[an error occurred while processing this directive]

SARs Newsletter

Number 19 Insert - March 2003

Adding Geography back to the SAR

Many users need to know about the district-level distributions of population characteristics not captured by planned standard census outputs. For this reason, they have requested district-level coding for the SAR. Unfortunately, as noted in the latest SAR newsletter, sub-regional geographic coding had to be sacrificed in the trade-off between SAR content and respondent confidentiality.

To help compensate for this loss, it is planned to add pseudo-LAD indicators to the SAR. These pseudo-LAD indicators will allow users to quickly and easily extract statistical best estimates of unknown district-level distributions from the SAR. The resulting estimates will normally be at least as accurate as those derived using Iterative Proportional Fitting, a statistical technique commonly adopted in such circumstances.

The proposed pseudo-LAD imputation strategy will identify the set of individuals in the SAR that 'best represent' each SAR district. Specifically, when aggregated at district level, the identified set of individuals will

  1. accurately reproduce a set of a dozen or so published district-level Census tabulations. These tables, acting as constraints on the imputation process, will involve the interactions between a dozen or so Census variables.
  2. provide best estimates for other unknown district-level distributions, provided that they involve interactions between table marginals used in (i).
Full details of the proposed methodology, including a statistical assessment of estimate accuracy, may be found at http://pcwww.liv.ac.uk/~william/microdata.

The initially proposed list of variables from which reliable district-level estimates can be derived are as follows:
Demographic Age
Sex
Marital Status
Household composition (adults/children)
Cultural Ethnic Group
Religion
Economic Economic activity
NS-SEC (2001 version of 'social class')
Cars in household
Housing Dwelling type
Tenure
Rooms in household

Please send suggestions for possible alternative/additional constraining variables to Margaret Martin, CCSR, University of Manchester, M13 9PL, or email: margaret.martin@man.ac.uk

ESRC Contact SARs Support | CCSR
These pages are maintained by the SARs support team.
Send us comments on this web page.