Sensitivity Analysis of the Empirical Literature on Civil War Onset

Abstract:
This paper applies a methodology for organized specification tests to establish the robustness of empirical findings in the quantitative literature on civil war onset. A review of the literature reveals that several empirical results are not robust or not replicable across studies, which creates uncertainty about the validity of inferences drawn from econometric studies of civil war onset. Given that most studies use different definitions of civil war and analyze datasets with different country-year coverage, readers cannot distinguish between variation in empirical results that is due to those factors and variation that is due to the measurement of independent variables or model specification when the same dependent variable and same country-years are analyzed. We reduce this uncertainty in this paper by systematically exploring how sensitive 88 variables used in the literature are to variations in the set of control variables.