RATIONAL OVERCONFIDENCE AND SOCIAL SECURITY: SUBJECTIVE BELIEFS, OBJECTIVE WELFARE.

We propose general principles for policy decisions in the context of diverse beliefs, principles which are strictly between those used in the traditional neoclassical approach and those championed by behavioral economics. Like the neoclassical approach to policy choices our principles are based on an assumption of individual rationality but broadens the requirements for efficiency to also include {\it societal rationality}. Rational overconfidence is present when equally informed agents hold diverse, confident rational beliefs. The fact that beliefs are diverse means that all of them cannot be correct, hence seen as a collective agents do not necessarily act optimally. In the presence of rational overconfidence, Pareto efficiency is no longer the natural criterion for comparing policies and we in stead propose a strengthened version of ex-post welfare optimality, according to which welfare evaluations are objective. We apply our general principles to a particular policy issue, namely whether Social Security increases social welfare. We show that an assumption of bounded rationality is not needed to explain Social Security and other mandatory pension plans when rational agents hold inconsistent expectations about their future wealth and productivity or about the returns to their investments. Based on rational overconfidence, we provide a rationale for two of the features that distinguish Social Security and many other state mandated pension plans around the world: (i) that a minimum level of savings for retirement is imposed on most citizens and (ii) that individuals cannot freely decide how to invest their contributions.