Abstract
The relationship between changes in delinquency prevalence over time and prediction error are developed and discussed. Published data from a 21-year British longitudinal study are used to exemplify the formal relationships that are derived. Statistically, as prevalence increases, there will be a relative increase in false negative errors and a decrease in false positive errors; the relationship is independent of prediction accuracy. Substantively, this means that as deliquency prevalence increases, imperfect prediction models will move toward "missing" more actual delinquents (increasing false negatives) and "mistaking" fewer actual nondelinquents (decreasing false positives). Some characterizations of differentially weighting the costs of false negative and false positive errors are presented in terms of intervention and nonintervention policies and decisions.
Get full access to this article
View all access options for this article.
