The Economist has reported a paper by Oya Celasun (IMF) and Jan Walliser (World Bank) which looks at the impact of unpredictable aid:
They show how unpredictable such aid flows are. The paper finds that the average absolute difference between aid promised and aid given was equal to 3.4% of each sub-Saharan African nation’s GDP between 1990 and 2005.
The paper is pretty interesting. It reiterates the difference between volatility and unpredictability; volatile-but-predictable aid (eg lumpy payments for a large infrastructure) is rarely destabilizing and is not problematic. It also reminds us of the positive reasons why we might want aid to be unpredictable (eg because project implementation is slow or because the government’s commitment to poverty reduction becomes uncertain) which they distinguish from negative reasons for unpredictability, such as bureaucratic delays, changes in donor political climate, or fickle interpretation of conditions.
The empirical findings are pretty striking. On their samples:
- on average annual aid disbursements deviated by 3.4% of GDP from aid commitments in sub-Saharan Africa (there is a trend decline in this deviation, to 2.8% in recent years, but it is still a massive economic disruption to a country to have that degree of unpredictability)
- no all unpredictability is a shortfall: sub-Saharan Africa received on average 1% of GDP more aid than was committed;
- up to 40% of the variations can be explained by changes in country circumstances; the remaining 60% is unexplained (the authors’ hypothesis is that this unexplained component is fickle donor behaviour such as administrative delays)
- budget aid disbursements fall short by about 1% of GDP from projections, representing about 30% of budget aid promised on average. Budget aid is less predictable than local tax revenues.
- Governments adjust to budget aid shortfalls by accumulating more internal debt and reducing investment spending; the losses in investment spending are not reversed in good times: budget aid windfalls lead to higher government consumption and some reimbursement of domestic debt. Thus the overall consequence of unpredictable budget aid is increased government current spending and reduced government capital spending compared to providing the same amount of aid predictably;
- if you believe (with Easterly etc) that public investment is positively and reliably correlated with long-term growth, then this means that lack of predictability of budget aid has a quantifiable and significant negative impact on long-term growth and poverty reduction.
I was struck that budget aid appears to be more predictable than other aid (though still woefully unpredictable). The mean absolute deviation in budget aid (1% of GDP, using the IMF projections) is much lower than the mean absolute deviation in overall aid (3.1% of GDP, using the DAC data). This is contrary to the conventional wisdom, which is that budget aid is more unpredictable than projects.
I also thought that Celasun & Walliser underestimate the harm done by unpredictability of budget aid. They focus on the resulting shift from government investment to government consumption, which may have costs (though I do not entirely share the fetish for investment over consumption). The costs of unpredictability go much wider: for example, within current expenditure, predictable aid could be used to restructure public service wages; whereas unpredictable aid is more likely to be used to buy in consultants to fill the gaps. There are also macroeconomic costs that are not included here (for example, higher borrowing leads to higher government interest payments, and also to higher interest rates which crowd out private investment). The agencies that give this unpredictable aid would complain like mad – and rightly so – if they did not have predictable budgets from their own Treasuries; so why would public services in developing countries be any different?
For the purposes of this paper, “predictability” is defined in a short-term sense (the gap between commitments for a given year, 0-6 months ahead, and disbursements that year). It is clearly a huge problem that in-year disbursements deviate from commitments by as much as 3% of GDP on average, and it is a problem which donors should do something to fix. But there are other dimensions of predictability – such as the ability to budget 3 or 5 years ahead – which may be just as important, or more so, and which are not discussed in this paper. My belief is that 3-year commitments which are sufficiently solid to be programmed in the budget are worth much more to developing countries than ten-year partnership agreements which are insufficiently reliable for them to be able to programme; if so, we should be emphasizing firmer commitments rather than longer time horizons.
While short term (in-year) unpredictability is a serious problem, the best solution to this may not lie in persuading donors to behave better (on which we have achieved little) but in helping to promote institutions that could help developing countries to smooth out the flows. This might take the form of insurance markets or mutual pooling arrangements, as well as improving access to financial markets. It may also be worth exploring the idea that donors could establish escrow accounts (“bathtubs”) through which aid would flow. Tackling medium term predictability, by contrast, probably does require changes in donor behaviour.
And finally, as you would expect, I strongly agree with Celasun & Walliser in highlighting the importance of greater transparency of aid data.