Promising Initial Results from a New Mass-Atrocities Early Warning System

15 Pages Posted: 24 Feb 2015

Date Written: January 26, 2015

Abstract

This paper describes the initial activity and performance of a system for eliciting and aggregating expert forecasts of mass atrocities and related events around the world. Our results so far confirm the power of aggregating predictive judgments. To date, our system's forecasts have proven quite accurate, even though the individual participants in that pool have demonstrated forecasting skill usually not much better -- and, in some cases, much worse -- than random guessing. The number of forecasting questions for which we have been able to observe eventual outcomes is still small, and nearly all of those questions have focused on events with a very low base rate. Still, we think our results add to the growing body of evidence that aggregated expert judgment can be a powerful predictive tool, even on processes as complex as the outbreak of large-scale political violence.

Keywords: prediction, forecasting, crowdsourcing, genocide, atrocities

Suggested Citation

Ulfelder, Jay, Promising Initial Results from a New Mass-Atrocities Early Warning System (January 26, 2015). Available at SSRN: https://ssrn.com/abstract=2568644 or http://dx.doi.org/10.2139/ssrn.2568644

Jay Ulfelder (Contact Author)

Independent ( email )

No Address Available

Register to save articles to
your library

Register

Paper statistics

Downloads
101
Abstract Views
503
rank
260,096
PlumX Metrics