Aid, income and dutch disease

One argument that aid sceptics like to use – often without really understanding it – is that aid damages recipient countries through a macroeconomic effect known as “Dutch Disease”.  The issue has been raised again in a new working paper: so let’s go back to basics and think about what is going on and whether recipient countries are being harmed by aid.  My conclusion is that it is highly unlikely that aid could be harmful overall to a country’e economic development through Dutch Disease, and that – notwithstanding how it is sometimes presented – the econometric research does not imply that it is.

Raghuram Rajan and Arvind Subramanian argue in a new CGD Working Paper that aid inflows reduce the relative growth rate of exportable industries; and they find some evidence that aid inflows cause a real exchange rate appreciation. They say that this effect helps to explain why there is, in their view, little robust evidence that aid leads to economic growth.  Raghuram Rajan and Arvind Subramanian are both smart people, but I don’t think the evidence they present leads to the conclusion they want us to reach.

Where we agree: trade is good for growth

I am pretty sure that everyone agrees with the following:

  • The growth of trade is good for growth and development.
    There are essentially no examples of economic development that have not involved rapid export growth.
  • In a static welfare sense, exports are a cost not a benefit.
    Exports are the price a country has to pay to be able to afford imports.  In a static world (just comparing two states of the world, rather than thinking about what causes economic growth) a country is better off if it can make fewer goods for export and still afford the same level of imports. (Normally this isn’t possible but it happens in the special case of a country receiving foreign aid.)
  • There are dynamic benefits from trade which might come from exports, imports, or both.
    We know that trading economies grow more quickly than closed economies.  The long-term compound effects of this growth are a key driver of development.  We don’t know if the dynamic benefits of trade come mainly from exports (e.g. because firms which compete in foreign markets are forced to raise their productivity) or mainly from imports (e.g. because capital goods like machines, and intermediate goods, are imported; imported products are copied locally; knowledge and skill are imported).   If exports and imports both go down, this is likely to be bad for the economy. But if exports go down and imports go up, we don’t know what effect this will have.

The direct impact of foreign aid

Foreign aid has both direct impacts (it increases particular kinds of spending and consumption) and indirect impacts (the changing composition of the economy may affect long-term growth; or the investments financed by aid may effect growth).  Let’s look first at the direct effects.

  • The immediate effect of using foreign aid must be to increase the real value of imports relative to exports.
    You don’t need a regression to tell you this: it is an accounting fact. Aid is a gift of goods and services from abroad, or foreign currency which can only be used to pay for imports now or in the future.  Aid means that people in the recipient country get to consume things made abroad that they do not have to pay for with exports.  That means that the direct effect of using aid must be  some combination of (a) additional imports and (b) a shift of productive resources (e.g. people and land) from producing for exports to producing for home consumption without forgoing imports.   So the ratio of imports to exports must rise compared to what it would have been without the aid: there is no way to stop this other than to stop giving or using aid.   It also means that exports must fall as a share of income as a direct result of the aid.  This is not the Dutch Disease effect: it is simple arithmetic.
  • The size of the effect on exports depends on the way aid is used
    How much imports will increase and exports fall depends on two things: the composition of the additional demand, and the extent to which there are unemployed resources in the economy.  Some of the aid arrives in the form of goods and services from abroad (e.g. food aid, foreign consultants).  Other aid is spent directly on additional imports (e.g. pharmaceuticals, project vehicles, building materials). These additional imports have no effect on exports.  But some aid is spent on “non-tradable” services, such as paying for teachers or construction workers, who are employed locally. If there are unemployed people who are able and willing do these jobs, then total labour supply increases and there is no effect on exports.  But if supply is scarce, as it might be of skilled labour, then the scarce inputs (e.g. labour) for these activities have to be pulled in from the tradable sector (e.g. export industries). This competition for scarce productive resources increases the price of non-tradables relative to tradables, and may result in a contraction of exports (compared to what they would have been).  This is the Dutch Disease effect.
  • These direct effects unambiguously increase income and welfare of the recipient country.
    All these direct effects are welfare-enhancing.  There is an increase in imports that does not have to be paid for  which increases incomes and consumption.  Exports go down but only because the resources that were needed to produce them are shifted to non-tradable production, while the imports they were needed to pay for continue.  National income goes up by the value of the aid. (Somehow people seem to forget that income equals what is produced locally plus the value of the aid; so even if local production fell, it would need to fall by more than the value of the aid for incomes to fall overall).

What are the indirect effects of aid on exports, GDP and income?

If the story stopped with the direct effects of aid, there would be no dispute.  A permanent inflow of aid worth 5% of GDP increases national income by 5% and the country is better off.  Now lets see what happens when that works through the economy.  Here are four indirect effects we need to consider:

  • Aid may raise productivity and so increase the growth of both tradables and non-tradables
    If aid is spent well it can increase productivity in lots of ways: by providing infrastructure, educating people, increasing security, reducing disease etc.   In principle these effects on productivity and growth could be large (for example, getting rid of malaria has been estimated to add about 0.5 percentage points to the growth rate in countries of high prevalence).
  • Aid may change the composition of the economy and so change the average growth rate
    If non-tradables increase as a share of total income, and if non-tradables tend to grow more slowly than tradables, then this will – as matter of arithmentic – reduce the average growth rate.  (This effect is an order of magnitude smaller than the above – at most about 0.05 percentage points on the growth rate).
  • A country with a smaller export sector may enjoy smaller spillover growth effects.
    An increase in aid results in higher imports and possibly lower exports compared to the economy without aid. Because we don’t know whether the dynamic growth benefits of trade come mainly from spillover benefits from imports or exports, we don’t know whether this increase in imports and fall in exports will increase growth, reduce growth, or cancel out.  It is theoretically possible that this will have a significant negative effect on growth; but it may also have positive effects.
  • There may be adverse political effects
    There is a set of worries about the political effects of aid: large aid inflows might weaken the social contract, sustain otherwise unpopular regimes, feed corruption, or disrupt the effectiveness and accountability of government. These are important and valid concerns, but they are distinct from the macroeconomic effects of aid through Dutch Disease, and we are not going to deal with them further here.

How big could the Dutch Disease effect be?

Suppose a country receives aid each year worth 5% of its GDP.  If the export sector had previously accounted for 10% of the economy, then even if exports are completely unchanged, the direct effect of the increase in the denominator is that exports drop to 9.5% of the economy. (The impact on GDP and GNI will differ according on how the aid is delivered and used.)

Subramanian and Rajan do not report whether or how they have adjusted for the denominator effect in their calculations.  If they have not adjusted for the increase in national income as a direct result of aid, this would explain why they find that exports grow more slowly relative to the rest of the economy when aid increases.  But since they are very smart people, we’ll assume that they have taken account of the denominator effect and simply forgotten to tell us about it.  So having adjusted for the denominator effect, the question is whether there are any long term effects on growth from any or all of the three economic possibilities described above (aid increases productivity; the composition of the economy changes; dynamic benefits of trade are lost).

Suspending our scepticism about cross-country growth regressions, and assuming they have adjusted for the growth of the denominator and forgotten to tell us how, let’s accept Subramanian and Rajan’s estimates of the impact of aid on the export sector. They find that an increase in aid of 1% of GDP leads to a fall in the growth rate of exportables of 0.5 percentage points.  Let’s look at what this means for the impact of aid on a country’s income and welfare.

For a country to be made worse off by an increase in aid of 5% of GDP, domestic production would need to fall by at least 5% (so that the loss of income from lower domestic production exceeds the benefit to incomes of the aid). How likely is this?

Well suppose a country without aid has exports at 10% of GDP; and then starts to receive aid at a constant level of 5% of its GDP. According to Subramanian and Rajan, the export sector will begin to grow at 2.5 percentage points a year more slowly.  Leave aside for now whether there is any impact on the growth of the other 90% of the economy, the impact on the exportable sector reductes total GDP growth by 0.25 percentage points a year.  It will therefore take about 20 years before the export sector shrinks by 5% of GDP, at which point total incomes have fallen to their without-aid levels.  Over that time, the country will have received benefits (higher income and consumption) from the aid, and it will take another 20 years of below-trend income before the country is worse off overall as a result of the aid.  So if the macroeconomic impact of aid were entirely on the exportable sector, and it were of the size that Rajan and Subramanian identify, in this example it would take about 40 years of aid before the country would be worse off overall.

Could the country be worse off as a result of aid?

So according to the Rajan and Subramanian regressions, the impact on exports would mean that a country could be made worse off overall as a result of receiving aid, but it might take of the order of 40 years before the negative impact through exports is big enough to offset the direct benefits of aid.

But this is only part of the story. The overall impact on the economy depends also on what happens to the non-exporting sector.

In principle, the effect on the non-exporting sector could be negative.  The contraction in exports could lead to a negative impact on the rest of the economy because the country is deprived of  some of the benefits of having to compete in the world economy.

For the overall impact to be negative, the non-tradable sector would need to be adversely affected by slightly lower exports and this effect would need to be larger than any benefits from higher imports and any benefits from how the aid money is spent.

This is theoretically possible, but it seems highly improbable.   The non-tradable sector must be growing to some extent (since that is what is crowding out the export sector).  At worst, it might be growing less fast than the export sector is contracting (so that the overall effects are negative after even more than 40 years.)  But there are good reasons for thinking that the aid would accelerate growth in the non-exporting sector, at least to some extent.  There will benefits from higher imports, including access to intermediate inputs, skills, and machinery, and benefits to the domestic economy of competitive pressures from imports.  And at least some of the aid is spent in ways that should increase productivity – for example, by improving education and skills, reducing sickness, enhancing security, building transport and communications infrastructure, and improving government administration.  For the overall effect on the non-tradable sector to be negative, all these benefits would need to be overwhelmed by the harm done to the non-tradable economy because the export sector is a little smaller than it would otherwise have been.

Even a very modest increase in growth in the non-tradable sector resulting from aid would be enough to offset the impact on GDP of the contraction in the export sector.  In the numerical example above, an increase in growth of just 0.05 percentage points would be sufficient to offset the negative impact on exports.

Empirical findings

The Rajan and Subramanian findings tell us that the export sector contracts as a result of aid (assuming they have correctly adjusted the denominator effect); but the effect is small.  Other things being equal, a country would have to receive aid continuously for of the order of 40 years to be worse off overall as a result of aid’s impact on export earnings.  But the overall effect on the economy plainly also depends on what happens in the non-tradable sector.  This is an empirical question:  the impact on the non-tradable sector could be positive (because of the benefits of the additional imports, and the impact of aid) or negative (because the positive spillover effects of a vibrant export sector are reduced).  You need to look at the effect of aid on growth overall, not just on the export sector.

The stastical power of cross country growth regressions is not good enough to say very much at all about the impact of aid on growth.  But if you hold your nose, the most likely interpretation of the data seems to be that there is no linear relationship between aid and growth, but there is a non-linear relationship in which aid increases growth but with diminishing returns.   If so, then the data tends to be consistent with the theoretical arguments: aid may reduce exports a little, but nothing like enough to turn the overall impact negative.


Rajan and Subramanian would like us to believe that aid can cause dutch disease and so make a country worse off.  But their research does not support this conclusion: the common sense view that a country benefits overall from being given free money is more likely to be right. Their empirical findings tell us that aid leads to exports being a smaller share of GDP.  This we already knew, since it is a matter of arithmetic; and it does not prove that the country is worse off immediately or in the long run. There is broad agreement that exports are critical for development, but it does not follow, as is often lazily implied, that the arithmetical effect of aid of reducing exports as a share of GDP harms a country’s prospects for development.  Looking at the impact on exports alone, their results suggest that it would take about 40 years for incomes to fall in aggregate as a result of aid.  But the overall effect depends on how what happens as a result of aid to growth in the rest of the economy.   There could be an indirect effect on the rest of the economy from the contraction of exports, but there should also be positive effects from the expansion of imports and the productivity effects of aid. We don’t know the relative size of these effects, but seems likely that the total effect is positive overall (it is the expansion of non-tradables that causes the contraction of exports, after all).  There is some rather flimsy empirical support for this from the  evidence of cross country growth regressions, and none for the alternative view that the effect is large enough to outweigh the benefits of aid.

(See also David Roodman’s review of this paper.)

4 thoughts on “Aid, income and dutch disease”

  1. Owen, I think this needs one correction, and it is actually in favor of your conclusion. As I explain in my post about this paper, and as Rajan & Subramanian well understand, the paper is only capable of discerning relative impacts on exporting sectors. As I wrote, “As far as this study goes, it is entirely possible that aid raised growth in all industries—just not quite as much in the most export-oriented ones….One could with equal logic celebrate this study for showing that aid is good for non-tradeable industries.”

    The reason it can only discern relative effects is that it includes fixed industry and country effects (to get technical). It should be said that this is also be a strength of the paper, because having these fixed effects makes it more conservative than most cross-country growth regressions. It is of course hardly iron clad.

  2. I’ll admit to be completely out of my depth when it comes to dealing with economic statistics. But this particular part of your critique does not seem quite right to me and thankfully I don’t think it is a matter of stats. You state “For a country to be made worse off by an increase in aid of 5% of GDP, domestic production would need to fall by at least 5% (so that the loss of income from lower domestic production exceeds the benefit to incomes of the aid).”

    This statement is obviously correct from a technical, short term, economic point of view. But I think it misses something. Because Aid is ideally temporary and domestic production, again ideally, not, I would suggest that if Aid has ANY (or least a decent measurable amount after accounting for the “denominator effect”) negative impact on domestic production, then we have a problem. Using aggregate GDP is misleading because it ignores the importance of the character of the different elements of which that GDP is made. When Aid inflows end it may not be possible for the domestic production to “bounce back” and so in the long term that Aid may have contributed to lowering a country’s domestic production and in general have a distorting effect on the composition of its economy. You might argue that economies need such distorting, but given the track record of Aid economists, I am a little skeptical that they know how to poke and prod correctly. Increasing domestic production is ESSENTIAL (that is one of the few things that we do know) for any country’s long term health, economic and otherwise, and we should be very careful that Aid doesn’t stand, even a little bit, in the way of that. In sum I think our “do no harm” standard should be a lot higher than being okay with a 5% Aid inflow that leads to a 4.9% decrease in domestic production.
    And this is probably unfair but many African countries have been receiving Aid continuously for 40 years. So even if we use your aggregate demand standard, it is possible that they are worse off.
    Let me know if I am missing something, the math makes my head hurt.

    Owen replies:

    Justin – you are right. But this is because I am not making the point sufficiently clearly. The point I am making is that aid has a direct benefit (extra income) which needs to be taken into account and is often forgotten in assessing the effect of aid.

    Consider this example. There is a lottery advertised on British TV at the moment which pays out £40K a year for the rest of your life. Suppose you won that lottery and, as a result, you decided to work part time because you wanted to voluntary work instead; and so your salary goes down by £10K a year reflecting your shorter hours. In this case, it would be very odd to say that you are “worse off” because your salary has gone down. Your total income is higher AND you are able to spend more time doing the things you want to do.

    Now of course, aid isn’t a permanent, guaranteed income for life. If you gave up work because you thought you had won the lottery, and then the lottery company went bust, you might end up worse off (depending how easy it is to get back in to work).

    So you are right that to say there is a fundamental difference between the income obtained from your own production and the uncertain aid from capricious donors. But my point, which is also true, is that at any given time, you are only “worse off” in terms of your income if your production has fallen by more than the value of the aid you are receiving.

    I hope that makes it clear?

  3. The following comment, even if it’s correct (which I’m not at all sure about – I always get in a muddle with this stuff) might just be a semantic nit-pick …

    “So the ratio of imports to exports must rise compared to what it would have been without the aid: there is no way to stop this other than to stop giving or using aid. It also means that exports must fall as a share of GDP as a direct result of the aid”

    That’s gross domestic product. If you are thinking about total quantity of resources at the disposal of the recipient country, then you’d be right: free gifts of imports (aid) would necessarily decreases the share of this total accounted for by exports (or the income from sale thereof). But (I think) imports are not part of GDP. If Rajan and Subramanian are measuring GDP, then increases in aid/imports do not mechanically (arithmetically) reduce the share of exports in GDP. That is to say, the imports that constitute aid are not counted in the denominator GDP. (If they were, you’d really have a job on your hands explaining why cross country regressions don’t pick up an increase in GDP resulting from aid).

    Owen replies: That’s right: GDP includes exports net of imports. Aid would count directly in GNI.

Leave a Reply to Luis Enrique Cancel reply

Your email address will not be published. Required fields are marked *