Study 1
Jalbert and colleagues (2011) used a quasi-experimental design to examine the effectiveness of the Reduced Probation Caseload in Evidence-Based Setting (Oklahoma City, Okla.) program on recidivism for high- and moderate-risk individuals on probation at the 2-year follow up.
Oklahoma City had implemented evidence-based practices in the assignment of individuals on probation to either administrative caseloads (where individuals are minimally supervised) and active caseloads (high- and moderate-risk individuals receive active probation supervision) and thus was an eligible site for participation. The study design originally used a random assignment of volunteer probation officers to conditions. About half of the volunteer officers were assigned to an experimental condition (reduced caseload probation officers), while the rest were initially assigned to a control condition (control probation officers); those who did not volunteer served as a comparison group (regular caseload probation officers). The random assignment deteriorated, however, after numerous officers dropped out of the control group. As a result, the control and comparison probation officers were combined into a single comparison probation officer condition, where officers had regular caseloads. Officers in the experimental condition shifted randomly selected individuals on probation off their active caseloads to achieve an ideal caseload of around 50. These officers were allowed to keep a small number of individuals on their caseload who were nearing the end of their probation sentences (within 3 months). Any other randomly selected case was shifted to a control condition officer. New probation cases were randomly assigned to experimental and comparison condition officers.
During the study period, probation officers in the experimental condition averaged 54 cases, compared with comparison condition probation officers, who averaged 106 cases.
Data were collected from several sources. Supervision outcomes (whether the individual on probation completed the term of supervision, had their term revoked, or was still under supervision as of August 2010) came from probation records. The Oklahoma Department of Corrections provided probation data for all supervision cases directly before and during the reduced caseload intervention. Criminal history data were obtained through a query of arrest records from the Oklahoma State Bureau of Investigation. For some individuals on probation, the agency was unable to match supervision records with criminal histories, so the analysis of arrests during and after supervision was limited to those observations that had matches with criminal history records. Assessment scores for individuals on probation were derived from the Level of Service Inventory—Revised (LSI—R) assessment tool that includes 54 questions on an individual’s attributes and their situations to assess their risk of recidivism and needs for treatment or intervention. Data on demographic characteristics, individual’s treatment programming (such as substance treatment), and supervision contacts were also collected.
Experimental condition probation officers with reduced probation caseloads monitored 1,054 individuals on probation, and comparison condition officers supervised 3,877 individuals on probation. The only statistically significant difference between the conditions was in individual history of alcohol and drug abuse, which was present in 41.8 percent of individuals receiving supervision from experimental condition officers, but in only 29.7 percent of individuals receiving supervision from comparison condition officers. Otherwise, the groups were similar on their LSI—R scores (with a score of 16.6 for individuals supervised by experimental condition officers, and 16.16 for individuals supervised by comparison condition officers) and in regard to gender (73.4 percent of individuals supervised by experimental condition officers were male, and 71.4 percent of individuals supervised by comparison condition officers were male). About 53 percent of individuals supervised by experimental condition officers had a high school diploma or higher, while 52.9 percent of individuals supervised by comparison condition officers reached the same level of educational attainment. Finally, 56.1 percent of individuals supervised by experimental condition officers had a prior conviction, incarceration, or probation, compared with 49.8 percent of individuals supervised by comparison condition officers.
Difference-in-differences estimators calculated using Cox proportional hazard models—a form of survival analysis—were used to examine the program’s effect on recidivism, defined as an arrest for any crime exclusive of a technical violation of the conditions of supervision, after 2 years. The analysis controlled for individuals’ LSI—R scores, gender, education, alcohol and drug use, and prior conviction, probation, or prison term. Additionally, the analysis used multiple imputation to deal with missing data in the LSI—R scores. No subgroup analysis was conducted.