Policy gambles and valence in elections

We study strategic policy experimentation by an incumbent politician when voters care about both policy and the candidates' valence. In our model, the voter does not know the location of her ideal policy and learns via experience, in turn, the officeholder uses policymaking to control voter learning. The incumbent thus faces a trade-off between implementing a policy close to his own ideal point, or one that induces the optimal amount of voter learning to win reelection. In equilibrium, how the incumbent solves the trade-off depends on his expected valence. We find that a trailing incumbent sometimes implements a safer policy than he would absent electoral incentives, despite needing to generate new information to win the election. Furthermore, increasing the incumbent's expected valence (and thus electoral advantage) can motivate him to gamble more in equilibrium. However, this relationship between electoral security and experimentation depends crucially on the amount of uncertainty on the valence dimension.