Can Big Data Forecast North Korean Military Aggression?

Abstract

Can textual analysis improve statistical prediction of risky geopolitical events? North Korea has been the most important source of geopolitical risk for South Korea due to the former’s unpredictable and secretive military actions against the latter. We find that the tone of English language news articles published by non-South Korean news media, especially U.K. news media, has significant predictive power about North Korean military aggressions. The usage of language tone improves the predictive power of the empirical model by as much as 47%.

Keywords: geopolitical risk, North Korea, Big Data, media, textual analysis

JEL Classification: A00, P16, P26