| Publication |
2014.
|
| Summary/Abstract |
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. In this article, a weekly risk index is derived by analyzing a comprehensive dataset of historical newspaper articles over the past century. News reports have the advantage of conveying information about contemporaries' interpretation of events and not having to rely on meaning inferred a posteriori with the benefit of hindsight. I applied this new index to a dataset of all wars within and between countries recorded since 1900, and found that the number of conflict-related news items increases dramatically prior to the onset of conflict. Using only information available at the time, the onset of a war within the next few months could be predicted with up to 85% confidence and predictions significantly improved upon existing methods both in terms of binary predictions (as measured by the area under the curve) and calibration (measured by the Brier score). Predictions also extend well before the onset of war - more than one year prior to interstate wars, and six months prior to civil wars - giving policymakers significant additional warning time.
|