Date of Award

1-21-2015

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Criminology

First Advisor

Bitna Kim, Ph.D.

Second Advisor

Timothy Austin, Ph.D.

Third Advisor

Alida V. Merlo, Ph.D.

Fourth Advisor

Jennifer Gossett, Ph.D.

Abstract

The creators of the LSI-R contend that this risk assessment is a `gender neutral' tool while feminist scholars remain skeptical as to the LSI-R's empirical ability to predict female recidivism as the tool was developed on male-centered theories. Research findings on the gender effect on the predictive validity of LSI-R are mixed. Very little research is available as to the effect of offense type on the tool's predictive utility. Using the disaggregated data by gender and offense types, this research aimed to determine the moderating effects of gender and offense type on the predictive utility of the Level of Service Inventory- Revised (LSI-R). This dissertation divided offense type into four categories: sex offense, person offense, property offense and drug offense. This dissertation used a sample of offenders (n=2,917) from the Kansas Department of Corrections (KDOC) who released in fiscal year 2008 (July 1, 2007- June 30, 2008). Data was collected for these offenders for a 36 month follow-up period to assess for any instances of recidivism. After controlling for offense type, logistic regression analyses showed that the LSI-R is the valid risk assessment for both male and female offenders. With the major research question of moderating effect of gender, this study found that the different subscales predict recidivism between genders. Regardless of offense type, the LSI-R total score proved to be a significant predictor of recidivism. Like the moderating effect of gender, the moderating effect of offense type in the predictive validity of LRI-R was supported. Though this dissertation found support for the predictive utility of the LSI-R across gender and offense type, no statistically significant subscale predicting recidivism for female property offenders was found. Furthermore, because statistically significant subscale predictors of recidivism varied across offense type and gender, it is recommended that future research further examines the predictive validities of subscales. Given the finding of this study that only few subscales reached the statistical significance to predict recidivism across offense type and gender, other factors should be considered to assess need and risk. Replication of this study using different samples is recommended.

Share

COinS