All inmates released from Florida prisons from January 1, 2005 to December 31, 2012 are included in the study with the following exceptions:
This methodology follows the definition of recidivism stated by the Association of State Correctional Administrators (ASCA) Performance-Based Measures System Counting Rules.4 For this study, survival analysis techniques are used to compute recidivism rates and to define the statistical models that determine which factors significantly influence recidivism rates. The basic rates for tables and graphs are computed from Kaplan-Meier estimates of the survival curve using right-censored data. The analyses of factor significance are conducted using Cox models (proportional hazards regression) of the same data. The analysis used a 5% level of significance to determine the factors to include, and a stepwise selection routine for determining the order of importance. The correlations between factors were considered during the stepwise routine to eliminate highly correlated variables from both being selected for inclusion in the model.
The following variables were considered for inclusion in the model:
Note that for the "worst" offense factors the hierarchy is: murder/manslaughter > sex offense > robbery > other violent offense. Each inmate can only be designated in at most one of the categories. For example, if he has committed both sex offenses and robbery, he will be considered in the "worst offense is sex offense" category, not in the "worst offense is robbery" category.
Also, for the criminal history factors, only those offenses for which the inmate received either a Florida prison sentence or Florida community supervision sentence are considered. Crimes in Florida that resulted in other sanctions, such as fines or county jail or federal sentences, are not considered. Arrests, supervision, or prison sentences outside of Florida are not considered unless they are part of the inmate's Florida sentence.
4“ASCA Performance-Based Measures System Counting Rules” (April 2013 Web) http://www.asca.net/system/assets/attachments/5685/PBMS%20KeyIndicators%204_3_13.pdf?1365012031