Griffith: Sound And Fury, Signifying Little

(With federal polling comments added Feb 11)

Of all the comments about the result of yesterday's Griffith by-election (won by Labor's Terri Butler but with a small swing to the Coalition) the one I liked best was the comment by Antony Green - that we would probably talk about the result for a day and then move on.  It might be just a bit longer than that, but in case it isn't, today's that day and here's my contribution.

By-elections are noted for often producing crazy swings and aberrantly strong results for minor-party candidates but this one is notable for the modest nature of all the party swings.  Not a single party contesting the election saw a swing exceeding 2% in either direction in its primary vote.  The LNP has currently gained 1.35% (although I expect this figure to increase slightly), Labor has currently lost 1.38% (this may also increase) and the Greens have currently gained 0.01% (this could become a small loss).  These small swings are against the backdrop of a slight change in the opposing candidate mix - the Palmer United Party which polled over 3% in the general election did not contest, while the Pirate Party is sitting in fourth place on 1.55%, having not contested this (or any other lower house) seat at the general election.  (A note in passing on the various fourth party attempts: while none were able to crack even 2%, it's interesting that the Pirates more than doubled their Senate vote from the same electorate last year while Katter's Australian Party could barely improve on theirs, even in the absence of Palmer United.)




PUP preferences split rather evenly between the Coalition (38%), Labor (32%) and the Greens (30%) in this seat in 2013, but some of the PUP voters would have voted for other fourth-party candidates anyway.  The Pirate Party would probably have drawn votes primarily from the Greens, and to a lesser extent from Labor.  With the slight differences in the field taken into account, I think the Coalition is effectively up half a point or so on its primary and Labor effectively down about two points, with the Green result more or less neutral.

The Coalition are rather churlishly refusing to concede the seat until the postals are counted, but they are over 3000 votes behind with about 8000 postals to come.  They would need about a 70:30 split of formal postal votes compared to 52:48 at the main election and this won't happen.  With the same margins on a general election night the seat would have been well and truly thrown in by now and it is hardly good form for a Government that hardly needs any more seats to be refusing to accept its failure to pull off what would have been a remarkable win anyway.

There is a lot of partisan spinning going on that tries to argue that Labor's win is either a gross indictment of Campbell Newman and Tony Abbott, or else a miserable failure because they should have won by so much more.  Compared to the general election, Labor had both advantages and disadvantages that should be taken into account:

Advantages for Labor

* Improved national polling since the election, especially with the weight of Gillard/Rudd infighting removed
* Traditional strong performance by oppositions in by-elections
* Unpopularity of state premier Campbell Newman in electorate
* Ability of candidate to focus on the electorate.  Kevin Rudd was singularly hampered in this regard in 2013 because he needed to not only run as Prime Minister but be heavily involved in a repair job for his party following a leadership stoush

Disadvantages for Labor

* Loss of personal vote of outgoing local member (who contested general election as Prime Minister)
* Blame for having caused by-election through Rudd's resignation (Rudd having said he would serve his full term if re-elected)
* Increased profile of opponent Bill Glasson through having recently contested general election

Readers may note I do not list Bill Glasson's popularity and qualities as a candidate as disadvantages for Labor.  The reason I do not is that they were already factors in the 2013 result in the state and hence are already included in the base result.

The partisan spinners can be detected by the way in which they take arguments from one side and argue that these should have generated a swing of several points, while ignoring the arguments from the other side.

What Are The Precedents?

Much has been made of the scarcity of 2PP swings to Governments at by-elections.  This is indeed an uncommon event in "classic" contests where the Opposition and Government are the last two candidates standing.  Peter Brent (here) gives some examples: Lindsay 1996, Fremantle 1994, Richmond 1984, and ACT 1970.  Earlier I find Bendigo 1969, Grey 1963, Calare 1960, and Bendigo 1960 as some other examples.  As is well known the only case of a government actually winning a by-election against the Opposition came in 1920, when Hugh Mahon was expelled from parliament for sedition and lost the subsequent by-election.  This is from about 57 by-elections (at a quick count) of seats vacated by Oppositions, of which around half have been contested by the Government of the day.

Spinning by the government about the Lindsay result being the last case of such a swing should be taken in this context: In the years since 1996, only once (Gippsland 2008) has a Government contested a seat vacated by an Opposition.  The Howard government never contested a seat vacated by Labor, while a 6.1% swing against it at its sole attempt to contest a Coalition seat dissuaded the Rudd government from trying again.  The Hawke/Keating government contested five in its first term (for four small swings against and one miniscule swing to) and then didn't bother again apart from the three-cornered Groom by-election in 1988.

Of the most recent cases, Lindsay was unusual in that the incumbent had been deposed by an eligibility issue rather than resigning.  Fremantle 1994 was also unusual in that the Labor candidate was former state premier Carmen Lawrence, who was not only high-profile but extremely popular even after her party's defeat at state level.  And Richmond 1984 had something of the flavour of the current contest, with a popular party leader (Doug Anthony) standing down not long after an election loss.

Exact precedents for the Griffith by-election are scarce since there are only two prior cases of a former Prime Minister resigning soon after an election defeat: Keating in 1996 (Government did not contest) and Fraser in 1983.  Fraser's resignation of his rather "safe" seat (9.7% margin) saw a 1.1% swing to the then Liberal Opposition, despite the strong federal polling of the new Hawke government.

Those interested in state precedents can see the Poll Bludger coverage here.  However, I note that the examples of state governments picking up swings in by-elections there all involved state governments that were in opposition federally. 

This by-election featured no neutrally-commissioned public polling.  Snippets of ALP internal polling were released, earlier examples of which had Glasson on too high a vote and later cases of which were reasonably close to the result but slightly underestimated the gap between the major party candidates.

All things considered I thought on the day that the par score for Labor in this by-election was a small swing to them, to around 54.5% 2PP.  Rather than trying to adjust the reasoning that led to this estimate too much post hoc,  I reckon that the real result actually is a very slightly disappointing margin for the ALP.  It may be telling us, for instance, that the party's current apparent lead in federal polling is a tad soft when it is put to the test.  It will also add interest to the results of the Newspoll believed to be now in the field, since if there has been a swing back to the Coalition in federal voting intention in the last week or two then that alone could account for this underwhelming result.

At the very least, at an electoral test that had the potential to demonstrate a backlash against the new Abbott government (or failing that the Queensland state government) has failed to do any such thing.  A win was never likely and would have been historic and remarkable; a narrow loss therefore was the best they could realistically expect.  As Peter Brent notes, these kinds of by-elections are historically all over the place as predictions of the next election.  Even if there is noise about this being a bit of a flop for Shorten's leadership, in the long term I don't think it will mean a lot.

But on that score, I found it notable that in Terri Butler's victory speech, when she mentioned the names Bill Shorten, Tanya Plibersek, and Anthony Albanese in that order, the cheering for Plibersek was about twice as strong as the cheering for Shorten, and the cheering for Albanese then twice as strong again.  Shorten did receive a better cheer the second time Butler mentioned him.  For all that he has started with some happy national polling figures, Bill Shorten has some way to go to receive the most rapturous love of one batch of his party's most ardent supporters. 

Update Monday 10 Feb: The addition of the first batch of postals has increased the swing to the Coalition to 0.94%.  Also the number of postal envelopes issued, initially shown as around 9000, now shows at around 11,000.  Glasson got a 58:42 split on the first lot of postals, slightly better than I expected, but even if that rate were to continue he would still finish up with around a 49:51 overall loss.  Meanwhile the Greens' pre-election prediction that they would beat 10% is threatening to narrowly fail.

Second Update: Further postals have actually broken slightly in favour of Butler, making something of an embarrassment of the Coalition's decision to hang out for them before conceding.  The swing has dropped back to 0.84%.  There could still be close to 7000 postals to throw.

Update Tuesday 11 Feb: Postals today headed back more in Glasson's direction than those from yesterday, increasing the swing against Labor to 1.32%.  But with Butler over 2500 ahead with not more than 3500 (if that) left to throw, Glasson has finally thrown in the towel.

Federal Polling Update

I'm putting this here because it is connected to what I discussed above about the result perhaps indicating Labor's polling lead was soft or had already softened.

Some possible corroboration of the idea that the vote reflects drift in Labor's national polling arrived with Newspoll down a point to 51:49 to Labor and the usually ALP-friendly Morgan (by last election preferences) down 1.5 points to the same reading.  As we know by now, Essential failing to move off 50:50 means very little, though it may well have company soon.  The net effect is a drop of 0.6 points for Labor in my polling aggregate to 50.7%. 

It's too early to say for sure that this is a real shift but if it is the question is why.  There has been the odd sighting online of a theory that the government's polling is improving because people are focusing more as the festive season ends, but this is a completely baseless view. In fact it is slightly more common historically (about 12 cases to eight, not statistically significant) for governments to lose support early in a polling year, as indeed happened to Labor under Gillard to a disastrous degree just one year ago.  My suspicion considering the evidence of Griffith as well as the polling is that Labor's lead is not what it was late last year and that the issue mix so far this year (boats, union corruption and welfare) has not been friendly turf for Labor.

Yesterday's Newspoll also came with a wallop for Bill Shorten whose netsat is now zero (35% satisfied, 35% dissatisfied, 30% uncommitted), a 17-point drop since the last Newspoll.  Essential confirmed this giving Shorten a -4 netsat; their second lukewarm reading for him this year and his first negative reading excluding ReachTEL (whose way of asking the question is not comparable).  In Newspoll to Newspoll terms the 17-point drop is the 12th worst netsat change for an Opposition Leader ever, but that is a misleading stat given that the usual break between polls is two or three weeks rather than two months.  When we look at falls over the same time period, this sort of thing happens every year or two on average, and has happened to every Opposition Leader in the Newspoll era excepting Rudd and the second term of Andrew Peacock (whose ratings in that term were too terrible to begin with to fall far.)  Most Opposition Leaders have seen this sort of fall repeatedly.  The more notable thing (though not necessarily as a predictor of anything) is the overall "meh!" factor in Shorten's current ratings.  This sort of profile (very high uncommitted with satisfied and dissatisfied roughly equal) was previously seen at similar stages of their careers from only Simon Crean and Brendan Nelson.

Meanwhile Abbott's ratings (his Newspoll netsat stable on -5) continue to add to the evidence that in the short-term at least, two-party-preferred polling is in lock-step with perceptions of the PM's leadership.  After five data points, variation in Abbott's net satisfaction score has explained 97% of variation in the Coalition's 2PP.  This will come down, but maybe not by much. 

Incidentally this year was the second-latest commencement of the Newspoll polling cycle ever; only in 1988 did they first venture out later than this.  (In some years their first poll release was later, but included polling over multiple weekends.)  That said it is still less than a week later than the last post-election-year return in 2011.