Estimating the Perceived Returns to College, Under Review
Abstract: The primary determinant of an individual's college attendance is their perceived lifetime return to college. I infer agents' perceived returns by exploiting the dollar-for-dollar relationship between perceived returns and tuition costs in a binary choice model of college attendance. This method has the advantage of estimating perceived returns in terms of compensating variation without assuming rational expectations on actual returns. Estimating the model using both maximum likelihood and moment inequalities, I find that the scale of the distribution of perceived returns is an order of magnitude lower than past work has found when assuming rational expectations on income returns. The low variance I find in perceived returns implies high responses to financial aid. I predict a 2.6 percentage point increase in college attendance from a $1,000 universal annual tuition subsidy, which is consistent with quasi-experimental estimates of the effects of tuition assistance on college attendance. Because I estimate the complete distribution of perceived returns, my results can be used to predict heterogeneous effects of counterfactual financial aid policies.
Peer Effects in University Housing: Evidence from Fuzzy Central Assignment, with Chao Fu, Jesse Gregory, and Victoria Prowse
Abstract: We estimate effects of roommates, neighbors, and dorms on academic outcomes at a large public four-year university. To address selection, we instrument for realized room assignments with simulated offers generated by a room assignment mechanism, while controlling for expected offers implied by the same mechanism. Our candidate assignment mechanisms fail to perfectly replicate the assignment offers made by the central planner, so we select the mechanism that best replicates assignments via a data-driven model selection procedure. We find that living in a fully gender-integrated dorm increases 4-year graduation by 10 percentage points for men, with no significant effects on women, while finding consistent (statistically insignificant) results for exposure to female neighbors within dorms. At the roommate level, we find that STEM roommates have significantly smaller negative effects on four-year graduation for other STEM students than they do for non-STEM students. Our findings suggest that universities with 70% gender-integrated housing (such as the university we study) could increase their male 4-year graduation rates by 3 percentage points by implementing full gender-integration in all dorms, with no offsetting negative effects on women.
Abstract: Many treatment variables used in empirical applications nest multiple unobserved versions of a treatment. I show that instrumental variable (IV) estimands for the effect of a composite treatment are IV-specific weighted averages of effects of unobserved component treatments. Differences between IVs in unobserved component compliance produce differences in IV estimands even without treatment effect heterogeneity. I describe a monotonicity condition under which IV estimands are positively-weighted averages of unobserved component treatment effects. Next, I develop a method that allows instruments that violate monotonicity to contribute to estimation of treatment effects by allowing them to place nonconvex, outcome-invariant weights on unobserved treatments across multiple outcomes. Finally, I apply the method to estimate returns to college, finding wage returns that range from 7% to 30% over the life cycle. My findings emphasize the importance of leveraging instrumental variables that do not shift individuals between versions of treatment, as well as the importance of policies that encourage students to attend “high-return college” in addition to those that encourage “high-return students” to attend college.
The Effects of Chicago’s Teacher Walkouts on Juvenile Crime, with Abigail Banan, Mary Kate Batistich, Jillian Carr, and Kendall Kennedy
Abstract: On September 10, 2012, Chicago’s teachers began a strike that forced over 400,000 students out of school for 8 consecutive school days. This study examines how this interruption in schooling affected crime in Chicago, both during and immediately after the strike. Using a variety of methods – interrupted time series analysis, synthetic controls, and place-based intensity of treatment analysis – we find that crime rates in Chicago neither significantly nor substantially changed after the strike, suggesting incapacitation has little role in the effect of education on crime.
The Effects of Peer Diversity on Entrepreneurship: Evidence from University Residence Halls, with Jon Eckhardt and Brent Goldfarb
Abstract: We estimate the effects of diversity in gender, major, test scores, and socioeconomic status among residence hall peers on entrepreneurship by combining administrative data from University of Wisconsin-Madison with a survey regarding students’ concurrent entrepreneurship activities and interest. To address nonrandom peer assignments, we instrument for peer traits with simulated peer traits computed using a quasi-random room assignment algorithm. We find that ACT Verbal score diversity between roommates reduces entrepreneurship, with evidence supporting reduced idea generation as a potential mechanism. Finally, we leverage portfolio theory to predict entrepreneurship under counterfactual policies that alter peer diversity in multiple dimensions.
A Survey Selection Correction using Nonrandom Followup with an Application to the Gender Entrepreneurship Gap, with Jon Eckhardt and Brent Goldfarb
Code with simulated data: https://github.com/clintmharrisecon/Survey-Correction-Code
Abstract: When subjects who respond to requests for data, such as in surveys or post-treatment follow-up, are not representative of the population as a whole, inferences drawn from the data can be misleading. We show that if subjects’ accumulated requests and responses over time are recorded and organized as panel data, requests can be used as instruments to correct for nonresponse bias even if total requests are not randomized between subjects. We demonstrate our method by estimating an 18-percentage-point gender gap in entrepreneurial career intentions using a survey of undergraduates at the University of Wisconsin-Madison.
Separate Identification and Estimation of Perceived Credit Constraints and Perceived Returns to College Using Observed Attendance Decisions
Looking out for Number 1,2,...,T: Decomposing Discounting into Identity Persistence and Altruism