News as Sources of Jumps in Stock Returns: Evidence From 21 Million News Articles for 9000 Companies

Stock prices exhibit large, discrete movements, typically labelled as "jumps'". A potential important source of jumps in stock returns can be material news events. In this paper, we collect 21 million news articles associated with more than 9000 publicly-traded companies from the Factiva database and use textual analysis to derive measures summarizing those news, including news frequency, tone and uncertainty. We find that these measures of news flow content are significantly related to nonparametric measures of jump intensity and jump size distributions and explain an important fraction of variations in the jumps across individual companies. Further, those nonparametric analyzes provide input for our time-series modelling of firm-level news processes. By modelling the observable news process explicitly and jointly with a latent jump process, we find that news are important drivers of the jumps in stock returns. Consequently, we are able to enrich the economic content of the widely-used econometric models of jumps used for applications such as option pricing and risk management where stock return jump models are frequently used.