APRIQ › How to read APRIQ
Every figure on this site carries labels — what kind of evidence it is, which geography it describes, when it became knowable, how much evidence sits behind it. This page is the key to those labels, in plain language. The method page holds the full register behind it: every source, licence and calculation, generated from the same database as the figures.
Two badges follow every figure around the site. They answer the first question worth asking of any statistic: who produced this number?
The number is exactly as its government publisher released it — a median the bond authority computed, a population the ABS estimated, a sale price the registry recorded. APRIQ did not touch the arithmetic.
Arithmetic on published figures — a yield from a published rent over a published sale price, a change between two published quarters, Tasmania’s medians from its published individual bond records. The inputs and the rule are always named beside the result.
What APRIQ never does: model a figure. No estimated sale price where none is published, no imputed rent, no score blending measurements into a number nobody measured. A famous exception proves the rule: the ABS’s population is itself an estimate — and it is shown as the ABS’s estimate, labelled “estimated residents”, never as a headcount.
A market dossier grades its own evidence along three named dimensions. They answer different questions, each keeps its own vocabulary, and no word appears in two of them — so two different readings on one page are two facts, never a contradiction. None of them measures whether a figure is right; that is a separate coherence check.
How much of the market’s picture was measured at this market’s own geography, rather than borrowed from a larger area.
| Broad | the demand evidence was published for this exact boundary and most components are present |
|---|---|
| Partial | derived through an official ABS correspondence — an estimate of this market assembled from measurements of others |
| Limited | the figures describe a larger area containing this market, or components are missing |
How many observations sit behind a specific rent or sale figure — a different question from whether the figure is right.
| Strong | 200+ observations | a figure resting on hundreds of real leases or sales |
|---|---|---|
| Moderate | 50–199 | reliable for direction; single buildings can still nudge it |
| Thin | under 50 | read it as a range, not a point — one unusual property moves it. A count the publisher withheld is unknown rather than small, and can never earn better than thin |
Rankings on this site admit only markets with at least 50 new leases behind the current figure, and a separate coherence check quarantines figures that cannot be right regardless of their count — volume is not accuracy, and neither substitutes for the other.
Which period a figure describes, how it sits against the newest quarter published anywhere, and whose clock dated it. Deliberately not a graded word: a label like “adequate” would hide exactly the dates worth checking. The freshness section below is the key to those dates.
A suburb is not a statistical area, a postcode is not a council, and a figure measured for one is not a figure about the other. Four badges say exactly how a number reached the page you are reading.
| Measured here | The publisher measured this figure for this exact market’s boundary. |
|---|---|
| Apportioned | Built from statistical areas using the ABS population-weighted correspondence — an estimate of this market assembled from measurements of others, rounded to say so. |
| Wider area | The figure describes an area containing this market — a suburb showing its statistical area’s population, named as such. |
| Not available here | Nothing can honestly be mapped to this boundary, and nothing is — rather than a neighbouring area’s number wearing this market’s name. |
The correspondence is used to link pages and label figures, never to compute one. Nothing on this site averages suburbs into a council or splits a council into suburbs and calls the result a measurement.
A quarterly median has a period it describes, a day it became knowable, and a day APRIQ retrieved it — and they can sit months apart.
| The period | What the figure describes: “June quarter 2026”, “30 June 2025”, “Census night, August 2021”. |
|---|---|
| Knowable | The earliest day anyone could have known it — the publisher’s release. Everything historical on this site is reconstructed against knowable dates, so nothing can quietly use information from the future. |
| Retrieved | The day APRIQ pulled the file, recorded with its content hash. A figure is never dated later than its own retrieval. |
Publishers run on different clocks. Victoria’s rent series is a moving annual median released on a slower cycle, so its newest figure sits behind the other states — the pages say so wherever it matters, because that is a fact about the publisher, not about this site’s freshness.
The most load-bearing rule on the site: what cannot be shown is said, with the reason, instead of being papered over.
The publisher suppressed a count — usually because it was too small to release safely. A withheld count is never treated as zero: it never satisfies a minimum-evidence filter and never sorts to the top or bottom of a ranking.
The publisher releases no such figure — Queensland’s sale prices, quartiles outside New South Wales, suburb rents in New South Wales. The page names which publisher, and why, and what evidence does exist instead.
Every market dossier ends its evidence section with the conclusions its evidence cannot support — no supply-adequacy verdicts without household formation, no price trends without a sale series, no forecasts before the gate. The list shrinks as gaps close; it is never hidden.
Publishers revise. Population estimates are restated for years; quarterly medians get corrected. APRIQ’s answer is the archive.
Every source file is stored exactly as retrieved, with a timestamp and a content hash, before anything reads it — and no archived file is ever overwritten. Re-pulling each quarter turns into revision detection: when a publisher restates a figure, APRIQ holds both editions and can show what a number said before it changed. Figures on the site are the most recently published edition, dated by the release that restated them.
Every forecast the method would have made, graded against what actually happened using only what was knowable at the time. The misses are as visible as the hits, on the scorecard, beside other forecasters graded by the same code.
No market page carries a prediction. A forward forecast ships only after the method has earned it in public, and will be a range with its own published record — never a single number, and never overwritten once issued.
Event-aware outlooks (“what would X mean for Y”) are a separate, later discipline with its own gates. Nothing on this site speculates about events.
Several common property phrases are banned here because they claim more than the evidence measures. What the site says instead is the measurement.
| APRIQ says | Instead of | Because |
|---|---|---|
| achieved rent | asking rent | bonds record what a tenant actually started paying; an advertisement records a hope |
| approved dwellings | counting approvals as finished housing | an approval is permission to build — some of it lapses, the rest finishes months or years later |
| observed sale median | the market median | a median of the sales that settled and were recorded — a sample the page counts, not the whole market |
| population growth relative to approvals | supply-adequacy verdicts | that claim needs household formation, which no publisher releases at this level |
Questions this page does not answer: the method page carries every source, licence, lag and calculation; Data use & access covers republication; and About covers who runs this. For anything else: data@apriq.com.au.
