Public trust and confidence, especially among Māori and Pacific men, looms as a problem for the census, internal documents reveal.
Already, Stats NZ has partnered with iwi in certain areas to collect data, and tried to garner support through sponsorships, like $350,000 spent on Te Matatini – the national kapa haka festival held at Ngā Ana Wai/Eden Park last month.
Contingency plans have been created to boost response rates if they’re low, like they were for the 2018 census.
But with budgets running thin, and areas affected by cyclone damage being given a time extension, there’s every chance Stats NZ, the government department behind the census, is bracing for another undercount.
Newsroom has been trawling through more than 1000 pages of census documentation, released to us under the Official Information Act last month. The material ranges from minutes from monthly census programme board meetings, to operational readiness assessments.
According to December research undertaken for the census to gauge public trust and confidence in the government, awareness of the census was ahead of where it was five years earlier. Trust in Stats NZ remained “steady” at approximately 67 percent.
However, a January report focusing on external risk factors, summarising the research, said there were no big shifts in the intention to participate.
“The research showed that there are sizable percentages of Māori men (19 percent) and Pacifica men (16 percent) not believing our ads and messaging on privacy and security.
“Funding has been approved for mitigation actions to alter campaign content, boost influence and support direct engagement with PRG [priority response groups], with a focus on Māori men and Pacifica men.”
An important goal for this year’s census is to get national response rates of more than 90 percent for the total population, people of Māori descent, and Pacific peoples.
The fear is there will be yet another undercount of Maori – a problem which dates back to 2013.
Census campaign tracking research from August last year surveyed 500 people online, and interviewed 100 people on the streets of South Auckland and Whangārei. It said the intention to participate had dropped to 51 percent in Māori and Pacific men aged between 30 and 49.
“I’m not sure it will be used to my benefit,” said a Pacific male, aged 30-49. Another Pacific man, aged between 18 and 29, told interviewers: “I personally don’t think it is worth my time.”
A Māori female, aged between 30 and 49, railed against negative statistical narratives. “I would rather exclude myself from the statistic.”
While more people had heard something about the census than the last survey, thanks to video campaigns, attitudes had declined and intentions to participate had dropped.
The report from August said: “Seeing the campaign is giving some people the impression that the census is imminent and making them think they’ll miss it because they’ve had no information about how to take part.”
However, it was noted, “community and customer engagement activity” would increase from October 1, including a regional tour of more than 40 locations, as well as “eight regional community engagement teams” with “tactical plans”.
Stats NZ’s iwi-led data analytics and collection initiative, known as Te Mana Whakatipū, is being run in the North Island communities of Tairāwhiti, Ōhua (the far north), and Te Whānau-ā-Apanui in the eastern Bay of Plenty.
Stats NZ’s January assessment of “Iwi/Māori sentiment” said relationships within Te Mana Whakatipū are being “managed closely”. The report described the Te Matatini partnership as successful, and noted the department had other iwi relationships.
“There is a risk that if Stats are unable to partner effectively this may impede progress. Anti-government sentiment is still prevalent [and] could cause risk to Māori trust and confidence and their participation.”
There were more illuminating comments in a peer review of the census preparations by the Australian Bureau of Statistics.
The report, completed in December, said the “barriers to participation” are highlighted by Stats NZ audience research, and the “shift to participation is not fully understood or how this will play out”. “Reaching priority groups (Māori and Pacific Peoples aged 18-49 years, young people, rainbow communities) is a known challenge.”
The Stats NZ social media team had flagged “low engagement and followers on its channels”, which ABS suggested could be lifted by a “suite of unpaid/organic social media content”.
Other risks to the census’s success included “Iwi and Māori engagement, anti-government sentiment and external lobby group voices”. (The appropriate response, ABS said, was to establish media and communications protocols with “key stakeholders, vendors and government agencies”.)
The census board remained confident of its collection model, however. A suggestion to put temporary collection boxes in places like supermarkets was binned.
Back-up plans laid out
A report presented to the census board in November discussed the prioritisation of households who hadn’t completed the census.
Tracking responses in line with those in 2018, reports stated the risk of not meeting national response targets would be flagged two days after census day. The risk of not meeting Māori response targets should be clear five days before census day.
“By nine days after the census: We plan to have visited all non-responding and partially responding households; this cycle will run twice weekly.”
It is through partial responses that Stats NZ might give itself some wiggle room to boost participation.
A paper in January updated the census board on something called “reduced data capture”, or RDC. While the first preference is for all people in New Zealand to answer all census questions, the paper said for some, especially those in challenging situations, “this is unrealistic”.
The paper mentioned rough sleepers or homeless people, and those in care facilities.
But a census oversight group called the design integration forum also approved the use of RDC “by collectors as a contingency option during operations, if approved by OLT [operational leadership team]”. It could also be used by team leaders if people persistently refuse.
A graphic in the report lists the “absolute minimum” responses as: name, date of birth, usual residential address, Māori descent, and iwi affiliation.
“Note, this contingency option has not yet been trained or communicated to field staff to prevent early usage,” the report says. “This option should not be triggered until at least two weeks have passed since census day (ie. 21 March). We should have strong evidence that we’re not on track to meet our response objectives.”
A paper to the census board meeting in January said a range of interventions had been considered if response rates are “lower than desired”, including “additional visits, additional field staff, additional reminders, additional community engagement and activation, additional localised advertising/marketing…”
The following passages were redacted.
Stats NZ’s ability to respond to low response rates has a question mark over it, in the form of a dollar sign.
Overall, the five-year census is forecast to cost $264 million, and budget commentary in December stated: “It is anticipated the remaining contingency of $6.4 million will be spent.”
The horror 2018 census cost just $126 million. (In saying that, a $1.65 million overspend was added to the 2023 census.)
Recruitment contract balloons
Just one example of cost escalations is the external contract with recruitment firm PersolKelly. We reported last year the original contract was worth $42 million.
But this fresh batch of documents state that with extra money approved by Cabinet for pay rises, and including expenses, mileage and allowances, that contract had ballooned to “circa $56.8 million”.
The situation forced Stats NZ to ask its chief financial officer, and Treasury, if chief executive Mark Sowden had the delegated authority from Cabinet to approve such a large sum. (The answer was: he did.)
A financial paper to the census board’s January meeting noted: “The programme does not hold any contingency for high-impact, low-risk events, for example, natural disasters and Covid-19 resurgences; [and] the programme does not hold any contingency for extending NRFU [non-response follow-up] as a result of low response rates.”
That’s not to mention disruption from potential cyber attacks.
An independent quality assurance review from last November stated: “Programme contingency is lower than desirable. Appropriate actions are being taken to free up contingency through tough spend control, reforecasting and seeking efficiencies.”
A paper from June last year, labelled “cost levers – not yet activated”, suggested delaying census data release dates beyond mid 2025, so costs would shift to the next census’s baseline funding. That old chestnut.
Another big headache for census officials covered in the documents is problems with the data delivery systems – the back-end systems used to process and analyse data.
“There is nervousness among some of the team that data quality issues might not get picked up before data is released to customers.” – ABS report
The Australian Bureau of Statistics report from December said: “With the focus on collection operations, there is a risk that data delivery is neglected.
“The inability to thoroughly test in the dress rehearsal and the planned release of systems so close to census night represent a significant risk.”
Prior to the creation of a dedicated project manager role for data delivery in August last year, “there was no detailed planning for the new data delivery systems”.
To make things worse, “there is no contingency available for unexpected complications that may arise in the months leading up to February”.
Time constraints mean final testing, including how systems work with each other, will be done “live”.
November’s independent quality assurance report described as “risky” the constrained data delivery implementation and testing window. “There is nervousness among some of the team that data quality issues might not get picked up before data is released to customers.”
In an emailed statement, Simon Mason, the deputy government statistician and deputy chief executive of census and collection operations, says at the time of November’s report Stats NZ was slightly behind self-imposed deadlines for systems readiness. “However, a dedicated team has worked to ensure all the systems that are required to support the collection of 2023 census data are completed.”
In summary, the report said census collection readiness was “good”, while the readiness of data processing, analysis and dissemination was “lower”. The census board approved an extra $1.8 million to the data delivery project.
The picture painted by this wodge of documents is unfortunate: A cyclone-affected census that finds itself squeezed financially, that might need to find more money for an urgent response if return rates are low.
(Although it did tuck away $500,000 of the $7.9 million budget boost from Cabinet for “field collections contingency”.)
Te Pati Māori president John Tamihere, speaking on Sunday’s Q+A programme, didn’t have much faith in the census, saying it hadn’t engaged with Māori communities and there’ll be “a significant undercount”.
Yesterday, census officials said only 1 million people had filled out their forms.
Organisers will be hoping for a big turnout today – census day – to turn its fortunes around.
* This story was updated on Tuesday morning with comment from Simon Mason.