Paul Collier and Anke Hoeffler at Oxford University have reported that countries with a higher percentage of national income from primary commodity exports have been more prone to civil war, an interesting finding that has received much attention from policy makers and the media. In this paper, James Fearon shows that this result is quite fragile, even using Collier and Hoeffler's data. Minor changes in the sample framing and the recovery of missing data undermine it. To the extent that there is an association, it is likely because oil is a major component of primary commodity exports and substantial oil production does associate with civil war risk. Fearon argues that oil predicts civil war risk not because it provides an easy source of rebel start-up finance but probably because oil producers have relatively low state capabilities given their level of per capita income and because oil makes state or regional control a tempting "prize." An analysis of data on government observance of contracts and investor-perceived expropriation risk is consistent with this hypothesis.