Why this session matters to us
A press conference about a report nobody can read yet is a strange object, and it is worth naming that before anything else. FAO’s chief economist spent twenty minutes with UN correspondents describing findings from a flagship that does not exist in public until 21 July. Every figure below was spoken aloud and cannot, today, be checked against the document it comes from. That is ordinary UN practice, and it is also a real limit on what a reader should do with any of it.
What makes the session worth its own page is a single admission buried in an answer about the Middle East. SOFI 2026’s hunger and cost numbers are 2025 data. The supply shock now working through the world’s fertilizer and energy markets began after the data closed. The report launching next week describes a world that has already moved, and its author said so, unprompted, to a room of reporters.
The companion HLPF special event on the cost of a healthy diet the day before covered what the report’s central metric is and how it is built. This page covers what the metric cannot yet see.
“It’s before the Strait of Hormuz”
Asked by a correspondent how the escalation around the Strait of Hormuz would hit food affordability, Torero Cullen drew the boundary of his own report first.
That sentence is the whole measurement problem in one line. A statistical flagship is a rear-view instrument by construction: it needs a closed year of country data, which is exactly why it cannot describe the thing currently happening. The projection to 2030 is the report’s attempt to reach forward, and a projection is a model, not a count.
His account of the shock itself is an argument about inputs rather than food. The closure removes, on his reckoning, somewhere between a quarter and a third of the key inputs into agrifood systems at once: natural gas, the sulphur that becomes the sulphuric acid that makes phosphate fertilizers, nitrogen fertilizers, and the energy used across the chain. Input prices rise. And then the farmer, not the shopper, makes the first decision.
The bind he described is specific. Cereals are nitrogen-intensive, and cereal prices are currently low because recent harvests were good. A farmer facing dear inputs and cheap output has margins near zero, and three bad options: absorb the loss, switch crops, or buy fewer inputs and accept lower yields. Torero Cullen pointed at a market signal suggesting many are choosing the third.
The commodities he named as already in trouble were rice and durum wheat. The risk he flagged as largest was the sulphur shortage feeding through to phosphates.
Three to six months, which is why nobody has noticed
The most useful thing Torero Cullen said was about lag, and it explains why a food shock is politically invisible at exactly the moment it becomes unstoppable. Wheat is a commodity; bread is a thing you buy. The distance between them is months.
A petrol price is a live readout of a supply shock. A food price is a delayed one, and the delay runs longer than most news cycles and some election calendars. The consequence he drew is that the window for acting on an input shock closes while the public evidence for it is still absent, which is a measurement problem wearing a political costume.
Pressed by another correspondent for the counterfactual, his answer was unusually blunt about what cannot be fixed. This is a supply shock, not a price shock, and unlike the war in Ukraine there are no high commodity prices to cushion farmers.
He cited an FAO simulation paper putting the hit at a fall in global welfare of about 1.7%, and said the growth effect it estimates is close to what the IMF was already projecting. The growth figure itself is not reproduced here; the recording does not render it clearly enough to print.
What a cost metric sees that a hunger metric cannot
SOFI has always been scored on two yardsticks: the Prevalence of Undernourishment (SDG indicator 2.1.1) and the Food Insecurity Experience Scale (2.1.2). Both count people. The cost and affordability of a healthy diet, added in 2020, prices a basket instead, and it is the newer lens that carries this year’s argument.
The reason is that a cost metric can be taken apart, and a headcount cannot. Torero Cullen’s method throughout was decomposition: cost against affordability against consumption, then calories against nutrients. As presented at this briefing, staple foods supply about half the calories a person needs but only around 13% of the cost of a healthy diet, while animal-source foods account for almost 30% of it. Both figures sit close to the ones FAO gave at the previous day’s HLPF event, which put staples near 16% and animal-source foods above a quarter.
The third share, for fruit and vegetables, we are not printing. FAO gave materially different splits on consecutive days: this briefing put fruit and vegetables at about 5% of calories and 16% of the cost, while the previous day’s event put them at roughly 12% of calories and some 40% of the cost, the single largest block in the basket. Both cannot be right, and the report that would settle it does not publish until 21 July, so this page prints neither share. The qualitative claim survives either version and is the one that carries the argument: the cheapest calories are the least nourishing, and the nourishing part of the basket is what makes it unaffordable.
Split that way, the number stops describing a problem and starts naming a lever. Because roughly 70 to 75% of the cost of a healthy diet is generated after food leaves the farm, in storage, transport, processing and wholesale markets, the binding constraint is logistics rather than production. And because the expensive part of the basket is the nutrient-dense part, the reflex policy backfires: subsidising cereals, FAO’s simulations say, can raise the relative cost of a healthy diet. Torero Cullen restated the point when a correspondent misheard it as an argument against food subsidies in general: the target is support for cereals specifically, and the proposal is to repurpose it toward high-value commodities.
This is the part worth being clear about. A metric changed a policy conclusion. The older indicators, which count hungry people, cannot generate that finding, because a headcount has no inside. That is the case for the newer measure, and it is a better case than the headline figure it produces.
The same decomposition produced the briefing’s most practical number: building the reference basket from locally available foods rather than a standardised global one cuts the cost of a healthy diet by about 34% globally and by almost 80% in Africa. Geography matters inside countries too, with fruit and vegetables costing more than twice as much in remote rural areas as in major cities.
The surprise in the regional table
Asked which country faces the worst affordability crisis, Torero Cullen went to regions and landed somewhere counterintuitive.
He called it a contradiction, and it is: this is a continent that exports cereals from Brazil, Argentina, Paraguay and Uruguay and high-value commodities from Peru, Chile and Mexico. His explanation was sequencing. Production has been oriented toward exporting more rather than building enough domestic supply to pull local prices down, and local varieties that would lower the cost are under-used. A region can be a food superpower and price its own citizens out of nutrition, and only a cost metric would ever show it.
On the food that never reaches anyone, he separated two ideas the debate usually merges. Losses run from producer to wholesale market; waste runs from retail to consumer. Losses are around 13.3% and waste around 19%, and the waste sits somewhere most campaigns do not point.
These two are the rare figures on this page that can be checked, and they hold. FAO’s Global Food Loss Index (SDG indicator 12.3.1a) puts losses at 13.3% in 2023, and UNEP’s Food Waste Index Report 2024 (12.3.1b) puts waste at 19% of the food available to consumers, with households the largest share. The split of the indicator between two custodian agencies is the reason the two halves are so rarely quoted together.
The risks the report does not price
A question about a pathogen spreading through salad crops drew out a strand of FAO work that sits outside SOFI. FAO has begun simulating the global impact of pests and diseases on specific commodities, and Torero Cullen’s example was soybean rust in the United States, which he said will reduce yields by 30 to 40%. His second point was the more interesting one, and it is about monoculture as a measurement blind spot.
Genetic uniformity converts a local outbreak into a correlated global one, and the insurance he named against it is the gene bank, the stock of varieties that might carry resistance. For maize and wheat that stock exists. Where the crop is uniform, it does not help. Asked whether the situation is worse than it used to be, he reached past agronomy: “we have trespassed seven of the nine planetary boundaries,” so the frequency and intensity of climate events, and with them the mutation of pests, are changing the ground the forecasts stand on. That count is right. The Stockholm Resilience Centre’s Planetary Health Check 2025 moved ocean acidification into the breached column, making seven of nine.
On a question about a proposal for city-owned supermarkets in New York City, his answer stayed on the economics: a subsidy delivered through such a channel is defensible so long as it is targeted at the people at the tail of the distribution who genuinely cannot afford a healthy diet, on the argument that the state pays later for the diet-related disease it does not prevent now.
Our read
The honest thing to say about this page is that it reports an unpublished document. The site’s standing rule is to verify figures against the primary source and to drop rather than repair what cannot be verified. Here the primary source does not exist yet, so nearly everything above is attributed to a man speaking from memory into a microphone, transcribed by software. Two figures survived that test independently, the food loss and waste numbers, because they were already published elsewhere. The rest is a preview, and a preview is a claim, not a record.
That is not a complaint about FAO, which is doing what agencies do. It is a note about what a launch cycle does to verification: the week when a report is most talked about is the week it is least checkable, and by the time the document lands the framing has set. The cost shares given here differing from the ones given a day earlier at the same headquarters is a small illustration of the cost of that gap.
Underneath the caveat there is a real finding, and it is the one worth carrying into 21 July. The cost of a healthy diet is doing analytical work the hunger headcounts cannot, because it decomposes, and decomposition is what turns a description into a decision. Set against that, the report’s own boundary is the sharper story: it stops at 2025, the input shock started later, and the transmission runs three to six months. The moderator closed by thanking Torero Cullen for an “interesting and somewhat depressing report.” The depressing part is not in the report. It is in the gap between when the world changes and when the statistics are allowed to notice.
Why it matters for the SDGs
This is SDG 2 (zero hunger), measured through 2.1.1, 2.1.2 and the newer cost and affordability lens that SOFI 2026 puts at its centre. It runs into SDG 12 through indicator 12.3.1, the loss and waste pair split between FAO and UNEP, and into SDG 3 through the non-communicable disease burden Torero Cullen ties to diets people cannot afford. The Hormuz thread reaches SDG 7, since fertilizer and energy are the channel through which an energy shock becomes a food price, and the pest and planetary-boundary material lands on SDG 13 and SDG 15. The counting lesson is narrower and applies beyond food: an annual indicator with a one-year lag is a good instrument for accountability and a poor one for warning, and the UN increasingly asks its flagships to do both jobs at once. SOFI answers by bolting a 2030 projection onto a 2025 count. Whether that is enough is a question the 21 July document, and not this briefing, will have to answer.
Watch & read
- The State of Food Security and Nutrition in the World, the FAO flagship page. SOFI 2026 launches in Rome on Tuesday 21 July 2026 and will be posted there. Nothing on this page has been checked against it.
- The cost of a healthy diet: what SOFI 2026 measures, the HLPF special event of 14 July that served as the report’s curtain-raiser, with the fuller account of how the metric is built and what the room contested.
- For an oil shock, a dashboard and a stockpile. For fertilizer, neither yet, the FAO side event of 14 July on agrifood risk, the hidden costs of the food system, and AMIS being extended to fertilizers.
- FAO Global Food Loss Index (SDG 12.3.1a) and the UNEP Food Waste Index Report 2024 (12.3.1b), the published sources behind the loss and waste figures.
- Planetary Health Check 2025, Stockholm Resilience Centre, on the seventh breached boundary.
- Full HLPF 2026 coverage.
Quotations are lightly edited from an automated (Otter.ai) transcript of the recording and should be read as close paraphrase. The recording is undated; 15 July 2026 is our inference from what the speaker says had happened “today or yesterday,” from the moderator starting a regular noon briefing minutes later, and from the speaker’s known commitments on 14 July. SOFI 2026 had not been published when this briefing was held, and it launches in Rome on 21 July 2026, so every figure attributed to the report here is as the FAO chief economist stated it aloud and could not be checked against the report. The food loss and waste figures are the exception and are linked to their published sources. FAO gave two materially different figures on consecutive days for the share of the cost accounted for by fruit and vegetables, so we do not print that share at all and say on the page why. One figure on the growth effect of the Strait of Hormuz disruption was too unclear in the recording to reproduce and is omitted. Journalists are cited by the outlet they named rather than by name.