Every weight.
Every source.
An index whose workings are hidden is a horoscope. An index whose workings are published is an argument — and arguments can be won or lost on the evidence. Everything needed to disagree with this one is on this page.
The two axes
Power level (0–100)
Where the nation stands today. A composite of six components, weighted as shown. The United States anchors the upper range at 88 rather than 100, leaving headroom — an index whose leader is already at maximum cannot record a rise.
| Component | Weight | Principal indicators |
|---|---|---|
| Share of world output | 25% | GDP share, PPP and nominal blended 50/50 (IMF WEO) |
| Manufacturing value-added share | 20% | UNIDO / World Bank MVA share of world total |
| Military capability | 20% | Expenditure (SIPRI), deployable force, shipbuilding and munitions throughput |
| Technology frontier | 15% | Semiconductor, AI compute, aerospace, pharmaceutical frontier presence |
| Financial centrality | 12% | Reserve-currency share, settlement share, market depth |
| Alliance depth | 8% | Treaty commitments weighted by partner capability |
Decline velocity (−10 to +10)
Not a GDP forecast. The first derivative of relative position, constructed in two equal halves:
- 50% trailing. The annualised five-year change in power level, scaled so that a one-point annual loss of level reads as roughly −1.5.
- 50% forward. A momentum score built only from commitments already locked in — birth cohorts already born, pension liabilities already accrued, debt already issued, retirement ages already legislated. Nothing discretionary, nothing forecast.
The forward half is what separates this index from a growth table. China’s trailing level rose between 2021 and 2026, from 70 to 74. Its forward half is deeply negative because the 2046 workforce is already born and already too small. Net: −3.4. A pure trailing measure would have scored China positive and missed the entire argument.
The five engines
Each engine is scored 0–100, where 100 is healthy. A low score is a failing engine, not a moral judgement. Engines are published separately and never averaged into a single figure, because averaging Nigeria’s demographic 82 with its institutional 26 destroys both numbers.
| Engine | Indicators | Primary sources |
|---|---|---|
| Demographic | TFR, working-age trajectory to 2050, old-age dependency, net migration rate, migration policy durability | UN WPP, OECD, national statistical offices (NBS, Rosstat, Destatis, e-Stat) |
| Fiscal | Debt/GDP path, net interest as share of outlays, reserve-currency privilege, external balance, revenue concentration | CBO, IMF Article IV, national treasuries, PGPF |
| Productive | Labour productivity growth, investment efficiency, industrial output vs peak, industrial energy cost, robot density | OECD, ifo, IMF, national industrial series |
| Institutional | State capacity, rule of law, succession clarity, policy reversibility, regulatory predictability | V-Dem, Freedom House, World Bank WGI |
| Cohesive | Institutional trust, polarisation, inequality, elite overproduction proxies (graduate unemployment), legitimacy indicators | V-Dem, national labour statistics, survey series |
The full index
All twenty nations, sorted by power level. Engine columns abbreviated: Demo, Fisc, Prod, Inst, Coh.
Confidence grading
Every figure on this site carries one of three grades, inherited from the underlying research methodology and applied consistently:
- High Official statistics, primary disclosures, treaty text, or two independent outlets in agreement.
- Medium Single credible source, reported-but-unconfirmed, or an official target rather than an outcome.
- Low Model output, long-horizon extrapolation, advocacy-cited figures, or this index’s own subjective judgement.
Two disciplines follow from this and are applied throughout. First, claims sourced to a single outlet are labelled as such rather than presented alongside confirmed ones. Second, ranges are published as ranges. Russian war deaths appear as three separate figures on three separate bases — a named-and-confirmed floor of 223,539, a probate model at roughly 352,000, and a GCHQ assessment near 500,000 — because collapsing them into one number would misrepresent all three.
Known limitations
Stated plainly, because an index that does not publish its weaknesses is asking to be trusted rather than checked.
- The forward half of velocity is a construction, not a measurement. It is defensible and the inputs are locked-in facts, but the weighting between locked-in demography and locked-in fiscal commitments is a judgement call.
- Power level weights are argued, not derived. There is no objective basis for 25% output versus 20% manufacturing. Different weights move the horizontal positions; they move the vertical ones far less, which is the more important axis.
- Institutional and cohesion scores rest on indices with their own methodological disputes. V-Dem’s recorded 24% one-year fall in the US Liberal Democracy Index — and a rank move from 20th to 51st of 179 — is a large swing that reasonable analysts contest.
- Probability bands on the conflict board are subjective. They are the author’s judgement, graded Low confidence, and published only because unscored intuitions are worse than scored ones.
- Nations are unequal units. The EU is disaggregated into Germany, France, Italy, the UK and Poland because the dispersion within Europe is larger than the dispersion between several continents. That is an analytical choice with consequences.
- An unrevised index is a wrong index. Quarterly re-scoring, published changelog, with the previous version left visible.
Revision policy
The index is re-scored quarterly. Every revision publishes what moved, by how much, and on what evidence. Prior versions remain accessible. When a score changes because the world changed, that is an update; when it changes because the index was wrong, that is a correction, and it is labelled as one.
Continue
The five engines and the literature ·
Editorial standards and corrections ·
Back to the index