Skip to Content
US Videos

Mean Reversion Has Slowed Down, not Vanished

The rise of giant capital light businesses has stalled the cycle of regression to the mean of corporate returns on equity, says GMO's Jeremy Grantham.

This video is the third of four with GMO's Jeremy Grantham. You can watch one, two, and four here.

Jeff Ptak: I know that you routinely introspect in the commentaries that you publish. I wondered if maybe you would reflect on the experience that you've had as an investor and allocator since the global financial crisis? 

Some of the maxims or rules of thumb that GMO has long used, mean reversion being one of them. Maybe it hasn't been quite as dependable and so I wonder sort of what you as an investor, what your firm takes away from that, and whether informs your approach going forward?

Jeremy Grantham: I think at the asset class level it's done fine. The real shortfall has been at the corporate level. This is my personal view rather than GMO formal view, but things are clearly quite different in this last 20-year window than they have been. 

We've seen the emergence of a handful of giant companies that don't really use capital in the traditional way. They're not building factories and machine tools. They're investing in advertising and brand building, building for the future, expensing it all, and it's unreasonable to expect that their nominal return on equity will look anything like we're used to. 

If you look at the return on equity by 10 percentile running through the last 20 years, curiously what you find is that the 80% in the middle have stayed fairly similar. Bouncing along. What is different is the top 20% that was the most profitable 20 years ago has widened its lead, as the Apples and the Amazons have done so well, and they have inflicted pain on the bottom 10% so that value traps, as we call them, have gone up a bit.

So the book companies get zapped. The retailers begin to get zapped. Industry after industry are being taken out by a new generation of disruptors and put through faster trouble than we were used to seeing in the database. The fairly dependable seven-year regression to normal has, I think, at the very least slowed down. 

I wrote a paper called "Not With a Bang but a Whimper," in which I argued perhaps counting on 20 years was better than seven because now we need more macro-level issues to mean revert and counting on not full scale reversion, but maybe going back two thirds of the way in 20 years was more appropriate. 

That's still a dismal return for the S&P, it's about 2 1/2% real for 20 years, which is nothing to wish for. Even in a world that is different, that stays higher priced, that has higher profit margins, and mean reverts more slowly, you still get some pretty ugly returns.