Poll Watch: Reid Research Poll 2018-1-31

Matthew Whitehead
6 years ago
Categories: greens, labour, national, nz first, polls
Tags: coalitions, MMP, newshub, poll watch, polling, pollwatch, reid research, threshold, trends

Newshub recently released the latest poll they commissioned from Reid Research, covering the period between Jan 18^th and 28^th. (I tend to refer to them by the day they’re released, however) Let’s dive into some analysis of what this means.

If you missed the actual poll findings, here is Newshub’s official tweet:

As some of you have known, I usually do some analysis of what these poll results mean using some custom spreadsheets I prepared during the election campaign, and have been posting them to twitter since before the election, but I thought I’d start sharing them on The Standard, too. The usual disclaimers about polls apply. Firstly, the most naïve way to read a poll is simply to look at what sort of parliament it would deliver if it were bang on. This does tell us the most likely result of a poll, but it’s not really super helpful in determining just how close some of these things are, but some people want that at-a-glance look, and being able to see whose lines cross the halfway point tells you who’s likely to be government in a very visual way.

You can see here that a simple reading of the Reid research poll tells us a few things: National is still the largest Party, but boy is the margin between them and Labour tight now, and with a result like this, they’d look even more foolish talking about a moral right to govern. New Zealand First is likely under threshold, and won’t be back without an electorate seat, which we can’t yet assume they’ll get given their spotty past record with electorate runs, especially as there’s no available polling on electorates this far out, and we don’t know which ones they’d contest.

Besides, it’s early days, this is just an indication of their support more than anything else.

Secondly, the more useful thing I do is plug these things into a giant, slow spreadsheet that randomly distorts the party vote within the margin of error hundreds of times, with an overall distribution that follows a normal curve to reflect the fact that more extreme deviations are rarer. I then classify each of these results as to which type of government they give us, and repeat them 2,000 times to get a good average before making a graph. That’s the one floating to the right. (I did manage to get a few outcomes pop out here where the two blocs each had 60 seats because ACT had won a list seat. A hung parliament with no NZF or Māori party to resolve the winner is a pretty terrible outcome, so let’s all breathe a sigh of relief it doesn’t seem likely at the moment)

This gives us a good idea of how likely each outcome is. As a general rule, you should start worrying if anything you don’t like gets over 20%, so a 4.9% chance of the government changing at this stage is not a big deal.

This stage is also where I calculate the chance that either of our medium-size parties will go under the threshold, by doing another 2,000 simulations specifically looking at that. At 6% and 3.8% respectively, both are within the range where that probability is relevant.

NZ First is actually polling just below the level where being over the threshold is still within the margin of error for them- this means the only way they’ve actually got a support of 5% or more is if this poll is a rogue, and that’s only a one in twenty chance given they have an even 1,000 respondents. Every single election simulation of these 2,000 had them popping up as under threshold.

The Greens have a minimal chance of dipping under threshold, in 2.2% of their extra 2,000 simulations they dipped under, and both the Greens and NZF being under threshold is the most likely way for the Nats to snatch victory from the jaws of defeat.

The final graph I do for people is a spread of the margin of error, showing the biggest swing leftward and biggest swing rightward possible from the poll results, with all the most optimistic or pessimistic assumptions for each. I’ve fallen out of love with this particular graph a little because we tend to naturally just look at the ideal scenario for our team and go “wow that’s good!” rather than using it to evaluate the spread between the two margins of error and remembering that reality is somewhere in-between. Remember, the further a result is from the first graph, the less likely it is. These two scenarios are almost vanishingly unlikely.

All of this analysis also is a static analysis of one poll. As I’ve said in the disclaimers, you should go back and compare this poll to previous polls to get the really important information, which in summary is this: The Greens are roughly where they were at election time. Labour’s support is up. National’s and ACT’s are down. New Zealand First’s is disastrously down, and if they don’t get some good publicity before the next election, they could be out of Parliament again.

We should also view polls in light of recent events- it’s entirely possible Labour has simply been buoyed by positive sentiment to Jacinda’s news that she’s expecting a baby, and that some of this support is soft. It’s also possible that National’s support is equally soft and based on a strong, united front they’ve been showing so far, when actually the caucus is pretty divided on a lot of the What Happens Next questions, such as who gets to lead when Bill’s done, a question which will probably be relevant if they can’t pull a near-majority of the party vote by the next election. The Greens staying still is probably not good news for them, but Reid Research has actually been one of the least friendly polls to them before the last election, with even Colmar Brunton being kinder, so it’s possible they have actually gained a little support.

But the big loser of this poll is New Zealand First- Reid Research was the most friendly pollster to them before the election, even though Peters was constantly rubbishing their coverage. If they’re polling at 3.8% with Reid Research, it’s likely they’ll poll even lower when the next Roy Morgan (who are suspiciously overdue on their December and January poll results, which is a huge shame because they were the only company that polled regularly) or Colmar Brunton poll is released, and then that begins to become something of a self-fulfilling prophecy if they don’t pull up.

I suppose there’s also some bad news in there for ACT, the Māori Party, and whatever’s left of TOP, but quite frankly, unless the former loses Epsom, or the latter two (re)gain an electorate at the next election, they’re completely irrelevant.