SAYit Blog
How many seats?

I thought I should start off by saying that the media commentary around polls is showing signs of improvement.  A while ago I criticised TV3's political reporters for over-egging the results of their poll (When does 4.9% become 5%), but last night's reporting was noticeably more cautious.  Whereas back then they rather breathlessly reported that 'NZ First were on 4.9% and therefore out of parliament', last night they reported that the Conservatives were 'close to the threshold' on 4.6%.  That's entirely reasonable reporting in my book - the actual 95% confidence interval is between 3.3% and 5.9%, and sample theory indicates that the bulk of likely results will be closer to 4.6% than either of those two extremes.  Personally I think other polls suggest that this might be a bit high for the Conservatives, but it's reasonable reporting given the information they have.

They're still a bit guilty, however, of making out that their poll is the only poll.  We certainly need other polls before we can properly judge what's happening, and the more we focus on individual polls the more confused people are likely to get.

More irritating though was the reporting of Maori TV's Te Tai Hauauru poll (by the same polling company as the TV3 poll, Reid Research).  While Maori TV's online article is headlined 'Polls show a close battle', the Dominion Post this morning made out that the Maori Party candidate was definitely in front.  A few things with that:

  • Maori TV's online article does not include the sample size, the margin of error or the dates of the poll, all of which are vital information.
  • The video posted online is in Maori so I don't know if they specified the dates or the sample size verbally, but it does at least include a graphic of the margin of error (+/- 4.38%). That tells me that the sample size was n=500.
  • The Maori Party's candidate Chris McKenzie leads Labour's Adrian Rurawhe by 32% to 29%.  The margin of error for McKenzie's vote is +/-4.1% (as I've mentioned before, the margin of error shrinks as you move away from 50% towards either 0% or 100%).
  • Maori seats are notoriously hard to poll in, because turnout is much lower than in the general seats and because the population is younger, poorer and therefore more mobile.

I don't think we can say with any certainty that the Maori Party are ahead in Te Tai Hauauru - just that it's a tight head-to-head battle between the Maori Party and Labour candidates.

That leads on to the topic I wanted to talk about - the impact of electorates on the overall results.  Last night's TV3 poll report predicted a 61-61 vote split between National & its current allies (ACT / UF / Maori) on one side, and Labour / Greens / NZF / Internet Mana on the other.  Their written report showed, however, that that 61 seats assumed that the Maori Party would win two seats, and I couldn't understand why.  They currently have three seats in parliament, two of which are represented by MPs who are retiring at this election.  To my mind when you're making these models, in the absence of other evidence, you should either assume that the Maori Party will retain all three seats they currently hold, or assume that they'll lose the two with retiring MPs but retain the one with the MP who's staying on (Waiariki).  So the assumption should be either three seats or one seat, but not two (unless, like the Dominion Post, they're taking the Te Tai Hauauru poll as absolute gospel).

  • If the Maori Party retained all three seats they currently had, then the TV3 poll had 62 seats for National + ACT + UF + Maori versus 61 seats for Labour + Greens + NZF + Internet Mana
  • If the Maori Party retained just one seat, then the poll had just 60 seats for National + ACT + UF + Maori, with the others still having 61,

This is really important.  If the Maori Party retained all the seats they have at the moment, then the current government could continue.  If they only won a single seat, then NZF definitely would hold the balance of power.

That applies elsewhere too.  On that poll:

  • If National gifted the Conservatives an electorate, then they'd have a majority without NZF but it'd involve four parties (Nat+ACT+UF+Con)
  • If National directed its Epsom voters to vote for Christine Rankin (Conservatives) rather than David Seymour (ACT), then Nat + UF + Cons is enough for a majority.  The equivalent would also be true in Ohariu.
  • If major parties won both Epsom and Ohariu, then Nat + NZF is an easy majority (65 seats).
  • If major parties won all seats currently held by minor parties, then National would have 59 seats in a 120 seat parliament.

My near namesake Jamie Whyte has pointed out that ACT's win in Epsom kept National in government in 2011, which is true to the extent that it meant that National + ACT + UF was a majority.  On the other hand, had they lost Epsom, then National (59 seats) would have been able to get a majority with either the Maori Party (3 seats) or NZ First (8 seats), so they would still have been in a pretty good position.  It's also reasonable to assume that ACT's party vote would have been lower if they hadn't been seen as likely to win Epsom, and if just one in 10 ACT voters had switched to National and ACT lost Epsom, National would have been entitled to 60 seats in parliament and formed a majority with one vote from Peter Dunne.  You can try it out yourself here - all I did was put in the 2011 election results and take 0.1% off ACT and put it on to National.  0.1% was 2257 votes in 2011, by the way.

The point I'm trying to make is that, thanks to the electorate vote exemption, results in the electorates can have massive effects on overall outcome of the election.  What's more, those effects are hard to predict and deals don't always work in the favour of the parties doing them.  There are, as I think I've demonstrated, logical situations where the major parties would be better off if they didn't do the deal.