Page 332 - Special Topic Session (STS) - Volume 3

STS547 Maarten C. et al.
Table 1: Census (A) linked to MOH (C). Covariate Ethnicity in A is denoted by
a and ethnicity in C is denoted by c, where a and c have levels ‘0’ (non-Māori),
‘1’ (Māori), ‘-’ (missing) and ‘x’ (not in register). Observed counts have been
randomly rounded to protect confidentiality. Source: Stats NZ.
Census DIA MOH MOE

non-Māori 3,690,913 3,668,349 3,782,239 3,665,099
Māori 736,308 758,872 644,983 762,122

Table 2: Summary of Census linked to DIA, MOH and MOE, estimated numbers

census DIA MOH MOE

=1| =1|

=1|
=1|
Class 1 0.826 0.004 0.012 0.003 0.014
Class 2 0.174 0.939 0.930 0.824 0.924
Table 3: Estimates of latent class model with two latent classes
3.3 Three registers without the Census
We also made estimations for three registers without the Census, see Table
4. We also present estimates derived only from the three administrative data
sources, so that we can see what would happen if the census were replaced
entirely by an administrative data-based system. The observed number of
individuals in at least one of the registers is 4,377,573. We estimate an
additional 24,058 individuals missed by all three registers. This leads to a total
population size of 4,401,631. This is somewhat less than the four register
estimate of 4,427,221.
DIA MOH MOE
non-Māori 3,599,611 3,760,211 3,625,453
Māori 802,020 641,421 776,179
Table 4: Summary of DIA, MOH and MOE, ignoring census, estimated numbers.

4. Discussion and Conclusion
Van der Heijden et al. (2018) presented an approach for estimating the
margins of auxiliary variables in the dual system estimation framework. They
suggested that more experience with applications of this methodology was
needed to be able to judge its usefulness. Here this approach is extended to
multiple system estimation with four registers, and a more complicated
missing data structure. We conclude that the methods of van der Heijden et
al. (2018) provide stable results that allow for detailed interpretation of the
processes of inclusion in the registers considered, and of recording Māori
status.

321 | I S I W S C 2 0 1 9