Sampling the 2001 SARs

The 2001 SARs were sampled from the One Number Census database for the entire UK. Thus individuals and households imputed as part of the ONC are included in the SARs.

Sampling was done within each of the 112 ONC design groups for the UK. Design groups represent groups of LADs, each with a population of approximately half a million people, which form an essential part of the ONC process. The CCS sampling was applied within each of the design groups to enable accurate direct estimates of under enumeration for 37 age-sex groups at the design group level.

There are 112 design groups in the UK (101 in England and Wales, eight in Scotland and three in Northern Ireland).

The sampling scheme for the household SAR is a stratified simple random sampling, where the strata are Enumeration Districts (EDs) (there is a very minor departure from simple random sampling in the sense that the sample size within EDs may vary randomly between two adjacent integers because of the random start from 1 to 10). Unlike 1991, there is no stratification within EDs. Random sampling is applied within each ED.

The sampling scheme for the individual SAR follows the 1991 approach of drawing from the population excluding the household sample. Stratification is again by ED. The Individual SAR sampled both private and communal persons, unlike the household SAR which only sampled only households.

As for the 1991 sample, variables in the SAR files show the effects of both stratification and clustering. Attributes that tend to be common across areas will be affected by stratification (for example, local authority housing tenure) and will therefore have a lower sampling error than that for a simple random sample. Other variables, where values tend to be the same for all household members, will be affected by clustering, leading to larger than expected sampling errors. For example individuals within the same household are likely to have the same ethnic group and social class. The effect of clustering is more pronounced for individual level variables in the household file, as all individuals in each household are selected for this sample.

There is no overlap between any of the SAR files. All records have been scrambled before release to ensure they cannot be identified geographically.

Last updated 25 October 2004