“A Mechanism for Selecting Public Goods When Preferences Must Be Elicited”, 1980-12 (; backlinks):
Decentralized provision of public goods provides individuals incentives to be free riders, which will lead to undersupply. If provision is centralized, individuals’ preferences are not known to the authority or mechanism choosing the public goods bundle; hence efficient decisions cannot be guaranteed. The authority can ask people to report their preferences, but strategic rather than truthful responses must be expected.
Is it possible to construct a procedure which simultaneously induces participants to report their preferences correctly and uses such reported preferences to select a Pareto-optimal bundle of public goods? We address this question in a model where the public goods are financed through an existing tax system. That is, an individual’s taxes depend only on the chosen bundle of public goods, not on anybody’s expressed preferences.
We have succeeded in constructing such a procedure. It employs a form of weighted voting. Each individual has an exogenously given endowment of “influence points”. A tentative decision is announced, and the individual then allocates these points among the various public goods and uses them to “vote” for an increase or decrease in the supply of each good. Influence points do not purchase votes for movement on a linear basis. As an individual spends more points on one good, the marginal value of an additional point decreases. Specifically, votes for movement equal the square root of the number of points expended. [quadratic voting]
This decreasing productivity of influence point expenditures induces participants to spread out their allocations in a way which truthfully reveals their marginal valuations of the different public goods, which are the relevant aspects of their preferences. If the votes cast in favor of increasing the supply of each public good exactly balance the ones cast in the opposite direction, an equilibrium is reached. This is the outcome of the procedure. It represents a Pareto-optimal decision.
Practical computation of such an equilibrium will involve all the problems associated with computing competitive equilibria in private goods markets. We present an algorithm which seems to work satisfactorily in a fairly general class of cases.
The procedure can be adapted to different sets of distributional objectives by varying the endowments of influence points assigned to individuals.