6224.0.55.002 - Improvements to Family Estimates from the Labour Force Survey, 2008  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 17/09/2008  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

CURRENT ESTIMATION METHOD


LFS PERSON-LEVEL PROCESSING

Currently, LFS family estimates are a by-product of LFS person-level estimates. To understand the current method of producing family estimates from the LFS, it is first necessary to understand the method of producing person-level estimates from the LFS.


Scope exclusions

The LFS is designed primarily to measure key characteristics of the civilian labour force in Australia. As a result, certain population groups are excluded from LFS person-level estimates, because they do not fall within the definition of the civilian labour force in Australia. Examples are permanent members of the Australian defence forces, children aged less than 15 years, and overseas diplomats living in Australia. People in these population groups are referred to as 'out on scope'.


Selection rule exclusions

People who fail to meet LFS selection rules are excluded from LFS person-level estimates. Selection rules are designed to associate each person in Australia with one and only one dwelling, to ensure each person has only one chance of selection in the survey. For example, persons away from their usual residence for more than six weeks are excluded from the LFS at their usual residence, but if the dwelling they are visiting is selected in the survey, they are included in the survey at that dwelling.


Exclusions due to partial non-response

People for whom key LFS information is missing are excluded from LFS person-level estimates.


Person weights

As part of LFS person-level processing, each person record which is in scope of the survey, fully responding, and meets LFS selection rules is assigned an expansion factor or 'weight' so that the distribution of the weighted sample reflects the in-scope Australian population in each state and territory and at a regional level. The weights are calculated so that they sum to independent estimates of the civilian population aged 15 years and over (referred to as population benchmarks).

Weighting the sample data to the population benchmarks compensates for people who did not fully respond to the survey. However, the population benchmarks explicitly exclude people who are outside the scope of the LFS.


FAMILY WEIGHTING

To produce family estimates, each family is assigned a family weight, so that the responses from the families in the survey sample can be inflated to represent all families in Australia.

Under the current method, the value of the family weight is the harmonic mean of the person weights of all family members aged 15 years and over.

Estimates for each characteristic of interest are then obtained by summing the weights of the families in the sample with that characteristic.


ISSUES WITH THE CURRENT METHOD

There are two main issues with the current method:

  • Some families in the LFS sample are not assigned a weight, and therefore make no contribution to LFS family estimates.
  • There is no use of independent population benchmarks, such as the Estimated Resident Population (ERP), in assigning the family weight. Therefore, no compensation is made for those families which make no contribution to LFS family estimates.

Further information about these excluded families is provided below.


Families in incomplete households excluded

Under the current method, a family weight is only assigned to families in households where all usual residents of the household aged 15 years and over have contributed to the person-level estimates. Families in households in which one or more members aged 15 years and over have not contributed to the person-level estimates (for example because they are permanent members of the Australian defence forces, did not fully respond to the survey, or failed to meet selection rules) are not assigned a family weight and therefore make no contribution to the estimates. Without the use of independent population benchmarks, no compensation is made for these excluded families.


Families in discrete Indigenous communities excluded

In discrete Indigenous communities, the LFS questionnaire does not collect information about relationships within households, due to the complexity and cost of doing so. Without information about relationships within households, family data cannot be derived from households in discrete Indigenous communities selected in the LFS.

Under the current family estimation method, families in discrete Indigenous communities make no contribution to the estimates, and without the use of independent population benchmarks, no compensation is made for these excluded families.


Effect of excluded families

The combined effect of these exclusions is that the current LFS family estimates do not accurately reflect the Australian population. The exclusion of some families in the LFS sample prior to taking the harmonic mean, without the compensating effects of independent population benchmarks, results in family estimates being lower than they would be if all families were included. Also, the family characteristics of the excluded groups, such as households containing defence force personnel, may be different from the included groups.

The lack of independent benchmarks also means that LFS family estimates are not comparable with other ABS estimates of families. The total number of children, adults and households are not constrained to equal the independent benchmarks, which means LFS family estimates differ from other ABS surveys.


SUMMARY

The current method of producing LFS family estimates excludes:
  • families in households in which one or more members aged 15 years and over have not contributed to the person-level estimates (for example because they are permanent members of the Australian defence forces, did not fully respond to the survey, or failed to meet selection rules); and
  • families in discrete Indigenous communities.

The exclusion of these families, without the compensating effects of independent population benchmarks, results in family estimates being lower than they would be if all families were included.