The Australian Bureau of Statistics published Average Weekly Earnings for May 2026 on 13 August, under the headline that earnings growth was the lowest since 2022. Inside the release, the agency names Queensland as the state with the largest six-month rise in full-time adult ordinary time earnings, at 2.52 per cent, and the Northern Territory as the runner-up at 2.15 per cent. In dollars, that is a difference of $8.40 a week between first place and second.
The same release carries the standard errors. On the Northern Territory’s May 2026 earnings level the ABS publishes $37.30; on Queensland’s, $19.80. These are the errors on the levels, not on the movement between them, and the ABS publishes no movement standard errors at all — which is part of the point. The only quantified statement of uncertainty available anywhere in the document is more than four times the gap the ranking rests on.
This is not a defect in the August release. Running the same comparison across every edition of the current series with complete state coverage — 28 releases, going back through the twice-yearly cadence to 2012 — the gap between the fastest and second-fastest state is smaller than the published standard error on the runner-up’s level in 24 of them. The exceptions are four releases in which one state moved far enough to clear the bar. Whatever the ranking is measuring, it is usually measuring it at a resolution the survey does not claim to have.
The choice of unit moves the order as well. The Australian Capital Territory’s full-time ordinary earnings rose $48.00 a week over the six months, more than the Northern Territory’s $41.80. The Northern Territory still ranks above it, by one hundredth of a percentage point, because its base is $1,942.30 against the ACT’s $2,245.30. That is definitional rather than an error — a percentage is a ratio and the denominators differ — but it means the second-placed jurisdiction in the release is not the one whose workers gained the most money.
The ABS is not silent about any of this. Its methodology page states plainly that "while AWE is not designed to produce movement in earnings data", users do so, and the release notes warn that changes in the composition of employment affect the average. The 13 August media release itself points readers to the Wage Price Index, which is the instrument built to measure wage growth and is unaffected by who happens to be employed in a given quarter. The caution is there, in the same document, in words. It is the numbers that are never brought together.
One figure in the release does clear its own uncertainty comfortably, and the ABS leads on a different one. The gender pay gap — the difference between male and female full-time adult ordinary earnings, as a share of male earnings — fell to 11.30 per cent in May 2026, the lowest in a series the agency has back-cast to November 1994 on the current basis. The four lowest readings in 64 half-years are all recent. But the move that set the record was 0.21 percentage points, worth about $4.50 a week on a male base of $2,181.00, and the ABS lists a sample redesign in May 2023 among its own significant changes to the survey, inside the window in which those four readings sit. The agency does not say the redesign created a break, and the record is arithmetically correct as published.
What follows from all this is narrow. The ABS has not published a wrong number, and every growth and record claim in the 13 August release checks out against its own tables. The gap is between what the survey is built to measure and what a ranking invites a reader to conclude from it — and the material needed to see that gap is printed in the release, a few tables apart, in two forms that are never multiplied together.