Page 63 - Special Topic Session (STS) - Volume 4

STS563 Pete Jones
2. Methodology
ONS have started testing a ‘master wave’ LMS/PCS survey that has the dual
purpose of assessing coverage of administrative data and collecting labour
market data. The proposal to merge labour market questions with population
coverage is desirable based on similarities underpinning the collections,
including sample size requirements for social surveys and population
estimation, the use of mixed mode collection and the inclusion of questions
that enumerate all household members.
While merging the two surveys will greatly reduce the operational costs of
running two large surveys, certain aspects of the integrated design need to be
developed and tested. To measure overall feasibility, ONS undertook an
operational test for three versions of an online survey in Autumn 2017. Tranche
1 of the test was designed to measure take up rate to an online version of the
new LMS, using different incentives. No additional follow up modes were
pursued with non-responding households that were invited to take part
online. Tranche 2 of the test was designed as a standalone PCS covering
questions typically used in previous CCS questionnaires. Non-responding
households to the online survey were followed up with face to face
interviewing and the option to take part in a telephone or postal survey.
Tranche 3 of the test was designed as an integrated household survey,
merging questions from both the PCS and the LMS. Insights from previous
cognitive testing was used to harmonise questions between the two surveys,
and a similar mixed mode approach was adopted as that used in tranche 2.
The results of the 2017 test are summarised in section 3.
Between November 2018 and April 2019, data has been collected to
support a statistical test comparing LMS and current LFS outputs. This data
has been collected using a similar version of the integrated survey described
in tranche 3 of the 2017 test, and results will be available later in 2019. In May
2019, a first attempt at using administrative data to support the integrated
sample design is being tested in the field, with the aim of increasing the
number of contacts with migrant households. The results of this test will be
used to determine whether a migrant boost should be incorporated into the
LMS/PCS survey to support transformation of migration statistics.
A collaboration between ONS and University of Southampton has
commenced to develop an integrated sample design. The current LFS is based
on equal probability sampling, whereas it’s assumed that a coverage survey to
support administrative data population estimates will require a stratified
design to adjust for varying levels of coverage error across geographic areas
and age-sex groups. Using derived variables from the SPD we have developed
a first iteration of a stratification index to predict over-coverage and under-
coverage propensities for small areas. To identify the optimum number of
strata required to measure under-coverage and over-coverage we have used
the Delanius-Hodges method of minimum variance stratification . We are in
7
52 | I S I W S C 2 0 1 9