‘Poll of polls’ of little use now

RNZ have updated their ‘poll of polls’:  Can Ardern lift Labour’s polling?

Three polls in late July dragged Labour down to a 24.0 percent average in RNZ’s Poll of Polls. This was from 26.5 percent in June and 29.4 percent in May.

At 24 percent just two MPs would come in off the list if electorate MPs all held their seats.

In March, after Ardern was elected deputy, Labour’s average lifted from 27.8 percent in four polls taken in February to 30.6 percent.

If she can double that lift as leader, Labour and the Greens might be back in the game.

There’s no way of knowing what effect Ardern will have on then polls. All we can do is guess.

It is interesting to see past trends and movements, but that is all polls and ‘poll of polls’ can tell us. They are good through the middle of a term, but they become useless at the business end of the term, during an election campaign.

Single polls, or clusters of polls as we have just had, can be more informative than a poll average over a longer timeframe, which is what we get from ‘poll of polls’.

As recent elections in the US and UK have shown, and as poll results here in the last week have shown, there can be significant shifts in measured voter preferences during an election campaign.

It isn’t new. Every election in New Zealand this century has featured late swings. Remember the surge of United Future in 2002 after Peter Dunne turned the worm? NZ First typically attracts late support, and Greens tend to do worse than polls suggest.

Even single polls can only tell us what some people thought a week or two ago. Opinion can react to polls, so as soon as they are published they can be out of date.

‘Poll of polls’ are more out of date because they average back over a longer period. And now that New Zealand has few polls there isn’t much for them to work with.

So making predictions based on ‘poll of polls’ is not a good idea, as much as poll aggregators and media like to try to do.

They tell us what has happened in public opinion up until recently, approximately.

They can’t tell us what people will think tomorrow, let alone on September 23.

They are interesting, but are hopeless at predicting the future.

Leave a comment


  1. Also interesting are Peter Green’s forecasts: http://ellisp.github.io/elections/elections.html

    But statistics can’t predict the future in a rapidly changing election campaign.

    • Alan Wilkinson

       /  3rd August 2017

      Yes and no. Statistics can be used to understand the factors that drive politics and that understanding can be used to predict the future. Trump’s win was predicted that way. Of course predictions always have uncertainty due to the unknown unknowns.


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 )

Google photo

You are commenting using your Google 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 )

Connecting to %s