From the Works Progress Administration of the New Deal to the Job Corps of the Great Society era, employment programs have been advanced to fight poverty and social disorder. In today's context of stubborn unemployment and neoliberal policy change, supported work programs are once more on the policy agenda. This article asks whether work reduces crime and drug use among heavy substance users. And, if so, whether it is the income from the job that makes a difference, or something else. Using the nation's largest randomized job experiment, we first estimate the treatment effects of a basic work opportunity and then partition these effects into their economic and extra-economic components, using a logit decomposition technique generalized to event history analysis. We then interview young adults leaving drug treatment to learn whether and how they combine work with active substance use, elaborating the experiment's implications. Although supported employment fails to reduce cocaine or heroin use, we find clear experimental evidence that a basic work opportunity reduces predatory economic crime, consistent with classic criminological theory and contemporary models of harm reduction. The rate of robbery and burglary arrests fell by approximately 46 percent for the work treatment group relative to the control group, with income accounting for a significant share of the effect.