The primary objective of this article is to advance the development of early warning of intrastate conflict by combining country-level structural and events data in a logistic regression model calibrated and validated using split-sample cases. Intrastate conflict is defined by the occurrence of one or more highly destabilizing events collectively termed a crisis of interest (COI). Two separate two-year periods between 1990 and 2005 were examined in twenty-five globally dispersed countries. COIs occurred in about 6 percent of all the half-monthly periods examined. While model accuracy (total correct predictions of COI and non-COI) usually exceeded 90 percent, the model did not generate sufficiently high and consistent precision (correct number of COI over total predicted) and recall (correct number of COI over total observed) for practical use.

crisis forecast, early warning, failed state, political violence, regression
dx.doi.org/10.1177/0022002712446129
Journal of Conflict Resolution: research on war and peace between and within nations
Norman Paterson School of International Affairs

Tikuisis, P. (Peter), Carment, D, & Samy, Y. (2013). Prediction of Intrastate Conflict Using State Structural Factors and Events Data. Journal of Conflict Resolution: research on war and peace between and within nations (Vol. 57, pp. 410–444). doi:10.1177/0022002712446129