Image for Election 2010: how Denison confused the nation

The indecisive outcome of the most recent federal election was always a possibility, but it surprised me that it actually happened.  With the government’s polled support generally around 51-52 in the final two weeks it appeared that the damage might not be so great, but in the end the swing in the last 24-48 hours went to the Opposition and not the Government, leading to a net loss of eleven seats from the government’s previous position.  Labor thus became the first first-term government since Scullin’s to not win reelection outright.  As I write it is attempting to herd the various crossbench cats in order to form a minority government that could be almost as messy as its previous term (oh, to beg for such scrapings when once so much was yours!) and it remains to be seen whether it will succeed. 

The post-night election count for the House of Representatives, for the most part, has not been that exciting.  In 2007, eight seats were decided with a two-candidate margin of 50.25-49.75 or closer, and one of these dragged on for weeks and was eventually settled in the courts.  In 2010, the closest seat (Corangamite) currently sits at 50.41 for Labor and is no longer in doubt; all others are outside 50.5.  There is a lot of life left in the Senate count, with a scrap between the DLP, Family First and the Nationals for a Victorian position still unresolved (though the DLP are well placed), as is one between Family First and the Liberals for a position in South Australia. But in the House of Reps count, our own electorate of Denison has at least livened up an otherwise boring post-count by providing a complicated count that was way over the head of virtually everyone.  As with the confusion over the national two-party vote, the media’s fumbling attempts to follow what the hell was going on in Denison have not only exposed a predictable level of mainstream reporting incompetence, but have also shown that the AEC and ABC websites have a long way to go in terms of making their output foolproof for popular consumption.

It all began on the night with 44 votes in the booth of Sandfly (Denison).  A semi-rural oddity on the edge of an urban electorate, the recently added Sandfly (Denison) booth is one of my favourites to observe as the figures come in.  As one of the smallest booths in Denison, Sandfly (Denison) is blisteringly fast in getting its figures through to the computer, and in this case, a mere 32 minutes after the polls closed it was up.  And it was showing 18.88% for Andrew Wilkie.

For months I had wondered (without thinking there was much more than a few percent chance that it would ever actually happen) about how Andrew Wilkie might win Denison, if it was indeed winnable by anyone but Labor.  The path to victory seemed clear enough: he had to beat either the Liberals or the Greens (probably the Greens), use the preferences of one of those to pass the other, and then win from second place.  If he could beat the Greens by even the barest of margins he had an excellent chance, but it seemed so unlikely that that would happen, when it required him, at the very least, to more than double his state vote.

So on election night I was watching the early booths to see how Wilkie would go, armed with matching data from the state election specifically so I could monitor the Wilkie vote booth-by-booth and see how he was going compared to the state election.  And it wasn’t the 18.88% in Sandfly (Denison) that made me wonder if we were in for an interesting week, but the change in his vote at that booth from the state election, where he had polled a mere 6%. 

One never knows what to make of a swing in a small and potentially unrepresentative booth; small booths can swing dramatically because of local issues under the radar of analysts unfamiliar with them, or because some bunch of voters from somewhere else happened to be in the area for some kind of community event.  But still, an eyebrow was raised, because if that swing to Wilkie was repeated across the electorate he could well be competitive.

After about five booths it became clear he was more than competitive.  Most of the booths that had been counted were northern suburbs booths, where Wilkie was obtaining primaries of just 3-6% in most booths in the state election, compared to double figures at most booths within Hobart City. If he had made further inroads in the south of the electorate, he would surely be brought back to ground in Glenorchy?  Yet this wasn’t happening at all.  Wilkie was, in fact, coming third on primaries, and in a winning position if he could only stay ahead of either Jackson or Couser.  For a while he even moved into second.

At this point it was time to sound the alarm that there was a boilover on the cards in Denison.  This was done by posts here and at Pollbludger just before 7pm.  At 7:24 pm national realisation that there was something going on erupted with a flurry of comments on the Twitter channel #ausvotes and the Pollbludger live feed.  Finally at 8:30 pm the ABC television broadcast, up til then assuming the seat was a safe win for Jackson based on the Liberal/Labor two-candidate preferred, cottoned on.  Even so, some other networks were still calling the seat for Jackson without comment at least another half an hour later.

By 9:08 pm the ABC were calling the seat won for Wilkie, in my view prematurely, based on a preference projection claiming that Wilkie would win by as much as 57:43 against Jackson.  Not only was that preference projection hard to credit, but there were still nuances of the order of exclusion to be dealt with.  But around the same time, while it was already “very likely” (in my view) that Wilkie would win the seat, Labor sources were trying to argue that they were more than an outside chance of holding on.  Among the first to insert foot in mouth was Paul Howes, who was claiming that Wilkie was slipping behind the Liberal candidate into third place (which he was) and therefore Labor still had good chances.  But in fact, Wilkie being in third place was irrelevant, provided that Wilkie recaptured second in the cut-up using Green and Socialist preferences.  Karl Bitar soon followed suit in disputing the ABC’s call of the expected result, and many erroneous comments about the potential relevance of the Liberal being second on primaries were made late into the night, some of them by people who were not actually Labor hacks and really should have known better.

Unfortunately, during the next day, the ABC’s computer went back to using the erroneous 2CP (two-candidate preferred) count between Jackson and Simpkins, presumably sourced from the AEC, and hence was calling the seat as an overwhelming win for Jackson when in fact that 2CP count was redundant. The next round of (more serious) confusion over the seat started when the Australian Electoral Commission made a correct decision to conduct a provisional two-candidate throw between Wilkie and Jackson, on the assumption that they would be the last two candidates standing.

This process involved reallocating all the votes received to that point to either Wilkie or Jackson, instead of reallocating them to either Jackson or Simpkins.  The booths were recounted in a more or less alphabetical order and as this was conducted, running tallies of this two-candidate count were being posted to the Australian Electoral Commission site, and picked up by the ABC.  And here, two peculiarities of the Denison electorate kicked into play.  The first is that the two-candidate preferred vote for Jonathan Jackson (ALP) compared to Andrew Wilkie (IND) was very strongly arranged along municipal lines.  Jackson won the 2CP over Wilkie in every booth that was firmly in the Glenorchy municipality, and lost in all the Hobart booths and booths further south, including the boundary outpost of Creek Road.  The second is that the Glenorchy and Hobart booths occur unevenly through the alphabet.  Thus, of the twenty booths starting with the letters A to G, twelve are in Glenorchy.  But of the 27 booths starting with the letters H to W, only eight are in Glenorchy, the rest being in Hobart.

The consequence of this was that Jackson jumped to an immediate and large two-candidate lead.  He won eight of the first eleven booths recounted and with a 2CP lead of several points, was showing as way in front with both the ABC and the AEC computers switching to calling the seat for him, and a flurry of inaccurate media and blog reports claiming that Wilkie wasn’t looking like a winner and calling the seat for Jackson.  I saw over 100 such wrong reports from various sources, but the most dunderheaded was a report printed in the Sydney Morning Herald and The Age, and credited only to “Staff reporters”.  Hopefully the (ir)responsible reporters and editors involved in this one have all been fired by now and are no longer staff.  Here we go then:

“ONLY ONE SEAT IN DOUBT AS WILKIE LOSES BID FOR DENISON

Labor appears to have won the crucial Tasmanian seat of Denison, with the Australian Electoral Commission dashing the hopes of the high-profile independent candidate Andrew Wilkie.

The former intelligence analyst, who turned Iraq War whistleblower, was initially tipped to steal the Hobart-based seat from Labor with a huge primary vote.

However, the AEC - having counted more than 5 per cent of the two-party preferred vote today - has just declared the seat for Labor’s Jonathan Jackson.”

Well, firstly, the AEC doesn’t declare seats until a little stamp with the word DECLARED on it appears next to the seat, and that isn’t done until the outcome is absolutely mathematically certain.  Secondly the reporters were unaware that the alphabetical foibles in the seat would mean the seat would turn around as counting continued (even without this, 5% is hardly enough votes to draw conclusions anyway).  Yes, the AEC and the ABC sites were now showing Denison as a win for Jackson, but as the AEC site notes, it automatically “calls” the seat for anyone with more than 50.5% 2CP, and this should not be considered final.

Yet I believe that the AEC and the ABC must take a large share of the blame for the media confusion at this point, never mind the fine print on the AEC page, for allowing their computers to still “call” the seat for Denison during this redoing of the 2CP count.  There needs to be a provision on both sites for a subjective manual override to declare a seat in doubt until the dust settles whenever the 2CP is being updated in a non-standard manner (such as through an alphabetical booth count).  Hopefully we will see more flexible AEC procedures in the future.

It was obvious that Wilkie was going to catch up to Jackson as booths more favourable to him were added in.  The question, early in the booth count, was would it be by enough?  Many observers attempted to model this using projections, but it was actually quite a difficult statistical problem.  If Wilkie could pick up 66.8% of all preferences from the Liberal, Green and Socialist Alliance candidates he would be home and hosed.  But the share he picked up varied from booth to booth.  In the first booths counted he was doing very well at Battery Point (78%) but poorly at Bowen Road (62), Chigwell (59) and Claremont (57).  And it wasn’t as simple as Wilkie doing badly on preferences in the strong Labor booths and well in the weak Labor booths, since Wilkie also didn’t do too well on preferences in weak Labor booths that were strong Green booths (like Cascades and Fern Tree, also early in the alphabet), because the Liberals preferenced Wilkie far more heavily than the Greens did.  Not only were the early booths counted distorting the two-candidate preferred count but they were also distorting projections of how much that count would change.  Even as late as 4pm on the first day of this process, Antony Green was still projecting a slim Labor lead over Wilkie on 2CP.

By far the best of the Denison projection attempts was by pollbludger member sykesie, who after eleven booths produced a linear regression model that projected Labor’s final 2CP value at 48.57%, which turned out to be only a fraction of a point out.  This model found that the Labor 2CP vote in each booth could be predicted as a function of the size of the Labor and Green votes (the latter probably just because the Green preferences were weaker for Wilkie than the Liberal ones and his own votes; it did not mean Green preferences were flowing for Labor).  So while Jackson had comfortably won big Labor booths, he was headed for big trouble in strong Liberal booths and in middle Hobart booths where neither the Labor nor Green votes were all that high. 

Yet after one night of the booth-by-booth reallocation, Jackson sat on 50.51% 2CP, which being just over 50.5% 2CP meant that the AEC and ABC computers were still calling Jackson winning and ahead respectively.  Those of us who knew the electorate knew that this lead would be blown away the next day.  And so it was, with thumpings for Labor in booths like Mt Nelson (63-37), Sandy Bay (67-33), Taroona (65-35), Waimea Heights (71-29) and the biggest hiding of all, Sandy Bay Beach (76-24 to Wilkie) Thus at the end of the reallocation of votes cast up til then, Wilkie led by 1375, with the 2CP just over 51:49 in his favour.

At this point, two new misconceptions crept in.  The first was that Wilkie only had to hold this lead on postals and he was home.  But actually that would only be one part of Wilkie’s challenge – the 2CP reallocation did not explain whether Wilkie would stay ahead of Couser, or whether he would overtake Simpkins.  The second misconception was that Jackson could do well on postals and get the 57:43-ish split he needed on postal and other non-ordinary votes to catch up.  An example of the degree of Labor insider shellshock by this stage was this post on Pollbludger from a poster claiming to be In The Know: 

“As for Dennison [sic], I understand that the libs did not preference wilkie in the postals. Everyone thinks 56% of the postal vote is achievable. The figure most people were concerned about was if he was 1700 or so in front. Confident that he can be mowed down.”

The misconception that Labor had a serious hope was based around two ideas: firstly that Wilkie would do more poorly on primaries on postals, and secondly that he would do more poorly on preferences.

The idea that Wilkie would do poorly on postal primaries was based on the view that independents often struggle on postals as they cannot organise postal vote campaigns as easily as incumbents.  But in the state election this had not hurt Wilkie greatly; he had polled 7.7% of postal primaries compared to 8.5% from other sources.  Furthermore, it soon became clear that not only had Wilkie run no postals campaign in Denison, but Labor had not run one either.  AEC figures for “party postals” (postal votes based on applications sent out with party advertising material, and therefore likely to favour the party sending them out) showed that Labor had targeted its (theoretically) marginal seats of Bass, Braddon and Franklin for a party postals campaign, ignoring Denison and Lyons.  The Liberals, perhaps anticipating this and with an eye to their Senate vote (not that that did them any good) had targeted Denison, Lyons, and to a lesser degree Bass, ignoring Braddon and Franklin.  Hence Labor had no postals campaign advantage over Wilkie and, furthermore, Labor tend not to do well on postals anyway.  The only party making an effort on party postals in Denison was the Liberals, and if that netted them more votes then that meant more preferences to Andrew Wilkie.

The idea that Wilkie would struggle on postal preferences was based on a key difference between booth voting and voting by post.  Booth voters are handed how-to-vote cards.  The Liberal how-to-vote card preferenced Wilkie at #2, netting him over 80% of Liberal preferences.  Would Liberal voters preference Wilkie anywhere near so strongly if they were not getting the How-To-Vote card?  I must confess to giving the chance that they might not some credence for a whole 15 minutes before looking at a relevant example – in the 2007 elections in the seat of Melbourne, the flow of Liberal preferences to the Greens was only eight points lower on votes not subjected to how-to-votes than to votes with them.  All other things being equal (which they wouldn’t be anyway) Labor needed a difference of more like 18 in the Denison case to make a difference.

It soon turned out there was scarcely any difference in how Wilkie performed on the postal preferences.  It seems that the few remaining Denison Liberal voters did not need how-to-vote cards to remind them of how much they hated Labor.  All the hype about how Labor would crush Wilkie down on postals eventually amounted to nothing whatsoever – at the time of writing the postals in fact advantage Wilkie, albeit by only eighteen votes.

However, the Liberals’ efforts on postals soon resulted in a new mini-wave of speculation, because their candidate was stretching the lead over Wilkie in second place on primaries, and maybe if Simpkins was a thousand votes or more up the road things might get interesting in terms of whether Wilkie made it into second following the three-way split of Couser’s votes between Wilkie, Simpkins and Jackson? 

I had already modelled this and found they wouldn’t; anything less than a 1500 vote lead would be blown away for sure.  The aborted Jackson/Simpkins 2CP results of the first night showed that Simpkins had done very poorly indeed from the combined Green and Wilkie preferences, and it stood to reason that he would do more poorly on the Green preferences than the Wilkie ones.  Also, the Wilkie/Jackson 2CP results, especially in booths where the Greens were strong and the Liberals weak, showed that Jackson couldn’t be hogging too many of the Green preferences himself.  It was hard to see why the Wilkie-Jackson-Simpkins split of Green preferences would be anything other than grossly unequal.  (I thought it might be something like 45%-40%-15%; in the end Wilkie got 53% to Simpkins’ 8!!) Thus, three days and four hours after noticing something amiss in the tiny booth of Sandfly (Denison), it was finally time to start telling anyone who cared that Andrew Wilkie had certainly won the seat.

Jonathan Jackson impressed me with his gracious acceptance of the result where he took full responsibility for an outcome that he is probably feeling quite bewildered and dismayed about.  It that was far more blame than he needed to take, because while the blame was partly his, it belonged largely to his party – and even if us pundits did not generally expect that Labor would actually lose the seat, the risk that they were taking was apparent all along.  Labor had handed Jackson the candidacy on a platter, as a result of a factional deal rather than through a competitive process, and in so doing had left the seat in the hands of a candidate lacking in political experience and whose CV seemed decidedly thin.  There was negligible effort to market Jackson to the electorate with adequate lead-in-time and to ensure that the transition from the 23-year-veteran Duncan Kerr to the new chum was sold to the electorate properly over a course of several months.  The general attitude of the party (perhaps inspired by the way Scott Bacon sailed into a seat in state politics without needing to show much of substance along the way) seemed to be that in such a safe seat it was enough to simply anoint a candidate who had a DNA connection to Tasmanian Labor politics, and make a very token effort.  Yes, even though that connection was to a former minister who by her final state election was so unpopular she had to be fished out of the slop on Jim Bacon’s preferences after polling tenth on primaries, behind three Labor neophytes and four of the Liberals.

Other responses to Jackson’s defeat were much less impressive and in some cases bizarre.  Antony Green’s blog received many angry complaints from posters clearly annoyed that a candidate could win from third on preferences, and Pollbludger even had a Labor supporter (albeit an idiosyncratic one) saying she saw merit in the first-past-the-post system.  The fact is that a party that lives all the time by the sword of preferences must now and then expect to die a little by it.  And Labor is certainly that party.  Labor won nine seats from behind on primaries in 2007, without which we would have had our hung parliament three years earlier and John Howard would probably still be prime minister.  Labor won the 1990 election on preferences, without which Andrew Peacock would have made it to the Lodge.  And if you really think that losing from a clear lead with a whole 36% of primaries (oooh!) is so unfair then take a step into your Tardis and set the dial for 1998; congratulations, you’ve just re-elected Pauline Hanson.

Indeed for those who think Denison is a good model of what shouldn’t happen in a preferential election, it is impossible to discuss it without reference to the defeat of Hanson in the seat of Blair in 1998.  Hanson polled 36% of the primary vote, 11 points ahead of the Labor candidate in second, with the Liberal, Cameron Thompson, third on 21.7.  But on the preferences of the Nationals (who finished fourth), Thompson moved ahead of Labor, and on the preferences of Labor, Thompson defeated Hanson and won the seat.  Of those who did not vote 1 for Pauline Hanson, 83% preferred the Liberal, and that’s why Hanson lost.  Denison is almost exactly the same story – with the difference that in Hanson’s case, all significant parties preferenced against her.  Among those who did not vote 1 for Labor in Denison, 80%,  from Socialists through to Sandy Bay Liberals, preferred a rather left-leaning independent who polled just 8% in the state election and three years ago could only get a gig as a Green #2 for Senate.  And that is a remarkable result. 

But perhaps the most peculiar response of all came from the State Secretary of the ALP, John Dowling, who in correctly pointing out that Labor’s slack postal campaign did not cause their defeat, sought to lay the blame for the loss of the safe Labor seat on … the Liberals! 

Yes, that’s right; according to Mr Dowling, “the postal vote campaign in Denison was quite successful.”, but the tories failed to pull their weight:

“The collapse in the Liberal Party’s vote was quite significant because that enabled the Independent to actually go in front of the Labor candidate.”

“I think the Labor vote has held up quite well in the northern suburbs but in conservatives [sic] booths where the Liberal vote has collapsed, that vote has in fact gone across to an Independent.”

There are actually only three truly conservative booths in Denison, and those are the three that returned 2CPs in favour of the Liberals in 2007: Lower Sandy Bay, Sandy Bay Beach and Waimea Heights.  Even the main Sandy Bay booth is a mixed bag because of the number of students in the area, returning a 2CP of 52.7 for Labor in 2007 – every other booth strongly prefers Labor over the Liberals at federal elections.

Secondly, the Labor vote didn’t hold up over the northern suburbs at all - if you look at what you should be looking at, the primary vote.  The median primary swing in northern booths was about 11 points, little better than the average for the electorate as a whole, and in the southern parts of the Glenorchy area (Lutana, Moonah, Bowen Road etc) the Labor primary was crunched by 14-15 points, above the electorate average, even though it was coming off a lower base than in the rest of Glenorchy. 

If you look at the two-candidate preferred result, the damage for Labor across the northern suburbs looks not too bad, with a median loss of seven points, and some booths only down by three or four.  But that impression is grossly misleading, because comparing a Labor/Liberal 2CP to a Labor/Wilkie one is not comparing like with like.  The major difference is that whereas Green preferences flowed overwhelmingly to Labor in 2007 (83:17 in Labor’s favour), in 2010 they tended to flow to Wilkie (57:43 of those not going to Simpkins).  Therefore, if Labor was dropping several points to Wilkie on a 2CP comparison in the northern suburbs, it was inevitable he would be trashing them in the Hobart municipality, where the Green vote is much higher, on the same measure. 

And so it transpired – the biggest hits to Labor’s 2CP on that misleading comparison between 2007 and 2010 came not in these mythical “conservative booths” Mr Dowling is complaining about, but in the radically Green booths and the mixed-left inner-suburban booths.  Cascades 37 points, Fern Tree 36.5, Hobart West 35.8, Hobart 32, Hobart South 31.6, Landsdowne 30.1, Swan 29.6 and so on. The rich Liberal booths in Sandy Bay delivered some of the lowest 2CP differences compared to 2007 in the whole Hobart municipal region, largely because Labor had so little left to lose to anybody else there anyway. 

So, with the Labor primary having crashed by 12 points, we’re supposed to believe that the Liberal primary crashing by seven was the real reason Jonathan Jackson lost.  That’s even though we already know from the state election that Andrew Wilkie rips at least four points from the Liberal primary by existing and by saying words like “poker machine” and “refugee”.  Indeed, the real reason Labor lost is because there was a massive swing from them direct to Andrew Wilkie across the entire electorate, that had not occurred in the state election.  In most booths the gain by Wilkie compared to his 2010 state election result was within 20% either way of the size of the primary vote loss from Labor compared to their 2007 federal result.  In those northern suburbs booths where the swing away from Labor was lowest, Wilkie made further inroads into the Liberal vote as well as trashing Labor, and in some of the northern suburbs booths what is actually occurring is that Labor is losing a lot to Wilkie and a little to the Greens, and Wilkie is gaining from the Liberals as well. 

There were also some booths where the swing away from Labor far exceeded the state-to-federal swing to Wilkie, and these were generally Hobart municipality booths where the Green vote is quite high, except for the booth of Montrose, where the Wilkie vote was unusually high for the area at the state election.  What seems to be going on in most of these booths is that Labor is bleeding votes not only to Wilkie but also to the Greens and in some booths even the Liberals, but the Green gains (as in the state election) are masked by them losing as many votes or more to Wilkie.  (My thanks to Dr Louise Crossley for sending me some spreadsheets concerning swings in the Green vote by booths, which show that the pattern is broadly similar to the state election – small gains where both Wilkie and the Greens are weak, but erratic results and even losses for the Greens in booths where Wilkie competes effectively for the historically strong Green vote.)

On the whole though, the main story of this seat is that Andrew Wilkie, who was previously only a threat to the Liberals and the Greens and didn’t bother Labor’s support base very much, has capitalised on discontent with a slack Labor campaign, a poorly-promoted little-known Labor candidate and a perception that Denison has been ignored during its time as a safe Labor seat, to take his support to a whole new level and make a massive direct raid on the Labor Party primary vote.  And this is Labor’s fault, and trying to blame those shifty Sandy Bay tories (a few of whom might have had the guile to cast a tactical 1 Wilkie vote just to get rid of Labor) is not exactly facing up to responsibility for the loss.  Unfortunately, this kind of response from Dowling is pretty similar to what we’re seeing from the Bitars, Arbibs, Howeses and so on as they try to deflect blame for a pretty shambolic federal campaign onto Kevin Rudd, the mysterious leakers, the meeja, the Liberals – pretty much anyone other than themselves.  Seriously, Labor has got to learn from this, or Lyons will be susceptible to a similar indie raid when Dick Adams retires. 

That’s not to say that the ALP campaign in Tasmania was bad – in all other respects it was a stunning success, with Bass, Braddon and Franklin no longer marginal and an overwhelming victory in the Senate.  Indeed, while EMRS have been justly shellacked for almost completely missing the size of the Wilkie vote in their polling (mainly because they didn’t mention his name), the firm’s polled projection of a massive 2PP swing to Labor in the state was only slightly overstated, refuting sceptics like me who thought the swing would be much less.

There is already much comment about whether Wilkie can retain his seat next time around, a good example being all this rubbish by Glenn Milne.  The truth is that Wilkie’s election depended on primary vote defections from all three parties, on both Green and Coaltion preferences, and on Labor stuffing up both their candidate selection and their pre-campaign.  With a very thin margin, if we just replay the tape with the Labor Party actually awake (decent candidate and big campaign), the seat automatically returns to sender anyway.  And furthermore if Wilkie went with the Coalition he would probably have annoyed both the Greens who preferenced him ahead of Labor, and the ex-Labor supporters from the northern suburbs whose defections to his cause were critical.  As far as I can determine, those whose preferences Wilkie relied on to beat Labor (including those who gave him their first vote) were probably split fairly evenly between the major parties.  By going with Labor he may very well have irritated the Liberal voters who supported or preferenced him, and some Libs may well be seething with disgust about that at the moment and saying all kinds of silly things. 

However, if the Libs preferenced Labor ahead of Wilkie at the next election they would be sending out a hilariously confused message by giving Labor an unconditional vote in parliament to punish Wilkie for giving Labor a very conditional, precarious and limited vote in parliament.  Not only would this be a thoughtless act of pique but it would be strategically stupid in that it would reduce the effort Labor needed to devote to recover the seat.  Sounder strategic logic has seen the Liberals preference the Greens in Melbourne and Denison (even Eric Abetz eventually failing to get his way on the latter) although they always knew there was no chance of the Greens ever repaying the favour.  I suspect that in Melbourne, a few smart-aleck Liberal voters even voted tactically 1 Green to ensure that Labor lost the seat – thus cancelling out the few smart-aleck Labor voters who were definitely voting 1 Liberal to try to save it. 

(Among the other strange claims in Milne’s article are the claim that Wilkie “will have to get 50 plus one per cent of the vote not to have to rely on Liberal preferences” (obviously Glenn hasn’t heard of Green preferences) and the claim that had the Greens got over Wilkie they would have won the seat (they might have done so, but it’s unclear either that the Wilkie voters would have put them over the Liberal or that they would have won the 2CP against Jackson).)

Wilkie’s election was so fortuitous that by the various fragile logics that got him (just) over the line, he is already a one-termer no matter what he does, and there’s no guarantee that term will be a long one.  The best he can do is ignore all backlashes and make the most of it for the things that he believes in and that he wishes to achieve.  But that baseline goes out the window, and the closeness of this result becomes irrelevant, if he is able to transform his support base and make further inroads into the support levels of the other parties, and turn himself into a local cult independent a la the rural indies or, a better model still, Ted Mack.  Having twice underestimated Wilkie’s mass appeal in the Denison electorate before, I will not be writing off his chance of doing that.

Dr Bonham voted 1 Wilkie to get rid of Labor out of Denison, and effectively voted for Lisa Singh in the Senate to get rid of Guy Barnett.  As usual he siphoned his Senate vote through someone who couldn’t possibly get elected, in this case the Secular Party, who once again finished last.