The paper considers both the social costs of youth crime and the personal risks and costs borne by the criminals themselves. It reviews various hypotheses as to the determinants of crime and presents three new sets of estimates. The first set of regressions uses the National Longitudinal Survey of Youth to explore the correlates of crime at the individual level. The second analysis focuses on census tract-level homicide data for the city of Chicago over a 30-year period. These data help to better understand the influence of social factors and local labor market conditions on youth crime. The final data set is a State-level panel covering 15 years. With these three sets of estimates as the data base, the paper seeks to determine the extent to which observed fluctuations in the correlates of crime can explain the time series pattern of juvenile crime over the last three decades. The paper contains strong evidence that juvenile crime is responsive to punishment. Findings also provide "some of the most compelling empirical evidence" to date of deterrence (as opposed to incapacitation). The paper highlights the importance of factors such as gender, family environment, and cognitive ability in predicting criminal involvement. Notes, tables, figures, references