CENSUS AND THE LABOUR FORCE SURVEY
COMPARISON OF LABOUR FORCE STATUS
The Census and Labour Force Survey (LFS) collect information about the labour market activity of people aged 15 years and over. However, differences in scope and methodology mean that estimates produced from these collections may not be the same.
The Census of Population and Housing is the largest statistical collection undertaken by the ABS. It aims to accurately measure the number of people in Australia on Census Night, their key characteristics, and the dwellings in which they live. On 8 August 2006, over 30,000 collectors delivered and collected Census forms from approximately 8.5 million households.
Due to varying complexity and processing requirements, 2006 Census data was released in stages. The first release was in June 2007 and consisted of demographic data such as age, sex, country of birth, marital status, and Indigenous status, among others. The second release, in September 2007, consisted of customised tables containing many of the more complex data items, including the following labour market items:
- Labour Force Status
- Hours Worked
- Employment Type
- Industry of Employment
- Number of Employees
- Government/Non-government Employer Indicator
- Community Development Employment Projects Participation
- Place of Work
- Method of Travel to Work.
An extensive range of Census data is available free online at <www.abs.gov.au/Census>. Further data from the second release will be available online from 25 October, 2007.
Many of the labour market variables collected in the Census attempt to measure the same concepts as those collected on a regular basis in ABS household surveys, such as the monthly LFS. Labour force status, hours worked, and employment type, for example, are collected monthly in the LFS, while industry and occupation data are collected quarterly (in February, May, August, and November).
As the Census and LFS collect similar labour force data for a similar reference period (i.e. August 2006), it could be expected that the data from the Census would align very closely with that from the August LFS. However, there are a number of factors which may lead to differences between the two sources.
This paper outlines some of those factors and compares Census and LFS data.
The Census and the LFS
The Census aims to collect information about every person in Australia on Census Night. This makes the Census a good source of information for very small population groups and small areas.
The LFS is a sample survey which has been conducted since 1960. The survey was run on a quarterly basis up to February 1978 and has been conducted monthly since then. The main purpose of the monthly LFS is to provide official estimates of the labour market activity of the usually resident civilian population of Australia aged 15 years and over. The current sample size for the LFS is approximately 60,000 people, covering around 1 in 224 (0.45%) of the population 15 years and over. A sample of this size provides reliable estimates of the key labour force statistics for the whole of Australia as well as each state and territory.
COMPARING CENSUS AND LFS
This section explains some of the main methodological and statistical differences between the Census and the LFS. While it is not possible to quantify the impact of each issue individually, it is likely that they account for much of the difference between the Census and LFS data.
The scope of any collection, be it the Census or a household sample survey, is defined as the population to which the data refer. While the scope of the Census is very similar to that of the LFS, there are slight differences which can have an impact on the data.
The LFS scope is limited to civilians. Members of the permanent defence forces are therefore excluded from the survey. The Census, on the other hand, includes members of the defence forces. This has the effect of increasing the number and proportion of employed people in the Census relative to the LFS.
The Census includes only those people who are in Australia on Census Night, while the LFS includes usual residents who are out of the country for less than 6 weeks. The ABS uses overseas arrivals and departures data to estimate the number of people overseas on Census Night in calculating the estimated resident population of Australia (ERP), but no adjustment is made to the Census data based on these estimates.
While the Census aims to capture information about every person in Australia on Census Night, it is inevitable that in such a large operation some people will be missed. To account for this, the ABS runs a sample survey a few weeks after the Census, known as the Post Enumeration Survey (PES), to estimate the number and characteristics of people who were missed.
On the basis of the 2006 PES, the Census net undercount rate was estimated at around 2.7%. That is, the Census count was estimated to be around 550,000 persons fewer than the actual number of usual residents who were in Australia on Census Night. Results from the PES are used in calculating ERP, but are not used to adjust Census data.
By contrast, as a sample survey, data from the LFS is weighted to independent population benchmarks (end note 1) based on the ERP (which has been adjusted for Census underenumeration). This means that LFS estimates add up to an independently estimated distribution of the usually resident civilian population aged 15 years and over regardless of any sample lost due to non-responding households.
For various reasons, a proportion of households that are occupied on Census Night either cannot or do not return a Census form. The non-response rate in the 2006 Census was 4.2%. To account for this, the number of persons in non-responding households is imputed, along with some demographic characteristics. Labour force characteristics, however, are not imputed. Imputed records are included in the 'not stated' category for labour force status, and in the 'not applicable' category for the remaining labour force variables.
The LFS, on the other hand, does not include non-responses, as only fully responding records (end note 2) contribute to the estimates. To account for non-responding households, data from the fully responding records are weighted to independent population benchmarks based on the (ERP). The population benchmarks currently used for weighting LFS data are based on data from the 2001 Census.(end note 3)
The Census uses a self-enumeration methodology where the householders are required to complete the Census form themselves. The LFS, by contrast, is conducted by trained ABS interviewers, either face-to-face or via telephone. Any responsible adult can provide responses on behalf of all members of the household.
Both methodologies can lead to (non-sampling) error if the respondent doesn't understand a particular question. However, this is much less likely to occur in the LFS as interviewers are able to clarify the meaning of questions which respondents have difficulty interpreting. Interviewers can also probe for additional information if a respondent gives a vague or ambiguous answer. An LFS interviewer may, for example, ask for additional information about a respondent's occupation or industry if their initial answer is not sufficiently detailed. This, of course, is not possible in the Census.
Census data are also affected when respondents don't answer all the questions applicable to them. This may occur when a respondent chooses not to answer a particular question, or when a respondent fails to follow the correct sequencing (i.e. order of questions). Sequencing errors are not possible in the LFS as interviewers record responses via notebook computers which proceed automatically to the next applicable question.
To minimise errors arising due to questions being misunderstood, the Census form is designed to ensure that questions are simple and easy to understand. Limits are placed on the number of questions asked as well as the complexity of the questions and the question sequencing. However, limiting the number and complexity of questions can result in some relevant information not being collected.
For example, in the 2006 Census, labour force status is derived from responses to four questions. In the LFS, labour force status is derived from up to 30 questions. Many of the additional questions in the LFS concern the circumstances where a person can be defined as 'employed' even though they were not at work during the reference week. The omission of these additional questions from the Census may result in employed people being categorised as either unemployed or not in the labour force. It is also possible that people who are unemployed or not in the labour force will be categorised as employed due to the shorter question module used in the Census.
All of the labour force questions in the 2006 Census relate to the week prior to Census Night (i.e. July 31- August 6). In contrast, LFS interviews are conducted over a two week period, with the survey reference period being the week prior to the week in which the survey is conducted. In August 2006, approximately 65% of the LFS sample was surveyed during the week of Census Night - meaning the reference period for the Census and the LFS was the same. The remaining 35% of the LFS sample was surveyed in the week following Census Night - meaning that the reference week was different to that of the Census.
Given the dynamic nature of the labour market, it is likely that some people's circumstances changed from the Census reference week, to the second LFS reference week. The number of hours people work, for example, can fluctuate from week to week. And in any given week many people around the country will start a new job, or a new business, or retire from the labour force.
Estimates derived from the LFS are based on information obtained from a sample of dwellings. Because the entire population is not sampled, the estimates are subject to sampling error. LFS estimates are therefore accompanied by information about the quality of the estimates, including relative standard errors (end note 4) (RSEs) and confidence intervals.(end note 5) While the sample is designed to ensure that the sampling error is as low as possible at the national and state/territory level, it can be high for small area data or for detailed population breakdowns. The Census, by contrast, is not subject to sampling error because the aim is to collect information from all people in Australia.
COMPARING THE DATA
There are advantages and disadvantages to using either Census or LFS data which should be considered when determining which source is appropriate for a particular purpose. The main advantage of the LFS is that data are available on a monthly basis (or quarterly for some items). Most LFS data are available in original, seasonally adjusted, and trend series. Seasonal adjustment is able to remove the effect of events which occur at the same time every year (e.g. Christmas), while the trend series reduces the impact of the irregular component of the seasonally adjusted series. Special care should be taken in interpreting data from the most recent months and quarters as some of the original and all of the trend estimates are subject to revision.
The main advantage of the Census is that data can be used to determine the labour force characteristics of small population groups and small geographic areas. The Census also allows analysis of industry and occupation data at a much greater level of detail than the LFS.
The following section compares Census data on five key labour market variables with original data from the August 2006 LFS. The variables compared are labour force status, hours worked, employment type, industry of employment and occupation. The differences between Census and LFS data for each variable will be discussed and, where appropriate, related back to the methodological and statistical differences outlined above.
Labour Force Status
As shown in the table below, both the Census and LFS indicate that in August 2006 around three in five people aged 15 years and over were employed, around 3% were unemployed, and about a third were not in the labour force.
However, the table also shows that there were considerable differences between the Census and LFS data. For example:
- the Census count of usual residents aged 15 years and over (15,918,100) was 664,500 less than the LFS estimate (16,582,600)
- the LFS estimate of employed people, particularly full-time employed, was much higher than the Census count
- the LFS estimate of people not in the labour force was appreciably higher than the Census count
- the Census count of unemployed people looking for part-time work was much higher than the LFS estimate.
There are a number of explanations for these differences. As outlined above, the Census undercount, estimated at around 2.7%, leads to an underestimate of the number of people aged 15 years and over. In contrast, the LFS totals are weighted to reflect independent population benchmarks and are therefore not affected by sample loss. Similarly, there are differences in the scope of the respective collections that may account for different totals.
The numbers in each of the Census categories are affected by the fact that the labour force status of around 6.5% (1,039,000) people counted in the Census was 'not stated'. Analysis shows that people aged 65 years and over account for around 26% of those whose labour force status was 'not stated', but only 16% of all people counted in the Census. As the vast majority of people aged 65 and over are not in the labour force, this suggests that the proportion of people not in the labour force may be underestimated in the Census.
Of those people counted in the Census whose labour force status was determined, 61.2% were categorised as employed, almost identical to the estimated proportion of people employed from the LFS (61.3%). Likewise, the proportion of people categorised as unemployed and not in the labour force were also quite similar. This may suggest that, despite the over-representation of people aged 65 and over, the characteristics of those whose labour force status was 'not stated' were not substantially different to those whose labour force status was stated.
Some of the differences between the two sources were also due to the categories used. For example, there were around 600,000 people categorised in the Census as 'employed, away from work'. This is not a category that features in the LFS as people in this situation are classified as either employed full-time or part-time based on their usual hours worked.
1. Labour Force Status
LFS, August 2006
|Labour force |
|Employed total |
|Employed, worked full-time |
|Employed, worked part-time |
|Employed, away from work |
|Unemployed total |
|Unemployed, looking for full-time work |
|Unemployed, looking for part-time work |
|Not in the labour force |
|. . not applicable |
|(a) In the LFS, people employed, but away from work are categorised as either employed full-time or part-time based on usual hours worked. |
|(b) Population aged 15 years and over. |
|(c) The number of people whose labour force status was 'not stated' in the Census (1,039,000) has been removed from the Census total of 15,918,100 to aid comparison. |
|(d) Civilian population aged 15 years and over. |
The remaining Census labour market variables to be discussed are only applicable to the 9,104,200 people whose Census labour force status was 'employed'. It is likely that many of the people whose Census labour force status was 'not stated' were actually employed. Therefore, the number of people employed is likely to have been underestimated in the Census. For this reason, comparisons between LFS and Census data for the remaining variables will focus on the proportions of total employed, rather than the total numbers of people.
The hours worked data in the Census refers to hours worked in the reference week and are therefore compared with the LFS data item 'hours actually worked', rather than 'usual hours worked'.
As the following table shows, a slightly higher proportion of employed people reported working between 1 and 34 hours in the LFS (34%) compared with the Census (30%). A higher proportion of those in the Census reported working 35-39 or 40 hours (36%) compared with the LFS (31%), while the proportion working more than 40 hours was roughly similar for both the Census and LFS.
A higher proportion of employed people in the LFS worked 0 hours (5.2%) compared with the Census (3.8%). It may be that some people who were employed but worked 0 hours in the reference week were incorrectly classified in the Census as either unemployed or not in the labour force.
A higher proportion of employed people in the Census indicated that they worked 35-39 or 40 hours, compared with the LFS. This may be due to a tendency for people in the Census (self-enumeration) to 'round' hours worked to their perception of full-time hours (e.g. 38, or 40 hours), while more precise estimates may be provided when responding to LFS interviewers.
Another possibility is that the 2.8% (253,600) of those categorised as 'employed' whose hours were 'not stated' in the Census were more likely to work less than 35 hours, thereby inflating the proportion of people in the 35-39 and 40 hours categories. However, despite the minor differences in the 35-39 and 40 hours categories, the distribution of hours worked for those stated in the Census closely aligns with the distribution in the LFS. It is therefore likely that the characteristics of those whose hours worked was 'not stated' were not substantially different to those whose hours worked was stated.
2. Hours Worked
LFS, August 2006
|0 hours |
|40 hours |
|(a) The number of people whose hours worked was 'not stated' in the Census (253,600) has been removed from the Census total of 9,104,200 to aid comparison. |
As shown in the table below, both the Census and LFS data indicate that around four in five employed people can be categorised as employees not owning their own business. The rest of the employed population is made up of owner managers of incorporated and unincorporated enterprises, and a small number of contributing family workers.
The proportion of employed people categorised as employees not owning their own business and owner managers of incorporated enterprises were similar for the Census and LFS. The proportion of employed people who were owner managers of unincorporated enterprises in the LFS was considerably larger than the Census count (12.5% and 9.7% respectively). There was also a considerable difference in the proportion of people categorised as contributing family workers. The Census indicates that 1.7% of all employed people were contributing family workers, compared with just 0.3% in the LFS. These differences may be due to a number of factors. It may be, for example, that the LFS questions, and the presence of the interviewer, allow people to be categorised more accurately than in the Census.
It should also be noted that the LFS data quoted in this article are based on an estimation method, known as composite estimation, which was introduced in June 2007. At this time, all LFS estimates back to April 2001 were revised on the basis of the new estimation method. In addition, data from the Survey of Employee Earnings, Benefits and Trade Union Membership (EEBTUM) are also used in calculating employment type estimates, and these have not been revised on the basis on composite estimation. This has a minor effect on the LFS data and may account for some of the difference between Census and LFS estimates.
The Census data is also affected by the fact that the employment type of around 1% (86,700) of those who were categorised as 'employed' was 'not stated'. However, apart from the differences with respect to owner managers of unincorporated enterprises and contributing family workers, the distribution of employment type for those who were stated in the Census is largely similar to the distribution in the LFS. This suggests it is likely that the characteristics of those whose employment type was 'not stated' were not substantially different to those whose employment type was stated.
3. Employment type
LFS, August 2006
|Total Employees |
|Employee not owning business |
|Owner managers of incorporated enterprises |
|Owner managers of unincorporated enterprises |
|Contributing family workers |
|(a) The number of people whose employment type was 'not stated' in the Census (86,700) has been removed from the Census total of 9,104,200 to aid comparison. |
|(b) Data on Employment Type from the Labour Force Survey was calculated prior to the introduction of composite estimation. Therefore the total will not equal the other LFS totals presented in this article. |
Industry of Employment
As shown in the following table, across both the Census and the LFS, the industries with the highest proportion of employed people were Retail trade, followed by Health care and social assistance, and Manufacturing.
The industry of employment of 2.6% (236,500) of employed people could not be determined in the Census as it was either ‘not stated’ or ‘inadequately described'. However, of those whose industry was determined in the Census, the proportion of people in each of the industry divisions was quite similar to the LFS.
LFS, August 2006
|Agriculture, forestry and fishing |
|Electricity, gas, water and waste services |
|Wholesale trade |
|Retail trade |
|Accommodation and food services |
|Transport, postal and warehousing |
|Information media and telecommunications |
|Financial and insurance services |
|Rental, hiring and real estate services |
|Professional, scientific and technical services |
|Administrative and support services |
|Public administration and safety |
|Education and training |
|Health care and social assistance |
|Arts and recreation services |
|Other services |
|(a) Classified according to the Australian and New Zealand Standard Industrial Classification (ANZSIC), 2006 (cat. no. 1292.0). |
|(b) The number of people whose industry of employment was either 'not stated' (123,000) or 'inadequately described' (113,400) in the Census has been removed from the Census total of 9,104,200 to aid comparison. |
Occupation of Employment
As the table below shows, the most common occupations as measured by both the Census and the LFS were Professionals, followed by Clerical and administrative workers, and Technicians and trades workers.
The occupation of 1.8% (165,600) of employed people could not be determined in the Census as it was either ‘not stated’ or ‘inadequately described'. However, of those whose occupation was determined in the Census, the proportion of people in each of the major occupation groups was quite similar to the LFS.
LFS, August 2006
|Technicians and trades workers |
|Community and personal service workers |
|Clerical and administrative workers |
|Sales workers |
|Machinery operators and drivers |
|(a) Classified according to ANZSCO - Australian and New Zealand Standard Classification of Occupations, First Edition, 2006 (cat. no. 1220.0). |
|(b) The number of people whose occupation was either 'not stated' (74,900) or 'inadequately described' (90,700) in the Census has been removed from the Census total of 9,104,200 to aid comparison. |
ACCESSING LABOUR FORCE SURVEY DATA
The LFS provides the official estimates of employment, unemployment, unemployment rate and labour force participation rate which are published (in original, seasonally adjusted and trend terms) each month in Labour Force, Australia (cat. no. 6202.0). Further data from the LFS is also published in the following publications which are available free online:
- Labour Force, Australia, Spreadsheets (cat. no. 6202.0.55.001)
- Labour Force, Australia, Detailed -- Electronic Delivery (cat. no. 6291.0.55.001)
- Labour Force, Australia: Labour Force Status and other Characteristics of Families -- Electronic delivery (cat. no. 6224.0.55.001)
- Labour Force, Australia, Detailed, Quarterly (cat. no. 6291.0.55.003).
1. Labour Force Survey estimates of persons employed, unemployed and not in the labour force are calculated in such a way as to add up to an independently estimated distribution of the usually resident civilian population aged 15 years and over. The independent population estimates (benchmarks) are the latest available estimates at the time the Labour Force Survey is conducted, but they usually differ from the official population estimates subsequently published in Australian Demographic Statistics (cat. no. 3101.0) because they are derived from incomplete information about population changes.Back
2. The Labour Force Survey receives a high level of co-operation from individuals in selected dwellings, with an average response rate of around 95% over the year to August 2006.Back
3. The population benchmarks used in determining LFS estimates will be rebased in February 2009 on the basis of data from the 2006 Census. The revised population benchmarks will then be applied to historical LFS data prior to February 2009. It is expected that this will change some of the LFS estimates for this period.Back
4. The relative standard error is the standard error expressed as a percentage of the estimate to which it refers, and is useful when comparing the variability of population estimates of different sizes.Back
5. A confidence interval is a range, centred on the estimate, with a prescribed level of probability that it includes the true population value.Back