Big data analysis of economic news

Abstract

We propose a novel method to improve the forecast of macroeconomic indicators based on social network and semantic analysis techniques. In particular, we explore variables extracted from the Global Database of Events, Language, and Tone, which monitors the world’s broadcast, print and web news. We investigate the locations and the countries involved in economic events (such as business or economic agreements), as well as the tone and the Goldstein scale of the news where the events are reported. We connect these elements to build three different social networks and to extract new network metrics, which prove their value in extending the predictive power of models only based on the inclusion of other economic or demographic indices. We find that the number of news, their tone, the network constraint of nations and their betweenness centrality oscillations are important predictors of the Gross Domestic Product per Capita and of the Business and Consumer Confidence indices.