By Albertino Francisco
On 13 December 2024, São Tomé and Príncipe left the list of least developed countries. Its gross national income per head had reached $2,271 against a graduation threshold of $1,306. Its Human Assets Index stood at 91.4 against a threshold of 66. However, its Economic and Environmental Vulnerability Index stood at 40.0, which should have been at 32 or below to pass. Two out of three. Under the rule the Committee for Development Policy applies, two out of three is enough, and so a small island state graduated with exactly the indicator that measures its exposure to shocks sitting eight points on the wrong side of the line.
This graduation happened notwithstanding that four months earlier, on 13 August 2024, the General Assembly had adopted the Multidimensional Vulnerability Index (MVI) built by a High-level Panel, precisely because existing measures were judged to understate the exposure of countries like this one.
Two arithmetics, one year
The interesting part is not that the two instruments disagree. It is that they disagree about arithmetic.
The MVI panel refused to add up its indicators and divide. It chose the quadratic mean — square each component, average the squares, take the square root — on the explicit ground that vulnerabilities do not substitute for one another. A country exposed to coastal inundation is thereby not less exposed to drought, and a country with neither is not the average of the two. Under a simple average, an acute weakness in one dimension is diluted by moderate scores elsewhere. Under a quadratic mean, it is not: the larger the spread between components, the more the aggregate leans towards the worst of them.
The graduation criteria work the other way. The vulnerability index itself is eight indicators — export concentration, export instability, remoteness, agriculture’s share of output, population in low-lying coastal zones, population in drylands, instability of agricultural production, victims of disasters — combined by simple averages, so each carries one-eighth. Then, one level up, its three criteria are not combined at all. They are counted. Meet two, and the third does not have to be met.
Counting is the most permissive aggregation rule there is. It sets the substitution rate between criteria to infinity: no amount of vulnerability can outweigh income and human assets, because vulnerability is simply not in the sum.
What the rule says out loud
None of this is hidden. The Committee’s own handbook states that countries with sufficiently high income may be recommended for graduation “even if they continue to have low human assets and are highly vulnerable” — and since the 2024 review, the income-only route requires income at three times the threshold, raised from twice, which is an acknowledgement that the route was doing too much work.
The safeguard is not in the formula. It is in discretion: the Committee explicitly recommended not to automatically follow the criteria, and allowed deferrence. Discretion is not a bad thing; a rule that ignores an obvious special case is worse than a judgement that catches it. But discretion has a property a rule does not. It is not portable. A finance ministry cannot plan against it, a creditor cannot price it, and a citizen cannot check it.
So the question is not whether the Committee got São Tomé and Príncipe right. It is whether a decision of this weight should rest on a convention nobody defends in public — a simple average — when the same organisation, in the same year, publicly rejected that convention for the same subject matter.
Averages are a political technology
This is not a technical footnote about a category with 44 countries. The additive form is everywhere in development measurement, and it always does the same thing: it converts a profile into a point, and in doing so it makes the worst number in the profile negotiable.
A national poverty rate falls while one province’s rises. A school system improves on average while the bottom decile does not move. A health index climbs on the strength of urban clinics. In each case the aggregate is not wrong; it is simply answering a question nobody asked. The question people actually have is whether the weakest part of the system is strong enough to hold — and no average can answer that, because averaging is the operation that removes exactly that information.
For a small island state the stakes are concrete. Graduation withdraws concessional finance windows, trade preferences and technical support on a timetable set by a date, not by a hurricane. A country whose vulnerability index says 40 is a country that expects a shock. The rule that lets it graduate says, in effect, that the expected shock is a matter for later, and later is when the terms of borrowing have already changed.
The cheap fix
The repair is not more indicators. It is a stated rule at the level where the decision is taken.
If non-compensation is right for the MVI, say where it stops. Publish the combination rule that applies between the criteria — a minimum standard on vulnerability that income cannot buy off, a quadratic mean of the three, or an explicit substitution rate, whichever the membership will defend. Then recompute the past two decades of graduation decisions under it and show where rule and judgement part company. If they never part, the Committee has been reproducing a rule it could have written down, and writing it down costs nothing. If they part, the divergence tells us which countries have been carried by discretion, and discretion can be withdrawn by a future committee in a way a published rule cannot.
Either result is worth more than the present arrangement, in which the arithmetic is conservative in one UN instrument and permissive in another, and the countries that live inside the difference are the ones least able to argue about it.
Albertino Francisco, Ph.D., is an economist and the founder of NOVENTIS, a publishing house in São Tomé and Príncipe. A former Minister of Youth and Sport and for a decade a national consultant to UNDP and the ILO, he writes on development measurement and state capacity. The argument above is developed at length in No Nation, However Small and The Development Equation (NOVENTIS, 2026), published in parallel English and Portuguese editions and available here.
Image: Ji-Elle on
Note: This article gives the views of the author, not the position of the EADI Debating Development Blog or the European Association of Development Research and Training Institutes