Structural Breaks in Online Investor Sentiment: A Note on the Nonstationarity of Financial Chatter

Abstract

Given the increasing interest in investor sentiment derived from social media platforms, we address one overlooked question - are there structural breaks in online investor sentiment? We cast the problem of break-point estimation in the dynamics of the sentiment series as a model selection problem. Considering 360 stocks, we detect structural breaks in most of the respective online investor sentiment series. A return prediction exercise illustrates the economic significance of the detected structural breaks. Our results call into question the widespread practice of using online investor sentiment series without taking into account the nonstationarity induced by structural breaks.

Keywords: investor sentiment, structural breaks, nonstationarity, group Lasso

JEL Classification: C22, C52, G40