Incorporating implicit knowledge into the Bayesian model of prior conviction evidence: some reality checks for the theory of comparative propensity

Abstract

The theory of comparative propensity, championed by the late Mike Redmayne, has been an influential theory underpinning normative models of the probative value of evidence of previous convictions in criminal trials. It purports to generalize an approximate probative value by means of a Bayesian model in which the likelihood of an innocent person having a criminal record is calculated by reference to general population statistics, and the hard evidence underpinning the prior probability is treated as unknown. The theory has been criticized on the ground that it fails to take account of bias against past offenders in the selection of cases for prosecution. This article analyses the model and these criticisms and concludes that both the model and the criticisms are flawed because they fail to address the evidence on which the prior odds are based. We find that, not only are such mathematical models unsound, but they can only be ‘repaired’ by making assumptions about the typical case which run counter to the legal presumption of innocence. Analysing the flaws in these models, however, does provide some insight into issues affecting the value of prior convictions evidence.