Last month I was privileged to participate in a conference on foreign aid at Oxford University.
Why do governments provide foreign aid? To which countries? And with what effects on economic development?
These are questions we considered. The participants included academics, like me, as well as practitioners - and people who straddle both worlds. Now, sometimes we academics have trouble expressing our nuanced ideas in a readily accessible format for a broad audience. This is why I was happy to see an innovative way to present the results of sophisticated data analyses with a picture - and the name is just perfect for The Vreelander: V-Graphs.
Two graduate students from Columbia University, Jordan Kyle and Elizabeth Sperber, presented research they've done with Jonathan Blake on "foreign aid fractionalization." What is "fractionalization" about? Well, sometimes donor countries give a lump some of aid to a developing country to be used for a broad range of purposes, but other times, donor countries "fractionalize" their aid into small packages for a number of very specific projects.
Just about everybody who studies foreign aid condemns "fractionalization." It increases transaction costs, places additional costly burdens on the bureaucracy, and limits the use local knowledge to respond to changing circumstances.
So why would a donor country "fractionalize" aid?
Blake, Kyle and Sperber suggest that it may be a question of how much the donor trusts the developing country's desire and ability to use the aid effectively. The more the donor trusts the recipient, the more of a "lump sum" that may be provided. If the donor doesn't trust the recipient, the aid may be "fractionalized" so that the disbursements can be micromanaged from afar. If a developing country has a highly-reputed effective bureaucracy, a donor may trust it to do a good job, but if the country's bureaucracy is known to be inept, the donor may "fractionalize."
The scholars test this hypothesis on a wealth of data from many different donors. They find that the effect of the developing country's bureaucratic quality influences different donors in different ways. Lately, the United States, for example, fractionalizes aid less for developing countries with high-quality bureaucracies, and fractionalizes more for the developing countries with low-quality bureaucracies, just as one would expect. But not every donor country follows this pattern. Many countries don't seem to care about the bureaucratic quality of recipient countries. And still other countries - like France - actually fractionalize more for the high-quality bureaucracies and less for the low-quality bureaucracies. (Why would the French do this? A mystery.)
I could go on listing the effects country by country, but Blake, Kyle and Sperber have a better way to present the results with a picture - the V-Graph!
Just a li'l bit of background: when we perform statistical analyses, we look for two things: (1) the direction & magnitude of a correlation (positive, negative, or neutral) and (2) how strong the relationship is - that is, whether it's "statistically significant." These two concepts are presented as the "coefficient" and the "standard error." Roughly speaking, the "coefficient" is a summary of the best fit of a relationship - sort of like a mean or average. The "standard error" is a sense of dispersion... or how many of our cases really come close to the best fitted relationship - sort of like a standard deviation. In the social sciences, we often look for relationships that are statistically significant at the 95% level... That means we're 95% certain that we would not have estimated a given coefficient if the real relationship is just zero... 95% significance is indicated when the size of the coefficient is about twice size of the standard error (more precisely, the ratio must be 1.96 or greater). Blah blah blah...
Whoah! Did I lose you in the previous paragraph? No problem. All you need to know is that the relationship between foreign aid fractionalization and bureaucratic quality might be either positive, neutral, or negative. The V-Graph summarizes all this in three partitions:
(Click the pic to make it bigger.)
If a country lands in the red part, the relationship is negative and statistically significant. (Higher bureaucratic quality => less fractionalization, like we might expect.) In the yellow part, the relationship is neutral or null - it's not considered statistically significant. (Bureaucratic quality doesn't influence these donors.) And in the green part, the relationship is positive and statistically significant. (Higher bureaucratic quality => more fractionalization - this is strange.) So, the graph conveniently summarizes the different ways donors consider the bureaucratic quality of recipient countries. You can easily see that for the USA the effect is negative - the United States "fractionalizes" its aid less for countries with high quality bureaucracies that it trusts to do a good job. The effect for IRE is neutral - Ireland doesn't seem to care about the bureaucratic effectiveness of recipient countries. And the effect for FRA is positive... strangely, France fractionalizes their aid packages more for the countries with high-quality bureaucracies. (If you're unfamiliar with the abbreviations for some countries, this page might help.)
From a substantive point of view, the question of fractionalization is important. Breaking up aid into small projects can overburden a developing country's bureaucracy and can be an inefficient way to provide aid. (See, for example, Nancy Birdsall's work on this.) So, we want to understand why donors choose to fractionalize. There is a lot more to it than I've touched on here - I recommend checking out the full paper.
And for statistically-minded folks who'd like to use V-Graphs to present their findings, the code is available here, on Jordan Kyle's webpage. So nice when a scholar provides a public good like this :-)