Ahead of its first big test since the great debacle of 2019, there are two reasons why attitudes towards Australian opinion polls should be revised from outright dismissiveness to a healthier scepticism that should have prevailed all along.
First and foremost is the track record that international market research giant YouGov has accumulated over three state elections since it took over from The Australian’s venerable Newspoll series in late 2019. Election eve polls in Queensland, Western Australia and South Australia were all broadly on the money, allowing for a margin of error that should be reckoned at about 3%.
The biggest errors were underestimates of the Labor primary vote by 2.6% in Queensland and 2.9% in WA and an overestimate of the Liberals by 2.3% in South Australia last month. If that doesn’t quite match the precision that caused mystical properties to be attributed to Newspoll before the wheels first started to wobble at the Victorian election in 2018, it is, so far at least, a record that any international pollster might envy.
The second positive sign is that there is little sense of the herding that rang alarm bells in 2019 even before the result laid bare the fact that the entire industry had followed Newspoll off a cliff.
All five polls of national voting intention in the last week of the 2019 campaign credited Labor with leads of 51-49 or 51.5-48.5, the latter being an exact reversal of the actual result. As shown in the charts below, comparing trend measures of the various pollsters’ Coalition two-party-preferred results over this term and the last, polls have behaved over the current term as polls should, with different methods revealing different biases.
While one might look askance at the recent form of the Nine newspapers’ Resolve Strategic series — which had some anomalously Coalition-friendly polls late last year but hasn’t done so since — the post-budget period has produced a fairly healthy spread from 53-47 in the weekend Newspoll to 57-43 in the latest Roy Morgan.
With that established, we can take a deeper look at what the pollsters are telling us with confidence that they are, however imperfectly, measuring something real.
All agree that a paradigm shift has occurred in WA, where Labor appears set to win the two-party-preferred vote for the first time since 1987 off a potentially double-digit swing. Knowing this is likely to cost them three seats and their majority, the Liberals have invested hope in a glut of Labor-held seats on threadbare margins in NSW, although you may not have known it from the way the party handled its preselections there.
However, none of the pollsters are offering much encouragement for this, recording Labor swings in the state of at least 4%.
If the swing off Labor’s high base in Victoria appears weaker, the consensus is that it should at least be sufficient to gain it the Melbourne seat of Chisholm, while keeping the wolf from the door in its own marginals.
Labor’s state election triumph in SA also appears to have fed into federal polling from that state, which doesn’t bode well for the Liberals in the marginal Adelaide seat of Boothby.
As ever, the great enigma is Queensland, which was ground zero of the polling failure in 2019 when polls underestimated the Coalition’s two-party vote there by 8%.
Home to four Coalition-held seats on margins of less than 5%, it’s the state where pollsters are offering the widest spread of results. Whereas a recent aggregated quarterly result from Newspoll had the Coalition leading 54-46, for a relatively modest Labor swing of 4%, the debut poll from Ipsos for The Australian Financial Review last week had it the other way around, allowing for a smaller sample and a 4.4% margin of error.
That at least was the state of play as the campaign began, with the Coalition’s hopes resting either on a second polling failure or a campaign dominated by strokes of fortune — like the one that happened to land in its lap on Monday.
Will you ever trust polls again? Let us know your thoughts by writing to letters@crikey.com.au. Please include your full name to be considered for publication. We reserve the right to edit for length and clarity.
