Driven by News Tone? Understanding Information Processing When Covariates are Unknown: The Case of Natural Gas Price Movements

Digitization promotes the instant dissemination of news in financial markets. These news represent unprecedented amounts of unstructured data. This paper applies Big Data analytics to financial news related to the natural gas market. To date, we find evidence on 16 different variables as drivers of the natural gas price. However, these fundamental drivers cannot explain a significant share of natural gas price volatility. Thus, we first apply a LASSO shrinkage method to identify the most significant control variables. Our feature-selection LASSO method suggests 4 out of the 16 drivers as relevant. Second, we investigate the effect of news sentiment on the Henry Hub gas price as a potential driver of volatility. We include the 4 most relevant control variables as of the LASSO method into our regression model. Our findings suggest a significant positive effect of news sentiment on the natural gas price.

Keywords: Information processing, human information processing, natural language processing, behavioral science, text mining, IS research

JEL Classification: Q30, Q31, C50, C51

Suggested Citation: Suggested Citation

Alfano, Simon and Rapp, Max and Pröllochs, Nicolas and Pröllochs, Nicolas and Feuerriegel, Stefan and Neumann, Dirk, Driven by News Tone? Understanding Information Processing When Covariates are Unknown: The Case of Natural Gas Price Movements (June 27, 2015). Available at SSRN: https://ssrn.com/abstract=2624106 or http://dx.doi.org/10.2139/ssrn.2624106