n.V1-9.04 | Theft and the Cost of Delivering

01| A dollar committed by a donor passes through several organisations before anything reaches a household. A bilateral agency funds a multilateral or a contractor, which funds an international NGO, which funds a national organisation, which runs the work. Each layer costs something. The obvious question is how much of the original dollar survives.

02| Figures circulate. Ten percent reaches the poor. Only a third arrives. Somewhere between a third and two-thirds, depending on the intervention. These get repeated widely and sourced rarely, which is the first warning.

03| The honest answer is that a general rate cannot be established, that this is itself the finding, and that the reason is a distinction the whole debate skips.

The hardest evidence anyone has produced

04| In the early 1990s the Ugandan government ran a capitation grant scheme: money sent from the centre to districts for primary schools' non-wage costs. Ritva Reinikka and Jakob Svensson built a survey that tracked what schools actually received against what had been sent.

05| Between 1991 and 1995, schools received on average thirteen percent of the grants intended for them. Most schools received nothing at all. The rest was captured en route by district officials and politicians.

06| The leakage was not evenly distributed either. Schools in better-off communities, with more bargaining power, claimed a larger share of their entitlement — so the money that did arrive went disproportionately to the least poor of the intended recipients. That is worse than most of the unsourced claims.

07| Two features of the study get left out when it is cited, and both matter more than the headline.

08| The first is what happened next. From late 1997 the government began publishing monthly transfers to districts in national newspapers, so a headteacher could read what had been sent for their own school. The follow-up work found this drastically reduced the capture. The leakage was not a fact of nature. It was a consequence of nobody downstream being able to see the number, and it largely ended when they could.

09| The second is what the study measured: a government-to-school transfer inside one country. Not a donor-to-NGO-to-community chain. It is powerful evidence that money in transit can be captured and that publication stops it. It is not a measurement of the international aid chain and should not be quoted as one.

The famous number is a misreading

10| The most repeated claim in this area is that around ninety percent of aid is skimmed before it reaches communities. The Center for Global Development has published a direct rebuttal, and it convinces.

11| The statistic being misread concerns direct payments to organisations based in developing countries. It is true that only a small share of aid takes that form. But the remaining money is not therefore lost. Much of it is spent on goods and services delivered in kind — antiretroviral drugs, vaccines, emergency food, equipment — which never appears as a payment to a local organisation because it is a shipment rather than a transfer. A dollar spent on medicines that end up in a clinic has reached the beneficiary about as directly as money can, and it counts as zero in the statistic used to prove it did not.

12| This is the central confusion in the genre, and once seen it is hard to unsee. Money not paid to local organisations and money that evaporated are different quantities. The evaporation argument depends on treating them as the same.

The distinction that would settle it

13| Two entirely different things happen to a donor dollar in transit, and they carry opposite moral weight.

14| One is leakage: money going somewhere it was not supposed to go. Capture, fraud, funds that disappear. Uganda measured this. It is theft.

15| The other is intermediation: money spent on the work of delivering. Salaries of the people running the programme, logistics, offices, audit, and the cost of complying with the donor's own reporting requirements. This money is consumed rather than transferred, but it is not lost. It is what makes the thing happen. A high intermediation rate may be evidence of inefficiency, or of several layers duplicating oversight, or simply of work that is hard.

16| The circulating figures do not distinguish between these, and in most cases do not appear to come from anywhere. Estimates of what each layer takes are widely quoted at fifteen to twenty-five percent for contractors, international NGOs and UN agencies alike, but tracing them back leads to secondary commentary rather than to any published accounting. That does not make them wrong. It makes them not evidence.

17| And the reason no better figure exists is plain. No organisation in the chain publishes its own intermediation rate, because none is required to and none benefits from it. The number that would answer the question is the one number nobody is asked for.

The same argument, running the other way

18| Now set beside all this the criticism most donors actually act on: that spending on the organisation itself is waste, and that low overhead is a sign of discipline. The reform literature inside the sector argues close to the opposite, and this is the sharpest reversal in the whole debate.

19| In 2009 Ann Goggins Gregory and Don Howard published "The Nonprofit Starvation Cycle," drawing on around 220,000 tax filings and roughly 1,500 surveys of organisations. Their argument is that the cycle begins with funders holding unrealistic beliefs about what it costs to run an organisation. Nonprofits respond by understating their costs and cutting what funders dislike paying for — financial systems, information systems, staff training — which produces exactly the weak organisations that then fail to deliver, which then confirms the funders' suspicion. Low overhead, in this reading, is not discipline. It is starvation.

20| Four years later the three main American charity rating bodies — GuideStar, the BBB Wise Giving Alliance and Charity Navigator — published a joint open letter under the title "The Overhead Myth," publicly disowning the overhead ratio as a measure of whether a charity is any good. The organisations that had popularised the measure were asking donors to stop using it.

21| So the sector contains two arguments that both say the money vanishes, and they point in opposite directions. One says the layers take too much. The other says the layers are starved and that starving them is why the work fails. Both cannot be the main story, and neither can be settled with the numbers currently published.

What the localisation data shows, and what it does not

22| There is one place where the sector has counted something close to this. At the 2016 World Humanitarian Summit, donors and agencies committed under the Grand Bargain to channel at least twenty-five percent of humanitarian funding to local and national actors as directly as possible. The commitment was made because everyone involved agreed too much money was passing through too many international hands.

23| Progress has been poor, and how poor depends on definitions. Assessments of funding reaching genuinely home-grown local organisations have put the figure below one percent, with one estimate under 0.2 percent. What counts as a local actor is disputed — a national office of an international agency may or may not qualify — and that dispute determines whether the target is nearly met or barely started. By 2021 the twenty-five percent figure had been softened from an absolute target to a commitment to increase.

24| What this measures is who receives money, not what fraction survives. A local organisation can be as expensive an intermediary as an international one. But it establishes one thing firmly: the sector's own flagship attempt to track where money goes has been undermined, in its own words, by a lack of reliable data on funding flows at country level. The people who set the target cannot reliably measure the target.

25| So: leakage is real, has been measured at severe levels in at least one rigorous study, and was largely fixed in that case by printing the transfers where the recipients could read them. The general evaporation rate cannot be stated. The numbers in circulation conflate theft with the cost of delivery, and the most repeated of them rests on a misreading. Which leaves a narrower formulation, and a harder one to dismiss: this is a sector that can be shown, when anyone bothers to look, to lose large sums in transit — and that has never built the accounting that would let anyone look without mounting a research project.