Regensburg 2013 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 22: Focus Session: Big Data (joint with jDPG)

SOE 22.4: Talk

Thursday, March 14, 2013, 11:00–11:15, H37

Geopolitical risk-index derived from 60 million news articles predicts war — •Thomas Chadefaux — Chair of sociology, modeling and simulation, ETH Zurich, Clausiusstrasse 50, 8092 Zurich, Switzerland

There have been more than 200 wars since the start of the 20th century, leading to about 35 million battle deaths. However, efforts at forecasting conflicts have so far performed poorly for lack of fine-grained and comprehensive measures of geopolitical tensions. Here, we developed a weekly risk-index by analyzing a comprehensive dataset of historical newspaper articles for 166 countries over the past century, which we then tested on a data of all conflicts within and between countries recorded since 1900. Using only information available at the time, we could predict the onset of a war within the next year with up to 85% confidence; we also forecasted over 70% of large-scale wars, while issuing false alarms in only 16% of observations. Predictions were improved up to one year prior to interstate wars, and six months prior to civil wars, giving policy-makers significant additional warning time.