Researchers have been long concerned about the nature and quality of New Zealand ethnic data when undertaking ethnic based health analysis. It is widely known that mortality data (at least during the 1980s and early 1990s) undercounted Māori and Pacific deaths.1-3Current problems include missing data and discrepancies in the way ethnicity is recorded in contributing data sources.4 People may identify themselves with more than one ethnicity and, while individuals may change their ethnicities over time and should have this recorded at each contact with service providers, in practice the ethnicity or ethnicities held against their health records may not accurately reflect their current preferences.Health researchers and epidemiologists continue to struggle with these concerns, particularly in regard to the choice of output strategies to handle multiple responses by people who have more than one ethnicity.Census has recorded multiple responses in a number of ways for many decades, earlier as combinations of ethnicities/ ethnic origins/ races and more recently with up to six ethnicities recorded per individual. Some other collections such as birth and death registrations now record multiple responses routinely in the same manner as done for census, but historic collections frequently restricted their recording to one ethnicity per individual, or, where more than one was collected, the options were only at high level and very restricted.When respondents to surveys are required to record only one ethnic group, reporting membership of ethnic groups is relatively straightforward provided respondents do in fact write a single response, necessitating selection during data processing. When more than one group is collected, then reporting is more complex.In the early period during which more than one group was recorded in New Zealand it was normal to output combinations but then to consider groups based on a half-or-more affiliation basis. This applied until around 1981. However, for the 1986 and 1991 censuses, the then Department of Statistics (now Statistics New Zealand) as well as most government agencies and researchers relied primarily on the prioritisation of ethnic groups in order to simplify the presentation of the data.Under this system, Māori had priority coding, followed by Pacific, then Asian, then other ethnic groups, with people of only European ethnicities last.The one advantage in using a system of prioritisation is that mathematically ethnic counts equal counts of the total population with specified ethnicity, making statistical analysis more straightforward. However, mathematical expedience should be secondary to the meaning of the data. In New Zealand this advantage was greatly outweighed by the disadvantages.The disadvantages are that (1) there is no underlying logic to the order of prioritisation except that it privileges the indigenous population, (2) it is not ethnically neutral (that is, it elevates one ethnic group over another), (3) it does not acknowledge the preferences of people, and (4) it biases population measures by misrepresenting the membership of all groups except for Māori.Moreover, a further complication was that questionnaire design often meant that it was not possible to identify the level of non-response, because the absence of a tick in the targeted tick-boxes was taken to indicate membership of a "non-X" group and then this was treated as though it were some kind of "ethnic group". Hence the correct denominator for calculation of rates was not available.14Increasingly, the biasing effects of both the process of prioritisation and the increase in non-response have become problematic in New Zealand in recent years, with the growth in the number of people, especially children, reporting ethnicities in more than one ethnic group.5In terms of reflecting the preferences of people, two sets of research confirm that imposed prioritisation is problematic. Using 2006 data from wave one of the Youth Connectedness survey of early adolescents, Kukutai and Callister found that three-quarters of youth who recorded more than one ethnic group were able to choose a main group when asked to do so.5Of the dual identified Māori-European children who could choose a main group, more chose a European ethnicity over Māori. Kukutai also found a similar pattern of self-prioritisation responses among women identified as Māori and European in the 1995 New Zealand Women: Family, Education and Employment (NZW: FEE) survey of women aged 20-59 years.6Among the 183 women who recorded Māori and European ethnic groups, 42% were identified mainly as European, 37% identified mainly as Māori, and the remainder could not choose. Table 1. More than one ethnicity reported, Census of Population and Dwellings, 1991-2006 Census year More than one ethnicity reported N %1 Total NZ, all ages Total NZ, 0-14 Māori, all ages Total NZ Total NZ, 0-14 Māori, all ages 1991 1996 2001 2006 166,158 536,757 324,090 400,428 77,172 181,338 145,194 164,262 111,351 249,894 231,552 266,934 5.0 15.52 9.0 10.4 19.3 45.2 34.2 38.1 25.6 47.8 44.0 47.2 1 Percentage of people with a valid ethnic group response. 2 In 1996 an "Other European" tick-box was included, along with a sub-list that specified English, Irish, Australian, Scottish and Dutch ethnic groups. This led to an increase in the reporting of those groups, though at the highest level of ethnicity this had much less effect than this table implies (see Table 2) because people with multiple responses within a group were counted only once in that group. The tick-box was dropped from subsequent census questionnaires. Source: Kukutai and Callister, 2009, drawn from Census of Population and Dwellings. Table 2. More than one Level 1 ethnic group reported, census of population and dwellings, 1991-2006 Census More than one ethnic grouping Numbers Percentages Total 0-14 Māori Total 0-14 Māori 1991 1996 2001 2006 138,171 308,154 282,825 363,402 69,939 135,288 131,580 150,318 111,351 249,894 231,552 266,934 4.1 8.9 7.9 9.4 9.0 16.9 16.0 18.0 25.6 47.7 44.0 47.2 Source: Census of Population and Dwellings. Disquiet with the system of prioritisation was already increasing from the early 1990s as the limitations and distortions to the data were being recognised. Throughout the 1990s, less use was being made of prioritisation in social science research, though it persisted among some economic, some education and most health analysts, largely because data were most readily available in this form. Following the 2004 Review of Ethnicity, Statistics New Zealand recommended that researchers and policy makers no longer use ethnic prioritisation. Despite this recommendation and despite an increasing proportion of health data being available in other formats, many health researchers continue to use the system. While many of the studies using prioritised data are published in overseas journals,7-10 recent examples can be found in this medical journal.11,12 So what effect does the continued use of prioritisation have? While each survey will have different characteristics, the five yearly Census of Population and Dwellings allows some tracking of the effect over time. Table 2 compares the difference in size of each grouping of ethnicities when total responses are used and when the data have been prioritised. For example, in 2006 the Pacific population is 14.9% larger than prioritised data would suggest even though the Pacific grouping has second ranking in the prioritisation system. The Table shows a loss across all age groups for every ethnic group except Māori, which is accorded the highest priority. The difference in numbers is a result of people who gave multiple responses.14 The greatest loss is experienced in the two youngest age groups and is shown to progressively extend into older age groups over time. This reflects the increasing number of children/younger people with multiple ethnicities due to ethnic intermarriage and changes in how people report their ethnic identification.15 More significantly though, this progressive extension shows that the effect has both age group and cohort implications. These data suggest that currently the main distorting effect of prioritisation is for studies that focus on young people, for example rates of child immunisation, but any continuation of a prioritised ethnic frame would have increasing significance for health monitoring of older age groups. Table 3. Percentage understatement of prioritised ethnic counts compared with total responses, 1991, 1996, 2001 and 2006 Censuses Ethnicity Year Age groups Total Under 15 15-19 20-24 25-29 30-34 35-39 40-44 45+ European 1991 1996 2001 2006 11.6 24.4 24.0 21.4 8.1 19.4 17.2 16.6 5.4 14.7 15.4 14.1 4.4 11.8 12.4 13.0 3.5 10.1 9.1 10.4 2.8 8.2 7.4 8.2 2.0 6.6 5.9 6.8 1.1 3.3 2.6 3.2 4.7 11.0 10.0 10.1 Māori 1991 1996 2001 2006 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Pacific 1991 1996 2001 2006 18.4 30.0 29.5 24.7 9.5 20.9 18.5 17.6 5.0 12.8 14.4 13.7 4.4 8.7 9.1 11.6 4.1 7.8 6.4 8.1 2.5 7.8 6.4 5.8 1.7 5.6 5.8 5.8 1.0 4.4 2.6 3.3 9.2 16.8 15.8 14.9 Asian 1991 1996 2001 2006 10.7 13.3 10.5 8.4 9.6 8.5 4.5 4.4 6.3 9.2 4.8 2.4 3.8 8.2 5.2 3.0 2.7 5.3 4.1 3.1 2.9 4.2 2.7 2.7 3.6 4.0 2.4 1.9 3.1 4.7 2.5 2.0 6.1 8.0 5.1 3.9 MELAA 1991 1996 2001 2006 13.9 19.8 14.4 11.7 9.9 16.9 8.1 6.2 5.1 12.6 7.8 4.4 4.9 8.1 5.5 2.8 3.0 6.7 3.4 2.2 2.3 8.9 4.5 2.2 1.2 7.4 4.1 2.3 2.6 5.3 3.8 2.8 6.5 12.0 7.7 5.6 Note: For consistency, "Other" (predominantly New Zealander responses) has been included with European for 2006 Census and the group labelled "Other" prior to 2006 is here labelled "MELAAR