by Bonnie Flaws

Newly released documents show expert advice was disregarded, and the admin-first approach to censuses, which ditches enumeration, was adopted for the next national census anyway.

The Future Census Independent Evaluation Panel produced its report evaluating five possible approaches in August last year, but it was withheld from the public until two weeks ago.

The excuse for not taking the advice of the panel was said to be cost, which is also one of the key arguments for taking this approach in general – censuses are apparently getting more expensive (see census 2018 and 2023). More on this later.

Similarly, a summary of public responses to a consultation that was run in May last year seems to have been sidelined by other ‘targeted consultation’. It too was held back and released two weeks ago in a tranche of official papers – significantly, after the ‘in-principle decision’ was made to go admin-first, and after cabinet had approved it.

The secrecy surrounding the advice and consultation has been slammed by former Government Statistician Len Cook as highly unusual.

“It is doubtful that the statutory obligation of impartiality has been met by the release to some of these plans. This obligation has been ignored, despite the constitutional implications for the conduct of elections that Parliament will now have to sort out,” he said in a Newsroom piece.

This is because the census supports the setting of electoral boundaries and determining the number of electorates for upcoming elections. Changes to the Electoral Act will need to be made as a result, although exactly what has been redacted in official papers. An amendment to the Data and Statistics Act 2022 will also be required to allow for the census to take place in 2030 instead of 2028, to give Stats NZ and other key agencies time to prepare for the new approach.

The explanation that has been provided for not releasing the consultation and advisory documents earlier, was that Cabinet needed to first consider and endorse the Government Statistician’s (GS) in-principle decision, and these were key decision-making documents for the GS.

They were therefore “released along with the key Cabinet decision documents and a range of papers, reports and other outputs related to Stats NZ’s work to modernise the census.”

Expert Review Panel recommended a full enumeration census in 2028

In its report, the review panel stated: “The panel recommends that a long-term strategic approach be taken for how Stats NZ move from the status quo to use of an admin data first enumeration of individuals and dwellings complemented with a census attribute survey which aligns with the other household surveys currently in use by Stats NZ. This process of change must be staged in order to reduce the risk of losing the trust and confidence of the people upon which a census depends, as well as the decision makers that rely on the data that these people provide.”

The Panel’s first recommendation was to conduct a final full-field enumeration census survey in 2028 (option 1), moving to an approach using a continuous sample survey in 2033 (option 4), and transitioning to a model where census data collection would be integrated into other household surveys (option 5).

The panel’s third recommendation was to proactively release its report. They also recommended it be attached to Stats’ cabinet briefing. Neither of these things happened.

Stats NZ claims ‘cost’ was the deciding factor

Going a bit deeper into the panel’s work, the consideration of ‘cost’ as a criteria “was unable to be included in time for the Panel’s assessment”, and it was removed from the initial Terms of Reference.

Stats then presented the panel with two cost-refined options based on the approach it had recommended in its report. However, the panel stood by its decision.

“The Panel was reconvened and invited to provide further insights to the Government Statistician based on the two revised options. The Panel considered that the recommendations in its report were robust and remained appropriate,” Stats NZ said in its response to the review.

Despite this, the GS did not adopt the Panel’s recommended approach and subsequently made its ‘in-principle’ decision to move to an admin-first census.

“… the approach recommended by the Panel was found to exceed the available funding. This meant the Panel's preferred option for 2028 Census could not be implemented as originally proposed,” it said.

Costs

I’ve read through most of the documents released by Stats NZ in mid-June, here’s what I found regarding costs.

The 2023 Census cost approximately $326 million over five years – a significant increase from the 2018 Census ($126m) and the 2013 Census ($104m).

A January 2025 cabinet paper says the funding envelope for the next census is $100m less than what was spent in 2023, so about $225m. Stats NZ estimated that running the next census on the current survey model would cost between $300m and $400m.

Going forward, Stats NZ has a five-year fund of $227m to deliver population and social statistics, and a yearly appropriation of $45.5m thereafter, from which Stats must fund its census work programme. There will be no contingency to cover unexpected costs.

In the medium term, there will be a financial impact on certain key government agencies and Crown entities that Stats NZ acquires data from, to bring them into alignment with how Stats NZ would like them to collect and transmit it. This will be mandatory, enforced via a joint directive from Cabinet and will impact the MoE, DIA, MSD, MBIE and Corrections, as well as the Tertiary Education Commission, Health NZ, ACC, NZTA and the Electoral Commission.

Panel member highlight’s social license

Dr Jesse Whitehead, member of the Future Census Independent Evaluation Panel and Senior Research Fellow at the Institute for Population Research at the University of Waikato said the announcement was disappointing.

“High quality data is important, and administrative data has its own challenges, biases, and inconsistencies, which means it is not the same quality as census data. Data from the census is important for checking the quality of administrative data, and this will no longer be possible under the proposed changes.”

There were also issues with social license, he said.

“It is not clear whether New Zealanders are willing to trade off not having to complete a census form once every five years, with having their data from service interactions, collected and used in this way. It is also unclear whether agencies are ready to provided the quality, and consistency of administrative data required to replace a full enumeration census, and whether these agencies are resourced to make the changes needed.”

Public submissions

When Stats announced its plans for a census-less census and a corresponding public consultation last year, there was some concern that the consultation period was too brief and the public awareness of the consultation was not widely advertised.

Reality Check Radio wrote to the Census consultation in June last year to request a longer time frame for submissions, another two months, given the significance of the proposed changes and and the limited time and public awareness about the consultation, but was denied.

In the end, as some suspected might happen, Stats NZ received a dismal 467 formal submissions.

It’s worth considering that 467 submissions to cover 5 million citizens is not a sufficient statistical sampling size – especially if we consider the scale of the changes being made to data collection, and the constitutional implications.

This feedback was only one stream the GS considered when making its decision about future censuses.

“We also shared potential options for the future of the census with key partners and stakeholders through intensive targeted consultation to understand their data needs and perspectives. Those groups included iwi and Māori, smaller population groups, community groups, government agencies, and researchers and academics.”

After publishing a feasibility report in September, outlining its technical readiness to make the shift towards an admin data-first census, the GS made his final decision to adopt an admin-first approach in October. Stats said the following points were considered:

• the data needs of our customers and stakeholders
• data quality, flexibility, and timeliness
• Māori data needs and governance
• stakeholder and public sentiment
• feasibility, including cost and design complexity
• opportunities and challenges
• the impact of change on our data users and suppliers, and the wider data system.

Trust

Public sentiment expressed in the summary of submissions were summarised by Stats NZ, including the following passages, which indicate there are some serious trust and transparency concerns with the new approach (emphasis mine):

Keeping people and their data safe

Data shared with Stats NZ needs to be safe from security breaches, including being appropriately anonymised so that individuals cannot be identified. Privacy and security are key concerns, with some submitters believing that data cannot be kept safe no matter how it is stored. Submitters want clear accountability should there be any breaches, and assurance that the state cannot use their data to target them. Not meeting these requirements could impact participation in providing data.

Consent and Choice

For any data that will be shared with Stats NZ, submitters want to know how their data will be used and have a choice as to whether they consent to it being used that way. We heard that some submitters may avoid using government services or provide false data if they can’t opt out of data sharing.

Mistrust in government and the data system

There are varying levels of mistrust, which have impacted how submitters feel about this proposal. Mistrust can be exacerbated for some submitters by the inclusion of census questions that conflict with their personal values, which can include specific variables or a general feeling of invasiveness.

A few submitters indicated that they do not (and will never) trust government.

New consultation in the second half of 2025

Stats NZ has confirmed there will be a second round of consultation this year – yet to be announced.

The second round will look at the content of the next census, which will be in 2030, and the way that information will be collected.

“We will advise the public of the consultation via our website and social media channels.

“The purpose of this consultation is to:

• highlight the information that may be collected in the next census, so people can understand the direction, scope, and constraints on content early in the design phases
• promote discussion and invite submissions from the people of New Zealand on what information we’ll be collecting in the next census.

  “The feedback will help inform final decisions about next census topics and also more broadly, the statistics we produce in the population and social statistics space.”

Sadly, there will be no turning back on decision to go admin-first, despite the overwhelmingly negative response to the decision so far by academics and experts, and a significant portion of submitters to the public consultation unless Len Cook’s call for an inquiry into how the decision was made becomes a reality.

Originally published on Byline Babylon.