Essays on Access to Education

This dissertation uses survey data and administrative data to explore persistent barriers in access to education. The first chapter explores how constraints on credit supply can impact the level and distribution of higher education, including access to selective and 4-year colleges. I exploit a 2003 Texas constitutional amendment that provided plausibly exogenous variation in access to home lending markets for Texas homeowners, without affecting credit access for renters, or homeowners in other states. By comparing outcomes between groups, I show that this led Texas homeowners to send their children to more selective colleges and spend $4,500 more in tuition (net-of-aid) per line of credit. In the presence of college supply constraints, homeowners’ increased demand for institutions higher in the college selectivity hierarchy forced some renters to attend less selective colleges, and others to forgo college altogether instead of attending less selective colleges. In addition, selective colleges capture some of the private credit supply shock through price-discrimination, raising tuition and shifting aid towards remaining renters. On net, the availability of home equity financing reinforced gaps in access to higher education. These results inform our understanding of how inequality in college access is generated and transmitted from parent to child: the availability of home equity credit reinforces gaps between homeowning and renting families, and it does so through two distinct mechanisms. First, constraints in credit access are relaxed for homeowners, allowing them to ascend the college quality hierarchy. Second, due to college supply constraints, the gains to homeowners crowd out some renters from making otherwise privately optimal investments. By documenting important distributional effects on renters, this paper informs our interpretation of previous research: increases in college choice for one group may come in part at the expense of another group. The results of the first chapter also demonstrate how the more selective colleges are able to capture some of the gains from cheaper credit by price-discriminating by homeownership status. The net effects of subsidized home lending markets and federal aid policy on college access are not immediately clear: on one hand, homeowners are sending their children to better colleges, but they are paying higher net prices at these colleges than they would in the absence of the private credit supply shock. On the other hand, tuition increases for renters who remain enrolled at selective colleges are offset by increases in institutional aid, but some renters are pushed down the college quality hierarchy and displaced from college altogether. The relationship between housing markets and access to higher education is also explored in chapter three, which examines the effect of metropolitan house price shocks on college enrollment patterns across cities. This chapter begins by presenting a simple theoretical model to illustrate the mechanisms through which parental housing wealth can ease educational borrowing constraints for their children. The model highlights how house price growth can reinforce inequality in future generations: increases in the value of parental housing collateral can ease educational borrowing constraints for children, but the indivisibility in owner-occupied housing limits exposure to this externality to higher-income families. The second part of this chapter presents estimates of the relationship between house price shocks and changes in college enrollment at the MSA level. The empirical results confirm that house price growth leads to higher college enrollment rates (and vice versa for declining house prices), but these effects are concentrated in metropolitan areas with lower house price levels. There is only weak evidence that house price growth leads to increases in housing-related employment among college-aged individuals. The second chapter of this dissertation examines the effect of policies used to reclassify non-native English speakers (English Learners, or ELs) in the Oakland Unified School District into mainstream classes. Despite a heightened policy debate surrounding appropriate instructional policies for the large and growing number of non-native English speaking students nationwide, policymakers have limited causal research available on the effects of reclassification policy. This paper addresses some of the gaps in the empirical literature on reclassification by exploiting exogenous variation in the probability of reclassification introduced by the multiple criteria students must meet to be eligible for reclassification. It begins with a conventional regression discontinuity (RD) design that estimates the short and long-term effects of reclassification for non-native English speakers who have met all reclassification criteria except potentially one. These students exhibit large jumps in the probability of reclassification around relevant test score cutoffs. The RD estimates suggest that reclassification has very limited effects on students at the margin, but that the timing of reclassification may indeed matter, though not necessarily through effects on student learning. There is some suggestive evidence of increases in SAT-taking and four-year college enrollment, but limited statistical power prevents definitive conclusions regarding small changes in long-term outcomes. Motivated by the limitations inherent to RD designs that estimate treatment effects for students at relevant cutoffs, chapter two also presents an extension to the conventional RD design in order to draw conclusions about the effects of reclassification for students whose reclassification scores place them well above the cutoff. The framework we present exploits the fact that some students who meet the first cutoff will remain untreated due to being below the cutoff for a second running variable. These untreated students provide additional information on the relationship between reclassification test scores and outcomes, which can then be used to inform our expectation of counterfactual outcomes for reclassified students in the absence of reclassification. More specifically, we can use this information for EL students who were not reclassified to estimate outcomes for reclassified students in the absence of reclassification under a straightforward separability assumption that can be examined in the data. We show this assumption holds in the data, before estimating the relationship between outcomes and reclassification test scores for non-reclassified students and then using these estimates to predict outcomes for reclassified students in the absence of reclassification. Estimates of the effect of reclassification for any score above the cutoff can then be obtained by comparing the prediction to the observed value for reclassified students. Beyond the immediate application to reclassification policy in Oakland Unified, the framework we introduce for estimating treatment effects above the cutoff can apply to any setting where treatment status is based on multiple criteria. The resulting estimates imply that for all students in elementary school who were above the cutoffs, the average effect of reclassification into mainstream classes on English language arts (ELA) scores is a 0.182 standard deviation increase in their ELA score in the following year. These results imply that the CST ELA cutoff should not be raised for students in grades 3 through 5, as benefits accrue to students above the current cutoffs and in mainstream classrooms. Without knowing how students below the current cutoff are impacted by reclassification, however, we cannot say whether policymakers should consider lowering the criteria