Distributive Justice for Behavioral Welfare Economics
AbstractSuppose the conflicting decisions of an individual at different decision-making frames are aggregated by unanimity into one preference. For example, an agent subject to status quo bias could place a value on a good that varies with the agent’s endowment. The behavioral preference that aggregates by unanimity would then judge the individual to be better off with the good only if all of the valuations exceed the agent’s payment. Since such behavioral preferences will typically be incomplete, they lead to a large set of Pareto optimal allocations: policymaking thus becomes problematic. This paper offers a utilitarian alternative that can discriminate more effectively among policy options.
The incompleteness of behavioral preferences manifests itself as sets of preferred bundles that are kinked and thus supported by multiple price vectors. The Pareto optima are therefore also supported by multiple price vectors and consequently, following small changes from optimal allocations, supporting price vectors will continue to be present. By the first welfare theorem, all allocations near to a Pareto optimum are optimal as well. Pareto optimality thus cannot locally screen policy decisions.
Utilitarianism can address the oversupply of optima without requiring a planner to supply the preference comparisons that individuals are unable to make definitively for themselves. But since classical models of utilitarianism do not allow preferences to be incomplete, the classical model must be reworked.
When behavioral preferences are incomplete, agents can still have well-defined utilities for 'groups' of goods. For example, agents can have von-Neumann-Morgenstern utilities for the goods delivered at particular states which they weight with probabilities that depend on the decision-making frame. For each group of goods, planners can cardinally compare the utilities of individuals for these goods and incorporate these comparisons into social welfare functions. Due to the incompleteness of the individual preferences, a planner will have multiple social welfare functions. There is nevertheless a unique utilitarian optimum when utilities are separable across groups (as in expected utility theory). If utilities fail to be separable, the precision of utilitarianism is reduced but the dimension of the utilitarian optima still falls markedly relative to the dimension of the Pareto optima.
Utilitarian optima can in principle fail to be Pareto optimal but we show that as preferences become more incomplete this possibility cannot occur.