Cross-section without factors: a string model for expected returns

Abstract

Many asset pricing models assume that expected returns are driven by common factors. We formulate a model where returns are driven by a string, and no-arbitrage restricts each expected return to capture the asset's granular exposure to all other asset returns: a correlation premium. The model predicts fresh properties for big stocks, which display higher connectivity in bad times, but also work as correlation hedges: they contribute to a negative fraction of the correlation premium, and portfolios that are more exposed to them command a lower premium. The string model performs at least as well as many existing linear factor models.