## Abstract

Deferred Action for Childhood Arrivals (DACA) is the first large-scale immigration policy to affect undocumented immigrants in the United States in decades and offers eligible undocumented youth temporary relief from deportation as well as renewable work permits. Although DACA has improved the economic conditions and mental health of undocumented immigrants, we do not know how DACA improves the social mobility of undocumented immigrants through its effect on educational attainment. We use administrative data on students attending a large public university to estimate the effect of DACA on undocumented students’ educational outcomes. The data are unique because they accurately identify students’ legal status, account for individual heterogeneity, and allow separate analysis of students attending community colleges versus four-year colleges. Results from difference-in-difference estimates demonstrate that as a temporary work permit program, DACA incentivizes work over educational investments but that the effect of DACA on educational investments depends on how easily colleges accommodate working students. At four-year colleges, DACA induces undocumented students to make binary choices between attending school full-time and dropping out of school to work. At community colleges, undocumented students have the flexibility to reduce course work to accommodate increased work hours. Overall, the results suggest that the precarious and temporary nature of DACA creates barriers to educational investments.

## Introduction

An estimated 11 million undocumented immigrants reside in the United States (Passel and Cohn 2016). They account for one-quarter of the foreign-born population and 5 % of the labor force, and they are responsible for almost 3 % of gross domestic product (GDP) (Edwards and Ortega 2016). Many of these immigrants are colloquially dubbed “dreamers,” referring to undocumented youth who were brought to the United States as children. Undocumented youth have a constitutional right to K–12 public education, but they come to face the realities of their illegal residency status as they transition into adulthood (Gonzales 2011; Gonzales et al. 2014). Without legal residency status, they cannot legally work or vote and are under the threat of deportation.

Recent efforts to reform immigration policies have focused on expanding opportunities for “dreamers” because public sympathy for them remains strong. Since 2001, legislators have attempted to enact the Development, Relief, and Education for Alien Minors (DREAM) Act, which offers legal status and pathways to citizenship for undocumented youth who entered the United States as children. In 2010, the DREAM Act failed to pass the U.S. Senate. In response, President Obama enacted Deferred Action for Childhood Arrivals (DACA) in June 2012 through executive action. DACA grants two-year renewable work permits and temporary relief from deportation to eligible undocumented youth. Of the estimated 1.7 million of eligible youth, more than 740,000 applications were approved for DACA as of 2016 (U.S. Citizenship and Immigration Service 2013).

As a stop-gap measure intended to offer temporary legal employment options to undocumented youth in the absence of viable options for legal residency, the program has been successful in increasing labor force participation among undocumented youth (Amuedo-Dorantes and Antman 2017; Pope 2016) and reducing poverty among households headed by DACA–eligible immigrants (Amuedo-Dorantes and Antman 2017). The positive effects that DACA has on labor force participation parallels the findings that other types of programs granting temporary work permits, such as temporary protected status, also improve the labor market conditions of undocumented immigrants (Orrenius and Zavodny 2015). In addition, research has shown that DACA increases the economic and social incorporation of recipients by allowing them increased opportunities to open bank accounts and obtain credit cards (Gonzales et al. 2014) and improves recipients’ mental health (Hainmueller et al. 2017; Patler and Pirtle 2018; Venkataramani et al. 2017).

The effect of DACA on higher education, however, remains unclear. Education has long been viewed as an engine for social mobility. Understanding the effects of DACA on the college attendance of undocumented students offers insight into how temporary work permits can affect the socioeconomic integration and well-being of their recipients. Do temporary work permits raise the returns to schooling and encourage college attendance? Or does the short planning horizon associated with two-year work permits distort educational decisions and limit the chances for upward social mobility? To date, an estimated 250,000 undocumented students are enrolled in postsecondary schools. Yet, despite the importance of these issues, we know very little about how DACA affects the educational choices and outcomes of undocumented college students.

Our study addresses these questions using a unique data set that accurately identifies legal status. We use a difference-in-difference approach to estimate the causal effect of DACA on the educational outcomes of undocumented students. We separately analyze community colleges and four-year colleges because these two types of institutions differ in terms of how students typically balance schooling and work: our results suggest that DACA has important effects on this trade-off. Our results also demonstrate that DACA increases the dropout rates of undocumented students attending four-year colleges and causes undocumented students attending two-year community colleges to switch from full-time to part-time attendance. These findings suggest that despite evidence of DACA’s positive economic effect on undocumented immigrants through increased labor force participation, the temporary and precarious nature of DACA status incentivizes work over schooling.

## Background and Prior Research

### Undocumented Students in Higher Education

Just like immigrants with legal status, undocumented students tend to be first-generation college-goers from low-income families, who struggle to graduate with their intended degree (Bailey et al. 2015; Suarez-Orozco et al. 2015). However, undocumented students face additional obstacles to college enrollment, attendance, and graduation. First, they attend college under the threat of deportation for themselves and their family members, which makes institutional interactions intimidating (Suarez-Orozco et al. 2015). Second, the cost of attending college is higher for undocumented students because they do not qualify for federal financial aid, and the returns are lowered by limited employment options (Conger and Turner 2017; Enriquez 2017). Third, undocumented youth face greater pressure to contribute to household finances (Gleeson and Gonzales 2012; Terriquez 2014) and are at greater risk of leaving school early. Finally, the returns to education are uncertain for undocumented youth because they cannot legally work. Thus, undocumented youth are less likely to enroll in college than their peers with legal status (Greenman and Hall 2013).

Despite facing great barriers to entry, an estimated 250,000 undocumented youth currently attend college in the United States (Passel and Cohn 2009). Yet, our understanding of the higher education experiences of undocumented immigrant youth is limited. Efforts to better understand their academic trajectories and outcomes are hampered by data constraints. First, the U.S. Census and most large-scale national surveys do not contain information on immigrants’ legal status. As a result, researchers need to rely on imputations of undocumented status. These imputation methods have evolved considerably over the last few decades (Passel and Cohn 2009; Warren and Warren 2013). However, some authors have shown that these imputation methods can lead to large bias in some applications (Van Hook et al. 2015).

Recent studies have either treated all foreign-born residents, including those who are legally authorized to be in the United States (i.e., legal permanent residents (LPRs)), as undocumented (Flores 2010; Kaushal 2008; Potochnick 2014) or treated students who hold student visas or who have refugee or asylum status as undocumented (Greenman and Hall 2013). Other researchers have employed online surveys as a tool for accessing the undocumented student population, but voluntary Internet surveys are very likely to suffer from selection bias, potentially excluding students who are less politically active or from low-income backgrounds (Gonzales et al. 2014; Suarez-Orozco et al. 2015). As a consequence, much of our knowledge of the experiences of undocumented youth is informed by qualitative studies focusing on specific populations (i.e., Mexicans in California) or students attending selective four-year colleges (Abrego 2006; Contreras 2009; Enriquez 2017; Garcia and Tierney 2011; Gonzales 2011).

The second important limitation is the lack of longitudinal data. Most studies have relied on cross-sectional surveys, such as the Current Population Survey (CPS) or the American Community Survey (ACS), and are likely to suffer from estimation bias arising from unobserved individual heterogeneity. Undocumented youth who enroll in higher education tend to be more positively selected; they are academically gifted, motivated, and resilient individuals with exceptionally high educational aspirations (Conger and Chellman 2013; Contreras 2009; Perez and Cortes 2011; Terriquez 2014).1 These characteristics likely correlate strongly with decisions to seek employment or to enroll in college. Failing to account for these unobserved differences would likely introduce omitted variable bias.

### Deferred Action for Childhood Arrivals and Higher Education

The effects of DACA on college attainment are theoretically ambiguous, and the existing empirical literature provides mixed evidence. On the one hand, DACA may have a positive effect on college attainment. DACA creates a “semi-legal” status (Menjívar 2006) in which recipients receive greater legal protections, legal employment opportunities, and greater certainty for the future. By extending work authorization to recipients, DACA also increases returns to a college degree, which may incentivize recipients to stay in school, focus on coursework and complete their degrees promptly. On the other hand, DACA may have a negative effect on college attainment. Nearly 70 % of families headed by undocumented parents subsist at or near the poverty line (Amuedo-Dorantes and Antman 2017). Undocumented immigrants are typically employed in low-wage, unstable jobs that do not offer benefits such as health insurance, sick leave, or overtime pay (Donato et al. 2008; Hall et al. 2010). Thus, families headed by undocumented parents commonly rely on all working-age members to contribute to the family income. By providing access to the legal segment of the labor market, DACA presents an opportunity to increase household earnings, which raises the opportunity cost of attending school. As a result, DACA status may lead undocumented college students to drop out of school in order to take advantage of the enhanced earning opportunities.

The existing empirical analyses of the effects of DACA on the educational outcomes of undocumented youth face the same data limitations that all quantitative efforts to study the undocumented population face. Internet-based surveys of undocumented college students tend to show that DACA enables recipients to pursue educational opportunities that they previously could not (Wong et al. 2015). However, respondents of online surveys are self-selected and likely to be higher-achieving and more motivated than the general population of undocumented students. Thus, it is unclear whether the findings based on online surveys can be generalized to the entire population of undocumented college students.

Two studies using nationally representative survey data examined the effect of DACA on employment and education outcomes among likely DACA–eligible youth: the ACS and the CPS (Amuedo-Dorantes and Antman 2017; Pope 2016). Lacking information on immigrants’ legal status, all noncitizens in a given age range were assumed to be undocumented. Both studies found positive effects of DACA on employment but mixed results on schooling. Whereas Amuedo-Dorantes and Antman (2017) found that DACA reduces college enrollment among probable DACA eligible students, Pope (2016) found no significant effect of DACA on schooling.

Our study extends the existing literature on the effects of DACA on college attendance by using unique administrative data from one of the largest public university systems in the country. These data allow us to overcome many of the data limitations plaguing previous studies because we can accurately identify students’ legal status, account for individual heterogeneity, and separately consider students enrolled at community and four-year colleges.

We can reliably identify legal status because the university is located in one of 21 U.S. states that offer in-state tuition to resident undocumented immigrants. Specifically, to qualify for in-state tuition, undocumented students must submit notarized affidavits attesting to their legal status and committing to the pursuit of legalization should it become possible. Accurately reporting legal status confers large financial incentives because in-state tuition is substantially lower than out-of-state tuition.

Another key feature of the data is that we can distinguish between community colleges and four-year colleges that offer bachelor’s degrees. Prior studies (Amuedo-Dorantes and Antman 2017; Pope 2016) did not consider how the effect of DACA on educational outcomes might vary by type of (college) institution. Distinguishing between community and four-year colleges may be important for understanding the effects of DACA on college attendance. Students at four-year colleges may face a sharp trade-off between studying and working, whereas community colleges provide greater flexibility to students who want to combine part-time enrollment with work. Nationally, less than one-third of four-year college students work, compared with nearly 70 % of community college students (Bureau of Labor Statistics 2017). As a result, we hypothesize that DACA will have larger effects on attendance at four-year colleges than at community colleges.

Finally, the data track students over time, even when they switch from one college to another within the UCS system. The longitudinal nature of the data allows us to improve on previous studies in two ways. First, we can estimate individual fixed-effects models to account for unobserved individual heterogeneity. Second, the longitudinal nature of the data allows us to restrict our sample to students who were enrolled in college during the passage of DACA and study the effects of DACA on this subgroup of students. Prior studies have used the proportion of undocumented students enrolled in college as their outcome variable of interest (Amuedo-Dorantes and Antman 2017; Pope 2016). This method conflates variation in enrollment patterns (i.e., temporal variation in the number and type of students enrolled in college every year) and variation in retention rates (i.e., currently enrolled students who drop out of school).

## Data and Methods

### Data

We analyze administrative data from one of the largest public university systems in the country. Because of data confidentiality agreements with the university system, we have anonymized the data source and refer to it as Urban College System (UCS). The UCS is set in a major metropolitan area and educates more 260,000 degree-seekers across 18 undergraduate campuses, 7 of which are community colleges. Nearly 80 % of undocumented college students living in the major metropolitan area attend UCS (DiNapoli and Bleiwas 2014). Therefore, our analytical sample of undocumented college students includes nearly the entire universe of undocumented students attending college in this large metropolitan area.

After individuals enter the administrative records, they are followed over 10 years. We analyze entering cohorts from fall 2009 to fall 2012. This analytical sample includes four cohorts of students who entered the university system during the four years immediately prior to DACA implementation. We exclude cohorts who enter post–DACA to avoid bias due to the possibility that undocumented students who entered college post–DACA may be differentially selected relative to those who entered college prior to DACA. Our analytical sample comprises 385,467 students: 198,986 attending two-year colleges, and 186,481 attending four-year colleges.

### Measures

We focus on two main outcome variables for our student population: a dropout indicator and a full-time enrollment indicator. Dropout is measured as a dummy variable indicating that a student who was previously enrolled is no longer enrolled (dropout = 1) in a given year (as opposed to remaining enrolled or having graduated). Full-time attendance is measured as a dummy variable indicating that a student completed 24 credits or more during the academic year (full-time = 1) and is defined only for the subset of students enrolled at each point in time.

Our main explanatory variable is the student’s immigration and legal status. Students are asked to self-report as U.S. citizens, LPRs, or undocumented immigrants at time of initial enrollment. Students must submit documentation to validate their own self-reports, and undocumented students must provide a notarized affidavit stating that they will pursue steps to obtain legal residency if such options become available.

Large financial incentives are in place for undocumented students to self-identify because UCS is located in one of 21 states that offer in-state tuition to undocumented students who graduated from a high school or obtained a GED from within the state. Out-of-state tuition for a full-time student at a UCS four-year college in 2016 was approximately $17,000, compared with$6,500 for in-state tuition. At community colleges, out-of-state tuition for a full-time student was approximately $9,500, compared with$5,000 for in-state tuition.

Legal status is measured as a dummy variable indicating that a student is undocumented at time of enrollment. Individuals who obtained their high school degree outside the United States and self-report as undocumented (N = 762) and individuals who obtained their high school degree in the United States but outside the state (N = 338) are excluded from the analytical sample. This step was taken to eliminate foreign students or out-of-state documented students who might self-report undocumented status to gain in-state tuition.

### Analytical Strategy

Our empirical strategy exploits changes in our outcome variables for undocumented students before and after DACA, relative to changes for documented students over the same period. Netting out the changes in outcomes for documented students allows us to purge the effects of unobserved factors that affected all students similarly, such as changes in local economic conditions.

Specifically, our difference-in-difference estimation is based on the following linear probability model:
$Yitc=αi+αt+αc+βPostt×Undoci+εitc.$
1
The dependent variable Yitc is the outcome variable for individual i in cohort c in calendar year t. Importantly, the specification includes individual fixed-effects, denoted by αi, that absorb all time-invariant characteristics of individuals (such as ability, motivation, race/ethnicity, and family background). Additionally, our specification includes dummy variables for calendar year (αt) and years since enrollment (αc). The former account for time-varying aggregate effects, such as local labor market conditions, and the latter set of fixed effects accounts for the differences in dropout rates (and full-time status) as a student progresses toward graduation. The dummy variable Undoci indicates whether student i reported being undocumented, and Postt is an indicator variable marking the rollout of DACA. Last, a disturbance term, εitc, captures all idiosyncratic variation in the outcome variable that is not picked up by any of the aforementioned regressors. When the outcome variable is full-time enrollment, the sample is restricted to currently enrolled students.

The key parameter of interest is β, the coefficient on the interaction term between Postt and Undoci. This coefficient is identified by the changes in the outcome variable for undocumented students before and after DACA, net of changes for documented students in the same period. In addition to the difference-in-difference estimation, we also estimate a more flexible specification that allows for time-varying gaps in outcomes between documented and undocumented students. The results of this specification will be useful to assess the validity of the identification assumption of common trends required to provide a causal interpretation of our estimates.

One important caveat is that we cannot determine DACA eligibility perfectly. To be eligible, undocumented immigrants need a high school diploma (or a GED, or having been honorably discharged from the Armed Forces), to have arrived to the United States before age 16, to be of continuous residence in the United States since 2007, and to have a clean criminal record. All students in our sample have fulfilled the first requirement, but we cannot determine whether they fulfill the other requirements. Nevertheless, it is likely that most undocumented students in our data are DACA–eligible.

Because we cannot know which students in our sample are DACA recipients, our estimates of β should be interpreted as intent-to-treat effects. Average treatment effects (on the treated) will likely be substantially larger because not all eligible individuals have applied for DACA. Nearly 50 % of eligible youth who reside in the state where UCS is located applied for DACA (Batalova et al. 2014), and nearly 95 % of those who applied were approved (USCIS 2017). This compliance ratio implies that the average treatment effect (on the treated) could be up to twice as large as the intent-to-treat effect.2

To further verify our claim that estimates should be regarded as intent-to-treat effects, we consider whether the effect of DACA varies across groups with higher and lower DACA application rates. To do this, we examine variation in difference-in-difference estimates across country of origin by including three-way interactions among dummy variables for country of origin, documentation status, and post–DACA years. Hipsman et al. (2016) reported substantial variation in DACA application rates by country of origin. Application rates among immigrants originating from Latin American countries are among the highest of any group. For example, the application rates among immigrants from Mexico, El Salvador, and Honduras exceed 70 %. In contrast, application rates among undocumented immigrants from Asian countries are substantially lower. For example, less than 1 % of eligible immigrants from China applied for DACA, and application rates for immigrants originating from South Korea, India, and the Philippines are all well below 30 %. Thus, we expect to see larger effects of DACA among students originating from Latin America and smaller effects of DACA among students originating from Asia.

Difference-in-difference estimates provide our main analytical results. In addition, we also estimate a flexible specification that allows for time-varying gaps in outcomes between documented and undocumented students:
$Yitc=αi+αc+αt+βtUndoci+εitc.$
2
The dependent variable Yitc represents a dummy variable for dropout or full-time status for individual i in cohort c in calendar year t. Terms αi, αc, and αt are fixed effects for individuals, years since enrollment, and calendar year, respectively. βt captures the difference in the dependent variable between undocumented students and their legal status counterparts for every year t. Regressions for full-time attendance are restricted to the subset of students who are enrolled at each point in time. We plot the estimated βt to assess parallel trends in the pre–DACA period.

### Timing of DACA Implementation

Understanding the date of implementation of the DACA program and when one should expect to see effects on academic outcomes is critical. On June 15, 2012, then-President Barack Obama announced the DACA program. Applications were accepted beginning on August 15, 2012, but very few cases were approved until after October 2012, with the vast majority of approvals occurring after December 2012 (Batalova et al. 2014). Figure 1 shows the number of approved DACA cases from the time when applications were first accepted to July 2013. For college students, this means that DACA is announced during their 2012 summer break. The vast majority of DACA applicants in college would have been approved during or after their spring 2013 semester. Thus, any anticipated effect of DACA should be observed during calendar year 2013 and beyond. Accordingly, indicator variable Postt takes a value of 1 for calendar years 2013 and onward.

## Results

### Descriptive Statistics

Summary statistics for outcome variables by legal status and college type are presented in Table 1. We also include other characteristics of students in Table 1 to provide a descriptive profile of students. However, they are not included as covariates in our regressions because they do not vary over time, and they are thus absorbed by the individual fixed effects.

The most striking result Table 1 is the positive selection of undocumented students relative to their peers with legal status. At community colleges, documented students have high school grade point averages that are 0.44 standard deviations below the sample mean, whereas undocumented students have grade point averages that are 0.29 standard deviations below the sample mean. At four-year colleges, the high school grade point averages for documented and undocumented students are 0.50 and 0.75 standard deviations above the sample mean, respectively. The positive selection of undocumented students relative to their legal status peers may explain why the statistics also show that undocumented students are less likely to drop out of college and more likely to attend college full-time than their counterparts with legal status.

Figure 2 plots the dropout rates for undocumented and documented students, as a function of years since enrollment, without making any adjustments to the raw data. The top four graphs correspond to four-year colleges, and the bottom four refer to community colleges. The figure reveals three noteworthy findings. First, the results offer strong evidence of parallel trends in the pre–DACA period, which is a key identifying assumption in difference-in-difference estimation. Namely, prior to 2012, the gap between the dropout and full-time enrollment rates of documented and undocumented students was fairly constant over time. Interestingly, dropout rates for undocumented students were lower than for their peers with legal status at both community and four-year colleges. These results are consistent with past studies showing that undocumented students are more positively selected in terms of ability and motivation than their peers with legal status (Conger and Chellman 2013). Second, coinciding with the rollout of DACA, we observe a sharp increase in the dropout rates of undocumented students at four-year colleges in year 2013 (top four graphs of Fig. 2). The increase in dropout rates is noticeable only for those students enrolled in college for up to three years but not for students close to graduation. Third, the dropout rates for undocumented students in community colleges do not display any changes around the adoption of DACA.

We now turn to full-time enrollment rates, plotted in Fig. 3. The figures in the top four graphs provide no indication of sharp changes in full-time status for undocumented students in four-year colleges around the adoption of DACA. In contrast, the bottom four graphs suggest a noticeable drop in full-time enrollment for undocumented students who have been enrolled for three years or less in community colleges.

Taken together, Figs. 2 and 3 suggest that DACA induced undocumented students in four-year colleges to drop out of school, while leading to a reduction in course load for undocumented students in community colleges. In both cases, these findings are suggestive of an attempt by these students to take advantage of the improved earnings opportunities opened up by the DACA work permits. It appears that students in community colleges—who were likely to be working already—exhibited an intensive-margin response, simply increasing their work hours without dropping out of school. In contrast, students enrolled in four-year colleges may have faced a sharper trade-off and responded by dropping out of college altogether.

Albeit strongly suggestive, the results shown in the previous figures may be driven by compositional changes. To provide a more formal analysis that accounts for individual heterogeneity and to test for statistical significance, we estimate a regression models that includes individual fixed effects and dummy variables for calendar year and years since enrollment. We begin with a more flexible version of the difference-in-difference model that allows for time-varying gaps in the outcome variable between documented and undocumented students (see Eq. (2)). The resulting point estimates and corresponding 95 % confidence intervals are plotted in Fig. 4. Each point in the graph can be interpreted as the adjusted mean gap in outcomes between undocumented and documented students. Full regression results are presented in upcoming Table 4.

Figure 4 offers two noteworthy findings. First, it provides additional evidence in support of the parallel trends assumption. We fail to reject the null of a zero-adjusted gap in all pre–DACA years in the four graphs of the figure. Second, the results confirm the descriptive results presented earlier in Figs. 2 and 3. The top two graphs clearly show that DACA significantly increases dropout rates among undocumented students in four-year colleges but has no significant effects on the decision to attend college full-time. In contrast, we find the opposite effect at community colleges. As illustrated in the bottom two graphs, DACA reduces full-time enrollment at community colleges but does not seem to induce dropping out of college.

### Effect of DACA on Educational Outcomes

Table 2 offers a simple quantification of the effects of DACA by reporting difference-in-difference estimates of the model specified in Eq. (1), referring to the decision to drop out (panel 1) and to attend college full-time (panel 2). For each outcome, we provide estimates separately by type of college. Column 1 presents estimates for the sample pooling community and four-year colleges, indicating a 3.7 percentage point increase in the dropout probability and a 2.8 percentage point reduction in the probability of full-time status. However, these estimates mask important composition effects. When estimating the models separately on the samples for community and four-year college students, we find that DACA increased the dropout rates of undocumented students in four-year colleges by 7.3 percentage points but had no effect on the dropout rates of undocumented students enrolled in community colleges. In contrast, DACA lowered the probability of full-time status by 5.5 percentage points for undocumented students at community colleges, while having no effect on the full-time status of undocumented students in four-year colleges. These findings underscore the importance of distinguishing between the effects of DACA by type of college, which probably reflects differences in students’ capacity to balance work and school in four-year and community colleges.

As noted earlier, our difference-in-difference estimates should be interpreted as intent-to-treat effects, which means that our estimates likely underestimate the true effects of actually receiving DACA status because only about one-half of those eligible actually applied for DACA in our sample period. To test this claim, we conduct an additional analysis that considers variation across groups with higher DACA application rates (undocumented students from Latin America) and lower rates (undocumented students from Asia). We anticipate that the effects of DACA will be stronger among Latino undocumented students than among Asian undocumented students because of variation in application rates.

Table 3 shows that undocumented students from regions that have the highest application rates primarily drive the estimated effects of DACA. For example, when looking at the effects on dropout rates at four-year colleges, we see that the three-way statistical interaction with Latino students is not significant, but the three-way statistical interaction with Asian students is significant and negative. Thus, undocumented students from Latin America drive the overall increase in dropout rates due to DACA. The effects are substantially smaller for Asians, presumably because relatively fewer undocumented students apply for DACA.

We see the same pattern for full-time enrollment rates in panel 2 of Table 3. At community colleges, undocumented students from Latin America drive reductions in full-time enrollment due to DACA. For Asian students at community colleges, DACA may actually increase full-time enrollment. At four-year colleges, DACA reduces full-time enrollment among undocumented Latino students but has no effect on the enrollment of Asian students. Overall, the results for full-time enrollment status also indicate that DACA has a larger effect on groups who are the most likely to be DACA recipients than groups who are the least likely to be DACA recipients.

The results from Table 4 indicate the effect of DACA vary by application rates. Thus, our estimates from Table 3 can be interpreted as intent-to-treat effects, which underestimate the average treatment effects because only a portion of undocumented students are DACA recipients. Average treatment effects on the treated can be derived from scaling the intent-to- treat effects with the compliance ratio (i.e., the proportion of undocumented college students who are DACA recipients). Unfortunately, we do not know what portion of undocumented college students are DACA recipients. What we know is that the compliance ratio for the general DACA–eligible population residing in our study state is 50 %. Therefore, assuming that all undocumented college students are DACA–eligible, the average treatment is a 14.6 percentage point (= .73 / .5) increase in the probability of dropping out at a four-year college, and an 11 percentage point (= .055 / .5) reduction in the probability of full-time enrollment at a community college. Overall, our findings indicate that DACA increases dropout rates within the ranges of 7.3 to 14.6 percentage points at four-year colleges and reduces full-time enrollment within the range of 5.5 to 11 percentage points at community colleges.

These estimates are quantitatively large yet consistent with the findings reported in previous studies showing that DACA increases the employment opportunities of undocumented immigrants. For example, Pope (2016) and Amuedo-Dorantes and Antman (2017) reported intent-to-treat effects on the probability of employment of noncitizen high school graduates in the range of 5 to 10 percentage points. Both qualitatively and quantitatively, these estimates reinforce our interpretation that DACA has led undocumented students to reduce their course work, partially or fully, to take advantage of the enhanced employment opportunities.

## Discussion

Our findings suggest that as a temporary work permit program, DACA incentivizes work over investments in human capital but that the effect of DACA on undocumented students’ decisions to remain in school depends on how easily colleges accommodate working students. We find that dropout rates for DACA recipients at four-year colleges increased between 7.3 to 14.6 percentage points, while leaving largely unaffected the full-time status of those who remain in school. In contrast, at community colleges where the vast majority of students work while attending school, DACA recipients reduced full-time enrollment by 5.5 to 11.0 percentage points, with no measurable effect on their dropout rates.

These patterns indicate that undocumented students at four-year colleges must make binary choices between attending school full-time or dropping out of school to work. Community colleges, on the other hand, are designed to help students balance schooling with work. For example, they offer more evening and weekend classes than four-year colleges. Course credits at community colleges are also significantly less expensive than at four-year colleges, allowing undocumented students more flexibility in course load to accommodate variable work schedules. As a result, DACA recipients at community colleges can simply reduce their course work to accommodate increased work hours.

This study is not without caveats. Although our findings indicate that DACA may cause some undocumented students to leave school to work, we cannot rule out the possibility that it may have allowed many others to remain in school by enabling them to work to finance schooling expenses. In fact, although the effect on dropout is sizable, the vast majority of undocumented students remain in school under DACA. Additionally, our study focuses on undocumented students who were already enrolled in college at the time of DACA’s passage. Thus, we cannot know how DACA may have affected college enrollment rates. The extension of legal work opportunities increased the returns to education and may have motivated some undocumented youth who would have otherwise never have pursed college to attend college under DACA. Our study cannot rule out this possibility.

Another limitation is that we cannot assess the long-term impact of DACA on educational attainment. Our results clearly show that DACA has led undocumented students to leave school or reduce course loads in order to take advantage of the (renewable) two-year work permit. What we are currently unable to assess is whether those students will eventually return to school and complete their degrees. One possibility is that students take advantage of temporary work permits by leaving school to work temporarily and return to school after their work permit expires. Because data collection is ongoing, as new data become available, we will be able to observe whether students who drop out of college eventually reenter.

Finally, our study is based on an analysis of administrative data collected from a large public university located in a large metropolitan area. The generalizability of our findings to other regions in the country may be limited as a result. However, our estimates closely align with estimates derived from nationally representative surveys (Amuedo-Dorantes and Antman 2017; Pope 2016), giving us greater confidence in the validity of our results.

Overall, the research to date demonstrates that “dreamers” are better off with DACA than without it. DACA improved undocumented immigrants’ mental health, economic well-being, and social integration (Amuedo-Dorantes and Antman 2017; Gonzales et al. 2014; Hainmueller et al. 2017; Patler and Pirtle 2018; Venkataramani et al. 2017). Yet, DACA offers only temporary protections. Our results suggest that as a temporary work permit program, DACA led many undocumented youth to myopically reduce educational investments because two-year work permits can afford only short-term planning horizons. New research demonstrates that extending pathways to legal residency through a policy like the DREAM Act would have larger positive effects on the educational attainment and wages of undocumented youth compared with temporary work permit policies, such as DACA (Ortega et al. 2018). Taken together, these findings suggest that immigration policies that offer students’ longer planning horizons and greater certainty for the future—such as pathways to permanent residency and citizenship—would introduce fewer distortions to educational investments and improve the long-term well-being of undocumented youth.

## Acknowledgments

This research was supported by the William T. Grant Foundation, the John D. and Catherine T. MacArthur Foundation and the Stanford Center on Poverty and Inequality. Thanks to Holly Reed; Sofya Aptekar; Thomas DiPrete; participants at the CUNY Institute for Demographic Research Center Seminar Series, Center for the Study of Wealth, and Inequality Seminar Series at Columbia University; and anonymous reviewers for insights on prior versions of this article. Keitaro Okura provided research assistance.

## Notes

1

Approximately 49 % of undocumented immigrants aged 18–24 attend college (Passel and Cohn 2009). By contrast, 70 % of the general population of high school graduates attains at least some college education (Bureau of Labor Statistics 2017).

2

We recognize that the compliance ratio among undocumented college students might be higher than 50 % because they are a selected group who are likely to be more motivated, more academically proficient, and of higher socioeconomic status compared with the general population of DACA–eligible youth. They may be more likely to apply and receive DACA. However, the compliance rate for DACA–eligible college students is unknown.

## References

Abrego, L. J. (
2006
).
“I can’t go to college because I don’t have papers”: Incorporation patterns of Latino undocumented youth
.
Latino Studies
,
4
,
212
231
. 10.1057/palgrave.lst.8600200
Amuedo-Dorantes, C., & Antman, F. (
2017
).
Schooling and labor market effects of temporary authorization: Evidence from DACA
.
Journal of Population Economics
,
3
,
339
373
.
Bailey, T. R., Jenkins, D., & Jaggars, S. S. (
2015
).
Redesigning America’s community colleges
.
Cambridge, MA
:
Harvard University Press
.
Batalova, J., Hooker, S., Capps, R., & Bachmeier, J. D. (
2014
).
DACA at the two-year mark: A national and state profile of youth eligible and applying for deferred action
(Report).
Washington, DC
:
Migration Policy Institute
.
Bureau of Labor Statistics
. (
2017
).
College enrollment and work activity of 2016 high school graduates
(Economic news release).
Washington, DC
:
U.S. Department of Labor
Capps, R., Fix, M., & Zong, J. (
2017
).
The education and work profiles of the DACA Population
(MPI issue brief).
Washington, DC
:
Migration Policy Institute
.
Conger, D., & Chellman, C. C. (
2013
).
Undocumented college students in the United States: In-state tuition not enough to ensure four-year degree completion
.
Education Finance and Policy
,
8
,
364
377
. 10.1162/EDFP_a_00101
Conger, D., & Turner, L. J. (
2017
).
The effect of price shocks on undocumented students’ college attainment and completion
.
Journal of Public Economics
,
148
,
92
114
.
Contreras, F. (
2009
).
Sin papeles y rompiendo barreras [Without papers and breaking barriers]: Latino students and the challenges of persisting in college
.
Harvard Educational Review
,
79
,
610
632
. 10.17763/haer.79.4.02671846902gl33w
Department of Homeland Security, U.S. Citizenship and Immigration Services. (
2013
)
Deferred action for childhood arrivals
DiNapoli, T. P., & Bleiwas, K. B. (
2014
).
Update: The NY DREAM Act
(Office of the State Comptroller, New York City Public Information Office, Report 11-2014). Retrieved from https://www.osc.state.ny.us/osdc/rpt11-2014.pdf
Donato, K. M., Wakabayashi, C., Hakimzadeh, S., & Armenta, A. (
2008
).
Shifts in the employment conditions of Mexican migrant men and women: The effect of U.S. immigration policy
.
Work and Occupations
,
35
,
462
495
. 10.1177/0730888408322859
Edwards, R., & Ortega, F. (
2016
).
The economic impacts of removing unauthorized immigrant workers: An industry- and state-level analysis
(Report).
Washington, DC
:
Center for American Progress
Enriquez, L. E. (
2017
).
A “master status” or the “final straw”? Assessing the role of immigration status in Latino undocumented youths’ pathways out of school
.
Journal of Ethnic and Migration Studies
,
43
,
1526
1543
. 10.1080/1369183X.2016.1235483
Flores, S. M. (
2010
).
State Dream Acts: The effect of in-state resident tuition policies and undocumented Latino students
.
Review of Higher Education
,
33
,
239
283
. 10.1353/rhe.0.0134
Garcia, L., & Tierney, W. (
2011
).
Undocumented immigrants in higher education: A preliminary analysis
.
Teachers College Record
,
113
,
2739
2776
.
Gleeson, S., & Gonzales, R. G. (
2012
).
When do papers matter? An institutional analysis of undocumented life in the United States
.
International Migration
,
50
(
4
),
1
19
. 10.1111/j.1468-2435.2011.00726.x
Gonzales, R. G. (
2011
).
Learning to be illegal: Undocumented youth and shifting legal contexts in the transition to adulthood
.
American Sociological Review
,
76
,
602
619
. 10.1177/0003122411411901
Gonzales, R. G., Terriquez, V., & Ruszczyk, S. P. (
2014
).
Becoming DACAmented: Assessing the short-term benefits of Deferred Action for Childhood Arrivals (DACA)
.
American Behavioral Scientist
,
58
,
1852
1872
. 10.1177/0002764214550288
Gonzalez-Barrera, A., & Krogstad, J. (
2016
).
U.S. immigrant deportations declined in 2014, but remain near record high
(Fact Tank report).
Washington, DC
:
Pew Research Center
Greenman, E., & Hall, M. (
2013
).
Legal status and educational transitions for Mexican and Central American immigrant youth
.
Social Forces
,
91
,
1475
1498
. 10.1093/sf/sot040
Hainmueller, J., Lawrence, D., Martén, L., Black, B., Figueroa, L., Hotard, M., & . . . Laitin, D. D. (
2017
).
Protecting unauthorized immigrant mothers improves their children’s mental health
.
Science
,
357
,
1041
1044
.
Hall, M., Greenman, E., & Farkas, G. (
2010
).
Legal status and wage disparities for Mexican immigrants
.
Social Forces
,
89
,
491
513
.
Hipsman, F., Gomez-Aguinaga, B., & Capps, R. (
2016
).
DACA at four: Participation in the deferred action program and impacts on recipients
(MPI issue brief).
Washington, DC
:
Migration Policy Institute
.
Kaushal, N. (
2008
).
In-state tuition for the undocumented: Education effects on Mexican young adults
.
Journal of Policy Analysis and Management
,
27
,
771
792
. 10.1002/pam.20366
Menjívar, C. (
2006
).
Liminal legality: Salvadoran and Guatemalan immigrants’ lives in the United States
.
American Journal of Sociology
,
111
,
999
1037
. 10.1086/499509
Orrenius, P. M., & Zavodny, M. (
2015
).
Does immigration affect whether U.S. natives major in science and engineering?
.
Journal of Labor Economics
,
33
,
79
108
. 10.1086/676660
Ortega, F., Edwards, R., & Hsin, A. (
2018
).
The economic effects of providing legal status to DREAMers
(IZA Discussion Paper No. 11281).
Bonn, Germany
:
Institute of Labor Economics
.
Passel, J. S., & Cohn, D. (
2009
).
A portrait of unauthorized immigrants in the United States
(Pew Hispanic Center report).
Washington, DC
:
Pew Research Center
Passel. J. S., & Cohn, D. (
2016
).
Unauthorized immigrant population stable for half a decade
(Fact Tank report).
Washington, DC
:
Pew Research Center
Patler, C., & Pirtle, W. L. (
2017
).
From undocumented to lawfully present: Do changes to legal status impact psychological wellbeing among Latino immigrant young adults?
Social Science & Medicine
,
199
,
39
48
.
Perez, W., & Cortes, R. D. (
2011
).
Undocumented Latino college students: Their socioemotional and academic experiences
.
El Paso, TX
:
LFB Scholarly Publishing
.
Pope, N. G. (
2016
).
The effects of DACAmentation: The impact of deferred action for childhood arrivals on unauthorized immigrants
.
Journal of Public Economics
,
143
,
98
114
. 10.1016/j.jpubeco.2016.08.014
Potochnick, S. (
2014
).
How states can reduce the dropout rate for undocumented immigrant youth: The effects of in-state resident tuition policies
.
Social Science Research
,
45
,
18
32
. 10.1016/j.ssresearch.2013.12.009
Suarez-Orozco, C., Katsiaficas, D., Birchall, O., Alcantar, C. M., Hernandez, E., Garcia, Y., & . . . Teranishi, R. T. (
2015
).
Undocumented undergraduates on college campuses: Understanding their challenges and assets and what it takes to make an undocufriendly campus
.
Harvard Educational Review
,
85
,
427
463
.
Terriquez, V. (
2014
).
Dreams delayed: Barriers to degree completion among undocumented community college students
.
Journal of Ethnic and Migration Studies
,
41
,
1
22
.
Van Hook, J., Bachmeier, J. D., Coffman, D. L., & Harel, O. (
2015
).
Can we spin straw into gold? An evaluation of immigrant legal status imputation approaches
.
Demography
,
52
,
329
354
. 10.1007/s13524-014-0358-x
Vaquera, E., Aranda, E., & Sousa-Rodriguez, I. (
2017
).
Emotional challenges of undocumented young adults: Ontological security, emotional capital, and well-being
.
Social Problems
,
64
,
298
314
. 10.1093/socpro/spx010
Venkataramani, A. S., Shah, S. J., O’Brien, R., Kawachi, I., & Tsai, A. C. (
2017
).
Health consequences of the US Deferred Action for Childhood Arrivals (DACA) immigration programme: A quasi-experimental study
.
Lancet Public Health
,
2
,
e175
e181
. 10.1016/S2468-2667(17)30047-6
Warren, R., & Warren, J. R. (
2013
).
Unauthorized immigration to the United States: Annual estimates and components of change, by state, 1990 to 2010
.
International Migration Review
,
47
,
296
329
. 10.1111/imre.12022
Wong, T. K., Richter, K. K., Rodriguez, I., & Wolgin, P. E. (
2015
).
Results from a nationwide survey of DACA recipients illustrate the program’s impact
(Report).
Washington, DC
:
Center for American Progress