The Global Conflict Risk Index (GCRI) regression model : data ingestion, processing, and output methods

The GCRI is a quantitative conflict risk model, developed by the JRC and based solely on open source data, providing quantitative input to the EU early warning framework, one input to the EU Conflict Early Warning System (EWS), developed by the European External Action Service (EEAS) in close partnership with the European Commission to enhance the EU's conflict prevention capacities. The GCRI distinguishes between three types of violent conflict a state may experience: civil war over national power, subnational conflicts over secession, autonomy, or resources, and conflicts in the international sphere. While the latter are not currently modelled by GCRI, for the first two the index quantifies the probability and the intensity respectively of national and subnational conflicts occurring in the next one to four years. Relying on historical data and a statistical model that includes political, socio-economic, environmental and security variables, it assesses the level and likelihood of future conflicts The GCRI is composed of two statistical models: the regression model and the composite model. Both models are based on twenty-four individual variables. This report presents the work done between February 2017 and September 2017, specifically focused on improving the documentation on the regression model. The present report describes on the one hand the regression model, including the input data and the model itself. On the other hand, it presents the statistical significance test, and the matrix of confusion, performed in order to get a highly detailed analysis of the performances of the model. The results of these analyses are presented in chapter 4 and 5. This report is part of a series of technical documents produced in 2017 aiming at improving the GCRI models with greater transparency and robustness. It is not a validation of the GCRI, but a contribution to exploit its potential.