A repeated moral hazard setting in which the Principal privately observes the Agentfs output is studied. It is shown that there is no loss from restricting the analysis to contracts in which the Agent is supposed to exert effort every period, receives a constant efficiency wage and no feedback...