Degree Name

Doctor of Philosophy


School of Psychology - Faculty of Arts


Offender risk assessments function to deconstruct criminal behaviours and identify individuals most at risk of engaging in future criminal activities. Correctional agencies and institutions rely on these instruments to provide the best care and management for offenders, ranging from placements into specific treatment/rehabilitation programs, to day-to-day inhouse human resource issues. The search for jurisdictionally appropriate and accurate risk assessments to cater for the heterogenous offender population, however, is ongoing and continues to be a major challenge for correctional institutions. There is, nonetheless, increasing consensus amongst international criminal justice professionals, academics and researchers that the Level of Service Inventory – Revised (LSI-R) is the risk assessment of choice in the understanding of offending behaviours, and the accurate discrimination between one-off and repeat offenders. This thesis comprises several papers evaluating the utility and predictive validity of the LSI-R for Australian offenders.

The first paper (Chapter 2) reviews the concept of risk and the historical antecedents of risk assessments. The development, or the generations, of risk assessments provide the chronological development of the measures of risk, with examples of risk assessments highlighting advancements whilst also addressing the limitations. The review concludes with the acknowledgement that within the current collection of varied risk assessments, the LSI-R is the most empirically developed and theoretically coherent risk assessment in the understanding of the risk and criminogenic need characteristics of offenders. The potential utility of the LSI-R for Australian offenders is then explored, with issues such as gender, minority offenders and the latent constructs of the assessment tool offered as thoughts for systematic research.

The second paper (Chapter 3) examines the normative statistics, criminogenic need profiles and the predictive validity estimates for Australian offenders using the LSI-R. Using more than 78,000 administrations spanning 4 years (2004-2007), this study explored LSI-R variation according to gender and differing types of sentence orders. The findings indicate that whilst male and female offenders do not differ on the LSI-R total score, idiosyncratic criminogenic need characteristics (i.e. LSI-R subscale differences) are apparent, with specific profiles also evident for offenders serving different sentence orders. This latter result is encouraging, suggesting the need for a third class of offenders in addition to the traditional community/custodial divisions. The predictive validity of the LSI-R was modest, with varying results for the different offender groups.

The third paper (Chapter 4) focuses on the sensitivity of the LSI-R with an overrepresented minority Australian offender population, namely, Indigenous offenders. With randomly matched non-Indigenous offenders (by gender), the results (N = 27, 822) indicate that Indigenous offenders generally score higher on every aspect of the LSI-R, with marked differences in criminogenic need characteristics as compared to non-Indigenous offenders. Gender differences within the Indigenous offender sample are also apparent. The discussion of these findings centres on the implications of criminogenic need profiles using generic risk assessments for minority offenders, which could exclude or ignore important factors such as embedded cultural differences between Indigenous and non-Indigenous offender groups.

The second and third paper (Chapters 3 and 4, respectively) raise concerns as to the stability of the latent structure of the LSI-R and its applicability for Australian offenders. Given that the origin of the LSI-R is Canadian, the final paper (Chapter 5) explores the relevancy of Canadian offender constructs for Australian offenders including Indigenous offenders. A review of the international literature reveals that not only are there very few studies that have explored the factor structure of this assessment tool, inconsistent findings are also apparent. Closer examination of these studies suggests inconsistencies may be the result of inappropriate statistical analyses and the inappropriate use of the LSI-R components (namely, the subscales). Correcting for these inappropriate statistical analysis and analysing the assessment tool at the item level, the newly revised, or recalibrated LSI-R, suggests factors, or subscales, with 28 items. The recalibrated LSI-R is then compared to the original version, in both subscale and total scores, for sensitivity and specificity predictions. The results indicate that across the different offender groups (by gender, Indigenous status and sentence orders served), the recalibrated version performs better with greater accuracies in the predictions of re-offending (and non re-offending). These results provide further evidence that constructs underlying generic risk assessments are not generally transferable across jurisdictions and should be reviewed and evaluated stringently. The potential implications for the use of the shorter assessment tool for Australian agencies in the care and management of offenders are also discussed.

The four papers from this thesis therefore suggest that there are specific criminogenic need profiles apparent for Australian offenders, especially when offenders are assigned to groups based on gender, Indigenous status and sentence orders served. The exploration of the dimensionality of the LSI-R provides insight into the inter-relations between the components, or factors, that contribute to the understanding and the prediction of risk and criminogenic need characteristics, as well as the potential risk in using an assessment tool not developed, normed and evaluated on the offender population of interest.

Although the findings from the thesis are encouraging, there exists a continuing need for further validation and investigation. Re-offending data from other judicial sources, for example, that include information such as police arrests and fines, may be helpful comparing between risk assessments (for example, predictive utility), as well as broadening current knowledge of the trends, characteristics and profiles in criminal behaviour.

02Whole.pdf (595 kB)