The Greens and Labour have been doing increasingly well in the special votes, and National increasingly poorly.
I did a projection, based on the 2011 results, that the Greens probably wouldn’t pick up the extra seat they often do from special votes. But I didn’t call it a prediction, partly because I didn’t intend it to be, but also because the numbers had the Greens only narrowly missing out on that 14th seat. So if the Greens had pushed just a tad further and made it over the line, I wouldn’t have been surprised. I said it was going to be close.
As it happened, it wasn’t close. The Greens had nearly 5,000 votes to spare.
|Party||Preliminary||Projected final||Project vs prelim||Actual final||Actual vs project||Actual vs prelim|
To be more specific: The Greens saw their 14th seat shoot not just past National’s 61st, but also past its 60th, to become the 119th quotient. This meant the Greens had room to spare: 4,873 votes less (with no redistribution) and they would’ve stayed on 13, or National would have needed 21,836 more to get to 61. (This is more than the ~18,800 “votes per seat” because National also has the 120th quotient.) For comparison, my projection had the Greens needing 1,139 more than projected to pick up a 14th seat.
Changes in context
What happened? I suggested there were reasons why the specials might swing further left in my projection post:
- The number of overseas voters doubled this election compared to 2011.
- Some universities had on-campus polling booths, where a lot of students would have cast advance special votes.
Further on the second point, as Jonathan Marshall suggested on Twitter, unlike 2011’s, this election was during the university semester so more students would have been outside their electorates in the first place.
But I certainly didn’t expect any combination of these effects to add to 5,000 more votes for the Greens alone, so I suspect there’s something else. Also, those reasons are just speculation. For all we know, it could even be related to migration patterns (including domestic migration), though I don’t know that those have actually been changing. The full statistics that the Electoral Commission will release in about two weeks might offer a bit more insight, but it’s hard to discern.
A growing swing since 2008
Beneath this swing, there might be an underlying trend. The Greens have traditionally done well on special votes, of course. But they’re doing increasingly better on specials than they used to, at least since 2008. (I couldn’t easily find preliminary results for 2005.)
The first set of columns is the special vote percentage divided by the preliminary result (so < 1 means a drop), and the second set is the difference between the specials (not total) and preliminary result in percentage points.
Now, take the sparklines with a grain of salt. The scales aren’t consistent between cells, and they each only have three data points anyway.* Also, the two sets of columns are just different measures for the same phenomenon, so the fact that it happens twice adds no weight. (I did both while trying to find a more consistent predictor of special vote impact, and couldn’t.)
But they do broadly indicate an improving trend for Labour and the Greens, and the opposite for National. And the trend isn’t small. In 2008, National lost 0.52 points from specials; this year, they lost 1.02 points. The Greens have gone from gaining 0.29 points to 0.67 points. One seat’s “worth of votes” is roughly 0.83 per cent, a bit less if you allow for wasted votes.
If I were in the National Party, I’d be concerned about this trend. As it stands, National is generally expected to lose a seat to the Greens on specials. But the Greens are getting increasingly good at this, and are picking up the seat even when the numbers suggest the increase won’t be enough.
For people who like numbers, here’s the spreadsheet I used for these calculations: Special vote impact projection 2014. It’s an extended version of the projection spreadsheet I posted earlier.
*Although there are only three data points, they are reliable since they are true by definition, not a sample. (Turnout is interesting for other reasons, but it’s the final result that matters here.) I’d be more worried about short-term circumstances giving the illusion of a long-term trend. The value of adding anything before 2005 might be questionable though, since the situation a decade ago might lose relevance to today.