| One of the prevailing assumptions in the design of incentive compatible mechanisms for implementing public decisions is that the mechanisms are informationally decentralized. However, where public decisions cannot be implemented by centralized actions by a planner, even when the planner possesses the knowledge of the agents? characteristics, he may still need to reach these decisions through decentralized decision mechanisms, in which the agents act in accordance with their incentives to select the public decisions. Given this requirement, even within a complete information model with an informed planner, one may need to find mechanisms that implement public decisions via decentralized decisions of the agents. We show that there exists a mechanism which resolves this problem. We then show that this result complements the possibility of implementation when the planner is partially informed. We demonstrate that, given partial information about the agents? preferences, there exist situations where the planner can nevertheless elicit the complete information on these preferences. We generalize the notion of implementation of social choice rules suggested by Maskin (1977,1979) to suggest a notion we call sequential implementation, under which a planner achieves the optimal outcomes under a social choice rule through a sequence of mechanisms that first elicits the correct preferences of the agents, and then implements the appropriate social outcomes. It is shown that a class of majoritarian rules can be sequentially implemented by a planner with a partial information on the agents? preferences. |