6202.0 - Labour Force, Australia, Sep 2004  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 07/10/2004   
   Page tools: Print Print Page Print all pages in this productPrint All

SEPTEMBER KEY FIGURES

Aug 2004
Sep 2004
Aug 04 to Sep 04
Sep 03 to Sep 04

Trend
Employed persons ('000)
9,676.2
9,681.2
5.0
2.0
%
Unemployed persons ('000)
575.5
577.1
1.6
-3.4
%
Unemployment rate (%)
5.6
5.6
0.0
pts
-0.3
pts
Participation rate (%)
63.5
63.5
0.0
pts
0.1
pts
Seasonally Adjusted
Employed persons ('000)
9,653.7
9,717.2
63.5
2.5
%
Unemployed persons ('000)
579.0
572.4
-6.5
-4.5
%
Unemployment rate (%)
5.7
5.6
-0.1
pts
-0.4
pts
Participation rate (%)
63.4
63.6
0.3
pts
0.3
pts

Employed Persons
Graph: Employed Persons

Unemployment rate
Graph: Unemployment rate



SEPTEMBER KEY POINTS


TREND ESTIMATES (MONTHLY CHANGE)

  • EMPLOYMENT increased to 9,681,200
  • UNEMPLOYMENT increased to 577,100
  • UNEMPLOYMENT RATE remained at 5.6%
  • PARTICIPATION RATE remained at 63.5%


SEASONALLY ADJUSTED ESTIMATES (MONTHLY CHANGE)

EMPLOYMENT
  • increased by 63,500 to 9,717,200. Full-time employment increased by 19,100 to 6,960,700 and part-time employment increased by 44,400 to 2,756,500.

UNEMPLOYMENT
  • decreased by 6,500 to 572,400. The number of persons looking for full-time work decreased by 4,400 to 402,400 and the number of persons looking for part-time work decreased by 2,100 to 170,000.

UNEMPLOYMENT RATE
  • decreased by 0.1 percentage point to 5.6%. Both the male and female unemployment rates decreased by 0.1 percentage point, to 5.5% and 5.6% respectively.

PARTICIPATION RATE
  • increased by 0.3 percentage points to 63.6%.


NOTES

IMPACT OF SURVEY RESPONSE RATES ON AUGUST ESTIMATES

The Labour Force Survey (LFS) usually achieves high levels of response from households selected for enumeration each month. Typically, the response rate for the LFS is around 97% nationally.


Operational difficulties during enumeration of the August 2004 survey resulted in a higher level of non-contact of households, and hence a lower than usual response rate - just over 94% nationally. Analysis conducted by the ABS subsequent to the publishing of August 2004 LFS results indicates that, due to this reduced response rate, the level of the Australian seasonally adjusted employment estimate for August may have been understated by between 15,000 and 25,000 persons. (As the standard error on the national employment estimate for August was 38,700 persons, an understatement of this size would be well within the bounds of expected sampling error.)


Any understatement in the estimated employment level for August would have affected the monthly movement in the seasonally adjusted estimate of employment between July and August as well as between August and September.


The reduction in response rates in August varied across the states and territories, but was most pronounced in NSW. Hence any understatement in August would have been greatest in NSW.


The response rate for the September LFS (close to 96% nationally) is still below the usual level. However, initial analysis of LFS data indicates that this reduced response rate has had no significant impact on the estimate of employment for September. The ABS will, however, continue to analyse data as response rates return to normal levels.


These changes to response rates have had no significant impact on the unemployment estimates.



INQUIRIES

For further information about these and related statistics, contact the National Information and Referral Service on 1300 135 070 or Michael Johnson on Canberra (02) 6252 6525.


SAMPLING VARIABILITY


INTRODUCTION

Labour Force Survey (LFS) estimates are based on information obtained from a sample of the population, and are subject to sampling variability (or sampling error), i.e. they may be different from the figures that would have been obtained from an enumeration of the entire population. The expected magnitude of the sampling error depends on the sample design, the sample size and the population variability. One measure of the expected magnitude of the sampling error of an estimate is the standard error.


There is approximately a 95% chance that an estimate will differ from the true value by less than two standard errors. Hence an approximate 95% confidence interval is the region covering two standard errors either side of a survey estimate. The following table shows 95% confidence intervals for key LFS seasonally adjusted monthly movements between August and September.

Movements in seasonally adjusted series between August and September 2004

Monthly change
95% Confidence interval

Total Employment
63,500
26,300
to
100,700
Total Unemployment
-6 500
-20 100
to
7,100
Unemployment rate
-0.1 pts
-0.3 pts
to
0.1 pts
Participation rate
0.3 pts
0.1 pts
to
0.5 pts



STATE AND TERRITORY ESTIMATES

The LFS is designed primarily to provide reliable estimates of the key labour force statistics for the whole of Australia and, secondarily, for each state and territory. The allocation of the sample across the states and territories is designed as a compromise between the accuracy of national estimates and state or territory estimates. The ABS routinely warns users that the estimates (especially for states and territories with smaller sample sizes) can have relatively high sampling variability.


Example

LFS estimates of employed and unemployed persons for the Northern Territory exhibit substantial sampling variability, reflecting the relatively small sample enumerated there. In the Northern Territory, approximately 580 selections (private dwelling and non-private dwelling households) are surveyed each month compared with New South Wales (for example) where the number surveyed is around 7500. The following graph shows the unemployed persons original series for the Northern Territory, together with approximate 95% confidence intervals for each month's level estimate.

NT Unemployed Persons, Original Series
Graph: NT Unemployed Persons (Original)



SAMPLE ROTATION

The LFS sample design provides for the replacement each month of around one-eighth of the sample. This rotation is generally to neighbouring households. Usually people in the new households have similar labour force characteristics to those of the people they are replacing, but occasionally the rotation is to a set of households with very different characteristics - this is a normal part of the random selection process, and can happen at any time. It contributes to the sampling variability of all LFS estimates, as reflected in the standard errors on pages 27 and 28 of this publication. However, especially when sample sizes are relatively small, the resultant changes in the estimates are not necessarily representative of real world changes.


Example

An example of the impact of sample rotation on sampling variability occurred in the Northern Territory between April and May 2004. In one particular area of the Northern Territory, the sample was rotated to a group of households with a much higher number of unemployed persons than is usually found in Northern Territory households.


The following graph shows the Northern Territory trend estimate of the number of unemployed persons as published, together with a trend estimate of unemployed persons as it would look if the persons in these households had the average characteristics of persons in the remaining sampled households. The vertical bars on each time point of the graph show a very approximate estimate of the 95% confidence interval around each point of the published trend series. The graph indicates the sensitivity of the trend estimate to some rotations of the sample.

NT Unemployed Persons, Trend Series

Graph: NT Unemployed Persons (Trend)




ENHANCEMENTS TO THE LFS

The ABS has introduced some changes to the LFS over the last year that have helped to reduce monthly volatility, such as the use of concurrent seasonal adjustment and regional population benchmarks.


The ABS is also looking at further reducing volatility in LFS estimates by using improved estimation techniques to adjust for households with unusual characteristics. This is expected to lead to more robust estimates for states and territories with smaller samples (such as the Northern Territory). It is hoped that these changes can be implemented by mid 2005.


PRINCIPAL LABOUR FORCE SERIES TREND ESTIMATES

EMPLOYED PERSONS

The trend estimate of employed persons generally rose from a low of 7,637,500 in December 1992 to 9,054,900 in September 2000. The trend then fell slightly to 9,033,700 in January 2001, before rising to 9,464,200 in March 2003. The trend then fell for three months before rising to stand at 9,681,200 in September 2004.

Graph: Employed Persons (Trend)



UNEMPLOYED PERSONS

The trend estimate of unemployed persons fell rapidly from 920,700 in September 1993 to 728,100 in July 1995. The trend then rose to 771,700 in February 1997, before falling to 583,500 in September 2000. After rising to 685,200 in October 2001, the trend has generally fallen to stand at 577,100 in September 2004.

Graph: Unemployed Persons (Trend)



UNEMPLOYMENT RATE

The trend unemployment rate fell rapidly from 10.7% in August 1993 to 8.1% in July 1995. The trend then rose slowly, reaching 8.4% in February 1997, before falling to 6.1% in September 2000. After rising to 7.0% in October 2001, the trend has since generally fallen to stand at 5.6% in September 2004.

Graph: Unemployment Rate (Trend)