The Leibniz Institute SAFE and the SAFE Visitor program organize a seminar on
Conditional factor models: first order vs. higher orders (joint work with Damir Filipovic, Mirela Sandulescu and Michael Weber)
Paul Schneider (USI Lugano, Institute of Finance)
Abstract: We study conditional nonparametric discount factors built around classical linear models through the lens of a large cross-section of U.S. asset returns. We find that while deviations from linearity yield only modest pricing performance increases out-of-sample, they largely dissipate certain asset pricing anomalies. Our framework allows us to discard coskewness as a first-order driver for low-risk asset pricing anomalies, and manages to explain them fully within our richer specification. The exclusion of arbitrage portfolios greatly regularizes the models across all specifications considered, leading to stable factor loadings.