Epic fails in 2015 elections: Does polling jeopardize fairness?
2015 has been a year of electoral polling failure and it could have major implications on election outcomes, say QUT economics researchers.
The recent definitive Conservative win in the recent British elections, when polling had produced the near universal prediction of a 'hung parliament', mirrors Labor's surprise win in Queensland where the election outcome was unrelated to the poll prediction, says Professor Lionel Page from QUT's Queensland Behavioural Economics Group (QuBE).
"The magnitude of the failure to predict these elections' outcomes left pollsters gobsmacked," Professor Page said.
"But beyond the surprise, a QuBE study suggests that errors in polling forecasts can actually bias the electoral results themselves.
"In the Queensland election, the poll indicated the LNP would be returned, albeit with a reduced majority, whereas in the British election all the opinion polls predicted a hung parliament.
"In a third 2015 example, Israel's pollsters were completely disproved - the re-election of Benjamin Netanyahu in March went against all poll predictions of an opposition victory."
Professor Page said the QUT behaviour economics study of how exit polling influenced French elections, published in the European Economic Review (July 2015), could provide answers.
"We found that the precise exit polls released in mainland France in the 2002 presidential elections influenced voters casting their vote French territories in later time zones around the world," he said.
"When the exit polls were released at 8pm on election day in France, it was only 2pm in Guadalupe and 9am in Tahiti which meant these voters knew the result of the election in mainland France before they had voted.
"France then changed the law in 2007 so that voting in its territories closed one day before the mainland election. We compared voter behaviour in the following presidential election with the 2002 one and found two significants effects.
"Firstly, in the 2002 election the turnout in the territories was 11 percent higher than in the subsequent election.
"Secondly there was the 'bandwagon effect' - when voters knew who was going to win, the voting share of the expected winner was higher but disappeared after the change in the law."
"This indicates people are less likely to vote when they expect their candidate to lose."
Professor Page said, using this insight, it could be that the record turnout of British voters was due to a fear of a hung parliament suggested by most pre-election polls.
""People were afraid that if there wasn't a definitive outcome, a hung parliament would give the Scottish National Party the balance of power in the UK.
"In Queensland, voters were motivated by the polls to avoid a hung parliament but never expected the ousting of the Liberal Government."
Professor Page said poor opinion polls could sway the outcome of elections.
"Because our study found that polls do influence voter behaviour, some elections could, therefore, end up being decided by the effect of random polling error," he said.
"Even UK Prime Minister David Cameron's advisor Lynton Crosby acknowledges: 'polls have become part of the political process so they're not an independent measurement that says this is what's going on, they actually influence what's going on.'.
"Indeed, it is said that Crosby knew from private polling that the Conservatives were ahead but kept it quiet because he knew fear of a hung parliament would motivate voters to vote for the Tories.
"What can be done? In the connected world a ban on polls in the last stages of a campaign is not a solution. Our results show that improving polling techniques and possibly regulating polling practices is important if we do not want elections to potentially end up being the result of a dice throw (ie the direction of a polling error)."