The Poll Bludger

In case you were wondering, The Australian reported on Monday that the first Newspoll since the election – indeed, the first poll on voting intention of any kind since the election, unless someone else quickly gets in first – will be published “very shortly”.

In the meantime, I offer what will be the first in a series of posts that probe deep into the results of the federal election region by region, starting with Sydney and some of its immediate surrounds. Below are two colour-coded maps showing the two-party preferred swing at polling booth level, with each booth allocated a geographic catchment area built out of the “mesh blocks” that form the Australian Bureau of Statistics’ smallest unit of geographic analysis (typically encompassing about 30 dwellings). The image on the right encompasses the core of the city, while the second zooms further out. To get a proper look at either, click for an enlarged image.

In a pattern that will recur throughout this series, there is a clear zone of red in the inner city and the affluent, established eastern suburbs and northern beaches regions, giving way to an ocean of blue in the middle and outer suburbs. The occasional patches of red that break this up are often associated with sophomore surge effects, which played out to the advantage of Mike Freelander, who had no trouble retaining Macarthur (more on that below); Susan Templeman, who held out against a 2.0% swing in Macquarie; and Emma McBride, who survived a 3.3% swing in Dobell (albeit there was little to distinguish this from a 3.1% swing in neighbouring, Liberal-held Robertson).

The second part of our analysis compares the actual two-party results from the election with the results predicted by a linear regression model similar to, but more elaborate than, that presented here shortly after the election. This is based on the correlations observed across the nation between booth-level two-party results and the demography of booths’ catchment areas. The gory details of the model can be found here (the dependent variable being Labor’s two-party preferred percentage). The r-squared values indicate that the model explains 76.5% of the variation in the results – and doesn’t explain another 23.5%. Among the myriad unexplained factors that constitute the latter figure, the personal appeal (or lack thereof) of the sitting member (if any) might be expected to have a considerable bearing.

Such a model can be used to produce estimates that hopefully give some idea as to where the two parties were punching above and below their weight, and where the results were as we might have expected in view of broader trends. The latter more-or-less encompasses Lindsay, which was the only seat in the Sydney region to change hands between Labor and the Coalition (the only other change being Zali Steggall’s win over Tony Abbott in Warringah). The table below shows, progressively, the model’s estimate of Labor’s two-party vote, the actual result, and the difference between the two.

The first thing that leaps out is that the current leaders of both parties did exceptionally well, with their margins evidently being padded out by their substantial personal votes. Beyond that though, patterns get a little harder to discern. The Liberal-versus-independent contests in Warringah and Wentworth appear to have had very different effects on the Coalition’s two-party margins over Labor, which reduced to a remarkably narrow 2.1% as voters turned on Tony Abbott in Warringah, but remained solid at 9.8% in Wentworth, suggesting Dave Sharma may have accumulated a few fans through two recent campaigns and a dignified showing in the wake of the by-election defeat. That there was nonetheless a 7.9% two-party swing to Labor illustrates that he still has a way to go before he matches Malcolm Turnbull on this score.

The modelled result further emphasises the particularly good result Labor had in Macarthur, a seat the Liberals held from 1996 until 2016, when Russell Matheson suffered first an 8.3% reduction in his margin at a redistribution, and then an 11.7% swing to Labor’s Michael Freelander, a local paediatrician. At the May 19 election, the seat defied the national pattern in which outer urban seats that responded had unfavourably to Malcolm Turnbull swept back to the Liberals, with Freelander in fact managing the tiniest of swings in his favour. In addition to Freelander’s apparent popularity, this probably reflected a lack of effort put into the Liberal campaign, as the party narrowly focused on its offensive moves in Lindsay and Macquarie and defensive ones in Gilmore and Reid.

The latest of Essential Research’s fortnightly polls, which continue to limit themselves to issue questions in the wake of the great pollster failure, focuses mostly on the Israel Folau controversy. Respondents registered high levels of recognition of the matter, with 22% saying they had been following it closely, 46% that they had “read or seen some news”, and another 17% saying they were at least “aware”.

Probing further, the poll records very strong support for what seem at first blush to be some rather illiberal propositions, including 64% agreement with the notion that people “should not be allowed to argue religious freedom to abuse others”. However, question wording would seem to be very important here, as other questions find an even split on whether Folau “has the right to voice his religious views, regardless of the hurt it could cause others” (34% agree, 36% disagree), and whether there should be “stronger laws to protect people who express their religious views in public” (38% agree, 38% disagree). Furthermore, 58% agreed that “employers should not have the right to dictate what their employees say outside work”, which would seem to encompass the Folau situation.

Respondents were also asked who would benefit and suffer from the federal government’s policies over the next three years, which, typically for a Coalition government, found large companies and corporations expected to do best (54% good, 11% bad). Other results were fairly evenly balanced, the most negative findings relating to the environment (26% good, 33% bad) and, funnily enough, “older Australians” (26% good, 38% bad). The economy came in at 33% good and 29% bad, and “Australia in general” at 36% good and 27% bad. The poll was conducted last Tuesday to Saturday from a sample of 1099.

Also of note:

• A referendum on indigenous recognition may be held before the next election, after Aboriginal Affairs Minister Ken Wyatt’s announcement on Wednesday that he would pursue a consensus option for a proposal to go before voters “during the current parliamentary term”. It is clear the government would not be willing to countenance anything that went further than recognition, contrary to the Uluru Statement from the Heart’s call for a “First Nations Voice enshrined in the Constitution” – a notion derided as a “third chamber of parliament” by critics, including Scott Morrison.

• A paper in the University of Western Australia Law Review keeps the Section 44 pot astir by suggesting 26 current members of federal parliament may fall foul by maintaining a “right of abode” in the United Kingdom – a status allowing “practically the same rights” as citizenship even where citizenship has been formally renounced. The status has only been available to British citizens since 1983, but is maintained by citizens of Commonwealth countries who held it before that time, which they could do through marriage or descent. This could potentially be interpreted as among “the rights or privileges of a subject or citizen of a foreign power”, as per the disqualifying clause in Section 44. Anyone concerned by this has until the end of the month to challenge an election result within the 40 day period that began with the return of the writs on June 21. Action beyond that point would require referral by the House of Representatives or the Senate, as appropriate.

Some unrelated electoral news nuggets to keep things ticking over:

• The Australian Electoral Commission has announced the deregistration of two right-wing minor parties, the more newsworthy of which was Cory Bernardi’s decision to decommission Australian Conservatives. This party owed its party registration to Bernardi’s position in the Senate, rather than its having 500 members, so the matter was entirely in his hands. In a sense, this also means an end to Family First, which won Senate seats at the 2004, 2013 and 2016 elections and had a presence in the South Australian upper house from 2002 to 2017, when it merged with Bernardi’s newly formed outfit. However, Family First appeared to lose energy as evangelical Christians increasingly preferred to direct their organisational efforts towards the Liberal Party, and was dominated in its later years by deep-pocketed former Senator Bob Day. Even further afield, the Rise Up Australia party, associated with controversial pastor Danny Nalliah of Catch the Fire Ministries, has voluntarily deregistered.

• JWS Research has released the latest results in its occasional series on issue salience, recording only one particularly noteworthy movement over the past three surveys: defence, security and terrorism, which only 20% now rate in the top five issues most warranting the attention of the federal government, down from 23% in February and 29% in November. “Performance index” measures for the government across the various issue areas have recorded little change post-election, except that “vision, leadership and quality of government” is up from 35% to 42% (which is still the fifth lowest out of 20 designated issue areas). The survey was conducted from June 26-30 from a sample of 1000.

• In the New York Times’ Upshot blog, Nate Cohn casts a skeptical eye over the record of online polling in the United States. It notes a Pew Research finding that YouGov’s “synthetic sampling” method achieves the best results out of the online pollsters, by which it “selects individuals from its panel of respondents, one by one, to match the demographic profile of individual Americans”. Another survey that performed relatively well, VoteCast, did so by concurrently conducting a huge sample phone poll, results of which were used to calibrate the online component.

Contrary to expectations it might put its head above the parapet with today’s resumption of parliament, there is still no sign of Newspoll – or indeed any other polling series, at least so far as voting intention is concerned. Essential Research, however, is maintaining its regular polling schedule, but so far it’s been attitudinal polling only. The latest set of results was published in The Guardian on Friday, and it encompasses Essential’s leadership ratings series, which I relate here on a better-late-than-never basis. Featured are the first published ratings for Anthony Albanese, of 35% approval and 25% disapproval, compared with 38% and 44% in the pollster’s final pre-election reading for Bill Shorten.

To put this into some sort of perspective, the following table (click on image to enlarge) provides comparison with Newspoll’s debut results for opposition leaders over the past three decades. The only thing it would seem safe to conclude from this is that Albanese’s numbers aren’t terribly extraordinary one way or the other.

Scott Morrison’s post-election bounce lifts him five points on approval to 48%, with disapproval down three to 36%, and he leads Albanese 43-25 on preferred prime minister, compared with 39-32 for Shorten’s late result. Also featured are questions on tax cuts (with broadly negative responses to the government policy, albeit that some of the question framing is a little slanted for mine), trust in various media outlets (results near-identical to those from last October, in spite of everything), and various indigenous issues (including a finding that 57% would vote yes in a constitutional recognition referendum, compared with 34% for no). The poll was conducted June 19 to June 23 from an online sample of 1079.

Elsewhere in poll-dom:

• Australian Market and Social Research Organisations has established an advisory board and panel for its inquiry into the pollster failure, encompassing an impressive roll call of academics, journalists and statisticians. Ipsos would appear to be the only major Australian polling concern that’s actually a member of AMSRO, but the organisation has “invited a publisher representative from each of Nine Entertainment (Sydney Morning Herald/The Age) and NewsCorp to join the advisory board”.

• A number of efforts have now been made to reverse-engineer a polling trend measure for the last term, using the actual results from 2016 and 2019 as anchoring points. The effort of Simon Jackman and Luke Mansillo at the University of Sydney was noted here last week. Mark the Ballot offers three models – one anchored to the 2016 result, which lands low for the Coalition in 2019, but still higher than what the polls were saying); one anchored to the 2019 result, designed to land on the mark for 2019, but resulting in a high reading for the Coalition in 2016; and, most instructively, one anchored to both, which is designed to land on the mark at both elections. Kevin Bonham offers various approaches that involve polling going off the rails immediately or gradually after the leadership change, during the election campaign, or combinations thereof.