Ruling Narrowly: Learning and Law Creation

I develop a dynamic model of law creation in which the court is uncertain about the ideal rule. The court learns about the ideal rule through the cases it hears, which are in turn the result of rational choices of agents responding to the court's previous decisions. Learning requires experimentation, and since agents choose optimally, learning is only possible experimentation is incentive compatible for the agent. The court provides incentives to the agent by setting penalties, and writing opinions that commit the court to sanctioning or punishing various actions. The model generates several predictions. First, the efficacy of opinion writing is asymmetric - the court has an incentive to write broad permissive opinions, but no cor- responding incentive to write broad restrictive opinions. Second, the court's learning is inefficient - it does not induce learning that minimizes the expected future cost of uncertainty. Instead, the court will induce experimentation that increases the likelihood that it can amend its permissive opinion - since this policy tool is more efficacious. Third, since the court cannot always amend an opinion, it has a incentive to preemptively write broad opinions.