Humanitarian Funding and Needs, 2021 to 2025: Coverage, Colonial Ties and Asylum

According to ALNAP's Global Humanitarian Assistance Report 2026, total international humanitarian assistance fell to $33.3bn in 2025, down 20% in one year and 30% from the 2022 peak (PRISM analysis). My earlier research (EUI School of Transnational Governance, Policy Paper 2026/02) traces the fall to the dismantling of USAID and cuts by Germany, France, the Netherlands and the United Kingdom. The Global Humanitarian Overview 2026 requested $33bn for 135 million of the 239 million people in need, with a prioritised $23bn for 87 million people, according to Lilly and Pearson (HPG, October 2026). Principle 6 of Good Humanitarian Donorship (2003) commits signatory donors to allocate humanitarian funding in proportion to needs, and according to the same HPG paper the United Kingdom, Sweden and the European Union use allocation models.

PRISM modelled the link between the number of people in need (PiN) and funding for the 154 country-years from 2021 to 2025 in which an HRP, HNRP or flash appeal carried a PiN figure, in 38 countries. In PRISM's model of OCHA FTS and HPC data, PiN accounts for 55% of the variation in funding received (log scale) and 1% of the variation in the share of requirements funded. In PRISM's donor-level model for 2021–2025, among 25 EU and EFTA donors (governments and the European Commission), a former colonial power gave 64% more to its former colonies at equal budget and crisis-year, and each doubling of a country's first-time asylum applications in the donor country went with 10% more funding from that donor.

Within a country over time, PiN changed with the country's GRFC Phase 3+ population (p < 0.001), and according to the HPG paper, from 2026 PiN counts only people in areas at JIAF Phase 3 and above. According to the Global Protection Update of October 2026 and OCHA's 2026 plans, protection PiN ranged from 9% of people exposed to protection risks in Niger to 244% in Ukraine. The data, code and models behind PRISM's estimates are published on GitHub, except ACLED figures, which ACLED's terms do not allow PRISM to redistribute.

Needs and Funding Received

In PRISM's models for 2022–2025, funding received rose in proportion to PiN when PiN stood alone (elasticity 1.09), and funding per person in need was unrelated to PiN (R² 1%). At equal INFORM Severity, violence, displacement and food insecurity, the elasticity was 0.40 to 0.57: a plan with twice the people in need received 32% to 48% more. PiN and INFORM Severity together account for 0% of the variation in the share of requirements funded across 110 country-years in 34 countries; within a country over time, the share funded fell as PiN rose (p = 0.006). Lilly and Pearson (HPG, October 2026) report that, between 2023 and 2025, the change in a plan's requirements accounted for 9% of the change in its funding (R² 0.09).

Funding per Person in Need

According to OCHA FTS and HPC, the 2025 drought flash appeals for Zimbabwe, Zambia and Malawi received $3 to $4 per person in need, against the people-in-need figures of their 2024 appeals, which they kept. The occupied Palestinian territory received $820 and Ukraine $131; the median across 27 countries was $37. Countries on DG ECHO's forgotten-crises list received a median of $42 per person in need in 2021–2025, against $70 for other countries. Being on ECHO's list or receiving a CERF Underfunded Emergencies allocation had no statistically significant association with a country's share funded the following year (p = 0.16 and 0.41); both selections draw partly on funding levels.

Factors Associated with Coverage

In PRISM's models for 2022–2025, adding violence, displacement, food insecurity and population raises the variation in coverage accounted for to 12%; adding ReliefWeb report counts, years under an appeal, donor mix, colonial ties, EU asylum applications and flash-appeal status raises it to 55%, and year effects to 68%. With year effects, coverage was 33 points higher for flash appeals (p < 0.001), 1 point higher for each consecutive year under an appeal (p = 0.003), 12 points higher for each doubling of ReliefWeb reports on the country the year before (p < 0.001), 2 points higher for each 10-point rise in the US share of plan funding (p = 0.03) and 3 points higher for each doubling of political-violence fatalities recorded by ACLED (accessed 12 November 2026; p = 0.006). A country's plan funding in the previous year accounts for 81% of the variation in its funding, consistent with the path dependency Mena and Hilhorst (2022) documented in South Sudan.

Colonial Ties and Asylum

PRISM's donor-level model holds each donor's yearly budget and every crisis-year constant. Among 25 EU and EFTA donors, governments and the European Commission, a former colonial power gave 64% more to its former colonies in 2021–2025 (p = 0.002); leaving out each former colonial power in turn gives 47% to 123%. Across 42 donors, including the United States and Japan, the figure is 45% (p = 0.07), and CEPII's narrower measure of colonial relationships after 1945 has no association (p = 0.82). Fink and Redaelli (2011) found a bias towards former colonies in emergency aid for 270 natural disasters, and my earlier research (EUI, 2026) lists colonial legacies, geopolitics and path dependency alongside need as factors in donor allocation.

Among the same donors, each doubling of first-time asylum applications from a country's citizens in the donor country went with 10% more funding from that donor to that country (p < 0.001; 9% using the previous year's applications). The association fits both migration-control objectives and diaspora ties. In a model with donor-year effects only, covering 42 donors, INFORM Severity was associated with a larger share of a donor's funding (p < 0.001); PiN, at equal severity, had no association (p = 0.64).

Measurement of Need

According to the HPG paper, the PiN in the Global Humanitarian Overview fell 34% between December 2022 and December 2025 while political violence incidents rose 45% and forced displacement 10%; from 2026, PiN counts only people in areas at JIAF Phase 3 and above, and plans covering only part of a country rose from 4 in 2024 to 12 in 2026. The same paper records “the lack of transparency over how the PIN number is calculated”. In PRISM's panel, PiN across countries rose with population and INFORM Severity, rose with food insecurity at the margin (p = 0.05) and fell with new conflict displacement (p = 0.006), with an R² of 81%; within a country over time it changed with GRFC Phase 3+ populations (p < 0.001) and with none of severity, violence or displacement. PiN in Malawi (2022), Mozambique (2021), Lebanon (2024) and Nigeria (2022–2025) was 52% to 74% below the value predicted from those measures.

According to the Global Protection Update of October 2026, 418 million people were exposed to protection risks, 139 million were in need of protection and 51 million were targeted. The GPC guidance note on estimating people exposed for the 2027 HPC proposes machine-learning methods such as XGBoost, with variables chosen by context and adjustments after prediction, and gives no model specification or code. The GPC data-poor tool prints its formula and weights: protection PiN is 10% to 90% of people exposed according to an expert-scored severity, with weights the note calls “illustrative defaults”. Neither note includes country inputs.