We provide an axiomatization of generalized utilitarian social welfare functions in the context of Harsanyi’s impartial observer theorem. To do this, we reformulate Harsanyi’s problem such that lotteries over identity (accidents of birth) and lotteries over outcomes (life chances) are independent. We show how to accommodate (first) Diamond’s critique concerning fairness and (second) Pattanaik’s critique concerning differing attitudes toward risk. In each case, we show what separates them from Harsanyi by showing what extra axioms return us to Harsanyi. Thus we provide two new axiomatizations of Harsanyi’s utilitarianism.
Keywords: D63, D71
JEL Classification: Generalized utilitarianism, Impartial observer, Social welfare function, Fairness, Ex ante egalitarianism
Savage motivated his Sure Thing Principle by arguing that, whenever an act would be preferred if an event obtains and preferred if that event did not obtain, then it should be preferred overall. The idea that it should be possible to decompose and recompose decision problems in this way has normative appeal. We show, however, that it does not require the full separability across events implicit in Savage’s axiom. We formulate a weaker axiom that suffices for decomposability, and show that this implies an implicit additive representation. Our decomposability property makes local necessary conditions for optimality, globally sufficient. Thus, it is useful in computing optimal acts. It also enables Nash behavior in games of incomplete information to be decentralized to the agent-normal form. None of these results rely on probabilistic sophistication; indeed, our axiom is consistent with the Ellsberg paradox. If we assume probabilistic sophistication, however, then the axiom holds if and only if the agent’s induced preferences over lotteries satisfy betweenness.
We provide necessary and sufficient conditions for a dynamically consistent agent always to prefer more informative signals (in single-agent problems). These conditions do not imply recursivity, reduction or independence. We provide a simple definition of dynamically consistent behavior, and we discuss whether an intrinsic information lover (say, an anxious person) is likely to be dynamically consistent.
We compare the Skiadas approach with the standard Savage framework of choice under uncertainty. At first glance, properties of Skiadas “conditional preferences” such as coherence and disappointment seem analogous to similarly motivated notions of decomposability and disappointment aversion defined on Savage “ex ante preferences.” We show, however, that coherence per se places almost no restriction on the structure of ex ante preferences. Coherence is an `external’ restriction across preferences whereas notions of decomposability in the Savage framework are ‘internal’ to the particular preference relation. Similarly, standard notions of disappointment aversion refer to ‘within act’ disappointments. Skiadas’s notion of disappointment aversion for families of conditional preference relations neither implies nor is implied by standard notions of disappointment aversion for ex ante preferences.
What is the relationship between an agent’s attitude towards information, and her attitude towards risk? If an agent always prefers more information, does this imply that she obeys the independence axiom? We provide a substitution property on preferences that is equivalent to the agent (intrinsically) liking information in the absence of contingent choices. We use this property to explore both questions, first in general, then for recursive smooth preferences, and then in specific recursive non-expected utility models. Given smoothness, for both the rank dependence and betweenness models, if an agent is information-loving then her preferences can depart from Kreps and Porteus’s (1978) temporal expected utility model in at most one stage. This result does not extend to quadratic utility. Finally, we give several conditions such that, provided the agent intrinsically likes information, Blackwell’s (1953) result holds; that is, she will always prefer more informative signals, whether or not she can condition her subsequent behavior on the signal.