with M. Schaelling
One goal of standardized tests is to measure aptitude across heterogeneous students with minimal bias. However, students differ in how they relate to different topics or characters that appear in exam content. We study how differential content relatability can impact testing outcomes using item-level data from reading comprehension exams in Texas. Using time-use data and natural language processing techniques, we first build a novel measure of race- and gender-based relatability to topics in the exams' text passages. A one standard deviation increase in exam-level topic relatability across race predicts a 0.05 standard deviation change in student performance, with null effects for gender. We find positive and separate estimates for the test score impacts of race- and gender-match of students to passage characters. Our estimates suggest that equalizing the relatability of passages in these standardized tests could reduce the Black-white and Hispanic-white test score gaps by up to 9 percent. We then counterfactually estimate close to 11,000 Black students and 37,000 Hispanic students during our sample period were designated to be at a lower reading comprehension level due to relatability.
Paper presented at: Society of Labor Economists (SOLE), Society for Research on Educational Effectiveness (SREE), Population Association of America (PAA), and Association for Public Policy Analysis and Management (APPAM)
with D. Cutler, L. Dafny, D. Grabowski, and C. Ody • September 2025
We examine whether vertical integration of hospitals and skilled nursing facilities (SNFs) could lessen competition by foreclosing rival SNFs’ access to lucrative referrals. We find that it could: among integrated providers, a one percent increase in SNF reimbursement for a given patient discharged from the upstream hospital increases the self-referral rate to the hospital’s downstream SNF(s) by 1.8 percent. We find no evidence of offsetting benefits for patients and payers: these increased self-referrals have an imprecisely estimated zero effect on patient outcomes and Medicare spending.
Paper presented at: American Economic Association (AEA), Federal Trade Commission (FTC): Microeconomics
This paper examines how labor market shocks influence college attendance decisions and subsequent earnings among high school graduates. Using comprehensive administrative data from Texas spanning 2004–2015, I study students’ responses to changes in local and industry-level wage conditions. I first implement a shift-share identification strategy that combines industry-specific wage shifts with students’ predicted industry propensities, and I validate the results using the mid-2000s fracking boom as a natural experiment. I find that positive labor market conditions reduce the likelihood of college matriculation: a one standard deviation increase in labor market strength leads to a 3.2 percentage point decline in college attendance. The students induced out of college are drawn broadly across the achievement distribution, even up to the second quintile of achievement. Despite entering a temporarily strong labor market, these students earn less within five years after high school graduation, suggesting limited long-term gains from substituting work for college.
Improving the take-up of preventive care is an important public health policy goal. I study the drivers of flu vaccination in the United States, where immunization is close to universally encouraged, but take-up is relatively low. I generate a plausibly random measure of vaccine effectiveness by combining cross-state and cross-season variation in the vaccine's antigenic match to circulating strains. Results among Medicare beneficiaries from 2013-2019 suggest that more effective vaccines are likely to reduce subsequent year take-up. I find evidence of persistence and stronger effects for currently vaccinated individuals. These results could indicate that salience about the condition may be a more important driver of flu vaccination decisions than the effectiveness of the vaccine.
Paper presented at: Population Association of America (PAA)
with D. Patel
with L. Dafny • April 2020
As the number of COVID-19 cases nationwide continues to grow, a number of hospitals will need to convert acute care beds into intensive care beds, and discharge stable patients to post-acute care settings such as nursing homes. In addition, nursing homes unable to care for COVID patients requiring intensive support services - or unable to isolate their COVID-positive residents - will require safe and high-quality options. We recommend designating specific nursing homes (also known as "skilled nursing facilities") to serve as "COVID-19 Skilled Care Centers" (CSSCs). Officials should identify these nursing facilities immediately, so the facilities can decline new uninfected patients and isolate/transfer uninfected longer-term residents. (see COVID-19 Skilled Care Center Scorecard)
with L. Dafny • April 2020
(see Designating Certain Post-Acute Care Facilities as COVID-19 Skilled Care Centers Can Increase Hospital Capacity and Keep Nursing Home Patients Safer)