Census Mess Can Be Resolved with a New One in 2021

I was commissioned by the ‘Dominion Post’ to write an opinion piece as part of their review of the anniversary of the 2018 Census. This is a slightly revised version of what they published. The main article is ‘365 days and still counting:  Census results still nowhere to be seen’.  An earlier ‘Pundit’ column is ‘The Census Flop‘. 

The test for a census has to be that it will stand up in a court of law. Its uses are so widespread, so fundamental to the governance of New Zealand, that there need be only one aggrieved party from a decision involving a census to end up in court. This has always been true but the 2018 Census of Population and Dwellings appears particularly flawed and therefore more likely to be litigated.

To avoid litigation the district health boards have decided to remain with the 2013 census. The Electoral Commission has no such discretion; the law requires them to use the latest census.

Some MPs have already said they have no confidence in any boundaries based on the 2018 census. An MP who disliked the Electoral Commission’s decisions may well go to court. Given the technical issues involved, the judicial process is likely to push the final determination of boundaries to a date too late for the 2020 general election. I am told, by people more informed than I am, that the courts may decide that the 2018 Census did not meet the standards required in the relevant statutes.

Every census has glitches, but the 2018 one had a major defect resulting in an exceptionally low ‘coverage ratio’, or the proportion who filled in their forms. Instructively, a year has gone by and we do not know that the exact ratio is but it is thought that about one in ten New Zealanders failed to file a return – double from what is normal.

Even Statistics New Zealand has no confidence in the outcome and is trying to patch the gap by using other data sources. For many purposes the resulting data-base may be adequate, although there may be larger margins of statistical error. However a court is unlikely to be convinced that the result meets statutory requirements.

The major defect was the inadequate provision of enumerators, the people who knock at your door (and find dwellings not on the public record), help you fill in the forms and go back to remind those who have not. Figures released by Statistics New Zealand suggest that only half the resources were provided for this purpose in 2018, compared to the 2013 Census.

The reasons for the reduction are beyond comprehension. So the government cuts back the resources; you tell the government you cannot do the job that it and statute ask of you. You think you can double enumerator productivity because you are expecting 70 percent to respond electronically. You do not have to be Einstein to know that the other 30 percent are those that require the most intensive work from the enumerators.

The anecdotal result seems to have been poorly trained and managed enumerators, and with fewer followups. The quantitative consequence of halving the resources has been a doubling of the coverage failure. (Twelve months after the failure and Statistics New Zealand has not provided a detailed review.)

What to do? When the failure became apparent nine months ago it was proposed that there be a new census run in March 2021 (just two years from now) and that the 2020 general election be run on the existing boundaries based on the 2013 Census. That has the additional advantage of realigning New Zealand census dates to the Western standard of years ending in 1 and 6; the sequence was disrupted by the Canterbury earthquakes.

Statistics New Zealand has claimed that it takes three years (instead of two years and nine months) to run a new census. But I am told that if they use the existing reliable parts of the 2018 census with the 2013 enumerator system, it will take 18 months to organise.

So it could start the task as late as this September. All it requires is leadership, although it may be sensible to bring back (sometimes out of retirement) the team that ran the 2013 enumeration.