Florida Department of Corrections
Michael D. Crews, Secretary

March 2006 Bureau of Research and Data Analysis, Florida Department of Corrections

The Washington State Institute for Public Policy (WSIPP) recently conducted a meta-analysis of adult corrections programs and policies to determine which programs reduce recidivism¹.  Analyses and studies conducted by the Florida Department of Corrections (FDOC) or other researchers using program data about FDOC inmates and offenders show program-specific reductions in recidivism similar to what was found in the WSIPP analysis.  That is, many of the same groups of programs that were found to have a demonstrated ability to reduce crime in the WSIPP study were also found to do so in previous analyses of adult corrections in Florida.  When cross-referenced, the WSIPP findings and the results of the Florida analyses indicate that selected programs work.  Similar findings from such a large-scale meta-analysis and from individual studies or analyses would tend to strengthen the validity of the respective studies.

By nature, the methodology used in a meta-analysis (which examines the results of numerous studies) differs from that used in studies or analyses using data from program results.  Yet, a review of key, applicable methodological similarities between the WSIPP meta-analysis and the FDOC-related analyses seems to further substantiate or validate recommitment/recidivism reduction findings for adult correction programs nationally and in Florida.  Similarities include:

1. The importance of using studies/analyses from both peer-reviewed publications and non-peer reviewed, non-published sources;
   
   
2. Use of a program and comparison group.  Studies/analyses using single-group, pre-post research designs are excluded.
   
   
3. Lack of a random assignment study does not preclude use of the program data (only about 20% of the studies examined by WSIPP were from well implemented random assignment studies).  Data from quasi-experimental designs were used as long as the treatment and comparison groups were comparable on pre-existing conditions like age, gender, and prior criminal history (see item 4 below).
   
   
4. Use of results from only Level 3 (or higher) analyses or studies, using a 5-Level evaluation scale developed by University of Maryland researchers².
   
   Thus, the 291 studies used by the WSIPP were either Level 3 (non-experimental with program evaluations where program and comparison groups were reasonably well-matched on pre-existing differences in key variables); or Level 4 (uses a rigorous quasi-experimental research design with a program and matched comparison group, controlling with statistical methods for self-selection bias that might otherwise influence outcomes); or Level 5 (a well-implemented random assignment of participants to a treatment and a control group that does not receive the treatment/program).³
   
   Analyses of Florida-related data presented in the comparison table are primarily Level 3 or higher.  Results from some Level 2 analyses (which have a comparison group that may be dissimilar to the treatment group, such as vocational certificate earners versus all other inmates) of Florida data are given to provide additional, similar data that also found a reduction in recommitment or recidivism rates as did the WSIPP study and the Level 3 and Level 4 analyses of Florida data.  The level of analysis or scientific rigor of the Florida data is clearly stated in the comparison table.
   
   
5. Two key formulaic or definitional matches exist.  First, the formula used to get the ‘percent reduction in recidivism rates’⁴ by WSIPP has been used with the data taken from all but two of the FDOC studies or analyses presented.  The formula is the difference in the comparison group recommitment/recidivism rate and the program group recommitment/recidivism rate divided by the comparison group recommitment/recidivism rate.  The example used in the WSIPP analysis is that a 49% recidivism rate for a comparison group compared to a 45% recidivism rate for the programs group is a four-point drop -- or an 8.2% reduction in recidivism rates.⁵ Second, neither the WSIPP analysis nor the Florida data analyses include technical violators in measuring recommitment or recidivism.  The intent in both cases is to examine “new criminal activity.”⁶

An important difference between the WSIPP meta-analysis and some of the Florida studies/analyses presented in the comparison table is the exclusion by the WSIPP of studies that had only program completers in the treatment group.  They adopted this rule because they believe that often self-selection factors “... distinguish a program completer from a dropout.”⁷  Most of the treatment groups in the Florida analyses include only program completers and such analyses are identified.  A key point is that despite this difference, both the WSIPP study and the Florida studies have found a reduction in recidivism rates for selected adult corrections’ programs.  In some of the Florida analyses presented, the data can be viewed as preliminary and suggest further study.