Legal Status and Educational Transitions for Mexican and Central American Immigrant Youth

Abstract

This study uses the Survey of Income and Program Participation to infer the legal status of Mexican and Central American immigrant youth and to investigate its relationship with educational attainment. We assess differences by legal status in high school graduation and college enrollment, decompose differences in college enrollment into the probability of high school graduation and the probability of high school graduates' enrollment in college and estimate the contributions of personal and family background characteristics to such differences. Results show that undocumented students are less likely than documented students to both graduate from high school and enroll in college, and differences in college enrollment cannot be explained by family background characteristics. We conclude that legal status is a critical axis of stratification for Latinos.

The rapid growth of the Latino population over the last several decades has fundamentally reshaped the racial/ethnic mix of the U.S. population. Latinos' success in navigating U.S. opportunity structures has important implications for both the pace of their incorporation and the nation's economic future. Given the enduring effects of schooling on status and economic attainment, understanding sources of inequalities in Latinos' educational transitions provides a lens through which to foresee the long-term ramifications of this demographic process. Thus, it is of considerable concern that this population suffers from serious educational deficits. Latino adolescents in general, and Mexican and Central American (hereafter “MCA”) youth in particular, have heightened rates of high school dropout (Fry 2005, 2011; Oropesa and Landale 2009; Perreira, Harris and Lee 2006) and are considerably less likely to enroll in college than other groups (Fry 2011).

While considerable research has explored individual-, family- and school-level explanations for the low levels of educational success observed for Latino youth, most prior empirical work is limited by its inability to account for legal status. Recent estimates indicate that there are approximately 11 million undocumented migrants living in the United States, of whom about 9 percent are younger than 18 years of age and nearly two-thirds hail from Mexico or Central America (Passel 2011; Passel and Cohn 2011). This group is not only large in size but also in share—one out of every 20 Latino children is unauthorized (Passel 2011)—and it is estimated that about 65,000 undocumented students receive a diploma from U.S. high schools each year and many more likely drop out before graduating (Passel 2003).

Children's legal status is likely to affect their educational progress both directly, due to state policies limiting access for undocumented immigrants, and indirectly through its association with family poverty and diminished expectations for the future. Because undocumented children have a legal right to K-12 education in the United States, children's legal status may become most salient as students leave high school and consider college enrollment, as undocumented students' access to higher education is limited in a many states (Diaz-Strong et al. 2011). Yet ethnographic work suggests that these barriers might influence earlier decisions about completing high school (Abrego 2006; Gonzales 2011; Suárez-Orozco et al. 2011).

Thus, the transition from high school to college represents a critical juncture at which legal status is likely to exert an important influence on immigrant children's educational attainment, as well as the pathways they follow during the transition to adulthood. As economic changes have rendered higher education increasingly necessary to enable today's young adults to support themselves and their families, pathways to adulthood that do not include higher education have become increasingly tenuous (Waters, Carr and Kefalas 2011). Given the effect of education on socioeconomic status, what happens to undocumented youth at this stage in the educational pipeline may have large, enduring effects on their future well-being and on patterns of MCA immigrant incorporation.

While some recent empirical work has explored these issues (Flores 2010; Kaushal 2008; Potochnick 2010), it is limited by a focus on a single point in the educational pipeline and by measurement issues in identifying the undocumented population. Our purpose in this article is to overcome these limitations by using a novel method for inferring the legal status of respondents in a nationally representative survey and examining the educational burden of “being illegal” for MCA youth at multiple points in their educational trajectories. We overcome the unavailability of direct measures of legal status by employing an approach developed by Hall, Greenman and Farkas (2010) that uses characteristics of migrants, including current and previous citizenship and visa status and public assistance enrollment, among others, to determine the documentation status of respondents in the 1996 through 2008 panels of the Survey of Income and Program Participation (SIPP).

Our research is guided by four goals: (1. to assess legal-status differences in high school graduation; (2. to assess legal-status differences in college enrollment; (3. to decompose differences in college enrollment into the portion attributable to high school completion and that due to high school graduates' matriculation; and (4. to estimate the extent to which the above result from observable differences in background characteristics between documented and undocumented students.

Background

Several theoretical models have been developed to explain youth educational attainment. A separate but related literature describes influences on educational attainment for youth in immigrant families. The effect of legal status has not been explicitly considered by any of these perspectives, but is theoretically relevant to all of them. Following Mare (1980), who characterizes educational attainment as a sequence of transitions, each influenced uniquely by social background, we focus on the transitions from high school enrollment to graduation and from high school graduation to college enrollment. We highlight three explanations for why undocumented status may limit educational progress: family socioeconomic disadvantage, the costs of attending college, and expectations for the future.

Family Socioeconomic Disadvantage

Undocumented students are more likely to come from socioeconomically strained households not only because immigrants who are able to cross the border legally tend to be more advantaged (Jasso et al. 2000), but also because undocumented migrants work in more dangerous settings, garner lower wages and are less likely to be provided with ancillary benefits (e.g., overtime pay, health insurance) than their legal counterparts (Donato et al. 2008; Flippen 2012; Hall, Greenman and Farkas 2010; Kossoudji and Cobb-Clark 2002). Furthermore, programs such as food stamps, which would otherwise reduce the effect of poverty on child outcomes, are not available to undocumented immigrants.

Extensive research has demonstrated the effects of family socioeconomic disadvantage on children's schooling success (Entwisle and Alexander 1992; Blau and Duncan 1967; Sewell and Hauser 1972). Both academic resources provided by parents, such as books in the home and extracurricular activities (Bodovski and Farkas 2008), and differences in parental practices, such as the complexity of the parents' speech when conversing with children (Lareau 2000), have been shown to mediate the effects of parental socioeconomic status (SES) on student achievement. Thus, educational disparities between documented and undocumented youth may simply be a reflection of differences in socioeconomic contexts. Additionally relevant to the impact of legal status on educational outcomes is that parents of undocumented students—who themselves are likely also undocumented—may be fearful of engaging with educational institutions and more likely to suffer from stress and financial uncertainty, both of which may diminish the quality of their children's academic environments (Yoshikawa 2010).

Direct Costs of College

The human capital model argues that students' educational decisions weigh both the costs and the benefits of school continuation. The costs of a college education include both the direct expenses, such as tuition and books, and the opportunity cost of other activities that time in college could be used for (Fuller, Manski and Wise 1982). Both lowering the cost of tuition (through policies such as in-state tuition) and providing aid to offset this cost (through financial assistance) are thus expected to increase enrollment (Schwartz 1985).

Current public polices raise the costs associated with college for undocumented students by restricting in-state tuition and financial aid. Policies governing whether undocumented students are eligible for in-state tuition in their home states have been among the most contentious immigration-related issues in recent years. As of August 2012 undocumented residents in 37 of the 50 states are ineligible for in-state tuition rates at state universities, with only 13 (California, Connecticut, Illinois, Kansas, Maryland, Nebraska, New Mexico, New York, Oklahoma, Rhode Island, Texas, Utah and Washington) allowing resident undocumented students to pay the in-state rate (National Immigration Law Center 2012). Several states in this group, namely, California and Texas, have large Mexican immigrant populations. Yet other states with large and/or rapidly growing ones, e.g., Arizona, Colorado, Georgia and South Carolina, have passed legislation that specifically prohibits undocumented students from receiving in-state tuition (Morse and Birnbach 2010).

Even for undocumented students who are eligible for in-state tuition, their ability to pay for college is restricted by their ineligibility for federal and most state financial aid. Private foundations and private colleges and universities are free to distribute aid to undocumented students, but many choose not to. Thus, the average direct costs of higher education are likely substantially higher for undocumented students and, in line with the human capital model, should reduce their odds of attending college.

Returns to Education and Expectations for the Future

Expected returns to the investment in education form the other half of the decision-making process under the human capital model. While such returns have traditionally been calculated in terms of the anticipated increase in earnings, researchers also recognize the nonpecuniary utility of a college education (e.g., prestige, health and longevity; Schwartz 1985). White collar and professional jobs normally aspired to by college graduates are less attainable for undocumented students, in part because they are either ineligible for them (e.g., government jobs) or run a high risk of being exposed as unauthorized, which in turn should lower the economic incentives to attain a college degree. While undocumented students can still reap the non-pecuniary rewards of a college education, their overall expected returns are thus considerably lower than those of other students.

Sociological models of status attainment have emphasized that students' expectations for the future are central to social mobility (Blau and Duncan 1967). The classic model articulates how family background characteristics, such as socioeconomic origins, affect educational attainment both by determining resources available to students, and by shaping how students' educational and occupational aspirations are influenced by significant others (e.g., parents, teachers, and peers; Sewell and Hauser 1972). Indeed, research indicates that undocumented students with access to adult mentors helping them navigate U.S. educational systems are more likely to make the transition from high school to college successfully (Gonzales 2010; Perez et al. 2009). The combination of limited professional job opportunities and low expectations for the future among undocumented students is likely to drastically undercut students' aspirations.

Educational and occupational aspirations are keys to accounts of the unique educational advantages of immigrants' children documented by many previous studies (e.g., Perreira, Harris and Lee 2006). Segmented assimilation theory, cultural-ecological theory (i.e., “oppositional culture”; Ogbu and Simons 1998) and the immigrant optimism hypothesis (Kao and Tienda 1995) all emphasize immigrant families' high levels of motivation and optimism as factors that promote educational success. However, these models do not explicitly consider differences between documented and undocumented immigrants. Even with highly motivated parents, “immigrant optimism” may be difficult to maintain for youth whose opportunities are structurally blocked by lack of legal status. The upshot is that undocumented students will have not only reduced rates of college enrollment but also heightened rates of high school dropout.

Existing Empirical Literature on Undocumented Students

Because of the sensitive nature of legal status and the danger to respondents in the event of disclosure, very little empirical work has directly ascertained legal status. Direct measurement of legal status is generally possible only in small-scale, qualitative research in which a high degree of trust exists between the researcher and respondents. Gonzales (2011) and Abrego (2006), for example, focus on educational transitions among high school-aged and college-aged youth and find that legal status takes on increasing importance as young people approach the transition to college (also Gonzales 2009; Suárez-Orozco et al. 2011). Both studies found that undocumented and documented immigrant Latino youth had similar schooling experiences until high school, with both groups suffering similarly from educational disadvantages such as low parental SES and enrollment in poorly funded, substandard public schools. In late high school, however, experiences such as applying for a driver's license or a part-time job force students to confront the realities of “being illegal” and the constraints that it places on everyday life (Gonzales 2011). Included here is recognition that their access to higher education—whether from costly tuition rates or ineligibility for financial aid—is severely limited (Abrego 2006).

Consistent with the idea that undocumented status lowers expectation for the future and increases the opportunity costs of higher education, qualitative research has found that undocumented students lack motivation in completing coursework in high school, reducing the chances of college enrollment and high school completion (Abrego 2006). Similarly, undocumented adolescents may also suffer from heightened anxiety resulting from their precarious state, making it more difficult for them to maintain their concentration on coursework (Potochnick and Perreira 2010). Such findings underscore the importance of considering this period in students' educational careers as a dynamic process, with decisions about high school continuation being influenced by expectations about higher education.

While no quantitative research has focused specifically on the linkage between students' legal status and educational transitions, a small number of studies focusing on the effects of state policies are relevant. Both Kashual (2008) and Flores (2010) attempted to assess the effect of in-state tuition policies on the college enrollment rates of undocumented Latino youth using data from the Current Population Survey (CPS). Because the CPS lacks information necessary to impute legal status of the foreign-born population, both use Latino noncitizens as a proxy for undocumented immigrants. Kaushal's descriptive statistics show that Mexican noncitizen youth are far less likely to attend college or graduate from high school than Mexican or Anglo citizen youth, while Flores' descriptive statistics indicate that Latino noncitizens are less likely to hold a high school diploma than Latinos as a whole.

These findings are suggestive of the importance of legal status for educational attainment, but are unable to distinguish the effect of legal status from that of citizenship and/or immigrant generation. Furthermore, both analyses include immigrants who arrived as young adults, even though they are unlikely to have ever enrolled in U.S. schools and are likely to have migrated for economic reasons. This contrasts with youth who arrived with their parents at ages in which they were both too young to work and obligated to attend American schools. Indeed, in studies of high school dropout behavior, inclusion of late-adolescent migrants has been shown to result in inflated estimates of dropout among Latino immigrants (Fry 2005; Oropesa and Landale 2009).

Another weakness of prior empirical research on this topic is that it has not evaluated the extent to which gaps in college enrollment are due to undocumented high school graduates being less likely to enroll in college or to their lower rates of high school completion. Potochnick (2010) provides an initial examination into this topic, by examining the effect of state higher education tuition policies on undocumented youths' likelihood of high school dropout, but using the same CPS-based measurement tool for identifying undocumented youth.

Data and Variables

We use data from the 1996, 2001, 2004 and 2008 panels of the SIPP—a panel study focused on U.S. households' employment and public program experiences—to explore the relationship between legal status and educational transitions. The SIPP design draws a large, nationally representative sample of U.S. households and interviews each household member every 4 months for 3 to 4 years. At each interview, respondents are asked a set of core questions and wave-specific topical questions that cover the reference month and 3 preceding months. In the 2004 panel, for example, respondents who completed all 12 interviews contributed 48 months of data. In cases where respondents are non-English speakers, SIPP provides translators.¹

SIPP is uniquely suited for this study for several reasons. First, these data contain necessary information on youths' enrollment in school and educational attainment. Second, the combined samples of the 1996, 2001, 2004 and 2008 panels include a large number of MCA immigrant youth. We included both Mexican and Central American immigrant youth in our sample because the 2008 panel of SIPP does not distinguish Mexican from Central American immigrants. Third, and perhaps most importantly, SIPP includes key variables, such as immigrant visa status, citizenship and federal program participation that can be used to assess the legality of MCA immigrants.

While SIPP follows respondents for 3-4 years, this is not a sufficient period to observe both high school dropout and college enrollment for a single cohort. Thus, respondents who are ever between 18 and 24 years of age are used for the college portion of the analysis, while those between 14 and 18 years of age are used for the high school portion. Given our focus on the progression of students within the U.S. educational system, we exclude immigrants who entered the United States primarily to work and who were never enrolled in U.S. schools by limiting our sample to those arriving to the United States before 14 years of age and having completed at least 8 years of schooling.

These criteria should exclude children who dropped out of school prior to migration because students typically complete seventh grade at 13 years of age; thus, an immigrant who arrived at 13 years of age and has completed at least 8 years of schooling would normally have completed at least 1 of these years in the United States. Therefore, the group of immigrant youth we evaluate here follows what is often referred to as the 1.5 generation. The high school-aged sample sizes used in our analysis are 728 for MCA immigrants, 3,177 for Latino natives and 17,917 for non-Latino white natives. For the college-aged sample, our data include 758 MCA immigrants, 3,877 Latino natives and 24,275 non-Latino white natives.

Our main dependent variables are high school dropout and college enrollment. Because SIPP provides monthly enrollment on respondents aged 15 years and older, we are able to track youths' educational progress throughout each panel. Children between 15 and 18 years of age who have yet to complete high school were determined to have dropped out if they were unenrolled in 3 consecutive months of the academic year. Because schools operate on slightly different calendars, and because SIPP interviewed respondents throughout a month, we conservatively define the academic year as spanning October to May. Thus, a student unenrolled in April and May would again have to be unenrolled come October to be classified as a dropout. To correct for students graduating early, we consider 17- and 18-year-olds unenrolled for 3 or more consecutive months but who report graduating from high school within 6 months as high school graduates. Unfortunately, we are unable to distinguish GEDs from high school diplomas.

We also examine high school dropout among our college-aged sample, using information on educational attainment. Our measure of college attendance is more straightforward, based on whether respondents aged 18-24 years are either currently enrolled in college or report having completed at least some college education. Unfortunately, SIPP does not contain information that would allow us to distinguish 2-year from 4-year colleges, public versus private institutions or institutional prestige, an important limitation given ethnographic evidence that undocumented students are more likely to enroll in community colleges than 4-year institutions (Teranishi et al. 2011).

Key control variables in our multivariate models include binary indicators of whether respondents are female or married. Whether the respondent is working for pay is measured with a dichotomous term. Age is expressed in terms of single-year dummy variables. For the high school sample, we include parental education, measured in years of schooling (averaged for children in two-parent families), monthly family income (averaged over the academic year), parental unemployment (whether a parent was unemployed during the academic year), and family structure (whether the child lived with both biological parents throughout the academic year). The number of older and younger siblings living in the household and whether high-school-aged children are living with their parents are also included.

In addition, we include age at immigration, categorized into those that arrived before the start of primary school (aged 0-5 years), those arriving during the elementary years (aged 6-11 years) and those arriving between 12 and 13 years of age.² To evaluate differential costs of attending college, we include a time-varying measure of whether respondents' states offer in-state tuition for undocumented immigrants.³ We also include year-specific state unemployment rates and percentage of all adults aged 25-64 years with some college education. Last, all multivariate models include dummy variables controlling for state and year. Estimates are weighted using longitudinal sampling weights to account for sample attrition.

Inferring Legal Status with SIPP Data

The focal explanatory variable in our analysis is legal status. While legal status is not assessed directly in SIPP, employing the approach developed by Hall, Greenman and Farkas (2010), we use available information to infer the legal status of MCA immigrant youth respondents. Among the questions asked of respondents aged 15 years and older at wave 2 are if they entered the United States with legal permanent resident (LPR) status, whether their status has since changed to LPR, and whether they are naturalized. Respondents who answer affirmatively to any of those questions are classified as legal.⁴

To further refine distinctions between remaining immigrants, we track participation in all federal assistance programs for which undocumented immigrants are ineligible (e.g., Food Stamps, Medicaid, Supplemental Security Income, Temporary Assistance for Needy Families). If immigrants report receiving benefits from one of these programs in their own name at any observation, then they are classified as legal. The remainder are either undocumented or fall into one of the following categories: refugees and asylees, students and exchange visitors, tourist/business travelers, temporary workers or diplomats and other political representatives (U.S. Department of Homeland Security 2011). SIPP does not sample tourists and other short-term visitors, and because our sample comprises children who migrated prior to 14 years of age, there will be few to none admitted on student, diplomat or temporary worker visas. (Children who have entered the country illegally are not eligible to later receive student or other visas.)

While very few Mexican immigrants have been granted asylum in the United States, immigrants from several Central American countries—particularly Nicaragua, El Salvador and Guatemala—have been admitted as refugees (or have been eligible to have their immigration status adjusted to asylee) after the conflicts in the region in the 1980s. Other Central Americans, including Hondurans, have been granted Temporary Protected Status after natural disasters during the late 1990s and early 2000s. Our imputation strategy may classify some such immigrants as undocumented when their legal status would be better described as “liminal legality” (Menjivar 2006). As Menjivar's (2006) work has shown, such temporary and provisional legal status is in many ways more similar to being undocumented than to being a legal immigrant. Overall, the number of refugees in our sample misclassified as undocumented is likely small given the numerical dominance of Mexicans among immigrants from the region.

For children who were younger than 15 years of age at the time immigration-related questions were asked, we infer their status based on that of their resident parents. Specifically, we assume children are undocumented if they were born abroad and if all of their biological, resident parents are deemed to be undocumented. Parents' documentation status is determined in a manner similar to that for older children, that is, they are deemed to be documented if they are U.S. citizens or LPRs, or if they “personally,” as opposed to “dependently,” receive federal welfare benefits. Immigrants who are potentially admitted with student visas are handled by classifying full-time students or married to full-time students as legal (student visas are contingent on full-time enrollment). We account for those admitted as diplomats by classifying MCA high-ranking public officials, as well as those married to high-ranking public officials, as legal. The residual group that we are unable to infer is temporary workers. Authorized temporary workers, however, form a comparatively small portion of MCA immigrants (Department of Homeland Security 2011). Nevertheless, our results should be assessed with some caution as the group we refer to as undocumented youth potentially includes a small proportion of the children of legal temporary workers.

Analytic Approach

Our analysis is carried out in two parts. The first provides individual-level estimates of differences in the likelihood of high school graduation by legal status, based on both our high school-aged and our college-aged samples. To do so, we use logistic regression to estimate the odds of lacking a high school degree among those old enough to have completed high school (aged 18-24 years). We adjust these estimates for differences in age, gender, age at immigration and state-level characteristics. Differences between documented and undocumented MCA immigrants not explained by such observable covariates provide an upper bound estimate of the barrier to high school graduation posed by legal status.

Next, we use the sample of high school-aged SIPP respondents (aged 14-18 years) to assess whether legal status affects high school dropout, and whether its influence persists with controls for personal and family background sociodemographic characteristics.⁵ We use discrete-time logit models to estimate the likelihood of high school dropout within single-age intervals, given that the respondent was still at risk of dropout at the beginning of the interval (i.e., had not yet graduated or dropped out; see Allison 1984). This research design allows observations to be included even if the outcome has not yet been determined by the end of the study, which will be the case for many respondents younger than 18 years of age at last observation.

Part 2 of the analysis investigates differences in college attendance by legal status. Using the subsample of 18-24 year olds, we use logistic regression to estimate: (1. legal status differences in the overall probability of college attendance; (2. differences in the probability of college attendance, conditional on having high school completion; and (3. the extent to which differences can be explained by observable sociodemographic and state-level characteristics.

Results

We start by examining overall differences between undocumented MCA immigrants, documented MCA immigrants, native Latinos and native whites in the means of family and individual control variables used in our multivariate models. The upper panel of Table 1 contains means for the high school-aged sample, which is structured in person-age format. The overall pattern in Table 1 is of large differences between MCA immigrants and native whites, but fewer and smaller differences between documented and undocumented immigrant youth. Especially of note are the very low levels of parental education for both documented and undocumented MCA immigrants. While parents of native white youth in our sample average about 14 years of education, those of documented immigrant youth average about 8.6 years, while those of undocumented youth average still lower at 7.8 years. Family income differences are similarly stark, with the average monthly family income of native white youth nearly twice as high as the averages of MCA immigrant youth. For both of these variables, native Latino youth fall squarely between immigrant and native white youth. The mean for parental unemployment is also notably lower for native whites than for any of the other groups. These results are consistent with those of past qualitative research, which suggest that the socioeconomic disadvantages of Latino immigrant adolescents are shared by both documented and undocumented youth (Abrego 2006). An exception to this pattern is family structure, with undocumented youth being less likely than documented youth to live with both biological parents. Finally, all three Latino groups have relatively similar patterns of geographic dispersion (and, correspondingly, state-level characteristics).

Table 1:

Sample Sizes and Means of Key Independent Variables, by Group

	Undoc MCA	Doc MCA	Native Hisp	Native White
14-18 Years of Age
Sample size (person-years)	380	1,430	8,054	46,053
Sample size (persons)	155	573	3,177	17,917
Female	.48	.50	.48	.48
Parental education (in years)	7.81	8.59	10.94	14.05
Lives with both parents	.66	.79	.67	.79
Single-parent home	.34	.20	.32	.19
Living alone	.01	.01	.01	.01
Number of younger siblings	1.72	1.63	1.30	.88
Number of older siblings	.26	.21	.21	.17
Parental unemployment	.14	.11	.09	.05
Teenager employed	.13	.13	.17	.28
Monthly family income (in $1,000s)	3079.23	3236.25	4196.27	6138.18
Age at immigration	6.85	6.05	–	–
State unemployment	6.68	6.68	6.65	6.04
State % some college	68.97	68.15	68.01	60.35
In-state tuition for undocumented	.44	.50	.52	.18
Lives in CA	.41	.38	.37	.08
Lives in AZ	.06	.04	.04	.02
Lives in IL	.06	.06	.05	.05
Lives in TX	.18	.23	.21	.05
18-24 Years of Age
Sample size (persons)	141	617	3,877	24,275
Female	.36	.46	.50	.50
Working at first observation	.51	.49	.53	.60
Married at first observation	.10	.14	.12	.10
Age at immigration	8.38	6.85	–	–
State unemployment	6.85	6.55	6.50	6.01
State % some college	69.50	68.23	67.47	60.39
In-state tuition for undocumented	.40	.40	.46	.17
Lives in CA	.52	.41	.32	.07
Lives in AZ	.02	.04	.04	.02
Lives in IL	.04	.04	.05	.05
Lives in TX	.11	.18	.22	.06

Because SIPP respondents living outside of their childhood homes cannot be linked to their families of origin, there are few variables available that are causally or temporally prior to the college enrollment decision for the 18-24-year-old (college-aged) sample. It is, however, noteworthy that only 36 percent of undocumented MCA immigrants in this age range are women, compared with 46 percent of documented MCA immigrants and 50 percent of the two native groups. This gender imbalance could be due to a higher rate of return migration among undocumented females or to the inadvertent inclusion of at least some labor migrants in the sample (despite our efforts to prevent this by restricting the sample to those arriving before 14 years of age).

Table 2 gives unadjusted group differences in the probabilities of high school completion, college enrollment and college enrollment conditional on high school graduation for the 18-24-year-old sample. There is a large, statistically significant difference by legal status in the percentage of MCA immigrants who have completed high school, with more than three quarters (77%) of documented immigrants having completed high school, but only two thirds (67%) of undocumented immigrants having done so. Both documented and undocumented youth also exhibit lower rates of high school completion than either Latino (84%) or white natives (91%). Thus, these results indicate that while legal status is an important source of stratification in immigrant youths' educational outcomes, even authorized MCA immigrants face substantial hurdles in completing high school in comparison to their native-born peers.

Table 2:

High School Graduation and College Enrollment Rates, SIPP Respondents Aged 18-24 Years

	% High School Graduates	% Ever Enrolled in College	% of High School Graduates Ever Enrolled in College
MCA Undoc Imm	66.8	13.0	19.5
MCA Doc Imm	77.3	21.8	28.2
Native Hispanic	83.5	34.4	41.2
Native White	90.9	42.8	47.1

Note: Differences between groups are all statistically significant at p < .05 (two-tailed t test).

There is an even larger difference by legal status in the likelihood of college enrollment. While about 22 percent of documented youth enroll in college, only about 13 percent of undocumented youth do so. Both immigrant groups have significantly lower college enrollment rates than either Latino or white natives. Because completing high school is required for college enrollment, we present results among only high school graduates in the third column. The difference by legal status remains strong, with 28 percent of documented high school graduates going on to enroll in college but only about 20 percent of undocumented high school graduates doing so. Both groups of immigrant high school graduates have significantly lower chances of enrolling in college than either native group. Thus, Table 2 suggests that legal status may act a stratifying force at multiple stages in the educational pipeline: during the transition from high school enrollment to high school graduation and from high school graduation to college enrollment.

Multivariate Models of High School Dropout

Table 3 presents results from bivariate and multivariate models of high school dropout. To better isolate the effect of legal status, we limit our multivariate models to MCA immigrants. We have also shifted our focus to dropout, rather than completion, to facilitate the use of event history models in the 14-18-year-old sample, many of whom were too young to have completed high school by the end of our observation periods. The left portion of the table contains results from the 18-24-year-old sample, while the right portion contains results from the 14-18-year-old sample. Given the observed (albeit small) differences in sociodemographic characteristics and geographic dispersion of documented and undocumented immigrants, it is possible that the legal status deficit in high school completion is simply a reflection of compositional differences between the two groups. The first column, consistent with the results in Table 2, shows a large, statistically significant association between documented status and high school dropout. The odds of dropout for documented youth are only 0.59 (e^−.525) times those of undocumented youth.

Table 3:

Multivariate Models of High School Dropout among MCA Immigrants, for 18-24-Year-Olds and 14-18-Year-Olds

	18-24-Year-Old Samplea				14-18-Year-Old Sampleb
					(1)		(2)
Documented Immigrant	-.525	(.204)*	-.477	(.276)+	-.392	(.275)	-.229	(.352)
Individual characteristics
Age 15	–	–	–	–	.412	(.526)	.253	(.518)
Age 16	–	–	–	–	.498	(.518)	.264	(.523)
Age 17	–	–	–	–	.899	(.502)	.592	(.528)
Age 18	–	–	–	–	1.452	(.477)**	1.065	(.535)*
Age 19 (at last obs.)	–	–	-.464	(.353)			–	–
Age 20 (at last obs.)	–	–	-1.375	(.409)**	–	–	–	–
Age 21 (at last obs.)	–	–	-1.299	(.472)**	–	–	–	–
Age 22 (at last obs.)	–	–	-1.306	(.465)**	–	–	–	–
Age 23 (at last obs.)	–	–	-1.873	(.522)***	–	–	–	–
Age 24 (at last obs.)	–	–	-1.651	(.461)***	–	–	–	–
Female	–	–	-.032	(.192)	–	–	.230	(.257)
Working	–	–			–	–	.714	(.302)*
Age at immigration, 6-11 years of age	–	–	.242	(.224)	–	–	.245	(.273)
Age at immigration, 12-13 years of age	–	–	.606	(.271)*	–	–	.596	(.431)
Household characteristics
Living alone	–	–	–	–	–	–	.366	(.324)**
Lives with both parents	–	–	–	–	–	–	2.381	(.874)
Number of younger siblings	–	–	–	–	–	–	.003	(.092)
Number of older siblings	–	–	–	–	–	–	-.236	(.408)
Parental education	–	–	–	–	–	–	-.069	(.036)
Parental unemployment	–	–	–	–	–	–	.926	(.302)**
Monthly family income (in $1,000)	–	–	–	–	–	–	.000	(.000)
State characteristics
State unemployment	–	–	.009	(.114)	–	–	.064	(.109)
State % some college	–	–	.018	(.022)	–	–	.056	(.033)
In-state tution for undocumented	–	–	-.616	(.524)	–	–	-.539	(.614)
In-state tuition X documented immigrant	–	–	.268	(.507)	–	–	-.698	(.637)
Intercept	-.699	(.181)	-3.01	(10.67)	-3.513	(.459)***	-12.58	(10.70)
Pseudo R-squared	.007		.135		.032		.160
χ2	6.59		80.51		17.93		120.28
N of persons	758		758		–		–
N of person-years	–		–		1810		1810

+ p < .10 *p < .05 **p < .01 ***p < .001

Note: Standard errors in parentheses; all models include controls for state and year. ^a Logit model predicting less than a high school education. ^b Discrete-time logit event history model predicting high school dropout at single age intervals.

Model 2, which presents results from the multivariate model, indicates that the observed covariates explain some, but not all, of the difference by documentation status. Specifically, net of controls, the disparity by legal status in the log odds of high school dropout remains moderately large and similar in size to that in the unadjusted model, but it remains statistically significant only at an alpha of .10.⁶ The drop in significance after the addition of covariates appears to be primarily due to the higher standard errors found in the multivariate models. The odds of dropout for documented youth are estimated to be about 38 percent lower (e^−.477-1) than those of undocumented youth.⁷ Respondents who were older than 19 years of age at the last available observation are also less likely to lack a high school degree. This is likely because of the fact that (1. GEDs are included in our measure of high school completion and (2. that a nontrivial number of young people older than 18 years of age finish high school .

Finally, immigrants who arrived at the United States at 12 or 13 years of age were more likely to have dropped out than those arriving prior to the start of first grade. Importantly, while the in-state college tuition variable and its interaction with documentation status operate in the expected direction—lowering the probability of dropout for undocumented students in states with in-state tuition policies for the undocumented—neither coefficient is significant.

We now turn to further exploring these differences in high school dropout. To account for family/parental characteristics, the sample we used for this stage of the analysis comprised respondents between 14 and 18 years of age whose records can be matched with those of a parent or guardian. Because many respondents are still younger than 18 years of age at last observation and therefore not old enough to have completed high school, the data are formatted into person-years and estimates are generated using discrete-time logit models.⁸ The outcome is whether high school dropout occurs within any particular academic year.

The results presented in the first model of the right half of Table 3 show overall differences by legal status in the likelihood of high school dropout among the 14-18-year-old sample, adjusting only for age. Here, we see that although the coefficient on the “documented” dummy variable is negative and about the same size as the corresponding coefficient in the college-aged sample, it is statistically insignificant (p = .15).

The second model for this group incorporates controls for family background and state-level characteristics. While the magnitude of the (nonsignificant) coefficient for documentation status is reduced by about two-fifths when these individual, family and state-level variables are introduced as controls, the difference between documented and undocumented MCA youth remains statistically insignificant. An examination of the other covariates included in the model indicates that working MCA immigrant youth, those living outside of the parental home and those with an unemployed parent have significantly higher odds of dropout, while those with better educated parents have significantly lower risk.

Multivariate Models of College Enrollment

Legal status is potentially important for not just high school completion but also college enrollment. Indeed, given differences in the costs of college for undocumented and documented immigrants, legal status may exert a more powerful effect on decisions regarding higher education than on primary schooling. The results in Table 4 explore this issue, presenting multivariate models of college enrollment using the 18-24-year-old sample. In addition to dummy variables for state and year, the models include controls for age, gender, whether the respondent was working or married at first SIPP observation and state-level characteristics. The in-state tuition variable is again interacted with documented status to allow the coefficients to differ for undocumented and documented respondents. The left panel predicts college enrollment for the sample as a whole, regardless of having obtained a high school diploma. This outcome can be thought of as a combined measure of the probability of having graduated from high school and the probability of enrolling in college, conditional on high school graduation. The results show a large, statistically significant difference by legal status. The odds of enrollment for documented young adults are more than four times those of undocumented young adults.

Table 4:

Logistic Regression Models of College Enrollment, SIPP Respondents Aged 18-24 Years

	College Enrollment (Entire Sample)		College Enrollment (High School Graduates Only)
Documented Immigrant	1.424	(.597)*	1.367	(.620)*
Individual characteristics
Age 19 (at last obs.)	1.209	(.465)**	1.137	(.521)*
Age 20 (at last obs.)	.455	(.554)	.084	(.600)
Age 21 (at last obs.)	1.036	(.607)	.729	(.646)
Age 22 (at last obs.)	.887	(.608)	.474	(.648)
Age 23 (at last obs.)	1.196	(.608)	.600	(.657)
Age 24 (at last obs.)	-.185	(.622)	-.699	(.666)
Female	.110	(.226)	.038	(.241)
Working at first obs	.330	(.228)	.234	(.238)
Married at first obs	-1.137	(.476)*	-.816	(.509)
Age at immigration, 6-11 years of age	.029	(.246)	.159	(.262)
Age at immigration, 12-13 years of age	-.660	(.343)+	-.649	(.361)+
State characteristics
State unemployment	.025	(.120)	.023	(.134)
State % some college	.038	(.025)	.055	(.032)
In-state tuition for undocumented	.892	(.751)	.733	(.825)
In-state tuition X documented immigrant	-1.255	(.708)+	-1.323	(.743)
Intercept	-10.620	(11.528)	-10.558	(12.685)
Pseudo R-squared	.197		.197
χ2	82.55		73.90
N of persons	758		552

+ p < .10 *p < .05 **p < .01 ***p < .001

Note: Standard errors in parentheses; all models include controls for state and year.

Is this difference by legal status simply reflecting the difference in the likelihood of high school graduation discussed earlier, or is there an additional difference in the likelihood of college enrollment, net of high school graduation? This question is addressed in the right panel of Table 4. The coefficient for documented status is still large and statistically significant, indicating that legal status is related to higher education decisions after the high school years. Among high school graduates, documented young adults' odds of going on to college are still nearly four times those of undocumented young adults. While the main effect of in-state tuition for undocumented students does not significantly predict college enrollment in either model, its interaction is marginally significant (p = .08), potentially suggesting that state tuition policies have a stronger relationship with college enrollment for undocumented than for documented youth.

Discussion

As highlighted by the ongoing debate over the DREAM Act and the more recent decision by the Obama administration to grant temporary deferrals and legal work permits to certain young, undocumented immigrants, policies directed at undocumented immigrant youth continue to be an important topic in political discussions. However, very little is known about how undocumented status shapes immigrant students' educational outcomes. Although current educational policies provide undocumented students greater access to education at the high school level than at the college level, there is reason to believe that being undocumented will make students uncertain about their prospects for the future—in terms of returns to higher education or access to professional jobs—and consequently influence decisions about whether to complete high school. Thus, any attempts to examine the influence of legal status on educational attainment must take a complete view of the educational pipeline, considering the transition from both high school enrollment to high school graduation and high school graduation to college matriculation.

Using samples of high school-aged and college-aged youth from the 1996, 2001, 2004 and 2008 panels of the SIPP and an imputation strategy that estimates the legal status of immigrant respondents, we examined whether “being illegal” is related to high school graduation and college enrollment for MCA immigrant youth who arrived in the United States as children. Not surprisingly, our results indicate that legal status matters: we find that the odds of college enrollment are about four times higher for documented immigrants than their undocumented peers. Among our older cohort (18-24 years of age), there is also evidence that undocumented students are less likely to complete high school; but after adjusting for covariates, this difference loses statistical significance. Among our younger cohort, we did not find evidence of a relationship between documentation status and high school graduation. This may suggest that documented and undocumented MCA youth share similar barriers to high school completion. Overall, our results indicate that although lacking legal authorization reduces the odds that MCA immigrant youth will attend college, it yields mixed conclusions on how legal status relates to high school completion.

Differences by legal status in college enrollment are not surprising, given that undocumented students face significant barriers to higher education. Most notable are access to federal financial aid and, in some states, reduced in-state tuition. Previous work has investigated the effects of state policies on both high school dropout and college enrollment for Mexican noncitizens, finding that in-state tuition is related to both lower risk of dropout (Potochnick 2010) and slightly higher rates of college enrollment (Flores 2010; Kashual 2008). We tested whether undocumented students' eligibility for in-state tuition attenuated the gap between undocumented and documented immigrant youth in college enrollment, but found little evidence of effect.⁹ However, our relatively small sample size, as well as the concentration of our sample in states that adopted in-state tuition policies, may make it difficult for us to detect statistically significant policy effects. Given that previous research using larger samples has found small effects of tuition policies on noncitizen Latino immigrants' educational outcomes, we would urge readers to interpret this result with caution.

Furthermore, it is important to note that having to pay higher tuition rates is only one of many significant barriers to college access for undocumented students. Given the very low family incomes of many undocumented students, even reduced in-state tuition may remain beyond the reach of many. Lacking access to federal financial aid—a policy that affects all undocumented immigrants—may in fact be a more significant hurdle than tuition rates. The fear of disclosure of undocumented status during the college application process and the logistical challenges associated with undocumented status (e.g., driving to campus without a license and paying for books and supplies without legal part-time work) (Gonzales 2011) likely serve as additional impediments to college enrollment.

Moreover, regardless of state education policies, in the absence of federal policies that normalize the status of undocumented youth, their returns to higher education will remain limited. In line with this idea, research has demonstrated that undocumented Mexican immigrants receive very few wage returns to schooling (Hall, Greenman and Farkas 2010).

How should our findings for high school graduation be interpreted? On the one hand, given that current educational policies guarantee access to K-12 education for undocumented students, it is not surprising that differences by legal status should be weaker for high school graduation than for college enrollment. On the other hand, some degree of difference makes sense in light of the weaker and/or uncertain labor market prospects facing undocumented workers. Most jobs available to undocumented immigrant workers involve manual labor or service work and require very low levels of schooling. Undocumented students may therefore expect to receive few economic benefits from staying in high school beyond the legally required age. In the current environment, it is thus rational for undocumented students to cease schooling prior to high school graduation unless they highly value the nonmonetary benefits of schooling.

While we focus specifically on educational transitions, our findings also have implications for other key components of the transition to adulthood. Waters, Carr and Kefalas (2011) emphasize the flexibility that young adults exhibit in overcoming challenges during the transition to adulthood: “Young adults innovate, they change, they retrain, they multitask, and they make their lives more livable even in the most difficult of circumstances.”(14) Yet it is likely that undocumented status limits this ability to adapt. Undocumented young adults have not only limited ability to engage in retraining or further education, but also face barriers to transitioning to stable, adult work roles because of the uncertain and contingent nature of many jobs available to undocumented workers. It is unclear how these limitations may affect other key transitions, such as those involving family formation. We plan to explore this topic in future research.

Our analysis contains several limitations. First, our measures of both outcome variables are somewhat blunt. For high school completion, we are unable to distinguish between respondents with regular high school diplomas and those with GEDs. Because previous literature has found that GED holders differ in many ways from high school graduates (Heckman and Rubinstein 2001), this distinction may be important. Similarly, our measure of college enrollment simply distinguishes between those who enroll in any type of college (and for any length of time) from those who did not. There is considerable evidence that the effects of college attendance on career outcomes vary greatly depending on factors such as whether the college is a 2-year or a 4-year institution (Louie 2005), whether enrollment occurs promptly after high school graduation (Kempner and Kinnick 1990) and whether attendance is continuous or interrupted (Goldrick-Rab 2006), all of which could not be assessed using SIPP data.

Thus, there may be crucial differences in the college attendance patterns of undocumented immigrants in addition to the lower enrollment rates we document here. More immigrant students in general attend community colleges than 4-year colleges (Teranishi et al. 2011), and undocumented immigrant students are probably particularly likely to do so given their lower cost. Such differences in postsecondary institution type can have implications for how much students benefit from postsecondary education (Louie 2005). The timing of college attendance may also be a critical difference between undocumented and documented students. Because of their relatively low placement on the socioeconomic scale, the college careers of undocumented students may be characterized by late starts and interrupted trajectories. A complete account of the role of undocumented status in the educational pipeline would need to capture such differences.

An additional limitation stems from SIPP's lack of information on prior educational achievement or educational aspirations. We suggested in the background section of the article that lower educational aspirations could serve as a mechanism linking undocumented status to lower educational attainment, but were unable to test this explanation with SIPP data. Because our theoretical framework suggests that legal status is causally prior to educational aspirations, we see this as a causal mechanism that deserves further exploration, rather than an omitted variable that may bias our results.

Likewise, prior educational achievement is a key factor given its strong association with ultimate attainment. Previous work has suggested that parents' lack of documentation status affects their ability to provide an educationally stimulating environment for children, resulting in lower academic achievement among children (Yoshikawa 2010). This implies that educational achievement could be another mechanism through which undocumented status could affect educational attainment. Alternatively, undocumented and documented immigrants could differ in terms of ability due to selection factors; in this case, our results could be biased by our inability to control for prior ability or achievement.

In sum, our findings confirm that unauthorized status likely limits college enrollment for MCA immigrant youth. Although our evidence is weaker for high school graduation, we do find some evidence that undocumented status may be related to higher high school dropout rates as well. Current policy discussions center on whether undocumented youth who arrived in the United States as children should have access to higher education and job opportunities after becoming young adults. These discussions rarely mention undocumented youths who are still minors, probably because of the unspoken assumption that high school-aged and younger students are not affected by such policies.

Although our ability to draw definite conclusions is limited by constraints of the data, our results are consistent with a scenario in which anticipation of blocked opportunities encourages high school dropout for undocumented students. This interpretation is bolstered by the findings of previous research showing that high school dropout for Mexican noncitizen youth is lower in states that adopted in-state tuition policies (Potochnick 2010). The consequences of the current restrictive policies may thus be more serious than previously suspected. Not completing high school predicts a wide range of important social and economic outcomes, including poverty, unmarried parenthood and incarceration (Sum, Khatiwada and McLaughlin 2009).