EFF: All the evidence I know says that market predictions are unbiased. It's understandable, however, that hedge fund managers are immune to this evidence since it's a threat to their existence.

KRF: It is hard to talk about bias without specifying the information available to measure the expected value. A drug company's stock price might be an unbiased estimate of its value based on all publicly available information, but the employee who just discovered a cure for cancer knows the company's price is too low. That is not to say all prices are unbiased with respect to publicly available information. The efficient market hypothesis is just a model and, like all interesting models, it is not literally true. There are mistakes in prices even if one considers just publicly available information and, since people use financial prices to help decide how to allocate resources, those mistakes must affect the underlying reality. Of course, the existence of mistakes does not imply they are easy to find.

People often misinterpret high average fund returns as proof that a manager does know how to identify pricing errors. Consider, for example, a hedge fund with an annual volatility of 20%. (To be more precise, the standard deviation of the fund's excess return with respect to the appropriate benchmark is 20%.) If the fund's average abnormal return is 5% per year over a ten-year period, many investors and financial reporters would conclude that the manager is truly gifted, with a real knack for identifying under- and over-valued securities. But they would probably be wrong. Suppose the manager's true alpha is zero, so he really has no skill beyond that needed to recover his costs. If we pretend his returns are normally distributed, the probability that his average abnormal return exceeds 5% per year for a ten year period is more than 20%. In other words, in a group of hedge fund managers with standard deviations of 20%, we expect one in five to have a ten-year average annual abnormal return of at least 5%—even if none actually have any skill. We expect one in twenty of the unskilled managers to produce a ten-year average annual abnormal return of at least 10%.

These issues are related to the results in our recent paper, "Luck versus Skill in the Cross Section of Mutual Fund α Estimates."