One way or another, Facebook’s money – all $789bn of it – comes from the idea that, by collecting information about all the things to which people pay attention, the social network can prove that advertising works. The same is true of Google, which is valued at $1trn. The business model of both companies is predicated on the idea that because they control what you see, they can influence the decisions you make. Between them, these two companies have created an economy of attention that pays for more than 80 per cent of all media.
When this information is used to influence a choice between consumer goods, it is mostly uncontroversial and in many cases useful. But the idea that these platforms could swing elections is contrary to our notions of political free will in a democracy. To counter this, people such as Nick Clegg, the former UK deputy prime minister and now Facebook’s vice-president of global affairs, are employed to give the impression that while Facebook has the power to move voters – a power that by February of this year it had already hired out for almost $800m in the 2019/20 political cycle – it is also Facebook’s deepest concern that democracy is protected.
At first, it appears that only one of these things can be true. Either Facebook controls a power that is profoundly undemocratic, or it doesn’t, and its users think what they like. But the 2020 US presidential election is the best example yet of the central truth of the attention economy – that while Google and Facebook do truly wield great influence, they have no control over it. It is directed by the rules of the system itself, and these rules favour extremes.
The fact that social media and search engines benefit a more polarising candidate is obvious from the number of interactions generated in the 2020 campaign. Over the past month, Donald Trump has averaged an 86 per cent “share of voice” among presidential and vice-presidential candidates on Facebook, and a 62 per cent share of interactions on Instagram.
The same is true of the favourable media that surround each candidate. The four most popular pages for political interactions on Facebook in the US over the last 30 days are, in order: Trump, Fox News, Dan Bongino (a pro-Trump activist) and Breitbart.
Every day, a Twitter account called Facebook’s Top 10 lists the ten most popular links posted on the site in the US. For months, this list has been populated by Fox News, Breitbart, pro-Trump commentators such as Sean Hannity, and Trump himself. On 23 October, eight of the most popular social posts in the US were shared by Dan Bongino.
Dr Jonathan Bright, a senior research fellow at the Oxford Internet Institute, is studying what he calls “junk news” during the 2020 election. This is not fake news, he told me, but news that has, like Trump, “freed itself from the constraints of the truth”. A junk message “can be written in a way to make it as engaging as possible”, conferring an inherent advantage: “if it’s not doing better than mainstream news, then they’re doing it wrong”.
[see also: Leader: Why Trump and his clan must go]
Polarising and divisive content is reliably effective, says Bright. “It’s one thing that generates engagement and attention on the internet. And we know that simply because people make money out of creating it.”
Controversy also extends the “reach” of a message. In a study published in June, an analysis of 17 million posts on Reddit found that more subjective content, posted by authors who scored higher for polarity of opinion, travelled faster and to more users.
Google’s data on how often people search for each candidate shows people also seek out Trump’s polarising content more than they do information about his opponent. Over the last year, there have been more than four times as many Google searches in the US for Trump’s name as for Joe Biden’s.
This clarifies something about how people use Google. The search giant’s mission statement – “to organise the world’s information and make it universally accessible and useful” – describes Google as a library, but the trends data from the 2020 election suggest it works more like a tabloid newspaper. People are not Googling Biden to find out how his policies differ from the incumbent president; they are Googling Trump’s latest outrage, for entertainment.
They are also, indirectly, paying the media to give more coverage to Trump. The outrage candidate, as a more popular search term, rewards any website that publishes articles about them – good or bad – with more traffic. In the free-to-use, ad-supported model that pays for almost everything Americans read, watch or listen to, giving more airtime to Trump directly results in more money.
In 2016, this was reflected in the wider broadcast and print media, which gave Trump 15 per cent more coverage than Hillary Clinton throughout the election. Polling appears to confirm that he enjoys the same “earned media” advantage in 2020.
The attention economy encourages polarisation not only because of the tendencies of its users, but because of the way it generates revenue from them. Google and Facebook have risen to command 70 per cent of the US digital advertising market by offering advertising that is targeted at specific users. Both companies have spent vast sums on building systems to find and advertise to the people most likely to respond to that advert, and to advertise to them in a way that is personal to them. When it’s used to reinforce a person’s desire to try a new brand of breakfast cereal, this is good business. When it is used to reinforce how a person feels about issues such as free speech or immigration, there is evidence that it creates stronger, more partisan political views.
Last December, researchers from Northeastern University copied real adverts from Trump and Bernie Sanders, and paid Facebook to target them at a selected group of tens of thousands of real people. Their findings were that Facebook’s delivery algorithms had made it more difficult – in some cases impossible – to reach a conservative audience with a liberal advert, or vice versa. “Facebook limits political advertisers’ ability to reach audiences that do not share those advertisers’ political views,” they concluded.
But this does not mean that targeted advertising is useless for politics. “Communicating with your partisans drives turnout”, explains Bright. “Even though these people weren’t going to vote for someone else, you want to make sure that they turn out. In increasingly polarised electoral systems, it might be that pushing turnout is more important than trying to encourage switching.”
It is also, as the Trump campaign found in 2016, useful to be able to identify your opponent’s base. “They were very pleased with how successful their negative ads were,” says Bright, “in persuading Clinton voters not to turn out”.
Political scientists cautiously agree that Trump is going to lose this election. The New Statesman’s own forecast model, which runs 50,000 simulations a day based on local and national polling, puts Biden’s chance of winning at almost 90 per cent. The betting markets rate Trump’s chances higher, but gambling companies wouldn’t be doing their job if they didn’t convince some people to put money on the wrong outcome.
Jonathan Bright says that while there is a “general consensus” that a strong social media presence can make a difference of a few percentage points, the 2020 presidential election “may well turn out to be a strong corrective” in our understanding of the relationship between digital media and elections.
But whatever the outcome of the election, the rules of the attention economy will remain the same. More than half of Americans (and 35 per cent of people in the UK) will continue to get their news from Facebook, despite the fact that most people don’t understand or try to adjust what news the network shows them. And as advertising becomes ever more effective at reinforcing consumer desire, it will have an ever more pronounced effect on political ideas.