A simple model of yardstick competition between jurisdictions is presented. Governments of jurisdictions face the alternative to choose between an old and a new policy with stochastic payoffs. The new policy is superior to the old policy in one state of the world, and inferior in the other. Governments are either benevolent, serving the interest of the voter, or rent-seeking. An equilibrium with yardstick competition is shown to exist where bad governments having a good government in their neighborhood choose the new policy more often compared to an equilibrium without relative performance evaluation. Overall, the probability of policy innovations is increased by yardstick competition. The model has a testable empirical implication saying that policy innovations should show spatial correlation.