Methodology

Overview

SALM presents estimates based on the Structure Preserving Estimation (SPREE) methodology.

The purpose of SPREE is to produce small area unemployment, unemployment rate and labour force estimates that reflect the regional disparities of the Centrelink data, while being consistent with Australian Bureau of Statistics (ABS) Labour Force Survey (LFS) estimates.

Given the level of disaggregation involved, the small area figures produced by SPREE are smoothed (i.e., averaged) over four quarters, to dampen the variability inherent in the small area estimates.

Source data

Four primary data sources are used to produce the SA2 and LGA estimates in SALM:

1. Current recipients of JobSeeker Payment and Youth Allowance (other) who are not on a zero rate of payment, by SA2;
2. ABS LFS data by Statistical Area Level 4 (SA4). The LFS samples private dwellings (approximately 24,000 dwellings) across Australia, resulting in a sample of around 50,000 people. More details about the methodology underpinning this survey are included in the ABS publication, Labour Force, Australia;
3. Participation rate data at the SA2 level from the Census of Population and Housing (see the section on Applying Census Benchmarks below for detail); and
4. The latest Estimated Resident Population (ERP) estimates for persons aged 15 and over by sex by SA2 from the ABS release Regional population by age and sex.

How SALM estimates are produced

Estimating unemployment

SALM produces unemployment estimates using an iterative process.

• Step one apportions ABS unemployment at the SA4 level across each of the SA2s within that region based on the distribution of Centrelink JobSeeker Payment and Youth Allowance (other) beneficiaries at the SA2 level by sex.
• Step two benchmarks the estimates produced in step one to ABS unemployment estimates by age, sex and marital status at the Greater Capital City Statistical Area (GCCSA) level.

These steps are repeated until the SALM unemployment estimates for the SA2s in each SA4 match the published ABS unemployment figures, and the age/sex/marital status groups in each GCCSA, for those regions and groups.

Estimating the labour force

Labour force size estimates are produced by:

• taking the participation rate for each SA2 from the Census (in instances where an SA2 does not have a participation rate, the participation rate for the Statistical Area Level 3 that the SA2 is within is used as a proxy);
• applying this to the latest available ERP data for persons aged 15 and over for the SA2, to produce a labour force weighting for the SA2 within the SA4; and then
• allocating the total labour force for each SA4 to the SA2s within that region according to the labour force weighting.

By using Census participation rates and the latest available ERP, the SA2 labour force estimates reflect any changes in the distribution of population within each SA4 that have occurred since the last Census.

Please note that the ABS generally only publishes up-to-date ERP estimates for the latest (2021) edition of the Australian Statistical Geography Standard (ASGS). Where ERP data for an earlier ASGS are needed but not published on these boundaries by the ABS (for example, 2022 ERP on the 2016 ASGS for SALM estimates prior to the September quarter 2023) a population-weighted correspondence is used to produce the required figures on the appropriate ASGS.

Calculating the unemployment rate

Consistent with ABS methodology, unemployment rate estimates are produced by calculating the level of unemployment as a proportion of the labour force.

Applying Census benchmarks

To smooth out the impact of changes in SA2 participation rates between each Census of Population and Housing (conducted every 5 years), SALM incrementally phases in the Census benchmarks between each Census. The Census benchmarks for each of the 19 quarters between two Censuses (which occur every five years, or 20 quarters) are a weighted average of both Census results, based on how far the quarter is between the Censuses. Estimates from the September quarter 2021 onwards will use 2021 Census benchmarks until 2026 Census benchmarks become available.

• For example, the benchmarks for the unsmoothed estimates (the Census benchmarks are applied before the smoothing process) for the March quarter 2017 are the sum of 18/20ths of the 2016 Census benchmark and 2/20ths of the 2021 Census benchmark for each SA2. This is because the March quarter 2017 is two quarters away from the 2016 Census (the 2016 Census took place in the September quarter 2016) and 18 quarters away from the 2021 Census (the 2021 Census took place in the September quarter 2021).

Producing LGA estimates

LGA estimates are produced by apportioning SA2 unemployment and labour force estimates using the latest available SA2 to LGA correspondence from the ABS.

Employment should not be derived from SALM

The SALM data are synthetic estimates based on ABS unemployment and Centrelink JobSeeker Payment and Youth Allowance (other) beneficiary numbers and labour force data from the LFS and the Census. The production of SALM does not involve the use of any data that enables reliable modelling of employment at the small area level. Nor does SALM attempt to estimate employment for either SA2’s or LGA’s. Accordingly, employment estimates should not be derived from SALM statistics. For more information, please email the SALM inbox at SALM@jobsandskills.gov.au.

Greater disaggregations not available

Due to the significantly more pronounced variability of the data disaggregated below the SA2 or LGA level, it is not possible to derive reliable estimates for smaller population groups.

Accordingly, it is also not possible to estimate reliable unemployment and unemployment rate estimates for particular groups, such as sex or age cohorts, within an SA2 or LGA.

Estimates are not available for all SA2s and LGAs

While most SA2s and LGAs have a complete time series of data, with smoothed SALM estimates available from the December quarter 2010 onwards, some regions have a shorter time series, while others have no estimates at all.

There are 3 reasons why data are not available for every SA2 and LGA:

• A ‘break in series’: some regions have a break in series, such that only a partial time series is available. Most of these series breaks were the result of SALM transitioning to a new edition of the ASGS, (see Geography, below). SALM moved from the 2011 to the 2016 ASGS in the June quarter 2019, and from the 2016 to the 2021 ASGS in the September quarter 2023. As smoothed estimates require four quarters of data, the first smoothed estimates after a break in series are available three quarters after the introduction of the new edition of the ASGS.
  
  • For example, for SA2s broken between the 2016 and 2021 ASGS, the earliest smoothed estimates are for the June quarter 2024 (three quarters after the September quarter 2023). The only available estimates between the September quarter 2023 and the March quarter 2024, inclusive, for these SA2s and LGAs are from the unsmoothed series (available below).
• The SA2 is considered too small: there are around 120 SA2s which JSA considers have too small a population to produce estimates of reasonable quality. As the reliability of estimates at the SA2 level are related to the size of the SA2, estimates are only published for SA2s that have an established labour force of 100 or more as at the September quarter 2023.
• Other factors: while the underlying methodology used to produce the small area estimates in SALM is robust for the vast majority of areas, in a small number of cases it can result in figures that do not accurately reflect labour market conditions within the region. This is why no estimates have been published for the SA2 and LGA of Aurukun since the June quarter 2017.

Impact of corresponded ERP data on the SA2 estimates

The ABS is no longer producing ERP data on earlier editions of the ASGS. As SALM SA2 estimates up to the June quarter 2023 are based on either the 2016 or 2011 edition of the ASGS, it has been necessary to use a correspondence to produce ERP estimates for those SA2s which have experienced any geographic changes between the earlier editions of the ASGS and the 2021 ASGS (even if those changes were not large enough to constitute a break in series in SALM).

While the use of corresponded data may have had a small impact on the labour force and unemployment rate estimates for a small number of the affected SA2s, analysis undertaken by JSA suggests this will not have been significant. For more information please contact SALM@jobsandskills.gov.au.

Unsmoothed SALM estimates

The smoothed SALM figures are JSA’s official estimates of unemployment and the unemployment rate at the SA2 and LGA levels. It is recognised, however, that some advanced users may need access to the unsmoothed estimates, which are used to produce the official figures, and these are provided.

The unsmoothed series can exhibit very high levels of variability. Accordingly, we advise exercising extreme caution when using the unsmoothed series, whether it is using point-in-time estimates or interpreting quarter-to-quarter (or even year-to-year) changes.

For changes that took place between the 2011 and 2016 editions of the ASGS (SALM moved from the 2011 ASGS to the 2016 ASGS in the June quarter 2019), please see the SALM 2016 ASGS Changeover User Guide – 2022 Update, which is available for download, below.

LGAs

There are around 540 LGAs in Australia.

• LGAs are based on the boundaries of the smallest government units (local councils) in Australia.
• JSA produces the SALM LGA estimates using the latest available SA2 to LGA correspondence from the ABS.

LGA boundary changes

LGA boundaries are determined by the governments of the states and the Northern Territory.

While there are often only minor, if any, changes year-to-year, sometimes there can be more substantial adjustments to this structure.

SALM transitions to a new LGA structure when a correspondence between SA2s and the new LGA structure is available from the ABS.

Boundary changes usually only lead to a small proportion of LGAs having a break in series. There can, however, be a larger number of breaks when moving to a new ASGS, as occurred between the June and September quarters 2023 (see the SALM 2021 ASGS Changeover Guide, above, for more details).

Smoothed data are not published for regions with a break in series until the fourth quarter after the break. 

SALM estimates have been based on the 2023 LGA structure since the June quarter 2023 release of SALM. There were only minor differences between this and the previous LGA structure – see the ABS ASGS page on the LGA structure for more details.

Estimates on earlier editions of the ASGS 

Smoothed SALM estimates based on the 2016 ASGS have been revised from the March quarter 2022 up to the June quarter 2023. If users need to obtain the revised SA2 or LGA estimates for past quarters based on the 2011 and/or 2016 ASGS (for example, for areas that have had a break in series), please email the SALM inbox at SALM@jobsandskills.gov.au.