Luke Malpass
In defence of the polls
As Labour basks in the glow of its win and tries to find a way to gracefully let the Greens know they won’t play a meaningful role in Government, the threeyearly raking over the entrails has begun.
There has been an emerging line of thought in the past few days since the election: that the polls got it wrong.
The public polls by election day had narrowed, putting Labour at around 46 per cent, National over 30 per cent, the Greens and ACT both on about 8.
In the event, Labour hit 49.1 per cent and National’s vote dwindled to 26.8 per cent. ACT got 8 per cent and the Greens 7.6 per cent. At this point these results should be treated with some caution. There are more than 400,000 special votes to come in and the dial could move around a bit – although based on recent elections, in the Labour and the Greens favour.
So, the public TV polls, which capture a snapshot in time of the views of about 1000 people or so, had National up a bit and Labour down a bit on what ended up happening.
Another difference was around the size of the wasted vote. The Newshub-Reid Research Poll had it that Labour would be able to govern on its own, while the Colmar Brunton Poll suggested the Greens would just be needed. The percentage of the vote going to minor parties who won’t get into Parliament doesn’t need to shift much to change this.
The last UMR corporate poll of the pre-election period was understood to have been more accurate still: Labour at 50 per cent, National at 29 per cent, ACT at 7 per cent, the Greens at 6 per cent. NZ First at 2.7 per cent. UMR also does Labour’s political polls.
Perhaps some of this confusion is around what a poll actually is. Polls do not have any sort of predictive power. They are only a snapshot of how
people say they would have voted if the election had been held during the period in which they were polled.
To put it simply, on the night a poll comes out on the news, it shows what people thought a few days ago. Over time, that builds up a trend, which is useful. The polls also pick up big movements.
There are a couple of other interesting factors in this election. The first is whether advance voters were less inclined to tell pollsters how they voted. Given that almost two million people voted prior to the day, there may have been a factor of people not wanting to reveal how they actually voted, as opposed to saying how they would have voted.
The other thing is that campaigns are actually designed to change voters’ minds.
Labour’s campaign, especially in the final week where it deliberately went after National voters who voted for John Key and Bill English, was designed to do just that.
Because New Zealand’s system is directly proportional, rather than having preference flows, there is always a risk that if a major party is clearly not going to be in government that some of its voters go elsewhere at the last minute, or don’t vote.
Polls can’t be expected to pick that up.
In the end, the ‘‘the polls got it wrong’’ discussion appears to have been shanghaied in to New Zealand from overseas with little regard to the overall facts: they all suggested the end result, or very close to it.
They all suggested that NZ First would be knocked out of Parliament, that ACT would do well and that the Greens would stay in. And they either suggested Labour would get a parliamentary majority or just short.
That isn’t too bad.
Polls do not have any sort of predictive power. They are only a snapshot of how people say they would have voted if the election had been held during the period in which they were polled.