Key points are not available for this paper at this time.
The conventional wisdom is that government decentralization promotes policy innovation because it allows for several simultaneous experiments by local governments. However, this ignores a learning externality: successful policy experiments provide useful information for all governments. Local governments will ignore this externality, but a central government should take it into account. This article uses a social learning model to compare policy innovation under centralization and decentralization. Centralization leads to more policy innovation if the local governments are relatively homogeneous or large in number. However, decentralization may induce more policy innovation if there are multiple experimental policies available.
Koleman Strumpf (Mon,) studied this question.