Australia’s great 2019 election polling catastrophe wasn’t that big a deal, according to one of the most esteemed and experienced public opinion survey junkies in the world. Nate Silver, the one-time New York Times number cruncher who now runs the election focused website and news outlet FiveThirtyEight, dismissed the whole thing with a word. Herding.
Discussing some recent US Democratic Party primary polling last week on his podcast, Silver said pollsters had a duty to publish what they found in their surveys.
What went wrong in the lead-up the May election?
“If you’re dealing with dozens or hundreds of polls in an election campaign, you’re going to have some weird polls,” said Silver. “It is absolutely a pollster’s job to publish the data that it produces as a result of the methodology it did ahead of time. Pollsters should not be saying, ‘I like what that poll says, I’m afraid I might get teased on Twitter. I better not publish that poll’.”
Silver then used the Australian experience to illustrate a case where people said polls shouldn’t have been published because they came up with what turned out to be incorrect findings.
Get Crikey FREE to your inbox every weekday morning with the Crikey Worm.
“Labor was one or two points ahead in every poll — they all showed the exact same thing which should have showed ahead of time that there was herding which means pollster were trying to not stick out, they’re trying to not move [from the herd],” said Silver.
This will be no surprise to readers of Crikey. Bernard Keane pointed to herding in his analysis of 2019 election polling here last month which noted there was no consensus on whether herding was the driver of the bad use of polling in our election, but it did stand out as a very likely suspect.
Silver, using the kind of blunt assessment that allowed him to call plenty of recent US electoral contests and give Donald Trump the highest published chance of victory just prior to the 2016 presidential contest, seemed to have nailed it to his satisfaction.
The polls keep rolling on
Silver’s observation went to air just as Australia’s Newspoll was back in the field for its third post-election survey, the results of which were published today.
On the face of it, not much has changed with Newspoll. They now give a published outline of who was surveyed and how. This week’s poll was of 1661 voters across city and country regions from September 5 to September 7 and included 956 online surveys and 705 robo-interviews. There was a margin of error plus or minus of 2.4%, giving a confidence level of about 95%. We’d like to assume the rigour of the methodology has been amped up.
Newspoll remains a trusted and mostly reliable source of longitudinal political opinion even though it’s gone through ownership and management changes which can affect how things are done.
Today’s poll shows not much has changed since the previous published survey published three weeks ago. The primary vote for both major parties increased one point, well within the margin or error and therefore statistically insignificant — this small concurrent change meant the calculated two-party-preferred vote remained the same.
What did change and was statistically very significant was voter satisfaction with Labor leader Anthony Albanese. On this measure he collapsed from a positive seven points to minus five — a turnaround of 12 points and almost five times the margin of error. This is known in the industry as a slam dunk.
What does this mean for Albo?
Albanese’s dive off the popularity cliff was matched by a rebound of a similar size in Prime Minister Scott Morrison’s rating — he’s back to a +10 points after having crashed from -15 to 6% up previously.
Of course, these surveys are being conducted years before the next election but the warning light should be flashing at least amber for Labor because right now sentiment about Albanese is in the baking phase.
Given overall support for Labor stayed stable — while the ALP was having disasters of various sizes in New South Wales, Queensland and Western Australia — the collapse of support for Albanese should be even more worrying for the party.
The likely explanation is Albanese did something that jarred with voters and here we can speculate with some surety. When the NSW politician became leader he avowed a change of heart on border protection and asylum seeker policy — saying he had made mistakes when supporting the push to water down Labor’s approach in the past.
Last week he looked like he didn’t mean those words, instead hitching his support to the cause of the Sri Lankan Tamil family the Morrison government is trying to deport because they don’t qualify as asylum seekers.
Regardless of the merits of the Sri Lankan Tamil family’s case — still to be tested in the courts — Albanese looked like was saying one thing and doing another. That’s a character matter and he ended up on the wrong side, as Newspoll demonstrated with next to no chance of error.
Do you still trust the polls? Send your comments to [email protected]. Please include your full name for publication.
Dennis Atkins is a freelance writer based in Brisbane where he was a national political editor during the Howard government. He is filling in for part of the time while Bernard Keane is enjoying a break.