Informing the Choice of Fractile Usage in Factor-Based Stock Screening Maria Fernandez David Marcus Matthew Meares Hiren Patel Hong Wan
Introduction Purpose: To study the trade-offs (risk-return) in factor-based screening by varying portfolio size Fractiles are used as the standard method for choosing portfolio size Potential to overlook the impact of fractile choice in portfolio formation FactSet offers limited defaults of 4, 5, 10 fractiles
Factor Rank Order Expectation Good factors will produce a rank order with a large negative slope Downward slope may exhibit linear relationship (R2)
Factor Rank Order Reality Factor is not sole driver of stock returns Unsystematic risk also exists
Fractile Portfolios for Diversification Placing securities into fractiles provides diversification of risk, but potentially at a cost…
Consequences of Diversification As portfolio size increases (# of fractiles decreases), portfolio is adding in securities that, based on factor rank, are expected to generate a lower return
Finding Risk/Return Balance The advantage of diversification diminishes rapidly in randomly selected stock portfolios Factor-ranking portfolios may not reduce risk as well, given non-random portfolio formation Source: http://mba.tuck.dartmouth.edu/pages/faculty/kent.womack/teaching/UnderstandingFF3Factor.pdf `
Methodology Two universes: Period: Factors for screening: 750 Biggest US Companies by Market Cap 750 Small Cap US Companies Equally weighted portfolio composition Period: Monthly Dec 30, 1994 till Nov 1, 2004; rebalanced every month Factors for screening: ROA Dividend Yield Price / IBES Standardized Unexpected Earnings (SUE) Price / IBES Mean EPS LTM EPS / Price Prior month trading volume LTM total return Cash / Price
Fractile Analysis - SUE SUE exhibited the one of the best factoring ranking characteristics
Fractile Analysis - SUE Long portfolio exhibits expected loss of diversification benefit Dynamics of long-short portfolio not completely understood
Fractile Analysis - SUE
Conclusion Similar dynamic exists across many factor ranked models and across market caps (see paper for details) Trade-off between return and risk directly applies to fractile selection and should probably be managed more effectively Areas for further study Application to value weighted portfolios Estimation of trading costs in small versus large factor ranked portfolios Stability across time intervals – perhaps measured with slope and R2 – and return periods