By Bonnie Flaws

Part four of a four part investigation by Bonnie Flaws. Read parts one, two and three.

• Census 2028 is doing away with the usual population survey in favour of using admin data instead, requiring a Statistical Register
• It’s initial consultation into an admin-data-first census was not widely publicised and did not disclose this fact
• Stats NZs increasing lack of transparency, illustrated by unpublished statistical reviews, has also been criticised
• Stats NZ says the Government Statistician has made an in-principle decision about how the next census will run (admin-first), which is under active consideration by the Government
• Another round of consultation is due to be launched in the the first half of this year
• An explanation of terms can be found here
  

The Admin Data first 2028 Census

I imagine part of the justification for building the Statistical Register, which I previously characterised as the eye of Sauron, watching all of us all the time, whether we want it or not, is that it is required for future censuses.

Certainly OIA responses reveal their plans to create a unique identifier for every citizen as part of this register, would be used to:

• Derive family relationships
• Identify a population of interest and track them over time
• Provide data for health research
• To inform urban development

A year ago the public was informed about plans for a ‘“census-less census”, which was then (barely) consulted on in preparation for the next census in 2028.

The thrust of this is that Stats NZ will do away with the traditional population survey, and use real-time administration data (data held by government agencies) to complete its five-yearly census, in what appears to be a world-first move.

Surveys will still be conducted but only to fill in the gaps where the government doesn’t already hold information. If a person has little contact with government agencies, then government’s won’t hold much data on them.

According to the March 2024 Cabinet Paper approving the change in direction, other countries exploring this model include Australia, the UK and Canada.

But former Government Statistician Len Cook, who remains well connected to the sector here and overseas, says he is not aware of another comparable jurisdiction attempting the same. And while the United Kingdom has floated the idea, the British Royal Statistical Society has only recognised a role for admin data as part of a portfolio of sources for offical statistics, and integrated with surveys, census and other types of data – not “admin first” as New Zealand proposes.

In Canada, plans have been paused due to privacy issues related to personal banking data. In 2018, complaints were made to the Privacy Commissioner because Statistics Canada requested several banks to provide it with the financial transaction information of hundreds of thousands of Canadians. The subsequent report found that necessity and proportionality has not been adequately justified. It recommended that the project be stopped, redesigned and that the data gathered be disposed of. It also called for far greater transparency from the agency.

Cook is also deeply concerned by the lack of transparency coming from Stats NZ.

“One of the problems I have”, he says, “is that Stats NZ is out of touch with what is going on in Australia, Canada and the UK. My sense is that the language that Stats is using about their conformance with other practice is fundamentally bullshit. It’s not really helpful but it makes someone who doesn’t know what’s going on – like an uninformed Minister – feel very comfortable.”

Crucially, the consultation never said or implied that the admin-data-first approach also means having to move towards a Statistical Register-based system. Please read part one of this series for more detail.

Instead it asked saccharine questions of the How much do you agree with this statement variety. Statements such as:

I trust Stats NZ to keep my information safe

or

I am ok with Stats NZ reusing information (like tax or housing information) I have given to other organisations, so that Stats NZ can produce data, statistics, or research that would benefit me and my community?

It all sounds very harmless. But what they don’t spell out is that a Statistical Register like the one they are building will link all a person’s previously siloed information. It will link the business register to a person register via IRD and tax data so that they know everything about everyone in real time.

Stats NZ would know your occupation, how much you earn, how much your business profits, your medical history, whether you are ‘up to date’ with your vaccines, your education levels, criminal record and if you are on a benefit.

And the list of data they can access is likely set to grow.

To learn more, see part three of this investigation.

Unpublished reports

An OIA response revealed that Stats NZ received 467 submissions to the consultation on the admin first census. This is a poor turnout for such a momentous change to statistical data collection, and one that indicates it largely flew under the radar of public awareness.

A report on the submissions has yet to be published, and the same goes for the report of the Independent Evaluation Panel appointed last year to review the admin-data-first approach.

However, Stats NZ General Manager, Customers and Partnerships, Sean Broughton said in a press release:

“We heard loud and clear through the consultation process that census data holds great importance and mana for many New Zealanders.

“The submissions provided a clear illustration of what matters to different communities across society. People also expect clarity and transparency about how their data is used or re-used, particularly in an administrative data-first approach.

“We have been given plenty of food for thought, and we look forward to continuing this dialogue with the public of Aotearoa New Zealand over the next few years.”

I followed up with a query to Stats NZ about why these reports remain unpublished. An unattributed response came back:

“The Government Statistician has made an in-principle decision about how the next census will run. This is under active consideration by the Government and announcements will be made in due course.

“Key documents relating to the decision will be released at that time, including the Panel’s report and a summary of public submissions.”

Why Stats is changing the way census are done

The admin-data-first census planned for 2028 is authorised under the 2022 Data and Statistics Act, but a Statistical Register must first be in place for it to function.

The rationale for this new approach to the census is laid out in a March 2024 Cabinet Paper, by Andrew Bayly, the previous Minister for Statistics (Shane Reti was moved into the role in the recent Cabinet reshuffle).

The paper argues that the traditional survey model of doing a census is “vulnerable to events like pandemics and natural disasters”, as well as being prohibitively expensive to undertake. It describes the traditional population survey model as “unsustainable”.

Cook says the ‘cost excuse’ probably arises from the cost-blowout on the 2023 census, which had its origins in repeating the failures of the 2018 census. The 2023 census is estimated to have cost $316 million dollars, and still fell short of the government’s target for a 90 per cent response rate.

Former head of Stats NZ, Mark Sowden, told Checkpoint last year that the cost would not come down again.

But Cook says it had become a myth that the “stupid amount of money” that was spent on the census is typical.

“The fact that we spent 2.1 times in 2023 than what we spent in 2018 … I mean [former Finance Minister] Bill Birch would have my guts strewn around the country. Now, two finance ministers, Robertson and Willis, haven't taken any notice of that cost. That would normally be a trigger for: ‘do you really understand what's going on?’”

He notes the fiscal blowout escaped scrutiny in the external review of the 2023 census as well.

And as for pandemics and natural disasters?

“I mean, I ran the British Census in 2001 in the middle of a foot and mouth epidemic. I find that argument pathetic,” he says.

Data Quality in a Population Storm

Moreover, Cook (and others), are concerned that the quality of admin data does not meet the standard of data collected through a census, and that this change in data collection is happening at a time when the quality of population statistics has never been more important.

This is because New Zealand is in the middle of what he describes as a “population storm”. A confluence of large and volatile migration flows with declining fertility and increased longevity.

“Population data are essential to know the different demographic characteristics (particularly age, ethnic and socioeconomic composition) of populations within geographic areas, for understanding the specific health care needs particular to each region,” he says.

Lack of transparency and proper due diligence

The risks of an admin-first approach remain unknown, according to Cook, who is appalled that both the reports from the public consultation on the admin-first census, and the independent evaluation panel commissioned last year, remain unpublished.

“We’ve had two external reviews of the 2018 and 2023 census. The one statistician involved in both those reviews left the country before those reports were available to other people, so there has never been any possibility of population experts meeting with them.

“The other two reviews – those should not be secret. The public and Ministers should be able to trust methodological advice from the government statistician, and one of the basis’ for that is that advice has been transparent and exposed to experts. The failure to have all of those reviews well-discussed and exposed is a failure of the department.”

Cook points out that a new group of Minsters are being asked to agree on what could be a third badly tested census model, one that influence nearly every spending decision that Ministers will have to make in the next decade.

“Without expert confidence the plan for Census 2028 will work, the next census should be in 2031,” he argues.

So, the confidence of Ministers and the Public is potentially at stake. With census completion rates at lows of 85% and 88% in 2018 and 2024, will even more New Zealanders distrust the government with their data? And with a Statistical Register in place, will trust and social license even matter?

A further round of public consultation is expected in the final phase of the census overhaul, scheduled for the first half of 2025, and will give the people another chance to have their say.

Watch this space.

Follow up questions were put to Stats NZ but they are now treating it as an OIA request and were not able to respond in time for publication. I will update readers with any response I receive in due course.

Originally published on Byline Babylon.