Tuesday, June 30, 2009

A Simple (and Impressive) New Three Factor Return Model

First, a little background on "factor models": The CAPM model for estimating expected returns is the oldest and most widely know of all finance models. In it, exposure to systematic risk (i.e. beta) is only factor that gets "priced" (i.e. that's related to expected returns).

In 1993, Fama and French showed that a three factor model (the CAPM market factor plus a size factor and a value/growth factor), did a much better job of explaining cross-sectional returns when compard to the "plain vanilla" CAPM.

Since the FF model became popular, a number of studies have come out that identify other factors that seem to be associated with subsequent returns, such as momentum (Jegadeesh and Titman, 1993), distress (Campbell, Hilscher, and Szilagyi, 2008), stock issues (Fama and French, 2008) and asset growth (Cooper, Gulen, and Schill, 2008).

Now, on to the meat of this post - another factor model. This one is based on q-theory (i.e. on the marginal productivity of a firm's investments). Long Chen and Lu Zhang (from Washington University and Michigan, respectively) recently published a paper "A Better Three-Factor Model That Explains More Anomalies", in the Journal of Finance. They propose a three-factor model", with the three factors being the aggregate returns on the market, the firm's asset-scaled investments, and the firm's return on assets). Their model significantly outperforms the Fama-French (FF) model in explaining stock returns, does a better job (relative to FF) at explaining the size, momentum, and financial distress effects (i.e. you don't need to add additional factors for these effects), and does about as well as FF in capturing the Value (i.e. Book/Market) effect. Here's a taste of their results:
  • The average return to the investment factor (i.e. the the difference between the low and high investment firms) is 0.43% per month over the 1972-2006 sample period. When measured only among small firms, the return difference between low and high investment firms is about 26% annually)
  • The average return to the ROA factor (the difference between returns to the firms with the lowest and highest ROA) is 0.96% per month over the sample period (with a high/low spread of about 26% for the smallest firms).
  • The differences in high vs. low portfolios persist (albeit in smaller magnitudes) after controlling for Fama-French and momentum factors.
It's definitely worth a read (in fact, it'll be on the reading list for my student-managed fund class). You can find an ungated version of the paper on SSRN here.

HT: CXO Advisory Group