A Bayesian Framework for the Precautionary Principle

A common misconception is that precautionary motives in public policy cannot be justified within the Bayesian rational choice framework and that decision criteria that appeal to ambiguity and pessimism are needed. This paper critically evaluates these claims, arguing that a rational policy process must be Bayesian in order to avoid paradoxical, even absurd, recommendations, such as policies that depend on sunk costs or that suppress costless information. The paper also argues that the distinction between measurable risk and fundamental, or Knightian, uncertainty can be made within the standard framework of Bayesian rationality. Finally, a simple model is proposed to highlight situations in which precautionary action may be normatively justified.