Turning worms?

Postscript (21st September): Anything could have happened, but what did happen was that the polls were not too far off the final mark, with no over-estimation (and in fact, for many an under-estimation) of National's election night tally. The result was that the same coalition was returned to government. Future posts will look at which poll got closest; in the meantime though, expect a return to some thinking about evaluation.

 

A fast update ahead of tomorrow’s election. In short, anything really could happen on Saturday 20th September. Make sure you have your say and vote.

In the peaks and troughs data-viz below, we can see what has indeed been reflected across all the polls. National have fallen well off their July peak to sit at 46% in the poll of polls. Bearing in mind the polls in 2011 over-estimated National by around 4% on average, tomorrow could be very interesting indeed. Labour however haven’t been the beneficiary of the decline; they’ve remained stuck around the 25% mark. The Greens have remained solid and aren’t far off their peak of the last three years.

Among the smaller parties, New Zealand First and the Conservatives seem to have taken much of National’s declining share, but the Conservatives are still shy 1% of the crucial 5% threshold. The Maori, ACT and Internet-Mana and United Future parties will all be crossing their fingers that they hold an electorate seat. Internet-Mana seemed to have momentum back in August, but are now well down from their peak.

And the shifting party fortunes have made the coalition options that much more interesting. National’s current coalition arrangements have slipped below 50%, and will in likelihood need to look to either the Conservatives (if they can cross the 5% threshold) or New Zealand First. The Labour, Greens and NZ First option, at 46%, is still shy of hitting a majority, but is a lot closer than at any time since May.

One note on the data. Up until now, I’ve been using a rolling average of the past four political polls (drawn from Roy Morgan, One News Colmar Brunton, 3 News Reid Research, Herald Digipoll and Fairfax-Ipsos), with a weighting towards the most recent polls. With the last five polls all coming at essentially the same time, the final poll of polls are a simple average of all five polls.

On election night I’m at a two-house party. One house is the party party, and its neighbour is the data geek party. No prizes for guessing which one I will be at.