If you're someone who doubts the influence of political opinion polling, you would do well I suggest to have a chat with former Labour leader Andrew Little.
It was a shock 1 NEWS Colmar Brunton Poll back in July showing Labour on just 24 per cent that prompted Mr Little to make an on-camera admission to me that he had considered resigning.
The 1 NEWS Colmar Brunton poll has built up a strong reputation over 22 years.
Source: 1 NEWS
That poll result closely mirrored Labour’s own private polling and a subsequent Reid Research poll.
Mr Little could see the public was telling him something through the polls and within a couple of days he stepped down and handed over to Jacinda Ardern.
It was the single defining moment in 2017’s general election campaign.
For me personally as Political Editor, the Andrew Little poll story is a very important reminder of the responsibility the media has, along with our polling companies, in presenting accurate polls and ensuring the methods we use are as good as they can possibly be.
As Andrew Little well knows, polls really matter.
So it’s with that sense of responsibility - as well as a look to the future - that 1 NEWS and Colmar Brunton have now decided it is time to change our polling methodology.
In future we will no longer just poll telephone landlines. It will be a 50/50 split of mobiles and landlines.
This is a significant step for us and one taken after thorough consideration.
1 NEWS and Colmar Brunton have been polling 1000 eligible New Zealand voters – from randomly chosen telephone landlines - for the last 22 years!
In that time the poll has built up a proud and very strong reputation for its reliability and credibility.
However, during the course of the past year we at 1 NEWS, along with Colmar Brunton, felt it was right to start exploring whether adding mobile phones was prudent, given the rapid changes we are seeing in communication habits.
The fact is, landlines are no longer used by as many people. The best information we have on this is Census data from 2013 which confirms only 86 percent of households had a landline compared to 92 percent in 2006.
Colmar Brunton believes that trend is only likely to have accelerated in recent years.
But Colmar Brunton’s main concern with the sole reliance on landlines is around the ability to reach young people who are far more likely to only communicate by mobile.
To compensate for this Colmar Brunton have weighted different age groups to ensure they are representative of the New Zealand population.
This means ‘weighting up’ its sample of younger New Zealanders, and ‘weighting down’ its sample of older New Zealanders.
However, getting the right representation of younger people aged 18 to 34 has been getting harder and Colmar Brunton says that since 2016 it’s been forced to repeatedly increase the weighting for that age group to ensure a proper representation.
So over the course of this year we've conducted two pilot polls using both randomly chosen landlines and mobile phones.
The first pilot was small and only used 150 mobile numbers, and the headline impact was also small.
But it provided enough evidence to suggest that if using both landline and mobile numbers, the sample could be made more representative. Especially among the young but also among some ethnic groups like the Pacific Island community.
So in September we conducted a more significant pilot to run alongside our last full poll before the election.
This pilot saw Colmar Brunton, in addition to its landline interviews, conduct 323 interviews using randomly generated mobile phone numbers.
Colmar Brunton says there is still some ‘weighting’ required; for example the mobile sample slightly under-represented some of the less populous regions such as Tasman, Marlborough and the West Coast and it needs its poll to be regionally representative.
What we got from the split poll were different results, but NOT a dramatic change from the landline only poll that we published.
In the landline only poll for example, National polled 46.5 percent versus 45.2 in the dual poll. Labour polled 37.5 percent in the landline only poll versus 38.3 percent in split poll.
The Greens were 7.7 percent in the landline only poll versus 6.6 percent in the dual poll. While New Zealand First was 4.9 percent in the landline, versus 5.2 percent in the dual poll.
However when compared to the final election result, the dual poll was more accurate. For example, the sum of the percentage point difference between the landline poll and the final result was 8.1 percent. For the dual poll the difference was just 5.6 percent.
On that basis we decided it was time to make the switch.
In Colmar Brunton’s view what the dual poll will do is ensure a more representative sample of New Zealanders.
That means there is then a reduced need for corrective weighting of the data for younger people and ethnic minorities. This allows for more confidence in the stability of the results.
So where to from here? This Sunday we will publish a new poll on Q+A of a 1000 eligible voters with a 50/50 split of landlines versus mobile phones.
The margin of error of plus or minus 3.1 percent will not change.
And no doubt it will be another fascinating result – and boy we’ve had some of those this year - given the wild ride politics has been in 2017.
While we won’t be able to compare Sunday’s poll to previous 1 NEWS Colmar Brunton polls as we typically do, we will be able to compare it with the final election result.
With mobiles included this really is a new poll and a new era for us and our polling partners.
That said, I have no doubt that like its predecessor the new 1 NEWS Colmar Brunton poll will quickly establish itself as the gold standard for political polling in this country.
* Poll results on Q+A on TVNZ1 from 9am on Sunday, and on 1 NEWS Now.