In 2014 I ran a projection of what might happen after counting special votes, based on what happened in 2011. I did it again in 2017, then kinda skipped 2020. But this year it seems interesting again, as the country holds its breath to find out whether National will need NZ First to form a governing coalition.

(I’m a bit slow off the mark here, for the same reason that I skipped 2020, basically that my priorities are nowadays elsewhere. But there’s still a week to go before official results are published, so I haven’t missed the boat completely 🙃)

Special votes aren’t counted on election night. The preliminary count just counts ordinary votes to give a rough idea. But special votes are of course included in the official count, after they’re all checked for validity and shipped to the right places.

We can’t really predict what will happen with special votes, and I’m not trying to. The following projection assumes that special votes will fall “similarly” (I’ll explain soon) to the 2020 election. Special votes often bias differently to ordinary votes; they tend to favour the left. In 2020, National ended up with two fewer seats in the final count than on election night, and Labour and the Māori Party picked up one each.

This is why Christopher Luxon is behaving as if he’ll probably need Winston Peters. A National–Act coalition has no room to spare in the preliminary count. If National loses a seat—and they normally do—then they’ll need NZ First to form a majority in Parliament.

**What does “similarly” mean?**

The key assumption I make is that the *percentage* of special votes* received by a party will go up/down by the same *factor* when compared to the preliminary count.

**Technical note: When I say “special votes”, I actually mean “difference between official and preliminary counts”, which can also arise from votes disallowed, corrections and the like.*

For example: In 2020, the Green Party got 7.57% in the preliminary count and 9.23% of special votes. This is a ratio of 9.23%/7.57% ≈ 1.2204. In 2023, the Greens got 10.77% in the preliminary count. So we project 10.77% × 1.2204 = “13.15%”, except that when we do that for every party it won’t add up to 100% (since every party has a different multiplier). The parties actually add to 96.16%, so we renormalise everyone to 100% and the Greens land on 13.67% of special votes. Adding these votes to the preliminary count (using the number of special votes estimated by the Electoral Commission), we project 11.38% (a bit higher than 10.77%).

It’s worth saying—there are a ton of different ways to do this. Instead of ratios, we could adjust using the difference (*i.e.* add 9.23% − 7.57% = +1.66 points). My choice was at least somewhat arbitrary. Graeme Edgeler did a projection last week and got different result from me, presumably due to some subtle difference in methodology.

(Though one big difference was actually because of a mistake I made, and I picked it up and corrected it by noticing the difference with Graeme’s! In 2020, the special votes saw the Greens overtake Act, so the party order wasn’t consistent between the preliminary and official counts, and I had attributed those vote counts to the wrong parties.)

### Projection

Okay, here’s what I got.

On this projection, National would lose three seats, one to each of Labour, the Greens and Te Pāti Māori, whose overhang would become just a normal seat. National and Act would have 58 seats between them, and would need NZ First to make up a coalition.

This would see Melissa Lee, Gerry Brownlee and Andrew Bayly all leaving Parliament (as the last three to be in on National’s list), to be replaced by Shanan Halbert (Labour), Kahurangi Carter (Greens) and the elimination of Te Pāti Māori’s overhang. But also remember that Port Waikato has a by-election, which is likely to see Mr Bayly returned as an extra MP, and Parliament back up to 121 MPs, for the remainder this term.

### Reasons to think this projection might be wrong

It’s worth interpreting these with caution.

The number of special votes is on the rise. They won’t know for sure until after they count, but the Electoral Commission estimates that 20% of votes were special votes, up from 17% in 2020. This could indicate a change in voting patterns among special votes, too.

The election flipped! Three years ago was a disaster for National; this year was a disaster for Labour. If ordinary voting patterns swung like that, maybe special voting patterns also swung in a way that you can’t just model as a multiplicative factor on ordinary votes.

Some electorates were close and for those involving Te Pāti Māori, this might affect the number of seats won (or the amount of overhang). Graeme Edgeler had a crack at calling these. I’m not going to try.