There is one thing more certain than pollies kissing babies on the campaign trail: You are going to see a lot of opinion polling.
But not all polls are equal, and often results shift from week to week by only small amounts, well within the margin of error.
[tABLE: recent polls]If you want to know what the polls are saying, the best thing to do is look at the trendline, rather than any individual result.
We're making that easier for you.
The ABC is working with Professor Simon Jackman from The University of Sydney to produce an average of the national polls for this election, using what we know about their sample sizes and margins of error to also calculate a margin of error for the combined trendline.
This model is one that Professor Jackman has used in Australia for more than 15 years, which the ABC is implementing.
It averages the results of the five pollsters conducting national polls: YouGov Galaxy, Essential, Resolve, Ipsos and Roy Morgan.
"While any given poll might have a plus or minus of two or three points, once we start to combine that information we can get down to something much tighter," Professor Jackman says.
The model is also an average of the information we have today, and not a prediction of how people will vote on election day.
And, while we can see what the polls are collectively saying, this model says nothing about how accurate those polls actually are.
"This is no magic bullet … it's better than relying on any given poll, but there are real limits to it," Professor Jackman says.
Remember that these are not the ABC's polls, and we aren't making any predictions about the election outcome.
This is just an effort to interpret the polling that is being published.
The lines in the polling average chart show the trendline generated by the model, while the shaded regions represent the margins of error on this trend.
Labor has a significant two-party preferred lead in the polls
The margin of error varies, depending on how many polls have recently been published, and their sample sizes, but currently it is plus or minus 1.2 per cent.
This margin is less than any individual poll, but still represents significant uncertainty.
Today, the polls are collectively putting Labor ahead — in two-party preferred terms — with between 53.6 and 56 per cent of the vote.
The Coalition, on the other hand, is sitting between 44.0 and 46.4 per cent.
If that was the result on election night, it would be a complete landslide and by historical standards.
The gap was not always so large.
There weren't many polls just after the 2019 election but, according to the ones that were published, the government was ahead until around November 2019.
Labor pulled briefly ahead during the black summer bushfire disaster, before the Coalition regained ground in the polls from February 2020 through the first year of the pandemic.
They started to slip in the polls around August 2020, and that slip continued until March this year.
In the past month, its downward trendline has flattened.
Polls ain't polls
Not all pollsters are publishing their polls in identical formats.
The model uses two-party preferred figures from each poll. These are published directly by Newspoll, Roy Morgan and Ipsos, but the ABC is calculating a two-party preferred figure for two pollsters.
For Resolve polls, this is done by applying preference flows from the 2019 election to its first-preference vote estimates.
Essential is reporting undecided voters in its figures, and ABC is converting this into a traditional, two-party preferred figure by excluding them.
Labor pulled ahead of the Coalition on primary votes in December 2021 polls
[graph2]On first preferences, Labor's support in the polls overtook the Coalition's in December, 2021.
This — combined with the vote for the Greens, which will favour Labor on preferences — is the reason for the opposition's current commanding lead in two-party preferred terms.
That gap will shrink
However, the campaign proper has not even begun.
It is normal to see a tightening in the polls in the weeks leading up to election day, by as much as a few percentage points.
"If history is any guide, there will be a narrowing," Professor Jackman says.
"There's been a narrowing in every cycle that I've been doing this, over the last 15 years.
"If we get the typical recovery in the Coalition vote that we've seen … then I suspect that it's going to be a closer election than the polls are [showing] right now … but probably not quite big enough to get the Coalition back above 50 per cent."
Weren't the polls wrong last time?
In short, yes.
Producing this model requires some assumptions.
The biggest one is that, collectively, the polls are unbiased. That is that, on average, the polls are accurately measuring the electorate's view.
"That assumption was manifestly incorrect in 2019," Professor Jackman says.
"The biases that the industry had were large by historical standards. It was an unusual miss, historically speaking."
Pollsters have done considerable work to adjust their methodologies since 2019 to correct for that bias.
There have been encouraging signs in state elections, including recently in South Australia, where polling came in quite close to the result.
But we will not know for sure where things stand, federally, until election night.
"If [the bias correction has not worked], and we get that Coalition recovery, not even is 50-50 on the table but, perhaps, even a stronger result for the Coalition," Professor Jackman says.
"We've got to be completely up-front about that.
"This could be a much closer election than the polls are perhaps letting on, certainly this far out."
What about seat polls?
The polling average considers only the national polls, but you can expect to see a flurry of other polls throughout the campaign.
Many of these will be robopolls of single seats, where voters in electorates are dialled and asked their views by an automated voice.
Dr Jill Sheppard from the Australian National University says single-seat polls are heavily derided by academics and commentators, "and they deserve all of that scorn".
"When seat polls get discussed in the media, I constantly encourage people to, if not disregard them, at least be fairly cynical about them," Dr Sheppard says.
"The biggest problem with polling is being able to identify the population that you want to ask … and then being able to contact them.
"That's hard on a national level, but it's almost impossible on a state level."
Perhaps the most-interesting thing about seat polls is the person who commissioned them.
"The fact that they've commissioned it, they've seen the results, and then they've chosen to release it is pretty illustrative," Dr Sheppard says.
"We've seen, lately, a lot of independent candidates and candidates associated with the Climate 200 group release single-seat polls that probably overestimate the likelihood that they will be elected, but it puts them in the national conversation.
"It paints them in voters' minds as being a viable candidate, and that's probably exactly what that candidate wants us to think."
Casey Briggs will have more on opinion polling on Insiders on ABC from 9am, or on iview.
