This paper tests the importance of changes in opportunity costs related to agriculture on the risk of violent conflict using data on locust swarms and conflict collapsed to annual 0.25 (approx. 28km2) grid cell observations across Africa and the Arabian peninsula. The identification exploits exogenous local variation in locust swarm exposure driven by patterns in swarm movements together with weather controls and grid cell and country-by-year fixed effects to identify causal impacts of these agricultural shocks. Locust swarms decrease the likelihood of violent conflict event in a given year by around 20%. Effects are driven by areas with crop and pasture land, and there is no evidence of conflict spillovers to nearby areas. The impacts are largest for swarms that arrive in the off-season or planting season for major crops, based on national crop calendars, and the patterns are not consistent with effects on conflict driven by changes in conflict opportunity costs related to agriculture. This points to the availability of non-agricultural livelihood opportunities and to alternative factors such as psychological impacts and relief efforts less often discussed in this literature as crucial in determining whether an agricultural shock increases conflict risk. In contrast to short term negative effects on conflict, cells affected by the 2003-2005 major desert locust upsurge were 62% more likely to experience any conflict in a given year afterward.Absolute impacts are increasing over time alongside a general increase in conflict in the sample countries, suggesting affected areas are made vulnerable to future shocks which precipitate conflict.