Crossing the Mexico-U.S. Border: Illegality and Children’s Migration to the United States

VIOLENCE, FAMILIES, AND ECONOMIC OPPORTUNITY

We review studies that consider the factors that push children to migrate. Much of this work relates to the recent growth of unaccompanied minors at the Mexico-U.S. border and the serious neglect of children in past migration studies. Wherever possible, we also review studies that mention children or adolescents in the process of Mexico-U.S. migration even if they are not central actors in these studies.

Despite a lack of clarity about the “precise combination of motives” that underlie recent child arrivals in the United States (Kandel et al. 2014, 12), some agreement has been reached about the key factors that underlie this migration (Chishti and Hipsman 2014). Three stand out: violence, family separation and reunification, and limited economic opportunity. Among these, drug-related and organized crime-style violence has been well documented. In Honduras, El Salvador, and Guatemala, for example, murder rates are among the world’s highest. Mexico’s homicide rates are somewhat lower, but between 2007 and 2012, they grew more quickly than in any country (from 8.1 per hundred thousand to 21.5). This rise correlates with the timing of Mexican President Calderón’s decision to go after all drug trafficking organizations in 2007, which subsequently intensified violence and geographically dispersed it across the entire nation (Guerrero-Gutiérrez 2011). In addition, in the last few years, although Mexico’s homicide rate has dropped somewhat, violence has become increasingly tied to organized crime groups whose influence infiltrates down into local community institutions, including the police (Heinle, Rodríguez Ferriera, and Shirk 2015). Reports also suggest that smugglers are recruiting young children to migrate to carry various forms of illegal contraband (Kennedy 2014).

One well-known case of organized crime and violence is the disappearance of forty-three students who were traveling by bus in the state of Guerrero in 2014. Initially detained by local police affiliated with an organized crime group, most of the students were taken away and presumably killed (their remains have yet to be found). As this occurred, the town’s mayor and his wife went into hiding; they, and some other believed perpetrators, were eventually found and charged for kidnapping and organized crime activity. A year later, no one had been charged with murder; an Inter-American Commission report suggested the attack was a coordinated effort between organized crime, local police, and federal security forces. Jo Tuckman also speculates that the reason for the students’ disappearance was corrupt police who thought the bus carrying the students contained heroin or drug money (2015). Although just one case, it illustrates how difficult it is to understand the extent of drug-related violence in Mexico and its impacts. Accompanying drug-related violence in Mexico are the many who have gone missing since 2006, albeit with far less publicity. Kimberly Heinle, Octavio Rodríguez Ferriera, and David Shirk estimate the number to be more than twenty thousand people between 2006 and 2012—including approximately 1,200 children ten years old or younger (2015, n. 60).

How violence and organized crime, real or perceived, affect the experiences of children and their behavior in Mexico is not well understood. Generally, studies suggest that children, no matter where they live, may develop symptoms related to post-traumatic stress disorder if exposed to violence (Martinez and Richters 1993; Berman et al. 1996; Osofsky et al. 2004). Exposure may also affect how parents relate to, and monitor, children including whether they permit them to go to school (Bryk et al. 2010; Harding 2010). In Mexico, Pedro Orraca Romano shows that in 2011 approximately 7 percent of students reported they stopped attending school because they feared becoming a victim of crime (2015, table 1). Using school-level data, he finds that homicide exposure at a young age reduced Mexican children’s academic achievement, more so in secondary than elementary schools. Valentina Duque reports a negative effect of violence exposure on children’s educational outcomes in Colombia (2013).

In 2011 and 2012, the United Nations High Commissioner for Refugees interviewed approximately four hundred unaccompanied children from Mexico, El Salvador, Guatemala, and Honduras to understand the reasons they left their homes for the United States (UNHCR 2014). Although findings show complexity in children’s reasons, approximately one-third of those from Mexico describe violence in their communities, 17 percent mention violence in their homes, and another 12 percent report both (UNHCR 2014). Moreover, unlike children from Central America, a sizeable share (almost 40 percent) of Mexican children in this sample reported being recruited into the human smuggling industry. Together, these findings suggest that children migrate to escape violence and are “in need of international protection” (UNHCR 2014, 6).

Not surprisingly, conditions of violence are likely to intensify “the desire for family reunification” between children in Mexican origins and their parents in the United States (Kandel et al. 2014, 15). In fact, many unaccompanied children who enter from Mexico have ties to parents in the United States. Descriptive results from the UNHCR report reveal that 22 percent of unaccompanied children entering the United States from Mexico had at least one parent living in the United States (2014). Katharine Donato and Blake Sisk also find a very strong relationship between the migration of children and parents (2015). Although the likelihood of a Mexican child making a first U.S. trip was quite low, it was practically nonexistent for children whose parents have no U.S. experience. For children with migrant parents in the United States, children’s cumulative chances of making a first trip by age seventeen were very high: 69 percent for trips made between 1970 and 1986, 75 percent between 1987 and 1996, and 55 percent between 1997 and 2011. Thus, although annual rates of Mexico-U.S. migration have declined, Donato and Sisk suggest continued and strong linkages between migrant parents and their children (Villarreal 2014; Donato and Sisk 2015).

In addition, parents’ attributes, especially legal status, may also influence children’s migration. Prior studies suggest that having a father as an undocumented migrant was positively associated with children migrating without documents. These studies mention children, but only in the context of understanding the larger process of Mexico-U.S. migration. For example, studies describe Mexico-U.S. undocumented migration from small rural villages as being passed down from one generation to the next, especially from fathers to sons (Reichert and Massey 1979; Massey and Liang 1989; Massey et al. 1987). Studies also note that children followed their fathers, who initially migrated without documents for agricultural work but subsequently obtained legal permanent residency, to reunify with their families (Reichert and Massey 1979, 1980; Massey, Goldring, and Durand 1994; Fonseca and Moreno 1988; Goldring 1990; Durand and Massey 1992; Donato 1993, 1994; Cerrutti and Massey 2001; Donato, Wagner, and Patterson 2008; Creighton and Riosmena 2013). Theoretically, Oded Stark and Richard Taylor recognize that children are involved in the migration process when they argue that households—and not individuals—make migration decisions to diversify household risks and costs (1991).

In contrast to these studies, Christine Tucker and her colleagues (2013) interviewed forty-seven Mexican youth, age fourteen through twenty-four, in two origin communities about their reasons for migration. Like many adults migrating northward, youth with U.S. experience reported that economic hardship and difficulty finding a job were their main reasons to migrate. Yet, the decision of most interviewed youth to migrate depended on their parents because they accompanied parents on U.S. trips. The remaining, smaller group of young adults whose parents had never migrated wanted to remain in Mexico. They had no plans to migrate because they envisioned economic opportunities in their origin communities and because they feared the difficulties related to crossing the border without authorization.²

Finally, in addition to personally threatening violent conditions and the motivation to reunify with parents, children in Mexico face limited economic and social mobility. Although economic opportunities are changing somewhat because of moderate annual rates of recent economic growth (World Bank 2015), Mexico ranks near the top of all countries for high income and wealth inequality. Income is highly concentrated among the top 1 percent of the population, and the wealthiest 10 percent control 64 percent of the nation’s total wealth (Esquivel Hernandez 2014). These indicators suggest children’s access to economic opportunities, and the skills and training needed to access them, will be—at best—uneven. Along these lines, student performance in Mexico’s schools remains well behind that of other Organization for Economic Cooperation and Development (OECD) countries, despite the government’s sizeable investments in Mexico’s educational system since the 1990s (Acevedo and Salinas 2000; OECD 2013). In 2012, although Mexico reported one of the highest rates of preschool enrollment, its effectiveness was challenged by high student-teacher ratios. In addition, although Mexico expanded compulsory attendance to the secondary level in 1993, upper secondary school graduation rates remain quite low (approximately 36 percent).

SHIFTING U.S. CONTEXT FOR CHILD MIGRATION

The United States has witnessed two key shifts likely to influence Mexican children’s propensities to migrate. The first is related to U.S. immigration policy. The second is related to the Great Recession; although the recession officially began in late 2007 and ended in June 2009, its shadow on the U.S. labor market continues to loom large (Shierholz 2014).

The U.S. policy provision that directly influenced children’s out-migration was the amnesty provision of the 1986 Immigration Reform and Control Act (IRCA). Approximately two million Mexicans received permanent residency after a five-year waiting period, when they became able to sponsor spouses and children for permanent residency. IRCA also set the groundwork for subsequent policies by dramatically increasing funds for enforcement and removal operations throughout the United States. Together, these policies have had the unintended effect of increasing settlement and family reunification in the United States (Massey, Durand, and Malone 2002). Thus, like many women entering immediately after IRCA, many children entered both with and without documents after 1986 (Donato 1993; Donato and Armenta 2011). Some were quickly able to adjust their status after their fathers received permanent residency, but many others continue to reside in the United States without documents, or with temporary legal status under the Deferred Action for Childhood Arrivals (DACA) program started in 2012. As a result, Roberto Gonzales argues that a broken immigration system has led to many child migrant residents unable to move forward and fully integrate into U.S. society (2015).

Together with record high deportations in recent years, unauthorized immigrants face higher risks of being deported and with their families and children have suffered substantial consequences (on deportations, Gonzalez-Barrera and Krogstad 2014; on consequences, Abrego 2014; Dreby 2012, 2015; Massey 2013). Thus, the last three decades have proved a labyrinth of restrictive policies that have made the lives of immigrants more difficult in the United States. One recent study suggests that border militarization has disrupted the lives of transnational families and lowered the odds that the children of Mexican farm workers reside in Mexico (Hamilton and Hale 2016).

Understanding the effects of these conditions is even more challenging given that certain migrant children are treated differently than others. For example, not all child migrants are unaccompanied minors encountered by officials at the Mexico-U.S. border. Provisions of the William Wilberforce Trafficking Victims Protection Reauthorization Act of 2008 (Public Law 110–457) mean that unaccompanied minors from Mexico and Canada are treated differently than those from Central America. Mexican minors are quickly processed and then deported if, after screening, they are not deemed trafficking victims or have asylum claims based on a credible fear of persecution or torture. In contrast, minors from noncontiguous countries are placed in formal removal proceedings, and released to parents or relatives who care for them until they appear in front of a U.S. immigration judge.³

Such a confusingly intricate state of affairs can easily create a complex set of perceptions about whether and how children crossing the border receive humanitarian protection once in the United States (Chishti and Hipsman 2014). For example, some may positively, but wrongly, perceive DACA as a form of permanent legal status even though it is temporary, subject to renewal, and available only to children who entered as unauthorized with their parents and meet other criteria. Others, after the Department of Homeland Security’s recent announcement to deport minors who recently received removal orders, may perceive that children have slim prospects of receiving protection and legal status (Markon and Nakamura 2015).

The second salient shift likely to affect children’s propensities to make a first U.S. trip is related to the U.S. economy. Before the Great Recession began in 2007, the United States witnessed four decades of sustained immigration and foreign-born workers made up approximately 16 percent of the U.S. labor force (Newburger and Gryn 2009). Late in 2007, however, the U.S. economy sank into a deep recession that was especially devastating for low-skilled workers, of whom approximately half were immigrants (Bean et al. 2012; Orrenius and Zavodny 2013). By 2010, those with less than a high school degree faced a 15 percent unemployment rate, versus the overall rate of 10 percent (Hout, Levanon, and Cumberworth 2011; BLS 2013). Sisk and Donato find that the recession’s impact on Mexican immigrant men was mixed (2017). Although they weathered the Great Recession well in some respects, for example, they were more likely than native whites and blacks to remain employed during the recession, Mexican men’s relative success was not without its costs as they were also more likely to become underemployed by transitioning into involuntary part-time work. Compounding this situation, especially for unauthorized Mexican immigrants, is a set of deteriorating working conditions, including lower wages and precarious working conditions, under way since the 1980s (Donato et al. 1992; Donato and Massey 1993; Phillips and Massey 1999; Donato and Sisk 2013; Hall and Greenman 2015; Durand, Massey, and Pren 2016; Massey, Durand, and Pren 2016).

In addition to worsened labor market outcomes, the Great Recession is also associated with fewer Mexicans migrating to the United States. Although some describe this downward shift as a standstill (Passel and Cohn 2011) and others as a decline (Warren 2016), the shift is unprecedented in recent decades and led to more recent immigrants entering from China and India than from Mexico (Chishti and Hipsman 2014). It is also associated with a decline in U.S. demand for Mexicans to work in industries, such as construction, which had employed them in large numbers in the past (Villarreal 2014). These conditions, then, are likely to lower the chances that Mexican children will make a first U.S. trip after 2006.

EXPECTATIONS

Based on this review of the literature, we argue that children’s unauthorized migration results from a set of decisions associated with conditions in Mexico and the United States. Thus we expect the following three hypotheses:

H1: Exposure to violence will affect children’s chances of making an initial trip; greater violence will stimulate children’s out-migration, with and without documents.

H2: Parents’ migration experience and legal status will influence children’s likelihood of making a trip because children’s migration is often linked to parents.

H3: Period of entry will positively influence the odds of children’s initial migration. We expect higher odds of authorized and unauthorized migration after 1986, when an amnesty program regularized the status of approximately two million Mexicans. However, after 2006, the Great Recession will be associated with lower likelihoods of making a first U.S. trip.

DATA AND METHODS

This study uses data from the Mexican Migration Project (MMP154), a collaborative ongoing research project launched in 1982 and based at Princeton University and the University of Guadalajara.⁴ Because the MMP is an ongoing data collection effort, it now represents 154 origin communities derived from interviews with more than 25,000 households and 160,000 people.⁵ Derived from randomly selected households in these communities, this is an established data source that contains, among other things, information about labor and marital histories, family composition and other demographics, and household assets. For each household member, the MMP154 includes data on first and most recent migrant trips to the United States, and attributes of these trips, including legal status, when they occurred, and duration.

Like most studies of migration, the MMP was not primarily designed to understand children’s migration. However, its substantial size, merged with data that describe conditions in origin communities, permit us to go beyond results from journalists and others, who often use ethnographic or U.S. Customs and Border Patrol data to examine unaccompanied children crossing the border. Therefore, like Donato and Sisk (2015), we analyze MMP data using sophisticated statistical techniques to examine children’s migration.

Although we construct two samples that correspond to different parts of the following analysis, both samples are restricted to children residing in two-parent households and to those with at least one biological parent in households from 149 of the 154 MMP communities. The first sample permits us to use event history models to examine children’s migration from Mexico. Given each child’s date of birth and year of the survey, we construct a year-by-year child life history up to the date of his or her first U.S. trip.⁶ The outcome measure is whether the child migrated within the person-year in question. If they did not migrate in a given year, the migration variable is coded 0; if they did, it is coded 1, and all later years of that child’s life are excluded from the file. In every year when migration occurred, we also record legal status (authorized or unauthorized). Legal—that is, authorized or documented—child migrants had valid U.S. documents to enter and reside in the United States; illegal—that is, unauthorized or undocumented—child migrants did not. Thus, the first sample is a large set of person years that refer to respondents who are age seventeen or younger at the time of migration, from households surveys conducted from 1984 to 2013.

Using this sample, we estimate person-year event history models, in which we regress the 0–1 migration variable on indicators representing legal status, gender, age, metropolitan type, parent’s migration, homicide rates, migration prevalence in origin communities, and period of first U.S. migrant trip. We measure legal status and gender as two dummy variables, where 1 = without documents, 0 = with legal documents, and 1 = female, 0 = otherwise. We include age as a set of dummy variables (two to eleven years = 1, 0 = otherwise; twelve to seventeen years = 1, 0 = otherwise) and use less than two years as the reference category. Metropolitan type captures the urbanicity of Mexican origins in a set of four dummy variables, with large urban as the reference category. Parent’s migration is measured in two dummy variables. The reference category is when a child’s parents had no migration experience, with two other dummy variables entered in the model (parents migrated in any year before the year the child migrated, and parents migrated in the same year the child migrated).

At the national level, we include national homicide rates per hundred thousand people per year (from 1972 to 2010) as a continuous variable.⁷ Following Douglas Massey and Steven Alvarado, we use homicide rates because they reliably measured across years (2010). We then merge these rates with MMP event history data to assess the effect of violence on the odds that children make a first U.S. trip. We also examine the effect of migration networks, included as the percentage of adults (fifteen years and older) in the origin community with U.S. migration experience of all adults in that community in a given year. Finally, to measure period differences, we use the year of the child’s first U.S. trip and construct four dummy variables: 1987 to 1996, which refers to the amnesty period; 1997 to 2006, when border enforcement activity significantly increased; 2007 to 2010, years that correspond to the Great Recession; and 1972 to 1987, the reference category.⁸

To better understand the specific ways in which unauthorized children enter relative to the migration of their parents and siblings, we create a second sample that contains everyone who reported being a child (biological, adopted, or stepchild) of the household head and made a first unauthorized U.S. trip before age eighteen. From a universe of 102,612 children of household heads in the data set, 4,286 (or 4.2 percent of all children) reported migrating on a first trip before age eighteen. From this total, we removed 719 children in female-headed households because children in these households may have different migration experiences.⁹ We also removed 167 children for whom we did not have complete migration information for either their mothers or fathers, twenty-two because legal status on first trip was missing, and twelve whose biological mother was not the spouse of the household. Thus, from the 2,497 child migrants for whom migration and legal status information for themselves and their parents is complete, we use a sample of 1,928 children who made a first U.S. trip without documents.

The analysis using this second sample examines characteristics associated with variation in undocumented children’s migration. We operationalize two dependent variables: whether children migrate without authorization but with at least one unauthorized parent in the same year, and whether children make an unauthorized trip alone (without parents or siblings migrating in the same year, including parents and siblings who did not report a U.S. trip). We coded each as one if the unauthorized child migrated on his or her first trip in that family arrangement, and zero otherwise. In these models, we control for gender, age, parent’s education, and parent’s age; we expect that boys will be more likely than girls to make an unauthorized trip with an unauthorized parent or alone. We also expect older children to be more likely than younger ones to make unauthorized trips alone, and younger children to be more likely than their older counterparts to make authorized trips. The chance that children make a first unauthorized trip will be negatively correlated with parent’s education and age.

We include mother’s and father’s prior unauthorized migration experience, and include these as controls in the form of two dummy variables. We expect that mother’s and father’s prior undocumented status is related to children’s unauthorized entry. Father’s prior unauthorized status will increase the likelihood that children make a first trip, but mother’s prior undocumented status will reduce the likelihood, especially for girls. Violence and migrant social networks will be especially important predictors of children migrating alone. Undocumented children will be most likely to enter with parents during the period of amnesty, between 1987 and 1996. Furthermore, relative to those making a first trip before 1987, children entering between 1987 and 1996 and 1997 and 2005 will be more likely to migrate alone than those entering afterward, between 2006 and 2010.¹⁰

Thus, our analysis strategy is as follows. First, we examine variation in children’s migration, separately for those making an initial trip with and without documents. We assess whether and how violence, as well as other characteristics, influence children to make a first U.S. trip. To illustrate differences by legal status and period, we calculate and present predicted probabilities that children make a first U.S. trip. Second, we estimate models that predict whether unauthorized children who migrate do so in the same year as an unauthorized parent, or alone (without parents or siblings). We estimate these models for all children, and separately for boys and girls, to investigate gender differences in children’s unauthorized migration arrangements.

CHILDREN MAKING A FIRST U.S. TRIP

Table 1 contains three regression models: one for all children, and two separate models for unauthorized and legally authorized children’s migration. The models that predict whether a child makes a first U.S. trip up through age seventeen. The first set of columns refers to all first trips, the next two sets to unauthorized and authorized trips. The separate legal status models are justified by Chow test results that reveal coefficients for the unauthorized differ significantly from coefficients for the authorized.¹¹

We begin by summarizing the coefficients in the pooled model. As we expected, net of other attributes, being unauthorized significantly reduces the likelihood that children migrate on a first trip. In addition, relative to boys, girls are more likely to migrate. Relative to those less than two years of age, being a young adolescent up to age seventeen significantly increases the chance of making a first trip. However, children from urban areas are no different from children originating in smaller places, whether towns or smaller ranchos.

Consistent with Donato and Sisk, having parents with U.S. experience, especially those migrating in the same year as their child, increases the risk of children’s out-migration (2015). Violence, as measured by national homicide rates, operates as a push factor and is positively associated with children migrating on a first trip. However, migration prevalence in origin communities significantly reduces that likelihood. In addition, when a child makes a first trip also matters. Compared with before 1987, the three coefficients for 1987 to 1996, 1997 to 2006, and 2007 to 2010 are positive and significant. Thus, children’s migration grows with time, and although it appears to peak between 1997 and 2006, it still remains substantially higher during the period of the Great Recession, between 2007 and 2010, relative to before 1987.

Comparison of the next two sets of models reveals a decidedly different process of first-trip migration for children making unauthorized and authorized trips. Beginning with the gender effect, we see that although being female is associated with a greater likelihood of making a first legal trip, it is not associated with making a first illegal trip. Age, too, operates differently across the two models. The risk of making a first undocumented trip is considerably higher for adolescents, but no different for those age two to eleven years relative to very young children. In contrast, no significant age coefficients predict the likelihood of making a documented U.S. trip. Moreover, although metropolitan area does not influence children’s unauthorized first trips, originating from a small town reduces a child’s chance of making an authorized first trip relative to those from large urban areas.

Whether we consider unauthorized or authorized first trips, children are much more likely to make a first trip in the year their parents migrate than children whose parents have never migrated. In both models, these coefficients are significant and large. In addition, although somewhat smaller in size, coefficients for parents migrating in the past suggest that children are also more likely to make an undocumented or documented first U.S. trip if their parents had U.S. experience.

Once again, rates of violence are also significant and positive, suggesting that exposure to violence can encourage children to migrate with or without documents. Interestingly, although the effect of social networks, as measured by the prevalence of migration in origin communities, affects the likelihood that children make a first trip, its effect has different signs in the two models. Having more adult migrants in a Mexican origin decreases the likelihood that children make an unauthorized first trip, but increases the chance that children make a first legal trip. T-tests of these two coefficients reveal a significant difference: that is, social networks operate differently when predicting the likelihood that a child makes a first unauthorized trip rather than an authorized one.

Coefficients for the period variables reveal that the risks of children’s first-time unauthorized and authorized migration has shifted over time. Relative to before 1987, the likelihood that children migrate on a first trip without documents is higher in two of the three subsequent periods: from 1987 to 1996 and from 1997 to 2006, but no different from 2007 to 2010. With respect to authorized migration, however, all three periods are significantly different from before 1987. Moreover, results from a t-test of the 2007 to 2010 coefficients across the two models (not shown here) reveal that this effect is significantly different across the two models. Together, these results suggest that children were less likely to initiate a U.S. trip without documents during the period corresponding to the Great Recession, but in the same period were more likely to do so with documents than before 1987.

Figure 1 presents predicted probabilities of children making a first trip from Mexico by legal status and period of entry, calculated from coefficients in the first model in table 1. These probabilities depict four important findings. First, before 1987, the chances that a child would make a first trip are negligible and we see no legal status differences. Second, from 1987 to 1996, immediately following IRCA’s passage, though they remain low, the chances that children make a first unauthorized or authorized trip grow somewhat but they are higher for children making an authorized versus unauthorized trip. Third, between 1997 and 2006, the overall chances of children making a first trip dramatically rise, up to 18 percent for all trips, 26 percent for authorized trips, and 15 percent of unauthorized trips. Fourth, between 2007 and 2010, the chances that children will make a first U.S. trip drop. They are highest among those making an authorized first trip, at just less than 10 percent; the chance of making a first illegal trip drops down to 5 percent.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Predicted Probabilities of Children Making a First U.S. Trip

UNAUTHORIZED CHILD MIGRATION ARRANGEMENTS

Table 2 presents descriptive statistics about children who made a first unauthorized U.S. trip, boys and girls, before their eighteenth birthday. From our analytic sample of 1,928 unauthorized child migrants, 75 percent were boys and 25 percent were girls. Approximately 23 percent of these entered in a year when at least one parent also migrated or when that parent was already in the United States; 21 percent were unauthorized children migrating in the same year that at least one unauthorized parent migrated, and 60 percent were at least twelve years old migrating alone, without parents or siblings in the United States. Among children making an unauthorized first trip, their average age was close to fourteen. Their parents combined education averaged less than seven years of school. Many more children had fathers (55 percent) with unauthorized migration experience than mothers (40 percent). Homicide rates were 17.4 deaths per hundred thousand, and approximately 24 percent of origin communities had adults with U.S. experience. As the table makes clear, fewer than half of the sample migrated in 1986 or earlier, another 40.1 percent migrated between 1987 and 1996, 13.3 percent from 1997 to 2005, and 1.1 percent from 2006 to 2010.

Key among the gender differences are the migration arrangements of undocumented children. They suggest that more girls are accompanied by, or migrate in the same year as, parents. Girls are significantly more likely than boys to enter in the same year as a parent, regardless of status. Among the authorized, 39.5 percent of girls entered in the same year as a parent, compared to 16.9 percent of boys. Similarly, 36.7 percent of unauthorized girls entered in the same year as a parent, compared to 15.4 percent of unauthorized boys. However, boys were much more likely than girls (66.4 to 41.8 percent) to migrate without parents. In addition, girls are younger than boys, and girls’ parents had significantly more education and were younger. Girls were also more likely than boys to have mothers with unauthorized migration experience (7.7 to 4.7 percent), and girls were also more likely than boys to have fathers with such experience (46 or 48 to 37.8 percent). The only other significant gender difference is in the share of children who migrated between 1987 and 1996, when boys were less likely to enter.

Multivariate logistic regression models that predict the likelihood of unauthorized children migrating in the same year as at least one unauthorized parent clarify the specific arrangements in which unauthorized children enter, and the extent to which violence and other factors matter (see table 3). The first column of table 3 refers to all children, the second two columns to boys and girls.

The first column reveals that girls are significantly more likely than boys to migrate on a first U.S. trip without documents in the same year as an undocumented parent. Unauthorized children with older parents are less likely to make a first trip with an unauthorized parent. In addition, the effect for mothers with prior unauthorized experience is negative but for fathers is strongly positive. Homicide rates are negatively associated with the likelihood that children make a first unauthorized trip with at least one parent. However, the migration prevalence ratio for a child’s place of origin has the opposite effect: the likelihood of first unauthorized trip with at least one parent increases with higher prevalence of U.S. migration. Moreover, as expected, unauthorized child migration is more likely in the amnesty period, immediately after IRCA, than before 1987. Furthermore, an examination of the models for boys and girls (columns 2 and 3 of table 3) reveals covariate effects that are similar across the two groups.

Table 4 presents models that predict unauthorized children, between twelve and seventeen years old, migrating alone. In the first model for all children, we see that girls are significantly less likely than boys, but older children are more likely than younger ones, to migrate alone. Once again, we observe negative effects for parent’s education and for parent’s age. We also see that having a mother who made a prior unauthorized trip reduces the likelihood that an unauthorized child migrates alone. In contrast to the findings in table 3, however, we also see that having a father who made a prior unauthorized trip also reduces the likelihood that an unauthorized child migrates alone. Moreover, the coefficient for mothers is at least twice as large as that for fathers. Although homicide has no significant effect, but social networks have a significant and negative effect, suggesting that larger networks in origin communities reduce unauthorized children’s chances of migrating alone. Of the remaining coefficients, only that for 1987 through 1996 is significant. Children were less likely than before 1987 to make a first unauthorized trip alone immediately after IRCA became law.

Examining the models separately for boys and girls reveals that, although children’s demographic attributes are mostly similar in their effects on the likelihood that unauthorized children migrate alone, having parents with prior unauthorized experience operates differently for boys and girls. Among boys, having either a mother or father with prior undocumented experience lowers the risk they migrate alone. Among girls, only mother’s prior undocumented status negatively affects this risk. Moreover, although violence has no effect, migrant networks lower the likelihood of migrating alone only for girls. Finally, period effects in the boys model suggest that, relative to those who migrated before 1987, boys were less likely to migrate alone in two of the three subsequent periods. Among girls, however, none of the period coefficients are significant. Results from tests that compare these effects for boys and for girls (not shown here) substantiate that the observed effects for period of first trip are significantly different. Thus, although the propensity to migrate alone did not shift over time for girls, it shifted downward for boys both during the amnesty period immediately after IRCA was passed and after 2003, when the Great Recession emerged.

DISCUSSION

In the context of contemporary U.S. immigration, children have become an important topic of study. Yet, although scholars have begun to examine how migration affects the lives of family and children, few researchers have systematically investigated the factors that push children to migrate (on how migration affects families, see Abrego 2014; Dreby 2007, 2010, 2012, 2015; Levitt 2009; Donato and Duncan 2011; Gonzales 2011, 2015; Mazzucato and Schans 2011; Nobles 2011; Adserà and Tienda 2012; Donato and Sisk 2015). This study attempts to fill that void by estimating a series of event history models to predict the likelihood that children from Mexico take an initial U.S. trip, with and without documents, as a function of violence, migrant social networks, period of a child’s first trip, and other child and parent attributes. In addition, we consider two of the ways that unauthorized children enter the United States, with parents or alone, and examine the factors that are associated with variation in these outcomes.

Using MMP data merged with national rates of homicide in Mexico, we find evidence consistent with our first hypothesis. Exposure to serious violence—in the form of homicide—operates to push children to make a first trip with or without documents. In addition, as expected, parent’s migration is strongly associated with that of children. Those migrating in the same year as or before their parents had higher odds of making an initial trip. Having larger migrant social networks reduced the overall likelihood that a child migrated on first U.S. trip, but reduced the odds of unauthorized and increased the odds of authorized migration. In addition, well-defined period differences in the likelihood that children initiate a first trip emerged. Children were much more likely to make a first authorized trip after 1987—in any period—than before. However, among those making a first undocumented trip, the odds between 2007 and 2010 were no different than those for before 1987. The Great Recession thus appears to have diminished the odds that children initiated unauthorized migration.

To visually display some of these effects, we present predicted probabilities. Two striking findings emerge from these results. First, the chances that children initiate a U.S. trip were dramatically higher between 1997 and 2006, after the initial period of amnesty but before the Great Recession. The chances of migrating on a first authorized trip was approximately 26 percent, up from 5 percent between 1987 and 1996. Correspondingly, the chances of making an initial unauthorized trip also rose to 16 percent from 3 percent between 1987 and 1996. Thus, although many migrants regularized their status after 1986, the odds that children made a first trip rose well after the initial legalization period. Second, the probabilities reveal greater variation by legal status over time. Before IRCA’s amnesty program, that is, between 1972 and 1986, the odds that children initiated migration with and without documents were near zero. Beginning in the 1987 to 1996 period, legal status began to stratify the chances that a child migrated. The largest difference appeared in the 1997 to 2006 period, and the second largest in the 2007 to 2010 period (25 to 16 percent, and 9 to 5 percent, respectively).

In the second half of our analysis, we consider the drivers of two types of unauthorized children’s migration, that is, whether entry was linked to that of parents and siblings, and within these arrangements, whether boys and girls differed. For example, the likelihood that unauthorized children migrate when an unauthorized parent migrates appears to be driven by children’s attributes (of themselves and their parents) and period of entry. As expected, we see higher odds for boys who make an illegal first trip with at least one parent, but not girls, from 1987 to 1996 versus before 1987. However, we also see that boys have a lower likelihood of migrating alone between 1987 and 1996.

In contrast, key predictors of the odds that children migrate alone also include mother’s and father’s prior unauthorized migration status, migrant social networks, and period of entry, though showing some gender differences across the models. In particular, father’s prior unauthorized migration does not matter when predicting girls migrate alone. What matters for boys, but not girls, is the prevalence of migration. Moreover, among boys, the odds of making an unauthorized trip alone were substantially lower after 1987 than before.

One limitation of this study is that it is limited to existing MMP data. As a result, our findings derive from an analytic sample of children defined by parents’ first and most recent U.S. trip, rather than their complete migration histories. Thus, when parents had made more than two trips, we are unable to discern whether children migrated with them and therefore exclude children from the analysis. A second limitation is that, although the MMP contains information about children who migrated earlier in the twentieth century, that is, before 1972, we restricted our analysis to children who migrated between 1972 and 2010 because we had Mexican homicide and migration prevalence rates only for these years.

Despite limitations, our findings suggest that legal status is an important stratifier of children’s chances of initiating migration. Legal status impacts are well documented for Mexican adult out-migration and for labor market outcomes (see Donato and Massey 2016). This analysis, however, reveals how legal status stratifies the odds of children initiating migration, and how exposure to violence, access to migrant social networks, period of entry, as well as other attributes of children and parents, operate differently to help explain variation in children’s propensities to make an initial migrant trip.

Throughout the analysis, we also see clear signs that children’s migration is part of a slow process of family reunification. Yet despite connections to family residing in the United States, most Mexican children do not receive special protections in the migration process—even though many arrive as a by-product of the strong labor demand provided by their parents. As Jacqueline Bhabha suggests, this situation does not result from a lack of awareness about the problems that child migrants face; rather, it reflects society’s ambivalence about integrating migrant children (2014). Thus, although presidential executive actions, like DACA, may temporarily resolve some of the problems that these children and young adults face, until political will to fix the immigration policy regime is strong, they and many other migrant children will remain either invisible or unequal or both.