This number does not show recessions, and that is the most useful thing about it

A measure of how much harder it is to be young ought to spike in a downturn. This one does not, and working out why tells you what it is for.

In 2009 the average adult unemployment rate across rich countries jumped 1.85 percentage points, the largest single-year move in thirty-four years of data. The median youth penalty moved 0.011. In 2020 adult unemployment jumped 1.33 points and the ratio moved 0.100, which is less than it moved in 2019 and less than in 2023, neither of them recession years.

Recessions hit both age groups, roughly in proportion, and the ratio holds still. What that means is that this is a structural measure and not a cyclical one: it describes the standing shape of a labour market rather than where that market sits in its cycle. Which is exactly why a thirty-four year trend in it is worth reading. The cycle is already stripped out.

It is not the unemployment rate in disguise. Across countries in 2025 the ratio correlates with the adult unemployment rate at minus 0.19, and with the youth rate at 0.08. Measuring how each country moves around its own average over time, minus 0.24 and 0.07. The relationship is weak and slightly negative, which is the mechanical effect of rising adult unemployment compressing the ratio. It is nowhere near strong enough for the ratio to be a restatement of the level. It is a separate dimension, and that is the argument for measuring it at all.

What this report does not claim.

It does not claim Britain's rise is caused by AI, or by any one thing. The series starts in 1991 and the rise runs continuously from the mid-1990s. Whatever is happening began long before the current wave of automation. That does not rule automation out of the recent years, but it rules out reading the whole trend as its consequence.

It does not claim a rising ratio is bad in itself. It can rise on good news. Britain's did not.

It does not claim the tenth place for Britain's increase since 1991 is a stable fact. It moves between 10th and 60th depending on the year you start from.

It does not claim apprenticeships cause small youth penalties. Three countries is a pattern worth pursuing, not a finding.

It does not cover countries at war. Ukraine, Sudan, Lebanon, South Sudan and the West Bank and Gaza all drop out of the modelling in the last four years, which means some of the worst cases in the world sit outside this table.

One more limit, on the age band. Youth here is 15 to 24, which in Britain includes large numbers of full-time students who are not looking for work and so are not counted as unemployed. The measure is a ratio of rates, so it is not distorted by how many students there are, but the youth rate is calculated on a smaller and differently composed labour force than the adult one. That is true of every country in the table, and it is part of why the ratio is above 1 almost everywhere. It is a reason to read change over time within a country as the stronger signal, and the level as the weaker one.