Wednesday, March 19, 2014

SA Election: Libs Fall Short Of Majority, Again!

(Note re Tasmanian state election: the seat postcount threads below will continue to be updated when signficant news is available, in the evenings only from Thursday onwards but hoping for more often once the cutups start mid next week.  Links to them will remain in the sidebar on the right)


Unfortunately I've been too busy with the largely-as-expected Tasmanian state election thrashing to even manage a proper devouring of some of my pre-election words on the South Australian election.  I thought that the Liberals had good chances to get a majority in SA since only one state poll (when properly interpreted) had pointed to a strong chance of a hung parliament, and even if the 2PP vote was below 53% to the Liberals, then they could have other avenues to majority victory.

As it turns out they've struck out on both their chances - the 2PP vote seems to have been closer to the final 52.3% Newspoll than to any other poll in the campaign (including the 55:45 final ReachTEL of which no other details have been seen) and the Liberals' performance in terms of seat-harvesting has been not much better or worse than random.  Their best hope of victory now is to scrounge a 23rd seat in late counting to place them on a better footing to deal with the two Independents, but even then, success is not assured.   If they can't flip any seat from the ALP's current lead (or even if they flip just one), then Labor may have pulled off an escape even more precarious than last time.  Of course, the Independents will have the final say about that, but it's much easier governing with a 25-22 majority than a 24-23 one.

You can see more detailed discussion on the SA Election Late Counting thread at Pollbludger but for now Labor seems to be winning 23 seats to the Liberals' 22 and two independents.  The Liberals have taken Mt Gambier from independent Don Pegler, and Mitchell, Bright and Hartley from Labor.

The seat of Colton has been in serious doubt but today's prepolls broke very slightly in Labor's favour. My current projections are that assuming the informal vote is as usual slightly lower on declaration votes than the level so far of 2.8%, and assuming the final turnout is the same as in 2014, then there are about 3407 formal primary votes remaining.  Paul Caica leads by 452, so the Liberals need about a 56.7% 2PP flow on declaration votes. They scored exactly this on the first lot of postal votes, but today's declaration votes (presumably including at least some non-postal declaration votes) were not so friendly.

Doubt also persists in some quarters concerning Ashford, but unless the turnout in this electorate increases markedly on the rather poor 91.8% last time, then there may be only about 3100 formal votes left, which is not enough to overturn a lead of 661.  Even if the turnout does increase, it's a big hill to climb with a 56% split only on votes cast so far.

In my leadup article for the election I discussed the difference between pendulum-based readings of a given 2PP result and probability-based readings.  I used the then in-vogue example of a 54% 2PP vote.  For the lesser 2PP vote that has actually happened (though it will be a while before we know exactly where it has settled) the methods give much more similar results.  On a uniform swing assumption, the swing should have picked up Hartley, Bright and Ashford and has instead picked up Hartley, Bright and Mitchell.

However, if an adjustment is made for retiring members, the uniform swing should have taken Elder as well. Elder is the seat in which Labor targeted Liberal candidate Carolyn Habib with the now infamous "CAN YOU TRUST HABIB?" attack pamphlet, the style of which was widely seen as a dog-whistle to anti-Islamic sentiment through the focus on Habib's surname, shadowy silhouette and choice of font and background masonry.  I am not sure whether or not it is refreshing that culture-war-type positions trumped partisanship with Labor's Ed Husic attacking the ALP pamphlet, and Liberal Cory Bernardi denying it was racist.  Whatever it was, it seems to have worked.

In the probability-based analysis I employed, a 52.5% 2PP should have given the Liberals on average 1.38 of the seats due to Labor; they have won either one or two depending on the fate of Colton.  The same 2PP should have given Labor on average 1.88 of the seats due to the Liberals; they have apparently saved two, Ashford and Elder.  The final 2PP (whenever it eventually arrives) should be a bit higher than 52.5% but similar conclusions are likely to follow.  The Liberals appear to have failed to get their majority because they did nothing special in the marginals and their overall 2PP was not high enough to overcome Labor's natural and largely geographic advantage.

The funny thing is that in one respect the Liberals were successful in managing the swing.  In their own seats (few if any of which were in danger, despite murmurs about Adelaide) the net swing was about zero while in Labor held-seats it is currently running at about 1.3% (and more or less static on average from the most to the least marginal).  Overall the seat swings are about as variable as expected with a standard deviation of 3.2 points, but a lot of that is caused by the swing of 13 points to the Liberals in Stuart.  If that seat is removed it comes down to +/- 2.6.

On the probabilistic model, given that swing in Labor seats they should have expected to take four or five Labor seats rather than three.  Here is a graph showing the swings at this election by previous margin for all the Labor-held seats (based on figures as of last night):

The closer to the thick black diagonal line, the closer the seat this time around.  As declaration votes are added more of the red dots will move up the page towards the black line, but it looks like none, or at most one, of them will cross it.

Really it is not the case that the Liberals' marginal-seats strategy did especially badly.  It is rather the case that the electoral geography of the state required them to win more of the 2PP to be highly confident of majority government, and that their seat results for the swing obtained were mediocre.  The presence of two independents occupying seats where the Liberals presumably won the 2PP doesn't help them either.

Of course there has been the usual wailing about Labor being in a competitive position when they were clearly thumped on the 2PP (by a margin similar to the Labor win in the 2007 federal election for example). But there is really not much that can be done when a state happens to have a split between a weakly Labor-leaning urban area and a smallish strongly Liberal-leaning rural population. (Other than, hmmm, for the Liberals to adopt policies that are more attractive to the capital city?) Rather than trying to force a single-member system to comply with a requirement it was never designed for, reformers need to look to an entirely different system.  The problem with that is that no form of proportional representation would be likely to give the Liberals a majority either.  An alternative might be a top-up system in which the side winning the 2PP was automatically granted extra list seats to give it a majority, but this would hardly be fair if voters actually chose to elect numerous independents or a third party and on average wanted a hung parliament.

In any case, the parties can only play the game as available to them, and Labor has a simple response to any claims that to win with such a low 2PP is outrageous: that had the system required them to win on 2PP they would have adopted a different election strategy approach and may have done so.

Finally even if the Liberals form a minority government this result is another strong argument against the predictiveness of betting markets.  The Liberals were at $1.01 to form government on election eve and Labor at $13 - even longer odds for Labor than in Tasmania where Labor were thrashed! Even those of us psephs who believed the Liberals were more likely than not to win, would not have written Labor off on election eve to that extent (an implied probability of just 7%.)

I will update this article in the evenings as time allows with any significant progress in the post-count or on the small matter of who actually forms government.  I will also have comments on polling accuracy when the final seat and 2PP results are known.

Freaky footnote:  Those chortling about the Liberals' apparent failure to secure enough seats (who might want to wait until a deal is done before getting too jubilant about that) may be amused to know that the Twitter-assigned URL for a link I posted on Twitter to this page - normally a jumble of completely random numbers and letters - in this case includes the letters "oWNED".

Thursday March 20: All seats are now looking done and dusted at 23-22-2 to Labor. Further counting today did not make any seriously close.

Sunday March 23:  Well, that didn't take long at all!  Geoff Brock has decided to support Labor, and with that, a $1.01 favourite has lost an election!


  1. Thanks Kevin for all your efforts and comments in regard to South Australia. Always outstanding observations. I can tell you one thing, Kevin, that not too many public servants voted Liberal in SA last Saturday. People have observed what has happened to the public service in the other states and I know in my workplace many people changed their votes from Liberal to Labor, especially in regard to job security and penalty rates. The SA public service is the most aged in the country and workers were terrified of ending their last few pre-retirement years on the dole. I hope the Independents back Labor!

  2. Quote from Kevin: "The funny thing is that in one respect the Liberals were successful in managing the swing. In their own seats (few if any of which were in danger, despite murmurs about Adelaide) the net swing was about zero while in Labor held-seats it is currently running at about 1.3% (and more or less static on average from the most to the least marginal). "

    Depending on how you view it, you could argue that the Liberals did even better than this at getting the swings in the right seats. According to the current ABC swing figures, there were 10 seats with margins less than 4% against the Liberals (9 Labor and 1 Ind). In 9 out of 10 of these seats the swing was to Liberal with Ashford being the odd one out. However in the 10 safest seats in the state (margins over 16% prior to the election) the swing was to the ALP in 6 of these seats. So the Liberals actually did well in terms of getting the swings to go in their direction in the marginals, but the swings weren't strong enough to get them over the line.

    Or possibly this is just a correction from 2010 when comparatively too much Liberal effort was placed in the safe seats?

  3. Yes my expectation was the Liberals might be able to get more of a correction than this from what happened in 2010.

    They did get more swings in marginals but the swings were often too small, any corellation between swing and marginality was not that steep.