How CensusAtlas works

Where the data comes from

All figures are sourced from the Australian Bureau of Statistics (ABS) Census of Population and Housing, used under CC BY 4.0. We use the published ABS DataPacks: the 2011 Basic Community Profile (the “B” tables) and the 2016 and 2021 General Community Profile (the “G” tables). CensusAtlas is an independent project and is not affiliated with, nor endorsed by, the ABS.

Regions, and comparing over time

Statistical area boundaries are redrawn at each census, so 2011 and 2016 can’t be laid directly over the 2021 map. We re-aggregate the older years onto 2021 boundaries using the ABS ASGS correspondences: counts are apportioned to each 2021 area and summed, while medians and averages — which can’t be added up — are only carried across where one 2021 area accounts for most of the old one.

Table numbers also shift between censuses, so we match statistics by their full ABS description instead — a figure means the same thing in every year it appears. Where a definition genuinely changed, or the statistic is new in 2021, the page says not directly comparable and drops the change-since callout rather than imply a trend that isn’t real.

Shares, medians and suppression

A “share” — the share of people renting, say — is a numerator over a denominator drawn from the same Census table. Where that denominator is fewer than 20 people (or dwellings, or families) we withhold the share: the ABS adds small random adjustments to protect confidentiality, and on a tiny base those can swing a percentage wildly.

For medians and averages a reported zero almost always means “not applicable” rather than a genuine $0 or 0 years — median rent where nobody rents, for instance. We treat those as no data, so they never drag down a map’s colour scale or a ranking.

Other open datasets

Other open datasets describe what a place is like beyond the Census. Most are published under CC BY 4.0; the routing-derived measures (walk, drive and transit travel times) are built on OpenStreetMap, which is ODbL. Each is attributed below, aligned to 2021 ABS boundaries and carries its own vintage, so it is never confused with a Census count.

• Socio-Economic Indexes for Areas (SEIFA) — Australian Bureau of Statistics (2021), source, CC BY 4.0.
• Regional Population (Estimated Resident Population) — Australian Bureau of Statistics (2024), source, CC BY 4.0.
• Counts of Australian Businesses & Building Approvals — Australian Bureau of Statistics (2025), source, CC BY 4.0.
• Australian Schools List — Australian Curriculum, Assessment and Reporting Authority (ACARA) (2025), source, CC BY 4.0.
• SILO gridded climate data — Queensland Government (Long Paddock) (2019–2023), source, CC BY 4.0.
• Public transport timetables (GTFS) — State & territory transport agencies (2026), source, CC BY 4.0 (WA: custom PTA licence).
• Road & path network (OpenStreetMap, via Valhalla) — OpenStreetMap contributors (2026), source, ODbL 1.0.
• Transit accessibility (GTFS + OpenStreetMap, via R5) — State transport agencies & OpenStreetMap contributors (2026), source, ODbL 1.0 + CC BY.
• Recorded crime statistics — State & territory police services and crime statistics agencies (2025), source, CC BY 4.0 (WA: CC BY-NC 4.0).
• Walkability (OpenStreetMap, GTFS, ACARA & G-NAF, via Valhalla) — OpenStreetMap contributors, state transport agencies, ACARA & Geoscape Australia (2026), source, ODbL 1.0 + CC BY.
• Geocoded National Address File (G-NAF Core) — Geoscape Australia (via data.gov.au) (May 2026), source, CC BY 4.0 (Geoscape G-NAF end-user licence).
• SRTM 1-arc-second digital elevation model — NASA / U.S. Geological Survey (via AWS Terrain Tiles) (SRTM v3), source, Public domain — "SRTM data courtesy of the U.S. Geological Survey".

Climate is sampled from the SILO ~5 km gridded record at each area’s centroid, so it reads best as an estimate for larger areas (SA3/LGA) than a precise value for a small urban SA2.

Schools come from ACARA’s Australian Schools List. Access figures — schools serving an area, schools within 3 km — include schools just over the boundary, so they overlap neighbours and shouldn’t be summed. Average ICSEA is enrolment-weighted and describes the school community rather than local residents: context, not a ranking of school quality.

Public transport stop counts, weekly services, routes and mode mix come from the state and territory GTFS timetable feeds. Coverage follows the feeds, so a blank means no scheduled service reaches the area — different from a low score where service exists.

Recorded crime has no single national source at suburb or council level, so the figures are stitched together from each state and territory police service — NSW (BOCSAR), Victoria (Crime Statistics Agency), Queensland (Queensland Police Service), South Australia (SA Police) and the ACT (ACT Policing) — and harmonised to ANZSOC-aligned offence groups. The headline figure is the recorded-offence rate per 100,000 residents against each area’s 2021 Census population, so places with big daytime or visitor populations (CBDs, retail and industrial precincts) read high per resident, and rates on a very small population are withheld. Crucially, offence definitions and counting rules differ between states, so comparisons across state lines are indicative only and rankings default to within a single state. Coverage isn’t national: WA, the NT and Tasmania don’t publish sub-state data we can align to ABS areas, and drug offences are absent from the SA and ACT releases.

The map basemap — roads, water and place labels beneath the choropleth — comes from OpenStreetMap, rendered in our own muted vector style. OpenStreetMap data is ODbL and is used here as on-screen map context only.

Measures we derive ourselves

A few measures aren’t published by anyone, so we build them from open data to put every place on the same basis. Here’s what they capture; the models themselves are our own work and aren’t published.

Walkability is a 0–10 score for how much of daily life sits within a walk of the average home. It weighs access to everyday destinations — groceries, food and drink, public transport, schools, parks, health, culture and services — along the real pedestrian network rather than as a straight-line radius, so the things people use most often count for most. Amenities come from OpenStreetMap, schools from ACARA, stops and their service frequency from the GTFS feeds, and walking is routed on an elevation-aware network, so hills genuinely shorten what’s in reach. Every home in the area (a G-NAF address) is scored and the area takes the average — which keeps big rural areas honest, since the town scores what the town deserves and empty paddocks don’t vote.

Travel times and transit reach. Walking time to the nearest stop and driving time to the capital are routed over the OpenStreetMap network; people reachable by public transport, and transit time to the city, combine that network with the GTFS timetables for a weekday morning. Drive times are free-flow, with no live traffic.

Reading them. Derived scores compare within a level — suburb with suburb, council with council — and that’s how the site frames them. Where the underlying map data is too thin to trust, mostly in regional and remote areas, we withhold the score and say so rather than publish a false near-zero. Walking times make no allowance for footpath quality, crossings, shade or perceived safety.

Attribution. These measures are built on OpenStreetMap (ODbL, on-screen context only) routed with the open-source Valhalla and R5 engines, plus the GTFS feeds, ACARA schools, Geoscape G-NAF addresses and NASA/USGS SRTM elevation — each credited in the list above.

Known caveats

• Confidentiality perturbation. The ABS randomly adjusts small cells, so very small counts may not sum exactly to their totals.
• Non-response. Not everyone answers every question; “not stated” responses mean shares are of those who answered.
• Small areas are noisier. The smaller the area, the more a handful of people moves a percentage — prefer larger areas for headline comparisons.
• Crime isn’t comparable across states. Each police service uses its own offence definitions and counting rules, so a crime rate is only safely compared with other areas in the same state.
• Vintages differ. External datasets refresh on their own schedules against 2021 boundaries, so a brand-new estate can have residents before it has schools, shops or a mapped street network.

Start exploring