The ABS releases two sets of regional Labour Force Survey estimates – the direct survey estimate from the Labour Force Survey and new, higher quality modelled estimates.
The ABS recommends using the modelled estimates over the direct survey estimates whenever possible in reporting on employment and unemployment rates for regional labour markets.
The modelled estimates are also the best source of data for analysis at the Greater Capital City Statistical Areas level. While estimates for capital cities and the balance of states have not been directly modelled, the aggregated modelled SA4 data are better than the direct survey estimates, particularly for the balance of states which generally have smaller populations and smaller sample sizes.
These modelled estimates will continue to be added to Labour Force, Australia, Detailed a week after each release. The ABS will move the release of the modelled estimates to the same day as the direct survey estimates later in 2024.
Advantages of the modelled estimates
The Labour Force Survey is designed for national and state and territory estimates with regional data produced as a secondary product. High levels of variability and extreme fluctuations are observable in the direct estimates. This can partly be because of real-world events, like floods, but are often due to the nature of sampling areas with smaller populations. As a result, the direct estimates are of lower statistical quality than the modelled estimates at the SA4 level, especially for sparsely populated rural and regional areas.
The Rao-Yu time-series model used to produce the modelled estimates provides less variable and more reliable levels of unemployment and employment as it draws on the strength of other administrative data and the strong relationship between that data and the labour force survey estimates. This administrative data includes de-identified Single-Touch Payroll data from the Australian Taxation Office for modelling employment and de-identified Youth allowance and JobSeeker recipients data from the Department of Social Services for modelling unemployment.
Further information on these regional estimates as well as other regional labour statistics the ABS publishes, is available in the Regional labour market data guide. For information on the development of these estimates, see Improving SA4 level estimates from the Labour Force Survey using administrative data models and Further refinements to modelled SA4 level Labour Force estimates using administrative data.
When to use the direct survey estimates
While ABS plans to produce modelled regional labour estimates by age and sex, these are not yet available.
When using the direct survey estimates (e.g. for insights at the regional level by age and sex, or for periods for which modelled estimates are not available), previous advice on using (moving) annual averages applies. Annual averages are important for understanding the current state of the labour market, and providing medium and long-term signals, as they can lessen the impact of the month-to-month variability associated with small sample sizes.
However, unusual, sudden and/or extreme changes in the labour market (e.g. changes to regular and seasonal patterns in employment, working hours, job search or availability as a result of COVID-19) can be difficult to account for in the short-term using a 12-month moving average. This should be considered before drawing any conclusions from these estimates.
In contrast, annual averages are not required when using the modelled data.
The ABS will provide further advice on how to use the modelled in conjunction with the direct survey estimates (for example, for age and sex analysis, or for comparisons over time where there are currently no modelled estimates), in the coming months.
How the modelled estimates compare to state and territory estimates
The modelled SA4 estimates are designed to be additive to the employment, unemployment and not in the labour force totals for each state and territory (and Australia) published in Labour Force, Australia. However, while the SA4 estimates in each state are 'calibrated' to the respective state or territory totals, in some states in some months prior to July 2022, the sum of people not in the labour force in SA4s may differ slightly from the published state total for people not in the labour force. Where these slight differences occur, there is no impact on the utility of the modelled estimates. The ABS intends to further refine the state-level calibration over the next few months, to ensure alignment with all state totals.