Diversified Policy Choice with Partial Knowledge of Policy Effectiveness (WP-09-02)
Charles F. ManskiAn important objective of policy research is to provide information useful in choosing new policies. Consider a planner, who must choose treatments for the members of a population. Policy analysts often ask how a planner should act. A standard exercise specifies a set of feasible treatment policies and a welfare function. The planner is presumed to know how persons respond to treatment. The goal is to characterize the optimal policy. Unfortunately, the practical relevance of this exercise is limited by the fact that available research typically yields only partial knowledge of treatment response. Hence, planners cannot determine optimal policies. Instead, they must choose treatments under ambiguity. This paper explains why research typically provides only part of the knowledge needed to choose optimal policies. Manski shows how planners can cope with ambiguity, making reasonable policy choices with the knowledge available. He also discusses how we can reduce ambiguity, enabling better policy choices.