How Different Sampling Methods Paint Vastly Different Pictures of Recidivism, and Why It Matters for Policy

In the national conversation around criminal justice reform in the United States, there are competing views about people with criminal histories. On the one hand, there is a long-standing view that people with criminal histories recidivate often and quickly. In the past several years, however, there has been a growing movement arguing that people should not be defined by their criminal histories and that people with criminal histories are not criminals in the persistent sense of the word.

It is important that these debates about public safety and the effects of interventions be grounded in facts. That requirement demands accurate measurement of how people with criminal histories behave, particularly how often they return to criminal activity. In this paper, the authors argue that the recidivism statistics cited most often in debates about the collateral consequences of criminal conviction are not appropriate to answer the questions inherent in those debates. In particular, the behaviors of criminal justice cohorts are too often mistakenly used to describe, or are entangled with descriptions of, behaviors of the overall population of people who have ever had a conviction, served time in prison, or experienced some other event in the criminal justice system. This confusion has consequences. This paper demonstrates that the most-cited recidivism statistics often are based on criminal justice cohort samples that disproportionately contain frequent participants in the criminal justice system and, as a result, have higher recidivism rates compared with the broader population of concern.