The OBR asked what happens if the wages stop. Then everyone went back to arguing about pensions — including us

On Tuesday 7 July the Office for Budget Responsibility published its fifty-year sums. The reporting that followed was about the triple lock. Pensioners' incomes, and who pays for them.

On page 86 of that report there is a box. It is called "The potential impact of AI on the labour share and tax receipts."

It is not a paragraph of throat-clearing. The OBR builds a scenario and puts a number on it.

Start with the baseline. Wages and salaries are 40 per cent of the economy, and the state collects about 38 per cent of GDP in tax. Now let AI do what a lot of people think it will do: the wage and salary share falls from 40 per cent to 20 per cent by 2075-76. The country produces exactly as much. Nothing is smaller. Only the shape of the income has changed.

The tax take falls to under 35 per cent of GDP. Nearly four percentage points, gone.

And the reason is one line, on page 87, which is worth more than everything the political class said this week. Across that projection the state takes an average of 46 pence in the pound on wages and salaries. On profits it takes 27.

Four points of GDP. Now go back and read the coverage. The argument was the triple lock, which the very same report says is worth a fifth of the rise in spending.

Nobody picked it up.

Nominations for the Labour leadership opened yesterday. Andy Burnham has 322 of them. I am not going to do the arithmetic on whether a challenger can still reach the threshold, because it turns on whether the Parliamentary Labour Party has 402 members or 403, and the wires cannot agree. What is true under either count is this: a challenger would now need the signature of every remaining Labour MP, or all but one, and three of those have already said publicly that they back him.

One correction, since the international wires have it wrong and it will matter to anyone counting days. MPs' nominations close at six o'clock on Wednesday 15 July. The 16th is the affiliate stage. Two different deadlines, one week, and most of the world's press has merged them.

So here is a man about to be Prime Minister. Three hundred and twenty-two colleagues put their names to him in a single afternoon. Not one of those signatures was conditional on an answer to the question on page 86.

I want to be careful, because it would be easy and dishonest to make this a Labour failure. It is not. I cannot name a Chancellor, of any party, who has stood up and said plainly what the state does when the tax base thins. The Conservatives did not ask it in office. Reform does not ask it now. The Liberal Democrats do not ask it. The Greens do not ask it. This is not one tribe failing a test. It is a whole political class that has not noticed the test exists, and a watchdog that has quietly started marking it anyway.

And now the part I like least.

We got this report wrong. Our economy piece yesterday morning carried three figures and a quotation lifted from a news agency's summary rather than from the report itself. One of them, the claim that the triple lock accounts for about a third of the rise in state pension spending, is not in the OBR's document at all. Another, the 300 per cent debt figure, is real enough in the world but is not in the report's text, and we had not opened the charts. The quotation we could not find at source has gone.

Worse than any of those: the piece said, in terms, that nobody was asking what happens to receipts when the work thins. That it was not on anybody's list.

It was on page 86. We had not turned to it.

And it nearly happened twice. By late this morning I had four drafts in front of me, all of them about that box, and not one of us had opened it. Our own contrarian had written a piece whose entire argument was that the OBR had not modelled a falling labour share and had not costed one. It had done both, on page 87, with the numbers above. He was wrong in the detail and completely right about the disease.

So I read it. Then I made the paper read it. That is why this edition is late, and it is the only good reason to be late.

One thing I will not do, having spent the day on this. The OBR calls that scenario highly stylised, and says the reduction it models sits at the upper end of what the research supports, and says plainly that the net effect of AI on the labour share is not clear. It has been stable, at around 55 per cent of GDP, since the mid-1980s. None of that is a reason to look away from four points of GDP. All of it is a reason not to tell you it is coming.

The correction is published in full, and it names what came out and why. It is the worst of the four errors because it was not a number. It was a certainty. We were so sure nobody had asked that we did not look.

That is precisely the failure we spend our mornings accusing everybody else of. A paper that runs on sourcing does not get to keep a soft spot for its own conclusions.

Read Elena, who has the absence in Westminster. Read James, who is going back into the report he got wrong, having now read it. And read Tomás, who thinks the three of us are already over-reading a box, and who may well be right.

— M.

**The Playbook**

**Check which state pension age applies to you, not the one in the headline.** It moves from 66 to 67 across 2026 to 2028, so the year you can claim depends on the month you were born.

**When a number is attributed to a report, ask whether anyone opened the report.** Ours came from an agency summary. The figure was wrong, and we printed it. Opening the source costs nothing.

**Ask a candidate where the tax comes from, not whether AI takes your job.** The second question gets a warm answer. The first one does not have an answer yet.

*The Playbook is general information. It is not financial, legal or career advice.*