Peer-Reviewed Publications

Forthcoming, Journal of Policy Analysis and Management

Online Appendix: Sub-Discipline Level Estimates   

Working Papers

Abstract: Community colleges are an important postsecondary sector, yet there are few causal estimates of their effect on student earnings. I estimate the labor market returns to a particularly important degree, the Associate’s Degree in Nursing, by leveraging random variation from admissions lotteries. I use student-level academic and earnings records across two decades for community college students in California. I find that enrolling in the program increases earnings by 55 percent. I also document substantial heterogeneity in these effects across nursing programs. I estimate that the benefits of expanding a program far outweigh the costs and are close to revenue neutral.

The Effect of Local Labor Market Downturns on Postsecondary Enrollment and Program Choice, with Andrew Foote (Revisions requested)
Abstract: We examine the response of postsecondary enrollment to local labor market downturns. We use institution-level data on enrollment and credential completion for most two-year colleges across the country over the past two decades. We leverage local area counts of mass layoffs as a measure of discrete and acute labor market shocks. We find that for every 100 workers involved in a mass layoff, first-time fall enrollment increases by 3 students within three years. We also find heterogenous responses by field of study: the largest responses to mass layoff events are in fields with higher expected labor market returns. 

This paper analyzes whether postsecondary training programs have kept up with shifts in the occupational structure of the labor market over the past decades. I show long-term trends in the distribution of degrees and certificates across occupation groupings in the nation’s largest community college system. Using an instrumental variables approach I then estimate that an occupation’s share of community college completions grows 0.47 percentage points for very percentage point increase in its share of employment. However, I show that this relationship is primarily due to increases in student demand rather than to colleges expanding capacity.

Selection into Identification in Fixed Effects Models, with Application to Head Start, with Doug Miller and Na'ama Shenhav 
Abstract: Many papers use fixed effects (FE) to identify causal impacts of an intervention. A leading example of this is family fixed effects (FFE), which compares the outcomes of siblings with different exposure to treatment. In this paper we show that when the treatment is binary, this design induces non-random selection of families into the identifying sample, which we term selection into identification. We document across multiple settings that this causes the identifying sample to be overly representative of larger families and that, in the presence of heterogeneous treatment effects, the estimated effect is biased relative to the average marginal treatment effect (ATE) for siblings. Additionally, we prove that for binary outcomes the linear probability model is unbiased, and provide a novel method for recovering ATEs from conditional logit specifications. We apply these insights to examine the long-term effects of Head Start in the PSID utilizing an expanded dataset of outcomes up to age 40 to update Garces, Thomas, and Currie (2002). Using FFE, we find that participation in Head Start increases the likelihood of completing some college by 12 percentage points. We estimate this parameter to be 50% larger than the ATE for siblings. We find no evidence of strong positive effects of Head Start on overall economic or physical well-being. We conclude that alternative research design strategies should be pursued to gain representative evidence about the long term impact of Head Start.

Abstract: In 2013, Colorado became the 15th state to pass a law effectively granting in-state tuition to undocumented students while keeping with a federal ban on specifically targeting them for financial aid. In this paper, we evaluate the effects of Colorado's Advancing Students for a Stronger Tomorrow (ASSET) legislation on the college application, enrollment, persistence, and credit hours of Colorado undergraduates using a differences-in-differences methodology. In the absence of information on legal immigration status, we construct a plausible treatment group from Hispanic non-resident non-citizens who attended high school in Colorado. We find evidence of an increase in the enrollment of likely undocumented students, and an increase in the the number of enrolled credits of new students. We do not find evidence of policy-induced changes in the persistence, full-time status or credit hours completed of continuing students, however.
Majority Builds: Direct Democracy and the Allocation of School Facilities Spendingwith Ross Milton
Abstract: The need to put local budgeting questions to a referendum vote is generally thought to reduce spending, but little is known about whether this comes from bureaucrats changing their behavior or voters rejecting their proposals. We leverage a policy change in California that lowered the share of votes required to pass a school capital improvement bond. We show that this loosening of the constraint on districts lowered the average vote share for school bonds by 3 percentage points. We find that part of the explanation for this effect comes from school districts increasing the amount of proposed per-pupil spending and thus lowering their support, especially for bonds closest to the old threshold. We also use detailed information on school construction projects to show how intra-district spending changed in response to the lowered vote threshold. 

Work In Progress

The Effect of Lottery-Based Admissions on Cohort Composition and Success: Evidence from California
Community College Capacity Constraints and Malleable Factors, with Michal Kurlaender and Ann Stevens
Does Specific Labor Market Information Affect Major Choicewith Matt Naven
School Infrastructure Spending and Academic Outcomes, with Ross Milton