US Election: What Can Trump’s Success Tell Us About Data?

Views, News & more

The biggest upset in US political history? Almost certainly.

Few, if any, thought he’d stand a chance. Donald Trump as president? No way! And yet, yesterday, there he stood as US president-elect, promising to “bind the wounds” of division and “come together as one united people” as a president “for all Americans”. Yes way.

Now, whichever side of the Trump fence you sit on (and you certainly must have to sit on a side), it certainly is an incredible victory. There is his complete lack of political experience for a start. His unusual claims (threatening to “cancel” the Paris Agreement on climate change, anyone?). Evidence of his past sexism and misogyny (that tape).

Somehow, positioned as an anti-establishment outsider, Trump weathered knock after knock and gaffe after gaffe to make his merry way to the White House in a surprising defeat over Hillary Clinton.

Now, if you’d read any of the political polls that have been doing the rounds these past months, you’d be forgiven for thinking you’d woken up on a different planet. Planet Trump.

But actually, perhaps we shouldn’t be too surprised at Trump’s ascension to the US presidency. Shock political results are becoming something of the norm: in the UK, the EU Brexit referendum result this summer took many by surprise. Back in 2015, many polls for the general election failed to point to the correct outcome - a majority Conservative government.

So why do some polls seem to get it so disastrously wrong?

Billionaire Trump’s victory hung largely on his success in key swing states including Florida, Ohio and Pennsylvania. He won more than 270 of the 538 electoral college votes - the number needed to secure the presidency. Of the estimated 129 million Americans who voted, who voted Trump?

Digging into the data, lots and lots of polls - the majority of polls - had Clinton as the clear winner. As the long campaign drew to its close, the respected polling website FiveThirtyEight gave Trump just a 30 per cent chance of victory. But, well, he trumped it.

When you look at it, a lot of it is down to factors that simply can’t be predicted or controlled. Writing in the Irish Times, Michael Marsh speculates that the polls got it wrong because they didn’t elicit properly random responses from samples of voters.

“Most [samples] are phone polls or some kind of on-line sample,” Marsh writes. “Face-to-face interviews are unusual. It is possible to adjust these samples so that they look like a random sample, but adjustments are limited by what is known about the sample and the underlying population.

“It is easy to ensure the right balance of gender, age, occupation and even past voting patterns and so on but this assumes that the subgroup actually in the sample accurately reflect those in the population.

“This was the big mistake in the UK in 2015. As it happens, the samples contained people more interested in politics than the actual population, and this factor led to a serious underestimate of the Tory vote. Something along these lines could be part of the explanation in the US.”

There is also the question of honesty. Some commentators think that some Trump voters were embarrassed to publicly admit they were pro-Trump, so they said Clinton instead.

“The very premise of polling is based on the idea that voters will be completely honest with total strangers,” Republican veteran Ned Ryun told Politico.

And there’s events, dear reader, events. A week is a long time in politics. Other observers have talked about the effect on voter intention the FBI had when it announced, less than a fortnight before election day, that it was reviewing fresh evidence around Clinton and her aides’ handling of sensitive information while at the State Department.

And there’s something else.

Trump and big data

There could be something more significant lying behind the pollsters’ failure to predict the correct outcome: the use of big data in political campaign planning and targeting. While blanket polls were asking voters straight yes or no questions, producing straight yes or no answers, Trump’s team had employed a big data firm to revolutionise the nominee’s campaign planning and strategy approaches.

Cambridge Analytica was employed to work on voter influencing, using data on millions of US adults to deliver specific, hyper-targeted communications, be that a social media post, TV spot, letter or something else, all nuanced to the individual and designed to influence their voting intention.

Gleaning data points like financial transactions, gym memberships, what people watch on TV, the company then turned these points into sophisticated messaging, with Cambridge chief executive Alexander Nix telling Sky News: "Today in the United States we have somewhere close to four or five thousand data points on every individual.

"The traditional model where 50 million people receive the same blanket advert is being replaced by extremely individualistic targeting.

"So we're able to identify clusters of people who care about a particular issue, pro-life or gun rights, and to then create an advert on that issue, and we can nuance the messaging of that advert according to how people see the world, according to their personalities."

So - are polls fundamentally unreliable, vulnerable as they are to subjectivity? Or was the Donald cleverly working his magic in the background all along?

Get in touch

 020 7847 4510

 info@hso.co.uk

Contact us

    • Head Office:
    • hSo, 50 Leman Street, London, E1 8HQ
    • Switchboard:
    • 020 7847 4500