Research

Job Market Paper

Between 2000 and 2014, yearly real average college tuition increased by 31% while the average student loan balance increased by 21%. I use administrative data to examine how an increase in the cost of higher education affects the likelihood of forming a business after graduation. I instrument for realized tuition by exploiting the variation in total four-year tuition due to differences in the length of exposure to increased tuition between students in different enrollment years at universities that experienced a large “sticker price” tuition shock. I find that for every $10,000 increase in tuition there is a 3.0 percentage point (41 percent) decrease in the likelihood of being an owner or early joiner of a business. This decline is 7.8 percentage points larger for individuals at universities that have larger student loan balances and 4.1 percentage points higher for individuals with parents who did not complete college. The negative effect on business formation does not exist for students attending universities with more generous financial aid. These results suggest that increasing higher education costs are deterring recent graduates from engaging in entrepreneurial activity. 

with Benjamin Collier and Sabrina Howell

NBER Working Paper No. 32326

Outstanding Paper Award from Wharton Research Data Services for the Best Financial Institutions Paper at the MFA 2024

Does emergency credit prevent long-term financial distress? We study the causal effects of government-provided recovery loans to small businesses following natural disasters. The rapid financial injection might enable viable firms to survive and grow or might hobble precarious firms with more risk and interest obligations. We show that the loans reduce exit and bankruptcy, increase employment and revenue, unlock private credit, and reduce delinquency. These effects, especially the crowding-in of private credit, appear to reflect resolving uncertainty about repair. We do not find capital reallocation away from neighboring firms and see some evidence of positive spillovers on local entry.

with Katharine Abraham

Brookings Papers on Economic Activity, 2023

Media Mentions: Reuters, Fortune, Bloomberg, Axios

Labor force participation and average hours of work both fell sharply at the beginning of the COVID-19 pandemic. Neither had fully recovered by the end of 2022. The drop in participation between December 2019 and December 2022 implies a loss of 3 million people from the labor force; the decline in average hours over the same period translates to the equivalent of 2.6 million fewer workers. Demographic and other trend factors that predated the pandemic explain most of the participation shortfall. Taken together, COVID-19-related health effects and the persistent (though shrinking) effects of the fear of contracting COVID-19 more than explain the rest. In contrast, pre-pandemic factors account for little of the shortfall in hours. COVID-19-related health effects account for perhaps 40 percent of that decline, but we are unable to explain the majority of the hours shortfall. We speculate that the lower level of hours in the post-pandemic period may reflect a shift in the desired balance between work and other aspects of workers’ lives

with Katharine Abraham and John Haltiwanger

Brookings Papers on Economic Activity, 2020

Media Mentions: Bureau of Labor Statistics, Washington Center for Equitable Growth

We construct a generalized measure of labor market tightness based on the ratio of vacancies to effective searchers. Our generalized measure exhibits substantially less volatility than the standard measure defined as the ratio of vacancies to unemployment. Effective searchers include not only the unemployed but also those who are out of the labor force and the employed. These groups account for a substantial share of hires and their presence mutes the effects of the pronounced countercyclical movements in unemployment. The effective searcher measure also distinguishes different groups among the unemployed. During protracted contractions, the distribution of unemployment shifts toward the long-term unemployed, a group with lower relative search intensities, contributing to the smaller proportional increase in effective searchers as compared to the simple unemployment count. The Beveridge curve constructed using effective searchers is much more stable than the standard Beveridge curve. Further, the matching function for hires based on our generalized measure outperforms the matching function based on the ratio of vacancies to unemployment. Our approach thus reduces the unexplained residual variation required in the matching function to be consistent with real world data.

with Michael J. Lamla and Damjan Pfajfar

We explore the consequences of losing confidence in the price-stability objective of central banks by studying the resulting inflation and deflationary biases in medium-run inflation expectations. In a model with heterogeneous household perceptions of an occasionally binding zero-lower-bound constraint and of monetary policy objectives, we show that the estimated model implied distribution of households' inflation expectations matches several characteristics of the empirical distribution when featuring both inflation and deflationary biases. We then directly identify these biases using unique individual-level medium-run inflation expectations survey data across nine countries and over time. Both inflation and deflationary biases are important features of the distribution of medium-run inflation expectations.

with Jorge Perilla-Garcia

Between 1991 and 2016, there was a steady increase in the number of states that passed targeted restrictions on abortion providers (TRAP) laws. In this paper, we study how different demographic groups of voters, among them young female voters, respond to these new restrictions. We geocode the distance from the nearest abortion clinic for each voter using information on residential addresses from each state's voter file.  We examine the difference in turnout for individuals who had a clinic close nearby before the election to those who had this happen after the election in a sample of four states: Florida, North Carolina, Pennsylvania, and Ohio. We find that, on average, there are no effects for any demographic group, which suggests that voters were not more motivated to participate in the democratic process after these events. For North Carolina, there is a positive turnout effect for young women, young men, and Democrats, with the first of these effects being the largest. This effect is significantly different from the effects for the other three states.