Monday, April 29, 2019

NZ’s Census disaster

The 2018 New Zealand census was an utter disaster—there’s no other way to fairly describe it. From record-high non-response rates, a high rate of non-completion, and data of such poor quality that Statistics NZ (Stats NZ) has had to “fill in the gaps” from other sources. Some of the data will be released in September, but the rest, such as it is, won’t be released until mid-2020, more than two years after Census Day, and less than two years until the next one. What’s the point?

One in seven people failed to complete the 2018 NZ Census, a fact that had to be dragged out of Chief Government Statistician Liz Macpherson under a threat of being held in contempt of Parliament for her continual refusal to release the figure. About 700,000 people didn’t complete the census, of whom 240,000 people began the form and failed to complete it, and 480,000 people missed it entirely (we, of course, completed the census).

How did this happen? We don’t know yet. The answers will come from an independent review—which isn’t due until July. In the meantime, it’s widely believed by us ordinary people that two of the factors were that the 2018 Census was, first, that this was the first conducted entirely online—there was no mass distribution of paper forms this year. Second, the previous National Party-led government, which drew up the budget that funded the 2018 Census, underfunded the whole thing (they deny this, of course).

The implications of this are that much of the data will be of little or no value, and that could include answers on things like smoking and religious identity. Those sorts of things are valuable indicators of the social changes in New Zealand, and the sort of thing that people like me use to work out where New Zealand society is headed. For example, the fact that New Zealand was becoming more and more secular meant that it was unlikely there’ll be any strong religious objection to marriage equality, and, in fact, there wasn’t.

The only data we have is based on the 2013 Census with some extrapolations since, and it shows that New Zealand is 47% Christian, which means the percentage that call themselves that, regardless of whether they actually have any affiliation with or practice the religion. Combined with the roughly 6% who have a non-Christian religion, roughly 53% of New Zealanders say they have a religion. At the same time, 41.92% of New Zealanders say they have "No Religion", a figure that obviously includes atheists and agnostics, but that also includes people who are religious or spiritual, but who have no affiliation with any organised or obvious religion.

There’s been speculation for years that that at some point New Zealand will become majority “no religion”, and it was expected to be reported in 2013 (it wasn’t), and then maybe in the 2018 Census. Now we know that whatever is reported is unlikely to be reliable. So, we now have to wait for the 2023 Census, by which time New Zealand will probably be majority non-religious.

Here’s the big issue, though: How do we know for sure that any of the other data is accurate? In some cases Stats NZ basically “patched” the data with information from other government sources. Somehow that just doesn’t sound very reassuring.

The Census data is used to decide where government money should be spent, to evaluate how effective government programmes are, where new schools should go, and local governments use the data to work out things like where parks should go. The data is also used to determine how electorate boundaries are drawn, leading some to demand that the current boundaries be retained until after the 2023 Census when—we hope!—the data will be more robust and reliable. I don’t know that’s necessary, not if Stats NZ can assure us that data is sound, but that could be a big ask.

There’s one final aspect of this that’s slipped through the cracks: This disaster has shown that we’re nowhere near ready to go online-only for important things like the Census—or voting. I’ve become more and more pessimistic that an online voting system can ever be made safe from hacking, and this debacle shows that we can’t even be sure that it’s robust enough to ensure every vote—or, rather, every voter—is counted. In any case, there’s clearly a lot more work that has to be done before we can move to online voting.

For me, this disaster is mainly just a matter of frustration and disappointment, because it means we won’t have the useful data on society and population that we normally get from a census. For all of us, though, this disaster will have big implications because it will often leave government in the dark as they make decisions, and that’s bad for us all.


rogerogreen said...

I'm worried about the US 2020 Census, which will be online and will likely have that damn citizenship question, if SCOTUS watchers are correct.

Arthur Schenck (AmeriNZ) said...

I was aware of the current regime's attempt to politicise the census, but I wasn't aware of the online aspect. That could result in an enormous undercount—probably by design.