Risk Factors That Matter: Textual Analysis of Risk Disclosures for the Cross-Section of Returns

Abstract

Using unsupervised machine learning, I introduce interpretable and economically relevant risk factors that characterize the cross-section of returns better than the leading factor models, furthermore, I do not use any information from the past returns to select the risk factors. I exploit natural language processing techniques to identify from the firms’ risk disclosures the types of risks that firms face, quantify how much each firm is exposed to each type of risk, and employ the firms’ exposure to each type of risk to construct a 4-factor model. The risk factors roughly correspond to Technology and Innovation Risk, Demand Risk, Production Risk and International Risk.

Keywords: Cross-Section of Returns, Factor Models, Machine Learning, Big Data, LDA, Text Analysis, NLP