Learning Through Policy Variation

Abstract

Rationalist analysis of policymaking, exemplified by cost-benefit analysis, ignores the variance in outcomes associated with policies and seeks to maximize expected outcomes. Burkeans, by contrast, view policy outcome uncertainty negatively. The Burkean approach is echoed in the precautionary principle, which argues that policies with hard-todetermine or high-variance outcomes should be avoided. Both approaches are the subject of vast literatures. This Article argues that both approaches are wrong. When policies can be reversed in future periods, variation in the outcomes associated with a policy is a good thing. Reversibility means that the downside risk of high-variance policies is limited; policies with unexpectedly bad outcomes can be changed in the next period. The upside of high-variance policies, by contrast, may last indefinitely, since policies with unexpectedly good outcomes will be retained. Thus, when policies are reversible, policymakers should deliberately choose policies with uncertain outcomes, other things equal. The Article also examines the assumption of policy reversibility. It shows that the most important source of irreversibility for policy analysis is irretrievable “sunk costs” rather than the potential for catastrophic outcomes or policy inertia. As a result, policies are more reversible than commonly appreciated. The Article then examines optimal policymaking under irreversibility. Under extreme irreversibility, conservatism of a particular sort, called the “real options” approach, constitutes the best policy. More generally, the Article argues that the appropriate attitude toward policy variance depends upon the reversibility of policy. This analysis illuminates many puzzles in constitutional law and institutional design, such as the puzzling difference between entrenched statutes, which are unconstitutional, and sunset clauses, which are permitted. The Article concludes with recommendations to encourage policymakers to use variance more effectively.