Feature Article - Labour Force Survey Regions
This article was published in the July 2004 issue of Australian Labour Market Statistics (cat. no. 6105.0).
To find maps of the Labour Force Survey regions, press the Details Tab from the top of this Page.
Estimates of the labour force status of people are available on a monthly basis for 77 Labour Force Survey (LFS) regions. These series go back to October 1982. LFS regions were established to meet user interest in small area data from the LFS.
This article discusses how LFS regions are formed and provides information on the quality of estimates for these regions, using the Inner Sydney Statistical Region as an example.
AUSTRALIAN STANDARD GEOGRAPHICAL CLASSIFICATION
Labour Force Survey regions are formed from elements of the Australian Standard Geographical Classification (ASGC). Therefore, to understand LFS regions, it is first necessary to look at the ASGC.
The ABS introduced the ASGC in 1984. The classification provides well defined structures of hierarchical geography. The ABS uses the ASGC as the geographical framework for the collection and dissemination of statistics about Australia's population and economy. The ASGC provides structures covering the whole of geographic Australia and also specific parts of interest.
Statistical local areas (SLAs) are building blocks for the ASGC. The SLA is the smallest spatial unit of the ASGC except in a census year, when the census collection district is the smallest unit. (Census collection districts aggregate to SLAs at census time.)
SLAs have a close relationship with local government areas (LGAs). An LGA can be composed of one or many SLAs. LGAs in total do not cover the whole of geographic Australia, but SLAs do add to Australia.
The ASGC has seven interrelated classification structures, each of which serves a specific purpose. The ASGC statistical region structure generally provides the geographical dimension of LFS outputs. (Information on the other ASGC structures is available in Australian Standard Geographical Classification (cat. no. 1216.0).) The ASGC statistical region structure is composed of SLAs that aggregate to statistical region sectors (SRSs) and in turn to statistical regions (SRs), major statistical regions (MSRs), then to states/territories and the whole of geographic Australia.
The LGA structure is the main driver of change to the boundaries of geographical areas of the ASGC. State and territory governments gazette changes to LGAs and these changes affect other structures in the classification through the link between LGAs and SLAs. These changes are the main reason the ABS publishes an updated version of the Australian Standard Geographical Classification (ASGC) (cat. no. 1216.0) every year.
The ASGC is reviewed periodically. The most recent reviews were held in 1996 and 2001. Such reviews can result in changes to definitions, concepts and boundaries.
LABOUR FORCE SURVEY REGIONS
LFS regions are contiguous regions composed of geographical areas defined in the ASGC statistical region structure current at the time of a Census of Population and Housing (for example 1996, 2001). There are currently 77 of these LFS regions across Australia, which include major statistical regions, statistical regions, statistical region sectors and one statistical division, and are based on the 2001 edition of the ASGC.
Maps showing these regions are available on the Labour theme page of the ABS web site <http://www.abs.gov.au>. A list of LFS regions is available in Information Paper: Labour Force Survey Sample Design (cat. no. 6269.0).
Because the size of an LFS region is determined largely by the sample size needed to produce accurate estimates, the smaller states/territories have few LFS regions. In the Northern Territory and the Australian Capital Territory, the only LFS region is the entire territory.
How LFS regions are formed
LFS regions were originally established after extensive consultation with major users of labour force data. Estimates for these LFS regions were first released in 1985. Factors in the design of the LFS regions included: the sample sizes required to yield reliable estimates; the need for consistency with the ASGC; and the need for comparability with other statistical collections.
LFS regions are determined, in part, by the expected sample size for each region. If the regions are too small, then the accuracy of estimates will not be acceptable: relative standard errors on estimates will be very large, and the estimates will not be reliable or useful. The regions represent a compromise between user interest in small area data and the design limits of the sample.
Why LFS regions change
The boundaries of LFS regions are revised as part of the five-yearly review of the LFS sample design. This review uses census data, and takes into account the most recent changes to the ASGC. Since the LFS regions are aggregations of SLAs, changes at the SLA level, including those flowing from LGA changes, affect the boundaries of the regions.
The regular revisions of the LFS sample aim to ensure that regions have sufficiently reliable estimates, and that the continuity of LFS region time series is maintained. The latest revisions were based on 2001 census data. The new sample design from the review was gradually implemented between November 2002 and June 2003. Details of the changes made are outlined in the ABS publication Information Paper: Labour Force Survey Sample Design (cat. no. 6269.0).
The ABS conducted a major review of LFS regions during 1996. The boundaries for many regions were changed to maintain consistency with the ASGC, and three new regions were created to meet user needs. More information on the changes resulting from the review is available in the ABS publication Information Paper: Regional Labour Force Statistics, September 1997 (cat. no. 6262.0). The ABS plans to conduct a minor review of LFS regions in 2006.
How data for LFS regions are used
State government agencies and others use labour market data for LFS regions to monitor the level of regional activity, assess regional development issues, and to inform the development of employment policies and programs. Some agencies are interested in the broad trends for regions as part of assessing the state's economic performance.
Limitations of LFS regions
The LFS regions have some limitations. Users generally want data for the smallest areas possible, so the main criticism is that the LFS regions are too big. The need to have regions with a large enough sample to produce reasonably accurate estimates has to be balanced with the desire of data users to have estimates for small specific areas.
While the ABS reviews the LFS regions as part of each five-yearly sample redesign, it is not possible to backcast estimates based on the new regions. As a result, changes tend to be kept to a minimum to preserve time series.
QUALITY OF LABOUR FORCE SURVEY REGION DATA
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 survey is not designed to provide accurate regional estimates. Since estimates for regions are based on much smaller samples they are subject to higher levels of sampling error.
Relative standard errors can be calculated for estimates from the LFS to give an indication of the percentage errors likely to have occurred due to sampling. (For more information on standard errors in the LFS, see Information Paper: Labour Force Survey Standard Errors (cat. no. 6298.0).) In LFS output, only estimates with relative standard errors of 25% or less are considered sufficiently reliable for most purposes. Tables showing the level at which LFS regions have a relative standard error of 25% are available on request.
Following each Census of Population and Housing, the ABS selects a new sample for the LFS, to ensure that the sample continues to accurately represent the distribution of the Australian population. Estimates for LFS regions may be more volatile at the time of sample re-selection because the new sample selected to represent each region may have different characteristics to the old sample. The most recent sample re-selection was implemented from November 2002 to June 2003. The way the new LFS sample is implemented means that regions in more remote areas are more subject to disturbances than those in less remote areas.
LFS estimates of people employed, unemployed and not in the labour force are calculated to add up to independent estimates (or benchmarks) of the usually resident civilian population aged 15 years and over. These population benchmarks are based on Census of Population and Housing data, adjusted for differences in scope and under-enumeration in the census. The benchmarks are produced monthly, based on projections of population estimates. The benchmarks are reviewed after each census (every five years), when revised population estimates are produced.
Before the February 2004 survey, population benchmarks in the LFS were classified by state/territory of usual residence, capital city/rest of state, age and sex. In addition to these benchmarks, from February 2004 the LFS also uses population benchmarks for LFS region by sex. LFS estimates at the region level were revised back to January 1999 with the release of February 2004 data.
The use of regional population benchmarks improves the quality of estimates for LFS regions, particularly estimates of employment, without compromising the quality of estimates at national, state and territory levels. However, the estimates for LFS regions are still based on small samples. Further improvement in the quality of regional estimates from the LFS would need a substantial increase in sample size (and hence in the cost of the survey and in respondent load).
Volatility of estimates for LFS regions
Labour force series for LFS regions are considerably more volatile than the corresponding state series. The volatility tends to be greatest for regions with the smallest populations, since these generally have smaller sample sizes in the LFS. In April 2004, the civilian population of LFS regions varied from about 86,000 (for the Mersey Lyell Statistical Region Sector) to about 500,000 for the largest LFS region (the Inner Eastern Melbourne Statistical Region). It would be difficult to estimate the seasonal factors for many of the LFS regions due to the volatility of the data, so seasonally adjusted data are not available for LFS regions. The ABS does not currently produce trend data for LFS regions.
The level of volatility means that little weight should be given to short-term movements. However, the LFS estimates for regions can be used to give an indication of longer-term trends. There are some simple methods that can be used to reduce the volatility, however the methods do have some disadvantages. The advantages and disadvantages of these methods are discussed in detail in A Guide to Interpreting Time Series - Monitoring Trends (cat. no. 1349.0).
One method is to compare an estimate for a period with the estimate for the corresponding period of the previous year, using original estimates. This performs a crude seasonal adjustment, because the comparison of like months reduces the impact of seasonal effects, although it will not allow for evolving seasonality. This measure will still reflect the variability of the estimates, and calculating year apart growth will rarely detect the timing of turning points accurately.
An alternative is to use a 13-term symmetrical moving average. This is quite easy to calculate in a spreadsheet. The moving average smoothes the series, but it also reduces major peaks and troughs. Another disadvantage is that this method does not provide a result for the final six observations in a time series (as shown in graphs 1 and 2, below). Appropriate asymmetric moving averages could be used to calculate smoothed estimates for these observations.
The Inner Sydney Statistical Region is a mid-sized (in terms of population, rather than geographical area) LFS region in New South Wales. It has been chosen to illustrate some of the issues when using labour force estimates for LFS regions.
Graph 1 compares the estimates of employment for the Inner Sydney Statistical Region (SR) with a 13-term symmetrical average of that series. In order to minimise phase shift (where the turning points in the averaged series are moved away from the turning points in the input series), the value of each average is placed on the central observation (i.e. the seventh observation of the 13-term span). The average has been calculated by adding 1/24th of the first and thirteenth observations to 1/12th of the intervening 11 observations (i.e. the observations have differing weights). This is a crude trend procedure. Ideally, it should be applied to seasonally adjusted data, but these are not available for LFS regions. As a result, the trends in graphs 1 and 2 may not reflect the true underlying direction of the series.
Graph 1, Inner Sydney SR - Employed: Original
The number of employed people in the Inner Sydney SR (and other LFS regions) is much greater than the number of unemployed people. Within the results of a sample survey, the larger an estimate is, the smaller the relative standard error. Therefore, the relative standard error is much lower for estimates of employed people than it is for unemployed people.
The relative standard error for monthly estimates of employed people in the Inner Sydney SR varied between 4% and 5% over the period April 1999 to April 2004. During the same period, the relative standard error for estimates of unemployed people in the Inner Sydney SR varied from 14% (on the largest estimate of 13,100 unemployed people, in April 2002) to 23% (on the smallest estimate of 4,700 unemployed people, in June 1999).
Between April 1999 and April 2004, the 13-term symmetrical average estimates for the unemployment rate were lower for the Inner Sydney SR than for New South Wales as a whole (graph 2).
Graph 2, Inner Sydney SR and NSW - Unemployment rate: Original
The estimates for the Inner Sydney SR unemployment rate vary considerably from month to month. Between April 1999 and April 2004 the largest monthly change in the estimated unemployment rate for the Inner Sydney SR was between January 2001 and February 2001, when the estimated unemployment rate doubled, rising 2.7 percentage points, from 2.7% to 5.4%. In contrast, the change for New South Wales for this period was an increase of 0.7 percentage points, from 5.9% in January 2001 to 6.6% in February 2001.
The LFS is designed primarily to provide robust national, state and territory estimates of the civilian labour force. LFS estimates are produced for 77 regions. These estimates can vary considerably from month to month. The smaller the region, the smaller the estimates and the greater the influence of very small sample size. Caution is needed when interpreting estimates for regions. The focus should be on trends rather than month to month movements.
A number of datacubes containing information for LFS regions are available in Labour Force, Australia, Detailed Electronic Delivery (cat. no. 6291.0.55.001).
For further information on estimates for LFS regions, please contact the Assistant Director, Labour Force Estimates, on Canberra 02 6252 6565.
For email enquiries, please contact Client Services on email@example.com.