In this paper, we analyze the effect of the criminal justice system on juvenile recidivism. Using a unique sample of German inmates, we are able to disentangle the selection into criminal and juvenile law from the subsequent recidivism decision of the inmate. We base our identification strategy on two distinct methods. First, we jointly estimate selection and recidivism in a bivariate probit model. In a second step, we use a discontinuity in law assignment created by German legislation and apply a (fuzzy) regression discontinuity design. In contrast to the bulk of the literature, which mainly relies on US data, we do not find that the application of criminal law increases juvenile recidivism. Rather, our results suggest that sentencing adolescents as adults reduces recidivism in Germany.