Estimation of a derived income variable for use with the LS analysis database
Karl Ashworth and Salah Merad, Office for National Statistics
[Project number 20087]
This proposal is to implement and provisionally 'field test' an imputed individual income variable in the LS. The methodology for imputing income from a national survey using multiple regression techniques has already been developed. The proposed project will produce:
a) One or more derived variables for 2001, which after testing will be
made available to all LS users as an experimental variable.
b) Documentation for these variables.
The income estimates developed require alpha testing, and, subject to resource availability, a further proposal for a beta test will follow. The model algorithm will be applied to each LS member in 2001 to impute an income variable and alpha testing criteria evaluated first. Criteria are identified below:
- Model is robust.
- Donor survey and 2001 LS census record variables are comparable in definition and categorisation.
- Income inequality makes sense for time at which it is derived.
- Income level plausible for each 2001 census sociodemographic variable included in the model.