6602.0 - Microdata: Longitudinal Labour Force, Australia, 2008-10

Quality Declaration

ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 10/12/2012 First Issue

Page tools: Print

Print Page Print all pages in this product

Print All

Contents >> Weighting and Benchmarks >> Population count [POPCNTC]

WEIGHTING AND BENCHMARKS

<< Previous Section: Retirement

Next Section: Conditions of Use >>

POPULATION COUNT

This data item provides the population count of the benchmarking group that a person belongs to for the current month. This allows customised weights to be calculated in order to create estimates relating to the Australian population, rather than the numbers of people in the sample.

These population counts are provided as discrete variables (whole numbers, no decimal places). The counts range from 1400 to 40 000, depending on the group. They are based on counts of the civilian population aged 15 years and over, as derived from the Estimated Resident Population (ERP) published in Australian Demographic Statistics (cat. no. 3101.0).

The population counts are broken down into the following benchmarking groups:

Sex [SEX] by
Age group [AGECB] by
State or territory of usual residence [STATURCE]

These population counts are the same as those used to produce the estimates in the main Labour Force, Australia publications (cat. no. 6202.0, 6291.0.55.001 and 6291.0.55.003), however, they are not broken down by Sub-state region of usual residence (see Weight for the current month [WEIGHT]). Furthermore, the population counts are not related to the benchmarks or estimation processes used in the supplementary surveys - these are based on different population exclusions and estimation methods.

For more information, refer to Population benchmarks and Estimation method in Survey Methodology and Weighting in File Structure.

For a respondent who is in the labour force survey for the full 8 months cycle, they will have a corresponding population count for their group in each month. These can be used to recalculate simplified single month cross-sectional weights that don't involve the sub-state regional benchmarking groups or the composite estimation techniques used in the Labour Force Survey (LFS). Estimates produced using these population counts will not exactly match official published estimates, but they will be at similar levels.

Image: Shows how a single month cross-section has all 8 panels unbroken across the single month span

The population counts are mainly provided to facilitate the re-calculation of customised weights for different longitudinal time spans, ranging from 2 to 8 months. For example, when considering the 3 month span between January 2009 and March 2009, 2 of the 8 panels are rotated out of the sample and replaced with 2 new groups, leaving only 6 of the 8 panels unbroken across the 3 month span. Furthermore, people who move out of dwellings in sample and people responding from special dwellings are also lost from the common sample across the span. If the Weight [WEIGHT] is used to produce estimates for this time span, they won't add up to represent the Australian population, as over one quarter of the sample is missing. Refer to Longitudinal structure for more information.

Image: Shows how a 3 month spane has only 6 panels unbroken across the span

By using the population counts provided, new weights for each of the remaining records in the common sample can be recalculated. Typically, the count for February 2009 (the middle month) would be representative for the time span, although the counts for January or March would also be equally valid, depending on the analysis (an average of the 3 months could also be used, but this is likely to closely match the February counts). The recalculated weights would be larger than the cross-sectional ('point-in-time') weights provided, as they need to be inflated to compensate for the lost sample. Using these new custom weights, estimates that represent the Australian population can be produced.

However, beware of the increased error associated with the reduced sample of a longitudinal span, and, more importantly, the significant bias in this kind of analysis. People who reside in private dwellings and consistently respond to the survey each month are generally more stably employed. People who drop out of sample, move to new dwellings or respond from special dwellings (particularly prisons, hospitals, nursing homes) are more likely to be NILF or unemployed. When these responses are lost from a longitudinal span and not included in the analysis, there is a significant bias towards employment.



Identifier:		POPCNTC



Level:		Person



Source:		LFS



Frequency:		Monthly



Population:		All



Categories:

XXXXX		Population count