• Open Access

Cross classification of the New Zealand population by ethnicity and deprivation: trends from 1996 to 2006


Correspondence to:
Martin Tobias, Ministry of Health, PO Box 5013, Wellington. Fax: 64 4 816 2340; e-mail: martin_tobias@moh.govt.nz


Objectives: To describe trends in the distribution of New Zealand's major ethnic groups by small area deprivation and trends in the ethnic composition of each deprivation category.

Methods: Data sources were the 1996, 2001 and 2006 New Zealand Census of Population and Dwellings. Ethnicity (Māori, Pacific, Asian or European/Other) was defined using total response output. Each person was assigned a deprivation score by geocoding their usual residence (as recorded in the census) to meshblock level. For each time period (1996, 2001 and 2006) the deprivation score for each meshblock was calculated by principal components analysis from nine socio-economic variables included in the corresponding census (the New Zealand Index of Deprivation).

Results: Throughout the observation period, Māori and Pacific ethnic groups were over-represented at the more deprived and under-represented at the less deprived end of the deprivation spectrum. The European ethnic group displayed less-marked skewing, and in the opposite direction, while the Asian ethnic group showed close to the expected uniform distribution. Neither the deprivation distribution of any ethnic group, nor the ethnic composition of any deprivation decile, showed any statistically significant change over the 10-year observation period.

Conclusions: Monitoring trends in the relative deprivation distributions of the ethnic groups helps assess progress towards social justice. Similarly, monitoring trends in the ethnic compositions of the different deprivation deciles is important in the formulation of social policy. Little change was found in either of these distributions over the relatively short observation period.

In many developed countries small area socio-economic or deprivation indexes have been derived from the national census and are widely used for monitoring social inequalities and for needs-adjusting capitation-based funding formulae.1,2 In New Zealand, the New Zealand Index of Deprivation (NZDep1991, NZDep1996, NZDep2001 and NZDep2006) has been applied at both local and central government level across a range of social domains since the 1990s.3,4 In the health domain the index is a component of the capitation formula used to allocate funds for personal health care services to district health boards (DHBs) and primary health organisations (PHOs), and is used to monitor social inequalities across a wide range of health outcomes including mortality, morbidity, disability and hospitalisation indicators. A relatively large body of literature attests to the validity and reliability of this index, particularly in the health domain.5,6

The NZDep index captures multiple dimensions of deprivation and is constructed by principal components analysis from nine socio-economic variables included in the five yearly Census of Population and Dwellings, each variable being operationalised as the age standardised proportion of the small area population with the relevant characteristic (Table 1). These variables capture eight dimensions of material and social deprivation, reflecting lack of income, employment, communication, transport, social support, educational qualifications, home ownership and living space. Demographic variables (such as ethnicity) are not included.

Table 1.  NZDep variables and weights; 1996, 2001 and 2006 indexes.
Variable nameVariable descriptionaWeight 1996Weight 2001Weight 2006
  1. Note:

  2. (a) age and sex standardised proportion

Income – benefitReceiving a means tested benefit0.3620.3610.371
Income – povertyEquivalised household income below threshold0.3510.3500.356
CommunicationNo access to a telephone (landline or mobile)0.3630.3360.314
TransportNo access to a car0.3480.3320.311
Social supportSingle parent family0.3450.3250.333
Educational qualificationsWithout any qualifications0.3270.3190.326
Owned homeNot living in own home0.2950.3120.334
Living spaceLiving in household below equivalised bedroom occupancy threshold0.2280.3090.318
Proportion of variance explained 57.4 %57.7%55.4%

The index has undergone some changes over time. From 1996, the changes have been minor, mainly involving shifts in the household equivalised income threshold – which is entirely appropriate as it keeps the threshold aligned with other measures of poverty and so will not introduce bias into time series analysis. Other minor changes have included changes in the range of means tested benefits included in the index, and in the age thresholds for these benefits – necessitated by policy shifts and corresponding variation to the relevant census questions. Again, these modifications to the index are needed for consistency over time and will not introduce bias. There was also a small change in the crowding variable between 1996 and 2001 (but not between 2001 and 2006), yet this has only a very small effect on the value of the overall index. However, significant changes in the construction of the index did occur between NZDep 1991 and NZDep 1996 (including dropping two of the 10 variables initially included in 1991 and adding one new variable – telephone access – in 1996). From 1996 onwards, however, population NZDep distributions can be safely assumed to have a nearly consistent meaning. This is confirmed by examination of Table 1: not only are the variables the same since 1996 (although some thresholds differ), but the variable weights (derived by principal components analysis and used to sum the individual variable values to give the final deprivation score) are also very similar, as is the total proportion of variance in the census data explained by the index.

NZDep may be based on two bespoke geographies: meshblocks (neighbourhoods of approximately 60 households or 100 persons on average – the smallest geographic units recognised by Statistics New Zealand) or census area units (CAUs, larger areas of approximately 2,000 persons on average, more or less corresponding to suburbs in urban areas). While CAUs offer some advantages (for example, more stable geographic boundaries over time, and – reflecting their much larger population sizes – more stable deprivation scores), they are also much more heterogenous socio-economically than are their constituent meshblocks. This implies that the meshblock will yield a more valid measure and is to be preferred whenever possible (calculation of the NZDep scores does inevitably require agglomeration of a very small proportion of meshblocks, to reach the minimum threshold of 100 persons per geographic unit – which is needed to reliably estimate values for the component variables).

NZDep scores are conventionally aggregated into deciles for analysis, with decile one representing the least and decile ten the most deprived 10% of small areas. However, continuous score distributions can also be analysed. This analytic approach has the advantage of being unaffected by the skewed distribution of the scores, which will inevitably introduce some bias into any analysis based on discretised categories (such as deciles). Note that the NZDep is a relative, not an absolute measure.

In New Zealand, social (including health) inequality is typically shaped by the conjoint effects of socio-economic location and ethnic identity.7 This implies that capitation-based funding formulae (for example) need to take this joint effect into account. This requires information on the distribution of each ethnic group across the NZDep deciles, as well as information on the variation in ethnic composition of each NZDep decile. Furthermore, trends in these distributions over time are of importance for strategic policy and service planning across a wide range of social policy domains.

The objectives of the present study are therefore to describe trends over time in:

  • the distribution of New Zealand's major ethnic groups by small area deprivation; and
  • the ethnic composition of each deprivation category.


Data sources were the 1996, 2001 and 2006 New Zealand Census of Population and Dwellings.

Ethnic populations (Māori, Pacific, Asian and European/Other) were defined using total response output (i.e. a person self reporting more than one ethnicity in the Census was classified into each group reported). Māori are the indigenous people of Aotearoa/New Zealand. Pacific peoples trace their origins to the Pacific island countries, in particular Samoa, Tonga and the Cook Islands. Asian peoples trace their origins to the Asian continent, predominantly China, Korea and the Indian subcontinent. The European/Other ethnic group is comprised mainly (>99%) of people who trace their origin to Europe, predominantly the British Isles. ‘New Zealander’ type responses were reclassified as European/Other (described as ‘European’ hereafter). Out of scope or missing responses were excluded (4.2%, 4.0% and 4.2% of total responses in the 1996, 2001 and 2006 Census respectively).

Each person was assigned a deprivation score by geocoding their usual residence (as recorded in the Census) to meshblock level. For each time period (1996, 2001 and 2006), the meshblock boundaries and NZDep formula relating to that period were used. Persons with missing deprivation score were excluded (0.31%, 0.20% and 0.12% in the 1996, 2001 and 2006 Census respectively).

Details of the calculation of the NZDep scores are provided in manuals for each of the 1996, 2001 and 2006 indices respectively, available on the Ministry of Health website http://www.moh.govt.nz.

Deprivation distributions of each ethnic group were described both categorically (by deprivation decile) and continuously (by actual meshblock deprivation score).

Trends in both deprivation and ethnic distributions over time are presented graphically. The Mantel-Haenszel trend test was performed to assess the change in deprivation by ethnic distribution over time. Odds ratios were calculated by logistic regression to estimate whether the probability of being in each deprivation decile was the same for all ethnic groups (the European ethnic group was selected as the reference group). All p values for trend tests were two-sided, and p<0.05 was considered significant. SAS software (version 9.1) was used for all statistical analyses.

In addition to this national-level analysis, a similar analysis was done at District Health Board (DHB) level. For use as a needs-based planning guide, DHB results are reported as absolute counts instead of proportions. The results have been re-scaled proportionately, in order to reduce the total number of responses to match the total number of people in the DHB (necessary because multiple responses to the ethnicity question are allowed in the Census).


Deprivation distribution of the ethnic populations

Distributions of the ethnic populations by deprivation decile and period are shown in Figure 1.

Figure 1.

NZDep decile distributions by ethnicity and period.

Figure 2 shows the same information, but now using the continuous distribution of deprivation scores rather than discretised categories (deciles).

Figure 2.

Continuous NZDep score distributions by ethnicity and period.

Figures 1 and 2 clearly demonstrate the extreme rightward skewing of the deprivation distributions for Māori and Pacific ethnic groups. The European ethnic group is also skewed, but much less so, and in the opposite direction. The Asian population is least skewed, closely approximating the expected distribution of 10% in each deprivation decile.

Superimposing the 1996, 2001 and 2006 ethnic distributions (from Figure 2) shows that little if any change has occurred in the position or shape of these distributions over the observation period. This is confirmed by formal statistical testing for trend (data not shown).

Odds ratios (ORs) for Māori, Pacific and Asian people being in any deprivation decile at any time, relative to their European counterparts, were estimated by regression modelling (Table 2). The odds of being in decile 10, for example, were 5.5, 5.9 and 5.4 for a Māori compared to a European in 1996, 2001 and 2006 respectively. These odds were even greater for the average Pacific person (8.5, 9.9 and 9.0 respectively). Interestingly, an inequality has emerged over time for Asian peoples: in 1996 the stereotypical Asian was as likely to be in deciles 1 or 2 as the stereotypical European (OR 1.0) but by 2006 this probability had declined to only about two thirds (OR 0.65). No other consistent time trends in the ORs were identified for any ethnic group.

Table 2.  Odds ratio for being in each deprivation decile, by ethnic group and period.
  1. Note:

  2. Reference group = European

Ethnic composition of the deprivation deciles

Figure 3 reverses the gaze, to show the ethnic composition of each decile at each census.

Figure 3.

Ethnic composition by deprivation decile and period.

Table 3 presents the same information, but now by deprivation decile rather than period, so allowing time trends in ethnic composition to be more readily seen.

Table 3.  Ethnic composition proportions by deprivation decile, 1996 to 2006.
  1. Note:

  2. Numbers in cells are the percentage of the total population of each decile contributed by each ethnic group at each census


Figure 3 and Table 3 show that Asian peoples make up an increasing proportion of all deprivation deciles over time, reflecting their increasing share of the total population. But the outstanding finding is the increasing proportion of Māori and Pacific peoples, and decreasing proportion of Europeans, progressively from decile one to decile 10. Furthermore, these proportions have hardly changed over the decade, once adjusted for the increasing proportion of Asian peoples (data not shown).

Thus in 2006, ‘New Zealand one’ (deprivation decile one, the least deprived 10% of small areas) is overwhelmingly (86%) European. The European population proportion declines progressively to ‘New Zealand 10’, where Europeans are an ethnic minority (36% of the population). The opposite trend applies to the Māori and Pacific ethnic groups, which collectively comprise only 7% of the population of ‘New Zealand one’ but 57% of the population of ‘New Zealand 10’.

DHB results are not reported here due to word limits, but may be accessed on the Ministry of Health's website http://www.moh.govt/PHI.


Key findings

Looking first at the 2006 distribution of each ethnic population across the deprivation spectrum, the outstanding finding is the heavily right skewed distributions of the Pacific and (to a lesser extent) the Māori ethnic groups. Thus 34.8% and 24.2% of the Pacific and Māori populations reside in decile 10 areas respectively, instead of the expected 10%. By contrast, only 1.9% and 3.4% of these populations live in decile one areas, respectively.

The European population distribution is also skewed, but to a lesser extent and in the opposite direction. Thus 12.3% of Europeans currently live in decile one areas but only 5.6% in decile 10 areas. The Asian ethnic group shows a distribution close to that of the total population (i.e. 10% in each decile), except for a small shortfall at both extremes of the distribution (8.0% instead of the expected 10% in both decile one and decile 10).

With regard to trends, the overall impression is one of stability over the decade. For none of the ethnic groups is there a statistically significant change in the position or shape of the distribution of deprivation scores over the (relatively short) study period.

Turning now to examine the ethnic composition of the different deprivation deciles, it is clear that ‘New Zealand one’ and ‘New Zealand 10’ are very different worlds, not just socio-economically, but ethnically as well. Thus in 2006, decile one was European-dominated, comprising 85.8% European, 7.7% Asian, 5.2%Māori and 1.3%o Pacific. By contrast, Europeans were an ethnic minority in decile ten, which comprised 36.0% European, 33.8% Māori, 23.1% Pacific and 7.1% Asian.

Despite changes in the sizes of the ethnic populations over the decade (for example, due to migration), the ethnic composition of the different ‘New Zealands’ (deciles) has changed relatively little. In fact, the only change of note has been the increasing proportion of Asian people in almost all deciles, with the result that the proportion of most other ethnic groups has declined slightly. In particular, the Māori population contracted on a relative scale in 2001, reflecting relatively high rates of both Asian and Pacific immigration around that time.


This study – like all studies – has its limitations, mostly related to trend analysis. First, concepts of ethnicity have changed over time and the wording of the ethnicity question varied in the 1996 Census from that asked in both the 2001 and 2006 Censuses. The major differences were a greater tendency to report multiple ethnicity in 1996, and a much higher rate of ‘New Zealander’ type responses in 2006. The former has been dealt with by using total response output, so variation in the rate of multiple ethnic identities should have little effect. The latter has been managed by re-coding all New Zealander-type responses to the European category, based on detailed analysis of these responses by Statistics New Zealand.8 Given these analytical approaches, it is likely that any residual bias arising from variation in ethnicity reporting in successive censuses will be small.

Second, the deprivation measure may not work equivalently in all ethnic groups, for at least two reasons. The relationship between some of the variables included in the index and the underlying latent variable ‘deprivation’ may differ by ethnicity. For example, crowding may not have the same ‘meaning’ for Pacific households as it does for European households, or even between different Pacific ethnic groups, reflecting differences in family size and housing tenure.9 Also, socio-economic heterogeneity within small areas (meshblocks) may be greater for some ethnic groups than for others. Thus, residential economic segregation may be relatively less pronounced for Pacific and (especially) Asian peoples, many of whom are recent migrants and tend to live in ethnic enclaves (mainly in the Auckland region) that may be economically heterogenous.10 Again however, any bias introduced is likely to be small.

Third, meshblock boundaries have varied over time. One option was to re-code all meshblocks to the 2006 boundaries; however, we have preferred to retain the census-specific boundaries. This means that the distributions at each timepoint (census) are accurate, but some bias is introduced when examining trends over time. However, over a decade most meshblocks undergo little if any boundary shifting, so this bias should again be small.

Finally, despite the stability of the NZDep index from 1996 onwards (see table 1), there are still minor inconsistencies between NZDep 1996 and NZDep2001, and to a lesser extent between NZDep2001 and NZDep2006. While these minor changes should not affect score distributions when aggregated to the level of deciles or major ethnic groups, trend analysis should always be interpreted with caution.

Policy implications

Our analysis was motivated by accumulating evidence that ethnic inequalities in health and other social outcomes reflect, in part, unequal distributions of economic, social and political resources across the ethnic groups.11 Inequalities in command over such resources are themselves rooted in historical and ongoing processes (specifically, colonisation, migration, acculturation and racism) that entrench the privileged position of dominant groups.12 Monitoring trends in the relative deprivation distributions of the ethnic groups thus provides a way of assessing progress towards social justice and anti-racism.

Monitoring trends in the ethnic composition of different deprivation deciles may also be useful – especially in relation to policies that claim to address social disadvantage irrespective of ethnicity. For example, ‘New Zealand 10’ is clearly a very different world from ‘New Zealand one’– not only is it relatively economically and socially deprived (relatively low incomes, high dependence on means tested benefits, high unemployment rates, high proportion of single parent families, few households with assets, high prevalence of crowded housing, little home ownership), but it is also predominantly (approximately 64%) non-European. This implies that policies directed to economic development of deprived areas and communities in New Zealand may need to go beyond income redistribution and active labour market policies to address issues of racial discrimination as well. More positively, the opportunity exists to design regional development policies that build on the unique strengths of these ethnic communities. Failure to seize this opportunity allows the conjunction of ethnic with economic residential segregation to amplify the health and other social inequalities experienced by these communities.

It is disappointing, therefore, that we have found no improvement in the ethnic distributions or deprivation decile compositions from the mid 1990s to the mid 2000s. However, there are several possible explanations for this. First, the observation period may be too short to show significant social change, since this depends (given our measure of deprivation) on residential mobility. Second, many pro-poor social policies (such as the reform of labour relations, the removal of market rents for social housing and the Working for Families income redistribution policy) were only introduced in the late 1990s or early 2000s and are still subject to lag effects.13 Finally, and most importantly, NZDep is a relative measure: ethnic distributions and compositions will be unaffected if all groups progress at similar rates. To achieve equity (i.e. 10% of each ethnic group in each deprivation decile, and each deprivation decile has the same ethnic composition as the national population), disadvantaged groups would need to make even faster economic progress than advantaged groups.

In conclusion, the information presented provides a useful input for needs-based planning in many social policy domains. The DHB-level analysis (available from the Ministry's website), which provides cross classified population counts by DHB, is only one example of the sort of planning guidelines that can be produced. The data can readily be produced for any bespoke geography used in any social policy domain in New Zealand, provided it is based on meshblock units. Furthermore, similar analyses could be undertaken for Australia, which also has well-established small area census-based socio-economic measures (such as the Index of Relative Socio-economic Disadvantage)14 and reasonably robust ethnicity data.


We owe an intellectual debt to Professor Peter Crampton and Dr Claire Salmond, the architects of the New Zealand Deprivation Index. We thank Teresa Wall and Gabrielle Baker (Māori Health Directorate) and Sandra Moore and Debbie Ryan (Pacific Health Team) for constructive criticism of an earlier draft of this paper. This report is published with the permission of the Deputy Director-General (Health and Disability Systems Strategy Directorate), New Zealand Ministry of Health. However, all opinions are the authors’ own and do not necessarily reflect the Ministry's policy advice.