Political polls like the RNZ-Reid Research poll look at how people may vote and what they think about current hot political topics. Photo: RNZ/Tim Collins
Explainer - Today sees the latest RNZ-Reid Research political poll, where we take the pulse of Kiwis on the state of the country.
But how does polling actually work, and what are some popular misconceptions about it?
How does the RNZ-Reid Research polling work?
RNZ joined with Reid Research in April 2025 to conduct political polling several times a year. Reid Research had previously worked with the now-defunct Newshub from 2009 to 2023 on polling.
The polls involve 1000 online interviews conducted nationwide from a pool of panelists assembled by Reid and augmented by other research firms. Eligible New Zealand voters aged 18 and over are polled.
RNZ's acting political editor Craig McCulloch said polling is the best way of testing the public temperature outside of an actual election.
"You have to remember: the major political parties shell out big money on private polls to guide their decision-making," he said.
"Public polls give us a glimpse behind that curtain, revealing not just what voters are thinking, but also what's shaping the choices politicians make."
The polling uses quota sampling and weighting to ensure representative cross sections by age, gender and geography based on the latest Census data. It includes rural, smaller urban and main urban areas.
The very nature of polling has changed a lot in recent years. Ngaire Reid, the founder of Reid research, started her company in 1996 after working with other firms including the Heylen Research Centre.
"In that time we've gone from doing the interviewing face-to-face, to telephone and now it's basically all online."
Surveys are typically conducted over the course of a week, including a weekend.
"The reason being that we need to make sure we are covering all times of day, weekdays and weekends, to collect a random sample of people - not just people who are at home during the day - where we'd just get a sample of people who are either elderly, unemployed, unemployable, shift workers, those with young families, etc," Reid said.
Reid Research's database collected over the years has grown to about 40,000 people.
All of the major polling companies in New Zealand - with the exception of Curia Market Research which has resigned its membership - are members of the Research Association of New Zealand and agree to abide by their NZ Political Polling Code.
The code, which can be read in full here, lays out best practice guidelines for how polls are conducted and how media should report them.
"Inaccurate polls or polls that are reported inaccurately can impact on voting attitudes and behaviours and thus influence the democratic process," the guide states.
Murray Campbell, polling spokesman for the Research Association, said that so far, the kind of deep political polarisation seen in other countries isn't as obvious here.
"New Zealand is still a relatively straightforward place to conduct a poll," he said. "There's still more homogeneity than divisiveness in our points of view."
Nicole Satherley, Lara Greaves and Andrew Sporle, academics and researchers with iNZight Analytics, wrote a paper, 'Understanding Public Opinion Polling in Aotearoa New Zealand', published in 2023 in the New Zealand Population Review which analyses much of the methodology that goes into polling in Aotearoa.
"Changes in technology are definitely making political polling harder, especially as the use of specific communication technologies is so different between age groups," Sporle and Satherley told RNZ in a joint statement about their polling research.
Polls aim to capture public opinion in a snapshot of time. Photo: Supplied / Wellington City Council
Who makes up the people who are polled?
Online panels are used for conducting all types of research, including market research.
"In our case we use our database for political polling, and other online surveys which can be on any topic," Reid said. Panels are also used for recruitment for group or one-on-one discussions as well.
People can approach Reid Research to be put on a research panel or they may be sought out.
Some marketing panels can include financial incentives to participate, but "those participating in political polls do not get any financial incentive," Reid said.
Polling such as the RNZ-Reid poll also uses other research organisations to supply respondents, but they must follow polling guidelines.
"Most companies which currently conduct polling and use a third party to supplement their own database of members, will always mention they use a third party, which adheres to the ESOMAR (The World Research Organization) guidelines and also the Research Association of New Zealand's guidelines," Reid said.
To ensure a diverse result, people polled in one RNZ-Reid poll will not be included in the next polling round.
"There is a stand-down period of six months usually," Reid said.
Why does how questions are worded matter so much?
Depending on how they're worded, poll questions can skew results.
For instance, a 2013 Gallup survey in America on physician-assisted suicide found very different results depending on if it was asked if people supported the idea to "end the patient's life by some painless means," or "assist the patient to commit suicide."
"If questions are poorly worded or ambiguous then results produced from that question become hard or impossible to interpret," Sporle and Satherley said.
"Another point is that consistent question wording makes comparison between surveys possible - whether it be comparing the same survey over time or between different surveys."
"It's not only the questions that are asked but how the answers are interpreted," Campbell said. "I think it's vital that RNZ or TVNZ or whoever do their very best to report the results as they stand and not sensationalise them."
Polling samples are carefully orchestrated to hit all demographics in a population. Photo: RNZ / Marika Khabazi
How does relatively small sampling get an accurate result?
Pundits often ask how a poll of 1000 people can accurately represent a country of 5 million people, but it all boils down to the well-established statistical theory behind random sampling.
"A thousand is not a small sample," Campbell said. "That sized sample is often used in political polls across many counties, with far larger voting populations."
The sampling methods used when conducting polls are critical to ensure the sample, and therefore results, reflect the underlying population of interest.
"A well-designed, smaller sample poll will always outperform a larger sample, poorly designed poll," Sporle and Satherley wrote in their paper.
"If everyone in the population has a known equal probability of being sampled, then even with small sample sizes we can achieve a reasonable estimation of what the population as a whole think."
A minimum of 500-1000 people is typically recommended for nationwide polls in New Zealand. The Research Association New Zealand requires its members to have a sample size of at least 500 for nationwide political polls.
Polls will also disclose the possible margin of error they have.
"Pollsters are neither hedging their bets or being slippery when they refer to the margin of error of a result from a statistical sample," Campbell said.
"The margin of error is a mathematical expression of the difference between a sample's opinion on an issue and the population's opinion from which the sample is drawn."
Smaller samples tend to have a larger margin of error, but the decrease in the margin of error "really slows down past sample sizes of around 1000," he said.
In the Reid-RNZ poll, a disclaimer in the June poll noted that there was a "maximum margin of error of +/- 3.1 percent at a 95 percent confidence level."
Campbell said that this means that, "If 50 percent of a representative sample of 1000 NZers of voting age said they would vote Labour if an election were held today, we can be statistically certain 95 percent of the time that between 53.1 percent and 46.9 percent of the population of NZ voters would have voted for Labour."
What do things like quota sampling and weighting mean?
Quota sampling is different from random sampling. With quota sampling, researchers are trying to create a sample to mimic the characteristics of a population - for instance, having the number of Māori in a sample reflect the number in the population.
Reid said that it is important polling panels represent the population based on official statistics.
"So when we set up, we know that roughly 48 percent of the population is males and 52 percent females. So we have to get 48 percent males, 52 percent females.
"We know that X percent of the population lives in Northland, Y percent of the population lives in in Auckland, et cetera. And we know the breakdown of ages, and get all that from the census data."
Weighting of a poll is a technique used to compensate for under or over-representation in a sample to achieve representative data. The weighting is achieved by making responses from under-represented respondents count more towards the results of the total poll.
"This means each response is given a higher or lower weighting based on whether the demographic characteristics of that respondent are under or over-represented in the survey, relative to the general population," Sporle and Satherley said.
"So, a single response may count as more or less than one response when the results when weighting is used to determine the survey results. This is common practice in political polls, but it should be reported alongside the survey results."
"You've got to be really careful with weighting," Reid said. "Any weighting that we do is very minor."
"We must get even the most difficult people to participate - the young people, particularly young males."
Sporle and Satherley noted that when it comes to polling, there are major age and socio-economic differences in access to and use of online resources.
"Another major challenge is created by New Zealand's rapidly changing demographics, especially with the youthful Māori and Pacific populations as well as large numbers of incoming migrants and outgoing migrants, with many of the latter still being able to vote."
The government's recent decision to abandon the regular five-year Census may affect polling in the future.
"Stats NZ will need to produce this Census data through other means now, although exactly how they go about that, and how it will compare to the original Census data as we knew it, remains to be seen," Sporle and Satherley said.
"This shouldn't influence survey sampling so much as things like the Electoral Roll and online panels are often used as sampling frames."
President Donald Trump's first victory in 2016 surprised most US pollsters. Photo: AFP
How accurate are polls, really?
Polling has come under fire in other countries, particularly the United States, where Donald Trump has consistently outperformed polls in the 2016, 2020 and 2024 presidential race. His victory in 2016 over Democrat Hillary Clinton was widely seen as a surprise and led to a sweeping post-mortem for pollsters.
One of the big findings in the aftermath was that polls were not properly weighted to account for people's education, as in every election Trump has won a majority of non-college educated voters.
But in New Zealand, polls have a fairly good track record predicting general election results in recent years.
Data by the Research Association found that the average percentage point difference of polling the week before an election tended to be off by only 1 point over the last several elections, except for 2020 when Labour's landslide result exceeded expectations.
In 2023, polling by Newshub / Reid Research and 1 News / Verian underestimated on combined average National by about 2.3 points and overestimated the Greens by about 2.8 points compared to the final result.
But consistently, going back to the 2002 general election, the final polls generally lined up to close to what the election results were.
"You have to trawl right back to 1993 or thereabouts to find that one of the polling companies published a result that was distinctly out of step of what happened on the night," Campbell said.
People in the polling profession do tend to make it clear that their work is a snapshot in time.
"I would never ever say, this is actually what the result's going to be," Reid said, noting that people can change their minds any time before an election right up until they tick the box.
Polling also has to overcome a general decline in trust many of the public have developed in recent years.
"If people lack trust in institutions and don't feel listened to or that their views are respected, they will be less likely to respond to a media request to be part of a survey," Sporle and Satherley said.
"If this happens we're less likely to get accurate data about people, so maintaining high levels of trust and cohesion is really important to avoid that disconnect."
And lastly, what about those self-selecting online polls you see on many news websites asking about the issues of the day? They're nothing like the more methodical polls that aim to survey a balanced population.
They're "a fun fact or a conversation starter poll," Campbell said. "They're more a reflection of that particular website's readership."
Political parties often see leadership changes due to poor polling, as seen with Jacinda Ardern and Andrew Little in 2017. Photo: Twitter / Michael Wood
Do polls make a difference or influence events?
Polls are not elections, but they have and will continue to have impacts on politics in New Zealand.
Most political parties also conduct their own private polling of how they're faring, which can be a gauge for where their support is. A few terrible rounds of polling can lead to major results.
Aotearoa political history is riddled with party leaders who have stepped aside partly as a result of a bad poll, such as Labour leader Andrew Little stepping aside for his deputy Jacinda Ardern only weeks before the 2017 election, or the National Party's Simon Bridges being ousted by Todd Muller in 2020.
"I think a good poll should challenge," Campbell said.
"Some people argue that polls should only reflect opinion and not drive it but in fact they can do both."
Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.
