Research In Progress

“Gender Differences in Learning: Evidence from the Emergency Department”

Gender differences in how workers perceive their own abilities may be an important channel through which gender gaps in performance appear and persist in the labor market. I use administrative data on every Emergency Department (ED) visit in New York from 2005-2015 to understand how doctors update their perceptions of their own abilities after a patient they treated dies. Using an event study design, I study how doctors change their treatment behavior in response to two types of patient deaths: when the patient had an ex-ante high(low) likelihood of death and the doctor was unlikely(likely) to be at fault. I explore how the effects of these shocks on a doctor’s treatment style, and the subsequent effects on the cost and quality of care they provide, differ by doctor gender. Preliminary results suggest that, even in situations where a patient’s death was unavoidable, male and female doctors respond by adjusting their treatment styles differently in ways that exacerbate existing gender gaps in performance.


"Habit Formation and Technological Deadoption”

Many clinical practices are ineffective or even harmful, yet they continue to be provided to patients. We still do not understand what causes healthcare practitioners to be slow to stop providing care that has been deemed low-value. I use administrative data on every hospital visit across the State of New York from 2005-2015 to understand how doctors de-adopt once-accepted practices that have been deemed harmful. I focus on two clinical practices: early elective deliveries for pregnant women and drug-eluting stents for blocked arteries, both of which were once common, then deemed harmful  in the early 2000s and early 2010s respectively. Both practices were de-adopted with significant variation across hospitals in speed and extent of de-adoption. I develop a model of physician habit formation as a within-physician productivity spillover, and test it as an explanation for slow technological de-adoption against other theories of information diffusion, risk aversion or institutional barriers.


“Decision Fatigue and Not-So-Grumpy Judges?”

How does decision fatigue affect high-stakes choices? Previous evidence from Israel shows that the likelihood of granting parole decreases dramatically as judges get fatigued and “grumpy”. I find evidence to the contrary. Using administrative data on every parole hearing held at the New Hampshire Parole Board between 2008 and 2011, I exploiting the alphabetical ordering of parole hearings at the New Hampshire Parole Board to estimate the causal impact of judge fatigue on hearing outcomes. I show that having a parole hearing 10 minutes later in the day increases the likelihood of receiving parole by 2.5-3.1pp. The results support the idea that decision fatigue is distinct from changes in the emotion of the decision-maker. I hypothesize that the “default heuristic” - when fatigued, judges take the decision taken most often - may reconcile these and prior findings, as 82% of New Hampshire hearings result in parole.