We study an adverse selection model, with a principal and several agents, wherecontracting is under asymmetric information. The number of agents is finite and types are "continuous" and independent. We analyze two settings. In the first one, the performance functions of mechanisms may depend on all the reported types. In the second one, each performance function depends only on the respective announced type.Under the standard hypotheses in the basic one-agent adverse selection model and theindependence assumption, there is not loss of generality if the principal considers onlymechanisms for which every agent reports his true type as a dominant strategy. We consider also the relaxation of the monotonicity hypothesis about the agents' welfare and we will prove that the former "equivalente" behveen the Bayesian implementation and the dominant strategy one stands firm in some cases. We examine the properties of the optimal mechanism, supposing that the principal's "virtual income" depends on the agents' performances only through the aggregate total performance (which is natural in the context of regulation of a good produced by an oligopoly), and also, assuming the frame of regulation of a monopolist with several independent divisions (or the one of a group of firms), each one producing a different good. Unlike the standard properties of the optimal mechmisiiis in the basic one-agent adverse selection model, in our model the optimal mechanism may ask very efficient agents for an individual performance higher than the one of complete information. We show also that if agents are symmetrical, the principal may prefer ex ante to hire more than one agent.
