■revenue year-by-year in 1995 US$ per capita. The weighted average for these countries over the
period was that aid amounted to $72.4 while government revenue was $279.7.In assessing the reliability of aid and revenue we need some measure either of volatility or
unpredictability. In principle, these are very distinct concepts: for example, if either has a regular cycle it could be highly volatile yet highly predictable. One approach would therefore be to modelboth aid and revenue as time series, determining how well they could be predicted
econometrically, and taking the regression errors as the measure of 'unreliability'. However, thisis not the approach followed Here. In practice, both governments and the IFI teams that assist them do not make such predictions, often for the good reason that the results would lack robustness. Hence, I rely upon a much more straightforward measure of unpredictability, namely,
volatility. The measure of volatility which scales for the mean level of receipts is the coefficientof variation (standard deviation/mean). I thus calculate the coefficient of variation of aid receipts and revenue both for each country and for the weighted averagte of all 36 countries. The results are reported in Table 1. The final column of the Table reports the normalised co-variance of aid
and revenue (co-variarice/the product of the means).The results are, of course, country-specific. However, in assessing the future riskiness of aid policy makers may well regard the information from the entire group of African countries a better guide that the historical experience of just their own country. After all, the move from a single country to the full sample represents the move from 25 observations to 900. The best guide to the full sample is the weighted average.
There are two key results. First, the coefficient of variation of aid is lower than that of revenue. Aid is more reliable than revenue, not less reliable as the aid dependency school believes. Secondly, the normalised co-variance of aid and revenue is negative. This implies that there is a further benefit from aid: it acts as a buffer to revenue shocks, tending to increase when revenue is low. Hence, on the aggregate evidence, a budget with a large component of aid would be more reliable than one with a small component of aid, both because the aid component is more certain than revenue, and because it tends to offset revenue shocks.
In some countries the contribution of aid to the overall stability of government resources is striking. For example, in Uganda the coefficient of variation of aid is only one tenth that of revenue. In such a case, if reliability was the criterion for inclusion in the budget the fiscal deficit
should be reported excluding' revenue, rather than excluding aid.
To conclude, this particular belief of the aid dependency school is demonstrably false.
There is no volatility basis for the exclusion of aid from the measurement of core fiscal resources.
Table 1: The Volatility of Revenue and Aid in 36 African Countries, 1970-95
Zambia
VL WILL AID RECEIPTS CONTINUE TO DECLINE?
Aid levels have fallen in recent years and this has produced an environment of aid pessimism: regardless of the above considerations, aid will continue to decline. While it is incontrovertible that aid budgets have declined in real terms, this should not be extrapolated.
By historical accident, both the USA and Western Europe have been through a phase of fiscal retrenchment at the same time: the USA because the budget deficit became a high profile political issue, and Europe because of the need to meet the 'convergence criteria' for monetary union. Unsurprisingly, during this phase aid budgets were reduced because its recipients, though poor, are not enfranchised. However, the American budget is now balanced and the European governments have nearly all met the convergence criteria. A further powerful effect reducing aid budgets has been the end of the cold war. During the cold war much aid was used to secure political allegiance. Now that it is over the value of allegiance has fallen and so the incentive for aid has correspondingly fallen. This may well account for why in the past bilateral aid was so unrelated to the policy environment: developmental effects of aid were secondary considerations to political alignment.
Both fiscal retrenchment and the ending of the cold war are one-off effects: aid drops like a step-function. Qualitatively offsetting these effects, the global economy is growing rapidly, and prospects for continued growth have seldom been so good. Global growth works to increase aid through three routes. First, the existing group of donors become richer and this enables all expenditures to rise. Secondly, as middle-income nations catch up the developed countries they can be expected to initiate aid programs as symbols of arrival at developed country status, thereby gaining participation in the institutions of donor clubs. Thirdly, as some low-income countries grow out of poverty, a given pool of donor funds can become concentrated upon a smaller group of countries.
The main jeopardy to this process of rising aid flows is not the one-off events discussed above, but the perception that aid has been ineffective. Paradoxically, this view has become prevalent at just the time at which research has identified not only that on average aid has been ineffective, but the circumstances in which aid is unambiguously effective, namely, a reasonable macroeconomic policy environment. In the past, partly because aid was allocated according to a political agenda, it failed to reflect this policy environment. Were donors to persist with such aid