Static risk assessment of sexual offenders in New Zealand: predictive accuracy, classification of risk, and the moderating effect of time offence-free in the community (2018)
AuthorsMoore, Lucyshow all
Sexual offending is regarded, both legally and morally, as among the most serious of crimes; therefore, accurate identification and reduction of recidivism risk for offenders who commit these crimes should be given considerable attention. The research in this thesis investigates whether any improvements can be made to the current sexual risk assessment measure for sexual offenders in New Zealand (the ASRS-R), and whether the categorisation of risk obtained by the static risk measure maintains the same level of accuracy once an offender has remained in the community without committing further offences after a certain period of time. Both of the studies undertaken in this thesis utilized the same large cohort of offenders; every individual convicted of a sexual offence that was released from a New Zealand prison between 1st January 1992 and 31st December 2002 (N = 5,895).
Study 1 attempted to create a static risk model that could provide comparable predictive accuracy to the ASRS-R, while removing some of the variables that have previously been found not to be significantly predictive of sexual recidivism. To generate the model, a cross-validation strategy was used; the model developed using one half of the sample, and validated using the other half of the sample. The computer-scored version of the model, named the Communicable Risk Measure for Sexual Offences (CRMSO), was able to provide similar predictive accuracy for sexual recidivism to the ASRS-R, while using fewer variables, obtaining AUC values of .712, .686, and .690 for 5-year, 10-year, and total sexual recidivism. The correlation between overall scores for the CRMSO and ASRS-R was r = .88 (p < .01). Additionally, the ability of the CRMSO to classify the relative risk of offenders was investigated, initially using the same 4 categories used in the ASRS-R, and then using a 5-category version, based on the guidelines proposed by Hanson, Babchishin, Helmus, Thornton and Phenix (2017), which also utilised the alternative risk category labels to standardize the communication of risk for decision-makers. The CRMSO demonstrated an increased sensitivity to those at the highest risk of recidivism when using the 4-category version, and the inclusion of a fifth category identified a very low-risk group of offenders; both of which have important implications for offender management.
Study 2 aimed to identify whether the level of risk attached to an offender at the time of their release remained predictive of their propensity to commit further offences after they had spent time offence-free in the community. Study 2 extended the previous research of Hanson, Harris, Helmus and Thornton (2014), who found that high-risk sexual offenders remained high-risk over time. Specifically, Study 2 extended the findings of Hanson et. al. (2014) to a nationwide cohort of offenders, and investigated whether violent and general desistance was also occurring, as well as desistance from sexual offending. The results from Study 2 indicated that the sexual offenders were in fact desisting from all types of criminal offending equally, and this finding was especially strong for the high-risk offenders. Overall, the findings of this research indicate that the classification and communication of static risk can be improved upon, but the overall predictive accuracy of different static risk measures remains stable. Additionally, it is apparent that static risk may be a valid, but time-dependent indicator of an individual’s propensity to commit further offences, and may affect sexual, violent, and general recidivism equally.