Given below is a summary of a submission I'm making to the National
Statistics Commission.   Comments would be welcome.  If you would like to
have an electronic copy of the complete document please let me know.

Ray Thomas, Social Sciences, Open University
Tel: 01908 679081 Fax 01908 550401
Email: [log in to unmask]
35 Passmore, Milton Keynes MK6 3DY

*************************************************************

THE USE OF ADMINISTRATIVE RECORDS AS PART OF THE SAMPLING FRAME FOR
GOVERNMENT SURVEYS -  A summary

The Labour Force Survey fails to produce accurate statistics on Claimants of
Job Seekers Allowance and thereby fails to produce statistics for LFS
non-claimants, i.e. those counted as unemployed according to the ILO/LFS
criteria who are not Claimants.    The Family Resources Survey likewise
fails to get proper coverage of JSA Claimants and also fails to get proper
coverage of most other recipients of government benefits.    This submission
argues that these failures should be dealt with through the use of
stratified sampling using administrative records as part of the sampling
frame for both the Labour Force Survey and the Family Resources Survey.

The basic sampling principle is that every member of the target population
should have an equal chance of being selected.  This principle is observed
within each stratum of a stratified sample, and is accommodated in the
aggregate results by multiplying the responses for each stratum by the
appropriate scaling factor.   Stratified sampling is commonly used in
government surveys of businesses and other organisations because of the wide
variation in size.   Use of a larger sampling fraction for large
organisations and a smaller sampling fraction for small organisations
increases overall coverage and reduces overall sampling error.

The prime aim of most large social surveys conducted by the Govenrment
Statistical Service is to obtain a sample representative of the population
as a whole.   But the use of the Postcode Address File as a sampling frame
in the case of both the LFS and FRS is associated with a failure to obtain
coverage of those in receipt of various state benefits.   Statistics on the
characteristics of recipients based on such limited coverage cannot be
assumed to be representative of all of those in receipt in the particular
benefit.

The use of stratified sampling using administrative records of recipients of
state benefits as part of the sampling frame can be expected to deal with
this problem.  Statistics produced on the basis of such a sample drawn from
records of recipients of a particular state benefit can be expected to
representative of all recipients of that benefit.  Such stratification could
be expected to remove such sources of bias and uncertainty deriving from
lack of proper representation of recipients of government benefits in both
the Labour Force Survey and the Family Resources Survey.

*************************************************************