- Comments and Discussion
Michael Kremer: This is an interesting and indeed provocative paper, arguing for a large increase in foreign aid to Africa to help the continent escape from a poverty trap. I agree that aid should be increased, but I will take issue with a few of the building blocks in the authors' analysis and will suggest an alternative case for aid. In particular, I will argue that the limited empirical evidence seems just as consistent with conventional growth models, in which government quality and policies determine steady-state income, as with the idea that Africa is caught in a poverty trap. This should not be taken as an argument against aid, however, because there is little evidence for the "new Washington consensus" that aid cannot work in areas with weak governments and that conditionality is a failure. In fact, the case for aid may be greater where governments are doing a poor job. I will conclude by arguing that planning development expenditure around the Millennium Development Goals (MDGs) risks generating distortions.
Poverty trap models suggest that if countries can get over a certain income threshold, they will take off economically. These models therefore suggest that aid can have a powerful effect by raising countries over this threshold. The basic microeconomic case for poverty traps typically involves nonconvexities. However, even if nonconvexities exist at the microeconomic level (for example, lumpy capital investments), it is not clear that they exist at the macroeconomic level. If a certain level of transport and communications infrastructure is needed for development, for example, one could imagine a country starting out by building this level of infrastructure in a single port or capital city and then expanding outward. This might eliminate nonconvexities at the macroeconomic level. [End Page 217]
Setting aside the theoretical issue, the question arises of how one would empirically distinguish a poverty trap model from a conventional model in which government policy determines steady-state income. In this model, countries converge to steady-state income from above or from below, depending on initial income. One can consider both microeconomic and macroeconomic approaches.
There are certainly some people in Africa with money to invest, and in some poverty trap models those people should be able to set up very prosperous enterprises, hire more and more labor, and expand their enterprises rapidly. However, we do not see that happening. In fact, many people with money in Africa move it to Europe or elsewhere rather than take advantage of the potentially huge returns available under poverty trap models to people who can reach a certain scale of investment.
At the macroeconomic level, poverty trap models suggest that African countries that attain a threshold income level should then take off. Again, this does not appear to be the case. In 1998 Gabon's GNP per capita was $3,870, but by 2002 its GNP per capita had fallen to $3,060.1 In 1980 Nigeria's GNP per capita was almost $1,000, but in 2002 it was only $300. Many argue that governments tend to waste oil revenue, but, in a pure poverty trap model, cases such as Nigeria are difficult to explain. Nor is the problem limited to countries with oil: consider the case of Zambia, where GNP per capita was $680 in 1981 but $340 in 2002. Nor is it limited to minerals in general: Côte d'Ivoire had for a while a relatively high income per capita for Africa, but then the economy collapsed. Zimbabwe had a GDP per capita of over $1,100 in 1982, but it had fallen to $480 by 2002 and is presumably below that now. South Africa's income per capita is certainly above the level associated with poverty traps in this paper, but it has not taken off economically.2 One relatively well off African country with good economic performance is Botswana, but with strong economic policies, democracy, and ethnic homogeneity, its experience is just as consistent with the story that good government is central as with the poverty trap story.
Under a conventional model, countries experiencing a large negative shock should grow quickly once conditions change for...