Skip navigation.
Home | About Us | Contact Us
Rick Scott, Governor
Florida Department of Corrections, Secretary Michael D. Crews

Florida Department of Corrections
Michael D. Crews, Secretary

Conclusion and Related Reports

III. Evaluations of prison inmates that use reoffense or reimprisonment rates as outcome measures should take certain (or similar) factors into account by controlling for their statistical effects on recidivism rates.

This report documents important and, in some cases, large effects of certain factors on reoffense and reimprisonment rates. These factors' effects must be accounted for when measuring and comparing recidivism rates. Failure to do so may bias recidivism measures and confound conclusions regarding whether differences in rates are attributable to certain correctional activities, functions, or programs. Evaluations using recidivism outcome measures for Florida state prison inmates should account for the following factors (or similar ones), measurable at the time an inmate is released:

  • Gender
  • Age at release
  • Race
  • Ethnicity
  • Number of prior recidivism events
  • Type of most serious crime ever committed
  • Number of property crimes ever committed
  • Number of drug crimes ever committed
  • Number of weapons crimes ever committed
  • Education or academic skill level
  • Custody level at release
  • Number of disciplinary reports
  • Length of stay in prison
  • Supervision after release

Several aspects of the relationships between these factors and recidivism should be noted.

  • Some factors influence both reoffense and reimprisonment in the same direction. For example, older inmates have both lower reoffense and reimprisonment rates than younger inmates. Of factors analyzed in this study, five raise both reoffense and recidivism for males and females, whereas seven lower both reoffense and recidivism for males and females.

  • Other factors influence reoffense and reimprisonment in different directions. For example, males with supervision after release have lower reoffense rates, but higher reimprisonment rates than those without supervision.

  • Some factors influence male and female recidivism to different degrees. For example, inmates with burglary as the most serious crime ever committed are more likely to reoffend and to be reimprisoned. However, this factor only raises male reoffending 11.6%, whereas it raises female reoffending 27.1%. Similarly, it raises male reimprisonment 28.7%, but female reimprisonment 58.6%.

  • Some factors influence male and female recidivism in different directions. For example, black males have higher reoffense and reimprisonment rates than non-blacks, but black females have lower reoffense and reimprisonment rates than non-blacks.

Not all factors affect recidivism with the same statistical significance. For males, 16 of 18 factors analyzed significantly influence reoffending and reimprisonment. For females, reoffending is significantly influenced by only 12 factors and reimprisonment by 10 factors. This variation in the effects of factors implies several conclusions.

  • Male and female inmates should be analyzed separately, because factors affect recidivism in these inmate subpopulations differently.

  • It is preferable to use both reoffense and reimprisonment measures of recidivism to gain a more complete understanding of recidivism.

  • When analyzing subsamples of male and female inmates, all factors should be examined for potential influence. However, some factors may not show a statistically significant effect on recidivism in particular subsamples. This can occur, for example, when the subsample has one or more factors with distributions quite different from the entire male or female cohort.

  • Similarly, in analyses of some subsamples of male and female inmates, the size and, possibly, the direction of a factor's effect may differ from that reported here. If this occurs, special attention should be paid to possible interactions between multiple factors in those subsamples.

A recent case from the research literature provides an example of problems caused by failing to account for the effect of an important factor that influences recidivism. A study published in 2002 on the relationship between education and recidivism (i.e., return to prison) of Oklahoma prison inmates (21,268 released from 1991 through 1994) found some unusual results. Among other findings, the study reported that: (1) earning a general equivalency diploma (GED) lowered recidivism, but that completing vocational training programs raised recidivism; (2) drug distribution offenses raised inmate recidivism, but drug possession offenses did not; and—most problematic—(3) older inmates recidivated at higher rates than younger inmates. However, there is scant theoretical reason to believe that vocational training actually increases recidivism. Further, most research shows inmates with drug possession offenses recidivate at rates at least equal to those with drug distribution offenses. Finally, virtually every study finds that younger inmates recidivate at higher rates than older inmates. The authors admit the odd findings may be due to the effect of repeat incarcerations, especially regarding the age finding, stating they "did not examine the effect of prior incarcerations." (See Brewster, Dennis R. and Sharp, Susan F. "Educational Programs and Recidivism in Oklahoma: Another Look" The Prison Journal, Vol. 82 No. 3, (September 2002) pp. 314-334.)

Had the authors of that research controlled for prior recidivism, a factor known to increase recidivism substantially, their results might well have been substantially different. Clearly, older inmates are more likely to have more prior recidivism events than younger inmates, so controlling for prior recidivism might have shown that older inmates recidivate less—a finding consistent with existing research. The Oklahoma Department of Corrections reports that drug possession inmates have a higher three-year recidivism rate (28.4%) than drug distribution inmates (23.5%), based on releases from 1985 through 1999; so controlling for prior recidivism could have allowed the authors to find results consistent with the agency's reports. More importantly, controlling for prior recidivism may have avoided the undesirable and theoretically problematic finding that vocational training increases recidivism, especially if vocational training completers in the sample were older and had more prior incarcerations. Finally, the desirable finding that earning a GED lowers recidivism would be credible, if the authors had not neglected to control for prior recidivism—a very important influence on recidivism.