The “out-of-sample” performance of long run risk models

Abstract

The performance of portfolio managers depends on market timing, volatility timing, and security selection. We develop holdings-based performance measures that adjust for risk using stochastic discount factors, display all three components in a consistent framework, and avoid strong assumptions about managers’ behavior. Previous models leave out some of the components of performance, and correcting for this we deliver better measures of selectivity. Sorting stocks held by funds on selectivity produces a quintile spread in four-factor alphas greater than 2.5% per year before costs and more than 1.7% greater than found using the Daniel, Grinblatt, Titman, and Wermers (1997) measure.