We present a rational theory of reform fatigue. At each instant a politician chooses to divide effort between reforms and the status quo, and this choice is modeled as a two-armed bandit problem. Reforms are expected to yield a higher rate of output to the voter than the status quo conditional on the politician being competent. We interpret competence as the administrative ability to ensure successful implementation of reforms. The politician’s competence is therefore unknown ex-ante to both the politician and the voter. In addition the voter is unable to observe the politician’s effort on reform, but only observes aggregate output. In equilibrium the voter gives the politician endogenous term lengths that depend on the timing of success. The executive experiments with reforms at the beginning of his first term, but gradually decreases the rate of reforms in the absence of early success. We call this gradual reduction in experimentation reform fatigue. The theory thus predicts that reform fatigue follows a political cycle. We provide empirical evidence of reform fatigue cycles in financial policies among presidential countries.