This time, with statistics! (Those make everything better.) All raw data come from the Election Results site, but the calculations thereon are mine, as are any transcription errors. I will also note up-front that there are a lot of factors in here, and in particular the electorate-level data do not necessarily correspond to individual voters, even collectively – the voting populations may be significantly different. They’re just interesting to look at.
(Edited to put the charts below the fold)
First up, because I was curious about the turnout: the 2008 turnout was 79.46%, while the referendum turnout was 54.04%. The electorate-by-electorate correlation is +0.806, for an r² of 0.650. That’s a middlingly strong correlation for data of this sort. As expected though there’s also a fairly strong negative correlation between turnout and voting yes: r = -0.480. Electorates with high No vote tend to have higher turnout as well.
Now for party vote correlations: as you’d expect, a high Green vote is strongly correlated with voting yes, with r = +0.788. National has the strongest negative correlation of the represented parties, at -0.280, followed by Act on -0.104. The Māori party has a negligible negative correlation as well, at -0.074 (skewed by their vote pattern), but most curious to me was United Future: a small, but noticeable positive correlation of +0.120. I don’t quite know what to make of that, but it wasn’t in line with my expectations at all. It’s not a terribly strong one, and the chart (see below) shows it may be more of an anomaly than anything else.
The combined Labour-Green-Progressive vote correlates at +0.533, while National-Act is at -0.266. Throwing United Future or the Māori Party to either side doesn’t make much difference.
Some charts showing the more striking results here:
Green Party vote vs Yes vote
I also have charts for the other parties and some others.
Finally, one non-referendum chart that I found intriguing: Green Party vote against United Future vote.
They have a positive correlation of +0.253, which isn’t strong but not at all negligible.
If anybody is interested in doing their own calculations, the SQLite database I have been using to store this is also available. The schema should be straightforward. I’ve been plotting the charts with GNUPlot, and a couple of trivial Python scripts to transform it appropriately for that.
Tags: charts, politics, referendum, statistics