Three releases today came with instructions for doubting them
Victoria Shinder2 min readA quiet pattern ran through today's releases, and it is worth naming because it is rarer than it should be: three institutions published a number and, in the same breath, told you what is wrong with it.
The ONS released UK regional GDP back to 1998 and shipped revisions triangles alongside — tables showing how each year's estimate has moved with every subsequent vintage. That is a statistical agency handing you the means to ask whether the gap between two regions is larger than the amount this series routinely revises by.
Governor Barr made his case on housing affordability and stopped to concede the standard objection first: on a quality-adjusted basis, rent increases do track improvements in housing quality — "but that is little comfort for someone who cannot afford the high rent burden that follows." He did not skip the adjustment that weakens his figure; he stated it and explained why it does not settle the question.
The RBI published an overnight weighted average rate of 5.11% and, in the next column, the range it came from: 3.01% to 6.30%.
Why this is the interesting part and not housekeeping
Every one of those numbers could have been published bare and none of them would have been wrong. A regional GDP series without triangles is still correct. An affordability argument that omits hedonic adjustment is still arithmetically true. A weighted average without its range is a weighted average.
What the accompanying material does is tell you which question the number can answer. The revisions triangle converts "region A grew faster than region B" from a fact into a claim with a confidence attached. The quality-adjustment paragraph separates "rents rose because apartments improved" — true — from "therefore affordability did not worsen" — not implied. The 3.01–6.30 range makes clear that 5.11 describes the middle of a distribution in which somebody borrowed at twice the rate somebody else did.
This matters in a week when the same desk covered an ECB discussion paper on AI in financial markets, whose whole subject is extracting more signal from more data. The constraint on that project has never been the amount of data. It is that a number produced cheaply and at scale arrives without any of this — without a triangle, without the conceded objection, without the range. The expensive part of a statistic was never computing it.
The practical version
For anyone reading today's releases: open the revisions triangle before quoting a regional growth rate. Read Barr's concession before using his affordability point. Look at the RBI's range before treating 5.11% as the cost of overnight money.
And when a figure arrives with none of that attached, the right response is not to disbelieve it. It is to notice that you have been given the answer without being told the question, and that the two are not the same thing.