Online panel polls
A couple of media outlets have reported this poll from Horizonpoll:
National has an 8% lead over Labour in the latest HorizonPoll voting intentions survey covering 3,673 New Zealanders.
National has 30% of decided voters (3% down on its 2008 share of all registered voters).
Labour has 22% (3% down on its 2008 share of all registered voters).
Even before one looks at what type of poll this is, warning bells should emerge over one claiming National has only 30% of “decided” voters when every other poll has them around 50%.
Also when they compare the percentage of “decided” voters to the percentage of “registered” voters. This would only be useful if there was 100% turnout.
But the big reason why the media should not report this as a poll, is it is not a phone poll, but an Internet panel survey.
At some stage I will do a fuller post on the pros and cons of Internet panels. In some circumstances they can be very useful – especially when surveying the opinions of a discrete group.
But generally they are unreliable when it comes to being a fair sample of all New Zealanders. Because they are only representing those NZers who have joined that online panel. And even with weighting, this does not mean it is representative. A weighted sample can still be unrepresentative.
Over time they will become more useful, as panels become larger. But in my opinion you need really really large panels. The only political poll I currently trust which is done through online panels is You Gov in the UK. They have a fine record of accuracy. Their large panels, means that every poll they do is not the same 3,000 or so people responding to it.
As I said, online panels do have some valid uses. And it is inevitable they will become more dominant over time, as the challenges for phone polling increase.
But my view is that at this stage in NZ, they are not a reliable indicator of New Zealand public opinion. I have seen many online panel polls like the above, which produce results massively different from those produced by phone polls. Normally I ignore them, but as some media reported this one, I thought it is worth making the point.
It would be useful if media did not report online panel polls just as “polls”, but always make it clear it is a “Internet panel survey”.