Australian Bureau of Statistics
6269.0 - Information Paper: Labour Force Survey Sample Design, Nov 2007 (Third edition)
Latest ISSUE Released at 11:30 AM (CANBERRA TIME) 01/10/2009
|Page tools: Print Page Print All RSS Search this Product|
The ABS reviews the LFS sample design every five years following the five-yearly Census of Population and Housing (Census). While the design has remained broadly the same since the introduction of the LFS, the review ensures that the survey continues to accurately reflect the geographic distribution of the Australian population, and remains efficient and cost-effective.
The 2006 LFS sample design, which took into account results from the 2006 Census, was implemented from November 2007 to June 2008. The 2006 sample design resulted in a smaller sample size, taking advantage of gains from the earlier introduction of the composite estimation methodology. Standard errors for key LFS estimates were maintained at the levels achieved under the previous sample design (except for the Northern Territory where they were improved). For details refer to the first edition of this Information Paper.
In July 2008, as one of a range of ABS savings initiatives for 2008-09, there was a 24% reduction in the LFS sample size relative to the June 2008 sample size. For details of the sample reduction refer to the second edition of this Information Paper.
On 13 May 2009, the Australian Statistician announced that the ABS would restore the 24% sample reduction.
2009 SAMPLE RE-INSTATEMENT
The ABS has made every effort to re-instate the sample in the shortest time possible, taking into account the potential impact on the volatility of key estimates. Three options were considered: an eight month phase-in period (the usual period for introducing new sample); a four month phase-in period; and a one month phase-in period. It was determined that the sample could be re-instated over a four month period. The chosen approach aims to accommodate the workload surge and to minimise volatility in the LFS estimates.
The private dwelling sample in larger urban centres and less remote areas, and the sample of Indigenous community dwellings, will be progressively re-instated each month from September to December 2009. This represents 87% of the sample being re-instated. The remaining 13% (in the more remote, less populated areas and for non-private dwellings) will be fully re-instated in October 2009.
The December 2009 LFS estimates, to be released on 15 January 2010, will be the first produced under the fully re-instated sample.
EXPECTED SAMPLE SIZE
Table 1 shows the number of fully responding persons (actual or expected) in the LFS sample at various points in time since the implementation of the 2006 sample design, specifically:
The sample re-instatement is expected to result in approximately 56,000 persons fully responding to the LFS in December 2009, covering about 1 in 315 (0.32%) of the civilian population aged 15 years and over. Thereafter, the sample size will gradually increase in line with population growth until the sample is redesigned following the 2011 Census.
NEW SAMPLING FRACTIONS
Table 2 shows the state and territory sampling fractions from the 2006 sample design, the July 2008 reduced sample and the December 2009 fully re-instated sample. The three sets of sampling fractions are an output from the 2006 sample design process. As can be seen, the sampling fractions for the fully re-instated sample revert to those for the 2006 sample design. Table 2 also shows the sampling fractions from earlier sample designs.
IMPACT ON QUALITY OF ESTIMATES
The most commonly used measure of the reliability of estimates is their relative standard errors (RSEs). The RSE is the standard error divided by the estimate. For the fully re-instated sample, the RSEs for estimates of employment and unemployment at the national, state and territory level are expected to be approximately 15% lower, on average, than those for the reduced sample. As the sample is re-instated from September to December 2009, the RSEs will progressively decrease.
Table 3 shows the target RSEs for the 2006 sample design, for the reduced sample from July 2008, and for the re-instated sample from December 2009. As can be seen, the target RSEs for the fully re-instated sample revert to those for the 2006 sample design.
IMPACT ON LFS PRODUCTS
Key monthly estimates from the LFS are published in Labour Force, Australia (cat. no. 6202.0). More detailed estimates are presented in Labour Force, Australia, Detailed - Electronic Delivery (cat. no. 6291.0.55.001) and Labour Force, Australia, Detailed, Quarterly (cat. no. 6291.0.55.003).
The full sample re-instatement will decrease the standard errors on LFS estimates by approximately 15%. This will be evident from the December 2009 issue onwards, where there will be a decrease in the number of estimates bearing the annotation "subject to sampling variability too high for most practical purposes".
Gradual improvement in the standard errors will also be evident in the September to November 2009 issues of LFS products as the sample progressively increases. An additional short term impact on the standard errors is expected in October 2009 when the sample is re-instated in more remote, less populated areas and for non-private dwellings (13% of the sample). This one-off re-instatement of 13% of the sample reduces the efficiency of the LFS composite estimation methodology as there will be a lower than normal proportion of common selections between September and October 2009. Therefore, the standard errors for month-to-month movement estimates will be slightly higher in October 2009 than in other months.
LFS standard errors are statistically modelled as a function of the monthly estimates. The standard error models will be modified to take account of the sample re-instatement. These modified models will be used to populate the standard error tables in Labour Force, Australia (cat. no. 6202.0). The modified models will also be incorporated in Labour Force Survey Standard Errors, Datacube (cat. no. 6298.0.55.001), to be released on 8 October 2009. This product allows users to calculate the standard error for any LFS estimate.
IMPACT ON ASSOCIATED SURVEYS
The ABS conducts supplementary surveys using the LFS sample. Each supplementary survey covers a different topic, and each topic is conducted on either an annual or less frequent basis. Table 4 shows the supplementary surveys conducted during the period of the reduced sample (from July 2008 to August 2009) and during the sample re-instatement (from September to December 2009).
For those supplementary surveys conducted during the period of the reduced sample, the sample size was up to one-third smaller than the sample when the survey was previously conducted. As a result, the standard errors on the estimates increased.
The sample re-instatement will decrease the standard errors on estimates for all supplementary surveys (with the exception of the Northern Territory which had a large increase in sample size under the 2006 sample design).
The ABS also conducts a multi-purpose household survey (MPHS) using the LFS sample. The MPHS includes a wide range of labour-related and other topics including: barriers and incentives to labour force participation, retirement and retirement intentions, work related injuries, crime victimisation, and household use of technology. The MPHS sample size will be retained, with the proportion of the LFS sample selected in the MPHS being reduced accordingly. The ABS expects the sample re-instatement to have a minimal impact on the MPHS.
LFS estimates are calculated in such a way as to add up to independent estimates of the civilian population aged 15 years and over (population benchmarks). The weighting methodology used in the LFS requires at least one respondent in each benchmark cell.
Under the reduced sample, the probability of some benchmark cells having no respondents increased (compared with the probability under the 2001 sample design). To address this, the ABS combined some single year age groups within the 15 to 24 years age range for the balance of South Australia, the balance of Western Australia and the city of Hobart. For details refer to the second edition of this Information Paper.
After the sample is re-instated, the probability of these benchmark cells having no respondents will be reduced, but not entirely removed. Therefore, these changes will remain in place until the sample is again redesigned (following the 2011 Census).
2006 sample design, refer to Information Paper: Labour Force Survey Sample Design, Nov 2007 (cat. no. 6269.0).
LFS sample reduction, refer to Information Paper: Labour Force Survey Sample Design, Nov 2007 (Second edition) (cat. no. 6269.0).
Implementation of LFS Statistical Regions, refer to Information Paper: Regional Labour Force Statistics, February 2009 (cat. no. 6262.0).
Maps of LFS Statistical Regions, refer to Australian Labour Market Statistics, July 2009 (cat. no. 6105.0).
LFS estimates, refer to Labour Force, Australia (cat. no. 6202.0).
< Back to top >
These documents will be presented in a new window.
This page last updated 1 October 2009