Prices went up 9 percent.
Your raise was 3.

Picture how you think your pay rise is set. A manager weighs what you did this year, what you would cost to replace, what the market says people like you are worth, and arrives at a number for you.

That is not what the payroll data shows. When you can see every wage change at tens of thousands of firms — which is what administrative payroll records let you do — a much blunter machine appears. Most firms pick one annual increase and apply it to the large majority of their continuing staff, in a single month of the year. Economists call the month the firm’s on-cycle month and the number its wage growth norm.

The clustering is remarkable. Among workers at a given firm who got a raise, about six in ten landed within half a percentage point of that firm’s own modal increase, and the great majority within a point and a half. Whatever your manager told you in the review, the number was mostly decided before the conversation started.

If firms really were re-optimising each worker’s wage against their own productivity and outside options, raises would land on a smooth spread of numbers. They do not. They pile up on round ones.

Fourteen percent of the workers in that range got exactly three point zero percent — the single most common raise in the American economy. Around four in ten of all increases under 6 percent sat within a hundredth of a point of a whole or half number. That is not a market clearing. That is a convention.

Here is the whole mechanism at the scale of one person. Set a starting wage, pick the norm at your employer, and watch the two lines separate. The blue line is what your payslip says. The red line is what that money actually buys, held in December 2020 dollars.

Notice what the shape of the blue line does to the argument. Nobody’s pay was frozen. Every year it went up. Every year there was a conversation in which the number sounded like good news. The erosion happened in the gaps, silently, and the raise that arrived each January was never big enough to buy the gap back.

Then flip the switch. Indexation is not exotic: it is the same norm plus whatever inflation ran above its normal level. The red line stops falling almost immediately.

Two-thirds of them ended the first year with less purchasing power than they started with. That part is unsurprising: almost every model of sticky wages predicts a bad year when prices jump. What is surprising is what happened next. The loss did not heal. Four years on, 43 percent were still behind, and the people still behind were further behind than ever.

That last figure is the quieter one, and arguably the more important. “Did your pay beat inflation” is a low bar. It ignores the fact that people’s wages normally rise as they get better at their jobs. More than half of these workers spent four years going essentially nowhere.

This was not a tail event concentrated on a few unlucky people. The whole distribution slid left, more or less evenly.

A firm’s norm binds you only while you stay. Walk across the street and your wage is set fresh, by a market that has noticed inflation. The payroll data shows this working almost perfectly: job-changers’ nominal wage growth tracked inflation nearly one-for-one, while job-stayers’ barely responded at all.

The catch is arithmetic. Switching employers is rare — a little over two percent of workers a month, barely up from before the pandemic — and for any one person it happens in one year out of several. Most people who changed jobs during this period did so once and spent the other three years back under a sticky norm. Adding every job-changer to the count moves the share of workers with a four-year real wage decline from 43 percent only to 37 percent — and 58 percent of all workers still finished below the path their pay had been on before the pandemic.

The other door: ask

Firms did respond to the inflation — just not by moving the norm. They moved individual people, one at a time, outside the annual review. The share of job-stayers receiving more than one base-wage adjustment in a year jumped from around 16–18 percent before the pandemic to about 27 percent in 2021.

On-cycle raises sit between two and four percent almost without exception. Off-cycle raises are a different animal: most exceed 4 percent, a third exceed 8, and nearly a fifth exceed 12. They are promotions, counter-offers, retention saves. They are what happens when somebody does something.

If sticky norms held real labour costs below where a fully adjusting market would have put them, the difference should show up somewhere on the other side of the ledger. The paper does the arithmetic in advance: labour is about 60 percent of costs, so a real wage shortfall of two to four percentage points implies a rise in the profit share of output of roughly 1.2 to 2.4 points.

Observed: a jump from a steady 11.4 percent across 2016–2019 to 13.1 percent across 2021–2025. A rise of 1.7 percentage points, squarely inside the predicted range, and the highest sustained level in half a century. It began when the inflation began and it has not come back down.

Keeping up required doing something: updating a résumé, sitting interviews, taking the disruption of a new job, or preparing an argument and making it to a reluctant manager. That effort produced no extra output. It was spent purely to stand still. Economists call that a deadweight loss: it is simply burned.

Not keeping up cost something too, but the money went somewhere. It stayed with the employer as a lower wage bill. That is a transfer, not a burn.

Both belong in any honest accounting of what inflation cost people, which is why the paper’s own summary is blunter than the numbers suggest: even the workers who kept up were made worse off by it.

Belgium took the same inflation as Germany, the Netherlands, Denmark and the rest of the euro area. The same energy shock, the same supply chains, the same Ukrainian refugee arrivals, the same tight labour market. On inflation, productivity, unemployment and vacancies it is unremarkable among its neighbours.

What is different is the wage-setting rule. Belgian wages are automatically indexed, through collective agreements, to a smoothed consumer price index. It reaches nearly every private and public employee and public pensions besides. During the inflation episode Belgian law also capped merit raises above indexation at zero — so Belgian workers got the indexation and nothing else.

Real wages fell about 8 percent everywhere between early 2021 and the end of 2022. Then Belgium climbed back to where it had started, during 2023, and stayed there. Nobody else did.

And Belgian consumer confidence came back with the wages — not with the disinflation. By 2024 it was within about three points of its pre-pandemic norm while Germany, the Netherlands and Denmark were still nine to eleven points down, on the same timetable of falling inflation.