- Date published:
4:00 pm, June 10th, 2019 - 27 comments
Categories: act, greens, labour, national, nz first, polls - Tags: hung parliament, labour government, labour-green government, labour-green-NZF government, poll watch, pollwatch
We had both major polls out yesterday, as noted in previous stories, and boy were the results a doozy! In addition to twice my usual talk about models, we’re going to get into the assumption some people are no doubt already making- that one of the polls is “wrong,” or more technically, rogue.
For those who want the raw results, I won’t summarize them in detail but will rather link them from here, except to say that Reid Research (Newshub) shows basically what we’ve come to expect recently: National performing poorly, Labour ascendant, the Greens improved slightly from last election, and NZF below threshold. Colmar Brunton (TVNZ) shows New Zealand First over threshold and just able to choose who would govern, but Labour on the verge of being able to choose between them and the Greens, so a slightly stronger result for the Government than the last election, but not as strong as recent polling has indicated, and fundamentally the same result if it were mirrored in 2020.
Also worth noting in leadership polling: Ardern continues to rate well in Reid Research’s net approval rating, (last image at this link, as usual the news continues to highlight PPM) at 56.9% higher approval than disapproval, while Bridges continues to be a drag on his party, at -38.6% net approval. (For context, Trump’s overall approval numbers are about twice as high as Bridges’) If we assume everyone who approves of Bridges is a National voter, this means that at least 20% of the country would vote for National but don’t approve of Bridges’ performance. Given that 7-8% of voters prefer Collins as Prime Minister in the media’s favourite but actually very-unuseful poll question, Preferred Prime Minister, we can conclude that there’s still at least 12% of National Voters that would like Anyone Else But Collins, and with Simon’s 17.4% approval rating, that 20% existing is pretty disastrous. He doesn’t command the majority support of voters in his own party.
On to the important part: The party vote results, and what they show in terms of probability. Colmar Brunton’s poll is interesting here, although the most likely result is New Zealand First being required for a Labour Government, it’s not the only result by far. My model has likely had a streak of luck hurting NZF in the above graph, as they were below threshold 51.3% of the time, (the Greens were below 2% of the time in these simulations) so likely there’s a 90-95% chance that if NZF were above threshold as this poll suggests, they would be necessary to govern, however when they’re below, it’s reasonable to conclude we’d get a Labour-Green government, as the Hung Parliament and National Government probabilities are so low.
Reid Research’s graph is so boring as to be barely worth talking about: We get a landslide Labour government based on those results, every single time, as the wasted vote is so high that the probability of Labour dipping under that 50% mark due to the margin of error is irrelevant. It is quite frankly too strong a result for labour, and we should hope that we don’t see many polls like this near the 2020 election, as I don’t really want to see what Labour’s like with that level of impunity- having a coalition partner be necessary is good for the health of our democracy. It is however worth noting that National also broke 50% a few times in mid-term polling, so we shouldn’t be entirely alarmed, as hitting it now and then is probably just normal noise from the margin of error, and of course, even ascendant parties tend to lose some support near an election, too, as kiwis aren’t too fond of majority government anymore. In this set of simulations, NZF were always below threshold, and the Greens were below 0.9% of the time.
As to whether one poll or the other is right: Most likely, the truth lies between these two. Colmar Brunton’s (TVNZ) poll gives an unlikely surge to National and New Zealand First, and Reid Research’s (Newshub) poll gives an unlikely surge to Labour. Here’s what my weighted average (mostly from these two polls and the one beforehand, but also including a tiny fraction of the three polls before that) looks like: A Labour-Green government with NZF out of the picture, and the opposition having 54 seats, assuming National doesn’t decide to run ACT out of Epsom. (I put the odds of Seymour losing his electorate at a nominal 5% in my model for exactly this reason, and there has been talk of doing this, however I hear the current policy is to continue their implicit support of ACT)
The trendline looks really similar to before: as usual, spikes of NZF going over and under threshold, spikes of the Nats or a hung parliament being non-zero chances, and spikes of Labour not needing the Greens, but the overall trend looks like the public continue to support results that lead to a Labour-Green government without New Zealand First.
If I absolutely had to pick one poll over the other, I would note that the Reid Research poll is actually closer to the trend, and that last election, it and Horizon were the only ones that really rated NZF correctly when they had what we can now tell was their late surge. I’m tentatively willing to credit that this might be due to the superiority of online panels supplementing landline polls, as the Colmar Brunton poll only calls mobiles and landlines to get its data, which might lead to it overestimating people who are more likely to answer phone calls. (who would likely tend to be older, richer, and more conservative voters, regardless of whether they answer landlines or not. Younger voters are less inclined to answer a call unless they know where it’s coming from) However, I think what’s likely happened is that Colmar Brunton has had a rogue poll, and Reid Research has had one that overstates Labour’s support a little bit, but is probably still within the margin of error- we’ll see how things look at the next poll, but this isn’t some Great Big Crisis or a matter of alternative reality, and both polls are useful data points we shouldn’t want to go away.