State of play

Posted: September 11, 2017 in elections, New Zealand
Tags: , , ,

A little bit of technical catch-up.

I have updated my metapolling data for 2017 to include the election season thus far. The results are as follows:

metapoll119National: 41.5% ±3.05%
Labour: 39.6% ±3.03%
New Zealand First: 8.8% ±1.75%
Greens: 6% ±1.47%
TOP: 1.6% ±0.77%
Māori Party: 1.5% ±0.75%
(Assume 2 electorates won)
ACT Party: 0.3% ±0.15%
(Assume 1 electorate won)
United Future: 0%
(Assume 0 electorates won)

Projected seats: 121
Majority requires 61 votes.

As this average stands, New Zealand First does most likely look to narrowly hold the balance of power if nothing changes, however that’s not entirely the reality of the situation, as I’ve stated before that polls are like taking a series of still pictures and using them to describe the motion of a dancer or a race. We have to consider possibilities like whether we could be viewing from a bad angle that makes the wrong racer appear to be in the lead, that perhaps one racer is behind but is rapidly gaining ground, or of course, to go to our other analogy, that the motion is so fast that we’re missing parts of the dance between polls. This is why looking at both averages and individual polls is important. That last one we won’t be so concerned about at the moment, given that the only large event that’s happened recently is Joyce’s now thoroughly-debunked claim of a gaping fiscal hole in Labour’s budget. Instead the hole was under his feet this entire time, and no other significant political events appear to have abbutted this one so it’s unlikely we need to worry about conflating two different causes for changes in polling. We’re not in a situation like where Metiria Turei and Andrew Little both resigned so close to each other it and it was difficult to tease apart the effects of one event from another.

metapoll119moeThe first thing we need to look at is the margin of error.

To the left is a nested donut chart, my traditional form to show MoE visually. The inner ring, as labelled, represents a best-case for potential left-wing coalitions assuming every party is out by the maximum MoE on their individual votes, and also the worst-case for parties that might consider a right-wing coalition. (As the balance has shifted to Labour in polling and NZF is considered undeclared, I have moved them to National’s bloc for the purposes of determining their size) The outer ring is the same scenario for the parties needed for a right-wing coalition. As you can see, we’re perched between on one side, a disaster scenario where the Greens are out of Parliament, NZ First are the kingmakers, and the left needs them and the Māori Party to work together to form a government, and on the other side, a dream team with an outright Labour-Greens majority and a friendly Māori Party potentially inside the coalition if Labour wants to secure extra votes in case it needs to hold by-elections this term, or ready to step in for help on conscience votes where NZF and National might be aligned and Labour might suffer defectors.

The other way to look at margins of error is to simulate elections using numbers generated within random ranges that reflect the margin of error. This approach is even better than simply looking at the two most extreme scenarios, as you can simply evaluate the number of seats to see which blocs can form governments in each simulation, rinse and repeat several times, and calculate probabilities based on the frequency of each outcome from there. I’ve actually started on this as well, (there is a LOT of copy-and-paste tedium to build a spreadsheet to do this. My current simulation runs 10 full MMP elections based on manually inputted electorate win probabilities (we have at least one electorate poll for all the critical races other than TTT and Nelson, even if none of those polls seem to be hugely reliable, but it’s better than assuming who will win) and party vote polling shares, with threshold cutoffs included) and under this metric we tend to have about 9:1 odds that New Zealand First is the kingmaker, using the results from my metapoll average. (the minority report being a Māori Party decision between Labour+Greens and National+ACT+NZF. I’ll get back to you more on this with more exact probabilities on social media once I’ve got my model built to simulate 100 or 1,000 elections at once, but for now I’m just running ten simulations ten times and doing the arithmetic mentally to get an impression of the odds) It’s main limitation is that it currently can’t handle small parties winning multiple electorates.


The other thing to consider is the direction of travel. As you can see to the right, the latest released poll by Colmar Brunton actually shows that the trend is towards the Māori Party, and not New Zealand First, being the potential Queenmaker. (I gendered the word this way because they have been clear that their members are indicating a preference for coalition with Labour, although it will require a formal post-election consultation with members as per their rules, so they are still open to working with National if they are not the deciding vote, and Labour’s preference isn’t completely in the bag)

I expect the average will soon start showing this as the overall expected outcome too, as I run a harsher formula to cut off input from old polls during election season, where I square the overall result of my weighting factors instead of just multiplying them all. (It’s still counting a couple of old polls back in June for a fraction of a percent, so it’s not unreasonably strict. I do this because there are far fewer polls run outside of election season) This is in addition to one weighting factor being towards polls closer to election day.

In terms of the modelling approach, going off CB’s numbers (which is roughly the direction where I expect the trend to lead us) we seem to have something like a 45/40/15 split in likely governments. In order, those numbers represent: NZF kingmaker, Māori Party queenmaker, outright Labour-Green government.

This is good news, and reason to feel optimistic. We just need a little more hard campaigning to push those Green, Māori, and maybe Labour Party numbers up a bit more in order to reduce the probability that Labour actually needs Winston, and we’ll have a strong progressive government. Ideally, we want that NZF kingmaker chance well below 20%.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s