Impact of Portfolio Formation Rule Variation on Factor Premiums

How sensitive are findings about the magnitude and reliability of equity factor premiums to differences in the rules researchers use in sorting stocks to calculate them? In their July 2022 paper entitled “Non-Standard Errors in Portfolio Sorts”, Dominik Walter, Rüdiger Weber and Patrick Weiss examine variations in factor premiums (non-standard errors) due to differences in 14 portfolio sorting decisions applied to each of 40 factors found significant in previous studies. The 40 factors cover well-known sorting variables such as size, book-to-market ratio, asset growth, gross profits-to-assets, momentum and idiosyncratic volatility. The 14 sorting decisions consist of seven sample construction decisions (such as firm size restrictions and exclusion of financial firms) and seven portfolio construction decisions (such as reformation frequency and number of portfolios). For each set of decisions and each factor, they compute a monthly premium as the average return difference in monthly returns between the two portfolios with the highest and lowest expected returns. They also look at effects of portfolio sorting decisions on monthly 1-factor (market) and 3-factor (market, size, book-to-market) alphas.  Using U.S. stock/firm data used in the relevant asset pricing studies as available during 1968 through 2021, they find that: (more…)