Corruption in aid has been a major topic of discussion in Washington in recent weeks. One of the questions reportedly from the Presidential transition team to the State Department was: “With so much corruption in Africa, how much of our funding is stolen?” During the nomination hearings for Rex Tillerson to be Secretary of State, Senator Rand Paul provided one answer: seventy percent of aid is “stolen off the top.” 

The question is a fair one to ask. The bad news is that the short answer is “we don’t know.” The better news is that the slightly longer answer is “nowhere near 70 percent.” And the best news is that if we spent more time tracking the results of aid projects, we’d have a much better idea of where corruption was a problem and if our efforts to reduce it were working.

Statistics about corruption are hard to verify and open to considerable dispute. I've written before about a recent Supreme Court case where the justices strongly disagreed about what counts as 'corrupt,' for example. But also, for obvious reasons, people don’t tend to advertise the fact that they are involved in corruption. That all makes measurement hard. Nonetheless, three indicators of corruption might help answer how widespread the problem really is:

  1. investigative cases of particular aid projects;
  2. survey evidence about bribe payments; and
  3. 'expert perceptions' about the general state of corruption in a country.

What does each indicator suggest about the extent of corruption?

1. Investigative cases

The World Bank’s Sanctions Evaluation and Suspension Office keeps track of cases where World Bank investigations have uncovered evidence of fraud and corruption. An analysis of cases between 2007 and 2012 found sanctionable fraud or corruption in 157 contracts worth $245 million, of which less than a third of contracts showed evidence of sanctionable corruption. The World Bank’s lending volume is about $40 billion a year, so this suggests less than a third of contracts collectively worth about 0.1 percent of volumes over the period involved discovered and sanctionable corruption.  

2. Survey evidence

Of course, that investigative cases only capture ‘discovered’ corruption is a huge issue. It is likely that the great majority of corruption isn’t uncovered by investigators. So what about asking firms that work on aid contracts?

Sadly, we don’t have survey evidence of corruption specifically involving businesses involved in aid contracting, but we do have surveys of corrupt payments by firms to government officials for government contracts in general from the World Bank. Enterprise surveys ask firm managers how much are the “gifts” expected in return for winning a government contract in their sector. In 22 countries, the average is more than 5 percent of the value of the contract. In 38 countries, the average is between 2 and 5 percent; in 52 countries the average is below 2 percent. You’d hope aid would be less affected by corruption than local spending given all of the oversight apparatus from investigators general to procurement experts involved, but in truth there is limited evidence they have a big impact, so maybe the numbers are about the same for aid-financed contracts. Given the $161 billion global aid business, an average of (say) five percent being lost to corruption adds up to around $8 billion—a real loss. But not the $113 billion that would be suggested by Senator Paul’s figure.

3. Perceptions measures

The trouble with measures of bribe payments, however, is that they only capture one form of corruption, and one of its impacts. They don’t capture officials simply stealing funds on their own account. And if the bribe goes to cover up substandard work which means the aid-financed road falls apart or the aid-financed drugs don’t work, the impact of corruption is far bigger than the bribe payment made. Again, we don’t have a measure of this broader corruption specific to aid flows, but you can ask a bunch of people how corrupt they think a country is in general as one potential indicator. Using one such ‘expert perceptions’ measure, Bill Easterly at one point calculated that 76 percent of US foreign aid went to countries judged to be corrupt.

Sadly, as Easterly himself noted, "It is inherently impossible to calculate the number 'x percent of aid is stolen' from the numbers that I use on how much aid goes to corrupt countries." And that’s not to mention the considerable problems with regard to the accuracy—and even the meaning—of a single measure of how a small group of people feel about overall corruption levels in a country.

So, using available corruption measures we’re left with three answers to “how much of our funding is lost to corruption”: one almost-certain underestimate of “less than 0.1 percent,” one possible underestimate of one part of the problem of “a few percentage points,” and one arguable interpretation that doesn’t really help that “a lot of aid goes to corrupt countries.” 

Measuring Lost Aid through Outcomes

Luckily there's another, better way to estimate how much aid is 'lost': look at outcomes. If the aid program manages to buy all of the things it is meant to buy and deliver them where they are meant to go at a reasonable price, the aid funds can’t have been lost to corruption. Of course, measuring missed outcomes would be an over-estimate of aid lost to corruption specifically—a lot of other things can mean aid fails to deliver from bad luck to incompetence.

But at least it might provide an upper-end figure.

One partial measure of outcomes is project ratings. Take the World Bank, where project outcomes are rated on a six-point scale from highly unsatisfactory to highly satisfactory. Assume all projects rated in the bottom three categories are ‘lost’ and that adds up to 461 out of 1,617 projects completed 2010-2015, or about 29 percent. Be a little more generous and assume only the bottom two categories are ‘lost’ and that drops to around nine percent.

Again, Sarah Dykstra, Justin Sandefur, Amanda Glassman, and I looked at evidence for waste in GAVI—an aid institution dedicated to expanding vaccination coverage in developing countries. While we found plenty of evidence that developing countries were spending less on vaccines themselves when GAVI gave them vaccines at no or low cost, we estimated actual waste (vaccines not leading to vaccination) at between 0.5 percent and 15 percent of vaccines delivered, depending on the vaccine.

Or following on through the causal chain from aid financing to outcomes, countries that were the focus of attention from the President’s Emergency Plan for AIDS Relief (PEPFAR) saw the annual change in the number of HIV-related deaths between 2004 and 2007 that was 10.5 percent lower than other African countries. If the money had been lost to corruption, we simply wouldn’t have seen these results.

All of the evidence we have—on corruption cases, bribes, and outcomes alike—suggests corruption is a problem for aid, but nowhere near as big a problem as suggested by Senator Paul. That said, we still don’t have enough evidence on results to come up with any conclusive overall numbers on ‘the percentage of aid that delivers the impact it was designed to.’ And we should. If the administration’s concerns over aid and corruption ended up improving the results focus of aid programs, that would be great news for many reasons—but not least because it would help reduce the real impact that corruption can have on development.