Mutual Fund Flows and Performance in Rational Markets

Jonathan B. Berk and Richard C. Green, JPE (2004).

Conventional wisdom among academics, even 12 years after this paper was published, is that the average mutual fund manager is unskilled–i.e., he cannot produce a high enough return differential vis-a-vis the market to justify his fees.  This is the case, the logic goes, because markets are efficient so that the average manager does not have private information.

Still, we know that money managers are highly paid, and presumably have a high level of skill that could be (better?) utilized elsewhere.  The conclusion to this story is that the market must be inefficient, or investors irrational.  Literature prior to Berk and Green (2004) tried to reconcile this puzzle by appealing to information asymmetry or moral hazard problems in mutual funds.  Berk and Green want to give the simple model of efficient markets and rational investors another chance.

The Question: In a model with no information asymmetry or moral hazard, what does investor rationality really imply for the returns to active fund management?

The Model:

Managers have differential ability to identify investment opportunities, but they face decreasing returns to scale.  Investors can costlessly reallocate funds between managers, but manager skill is initially unknown both to investors and to managers.  When a manager earns a high return, investors make inferences about his ability in a learning model. They then give him more funds to invest until his expected return going forward is no better than the market return.  That is, if the manager is expected to return positive (negative) alpha, a rational investor will allocate (withdraw) money to (from) his fund.  When a manager has more funds than he can actively manage, he indexes the surplus.  This means that investors do not profit by allocating funds to a good manager; the manager is the one who captures the rents.  It also undermines low returns to active management as an argument that paying mutual fund managers a lot of money is socially wasteful.  In this paper, skilled managers allocate funds to their most efficient use, and they are highly compensated for doing so.

Sirri and Tufano (1998) argue that when current performance is most predictive of future returns, then fund flows are most sensitive to performance.  In contrast, Berk and Green (2004) find the opposite result: when fund flows are more sensitive to performance, current performance is least predictive of future returns in a rational/efficient market without frictions, because funds reallocate so efficiently that return predictability is wiped out.

The outline of the model is as follows:

  1. build a model of fund returns and flows based on learning about managerial ability
  2. Derive expressions for two features of mutual funds: (1) survival probability and (2) fund flows
  3. Calibrate the model, then select the two free parameters to match the data according to the two moments named above.
  4. Examine the distribution of manager skill implied by the two parameters selected in the step above.


  1. Managers’ future performance can’t be inferred from their past returns.
  2. Investors in mutual funds can’t beat the market–not because the managers aren’t good, but because the investors compete so perfectly for managerial talent.
    1. The rents to good management are captured by the managers, not by their investors.
  3. If returns in a sector are predictable, then investment in the sector must be inefficient.
  4. 80% of managers have enough skill to earn their high pay.
  5. Successful managers have more asset under management (AUM) and lower idiosyncratic risk
    1. successful managers have large funds, and are more likely to index surplus funds.  As they do so, their returns become more strongly correlated with the market.
  6. Managers of younger funds have fewer AUM, index less (or not at all), and have higher idiosyncratic risk and so more extreme outcomes.
  7. Fund flow is more sensitive to performance when
    1. the fund is newer — investors have less information about manager skill
    2. the fund is large — absolute flows are more sensitive to returns in large funds, but flows as a percentage of fund size are independent of fund size
    3. observed returns are less noisy — investors can make better inferences about manager skill
    4. fund fees are higher — high fees make funds less attractive to investors


The takeaway from this paper is that a model with efficient markets and rational investors is consistent with manager skill appearing prima facie to be both lacking and not persistent, even when managers do have skill.  In a market for a scarce resource (here, managerial skill), competitive buyers will bid up the price until the return from purchasing the resource is no better than the return to purchasing an alternative.  This does not have to mean that the resource has no value and that buyers are irrational.