We note and correct a flaw in the analysis of Ferejohn (1986)'s seminal model of electoral accountability. In the original solution, it is supposedly optimal for the voter to impose a stationary path of performance targets on officeholders. We show that, in fact, stationary paths are suboptimal; the voter can do better by choosing a path of performance targets that become increasingly lenient over time, which extracts more effort from the politician earlier on. We explicitly solve for the optimal performance targets for a class of examples and discuss the substantive implications of our exercise.