The Stability of Ethnic Identity and Reporting

Note written for circulation, December 2023

Ethnicity is not a well-defined notion for the majority of the population, but when asked for ‘official’ purposes (usually with a choice of tick boxes) most can ethnically identify themselves.  (Typically, they may check as many ethnicities as they wish.)

In the Population Census data used here there is no definition of ethnicity. I mean by ethnicity a grouping of people who identify with each other on the basis of perceived shared attributes that distinguish them from other groups. Those attributes can include common sets of ancestry, traditions, language, history, society, religion, or social treatment. There can be considerable heterogeneity within an ethnic group. For instance its young and old may have different views on many matters.

Ethnicity is not the same as ‘race’, which refers to genetic ancestry. There may be some overlap. Probably some people confuse the two notions.

The four key ethnicities in New Zealand statistics are ‘European’ (although this paper uses ‘Pakeha’, in order to make it clear that this is not a racial concept), ‘Maori’, ‘Pasifika’ and ‘Asian’ (although this is quite a polyglot since it includes large numbers of Chinese and Indians who would claim to have little in common). The data bases used here also distinguish ‘Middle East-Latin American-African’ (MELAA) and ‘Other’, which are also polyglots.

One’s ethnicity is subjective in the way one’s ancestry is not. You cannot change your ancestors, but you can change your ethnicity. Critical to the use of ethnicity in public discussion is the degree to which an individual’s notion of their ethnicity is stable or whether it changes over time and with changing circumstances. Public discussion assumes that it is largely stable. This paper explores the issue both in terms of self-identification and in the public record.

(The data base on which this discussion is based was supplied by Statistics New Zealand. The definitions and caveats associated with it are set down in an appendix (yet to be written). While kindly supplying the data base, neither Statistics New Zealand as an institution nor any of its people are responsible for the analysis on which the paper is based nor the conclusions drawn from the analysis.)


As far as I know, there is no survey of individuals’ ethnic identity which enables us to monitor how they classify themself on a regular – say, daily or monthly – basis.[1] The best we have available is the self-identified ethnicity in the 2013 and 2018 population censuses.

However, it has been possible to compare the responses for 1.9m people who answered the 2013 and the 2018 censuses. This is only a selection of all the 2013 census respondents. The missing would include those who had died or migrated in the intervening five years and those for whom we have responses for both censuses but whose records could not be matched.

The data was supplied in 11 categories:


Pakeha & Māori

Pakeha, Māori & Pasifika

Pakeha & Pasifika

Pakeha & Asian


Māori & Pasifika





Slightly under 90 percent (89.9%) identified themselves in the same ethnic category in 2018 as they chose in 2013.

If we extend the categorisation to explore those who chose a self-identification of Pakeha in any of the five relevant categories, Māori or Pasifika in any of four, or Asian in any of two we find the successful identification rises to 93.1 percent. (There are some combinations such as Māori -Asian which appear in the data as other.

That means than 3.2 percent modified their ethnicity between 2013 and 2018 but retained at least one of their identifying labels. That also means that almost 7 percent (one-in fourteen) chose a category in 2018 which had nothing in common with their 2013 choice.

The implication of this result is that there is some long-term stability of ethic self-identification but a not insignificant minority change theirs.

Recording Ethnic Classification

Statistics New Zealand also provided a comparable tabulation between the 2018 census ethnic classification and the classification in the following six administrative data bases in the same (2018) year:

            The Statistics New Zealand Administrative Population Census (APC);

            A Ministry of Health data base (MOH);

            A Ministry of Social Development data base (MSD);

            A Ministry of Education (Tertiary) data base (MoET);

            A Department of Internal Affairs data base of children (DIAC);

            A Department of Internal Affairs data base of parents (DIAP).

In this case the administrative categorisation may not be by the individual being classified. Depending on the circumstances it could be by family or a friend, the person administering the record or an outsider.

The comparable ratios are in the following table, with 2013 included.

Matching between 2018 Population Census self-categorisation and categorisation in an Administrative data base.

<>   TOTAL Exact Match Some Match 2013 1921765 0.899 0.931 APC 1590232 0.810 0.870 MoH 3605252 0.923 0.939 MSD 1540228 0.887 0.920 MOET 1908985 0.880 0.914 DIAChild 820744 0.892 0.940 DIAParent 487172 0.928 0.962

In summary, the match proportions are similar to the census matches despite being in the same year. But different administrative data bases show differences. The least successful is the Administrative Population Census.

The matches for the four major ethnic classifications also show inconsistency.

Exact Match between 2018 Population Census self-categorisation and categorisation in an Administrative data base.

<>   Pakeha Māori Pasifika Asian 2013 0.945 0.812 0.933 0.979 APC 0.711 0.760 0.896 0.980 MoH 0.981 0.834 0.943 0.959 MSD 0.972 0.425 0.594 0.916 MOET 0.966 0.835 0.920 0.870 DIAChild 0.951 0.715 0.921 0.976 DIAParent 0.979 0.800 0.956 0.989

Partial Matching between 2018 Population Census self-categorisation and categorisation in an Administrative data base.

<>   Pakeha Māori Pasifika Asian 2013 0.945 0.905 0.946 0.975 APC 0.857 0.887 0.932 0.957 MoH 0.970 0.932 0.928 0.977 MSD 0.975 0.866 0.868 0.985 MOET 0.958 0.934 0.896 0.986 DIAChild 0.960 0.923 0.942 0.941 DIAParent 0.966 0.912 0.960 0.990

The matching seems reasonable for Pakeha and Asia, but much poorer for Māori and Pasifika, especially in the case of exact matching. The APC, MSD and DIA all do poorly.

Notice that the partial matching is better than the exact matching. What is happening, according to the data, in 2018 the MSD classified 15.3 percent of the sole Māori in the 2018 Population Census as sole Pakeha and 36.3 percent as Pakeha and Māori so that it had only 42.5 percent who were sole Māori. (Until we know how the MSD data base is compiled – this is the worst example – it is idle to speculate on the source of the discrepancy. It is not because the MSD population is unrepresentative of the national population; the comparison is between the MSD population and the same census population.)


The above data provides evidence that individuals change their ethnicity sufficiently often to be a potential concern. It also suggests that administrative data bases may be applying definitions or determining ethnicity differently from the self-categorisation approach of the Population Censuses. We should therefore be careful when we are comparing the aggregate ethnicities from administrative data with national proportions. More generally it is unwise to treat everyone’s ethnicity as fixed; for many it is fluid.


[1] The Wellington School of Medicine New Zealand Census Mortality and Cancer Trends Study found there was an imperfect match between Census ethnic responses and those on death certificates. Presumably ethnicity recorded on a death certificate is not a self-categorisation.