What the Census says about house prices

We tested every figure in the ABS Census against house prices: all 16,923 items published for each area in the 2011, 2016 and 2021 Censuses for every SA2 and council in Australia, and 187 measures aligned across six Censuses from 1991 for each state. We matched them with suburb sale prices from Victoria’s Valuer-General and South Australia, 30 years of Victorian council prices, state price indexes from 1986, and migration by country of birth (ABS Census DataPacks; Valuer-General Victoria; data.sa.gov.au). Every finding below was tested a second time: in another state, or in a later period the analysis had not seen. The models, every table and the full lists are on the method page; the maps are of every council, and of Melbourne and Adelaide by suburb.

New South Wales is not in the suburb results: its sales data is licensed for non-commercial use only. Queensland, Western Australia, Tasmania, the Northern Territory and the ACT publish no open suburb sales. Where we needed every council, we used the Census’s own housing costs, which track prices closely (below).

Where prices are high: the Census explains it

Of 1,580 Census measures for each SA2, 945 line up with the price level in Victoria and again, the same way, in South Australia; 84% of the relationships found in Victoria have the same sign in Adelaide. The five strongest, by rank correlation with the 2021 median house price (method):

The Census’s own housing costs say the same: an area’s median mortgage repayment ranks it by price at 0.90 across SA2s and 0.96 across Victorian councils (method). Over ten years, a council’s repayment growth tracks its price growth at 0.67; over five, not at all, because repayments follow interest rates and when loans were written.

Where prices will grow: it mostly does not

Measured against growth from 2016 to 2021, most Census measures look significant in Victoria, but in Adelaide only 45% point the same way: what a coin gives (method). Victoria’s slowest growth was in expensive inner and eastern suburbs, so whatever marks those suburbs looked like a cause of slow growth. In Adelaide the same measures went the other way.

Thirty years of Victorian council prices show why. Whether dearer councils grew faster or slower over the next five years has flipped three times since 1995 (Valuer-General Victoria):

Of 2,307 Census measures for Victoria’s 79 councils, 126 kept the same sign of relationship with growth across 2011–16, 2016–21 and 2021–25; about 577 would by chance (method).

States and the long run

Between states, comparing each year with the others, price growth moves with population growth (correlation 0.44), net interstate migration (0.43) and, more weakly, net overseas migration (0.23), from 1988 to 2025, each well beyond what shuffling the years gives (ABS population; ABS dwelling prices). Once population growth is known, overseas migration adds nothing: growth from another state and from another country goes with price growth alike.

Across six Censuses from 1991, of 181 state measures at the start of each five years, six were linked to the next five years’ price growth after allowing for testing so many; none was once the Northern Territory, whose price series is the least reliable, was left out. The clearest is catch-up: states with lower incomes grew faster next.

Country of birth

This is about where populations grow and settle, not about any group of people. Taking each of 50 countries of birth, migration from that country as a share of a state’s population had no link with the state’s price growth once total population growth was allowed for, in any version we ran (ABS Overseas Migration). By SA2, changes in six countries’ shares went with slower growth in 2016–21; none held in 2021–25, and four reversed. That was the pattern of student and inner-city areas through COVID, not a lasting relationship.

Tested on the past, and the next three years

We fitted each model on 2016 Census values and price growth from 2016 to 2021, then gave it the 2021 Census and compared its picks with what happened from 2021 to 2025, a period it never saw. Scores are for growth relative to the state’s average, since the states’ own averages swing with rates and the cycle (method).

The Census beat the state average and past growth, but the price level alone beat the Census: in both periods, areas dear for their state grew more slowly. No birthplace, ancestry, language or religion measure was allowed into any model. Thirty years of council prices show that this pull to the middle has reversed before, so treat it as the pattern of the last decade, not a law.

Illustrative only. Not a valuation, not a prediction to rely on, and not advice. We will score these against what happens.

For 2025 to 2028, by suburb, from the price level alone, the model that won the test (maps of Melbourne and Adelaide). Each figure is growth against its state’s average, with a typical range of −2.07 to +1.93 points a year around it. Highest in Melbourne: Melton (Vic.), Kurunjang, Melton South; lowest: Toorak, Brighton (Vic.), Hawthorn (Vic.). In Adelaide, highest: Elizabeth Downs, Elizabeth North, Smithfield (SA); lowest: Walkerville (SA), Vale Park, Medindie.

By state, a model of growth on population growth, fitted to 1988–2015, explained 21% of the differences between states in 2016–2025 (Centre for Population projections; method). On the projected population growth for 2025–26 to 2027–28, it gives:

Each state also carries its own long-run constant, which is why Queensland comes out low despite fast population growth. The model leaves four-fifths of the differences unexplained; read it as a direction, not a number.

What it means for a lender

• Value by peers, not by stories. The Census tells you what an area should cost; a valuation far from what its income, education and work suggest deserves a second look.
• Don’t buy growth on demographics. No Census measure reliably picked growth in another state or a later period.
• Watch population growth. It is the one driver that held across states and four decades, and it reaches prices whether people arrive from overseas or interstate.
• Use Census housing costs where there is no open price data, for levels and for ten-year change, not five.

Sources

1. ABS, Census DataPacks: Time Series Profiles 2021 (2011, 2016, 2021 on 2021 boundaries; SA2, LGA, state, Australia), 2006 and 2001 (1991–2006), 1996 Basic Community Profile and 2021 General Community Profile. CC BY 4.0.
2. ABS, ASGS Edition 3 allocation files and Mesh Block Counts 2021.
3. Valuer-General Victoria, property sales statistics: median house prices by suburb (2014–2025) and by council (1995–2025). CC BY 4.0.
4. Government of South Australia, Metro median house sales (2014–2026). CC BY 4.0.
5. ABS, Residential Property Price Indexes, Total Value of Dwellings and House Price Indexes (6416.0, from 1986).
6. ABS, National, state and territory population, Overseas Migration and Australia’s Population by Country of Birth.
7. Centre for Population, 2025 Population Statement, state projections. CC BY 3.0 AU.