Opinion polls in Australia, like in the USA and England, exercise a powerful influence upon our elected leaders and the media. The weekly and fortnightly polls shape the media landscape, as journalists breathlessly report on the horse-race: are they up this week or down?

And there’s no doubt that consistently poor opinion polls have been the catalyst for political party leadership change on both sides of parliament. Brendan “Mr 9%” Nelson was undone by the long streak of low polling numbers; Crean was toppled after he reached his low of 14% and Kim Beazley in 2006 was replaced by Rudd after his polling slumped to 18% (for preferred PM). Rudd himself was replaced by Gillard in 2010 with the parlous polling numbers and focus groups used as a primary reason for the spill:

On June 16, after the Queen’s birthday weekend, Shorten went to see Gillard in Canberra. A left-winger and pragmatist with the lawyer’s flair for detail rolled up in a highly-engaging personal style, the deputy prime minister heard Shorten out as he told her that he thought there was a real chance that Labor faced a wipeout. The implications were clear. Gillard had plenty to gain but she had plenty to lose, too. She held her counsel. Publicly, she continued to swat away rumours that she could challenge Rudd.

Party polling undertaken on June 16 and 17 in four key NSW marginal seats by long-term Labor pollster UMR highlighted the storm clouds. The four seats were the bellwether Eden-Monaro, Greenway, Hughes and Page. Across the four seats, Labor’s primary vote was 35 per cent against the Coalition on 47 per cent. The Coalition’s two-party preferred vote across these seats was 55 per cent, with a seven per cent swing against Labor. The polling showed 62 per cent of voters dissatisfied with the federal government. Opposition Leader Tony Abbott had closed the gap on the question of preferred prime minister – with Abbott on 37 per cent and Rudd on 42 per cent.

Now, these same polls have fed the 18 month media campaign of leadership speculation against Gillard.

A sharp drop in support for the ALP in February, which sent shockwaves through the party and fuelled leadership anxiety, is now showing signs of becoming entrenched with less than six months to go to the federal election on September 14.

The latest Age/Nielsen poll has confirmed Labor’s share of the primary vote is languishing at a landslide-losing 31 per cent – up a statistically insignificant 1 point from February.

This compares with support for the Coalition being unchanged on 47 per cent.

In fact, chief Rudd backer, Joel Fitzgibbon confirmed poor polling was the primary excuse behind the recent still-born leadership challenge against Gillard:

However Mr Fitzgibbon told Fairfax: “It would be silly to tell people watching your program that there is nothing going on.

“Obviously, internally people are looking at the polls and they are expressing concern about the future of the government and indeed the party and you’ll get conversations and those conversations are, unfortunately, making their way into the media. We should keep them internal.” Ms Gillard’s hold on the leadership could be bolstered if she manages to negotiate through her media reform package, which appears increasingly likely.

Polling is a ubiquitous element of political commentary and reporting. The polls attract disproportionate attention given that they numbers themselves don’t change much from week to week.

There is always a debate over the legitimacy of polling. Party partisans decry the polls as tainted, biased or just plain wrong.

In the USA, the Republicans obsessed about “skewed polls”, and even rewrote polls to be “unskewed”.

There were even articles by Republican bloggers arguing that skewed “pro-Obama” polls were a method of voter suppression against conservatives:

This year, instead of producing too many votes, pollsters are allegedly doing the opposite – making sure fewer people cast a ballot on Election Day. Teaming up with the media, pollsters are suppressing the vote by concocting phony results; by skewing the data. That drumbeat of results is supposedly designed to “depress Republican enthusiasm,” which in turn hands victories to the Democrats.

I think this issue is an interesting one, mainly because there is research that suggests that widely reported public opinion polling can influence the opinions of the public.

There is growing evidence since the 1980s that published opinion polls do exercise an influence on public opinion. This of course is rejected by most pollsters, who claim to be careful to avoid influencing the outcome of a poll through priming and order effects in the survey design.

However, the wide dissemination of poll results does have an impact.

Research suggests that polls influence people’s heuristics regarding politics and civil society. For people without much information about an issue or public figure, opinion polls help them quickly gauge what the “majority” feel (the bandwagon effect). For people who already have formed opinions, opinion polls often reinforce those beliefs (consistency principle).

This is not to say that a poll showing that party or leader X with a certain amount of support will automatically cause that party to receive the published support. We have countless examples where parties polled well in the lead up to an election, only to see modest or no gains on an election day. For example the Greens Party or Lib Dems during the last Australian and UK elections.

This line of reasoning is overemphasised because it focuses on the predictive power of polls.

The research I have linked to above by Vicki G. Morwitz and Carol Pluzinski suggests that prolonged exposure to polls can influence voters’ support for candidates or public figures. In some cases, this can result in behaviour changes — e.g. in actual support on election day.

This influence exists even without pollsters who write questions in a calculated manner, or who are beholden to a media narrative or news cycle. We should all be concerned about polling organisations in bed with major news outlets who use widely publicised polls to run campaigns or pursue vendettas — even if those polls are scientifically conducted and statistically rigourous.