How the ABS quality assures Labour Force data during times of change

How the ABS will approach quality assurance of Labour Force Statistics during the upcoming Modernisation Program

Released
16/04/2026
Released
16/04/2026 11:30am AEST

The ABS is improving how we collect the Labour Force Survey (LFS) so we can continue producing high-quality labour force statistics, maintain strong response rates, provide a better experience for respondents, and run our business more efficiently. These changes are being progressively implemented between April and August 2026.

The ABS undertakes thorough quality assurance processes for all aspects of the Labour Force Survey cycle. This includes monitoring how the data collection is tracking, and ensuring that data is accurate and is representing real-world economic changes. When improvements are introduced to the Labour Force Survey, significant testing activities occur prior to the change being implemented. For example, the ABS undertook a 2,000-household field test in October and November 2025, to successfully test the new systems and questionnaire ahead of the upcoming Labour Force Modernisation program. 

This article provides details and insights into how the ABS maintains the high quality of its Labour Force statistics. While the description is not exhaustive, it illustrates the general approach taken to identify, monitor, and address potential data quality issues. 

Introduction to Labour Force Methodology

How the Labour Force Survey is collected

The LFS collects information about the labour market activity of Australia’s resident civilian population aged 15 years and over. The LFS is a sample survey, meaning that the ABS does not interview every person like the Census does, and instead uses approximately 24,000 dwellings comprising approximately 50,000 people. These households are selected as representative of Australia’s population for key measures such as the number of employed persons and the unemployment rate. 

Rotation groups

The labour force sample is split into 8 groups, with each group remaining in the survey for eight months. Each month, one rotation group is rotated out of the survey, with a new incoming rotation group selected to replace it. This replacement group generally comes from the same geographic area as the outgoing rotation group. 

This rotation model is designed to: 

  • Provide reliable measures of monthly change in labour force status given seven-eighths of sample is retained from one month to the next.
  • Ensure the sample reflects changes in the population over time, such as the formation of new households, or shifts in household composition.
  • Reduce respondent burden by rotating households out of the survey after 8 months, helping to maintain higher response rates compared to long-term participation.

Monthly variability and standard errors

All surveys are subject to sampling error which depends on the sample size. If the sample size is larger, it is likely to be more representative of the entire population it is intending to capture, and the sampling error is likely to be smaller. Conversely, if the sample size is small, the sampling error is likely to be larger.

As such, whilst the LFS enumerates a suitable sample size, the results can sometimes vary from month-to-month, based on the characteristics of who responded to the survey and how they responded. 

The LFS is designed to have low error for key headline estimates at the national level. However, it can have notable month-to-month variability in state and territory estimates, as well as other smaller breakdowns, where the sample counts are much lower than for the headline estimates. 

The following graphs shows the monthly change in seasonally adjusted employed persons and the unemployment rate, compared to their decade averages. 

*2020s average excludes April 2020 – May 2022

*2020s average excludes April 2020 – May 2022

Standard errors quantify the amount of uncertainty associated with an estimate derived from sample survey data. They measure the extent to which an estimate could vary from sample to sample, if different households were selected to represent the same population under the same conditions. 

The seasonally adjusted unemployment rate typically has a two standard error of 0.2 percentage points. This means that if the ABS ran two separate Labour Force Surveys in a month, it is most likely that the unemployment rate for the two surveys would differ by up to 0.2 percentage points. 

The following table shows data on two standard errors on the March 2026 LFS results. 

Mar-26Two SE on levelsMonthly movementTwo SE on movement
Employed total ('000s)14,767.793.417.966.2
Unemployed total ('000s)656.336.8-3.737.2
Unemployment rate (%)4.30.20.00.2
Participation rate (%)66.80.4-0.10.2

It is important to take into account standard errors and month-to-month variability when interpreting movements in survey data. The ABS recommends that trend estimates be used as they provide a more reliable guide to the underlying direction of labour market data, especially for more detailed series which are subjected to larger degrees of variability and higher standard errors. 

 For more information on Labour Force Methodology, see Labour Force, Australia methodology.

Changes to Labour Force Methodology during transition

The ABS is making some changes to how the Labour Force Survey is conducted. Key questions in the Labour Force Survey that are used to calculate headline estimates will not be changed. Some small improvements will be made to the supporting information that provides guidance to respondents about how to accurately answer the survey questions. Some detailed questions will also either change, be removed, or new questions added. Furthermore, the systems used to collect the Labour Force Survey will be changing. 

These changes will be implemented progressively between April and August 2026. The timeline of these changes are:

  • From April 2026, the incoming rotation group will complete the survey on the new system and model.
  • In May, June and July 2026 two incoming rotation groups will also be onboarded to new system and collection model each month.
  • In July 2026, there will be seven rotation groups, instead of the usual eight.
  • By August 2026, all eight rotation groups will be using the new survey system and collection model. This will also be the first month the new Labour Supplementary Survey will be collected. 

The following diagram shows a visual representation of the transition plan for the modernisation of the Labour Force Survey. It also shows where there will not be continuation in the matched sample between months, and therefore the gross flows data should be used with additional caution during this time.  

Timeline of changes

Transitioning to the new systems and questionnaire in phases will allow the ABS to carefully monitor for any statistical impacts and enable any unforeseen system issues to be isolated and addressed. This approach will enable the ABS to continue publishing high quality estimates without disruptions to the release of key labour market information.

The full details of the changes can be found at Modernising the Labour Force Survey.

Labour Force quality assurance process

The below section provides an overview of some of the key quality assurance processes that ABS conducts for the Labour Force Survey. These processes are designed to ensure that the data is accurate and remains of a high quality for informed decision making. The below details are not exhaustive, but illustrate the general approach taken to identify, monitor, and address potential data quality issues. 

Collection checks

During the collection of LFS data, the ABS frequently monitors response counts at SA4 level, by mode and by rotation group. Where there is low response count during the enumeration, we investigate why this is so and undertake targeted follow-up with those who haven’t responded where possible. 

Where there is a disruption to collection in a particular month, such as a natural disaster, the ABS will also allocate additional staffing resources in future months to re-engage respondents when safe to do so (i.e. after the natural disaster ends). 

Representative checks

After collection, the ABS calculates coverage rates to indicate how well the responses represent the target population. These are calculated for each state and territory by rotation groups. This check also provides a measure of the representativeness of the incoming rotation group compared to other rotation groups. 

The ABS monitors response rates at state and territory levels by predicted characteristics of the addresses selected in the sample. These characteristics are predicted from models that use information from the Personal Level Integrated Data Asset (PLIDA). This information is monitored with the potential to detect any change to response patterns during a period of change. 

The ABS also analyses key estimates by rotation group, sex, and state and territory to see if any rotation groups’ results are notably different from other rotation groups, which can indicate that a particular rotation group is not accurately representing the population. In such a case, this could be due to collection disruptions or abnormal non-response. 

Rotation group bias checks

Analysing key estimates, as outlined in the aforementioned checks, can also suggest where any bias may have been introduced from a particular rotation group. During the transition, the ABS will pay particular attention to analysing the results between the old and new systems to ensure no bias is implemented due to the transition. 

We also calculate how much each rotation group influences the overall results for key estimates to determine its contribution. 

Confrontation with modelled estimates and administrative data

The ABS models the current month’s labour force data for further confrontation and quality assurance purposes. Each month, we use an Autoregressive Integrated Moving Average (ARIMA) model as a guide to indicate if results are outside forecast standard errors. This model uses only historical labour force data to predict the current month and so does not detect any real-world changes in the economy. 

We also use aggregated administrative data from Department of Social Services and payroll tax data from the Australian Tax Office to create forecasts using the Seemingly Unrelated Time Series Equations (SUTSE) model. This model utilises both historical data and real-time administrative data to forecast the current month’s labour force estimates, and can be used as a confrontation source for LFS estimates.  

These two models are designed for internal quality assurance purposes and are not used to forecast future results. 

Confrontation with external insights

As an additional quality assurance layer, the ABS confronts its labour force estimates against other publicly available reports and data on the current state of the economy. However, as a leading economic indicator, the LFS is often at the forefront of information showing changes in the economy. 

Adjustments to data

In rare cases, if there are issues in the collection of Labour Force data, the ABS has a range of methods to adjust when necessary. This can include changing the weight each respondent has, or drawing information from previous responses or respondents with similar characteristics. The ABS will highlight in the Labour Force release if any intervention has been made in a particular month. For example, in November 2025, the ABS adjusted the weights for NSW, as outlined in Labour Force, Australia, November 2025

 

For further details on the Labour Force Modernisation plans, see Modernising the Labour Force Survey and Upcoming changes to Labour Force Survey outputs.

If you would like to provide feedback or seek clarification, please email labour.statistics@abs.gov.au

Back to top of the page