The Pennsylvania State University. The Graduate School. College of Education MIGRATION, FAMILY TYPES, CHILDREN'S EDUCATION AND WORK

Transcription

1 The Pennsylvania State University The Graduate School College of Education MIGRATION, FAMILY TYPES, CHILDREN'S EDUCATION AND WORK PARTICIPATION IN MEXICO: WHO LEAVES, WHO STAYS, AND DOES IT MATTER? A Thesis in Educational Theory and Policy by Xiao Yu 2013 Xiao Yu Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Arts May 2013

2 The thesis of Xiao Yu was reviewed and approved* by the following: Katerina Bodovski Assistant Professor of Education Thesis Advisor David Post Professor of Education Gerald K. Letendre Professor of Education and International Affaires Head of the Department of Department or Graduate Program *Signatures are on file in the Graduate School

3 iii ABSTRACT Labor migration and marital dissolution both contribute to family separation and, through separation, affect children's decisions about education and work. This paper compares the impact of different types of family separation, including labor migration and marital dissolution, on educational enrollment and work participation among older adolescents in Mexico. It also compares the educational expectations, especially those of younger left-behind children in Mexico depending on different types of family separation. Using the nationally representative 2005 Mexican Family Life Survey, I first compared children aged who lived in non-migrant, twoparent households with children who experienced labor migration and children who lived in lone-parent households. Both children living in households of family separation are less likely to attend school, even though the results are not statistically significant when further controlling for socioeconomic status. In addition, children living in migrant households are more likely only to work without enrolling in school, and they are also more likely to report neither work nor schooling. Children living in single-parent households are more likely to combine school and work. In terms of educational expectations, after identifying who stays and who leaves the household in both types of family separations, I find that exposure to migration (either sibling migration or paternal migration), is negatively associated with aspirations for postsecondary education. Children ages who attended school, and who lived in households with migrating members, were even less likely to aspire to college than were children living in single-mother households. In addition, and surprisingly, children living with divorced/separated mother were more likely to aspire to postsecondary education.

4 iv TABLE OF CONTENTS LIST OF TABLES... v ACKNOWLEDGEMENTS... vi Chapter 1 INTRODUCTION... 1 Chapter 2 THERETICAL FRAMEWORK AND HYPOTHESES... 5 Household Migration: Potential Benefits vs. Negative Effects... 5 Labor Migration and Marital Dissolution... 9 Who Stays and Who Leaves Chapter 3 DATA AND METHODS Data and Sample Measurements Explanatory Variables Control Variables Dependent Variables Methods Chapter 4 RESULTS Descriptive Statistics School Enrollment of Older Adolescents School and Work Status of Older Adolescents Educational Expectations Chapter 5 CONCLUSION AND DISCUSSION Bibliography... 52

5 v LIST OF TABLES Table 4-1: Descriptive Statistics of Individual, Household and Community-level Characteristics by Family Types Table 4-2: Descriptive Statistics of Dependent Variables by Family Types Table 4-3: Logistic Regression on School Enrollment by Family Types, Children Aged Table 4-4: Relative Risk Ratio of Multinomial Logistic Regression on School and Work Status by Family Types, Children Aged Table 4-5: Probabilities of Only School Without Work By Family Types Table 4-6: Odds Ratio of Logistic Regression on Educational Expectation for College by Family Types, Children Aged Table 4-7: Odds Ratio of Logistic Regression on Educational Expectation for College by Family Types, Children Aged

6 vi ACKNOWLEDGEMENTS I owe my great thanks to my committee, who generously offered their time, support and feedback. I would like to thank my advisor, Katerina Bodovski, who has patiently helped developed this study design during a summer individual study. As a novice stepping into the field of sociology, it was my first time to deal with a real data set. She had been generously devoting her time to the development of literature review, methodological consideration and the refinement of measurements. In addition, as my program advisor, she had constantly supported me throughout the two years, assisting me to planning for graduate years, encouraging me with her warm words and helping me weather some the most difficult times when I felt uncertain and unconfident. A special thank goes to Professor Post, David, who has been a continual source of support and encouragement ever since I took his course of Educational Mobility and Stratification. There is nothing adequate I could say to fully express my gratitude to him. He had devoted his spare time even at weekends, even after long trips back from abroad, to checking my progress, reading the results of data analyses and editing the final paper for multiple times. He had even put me in contact with professors who previously utilized this data set as a source of assistance. His unwavering friendship and support has motivated me to strive harder for a scholar and made me feel no more helpless. I have been very lucky in having him as my professor. I cannot imagine how I was able to overcome all the difficulties without his support and care. I would also like to thank Professor Zhang, Liang. Without his course in Data Analysis in Educational Research, I would not have gained training in learning and using STATA. He has also generously spent his office hours helping me figuring out how to deal with this complicated data set, which all contributed to the final accomplishment of this paper.

7 vii I would also like to thank my parents, who gave me unconditional love and support even they were far away from the United States. They have sacrificed a lot for supporting my dream and ambition. I owe my greatest gratitude to them.

8 Chapter 1 INTRODUCTION Mexico-U.S. migration has been the focus of researchers and policy-makers for many years. In addition to a large body of research examining the motivations and networks of migrants themselves, as well as studies on assimilation patterns in destination countries, researchers have considered the impact of migration on the origin country. Aside from macro-level studies on national development from remittances and debates between "brain gain" and "brain drain", most research has focused on the individual lives of the non-migrants, because "the actions of migrants can have a major impact on the life chances of household members left at home" (Townsend et al, 1997). In this investigation, I examine how household migration may affect the well-being of left-behind children in Mexico. Most researchers suppose that potential benefits from remittances, sent home by migrating household members, can be offset by disruption of family life due to parentchild separation. Father-child separation due to migration has been common in Mexico since the Bracero Program since 1940s. With persistent inflows of Mexican migrants, it is estimated nationally that nearly as many Mexican children live in lone-mother households due to absent migrant fathers as live with a single mother following union dissolution. Overall, seventeen percent of children born into two-parent homes are expected to experience a migrating father at least once during childhood (Nobles, 2006). In addition, at the turn of twenty-first century, female migration has also grown. An

9 2 increasing number of unmarried, widowed or divorced Mexican women are heading north for better opportunities (Cerrutti and Massey, 2001; Chavez, 1992; Donato, 1993; Knaiaupuni, 2000), which results in mother's absence. The divorce rate has nearly doubled, to 14 in 100 marriages, over the past twenty years. In addition, non-marital fertility is also on the rise (Nobles, 2011). Both trends account for the decline of parentchild co-residence. However, research indicates that a present mother and a present father do not have the same meaning. Since father-child separation is quite common among Mexican children, it is the caregiver's absence (usually the mother's) that has the most deleterious effects for children's well-being (Lehaie et al, 2009; Heymann et al 2009; Levision et al, 2008). Research on school attendance and educational achievement suggests that "Mexico is lagging behind given its level of economic development" (IADB 1996; Levision et al, 2001). About half of all school-aged children are experiencing early transition into adulthood by leaving school and entering labor force before the age of seventeen, especially for boys (Giorguli, 2006). Although Mexican law states that children younger than the age of 14 are not permitted to work, and children aged 14 or 15 are allowed to work up to six hours provided that the job is not dangerous and poses no interruption for secondary education, both labor and education laws are widely violated (Levision et al, 2001). Family structure and context both shape the transition into adulthood, and in particular they affect education outcomes and future labor force participation, which are crucial for individual trajectories of social mobility. Therefore, this study addresses three questions:

10 3 First, to what extent does family separation, including both labor migration and marital dissolution, contribute to children s education and work participation? Second, does family separation resulting from migration differ from separation of marital dissolution, particularly in terms of who stays and who leaves? Third, under what contexts will children benefit from family migration, for example, will children benefit from migration if no parent-child separation is experienced? Addressing these questions can illuminate several issues. By exploring the education and work outcomes of children living in different household arrangements through multivariate analyses, this study sheds light on whether migration and marital dissolution are distinctive experiences for adolescent children in Mexico, and can also reveal the corresponding determinants involved in these processes. Second, previous studies on family migration and marital dissolution rarely take into account who stays and who leaves. By identifying effects of migration and absence by particular family members, this research shows how children fare under specific migration and marital dissolution contexts. Third, by introducing indicators of work participation and educational expectation in addition to school enrollment, this study tests competing theories of the effects of household migration on children's activities in home institutions. Is it due to a "culture of migration" (Kandel and Massey, 2002), a culture where children are provided with alternative channels of economic mobility, which leads to their orientations towards opportunities outside Mexico and reduces their likelihood of

11 participating in the local labor market? Alternatively, is it because of the loss of domestic labor that work participation increases? 4

12 5 Chapter 2 THERETICAL FRAMEWORK AND HYPOTHESES Household Migration: Potential Benefits vs. Negative Effects Previous studies on the effects of household migration on left-behind children are not conclusive, due to various unobserved factors simultaneously affecting children's school and work outcomes. Conceptually, the potential benefits from household migration can result from three mechanisms: released household credit constraints from remittances, acknowledging the sacrifices of parental migration by children, and the shift of decision-making power for left-behind mother. According to New Economics of Labor Migration theory (Sana and Massey, 2005), household migration is used as a strategy to manage household risks and future investment. Therefore, the remittances from a parent s migration can release household credit constraints and allow families to more freely invest in children's education and to less depend on child labor to contribute to household economy. Previous research suggested that remittances help school retention, increase schooling of children and reduce incidence of child labor in developing countries from Pakistan, El Salvador and Mexico (Mansuri, 2006; Acosta, 2006; Hanson and Woodcuff 2003; Edwards and Ureta, 2003; Taylor and Mora, 2006; López-Córdova, 2005; Antman, 2012). However, because remittances may promote new family businesses, children of migrant parents may be required to work in a family business. In addition to potential benefits from remittances, children of migrant households could possibly acknowledge the sacrifices of parental migration and thus have more

13 6 incentives to study hard. Due to the tightened U.S. immigration policy, most Mexican migrants enter into U.S. without legal documents, ending up doing low-status jobs and experiencing downward social mobility. Therefore, the sacrifices made by parents through migration may motivate children to adopt a pattern of "compensatory achievement" (Suarez-Orozco, 2002). Using surveys and interviews in Mixteca of Mexico, Dreby (2011) indicated that children of unmarried migrating mothers have the highest educational aspiration by acknowledging the sacrifices, compared to children of both migrating parents and those of a migrating father. Further, the modal migration pattern in Mexico is one where the father migrates and mother remains (Dreby, 2011). When the father is absent, women will have a greater say in the decisions of distributions of household resources (Giorguli 2006), which may have beneficial outcomes especially for daughters. Some have suggested that highly educated mothers do not favor boys over girls when making educational investments, but that fathers with more education tend to favor boys (Meyerhoefer and Chen, 2011). Mothers may also distribute household work equally across gender of the children. On the one hand, because of the presence of mother and her decision-making power, girls may benefit from a more equal investment in education and distribution of household labor. On the other hand, the potential benefits from household migration can be offset by disruption of family life. By using innovative methodology such as instrumental variables or endogeneity switching model to control for selectivity issues, some researchers find that parents' migration has an overall negative consequence for the wellbeing of left-behind children (Frisancho and Orepesa, 2011; Antman, 2010; Mckenzie

14 7 and Rapoport 2011). Potential negative effects on children's school and work outcomes from household migration may function through left-behind children's compensation for loss of a domestic labor, lack of supervision and role model, as well as cultural diffusion by exposure to migration. Sending a parent to migrate costs money and affects children's time allocation, since a parent's absence can "translate into less education acquisition, requiring the child to undertake housework or work to help meeting short-term labor and cash shortages" (Mckenzie and Rapoport, 2011). By examining the short-term paternal migration on children's time allocation, Antman (2010) found that a father's migration leads to reduced study hours among both boys and girls, and increased the working hours outside the home among 12 to 15 year-old boys. This implies that boys usually assume financial responsibilities for the household in response to a father's migration. Parents' migration can also have negative effects for girls' schooling by re-allocation of time towards home production. Mckenzie and Rapoport (2011) found that although Mexican rural children living in migrant households and children in non-migrant households do not significantly differ in work outside of home and work in family business, girls' education is more likely to suffer from increased household chores in migrant families, which is also widely supported by comparative research from China and Albania (Meyerhoefer and Chen, 2011; Giannelli and Mangiavacchi, 2010; Feng Hu, 2011). In addition, according to socialization and learning theories (Biblarz and Raftery, 1999), parents' migration and marital dissolution may have the same deleterious effects for children because of lack of both supervision and a role model. Research suggests that parental absence, whether due to legal separation or migration, is associated with an

15 8 increased likelihood of having children with behavioral and academic problems, compared to children living in two-parent households (Lehaie et al, 209). However, this may depend on the gender of the child. Booth (1995) discovered that in one-parent households in Swazi, boys had reduced educational achievement because of a lack of role models, but no such effects were found among girls. Exposure to migration also offers the children with alternative opportunities. By receiving material goods and gifts along with the social network from migrating household members, children may be exposed to alternative pathway mobility by "looking northward rather than locally for opportunities and social mobility" (Kandel and Massey 2002). This could lead to lowered educational aspirations to attend university in Mexico and higher aspirations to work in the U.S., especially in the case of a father's migration (Kandel and Kao, 2000). Researchers have also found that children in migrant households had reduced likelihoods of participating in the local market, thus supporting a "culture of migration" theory (Andrew, 2011). Further, a parent's migration may suggest to children that Mexican education is not well-rewarded in the U.S. (Brasberg and Ragan, 2002; Gonzalez, 2003; Freidberg, 2000) and thus reduce children s aspirations for further study. By exploring what else children are doing instead of school, Mckenzie and Rapoport (2011) concluded that boys in migrant households are more likely to migrate themselves compared to those in non-migrant families. Based on the competing mechanisms from "remittances" and "cultural diffusion", to gain insights into the potential benefits and negative effects of household migration, I hypothesize that: H1: Compared to children living in two-parent, non-migrant households, older children living in migrant households are less likely to enroll in school, and

16 9 will be less likely to work in the labor market; younger children are less likely to aspire for postsecondary education. Labor Migration and Marital Dissolution Labor migration and marital dissolution both contribute to family separation. However, whether labor migration and marital dissolution are distinctive experiences for children's school and work outcomes and under what mechanisms are these two experiences are different, are both open questions. In contrast to children living in migrant households with potential remittances from migration, children living in a household with a divorced single parent generally suffer from economic deprivation (Biblarz and Raftery, 1999). This may require children to drop out school and participate in the labor market early in their lives. As a result, from an economic perspective, compared to children living in migrant households, children of divorced parents will have fewer economic resources for human capital investment, and will need to work. Moreover, in terms of the psychological and emotional costs, children living in lone-parent households may also encounter the same lack of supervision and role model absence as do children living with a single-parent because of father s migration. Similarly, the struggles and difficulties faced by single-parent may also have similar effects as parents' migration on children's acknowledgement of parental sacrifices, which is supported by Booth's finding in Swazi (1995) that girls excel in school by understanding the mother's financial difficulties. Therefore, for a comparison between single households and intact households, I hypothesize that:

17 10 H2: Compared to children living in two-parent, non-migrant households, older children living in single households are less likely to enroll in school, and more likely to work outside of home; younger children are less likely to aspire for postsecondary education. The simple assessment of migration effects is complicated by the fact that there are actually three distinctive Mexican migration patterns: married fathers migrating alone, both parents migrating, and unmarried mothers migrating without their children (Dreby 2006, 2010). Since the migration pattern that mother migrates while father stays is not common, when it comes to comparing family separation between migration and marital dissolution, it is necessary to compare a migrant father with a divorced father in lone-mother households. By employing a longitudinal model, Creighton et al (2009) found that upper-secondary students living without a father because of either international migration or marital dissolution are at greatest risk of dropping out, relative to children living with two-parent households. However, although a father's migration can contribute to emotional stress in the relationships between spouses and between child and father, some research indicates that parents' migration and divorce are distinctive experiences. A migrant father, compared to a divorced father, engages more frequently in children's lives through both financial and parenting efforts and positively contributes to children's schooling outcomes (Nobles, 2011). Alternatively, with exposure to migration through "cultural diffusion" in migrant households, children living in divorced families may lack a social network and alternative economic mobility. They may thus also have relatively less orientation towards foreign labor markets and more orientation towards

18 11 home institutions compared to children living in migrant households. To explore these possibilities, I hypothesize that: H3: Compared to children living with a present mother and a migrant father, older children living with a single mother and a divorced father are more likely to work in home institutions but may not have difference in terms of school enrollment; however, younger children are more likely to aspire for college. Who Stays and Who Leaves Extensive research on children's well-being points to the importance of parental availability (Coleman et al, 1966; Epstein, 1983; Gordon, 1979; Henderson, 1987). However, is it an important role conditioned by who stays and who leaves in a family separation? Is a present mother distinctive from a present father for children's education and work outcomes? Will family migration benefit left-behind children if no parent-child separation is experienced? Numerous studies have suggested that a present mother does have the same effect as a present father (Townsend et al, 2002; Biblarz and Raftery, 1999). In Mexico, although the gendered division of domestic labor has changed in recent years, with more wives assuming house-head roles (Chant, 2002) and more husbands devoting time to children's daily needs (Garcia and Oliveira, 2005), evolutionary psychology theory suggests that the mother, who is usually the main caregiver, is still more child-centered and distributes investment more equally between sons and daughters compared to a father. Comparing children's study and work between households with a present mother

19 12 and households with a present father in Mexico, Levision et al (2002) found that although having a mother present and a father present has similar effects, both leading to children's decreased likelihood of specialization in work and increased probability of school attendance, the effect is sizably larger for a present mother. Therefore, in contrast to the absence of father, a present mother leads to more decision-making power that children will receive more education. Furthermore, research suggests that after controlling for socioeconomic factors, mother-only arrangements and traditional two-parent families are of not different in Mexico. Both contribute to positive school enrollments and free girls from home production and labor force participation. By contrast, lone-fathers may have negative effects for girls' schooling, with more work in household labor (Giorguli, 2006; Townsend et al, 1997). In addition, single mother households tend to be smaller, and thus household chores might be less demanding than in other types of households (Giorguli, 2006; Biblarz and Raftery, 1993). In Mexico, single mothers tend to live with an extended family member, especially a grandmother (Nobles, 2006). Extended family members can also contribute to household chores, which is conductive for girls' concentration on study or participation into formal labor employment other than housework. When further exploring the presence and absence of family member under the contexts of labor migration, Dreby (2011) indicated that, compared to the children living in non-migrant families, children of fathers' lone migration have no significantly different educational aspirations, since paternal migration is quite common in Mexico. When a mother stays in Mexico, the children's lives do not change so drastically as to alter their

20 13 educational goals. However, mother's migration and absence may come with an emotional cost. When a caregiver-spouse migrates to the U.S., children who remain behind are frequently those experiencing the worst academic, behavioral and emotional outcomes (Heymann et al, 2009; Lahaie et al 2009). However, Dreby (2011) also provided contradictory evidence that children of the unmarried single migrant mothers have the strongest educational motivation because they acknowledge the sacrifices their mothers have made. Townsend et al (1997) also suggest that having a father who is away from home as a migrant appears to benefit older children, whereas, for girls ages 11 to 15, having a mother who is a migrant lowers educational attainment. However, when a mother migrates with a father, the emotional costs can be blamed for lowering children's educational aspirations. Therefore, based on the gendered presence of parents, I hypothesize that: H4: Children living in lone-father households are less likely to attend school and participate in work, and are less likely to aspire for college compared to children living in lone-mother households. Finally, literature on internal migration of other developing countries such as rural China suggest that sibling migration generates benefits for children's education, especially for older girls, while parental labor migration may not offer significant advantages for the educational prospects of left-behind children (Lu, 2012). Furthermore, the effects of presence of family members are subject to community-level characteristics. Amuedo et al (2010) decomposed remittance-receiving households in Haiti into those with a migrating household member and those without a migrating household member (e.g. distant relatives, friends). They found that receipt of remittances raises school

21 14 attendance for children only in communities characterized by an "abundance of private, parochial and public schools", regardless of whether they have household members abroad. In contrast, for communities lacking easy school access, the positive effects from remittances only apply to children residing in households without experiencing absence of close family members. Therefore, based on the presence of specific household member in migration households, I hypothesize that: H5: Children living with two present parents and migrant siblings are not different in their in work and school outcomes, as well as in their aspiration for college, compared with children living in two-present parent, non-migrant households.

22 15 Chapter 3 DATA AND METHODS Data and Sample For this study, I drew on data from Mexican Family Life Survey (MxFLS). The MxFLS is a longitudinal, nationally representative database, including rich information on the social, economic and demographic aspects, as well as the dynamics of Mexicans who decide to migrate to the United States during the first decade of the 21st century. The first field work was conducted in The original sample drew on 35,000 individual interviews in 150 urban and rural locations throughout the country. The second round re-contacted 90% of the households in 2005, and re-contacted those who migrated to the United States at a rate of over 91%. Due to sample size considerations, this study only uses the second wave conducted in the year of In order to explore how different types of family separation affect left-behind children's school and work outcomes, in this study I limit my sample to children ages who were single, were not household heads, and were still living in their original household at the time of survey. I divided the ages range into two groups for some analyses. Since the 1993 introduction of nine years of compulsory education (Creighton & Park, 2010), labor laws have forbidden children under 14 from working. Therefore, I test only the school and work outcomes for children ages 15 to 18. It is during this age range that children start (and finish) upper secondary education and begin legally to participate in the labor force. In addition, I first explore the educational expectations for the whole population, and then explored the educational expectation especially for

23 16 younger children, ages 11 to 14, since the question on educational expectation was only posed for children who were currently attending school (91.5% of those 11-14). In addition, in Mexico, when children enter "preparatoria" (high school), they have been preselected for postsecondary education because most preparatorias are operated by universities. Therefore, they are assumed to be likely to go to aspire to postsecondary. Thus, exploring educational expectations of children before they enter preparatoria is more relevant to the aim of this study. In the 2005 Mexican Family Life Survey, there were 5,677 children ages in total, who are single, non-household heads and currently living at the households. After excluding the 18% of children living with no parents and children living with unmarried and widowed mother and other complicated types of household arrangements, there remained 4648 children. For younger children aged from 11 to 14, there remained 2429 children, of whom 2219 (98.7%) were currently attending school. For older adolescents ages 15 to 18, there were 2219 children in total. Measurements Explanatory Variables The main explanatory variables of interest are the different types of household configurations for children. These consist of five categories: 1) children living in twoparent, non-migrant households; 2) children living in households with two present parents and migrant siblings; 3) children living in households with a present mother and a migrant father; 4) children living in households with a present mother and a

24 17 divorced/separated father; 5) children living in households with a present father and an absent mother. I construct the family structure based on information on household member's marital status, whether the child and the parent still lived in the household at the time of survey and where parent migrates at the time of the survey. Children living in two-parent, non-migrant households could be determined if two parents both stated consistently they were with each other as in "marital/cohabiting", and they were still living in household at the time of survey. In addition, in this category there is no current international migrating family member in the household. Children living in households with two parents and migrant siblings are defined as those who live with two present, married/cohabiting parents while there are siblings currently migrating internationally. Children living in households with a present mother and a migrant father are determined by the mother's statement that her spouse has migrated internationally and who is currently living in the household with the children. Children living in households with a present mother and a divorced father are defined as those whose mother's marital status is divorced or separated, who still lives at this household while her partner did not at the time of survey. Children living in households with a present father and an absent mother are those whose father is stilling living at this household while whose mother's status is either deceased or not living at this household. The limitations of the data are that the information of whether a divorced father was migrating or not at the time of survey is unavailable. Thus, we cannot disentangle the effects of migration and marital dissolution. Similar to previous research (Nobles, 2011), I categorize children of a divorced father in the category of "divorced single households" even if his father was migrating. In addition, no information is available on the migration

25 18 of unmarried, divorced/widowed mother and of both parents. As a result, I cannot compare the dominant migration patterns in Mexico for 2005 and I am unable further to examine the context of who stays and who leaves. Control Variables Three levels of control variables are included in the analysis. For individual-level characteristics, I include age, gender and ethnicity. Age is expected to be positively related to work participation and negatively to school enrollment. Gender is a dichotomous variable with male as reference. It is supposed that sons were more likely to take financial responsibility due to the expectations of gender roles in Mexico. However, whether there is a gender gap in school enrollment is an empirical question. "Ethnicity" is constructed using language ability for older children aged from 15 to 18, and by the question of "whether part of indigenous group" for children aged from 11 to 14. Membership in an indigenous group is the reference category. And the effect of belonging to an indigenous group is remained to be examined. For household-level control variables, two sets are employed: family structure and socioeconomic status. In terms of family structure, whether or not grandparents are present is an important control variable. For single-parent households, the presence of grandparents may positively relate to girls' outcomes by releasing participation in domestic chores and by providing benefits for education and work outside home. Aside from the presence of grandparents, the number of co-resident siblings is also included. These are five dummy variables, including none, one, two, three, four and more, with the number of one co-resident sibling as the reference category. The effect of family size has

26 19 been frequently tested in family studies since larger family size may mean investment delusion for children's investment. For socioeconomic status characteristics, three indicators are introduced including household income, employment of mother and parental education. Household annual income is constructed by summing up all family members' annual income from labor participation in the last 12 months. However, the income of the child is not included because earnings from child labor are endogenous and cannot simultaneously be used to predict the child's work status. Missing values are assigned the means of all eligible households. The natural log of this variable is included in the model. Parental education is chosen for the present parent's level of education in lone-parent households. The parent who has the highest educational level in two-parent households is used for parental education. This is constructed using three dummy variables: less than primary (including no education, preschool or kinder, elementary), lower secondary (including secondary and open secondary), upper secondary or higher (including high school, open high school, normal basic, college, graduate). Lower secondary education is used as the reference category. The employment of parents is based on the question of "whether worked in the last 12 months". Mother's employment status is used for all household types except for lone-father households, which are used father's employment status instead. This is a dichotomous variable where "1" equals "yes" and "0" equals "no". It is supposed that when the mother works it will have positive influence on children to combine work and school (Giorguli, 2006). Therefore, I hypothesize that mother's employment is positively associated with children's participation into work.

27 20 Local-level characteristics can distinguish rural areas from urban areas. By merging 2005 household data with community data and further merging with the first wave of 2002, households in 2005 can be matched to community characteristics in 2002, with information on the population size of each 150 communities. Rural communities are constructed by population size less than 2,500, and urban communities are constructed by population size more than 2,500. This is a dichotomous variable with rural coded as "1" and urban coded as "0". Dependent Variables For older adolescents aged from 15 to 18, variables containing information on school enrollment and work participation are constructed for analyses. Based on the question "currently attend school", children's current school attendance is coded as "1" if he/she still attends school currently, and coded as "0" if not. This is a dummy variable. In terms of the combined school and work status, Based on children's current school attendance and whether the child worked in the last 12 months, four categories are constructed: children who are currently attending school and did not work in the past 12 months (only school and no work); children who are currently attending school and had worked in the last 12 months (both school and work); children who are not currently attending school but worked in the last 12 months (only work and no school); children who are neither currently attending school nor worked in the last 12 months (neither school nor work). This is a four-scale variable with children who are currently attending school and did not work in the last year as the baseline category in multinomial logistic regressions.

28 21 Among children aged from 11 to 18 and from 11 to 14, I modeled the determinants of their postsecondary educational aspirations. Based on the question "until what grade are you expecting to study" for children who are currently attending school, a dichotomous variable was constructed to indicate whether or not the child aspired to study until "college or higher" or not. Ordered logistic regression is employed, which shared similar results with logistic regression. Therefore, I use the educational expectation as a dichotomous variable. Missing values are assigned at mean (upper secondary but lower than college) since most of the missing value (over 86%) are coming from children living in non-migrant, two-parent households. I also estimated the regression equation by deleting missing values, and found similar results. Therefore, I kept means for missing values. Methods For school enrollment of children aged 15-18, logistic regression is employed. The model is formulated as: ln[ P /(1 P)] = β + β Ftype + β X + ε i 0 1 j 2 where ln[pi/(1-p)] is the log ratio of probability of attending school currently. Ftype is the family types (including migrant households, single households, and nonmigrant, two-parent households as the reference). X is a vector of multi-level control variables from individual, family structure, household socioeconomic status and local characteristics. ε is the error term. Since the analysis is cross-sectional and not longitudinal, and self-selectivity may violate the assumptions of OLS model and may be

29 22 an issue for children living in different households. There are limitations of this method, as discussed later. For school and work outcomes of children aged 15-18, multinomial logistic regression is employed with "only school and no work" as the baseline category. No children living in households with a present mother and a migrant father, or in households with lone-fathers, were in the category of "both work and school" (see Table1a). Multinomial logistic regression would be difficult in this case. As a result, children living with migrant siblings and those living with a migrant father are aggregated as "migrant households", while children living with single mother and divorced father and those living with single father are aggregated as "single households". The model is formulated as: ln( P / P ) = β + β Ftype + β X + ε i base 0 1 j 2 where ln(pi/pbase) is the log ratio of probability of one school and work status (from only school, both, only work) relative to baseline category of only school and no work. Ftype j are the family types (including migrant households, single households, nonmigrant, and two-parent households as the reference). X is a vector of multi-level control variables from individual, family structure, household socioeconomic status and local characteristics. ε is the error term. Relative risk ratios are calculated. For educational expectation of younger children aged and 11-14, logistic regression is employed and odds ratio are computed. Both family types as aggregated form and as disaggregated form by who leaves and who stays are included. The model is formulated as:

30 23 ln[ /(1 P)] = γ + γ Ftype + γ X + µ P 0 1 i 2 where ln[p/(1-p)] is the log odds of aspiration for college, and Ftype i are the family types, including all the five types discussed above. X is a vector of control variables from multi-level characteristics. µ is the error term.

31 Chapter 4 RESULTS Descriptive Statistics Table 4-1 reports the distribution of the individual, household, and community characteristics by different family types. According to the table, the majority of children live in non-migrant, two-parent households, with 87.4% of the younger cohort and 86.5% of the older cohort. For the rest of children, slightly more children are living with single mothers (6.2% for the younger and 7.1% for the older), followed by children living in two-parent and migrant-sibling households. The proportions of children living with a present mother and a migrant father and those living in lone-father households is small (less than 2%), which may raise selection issues. Therefore, interpretation of the results requires caution. In terms of individual characteristics, children do not vary in average age, gender by types of family. However, fewer children living in single households were of indigenous origin (7.6% for single mother and 8.7% for single father) compared to other types of family. With respect to family structure, children living in single-parent households, have more presence of co-resident grandparents, which are 20% of single-mother households and 36.3% of single-father households respectively. In addition, children living in single households have a smaller family size. Nearly 50% of them (including both households

32 25 of single mother and single father) have none or one co-resident siblings, while half of children living in migrant households have more than three co-resident siblings. When it comes to socioeconomic status of the family, children living in all other family types have less household annual income compared to children living in two-parent and nonmigrant households. However, the differences are not significant. It may because that the measurement of household annual income is used by income from work or employment, instead of assets, which can be accumulated from international remittances. Therefore, the annual income may not show the potential benefits from remittances in migrant households and possible economic deprivation of single households. In terms of the employment of the mother, children living in two-parent, migrant-siblings households have the lowest proportion of working mothers (10.4%). Children living in single-mother households have the highest (83.5%), and are similar to the proportion of children living with working fathers (88.6%). This is consistent with previous studies showing that single mothers in Mexico are more likely to enroll in labor force. In addition, children living in intact, non-migrant households also have smaller proportions of working mothers as compared to children living in only-mother, and migrant-father households, with 29.8% and 36.5% working mothers respectively. In terms of parental education, children living in migrant households, including both households with migrant-father and migrant-siblings, have less educated parents. Only 4.9% and 3.1% of these children had a parent who completed upper secondary or higher education. In contrast, nearly (some more than) 20% of children living in other types of households had parents with at least upper-secondary education.

33 Table 4-1: Descriptive Statistics of Individual, Household and Community-Level Characteristics by Family Types Non-migrant Migrant HH Single HH Two-parent HH Migrant siblings Migrant father Single mother Single father Age groups: ,939 (87.4%) 79 (3.6%) 30 (1.3%) 138 (6.2%) 33 (1.5%) ,920 (86.5%) 73 (3.3%) 30 (1.4%) 158 (7.1%) 38 (1.7%) N=2, 219 for either age group Individual characteristics Age (mean) Gender Male 50.6% 43.6% 44.4% 50.5% 60% Female 49.4% 56.4% 55.6% 49.5% 40% Ethnicity Indigenous 11.9% 18.4% 15.9% 7.6% 8.7% Non-indigenous 88.1% 81.6% 84.1% 92.4% 91.3% Family Structure Presence of grandparents Yes 8.3% 5.5% 14.3% 20% 36.3% No 91.7% 94.5% 85.7% 80% 63.7% Num. of co-resident siblings 0 4.9% 10% 3.3% 13.9% 28.2% % 17.1% 20% 28.5% 20.5% % 12.9% 23.4% 29.7% 23.1% % 28.6% 20% 14.6% 12.8% % 31.4% 33.3% 13.3% 15.4% Socioeconomic status Log annual per capita income Mean (st.) (1.08) (1.04) (1.38) (0.93) (0.90) Mother employed a Yes 29.8% 10.4% 36.5% 83.5% 88.6% No 70.2% 89.6% 63.5% 16.5% 11.4% Parental Education Less than primary 28.0% 17.2% 31.8% 29.2% 11.4% Lower secondary 47.2% 77.9% 65.1% 53.0% 68.6% Upper secondary or higher 24.8% 4.9% 3.1% 17.8% 20% Local characteristics Rural (<2,500) 51.1% 66.9% 71.4% 32.7% 50% Urban 48.9% 33.1% 28.6% 67.3% 50% Note: a for all family types except lone-father households, mother's employment is used; for single-father HH, father's employment is used 1

34 27 Further, with respect to local characteristics, nearly 70% of migrant households, including both households of migrant-father and migrant-sibling, came from rural areas. By contrast, only 32.7% of children living in single-mother households were from rural areas. Half of the children living in households of single-father and those of two parents and non-migrants reside in rural areas. Table 4-2 (aggregated and disaggregated) reports the distribution of school enrollment, school and work statuses for children ages 15 to 18 in different family types, and educational expectation for children aged and respectively. In terms of school attendance, 47.6% of children living in migrant households currently attend school, compared to 60.3% children in intact, non-migrant families and 57.1% of children. The descriptive statistics showed that children living in migrant households are less likely to currently enroll in school compared to children living in non-migrant households or single households. However, children who have migrant siblings (49.3%) and who have a migrant father (43.3%) do not differ much in school enrollment, so do the difference between children living in single-mother households (58.9%) and those living in single-father households (59.4%). In terms of school and work outcomes, 42.7% of children living in migrant households currently attend school without work participation in the last year, compared to 57.3% children in intact, non-migrant families and 52.0% of children in single households. Further, children living with single-mothers and those living with single-father do not differ much in terms of current school attendance without work participation in last year (52.5% vs. 50%), so did children living in migrant-siblings households and migrant-father households (42.5% vs. 43.3%). In terms of combining school and work, slightly more children living in migrant (4.9%) and

35 28 single households (5.1%) fall into this category than children from intact, non-migrant families (3.1%). With regard to only work without school enrollment, there are 18.4% of children living in migrant households, compared to 14.8% for the other two types of households. In addition, children living in migrant households have largest proportion of "neither school nor work" (34.0%), followed by single households (28.1%) compared to children in non-migrant, intact households (24.8%). The descriptive statistics showed that children living in migrant households are less likely to currently enroll in school and more likely to have worked in the past year. They were also more likely to be doing nothing compared to children living in non-migrant households or single households. With respect to the educational expectations of the younger cohort who currently attended school, large variations emerge. Only 16.7% of children living with a present mother and a migrant father, and 27.9% of children living with both parent and migrant siblings stated an expectation to study for a grade until college or more. In contrast, nearly 50% of children living in other types of households aspired for college. Even more children living with a divorced/separated mother aspire for college (55.8%) compared to children living in intact, nonmigrant households (49.2%).