Optimal Compensation for Fund Managers of Uncertain Type: Informational Advantages of Bonus Schemes

Abstract

Performance-sensitivity of compensation schemes for portfolio managers is well explained by classic principal-agent theory as a device to provide incentives for managers to exert effort or bear the cost of acquiring information. However, the majority of compensation packages observed in reality display in addition a fair amount of convexity in the form of performance-related bonus schemes. While convex contracts may be explained by principal-agent theory in some rather specific situations, they have been criticized, both by the financial press as well as the academic literature, on the grounds that they may lead to excessive risk-taking. In this paper, we show that convex compensation packages, though likely to be myopically not optimal, may serve as a device to extract information about the ex-ante uncertain type of portfolio managers. Optimal contracts are thus determined by the trade-off between maximizing short-run expected returns on one hand, and long-run informational benefits on the other. In a discrete-time model, combining dynamic principal-agent theory with the theory of learning by experimentation, we characterize optimal incentive schemes and optimal retention rules for fund managers, consistent with empirical observations.