Abstract
Behavioral economists increasingly argue that violations of rationality axioms provide a new rationale for paternalism - to "de-bias" individuals who exhibit errors, biases and other allegedly pathological psychological regularities associated with Tversky and Kahneman's (in Science 185:1124-1131, 1974) heuristics-and-biases program. The argument is flawed, however, in neglecting to distinguish aggregate from individual rationality. The aggregate consequences of departures from normative decision-making axioms may be Pareto-inferior or superior. Without a well-specified theory of aggregation, individual-level biases do not necessarily imply losses in efficiency. This paper considers the problem of using a social-welfare function to decide whether to regulate risk-taking behavior in a population whose individual-level behavior may or may not be consistent with expected utility maximization. According to the social-welfare objective, unregulated aggregate risk distributions resulting from non-maximizing behavior are often more acceptable (i.e., lead to a weaker rationale for paternalism) than population distributions generated by behavior that conforms to the standard axioms. Thus, psychological theories that depart from axiomatic decision-making norms do not necessarily strengthen the case for paternalism, and conformity with such norms is generally not an appropriate policy-making objective in itself.