Laboratories of Democracy : Policy Experimentation under Decentralization ∗

We examine the extent of policy experimentation under decentralization and centralization. More speci cally, we study the incentive for experimentation in an environment where there is uncertainty about the payo of policies as well as the competence of policymakers. Voters form beliefs about the policymakers upon observing noisy signals. Policymakers are concerned about their reputation to the voters, where in the worst case, they are voted out. In equilibrium, policymakers adopt a greater variety of policies under decentralization than under centralization. This rationalizes the popular notion of policy laboratories. We also examine the welfare value of information on policies (policy learning) and on policymakers (electoral accountability). We show that district homogeneity is an important factor in determining voter welfare. In particular, decentralization delivers greater welfare than centralization when the districts are homogeneous, while the reverse holds when the districts are heterogeneous. ∗We thank Scott Ashworth, Daniel Diermeier, Georgy Egorov, Tim Feddersen, Matthew Jackson, Samir Mamadehussene, Roger Myerson, Wojciech Olszewski, Nicola Persico, Bruno Strulovici, Yiqing Xing and participants at the NSF/CEME Decentralization Conference 2015, the Econometric Society World Congress 2015, and Northwestern Theory Brown Bag Lunch for helpful comments. All errors are our own. †Kellogg School of Management, Northwestern University, [email protected] ‡Department of Economics, Northwestern University, [email protected] §Corresponding Author