Voter volatility behind shock election result

3 September 2019

Research from The Australian National University (ANU) has shed new light on how voter volatility was behind the Coalition's shock win in the 2019 Federal Election.

As part of the ANUpoll series of surveys, Associate Professor Nicholas Biddle examined voter volatility - or changes in who Australians said they would vote for in the lead-up to the May election, and who they actually voted for on Election Day.

He found more than a quarter of those surveyed ended up voting for a different party than the one they'd indicated.

According to Associate Professor Biddle the data is the first time voter volatility in an Australian election has been measured.

"We saw a swing towards the Coalition during the election campaign that came mainly from those who'd intended to vote for minor parties, or didn't know who they would vote for. Obviously, this swing was not picked up by the polls," Associate Professor Biddle said.

"The 'other' and 'don't know' voters appear to have had a big impact on the May election," he said.

"Data like this is far and away the most powerful tool for understanding voter volatility and changes in voting intention and helps explain why the polls got it so wrong. In the past, we haven't been able to measure voter volatility as a key driver of results. Clearly this election results shows that we need to."

The data revealed some of the characteristics of those who did change their vote, and why they did so.

"Females, the relatively young, and those in more disadvantaged neighbourhoods were most likely to change their votes," Associate Professor Biddle said.

"The most common reason given was their views on the local candidate had changed."

Associate Professor Biddle said this also provides a fascinating insight into those who swung towards the major parties.

"We can learn a lot about the result and the failure of polls to predict election outcomes by looking at those who didn't intend to vote for the Coalition, but ended up doing so," Associate Professor Biddle said.

"These people were more likely to be female, at the upper end of the age distribution, non-Indigenous, without a university education, and living outside the most disadvantage areas in Australia."

While some of these characteristics align with traditional models of voting behaviour, others, such as low education, were more of a surprise.

"These individuals also tended to be less supportive of population growth. And despite the popular narrative, they were no more likely to support populist views," Associate Professor Biddle said.

"Interestingly, those swinging towards Labor tended to be the least risk averse. If Labor had been able to convince a slightly larger percentage of those who were relatively risk averse to change their vote to Labor, the election outcome could've been quite different." 

Associate Professor Biddle said these results prove who a person says they'll vote for on a particular day is a far from perfect predictor of who they'll end up voting for. This has clear implications for polling in Australia.

"More care and transparency about how these groups are treated should be a real focus of any adjustments to polling in Australia," he said.

"In the end, polls are just sample surveys, often undertaken on a small percentage of the population. There's a considerable skill and science around turning that into something more meaningful. Election polls tend to get pretty close, but like any survey, are prone to errors."

The results from ANUpoll have been published as a working paper.

The full report is available here: