Separating state dependence, experience, and heterogeneity in a model of youth crime and education

We study the determinants of youth crime using a dynamic discrete choice model of crime and education. We allow past education and criminal activities to affect current crime and educational decisions. We take advantage of a rich panel dataset on serious juvenile offenders, the Pathways to Desistance. Using a series of psychometric tests, we estimate a model of cognitive and social/emotional skills which feed into the crime and education model. This allows us to separately identify the roles of state dependence, returns to experience, and heterogeneity in driving crime and enrollment decisions among youth. We find small effects of experience and stronger evidence of state dependence and heterogeneity for crime and schooling. We provide evidence that, as a consequence, policies that affect individual heterogeneity (e.g., social/emotional skills), and those that temporarily keep youth away from crime, can have important and lasting effects even if criminal experience has already accumulated.