April 17, 2026

$325 million census scrapped to save money on a system that costs more

Tracing spiral arms in infrared (potm2411a)

A budget cut dressed as modernisation

New Zealand has legislated the end of its traditional census. The Data and Statistics (Census) Amendment Bill came into force on 2 March 2026, replacing the five-yearly population count with a combination of administrative data and small annual surveys. The next full enumeration will not happen until 2030 at the earliest.

Statistics Minister Shane Reti pointed to ballooning costs as the primary justification. The 2023 census cost $325 million, up from $104 million in 2013, with the 2028 edition projected at $400 million. But the government’s own Regulatory Impact Statement reveals what actually happened: all five options initially considered “significantly exceeded” available funding, forcing officials to develop two cheaper alternatives. This was not a modernisation decision. It was a fiscal one.

The cost argument collapses further when you look at returns. Stats NZ’s own analysis, surfaced in the NZ Council for Civil Liberties submission, shows the 2023 census produced $2.8 billion in value against its $326 million price tag. That is an $8.50 return per dollar spent. The Reserve Bank estimated even slight improvements in data-driven economic decisions could be worth $100 million per annum.

Stats NZ’s own numbers say the replacement is not ready

The Technical Feasibility Study contains 27 explicit admissions of unpreparedness. Stats NZ states directly that it has “not yet demonstrated the ability to provide a full census of dwellings” without traditional field enumeration.

The accuracy gaps are not marginal. The 2015 quality standard requires measuring the Māori population to within 0.5%. The new model achieves only 0.9%. For age-sex population groups, the standard demands 90% of groups fall within 1.5% accuracy. Stats NZ manages 72%. The agencies meant to feed data into the new system have warned they “do not expect to be able to make changes” within the required timeframes.

Former Government Statistician Len Cook has written to every MP warning that New Zealand will be alone among comparable countries in adopting this method without a compulsory national registration system. The Nordic nations that successfully use administrative data have exactly that infrastructure. New Zealand does not.

Errors compound without a reset

The Infrastructure Commission identified a technical risk that should concern anyone making location-based investment decisions. Historically, administrative data errors were corrected every five years when the census provided a full population reset. Without that anchor, errors drift and compound invisibly.

The Commission called the census “the single most important collection of data supporting our understanding of New Zealand’s population” and warned that infrastructure demand forecasting would be “challenging, potentially compromising investment efficiency and value for money.”

For insurers modelling risk by postcode, retailers sizing catchments, developers timing residential projects, or aged care operators forecasting demand, compounding data drift is not abstract. It is the difference between a sound investment and an expensive mistake.

The worst possible timing

Cook describes the current demographic environment as a “population storm” where falling fertility, rising life expectancy and volatile migration flows are hitting simultaneously. He estimates the fiscal impact of demographic change from static policy settings will match a fall in government revenue of more than 30 percent by 2045.

The housing sector is particularly exposed. The Ministry of Housing and Urban Development says census data is “crucial to its ability to forecast, plan and invest” in housing and urban development, with no administrative data available for critical variables like access to basic amenities. Severe housing deprivation estimates depend entirely on traditional census collection.

Meanwhile, the overhaul creates a constitutional wrinkle. Electoral boundary reviews must follow each census under provisions requiring a 75% parliamentary majority to change. Delaying the census to 2030 means no boundary review before the 2029 election, with some electorates already 14.3% outside average population.

Business is absent from a debate it will pay for

The government has a real problem. Two consecutive Government Statisticians resigned. Costs tripled. Response rates fell. But the solution its own documents admit cannot meet basic accuracy standards is not a fix. It is a transfer of risk from the public balance sheet to every private decision-maker who relies on population data.

As Cook told RNZ: “There will be turmoil if decisions are driven by political instincts rather than sound statistics.” The businesses that will feel this most, from insurers to developers to infrastructure investors, are largely absent from the debate. They are losing a public good they do not yet know they depend on, and by the time the data gaps become visible in bad forecasts and mispriced risk, the damage will already be done.

Sources

Subscribe for weekly news

Subscribe For Weekly News

* indicates required