Sentiment and Covariance Characteristics

Abstract

We propose a bridging model that connects risk-based factor models to sentiment models by using characteristics. Investors use stock characteristics as information to form their biased view and hence creating mispricing. Characteristics also serve as the proxy for the covariance risk to a latent factor. The α from our factor model of mispricing ranges from 0.70% to 1.38% monthly after controlling for other common factors and mispricing measures. Well-known anomalies are only represented in either underpriced or overpriced stocks but not in all the cross-section.

Keywords: Characteristics, Sentiment Model, Factor Model, Risk