Tests the model has to pass

Read from each release's committed output. A release with fewer than nine dimensions had too few countries with all nine scored, so it reads the dimensions scored for at least 90% of countries.

Does what is left move together? Reads structure

Take out of each capability the part its income predicts, and the leftovers still move together. Across the 110 countries with all nine, one factor carries 29.3% of what is left, where unrelated leftovers would give 16.3% by chance and 18.0% in 19 draws out of 20. It runs mostly through Anticipation, Learning, Coordination and Trust.

First-factor share 29.3%. Eigenvalues, largest first: 2.64, 1.48, 1.17, 1.11, 0.65, 0.60, 0.50, 0.48, 0.38. Chance at 109 countries, one fewer because the income line uses one up: 16.3%, 95th 18.0% (2000 draws, seed 20261001). Each residual column shuffled on its own: 16.2%, 95th 17.8% (2000 draws, seed 20261002). A share above the chance 95th reads structure.

Do countries at the same income share a shape? Reads shapes differ

Countries at the same income do not share a shape. A country's shape sits about as far from its 10 nearest neighbors in income as from countries picked at random (a mean distance of 1.44 against 1.48), and the shapes line up along patterns that leftovers dealt out at random do not produce: one pattern carries 22.6% of them, against 19.4% at the random 95th percentile. Neighbors still differ less than leftovers dealt out at random would (1.44 against 1.78), because part of what is left is a level: a country above its income line on one capability tends to sit above it on others.

Each residual is divided by its spread so every capability counts the same, and a shape is a country's nine values minus their own mean. Peer distance is the mean squared gap between a country's shape and each of its peers', averaged over countries. One null keeps every profile whole and deals the incomes out at random, so peers become countries picked without regard to income. The other deals each capability's residuals out across countries, which keeps their spread and breaks any link between one country's nine. 2000 draws each, seeds 20261003 and 20261004.

Does the order hold still? Reads mixed between releases, robust without one country

The order of the leftovers mostly holds between releases. Across 41 capability-by-release comparisons where an indicator change moved them, the rank correlation with the release before fell as low as 0.72 on Trust (n 36), and never below 0.50, where the page would say the leftovers read the indicators more than the countries. Dropping any one country moves an income line by at most 0.47 standard errors, and that country's own leftover by at most 0.11 of the spread around the line, so no single country carries a line.

Rank r is the Spearman correlation of a capability's residual order with the release before, over the countries in both, counted only where a release kept the same countries and moved some residual on that capability by 0.05 points or more. Read across 34 dataset releases, 30 pairs with the same countries. Every lowest value at 0.8 or more reads stable, any below 0.5 reads churning. The shifts are the largest when any one country is dropped and the line refitted: of the slope in its standard errors, and of that country's own residual in the spread around the line, where more than 1 reads fragile.

How much of a profile is income? Reads part

For the typical country, income accounts for 25.3% of how far its profile sits from the average profile, and 25.5% on average across 110 countries. Income is part of a capability profile, and most of it is something else.

For each of 110 countries, one minus its nine squared residuals over its nine squared distances from each capability's average. Mean 25.5%, median 25.3%, pooled over all countries 38.7%. A mean of 50.0% or more reads most, 25.0% or more reads part.