In 2000 the relaunched survey introduced a dual weighting and grossing scheme (a full description of this major methodological change can be found in the 2000 report).
Firstly, weighting to compensate for non-response in the sample based on known under-coverage in the Census-linked study of non-response. Secondly, the sample, which has been weighted for non-response, has been grossed up to match known population distributions, as used in the Labour Force Survey, LFS.
During the writing of the 2001 report, the results from the 2001 Census were published. They indicated that previous mid-year estimates of the total UK population were around 900,000 too high, with disparities being most apparent among men aged 25 to 39.
These disparities were larger than expected, so it was decided that the 2001 weighting would need to be revisited post publication as the reporting process was too far advanced for them to be incorporated.
In February 2003, the ONS published revised mid-year population estimates for 1991 to 2000. This brought them into line with the post 2001 Census-based mid-2001 population estimates (hereafter referred to as the 2001-based intermediate population estimates).
In September 2003, there was a relatively small upward revision to the mid-2001 population estimate of men (mainly in those aged 25 to 34 of around 190,000). There have been, and will be, further revisions to some local authority population estimates.
ONS examined the effects on GHS estimates of re-weighting the data using the 2001-based intermediate population estimates. The effects were sufficiently small that it was decided that it was unnecessary to issue revised weighting post-publication.
However, GHS 2001 data with weights derived from the 2001-based intermediate population estimates are available from the UK Data Archive. The substantive chapters of the report present data for 2002/2003 in weighted form only. For further details of the weighting see Appendix D.
Weighted and unweighted bases are shown in the tables in this report. Weighted bases should primarily be considered as bases for the percentages shown rather than estimates of population size. Details of data presentation in report tables can be found in ‘Notes to tables’.
Inclusion of extra households
Each year on the GHS there is a small proportion of responding households where one or more members do not complete the individual questionnaire. For these households there is a complete set of data at household level and a full individual interview for at least one adult in the household.
Before 2001 these households were excluded from the analysis, resulting in the loss of some data each year. Since 2001 these households have been included in the analysis. This increases both the household sample size and the individual sample size.
The non-response weighting was calculated based on a responding sample that excluded these households. ONS methodologists judged that the effect of including these households was negligible since the numbers involved are small (around 240 households).
The middle response rate quoted for 2002/2003 excluded these households to provide direct comparison with previous years. They are included in the overall response rate - see Appendix B for full details.
For further background information about the General Household Survey, please download the Introduction PDF on the right-hand side of the page.
Source: Living in Britain 2002, published 2004
Note: Foster K (1994) The General Household Survey report of the 1991 census-linked study of survey non-respondents. OPCS (unpublished paper)
printer-friendly
About the survey