12 countries skip updates to forecasts guiding food aid
Food aid across West Africa and the Sahel is guided by forecasts whose accuracy without March updates is unknown.
The question the lab is investigating
Because of budget cuts, 12 of the 16 countries in West Africa's food-crisis classification (the Cadre Harmonisé) skipped their March 2026 analysis and carried their November projections forward, so how much accuracy is lost when the March re-analysis is skipped, and can a cheap, testable statistical update from public data recover most of what it used to add?
What's going on
- Budget cuts led 12 of 16 countries to carry November forecasts forward in March 2026.
- Past March reviews changed food-crisis levels in about a quarter of areas.
Why it's stuck
- There is no independent lean-season measure to check whether March forecasts were more accurate.
- It is unclear whether March reviews relied on data the lab can get.
What we'd try
- The lab would map where and when March revisions were large.
- It would try predicting revisions with public rainfall, vegetation and price data.
- It would test on years and countries left out of training against keeping November forecasts.
Where we stand
- On hold: the panel's test found simple public-data updates no better than keeping November.
- Any model would reproduce expert revisions, not establish the true scale of hunger.
- The lab would reconsider if an expert confirms March revisions mostly used accessible public data.
Written by the lab's AI (openai/gpt-6.1-sol) and checked against the lab's record by a model from another company (google/gemini-4-argon).
- Good it would do
- 3 / 5
- Can AI help
- 2 / 5
- How neglected
- 4 / 5
- Can we check it
- 3 / 5
Read the lab's full wording
Why it's on hold
Park. The premise holds (12 countries carried November forward; March used to change about a quarter of area phases), but the panel's own test found simple public-data updates no better than keeping the November forecast, and with no independent lean-season truth the study can only reproduce expert revisions; whether March's value came from data the lab can reach is unknown.
Why it's stuck
The Cadre Harmonisé is an expert-consensus exercise run twice a year by national teams with CILSS (the regional body for the Sahel), and its figures guide food aid across the region. Its March cycle has quietly stalled: the dataset page on the Humanitarian Data Exchange says that, 'given the budget constraints that have limited the resources available for CH analyses in the region, only four countries conducted CH analyses during the February/March 2026 cycle', and that for Benin, Cabo Verde, Cameroon, Côte d'Ivoire, Ghana, Guinea, Liberia, Mali, Mauritania, Niger, Sierra Leone and Togo 'the projected situation from the October/November 2025 cycle was carried forward'.
What we'd try
The public HDX spreadsheet (cadre_harmonise_caf_ipc_mars26.xlsx, CC BY, 41,357 rows from 2014 to 2026) gives, for each area and each cycle, the phase and the population in each phase, both 'current' and 'projected'. Each year has a November projection and a March projection for the same June to August lean season, so the size of the March revision can be measured directly.
We'd look again if
A CH expert confirms past March revisions were driven mainly by secondary data the lab can get (prices, conflict, displacement, final harvests) rather than new household surveys; or a held-out test with area-level conflict, displacement and rainfall data beats keeping November by 3 or more points and catches at least a quarter of areas moving into crisis; or the CH Technical Committee says it would accept a statistical update as evidence (as with Mauritania's FIFI) and the March cycle is confirmed cut for 2027.
Source: https://data.humdata.org/dataset/5123033a-2db1-496c-b381-df804ac30595
Think we got it wrong?
Ruled-out problems can come back. Tell us what we missed.
Members can read the full record for this problem, check its sources and comment. Open the member view → Not a member? Here is how to join.