The citizen-candidate approach, proposed to study the performance of representative democracies, builds on a multi-stage game where the same agents are asked whether or not to become a candidate and, successively, to vote. Consistently, the solution concept adopted in Besley and Coate (1997) conforms to backward induction rationality. In this note we remark that it does not conform to forward induction rationality. Some results on stable sets are then obtained.