frogblog

Before the election, frogblog ran a Latest Polling Averages graphic. This page explains how it was calculated.

This Latest Polling Averages table is intended as a snapshot of public opinion, aimed at pulling together all the public polls into a single number for each party. It is in no sense intended to be my prediction of how the election will ultimately play out. It’s a picture of how opinion is right now, not a prophecy of how I expect it to be on election day.

How do you get the party vote percentage?
There are six publicly-available polls: those published by Fairfax, the National Business Review, TVNZ, TV3, the Sunday Star-Times and the New Zealand Herald. The party vote percentage in the graphic is an average (of sorts) of the latest poll released by each organisation. So, when a poll is released, its figures replace, in my calculations, the figures of the previous poll released by the same media organisation.

The party vote percentage is not a simple average of these six polls, however. I weight the polls in two ways.

First, I give more recent polls more weighting than older polls. The weighting is as follows:

• A weighting of 1.0 for polls that are less than a week old.
• A weighting of 0.5 for polls that are more than a week but less than two weeks old.
• Polls that are more than two weeks old are excluded from our analysis.

I take the age of a poll to start from the date it is published. So, a poll published on August 1 will move from a 1.0 weighting to a 0.5 weighting on August 8, and then taken completely our of the analysis on August 15.

Second, I weight polls according to sample size. A poll with a sample size of 1000 has a 1.0 weighting, a poll with a 750 sample has a 0.75 weighting, and so on.

So, my party vote percentage is our graphic is an average of the six polls mentioned above, after they have been weighted for age and sample size. The only exception to this rule if for the Progressives, which I explain below.

How do you get your Parliamentary seat allocation?
I simply plug the party vote percentage, to two decimal places, into the Electoral Commission’s seat allocation calculator. In doing so, I make the following assumptions:

• Jim Anderton will win Wigram. While the Progressives are barely rating in most opinion polls, I will set their party vote in my calculations as 0.7% - or, just enough party votes to prevent his seat becoming an overhang seat. This second assumption is based on the fact that, while the Progressives failed to register in most opinion polls before the last election, Jim Anderton’s personal success translated into a party vote success that national opinion polls didn’t pick up. While I don’t expect the Progressives to get as many party votes as last time (1.7%), I do expect them to get more than the 0% that most opinion polls have them at.
• Peter Dunne will win Ohariu-Belmont. This assumption is based on the fact that Mr Dunne won a safe majority of over 12,500 votes in this electorate in 2002, and there is no polling evidence to suggest he won’t repeat this victory in 2005.
• The Maori Party will win four Maori electorates (Te Tai Tokerau, Tai Hanauru, Tamaki Makaurau and Waiariki). This is the most difficult assumption to make, and is based on the very limited polling data that is available for the Maori electorates so far. I may change this assumption as new polling emerges.
• Rodney Hide will not win Epsom, and no other Act candidate will win an electorate. This assumption is based on a recent Herald on Sunday poll which had Mr Hide in a distant third in this electorate.
• The Greens will not win an electorate seat. This assumption is based on the fact that the Greens are campaigning primarily for the party vote in 2005.
• Winston Peters will not win Tauranga.

So, using these assumptions, and the party vote figures I have calculated, the Electoral Commission’s virtual election calculator throws out a make-up of Parliament. I reserve the right to alter any of these assumptions if pollling evidence suggests I ought to.

How do you choose what is the most likely government?
I will use the following assumptions, based on what the parties have said publicly and what I have discerned of their political strategies post-election:

• Act will not support a government led by Labour.
• The Greens and the Maori Party will not support a government led by National.
• Jim Anderton will be, de facto, a Labour MP, so his seat will automatically be included with Labour’s when I am naming the most likely government and the number of seats it has.
• Labour and National will both prefer the “cleanestâ€? governing arrangement possible - that is, involving the fewest support parties possible.
• Labour will prefer to deal with the Greens and/or United than the Maori Party.
• National will try to deal with Act and/or United.
• NZ First will allow the largest party in the next Parliament form the next government by pledging to abstain on confidence and supply motions. However, NZ First will not offer such a pledge to a government that has the Greens or Act as part of a formal coalition.
• National and Labour will not enter into a governing relationship with one another.
• United will support a government including the Greens if that is the only way that a government can be formed in the next Parliament - as Peter Dunne attempts to position himself as the guarantor of “political stability”.

I reserve the right to alter these assumptions, if parties’ public statements suggest I ought to. If these assumptions, together, don’t offer definitive guidance on the “most likely government�, then I will revert to my own instincts.

When will the data be updated?
Every time a poll is released and every time an old poll moves from one weighting period into another. If you’re interested in the raw data, it is here.

What do the arrows mean?    
The arrows indicate simply whether, in the latest update of the party vote percentage, a party’s score has gone up or down.