Assessing Recidivism Risk Across Female Pathways to Crime

Abstract

Actuarial tools, such as the Level of Supervision Inventory—Revised (LSI‐R), are regularly used to classify offenders as “high,” “medium,” and “low” recidivism risks. Its supporters argue the theory upon which the LSI‐R rests (i.e., social learning theory) accounts for criminal behavior among men and women. In short, the LSI‐R is gender‐neutral. Feminist criminologists question the LSI‐R’s validity for female offender populations, especially women under community supervision. Guided by Daly’s (1992, 1994) pathways to crime framework, we use a sample of women under community supervision in Minnesota and Oregon to evaluate the LSI‐R’s performance across offender subgroups. The results show that the LSI‐R misclassifies a significant portion of socially and economically marginalized women with gendered offending contexts. Predictive accuracy was observed for women who did not follow gendered pathways into criminality, whose offending context was similar to males, and who occupied a relatively advantaged social location.

Keywords:

• community corrections
• LSI‐R
• recidivism
• risk
• women offenders

Acknowledgments

This research was supported by Grant #96‐IJ‐CX‐0021 by the National Institute of Justice, Office of Justice Programs, US Department of Justice. Points of view in this document are those of the authors and do not represent the official position or policies of the US Department of Justice. A previous version of this manuscript was delivered at the Societies of Criminology 1st Key Issues Conference, Paris, France, in May 2004.

Notes

1. Schmidt, Hoge, and Gomes (2005) report the following bivariate correlations between scores derived from the youth version of the LSI‐R (i.e., the Youth Level of Service/Case Management Inventory) and recidivism for their sample of 38 “disturbed” and “serious” female juvenile Canadian offenders: r = .14 for reoffense, r = .35 for serious reoffense, and r = .20 for number of new offenses (p. 338; also see Jung & Rawana, 1999).

2. The original research design called for three follow‐up interviews at six month intervals. Financial constraints prevented completion of all three interviews for every participant. As a result, the average observation period for the women included in the analyses that follow was eleven months (SD = 3.3 months).

3. Officer surveys were administered after each study participant was interviewed. Community corrections officers were asked a variety of questions about their clients, but most importantly for our purposes officers were asked about official recidivism events (e.g., supervision violations and rearrest). Logs consisted of correctional officers’ notes regarding study participants. This source provided information regarding participants’ life circumstances (e.g., familial and intimate relationships) prior to the felony conviction that made them eligible for the study.

4. Our method of constructing biographies closely resembled Daly’s (1992, pp. 18–19) strategy. Only the information related to circumstances occurring prior to a woman’s felony conviction was incorporated into her biography. While Daly relied solely on PSIs, we accessed data from multiple sources (e.g., correctional staff, PSIs, and participants).

5. The sample mean is similar to those reported in other studies assessing the LSI‐R’s predictive accuracy of recidivism for women, such as Coulson et al. (1996, p. 432) (mean = 15.50) and Lowenkamp et al. (2001, p. 552) (mean = 25.05).

6. Although examples litter the classification research literature, the practice of collapsing continuous measures (e.g., the LSI‐R) into ordered categories unnecessarily discards information and can produce misleading results (see Reynolds, 1977). To guard against potential bias, we present correlations between recidivism and two versions of the LSI‐R (i.e., coded as both categorical and continuous variables). When the items used to construct our recidivism outcome are assessed individually, we observed the following bivariate associations with the LSI‐R (coded as a continuous variable): violation of supervision conditions (r = .03), rearrest (r = .05), reconviction (r = −.01), and revocation of community supervision (r = .05).

Additional information

Notes on contributors

Michael D. Reisig

Michael Reisig is an Associate Professor in the College of Criminology and Criminal Justice at Florida State University. His corrections research has appeared in a variety of journals, including Criminology, Criminology and Public Policy, and Punishment and Society.

Kristy Holtfreter

Kristy Holtfreter is an Assistant Professor of Criminology and Criminal Justice at Florida State University where she is also affiliated with the Women’s Studies program. Her research examines white‐collar crime, fraud victimization, and female criminality. Recent articles have appeared in Criminology and Public Policy, Journal of Criminal Justice, and the Journal of Financial Crime.

Merry Morash

Merry Morash is a Professor of Criminal Justice at Michigan State University. Her research projects include programming to meet the needs of women offenders, and domestic violence among different racial and ethnic groups. She has conducted research with funding from the National Institute of Justice, the National Science Foundation, and numerous other sources.