Building Criminal Capital vs Specific Deterrence: The Effect of Incarceration Length on Recidivism

Document Type


Publication Date



In evaluating the efficacy of most modern criminal justice systems, a vital relationship to understand is that between incarceration length (and likelihood) and recidivism. Because most previous attempts to estimate this relationship suffer from omitted variables bias, even the sign is unknown. In this paper, I build on previous work identifying substantial heterogeneity in attorney ability in a public defender office with random case assignment. I make use of this variation to address the omitted variables problem by instrumenting for sentence length and incarceration rate using the randomly assigned public defender. A negative relationship between recidivism and sentence length goes away when instrumenting for sentence. Similarly, a positive and statistically significant relationship between recidivism and incarceration becomes insignificant in the IV regressions. However the regression results do not reveal the full story, as the relationships are rather nonlinear. A graphical examination reveals a negative relationship between recidivism and sentence length and also recidivism and incarceration rate, particularly for shorter sentences and lower incarceration rates. In addition, longer sentences tend to lead to more severe crimes upon offender release. Put together, these findings provide some evidence for a mild specific deterrent effect, but one that rapidly diminishes.

Full text not available in Penn Law Legal Scholarship Repository.