Working papers
Student Loans and College Majors: The Role of Repayment Plan Structure [Latest Version] [Appendix]
This paper highlights the role loan repayment plan structure has in students’ human capital investments. I link academic records from a major public university to credit records to assess the empirical evidence for shifting major selection. After an expansion in Income Driven Repayment (IDR) options increased generosity and use, borrowers are more likely to select majors with worse initial labor market outcomes but higher wage growth, consistent with theoretical predictions. These results are robust to specifications that account for nonrandom selection into borrowing status as well as compositional shifts in borrowers over time. This sheds light on how changes in student loan repayment plans affected major selection and, given the new SAVE plan, how students may respond in future human capital investments.
Revisiting Difference-in-Differences with Compositional Changes using Propensity Score Reweighting: a comment on Hong (2013) [Latest Version]
This paper revisits the measurement of Napster's effect on record sales. Recognizing that there are compositional changes in Internet users, and that such compositional shifts render the estimates of the standard reweighting approach to difference-in-differences inconsistent, I extend the reweighting approach to explicitly account for compositional changes in treatment and comparison groups. This extension of the reweighting approach allows for identification of causal parameters in settings with compositional shifts in groups over time. I compare my results to existing results that use multi-dimensional matching estimators. In contrast to multidimensional matching, my proposed procedure is computationally simple, requiring only fitting a binary choice model and estimating weighted means.
College Finance, Repayment Regimes, and Job Search [Latest Version]
I examine a directed search model with educational investment and financing decisions. Repayment plans alter search behavior: under Fixed Repayment (FR), higher payments push high ability people into higher-wage submarkets to ease the "repayment burden" whereas lower ability people move to lower-wage jobs to avoid the "default penalty." These heterogeneous responses mirror two strands of current literature within one model. Under a simple Income Driven Repayment (IDR) plan, individuals search in higher wage submarkets relative to the no-loan case, with the difference depending plan parameters, including time remaining in repayment, the income disregard, and limits on maximum repayments. As existing literature has argued, under a graduate tax optimal search behavior coincides with the no-loan case. However, here this is true if borrowers have constant relative risk aversion and may not endogenously save. If there are any features more common to IDR plans introduced or borrowers may save, this no longer holds generally. I outline conditions on efficient loan offerings, which mitigate differences in optimal search behavior between borrowers and non-borrowers and encourage optimal college attendance.
Competing Risks Models, Assumptions, and Implications: Examining the Opioid Epidemic in the United States
(with John Bound, Timothy Waidmann, and Arline T. Geronimus)
From the turn of the 21st century up to the beginning of the global COVID-19 pandemic, life expectancy trends among U.S. adults stagnated or fell. Dramatic increases in drug and alcohol-related deaths, especially deaths caused by opioid overdose, have been identified as a key driver of those trends. Estimating the total contribution of rising opioid use to population mortality using standard demographic techniques and vital statistics data includes the assumption of independence between competing risks. This assumption is made more for lack of practical alternatives than out of conviction. Alternative methods have been developed that rely on spatiotemporal correlations among causes of death to identify the total impact of a specific cause of death, including causal relationships between the cause in question and other causes recorded on death certificates. However, these methods require assumptions about causal pathways. In this paper, we relax several of these assumptions. Specifically, we introduce economic distress as a potential common driver of opioid overdose and other types of mortality that may drive correlations, and disaggregate mortality trends by population, time period and type of opioid involved. We find evidence that correlations driven by economic conditions affect prior estimates of spillover, and also that prior estimates are sensitive to time, drug and populations studied, substantially overstating recent contributions of opioid use to life expectancy trends.
Works in progress
The Determinants of Life Expectancy in the 20th Century U.S.
(with Martha J. Bailey)
The Long-Term Health and Economic Benefits of Community Health Centers
(with Martha J. Bailey, Valentina Duque, and Andrew Goodman-Bacon)
Other work
Private Student Loans Remain a Significant Source of Educational Debt (2025) CFPB Office of Research Working Paper No. 25-08
(with Cooper Luce)
This paper uses existing data to examine dynamics in the private student loan market and revisits several findings from earlier work done by the CFPB on characteristics of private student loan borrowers. We show that, after the rapid decline in in-school private student loan originations between 2007-08 and 2009-10, there has been steady growth in this market over the last decade. Furthermore, the overall growth in private student loans is likely significantly larger due to the increase in private refinance student loans over this same period. Finally, we show that many of the findings from the 2012 CFPB report on who takes out private student loans remain true today. While most student loan in the U.S. are federal, the private student loan market remains an important component of the higher educational financing system and better understanding who borrows, the performance of loans, and interaction between the private and federal loan systems will contribute to a more complete understanding of the student loan system in the U.S. However, a lack of comprehensive data on private student loans remains a hurdle in this research area.
Insights from the 2023–2024 Student Loan Borrower Survey (2024) CFPB Office of Research Report
(with Isabelle Caldwell, Thomas Conkling, West Garrett, Christa Gibbs, and Cooper Luce)
This report provides initial survey findings related to student loan borrowers’ difficulties with repayment and the make-up of borrowers applying for and receiving loan forgiveness and discharges. The survey was fielded between October 2023 and January 2024, as the payment pause on federal student loans ended and many borrowers were returning to repayment.
Mortgage Refinancing during the Great Recession: The Role of Credit Scores (2016) Chicago Fed Letter No. 355
(with Gene Amromin and Benjamin J. Keys)
This article examines whether deteriorating credit scores may have posed a barrier to mortgage refinancing during the Great Recession of 2008–09 and its immediate aftermath. The authors find that in general, as long as borrowers kept up with their mortgage payments, their credit scores did not fall significantly over this period. Hence, credit scores are not likely to explain why certain borrowers with sufficient home equity did not refinance their mortgages.