View From a Height

Live-Blog: The Problematic Style Grid

The talk will be based on a paper by Callahan and our professor, Tom Howard, about the dictatorship of the Morningstar "style" grid. You know: large-cap growth, small-cap value, etc. These began as descriptive, but have now become normative, and their claim is that it hurts returns.

Specifically, their claim is that these boxes don't really represent style or asset classes, but characteristics of the portfolio. What had been intended as guides to the investor have now become strait-jackets to the manager.

5:36 - The boxes in the value-size grid are highly-correlated. Why, then, are they considered asset classes?

5:38 - It also means that an all-stock portfolio becomes a hard-sell, since it doesn't fit into these boxes.

5:39 - "Style" really should mean the method of investing; what do you look for in a company?

5:41 - Stocks that meet a manager's style may not fit neatly into a box. It only makes sense to keep managers in a characteristics box if the following 3 conditions obtain: 1) a manager's screened stocks need to fit into a single box; 2) the screened stocks themselves cannot drift to other boxes; 3) Multi-specialists outperform those who let their portfolio characteristics drift.

5:42 - This system, according to a literature search, did not evolve from empirical data, but from convenience and the above 3 assumptions.

5:45 - First, Russ Wermers (U.Md.) did demonstrate in a August 2000 Journal of Finance paper that stock-picking ability does exist.

5:46 - Wermers also found that small-cap, growth, and good stock-pickers drifted the most, and had a 2.9% excess alpha. (July 2002 working paper.)

5:51 - Asness in Financial Analysts Journal, M/A 1997. A value-momentum specialist actually would perform better.

5:53 - Back-test Graham, William O'Neil, T. Rowe Price, and Neff. Rigidly apply each of these methodologies regardless of the stock over the last 20 years, and see how well the best stocks fit into each characteristic box. In order to avoid hindsight, use only forward-looking earnings estimates.

5:57 - In Graham's case, 53% of the top stocks fell outside the small-cap value box, the best box for his style. The % in value move all over the place, from less than 60% to over 90%. The % is small-cap also drift from about 35% to 80%. Finally, Graham outperforms the Russell 3000 by 8% a year, 1995-2003.

6:02 - You need to have the top 20 Graham picks to get the full benefit of his style. This averages out to his 10th best pick. In any given characteristic box, the best box averages to the 35th best pick. Which throws away almost all the benefit of their styles.

6:05 - Overall portfolio tilt? By combining the 4 styles, you actually get a very balanced portfolio, slightly tilted towards growth, and slightly tilted towards small-cap.

6:11 - The peer groups are M(arket) - S(ize) - V(alue) - M(omentum), rather than Growth-Value/Size. Measuring against a characteristic peer group becomes more complicated, requiring weighting for the boxes, rather than a single box.

6:16 - The correlation coefficients among the "style" boxes are huge, averaging 0.86. The correlation coefficients among sectors average 0.55. If the job of the portfolio manager is to look for low correlations, calling the "style" boxes different asset classes makes no sense.

And for the core paper, that's it. I'm going to try to re-start comments and trackbacks, hoping I've shaken most of the spam.