When Malcolm Turnbull challenged Tony Abbott for the leadership of the Liberal Party, one of his main justifications for the move was because the party had lost 30 Newspolls in a row. Given we’ve lost more prime ministers to bad polling than at the ballot box over the past five years, who are these deciders of democracy? And in an election campaign, how much influence do the pollsters have?
As Crikey’s Bernard Keane noted when examining pollsters in 2013, it’s not entirely clear that polls have a direct effect on how people vote. Academic studies of whether people support the underdog or are impressed by political momentum show mixed results. Politicians love to say that the only poll that matters is the one on election day, and Turnbull now insists that he will leave commentary on polls to the the commentators, but between now and July 2, party strategists, candidates and the media will all be intensely focused on a variety of opinion polls, both public and private, to shape policy and the direction of the campaign over the next seven weeks.
Polls matter to the media. Australia’s major outlets love to get their hands on exclusive polling. The Australian, for example, builds a lot of its political clout on exclusive publication of Newspoll and will run copious coverage about why each individual result is significant. Fairfax does the same for its polling — by Ipsos — as does Crikey with Essential. ReachTEL has an association with Channel Seven. Morgan is the only major poll without a media backer.
These polls shape the opinion pages. A party riding high in the polls will be assumed by commentators to be a good one. When a party plummets, commentators will adjust their analysis to explain why.
The public can pay curious attention to polling too, says, Osman Faruqi, founder of new polling company MetaPoll, which combines its own online political polling with aggregated results from other polls to come up with what it bills as a more reliable poll. He says pollsters often find the people they are polling quoting their own polls back to them in response to questions about how one politician is tracking against another.
“Most Australians, when they do tune in [to politics], it’s often because a poll has been released and it gives them some sense of how people are tracking. [And] all politicians pay attention to the polls. If they say they don’t, they’re totally lying to you because the major parties spend enormous sums of money doing their own polls to work out where votes lie, and how to switch those votes.”
The methods of polling vary, and they have changed a lot in recent years. Polling used to be mainly done to landline phones, for example. But as Australians increasingly abandon landlines, polling companies have begun to call mobiles. Some use text messages or online surveys (as Essential does). Or pollsters use a mix of these methods.
Newspoll is the undisputed gold standard of polling, says polling expert Peter Brent, for both its longevity and reliability. But it, too, has changed, probably more than most people realise. Last year, while keeping the brand name, News Corp turned to Galaxy Research to conduct its polling. Unlike the old Newspoll, Galaxy relies on a mix of robocalls and online polling to come to its results — it’s cheaper this way.
Live phone calls (as opposed to robocalls) are seen as the most reliable method of gauging voting sentiment. But the Galaxy Newspoll switch, Brent says, has worked. “The whole corporate reputation has carried over to new outfit. The political class still hangs off it — it’s been a very successful change.” And Galaxy’s results have been stable — overcoming the volatility that was characteristic of the old Newspoll, he adds.
Despite the change, psephologist Kevin Bonham expects Newspoll will remain as reliable. He thinks people overstate the impact of things like live calls on the reliability of the results. Political polling isn’t a quiz. Pollsters carefully calibrate and weigh the answers they receive to create a poll representative of Australia’s demographics. The process is highly complex and subjective — pollsters have their own “secret sauce”.
Can robocalls still give a good result? “It depends how well you do it,” Bonham said. “The scaling, sampling, other methodology decisions … these are more important choices than whether or not you’re live calling.”
Perhaps the only outfit still relying on live calling is one of the newest: the Fairfax-Ipsos poll. But Ipsos has yet to garner much influence, Bonham says. “It has the disadvantage of being relatively new. And I think people are a bit sceptical of some of its behaviour. It tends to have a very high Greens vote, for example.” Like Newspoll, Fairfax’s major poll is new this election — the media company had a long association with Nielsen until 2014, when the pollster decided to stop political polling in Australia.
In a bid to overcome the biases of any one poll, most psephologists turn to aggregated polls. This is what MetaPoll does, for example, as does Crikey psephologist Poll Bludger. Such aggregated polls are a staple of American political coverage — the most famous aggregate poll is Fivethirtyeight, done by Nate Silver, previously of The New York Times (he is now associated with ESPN).
In Australia, aggregate polls haven’t tended to make as much of a splash. “There’s still far too much of this tendency for media sources to report their own houses polling as if it’s the be-all and end-all,” said Bonham. The exceptions to this ares The Australian Financial Review, which has for years been giving prominence to its Poll of Polls. The Guardian published MetaPoll’s first poll in March, though the pollster hasn’t signed with a media outlet.
The polling momentum this campaign appears to be with Labor. But Faruqi says he believes many of the current polls are underestimating the preference flows to the Coalition. Currently, MetaPoll predicts the Coalition will win the election 51.1% to 48.9%. Faruqi said this has been determined because he believes voter preferences to the Coalition will be higher than many predicted as voter preferences from 2013 are unreliable. There were a record low number of preferences to the Coalition in 2013, Faruqi says, due to the unpopularity of Tony Abbott as the Liberal leader, and preferences not flowing from Palmer United Party, which are now expected to return to the Coalition.
“We think that a decent chunk of voters, more independent and minor party voters, are going to preference the Coalition this time around.”
And before election day? Polling will impact the coverage. But Brent is sceptical of the power of polling in and of itself to change the result. This election, it appears the Labor Party has the momentum, at least as far as the polls are concerned. Will this make it easier for Bill Shorten?
“Imagine all the pollsters got together last election, decided to play a joke on us, and said Labor were comfortably ahead. It would have changed the reporting. Suddenly Wayne Swan would have been a sublime communicator, and Gillard the most popular politician since Bob Hawke. But they’d still have lost badly on election day.”
Perhaps the power of pollsters, or otherwise, is illustrated by what happened to Morgan in 2001. The pollster emphatically predicted an ALP victory — and was wrong. Shortly after, The Bulletin ended its contract with Morgan Polling. The pollster has never quite regained its influence, says Brent.
Pollsters can also damage their reputation in other ways. While political polling brings clout and coverage, most outfits make their money through private polling — commissioned polls by businesses, lobby groups or political parties aimed at finding the answer to something or selling a particular position. Some polls, like economic modelling, are commissioned just for the media coverage they’ll engender. Outfits with a barrow to push can ensure punters give the answers they want through loaded questions, which pollsters may be reluctant to push back on.
“Some polling outfits are more careful and will say, ‘Our reputation would not allow us to use this wording, this is beyond the pale’. But some are more easygoing about that kind of stuff,” Brent said. Polling is often not very transparent and can be easily manipulated and misused. Party political polling can play a similar purpose. While political parties will commission marginal seat polling, for example, to aid their decision-making, these polls can also be leaked for political advantage.
So sure, the pollsters are influential. In times of government, they unseat prime ministers. During this election, their finessing of the raw data will shape the coverage, and the tone of the campaign. But the ball is in the voters’ court — and they’re not paying as much attention as the politicians or journos, who study this stuff minutely. If a pollster gets it wrong, they may find the hit to its reputation hard to overcome.