Detecting Serial Rape The Role of Offence Behaviours in Case Linkage

  • Serena Davidson

Student thesis: Doctoral Thesis

Abstract

Rape is an under reported and under-researched issue within Australia,causing serious harm to victims and society, with serial stranger rape presenting unique investigative challenges. Behavioural investigative tools and techniques such as case linkage have been developed to assist proactive policing in the early identification of serial offenders. Case linkage is based upon the principles of offender consistency and distinctiveness, which state that a serial offender will remain relatively consistent across his or her offences, yet distinct compared to other serial and non-serial offenders. However, the understanding of serial versus non-serial rapist behaviours generally and within Australia specifically is limited.

This thesis examined the behaviours of both serial and non-serial offenders across 250 stranger rapes extracted from Queensland Police Service databases. The purpose of this examination was threefold. First, to determine whether serial rapists engage in behaviours that are distinct to non-serial rapists. Second,to contribute to practical rape investigation by examining whether offence behaviours can be used to discriminate between serial and non-serial rape offences. Finally, to test case linkage theory by exploring whether serial rapists display offender consistency and distinctiveness.

Chi square analysed offence behaviours of both serial and non-serial rapists. Twenty-four significant variables were then included in a binary logistic regression to determine whether serial and non-serial offences could be distinguished. To examine case linkage theory and practice, all possible crime pairings were created, and cross-crime similarity was assessed using Jaccard’s coefficient. Receiver Operating Characteristics (ROC) analysis was used to examine the ability to distinguish between linked and un-linked offence pairs.

Sixty-seven variables were found to be significantly different between serial and non-serial rape offences. Of those, 24 were included in the logistic regression, chosen based on phi values and previous research. Eight variables contributed significantly to the model, which correctly classified 87.8%of offences. Through an examination of cross-crime similarity coefficients between serial linked, non-serial unlinked, and serial unlinked offence pairs, serial linked offences were found to have the highest Jaccard’s coefficient (M = .46, SD = .14).Two ROC analyses were run on serial-only offence pairs and all offenders' offence pairs, resulting in an AUC of .919 and .913, respectively. This indicates an excellent level of accurate discrimination between linked and unlinked cases based on Jaccard’s coefficient.

These findings have both academic and practical implications. By supporting case linkage theory, this thesis contributes to the growing body of knowledge and evidence highlighting the usefulness of case linkage as an investigative tool. Furthermore, by identifying unique and predictive behaviours of serial rapists, these findings have positive implications for investigations. The early identification of serial rapists can contribute to jurisdictional collaboration and the minimisation of further victimisation.
Date of Award21 Oct 2020
Original languageEnglish
SupervisorWayne Petherick (Supervisor) & Terry Goldsworthy (Supervisor)

Cite this

'