Assistant Professor in Microeconomics, Section AE1 Department of Economics, Maastrichut University, The Netherlands.

This paper studies a principal-agent problem where a risk averse agent is able to exert costly effort to improve the distribution of his type, which in turn affects his value to the market. The principal is able to observe the agent's type, but not directly observe effort, and can then provide information to the market, which is competitive and pays the agent his expected valuation. The principal's goal is to pick an information scheme to maximise the agent's utility. For example, the principal could provide full information to the market, but this would expose the agent to the full amount of risk. Alternatively, the principal could provide no information, leading to a market wage that is constant, but leaves the agent with no incentives to exert effort. Thus, the principal faces a trade off between providing incentivizes the agent and smoothing the agent's risk. For an agent with binary types, the optimal information scheme is charaterized for a large class of utility functions. The crucial determinant of the optimal policy is prudence. It is shown that there exists a threshold, such that if the agent's prudence lies above this threshold, then the optimal policy is to reveal a fraction of good types and then pool the other agents. If the agent's prudence lies below this threshold, then optimal policy is to reveal a fraction of bad types and then pool the other agents.