Page 327 - Special Topic Session (STS) - Volume 3

STS547 Maarten C. et al.
in only one register (item missingness). The linked data are used to estimate
the size of the population that is not present in either register, and for these
unobserved records all the variables are missing (unit missingness). There is
an extensive line of research that studies this problem from a missing data
perspective, starting with Zwane and van der Heijden (2007), and summarized
in van der Heijden et al. (2018). The latter paper concluded that further
practical experience with these methods is needed to demonstrate their
usefulness in a variety of situations and encourage their wider application.
Here we consider the methods for estimating the size of the Māori
population in New Zealand. Ethnicity is the principal measure of cultural
identity in New Zealand, and is used across the official statistics system. The
2005 New Zealand statistical standard for ethnicity states that ethnicity is self-
perceived and a person can belong to more than one ethnic group. Identifying
the indigenous Māori population is of particular importance.
Ethnicity is regularly included in data collections because of its importance
in defining groups of policy interest, for example on health outcomes for
indigenous people in New Zealand. However, differences in questions,
differences in self-perception depending on the context, and changes over
time, can all affect how ethnicity is recorded in these data sources (Statistics
New Zealand 2005). Ethnicity is collected independently in a number of
administrative sources as well as through the census and household surveys.
People do not always report the same ethnicity in each source. Also, people
do not always report their ethnicity, so there is an additional missingness
problem to deal with.
Official population estimates and projections for major ethnic groups in
New Zealand are based principally on the responses people provide in the five
yearly census, adjusted for non-response using a post-enumeration survey. As
part of its census transformation programme, Statistics NZ is exploring the
feasibility of a census based on administrative data (Statistics New Zealand,
2012, 2014). The ability to produce ethnicity data from administrative sources
is a key consideration. Using ethnicity information from linked administrative
data sources may also improve the current production of official ethnic
population estimates.
The aim is to use ethnicity information from linked administrative data to
improve official ethnic population estimates in New Zealand. In support of this
we analyse a variety of census and administrative sources using the approach
of Zwane and van der Heijden (2007), with a specific focus on the estimation
of the size of the Māori population at the time of the 2013 population census.
The analysis requires the extension of the methods to deal with multiple
registers and with a variety of different types of missing data. The
methodology falls within the area of data integration of multi-source statistics,
see de Waal et al. (2017) and Zhang and Chambers (2018).

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