On 24 February 2021, a shipment of 600,000 COVID-19 vaccine doses arrived in Accra, Ghana. The AstraZeneca/Oxford vaccine had been manufactured by the Serum Institute of India and dispatched from Pune the previous day. It was COVAX’s first delivery in Africa.^[1]

An arrival is a useful place to begin. A shipment has a quantity, a route and a date that can be checked. After the fundraising and negotiations, something that had existed in plans could now be counted in boxes.

But an airport delivery is not a vaccination record. Nor does one completed shipment tell us when the next will come. Those are different achievements, even when a press announcement can make them feel like consecutive frames of the same film.

COVAX brings that distance into focus: the distance between agreeing who should receive something and being able to place it in their hands. Fair allocation is difficult. Obtaining the thing being allocated is another difficulty, governed partly by decisions outside the allocation scheme.

A share before a bidding war

In 2020, no purchaser knew which vaccine candidates would succeed, when they would be authorized or how much could be manufactured. Pooling purchases offered a way to spread that risk. It also offered countries with less purchasing power something better than waiting until everyone richer had finished shopping.^[2]

COVAX was not a world health government. Its machinery connected different organizations: Gavi’s purchasing and financing arrangements, CEPI’s support for development and manufacturing preparation, WHO’s technical and allocation work, and UNICEF’s procurement and delivery role. Governments and manufacturers retained powers of their own. Connecting these actors did not place every decision under a single command.^[1][2]

WHO’s initial 2020 framework envisaged proportional allocation, beginning with a benchmark sufficient for roughly 3% of national populations and progressing towards 20%, with priority groups in view. Twenty per cent was an initial floor, not a permanent ceiling on vaccination, and certainly not a guarantee that deliveries would arrive by a particular date.^[2]

There was a readily intelligible reason for this approach. Access should not begin only where the strongest buyer had secured the earliest contract. When everyone was short, a country’s ability to start protecting people should not depend entirely on its place in a bidding war.

Population shares nevertheless set aside important differences for the purpose of reaching a workable agreement. Age profiles, health systems and the timing of outbreaks were not identical. A common percentage could make coordination possible without making those differences disappear. It was a starting method, not a final description of equal need.

And the scheme could allocate only what it could obtain. An allocation table did not itself control every manufacturing line, export decision or bilateral purchase. The moral ambition reached further than the partnership’s command over supply. That gap was not an obscure drafting defect. It was central to what the arrangement could deliver.

This does not make the ambition fraudulent. It makes its practical test more demanding. Saying that no country should be left behind remains a reason to act. It does not settle who can obtain doses already promised elsewhere, or what happens when those with the best position in the queue have little incentive to move.

A revised forecast does not restart the clock

On 8 September 2021, the partnership published a reduced supply forecast. Its most likely scenario was access to 1.425 billion doses that year, with the two-billion-dose milestone for release for delivery postponed to the first quarter of 2022. This was a forecast made at the time, not a subsequently audited total of actual deliveries.^[3]

The joint statement identified early purchases by wealthy countries, export restrictions, the prioritization of bilateral deals, production difficulties and regulatory delays. This was the partners’ public account, not an adjudication of liability under each contract. Even so, it showed why an agreed allocation method could coexist with a very different delivery timetable.^[3]

In a retrospective chart, a postponement can look like a movement from one column to another. During an epidemic, the interval is not empty. Protection that arrives later may still be valuable. It cannot travel backwards and supply protection during the period when people were waiting.

That observation does not establish that every infection or death during the wait was caused by COVAX. Transmission, domestic policies, health services and other procurement routes also mattered. Acknowledging the cost of time is not the same as possessing a defensible ledger assigning every loss to one organization.

Yet uncertainty about that ledger cannot make the delay irrelevant. Planning vaccination requires staff, sites and arrangements for further doses. When expected supply changes, plans must change too. This is a practical implication of unreliable delivery, not an invented account of a particular nurse’s working day. The inconvenience does not remain where the forecast was revised.

A dose therefore has more than a delivery count. Its arrival date, remaining shelf life and suitability for use affect what help it provides. A donation that is convenient to dispatch can still be difficult to use. If the recipient bears all the risk of a sudden change, generosity in the shipping record need not translate into a comparable gain in usable protection.

The September statement itself asked for more predictable donations with longer shelf lives.^[3] Such requests may sound like a retreat from lofty principles into logistics. In fact, they make the principle concrete. Who carries the work created by the gift? Can the receiving health system say when and how it can use what is being offered?

The Accra shipment deserves to remain an achievement. Later shortages did not unmake its arrival. Precisely because it was real, however, it should not be required to stand for every delivery that had not occurred. A completed link in a chain is not evidence that the whole chain has held.

When the question changes

By the allocation mechanism’s second phase in 2022, conditions had changed. WHO’s explanation moved towards allocations driven more by participant demand, considering supplies obtained outside COVAX, capacity to use doses and rolling arrangements under uncertainty. This was an adaptation to changed circumstances, not simply the execution of every detail envisaged for a later phase in 2020.^[4]

Earlier, the urgent problem had been securing scarce supply. Later, it was also necessary to ask whether a proposed delivery matched what could actually be used. These are not contradictory obligations. The reason for protecting people can remain while the way of pursuing it changes. Equal concern need not mean repeating yesterday’s method indefinitely.

Listening to demand, however, cannot mean treating every small request as proof of small need. Staffing, budgets and service capacity can restrict what a country is able to ask for. An inability to use a larger consignment immediately is not evidence that everyone who could benefit is already adequately protected. A demand-led arrangement still needs to hear what makes demand difficult to express.

The converse matters too. An outside organization cannot infer from those constraints that it always knows better than the recipient. Helping a health system do more is different from overruling its plans. Sometimes another conversation about timing is more cooperative than another shipment. Neither a universal formula nor a declaration of local ownership makes that judgment unnecessary.

Individuals must also remain visible. In 2021, the COVAX AMC established a no-fault compensation programme for eligible serious vaccine-related harm in its 92 supported lower-income countries and economies. The programme had eligibility conditions and procedures; it was not a promise to compensate every kind of health loss.^[5]

It would therefore be wrong to say there was no individual compensation mechanism at all. But vaccine-injury compensation does not answer a procurement question about a delayed consignment. One route to redress cannot stand in for every other question. A government’s participation likewise cannot supply an individual’s informed agreement to a particular medical intervention.

In December 2023, the partners reported that COVAX had supplied nearly two billion doses to 146 economies. They announced its closure on 31 December, with continuing support moving into regular immunisation arrangements. These were doses delivered, not two billion people who had completed a vaccination course.^[6]

Nearly two billion doses should not disappear behind an effortless verdict of failure. Nor should an accumulated total absorb the earlier waiting into an equally effortless verdict of success. Cooperation expanded access without making purchasing power, productive capacity or the power to set priorities equal. Both belong in the account, without being converted into competing points on a scorecard.

The useful question is not which final adjective COVAX deserves. It is how to read an allocation promise: what it aims to protect, what resources it actually commands, which decisions still depend on someone else, and what the calendar means to those awaiting delivery. An aggregate can describe a considerable achievement while leaving important promises unfinished.

On 24 February 2021, 600,000 doses arrived in Accra. That sentence should be allowed to finish. So should the question that follows it: when will the next shipment arrive?