How reliable are New Zealand’s polls? A look at who’s asking and how, as the country heads into election day

Nobody watches polling closer than a politician who says they don't care about the polls.

The only poll that matters is on election day, they say on a bad poll day. Our hard work and good policies are resonating with the public, they say on a good poll day. 

They are right in the sense that the poll that determines the leadership of the country, the size of their caucus and the job prospects of their political colleagues is that on election day, but the snapshots of voter sentiment aren't simply ignored. 

Political leaders risk being rolled as their MPs study the numbers and backbenchers conduct the internal maths of whether or not they look like they could lose their seat at the next election.

As well, things are slightly more complex when it comes to polling this election. A major shift in polling has been the introduction and roll out of the use of online panels as largely the main source of contact with voters in New Zealand. (Curia and Verian still partially using phones).

An online panel can be a pre-recruited group of people who agree to take part in digital market research studies, surveys, or focus groups over time.

Overseas, the New York Times/Siena Poll experimented polling with text messages, sending mail with financial incentives, data from online panels - but went back to phone surveying.

"The phone, while expensive, is still much faster and cheaper than high-incentive mail surveys. And the phone still seems to reach a more representative population - based on measures like educational attainment and record of voting in recent elections - than supplementing the phone with text-message links to surveys or online panels," New York Times chief political analyst Nate Cohn wrote.

And advanced voting - although not new - also puts a spanner in the works when it comes to predictions. In New Zealand, exit polling or polling someone on how they voted is not allowed. However, pollsters can ask if a person has cast an advanced vote, and also ask the normal party support question along the lines of who they would vote for if voting today.

Because polling methods are quite different, with different polling lengths (such as the Curia and Verian polling periods being shorter than Reid Research and Freshwater), contact methods, weighting and margins of errors, rather than pulling the averages, we compared the party support suggested by each poll individually for the major parties, then across the political spectrum for each of the major polling companies. The polls compared here had at least 1000 participants and public information regarding methodology.

Labour and National support in political polling in 2026

(Please note the x-axis scale is non-linear and aggregates data by month before shifting to specific dates within October. Exact polling dates for each company are listed in the graphs below). 

Taxpayers' Union - Curia Poll

Curia Market Research runs the poll for the Taxpayers’ Union of 1,000 adult New Zealanders, weighted to the overall adult population.

Curia runs its polls by phone, about 700 landlines and mobile numbers, and 300 by online panel. That comes from a "random selection of 7,000 NZ phone numbers (landlines and mobiles) and a random selection from the target population from up to three global online panels (that comply with ESOMAR guidelines for online research)", it said in its last poll. 

"If the call is to a landline, the person who is home and next has a birthday is asked to take part. Those who take part through an online panel are excluded from further polls on the same topic for six months."

1News Verian Poll

The 1News Verian Poll (formerly the Kantar and Colmar) splits its poll by mobile phone (about 500) and online panels (500). It is weighted to align with Statistics NZ population counts for age, gender, region, ethnic identification and education level.

Verian polls voters via mobile by randomly dialing, "using probability sampling, and the online sample is collected using an online panel".

RNZ-Reid Research

Methodology from 2025 stated Reid Research used "one of several Online Research Companies’ Panels which comply with the ESOMAR guidelines for online research". Of its margin of error (most polls just disclose it is about plus or minus 3.5% at the 95% confidence level), RNZ Reid Research breaks it down further:

10% or 90%: +/- 1.9%
20% or 80%: +/- 2.5%
30% or 70%: +/- 2.7%
40% or 60%: +/- 3.0%

It had a maximum margin of error of +/- 3.1% at the 95% confidence level for its latest poll. It also uses quota sampling and weighting by age, gender and geography. 

The Post/Freshwater

In its February poll, Freshwater conducted its survey online by sending invitations to members of online panels to participate. 

Non-response from different demographic groups was taken into account, and results were weighted to reflect the profile of New Zealand voters. 

"After base vote intentions are produced, 'lean' responses are included, remaining undecideds are removed to produce a 'valid' result, then finally results are filtered for likely voters," it said. 

The margin of error was plus or minus 3% at the 95% confidence level.