[Empirical Finance Research Blog] A Better Three-Factor Model that Explains M...

[Wesley R. Gray]

Investment Potential Rating: 7/10 (1 worst, 10 best)

============

A Better Three-Factor Model that Explains more Anomalies

Long Chen
Washington University St. Louis

Lu Zhang
University of Michigan

Journal of Finance, forthcoming, 2001 http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1418117

Abstract:


A new three-factor model consisting of the market factor and common factors formed on investment and return on assets goes a long way in summarizing the cross-sectional variation of expected stock returns. The model substantially outperforms traditional asset pricing models in describing average returns across testing portfolios formed on short-term prior returns, financial distress, net stock issues, asset growth, and earnings surprises. The model also performs roughly as well as the Fama-French model in accounting for average returns across portfolios formed on valuation ratios, industry, and CAPM betas. The model's performance, combined with its economic intuition, suggests that it can be used to obtain expected return estimates in practice.

Data Source:

Data for this study come CRSP/COMPUSTAT.

Data Specification:

Linear factor models are extensively used by empirical finance researchers--and for good reason. Unlike pure scientists, who have the ability to run repeatable controlled experiments, financial economists often have ONE experiment. Linear factor models are a way researchers can "control" the environment in which they are conducting tests. In the context of finance, the biggest environmental factor researchers need to account for is the risk level.

The CAPM was supposed to be the game-changer that finally gave economists a way to control for "risk" and understand how asset prices "should" be priced.

But let's face it--the empirical evidence suggests the CAPM stinks.

The CAPM theory is interesting, the concepts are intuitive, bankers continue to use it, and professors love teaching the subject, but empirically it doesn't hold any water. (although many devotees argue that researchers do not have the data to really test the CAPM.)

Fama and French have since come to CAPM's rescue with their 3-factor model, which isn't as theoretically appealing as the CAPM, but has been relatively effective empirically speaking. Unfortunately, the 3-factor model has shown signs of wear and tear over the years. Since the landmark 92/93 Fama and French papers, researchers have found plenty of evidence that seems to suggest there are a wide range of anomalies in the marketplace: momentum, accruals, asset growth, earnings surprises/PEAD, piotroski F-score, etc.

Now, Chen and Zhang have come to the rescue for linear factor models--and this one looks pretty solid. The Chen and Zhang 3-factor model is fairly intuitive, easy to implement, and does a solid job explaining the cross-section of expected returns. I'll leave it up to our audience to decide whether the factors they identify are actually "risk" factors or "alpha" factors.

Investment Strategy:

The Chen/Zhang 3-factor model maintains the market excess return factor, but scraps SMB and HML for INV and ROA. Inv is the return spread between low investment and high investment stocks, and ROA is the spread between high ROA stocks and low ROA stocks. (for an analysis of ROA as a stand alone factor see our analysis here).

1. Form an INV factor and a ROA factor--invest how you see fit.

Facts:
1)The long/short INV portfolio earns .43% per month, and the long/short ROA portfolio earns .96% per month
2)A long/short high INV/High ROA minus a High Inv/Low ROA portfolio earns 1.65% a month--not too shabby.

Implementation Issues and Remarks:


Lu Zhang is an "all-star" in academic finance and his research is spectacular--this paper only strengthens his reputation. From a practical standpoint, the Zhang/Chen 3 factor model is extraordinarily useful (as was the FF 3 factor model): it's easy to create, helps determine discount factors, and has an intuitive appeal. The key difference between Zhang/Chen and FF is that Zhang/Chen works better!
Investment managers can use the new 3-factor model to form better DCF estimates, corporate managers can assess discount factors more easily, and fund of funds can assess manager performance more accurately. I'm sure DFA or the ETF/mutual funds will soon come out with INV and ROA type portfolios, but until then, you may want to investigate how low INV and high ROA firms can increase your portfolio returns...

Investment Potential Rating: 7/10

--
Posted By Wesley R. Gray to Empirical Finance Research Blog at 12/13/2009 12:09:00 PM