As I write, it remains unclear whether independent Andrew Wilkie will win a seat in Denison at the expense of the Liberals, but there is no doubt he has done much better than most observers, including me, expected. The 8.4% he has polled is just above the threshold I suggested he needed to have any chance of election in the cutup, and he has been assisted by a softer than expected Green vote and by a fairly helpful distribution of the votes between the major parties.
In a comment I failed to notice and respond to earlier, James Boyce (Comment 9, Denison could be closer than it looks, HERE)
questions my previous “suggestion that Andrew Wilkie’s candidature has made it less likely that a second Green candidate will be elected in Denison”. Boyce argues that “Kevin does not suggest that there has been a collapse or splitting of the Green vote which has been solid but faced the predictable obstacle that, however popular the individual candidates, roughly three quarters of Denison voters will just not vote Green.” He also says: “One of the most important blocs of Wilkie votes probably came from former Labor voters especially in the northern suburbs. There are many people who were disillusioned with the Labor Government but for entrenched cultural reasons will never vote Green or Liberal.”
Now that the booth-by-booth provisional figures are available, it is possible to see exactly where the Wilkie votes indeed came from. It turns out that the highest Wilkie votes were not where all the evil pokies are at all. Indeed, in virtually every Hobart City booth except Lenah Valley, Wilkie broke 10% and in virtually every other booth he didn’t. His highest votes were at Fern Tree (15.6), Albuera (14.7), Taroona (14.7), Sandy Bay (13.9), Dynnyrne (13.7), Hobart (13.2), Sandy Bay Beach (13.1). His lowest votes were all at northern suburbs booths, eg Roseneath (2.3), Abbotsfield and Elwick (3.3), Austins Ferry and Cosgrove (3.7), Claremont and Goodwood (3.9), Windermere (4), Chigwell (4.1).
I did indeed argue that the Green vote in Denison would not break high twenties to low thirties and that those polls showing it in high thirties to forties were overcooking it, however I never argued that it had maxed out at around one in four, and I was quite surprised that it did not even significantly increase at all. Competition from Wilkie was a definite part of this. Looking at the swing to the Greens booth-by-booth as a linear regression of the Wilkie vote there is a strongly statistically significant relationship:
Swing to Greens = 4.23 (+/- .92) – (.46 (+/-.09) * Wilkie vote.)
Correlation doesn’t equal causation, and it’s possible that without Wilkie those people living in areas where Wilkie happened to do well would have all switched from Green to Labor or Liberal had Wilkie not run. But given that the Greens achieved hefty swings in every other electorate, nobody’s going to believe it. Most likely the Greens would have gained about 4 points of swing in Denison had Wilkie not run, and close to half his voters would have otherwise voted Green. (At the low end of the 95% confidence level the values are 2.4 points of swing without Wilkie, and .26 lost for every point gained by Wilkie.)
Indeed there were many northern suburbs booths where the Green vote went up substantially despite swings to the Liberals and the presence of Wilkie. The big swings against the Greens all came in booths where they normally break 30 – they were down in nearly all of these (up in Cascades only).
What’s more, in those booths where the Green vote is normally less than 20, the Wilkie vote either had no impact on the Greens vote or else moved with it. But in booths where the Green vote is normally more than 20, there was a very sharp correlation between the Wilkie vote and the fortunes of the Greens, to the point that the strength or otherwise of the Wilkie vote may have been the main driver of swings to or away from the Greens in such booths.
I think the claim that “roughly three quarters of Denison voters will just not vote Green” is therefore incorrect; without Wilkie’s candidacy, the Green vote in Denison would have been about 29%, just taking into account those Green voters who switched to Wilkie. Whether there would have been others still who declared a pox on both Greens and Wilkie following the falling-out between them is unknowable, though I think that that number would not have been very large.
Of course, to get an idea of what the election might have looked like without Wilkie, it is necessary to also examine how his vote interacts with swings to and from the other parties.
The Labor Party suffered a swing against it in every booth in Denison. The smallest swings were at Swan (4.5), Cosgrove (4.7), Elizabeth St (6.7), Sandy Bay (7.2), Goodwood (7.9), Tolosa (8.1), Dynnyrne (8.6), Sandy Bay Beach (8.7), Glenorchy and Battery Point (8.8). The biggest thumpings were at Hobart Central (17.6), St Peters (15.3), Fern Tree (14.4), Rosetta (13.8), Moonah (13.7), Lutana (13.4), Austins Ferry (13.2), Waimea, Mt Nelson and New Town (12.9). Thus the swing against Labor in Denison appears to have had a complex and almost demographically random nature. Furthermore, the relation of the booth-by-booth swing against Labor to the Wilkie vote is not statistically significant, and if it exists at all is probably minor (perhaps around a tenth of a point increase in the Wilkie vote for every point lost by Labor).
The Liberal Party gained swings in almost every booth in Denison but lucked out at Sandy Bay Beach (-4.3), Creek Road (-3.1), Taroona (-1.5), Sandy Bay (-0.7), Mt Stuart (-0.5), Cascades (-0.1) and didn’t get much at Lower Sandy Bay (0.2), Hobart South (0.8), Swan (1.1), Newdegate (1.2), Cross Street (1.6), Dynnyrne (1.7) or Waimea (1.9). Southern suburbs where the Wilkie vote was through the roof are prominent in this list. Their best swing booths were Austins Ferry (9.4), Collinsvale (8.6), Roseneath (8.3), Hobart Central (8), Lutana (7.5), Fern Tree (7.3), Elwick and Moonah (7.1) – quite a lot of northern suburbs in this lot. There turns out to be a very strong inverse correlation between the Liberal swing and the Wilkie vote:
Swing to Libs = 7.3 (+/-1) – (.42 (+/-.1) * Wilkie vote)
Considering this result and the Greens result, it appears that Wilkie primarily harvested votes that would otherwise have gone to the Greens or Liberals, knocking around four points off what would otherwise have been the swing to the Greens, and around three and a half points off the swing to the Liberals. But the lack of significant correlation between the Wilkie vote and the swing against Labor suggests that Wilkie was probably not taking any great number of votes off Labor himself – perhaps a point or so, if that. As further evidence for this, there are extremely strong direct correlations between the Wilkie vote and both the Green and Liberal votes from the 2006 election, and a strong inverse correlation between the Wilkie vote and the 2006 Labor vote. Labor did suffer big swings in some of its traditional high-voting areas but these went mostly to the Liberals rather than Wilkie. He was most effective in areas where the Labor vote was already low. Look at the Sandy Bay booths for instance: in these booths the Wilkie vote massively exceeds the net swing against Labor. It also exceeds or at least matches it in a number of western and northern Hobart booths.
It seems that Wilkie was taking votes mainly off the Greens and the Liberals, especially in areas where they were already strong before this election, and most likely largely from people who had already been voting for them. This conclusion backs the candidate’s own reported view that the votes he was taking were coming from Liberal as well as Green voters, which also has implications for his preference distribution in the (unlikely) case he is excluded before Helen Burnet. It also has implications for the struggling future of fourth-party politics in this state, because it makes the point I have been making for years: as oppositions to Labor, both the Greens and Liberals have extreme elements in their support base that can make centrist voters (especially me!) decidedly nervous about voting even for very good candidates from such parties. Furthermore, the right kind of campaign can appeal both to those who consider the Greens the least of the three main evils, and those who hold such a view of the Liberals. The interesting thing is that Wilkie has knocked such a dent in the Liberal vote despite running a campaign that I for one thought was off target as it seemed too left-skewed (and too intellectualised and policy-driven, with some seemingly rather arbitrary and in cases easily questionable policies at that) to really much harm either major party. There is a message for the state Liberals in this about what happens when they fail to adequately reassure the small-l Greg Barns types who are progressive on social and international issues and possibly green-tinged, but still distrustful of Labor and Green economics. They may be only a couple of percent, but the Liberals cannot take this vote for granted. In terms of the pokies dimension, it appears that the typical Wilkie voter moved by that issue was not a northern suburbanite trying to save their own community from the dreaded one-armed bandits, but rather a southern-suburbs middle-to-high-income earner voting for the bogans and battlers to be rescued from themselves (perhaps with a semi-conservative view to thus reducing the crime rate, perhaps not). In the end, it appears that what we have here is not a new Tasmanian electoral phenomenon, but an old one re-emerged: the Wilkie vote is an Australian Democrat vote (pale Greens and left Liberals) on steroids, perhaps because it isn’t hampered by the structural flaws, internal bureaucracy and historical baggage that caused that party’s effective demise.
This leaves the central question of whether the Greens would have won two seats without the Wilkie factor. And while I don’t agree with James Boyce’s assessment that the Green vote in Denison is capped at around 25%, he is right about the key point: without Wilkie, Denison probably really would have been a solid 2-2-1. In my initial impressions (written without the benefit of the booth figures) I overestimated how much of the Wilkie vote had come from the Greens compared to the Liberals. Remove Wilkie from the mix and most likely the Greens would have had about 29% (1.74 quotas), but the Liberals would have had almost exactly 2 quotas and Labor the rest, and the Greens’ extra three-quarters of a quota would have done them no good at all since there would have been no preferences to make use of it with. As it turns out, given the actual size of the swing against Labor and the nature of the Wilkie vote, Wilkie’s candidacy has in fact most threatened the Liberals, and the Greens very probably would not have won anyway.
How much it has threatened them we will really find out next week when we see whether the Liberals can yet hold him off for the final seat or not.
PS re Lyons: Given the Australian Story “toxic trees” items, the booth of St Helens was one of particular interest to me, especially after some of the spectacular patterns generated by the pulp mill proposal in the Tamar Valley. However, it turns out that in St Helens, the swing against Labor was just 3.4 points higher than in the rest of the sprawling and diverse Lyons electorate, and the swing to the Greens was 2.8 points higher than the Lyons average. All a bit ho-hum really; the thought that all that fuss and bother only shifted a net fifty or so votes in the town most supposedly impacted by it does make me somewhat sceptical that the issue had any real electoral impact anywhere – but perhaps a full analysis of Lyons will reveal a more complex picture.
