This post originally ran on Epolitics.com on October 4, 2017. Let’s reread it now in the light of this week’s Manafort redaction disaster, in which we learned that Trump’s onetime campaign manager had passed polling data to a suspected Russian intelligence agent. Go figure.
A number of Russian-linked Facebook ads specifically targeted Michigan and Wisconsin, two states crucial to Donald Trump’s victory last November, according to four sources with direct knowledge of the situation.
Some of the Russian ads appeared highly sophisticated in their targeting of key demographic groups in areas of the states that turned out to be pivotal, two of the sources said. The ads employed a series of divisive messages aimed at breaking through the clutter of campaign ads online, including promoting anti-Muslim messages, sources said.
We already knew that Russian groups used social media to pound on the raw nerves of American political discourse around race and religion. But this revelation leads directly to the question haunting all the Trump-Russia talk: what did the campaign know about what the Russians were doing, and what (if anything) did it do to help them?
The Michigan/Wisconsin angle may turn out to be a significant hook in the investigation. In October of 2016, most campaign analysts agreed with the Clinton campaign that those two states were already in her bag, despite warning signs from the Democratic primary and pleas for help from grassroots organizers before Election Day. The Trump team, which relied heavily on highly targeted Facebook content for voter contact, by contrast shifted resources to Rust Belt states like Michigan, Wisconsin and Pennsylvania in the last weeks of the ccampaign. Did they also pass the word to outside actors working on their behalf?
Timing is the critical question: if the Russian ads ran months before Election Day as part of a package of content aimed at traditional U.S. battleground states, they become less of a sign of potential collusion. If instead they tracked the Trump campaign’s GOTV targeting tightly, that gun starts to smell like smoke.
How would a campaign pass targeting parameters to outside actors? I explored the options for Campaigns & Elections back in August:
…Investigators will surely look for evidence more subtle than an email announcing dirt from a foreign government on a political rival.
One potential angle? Data and targeting information left online for someone to find, perhaps sitting in the open behind an unprotected (but unpublicized) URL or IP number. We’ve seen this story before with other kinds of campaign assets, most famously when long clips of high-quality, TV-friendly Mitch McConnell b-roll video turned up online and got the Daily Show treatment.
Campaign data wanders online surprisingly frequently, too, often without clear signs of who put it there. For example, a national voter file popped up on a bare IP number late in 2015, and a huge trove of voter data sat out in the open for two weeks in June of this year. The firm responsible for the latter incident, Deep Root Analytics, owned up to it—blaming it on human error. But how often does someone let it all hang out on purpose?
An organization trying to help a campaign by targeting content via a bot or orthodox content-promotion wouldn’t need something as elaborate as a full voter file, in any case. Trump’s campaign made a point of aiming Facebook content at specific slices of the electorate, and a simple spreadsheet of counties or zip codes and demographic parameters would allow outside actors to hit the same voters with positive or negative messages.
Investigators apparently want to see if anyone outside the Trump campaign used Facebook audiences directly created by the campaign, but independent groups could get the same results just by knowing the targeting model.
With over three thousand separate Russia-linked Facebook ads already in investigators’ hands, we have not heard the last of this story. I suspect we’ve barely heard the beginning.