Post-Secondary Correctional Education and Recidivism: A Meta-Analysis of Research Conducted 1990-1999

Journal of Correctional Education, Jun 2004 by Chappell, Cathryn A

Recidivism is only measured when it occurs within the same state as the inmate was originally incarcerated and is therefore a conservative estimate of actual behavior. At this point, a national database is either not available, or is not used, by states in tracking their releasees. However, it should be noted that this was the case for virtually all the studies included in the meta-analysis, and is therefore less problematic than if this was not the case.

Selection bias and lack of matched control groups are issues in any study of corrections education. Also, relatively few studies looked at the inmates' achievement in the PSCE programs (grades, attendance, motivation, et.) in reference to recidivism. Differences in PSCE participation and recidivism construct and length also exist but may be tested as possible moderators.

Meta-analysis as a technique is also subject to several limitations. As with any research, the results are only as good as the input. The assumption is made, when selecting the final set of studies for inclusion, that rigor and accuracy of reporting are present. The fact that much research of any kind that fails to show significance is often relegated to the file drawer of the primary researcher is also a limitation to meta-analysis.

Although several possible moderators will be examined, there is the very real possibility of unknown or unidentified moderators. Perhaps if this research becomes widely distributed, other ideas will be suggested. It is also possible that the researchers of the original 15 studies may have input or suggestions as well.

The choice of specific data analysis was a subjective decision. This researcher chose the Arthur, Bennett & Huffcutt (2001) guided text of the Hunter and Schmidt (1990) approach using the SAS PROC MEANS based on recommendations from several respected statisticians. The decision to choose a metaanalysis of correlations, rather than effect sizes, was also a matter of judgment. As always when using correlational data, there is the danger of implying causation. It is hoped that emphasis on the relational quality of the constructs of this meta-analysis was both obvious and ubiquitous, so that an unintentional implication of causation was not inferred by the reader.

Results

Due to the selection process used in the final set of studies, only 15 studies were deemed appropriate for a meta-analysis of post-secondary correctional education and recidivism for the 1990-1999 time frame. Data extracted from each study included: N, PSCE (type of courses, i.e. academic or vocational, and amount, i.e. minimum number of classes taken or degree completed), recidivism construct (re-incarceration, re-arrest etc.), and recidivism length (the length of follow-up for checking each releasee for recidivism)(see Table 1). Numerical data was also extracted and converted, if necessary into correlation, variance and effect size statistics (see Table 1). In subset 4, a smaller meta-analysis, the same calculations were conducted on the three studies with control groups, using the control group recidivism rate rather than the institutional rate. Sample sizes were also adjusted according to the size of the treatment and control groups (see Table 2). A total of 15 studies were included with a total sample size of 7320 subjects.