Characteristics are variables that can be computed using individual-firm data; for example, the size of a firm or the volatility of its stock return. Hundreds of characteristics have been proposed to explain the cross-section of stock returns; these are described in the comprehensive analysis undertaken in “Replicating Anomalies” by Kewei Hou, Chen Xue, and Lu Zhang. John Cochrane in his presidential address and Amit Goyal in his paper “Empirical Cross-Sectional Asset Pricing: A Survey” ask whether we actually need all these characteristics to explain stock returns or whether a small set of characteristics subsume the rest.
In the paper “A Transaction-Cost Perspective on the Multitude of Firm Characteristics” Alberto Martin-Utrera, Victor DeMiguel, Javier Nogales, and I examine how transaction costs influence the number of significant characteristics. Counter to what intuition would suggest, we find that transaction costs increase the number of significant characteristics from 6 to 15. This is because combining characteristics reduces transaction costs, and hence increases the investor’s utility. The reduction in transaction cost happens because the rebalancing trades in the stocks underlying different characteristics often cancel out.
We also find that the portfolios we design based on optimally selected characteristics perform well in terms of out-of-sample Sharpe ratio. Moreover, none of the three prominent factor models in the literature (Fama-French-Carhart model, Fama-French 5-factor model, and the Hou-Xue-Zhang 4-factor model) can fully explain the returns of our portfolio strategies, which achieve an economically and statistically significant abnormal average monthly return of about 1% relative to these three models. This work suggests that when investing in characteristics, it is important to account for transaction costs. And, the number of significant characteristics may be greater than what one may have thought.