2PP Aggregate 52.4 to ALP (+0.5 in a week, -0.1 since two weeks ago)
Labor would very probably win election "held now" with small majority
This fortnight's polling has reflected a mixed news cycle for the major parties. The first half was dominated by Bill Shorten's mildly embarrassing appearance at the Royal Commission into trade unions, and a leak of internal ALP discussions about carbon pricing. The second half was dominated by an expenses scandal involving speaker Bronwyn Bishop, initially covering her unnecessary helicopter trip to a golf course for a party event, and also extending into aspects of her overseas travel and other expenses.
None of these things are anywhere near as large movers of voting intention as some people might expect, but that doesn't mean they have no impact at all. While the various moves in the polls over the last week might just have been random fluctuations, they are fully consistent with the Coalition having made small gains with Labor in the spotlight but then lost them all as the Bishop scandal extended to messy recriminations between Bishop and Joe Hockey. Hockey's misfortune was to be the first frontbencher to say what the Coalition generally including PM Abbott are now admitting: that the Speaker's behaviour has hurt them. Bishop's response of dredging up Hockey's past gaffes was a remarkably dumb case of shooting the messenger.
Our four polls for this fortnight are two Essentials (both 52:48 to Labor off identical primaries), a Morgan at 51:49 to Labor last week, and the second ever Galaxy-conducted Newspoll at 53:47 yesterday.
After considering the primaries I aggregated both Essentials at face value, the Morgan at 50.1 to Coalition and the Newspoll at 53.1 to ALP. The current result is little different from a fortnight ago and continues the pattern of nothing major happening in voting intention for at least four months:
A bit of wonky stuff here: there was a lot of variation in the readings on various polling aggregates last week, with Bludger Track on 51.0, Phantom Trend as low as 50.6, Mark the Ballot at 52.0 and mine at 51.9. A major cause of this unusual divergence was our different treatments of sample size issues, with other issues like inclusion or not of the new Newspoll, Poll Bludger's access to Essential weekly samples and the different rate at which we age data also having some effect. The Morgan had a larger influence on those aggregates that scale results compared to sample size, compared to those that treat Morgan as having a lower sample size (MTB) or ignore sample size entirely unless it is too small (KB). At present Phantom Trend is on 52.7, MTB on 52.3. Bludger Track is giving the new Newspoll a low weighting and is on 51.5; Andrew Catsaras is on 52.1.
Ignoring sample size was something I initially did because I liked to keep things simple, but as my aggregate has become slightly more complex it's a feature I've retained for two reasons. Firstly I find that the apparent differences in the ability of different polls to follow my aggregate - which should ideally be a proxy for how reliable they are - don't really have a lot to do with sample size and aren't all that large anyway. (I think this is partly because polls that aren't bouncy, like Essential, also have a tendency to go off course compared to the consensus of the others). Secondly I expect the quirks of varying polling methods may produce inaccuracy in some polls given certain issues mixes, but in others when the issues mix is different. For instance, I suspect that Morgan responds unusually, perhaps even unduly, strongly to "news cycle" items. I don't like to have any one pollster carrying much weight at any time, even when that pollster is most recently in the field and with a massive sample.
Mark the Ballot's post last week is useful reading for those into the technical question of which polls are more or less bouncy than their sample size predicts. (He doesn't use Essential, which I've previously found to be especially non-bouncy and which lately seems to be even sleepier than it used to be.)
It was popcorn time again with Newspoll delivering some more terrible netsats for both leaders. Prime Minister Abbott remained at -27 (33% satisfied, 60% dissatisfied) while Opposition Leader Bill Shorten declined to a new personal low of -32 (27-59). Bearing in mind the recent Newspoll methods change - which in my opinion had nothing at all to do with it - this represented Shorten's lowest satisfaction and highest dissatisfaction so far. Shorten also equalled Tony Abbott's lowest satisfaction when Abbott was Opposition Leader, though Howard I (23), Peacock II (22), Hewson (22), Downer (20), Crean (22) and Beazley II (26) all at some stage were even lower.
The combined netsat for both leaders was a new record for this term at -59, the lowest for three years and in the bottom 3% for the whole Newspoll era. Another statistic I like measuring is what I call the "disconnect" score - the mathematically minimum possible percentage of voters who appear to be dissatisfied with the leader of the party they would prefer on a 2PP basis. The combined disconnect score for both leaders this week (19%) was the highest for three years.
Essential last week had Abbott at an unremarkable -16 (37-53) but Shorten at a new personal low of -25 (27-52). A massive 29% of Labor voters disapproved of Shorten's performance.
Both Essential (37-30) and Newspoll (39-36) have Abbott ahead in beauty contest polling, though the Essential result should be seen in the light of a possibly friendly weekly sample. Shorten used to punch above the weight implied by Labor's 2PP on this score but not any more, as his own unpopularity is now dragging him behind.
A common question at the moment is: if Shorten is even less popular than Abbott and Abbott is the preferred Prime Minister, then how come Labor is leading 53:47 in Newspoll? The answer, as always, is firstly that preferred PM ratings skew strongly to incumbent PMs, and secondly that Opposition Leader netsats have much less to do with 2PP voting intention than Prime Minister netsats do. Far from suggesting that the 2PP is unreliable, the leadership figures show that the 2PP is about where it historically should be.
In one of my very first pieces on this site, The Abbott Factor, I presented an, in places, overly optimistic assessment of Gillard Labor's chances of making something of Abbott's then very poor ratings. At that time I noted that for Opposition Leaders prior to Abbott there had been a weak but real relationship between netsat and 2PP once the leaders were well-established in the role. I checked that relationship again this week and found that three years of Abbott and Shorten as Opposition Leaders has completely destroyed it. The only reliable link involving Opposition Leader netsats now is that when Opposition Leaders are really popular (netsat > +30, which mostly means Kevin Rudd in 2007) then their parties have unusually high 2PPs.
A by-election is very likely in the seat of Canning following the very sudden death of Liberal MP Don Randall at the age of just 62. If the by-election goes ahead (assuming there is not an early federal election) then this will be the first time a federal government has had to face a by-election since 2001, which was also the last time a sitting MHR passed away in office. Speculating about the by-election so soon after what has happened may seem tasteless to some, but there is bound to be much public interest in the outlook, so here goes.
By my rough count governments since Federation (including junior coalition partners) have won 54 by-elections for government-held seats and lost 20 (three of these to third parties or independents). The average 2PP swing against a government for those by-elections for which it can be calculated is about 6.3 points (standard deviation of +/- 4.7) and by-elections caused by the death of the sitting member display exactly the same pattern as the rest.
Canning on 11.81% is hence well above the average by-election swing, but by no means out of reach. I found only one previous case of a seat on a higher margin being taken (Wakefield 1938 with a 20.1% swing overturning a 13.3% margin), but also three other cases (Parkes 1931, Bass 1975 and Canberra 1995) where a by-election has been won with a swing above 11.8%, and two other cases with a swing as high or higher in which the government retained the seat. The Parkes, Bass and Canberra cases were all famous preludes to changes of government, while the Menzies government retained in minority at the election after losing Wakefield - but not for long.
Add in the general state of federal polling, especially in Western Australia, and the fact that Randall takes with him 19 years of personal vote and a larrikin image, and it would be unthinkably daft for the ALP not to contest the by-election. (It would also be tactless to announce they were running too soon.) Win or lose, it will be a valuable test of how solid and serious Labor's lead in the polls really is when translated to an election setting, and it is effectively a free hit since a substantial swing to the Opposition is more or less certain.
No-one should treat Canning as a safe seat for this by-election on account of its current margin, nor be greatly shocked if it actually falls. Based on current polling the Coalition would take a win in this seat by any margin, however small, while Labor should be disappointed and concerned if they can't get a swing of at least 10%.
PS: Antony Green in some background information here notes that there is a recent tendency for by-elections to see smaller swings when held in the first term of a sitting government. However seven of the nine first-term by-elections he refers to were for Opposition vacancies, meaning that the Opposition was losing the personal vote of the retiring member and possibly copping the backlash for dragging the electorate to the polls. One involving a government was the Lindsay by-election in which Jackie Kelly's win was annulled by the courts on eligibility grounds and Kelly received a record swing to her as a consequence. I suspect that the popularity of the government at the time of a by-election is a much more significant factor than its age.
A Flock Of Lambies?
This fortnight also saw some interesting claims about internal Liberal polling made by Victorian Liberal state president Michael Kroger. Kroger claimed on Sky Agenda that the Jacqui Lambie Network "could arguably win a minimum of two and possibly three seats in the Senate on current polling down there, she could poll around 20, 22, 24 percent of the vote". Sean Ford also reports the JLN (apparently that's a thing) is believed to be polling well in all other mainland states except South Australia where Lambie may not wish to compete with Nick Xenophon. Labor's Peter Beattie agreed with Kroger but did not describe any polling evidence in the process.
Many quotes of Kroger's comments in other sources have been inaccurate and have made it sound like he said she actually is polling those sorts of numbers. However, that's not exactly what he said, and even by the poor standards of internal polling rumours this one is very sketchy on details. Senate polling is a difficult enterprise at the best of times and I don't even know yet whether this is a proper poll with a readout of multiple options or just some kind of predictively meaningless "would you consider voting for X" type of polling. But if it's the first, I think that to effectively poll the support of a group like JLN in the Senate, especially outside Tasmania, you would need to include a large number of other parties in the readout as well as the obligatory Others/Independents.
Whatever the quality of the polling and the motivations of the source in referring to it, these reports have created concern that an army of Jacqui Lambies (a lambarmy?) would storm any double dissolution and invade about a tenth of the Senate. It's more or less clear based on Lambie's own vision for the JLN as more an umbrella brand than a party - quite aside from her previous involvement in PUP's ghastly state preselections - that rather than this creating a new, cohesive populist power bloc, it would really just flood the Senate with yet more random Senators (some with obnoxiously xenophobic views) who would sooner or later effectively be local indies much as Lambie and Lazarus are now. They would be uncontrollable even if Lambie wanted to control them, and having a bunch of them upstairs would probably be not much different to having any other bunch of randoms that would be elected in a double dissolution under the current system.
However what the Lambie issue adds to the mix is the possibility that even if Senate voting is reformed, a double dissolution might still produce an unstable crossbench. As this claim of a Lambie plague (a bit like a zombie plague, except zombies go looking for brains) is now an issue in debate about whether to go to an early election and in projecting the outcomes of Senate reform, I make the usual call for the full details of this polling to be released so that those debating these matters can see whether it all stacks up.
In the absence of any kind of useful polling I really can't say how Lambie is travelling in Tasmania. I don't subjectively sense that she has become the cult figure Kroger refers to, but perhaps as a Hobart inner-city dweller I don't move in the right circles to sense that. I suspect she has turned a corner in terms of cutting her public self-embarrassment rate and projecting a slightly calmer public persona and that she might well be re-elected and drag one running mate in with her in Tasmania in the case of a double-dissolution. However the idea that she has similar appeal in other states that have their own curious crossbenchers is not very convincing to me and I'd want to see proper data on it.
Not a huge amount of interest (to me anyway) here this fortnight. Essential found voters across parties divided on whether Australia's gun laws strike the right balance or are too lax, with only a fringe believing they are too harsh. Essential last week found that voters are divided over who will win the next election, with major party supporters generally thinking their own party will win as per normal. (Questions of this sort really need to stress something like "This question is asking who you actually think will win, not necessarily who do you want to win", to try to cut down the "barracking" response.) Essential this week found that Coalition voters think the government will run full term, while lefties think it will go early. A late-released ReachTEL question from their most recent national poll showed also that Coalition voters want the government to go close to full-term, while opposing voters want it to go to an early election, probably in the hope of getting rid of it.
I also saw some late-breaking Newspoll standard-of-living figures showing that Labor voters are more pessimistic than Liberal voters about changes in their own standard of living over the next six months. The history of this question has a near-perfect correspondence to the party of government at the time. If we find a net rating by subtracting those who expect their standard of living to get worse from those who think it will get better, then the net rating among supporters of the government of the day has been higher 57 times out of 58 (by an average of 13.8 points) with the sole exception being the July 1995 survey. The current gap between Liberal and Labor supporters is 20 points and the all-time high was 40 points in June last year. Mainly this is because Coalition supporters have been more optimistic about their own standard of living under this government than they were for the vast majority of the Howard years, which is quite a surprise given how much of the Howard era was an economic boom.
Thursday Update: Morgan Leader Polls
Morgan has issued one of its infrequent leadership phone polls, based on a sample of 587 voters. Unfortunately no voting intention figures were released, but all the same, some rather stark results.
Both leaders were found to be unpopular with Abbott on a -25 netsat (the same as last June) and Shorten on a new personal worst (from any pollster) of -36 (24% approve, 60% disapprove). The party breakdowns are the big ticket item here with Abbott enjoying a +51 net rating among Coalition supporters (71-20) while Shorten is -12 among Labor supporters (36-48).
The latter figure should be treated with caution given the small sample size (in isolation the 48% disapproval has a margin of error of 6.5 points, but since I'm cherrypicking it as the most notable among a bunch of figures, treat the MOE as more like 10). Still, even from a sample of 230 or so Labor voters it's pretty amazing. I thought it might be a record, but it isn't: in October 2009 (shortly before he was rolled) Malcolm Turnbull polled a -23 netsat (31-54) among Liberal voters - the last time until now that a leader polled a negative net rating from their own party's voters. Tony Abbott's highest disapproval rating among Liberal voters ever polled was 36% in the January 2012 sample.
A very big difference between Shorten and Turnbull there is that Shorten is at the helm of a party with a solid 2PP lead while Turnbull was in charge of a party being thumped 42:58.
Morgan also has its usual smorgasbord leadership polls. Again note that the question here is about who a voter would choose as a party's leader for the next election if they voted for that party, a more complex question than simply who the voter prefers as leader. On the Liberal side it's Turnbull 44 leading Julie Bishop 15, Abbott 13 with Morrison, Hockey and Joyce all on 5. Among Coalition voters, Turnbull again leads Abbott 32-26 with Bishop 16 and Morrison 13. Bishop is down 12 points since April and a common suggestion is that some respondents are confusing her with, or else blaming her for, Bronwyn Bishop.
On the Labor side Plibersek 26 leads a resurgent Albanese 19 with Shorten down from 21 to 12, Wayne Swan 10 and Bowen 7. Among Labor supporters Plibersek 32 Shorten 17 Albanese 16.
The poll was taken on the nights of July 20-22, with Bill Shorten's 7:30 interview announcing support for boat turnbacks probably just catching some of the third night of the sample.