Why do population numbers change?

The two Census measures, usual residence and enumerated, overlap for about 95 per cent of people, who are counted at home anyway.

The Census is never a 100% accurate count of the population, and 2021 shows why in two different ways.

Around 96 per cent of households responded (96.1% to be precise). The ABS imputes the other four per cent based on other information about who is likely to live in a ‘non-responding dwelling’.

That imputation is itself a source of double counting. Some non-responding dwellings turn out to have been unoccupied at the time, with their usual occupants already counted somewhere else, at a holiday house, with family, or travelling.

How does the ABS make up for the known undercount on Census night?

About a month after Census night the ABS runs a Post-Enumeration Survey, a random sample of households used to work out the net effect of all this missing and double counting. It is not used to identify or fine anyone who missed the Census, only to measure the gap.

For 2021 the result was a net undercount of 0.7 per cent nationally, one of the most accurate Censuses in the world. Most states showed a net undercount. The ACT had a net overcount, meaning more people were counted twice there than were missed altogether.

Adding in the undercount, plus people known to be overseas on Census night from passenger card data, a group that was unusually small in 2021 given COVID travel restrictions, gives a Census-night ERP.

That figure is never actually published. Because ERPs are always dated to 30 June, the ABS then adjusts it back about five to six weeks, subtracting the births, adding back the deaths, and removing the net overseas migration that occurred in between, to land on the 30 June figure.

How does the ABS calculate the ERP between each Census?

Each year after that the process runs in reverse. The ABS takes the previous year's ERP and rolls it forward by:

• adding registered births
• subtracting deaths
• adding known overseas arrivals
• removing known departures, long-term and permanent only, so you are still counted if you are overseas short-term
• adding an estimate of internal migration, mainly from Medicare change-of-address records, now supplemented with ATO address data

The result is the annually revised ERP you see on the Informed Decisions sites. Each year also brings minor revisions to the previous year or two's figures, but none of it is final until after the next Census.

What will the 2026 Census do to my numbers?

In 2027, following the 2026 Census, there will be two versions of the population for the same date, 30 June 2026, at small-area level: one pre-Census, and one rebased on the 2026 Census. The timeline runs like this.

March 2027

The usual annual update, rolling forward the previous year's figures, themselves built by adding five years of change to the 2021 Census-based ERP. This figure does not reference the 2026 Census results at all, but is still built on the 2021 base, which can cause a lot of confusion as it comes out well after the Census was actually run. It is also the furthest point from a Census in the whole cycle, which makes it the least reliable figure in the series.

Around July 2027

After the main 2026 Census data release, which gives usual-residence counts rather than the official population (ERP), the ABS releases a preliminary rebased set: new 30 June 2026 figures based on the 2026 Census itself.

At the same time, the previous five years, back to 2022, are rebased as well. We update the Informed Decisions sites to match. Expect the Census counts (usual residence) and the ERP for the same year to differ noticeably when you first put them side by side, before the re-based numebrs are published by the ABS. That is normal.

About a year later, in 2028. The ABS makes final, minor tweaks to the 2022 to 2026 figures, which then become final and never change again. The first annual update for 2027 arrives around the same time, in March 2028, and the cycle starts over.

Our 2026 Census hub has the full release timeline.

Why does it have to be so complicated?

If that sounds convoluted, it is. But it produces the most reliable population figures (albiet in hindsight).

It also means a given year's ERP can keep changing for up to six years after the Census it is based on. The 2022 figure, for instance, can still move right through to 2028, though usually only by a small amount.

How much is my areas population count likely to move after the re-base?

Of course, we don’t know for sure until the numbers are out. The Census-based revision can be substantial, but it is worth knowing what normal looks like. At state level a rebase is usually within the tens of thousands, but rarely as much as 100,000.

The 2021 round was an unusually large one, distorted by COVID: Tasmania's population was revised up by 27,000 people, or 4.9 per cent of the state, which is an exceptional result rather than a typical one, while Victoria's was revised down by 101,000, or 1.5 per cent (Centre for Population, Rebasing Australia's population estimates using the 2021 Census).

That flows through to every local area estimate in those states, and some individual councils moved a long way. The important thing to note is these are revisions to previous estimates, not a sign of actual population change in each place.

Your area may buck the wider trend

The part that matters for you is that every level of geography gets revised, and they do not all move the same way. After the 2021 Census New South Wales was revised down by around 50,000, and yet Blacktown, one of its largest Local Government Areas, was revised up. Your state's headline number tells you very little about what will happen to your own figure. Do not read the state result and assume yours follows it.

This chart from Community Profile shows the City of Maribyrnong in Melbourne, and 20 years of population estimates. It shows the clear dip when populations declined in Victoria during COVID, a dip that wasn't picked up until after the 2021 Census. On this chart, the numbers for 2022 to 2025 are still subject to change after the 2026 Census release.

So what should I do with each release?

Over the next year, two different population figures will be released for your area as at 30 June 2026. The first is based on the 2021 Census count, plus adjustments. The second will be based on the 2026 Census count (also with adjustments).

The two numbers will likely not align. This can cause some confusion within your organisation, espeically if you jump to publishing and promoting documents based on the first release. Here’s how we suggest you approach it:

March 2027, the annual update

Treat it as a first cut. It is a sound figure for routine reporting and for tracking year on year, but it is the furthest point from a Census in the whole cycle. If it is going into a strategy or a funding submission, name the release you used and note that a rebase is coming (you can link to this article if you need an explainer).

Around July 2027, the preliminary rebased estimates

This is the one to plan around. It resets 2022 to 2026 in one go, so your recent trend can change shape as well as the current population count.

Once the rebased figures are out, you can refresh the baseline figures in your strategies, your monitoring frameworks and anything you report to councillors or other stakeholders.

2028, the final figures

These never change again. Use them for evaluation, for anything that needs a stable series, and for the long-run comparisons you do not want to have to redo.

How does this affect my population forecasts?

Informed Decisions Population Forecasts always match the population at the base Census year, currently 2021, within small rounding adjustments.

Forecasts are completed progressively around the country, so each one will generally match the latest ERP that was current when it was completed.

Because of the small annual revisions to the ERP, there can be differences between a forecast and the current ERP simply because the ERP has moved since the forecast was published, or because the underlying assumptions in the forecast have changed as conditions have changed or new information has become available, produce different population outcomes.

In some cases we cannot match the interim ERP figures, but often find we have matched the Census-based revision better than the preliminary numbers once it comes through. None of these small differences change how useful the forecasts are for planning.

The best source of truth for the current population count? The community profile for your area.

If you would rather not track all of this, you do not have to. We update the Informed Decisions sites as each release becomes available, including the full rebase, so the population figure on your Community Profile is always the current official one.