Protection Funding 2021–2026: The Cuts Reached the Evidence

Protection is the part of humanitarian response that is hardest to defend in a shrinking budget and hardest to rebuild once cut. It does not deliver a tonnage figure. Its outputs are presence, continuity, case files and analysis, and those are the first line items to go when donors narrow their funding around visibly life-saving delivery.

The six-year record is unambiguous. Across the 19 crisis contexts that published a protection cluster requirement between 2021 and 2026, humanitarian appeals asked for $15.71 billion and received $7.22 billion, or 46.0% of stated need. The shortfall is not spread evenly. Ten of the 19 contexts sat below 40% coverage across the entire period, and the worst funded, Somalia at 22.5%, has been running a protection response on under a quarter of what it said it required, cumulatively, across six years.

What has changed since 2025 is not the underfunding but its composition. In June 2026 the Global Protection Cluster surveyed 747 protection partners across 22 operations and found that the functions cut hardest are the ones that generate the evidence: protection monitoring, protection analysis, early warning, information management. This deep dive works through three layers in turn, the money, the capacity it buys, and the capacity to know, and argues that the third is where the damage compounds.

Six years, 19 contexts, and a requirement that was never met

The ranking below applies the Humanitarian Funding Forecast's Index of Underfunded Crises methodology to protection cluster funding alone. That method ranks contexts by how much of their published requirement was actually funded, cumulatively, rather than year by year. The method is theirs; the protection-only application, the six-year window and the figures are PRISM's. Ranking this way strips out the single-year noise that makes appeal coverage look volatile and exposes something steadier: which responses are structurally underwritten and which are structurally not.

The bottom of the table is a list of protracted conflicts in the Sahel and the Horn of Africa, plus Myanmar and Haiti. Somalia (22.5%), Mali (23.3%), Niger (29.3%) and Myanmar (29.7%) all received under 30 cents on the dollar. Syria is the sharpest case of scale meeting neglect: the second-largest protection ask in the dataset at $2.1bn, funded at 34.8%.

Read against the top of the table, the pattern is about attention rather than assessed need. Ukraine carries the largest protection requirement in the dataset, $2.24bn, and is funded at 70.6%. Syria carries the second largest and is funded at roughly half that rate. South Sudan (75.2%) and Afghanistan (74.9%) are the best covered, and no context in the index reached 80%, the threshold at which the index would call underfunding minimal.

The annual series shows why 2026 matters. Coverage climbed steadily from 31.9% in 2021 to a peak of 54.2% in 2024, then turned: 51.9% in 2025 and 46.0% so far in 2026. Requirements have been falling since 2023 as appeals are cut back and prioritised, so the recent coverage decline is happening against a smaller ask, not a larger one.

What the contraction does to capacity

In June 2026 the Global Protection Cluster published Protection Capacities Under Pressure, a consultation with 747 respondents across 22 operations. It is the first systematic account of what the 2025 funding cuts did inside protection programmes rather than to their budget lines, and it should be read alongside the funding record above.

The macro backdrop it sets out is severe. Official development assistance fell 23.1% and humanitarian ODA fell 35.8%, to US$15.5bn. The global appeal was rescaled from US$33bn for 135 million people to US$23bn for 87 million, a formal decision to plan for fewer people rather than a shortfall discovered after the fact. Protection contracted more sharply than comparable sectors, including health, which fell 19%.

Two features of the 2025 fall deserve attention. Protection recorded the third-largest drop in funding received between 2024 and 2025, US$516 million, against only a modest reduction in requirements, which is why the coverage decline was the sharpest of any sector at 39.2%. And the contraction was concentrated: the five largest DAC providers accounted for 95.7% of the total ODA decline in 2025. Donor concentration in protection itself has loosened but from an extreme base, with four funders providing more than two thirds of protection assistance in 2023 and the top five providing 38% in 2025.

Early 2026 offers no recovery signal. The GPC records US$610 million received as of April 2026 against a coverage rate of 21%.

Inside programmes, 69.1% of partners report a direct impact on their capacity. When the survey isolates only the partners who actually implement each activity, removing the distortion caused by counting non-implementers as unaffected, the impact levels are consistently very high and remarkably flat across the response chain.

What the documented record shows

PRISM's protection archive, which preserves ACAPS Protection Risks Monitor events and adds PRISM-documented events curated against the same taxonomy, holds 34,651 documented events across 107 countries, classified against the 15 protection risks of the Global Protection Cluster's analytical framework. The shape of that record is concentrated. Attacks on civilians and other unlawful killings account for 31.2% of all documented events, nearly a third of the record on their own.

Severity, scored by protection partners across the same 15 risks on a one-to-five scale, ranks the contexts differently from the funding table. Sudan sits at 4.62, the highest mean severity of any monitored context, and Palestine at 4.38. Both are funded better than Somalia, which combines the fifth-highest severity score with the worst coverage rate in the index.

The caseload figures need care, because two credible numbers circulate and they measure different things. The GPC's published 2026 global figures are 351 million people exposed, 139 million in need and 51 million targeted. PRISM's own sum from OCHA plan data is considerably narrower.

The gap between the two is not a contradiction and should not be read as one. It measures the distance between the protection problem as assessed and the protection problem as planned and costed. Roughly four in 10 of the people the GPC identifies as needing protection sit outside a plan that carries a protection requirement, which means they sit outside the funding record analysed above.

The functions that were cut hardest are the ones that see

Everything above rests on a prior condition: that protection risks are detected, documented and analysed in the first place. This is the layer the GPC calls system-enabling, and it is the layer the survey finds most consistently eroded. Around 7 in 10 partners report direct impact on protection analysis and early warning. Among implementing partners, protection monitoring stands at 83% and protection analysis at 71%.

The workforce evidence points the same way. Partners report loss or reduction of knowledge, technical expertise or capacity (44.1%), of technical advisory or supervisory staff (42.7%) and of coordination and information-management staff (40.6%). Only around two in 10 partners retained capacity strengthening or targeted advocacy, and only three in 10 retained locally-led early warning mechanisms. Staff fully dedicated to child protection functions fell from 72% to 52%, and human rights monitoring work was reduced by over 50% with operational presence significantly scaled back.

The GPC states the consequence plainly. Partners describe losing information because short project timelines and reduced analytical capacity mean, in the report's words, that much of what is happening is no longer captured or analysed. That is a different kind of loss from an unfunded activity. An unfunded activity is visible in a coverage table. Uncollected information is not visible anywhere, including in the tables used to argue for the next appeal.

A live demonstration: the ACAPS Information Landscape

The clearest available illustration is not a projection but a dataset that has just stopped. The ACAPS Information Landscape tracked information freedom, access to information, information trends and the vulnerability of specific groups accessing information. It ran from 1 January 2024 and recorded 3,368 entries across 102 countries. Collection was discontinued on 31 July 2026 for funding reasons.

Its own collection rate is the demonstration. Entries fell from 178.9 per month in 2024 to 79.6 in 2025 and 38.0 in 2026, a 79% decline before the dataset closed entirely. Every one of the 19 contexts in the funding index above was covered by it, from Ukraine at 248 entries and Palestine at 234 down to CAR at seven. The monitoring that closed was monitoring of exactly the contexts where protection funding is thinnest.

This is the practical form the GPC's warning takes. Four in 10 partners lost information-management staff, only three in 10 retained locally-led early warning, one long-running monitoring dataset closed, and the record of what is happening to civilians thins accordingly. None of that registers as a fall in protection needs. It registers as a fall in the ability to state them.

What follows for the record

The three layers resolve into one argument. A protection response funded at 46.0% of its own stated requirement for six years was already operating on assumptions rather than coverage. The 2025 contraction did not simply take more away; it took disproportionately from the functions that establish what is happening, which is why the next six years of protection data will understate the problem by an unknown margin.

The response to that is unglamorous and largely institutional: treat monitoring, analysis and information management as protected core rather than overhead, and preserve the series that already exist. PRISM's own contribution is the second half of that. The protection archive behind this analysis runs from 2014 to the present precisely because upstream sources have rolling windows and finite funding, and archived series remain readable after the collection that produced them has stopped.

Readers who want the narrative version of this evidence can work through Protection in Focus, PRISM's scrollytelling read on the same archive, which walks the 15 risks and the contexts behind them. Readers who want the current numbers rather than the August 2026 snapshot used here should go to the protection severity dashboard, where severity scores, funding coverage and the documented event record update from source.

"In some contexts, this means that people may still be counted as 'reached', while the quality, continuity and safety of the support available to them has significantly changed." — Global Protection Cluster, Protection Capacities Under Pressure, June 2026