Submission to the Justice Select Committee

Data and Statistics (Census) Amendment Bill

Research Association of New Zealand (RANZ)  |  April 2026

1. About the Research Association of New Zealand

The Research Association of New Zealand (RANZ) is a not-for-profit organisation established in 2014 through the merger of the Market Research Society of New Zealand (MRSNZ) and the Association of Market Research Organisations (AMRO). RANZ exists to promote the value, use, evolution and development of research practices, disciplines, and ethics to the wider community, and to provide industry leadership within the New Zealand research community.

Our tikanga commits us to honesty and integrity in research, the promotion of ethical research practices, and recognition that New Zealand is a diverse society. Our membership includes most market research agencies in New Zealand as well as many commercial and government in-house research and insight teams.

Much of our work sits at the heart of local, national and global decision-making. This could be evaluating the need for, and impact of social policies, building investment cases for business expansion, understanding New Zealand’s appeal among possible overseas visitors, launching a new product into a global market, or helping the local sports club with the evidence required for their community investment programmes. Good research drives good decisions and great outcomes for all New Zealanders.

We maintain close links with relevant international industry bodies including The Research Society (TRS, Australia), ESOMAR (the European Society for Opinion and Marketing Research), the Market Research Society (MRS, UK), the Insights Association (IA, USA), and regional associations in Asia and Canada. We have a well-established Code of Practice and a complaints process for handling member and non-member complaints.

This submission is made in the spirit of constructive engagement with the select committee, drawing on our members' professional expertise in data collection, survey methodology and statistical analysis.

 

2. Census Data, Our Members, and the Challenge of Falling Response Rates

Accurate, granular census data is not an abstraction for RANZ members. It is the foundation on which the work of New Zealand's research and insights industry is built. Market researchers, social scientists, public health analysts, and policy evaluators rely on census outputs as the primary benchmark for understanding who New Zealand's population is, how it is distributed, and how it is changing. Census data underpins survey sample design, population weighting, small-area estimation, demographic profiling, and the validation of findings drawn from other data sources. Without a robust population baseline, the reliability of virtually every piece of commissioned and independent research conducted in New Zealand is affected.

This is not a theoretical concern. When census data are incomplete, delayed, or methodologically inconsistent, the consequences ripple through the entire research ecosystem. Businesses make investment decisions on flawed market size estimates. Government agencies design interventions on the basis of demographic profiles that do not reflect reality. Iwi and community organisations cannot make a credible case for resource allocation when the population counts underpinning their arguments are in doubt. For RANZ members, the integrity of the census is, quite directly, the integrity of their work.

We want to acknowledge directly, and without reservation, that declining response rates are a genuine and serious problem. They are a global challenge that our members encounter in every market in which they operate, and one that our industry has invested considerable effort in understanding and addressing. Through our connections with ESOMAR, the Market Research Society, The Research Society, and the Insights Association, RANZ members have access to, and actively contribute to, a substantial body of international evidence on the drivers of non-response and the interventions most effective in recovering participation.

That experience, however, also tells us that New Zealand is unusually well-placed to address this challenge. New Zealand has produced some of the most successful population-level social marketing programmes in the world: from road safety campaigns that have fundamentally changed driver behaviour, to anti-smoking programmes, mental health initiatives and Covid-19 public health communications that were studied internationally for their effectiveness. These campaigns succeeded not through compulsion but through a combination of trusted institutions, compelling communication and a genuine sense of working towards a shared national good. That same cultural infrastructure, properly mobilised, is the most powerful tool available for reversing participation decline in the census. It is not something that any other country can simply replicate.

Declining response rates are a challenge that can be managed with the right investment, design, and community engagement. They are not, in RANZ's view, a sufficient justification for abandoning universal enumeration. The populations most likely to disengage from surveys are also, as we discuss below, the populations most likely to be poorly represented in administrative datasets. Substituting one incomplete data source for another does not resolve the underlying problem; it relocates it, and often makes it harder to see.

RANZ would welcome the opportunity to contribute its members' experience in survey design, response recovery and community engagement to the development of a modernised census approach that preserves the strengths of universal enumeration while addressing its practical limitations.

 

3. Executive Summary

RANZ respectfully opposes the Data and Statistics (Census) Amendment Bill in its current form. While we acknowledge the genuine practical pressures that have prompted this reform, we believe the proposed shift to an administrative data-first, person-centric model introduces risks that outweigh the projected efficiencies. Our concerns are presented in order of significance. The most serious is the impact on small-area data, which we regard as a matter of direct consequence for every person in New Zealand who depends on good local decision-making. This is followed by the risk of permanent breaks in our national time series, the absence of the administrative infrastructure that makes similar models work in comparable countries, the threat to international data comparability, and the inadequacy of the current proposal's response to the challenge of counting marginalised communities.

We also believe the proposed changes have not yet been subject to the independent expert scrutiny that a reform of this magnitude warrants, and we call on the committee to require that scrutiny before the bill advances. In the event the bill passes in its current form, we set out the safeguards that we believe must accompany it.

 

4. Our Position: Retain the Current Census as the Primary Method

4.1  Small-Area Data: The Most Direct Risk to Local Communities

The most immediate and consequential risk in this bill is the one closest to home: the proposed annual survey of approximately 5% of the population cannot produce reliable population estimates at the level of geographic disaggregation that local decision-making requires.

Every decision about where to build a school, how to staff a health clinic, whether to fund a community social service, how to zone for housing, or how to allocate local infrastructure investment depends on knowing, with reasonable precision, who lives where and what their characteristics are. These decisions are made at mesh-block, area unit, territorial authority and regional council level. A 5% annual sample produces margins of error that render estimates at these scales statistically unreliable for the populations and geographies that matter most. Stats NZ's own regulatory impact statement acknowledges this, noting reductions in data accuracy, detail and coverage, particularly for small population groups and local areas.

The practical consequences are not abstract. The Ministries and local authorities involved with planning a new primary school need to know how many children under ten live in a particular catchment, not an estimate that is accurate to within thirty percent. Te Whatu Ora allocating mental health services needs a reliable count of the population at risk, not a modelled figure with wide confidence intervals. An iwi governance body making the case for resource allocation needs population data that will withstand scrutiny, not an administrative data estimate whose limitations are opaque to the people reviewing it. These are the decisions that shape the daily lives of New Zealanders, and they are precisely the decisions that are most vulnerable to the data quality reduction this bill proposes.

Designing local services from a 5% annual sample is like a builder being asked to fit out twenty different houses but only being permitted to inspect one of them beforehand. For the house they inspected, the result may be good. For the other nineteen, they are working from assumptions about what they might find, and the further those houses are from the average, the more consequential those assumptions become.

For small population groups at local level, the problem compounds. A 5% sample in a small territorial authority may yield only a handful of observations for minority ethnic communities, people with disabilities, or specific age cohorts. Even multi-year pooling of the annual survey cannot consistently produce the precision that these groups require for evidence-based advocacy and service design. The people least able to make their case through other means are the most dependent on the census to make it for them.

 

4.2  Time Series Continuity: The Risk of Losing Our Statistical Memory

The 2023 Census anchors a time series stretching back to 1851. This is not merely a historical curiosity; it is the analytic foundation that allows New Zealand to understand change, measure progress, and evaluate the effects of policy over time. The ability to answer questions like whether housing affordability has worsened, whether educational outcomes for Māori have improved, whether income inequality has widened, or whether regional population patterns are shifting depends entirely on the continuity and comparability of the underlying data series.

Transition to an administrative data-primary methodology will introduce conceptual differences in how ethnicity, household composition, tenure status and other key variables are defined and captured. These differences will break the comparability of statistics across the series at the point of transition. Stats NZ has signalled its intention to produce bridging analyses, but bridging analyses are approximations. They introduce uncertainty that compounds over time, and no methodology can fully reconstruct the information that is lost when the underlying collection changes.

Researchers, local authorities, iwi, and central government agencies that rely on long-run trend data will face a material reduction in analytical capability that no bridging methodology can fully restore. For programmes and policies designed to close persistent gaps in health, education, housing and income, the loss of a consistent time series is not a technical inconvenience; it is a direct impediment to accountability.

 

4.3  Infrastructure Readiness: New Zealand Is Not Yet Scandinavia

Proponents of the bill frequently cite the Nordic countries as evidence that register-based census systems work. This is true, and we do not dispute it. What the comparison obscures is the investment horizon required to make them work. Sweden, Denmark, Norway and Finland have spent fifty or more years building the comprehensive, interoperable, high-quality population registers that underpin their statistical systems. These registers are not a product of legislation and goodwill; they are the result of sustained institutional investment, rigorous data governance, and a legal framework that mandates consistent administrative recording across all public agencies.

New Zealand does not yet have this infrastructure nor, we would argue, the social buy-in required for a national identification system. Our administrative datasets are fragmented across agencies with different collection standards, different definitions, different update cycles and different coverage rates. The bill proposes to use these datasets as the primary source of population statistics from 2030, a period of just four years from now. That is not sufficient time to build the underlying data quality that register-based census systems require, as Stats NZ's own regulatory impact statement implicitly acknowledges when it concedes early-year reductions in coverage and accuracy.

The question is not whether New Zealand could, in principle, eventually develop Nordic-quality administrative registers. It is whether the 2030 timeline is realistic, and whether it is responsible to retire the traditional census before that infrastructure is demonstrably in place. Our answer to both is no.

 

4.4  International Best Practice: New Zealand Would Become an Outlier

The United Nations Principles and Recommendations for Population and Housing Censuses (Revision 3) recommend that Member States conduct a population and housing census during the 2025 to 2034 round. Data from the UN Statistics Division show that 71.8% of countries that participated in the 2020 census round used a traditional, full-field enumeration methodology. Administrative register-only approaches remain exceptional globally, and are associated almost exclusively with countries that have the long-established register infrastructure described above (UNSD, 2024).

Australia, the United Kingdom, Canada, and the United States, New Zealand's closest comparators in terms of research relationships, trade, and demographic structure, all retain five-yearly or decennial universal censuses as the cornerstone of their population statistics systems. The UK Office for National Statistics, following an extensive review of register-based alternatives after the 2011 Census, concluded that a traditional enumeration should proceed in 2021 and committed to retaining a hybrid approach for 2031, citing irreplaceable small-area data and international comparability as key justifications (ONS, 2023). The Australian Bureau of Statistics similarly reconfirmed five-yearly full enumeration as foundational to the national statistical system (ABS, 2021).

Many international datasets maintained by the OECD, the World Bank, the UN Population Division and the International Labour Organization rely on national census outputs as their primary population reference. These organisations publish under the assumption that member country statistics are produced using broadly comparable methodologies. When New Zealand publishes administrative data-derived population estimates in place of census-based ones, without the register quality assurance that justifies this approach in Nordic countries, the risk of non-comparability in cross-national benchmarks is material. For New Zealand researchers, businesses and government agencies that routinely use international benchmarks to contextualise domestic findings, this is a practical and ongoing cost.

 

4.5  Marginalised Populations: Addressing the Strongest Counter-Argument

Supporters of the bill argue that marginalised communities are already systematically undercounted by the traditional census, and that administrative data does not create a new problem but simply makes an existing one visible. This is a fair point and one we take seriously. The 2023 Census did have coverage gaps, and some communities are demonstrably harder to reach through mandatory address-based enumeration.

But the argument proves too much. The existence of undercount in the traditional census does not justify replacing it with a system that has the same problem in a different and less transparent form. People who are undercounted in the census because they distrust government data collection, live in informal housing, or have limited civic engagement are precisely the people with the least complete, least accurate, and least current administrative records. The administrative data that the new model proposes to use as its primary source is not a neutral mirror of the population; it reflects the population as it appears in government systems, which means it is systematically thinner for those who interact least with those systems.

The right response to undercount in the traditional census is targeted investment in outreach, trusted community intermediaries, and co-designed collection approaches for populations that are hardest to reach. New Zealand has demonstrated, through its social marketing and public health communication programmes, that it can change population behaviour at scale. The same approach, applied to census participation with proper funding and genuine community partnership, offers a more durable solution than substituting one incomplete data source for another. The challenge of counting marginalised communities is an argument for doing the census better, not for doing something different.

 

5. A Call for Independent Expert Review Before Proceeding

RANZ believes that a reform of this scale and permanence should not proceed without rigorous, independent external expert scrutiny. This is not a delaying tactic; it is a standard that comparable countries have applied to equivalent decisions.

When the United Kingdom considered moving to a register-based census model following the 2011 Census, it commissioned an independent Beyond 2011 programme that ran from 2011 to 2014, involving multiple external academic and statistical reviews, before concluding that a traditional census should proceed in 2021 (ONS, 2014). When Australia undertook its post-2016 Census review following the significant operational difficulties of that year, it commissioned an independent review panel whose findings directly shaped the design and resourcing of the 2021 Census (ABS, 2017). In both cases, the independent review strengthened rather than delayed the reform process, because it identified specific gaps and provided the public confidence that a change of this consequence requires.

New Zealand should apply no lesser standard. We recommend that the select committee require, as a condition of the bill's progress, the commissioning of an independent panel comprising international statistical experts, Māori data sovereignty specialists, and representatives of data user communities, to assess and publicly report on the following before the bill is enacted:

• Whether the administrative data sources proposed as the primary input to the new census model are of sufficient quality, coverage and consistency to produce nationally reliable population statistics, including for small areas and small population groups.
• Whether the proposed transition timeline is realistic given the current state of New Zealand's administrative data infrastructure, including known gaps for marginalised and hard-to-reach populations.
• The adequacy of the proposed bridging methodology for preserving longitudinal comparability with the 2023 Census and earlier data.
• The implications of the proposed model for New Zealand's standing in international statistical frameworks, including the OECD, the UN Population Division and comparable national statistical offices.
• Whether the privacy and data governance framework proposed is sufficient to maintain public trust and social licence, including specific protections for sensitive variables such as ethnicity, disability status and sexual identity.

If the proposed model is as sound as its advocates believe, independent scrutiny will confirm that and provide the public confidence that a change of this magnitude requires. If gaps are identified, the review provides the opportunity to address them before they are locked into legislation.

 

6. Equity and Coverage Concerns

RANZ is particularly concerned about the differential impact of the proposed approach on communities that already experience data gaps and service inequities. People with limited engagement with government services, including recent migrants, undocumented residents, people experiencing homelessness, some disabled people, and those in informal or overcrowded housing, are systematically less likely to appear accurately and completely in the administrative datasets that the new model proposes to use as its primary source.

The irony of the proposed approach is that it risks making population statistics less accurate for exactly the people and communities for whom accurate data is most urgently needed. A universal census, whatever its logistical imperfections, imposes an enumeration obligation that at least attempts to reach every household. An administrative data system reflects, and may amplify, the institutional invisibility that already disadvantages marginalised communities.

We note that iwi and Māori organisations, Pacific health advocates, and disability sector groups have raised similar concerns in their own submissions. RANZ supports these concerns and believes the select committee should weigh them seriously before advancing the bill in its current form. 

  

7. In the Event the Bill Passes in Its Current Form

While RANZ's primary position is that the bill should be amended to preserve universal census enumeration as the foundational method for New Zealand's population statistics, we recognise that the legislative process may proceed. In that event, we urge the committee to ensure that the following safeguards, drawn from our December 2025 submission to Stats NZ on the proposed data collection approach, are embedded in the legislation or in binding operational requirements placed on Stats NZ.

Specifically, in the event the bill passes, RANZ supports the following measures being mandated or strongly encouraged:

• Sustained investment in administrative data quality and coverage, with particular focus on ensuring completeness for populations that have limited contact with government services. This should include an All of Government data dictionary and mid-to-long-term working groups with agency research teams to align definitions and collection processes.
• Transparent quality grading for all published outputs, clearly indicating whether figures are administratively derived, survey-derived or modelled, and publishing associated uncertainty estimates and quality flags by variable, community group and geography.
• Bridging analyses against 2023 Census outputs to preserve longitudinal usability, allowing researchers and planners to understand and account for methodological discontinuities over time.
• A genuine co-design process, not merely consultation, with Māori, Pasifika, disabled communities, LGBTIQ+ communities and other priority groups, with ring-fenced funding and clear deliverables through to and beyond 2030.
• Oversampling and booster sample approaches for small population groups in the annual survey, alongside multi-year pooling of results, to provide statistically robust estimates at the level of disaggregation these communities require.
• An ongoing independent user and community advisory mechanism, including Māori data sovereignty partnership, to provide governance oversight of the new model's performance against equity and data quality benchmarks.
• A committed parallel-running evaluation period in which the new administrative data model's outputs are systematically compared against alternative estimates, to build the evidence base for future methodology decisions and to signal to international partners the basis on which New Zealand's statistics are produced.

 

8. Conclusion

RANZ values the work of Statistics New Zealand and recognises the genuine challenges that prompted this reform. We are not opposed to modernisation. Our December 2025 submission to Stats NZ acknowledged real opportunities in the better use of administrative data, and we remain committed to contributing our members' expertise to a census approach that is fit for the twenty-first century.

But modernisation and replacement are different things. The arguments presented in this submission, led by the direct risk to small-area data quality that underpins local decision-making across New Zealand, point consistently in the same direction: universal census enumeration should be retained as the primary methodology, with administrative data playing a valuable complementary role. New Zealand has the social infrastructure, the communication capability and the institutional trust to address declining response rates without abandoning the model that declining response rates have put under pressure. We should use those strengths, not concede defeat to a trend that we are, in fact, better equipped to reverse than most.

We also call on the committee to require independent expert review before the bill advances. The standard applied to this decision should be no lower than the standard the United Kingdom and Australia have applied to equivalent decisions about their own national statistical foundations.

RANZ requests the opportunity to appear before the committee to discuss these matters further.