Shackled labor markets: Bounding the causal effects of criminal convictions in the U.S.

Abstract

This paper examines the causal effects of criminal convictions on labor market outcomes in young men using U.S. data from the National Longitudinal Survey of Youth 1997 cohort. Unlike previous research in this area which relies on assumptions strong enough to obtain point identification, this paper imposes relatively weak nonparametric assumptions that provide tight bounds on treatment effects. Even in the absence of a parametric model, under certain specifications, a zero effect can be ruled out, though after a bias correction this result is lost. In general the results for the effect on yearly earnings align well with previous findings, though the estimated effect on weeks worked are smaller than in previous findings which focused on the effects of incarceration. The bounds here indicate the penalty from convictions, but not incarceration, lowers weeks worked by at most 1.55 weeks for white men and at most 4 weeks for black men. Interestingly, when those ever incarcerated are removed from the treatment group for black men, there does not appear to be any effect of convictions on earnings or wages but only on weeks worked.