Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Timothy Austin, Ph.D.

Second Advisor

Erika Frenzel, Ph.D.

Third Advisor

John A. Lewis, Ph.D.

Fourth Advisor

Shannon W. Phaneuf, Ph.D.


The classification system in corrections seeks to accurately predict inmates’ future behaviors by utilizing risk assessment. Of actuarial risk assessment inventories, the Level of Service Inventory-Revised (LSI-R) has been widely utilized to classify offenders for treatment and prevention of re-offense. The LSI-R has been considered a ‘gender-neutral and culture-responsive’ risk assessment inventory. Even though multiple studies assert the validity of the LSI-R beyond gender and culture, there has been a concern that such research has been conducted predominantly with white male inmates. Recognizing such concern, the primary purpose of this study was to assess the validity of the LSI-R based on gender, offense type and race. The study also documented the historical scheme of the evolution of classification systems, the risk assessment inventories, and the theoretical underpinning of the LSI-R. The study sample consisted of 12,975 male and female offenders who were released from the Pennsylvania Department of Corrections in 2004. The data were collected about the initial LSI-R score, the recidivism record for 36 month follow-up period, and other demographic characteristics. Nine hypotheses were tested by conducting a series of statistical analyses including two-way ANOVA and logistic regressions. Relationships tested include the impact of gender on LSI-R scores and recidivism, and the impact of gender and offense type impact on LSI-R scores and recidivism. Also tested was the impact of race on LSI-R scores and recidivism. Research findings supported the impact of gender on the LSI-R scores, but did not support the offense type impact or the interaction impact of gender and offense type on LSI-R scores. The results also supported the prediction that the LSI-R score was a reliable measure in forecasting recidivism for each racial group. However, the results called into a question about the predictability of the LSI-R subscales for the violent female offender group by failing identification of salient factors of such subscales. The findings of this dissertation suggest caution in using the LSI-R to predict recidivism for violent and nonviolent female offenders. This research has clear implications for development of gender-specific risk assessment tools and provides new empirical evidence for risk assessment practices.