Page 283 - Invited Paper Session (IPS) - Volume 1
P. 283
IPS153 Christine B. et al.
Admin enumerations in meshblocks
After deriving admin households for non-responding dwellings where
possible, the next phase places admin people into the smallest New Zealand
geographic area (meshblocks). The admin address ID provides an x,y location
coordinate, but the person will be excluded from census households.
The remaining IDI-ERP population eligible for inclusion in meshblocks are
those who have not been linked to a census respondent, and are not already
included at a dwelling. We then explicitly adjust for potential coverage errors
that could result from over-coverage in the IDI-ERP, and for duplication
caused by missing linkages between the census and the IDI. We draw on new
methods Stats NZ have developed to remove over-coverage and adjust for
missed matches in the context of a new Dual System Estimation (DSE)
population estimate based on the same linked census and IDI-ERP
administrative population sources. DSE is a well-established methodology
used in population estimation, often applied in the context of a full
enumeration census and a coverage survey. Over-coverage is typically avoided
by design of the coverage sample, and may be removed from the census, and
extensive checks are made to meet the DSE assumption of no linkage error.
Neither of these approaches are possible in this situation.
Removing over-coverage in the IDI-ERP: Our approach is designed to
remove over-coverage caused by people selected in the IDI-ERP who are not
NZ residents at the time of the census. As we cannot directly find over-
coverage, we will also remove valid records. The goal is to effectively remove
as much over-coverage as possible, while also minimising the amount of
under-coverage being introduced. We apply a more rigorous selection criteria
for inclusion, whereby people must have activity in at least two data sources:
tax or health, plus one other. Tax and health datasets have high coverage of
the population, and requiring an additional activity provides stronger evidence
that people are in fact residents. This approach targets all age-sex groups,
combines factors found to predict over-coverage in previous research, and is
easy to apply and explain.
Accounting for missed linkages: We also adjust for duplicates caused by
missed matches when we link the census file and the IDI spine. If we have not
linked a census record to the IDI spine when we should have, it will still be in
the eligible IDI-ERP dataset, and would be counted again. We are matching
around 4 million census records to an IDI spine consisting of nearly 10 million
records. We do not at present have the time or resources for checking even a
sample to estimate missed matches.
We estimate missed matches on the basis of responses to census questions
that closely match the criteria for inclusion in the IDI spine. A subset of the
census, M*, is created through applying strong requirements of membership in
two or more of the datasets that make up the IDI spine (Choi, 2019). Those in
272 | I S I W S C 2 0 1 9