Atrocity Forecasting Project


The project has the overall purpose of enhancing capacity for forecasting mass atrocities and genocide globally and in the Asia-Pacific region. The specific aims are to:

  1. develop sophisticated, appropriate, and cutting-edge quantitative forecasting models,
  2. improve understanding of the causes of political instability and conflict which greatly increase the probability of mass atrocities or genocide,
  3. improve understanding of the crucial causal processes which lead from instability to mass atrocities or genocide, and
  4. produce forecasting software and reports which are useful as early warning tools for protection of vulnerable populations.

The aim is to provide tools for analysts and policy makers. The project builds on the current academic literature, and employs econometric and machine-learning based forecasting techniques, which can greatly enhance analytical capacity in combination with standard qualitative and quantitative social science methods. The forecasts are intended to be used in combination with other quantitative and qualitative analysis and expert knowledge.

Funding for this project from AusAID’s Responsibility to Protect Fund, via the Asia-Pacific Centre for the Responsibility to Protect, is gratefully acknowledged. However, the projects’ Chief Investigators, based at the University of Sydney and the University of New South Wales, take full responsibility for the analysis and forecasts presented here and in the project's reports and publications.

New Forecasts, 2016-2020

Using updated data and the same method we used for our 2011-2015 forecasts, we have produced a new set of forecasts for 2016-2020 Given our reasonably good performance for 2011-2015, we expect similar performance from these forecasts. For more information on our evaluation of the 2015-20 forecasts, see our report: New Forecasts for 2016-2020 & Evaluation of our Forecasts for 2011-2015.

Map of Predicted Genocide/Politicide Risk, 2016-2020

The risk categories correspond to the following: Highest Risk = top 1% of states ranked by risk, High Risk = top 5% of states ranked by risk, Moderate Risk = top 10% of states ranked by risk, Low Risk = top 25% of states ranked by risk, Lowest Risk = outside of the top 25% of ranked by risk.