Using unique data on conflict-related incidents in the Philippines, we exploit seasonal variation in the relationship between rainfall and agricultural production to learn about the mechanism through which rainfall affects civil conflict. We find that an increase in dry-season rainfall leads to an increase in agricultural production and dampens conflict intensity. By contrast, an increase in wet-season rainfall is harmful to crops and produces more conflict. Consistent with the hypothesis that rebel groups gain strength after a bad harvest, we find that negative rainfall shocks lead to an increase in conflict incidents initiated by insurgents but not by government forces. These results suggest that the predicted shift towards wetter wet seasons and drier dry seasons will lead to more civil conflict even if annual rainfall totals remain stable. We conclude that policies aimed at mitigating the effect of climate change on agriculture could have the added benefit of reducing civil conflict.