How Political Polls Work And Which Ones Actually Matter
With election season in full swing, we're being assaulted with numerous poll numbers -- which one should you pay attention to?
The two best-known polls, Ipsos and The Australian's Newspoll, survey 1600 and 1200 people respectively for their regularly published surveys.
Each company has its own methods of choosing respondents and administering their questions, and back their results as an accurate snapshot of the nation's mood -- but there are other polls and numbers out there which aren't as useful.
Newspoll: Online surveys and automated 'robocalls'
"Newspoll has the benefit of regularity, giving the pulse of the nation, generally on a fortnightly basis," David Briggs from You Gov Galaxy, which administers the Newspoll, told 10 daily.
He said Newspoll uses a mixture of online polling and 'robocalls' to landlines and mobiles to get results.
"We need a good cross-section of the population. Online gives you that, as people can complete surveys when it's convenient for them."
He said Newspoll has never called an election wrong.
Ipsos: Human beings dial randomly-generated numbers
Ipsos director Jessica Elgood said her poll was "a higher quality" than others due to their methodology of contacting respondents.
Ipsos checks the demographic status of each person after the call, and keeps calling until they get a representative sample.
"If you pick up the phone, and it's a robocall, a lot of people hang up straight away," Elgood told 10 daily.
"So the mix of people [with robocalling] is not typical, despite the aim being to get the most typical bunch of Australians as possible."
How similar are they?
The two companies often have nearly identical results, but they can occasionally produce something different to one another.
The latest Newspoll has Labor leading the two-party preferred over the Coalition 52-48, while Ipsos has 53-47.
But back in November, Ipsos had Labor at 52, with Newspoll at 55.
Newspoll and Ipsos both say they have a margin of error of roughly three percent -- meaning that, if a party's numbers change just one or two points between surveys, that should be taken with a grain of salt.
"We love to look at polls to see a few points difference, but reality is that it's often within the margin of error," Elgood said.
Polling Ins and Outs
William Bowe, editor of Australian electoral studies blog The Poll Bludger, explained the basic principals of polling:
"If you get about 2000 people out of even a very large population, you can get a margin of error of about two percent," Bowe said.
"With the basic mathematical theory of random sampling, then there's a 95 percent chance you'll get a result within two percent of the actual public opinion."
The polls must be representative of the demographics of Australia -- age, gender, socioeconomic status, location -- so that the results can be extrapolated.
"A real opinion poll must make a conscientious effort that represents a random sampling of the population," Bowe said.
The Polls To Take With A Grain Of Salt
Bowe said polls on social media are the ones to definitely take with a grain of salt. He also advised to largely ignore any polls commissioned by an interest group on specific topics, or polls on social media.
"Seat polls don't perform as well as national polls," Bowe said.
With lots of polling still to come in the next four weeks of the campaign, it pays to check the source of the poll, to look at the size of the sample, and look into what methodology has been used.
"I would look at any polls commissioned by an interest group, and ignore it," Bowe said.
"Questions will be asked to load into a press release to turn into a news story favourable to their point of view. I wouldn't pay attention to polls commissioned by a group with skin in the game."
Life Beyond Polls
Bowe, despite his work focusing on polling, said there is sometimes far too much focus given to them. He said politicians pay too much attention to poll and "lose their nerve in the face of them".
"We'd probably be healthier if we paid less attention and regarded polls as mildly interesting... governments need to behave as if there isn't an election on the weekend. They shouldn't be taken all that seriously."