Thesis submitted in partial fulfilment of the requirements for the degree of Master of Science in Psychology at the University of Canterbury. Available for loan from the University of Canterbury library.

The main aims of the present study were to identify variables that predict future sex re-convictions among child sex offenders, and to develop a statistical model for predicting sexual re-convictions. The data was provided by the New Zealand Department of Corrections, and consisted of two cohorts of extrafamilial child sex offenders. The first cohort, the estimation sample, comprised criminal histories of 741 male sex offenders who were convicted between 1988 and 1997 of sexual offences against children. The second cohort, the validation sample, comprised criminal histories of 156 male sex offenders, who were convicted between 1983 and 1988 of sexual offences against children. The variables were grouped into three categories: offender's age and race, offence history, and victim's characteristics. The predicting power of each variable was tested against re-conviction data in the estimation sample using Cox regression. Next, Cox regression and logistic regression prediction models were constructed utilising those variables that predicted re-conviction. The predicting power of both models was then tested and adjusted against the validation sample. Based on Cox regressions and logistic regression analysis, five significant covariate predictor variables were identified (in order of magnitude): preference for a male victim(s), diversity of offending based on criterion convictions, presence of a female victim under the age of 12, race, and age at first sexual conviction. However, both presence of a female victim under the age of 12 and diversity of offending did not show a similar relationship with re-convictions in the validation sample and thus were not included in the final prediction models. The final adjusted prediction models included three predictor variables (in order of magnitude): sex preference for male victim, race, and age at first sex conviction. The above models indicated that those offenders who were convicted for offences against males only, who were white, and who were younger at first sex conviction were most likely to be re-convicted in the future. The results of this study demonstrate that, for some samples of child sex offenders, a number of variables significantly correlate with future re-convictions and thus can be used to predict likelihood of sexual recidivism.--AUTHOR'S ABSTRACT