Searching for Good Policies : Repeated Elections , Learning , and Policy Dynamics ∗

I study a model of dynamic policy making in which citizens do not have complete knowledge of how policies are mapped into outcomes. Citizens are able to order policies in the single-dimensional space according to the outcomes they are expected to produce but not according to the realized outcomes they do produce. They learn about the mapping through repeated elections as policies are implemented and outcomes observed. I characterize for this environment the policy trajectory with impatient voters. Although the trajectory is path-dependent, I show that basic patterns emerge. Policy making passes sequentially through three phases: a monotonic phase, a triangulating phase, and ultimately a stable phase. Policy making generally stabilizes at outcomes close to the median voter’s preference, although I show that it can at times get stuck, in which case any outcome can prove stable. I also consider how the structure of political competition affects experimentation and learning. ∗Preliminary version. Comments welcome. I thank Martin Osborne and Abhinay Muthoo for helpful discussions. †Managerial Economics and Decision Sciences, Kellogg School of Management, Northwestern University, Evanston, IL 60208; [email protected].