A recent study by two scholars at the University of California Irvine asks how quickly investors learn about the skills of their asset managers, including their hedge fund managers.
Christopher Schwarz and Zheng Sun, both associate professors of finance at The Paul Merage School of Business at UC Irvine, had the novel idea of examining empirically how disagreements among investors as to the skill level of managers converges over time. This study, they contend, allows “inferences about how fast investors learn about uncertain financial parameters.”
The premise is the plausible one that investors put their money into funds that they believe have skilled managers and take money out of funds that they believe have unskilled managers. If investors have converged on the opinion that ABC Fund LP has unskilled managers, there will be a lot of redemptions and little inflow. On the other hand, if the convergence is of the opinion that ABC Fund has extraordinarily skillful traders, the flow will be in, not out. If there is a lot of movement in both directions, the investors disagree about the skill level, there has not yet been a convergence.
A Bayesian Approach to Learning
The key number for understanding the degree of convergence, then, is the minimum of inflows/outflows. The larger the minimum, the greater the disagreement about managerial skill.
Schwarz and Sun take a Bayesian approach to learning. In the operation of Bayes’ theorem, an investor enters a marketplace with an initial conviction as to managerial skill. Facts, as they arrive, change this prior conviction into various beliefs over time. But the amount of difference that any new fact will make to the settled conviction diminishes as the number of facts already digested increases. In other words, “the rate of convergence in beliefs has a convex shape with respect to time.”
One fascinating conclusion they draw when they look at the data is: hedge fund investors learn with the speed that Bayes’ theorem suggests they “should.” But mutual fund investors learn (that is, their views as reflected in the fund flows converge) much less quickly than the Bayesian pace. The difference is significant and exists despite the lesser transparency of hedge funds—which, one might expect, would make informed judgments about skill more difficult for such funds.
An Objection and a Rejected Interpretation
One obvious objection to the method of the study would be that the development of judgments about skill cannot be the only cause of fund flows in either direction They may accidentally be measuring something very different from what they think they are.
Schwarz and Sun address this point, saying that liquidity needs and other matters also drive fund flows, but … that those other matters “do not go up with fund age.” Investors’ level of “informedness” about the skill of a funds’ managers would go up with fund age, so it would have an impact that the other factors would not mask.
So, take it as a given (until such time as other scholars may challenge the findings of the paper), that investors in hedge funds learn more quickly about the skill of their managers than do investors in mutual funds. Why might that be? One possible read: it could be that different performance measures are involved, and that mutual funds investors are less decisive about or amongst the diverging metrics that are their alpha-qualified counterparts. This is an attractive notion in some ways, but it is wrong.
Say Schwarz and Sun, “funds with similar performance profiles across various performance measures have the same speed in the decline of disagreement as those whose performance measures disagree.”
Conclusion
In the end, they explain their data way of “bounded rationality.” It is rational for mutual fund investors to pay less attention to their fund results because such attention doesn’t prove valuable enough for them once obtained. There result, then, is twofold. On the one hand learning can occur in the manner presumed by Bayesian theory, and thus at the rate it would indicate; on the other hand, learning can be slower if the learning doesn’t provide value to market participants.