Criminal law and economics rests on the expectation that deterrence incentives can be employed to reduce crime. Prison survey evidence however suggests that a majority of criminals are biased and may not react to deterrence incentives. This study employs an extra-laboratory experiment with criminals in a German prison to test the effectiveness of deterrence and compares it with data of student subjects. Subjects either face potential punishment when stealing, or they can steal without deterrence. We confirm Gary Becker’s deterrence hypothesis that deterrence works for criminals (and similarly for students). We observe significantly more risk-seeking criminals than students, although the vast majority (80.77 %) of criminals behaves risk-neutral or risk-averse.