This article is part of our special report Big data meets politics.
US politicians are the acknowledged forerunners when it comes to using digital technologies in election campaigns. But Europeans are making strides in their attempt to catch up, with the 2014 EU election providing a testing ground for big data analysts.
The Howard Dean campaign for the 2004 US presidential election is considered as a kind of year zero of data analytics applied to political campaigning. Although he lost the Democratic party nomination to John Kerry, Dean is remembered for pioneering internet-based fundraising and grassroots mobilisation, which inspired others after him.
Echoes of the Dean campaign are still being heard today, but with an added twist called big data analytics. Political campaigners have long sought to segment their electorate according to income, ethnic origin, age group or any other socio-demographic factor that can play a role in an election.
This ability to identify and target specific parts of the population was taken to a new level during the 2012 US election. And it was made possible by big data analytics.
“The Obama 2012 campaign used data analytics and the experimental method to assemble a winning coalition vote by vote,” writes Sasha Issenberg, author of The Victory Lab: The Secret Science of Winning Campaigns.
“In doing so, it overturned the long dominance of TV advertising in US politics and created something new in the world: a national campaign run like a local ward election, where the interests of individual voters were known and addressed,” Issenberg wrote in the MIT Technology Review.
Lessons from US elections are usually quick to cross the Atlantic. In Brussels, the digital methods of the Obama campaign provided inspiration for the Party of European Socialists (PES) during the last EU Parliament elections in 2014.
Brian Synnott was press and web communications coordinator for Martin Schulz, the PES frontrunner in 2014. He says big data analytics offered a whole new way of identifying and mobilising voters.
“Unlike a national election where you’re fighting for floating voters, you’re looking at people who would naturally agree with your line but need to be mobilised to just go out and vote,” Synnott recalls.
‘Knock the vote’ campaign
The Schulz team had done their math before the campaign, based on turnout figures, to measure the size of their electorate. Whereas national elections register between 65 and 80% turnout, European elections tend to score 43 to 50% at best, Synnott said. “So that opened up a much higher percentage of the electorate that would be your core base,” he told EurActiv.com.
Based on these calculations, the PES launched what they called the ‘knock the vote’ campaign. “‘Knock the vote’ was about how you digitally go knocking on people’s doors – not just to encourage them to go and vote but also to campaign and mobilise,” Synnott said.
“A lot of the practical steps involved Facebook,” he said, explaining that “Facebook is excellent if calibrated towards grassroots mobilisation”. Twitter, on the other hand, is better suited for sending out information to targeted audiences.
“Put simply, you could say: Facebook for mobilisation, Twitter for information,” Synnott said.
But how do you identify those potential voters?
One big difference between Europe and the US is access to voting records. In the US, campaigners can consult registers of previous elections, allowing them to identify neighbourhoods – individual houses, and people – who are likely to back their candidate.
These records contain information such as age, gender, and income, but also participation in previous elections, including voting preferences. For some individuals, these records can be extremely detailed. Digging through this data allows predicting the probability that individuals will vote for a candidate and target undecided voters in the hope of turning around an election.
“This is really big data: a table with one million people, and each of them with 7,000 entries,” says a Brussels insider with knowledge of digital campaigning tools and methods.
Matching these voter records with profiles developed by data brokers allows campaigners to grow a database of potential voters who can then be approached in several ways: via social media, on the phone, or in person, knocking on their doors.
“If you have information about how old they are, how much they earn and where they live, then you look for people who are similar and approach them,” the Brussels insider explains.
In Europe, most countries don’t allow access to voter records, so campaigners look for their data elsewhere.
It can be obtained from electoral polls, but also from “surveys, interviews, door to door, focus groups, applications in social networks, web statistics, data obtained by local campaign staff, etc.” says Xavier Peytibi, a political consultant who has run various campaigns for presidential, regional, and local elections in Spain and Latin America.
However, this can prove legally borderline, for example when individuals are identified by website “cookies” tracking their online activity, sometimes without them realising.
“Data can be misused in many different ways,” warns Aurélie Valtat, an official in charge of digital strategy at the European Commission’s development directorate (DG Devco), which manages multi-million international assistance programmes in the world’s poorest countries.
“It can be misused because it hasn’t been captured in an honest and legal way,” she told a EurActiv event on big data mining, held in March. “It can also be misused because the people supplying the data are not representative of the population that you’re trying to capture,” she said. Another challenge, Valtat pointed out, is whether the private records of individuals can be stored legally or not, and for how long.
Diego Naranjo, a privacy campaigner for the European Digital Rights initiative (EDRi), is more alarmist. “Data brokers have our personal data and they are already trafficking with it,” he told the EurActiv event, pointing to what he calls “pregnant data” stored on databases and ready to be sold without consideration for privacy or fundamental rights.
For political campaigners desperately looking to expand their electorate, the temptation to buy dubious databases can be difficult to resist.
But surprisingly, a lot of the personal information is also provided voluntarily by grassroots campaigners or party sympathisers on social media.
“We got data from Facebook but also from a tool called NationBuilder, which allows people to sign up for the campaign,” Synnott told EurActiv. “And you have the possibility of building an electoral picture based on that.”
The trick is to use those verified – and legally obtained – personal records as a basis for growing a database of people with similar socio-demographic attributes. “By building a grassroots campaign through a tool like NationBuilder, you’d be able to build up your profiles that way and then apply them on different platforms,” Synnott explains.
From data to persuasion
For some, data has turned into an obsession, if not an addiction.
“Big data, today, is an essential tool for an election campaign,” says Xavier Peytibi. “In any city, in any election, it is essential to meet your voters, and it is also essential in order to get your message across to impact them in a more direct way,” he told EurActiv.
“Data is information and information is power to discern what message by a party or candidate can get closer to a supporter or a voter, either to mobilise him or to get him to vote,” he said.
The hunger for data by political campaigns has fostered a growing industry of analytics start-ups and political consultancies. One of them is Liegey Muller Pons, a Paris-based firm, which claims to be the first campaign technology start-up in Europe.
As you would expect, the firm draws much of its inspiration from field trips it made to the US. “The most recent political campaigns [in the US] were targeting mainly two groups: the democrat ‘abstentionist’ and the hesitant ‘participationist’,” Liegey Muller Pons wrote in a blog post reflecting on the 2012 Presidential election.
“Today, the latest innovations are aimed at improving persuasion models to identify even more precisely what fractions of the electorate can change their minds after a campaign contact and what messages and arguments are able to produce that change,” the firm wrote.
Liegey Muller Pons is close to the Socialist family and advised the PES during the 2014 EU election campaign. It currently counts French Economy Minister Emmanuel Macron among its most prestigious clients.
But the centre-right European People’s Party (EPP) is equally keen on digital tools.
Juncker campaign goes digital
During the 2014 election, the campaign team of EPP candidate Jean-Claude Juncker – now President of the European Commission—made heavy use of Twitter, Facebook, Instagram and Flickr to spread its message.
“The Juncker campaign had a team of 25 e-campaigners who sat in the ‘war room’ of the EPP headquarters and oversaw the campaign’s numerous Facebook pages and Twitter,” recalls Juncker’s spokesperson Natasha Bertaud, who was among the key figures in the Juncker team.
“Contrary to the other campaigns, the EPP had Twitter accounts and Facebook pages for every EU member state we were campaigning in,” allowing targeting local audiences in their native language, Bertaud told EurActiv.
The multilingual dimension of the Juncker campaign allowed “generating the most ‘buzz’ around the EPP candidate,” Bertaud claims. Within weeks, Juncker’s Twitter account had already gained 20,000 followers, she said.
Limits of big data
But Bertaud is also quick to remind the limits of digital campaigning, especially in a European election context, where interest is lower than for national elections.
“You have to be realistic about what you can expect in terms of reach – this was not a US presidential election and Juncker was never going to rival Justin Bieber for followers,” she said.
Social media buzz is indeed no accurate proxy to predict who will win an election, according to Wired, America’s longest-running tech culture magazine. Still, online activity “does give us a previously unseen window into the types of issues Americans care about most,” the magazine writes.
“It’s trends,” Valtat summarised. “Big data mining can give you information about trends in elections, trends in public opinion and what it means for the people involved. But it’s not a reflection of the real world – it’s just an image through social media”.
Synnott agrees, and cautions about focussing election campaigns too heavily on digital and big data.
“‘Digital’ is not something that exists in isolation,” he said. “It’s a tool that’s part of an overall strategy. So you have to combine it with grassroots, with face-to-face, in a way that people can relate to it”.
He is also keen to assuage concerns by conspiracy theorists who see big data as a dangerous tool for government surveillance or electoral brainwashing.
“Some people get scared and see this just as big data. But it’s also genuine engagement with real people who can get a sense of their own stakes in a political campaign,” Synnott said.
“Most importantly, I think there was a template established that I think could be used in 2019.”