Many equity market researchers assume conventional three-factor (excess market return or beta, size, book-to-market ratio) and four-factor (plus momentum) models as standards of comparison for discovery of new sources of abnormal returns. Are they the best standards? In their November 2008 paper entitled “Fishing with a Licence: an Empirical Search for Asset Pricing Factors”, Soosung Hwang and Alexandre Rubesam investigate the empirical power of 12 previously identified asset pricing factors using a Bayesian variable selection method called Stochastic Search Variable Selection (see the paper for a description). The factor candidates are: excess market return, liquidity, coskewness, cokurtosis, downside risk, size, book-to-market ratio, momentum, asset growth, idiosyncratic volatility, volume and long-term reversal. Using data for thousands of individual U.S. stocks and associated firm characteristics, 25 factor-based portfolios and 30 industry portfolios over the period 1967-2006, they conclude that:
- Based on analysis of returns using individual stocks:
- There is strong and robust evidence that a linear factor model for individual stock returns should include the excess market return, size and liquidity factors.
- There is only weak evidence that the book-to-market and idiosyncratic volatility factors matter for individual stock returns.
- Based on analysis of returns using portfolios of stocks:
- Across the 30 industry portfolios, the size factor remains strong and there is evidence that the idiosyncratic volatility, liquidity and downside risk factors matter.
- The book-to-market factor is reliably important only for portfolios formed on the book-to-market ratio, suggesting that data snooping accounts for its isolated predictive power.
- Differences in results between analysis of individual stocks and analysis of portfolios suggest that portfolio formation assumptions may incorporate factor-related biases.
In summary, investors should probably use the excess market return (beta), size and liquidity factors in explaining and predicting individual stock returns, but not the book-to-market ratio (value factor) or other commonly used stock/firm-specific factors.
