The Last Third

Executive Summary

The SDG record since 2015 is not uniformly bad. It is unevenly good, and the unevenness follows a rule that tells us where the remaining years should be spent.

Two-thirds of the way through the 2030 Agenda, the United Nations' Sustainable Development Goals Report 2026 finds that of 139 targets with trend data, 36 per cent are on track or making moderate progress, 49 per cent are advancing too slowly, and 15 per cent have regressed below their 2015 baselines. That is a slight improvement on the 2025 assessment and a long way short of what was promised.

Aggregate scores invite aggregate explanations — insufficient finance, insufficient ambition, the pandemic, conflict. All are real. But the variation between targets is more instructive than the average, because it shows that the same countries, in the same years, under the same shocks, made rapid progress on some goals and none on others. The difference lies mostly in what kind of thing each target asks for.

Section 1 sets out the overall position. Sections 2, 3 and 5 take the three categories in turn, and Section 4 explains why services stall. Section 6 shows that even the product successes are now meeting their limit, and Section 7 places the "sustainable" half of the agenda in context. Sections 8 and 9 give an agenda and conclusions.

1. Two-Thirds Through

The headline assessments agree on direction even where they differ on method: some progress, nowhere near enough, and a widening gap between the richest and poorest.

Other assessments are harsher. The Sustainable Development Solutions Network's Sustainable Development Report 2026 projects that only about 16.5 per cent of targets will be achieved by 2030 and that none of the seventeen goals is on track; its most off-track goals are cities, oceans, land and peace and institutions. UNDP's Human Development Report 2025 recorded the smallest rise in the Human Development Index since 1990, excluding the crisis years of 2020 and 2021, and a fourth consecutive year of widening inequality between low- and very-high-development countries — reversing two decades of narrowing.

The regional picture adds a qualification to the gloom. SDSN finds that East and South Asia have outperformed every other region since 2015, with Nepal, Cambodia, Bangladesh and the Philippines among the fastest-improving countries. Progress at speed is possible within the current period, in poor countries, under the same global conditions. The question is what those countries progressed on.

Table 1 — Three kinds of target

The classification is the author's, applied to the evidence in Sections 2 to 5. Many targets combine elements of more than one kind; the table assigns each by the element that has proved binding.

2. What Moved: Products

The goals that advanced fastest were the ones that could be met by something bought or installed once, by a supplier whose costs were falling.

The UN's 2026 assessment highlights the same successes that SDSN ranks as the most on track: internet use rose from 40 to 74 per cent of the world's population over the Agenda period; electricity now reaches 92 per cent; renewable capacity per capita is 2.2 times its 2015 level; new HIV infections fell by 30 per cent between 2015 and 2024; and nearly a billion more people gained safely managed drinking water. SDSN's five most on-track targets are mobile broadband subscriptions, adolescent fertility, new HIV infections, internet use and electricity access.

What these share is not merely that they are technological. It is that each can be delivered once — a handset, a subscription, a connection, a solar kit, a course of antiretrovirals — by a provider whose unit cost was falling fast, and frequently paid for by the household rather than the state. A mobile network expands because each new subscriber is profitable; a solar kit spreads because pay-as-you-go turns its price into a daily payment, the model examined in a recent report in this series. Neither requires a ministry to find recurring budget for every additional person served.

East and South Asia's outperformance fits the same reading. The fastest-improving countries combined rapid growth with heavy uptake of exactly these goods — connectivity, power, mobile finance — and, in several cases, manufacturing employment that raised household incomes enough to buy them. A product target is one a growing economy can partly meet by growing.

3. What Stalled: Services

The goals that need someone paid every month, indefinitely, moved slowly or not at all — and the gap between income groups is enormous.

Social protection is the cleanest illustration because its coverage is measured consistently. The ILO's World Social Protection Report 2024–26 finds that, for the first time, 52.4 per cent of the world's population receives at least one cash benefit, up from 42.8 per cent in 2015. But coverage is 85.9 per cent in high-income countries, 71.2 per cent in upper-middle-income, 32.4 per cent in lower-middle-income and 9.7 per cent in low-income countries — where it has barely moved since 2015. Low-income countries spend about 2 per cent of GDP on social protection; high-income countries spend almost 25.

Education tells the same story in a different metric. Before the pandemic, the World Bank estimated learning poverty — the share of ten-year-olds unable to read and understand a simple text — at 57 per cent in low- and middle-income countries and 86 per cent in sub-Saharan Africa; its 2022 simulations put the post-closure figure at around 70 per cent, an estimate rather than a measurement (World Bank). The UN's 2026 report counts 273 million children and young people out of school and maternal mortality at nearly three times the global target.

Enrolment, notably, is the part of education that behaves like a product — a school place is built once — and it improved. Learning, which depends on a trained teacher paid every month and supported by a functioning system, is the part that did not.

4. Why Services Stall: the Interest Bill

A service is a recurring claim on the budget. In much of the Global South, the first recurring claim on the budget is now the creditors.

UNCTAD's A World of Debt 2026, as reported on publication, finds that developing countries paid US$995 billion in net interest on public debt in 2025, nearly three times the 2010 level, and that 3.7 billion people live in 51 developing countries where interest payments exceed government spending on health or on education. The median developing country now allocates 8.1 per cent of government revenue to net interest, almost double the 2010 share. UNCTAD estimates that borrowing at rates comparable to advanced economies would save developing countries around US$500 billion a year.

The previous edition gave the regional comparison that makes the mechanism concrete. Between 2021 and 2023, African governments spent about US$70 per person on interest, against US$63 on education and US$44 on health.

The external side tightened at the same time. The UN's 2026 report records external debt of low- and middle-income countries at a record US$8.9 trillion in 2024, and official development assistance falling by a record 23.1 per cent in 2025. The financing gap for the Goals is put at about US$4 trillion a year.

5. What Reversed: Systems

The goals that require whole systems to behave differently — food, land, norms, peace — are the ones that did not move or went backwards.

The UN's 2026 report is direct about these. The global extreme poverty rate is projected to reach only 10 per cent in 2026 — three percentage points below its 2015 level, far short of ending it. Around 2.3 billion people face moderate or severe food insecurity, and more than 150 million children remain stunted, although that number has fallen. None of the gender equality targets is on track. The number of people affected by climate-related disasters has more than doubled since 2015; violent conflict is at its highest level in decades, with more than 117.8 million people forcibly displaced; and extinction risk is worsening across all species groups.

What these targets share is that no single delivered thing meets them. Hunger is a property of food systems — production, storage, prices, incomes, conflict, climate — rather than of any one input. Biodiversity depends on how land is used everywhere. Gender equality depends on norms, laws and labour markets together. Extreme poverty, once the easiest gains from growth are taken, persists among people outside the markets that product delivery reaches and the services that budgets can no longer fund.

System targets are not unreachable, but they respond to governance rather than procurement: coalitions that hold over years, rules that change incentives, institutions that coordinate across ministries. Those are slow to build and easy to disrupt, and the disruptions of the period — pandemic, debt, conflict, climate — have fallen precisely on them.

6. The Frontier: When Products Reach People Who Cannot Pay

Even the success category is now meeting its limit, and the limit turns product targets into service targets.

Electricity access is the clearest warning. Tracking SDG7 2026 reports that 89 million people gained electricity in 2024, yet the number without access fell by only 11.5 million, to 655 million, because population grew almost as fast. The global access rate stagnated at 92 per cent. In sub-Saharan Africa the rural deficit grew from 376 million in 2010 to 447 million in 2024. About two billion people still lack clean cooking.

The households left are those a product market does not serve on its own terms: rural, dispersed, poor, and often unable to make even a small regular payment. Reaching them requires a subsidy, a public connection programme or a guarantee — that is, it requires the product to be partly converted into a service, with all the fiscal exposure Section 4 describes. The easy phase of product-led progress is ending, and the last stretch of every product target looks increasingly like a service target.

7. Where the "Sustainable" Half Fits

For most of the Global South, the binding half of sustainable development is still development — under a sustainability constraint that has become far cheaper to respect.

The most systematic evidence on the relationship between meeting needs and respecting limits comes from the safe-and-just-space framework. O'Neill and colleagues found in Nature Sustainability (2018) that no country meets the basic needs of its residents at a level of resource use that could be extended sustainably to everyone — but also that basic needs such as nutrition, sanitation and the elimination of extreme poverty could most likely be met for all without exceeding global environmental limits. Fanning and colleagues' follow-up (2022), tracking more than 140 countries from 1992 to 2015, found that countries tend to transgress biophysical boundaries faster than they achieve social thresholds, with wealthy countries driving overshoot and poorer countries falling short on social outcomes while largely within ecological limits.

Two implications follow for the Global South. First, the agenda's two halves are not symmetric there: the large deficits are social, and the ecological overshoot is mostly imported through global consumption and emissions. Second, the sustainability constraint on meeting those deficits has loosened sharply because the cheapest new electricity is now clean, a point made at length in the first report in this series. The product category in Section 2 increasingly is the sustainable option, which is part of why it moved.

The sustainability failures the Global South does carry — deforestation, depleted aquifers, collapsing fisheries, urban air pollution — sit overwhelmingly in the system category. They are the same kind of problem as hunger and gender, and they respond to the same kind of remedy.

8. An Agenda for the Last Third

Each kind of target needs a different kind of effort. Treating them as one undifferentiated shortfall is how the services and systems keep losing.

8.1 Products: extend the wins to the frontier

• Pay for the last households explicitly. Connection subsidies, results-based payments for off-grid providers and lifeline tariffs are the bridge from a product market to universal coverage; without them, access rates plateau where the market stops, as electricity has at 92 per cent.
• Keep the cost curves falling. Import treatment, spectrum policy and standards that let cheap clean hardware in are a large part of why products moved; protecting them is cheap and effective.

8.2 Services: defend the fiscal base

• Treat debt as an SDG issue, not a separate one. Where interest exceeds health or education spending, the most effective education and health policy may be a restructuring, a maturity extension or a shift to cheaper borrowing — UNCTAD puts the potential saving from borrowing at advanced-economy rates at around US$500 billion a year.
• Raise and protect domestic revenue, and ring-fence service budgets in fiscal rules, so that the next shock does not fall first on teachers and clinics.
• Use delivery technology where it genuinely lowers the recurring cost — digital payments for cash transfers, structured pedagogy, community health workers — without pretending it removes the need to pay people.

8.3 Systems: govern, rather than procure

• Build coalitions that outlast an electoral cycle around food systems, land use and gender, where change depends on many actors moving together.
• Change incentives through rules — land tenure, deforestation-free procurement, inheritance and labour law — rather than funding projects that operate around unchanged incentives.
• Protect the institutions that coordinate across ministries, because system targets fail at the boundaries between them.

8.4 For development partners

• Stop funding what markets now fund. With ODA down by almost a quarter in 2025, scarce grant money belongs at the frontier of product targets, in the recurring costs of services and in the long work of system change — not in the categories private capital already serves.

9. Conclusions

The SDG record is a diagnosis as well as a scorecard. It shows which kinds of progress the current arrangements can deliver, and which they cannot.

Two-thirds of the way through, 36 per cent of assessable targets are on track or moving moderately and 15 per cent have gone backwards. Behind those numbers is a pattern that held across regions and shocks. Where a target could be met by a product delivered once at falling cost, progress was fast: internet use from 40 to 74 per cent, electricity to 92 per cent, renewable capacity per head more than doubled, new HIV infections down by 30 per cent. Where it needed a service paid for every month, progress stalled, most visibly in social protection coverage of 9.7 per cent in low-income countries and learning poverty of 86 per cent in sub-Saharan Africa before the pandemic. Where it needed a system to change, it reversed or barely moved: extreme poverty at 10 per cent, three points below 2015, and no gender equality target on track.

The service category is where the explanation is most concrete and the remedy most neglected. A recurring public function is a recurring claim on revenue, and in the countries that need these services most, interest now comes first: 3.7 billion people in 51 developing countries live where interest exceeds health or education spending, net interest reached US$995 billion in 2025, and ODA fell by a record 23.1 per cent. The SDG slowdown since 2020 is, to a large extent, that arithmetic made visible.

The product category, meanwhile, is reaching its frontier. When 89 million people gain electricity in a year and the number without it falls by only 11.5 million, the market has done what it can and the last households require public money — which brings them back into the fiscal squeeze. And the "sustainable" half of the agenda, for most of the Global South, constrains less than it once did, because the cheapest way to meet many basic needs is now the clean one; the sustainability failures that remain are system problems, with system remedies.

The last third of the Agenda will not be won by doing more of everything. It will be won, to the extent it is, by matching effort to the kind of target: subsidy at the frontier of products, fiscal space for services, and the slow work of governance for systems. The first is cheap, the second is a question of debt and revenue, and the third is a question of politics. The record since 2015 says that is where the remaining progress will be decided.

References

Every quantitative claim above is attributed inline. The principal sources are collected here.

• United NationsThe Sustainable Development Goals Report 2026 — 36 per cent of 139 targets on track or moderate, 49 per cent too slow, 15 per cent regressed; internet use 40 to 74 per cent; electricity 92 per cent; renewable capacity per capita 2.2 times 2015; new HIV infections down 30 per cent; extreme poverty projected at 10 per cent in 2026; 2.3 billion food insecure; 273 million out of school; external debt US$8.9 trillion; ODA down 23.1 per cent in 2025. Summary at UN DESA; the 2025 assessment (35 / 47 / 18) via the Secretary-General's progress report.
• Sustainable Development Solutions NetworkSustainable Development Report 2026 — about 16.5 per cent of targets projected to be met by 2030; none of the seventeen goals on track; most on-track targets and East and South Asia's outperformance. The 2025 edition is here.
• UNDPHuman Development Report 2025 — the smallest HDI rise since 1990 excluding crisis years, and a fourth year of widening inequality between low- and very-high-HDI countries.
• ILOWorld Social Protection Report 2024–26 — 52.4 per cent covered globally, 9.7 per cent in low-income countries; spending of 2 per cent of GDP against nearly 25 in high-income countries. Income-group detail via ILOSTAT.
• World BankThe State of Global Learning Poverty: 2022 Update — 57 per cent pre-pandemic in low- and middle-income countries, 86 per cent in sub-Saharan Africa, and a simulated post-closure figure of about 70 per cent.
• UNCTADA World of Debt 2026 — US$995 billion of net interest in 2025; 3.7 billion people in 51 countries where interest exceeds health or education spending; median 8.1 per cent of revenue; potential US$500 billion annual saving — as reported by Anadolu Agency on publication. The Africa per capita comparison is from A World of Debt 2025.
• IEA, IRENA, UNSD, World Bank, WHOTracking SDG7: The Energy Progress Report 2026 — 655 million without electricity in 2024; 89 million gained access while the deficit fell 11.5 million; access stagnant at 92 per cent; about two billion without clean cooking. See the WHO release and SEforALL on the rural deficit.
• O'Neill et al.; Fanning et al.A good life for all within planetary boundaries, Nature Sustainability (2018); The social shortfall and ecological overshoot of nations, Nature Sustainability (2022).
• H HeuristicsRelated reports in this series: from payments to power; financing the synergy; convergence without absorption; and the clean transition as mitigation infrastructure.