[Update: Arvind Subramanian has replied on Aid Watch to this post, and I have commented there.]
If the United States gives Uganda a billion dollars, Uganda becomes richer. However, Uganda can only enjoy that money by spending it; and since the money is dollars Uganda must buy goods and services from the United States. So when the United States gives Uganda a billion dollars, exports follow. With enhanced access to products easily tradeable across national borders, Uganda can devote more economic energy to making non-tradeables such as restaurant meals and bicycle repairs. With the wealth, then, comes economic change.
That's a simplified picture of the impact of aid on a receiving economy. One can imagine many hidden complications: Uganda might stash the money in a New York bank, deferring the spending. Or it might buy yen with the dollars so that Japan would import California wines while Uganda got Sony TVs. Most of these complications do not change the essential picture, however. Aid is free imports, so while making the recipient richer, it also compresses the recipient's production of tradeable goods. This compression is known as Dutch Disease because something like it happened to the Netherlands after natural gas was discovered there. Suddenly foreigners were clammoring to hand foreign currency to the Dutch in exchange for the gas. The country's hydrocarbon industry grew, and since there were only so many workers and so much capital, other industries got squeezed down, disrupting the lives of many. You can imagine the channels through which these changes played out. The natural gas company started hiring, for example; that bid up wages; that forced other companies to lay off people or completely shut down. Other ripples ran through exchange rates.
Perhaps because of the name, "Dutch Disease" is often and incorrectly assumed to make a country poorer in aggregate. It need not. In the opening scenario, Uganda is still richer. Yet there is one important way in which Dutch Disease could do aggregate harm: if it suppresses industries that are dynamic, most apt to foster the constant churning and innovation that historically has been the main force that lifts productivity and reduces poverty. If Japan had discovered oil in the 1950 and abandoned manufacturing in favor of extraction and export, would it be as rich today? Perhaps a short-term gain in wealth would have come at a long-term cost. Aid could do the same thing, making countries worse off long-run by stymieing their entry into manufacturing.
Back in 2005, my colleague Arvind Subramanian and Raghuram Rajan, both then at the IMF, wrote a creative paper to partially check for this possibility. The respected Journal of Development Economics has just accepted the final version, and Arvind today blogged it at Aid Watch as a guest of JDE co-editor Bill Easterly.
Rajan and Subramanian indeed find evidence that foreign aid causes Dutch Disease. In the 1980s and 1990s, the more aid a country received, the less growth (or more shrinkage) it saw in industries that tend to export the most. Last month I blogged my broad skepticism of attempts to gauge the effects of aid by analyzing cross-country data. How much do I trust this one? The best answer I can give is weak and relative: more than most. The method is clever and the conclusion plausible.
A big concern with most cross-country aid impact studies is that subtle, unmeasured national characteristics---culture, history, quirky economic circumstances---simultaneously affect both variables of interest. Maybe something that caused Uganda to grow faster, like having nowhere to go but up after a civil war, also attracted more aid, making it look as if aid caused growth. Perhaps uniquely, Rajan and Subramanian attack this problem by studying not countries, but industries within countries, via a clever technique originating in Rajan and Zingales 1998. This allows them to abstract from all unmeasured characteristics that simultaneously affect aid and national economic growth. It is a big step up, but it still works only with a proviso: if we believe those factors have the same effect on all industries within a country---e.g., if Uganda's starting from next-to-nothing after civil war gave the same natural lift to growth in all industries, say 1%/year. As always, you have to assume something to prove something. Since there can be no randomization in such cross-country studies, the assumptions needed are always substantial.
Another source of caution is that the history of the statistical study of aid's impacts can be seen as cyclical, in which clever new techniques are introduced with hope, only to be undone by closer scrutiny. I have not scrutinized Rajan and Subramanian's regressions closely, but Aart Kraay at the World Bank did four years ago and his critiques may still apply.
A final caveat---actually a compliment for modesty of ambition. This study does not and cannot assess the overall impact of aid on economic growth. (Another Rajan and Subramanian paper tries to and you can guess my thoughts about it.) The study at hand looks only at whether certain industries in aid-receiving countries grew more or shrank less than others. This is a subtle point and it makes the paper easy to mischaracterize. 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. That is not the impression you'd get from this tweet:
One could with equal logic celebrate this study for showing that aid is good for non-tradeable industries. In October I regretted an instance in which Bill seemed to hold his dissenters to higher intellectual standards than his supporters. Lord knows no one is perfect in this regard. Indeed, since this paper does not contradict my thesis about the difficulty of proving with cross-country data that "aid works," I was tempted to let it pass without comment. But having skeptically blogged a paper that does contradict my thesis, I decided to strive for my own standard of equal-opportunity criticism.