Implementing action profiles when agents collude

In this paper we consider the principal-multiple agents problem when collusion among agents is possible. Collusion is captured by the use of equilibrium notions allowing for coalitional deviations. We first analyze the constraints that collusion puts on feasible wage schemes. Differently from the case of subgame perfect implementation the first best is not implementable. We provide a necessary condition that compensation schemes achieving the second best must satisfy. Roughly stated, the condition says that wage schemes that are collusion-proof are of the type ''basic fixed wage plus collective bonus.'' We next provide sufficient conditions for unique implementation. The conditions are analogous to those previously found for subgame perfect implementation, but an extra condition of Pareto-optimality (from the agents' point of view) is needed. The mechanism we propose involves only a finite set of messages. No ''integers game'' or ''tail chasing'' construction is used. (C) 1997 Academic Press.