Incentives, Informational Economies of Scale, and Benchmarking

In this paper, we analyze the problem of providing incentives when there are more than one project. A principal has access to two (possibly correlated) projects which are managed by a single agent. Before undertaking a project, the agent-manager can spend some resources to investigate its quality, namely its probability of success. Only projects with a high probability of success are profitable, and therefore should be invested in. There are two classes of strategies. First, the manager may investigate only one project to save on investigation costs, but still use the acquired information to make an investment decision on the noninvestigated project. Second, the manager could investigate both projects, and make the investment decision based on the acquired information on each project. Comparing these two strategies, it would seem that the more correlated are the projects, the better it is to investigate only one project and use the acquired information to learn about the other project. And, when correlation is low, both projects should be investigated. There is, however, a third strategy that the principal could use. He could hire two agents, each managing one project. By making each agent's compensation dependent on the outcome of the project of the other agent, the principal creates some form of competition between them. When projects are highly correlated, endogenous competition provides incentives but duplicate investigation costs, while having one manager investigating only one project exploits informational economies of scale by economizing on investigation costs, but does not always yield the best investment decision since information on the noninvestigated project is not perfect. We assume that the investigation decision is private to the manager (moral hazard), as well as the information obtained doing so (adverse selection). We then show that the optimal structure depends, among other things, on the degree of correlation between the returns of the two projects. In general, delegating to one manager and investigating both projects is optimal when the projects are weakly correlated; delegating to two managers is optimal for intermediate values of the correlation coefficient, while delegating to one manager and investigating only one project may be optimal when projects are strongly correlated, depending on parameter values. We show that, for some parameter values, it may never be optimal to delegate to one manager and investigate only one project, and this even when projects are perfectly correlated. Endogenous competition is then optimal as it minimizes the cost of providing incentives to the managers, even though investigation costs are duplicated. In that case, delegating to two managers becomes optimal for intermediate and high values of the correlation coefficient.