Have pollsters created an ‘alternate universe’? 

It’s impossible to overstate how central polling is to the political class in Australia. It is commonplace for politicians and staffers to obsess over every poll for evidence of how they will go at the next election and how their party and leader are performing — and that the media love reporting polling stories. The heavy reliance on polling is reinforced by existing trends in the media, in which outlets have cut back on journalists, forcing those left behind to cover far more with less time, which encourages race-calling journalism rather than more resource-intensive policy analysis.

Veteran Liberal staffer and Abbott loyalist Peta Credlin called our obsession with polls damaging to democracy. “It’s not good the way that we fixate on these fortnightly polls,” she said in the wake of the election, “and we can see now how far away they are from the actual results.”

However, it’s not just politicians and journalists. Lobbyists worked out during the Howard years that if you could present polling that backed your case, you’d get much further with government ministers than if you merely had evidence or powerful connections on your side. Unions and activist groups like GetUp now use polling to drum up publicity from poll-hungry media outlets, usually with little rigour around their methodologies.

How polling has changed

High-quality polling has become more problematic as voters become harder to reach and communications become fragmented. The wholesale abandonment of landlines for mobiles — and the lack of any independently verified list of mobile numbers — has rendered live phone polling far more difficult. Robo-polling has a tiny response rate and is despised in the community. Online polls skew young and urban. Face-to-face polling is very expensive — costing perhaps $40,000 for a national survey, compared to $2000-$3000 for robo-polling. No one makes money from political polling, and with Nine-Fairfax dropping Ipsos after the election, the financial pressure on the industry will only increase.

Industry figures INQ spoke to thought that the 2019 failure was actually overdue given how the problems of obtaining an accurate sample had increased in recent years. Australia has indeed seen extraordinarily accurate polling at the federal level up to 2016; Andrew Bunn, who ran Essential’s poll for years, said he’d warned a polling conference in 2013 that the good run of Australian pollsters couldn’t last. The growing problems around state polling recently might have suggested a reckoning was finally coming: pollsters significantly underestimated the Coalition’s victory over Labor in NSW earlier this year, the Liberals’ victory in Tasmania in 2018 and Labor’s performance in 2o17 in WA.

Robo-polling, though, has brought low-quality polling within the reach of community groups and activists. GetUp and unions have mastered the art of commissioning cheap robo-polls, often at a seat level, with loaded questions favouring the issues they want to highlight, and coupling it with voting intention questions.Because mainstream media outlets love polling stories, these are usually run uncritically by journalists. For just $2000-$3000, groups can get some free publicity for themselves and their issues, even if the robo-polling proves wildly inaccurate.

But what if it is all inaccurate? What if we had to accept that polls might be wrong by at least as much, and maybe more than, the two-party preferred gap between the parties, misleading us about who was really winning? A government trailing 49%-51% and facing defeat instead could enjoy an election-winning lead. A prime minister considered doomed in fact might be heading for re-election. What if, as per the Mark the Ballot analysis, the Turnbull government had enjoyed a small but consistent lead over Labor for all of the last parliamentary term? In his final media conference after being forced out, Turnbull said the Liberals’ internal polling suggested the party was about even or a little ahead of Labor.

Who are the main players?

Newspoll is a brand owned by Galaxy Research, which itself was acquired by British polling firm YouGov in late 2017. YouGov, which did not respond to requests for comment, was also hired by Labor’s campaign team to provide its internal polling, making it Australia’s dominant pollster in what is now a highly concentrated political polling market, with only Ipsos and Essential Research providing competition.

Nine-Fairfax announced after the election that it would no longer be running Ipsos’ polling. However, Essential’s polling continues to appear in The Guardian but, temporarily, has dropped its voting intention results (and Essential always emphasised its issues-based results rather than voting intention).

With French-owned Ipsos no longer running at Nine, Newspoll — assuming The Australian restores regular coverage — will be further entrenched as the dominant political opinion poll, despite getting it badly wrong in the most recent election. Newspoll also has the greatest uncertainty about its methodology. Essential is an online poll with the results weighted to ensure the sample better reflects the population — although which kind of weighting, given there are a range of possible approaches, isn’t known. Ipsos — which is most transparent about its methodology — is a phone interview poll. Newspoll is now a combination online and robo-call poll, but how those disparate sources are combined and weighted (online polling skews young, robo-polls skew old) isn’t at all clear. As Emeritus Professor of politics at Macquarie University Murray Goot explained in his history of Australian polling earlier this year, lack of methodological clarity in polling is a persistent problem.

That makes Newspoll’s dominant position in the Australian political landscape problematic.

The power of polls

Party machines invest heavily in internal polling during election campaigns, conducting nightly polls to track how their leader, campaign and policies are faring with voters (in Labor’s case this year, it seems, they were given a bum steer by YouGov. But they also conduct internal polling outside of campaigns — internal polling (not one, but two, sets) played a key role in the ousting of Kevin Rudd in 2010.

Turnbull doesn’t have a lot to complain about when it comes to losing his job. He famously invoked Newspoll as the reason for removing Tony Abbott. Turnbull’s own persistent lack of success in Newspoll after the 2016 election became a key reason for his ouster.

Newspoll results are magnified by The Australian‘s front-page coverage, the demand among press gallery journalists for polling stories, and the authority attributed to it by politicians. However, it can not merely make or break prime ministers, but routinely sets the political agenda, particularly around set-piece political events like the budget or major policy announcements, and in the second half of each term of parliament as politicians start to focus on the next election. But as the 2019 experience demonstrated, Newspoll in fact created an entire alternate political reality within which politicians and journalists were operating without realising it was significantly inaccurate.

Turnbull ousted Tony Abbott because Abbott had been behind in 30, presumably accurate, polls. But a Turnbull government that was persistently ahead in the polls would have had a very different dynamic than it did in reality — and might have dramatically changed Labor’s political strategy during the last term, especially in relation to the leadership, and to potentially damaging policies like its commitment to end the franking credits rort. It would have also strengthened the hand of Turnbull against Tony Abbott and the reactionary rump in the party determined to destroy him, and the lunatic fringe in the far-right media that attacked his every move.

While Turnbull might have actually been better placed than the polls suggested, he was also unlucky — implausibly unlucky. Despite his narrowing the gap with Labor in all published polls from early 2018, not once did he benefit from a rogue poll that gave his government a 50-50 result or a narrow lead, as should have happened several times given the large number of polls that had him close to Labor. Herding, or pollsters massaging results to ensure they’re not too far off other polls, may not have just been a problem in the lead-up to the election, but might have hurt Turnbull as well.

But imagining much of the last term of parliament through the lens of a Turnbull government ahead in the opinion polls is so difficult precisely because opinion polling has become so fundamental to the way both participants view what is happening and how the media reports what is happening in politics.

Flawed figures and fragile facts

Interest groups and activists will continue to use dodgy robo-polls for media releases. And many in the media will continue to eagerly report polls — the most recent Newspoll was reported by some other outlets without any acknowledgement of its election failure, and not just at right-wing outlets like Sky.

If journalists, editors and producers can’t wean themselves off polling, then the risk of us again living in a Matrix-like political fiction will persist. The only way to combat that is for journalists to become significantly more sceptical, and more expert, about how polling works, about the methodologies of polling firms, and about the possibility that they’re wrong.

That’s a tough ask for journalists, but Australia has a number of respected psephologists — academic, media and online — who have been poking and prodding under the hood of political polling for years and who should be drawn on much more consistently by media outlets.

And, ultimately, polling companies will have to accept that if they want the profile and influence that comes with political opinion polling, then they have to accept the obligation of greater transparency about their methods.