Last Tuesday (September 9, 2025), the US Bureau of Labor Statistics published a news release – Preliminary benchmark revision for March payroll employment is -911,000 (-0.6%) – which told us that its employment estimates for the current year are likely to be significantly overstated. Given that the BLS has been under intense political scrutiny in recent months, with the US President recently sacking the Bureau’s head, I expect some noise from the conspiracy types to accompany this preliminary statement from the BLS. The fact is that when we undertake the adjustment process that the BLS deploys (explained below), the average monthly change in non-farm employment between March 2024 and March 2025 will turn out to be around half the current estimate – 71 thousand as opposed to 147 thousand per month. In other words, when the revisions are finalised in February 2026, the labour market will be assessed as having started slowing considerably in 2024 and continuing into 2025. I explain all this in the following discussion but emphasise that the process of revision is not some arbitrary act to make some politicians look bad. It is actually a process that upholds full transparency and is a regular activity that national statistical agencies undertake to make the data they publish as accurate as possible.