Regular Article
Consistent Bayesian Aggregation

https://doi.org/10.1006/jeth.1995.1044Get rights and content

Abstract

We investigate the aggregation of first nonatomic probabilities and second Savagean orderings, subject to the following consistency constraints: (i) the aggregate is a subjective probability or a Savagean ordering, respectively; and (ii) it satisfies the Pareto principle. Aggregation is viewed here as a single-profile exercise. We show that affine rules are the only solutions to the problem of aggregating nonatomic probabilities. Assuming weak Pareto conditions and technical restrictions, we show that dictatorial rules are the only solutions to the problem of aggregating Savagean orderings. No solution exists when strong Pareto replaces weaker conditions. Without technical restrictions of affine independence, nontrivial solutions may exist. Journal of Economic Literature Classification Numbers: D70, D81, C11.

References (0)

Cited by (151)

  • Bayesian social aggregation with accumulating evidence

    2022, Journal of Economic Theory
    Citation Excerpt :

    This raised a question: does Harsanyi's result remains valid when the individuals and the planner are Savage-style subjective expected utility maximizers, perhaps with different beliefs? In an influential paper, Mongin (1995) answered this question in the negative: in the Savage framework, the planner can satisfy the ex ante Pareto axiom if and only if all agents have the same beliefs. Since such homogeneity of beliefs is empirically implausible and perhaps even normatively undesirable, Mongin interpreted this result as an impossibility theorem.

  • Belief-averaging and relative utilitarianism

    2021, Journal of Economic Theory
    Citation Excerpt :

    They show that no social choice function satisfies the following properties: acts in the choice set are expected utility-maximizing for a belief and utility function of society, this belief (utility function) of society depends only on the agents' beliefs (utility functions), Pareto-dominated acts are never chosen, and no agent can dictate the belief. Much of the literature on single-profile results is inspired by Mongin (1995) and Gilboa et al. (2004) and focuses on the Pareto condition. Recall that Mongin proved that the Pareto condition can hold only if all agents have the same belief.

  • Smooth aggregation of Bayesian experts

    2021, Journal of Economic Theory
    Citation Excerpt :

    To preserve Theorem 1, one would have to directly make an assumption analogous to (3). Gajdos et al. (2008) obtain a negative aggregation result that generalizes Mongin (1995). They consider a planner and agents whose preferences belong to a large class that includes most known models of decision under uncertainty.

  • A new puzzle in the social evaluation of risk

    2022, Economics and Philosophy
View all citing articles on Scopus
View full text