Members of the NBER's Program on Economics of Education met on April 13 in Cambridge. Program Director Caroline M. Hoxby of Stanford University organized the meeting. These researchers' papers were presented and discussed:
Stephanie Cellini, George Washington University and NBER; Rajeev Darolia, University of Kentucky; and Lesley J. Turner, University of Maryland and NBER
Where Do Students Go when For-Profit Colleges Lose Federal Aid?
Recent policy debates have focused on whether restricting for-profit institutions' access to federal student financial aid could reduce student loan defaults without restricting prospective students' access to higher education. Cellini, Darolia, and Turner examine the effects of similar restrictions imposed on over 1,200 for-profit colleges in the 1990s. Using variation in the timing and magnitude of sanctions linked to student loan default rates, the researchers estimate the impact of the loss of federal aid on the enrollment of Pell Grant recipients in sanctioned institutions and their local unsanctioned competitors. On average, sanctioned for-profit colleges experience a 40 percent decrease in annual enrollment in the five years following sanction receipt. Enrollment losses due to for-profit sanctions are offset by enrollment increases within local community colleges. For-profit sanctions also produce negative enrollment spillovers on unsanctioned for-profit competitors, and the researchers provide evidence that these effects are likely due to improved information about local higher education options and/or reputational spillovers to for-profit institutions offering similar programs. Given these offsetting effects, Cellini, Darolia, and Turner estimate that within the average county, the public sector absorbs 40 to 60 percent of the total enrollment decline generated by an additional for-profit sanction. Overall, market enrollment declines by just 3 percent. Finally, the researchers provide suggestive evidence that students induced to enroll in community colleges following a for-profit competitor's sanction are less likely to default on their federal loans.
Peter S. Bergman and Magdalena Bennett Colomer, Columbia University
Better Together? Social Networks in Truancy and the Targeting of Treatment
Truancy predicts many risky behaviors and adverse outcomes. Bergman and Colomer use detailed administrative data to construct social networks based on students who miss class together. They simulate these networks to show that certain students systematically coordinate their absences in the observed data. Leveraging a parent-information intervention on student absences, the researchers find spillover effects from treated students onto peers in their network. Excluding these effects understates the intervention's cost effectiveness by 19%. They then show there is potential to use these networks to implement costly interventions more efficiently. The researchers develop an algorithm that incorporates spillovers and treatment-effect heterogeneity identified by machine-learning techniques to target interventions more efficiently given a budget constraint.
Rebecca A. Johnson, Princeton University, and Dalton Conley, Princeton University and NBER
Tags and a Leaky Pipeline in School Districts' Allocations to Students
A variety of social policies use "tags" -- labels that mark a class of recipients as deserving of extra resources -- to direct resources towards individuals. In the case of children and education, important tags include those indicating the child qualifies for Free or Reduced Price Lunch (FRPL), English Language Learners (ELL), or students with disabilities (IEPs), and more. A wealth of research studies how tags structure between district resource allocations. Researchers investigate how state legislatures, often under pressure from school finance litigation, change the weights they attach to certain tags like poverty status to distribute state budget allocations between districts, researchers then study how these changes affect student outcomes. Much less research studies how tags structure within-district resource allocations. Johnson and Conley ask: is there a "leaky pipeline" where resources intended for students with certain tags "leak out" to students with a tag accompanied with stronger legal tools to influence allocations? The analysis proceeds in two steps. After showing how the tag of disability status (IEP) is associated with stronger legal tools for parents to influence district resource allocations than other tags, the researchers use a unique exogenous increase to funding allocated to students with disabilities -- the American Recovery and Reinvestment Act (ARRA) causing a one-time doubling of federal allocation to districts to cover the costs of disability services -- to show that the legal tools attached to the disability tag are primarily used to influence district resource allocations. After establishing that the disability tag is associated with stronger legal tools than other tags, and the legal tools are used primarily to influence district resource allocations, they investigate the consequences for between-student allocations. They use a regression discontinuity design (RDD) caused by a threshold in a new California grant where districts above a threshold for grant receipt receive resources intended for students with tags other than disability status. The researchers find that despite this targeting of resources towards students with other tags, some of the resources appear to "leak out" to students with the disability tag as shown by decreases in parent complaints over services. These reduced complaints indicate less contention over resources and the money "leaking out" from its intended recipients (students with other tags like ELL) to special education students, a tag with stronger legal tools.
Karthik Muralidharan, University of California, San Diego and NBER, and Abhijeet Singh, Stockholm School of Economics
Understanding the Flailing State: Experimental Evidence from a Large-Scale School Governance Improvement Program in India
Evan Riehl, Cornell University
Fairness in College Admission Exams: From Test Score Gaps to Earnings Inequality
Riehl asks whether socioeconomic gaps in college admission test scores also translate into greater earnings inequality. He shows that the link between test scores and labor market outcomes depends on how well the exam measures students' potential earnings returns to college quality. He estimates the distribution of these returns by exploiting a natural experiment from a redesign of the Colombian national college admission exam. He find substantial heterogeneity in the returns to college quality including negative returns for some low-income students. This suggests that raising low-income students' test scores may reduce their future earnings if the exam cannot identify which students would benefit from attending a better college.
Meltem Daysal, University of Southern Denmark; Todd Elder, University of Michigan; Judith K. Hellerstein, University of Maryland and NBER; Scott A. Imberman, Michigan State University and NBER; and Chiara Orsini, London School of Economics
Parental Human Capital Traits and Autism Spectrum Disorder in Children
Daysal, Elder, Hellerstein, Imberman, and Orsini examine whether and to what extent parental characteristics are related to autism spectrum disorder (ASD) diagnoses in children. A key insight they use is that parents' underlying traits are reflected in their choices of occupation and educational field, opening new avenues to understand intergenerational transmission of developmental disorders. To this end, the researchers combine data from multiple administrative datasets from Danish registers and use saturated regression models that include information on parental occupations, education, family relationships and ASD diagnoses of children born between 1995 and 2010. The study makes two substantive contributions. They add to existing evidence on the causal relationship between characteristics of parents and ASD prevalence in children, and provide the first large-scale empirical assessment of medical theories linking assortative mating to ASD prevalence. The researchers follow Baron Cohen et al. (1997) by characterizing educational and occupational choices as reflecting a "systemizing" trait. They show that systemizing of fathers, but not mothers, is related to ASD diagnosis rates, even conditional on parental age at birth, education level, income, geographic location, and other controls. The researchers also find similar results when they examine the link between systemizing of paternal and maternal grandfathers and ASD diagnoses of grandchildren. Given that this relationship appears to operate exclusively through the paternal line, they also find no evidence that assortative mating on the basis of systemizing is linked to ASD prevalence.