Identifying the effect of climate variability on communal conflict through randomization

In recent years, the focus of quantitative climate-conflict research has shifted from studying civil wars to studying different types of conflicts, particularly non-state and communal conflicts, based on the argument that these local-level conflicts are a more likely consequence of climate variability than civil war. However, the findings from previous research do not paint a consistent picture of the relationship between climate and communal conflict. We posit that a research design treating the climate variable as randomized is a better and more convincing strategy for estimating the relationship between climate variability and communal conflict compared to the conventional control method to account for confounders. In this paper, we ask two questions: (1) what type of research design allows us to treat climate variability as randomized and (2) what can we say about the relationship between climate variability and communal violence using this new design? To answer these questions, we analyze six large subnational areas, at a monthly time scale, and calculate the standardized precipitation index for each area for each month. We find that both short, unusually dry intervals and long, unusually wet intervals increase the likelihood of a communal conflict event.