Investigating whether a crime reduction measure works

Abstract

Crime is a serious business. It causes great distress and fear. It costs a lot to deal with its consequences. In these regards crime shares much with the problem of ill-health and disease. The application of sound science and statistics has allowed great strides to be made in dealing with problems of ill health. Medical statistics is one of the recognised, established disciplines involved in researching healthcare.

The parallels between research in crime reduction and in healthcare do appear to differ in terms of quality. Although there is still room for considerable improvement in researching health-care, an investigation into the underpinning of statistical methods used indicates that the problems are substantially worse in the study of crime. The consideration given to statistics in crime studies seems rather flimsy, yet important claims are made which are statistical at source and may affect policy, and so can have considerable costs attached. Therefore, for example, it is important to know whether the underlying crime level has really changed, rather than just being the result of perhaps sampling variation or some artefact giving rise to statistical bias or systematic error. This is necessary when trying to determine whether a Crime Reduction Intervention (CRI) has actually worked.

I started examining the scientific basis of the claim for the effectiveness for one particular CRI, basically because I was concerned about negative side effects and I thought the claim implausible. I remain concerned and unconvinced. The statistical issues and concerns I raise apply also to investigating other CRIs and to existing published analyses.

This piece extends work presented in Marchant (2006); earlier work on the statistical issues involved can be found in Marchant (2005a, b; 2004).

The full text can is available here.