IMPACT ON STANDARD ERRORS
Standard errors, including those for employment and unemployment estimates, associated with the 2011 sample design are predicted to generally align with the target 2006 design standard errors. For unemployment the 2011 design (target) standard errors will generally be slightly higher than those that were actually achieved under the 2006 sample. In the case of employment estimates, the standard errors, under the 2011 design, are predicted to typically be lower than standard errors predicted at the time of the sample design of the 2006 sample. The change in standard errors will be fully evident once the transition from old to new sample is complete, i.e. from September 2013. The transition itself will have a short-term impact on the standard errors of the LFS movement estimates.
The sample is being phased-in over the four month period May 2013 to August 2013. Over the four months, the LFS sample will become progressively smaller because the 2011 sample is smaller than the current sample size used in the first half of 2013. The standard errors will progressively change from the current levels to the new levels by September 2013.
There will be an additional short-term impact on the standard errors during the phase-in of the new sample. Introducing two rotation groups from the new sample per month (compared to the usual introduction of one rotation group each month) reduces the proportion of common selections each month over the period between May and August 2013. As a result, the standard errors on month-to-month movement estimates are predicted to increase by approximately 10% during this period and will only have a marginal impact on the quality of level estimates. This is an improvement on the 22% increase in movement standard errors reported in the July 2012 Labour Force, Australia, Jul 2012 (cat. no. 6202.0) because the multipliers used in composite estimation have since been optimised during the phase-in to reduce the impact of the increase in rotation.
The standard errors are published each month for key estimates in Labour Force, Australia (cat. no. 6202.0). These published standard errors have been statistically modelled as a function of the estimate itself. The ABS makes these models available on the ABS web site <www.abs.gov.au> to allow the estimation of the standard error of any LFS statistic.
The updated standard errors will be incorporated from the May 2013 issue of Labour Force, Australia (cat. no. 6202.0), which will be released on 13 June 2013. The revised models will be made available through the product Labour Force Survey Standard Errors, Data Cube (cat. no. 6298.0.55.001), which will also be released in June 2013. The Information Paper: Labour Force Survey Standard Errors, 2005 (cat. no. 6298.0), has not been updated but contains relevant information on standard errors for the LFS.
RELATIVE STANDARD ERRORS
The table below shows the target RSEs for the 2006 sample design, the RSEs actually achieved during the 2006 sample design, and the predicted RSEs for the 2011 sample design.
The table shows that, averaged over the life of the new 2011 sample design, the predicted RSEs for employment at the national, state and territory level are expected to be generally the same or similar to those achieved under the previous sample design, but slightly lower than the target RSEs from the 2006 sample design. Predicted RSEs for unemployment at the national, state and territory level 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 2011 sample design aimed to attain RSEs comparable to the target RSEs from the 2006 design. However, analysis of recent LFS data suggests that a larger sample size is required to achieve the unemployment 2006 target RSEs compared with the sample size required to achieve the employment 2006 target RSEs. A hybrid measure of sampling error which combines the sampling error for estimating employment and unemployment was used to define a single target RSE. The sample size required to achieve this hybrid target RSE has resulted in the predicted RSEs for unemployment generally being higher than the target RSEs from the 2006 design.
TABLE: STANDARD ERRORS TARGET AND ACHIEVED, 2006 and 2011 Sample Design
Employment RSE (%)
Unemployment RSE (%)
|New South Wales|
|Australian Capital Territory|