Cautious optimism

Posted: December 21, 2016 in democracy, elections, New Zealand
Tags: , ,

So, the first post-resignation Roy Morgan poll came out today. (this is of course the only regular poll in New Zealand at the moment, and it’s not even run by a New Zealand firm…)

Their analysis is somewhat on-point, although I’d be very cautious about attributing the good performance from Labour to the by-election or the resignation, rather those seem reasonable explanations for National’s slump. Even though this is a dramatic change from the last poll that would usually make it suspect of being a rogue1, it follows dramatic political events and moves in the expected direction, so I believe Roy Morgan deserve the benefit of the doubt on this one that their poll is probably on-trend.

Roy Morgan doesn’t show the results in terms of seats in Parliament, you may be interested as to what the poll indicates if it’s bang on. (NZF holding the balance of power, with National and Act having a couple extra MPs compared to Labour and the Greens) But that’s a little superficial, as polls don’t really work that way.

What a poll does is give you, 19 of every 20 tries, a range within which we expect each party to perform. As such, I need to introduce you to a rather odd type of graph to show what the RM is really telling us. This graph is a “doughnut hole,” or essentially a way to compare two pie graphs that show different scenarios or results from different time periods, but with the same categories.

On the inside, we have the leftmost Parliament we can expect. On the outside, we have the rightmost Parliament2 that this Roy Morgan poll indicates we could have got if an early election were called.

rm-poll-2016-12-11-margin-of-error

If a general election were held back on the 11th, Parliament should lie somewhere between the results shown in those two rings. That is, we could possibly get either a Labour-Green coalition that can flex between the independent Māori Parties and NZ first for support, or a National-Act coalition that can flex between the Māori Party and NZ First for support, or we could end up with a result where Winston is kingmaker if the result is somewhere between those two scenarios. As the first two scenarios are literally the most extreme ones, we should expect that the likeliest thing this poll indicates is that the blocs are pretty evenly balanced at the moment, (even though Labour and the Greens are polling lower, they may be in a position to have more friends on their side, too) but that New Zealand First will decide who governs unless New Zealanders tip in one direction or another.

If this RM is a valid starting point for the new trend with Bill English as Prime Minister, then the race has essentially started in a dead heat. It’s possible that future polling will show a difference, (I’m waiting for a poll other than Colmar Brunton and Roy Morgan to show up to give us some better data)

1 As I mention later, polls usually sample enough people that 19 of every 20 should have every result within its margin of error. That remaining poll in every twenty is called a “rogue,” and it’s a term to be reasonably careful of. Generally you can only tell rogue polls in hindsight, and they need to be off by a significant amount from the trend line (ie. so much that the margins of error for the poll and the trend line don’t overlap if you draw them on a line graph) and without a reasonable explanation for the variance.

I fully expect that people who know just enough about polling, statistics, and maths on the right-wing of New Zealand politics to be dangerous will becalling this result a rogue way too early. Right now every indication is that this is a result that will be on-trend in the future, say in March or so, when we’ve had a few more polls and can check it against a trend line.

2 My assumptions in calculating these two scenarios are as follows: Start with a set of party vote results, either from a poll or an aggregation. Sort the parties into three blocs, left, undeclared, and right. Calculate expected margins of error for each party, assuming a polled sample of 1000 voters. Add this margin to the party’s result in the scenario that favours that bloc, subtract the margin in the scenario that doesn’t. As there is only one undeclared party, it gets the remainder of the party vote. (So NZ First is actually a little inaccurate in the right-leaning scenario: in reality their share of the vote should lie between 5.87%-9.13%, and their 14 seat result came from a 11.58% share of the party vote. I’ve got an idea about what to do to fix that in the future, but it’s a fair amount of work to crunch all these results at the moment as I haven’t automated the maths for figuring out seats from party vote results, so I have to plug them into the Electoral Commission’s website each time, so I didn’t really want to start over to fix one party being a bit high in one of the scenarios)

I also assumed that for electorates in which a former MP is running, that they can unseat the incumbent if it’s favourable in the scenario I’m running, hence why Mana shows up. Normally I simply assume that last election’s electorate winners will win again, but that seemed too simplistic for this analysis.

There are overhangs in both scenarios- 1 seat for the Māori Party in the right-leaning Parliament, and 1 seat each for Act and UF in the left-leaning Parliament.

Advertisements

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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