The release of data from this year’s census has been delayed because not enough people participated in it:
Stats NZ has revised the date for first release of census information from October 2018 to March 2019.
We will confirm the exact response and the coverage rates for the census after we complete our reconciliation processes. Stats NZ’s interim calculations show that full or partial information for at least 90 percent of individuals was received, compared with 94.5 percent for the 2013 Census.
As with previous censuses, we will use statistical methodology to compensate for missing data. For the 2018 Census we are revising this methodology because of the lower-than-expected response. We are discussing this new methodology with our technical customers. We’re also undertaking analysis on how to improve data for small populations, subgroups, and small geographies. The new date for our first release will give us time to develop revised methodology for processing and analysing census data. We are committed to delivering a high-quality and accurate dataset.
There is a long term, international trend of declining census response rates. Because of this we have made a strategic decision to use more administrative data to improve the quality of census data.
Stats NZ is in a good position to adopt this approach as we have been investigating future census models that would supplement census data with administrative data.
How significant is the drop?
Over at Kiwiblog David Farrar says:
. . .The Minister of Statistics should call for an independent review of this failure, to ensure the next census has a much higher participation rate.
Also we should not be given spin for months about how great the census went and then find out only now, how bad the participation rate was.
The last Australian census had a 96% response rate. They regarded 93.3% as the minimum required.
The Canadian census had a 98.4% response rate.
A better way to look at it is the non response rate. In Canada is was 1.6% and in NZ it was 10% – six times higher.
The move to on-line forms was supposed to make it easier to complete the census.
We won’t know if completion would have been worse if Stats NZ had stuck to the paper-based system but there were lots of complaints from people about the difficulty of dealing with the call centre by those requesting paper forms.
We were in Queenstown on census night. Nothing was said at the hotel when we checked in but there were papers under the door when we left at 6:30 next morning. I picked them up, stuffed them in my bag and forgot about them until I got home when I found forms at the door.
We had breakfast with around 50 farmers from Australia and New Zealand so I did a census on census completion and asked everyone if they’d done it.
One of the hotels had run out of forms, none of the Australians had completed them and all of the New Zealanders had done theirs at home, as if they were at home.
A woman who gave forms to tourists staying at her B&B was told where she could put them.
A friend has a holiday home in Wanaka with two houses and one mail box. She presumes her tenant would have got the letter fromStats NZ that went to every household, but would have filled it in for only one of the houses.
Anecdotes don’t make good data but they do illustrate problems with this year’s census.
With the old system someone visited every house and some census staff went many extra miles. A friend was climbing in the Southern Alps on census night and was presented with forms by a worker who came to the hut.
The on-line census was easy if you have a computer and are comfortable using it but that’s not everybody and problems with the call centre didn’t help.
It would have been more expensive to have people calling on every house as they used to do, but it would have ensured a better count which is important for planning and funding.
It’s also necessary for working out electorate boundaries and the delay in the data release will delay the final release of new boundaries which in turn will delay the candidate selection process.
As National’s Statistics spokesman Nick Smith points out:
. . .There is over $10 billion of health funding allocated to the twenty DHBs each year based on census population data. The funding formula for the operating grants for our 2500 schools is derived from the census as are decisions about the allocation of resources in social services, police, sports, transport and many other services.
“It also has major implications for the Representation Commission. The number of general and Maori electorates in Parliament are determined by the Census and the process for determining the new boundaries was due to start in November.
“Changes in population figures as small as 1 per cent can impact on whether there is, for instance, an extra or the removal of one of the Maori electorates. This process will now not be able to start until April next year and the compromised statistics will affect the integrity of the make-up and boundaries for the 2020 and 2023 elections. . .
The significant drop in completion rates and consequent delay in releasing data are concerning.
Having 10% of the population not counted is serious, especially when it is likely to include more people who for example have intellectual or mental disabilities, don’t speak English, are illiterate or for other reasons are more likely to be in need of publicly funded support.
A review must determine what went wrong and why, and ensure that it doesn’t happen again in 2024.