Aid and the Millennium Development Goal Poverty Target: How Much is Required and How Should it be Allocated?

Abstract

This paper uses econometric estimates of the link between aid and economic growth to ask how much additional aid is required to meet the Millennium Development Goal of halving global poverty by 2015, and how this aid should be allocated across countries. It first shows that a large increase in existing aid levels can be justified to halve $1-a-day poverty by 2015, on a country-by-country basis, under the econometric estimates obtained by Hansen & Tarp (2001, Journal of Development Economics, 64, pp. 547–570) and Lensink & White (2001, Journal of Development Studies, 37, pp. 42–65), but not those of Collier & Dollar (2002, European Economic Review, 46, pp. 1475–1500). The paper then shows that, even where an increase in existing aid levels can be justified, a much larger number of people—up to around 70 million—could be lifted out of poverty by 2015 if aid was instead allocated on a poverty-efficient basis. This cautions against the use of a country-by-country target approach when allocating aid across recipient countries.

Notes

¹ Good surveys of such attempts have also been provided by Vandermootele (2002), Naschold (2002), Christiansen et al. (2002) and Clemens et al. (2004b).

² The smaller figure ($54 billion) is based on the assumption that 22 countries with current poor policy environments improve their policy to the average level in the 43 countries with current good policy environments (which are assumed to remain unchanged). The larger figure ($62 billion) is based on the assumption that there is no such improvement. They assume that donors do not allocate any of the additional aid to countries that are currently on track to meet the poverty MDG, or in which aid is small compared with the size of the economy.

³ The latter explanation would be consistent with the findings of Ravallion (2003), who showed that, based on cross-country regressions, rates of growth in mean household expenditure per capita, as measured in household surveys, are lower on average than those in mean household expenditure per capita, as measured in the national accounts data. The explanation offered for this finding is overestimation in the growth rate of household expenditure in the national accounts data.

⁴ Equation (2) is the solution to the following equation: . There are two solutions to this equation; we chose the lower as the higher is inefficient.

⁵ The figures for low-income countries are 89 and 80%, respectively, while those for middle-income countries are 95 and 76%, respectively.

⁶ http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp

⁷ The exceptions are India and China, for which PovcalNet provides poverty data separately for rural and urban areas. In each case we combined the two sets of data to create nationally representative estimates.

⁸ While the beta Lorenz curve specification, by providing a better fit at the lower end of the distribution (see Datt, 1998), usually gives more accurate poverty estimates for countries with low incidence, in the vast majority of cases beta and generalised quadratic estimates lie within one percentage point of each other. In all cases in which the beta specification would otherwise be preferable to generalised quadratic, according to Datt's (1998) method of distinguishing them, the estimates of the poverty headcount are within 2 percentage points.

⁹ There are three main reasons for these differences. First, all of our estimates are based on group-based tabulations, while some of the Chen & Ravallion (2004) estimates are based on micro-data. Second, of the group-based tabulations, all of our estimates are based on the generalised quadratic Lorenz curve specification, while Chen & Ravallion (2004) used the generalised quadratic or the beta Lorenz curve specification, depending on which provided a better fit to the group-based data. Third, when estimating poverty in 1990, we used household data only from the survey year closest to 1990, whereas Chen & Ravallion (2004) used household survey data (where available) from surveys either side of 1990. Finally, the estimates in Table 2 refer to the sample of countries with nationally representative household surveys only, whereas the Chen & Ravallion (2004) estimates refer to all developing countries, on the assumption that the incidence of poverty in countries without nationally representative household surveys is equal to the average incidence of poverty in the regions in which they are located.

¹⁰ We estimate the immediate required increase to lie between a factor of 1.8 (using the CD estimates) and 7.1 (using the LW estimates) (details available on request). A much larger volume of aid would be required to halve poverty in each middle-income as well as low-income country, mainly because higher levels of GDP per capita in middle-income countries raise the amount of aid dollars required to raise aid as a share of their GDP, which is necessary if aid is to increase economic growth and reduce poverty. This offsets the fact that middle-income countries have a higher average CPIA score than low-income countries, which (at least according to the CD estimates) reduces the level of aid as a share of GDP required to reach a given growth target.

¹¹ These are as follows: Vietnam (East Asia and Pacific), Nicaragua (Latin America and Caribbean) and Zambia (sub-Saharan Africa).

¹² This figure differs across individual donors. The UK Department for International Development sets a more explicit target that no more than 10% of its own bilateral aid should go to middle-income countries (DFID, 2004).

¹³ This figure lies in the middle of the range of estimates of the increase in total aid required to halve poverty on a country-by-country basis by 2015 (see Section 5.1).

Additional information

Notes on contributors

Edward Anderson

The work carried out for this paper was funded by the UK Department for International Development (DFID). The authors are grateful for helpful comments and suggestions from Iain Jones of the DFID Aid Effectiveness Team, John Roberts, Andrew Rogerson, Finn Tarp, Howard White and Adrian Wood, participants at the WIDER conference “Aid: Principles, Policies and Performance” held in Helsinki in July 2006, and an anonymous referee. The views expressed are those of the authors and should not be attributed to the organizations for which they work or DFID. Annexes 1–4 referred to in the paper are available on request.

Hugh Waddington

Hugh Waddington, Development planning unit, Ministry of Finance and Economic Planning, Government of Rwanda. E-mail: [email protected]