Pollsters’ fingers crossed

THEY got it wrong in 2019, and Australia’s polling companies know their own reputations – and that of the industry itself – are on the line in Saturday’s federal election.

Australian pollsters uniformly predicted a victory for Bill Shorten at the last election, with a subsequent inquiry by the Association of Market and Social Research Organisations finding their errors were caused by over-representing politically engaged and better educated voters in their samples.

Three years on, many of the research companies say they have changed their approach and can be trusted to get it right, but with their polls predicting quite different outcomes, it’s clear not all of them are actually correct.

Dr Campbell White from Yougov said the company “went back to basics” after 2019, and getting it wrong then “provided us an opportunity to modernise the way Newspoll was done”.

“The poll was previously a mix of robopolls and online data, and we shifted the methodology to a purely online basis … and we are able to better represent the population through that method ” Dr White said.

Since then the company has published polls for the Queensland, WA and SA state elections, he said, and accuracy had improved each time.

Jessica Elgood from Ipsos said they had also changed methodology – their surveys were now based on 20 per cent telephone calls and 80 per cent online – and they’ve also changed how they “weight” their respondents.

Roy Morgan chief executive Michelle Levine said the company had not changed its polling methodology, but focused on “all of the things that tend to correlate” around how people feel about politics.

“When things like consumer confidence are high, you’ll generally find people are supportive of the government of the day,” she said. “If consumer confidence is low they tend to be a bit annoyed with the government of the day.” But there was a common misconception about what polling was.