Correspondence (ABS correspondence file)
A table published by the ABS for converting data from one set of geographic areas to another, such as 2021 SA2s to 2026 SA2s. Each row gives the share of a 'from' area that falls in a 'to' area, weighted by where the population lives.
What a correspondence contains
The ABS publishes correspondences with each edition of the ASGS so that data coded to one set of areas can be converted to another. It describes them as files to mathematically convert statistical data between geographic regions, for use when more accurate methods, such as geocoding addresses or point-in-polygon allocation, aren't available.
Each file is a CSV with one row for each pair of overlapping areas. A row gives the code and name of the 'from' area and the 'to' area, and a ratio between 0 and 1 (RATIO_FROM_TO) recording the share of the from area allocated to the to area. To convert a count, multiply it by the ratio on each row and add the results for each to area. File names describe the conversion, so CG_SA2_2021_SA2_2026.csv converts 2021 SA2s to 2026 SA2s.
Population-weighted and area-weighted
The ABS's standard correspondences, whose file names start CG, are population weighted, with weights calculated from population data modelled to residential address locations, so each from area's share goes where its residents live. Area-weighted correspondences (CA) split a from area by land area, which assumes the data is spread evenly across it. Among the Edition 3 files, only the Australian Drainage Divisions correspondence is area weighted.
The ABS's 2012 information paper on converting data shows why the weighting matters. Where one SA2 holds nursing homes and its neighbour is an industrial estate, a population-weighted correspondence allocates the whole population to the SA2 with the nursing homes, which is more accurate than dividing it by area. The paper recommends area weighting for data that tends to be uniform, such as agricultural or environmental data. Area weighting covers these methods in more detail.
Quality indicators and the ABS's cautions
Each row carries a quality indicator for its to area, and each file carries an overall indicator, both rated Good, Acceptable or Poor. The ABS expects a Good correspondence to convert data to a high degree of accuracy and an Acceptable one to a reasonable degree of accuracy, with caution because quality will vary. Poor means there is a high likelihood the correspondence will not convert data accurately, so the results may not reflect the actual characteristics of the areas involved. The ABS publishes no numeric thresholds for the grades, and it doesn't release correspondences it judges not fit for purpose.
The indicator measures quality against the weighting unit, so it can't account for the data being converted. Data distributed like the weighting unit converts best. In the ABS's 2012 paper, address counts converted with a dwelling-weighted correspondence matched closely, while deaths data converted with the same correspondence did less well.
- Rows below a minimum output size, typically a ratio under 0.01, are flagged as statistically insignificant in the BMOS_NULL_FLAG column.
- Null rows mark areas with no counterpart, such as islands included in one set of boundaries and left out of the other.
- Converting small areas into larger ones generally gives more accurate results than converting between areas of similar size or from larger areas to smaller ones.
A correspondence converts counts, so a median or a percentage has to be rebuilt from converted counts, and converting an earlier Census onto newer boundaries this way produces a carried figure.
Where to get them
Current files are on the ABS's ASGS correspondences pages, grouped into Main Structure, Indigenous Structure and Non ABS Structures. Edition 3 to Edition 4 files, including mesh blocks, SA1s and SA2s, Suburbs and Localities, Postal Areas and Local Government Areas from 2021 to 2026, are being released progressively from July 2026 as each Edition 4 structure is published. Edition 3 files, mostly converting 2016 areas to 2021 areas and successive council and electoral updates, stay on the Edition 3 pages, and older ASGC and ASGS correspondences are on data.gov.au.
Correspondences and Censible
Censible carries the 2011 and 2016 Censuses onto the 2026 boundaries with the ABS's population-weighted census-to-census correspondences, chained from 2011 to 2016, 2016 to 2021 and 2021 to 2026. SA1s, councils, postcodes and suburbs are each carried with their own correspondence, and SA2s, SA3s, SA4s and city regions are then added up from SA1s, so they're consistent with the SA1s inside them.
Where a 2026 area is exactly one earlier area, its earlier figures are carried as counted, and anywhere else they're apportioned, marked as estimated in the app and in API responses, and described in the carried figure entry.
Sources: ABS, ASGS Edition 4, Correspondences; ABS, Information Paper: Converting Data to the ASGS, 2012 (Geographic Correspondences); ABS, Information Paper: Converting Data to the ASGS, 2012 (Quality Indicator).
Carried figure
A count from an earlier Census moved onto newer boundaries with the ABS's population-weighted correspondences, so places can be compared across Censuses on the same map. Where a boundary didn't change, the carried figure is the ABS's own count; where it did, it's an estimate.
Australian Statistical Geography Standard (ASGS)
The ABS's framework of boundaries for publishing statistics, from mesh blocks of a few dozen dwellings up to states and territories, plus the ABS's versions of councils, suburbs and postcodes. It's redrawn every five years; Edition 4, released from July 2026, is the one the 2026 Census will use.
Area weighting (apportioning)
A way to estimate figures for an area that doesn't match published boundaries, by giving it a share of each overlapping area's counts. The share can follow land area, population or dwellings, and each assumes the figure is spread evenly within what it measures.
Statistical Area Level 2 (SA2)
A medium-sized area defined by the ABS to represent a community that interacts socially and economically. Most have between 3,000 and 25,000 residents.
Median
The middle value when everything is put in order: half the values are above it and half below. Census figures for age, income, rent and mortgage repayments are published as medians because a few very large values would distort an average.
Questions people ask
A CSV published by the ABS for converting data from one set of geographic areas to another, such as 2021 SA2s to 2026 SA2s. Each row gives the ratio of a 'from' area allocated to a 'to' area, and multiplying counts by those ratios converts them.
A population-weighted correspondence (file names starting CG) splits each area by where its residents live, using population modelled to residential addresses. An area-weighted one (CA) splits it by land area, which assumes the data is spread evenly and suits uniform data such as agricultural or environmental figures.
The ABS expects a Good correspondence to convert data to a high degree of accuracy and an Acceptable one to a reasonable degree, with caution. Poor means there is a high likelihood the conversion is inaccurate. The ABS publishes no numeric thresholds for the grades.
From the correspondences page of each ASGS edition on the ABS website, where Edition 3 to Edition 4 files are being added progressively from July 2026. Older ASGC and ASGS correspondences are on data.gov.au.
Only indirectly. Convert the underlying counts, such as households in each income range or the numerator and denominator of a percentage, then work out the median or percentage for each new area from the converted counts.