This is from 1News last night (I didn't watch 1News, obvs) but this is from their website – both National and Labour have slid in the latest 1News-Verian poll, while New Zealand First have moved to their strongest position in eight years. If an election were to be held today, the right bloc of National, ACT, and New Zealand First would have 63 seats —enough to form a coalition— while the left bloc of Labour, the Greens, and Te Pati Māori would have 58 seats. So that's from 1News and their Verian poll.
This is from Radio New Zealand – after the budget and pay equity changes, the left bloc would have the support to turf the coalition out of power, the latest RNZ-Reid Research poll shows. The preferred Prime Minister and leadership ratings are also bad news for the government, with the exception of Winston Peters, who's seen his highest results since 2017. The ratings of the government's general performance have also continued to slide, with Labour, the Greens and Te Pati Māori all gaining compared to the previous poll, they would have a majority with 63 seats between them, compared to the coalitions 57. A direct opposite of what 1News-Verian said.
How can this be? And it's always headline news. You've got 1News talking about the right bloc being able to hold on to power, but only just, and look out. You've got RNZ crowing about the fact that the coalition government would be turfed out of power with the left gaining hold. And both lead with it, and it leaves me scratching my head and doubting both of them.
How do you imagine the pollsters collect their data? Random phone calls of 1000 people? No, no, no. It's far more tricky than that, and they put it in every story. It must be an obligation on the part of the media company to say how the data was collected. From TVNZ: Between May 24 and May 28, 1002 eligible voters were polled by mobile phone (500) and online, using online panels (502). What are online panels? Are they things you sign up to yourself? Who knows? The maximum sampling area is approximately plus 3.1%. Party support percentages have been rounded up or down to whole numbers. The data has been weighted to align with Stats NZ population counts for age, gender, region, ethnic identification and education level.
So what does that mean? If I'm a numpty, am I worth 2 points as opposed to somebody who leaves school worth NCEA and that's worth one? What does that mean? If I'm 18 and I respond, does that mean because there are fewer 18 year olds who respond, does it mean that my reckon is worth double that of somebody who's 50+. How can you weight the information? And not all 18 year olds think the same way. If you're looking at ethnic identification, not all Māori, not all Pakeha, not all Pasifika, not all Chinese people, think the same way. The sample for mobile phones is selected by random dialling using probability sampling. Online sample is collected using an online panel. So that's from 1News.
This is from RNZ: This poll of 1008 people was conducted by Reid Research using quota sampling and weighting to ensure representative cross section by age, gender and geography. The poll was conducted through online interviews between the 23rd and 30th of May 2025, has a maximum margin of error of +/- 3.1%.
I'm of a mind to never discuss the polls again. The statisticians and the research pollsters and the companies all say, oh, no, no, no, it's terribly scientific. Is it really? When you've got two polls conducted over the same time, presumably using the same scientific methods, coming up with two completely different results. If the polls were scientific, surely you'd see a consensus of opinion. You wouldn't go sniffing like a truffle hunter looking for respondents that agree with your particular version of the way things should be. It's like you're researching into an echo chamber. It's not worth the time and the money.
If this is what RNZ is spending their money on, given that they are funded by the taxpayer, I'd rather they spend it on training up young reporters or allowing a veteran reporter to spend some time doing some investigative journalism, rather than coming up with a poll that supports their worldview. And which is in direct contrast to the other taxpayer funded organisation, which is kind of paying its way at the moment, which is 1News. What is the point? How on earth can we take them seriously when they come up with completely different results and when all the data is weighed, quotas are taken, samplings adjusted. It's an absolute crock. The emperor is stark naked and shouldn't be taken seriously at all.
Kerre McIvor, is a journalist, radio presenter, author and columnist. Currently hosts the Kerre Woodham mornings show on Newstalk ZB - where this article was sourced.
This is from Radio New Zealand – after the budget and pay equity changes, the left bloc would have the support to turf the coalition out of power, the latest RNZ-Reid Research poll shows. The preferred Prime Minister and leadership ratings are also bad news for the government, with the exception of Winston Peters, who's seen his highest results since 2017. The ratings of the government's general performance have also continued to slide, with Labour, the Greens and Te Pati Māori all gaining compared to the previous poll, they would have a majority with 63 seats between them, compared to the coalitions 57. A direct opposite of what 1News-Verian said.
How can this be? And it's always headline news. You've got 1News talking about the right bloc being able to hold on to power, but only just, and look out. You've got RNZ crowing about the fact that the coalition government would be turfed out of power with the left gaining hold. And both lead with it, and it leaves me scratching my head and doubting both of them.
How do you imagine the pollsters collect their data? Random phone calls of 1000 people? No, no, no. It's far more tricky than that, and they put it in every story. It must be an obligation on the part of the media company to say how the data was collected. From TVNZ: Between May 24 and May 28, 1002 eligible voters were polled by mobile phone (500) and online, using online panels (502). What are online panels? Are they things you sign up to yourself? Who knows? The maximum sampling area is approximately plus 3.1%. Party support percentages have been rounded up or down to whole numbers. The data has been weighted to align with Stats NZ population counts for age, gender, region, ethnic identification and education level.
So what does that mean? If I'm a numpty, am I worth 2 points as opposed to somebody who leaves school worth NCEA and that's worth one? What does that mean? If I'm 18 and I respond, does that mean because there are fewer 18 year olds who respond, does it mean that my reckon is worth double that of somebody who's 50+. How can you weight the information? And not all 18 year olds think the same way. If you're looking at ethnic identification, not all Māori, not all Pakeha, not all Pasifika, not all Chinese people, think the same way. The sample for mobile phones is selected by random dialling using probability sampling. Online sample is collected using an online panel. So that's from 1News.
This is from RNZ: This poll of 1008 people was conducted by Reid Research using quota sampling and weighting to ensure representative cross section by age, gender and geography. The poll was conducted through online interviews between the 23rd and 30th of May 2025, has a maximum margin of error of +/- 3.1%.
I'm of a mind to never discuss the polls again. The statisticians and the research pollsters and the companies all say, oh, no, no, no, it's terribly scientific. Is it really? When you've got two polls conducted over the same time, presumably using the same scientific methods, coming up with two completely different results. If the polls were scientific, surely you'd see a consensus of opinion. You wouldn't go sniffing like a truffle hunter looking for respondents that agree with your particular version of the way things should be. It's like you're researching into an echo chamber. It's not worth the time and the money.
If this is what RNZ is spending their money on, given that they are funded by the taxpayer, I'd rather they spend it on training up young reporters or allowing a veteran reporter to spend some time doing some investigative journalism, rather than coming up with a poll that supports their worldview. And which is in direct contrast to the other taxpayer funded organisation, which is kind of paying its way at the moment, which is 1News. What is the point? How on earth can we take them seriously when they come up with completely different results and when all the data is weighed, quotas are taken, samplings adjusted. It's an absolute crock. The emperor is stark naked and shouldn't be taken seriously at all.
Kerre McIvor, is a journalist, radio presenter, author and columnist. Currently hosts the Kerre Woodham mornings show on Newstalk ZB - where this article was sourced.
4 comments:
At the risk of repeating myself ad nauseum a major lesson from Covid was that Pollsters and Modellers are just guns for hire and their outputs are best treated with derision and contempt.
Kerre , In the 2024 US election the public polls were going everywhere , however the betting market ( where people put their money where their mouth was and placed a bet on who would win ) were very close to the final result. Not 3.1% either way which is really a 6.2% discrepancy.
Polls are 'scientific' in the way they process sample data. Not that that would mean necessarily getting a consensus between polls, as a poll is an exercise in measuring population attributes based on a sample. As an analogy, you can apply 100% 'scientific' sampling processes - i.e. wholly random sampling, perhaps with some sample stratification if you are dealing with, say, a huge number - to a barrel of several thousand marbles of different colours and up with a slightly different estimate each time.
But the vagaries of probability aside, polls are not in the same league as marbles because people can refuse to participate or lie when answering. And before you even have a sample there are biasing factors at work such as the number of occupants per building (associated with social class, therefore with political polling) and mobile phone ownership (not universal).
People, unlike marbles, use polls to send messages to the govt - "ho on like this and I'll vote for the other lot next time" - even if they have no real intention of doing so.
Polling and sampling from a barrel full of marbles have a lot in common, principally that both are exercises in taking samples for the purpose of inferring the attributes of the population from which they are drawn. But they also differ significantly, principally in relation to the impossibility of ensuring sample representativeness in the case of warm bodies with wills.
Basil. Unfortunately, NZ Polls are manipulated for political purposes.
Which explains why the most accurate 'NZ politics' polling company is Australian.
Post a Comment