Aurélie Ouss is an Assistant Professor in the Department of Criminology at the University of Pennsylvania. Aurélie's research examines how good design of criminal justice institutions and policies can make law enforcement fairer and more efficient. 

Abstract 

This paper examines the effects of conviction and incarceration on recidivism using quasi-random judge assignment in a multiple-treatment setting. We first present estimates based on 2SLS regressions with judge stringency instruments. If given a causal interpretation, these estimates imply that receiving a felony conviction results in large and sustained increases in recidivism. In contrast, being incarcerated reduces recidivism in the first year, likely due to incapacitation, but has no detectable longer-term effects. We show that, in multiple-treatment settings, some models of judge decision- making facilitate interpretation of the 2SLS estimands as a causal and margin-specific treatment effect while others do not. We specify testable implications for the models we consider and reject models that support a causal and margin-specific interpretation of the 2SLS estimand. We characterize the resulting bias and argue that it is likely to be small in our setting. Finally, we describe and implement an alternative strategy for identifying causal and margin-specific treatment effects. This analysis yields estimates qualitatively similar to those based on the 2SLS estimates, although they are sometimes less precise. Taken together, our results suggest that conviction may be an important and potentially overlooked driver of recidivism, while incarceration mainly has shorter-term incapacitation effects.