Heather Roy

2 August 2020

VoterTorque Podcast 6 is about Political Polls. We don’t talk about any particular poll and the numbers, but my TorquePoint business partner Simon Ewing-Jarvie and I cover Polling 101, the basics.


We cover the basics so listeners are better informed about how to read the polls. For example, one poll is merely a snapshot in time, but trends start to emerge when you have three data points (i.e. 3 separate polls) or more. In fact the best information is gleaned from a ‘Poll of Polls’ where all recent polls are plotted and results averaged out. Closer to the election we can expect to see these from Radio New Zealand and Curia.

It’s important to understand how to interpret the polls because they do have an influence on how people vote.

In New Zealand there are three big polling companies and many smaller ones. Those most commonly known are:

  • TV1 Colmar Brunton poll which surveys 1000 people (by landline and cellphone)
  • Newshub Reid Research poll which surveys 1000 people (by a mix of landline, cellphone and internet panel)
  • Roy Morgan which surveys around 900 people (by landline and cellphone)

The two most recent polls have thrown up interesting results. The Newshub Reid poll had Labour at 60% and National at 25% and it was widely regarded as an outlier or ‘rogue’ poll. Ever poll has a margin of error and the polls with over 800 participants usually have 95% confidence that the actual result will be within 3 percentage points of the poll result. This means 5 percent of the time the result won’t be within this tight range – that’s one in 20 polls that on average will have outlier results.  The more recent TV1 Colmar Brunton poll had the two big parties closer (a 21% gap) and felt much more realistic. A discussion on this poll can be found here.

Polling really is a science and for those with an interest there is good information to be found on NZ website Research Association and FiveThirtyEight – a US site which focuses on American politics but also talks about polling generally.