A growing literature suggests that juvenile arrests perpetuate offending and increase the likelihood of future arrests. The effect on subsequent arrests is generally regarded as a product of the perpetuation of criminal offending. However, increased rearrest also may reflect differential law enforcement behavior. Using longitudinal data from the Project on Human Development in Chicago Neighborhoods (PHDCN) together with official arrest records, the current study estimates the effects of first arrests on both reoffending and rearrest. Propensity score methods were used to control differences between arrestees and nonarrestees and to minimize selection bias. Among 1,249 PHDCN youths, 58 individuals were first arrested during the study period; 43 of these arrestees were successfully matched to 126 control cases that were equivalent on a broad set of individual, family, peer, and neighborhood factors. We find that first arrests increased the likelihood of both subsequent offending and subsequent arrest, through separate processes. The effects on rearrest are substantially greater and are largely independent of the effects on reoffending, which suggests that labels trigger "secondary sanctioning" processes distinct from secondary deviance processes. Attempts to ameliorate deleterious labeling effects should include efforts to dampen their escalating punitive effects on societal responses.

Most deterrence research has investigated how perceptions about sanction threats influence decisions to offend. Far less scholarship has investigated the processes in which sanction threat perceptions are formed and modified. In this study, we advance and test a theoretical framework in which perceptions of the certainty of punishment are a function of the offending experiences and consequences of both the actor and others. Some of the empirical implications of this framework are tested with data from the National Youth Survey. The findings include: (1) Arrests had little effect on perceptions of the certainty of punishment for stealing and attacking; (2) In contrast, offending corresponded with decreases in the perceived certainty of punishment for both offenses; (3) Peer offending produced decreases in the perceived certainty for stealing, but not for attacking; (4) Prior offending experience did not diminish the influence of more immediate offending experience on risk perceptions; and (5) Moral inhibition reduced the effects of offending experience on risk perceptions. The implications are discussed for refining theories of offender decision-making.Greg Pogarsky is an Associate Professor in the School of Criminal Justice at the University at Albany. His research focuses on refining models of offender decision-making by revisiting their assumptions about human nature and by appealing to recent advancements on human decision-making in economics and psychology. KiDeuk Kim is a doctoral student in the School of Criminal Justice at the University at Albany. His current research interests include rational choice theories, perceptions in the criminal justice system, and quantitative methodology. Ray Paternoster is a Professor of Criminology at the University of Maryland. His research interests include criminological theory, quantitative methods in criminology, and the death penalty. Correspondence to:

Recent research has produced mixed results as to whether newer machine learning algorithms outperform older, more traditional methods such as logistic regression in predicting recidivism. In this study, we compared the performance of 12 supervised learning algorithms to predict recidivism among offenders released from Minnesota prisons. Using multiple predictive validity metrics, we assessed the performance of these algorithms across varying sample sizes, recidivism base rates, and number of predictors in the data set. The newer machine learning algorithms generally yielded better predictive validity results. LogitBoost had the best overall performance, followed by Random forests, MultiBoosting, bagged trees, and logistic model trees. Still, the gap between the best and worst algorithms was relatively modest, and none of the methods performed the best in each of the 10 scenarios we examined. The results suggest that multiple methods, including machine learning algorithms, should be considered in the development of recidivism risk assessment instruments.

No abstract

No abstract