Daniel Mejía.

This paper estimates the causal effect of pretrial detention on subsequent criminal behavior using linked police and judicial administrative records from Bogotá, Colombia, for the period 2008–2018. To address selection into pretrial detention, we exploit the quasi-random assignment of cases to judges who differ in their propensity to detain individuals pretrial. Pretrial detention reduces the probability of rearrest by 9 percentage points and the probability of new charges by 5.4 percentage points, corresponding to reductions of approximately 32 percent and 35 percent relative to the respective baseline means. The estimated reductions in recidivism are concentrated among property crime defendants and repeat offenders. The estimated effects are also larger among defendants with longer detention spells, suggesting that pretrial detention may increase the salience of punishment or disrupt criminal routines and networks. These findings contrast with much of the existing U.S. evidence, which typically finds no net effect on crime.