Are command systems that rest on coercion inherently unstable, and did the Soviet economy collapse for this reason? Postwar evidence is inconsistent with the hypothesis that the Soviet economy was unstable. If it was not unstable, why did it collapse? A repeated game of coordination between a dictator and producers shows that a high level of coercion may yield a stable high–output equilibrium, that the command economy contains a time–consistency problem for central planners, and that a transition to a low state of coercion and performance in which everyone’s income falls may be brought about by rising monitoring costs and the dictator’s loss of reputation. The facts of the Soviet case are consistent with a collapse triggered when the dictator threw in the towel.