It seems to be widely accepted that Jensen alpha fails to detect successful market timing funds spuriously indicating poor fund performance. Jensen (1972), Admati and Ross (1985), Dybvig and Ross (1985), and Titman (1989), (1995) attribute that to an upwards biased estimate of the systematic risk of successful market timers. Therefore, they recommend not to use alpha in external performance evaluation. In this paper, we show that this conclusion is misleading. We set up a theory of delegated portfolio management in a mean variance framework with asymmetric information. Within this model we prove that alpha is an unbiased performance measure even for market timing funds. We show that the systematic risk for a fund investor consists of two parts: benchmark risk and management risk resulting from the uncertainty about the skills of the fund manager. We show that the extent of management risk depends on what fund investors know about the fund manager's trade record. Therefore, the performance of mutual funds depends not only on the skills of the fund managers, but also on whether they publish their trade record or not. (JEL classification: G11, G23). * We are grateful to Olaf Korn, Christoph Memmel, and Eric Theissen for their helpful comments. Both authors are from the