Canonical work in political economy argues that order promotes private investment: conflict generates fears that future returns will be destroyed or expropriated. While most studies support this claim, the empirical record is mixed: a third of the studies in our systematic review report null or mixed correlations between instability and investment; a smaller number of studies suggest that conflict can actually boost investment. We help to reconcile these disagreements, arguing that conflict has divergent effects on firms’ investment decisions depending on how exposed a firm’s operations are to fighting. Conflict may deter investment by either disrupting production or raising policy uncertainty or reputation risk. By contrast, conflict may encourage investment where it hampers state taxation and regulation. We argue that these mechanisms operate at different geographic scales. We collect unique, firm-level panel data from the mining sector and estimate how firms with differential exposure to armed conflict change their investment. Our analysis reveals divergent responses: we find that firms with operations at conflict sites dramatically reduce their investments following violence. Yet, firms operating in the territory surrounding conflict, but at a remove from the fighting, actually increase their investment. Firms well-removed from violence see a small negative effect. We suggest that mixed results in past studies stem from an ecological inference problem: overwhelmingly, they rely on country-level analyses that mask the cross-cutting effects of conflict.