NZ Council for Civil Liberties

Submission: Data and Statistics (Census) and Electoral (District Boundaries) Amendment Bills

About the New Zealand Council for Civil Liberties

  1. The New Zealand Council for Civil Liberties (‘the Council’) is a voluntary, not-for-profit organisation founded in 1952 which advocates to promote human rights and maintain civil liberties.
  2. In 2015, the Council merged with Tech Liberty New Zealand. Since then the Council has had technology governance and policy as a focus area.
  3. The Council campaigned against the Data and Statistics Act 2022. We have since kept a close eye on Stats NZ, including responding to their consultations and attending quarterly meetings with Stats NZ for the past several years. 
  4. We wish to make an oral submission to the Committee.

Introduction

  1. The Data and Statistics (Census) Amendment Bill changes our laws to move our census from a full enumeration survey, what every ordinary person understands the word census to mean, to calculations on government records collected for other reasons.
  2. The Electoral (District Boundaries) Amendment Bill makes changes to the frequency of redrawing boundaries between electoral districts, supposedly to align them to the change in frequency of the census.
  3. Our Government makes countless decisions based on the census. The census is the map by which we understand ourselves. Diminished census quality will create an avalanche of government inefficiencies and poor decision making which will affect people’s lives.
  4. The census may be the last vestige of leave-no-one-behind thinking in government. Yes it is hard, and of course not everyone was contacted, but the effort is worth it. Now is a time when our government needs to be investing in rebuilding public trust.  Ending the census as we know it does the opposite.
  5. The Council opposes both bills. The entire future census programme will cause widespread harm. We oppose them primarily for the same reasons as the UN Committee on the Elimination of Racial Discrimination opposes: data will inevitably decline in quality, especially data about Māori, Pasifika, and other minorities.1

A full enumeration survey in 2028 will work

  1. Some people have argued that it is too late to run the 2028 census in the same model that was used for the 2023 census.  The Council disagrees.
  2. First we hope that the Committee has the sense to appropriately deal with people who are trying to rob it of its agency by telling it that their preferred outcome is inevitable.  If nothing else, these tactics should not be rewarded.
  3. Stats NZ have not provided evidence in response to the Council’s repeated requests for why there is insufficient time to conduct the 2028 census.

Recommendation 1

The Committee should require Stats NZ to provide detailed schedules demonstrating why it can not conduct the 2028 census.

Future Census Independent Evaluation Panel (FCIEP)

  1. In July 2024, Stats NZ established the FCEIP to evaluate its plans. The FCEIP were asked to pick between five options, all of which were for the change being proposed.  The 2023 approach was not an option presented, nor was there an option to move back to using even more of a traditional census.2 
     
  2. The FCEIP recommended that Census 2028 be conducted exactly like Census 2023 saying:3 

the government’s data and statistics system would not be ready to move to, and successfully implement, a more transformative approach by 2028. 

  1. On September 30, 2024 the FCEIP were asked to change their decision and refused to do so.4
  2. Despite repeated calls from notable observers, including former Government Statistician Len Cook, there have been no supplementary external reviews.

Recommendation 2

The Committee should use its powers under section 198(2) of the Standing Orders to commission its own independent study of the feasibility and consequences of performing an admin-data-first “census”.

The admin-data-first 2030 “census” is risky

  1. The Council is skeptical that Stats NZ will be able to accomplish the admin-data-first “census”. The idea that an organisation already starved of funding will accomplish a large, novel, and complex transformation without an increase in funding would be comic if it wasn’t so tragic.  This is not a risk worth taking, even if there was a strong case for change.
  2. In September 2024, Stats NZ published a technical feasibility study (TFS). This report was updated in December 2025. In the 2024 version of the TFS, Stats admits that they are not ready for the shift to an admin-first census 27 times in addition to the 26 bullets on pages 48-49. The current TFS correctly states “Stats NZ has made significant methodological and infrastructure progress.”5 However, it also reports under the heading “critical areas to progress” that “we have not yet demonstrated the ability to provide a full census of dwellings and their attributes without field enumeration”.
  3. Stats NZ acknowledges that, at best, if we are to switch to an admin-first census, we are changing what we mean by the word census:6

At the same time, there are constraints and risks associated with using admin data to produce census-type information. The key challenges are related to the comprehensive nature of census information and the high-quality standards needed to meet users’ needs. Difficulties may arise because admin data is not comprehensive by its nature (for example, certain data may only include a subset of the population) or may include people who are not part of the target population.

  1. With a survey-first census, when Stats NZ is willing to acknowledge a change in society, it can change the way that it measures us in the next census.  With admin-data first, it has the impossible mission of changing every application in every government agency to, for just one example, acknowledge the existence of non-binary people.  The massive delays this will cause will cause real harm to marginalised communities.
  2. The TFS acknowledges that the admin-first “census” reduces our ability to track important characteristics:7

The collection of representative and complete information for rainbow communities is limited in admin-data settings, and may continue to be, because in many contexts it will not be necessary or appropriate to collect. If the census model moved away from a full enumeration approach, the loss of population-level identifiers for rainbow communities would reduce the quality of detailed population counts, the ability to report on attributes collected in the census, and the ability to link to admin data to report on measures such as health outcomes, income, and education. A census attribute survey would provide support for measuring these outcomes, but it would not provide the same level of detail as a full enumeration census, particularly for small populations such as trans and non-binary populations. The development of a census survey programme does offer potential to deliver a wider set of social measures, such as wellbeing, that are not able to be provided by existing surveys due to sample size.

  1. To be entirely clear, these bills legally bind an agency to perform a task which it has clearly and repeatedly stated may not be possible. The only sensible path for the Committee is to reject the bills.

Data Quality

  1. Information from the admin-data-first “census” will be less accurate than a full enumeration census would be.
  2. The TFS acknowledges the fundamental problem with admin data:8

In addition to coverage issues, if information is not important or necessary for an administrative purpose, data may not be collected or may be of poor quality (Stats NZ, 2014). For example, as agencies move services online, residential addresses may become less important as a contact point and may not be captured well. This affects the ability to locate people where they live. Admin data only collects what is needed to deliver a service or to record an event. It does not capture everything we want to know from a census (for example, information needs for rainbow communities, iwi, and the disabled population). Further, admin data is susceptible to changes in definition or collection, which can adversely affect time series comparisons and the relevance of the data.

and:9

 For population groups that are very small (say fewer than 9,000 members) the quality of data from even a large sample survey like the CAS will be poor in comparison to a full enumeration census.

  1. We note with considerable alarm that the “smaller population groups” who are most badly impacted by the reduction in data quality are the very groups our government should be most concerned about: Māori, Rainbow, Pasifika, “Ethnic Communities”, disabled people, migrants and refugees, and homeless people.10
  2. In 2015 Stats NZ commissioned McNally & Bycroft to create standards for assessing the quality of statistics produced from admin data. After noting that these standards set the bar too low, the TFS details how Stats NZ has failed to reach them.11 For example:
    1. The 2015 standard sets a target for measuring the Māori population to within 0.5%, the best Stats NZ can do is 0.9%.12
    2. The 2015 standard calls for 90% of the 5-year-age-and-sex groups to be within 1.5%, but Stats NZ can only get 72%.13
  3. Stats NZ expects an 8% to 10% error rate from the admin-first model in identifying Māori in 2028, and they acknowledge that that rate will get worse with time:14

Further data sources would be required to improve the coverage of Māori descent. Without a full enumeration census, the combination of Māori descent information collected in electoral roll data and DIA birth registrations would provide the Māori descent indicator for between 90 and 92 percent of the population. In the short term, historic census data will further reduce the amount of missing data.

  1. The TFS identified ethnic communities as another area where declines in data quality would be significant:15

Small ethnic communities rely on census for population counts and demographic summaries. These populations are unlikely to be well served by deriving this information from a sample survey due to population-size and sample-size restrictions. Improvements in the collection and coding of ethnicity collected in admin settings is likely to add significant value to the ability of different agencies to monitor outcomes within their own data system, but improvements will take time and significant resource by the agencies involved. Identifying where to prioritise quality improvement will need to be done in partnership with the agencies involved to ensure the work aligns with their data strategy. Given the coverage of the population, the health and education sectors are likely to be key areas for focusing improvements.

and:16

Improvements in the quality of ethnicity data collected by key agencies to support level 4 ethnicity will be critical to an admin-first approach that does not include the use of a full enumeration survey. Without this information, Stats NZ will not be able to produce detailed population counts and summaries for these communities.

  1. As difficult as it might be for the uninitiated to believe, Stats NZ is unable to determine where people live from admin-data. They call these the “households”, ”families”, and “dwellings” problems.  On page 7 of the 2024 version of the TFS they claim 82% accuracy as a success. These problems have been so bad for so long that misidentifying the homes of nearly 1 in 5 people seems like a victory to Stats NZ. This is the future that these bills bring to us. Stats NZ have been working on the dwelling problem for more than a decade. They carefully avoid claiming that they will have it resolved by 2030, as they know they will not.

There is no case for change

  1. The reasons provided for this change do not stand up to scrutiny. The bill’s explanatory note says:

The transition to an admin-data-first approach to future censuses is required to respond to declines in affordability, efficiency, data quality, and timeliness under the full field enumeration survey approach. The full field enumeration survey approach is no longer financially sustainable. There is growing resistance to completing census forms and rising costs in maintaining survey response rates to preserve data quality. Increased investment for the 2023 census lifted response rates but, despite this, target response rates for some populations were not met.

Alongside this, a full field enumeration census model based on a 5-yearly survey is not delivering data at the frequency that users need to make well-informed decisions.

Data quality was addressed in the previous section.

Affordability

  1. Stats NZ calculated that the 2023 census produced $2.8B in value for a cost of $326M.17 More than $8.50 of benefits resulted from every dollar spent.
  2. Stats NZ has argued in economic terms, however it has not provided cost estimates for the alternatives, nor estimates of the benefits. The Council rejects the notion that the 2023 census was too expensive, or that the 2028 census would be. We believe that a census conducted in 2028 in the same approach 2023 census has a higher net benefit than the new approach does and that any decrease in costs in the new approach would be accompanied by an even greater decrease in benefits.  Former Government Statistician Len Cook said about this proposal:18

The opportunity cost from potential investment failures far exceeds the statistics budget.

and

First, neither Statistics New Zealand nor the Treasury are likely to be aware of the majority of the uses to which official statistics are put. Second, those who use official statistics do not invest enough in influencing how statistics can be developed to provide further value. Expert users themselves may be unaware of the scale of decisions in their sector which depend on the scope, frequency and quality of population statistics.

  1. When Stats NZ, or its Minister, has argued for this change, they have made a misleadingly narrow economic argument. Stats NZ costs to run the census might decline with the new approach. However, every other agency in the government will see their costs rise and their effectiveness decline as they make decisions with less accurate information. As will countless bodies outside of government. No attempt has been made to quantify nor balance the cost of these externalities.
  2. Independent reviewers Murray Jack and Geoff Bowlby said moving from a traditional census to administrative data would involve a large and complex work programme, and “great care and caution” should be taken.  “Great care and caution” is likely to increase, rather than decrease cost for the first census under the new model.
  3.  The TFS says:
  • “Collecting high-quality survey data is increasingly challenging and costly”19
  • “[Admin-based misclassification correction] in the case of two large administrative lists would lead to a much larger number of records going to resource intensive clerical review”20
  • “Investment in survey methods is needed for both an attribute survey and our household survey programme more generally“21
  • “The systematic use of model-based estimation in a production setting would require significant methodological investment”22
  • “While improvements in the collection and coding of ethnicity is likely to add significant value to the ability of different agencies to monitor outcomes within their own data system, this will take both time and significant resource by the agencies involved.”23
  1. We should note that the proposed budget does not result from any sort of effort estimate at all. It is merely the mid-point between the previous two censuses.24  Notably it’s not even an inflation and population adjusted midpoint. There’s no reason to believe that the target will be reached.
  2. It should also be noted that it is an extraordinary claim that a new, and more complicated, process will be less expensive than the process which we already know how to do.  Again, no evidence supports this extraordinary claim.
  3. The Council will again note that Stats NZ have still not published an estimate of either the costs or the benefits of either the 2028 census or their new proposal. Stats NZ refusal to provide evidence despite the Council’s repeated requests should greatly concern everyone.

Recommendation 3

The Committee should request that Stats NZ provide budgets for both approaches accompanied by estimates of the benefits of each approach, and the costs borne by other government agencies by a switch to the new approach.

  1. Earlier this year the Council was informed by Stats NZ that detailed costs for the admin-data-first census would be published in this year’s budget and that these bills would appear after the budget.

Recommendation 4

The Committee should delay deliberation of these bills until it has been able to examine the relevant materials from the budget.

Timeliness

  1. The third reason which the bill cites as a reason for change is improved timeliness. Improved frequency is not a quality inherent to the new approach. A census on the old approach could be run every year, or the new approach could only run once a decade. Timeliness is directly related to affordability, which has not been demonstrated.

Response Rates

  1. Finally, the bill proposes to resolve the problem of falling response rates by asking fewer people. This is the opposite of the statistically valid approach.  The smaller the sample size, the bigger the problem caused by low response rates. 
  2. The Council further notes that Len Cook disputes “the idea that New Zealanders are not willing to complete census forms”. As intuitive as that conclusion may be from falling response rates, there are numerous other factors involved.  We agree that Stats NZ have again failed to provide evidence for a key claim.

These bills hide decisions of constitutional significance

  1. Pages 11 and 12 of the TFS describe the removal of legislative “barriers” to the admin-data census. There is no acknowledgement anywhere that those laws served important purposes, nor an explanation of why these safeguards were no longer necessary.
  2. IPP 13 of the Privacy Act makes a government-wide identifier, such as those present in other countries, illegal. IPP 13(2)(b) reads “the unique identifier is to be used by A for statistical or research purposes and no other purpose.”25
  3. IPP13 is the current version of legal safeguards which were put in place to deter a recurrence of the Holocaust. Counting people has been, and remains to this day, an obsession for those considering genocide. The Council is deeply disturbed that Parliament is choosing this moment in history, with genocide back on the world stage after generations of peace, to weaken these protections.
  4. The TFS states that a functional admin-first census “requires” “an enterprise register-based statistical system”.26 Stats doesn’t describe this system.  For the paragraph to hold together logically, the “enterprise” in question is the entire government. Stats NZ knows that the correct term for such a system is a population register.
  5. In 2015, Stats recommended against a population register.27 This recommendation officially remains unchanged.
  6. The Council regularly discusses population registers with Stats, who maintain that this is a change of constitutional importance, which would require a referendum accompanied by a public education campaign. They are relying on the detail that these bills do not entirely implement a population register.
  7. In countries with population registers the government wields more detailed and direct control over individuals. Whatever the academic merits were of the Soviet Union, such a system is in opposition to civil liberties. If we are to adopt such a system, it should be an explicit choice, not the consequence of technical laws.

Data and Statistics (Census) Amendment Bill

Clauses 11 and 13

  1. Clause 11 repeals section 38 of the Data and Statistics Act, which requires ordinary people to do all of the work to determine “how to access and respond to … request[s] made by the Statistician”. Clause 13 repeals section 89 which establishes an offence for failing to figure out how to obtain a “request” from the Statistician.28
  2. The Council was highly critical of the census offences when the act was passed.29 We support the repeal of these sections.

Clause 12

  1. Clause 12 repeals section 76(2) of the Data and Statistics Act 2022. The Council continues to maintain that the 2022 Act is riddled with unjustified limitations of everyone’s liberties.30 Earlier this month we recommended that Stats add more safeguards in this area. 
  2. Recommendation 6.4 of our 2022 Briefing on the Data and Statistics Act was to remove the word “not” from section 76(2), thereby inverting its meaning. Whilst removing section 76(2) effectively accomplishes the same thing, greater clarity results from stating that not being told “how to access and respond to, a request made by the Statistician under section 23 as part of the census” is a defence to charges under that section.

Recommendation 5

Replace clause 13 to amend section 76(2) to put an obligation on the Statistician to prove that the person being charged under section 76 was given the opportunity to fill in the census. 

Electoral (District Boundaries) Amendment Act

  1. The Electoral (District Boundaries) Amendment Act is temporal gerrymandering. It seeks to subvert the course of democracy by changing when the lines are drawn, rather than where. By the government’s own admission, if the boundaries were adjusted before 2029, then there would be an 8th Māori electorate.31 

Recommendation 6

If these bills proceed, and we therefore have high quality population counts annually, we should adjust the boundaries before every election, so that they are as accurate as they can be. 

  1. Further, section 35 of the Electoral Act, combined with the definition of “general electoral population” in section 3, results in general electorates being established based on the census. Section 45 of the Electoral Act, combined with the definition of “Māori electoral population” in section 3, results in Māori electorates being based on enrolment on the Māori electoral roll, not on the census.  Given that great numbers of Māori people have become politically active recently, it is entirely reasonable to suspect that more than one Māori electorate might be added soon.

Recommendation 7

Regardless of whether these bills proceed or not, given that Māori electorates are based primarily on voter enrolment, not census data, the Māori electorates should be adjusted before every election. 

  1. The Council thanks members of the Committee for their time and consideration of our submission.
  1. CERD/C/NZL/CO/23-24 at paragraph 5 “It is also concerned about the possible implications of this shift for decisions that have traditionally relied on census-based data, such as those relating to affirmative action measures, quotas and other targeted policies” ↩︎
  2. Decision paper paragraph 12 ↩︎
  3. Decision paper paragraph 20 ↩︎
  4. Decision paper paragraphs 42-45 ↩︎
  5. TFS page 48 second paragraph ↩︎
  6. TFS page 16 first paragraph ↩︎
  7. TFS page 9 fourth paragraph ↩︎
  8. TFS page 16 second paragraph ↩︎
  9. TFS page 40 third paragraph ↩︎
  10. Regulatory Impact Statement page 25
    The Council takes the opportunity to note that Stats NZ has high confidence for most things on pages 34-37 despite this being an untried and underfunded programme from an agency with a poor track record. Further, there is no modelling in the RIS of the possibility of failure.  All of the “risks” in the RIS are about the plan going better than planned. ↩︎
  11. TFS page 17 second last paragraph ↩︎
  12. TFS page 20 seventh paragraph ↩︎
  13. TFS page 20 eighth paragraph ↩︎
  14.  TFS page 7 second paragraph ↩︎
  15. TFS page 24 last paragraph ↩︎
  16. TFS page 42 bullet 5 ↩︎
  17. paragraph 12 of MM2826 ↩︎
  18. https://ojs.victoria.ac.nz/pq/article/view/9484/8393 ↩︎
  19. TFS page 6 fifth paragraph ↩︎
  20. TFS page 35 third paragraph ↩︎
  21. TFS page 38 third paragraph ↩︎
  22. TFS page 38 last paragraph ↩︎
  23. TFS page 43 sixth bullet ↩︎
  24. Decision paper paragraph 25 ↩︎
  25. Privacy Act s29(2) exempts Stats only from IPP7 ↩︎
  26. TFS Page 10 second paragraph ↩︎
  27. TFS page 11 second bullet ↩︎
  28. Failure to respond to a request is separately an offence under section 76. ↩︎
  29. https://nzccl.org.nz/submission-data-and-statistics-bill/ at paragraph 83 ↩︎
  30. https://nzccl.org.nz/submission-data-and-statistics-bill/ at paragraph 84 ↩︎
  31. Regulatory Impact Statement paragraphs 33 to 43. ↩︎