View From a Height

Morningstar's Style Boxes- II

One of the claims that Morningstar makes is that its style boxes have a lower correlation that their competitors'. That's true, but it depends strongly on what time period you use.

Using Morningstar's own numbers for annual returns, years 2000-2004, the average correlation among the style boxes is 0.927. However, if you include the years 1997-1999 in the calculation, the correlation drops to 0.606. Holding Size constant, the average correlation drops from 0.927 to 0.677. Holding VG constant, it drops from 0.955 to 0.756..

Why?

Well, for one reason, the years 1998 and 1999 provided a huge differentiation between value and growth (mostly Internet) stocks, and between small (mostly Internet) and large stocks.

But I think the more important reason is that Morningstar introduced their new style boxes in 2002. That meant that they were researching the algorithm as far back as 2000, and all they had to use was restrospective data. So it's no surprise that any formula would, like an Italian tank, work better in reverse.

I don't believe Morningstar is guilty of anything here other than marketing. But the fact is, the correlations among their style boxes are probably a lot higher than they would lead you to believe.

Here's the data, for those who want to check.