- BRW Lists
Published 14 February 2013 00:50, Updated 10 April 2013 09:43
Barack Obama’s presidential win in the US has but big data firmly on the agenda for Australia’s 2013 federal poll. But the last thing the major Australian political parties would want to do is give voters the impression that Big Brother is watching them, says Hill+Knowlton’s Simeon Duncan. Illustration: Karl Hilzinger
Companies wanting to put big data strategies into action in 2013 could find the federal election makes the battle for technical talent even tougher than usual.
While the exact campaign strategies coming up in this election remain closely guarded secrets, observers believe both the main political parties are looking to borrow social media and data analysis techniques from the most recent presidential race in the United States.
The sophisticated data strategies used by the Democrats in particular are credited with the fact the campaign for Barack Obama’s re-election raised $1 billion and ultimately ensured his re-election. The Republican Party also had a big data strategy, dubbed Project Orca, although it was not nearly as successful.
Customer experience consultancy Fifth Quadrant director Catriona Wallace, who specialises in big data, says the recruitment process for the Australian election would have already started, but the details are unlikely to be made public until after the fact.
They’ll be targeting reasonably high-profile people in Australia, and the parties will be looking to bring in international consultants, too.
“This is now a competitive advantage because they’re building strategy, so they wouldn’t want anything of this leaked,” Wallace says.
“I expect the process is well under way, if it’s not already done. They’ll be targeting reasonably high-profile people in Australia, and the parties will be looking to bring in international consultants, too.”
The Obama campaign was able to handpick experts from Silicon Valley, such as Teddy Goff, Harper Reed and Dan Wagner, and had a total of 300 people in its digital, technology and data analysis teams. This enabled them to effectively use big data – pulling data from different sources into one system – to target people at an individual level via email, phone or advertising.
“They had a single system and could merge information from pollsters, fund-raisers, field workers and consumer databases, as well as social media and mobile contacts,” Wallace says. “Data strategies allowed analysts to predict which type of people are most able to be persuaded or to persuade others.”
She says all the main components of a big data strategy could be done just as easily in Australia, despite a population about 15 times smaller, although it would have to be done centrally by the parties rather than by MPs in individual electorates.
However, any use of big data in the Australian election would be on a smaller scale to the US since neither political party has a budget anything like that of the Obama campaign, nor of Republican opponent Mitt Romney.
Wallace also suggests some cultural differences would come into play because Australian voters tend to be less vocal about their political beliefs online, so the data would be more about demographics than voter intentions.
She says Australians are likely to be surprised by the level of targeting in this election but adds that they might also react more negatively than Americans if it were perceived as an invasion of privacy.
A former Liberal Party staffer who left politics at the end of 2011 agrees that public sensitivity would be a big concern. Simeon Duncan, now director of public affairs at Hill+Knowlton in Canberra, says both parties would be keen to avoid being “too intrusive”.
“The last thing they would want to do is give [voters] the impression that Big Brother is watching them,” he says.
The main database used by the Liberal Party is called Feedback; the Labor Party has a similar software system called Electrac.
Peter van Onselen, a former Liberal Party staff member who completed his doctorate on the topic in 2005, says the main component is public data from the electoral roll augmented by information saved when an elector deals with the MP’s office.
He expects, but doesn’t know, that the systems would be greatly improved since then, with greater use of information from commercial lists and social media.
During the US election, the Nate Silver phenomenon came to the fore. Silver predicted the outcome of the election with uncanny accuracy simply by aggregating data from all the polls and crunching the numbers.
However, political blogger Peter Brent of Mumble Blog says there are statistical limitations to what can be done in Australia compared with the US. He says a survey with a sample size of between 1000 and 2000 is about as strong in Australia, with a population of 22 million, as it is in the US. Pollsters in the US can concentrate on 10 swing states since most states are winner-takes-all for the electoral college votes.
“The equivalent would be to poll every electorate that might arguably be in contention and if you’re doing 60 to 70 seats with a statistically valid sample size of 1000 people in each, not just once but again and again, it would cost an absolute fortune,” Brent says. He also suggests the US has many more polls than Australia, not just 15 times as many but a larger factor.
This has implications not just for the reliability of polls and the potential success of a Silver-like collation of data, but also the detail of any big data strategy in play by the political parties.
But Roy Morgan managing director Michele Levine contests this. She says the polling data in Australia is just as good as the US and that there are possibly more polls in Australia, when you include those conducted online.
Levine says the job is easier in Australia because it is compulsory to be on the electoral roll and to vote.
“The big deal that you have to get right in America is predicting who will vote and what the voter turnout is going to be,” she says. “What you do is poll on a national scale – it’s really unusual that you would get a swing to the left and the key marginal seats would swing to the right or vice versa.”
She says if you need to drill down to the level of a single electorate, a sample size of 250 is enough.
Levine says governments in Australia are good at using poll data to get detailed information to inform policy on social issues such as unemployment or drug addiction, but don’t tend to be as sophisticated when it comes to using poll data in campaigning. She says organisations that do it well include Virgin Australia and Tourism Victoria.
Levine expects both parties will be engaging with social media a lot more during this election. “It will be interesting to see how social media kicks in this time – but you can go quite crazy trying to see what is being said about you and thinking about whether to respond or not . . . so I hope they don’t get too distracted,” she says.
Labor is certainly ahead in social media. Julia Gillard has 154,000 likes on Facebook and 341,500 followers on Twitter; Tony Abbott has 29,000 likes and 99,100 Twitter followers.
Unfortunately for Gillard, it might not make a great deal of difference. “It’s preaching to the converted,” says Duncan of Hill+Knowlton.