We consider a dynamic game of electoral competition with adverse selection, moral hazard, and imperfect monitoring. We show that this dynamic game can be estimated using a flexible maximum likelihood estimator. We implement the estimator using data from recent mayoral elections in large US cities with binding two-term limits. Our empirical findings suggest that there are large differences in performance among different types of mayors. We find an economically important degree of policy responsiveness, with effort accounting for a larger fraction of the total effect than selection. Finally, we evaluate several institutional reforms that promise to increase policy responsiveness.