An Examination of the Dynamic Predictive Validity Thesis

The “dynamic predictive validity thesis” is a thesis regarding the properties of offender risk assessment instruments that was developed over the space of about a decade by Andrews, Robinson, Bonta, and Motiuk (Andrews & Robinson, 1984). This thesis became a popular measure of the quality of offender risk assessment instruments and dynamic predictive validitywas listed as a required attribute by several state level criminal justice organizations. The present article will demonstrate that the dynamic predictive validity thesis, as it is currently being practiced, is both conceptually and methodologically flawed. Moreover, it will be demonstrated that dynamic predictive validity, as it is currently being measured, is probably an indicator of poor performance for an offender risk assessment instrument. It appears that anoffender risk assessment with dynamic predictive validity may simply be too subjective and is in need of simplification to avoid subjectivity in rating. These two issues will be examined using evidence collected from offender reassessment scores and the principles of psychometric theory. The logic model behind the dynamic predictive validity thesis is simple and appealing. The model proposes that criminal offender recidivism risk is “dynamic” and that the predictive validity of offender risk assessment instruments improves between two points in time because the second assessment measures the “newer” risk level more accurately than the “older” risk assessment. This thesis is flawed on two levels. First, the concept of “dynamic” is flawed. The propensity for recidivism risk is much more dynamic that previously supposed. Second, tests of dynamic predictive validity are confounded by the potential for changes in two separate measures, a) actual recidivism risk, and b) perceived recidivism risk. An expanded examination of the data suggests that dynamic predictive validity is a function of changes in perceived risk by raters rather than systematic improvement in risk rating accuracy through reassessment.