The Fragmentation Paradox

448 reporting organizations in Kenya. Four of them account for half the money. And the first version of this analysis got the headline wrong.

Correction. The version published on 7 September reported that two organizations — the U.S. Department of State and USAID — accounted for 60–67% of IATI-reported spend in every country. That was an artifact of the method: each activity’s full lifetime spend was attributed to every country it was tagged with. One activity, the U.S. Foreign Military Financing program ($90.2 billion, tagged to 131 countries, 53% Israel), was counted in full as aid to Kenya, Uganda, Tanzania and Ethiopia. The method note claimed the inflation “affects all organizations proportionally”. It does not. This version weights every activity by its declared recipient-country percentage. All figures have been recomputed. What changed ↓

The development policy conversation worries about aid fragmentation — too many donors, too much coordination overhead. Weigh the donors by spending instead of counting them and the picture changes: a handful of funders, led by the multilateral banks, control most of the money. It is a concentration story. It is just not the concentration story first reported here.

4
organizations account for 50% of weighted spend in every country
0.93–0.95
Gini coefficient of donor spend (unweighted version: 0.96–0.97)
69–79%
of naively attributed spend belongs to other countries

The paradox: counting vs. weighing

Count organizations reporting IATI activities: you see fragmentation — 253 to 448 per country. Weight by spending, with each activity credited to a country only in the proportion its publisher declares: you see concentration — four organizations reach half the money, 12 to 15 reach 80%.

Organizations vs. concentration
Total reporting organizations (blue) vs. number that account for 50% of country-weighted spend (orange)
Reporting organizations
Orgs for 50% of spend
Kenya
448
4
Uganda
427
4
Ethiopia
358
4
Tanzania
290
4
Rwanda
253
4
CountryOrgs50% in80% inTop 2 shareGiniOrgs <0.1%
Kenya44841541.6%0.95088%
Uganda42741435.5%0.94989%
Ethiopia35841538.0%0.94587%
Tanzania29041233.5%0.94287%
Rwanda25341232.3%0.93281%

What the weighting changes

IATI activities can be tagged to many recipient countries, each with a percentage. A naive country filter returns every tagged activity at full value. Weighting by the declared percentage removes 69–79% of the apparent spend in these five countries — $1,156 billion becomes $274 billion — and it removes it unevenly. Publishers of large global programs lose most; publishers of country projects lose nothing.

Top-2 share of spend, before and after weighting
Share of a country’s IATI spend held by its two largest reporting organizations. Grey: unweighted (as first published). Blue: weighted by recipient-country percentage.
Unweighted (7 Sept version)
Weighted (corrected)
Kenya
66.1%
41.6%
Uganda
64.7%
35.5%
Ethiopia
59.4%
38.0%
Tanzania
66.5%
33.5%
Rwanda
63.6%
32.3%
CountryTop 2, unweightedTop 2, weighted#1 unweighted#1 weightedSpend belonging elsewhere
Kenya66.1%41.6%Dept. of State (44%)World Bank (26%)79%
Uganda64.7%35.5%Dept. of State (43%)AidData* (23%)79%
Ethiopia59.4%38.0%Dept. of State (36%)World Bank (23%)69%
Tanzania66.5%33.5%Dept. of State (43%)World Bank (22%)78%
Rwanda63.6%32.3%USAID (48%)World Bank (22%)79%

The U.S. Department of State falls from first place in four countries to outside the top six in all five. The World Bank becomes the largest reported funder in four of five. USAID stays second or third, on a much smaller base. The corrected landscape is multilateral-led with a strong U.S. bilateral presence, not U.S.-dominated.

The same donors everywhere

Eight organizations appear in every country’s weighted top 15. The aid landscape looks diverse when you count organizations, but the dominant funders are the same across the region.

Top donors across all five countries, weighted
Combined country-weighted IATI spend (lifetime, $ billions), colored by number of countries in whose top 15 the organization appears
All 5 countries
4 countries
3 countries
World Bank
$57B
USAID
$38B
AidData*
$28B
African Dev. Bank
$19B
UK (FCDO)
$15B
Global Fund
$14B
Sweden (Sida)
$8B
UNICEF
$7B
HHS (CDC, NIH)
$7B
Germany (BMZ)
$6B
UNHCR
$5B
Gavi
$4B
Denmark (MFA)
$4B
OrganizationWeighted spendCountriesNote
World Bank$56.8B5/5Largest in KE, ET, TZ, RW
USAID$37.8B5/5
AidData$27.8B5/5Research org; reports third-party data. Largest in UG
African Dev. Bank$18.8B5/5
UK (FCDO)$14.7B5/5
Global Fund$13.7B5/5
Sweden (Sida)$7.5B5/5
UNICEF$6.8B5/5
HHS (CDC, NIH)$6.7B4/5Not in Ethiopia top 15
Germany (BMZ)$5.8B5/5
UNHCR$4.6B4/5
Gavi$3.8B3/5
Denmark (MFA)$3.7B3/5
U.S. Dept. of State0/5 in top 6Was #1 in four countries before weighting

AidData: the ghost in the data

AidData, a research lab at William & Mary, appears among the top five funders in every country and is the largest single “donor” in Uganda. It is not a donor — it publishes historical aid data into IATI from funders (especially China) that don’t publish themselves. Its entries represent real flows, reported by a third party, and may overlap with activities reported by the original funder.

Weighting does not remove this artifact, because AidData’s activities are mostly single-country. It is a different measurement-layer problem: the identity of the reporter is being read as the identity of the funder.

Two problems, not one

Fragmentation counts organizations. More organizations mean more coordination overhead for the recipient government, regardless of how much money each brings. This is real: Kenya’s government must engage with 448 organizations reporting activities.

Concentration measures the distribution of resources. A few large funders dominate the financial landscape, creating dependency risk. This is also real: four organizations control half the money.

The paradox: the same data produces opposite diagnoses depending on whether you count heads or dollars. The corrected numbers move the concentration story from “two funders” to “a handful”, and change who they are — dependency on the World Bank, the African Development Bank and the Global Fund is a different risk from dependency on one bilateral donor’s diplomatic budget.

How easy this was to get wrong

The uncorrected version passed a full analysis, a written brief, three charts and publication. The artifact was found only when the same tool was pointed at Pacific island states, where a $90 billion “donor” to Tonga (national GDP under $1 billion) was impossible to miss. In East Africa, where U.S. spending genuinely is large, the same artifact produced numbers that looked plausible.

Nothing in the output flags the omission. A country filter that returns “$262 billion of aid to Kenya” is not wrong in any way the database can detect. Only a reader with an external sense of scale can catch it — and for East Africa, that sense of scale was not enough. This is the measurement layer operating on the analyst. See the companion analysis for the Pacific: Who Funds the Pacific?

Method

Data from d-portal.org, activity and country tables joined on activity identifier. ~97,000 IATI activities with positive spend across five East African countries (Kenya, Uganda, Rwanda, Tanzania, Ethiopia) from ~1,700 reporting organizations. Weighted spend = activity lifetime spend × declared country_percent / 100; single-country activities carry 100%. 19–29% of activities per country are multi-country; they account for 69–79% of unweighted spend. Spend values are lifetime totals, not annual. Organizational identity uses reporting_ref — PEPFAR spending is attributed to the Department of State (the IATI publisher). Unweighted figures are retained for comparison. Tool: donor_concentration_weighted.py.