Prediction of crime and early interventions

This paper presents a prospective longitudinal study that attempts to predict juvenile delinquency measured by first contact with the police (arrest, pre-trial detention or charges of crimes) taking a complete cohort of all children born in Denmark in 1984 (N=54,458). The children are followed from birth to early adulthood in 2006. The predictors represent the major crime reduction paradigms, such as family circumstance and individual skills. A discrete-time Cox model is used to allow for changing covariates over time. The population had 6,075 first time contacts with the police over the 300,591 person-years available. More than twenty risk factors were significantly predicting first-time contact with the police. Predictions were substantially more accurate than chance. Results also showed that the proportion of ‘false-positives’ were about 77 % of estimated high-risk subjects when observing the following calendar year. One out of eight high-risk youths will not get in contact with the police during any of the years under observation.