The Best Ones Come Out First! Early Release from Prison and Recidivism A Regression Discontinuity Approach *

There is strong evidence that incarceration has a general deterrent effect on individuals on the margin of crime. The impact of the experience of incarceration on future criminal behaviour, self deterrence, is more controversial. It is becoming a pressing issue in view of the large increases in the prison population of past decades. A main question is if harsher or more lenient sentences are more efficient in reducing future recidivism? The fact that offenders with different criminal profiles are treated differently makes it difficult to answer. The best behaved and least dangerous inmates are, for example, more likely to be selected for an early release programme. These characteristics will also influence their future offending behaviour making it difficult to identify the impact of time spent in prison. In this paper we exploit an administrative rule which makes offenders sentenced to less than three months in prison ineligible for the Home Detention Curfew (HDC) scheme in England and Wales to estimate the impact of early release on recidivism using a regression discontinuity (RD) approach. We have access to detailed data on all prisoners released between 2000 and 2006 and their past and future criminal history. We first obtain estimates controlling and matching on observable characteristics which find that the policy reduced recidivism by about 9 percent. The RD methodology takes into account the potential importance of unobservable characteristics. We find that the policy impacts remain relatively unchanged. However, when taking into account prison establishment unobserved characteristics, our results are weakened but still suggest that early release on electronic monitoring can reduce the likelihood of future arrest by 5 to 7 percent.