Quantifying the Diversity of News Around Stock Market Moves

Abstract

The dynamics of news are such that some days are dominated by a single story while others see news outlets reporting on a range of different events. While these large-scale features of news are familiar to many, they are often ignored in settings where they may be important in understanding complex decision-making processes, such as in financial markets. In this paper, we use a topic-modeling approach to quantify the changing attentions of a major news outlet, the Financial Times, to issues of interest. Our analysis reveals that the diversity of financial news, as quantified by our method, can improve forecasts of trading volume. We also find evidence which suggests that, while attention in financial news tends to be concentrated on a smaller number of topics following stock market falls, there is a "healthy diversity" of news following upward market movements. We conclude that the diversity of financial news can be a useful forecasting tool, offering early warning signals of increased activity in financial markets.

Keywords: complexity science; computational social science; latent Dirichlet allocation (LDA); financial news; financial markets

JEL Classification: A10, B40, C10, C20, C22, C53, C90, D70, D79, D83, J10, J11, O40, O47