Macroeconomic Content of Characteristics-Based Asset Pricing Models: A Machine Learning Analysis

Abstract

We explore whether the empirical success of seven characteristics-based asset pricing models can be explained by their ability to identify macroeconomic risks. We find that although the stochastic discount factors (SDFs) of some models are weakly related to macroeconomic shocks, the SDFs' non-market components are totally unrelated to them. The result also holds for macroeconomic news and real-time macroeconomic shocks. Our analysis involves a comprehensive set of more than 100 macroeconomic indicators and uses machine learning techniques to mitigate the overfitting problem caused by a large number of explanatory variables. Our paper illustrates how machine learning can be used for analyzing the explainability of one variable by many others.

Keywords: asset pricing, stochastic discount factor, machine learning, elastic net, macroeconomic shocks

JEL Classification: G12, C58