The Effectiveness of Aid: The Impasses of the Macroeconomic Approach and the Idea of Selectivity.
On the occasion of the Paris Forum on the Effectiveness of Aid (February 28 - March 2), jointly organized by the OECD, the World Bank and the United Nations at the invitation of France, the Notes du Jeudi (Thursday Notes) are reviewing some key factors of the question of effectiveness. After a general presentation on the main themes involved (NdJ nº 24), the macroeconomic approach is examined here.
The question of the macroeconomic effectiveness of aid fills a large part of the literature on ODA (Official Development Assistance). There is, in fact, nothing absurd about seeking to show the impact of aid on the well being and living standards of populations in general, above all when the volume of this aid is significant at the level of the country as a whole.
1. Search for the Macroeconomic Results of Aid
To measure the macroeconomic impact of aid, it is tempting to isolate, in a complex process, a relation of cause to effect between volume of aid and economic performance, in general, the rate of growth. For that, there are three possibilities:
comparison with what would happen without aid
comparison with the expected objectives of the aid
search for correlations between aid and growth in a sample of situations.
The first way is a counterfactual method. It is necessary to compare the situation of the country with aid to a fictional situation without aid. From a macroeconomic point of view, this counterfactual can be conceived only by modeling the beneficiary countries and the effects of the aid. This appears difficult to do and has only rarely been explored up to now. It undoubtedly would be more reasonable at intermediate levels, by constructing counterfactual sectoral scenarios.
The second requires general objectives for the aid. This approach is stimulated by the adoption of the Millennium Development Goals (MDG) and by the attempts to evaluate the cost of attaining these results. Thus, for the first time in the history of aid, there would be a connection between the means of the ODA and its objectives. This connection would be found at the level of the beneficiary countries through macroeconomic objectives that are consistent with the sectoral policies that allow the quantified objectives to be attained.
It is no longer a question of measuring only the correlation between the aid and the rate of growth, but between the aid and the different areas of the MDGs. Monitoring the indicators of the different objectives at the local level as well as aggregated at the world level is a significant challenge for the donor community, but this work has only just begun and straight away is encountering serious obstacles concerning the availability of data.
Up to the present, the comparative spatial and/or temporal method was the one most used. This method is an opportunity for the use of econometric analyses:
- in cross sections (different countries)
- in longitudinal sections (different time periods)
- both together: sample group methods (different countries at different time periods, which makes it possible to multiply the number of observations in the samples).
Dozens of general empirical econometric analyses have been carried out since the 1970s on the growth of third world countries by including a measure of the impact of aid, and numerous specific studies of the impact of aid have also been attempted, above all since the beginning of the 1990s.
The results of the first studies tended to support the idea of the ineffectiveness of aid, or the impossibility of proving its effectiveness. Even at the macro social level (infant mortality, level of schooling), there did not appear to be any correlation between the volume of aid received and improvement in the corresponding indices. Recent studies go in the opposite direction, asserting that aid and growth are, in general, positively correlated.
Likewise, the link between aid and internal savings (and thus investment) is viewed differently depending on the study. On the other hand, it is generally admitted that aid results in an increase in consumption and even more, an increase in public expenditures, because of the expenditures linked to foreign aid in the areas of health and education.
A strongly disputed study published in 1997 by Burnside and Dollar, and dealt with later, supported the idea that a correlation between aid and growth well and truly existed “as soon as ‘good policies’ exist”. In the presence of favorable situations, aid would have an obvious impact. This result has received considerable publicity. It was taken up right away by the World Bank (Assessing Aid in 1998) and the OECD (DAC Dossiers, 1999 Report) and made use of to propose a selective allocation of aid: “It is in the countries that follow sound policies that aid will be the most useful”, adding that the others would not be affected by reduction in aid that has little effect. We will return in the 3rd part to this idea of selectivity of aid allocation.
2. Conceptual Problems
Whatever the results of the empirical studies cited above may be, the search for an econometric relation of the type “more aid implies more growth” poses so many conceptual and methodological problems, without even leaving the field of macroeconomic analysis, that this analysis itself can be contested.
Common and interconnected causalities. Some factors, some events, can influence the level of aid and the level of growth at the same time. This is, for example, the case with an exceptional natural disaster, which is going to drain significant quantities of aid while negatively influencing the indices of growth.
Furthermore, the level of aid is not a totally exogenous datum. There are two reasons for that:
On the one hand, the level of growth can itself influence the amount of aid through the working of allocation criteria. It has a negative influence if the allocation criterion is based on need and a positive influence if it is effectiveness. The balance between these two criteria is difficult to evaluate and it has, moreover, undoubtedly shifted recently. It is therefore difficult to draw any conclusions.
On the other hand, the general level of development of a country improves the absorption capacity for aid and facilitates its disbursement. The ODA measurement (by the DAC of the OECD) is made, not on allocated quantities (“commitments”), but on quantities effectively disbursed (“disbursements”). Therefore, more development allows more aid.
In the two cases, there is an interconnected causality between aid and growth. Serious econometric studies could deal with these complex relations, but they require the existence of much more reliable and detailed data than what is currently available.
Objectives. The objectives of the ODA are overall to ensure economic development. Nevertheless, the objectives are varied. This can be seen in the case of the MDGs. Even economic themes can be a long way from the search for growth alone, to the point of appearing as completely distinct. For example, the first adjustment plans in some countries had the objective more of avoiding the violent collapse of the State than restoring equilibrium to the economy. A correlation between aid and growth can exist without growth being the primary and direct objective. Yet the correlation can be weak without necessarily implying failure.
Temporality. Empirical studies search for a link between economic growth and disbursed aid over the course of the same period of time (generally, 4 or 5 years according to the sample groups used). That comes down to considering the effect of aid to be immediate, which is, obviously, contrary to reality. Investments in human capital, for example (health, education), can have an effect only in the long or very long term.
It would be necessary to be able to determine the horizon of the expected returns from the activities financed by different forms of aid and search for correlations between current growth and past aid. The difficulties of such an undertaking can be imagined.
Multiplier effect. By searching only for the immediate effects of aid on growth, the analysis implicitly chooses to overlook the qualitative content of long term actions, to search only for their immediate, direct profitability, which is inherently weak, on the one hand and for their Keynesian effect, on the other hand. The teacher is thus taken into account by the redistribution of his salary, the expatriate technical assistant by the portion of his income spent on the spot and a road by the required sub-contracting and wages paid.
However, a road indeed has a measurable economic profitability, but over a long period of time. As for support to education or institution building, the profitability is not in question, but neither its measurement nor its term is known.
The portion of aid that could lend itself to this multiplier effect (local spending) is not, for all that, negligible, although it is difficult to estimate. There is, then, a real question on this subject: why does this effect, even partially, not appear more clearly? The reasons are undoubtedly the same as in the cases of other countries where the Keynesian multiplier is barely apparent:
If the variation in incomes (resulting from the investments of public aid) is strong and quick, internal supply follows with difficulty and it is imports that benefit the most. Increased inflation is likely.
If the variation in incomes is weak, that is, if the level of public spending allowed by the aid is sustained, it is difficult to detect the correlations. Now, the general level of disbursement of aid in each country is rather stable, which explains the absence of monetary perturbations specific to the ODA.
Comparability. The empirical studies mentioned above concern, for the most recent of them, a number of countries, varying from 50 to 60 over 5 or 6 periods, or about 300 samples. While this number appears relatively high, it hides a double reality:
- most of the countries are not in any way comparable (example: China and Sao Tomé)
- the countries that could be subject to comparisons, because their general conditions are similar, are subjected to the same criteria for aid allocation and thus experience similar situations in this regard also. In the Sahel region, every country that can receive aid does so, at a level that is undoubtedly close to saturation. In countries where there is no aid, this absence is linked to a particular political situation (example: Burma), whose effects can be much more significant than those resulting from the absence of aid. What is more, the levels of aid to the Sahel seem relatively stable over time, which reduces comparability between periods.
Nonlinearity. Supposing that an effective relation between volume of aid and rate of growth is conceivable, nothing argues in favor of this relation being linear, as sought by Burnside and Dollar, for example. Recent studies suppose it to be, rather, a decreasing marginal return, which appears much more likely. Threshold effects are also possible, in connection with the idea of absorption capacity (Cf. Note du Jeudi [Thursday Note] nº24). If the volume of aid is greater than this absorption capacity, the effectiveness of the marginal dollar can become nil. By introducing qualitative factors, nonlinearity becomes even more obvious, which again complicates the econometric approach.
Aggregation by country. The level of aggregation for the aid used is the country. It is at the level of countries that aid and growth are measured and a correlation sought between them. This is unsatisfactory for two reasons:
- On the one hand, differences in demographic size between some countries make some comparisons meaningless;
- On the other hand, nothing proves that the type of aid, or the distribution by type of aid, is the same in each country. The type of aid appears to be a cause of possible variation in effectiveness at least as important as the country where it is distributed.
Processing of the data. The scientific quality of empirical studies is open to criticism. For example, by taking the same data as Burnside and Dollar, but adding to it, among other considerations, fixed effects by country, Hansen and Tarp challenged every conclusion of the Burnside and Dollar’s work. That researchers end up with opposite results based on the same data should call attention to the fact that initial hypotheses are not always made explicit or justified in the presentation of analyses.
Origin and quality of data. Correlations are sought between two magnitudes whose measure is not obvious. Concerning aid, uncertainties exist, above all at levels that are not very aggregated. Concerning growth of GDP, doubts are even greater, given the weakness of national statistical systems, particularly for capturing the activity of the so-called informal sector or sectors. Even if it is possible to conclude that the same structural errors are repeated year after year, it is clear that the growth figures for GDP are far from reflecting the economic life of Third World countries.
Moreover, the ODA aggregates things as different as debt cancellation, budgetary aid, food aid, humanitarian and emergency aid, training, or development projects, with or without technical assistance. And among development projects, things as diverse as infrastructures, productive capital, services, and consulting are aggregated, and this in areas as varied as education, health, culture, industrial and agricultural production, urban and rural planning, the environment, etc. Aid is a flow of money, materials, information, thought; an aid to financing, process, governance.
Hence a great diversity in results is expected. The question of data thus returns to quantitative considerations, which enter into the framework of the econometric approach, as much as much more qualitative considerations.
To search for an empirical link between the macroeconomic data concerning financial quantities of aid received and the rate of growth, beyond all the reservations cited above, is to consider that aid, by itself, is a factor of growth. That is part of the original view where economic growth is reduced to the capacity to save and where the only object of aid is to fill financial gaps. In reality, “the goal of aid cannot only be to act on the capabilities and opportunities of the beneficiaries in order that the latter can fulfill the development objectives by themselves. The relations between the growth of human and physical capital and macroeconomic objectives such as growth and reduction in poverty are complex and very long term. No current measurements of the macroeconomic impact of aid can pretend to take this relation into account.” (J.D. Naudet).
3. Selectivity of Aid: an Ambiguous Idea
To support their assertion, Burnside and Dollar presented a model for 40 poor countries, of which 15 had attained at that moment a “policy level” adequate to make the aid effective. According to this econometric model, by reallocating the aid received by these 40 countries to the 15 that could make good use of it (in the latest period), the average annual rate of growth per inhabitant of all 40 countries would have passed from 1.10 to 1.44%.
It was, moreover, understood that the countries removed from aid would be encouraged to improve their “policy level”, in order to regain higher levels of aid.
Armed with these results, to which unprecedented publicity was given, the subject of selectivity again appeared in the ODA. Implemented gradually by most of the agencies for around five years, it has replaced the one that had been more or less applied up to then, based on need (level of poverty). 
However, the results presented are not convincing. The weakness of the argumentation, the fragility of the data and interpretations, the very strong critiques that have followed, do not make it possible to explain the success of the concept of selectivity by these results of macroeconomic analysis alone.
The causes of this success lie elsewhere. The concept of selectivity is above all a common sense idea. Wanting to prove it by macroeconomic analysis was a challenge to common sense and it turned out to be trickery of the first order.
Why has selectivity been imposed? Because it responded to real needs, beginning in the middle of the 1990s, with the “exhaustion of aid”. It was imperative to:
Leave the logic of the standing order, which prevailed, for the most part, until then, notably vis-à-vis governments that were notoriously corrupt and incompetent.
Renew the image of aid.
From this point of view, the introduction of the principle of selectivity was necessary, and, it turns out, a success. This must not, however, overshadow certain realities.
First of all, there are other ideas of common sense, a bit different, that have also been supported (with the same doubts and limits) by macroeconomic analyses. Hansen and Tarp, for example, have shown that policies considered as good lead to growth, but they are neither sufficient nor necessary to make aid effective. Durbarry concludes that they improve the effectiveness of aid, but do not influence it. He adds that there are, above all, general conditions already satisfied, and particularly superior human capital, more than supposed “good policies”, that influence the effectiveness of aid.
In other words, aid is undoubtedly less effective, in the sense that it presents fewer discernible results through traditional analysis, in countries where the conditions are more difficult: low level of literacy, prevalence of AIDS and weak governance. That is not to say that it is less useful.
On the other hand, against the current of the ideas of alignment and ownership, selectivity can be analyzed as a prior super conditionality, concerning not only governance in the narrow sense (level of corruption) but also all the country’s policies (economic opening in particular). “Sound policies” refer to standards elaborated outside of the country. For countries removed from the ODA, it can be observed that these prior conditionalities imposed by the donors work (their weak ability to offer incentives is, in general, recognized).
Conversely, aid resources are certainly limited, but agencies nevertheless need to disburse the resources they possess. They thus need so-called virtuous countries, to which labels of worthiness are sometimes awarded a bit too quickly. Aid, assimilated to a reward to governments, thereby loses, in part, one of its principal roles, i.e., its support for internal change.
Finally, supposing that selectivity brings increased effectiveness to all aid, there remains the question of justice, opportunely recalled through the definition of the MDGs: the objectives defined in the autumn of 2000 should not only be attained at the world level, but also in each country. A uniform selectivity by every agency leads to an over-allocation of aid to certain countries, declared virtuous, and the concomitant existence of “orphan” countries , whose populations suffer from a double deficit of aid and governance. Aid fails in its vocation if it no longer attempts to aid the neediest countries also, even if they are poorly governed.
The macroeconomic approach has essentially been used lately to justify the concept of selectivity, which rests, in fact, on other analyses and other needs. The use of performance criteria in the allocation of aid is justifiable, but cannot be based on this type of analysis. It can and must be justified otherwise, and taking account of the contradictions that it includes.
The classical macroeconomic approach is wrong in numerous poor countries, where the economic fabric is not very complex, the State apparatus significant, institutional dysfunctions numerous and aid sizeable in relation to the weak GDP. It is hampered for numerous technical reasons and also because, more fundamentally, decisions are made by fewer actors. Their individual behaviors have more significance than structural factors. Approaches in terms of analyzing behaviors would undoubtedly be more useful.
In countries where there are few actors, a weak absorption capacity and large and very diverse aid (thus difficult to aggregate), aid cannot simply be assimilated to its disbursement volume, and the macroeconomic approach is an inadequate form of analysis. Qualitative factors predominate and it is thus necessary to use tools that describe the behaviors of actors, more particularly when they are macro-actors.
 Except, paradoxically, for debt cancellations, since the higher volumes of cancellations, which tap large parts of the ODA, reflect significant outstanding payments and thus poor quality governance.
 From this point of view, selectivity must not be confused with the principle of complementarity, where the donors divide up sectors, even countries, according to their comparative advantages or their respective priorities, without the beneficiary countries, in principle, seeing their overall amount of aid changed.