Despite the surge in quantitative research examining the link between climate variability and conflict, a lot of uncertainty exists concerning whether there is a link. One shortcoming of the current literature is that it focuses mainly on statistical inference in order to establish causation with little attention for the predictive performance of the model. In contrast, this study extends the current literature by focusing on the predictive accuracy of a model linking droughts to communal conflict using data for Nigeria for the period 2006-2014. Using a number of different model specifications and estimation methods to test the robustness of the results, the analysis shows that although the regression results show a positive link between the occurrence of droughts and communal conflict, the predictive accuracy of the model is relatively low. In contrast, accounting for the temporal and spatial dynamics of conflict leads to better forecasts compared to the climate variable.