference in university completion. In 2006, among young adults aged 25 34, regardless of generational status, women have much higher university completion rates than men. Both female and male children of immigrant parents (i.e., including 1.5, 2nd, and 2.5 generations) have higher university completion rates than their thirdand-higher generation counterparts. The pattern of gender differences is the opposite among their parents generation, where women had lower university completion rates than men for both immigrants and the Canadian-born. The remarkable reversal of women s relative position in university completion Table 1 Percent with University Degrees, by Sex and Generation Status Women Percent Men Difference (women men) Percentage point Ratio (women/men) Sample size Youth aged 25 34 in 2006 1.5 generation 37.2 30.3 7.0 1.23 44,106 2nd generation 43.7 32.1 11.6 1.36 77,699 2.5 generation 37.8 27.6 10.2 1.37 57,712 3rd-and-higher generation 27.1 18.4 8.7 1.47 503,243 Parent generation in 1996 Immigrants 21.3 28.7 7.4.74 283,480 aged 40 49 Canadian-born aged 40 49 15.3 17.7 2.4.86 761,917 Sources: Statistics Canada, the 1996 and 2006 Census of Population, 20 percent sample microdata file.

142 TERESA ABADA ET AL. clearly accentuates the widespread improvement of women s socioeconomic status over the last few decades. The observation that 1.5-generation women s advantage in university completion is the smallest among the four groups of youth by generational status may reflect the possibility that women s advancement in educational attainment is lower among children of immigrants from certain source regions where norms and attitudes toward women s role are less favorable than in Canada. In other words, source-country gender roles may have an impact on the 1.5-generation youth. If this is true, we would expect a large variation in the gender differences in university completion among 1.5-generation youth from different source countries, and this variation should be associated with source-country gender role indicators. The data in Table 2 confirm a large variation in the gender difference in university completion among 1.5-generation youth by source country (left panel). To save space, Table 2 only shows university completion by gender in 2006 for 1.5-generation youth groups with a minimum sample size of 300. With the exception of those from Pakistan, young women have higher university completion rates than men across all listed 1.5-generation groups (left panel, Table 2). 6 These patterns are in sharp contrast to the gender differences among their parents generation (right panel, Table 2). Among their parents generation, adult immigrant women from Iran, South Africa, Malaysia, Hong Kong, and Taiwan had university completion rates that were lower than immigrant men by at least 15 percent points. Only adult immigrant women from Chile, Germany, and France had a slightly higher university completion rate than men. These results indicate that the large advancements in women s education relative to men from the parents generation to the 1.5 generation occurred in essentially all immigrant groups regardless of the source country. However, the gender difference in university completion does vary considerably between the groups of 1.5-generation youth. Among the large groups listed in Table 2, 1.5-generation women from South Korea, Malaysia, Italy, Poland, Romania, France, the United Kingdom, and Taiwan lead men in university completion rates by 10 percent points. In the case of 1.5-generation Italian immigrants, the gender gap is driven mostly by a very low university completion rate among men. This is likely related to the fact that Italian immigrant parents had very low educational levels (Abada, Hou, and Ram 2009). The gender differences are small among those from India, Mexico, Pakistan, and the United States. The group variation in the gender difference in university completion rates among the 1.5 generation does not correspond to the group variation among their parents generation. For instance, while the gender difference greatly favored men among adult immigrants from East Asia (Hong Kong, Taiwan,

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 143 Table 2 University Completion Rates by Sex and Generational Status Proportion with university degrees among 1.5 generation aged 25 34, 2006 Proportion with university degrees among adult immigrants aged 40 49, 1996 Women Men Difference (women men) Women Men Difference (women men) Source country Percent Percentage point Percent Percentage point The United Kingdom 41.1 30.6 10.5 14.3 24.0 9.8 Viet Nam 36.9 29.5 7.4 9.4 15.7 6.3 Hong Kong 69.5 59.7 9.8 15.7 32.0 16.3 Philippines 25.3 17.6 7.7 40.3 42.5 2.2 Poland 42.8 29.0 13.8 26.2 29.7 3.5 India 29.3 27.4 1.9 34.6 40.2 5.6 The United States 34.2 31.4 2.7 45.8 55.1 9.3 China 53.2 44.0 9.2 13.8 24.8 11.0 El Salvador 14.2 11.8 2.4 7.8 10.7 2.9 Taiwan 67.9 57.6 10.3 28.1 44.0 15.9 South Korea 59.3 42.6 16.6 29.1 42.6 13.5 Iran 52.6 44.8 7.8 35.3 53.3 18.0 Mexico 7.7 6.3 1.3 15.5 23.2 7.7 Pakistan 39.0 41.1 2.1 35.7 45.0 9.2

144 TERESA ABADA ET AL. Table 2 (continued) Proportion with university degrees among 1.5 generation aged 25 34, 2006 Proportion with university degrees among adult immigrants aged 40 49, 1996 Women Men Difference (women men) Women Men Difference (women men) Source country Percent Percentage point Percent Percentage point Trinidad and Tobago 30.3 21.6 8.7 9.5 12.1 2.6 Germany 32.4 23.7 8.7 22.7 21.3 1.5 Chile 27.1 20.8 6.3 18.5 14.5 4.0 Romania 48.9 37.0 11.9 43.5 50.9 7.4 South Africa 55.1 45.1 10.0 30.4 48.0 17.5 France 44.0 33.1 10.8 33.7 33.4.3 Guatemala 12.6 9.7 2.9 4.9 15.3 10.4 the Netherlands 35.5 26.3 9.2 11.5 19.1 7.7 Malaysia 61.3 46.2 15.2 14.2 31.3 17.1 Italy 32.2 18.2 14.0 5.8 7.2 1.4 Peru 29.0 20.2 8.7 22.3 34.7 12.4 Sources: Statistics Canada, the 2006 Census of Population, 20 percent sample microdata file. Notes: The sample size for 1.5 generation ranges from 311 (Peru) to 4,507 (the United Kingdom), and from adult immigrants in 1996 from 207 (Guatemala) to 5,095 (the United Kingdom).

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 145 South Korea, Hong Kong, China), the pattern reversed among the 1.5 generation. Indeed, the correlation between the gender difference among the 1.5 generation and the gender difference among their parents generation is only.08 and not statistically significant. The lack of a strong association between the differences in educational attainment among 1.5-generation youth and their parents generation likely indicates that source-country attributes had different effects on adult immigrants and their children. Further evidence is provided in Table 3 which shows the bivariate correlations among the gender difference in university completion rate among 1.5-generation groups, the gender difference in university completion rate among their parents generation groups, and four source-country attributes. Among the parents generation, the gender difference is significantly correlated with source-country female/male ratio in tertiary education access and gender role attitudes, but not with source-country female/male ratio in labor force participation rate and the level of economic development (log GDP per capita). Among the 1.5 generation, the gender difference is significantly associated with source-country female/male ratio in labor force participation rate and the level of economic development. These bivariate correlations suggest that sourcecountry attributes are selectively passed on to childhood immigrants. The correlations among the four source-country attributes are generally small or moderate. The multilevel linear model results in Table 4 also confirm that source-country female/male ratio in labor force participation and source-country economic development tend to increase women s advantage in university completion, while the effects of source-country female male tertiary education access and sourcecountry gender roles are not significant. Model 1 in Table 4 shows that 1.5-generation female youth have a higher university completion rate than male youth even after controlling for age at immigration, years since immigration, racial minority groups, mother tongue, and geographic location. Younger age at immigration, more years since immigration, and speaking English or French as mother tongue are associated with a higher likelihood of finishing university education. Relative to whites, 1.5-generation Asian youth have higher completion rates while black youth have lower university completion rates. Model 2 adds group-level parental characteristics and source-country attributes to predict the variation in the intercepts (the average university completion rates among 1.5-generation groups by source country) and the variation in the slopes for women (the lead of women over men in the university completion rate among 1.5-generation groups by source country). Among the four group-level measures of parental characteristics, only parental education is significantly associated with a higher likelihood of completing university education.

146 TERESA ABADA ET AL. Table 3 Simple Correlation Between Gender Differences in University Completion Rates and Selected Source-Country Characteristics 1 2 3 4 5 1 Gender difference (women men) in the share with university degrees among 1.5 generation in 2006 2 Gender difference (women men) in the share with university degrees among adult immigrants in 1996 3 Female/men ratio in labor force participation rate.250*.183 4 Female/men ratio in tertiary education access.135.443***.061 5 Gender role attitude score.193.322**.249*.205 6 Log GDP per capita.296*.086.134.345**.568***.081 *Significant at p <.10; **p <.01; ***p <.001. The total sample size is 44,106. Sources: Statistics Canada, the 2006 Census of Population, 20 percent sample microdata file.

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 147 Table 4 Multilevel Logistic Regression Predicting University Completion Model 1 Model 2 Coefficient Standard error Coefficient Standard error Fixed effects Intercept 1.929***.146 3.474***.936 Women.401***.021.144.199 Individual-level characteristics Age at immigration.021***.004.020***.004 Years since immigration.019***.004.032***.004 Blacks.415***.097.392***.097 Asians.574***.049.570***.049 Latinos and other.163*.065.178**.065 racial minorities English/French mother tongue.144***.033.141***.033 Living in metropolitan areas.781***.050.779***.050 Parental characteristics The share of mothers with a university degree 2.624***.604 The share of mothers.761.977 speaking an official language The share of lone mothers.393.852 The share of city population from the same immigrant group.180.190 Source-country attributes and cross-level effects Source-country female/male ratio in labor force participation Women*source-country female/male ratio in labour force participation Source-country female/male ratio in tertiary education access 1.070***.214.494***.129.001.067

148 TERESA ABADA ET AL. Table 4 (continued) Model 1 Model 2 Coefficient Standard error Coefficient Standard error Women*source-country.082.077 female/male tertiary education access Source-country.242.207 gender role attitudes Women*source-country.151.090 gender role attitudes Source-country GDP per capita.084*.040 Women*source-country.067***.019 GDP per capita Female/male ratio in university.876*.343 completion among parents Women*female/male.063.131 ratio in university completion among parents Random effects Intercept.014***.009*** Women.002**.001* Residual 1.084*** 1.087*** *Significant at p <.10; **p <.01; ***p <.001. The sample size is 44,106. Sources: Statistics Canada, the 2006 Census of Population, 20 percent sample microdata file. The main effects of source-country factors show that source-country female/male ratio in labor force participation and source-country GDP per capita are significantly and positively associated with average university completion rates among the 1.5-generation groups. The negative effect of sourcecountry gender role attitudes reflects the fact that university completion rates are not particularly high among 1.5-generation youth from the United States and Northern and Western Europe where gender role attitudes favor equality

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 149 between men and women, while university completion rates are high among 1.5-generation groups from East, South, and West Asia where gender role attitudes are much less favorable to women. The cross-level effects between source-country factors and women show that women s advantage tends to be stronger among 1.5-generation groups from countries with higher source-country female/male ratios in labor force participation and higher GDP per capita. Source-country gender role attitudes and source-country female/male ratio in tertiary education access are not significantly associated with gender differences in university completion rates among the 1.5-generation groups. The female/male ratio in university completion among parents generation is included as a way to control for immigration selectivity and to test whether the effects of source-country attributes on 1.5-generation youth are mediated through parents. Its main effect was negative and statistically significant. This is due mostly to its high correlation with the measure of parental education (the share of mothers with a university degree). When the latter is not included in the model, the main effect of the female/male ratio in university completion among parents generation is positive but not significant. Its cross-level effect is not statistically significant with or without the control of the measure of parental education. Discussion and Conclusion This study examined the extent to which source-country characteristics continue to play a role in the educational outcomes among immigrant children. Analyses of 1.5-generation youth in Canada revealed that among the 73 groups studied, with the exception of child immigrants from Pakistan, women achieved higher rates of university education than their male counterparts regardless of the gender equality status in their source countries. Among the parents generation, immigrant and native-born women had lower university completion rates than men. The pattern was reversed with the passing of one generation for both immigrants and nonimmigrants. This reversal in the gender gap is observed not only in North America, but also in most industrialized countries and in other parts of the world (Diprete and Buchmann 2013; Feliciano and Rumbaut 2005; Vincent-Lancrin 2008). There was also large ethnic variation among 1.5-generation groups. The largest gender difference was observed among immigrant children from South Korea, Malaysia, Italy, Poland, France, Romania, and Taiwan. Female child immigrants who come from countries with high female labor force participation and high levels of GDP have an advantage over their male counterparts in university education. These findings are in line with Diprete and Buchmann (2013) argument about the returns to women s education and opportunities for women to succeed in the labor market. Higher levels of female labor force participation may condition individuals to believe that women can also compete in the labor market and are as capable

150 TERESA ABADA ET AL. of financial independence as men. Immigrants may then carry these attitudes over to their destination countries, subsequently shaping their educational expectations for their children. When these beliefs interact with the economic and social conditions in the host country, it is not surprising that parents will push their children to take advantage of opportunities that will enhance their chances for upward mobility. Immigrant parents who come from countries with high levels of GDP may also have a sense of collective outlook that makes them feel entitled to mainstream success in the host country (Feliciano 2006; Ogbu 1991). This entitlement can include aspirations for their children to succeed in school. The results of this study demonstrated that while source-country gender role attitudes played a role in university completion rates for immigrant parents, this was no longer the case among child immigrants. Female immigrants, especially those who come from countries with more traditional gender ideologies, may no longer adhere to the traditional norms that had previously placed them at a disadvantage, and may instead welcome the egalitarian gender attitudes of their host country (R oder and M uhlau 2014). Moreover, it is possible that the erosion in gender stereotypes in the host society may lead to a decline in traditional gender role attitudes within immigrant families leading parents to be supportive of the education of their daughters (Vincent-Lancrin 2008). Even among groups in which 1.5-generation women still lagged slightly behind men in university completion, such as Pakistani, it can still be argued that the opportunities found in the host country may trump traditional gender attitudes. Given the strict gendered norms represented within Pakistani society, the daughters have achieved a great deal of educational mobility in Canada. The high educational achievement in university completion among female immigrant youth may reflect the degree to which these patriarchal ideologies have been challenged. Another reason for the diminished impact of source-country gender ideology may be related to possible selective immigration from countries with more traditional gender roles. For example, the political and social changes following the Islamic revolution in Iran promoted changes that decreased educational opportunities for women, emphasized women s domesticity, and enforced obedience to male kin (Hakimzadeh 2006). This provided some with the impetus to seek better economic opportunities elsewhere and a more favorable social and political climate that is conducive to elevating women s position in society. Immigrant women who come from countries that hold less egalitarian attitudes may be more open to feminist ideals especially if it serves their individual interests and if it enhances their chances for upward mobility (Davis and Robinson 1991). This new found independence and the belief that education can enhance one s bargaining position within traditional households likely play a role in their children s academic pursuits, particularly

GENDER DIFFERENCES IN EDUCATIONAL ATTAINMENT 151 among their daughters (Zhou and Bankston 2001). Further, many immigrants cite greater opportunities for their children as a key motivation for moving to Canada (Houle and Schellenberg 2010). Thus, immigrants who migrated for this reason are more likely to be supportive of their daughters pursuing higher education. In the literature, origin-country effects are not deemed as salient for immigrant children as they are for immigrant parents. Our study makes two important contributions in understanding the effects of source-country effects. First, this study has shown that the structural factors were more important than gender role attitudes in ascertaining university completion rates. The characteristics of the origin countries provide a glimpse into the differences in gender attitudes among immigrant groups, specifically the perception of the role of women in the household and in the workplace, and how this ultimately conditions what is achievable in the host country. The female advantage observed among the 1.5 generation in tertiary education reflects the trend in most Western countries. Our study found that source-country gender role attitudes were not significant in the gender differences observed for the university attainment of the 1.5 generation. This supports the theory that sourcecountry proscriptions regarding women s role in society may play a more inherent role in areas that are considered to be private family matters, such as the household division of labor. Future studies could further investigate this possibility by examining whether source-country gender norms also play a role in these aspects of integration including conflicts related to issues of housework and work life balance. Second, it is critical to distinguish 1.5 generation and second generation among children of immigrants in studying the effects of source-country characteristics. When we replicate our analysis for the second generation, we find that there is little variation in the gender difference in university completion by parents origin country, and this little variation is not significantly associated with any of the three origin-country traditional gender roles and with origin-country GDP per capita (tables are available on request). Other Canadian studies show that immigrant parents of the second generation immigrated at an age that is, on average, 10 12 years younger than parents of the 1.5 generation, and many of them are 1.5 generation themselves (Bonikowska and Hou 2010). Therefore, there are large differences between parents of the 1.5 generation and second generation in their exposure to the origin society and socioeconomic integration into the receiving society. Furthermore, the 1.5 generation lived some of their childhood years in their source country and thus are likely more influenced by the gender norms and socioeconomic conditions of their source country than the second generation who are born and raised in the receiving country.