6269.0 - Information Paper: Labour Force Survey Sample Design, May 2013  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 30/05/2013   
   Page tools: Print Print Page Print all pages in this productPrint All



To maintain the efficiency and effectiveness of the LFS over time, the 2011 sample design has the following key aims:

  • to achieve a level of accuracy for employment and unemployment estimates at the national level and for each state and territory comparable with the target accuracy established for the previous sample design;
  • to align the multi-stage area sampling with the geography defined by the ASGS;
  • to contain the costs of collection for the LFS sample, and;
  • to provide sufficient sample for the LFS for the following five year period.


A number of improvements were considered in developing the new design and the following changes have been implemented:
  • using the Statistical Area Level 4 (SA4) classification from the Australian Statistical Geography Standard (ASGS) to define the geographic sampling strata as a result of the ASGS replacing the Australian Standard Geographical Classification (ASGC);
  • using the boundaries of mesh blocks in the ASGS to create the area sampling units within which groups of dwellings are selected, this replaces the Census Collection Districts used under ASGC;
  • independent selection of two master samples of areas to provide separate available samples of dwellings for MPS and other ABS household social surveys. This is designed to better target Special Social Surveys; and
  • a reduction in the initial sample size of around 4% compared with that initially implemented at the start of the 2006 design. This reduction is due to efficiencies of the 2011 sample design targeting the 2006 target Relative Standard Errors (RSEs). Further details are outlined in the in the Impact on Standard Errors chapter.


There have been some minor changes to the relative sample allocation between the states and territories for the 2011 sample design. The relative allocation of the national sample across the states and territories has increased in Tasmania, Northern Territory and New South Wales, while the relative allocation in the remaining states and Australian Capital Territory has decreased as a result. Analysis has shown the target accuracy for employment and unemployment estimates in Western Australian and Australian Capital Territory can be attained with a smaller sample size than the 2006 sample design.


The sample design is specified in terms of selecting a proportion of dwellings within the state/territory. This is the sampling fraction. Traditionally the sampling fraction is not changed during the five-year life of the sample design, meaning that the sample size increases over time as the population size grows, resulting in a gradual increase in the number of persons enumerated during the life of each sample design. While the outcome of this is some improvement in the accuracy of the survey results, the improvement is partially offset by a deterioration in the efficiency of the sample in the period since the previous sample design. Further, as more dwellings are added to the survey over time, the operational costs of collecting the data increase. At each sample design, a new sampling fraction is determined to produce estimates at the desired level of accuracy and this usually results in a decrease in sample size. One notable change to the level of accuracy occurred from July 2008 through to November 2009, when there was a 24% reduction in the size of the LFS sample as part of a savings initiative. More detail is available on this reduction in July 2008 and reinstatement in December 2009 in Information Paper: Labour Force Survey Sample Design, Nov 2007 (Third edition) (cat. no. 6269.0).

In the 2011 sample design, the initial sample size is expected to be about 4% smaller than the sample at the start of the 2006 design. The new sample has been designed with the aim of achieving similar levels of sampling error as the target levels established for the 2006 sample design. Generally the previous design achieved lower levels of sampling error than the sample was designed for, allowing for a reduction in the 2011 sample size to match the 2006 targets. The reduction in sample size is predicted to generally result in RSEs for employment at the same or similar level to those achieved under the previous sample design, but slightly lower than the target RSEs from the 2006 sample design. Predicted RSEs for unemployment are expected to be generally slightly higher than those achieved under the 2006 sample design (although Tasmania and Northern Territory are predicted to be slightly lower). The Impact on Standard Errors section of this information paper has further information on expected RSEs.

When the new sample is fully implemented in August 2013, the LFS is expected to comprise about 26,200 private dwellings; 400 non-private dwellings; and 150 dwellings from discrete Aboriginal and Torres Strait Islander communities. This is expected to result in about 52,200 persons responding to the survey, covering about 1 in 312 (0.32%) of the civilian population aged 15 years and over.


Table 1 gives the sampling fractions for each state and territory as a result of each sample design from 1976 through to 2011.

TABLE 1. SAMPLING FRACTIONS, By year of sample design

New South Wales
1 in 200
1 in 200
1 in 230
1 in 277
1 in 300
1 in 321
1 in 380
1 in 419
1 in 200
1 in 200
1 in 230
1 in 242
1 in 257
1 in 270
1 in 336
1 in 390
1 in 140
1 in 140
1 in 160
1 in 195
1 in 222
1 in 239
1 in 315
1 in 369
South Australia
1 in 100
1 in 100
1 in 115
1 in 139
1 in 147
1 in 149
1 in 184
1 in 209
Western Australia
1 in 90
1 in 100
1 in 115
1 in 146
1 in 160
1 in 165
1 in 246
1 in 295
1 in 60
1 in 60
1 in 70
1 in 75
1 in 83
1 in 90
1 in 103
1 in 99
Northern Territory
1 in 100
1 in 100
1 in 115
1 in 75
1 in 85
1 in 98
1 in 54
1 in 52
Australian Capital Territory
1 in 100
1 in 100
1 in 115
1 in 75
1 in 85
1 in 86
1 in 117
1 in 149

Sampling fractions have changed from the 2006 design due to:
  • a reduction in the 2011 initial sample size to maintain the same 2006 sample design target accuracy measures; and
  • population increases since the 2006 sample design.


To reduce the potential impact of the change in sample on labour force statistics, the new sample is being phased-in progressively.

The private dwelling sample in urban centres and non-remote areas, will be phased-in over a four month period from May 2013 to August 2013. In each of these four months one-quarter of the new sample will be introduced, This represents almost 90% of the total sample.

The rest of the sample (in the more remote, less populated areas and for non-private dwellings) will be introduced in two stages: in June 2013 for Tasmania, the Northern Territory and the Australian Capital Territory, and in July 2013 for New South Wales, Victoria, Queensland, South Australia and Western Australia.

The sample of dwellings in discrete Aboriginal and Torres Strait Islander communities will continue to be maintained from the 2006 design in the short term and new selections are planned to be implemented by the end of 2013.

This method of implementation means that most of the changes to labour force statistics due to differences between the two samples will be spread over the four months. Previous redesigns have introduced the new sample over periods of one, four or eight months. The 2006 sample was phased-in over eight months.

The increased sample rotation during the phase-in of the new sample over the four month period will have an impact on the quality of estimates. Movement standard errors will increase by approximately 10%. Due to the use of composite estimation, there will only be a marginal impact on the quality of level estimates. Gross Flows analysis will be impacted by the phase-in with between 60% and 70% of the sample available for matching between the current and previous months, instead of the usual 80%. After the phase-in to the new sample, the quality of level and movement estimates will align with those to be expected for the 2011 design. For more detail see table in the Impact on Standard Errors section.