Performance Perspectives Blog

The meaningfulness of statistical measures

by | Mar 12, 2014

When we conduct GIPS(R) verifications, it is not uncommon to see a footnote appear on a presentation where the number of accounts in the composite are five or fewer. The Global Investment Performance Standards do not require firms to disclose the number of accounts or dispersion if there are five or fewer accounts (at year or end or for the full year, respectfully), and most firms avoid doing it. Footnotes often include wording such as “statistical measures of internal dispersion are not considered meaningful and therefore not presented.”

Not meaningful. What is meant by that? Statistically significant? Doubtful.

If there are six accounts present for the full year, is dispersion “meaningful”? Apparently, yes, though I believe statisticians would probably argue that they are not. For standard deviation to have value, we’d probably want 30 or more observations, which is a threshold that is well beyond what is permitted. 

Simply indicating “≤ 5” should suffice for number of accounts, and “n/a” for dispersion, though a footnote indicating that there were less than six accounts present for the full year will work, too.

Free Subscription!

The Journal of Performance Measurement

The Performance Measurement Resource.

Click to Subscribe