The “resource curse” refers to a set of cross-national relationships between resource dependence on the one hand and economic growth and civil conflict on the other. Sachs and Warner were the first to document the negative relationship between resources and growth. The notion that countries blessed with abundant mineral resources tend to suffer slow economic growth was just counterintuitive enough to appeal to most economists, and the idea quickly became a piece of conventional wisdom.
Subsequent research delved deeper without questioning the basic relationship. Was all production of primary commodities associated with slower growth, or just mineral production? (Evidence suggested the latter.) What was the causal mechanism: “Dutch disease” via appreciation of the real exchange rate, a political economy explanation such as corruption or rent-seeking, or civil conflict? (Sachs and Warner argued for the first, Robinson, Torvik, Verdier, Moene, and colleagues argued for the second, and many political scientists and economists associated with the World Bank’s Economics of Conflict program argued for the last.)
The research on resources and civil war became a massive literature in its own right. Collier and Hoeffler kick-started the program in 1998 with their “rational rebel” model of civil war. For them, resources promoted conflict by rewarding looting. Fearon and Laitin found that oil exporters were more likely to see new civil wars. Some research also found a diamond curse in the 1990s, but Ross’ work has found that oil (especially onshore oil) is the only robust channel by which resources promote intrastate conflicts. To my knowledge, no one has yet quantified the deleterious consequences that resources have on growth via the oil-civil war link or determined whether there is a growth residual that civil war does not explain.
However, some of the latest research to come out is questioning the very basis of the resource curse, the cross-national negative partial correlation between resource dependence and economic growth. The criticisms of prior research have focused on measurement and modeling issues. Typically, resource dependence has been measured as resource exports divided by GDP. In the numerator, total resource production is not available for many country-years, and exports are therefore preferred. The main problem is that re-exportation of minerals is apparently counted in some datasets, and export of processed minerals (e.g., smelter production) is definitely included. What we ideally want is mine production of raw minerals. In the denominator, GDP is used rather than population in order to test the Dutch-disease, crowding-out argument. If resources are important relative to the economy, they may draw capital and labor out of more “productive” sectors that generate dynamic gains and drive up nontradables prices.
The problem is that an economy may develop resource dependence because of growth-retarding institutions that have persisted for a long time, institutions that discourage capital accumulation and the manufacturing and service sectors dependent on abundant physical and human capital, thus generating a negative partial correlation between resource dependence and growth that is spurious due to “endogeneity.” Endogeneity simply means that the errors in the growth model correlate with one of the regressors, in this case resource dependence.
To try to address this problem, Sachs and Warner included some geographic controls, on the assumption that geography would be the main omitted variable that could drive spurious results. However, they found no effects for their geography controls. Since then, we have discovered that geography as a determinant of growth has been vastly overrated. Once Acemoglu, Johnson, and Robinson include political institutions (instrumented by settler mortality rates) in their growth models of post-colonial countries, geography is no longer statistically significant. What this tells us is that endogeneity may still be a problem.
In two recent papers, Brunnschweiler and Bulte argue that resource dependence is a result of poor institutions and conflict, and that when instrumented, it no longer affects either economic growth or civil war. They report that a very broad measure of resource abundance (net present value of natural capital, including mineral, agricultural, and environmental assets) is positively related to growth and therefore indirectly related to less conflict risk (richer countries experience less civil war). Whether these new papers shake the conventional wisdom or not remains to be seen. Certainly, instruments can be questioned (presidentialism and trade/GDP play a prominent role, but these do not strike me as obvious choices), and their measure of resource abundance is fraught with measurement error, but at the moment, the evidence is not looking good for the conventional resource curse.