Introduction and following the US
Political science is, at last, becoming a ‘proper’ science. Data analysis techniques are spreading throughout the political class as big data is used in this year’s General Election like never before.
No longer are those gaggles of people going door-to-door on a hit-and-miss mission – if there are party activists in your area, then you can bet that an analytical engine has told them that your constituency, your area and your street is of critical importance in the outcome of the election. Is this a new approach? Is it too cynical an approach? Only you can decide, but just know that whichever way you vote on May 7, some software has probably already predicted it.
"In previous elections it was the spin doctors and the media that influenced the election outcome," says Mark Morley, Director of Industry, OpenText. "Perhaps in 2015 it will be the data scientists who have the most influence."
Is this the UK’s first truly data-driven vote?
"Data has been used for decades to understand voters’ preferences and habits, though this is the first time political parties are using it in earnest to communicate," says Jed Mole, European Marketing Director at Acxiom.
Big data and data analytics have become much more mature since 2010. "Political parties have learnt that it is not just how you gather and archive information that counts, but how you use it to develop an action plan and strategy," notes Morley, who says it’s all about taking advantage of digital information to react strategically in near real-time.
Which analytical engines are being used?
Analytical engines are being used by all three main political parties; the Lib Dems use one called Contact Creator, Labour has Voter ID and the Tories use Merlin. They calculate the likelihood of voters choosing each party in every UK constituency, though the raw data behind all of them comes from the Mosiac database of UK demographics.
"Mosiac was one of the first segmentation classifications in the UK," says Mole, who explains that it’s stereotyping based on information. On its own it’s a really blunt tool. For instance, it will use someone’s postcode to determine the type of voter they are likely to be – where you live determines whether you’re communicated to as ‘Mondeo Man‘, or not. "Individuals, regardless of whether they are voters, are now far more complex and need to be understood and communicated to on an individual level," says Mole.
Mosaic is the base layer of information needed to make predictions about the voting population, but as we’ve become more diverse and unpredictable, a new layer of analytics is being used. It was pioneered, perhaps not surprisingly, in the USA.
Is this the ‘Obama-isation’ of politics?
"If we use the US 2012 presidential election as the yardage stick for a truly data-driven campaign, then it doesn’t seem like we’ve yet seen the same from a major UK campaign," says Sean Owen, Director of Data Science, at big data analytics firm Cloudera. "I believe this will be the first General Election when we can point to a decisive effect from analytics."
In the US 2012 presidential election the Obama campaign used analytics to centralise scraps of information from campaigns, and merge them with demographic databases to understand voters. "Analytics was used to precisely identify ‘swing’ voters who are receptive to a political message, and intelligently buy media time that precisely reaches them – i.e. many, cheap, focused TV slots, not simply a couple of expensive prime-time slots," says Owen.
However, the UK hasn’t reached that stage yet. "The UK’s political parties are nowhere near as advanced in their use of analytics platforms and databases compared to the US," says Mole. "The main UK parties will of course have access to proven and industry-leading analytical software such as SAS, SPSS, but the most important part of any insight-led analytics campaign is the data that goes into it."
"Imagine your analytical software as an engine; the data is the fuel that drives it, and it simply wouldn’t run without it.
Analytical engines and social media
How data-savvy are political parties?
Err, not very. They’re a long way behind retail, that’s for sure. "If high-street brands had five years to build a multi-channel campaign for one day’s worth of voting, we’d see a much greater level of spend in data-driven targeting," says Mole. "Political parties are typically richer in data compared to brands, but they don’t have big budgets or the expertise at their disposal to use this data effectively and creatively."
How are advanced analytical engines being used?
The use of analytics in electioneering isn’t that different to online advertising. "In online advertising we collect information about customers, understand what it is about customers that indicates they respond to your product, and reach them in the most targeted and cost effective way," says Owen. "That means modelling around why customers buy or don’t buy, leading to a great deal of data collection work, which then results in a lot of number crunching to calculate exactly which ad buys are most effective." It’s all about identifying swing voters.
Are voters really that predictable?
Apparently, yes. "Everyone is predictable to a certain degree, we can see that in data," says Mole. "Supporters of all the three main parties have distinct and unique traits of demographics and behaviour."
While individual voters are notoriously hard to pin down, aggregates and broader voting trends are surprisingly foreseeable. But is being able to predict voting behaviour on an unprecedented level really that desirable?
"By knowing voters better it will lead to better, more responsive government, and parties that would vie to embody our desires perfectly," says Owen, but he wonders whether the ultimate outcome is positive – do we want politics to be even more poll-driven? "We may find ourselves inexorably tempted to pander to the majority view," he says. Good data could have bad consequences.
Could the use of big data decide the results?
It definitely helped Obama win re-election in 2012. "Obama’s campaign was more finely targeted where it counted," says Owen, though he admits that the US electoral system has its own characteristics that make comparison with the UK difficult. "It exacerbated the tendency to focus on swing states and swing voters, which are unduly important in the US," he says. "In a way, big data amplified how elections are decided, which is a double-edged sword."
In the UK, data could still make all the difference. "It can certainly influence the results, and these days – with such tight margins and talk of a second coalition – it could help tip the balance to the right or left," says Mole. "It’s all about getting the right message to the right person."
What about social media?
We don’t vote on Facebook and Twitter – not yet – but it’s where many more of us will not only get our politics news in the run-up to the election, but where we’ll react to it, too. "Social media listening tools enable trends to be identified and analysed, allowing the messages to be promoted to amended as required," says Morley.
South Wales-based Blurrt create political Twitter Worms, Twitter Polls and Twitter Wordles based on real-time data from Twitter. "Twitter is an incredibly powerful platform for capturing people’s views on the political parties, their policies and leaders," says Blurrt’s Jason Smith. "We captured and analysed the Twitter reaction for both of the Clegg vs Farage EU debates and both Scottish Referendum debates, and we were successful in calling the result of all four debates within seconds of them ending."
Blurrt collects and analyses tweets in real-time for sentiment on a scale of -5 to 5, and the results are plotted out on a graph over time to create a Twitter Worm. "For this general election we’re collecting the tweets for all the main political parties and running a live leaderboard," says Smith. "The leaderboard ranks the parties based on Twitter reaction – volume and sentiment combined – and the number of tweets."
"In the last election, parties dabbled with social, but this time round, they’re engaging with it in a more mature way and using data to power meaningful conversations based on rich voter insights," says Mole.
Morley thinks that data will play an even bigger role at the next election. "Politics will become increasingly data-driven, and the party that best embraces a digital-first philosophy, leverages technologies such as big data and analytics, will certainly have a much better chance of success than relying on blunt tools or traditional voting indicators."