If we don't count, we can't help: How ending the census obscures poverty

“Tātou, tātou – We all count” was Statistics New Zealand’s slogan for the last Census. Now the Census has been scrapped, independent researcher and housing analyst Greg Waite has a simple question: What’s the plan for counting those who are homeless, marginalised, or who simply have no means to access the system?

On June 18, 2025, the Government announced it would no longer conduct a national Census, the five-yearly survey of all New Zealanders. Instead, it plans to rely on administrative data already held by government agencies, supplemented by smaller annual surveys and targeted data collection.

In my research I use the Census in combination with administrative date to analyze income poverty and housing affordability. I would like to clearly note some serious negative consequences of this surprise cabinet decision to end the Census.

I currently use the Census as a source of data on:

1. household and family structure,

2. the full tenure spectrum, including informal rental (no bond),

3. unoccupied dwellings

4. rents, including informal rents

5. bedrooms and overcrowding

6. substandard dwellings; damp and mould, lack of essential services

7. homelessness

As I understand it, 1-4 will be less complete with administrative data in 2030, while 5 and 6 will be available less accurately from surveys, and the homelessness measure will be lost.

As Statistics NZ research has documented, the administrative record of household and family structure is partial. Knowing the makeup of households is essential for many researchers, and essential for affordability analysis, so my future results for Child Poverty Action Group will have reduced i.e. partial coverage.

The decision papers submitted to cabinet include these statements “Improving admin data will require collective effort across government,” “Other aspects of data quality, including accuracy, granularity, and coverage, will improve over time,” and “I propose to develop two joint directives concurrently: a Cabinet directive applied to relevant government agencies, and a statutory directive applied to Crown entities”.

In my experience with NZ administrative data, the gaps are a consequence of the naturally incomplete coverage of administrative data collections. These are typically not easy to change.

Related to this, the cabinet papers claim: “Our desired shift towards an administrative data-first approach aligns with a growing international trend of National Statistics Offices (particularly in OECD countries) moving away from only using traditional data collection methods.” Those countries typically have compulsory population registration processes, which NZ does not. Countries like NZ - Australia, Canada, UK -  have not committed to an administrative census.

Most critically, the decision to end NZ’s census overrides the advice of Statistics NZ’s independent evaluation panel, which recommended a full census in 2028. This cabinet decision is described as “an in-principle decision on the manner of taking the next census.” The principle here is simply making operational decisions to stay within the reduced budget which cabinet has set: “we have found that the five options considered by the Panel significantly exceed the available funding”.

The supporting papers also claim: “The new approach will immediately improve the frequency and timeliness for much of the data that is provided through the census.” This is only true for published statistics. Survey data is much less suitable for social research. The census has always been widely used by researchers and community advocates, whereas survey level is less suitable because of its limited accuracy and drill-down.

Finally, the risk analysis appears to make some assumptions in favour of a pre-determined preference for budget cutting. In Appendix 1, ‘Delivery risk - other’ includes low public engagement. Continuing the census is rated  ‘high risk, low control’ while the change to a primarily administrative collection is rated ‘medium risk, some control’. The proposed supplementary Census Attributes Survey, where the sample is described as up to 5% of the population annually, is likely to face the same or perhaps greater public scepticism, resulting in further reductions in the response rate. Also, without the publicity which accompanies a census, public engagement may be reduced. Perhaps an alternative, shorter census might have improved both cost and response-rate outcomes?

The rising scepticism about government data collection we have seen through Covid and after is not a given. It is a product of the misinformation paid for by conservative advocates of various kinds, supported by the growth and profit focused US social media corporations. Responding to this by giving in is counterproductive. Governments need to act in support of openness, to rebuild public confidence.

Addressing Data Quality Gaps If The 2028 Census Is Not Reinstated

Since a reversal of this decision is unlikely under this government, I therefore emphasize the importance of prioritizing development of standard coding and data in the IDI (research data repository) which optimizes the  partial coverage of household type and family type, tenure and landlord type, and private rent.

In particular related to rent, there are significant technical challenges to replacing the census with administrative data on rents recorded when the bond is lodged. The data is partial, omitting NZ’s large informal rental market (family, friends and unregistered landlords of lower quality homes). The bond rents are only accurate at the time tenancies start, but many ongoing tenants pay rents up to 30% below new tenancies.

Additionally, to fill gaps after the census ends, SNZ survey data needs to include bedrooms and overcrowding, substandard housing (damp, mould, lack of essential services) and homelessness.