by Philip Bump Washington Post Jan 9, 2019
The revelation on Tuesday that Donald Trump’s 2016 campaign chairman, Paul Manafort, shared polling data with a colleague in Ukraine who had ties to Russian intelligence predictably kicked up a furor of speculation about the significance of the move.
This is what one of the Russia-Trump collusion scenarios looks like: someone from Team Trump passing data to the Russians the latter group could use to target voters and influence the election. After all, the common understanding is Russia’s interference efforts included sophisticated targeting of specific voting groups on Facebook, which could have made the difference in states that Trump narrowly won on his way to an electoral-vote victory.
That understanding about Russia’s sophisticated targeting, though, is not supported by the evidence — if it’s not flat-out wrong.
Let’s assume that your goal is to win an election using social media. What Facebook offers campaigns is the ability to target voters in very specific ways, including by age, interest and location. The more refined your target, the more expensive the ad. There are other ways to leverage social media, of course, including creating your own Twitter or Facebook accounts and building up followers or broader outreach efforts that you might see from a corporation — AARP targeting people over 65, for example.
If we’re talking about specific poll data being passed from the Trump campaign to Russia, though, the presumption is that the Russians would receive information that allowed very specific targeting of voters in places that would have had the biggest effect on the 2016 election.
There are a lot of ways in which even that broadly stated paragraph doesn’t match well with what’s known about the Manafort situation. According to the New York Times, the information passed from Manafort included some proprietary information but, for the most part, was public, obviating the need for much cloak and dagger. The data was passed to Manafort’s colleague Konstantin Kilimnik in the spring of 2016, before Trump had been nominated by the Republican Party. It’s data that, by Election Day, would be several months out of date.
The most important way in which that paragraph doesn’t match reality is in the specificity of that targeting. Yes, the Russians at times targeted specific age groups that matched specific interests within a specific geographic area, but those tailored ads often shared specific characteristics: targeting adults who had expressed an interest in issues related to African American politics in or near cities with large black populations.
It’s important to remember how important geography is to U.S. presidential politics. You know this, of course; had the popular vote determined the 2016 winner instead of geography, Donald Trump would be in negotiations to renew “Celebrity Apprentice.” But it means that targeting, say, 30- to 40-year-olds across the country with an interest in jazz music who own guns is probably less helpful for winning an election than targeting New Hampshirites, even if it’s much more specific.
Most of the ads purchased by the Russians didn’t specify a geographic target smaller than the United States on the whole, according to a Post review of the ads released by the House Intelligence Committee. Those that did target specific states heavily targeted those that weren’t really considered targets of the 2016 election, such as Missouri and Maryland. And of those ads that did target specific states, most happened well before or well after the final weeks of the campaign.
We looked at this broadly in May when the data were first released. The Russian effort included launching Facebook (and subsidiary Instagram) ads each month from June 2015 to May 2017 (plus some in July and August of that year).
More than half of the actual clicks on ads, in fact, came after the election.
What we hadn’t done, though, is overlay when ads ran with geographic targeting. The map below does precisely that, counting campaigns that ran in full or in part in any given month, regardless of the success of the campaigns.
A few things to notice immediately: There was an early push to target ads in Missouri, Maryland and Ohio. That’s in part because the Russians had started assessing American politics in 2014 and seized on the polarized Black Lives Matter movement as a focus of energy. A lot of the ads in those states in that time period actually targeted specific cities or regions, like Baltimore and Ferguson where there had been days-long protests.
The black bars indicate campaigns that ran in October 2016 or in the first eight days of November; that is, until Election Day. You’ll notice that two of the states where the most ads ran were Texas and New York — neither of which was considered close in the final weeks of the campaign. States that Trump won narrowly — specifically Pennsylvania, Michigan and Wisconsin — saw very few campaigns and, per our count, were seen by fewer than 1,000 people in the last five weeks before the election.
The most successful of the ads that ran in those three states at the end of the campaign, it seems, was this one — which ran not only in Michigan, but also California, Illinois, New York and Texas.
Not exactly a strong exhortation to vote for Trump.
Most of those ads that ran in Minnesota at the end of the campaign, the sole state where there was overlap between a narrow race and a heavier rotation of ads, targeted Minneapolis with messages related to African American issues. Those were more successful — but were still viewed only about 7,000 times.
Of course, no one knew with certainty that the states that ended up being close were going to be the ones that ended up being close. Part of political targeting is guesswork, estimates based on polling and prior elections. But that there were so few ads in, say, Nevada or Florida in the weeks before the election gives the lie to the idea that the Russian effort was focused on targeting specific voters.
These ads were not the only way that the Russians targeted American voters, of course. As mentioned above, there were hundreds of ads targeting broad interest groups like gun rights supporters or the LGBTQ community. But those efforts were obviously more focused on the stated intent of fostering political division than on ensuring a particular outcome in the 2016 election.
It has been argued that the focus on messages disparaging Hillary Clinton in the black community might have been effective at tamping down turnout in important ways. What that argument misses, though, is that the ads at the tail end of the campaign didn’t really focus on suppressing turnout that heavily, and that it was generally much broader outreach than targeting people in particular places.
Could the Russians have deployed sophisticated social media tracking tools to follow users across sites and determine where they were based, then tailor political messages to the broad groups or Twitter accounts they created? In theory, sure — but it seems like an awfully indirect way of targeting specific geographies, especially when they were already buying ads with specific geographic targets without raising any eyebrows.
Even if Russia had deployed a particularly insightful, strategic effort to influence specific voters in specific places, it’s not necessarily the case that they would have had to rely on data from the Trump campaign to do so. As the New Yorker reported last year, hackers believed to be linked to Russian intelligence stole information from the Clinton campaign that could have revealed where they were most concerned about turnout, giving the Russians a map for where to deploy their resources the most effectively.
But, again, there’s no evidence that they did any particularly sophisticated targeting. Russia’s hacking of the Democratic National Committee and Clinton’s campaign chairman remains the more obviously successful and effective influence effort of the 2016 campaign.
It’s just not as enticing as the idea of savvily deploying American social media users against themselves.