Public-good provision with macro uncertainty about preferences : efficiency, budget balance, and robustness

The paper studies efficient public-good provision in a model with private values whose distribution depends on a macro shock; conditionally on this shock, values are independent and identically distributed. A generalization of the Bayesian mechanism of d'Aspremont and Gérard-Varet is shown to implement an effi cient provision rule with budget balance. However, first-best implementation and budget balance are incompatible with a reqruirement of weak robustness whereby incentive compatibility of the mechanism is independent of the stochastic specification within the class of specifications defined by the structure of the model. Budget imbalances with robust implementation are small if there are many participants, as surplus from the Clarke-Groves mechanism converges to zero in probability when the number of participants becomes large. In the limit, with a continuum of agents, a first-best provision rule with equal cost sharing is robustly incentive-compatible. In this limit, information about the macro shock, which is the only thing that matters for public-good provision, can be elicited without any efficiency loss.