We created a set of factors within the Russell 1000 to understand the relationships that drive stock prices. For the most part, our factors are derived from the familiar concepts employed by fundamental analysts when they evaluate stocks. For all Russell 1000 stocks, each score on a particular factor is calculated from several sources (CompuStat, I/B/E/S, and IDC pricing databases) and then ranked cross-sectionally at a point in time. Factor returns are generated by calculating the subsequent performance of a portfolio that is long the highest scores (i.e., the top decile containing approximately 100 stocks) and short the decile with the lowest scores (rebalanced monthly).
The complete set of 45 factors is defined in Exhibit 15.1. The historical returns to these factors are shown in Exhibit 15.2 and they are sorted by their 2006 to August 2010 returns and ranks. A subset of the 45 factors is shown in Exhibit 15.3 to illustrate how some popular strategies have failed while others have emerged as winners. Consider the cumulative factor returns for the five strategies displayed in the Exhibit 15.3:
- Dividend yield (dividend per share divided by price).
- Estimate dispersion (standard deviation of I/B/E/S fiscal year (FY) 1 EPS estimates divided by ABS (mean of I/B/E/S FY1 estimates)).
- Earnings quality (accruals).
- Up to down revisions (number of FY1 up revisions minus number of FY1 down revisions divided by total number of estimates).
- One-year price momentum (12-month total ...