Dissecting Firm-Level News: A New Measure to Capture the Time-Varying Risk of the Company

Abstract

The volume of papers investigating the relationship between firm-level news and stock prices has shown a marked increase. While the vast majority of these studies use sentiment measures or extensive coding rules to shed light on the implications of qualitative information for financial data, not much work has been done on the role played by the associated meaning of words in enhancing our understanding of the company systematic risk and return volatility. Using recent advancements in textual analysis, I derive a novel measure from the word flow of firm-level news. This measure, which I call Market Similarity, overcomes most of the limitations of classical text analysis and does not require to be trained on any quantitative variable. I show that this proxy captures the change in firm and market conditions, and explains the variation in the Market Volatility Index (VIX). A one percent increase in Market Similarity leads to a 1.89% increase in the VIX level, a much larger effect than that documented by Manela and Moreira (2017) in their NVIX measure. Furthermore, this news-based measure seems to drive the time variation in the CAPM market beta. Given the effect of the Market Similarity in the R-squared of the market model, in the return volatility and systematic risk of a company, I conclude that this proxy measures market and company uncertainty.

Keywords: Uncertainty, Firm-Level News, Natural Language Processing, Systematic Risk, Return Volatility

JEL Classification: C43, C45, C55, C58,G11, G12, G32