The resource curse: which institutions matter?

Abstract

Two types of models are dominant in the current resource curse literature. One type of model studies the selection of entrepreneurs into rent-seeking versus productive activities. The other type analyses the use of patronage by politicians seeking re-election. The policy implications of the two models are quite different. The first model suggests that institutions governing the private sector ought to be improved. The second model suggests that institutions governing the public sector should be emphasized. This article empirically tests the impact of the private versus public sector institutions on the resource curse, using cross-country data from Sachs and Warner (1997a) and Polity IV. The main result is that only improved private sector institutions ameliorate the resource curse.

Acknowledgements

The author thanks Tina S⊘reide, Line T⊘ndel, Arne Wiig, Espen Villanger and participants at the Nordic Conference on Development Economics in Copenhagen, June 2007, for comments.

Notes

¹See also Sala-i-Martin and Subramanian (2003) for a more extended test of robustness. However, Stijns (2005) shows that using reserve and production data on resources, rather than export shares, yields no effect of natural resources on growth.

²Damania and Bulte (2003) include an index of the rule of law in their regressions; however, they do not include the interaction between the rule of law and natural resources required to test the hypotheses against each other.

³Mehlum et al. (2006) use data from Sachs and Warner (1997b) in their main text and Sachs and Warner (1997a) in an appendix, but qualitatively the results are the same.

⁴Damania and Bulte (2003) include both the Polity IV Democracy and Autocracy indices simultaneously in their regressions. This is problematic because the two are highly correlated (0.915 for our sample, 1970 data). They argue that the resource curse is related more to autocracy than democracy, but given the multicollinearity problem, their material does not provide a basis for this conclusion. Nevertheless, we tested whether our results would be different if we used the autocracy index rather than the democracy index, and they were not.

⁵Multicollinearity problems in including both institutional indices do not appear too severe, cf. the correlation matrix in the Appendix.