In recent decades, there has been assertions that climate change triggers conflict via multiple pathways, including food shortages, pest and disease incidence expansion, and water scarcity. However, broad empirical studies on the link are still lacking. This study aims to quantitatively explore that linkage using a global dataset. This involves development of a model that predicts the probability of conflict incidence given climate variations. We apply both parametric and semiparametric techniques in a rolling window scheme, which allows for a system that evolves over time. Two criteria are employed to evaluate out-of-sample predictive capability of the estimated models. Our investigation suggests that precipitation variation has a statistically significant effect on conflict. Generally we find the more that this year’s precipitation is smaller than last years the more likely is civil conflict.