How to check the source of a property statistic

Write the answers before you quote the number.

1. Who produced it? A government agency, a registry, a publisher, a portal, an agent, a model. Each has a different relationship to the record.
2. From what? Recorded transactions, listings, a survey, a model, another publisher. A median of recorded sales and a median of asking prices share a name and not an object. See ask versus sale.
3. What period and boundary? Twelve months or three. Suburb, postcode or statistical area. Comparisons need both to match.
4. How many observations? The sample size decides whether the figure is a summary or a coincidence. See medians.
5. When? Observation date, not only publication date. A planning layer retrieved two years ago may miss a rezoning. A Census figure describes Census night.

Brunswick (Vic.) SAL20361, Census night 2021: 1,594 unoccupied private dwellings, 12.6 per cent of private dwellings. That is an official count of dwellings empty on the night for any reason: holiday houses, dwellings between occupants, vacant rentals, and others.

A later rental vacancy rate for Melbourne, published from bonds or from a listings sample, is also a real statistic if its publisher states the denominator and the window. It is not the Census unoccupied share. Putting 12.6 per cent next to a 1.4 per cent vacancy rate and calling the gap “tightening” is a definition error. Both numbers can be correct. The row is unmatched. See listings versus vacancy. A second pair: a house median and an all-dwellings median for the same suburb and quarter. Both can be right. They still do not compare.

Precision without a sample. A growth forecast with no method. A suburb score with no inputs. An “estimated value” with no range. Identical figures across sources because one copied the other. A listing claim labelled as a planning fact.

These mistakes judge the number by whether it looks reasonable.

• Reading the publication date as the observation date.
• Treating two sources agreeing as confirmation when one copied the other.
• Comparing an asking-price median with a recorded-sale median.
• Accepting “the suburb” as the boundary when SAL, SA2 and postcode differ.
• Using Census unoccupied dwellings as a rental vacancy rate.

It tells you what Districts checked and when. It does not certify that the source keyed the sale correctly. Districts cannot audit a third-party forecast or suburb rating it does not hold, beyond showing it as that publisher’s figure with its date. Planning and statistical geographies also differ by state and collection. The five questions stay yours.

Every figure carries a source line: label (source data, calculation, derived), source, period, date checked, and a methodology link. Suburb medians show the sales count. Unknown figures show why. Where the five questions cannot be answered, the number is not shown. See facts, calculations and unknown.

Follow the source line to the originating record. For a third-party figure, ask the publisher the five questions. A publisher who cannot answer them is telling you something about the figure. Planning layers, Census tables and sale medians also update on different clocks. A figure that is “current” on a blog may be a 2021 Census row with a 2026 publish date. Write both dates before you quote the number.