Poll Roundup: Shorten's Struggles Aren't Shifting Votes

2PP Aggregate: 52.3 to Labor (unchanged since end of last week)

Labor would probably win election "held now" with small majority

It's been another fortnight of sound and fury, dominated by speculation over Bill Shorten's leadership (amid worsening personal polls, unflattering reflections from The Killing Season, and concerns about aspects of his union background), still more anti-terror laws and another overheated culture war about some goose let loose on QandA.  Yet in this week's voting intention polls, very little has changed.

A ReachTEL late last week came out with a 52:48 to Labor headline, followed by a 53:47 by last-election preferences (53.5:46.5 respondent-allocated) from Morgan and a 53:47 from Essential.  The latter broke the run of seven consecutive 52s, as Essential finally caught a touch of the Green surge being seen by all other pollsters.  There was no Newspoll as the brand is in transition to operation by Galaxy.  (Meanwhile former Newspoll staff have started Omnipoll, which doesn't look like it will be doing regular voting intention polling, but which will continue many of Newspoll's other operations in an online format.)

After adjustments for the primaries and Morgan's house effect, I aggregated the ReachTEL as 48.2 to Coalition, the Morgan as 48 and Essential as 47.3, and the net impact is diddly-squat: a 0.1 point improvement for the Coalition on two weeks ago:

I should note that while the different aggregators have been pretty close together in recent weeks, PhantomTrend now has it at just 51.5, from which the "model's best guess" is an Abbott-Katter coalition. Mark The Ballot has 52.2 and I'll add in BludgerTrack when it's updated (update: it's 52.0, with a projection of 77 seats for Labor).  Labor's lead might look pretty solid but when translated into seat terms, it's a sliver.

Leaderships

ReachTEL has been showing remarkably flat net ratings for Shorten, but not any more.  Shorten's net rating crashed over ten points down to -26.2, making this another poll in which he has recently recorded a career-low rating.  There was a seven point rise in the number of respondents rating Shorten's performance Poor or Very Poor with falls in all of Very Good, Good and Satisfactory.  (Abbott also dropped a few points to -25, with a 3.9-point increase in Very Poor.  For those who like much prettier graphs than mine, check out Mark the Ballot's review of leader polling).

ReachTEL had Shorten leading Abbott as preferred Prime Minister, 56.3% to 43.7%, but this lead is a product of ReachTEL's forcing method, and is not comparable to polls that don't use forcing (which have lately tended to have Abbott leading.) As noted before I actually prefer ReachTEL's method, but Shorten's figure would include far more soft preferences from voters who dislike both leaders than would Abbott's.  Others voters strongly prefer Shorten, my estimate being 66:34 in his favour.

Speculation about Shorten's leadership has built in the last fortnight.  The reforms brought in on the reinstatement of Kevin Rudd were supposed to solidify the party leader's position, but in fact they have undermined Rudd's successor's authority by making him the first Labor leader in history to be clearly not supported by the rank and file.  The original protections against rolling the leader (excessive in the case of Opposition Leaders in my view) are also not yet constitutionally entrenched  (Article by Troy Bramston, title "Bill Shorten's fate back in the hands of Caucus") and it remains to be seen if they'll become so.  All the same, none of the Shorten stuff seems to be having the slightest impact on voting intention.  It's quite possible that a lot of voters who disapprove of Shorten just don't care who leads Labor so long as that person beats Abbott.

Issues Polling

ReachTEL asked a question about "confidence in the Australian political system".  It's no surprise that the result is slightly more high than low, and amusing that the breakdown by party correlates so cutely with what happened in the last election.  

ReachTEL found the Coalition favoured on handling of national security, but given that this is a Coalition strength area and especially given the PM's constant hammering of the issue, the narrowness of the margin (52.6:47.4) is unflattering, even for a forced-choice question.  (With a don't know option thrown in the lead would be somewhat greater.) 

Essential showed that respondents like the idea of the federal government handling all school funding and strongly dislike the idea of it funding private schools and leaving public schools to the states.  There was also strong disapproval for paying people smugglers to return to Indonesia, with a predictable partisan breakdown.  A question on trust in journalists showed Laurie Oakes both widely recognised and far more trusted than other options, with Alan Jones dragging along the bottom.

The AFR had a useful wrap of the most recent JWS True Issues release, showing that the government rates best on issues voters aren't so prone to rate as important. Comments under the heading "obsessional culture war" are especially relevant - hardly any voters consider ABC bias an important issue - and I wonder whether the Coalition might have made more headway off Shorten's miseries had they not been preoccupied with side-issues.

Spurious Concept of the Week: The October Curse!

I'm not going to even try to link to it as it has been heavily paywalled, but Phillip Hudson in Monday's Australian had an article about the idea that first-term Labor leaders are blighted by an October Curse.  This is based around the defeats of Curtin (1937), Whitlam (1969), Hayden (1980), Beazley (1998) and Latham (2004), all first-election ALP leaders who lost in elections held in October.  The suggestion was even that Tony Abbott should be frantically scurrying for double dissolution triggers to exploit this blight, even though (as Hudson basically admits) there would probably not be time to get Senate reform implemented properly by the AEC.

The first problem with this concept is that Labor actually won an election in October once, under Scullin in 1929.  That is, however, supposedly different, because Scullin was a second-term Labor leader.  Absent of any explanation as to why that should matter, this is simply special pleading (the explaining away of a data point that doesn't fit the pattern by referring to an irrelevant difference.)  It's hardly as if Scullin's victory in 1929 was a result of a vast increase in political skill obtained in the eleven months since losing the 1928 election; indeed, at the time of his 1929 win, Scullin had been in the job for less time than most first-term losers.  Nor is the record of second-election Labor opposition leaders by any means stellar.  Rather, the main reason Scullin won big in 1929 was that the Government he opposed had collapsed.

When looking at the supposed curse on first-term Labor Opposition leaders going to the polls in October, the important thing there is that first-term Labor Opposition leaders don't do well in general.  Excluding those who had already been Prime Minister their strike rate is 2 from 12; if we add in past PMs it's 3 from 13.  It's hardly statistically significant if a month that happens to contain five such elections produces a 0-5 split. 

Furthermore, of the five first-election October Labor losers, most of them actually polled reasonably well.  Beazley and Whitlam both won the 2PP, with Curtin and Hayden breaking 49% and only Latham (47.3) really letting the side down.  If we had 2PPs for the first up losses by Tudor and Charlton it would turn out that the average 2PP of first-term Labor Opposition Leaders at October elections is no worse than at any other time.  The real problem is that coincidentally the two of them who polled the best (Whitlam in 1969 and Beazley in 1998) could not convert their good 2PP performance into enough seats, because of personal-vote legacies of the Coalition's crushing victories in the elections immediately before.  

The "October Curse" then, is just a random clustering of Labor first-term-leader defeats that has no more merit to it than the meaningless precedents that infest US election punditry.   There is really not much variation in how well governments poll at different times of the year.  

That wasn't anywhere near the worst thing I saw in psephology this week, but time is short and work is uncommonly abundant.  I'll save the other one for a separate article sometime.

Not-A-Poll Update: Another deadline has passed and Abbott is still PM; as of the end of June 2015, 140/543 (25.8%) of voters had incorrectly predicted he'd be gone by now, with the rest still in the game.  151 votes (mostly old rope from the spill-vote phase) still say he won't make the election (or it will be held this year and he'll lose it), 129 say he will lose the election next year, and 121 say he will win it.  This includes 44 votes (8.1%) who predict him to have at least as long a career as Howard.