We consider the applicability of the Revelation Principle under the possibility of collusive behavior among players in some Bayesian framework. In doing this, since the coalition formation itself suffers information asymmetry problems, we assume that the coalition is formed if the colluding parties can successfully find some coalitional mechanism whose outcome is a set of messages in the original mechanism. Recently Cremer [1986] proposes a coalitional mechanism in the framework of the well known Vickrey-Clark-Groves mechanism. We assume that the agents successfully collude if they can find coalitional a mechanism such that (i) coalitional mechanism is incentive-compatible and (ii) the payoff of this mechanism is strictly Pareto-improving in terms of the agent’s expected utility. Our analysis is undertaken in a one principal/two agent framework. We first ünd that the Revelation Principle is still applicable in the pure adverse selection model. We then extend this result to a model with both adverse selection and moral hazard aspects. Finally, we consider a three-tier principal/supervisor/agent hierarchical organization, as in Tirole (1986). We explicitly present the coalitional mechanism as a side-contract between the supervisor and the agent. We apply the previous result of applicability of the Revelation Principle and characterize the coalition-proof mechanism. We find that the principal can design an optimal collusion free contract with some additional cost by specifying proper individual and coalitional incentive-compatibility conditions and individual rationality conditions. Moreover, we find that the results of Tirole (1986)’s paper hinge on the fact that he considers only “hard,” verifiable, information. / Ph. D.