On Tuesday, I published estimates of the Senate vote at the 2016 federal election for the Nick Xenophon Team, broken down by South Australian state electorates.
While writing that article, I became curious about whether the similar dataset (built by Alex Jago) for One Nation in Queensland had done a good job of predicting the One Nation vote at the 2017 state election – we’d used the data regularly before the election, but after 6pm on election night we threw it out.
Well here is a chart comparing the 2016 Senate vote for One Nation (after distribution of minor preferences) to the party’s primary vote in the recent state election, in the 61 seats where One Nation stood candidates.
The trend is pretty strong – while the vote at the state election was slightly higher, it did not vary tremendously from the trend.
The correlation between these two datasets is 0.796 – in other words, the two datasets correlate by about 80%. It’s not a perfect correlation, but a very strong trend.
The biggest exception was in party leader Steve Dickson’s seat of Buderim. One Nation polled 12.7% at the 2016 election, but managed 28.85% at the recent state election. This makes sense – the party had an incumbent MP and put a lot more effort into that seat. Apart from that seat, the ratio was not more than 2:1 in any other seat.
Hey Ben
Just wondered if you removed Buderim (and one or two others which look like outliers), how much does the correlation go up by?
Thanks
Craig
Craig, if you remove Buderim the correlation increases from 0.796 to 0.832.
Comments are closed.