A long stream of papers documents robust correlations between firm characteristics and future stock returns. Most of these characteristics involve accounting numbers. Usually, characteristics have been identified from data analysis, resulting in a proliferation of characteristics. A 2013 survey of published papers and working papers found 186 predictors — a number that the authors said likely under represents the total.
A standard approach identifies firm characteristics that predict equity returns in the data, and then builds "factor-mimicking" portfolios based on these characteristics. The first stage identifies a "characteristic model," the second stage a "factor model." This empirical approach was pioneered by Eugene Fama and Kenneth French, resulting in their three-factor model based on the market return, book-to-market, and size. Subsequent research expanded the set of characteristics to include momentum, investment and profitability, among others.
Without added structure, however, data correlations provide little insight. Consider, for example, the book-to-price ratio (B/P), a prominent characteristic that emerged in empirical asset pricing research. This ratio first surfaced through data mining by Fama and French, who then developed a factor model with a "book-to-price factor." However, there is no clear understanding of why B/P predicts returns.
The characteristic model of this paper provides a structure to interpret these observed correlations.