We investigate the relationship between employment and crime exploiting individual-level data on the universe of (male) workers and criminal cases filed in Brazil over the period 2009-16. We preliminarily document a wide gap in the probability of committing crimes by employment status. At the peak of the crime-age profile (21 years) the probability of committing crimes is twice as large for job losers than for employed workers. We then estimate the effect of losing a job on the probability of being prosecuted for having committed a crime exploiting variation in layoffs across individuals and over time. Difference-in-differences estimates allowing for dynamic and heterogenous treatment effects as well as for city and sector specific shocks suggest that the probability of being prosecuted for a crime increases by as much as 70 percent after losing a job. The effect is larger for economically-motivated crimes than for violent crimes, and for workers with lower salaries and lower tenure – and, thus, lower severance payments. These estimates are unaffected when focusing only on mass layoffs, thus reducing the scope for omitted confounders and reverse causation at the individual-level. These results are consistent with Becker’s rational model of crime. More generally, they provide the first individual-level analysis of the economic determinants on crime for a country with high levels of criminal activity. |
