Dataset Limitations

The analysis dataset includes three limitations, acknowledged by the FDOC,²⁰ but ones that should not impair their use in this study. First, some inmates released from prison in Florida who commit subsequent felonies are sentenced to local jails rather than returned to the department’s jurisdiction. Felonies committed by these offenders do not appear in the recidivism analysis dataset. Based on analysis of data from another source that includes data on most felony sentences to jail, the FDOC reports: “At most, including jail data would raise the re-offense rate estimates by an average of 1.2 points with a maximum increase of 1.6 points in the 3-year range. This minimal effect reflects the low probability that former prison inmates are sentenced to jail instead of prison for subsequent felonies.”

Second, inmates released out-of-state are removed from the analysis. This is necessary because the FDOC does not have access to corrections data from other states. The department notes that, “Using only Florida data to determine whether an inmate who was released out-of-state re-offended and returned to the department’s custody would skew the recidivism rates lower.” However, based on estimates of out-of-state arrests reported by the U.S. Bureau of Justice Statistics and the ratio of reconvictions to arrests, the FDOC reports that including releases outside Florida would “increase these re-offense rate estimates by an average of .85 points, with a maximum of 1.3 points at 60 months.”

A third limitation is that inmates who die subsequent to release are not excluded from the analysis. However, post prison release deaths within 60 months occur in a small minority of cases, and are unlikely to affect rates substantially. Since this research focuses on comparing recidivism rates between inmates housed in public and private facilities, the only way these data limitations could bias the results is if the likelihood of jail sentences for new offenses, out-of-state releases and re-convictions, or post-prison death are different for public and private prison inmates. The authors are unaware of any theoretical reason why these factors would differ between the treatment and control groups analyzed for this study.

20. FDOC, 2003.