LABOUR FORCE SEASONAL ADJUSTMENT AFTER REGIONAL BENCHMARKING AND CHANGE TO FUTURE STARTERS
From February 2004, Labour Force Survey (LFS) original estimates have been compiled using the new benchmarks derived from the 2001 Census. This is referred to as a benchmark. The revised benchmarks now include population benchmarks by 68 regions covering Australia. At the same time, the Australian Bureau of Statistics (ABS) has introduced a change to the definition of unemployed persons. Future starters not actively looking for work (FS) are now reclassified from the 'not in the labour force' category to the 'unemployed' category in line with the United Nations definition. The category changes were implemented by recompiling LFS original estimates from April 2001 onwards.
Changes in LFS original estimates will impact on related seasonal adjusted and trend estimates. To assist in the 2004 annual seasonal reanalysis of LFS estimates, the Time Series Analysis Section (TSA) investigated LFS time series estimates to assess the impact of the category change of FS, and regional benchmarking.
A possible impact of applying regional benchmarking is increased revision of the historical original estimates and, consequently, the seasonally adjusted estimates of Australian and State Labour Force. Our investigation showed that regional benchmarking did not introduce significant trend and seasonal breaks in the 64 directly adjusted LFS series, covering the 27 month period from January 1999 onwards.
A significant level shift (about 0.3% on average and ranging up to almost 0.8%) was detected in each of the National level seasonally adjusted unemployment component series. In addition, seasonal pattern changes were also identified for most female unemployment series. The common feature is that January female unemployment is seasonally higher under the new category definition.
Not all State level unemployment component series have shown significant FS induced changes in their underlying level. For small states and territories (Tasmania, ACT and NT), it was found that the FS original estimates were not reliable because FS were represented by only a handful of persons within the relatively small LFS samples. This resulted in volatile original estimates of FS, making it difficult to reliably estimate trend and seasonal break corrections for these states and territories. The trend and seasonal break correction factors at national level by gender were applied across all States and Territories first for a seasonal adjustment quality assessment. The corrections were then implemented only after improvements in seasonal adjustment quality were confirmed for the state level series.
The resultant estimates of trend and seasonal break corrections for FS were applied to Australian and State Unemployment component series prior to the 2004 annual seasonal reanalysis. The corrections adequately took account of the systematic (seasonal) and underlying (level shift) effects of the category change of FS.
For further information, please contact Mark Zhang on (02) 6252 5132, or Email: email@example.com.