How Donald Trump campaign used data scraped from Facebook to win presidency

From harvesting personal information to Russian troll ads, there appears to be many ways in which social media giant inadvertently assisted head of state

Something that you learn when you get ready to write about political campaigns: the expression “victory has a thousand fathers but defeat is an orphan” is apparently attributed to Galeazo Ciano, an Italian count who wrote it in a 1942 diary that was published posthumously. The guy knew something about defeat, given that he was Benito Mussolini’s son-in-law and a senior official in Mussolini’s government.

Ciano’s point wasn’t about politics as such, but the phenomenon he captures is hugely resonant in that context. Perhaps we can swap “fathers” for “consultants” to increase the specificity, but otherwise the idea that everyone takes credit for a winning campaign while burning bridges after a losing one is ‘perfezionare’.

Donald Trump’s 2016 election scrambled that calculus, though. He won – but he also lost the popular vote. On election day, Kellyanne Conway was reportedly seeding the idea that an expected loss was the fault of the Republican National Committee. After Mr Trump was elected, Cambridge Analytica both denied and embraced the idea that they’d played a pivotal role in the outcome.

We’re back at Ciano’s point: the end result of any campaign is that it’s often hard to tell who deserves credit – positive or negative – for the outcome. After Mr Trump’s squeaker of a win, it’s harder than normal, given that a thousand small things might have made the difference.

In the wake of revelations that Cambridge Analytica used data vacuumed up from Facebook by a researcher in 2014, there’s been a lot of finger-pointing at the firm suggesting that their misuse of that information was critical to Mr Trump’s election. Cambridge Analytica hasn’t done itself any favours, given that its marketing plan for years was to position itself as the nebulous magicians who’d cracked the code on victory. But there’s little actual evidence that what the firm did with that Facebook data was actually all that fruitful. Meanwhile, the campaign’s more direct use of Facebook in its efforts have been generally disregarded, given that they don’t involve Bond-villain-style antagonists complete with British accents.

There are at least five ways in which Facebook’s network and data were used to bolster Mr Trump’s campaign. We’ve done our best to rank them from least to most important in his victory. But even this analysis is hampered by the various parties claiming to have fathered Mr Trump’s victory.

5. Cambridge Analytica’s scraped Facebook data

It’s important to understand two things here. The first is how Cambridge Analytica got the Facebook data. The second is how it was used in the Trump campaign.

We’ve answered the first question previously, but a quick review is useful. Cambridge Analytica was founded a few years ago as an offshoot of a British firm called Strategic Communication Laboratories Group to pitch campaign services in the United States. The firm’s value proposition was that it could offer insight into the psychology of individual American voters – but gathering that information on a nation of 320 million people was daunting.

So, Cambridge Analytica apparently hired a British researcher who created a tool that gathered data from people’s Facebook accounts. At the time, in 2014, Facebook’s permissions for accessing that information were more lax than they are now and information was gathered not only from the users participating with the researcher, but also from their Facebook friends. Ultimately, some 50 million people’s information was added to the data set. That was then used in some still-unclear way by Cambridge Analytica, though it may have been used to create general categories of belief on key political subjects.

When the Trump campaign hired the company in June 2016, it was only after an internal struggle. Hiring Cambridge Analytica was seen as a way of currying favour with Robert Mercer, the heavy-hitting Republican donor who helped found the company. Cambridge Analytica has asserted that it was never able to actually create the psychographic profiles for which it’s known, given the short time frame between June and 8 November of that year. The Trump campaign has agreed that it didn’t use Cambridge’s data.

So, in other words:

Cambridge Analytica says it didn’t have good psychographic data that it could use for the Trump campaign.

The Trump campaign says that it didn’t use Cambridge Analytica’s data anyway.

Regardless, the data was two years old.

We are certainly taking the word of Cambridge Analytica and Mr Trump at face value. But Mr Trump’s team has pointed out that it had access to data from the Republican National Committee, which has been building out profiles of American voters for years, including vote history. That data set is certainly enormously useful and there’s little reason to think that it wasn’t the primary tool Mr Trump’s digital team used.

What’s more, Cambridge Analytica could (as they did in hidden camera footage obtained by Channel 4 in Britain) argue that their magic made the difference in a close race. But, publicly, they didn’t. They downplayed their role.

There are certainly other ways in which Cambridge Analytica may have affected the Trump campaign. But the use of that Facebook data? There’s no suggestion I’ve seen that it played any role, much less a significant one.

4. Russian troll Facebook ads

Before Cambridge Analytica, there were the Russian trolls – at least in terms of speculation about nefarious influencers on Mr Trump’s campaign effort.

We know, thanks to data released by Facebook and an indictment issued by special counsel Robert Mueller, that Russians associated with a group called the Internet Research Agency bought ads on Facebook to promote divisive content. Those ads included messages promoting Mr Trump and disparaging Hillary Clinton, including by suggesting that black voters shouldn’t support her candidacy.

The scale of those ads, though, has been vastly overstated. Democrats on the House Intelligence Committee released data showing a number of those ads. Of the 30 ads they released, only seven ran in the last month of the race.

Overall, 10 million people saw ads bought by the Russians – but more than half were seen after the election was over. About 440,000 Facebook users saw the ads before election day. A quarter of the ads, Facebook estimates, were never seen by anyone.

So the ads were seen, certainly, hundreds of thousands of times. But each month in the United States and Canada in the fourth quarter of 2016, 231 million people used Facebook. Meaning that about 0.2 per cent of Facebook users would have seen the Russian ads had all of them been seen in October or November of 2016.

The effect of that, it’s safe to say, was small.

3. Russian troll Facebook engagement

That analysis skips over two caveats. The first is that the data includes only Facebook ads, excluding Facebook-owned Instagram, where about 5 per cent of the ads ran. Facebook hasn’t released data on the scale of Russian activity on that platform.

The other caveat is that ad-buying was only part of what the Internet Research Agency did. Much of its activity was aimed at organic social activity, which is to say, posting things on Facebook and hoping they went viral. The Russians used Facebook as a platform to organise and promote real-world events aimed at bolstering Mr Trump and sowing dissent. They engaged American activists over Facebook. They posted memes and comments aimed at the same outcome.

Facebook estimates that as many as 126 million people saw content posted by the Russians.

In written testimony to Congress, Facebook general counsel Colin Stretch downplayed that number.

“This equals about four-thousandths of one per cent (0.004 per cent) of content in News Feed or approximately 1 out of 23,000 pieces of content,” he wrote. “Put another way, if each of these posts were a commercial on television, you’d have to watch more than 600 hours of television to see something from the IRA.”

We have gotten to a point where we need to ask a broader question. To what extent does seeing a meme or an ad actually affect someone’s vote? Much of American politics these days is predicated on partisan loyalty: Republicans mostly voted for Mr Trump and Democrats mostly for Ms Clinton. A Russian ad – or a campaign ad – wasn’t probably going to change that. It gets trickier when you consider voter suppression, efforts like those by the Russians (and the Trump campaign itself, according to Bloomberg reporting) to encourage black Ms Clinton voters not to cast ballots. That may have been more effective, but it’s not simple to measure.

2. Trump campaign integration with Facebook

Before the 2016 election, Facebook was very upfront about how it might be a useful tool for political campaigns. Its ability to connect voter files to individual behaviour – and then show targeted ads to those voters – was part of its sales pitch. (In the wake of questions about how Russians used the platform, Facebook buried that bit of self-promotion.)

What Facebook promised, in other words, was a version of what Cambridge Analytica claimed to be able to do. Except that, unlike Cambridge Analytica, it could actually demonstrate that the tools worked as advertised.

The Trump campaign has been explicit that Facebook was central to its victory. Brad Parscale, who managed Mr Trump’s digital efforts, has touted the campaign’s use of Facebook repeatedly. Some 80 per cent of the digital ad budget was spent on Facebook, he’s said, with the campaign creating thousands of iterations of slightly-tweaked ads meant to test which sale pitch was most effective. Both Mr Parscale and representatives of the Clinton campaign have acknowledged that Mr Trump team’s rhetoric and imagery was a strong fit for Facebook’s network and for Mr Trump’s base.

What’s more, Facebook had staff that actually worked alongside Mr Trump’s team, helping them figure out how to maximise the site’s toolset to increase donations and reach. And, of course, to spend as much of their budget as possible at Facebook. Ms Clinton’s campaign worked with Facebook, too, but Facebook itself had much less influence over how the Clinton campaign approached digital marketing than it did over Mr Trump’s.

Both Facebook and Cambridge Analytica worked alongside Mr Trump’s digital team, day in and day out. Mr Trump’s campaign used Facebook’s tools and it championed Facebook’s results. It showed well-honed ads to the people it wanted to see them.

Facebook said in September that the Russians spent $100,000 (£70,900) on Facebook ads. The Trump campaign spent about 700 times as much.

1. Trump supporters’ Facebook activity

All of this skips over what’s probably the most important role Facebook played. It was most useful to Mr Trump, it’s safe to assume, when it was doing what it was designed to do: let people share and comment on content.

We’ve talked a lot about how false or misleading news stories were rampant on Facebook before the election. The site began experimenting with how to slow or block the flow of fake news on the site, without a whole lot of success to date. Those fake stories were among the most popular items on Facebook at the tail end of the campaign.

This, too, underplays the extent to which real news shared among family and friends probably influenced the vote. Again, we don’t know how much of an effect any ad or news story affects someone’s vote, but it’s fair to assume that organically shared content on the network reached a lot more people than Mr Trump’s ads or the Russians’ ads did.

We do know, though, that Facebook can help boost voter turnout – thanks to research conducted by Facebook. In 2010, the site ran a test to see if raising awareness that friends and family had voted would increase a user’s likelihood of voting. It did. About 340,000 more people voted after seeing Facebook’s nudge than would otherwise have. It’s not much of a stretch to assume, then, that enticements and stories shared by friends and family informally might also have had an effect.

This is not sexy, in terms of storytelling. “Facebook had the most influence in the 2016 election by simply being Facebook” isn’t as compelling as “data stolen by a rogue British firm was used to create psychographic profiles that gave Trump a slim win.”

But the story that should be told is the one that’s the most accurate. It’s not always the flashiest fathers claiming credit for a victory who deserve it the most. It usually isn’t.

The Washington Post