Electoral Competition, Control and Learning∗

Abstract. We study an agency model that incorporates three world elements: voters learn about an incumbent’s competence from his record in office, and about his opponent via a campaign; the former is more informative than the latter; but it can be costly for the voter to learn about the incumbent when policies he implements are risky and go wrong. In our two period model an incumbent chooses between implementing a risky reform policy or a safe option in each period. The former is successful only if the incumbent is competent. After the first period policy has been implemented an opponent can choose whether to inform the voter of his competence. The first best level of reform balances the associated costs/benefits with optimal learning. The voter has a simple retention rule that she uses to control the incumbent by aligning his incentives with her own and to learn about his competence. In the absence of an informative campaign by the opponent the voter can attain first best, though not in its presence: then the degree of reform is too high from the voter’s perspective; this can have negative consequences on welfare. These inefficiencies arise due to the different communication technologies available to the incumbent and his opponent. Our central results are robust to subtle changes in our core model specification