There is also support for the notion, observed by others, that manager skill is more valuable during challenging economic times. Fund alpha averaged 232bp over the eight economically difficult years of 1982, 1983, 1990, 1991, 2000, 2001, 2007, and 2008. In contrast, average fund alpha plummeted during the supercharged economy of the late 1990’s. In fact, the fund underperformance observed over 1982 - 2008 can be turned into superior performance when 1995-99 is eliminated, resulting in an average fund alpha of 20bp. While the late 90’s were an amazing time economically, they were not at all kind to equity fund managers.

Below is a graph of the average fund 36 month trailing annual standard deviations as of the end of each year, along with the time trend. Note that there is a downward trend (-8bp annually), implying average fund risk declined over this period.

Together, these two graphs reveal average fund performance has improved over this nearly thirty year time period while average fund volatility has declined. This represents a strong case for increasing manager skill.

So why did BSW come to such a different conclusion?

I and BSW are using essentially the same data over the same time period, so our diametrically opposed conclusions are driven by different methodologies. My results are based on straightforward, intuitive calculations. BSW employs complicated statistical techniques that distort the conclusions reached, as I argue below.

• In using the Carhart four factor risk model, BSW needs at least 60 months of historical data to estimate required parameters. Thus they eliminate all fund/month observations that are preceded by less than 60 months of history. This dramatically reduces sample size. For example, if I were to follow this procedure, 44% or 210,640 of my fund/month observations would be eliminated. Essentially BSW have gutted their sample in order to feed the data voracious Carhart model.

• It turns out that funds with less than five years of history generate very attractive returns. In my sample, the fund/months with less than 60 months of history generate an average annual return 50bp higher than those with a longer history. By eliminating short history observations, BSW injects a downward bias into their results.

• In addition, there is a negative relationship between returns and fund history. That is, the longer the fund history, the lower the average return. This is a counterintuitive result in that most feel longer histories imply a better manager and thus higher returns. In fact, many in the industry demand long performance histories before a manager is even considered. But the evidence portrays the opposite, the shorter the history the better the performance. Thus any search for manager skill needs to include short history funds, which BSW does not.

• BSW compound their mistake by looking at life-of-fund alphas for evidence of managerial skill. This is consistent with the widely held belief that skilled managers perform consistently over time and skill cannot be detected without long performance histories. The data reveals a different reality. On average, superior manager skill exists early in the life of a fund, with declining performance thereafter. This means that it is easier to find a skilled manager among funds with short histories than among those with long histories. (Note that this is just the opposite of the approach used by most investors when looking for skilled managers). Any search for manager skill should consider these realties. BSW fail at this.

• The combination of shorter history funds performing better and deteriorating performance with lengthening fund history produces an undesirable end-of-sample bias in the BWS study. Over the last five years of their study (i.e. 2002-2006), no new funds are included and the funds that are included have ever lengthening histories. Thus there was a growing downward bias in their results as the last sample year approaches. This may help explain why they observe a decline in manager skill at the end of their sample. Note that extending the sample will not remedy this bias. BSW must change their methodology in order to eliminate this bias.

• As a final comment, BSW cannot make the statement of having precise estimates of the number of lucky managers. This is because such estimates depend critically on the underlying statistical model being assumed. BSW assume three distributions, one each for unskilled, zero alpha, and skilled managers. This allows them to estimate precisely the number of lucky managers. But what if the truth is many distributions, one each for managers with slightly differing skills. Then it becomes impossible to estimate precisely the number of lucky managers. BSW overstate their level of precision.

Manager Skill is Increasing

The number of skilled managers is a direct function of the overall average fund alpha. The higher the average alpha, the larger is the number of skilled managers. The results I have presented reveal an upward trend in fund alpha combined with a decline in fund volatility. These observations point to increasing managerial skill rather than decreasing skill as claimed by BSW. Today, the average active US equity fund is delivering a solid value proposition to its investors.

Display article as PDF for printing.

Would you like to send this article to a friend?