The Government's top statistician, Stats NZ chief executive Liz MacPherson, resigned on Tuesday over what was the worst public response to a Census in decades and a damning report.

So what went wrong, why does it matter and what does it mean for this Census and the next?

Why do we need the Census?

The Census is a statistical survey of all people in the country that is usually carried out every five years.


It's the most accurate picture of who is in the country and where, and captures a range of information to help the Government make a range of decisions, including where to put education and health funding, as well as helping determine what electorates for an election should look like.

The data is considered of the highest standard and is widely used by researchers and organisations across a range of fields.

What was different this year?

The last Census was in 2013. In 2014, the Government approved a plan to "modernise" the process. The questionnaire hadn't been update since 2001 and it needed to be aligned with changes in international best practice. A plan was laid out by Stats NZ in 2016 and the Census was carried out in March 2018.

The biggest change was moving to primarily collecting data online, rather than the previous method of focusing on collecting paper forms.

The moves also aimed to reduce the costs of collections and increase the use of "administrative data".

Administrative data is information Stats NZ can get from other Government records - such as drivers' licences or tax records - to fill gaps where it might not get all the results.

So it went badly?


The response rate from the public was significantly worse than in recent years, with the department calling it "unacceptably low".

For example, the overall national population response rate was 83.3 per cent. That's compared to a return of 92.2 per cent in 2013 and 94.5 per cent in 2006. It was the worst rate of reply in about five censuses.

The return from people aged 15 to 29 was down from 88.5 per cent to 75 per cent.

But the department scrambled to fix the gaps once it became obvious the results weren't coming in as expected. It plugged some major gaps using significantly more "administrative data" than originally planned.

The overall interim "coverage rate" – the total number of people counted using all data sources – was 98.6 per cent – which is actually higher than the 2013 rate of 98 per cent.

That means that the data should be able to broadly serve its function of telling authorities who is living where for some resources and for electorates to be set, according to Stats NZ.

The whole process has significantly delayed the release of the information.

Then what's the problem?

Administrative data is good for some kinds of information – like where people live or how old they are.

But it's not very useful for some personal and detailed information.

In the most notable example so far, we won't be getting official-standard data on how big iwi are.

That's because the response rate among Maori was 68.2 per cent (down from 88.5 per cent in 2013) and other Government departments don't collect enough information about iwi affiliation.

And there could be plenty of other data sets that won't work either, meaning researchers and the Government could be left without some important details.

The statisticians are currently working through those to figure out what they can repair and what will be a write-off.

Some of the data may still be useful for research, but not reach the "gold standard" required for a census and it's not yet clear how it might affect Government policy-making in years to come.

We'll start to learn exactly what is missing when the first data starts coming out in September 23.

What went wrong?

An independent investigation was launched to answer that question and delivered its results on Tuesday, saying while the overall plan was feasible, the problem was in the execution.

It found too much of the focus had shifted to the online work, both in terms of how the Census was communicated to the public and in terms of backup plans, leaving the paper part of the process struggling.

The number of staff on the ground was only about 40 per cent of those in 2013, and was described as "inadequate".

That left some areas and parts of the population, including Maori, ill-equipped to take part. There were only 4800 bilingual forms initially printed.

The review also concluded that as it became clear the responses by paper weren't up to scratch, staff and management were overly optimistic, maintaining they expected the results to come in eventually.

And concerns weren't properly escalated up an overly complex chain of command.

The impact of the 2016 Kaikoura earthquake – that closed Stats NZ's offices – was also underestimated - and as a result staff were left scrambling to do last-minute work on already-late IT systems created for the new processes.

The report said while the budget given for the process ($119 million) was big enough for the original plan, more money was needed in hindsight to cover the risks from the shake-up.

Overall, it found there had been too many changes at once.

What happens next?

Stats NZ will begin releasing its data from September, with an aim of having it all out by the first quarter of 2020.

We'll start learning where the gaps in the data are then.

The report advised against moving the next Census to 2021, instead saying it should continue in 2023 to allow for more planning and fixes.

It made a raft of recommendations for future censuses, saying they need to build on the same changes, but with more funding for contingencies.

It's also called for the creation of an independent body to advise the department on ethics, privacy, security relating to the use of alternative data sources, the establishment of a Design Authority to help set up future surveys, and a Census Board to be installed in the department.

The Government is now considering the recommendations.