Method note

How we will measure fuel excise pass-through

Fuel excise returned to the full rate of 53.7 c/L on 3 August 2026, a rise of 17.1 c/L, or 18.81 c/L at the pump once GST is applied. We are measuring how much of that step actually appeared in retail petrol and diesel prices, in each state and territory. This note sets out exactly how, and it is published before the measurement is taken.

Published 7 August 2026. Run 16 August 2026: no result published. Re-running 24 August 2026.

The reason for the gap is the point of the note. A method described after its own result has been seen can always be the method that produced the most interesting number. This one was fixed in public first, and it was run on the date it said it would be run.

It returned nothing we are willing to publish. The measurement was taken on the evening of 16 August 2026, exactly as section 01 describes it, and it met the no-publish condition in section 05. What that means in practice is set out below. The short version: retail prices across every state and both fuels were already climbing steeply before the excise changed, so the before-and-after comparison cannot separate the tax from the trend, and the two baseline lengths this note commits to disagree with each other about every single state. We had said in advance that we would publish nothing in that case, so we have published nothing. The measurement runs again on 24 August 2026, with another price cycle behind it.

01FuelCrisis-derived

The comparison windows

A 14 day baseline ending the day before the step, against every day of data from the step onwards. Both windows are reduced to medians, not means.

Baseline: the 14 days to 2 August 2026. Long enough to average over most of a discounting cycle in the eastern capitals, short enough to stay clear of the methodology seam described in section 03.
Post-step: 3 August 2026 to the last day of data. The measurement of record will be run on the evening of Sunday 16 August 2026, giving 14 days after the step. We will not publish a shorter post-step window as a result.
Sensitivity: the same measurement at 20 days. 20 days is the longest baseline available to us, because a longer one crosses the seam. We will run both, and if the two baselines disagree materially about any state we will publish both figures for that state rather than choosing the more striking one.
Overnight: 2 August 2026 against 3 August 2026. Reported separately for every state. It is the least cycle-sensitive read available for a tax that applies at a stroke, and it is also a single pair of days, so one bad feed day moves it. It will always be published beside the window figure, never instead of it.

02Official

The data

Per-state daily medians, from state government retail price feeds only. Wholesale prices are excluded, and any state without retail data is named rather than dropped.

Retail only. The state and territory government fuel price feeds, the same sources listed on our methodology page. New South Wales FuelCheck also serves the Australian Capital Territory and Tasmania.
Wholesale is excluded, deliberately. We also hold wholesale terminal gate prices, and roughly half of all state, fuel and date combinations have both. Terminal gate prices are excise-inclusive and move with the tax essentially on the day. Blending them into a retail pass-through measurement would push the answer towards 100 per cent by construction, which is a result about arithmetic rather than about the market.
Medians, at both levels.Every price recorded in a state on a day is reduced to that day's median; each window is then the median of its days. Medians, not means, because a handful of very high or very low sites should not move a state figure.
Nothing drops out quietly. Any state and fuel combination with no retail rows in the window is named in the published output. The same applies if a new price source appears in our data that the measurement does not recognise: it stops rather than guessing whether the source is retail or wholesale.

03FuelCrisis-derived

Three things that could make this wrong

Named in advance, with what is done about each. The third one is not solved, and will not be presented as if it were.

1. Our own methodology seam, 14 July 2026
Our retail figures before that date are mean-based, and from that date they are median-based. A baseline window reaching back across it would compare two different statistics and manufacture a step that is pure methodology. The measurement refuses to run if the requested baseline starts before the seam. That is why 20 days is the longest baseline available, and why the sensitivity check in section 01 stops there.
2. Wholesale prices posing as retail
Handled by the retail-only rule in section 02. The measurement uses an explicit list of retail sources and stops on any source it does not recognise, so a new feed cannot slip into the answer without a person deciding which side of the line it sits on.
3. Retail discounting cycles, which are not solved
Australian capital cities run discounting cycles: roughly weekly in Perth, three to five weeks in the eastern capitals. A step of about 18.8 c/L sits well inside the amplitude of those cycles, so a before-and-after difference can be swamped by cycle position. We do not have a fix for this, and we will not write as though we do. What we will publish instead, for every state: the baseline minimum, maximum and standard deviation next to the observed change, so the noise is visible beside the signal; and a flag on any state whose move does not clear two standard deviations of its own baseline variation.

04Editorial

What will be claimed, and what will not

An observed price change and an estimate of pass-through are two different things. They will be presented as two different things, in separate columns, with separate language.

Claims that will and will not be made when the pass-through measurement is published.
We will say	We will not say
Median retail prices in a named state rose or fell by a stated number of cents per litre between two stated date ranges.	That retailers raised prices by that amount. The measurement observes prices, not decisions.
That movement was a stated share of the 18.81 c/L pump-side step, described as an estimate, with the other things that moved reported alongside it.	That a stated percentage of the excise rise was passed on, as a settled fact. Nothing here controls for crude oil, the exchange rate or cycle position.
Where a state was cycling harder than the tax moved, that its figure carries little information on its own.	Any characterisation of retailer conduct or motive. The words gouging and profiteering will not appear, because this measurement cannot support them.
Which states and fuels were measured, and which were excluded and why.	A single national pass-through percentage presented as the headline. Eight state figures summarised eight ways is not a national retail average and will not be quoted as one.

One further limit, stated in advance because it is the first thing an informed reader will raise: excise changes at the terminal gate on a fixed date and historically reaches the board in days. A crude oil movement takes four to six weeks to reach the pump. Those two lags are not interchangeable, and a Brent move inside the measurement window has largely not arrived yet.

05Editorial

When we will publish nothing

Committed to in advance, so that a disappointing measurement cannot quietly become an interesting one.

We will not publish a result if the other things that moved over the window can account for a large share of the observed change, or if no sentence survives the claims rule in section 04. In that case nothing goes out, we say so to anyone who has asked, and the measurement is re-run a week later with one more price cycle behind it. A measurement that finds little is a legitimate outcome of this method, and it is the outcome this note exists to keep available.

16 August 2026: this condition was met, and nothing was published.

The measurement was run on the evening of 16 August 2026 as described, on 14 days of data either side of the step. Three things went wrong with it, and any one of them would have been enough on its own.

Prices were already climbing before the tax moved. Every state and both fuels rose between 1.1 and 2.7 cents a litre per day through the baseline window, which is 11 to 33 cents over the fortnight before the excise changed at all. A before-and-after comparison built on a baseline that is itself a steep ramp measures the ramp as well as the tax, and cannot say how much of the answer is which.
Something large was moving underneath it. Brent crude rose 6.27 US dollars a barrel over the same window, worth roughly 6.1 cents a litre at the pump once the exchange rate and GST are applied, against a tax step of 18.81 cents. That is about a third of the step, and it is the confounder this section was written for.
The two baselines disagreed about every state. Section 01 commits to running 14 days and 20 days and publishing both where they differ. They differ everywhere, by an average of about 27 percentage points, and on the longer baseline 13 of the 16 state-and-fuel series come out above 100 per cent, which is not a pass-through reading at all. Publishing both figures was written for a handful of disagreeing states. When every series disagrees and one whole window is uninterpretable, printing both is not disclosure, it is printing a number we know is wrong beside one we cannot defend.

6 of the 16 series also fall below the two standard deviation threshold in section 03, meaning those states were cycling harder than the tax moved. The measurement runs again on 24 August 2026 with another price cycle behind it. If it is still this unstable then, the instability is the finding and we will publish that instead.

If anything in this note changes before publication, the change will be added below with its own date, and the original text will stay readable. Errors found after publication are handled under our corrections policy.

06Editorial

Changes to this note

Dated entries only. An empty list means the method published on 7 August 2026 is the method that was used.

No changes. Published 7 August 2026, and run unaltered on 16 August 2026. The outcome of that run is recorded in section 05, where the condition it met is stated. Recording it there rather than here is deliberate: the method did not change, it simply returned a result we had committed in advance not to publish.

Source: ATO excise duty rates; state government retail fuel price feedsThe excise rate and how it is set How we measure this

The excise figures on this page are the same constants the rest of the site renders from: 36.6 c/L to 53.7 c/L is a rise of 17.1 c/L, which is 18.81 c/L at the pump once GST is applied.

How we will measure fuel excise pass-through

Published 7 August 2026. Run 16 August 2026: no result published. Re-running 24 August 2026.

We will say

We will not say

Median retail prices in a named state rose or fell by a stated number of cents per litre between two stated date ranges.

That retailers raised prices by that amount. The measurement observes prices, not decisions.

That movement was a stated share of the 18.81 c/L pump-side step, described as an estimate, with the other things that moved reported alongside it.

That a stated percentage of the excise rise was passed on, as a settled fact. Nothing here controls for crude oil, the exchange rate or cycle position.

Where a state was cycling harder than the tax moved, that its figure carries little information on its own.

Any characterisation of retailer conduct or motive. The words gouging and profiteering will not appear, because this measurement cannot support them.

Which states and fuels were measured, and which were excluded and why.

A single national pass-through percentage presented as the headline. Eight state figures summarised eight ways is not a national retail average and will not be quoted as one.