n.V1-5.07| TWO HUNDRED THOUSAND PIECES

01| In 2019 official finance from richer countries to poorer ones moved through more than 222,000 separate transactions. The average one was worth 1.4 million dollars. That was a twenty-four percent jump in a single year to an all-time high, and it sits inside a longer trend: the average size of an aid activity has fallen roughly thirty-five percent since 2000.

02| More money, cut into more and smaller pieces. On the receiving end it arrives as administrative load. Ethiopia has at times dealt with more than 250 separate donor agencies, each with its own reporting format, mission schedule, procurement rules and financial year.

03| Nobody chose this. There is no committee that decided aid should be delivered in two hundred thousand fragments, and no interest is served by it. It is the aggregate of a very large number of individually sensible decisions, which is why it is worth studying rather than denouncing.

04| The literature separates two things that get confused. Proliferation is the growth in the number of donors — more countries giving, more agencies within each, more foundations and international organisations entering. Some of that is straightforwardly good; newer donors from outside the traditional Western group have widened who is at the table. Fragmentation is what each donor then does with its money: spreading it across many countries, sectors and small activities rather than concentrating it.

05| William Easterly and Tobias Pfutze examined this across 48 agencies in the Journal of Economic Perspectives in 2008, comparing specialisation, selectivity, use of ineffective channels and overhead. Their conclusion on the first was blunt: the aid effort is remarkably fragmented, the worldwide budget split among many small bureaucracies — and even small agencies fragment their own effort across many countries and sectors, producing coordination problems and high overhead for donors and recipients alike.

06| That last clause rules out the easy explanation. If only large agencies were spread thin you could call it overreach. Small agencies with limited budgets scatter too, which means something other than capacity is driving it.

07| The mechanism is not mysterious and requires nobody to behave badly. A donor agency answers to a domestic constituency — a parliament, a minister, a public, a board. That constituency wants to see what its money did, which means the agency needs results it can attribute to itself. Attribution is easy in a small project the agency funded alone and hard in a pooled fund where its contribution is indistinguishable. Visibility therefore favours having your own thing in many places over a large share of one thing somewhere.

08| Add that each donor's constituency has different priorities, and that this year's cause is not last year's, and the aggregate becomes a very large number of small, separately branded, separately reported activities, distributed according to the preferences of giving publics rather than any assessment of where need or absorptive capacity lies.

09| Every step of that is rational for the actor taking it. The outcome is one nobody wanted.

10| The cost has been measured, once, carefully, by Stephen Knack and Aminur Rahman in the Journal of Development Economics in 2007, and their argument runs through the labour market, which is what makes it concrete.

11| Each donor wants its own projects to succeed, so it recruits capable local administrators into its own project implementation unit, paying more than the civil service can. A donor holding a large share of a country's aid has some reason to care about the state's overall administrative capacity, because its projects depend on the country working. A donor with a small share has almost no such reason: the cost of hiring away good staff falls mostly on other donors' projects and on the state.

12| So the more donors there are, each with a smaller share, the stronger the incentive on every one of them to strip capability out of the state, and the weaker the incentive on any of them to build it. Tested across recipient countries on an index of bureaucratic quality from 1982 to 2001, that is what they found: declining bureaucratic quality associated with higher donor fragmentation, and with a smaller proportion of aid moving through multilateral channels.

13| That is the sharpest available answer to what fragmentation costs. Not waste in the accounting sense — the systematic erosion of the recipient state's ability to do anything itself, produced by donors competing for the same scarce local talent.

14| One further figure makes the direction concrete. As of 2022, four out of five projects funded by official financial flows were run by non-governmental entities rather than through recipient governments, and more than seventy percent of the money continues to be delivered as project-type interventions. Most of it therefore never passes through a recipient country's budget, and never passes under the eye of its parliament or its planning process. Whatever anyone thinks of any particular government's capability, this means the aggregate of what is spent in a country is not visible to, or steerable by, that country's own institutions.

15| The defence is real and should be stated properly. A recipient dependent on one dominant donor is in a weaker bargaining position than one dealing with thirty, and is easier to coerce. Multiple donors mean multiple approaches tried in parallel, which is how anything new is ever discovered. Concentration would put enormous discretionary power in whoever ended up concentrating it, and nothing in this book part suggests large institutions choose well when nobody can check them. And part of the proliferation is simply more countries participating, which widens the field rather than degrading it.

16| The honest position is that fragmentation has a documented cost and an undocumented benefit, and nobody has established the trade-off. What is established is that the pattern is emergent rather than chosen — which is exactly why the successive agreements written to reverse it did not.