Linking crimes is a decision making problem faced by investigators and practitioners. Deciding if a number of crimes are committed by a single offender (linked) or multiple offenders (unlinked) has implications for how investigations are structured and impacts on the success of the investigative outcome. The police are sometimes criticised for not making seemingly 'obvious' links between offences. However viewed prospectively the linking problem is far more difficult. It is not obvious what type of information should be used in order to make a correct linking decision. The police deal with large volumes of information, sometimes over long periods of time, not all of which may be relevant to the linking task. In order to successfully link crimes, offenders must behave with consistency. Consistency has two components: similarity and distinctiveness. Offenders must demonstrate a degree of similarity across consecutive offences and must also exhibit behaviours distinct from that of other offenders. Differentiation is the product of consistency. The more consistent an offender behaves the more readily their crimes can be differentiated from those of other offenders. Employing binary logistic regression and ROC analysis, the literature on crime linkage or 'Linkage Analysis', reports encouraging findings across a variety of crime types. Offenders are universally found to be highly consistent in their spatial behaviour, resulting in excellent discriminative accuracy. Some crime types also attract highly consistent control type behaviours. Research demonstrates the existence of consistency and reports variation in consistency and differentiation, as a product of the type of offence behaviour employed in making the linking decision. Although important steps have been made in determining that behavioural consistency exists the vast majority of research has neglected to consider the conditions that foster consistency. This is curious considering the current depth of theoretical knowledge concerning consistent behavioural expression. Extensive research in the personality literature supports the premise that behavioural expression is contingent upon individual and situational factors. Evidence suggests that similar situations are likely to give rise to consistent behaviour, while dissimilar situations are not. Further conditions of consistency are also evidenced in the personality literature: events that are temporally proximate tend to be more consistent; and frequent exposure to a situation results in increased behavioural similarity. In an attempt to incorporate the current theory of behavioural consistency with crime linkage analysis, the present study tests three conditions of consistency on a naturally occurring forensic sample, serial residential burglary. The first condition is situational factors. It is hypothesised that consecutive offences committed in similar situations will be more consistent than consecutive offences committed in dissimilar situations. The second condition is temporal proximity. It is hypothesised that greater consistency will be expressed in consecutive offences that are temporally proximate. The third condition is the effect of experience. It is hypothesised that consecutive offences committed at the end of prolific series will be more consistent than consecutive offences committed at the beginning. Abstract: The sample comprises 1991 residential burglary offences committed in the Lancashire area by 414 offenders. Data includes spatial information in the form of offence location geo codes and temporal information comprising date of offence. Binary behavioural data include: Planning and Target behaviours; Entry Behaviours; Internal Behaviours; Exit and Escape behaviours; and Stolen Property information. Similarity scores are calculated between linked and unlinked crime pairs. Inter crime distances comprise the similarity scores for spatial behaviour. Inter crime time differences comprise the similarity score for temporal behaviour and Jaccard's correlation coefficients for Behavioural information. Similarity scores provide the basis on which binary logistic regression models are constructed. Regression models ascertain the predictive accuracy of behavioural domains. In order to ascertain which combination of predictors combined to produce the most predictive model, optimal binary logistic regression models are constructed. The predicted probabilities of the binary logistic regression models provide the measure upon which ROC analysis models are constructed. ROC analysis provides a separate measure of consistency and discriminative accuracy, allowing for improved utility of the measure. A baseline linkage model is constructed using the first two crimes of each series. In order to test the first condition of consistency, situational similarity, the sample is divided into subsamples relating to varying situational contexts. Eight burglary 'contexts' are identified, namely: occupied and unoccupied offences; offences where the victim is known to the offender 'and offences where the victim is a stranger; offences which were disturbed and those which were not; and lastly offences committed in conjunction with a eo-offender and those committed alone. A linkage analysis is constructed for each of the situational subsamples. Subsamples are characterised by consecutive offences committed in similar situations. The linkage analysis of each situational subsample is compared to that of the baseline model. The baseline model is constructed from a sample characterised by dissimilar consecutive offences. In line with the hypothesis, results indicate an overall increase in consistency as a product of situational similarity. The second condition of consistency, temporal proximity, is tested by correlating similarity scores and inter crime time differences. In order to control for the variance introduced by situational factors, the analysis is conducted for the complete sample and also within each of the situational subsamples. In line with the hypothesis, a statistically significant relationship is observed between temporal proximity and behaviour and spatial similarity. Furthermore by controlling for the influence of situational variance, the associations are slightly exaggerated. The third condition of consistency, effect of experience, is tested by comparing the similarity scores of early and late career offences in prolific offenders. In order to control for the variance introduced by situational factors, the analysis is conducted on the full sample and, sufficient sample size permitting, the situational subsamples. In all samples, statistically significantly greater consistency is observed in late career offences of combined behaviours, planning and target behaviours and temporal behaviour. Minimal increases in differences are observed as a product of controlling for situational factors. The findings of this research resonate with both the forensic and personality literature. The operational implications for Linkage analysis are discussed and strategies for future research presented.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:569260 |
| Date | January 2012 |
| Creators | Goodwin, Louise |
| Publisher | University of Liverpool |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0026 seconds