Tradergy
GUIDES

How the trimmed mean and weighted median are built

Australia's underlying inflation measures do not exclude a list of volatile categories. They exclude positions in a distribution, recalculated every period, which is why nothing is permanently in or out of them.

Reviewed by the Tradergy editorial team · September 2026
In short

The trimmed mean is a weighted average of the middle 70 per cent of the distribution of price changes: the Bureau of Statistics orders every expenditure class by how much its price moved, weights them by household spending, and excludes those "with a cumulative total below 15 per cent and above 85 per cent" of that distribution. The weighted median takes the single expenditure class at the 50th percentile instead. Because both select by position rather than by category, what is trimmed changes every period.

Two inflation figures come out of the same release. One is reported as the inflation rate. The other is the one policy is set against, and it is built by throwing away part of the first. Which part, and on what rule, is less widely understood than the fact that it happens.

Key Takeaways

  • The trimmed mean is a weighted average of the middle 70 per cent of the distribution of price changes, trimming below the 15th and above the 85th percentile.
  • The weighted median takes the expenditure class at the 50th percentile instead of averaging a range.
  • Membership of the trimmed set changes every period, because it is a position in a distribution rather than a list of categories.
  • The measures were developed by the Reserve Bank; the Bureau of Statistics took over publication in 2007.
  • Both are built on seasonally adjusted movements, so both are revised as new periods arrive.

The problem underlying measures were built to solve

A price index is an average of many separate price movements, and in any given period a few of those movements are enormous. A cyclone takes out a banana crop. A fuel excise change lands in one quarter. A government rebate arrives and then expires. Each of these is a real price change that households genuinely paid, and each tells you almost nothing about whether prices in general are accelerating.

That distinction matters because of what the figure is used for. Monetary policy operates with a long lag, so it has to be set against the part of inflation that is likely to persist, not the part that will reverse on its own before any decision could take effect. An underlying measure is an attempt to separate the two using the shape of the distribution rather than a judgement about which items to exclude.

The alternative approach, used elsewhere, is to name the volatile categories in advance and publish an index without them. That has the advantage of being simple to explain and the disadvantage of being a permanent decision: an excluded category stays excluded in the quarters when it is behaving perfectly normally, and a category never on the list stays in even when it moves violently.

How the trim is actually taken

The trimmed mean does not exclude named categories. It excludes positions in a distribution.

Every expenditure class in the basket produces a price movement for the period. Order those movements from the largest fall to the largest rise, and weight them by how much households spend on each. The Bureau of Statistics then takes "a weighted average of percentage change from the middle 70 per cent of the distribution", so that expenditure classes "with a cumulative total below 15 per cent and above 85 per cent are excluded (trimmed) from the calculation".

The consequence worth holding on to is that membership of the trimmed set changes every period. An item is not on a list. It is at a position, and its position is decided by how it moved this time relative to everything else. A category that has been trimmed for several periods because it was moving sharply will re-enter the calculation the moment the rest of the basket catches up with it, and a normally sedate category will be trimmed out in the period it happens to jump.

The weighted median does something similar with a blunter instrument. Rather than averaging the middle range, it takes the single expenditure class sitting at the 50th percentile of the same distribution. Where the trimmed mean averages away the extremes, the median simply steps over them.

Where the measures came from, and why they revise

Neither measure originated with the statistical agency that now publishes them. The Reserve Bank developed the trimmed mean and the weighted median, and the Bureau of Statistics assumed responsibility for calculating and publishing the series in 2007. That history explains the shape of them: they were built by the institution that needed a persistence-focused number to set policy against, and were later adopted into the official release rather than designed as part of it.

It also explains an inheritance that surprises people. Both are calculated from seasonally adjusted movements rather than from the raw index, and the adjusted series is rebuilt as new periods arrive - so the underlying measures are revised as a matter of routine, while the headline index they are derived from is not. The mechanics of that asymmetry, and what it means for a figure quoted from an old release, are set out on the inflation rate page.

What this means for reading a series

Three practical consequences follow, and none of them is that the underlying measures are unreliable.

The first is that a quoted underlying figure carries a date in a way the headline does not. A trimmed mean for a past period, quoted from a release at the time, is not necessarily the trimmed mean now recorded for that period. If two sources disagree about a historical underlying figure, the likely explanation is that they were published at different times rather than that one of them is wrong.

The second is that the gap between headline and underlying is information rather than error. A wide gap means the distribution had large tails in that period: something specific moved a long way. A narrow gap means the movement was broad. Neither is inherently the truer number, because they answer different questions.

The third is that trimming has a genuine cost for a household. An expenditure class trimmed out of the calculation is still a bill somebody paid. The measure is deliberately built to be less representative of any particular period's experience, in exchange for being more representative of the trend, and that trade is the whole point of it rather than a flaw in it.

Where to look next

The current published figure and its history sit on the inflation rate page, the consumer price index entry defines the underlying measure, and what a comparison in real dollars can and cannot tell you covers the separate question of how the index is built and re-weighted across long periods. The cash rate is the policy rate these measures are read for.

Sources

Not financial advice, and not a recommendation of any product or institution. Reviewed by the Tradergy editorial team, September 2026.

Common questions

What is the trimmed mean, in one sentence?

A weighted average of the middle 70 per cent of the distribution of price changes. The Bureau of Statistics describes it as excluding expenditure classes "with a cumulative total below 15 per cent and above 85 per cent" of that distribution.

How is the weighted median different?

It takes a single point rather than an average of a range. The Bureau defines it as the seasonally adjusted quarterly percentage change of the expenditure class at, or immediately over, the 50th percentile of the distribution.

Why trim by position rather than by naming volatile categories?

Because a named list is a permanent decision. An excluded category stays excluded in the periods when it is behaving normally, and a category never on the list stays in even when it moves violently. Trimming by position lets membership follow the data each period instead.

Does trimming mean the excluded prices do not count?

Not permanently. The trim is applied to each period's distribution separately, so an expenditure class is excluded from one period's calculation rather than removed from the index. A category that keeps moving sharply in the same direction stops being an outlier and re-enters the middle of the distribution.

Who created these measures?

The Reserve Bank developed the trimmed mean and weighted median series. The Bureau of Statistics assumed responsibility for calculating and publishing them in 2007.