Aside from its Orwellian doublespeak, the proposed legislation to change how we count New Zealand requires an expert review.
In early April, the world’s biggest census ever began; it is a $2b exercise covering 1.4b Indians. It is so complex it will take a year, as first they count the households, then the persons living in them. The results will have controversial impacts on both the parliamentary balance of seats and on fiscal redistribution. Nevertheless, despite the implications and despite the cost, the Indians are proceeding with their big count.
New Zealand seems to be going in the opposite direction if our Parliament passes the Data and Statistics (Census) Amendment Bill before it. By way of background, the Data and Statistics Act 2022 currently requires the Government Statistician to take a census of the population and dwellings of New Zealand in every fifth year after the previous census.
The invitation to make submissions to the select committee states: ‘The Act implicitly [sic] assumes that it will be a large-scale survey involving the total population of New Zealand. The Data and Statistics (Census) Amendment Bill would amend the Act to update the Statistician’s census obligations under a modern [sic], administrative-data-first census [sic] approach. Under this approach, future censuses would be taken using administrative data as the primary source, which would be supported with survey data.’
There are weasel words here. The ‘implicitly’ is unnecessary. Any dictionary will confirm that the normal meaning of ‘census’ is ‘a large-scale survey involving the total population’. A count based on ‘administrative data’ is not a census in the normal meaning of words. As for ‘modern’, apparently those poor benighted Indians, and just about every other jurisdiction in the world, are presented as living in the past.
The bill’s is Orwellian doublespeaking – a practise of many politicians and raised to an art by Donald Trump – where the definition of a word is twisted to mean its opposite. The proposal is not about modernising the census but replacing it with a different method of counting the population. We’d prefer our MPs were not hypocrites nor that the courts would get into a tangle over the meaning of words such as ‘census’. Parliament should make that honestly clear in the proposed legislation that it is replacing the Population Census.
Doublespeak aside, should we replace the census as our main form of a national count?There are some good reasons.
First is that a proper census is expensive. The 2023 census cost more than $300m. It is claimed the new approach will be cheaper, although the ‘supported with survey data’ gives no indication of how many surveys will be necessary. The fiscal requirements of the new approach will give the government the opportunity to control what is known about New Zealanders, which may not be the same as what we want to know, nor what good government requires.
The second is that the last two censuses have been botched with unsatisfactorily low response rate. (See here for the 2018 census.) Part of the problem has been that the State Services Commission has been appointing Government Statisticians who were generic managers with little experience in statistics or its culture, let alone in managing a population census. To protect the SSC from perceptions of further failure, we have descended into doublespeak, proposing to abandon what their generic managers cannot manage.
It is true that both the 2018 and 2023 censuses were partially rescued by recourse to administrative data. But that is not evidence that an administrative data system alone will work.
I’ve had a bit of experience of the problem. Historically, Statistics New Zealand used administrative data to update the population statistics after the census was taken. At the next census it would be found they were not quite right and the interim years had to be realigned. Fair enough, but I can report it was a tedious exercise for the serious researcher to have to go back and change all the data that they had been working with.
Allow me a personal note. I have been using census data for more than 60 years, including censuses as far back as the first ones in the middle of the nineteenth century. I know quite a bit about their difficulties. The historian in me has been particularly frustrated by their absence in 1931 (the middle of the Great Depression) and in 1941 (the middle of the Second World War). The deficiencies in the 2018 census have upset my contemporary work – and made it difficult to check the 2023 census. I guess I am a bit of a conservative, although I try very hard to come to terms with contemporary social change. Ironically, I use the Population Census as an important means for studying the growing diversity.
One of the issues which arose on the 2023 census was that the enumerators experienced a far higher level of abuse – some physically threatening – than in previous censuses. Why? It may be because of the poor publicity before census day. (Apparently they relied on social media so I didn’t see it.) Or it may be an ongoing breakdown in social coherence – we need a decent census to help track it.
We don’t know whether a new system based on administrative data will work. It will not cover all the traditional variables in the census – religion for instance, or iwi membership – and other variables are problematic – people change their ethnicity depending on circumstances and the administrative record ought to reflect this but it will muddle the population count.
It is unlikely that the new approach will give the precision needed. Precision depends on the size of the samples necessary to supplement the administrative data. Especially for small groups in the population, the need is for very large (and correspondingly expensive) samples. I do not see any future Minister of Finance looking favourably on every request for such funding. Yet understanding these small groups is critical for understanding the way that Aotearoa New Zealand is evolving. I am not even sure that a cheap sample survey will give the precision some Māori; one of our largest minorities, will require.
What is making an assessment difficult is that there is little published material which demonstrates that the new approach will work. The 2018 and 2023 censuses introduced untested methods without sufficient prior validation – they did not work.
That is why there is a call that the bill not proceed until the select committee has had advice from an independent expert panel. (To make it clear; by ‘expert’ is meant they will have run successful census operations – some will have to come from overseas – and are not generic managers or political/woke appointments.)
That could delay the implementation of the big count by one year. Currently the bill says 2030. Any competent social statistician will ask why not wait to 2031? That is the year most countries do their (proper) population censuses; we would get back to alignment with them. That this did not seem to occur to those promoting the bill tells us something about their lack of expertise.