Civil wars inflict considerable costs on countries which may be trapped in vicious cycles of violence. To avoid these adverse events, scholars have attempted to identify the roots of civil wars. Valuable minerals have been listed among the main drivers of civil conflicts. Yet, despite the large body of literature, the evidence remains mixed. This paper provides a spatially nuanced view of the role of mineral resources in civil wars in the particular case of the Democratic Republic of the Congo. We estimate the impact of geolocated new mining concessions on the number of conflict events between January 1997 and December 2007. Instrumenting the variable of interest with historical concessions interacted with changes in mineral international prices, we unveil an ecological fallacy: Whereas concessions have no effect on the number of conflicts at the territory level (lowest administrative unit), they do foster violence at the district level (higher administrative unit). We develop a theoretical model wherein the incentives of armed groups to exploit and protect mineral resources explain our empirical findings. A spatial analysis of the effect of mining concessions on conflict backs our proposed theoretical explanation.