Tweets were more accurate than polls in predicting the Australian election result, an expert says.

Professor Bela Stantic, director of the Big Data and Smart Analysis Lab at Griffith University in Queensland, says an analysis of more than 2 million tweets on Twitter showed that the Liberal-National Party Coalition would win Saturday's election despite all other polls tipping a Labor win.

Stantic's analysis also correctly predicted Brexit and Donald Trump's win in 2016.

He told Mike Hosking on Newstalk ZB today that tweets provided far more data than traditional polls based on calls to about 1000 landlines.

Advertisement

"They are providing huge amounts of data - particularly in this case I collected in the last week about two million tweets," he said.

He said the trend in the past two weeks was "very clear".

Asked by Hosking about whether his analysis could be undermined by increasing numbers of early voters, Stantic said his analysis was now taking that into account.

He told the Daily Mail that it appeared that "people are more honest when talking to friends and social media than answering polls".

"Such analytics can provide much more accurate information than telephone polling, especially in a day and age where people have caller ID and don't have landlines," he told the British newspaper.

"The amount of data that all of us generate is truly staggering, and it is continuing to grow. This publicly available data is secret treasure of information if we know how to discover it."

Leading opinion polls had consistently put Labor ahead of the Coalition for two years before the election.

Professor Bela Stantic used Twitter to correctly predict Brexit, Trump's win and Scott Morrison's upset victory. Photo / Griffith University website
Professor Bela Stantic used Twitter to correctly predict Brexit, Trump's win and Scott Morrison's upset victory. Photo / Griffith University website

One poll, Newspoll, had accurately predicted the winners of every Australian state and federal election since 1985 until Saturday's upset Coalition victory.

But Sydney University political scientist Stewart Jackson said the polls were too consistent for too long to be credible.

"That indicates 'herding,' where the pollsters themselves are getting results that they don't think are right and are adjusting them," Jackson said. "Because statistically, polls should never come up like that."