The validity of reconviction as a proxy measure for re-offending: Interpreting risk measures and research in the light of false convictions and detection and conviction evasion skills (DACES) and processes

Abstract

Risk assessment tools and intervention efficacy evaluations typically use reconviction as an outcome that is assumed to be a valid measure of the return to offending (RTO). Reconviction is however problematic as a measure of RTO because a significant amount of offending goes unreported, undetected and/or unconvicted. The consequences and implications of this measurement problem are significant for the forensic practitioner. In this chapter we outline the nature of this problem, highlighting one of the key differences between clinical formulation and actuarial assessment being that the former develops a causal model of offending behaviour whilst the latter is a largely atheoretical statistical account of factors correlating with reconviction (which is fundamentally different from RTO). We explore how clinical judgement may be predicting RTO, whereas actuarial assessment predicts reconviction (a smaller subset of those re-offending). The literature supports the idea that biases, such as racism and unequal detection and conviction rates for different groups of people, underpin convictions which are inevitably “baked in” (e.g., Mayson, 2019) to actuarial assessment; hence risk assessments are predicting outcomes that can be biased. The need to assess individual and systemic detection and conviction evasion skills and processes as part of assessment is highlighted, and a preliminary model for analysing systemic detection and conviction evasion skills and processes is presented. The importance of specifying a measurement model before interpreting reconviction as a “valid” outcome measure is highlighted