More than half of Facebook pages that displayed U.S. political ads during a recent 13-month period concealed the identities of their backers, according to research reviewed by POLITICO — a tide of deceptive messaging that raises new questions about the social network’s promises of transparency.
The stealth political ads were worth at least $37 million, equivalent to 6 percent of all the money spent on Facebook ads in the U.S. during the research period, from May 2018 until June 2019, according to estimates from New York University researchers.
The academics also found evidence that partisan groups across the political spectrum had created 16 clusters of inauthentic communities that bought ads aimed at swaying potential voters, borrowing from tactics Russian operatives had employed during the 2016 U.S. presidential election campaign.
The findings cast doubt over Facebook’s ability to enforce its own transparency rules across its global platforms, which are already playing an outsize role in spreading political messages to voters ahead of the 2020 U.S. presidential election. The rules require buyers of political ads to identify who is paying for them after criticism of Facebook’s role in the 2016 vote.
Mark Zuckerberg, Facebook’s chief executive, had assured Congress that his company was ratcheting up its efforts to root out interference ahead of November’s presidential election.
Facebook’s platforms have sold more than 7 million political ads in the U.S. worth more than $1.1 billion since May 2018
But political parties and digital campaign groups are growing increasingly sophisticated in gaming the rules and eluding Facebook’s attempts to uproot underhand behavior, the NYU researchers warned.
As U.S. voters gear up to head to the polls later this year, campaigns and online activists are already reworking their tactics, including paying social media influencers to promote partisan messages, to bypass Facebook’s rules on online political ads.
“There’s very little to stop someone from opening a Facebook page, running political ads, getting caught and repeating that cycle,” said Laura Edelson, a researcher at NYU’s Online Political Ads Transparency Project, who co-authored the study of ads on Facebook and Instagram, its photo-sharing platform. “We’re not really in a better position to know what’s going on now than we did in 2016.”
In response, Facebook said its political transparency tools were better than those offered on television or radio, and pointed to a series of steps it had taken to clamp down on wrongdoing.
They include efforts announced late last year aimed at clamping down on digital misinformation, labeling content from government-backed news outlets like Russia Today, and making it more clear to voters that political groups had bought messages on both Facebook and Instagram.
The social network, though, has rejected a complete ban on all political advertising — a step taken by Twitter — and lawmakers have criticized the company for refusing to remove deceptive ads posted by candidates, including Donald Trump.
Facebook’s platforms have sold more than 7 million political ads in the U.S. worth more than $1.1 billion since May 2018, according to the company’s online ad-tracking library — dwarfing the $216 million that Google and its YouTube video platform have sold over the same period, according to the search giant’s figures.
Top spenders on Facebook include former U.S Democratic candidate Mike Bloomberg, who has spent almost $60 million since he joined the presidential race, and Trump, who has forked out $26 million over the same period.
At congressional hearings and in a round of interviews in late October, Zuckerberg said the tech giant had turned a corner in its fight to keep bad actors from using its network to deceive voters.
“We’ve gone from being on our back foot to now proactively going after some of the biggest threats that are out there,” Zuckerberg told reporters. “We’ve built systems to fight interference that we believe are more advanced than what any other company is doing — and most governments.”
Still not enough
But the NYU researchers discovered several loopholes that allowed Facebook pages that had bought more than 350,000 political ads, worth at least $37 million, to avoid ever having to disclose who was behind them. That represented 54.6 percent of all Facebook pages whose ads were included in the company’s online register of political messages.
The dollar figure is an estimate and may be higher because Facebook does not publish the exact figures spent on online ads across its platforms.
Following the 2016 U.S. presidential election, Facebook imposed rules that required political advertisers to include a message on all paid-for posts disclosing who was behind them. The goal was to allow voters to understand which groups were targeting them on social media.
The company also created an online register of all political ads, though the database remains difficult to access and fails to capture all paid-for messages bought by political campaigners. It often does not immediately know when partisan ads run on its global platforms, according to the NYU research.
To analyze how actors were buying political messages on Facebook and Instagram, Edelson and her co-authors, Damon McCoy and Tobias Lauinger, created a database of all available paid-for messages that had run on the company’s platforms in the U.S. for 13 months ending June 1, 2019. The researchers incorporated the company’s own ad library with an independent dataset pulled directly from Facebook’s online database of political ads.
The NYU academics then combined all ads from the same buyer to avoid duplication, cross-referenced their dataset with Facebook’s records and tracked how long it took the social network to add undisclosed political ads to its ad library.
The research represents the largest effort, so far, to determine how successful Facebook has been in meeting Zuckerberg’s promises to clamp down on the worst offenders, including potential foreign actors targeting the upcoming U.S. presidential election and elsewhere.
Facebook found it difficult to keep up with the pace of stealth political activity on its network, according to the NYU research.
During the 13-month research period, for instance, the academics said that the number of partisan paid-for messages whose backers were not disclosed did not decline, despite Facebook’s multi-million dollar investments targeted at those who tried to circumvent its systems.
Among the groups that repeatedly did not disclose their political affiliations were several U.S. government agencies, as well as Xinhua News, the media outlet owned by the Chinese government. Despite Facebook outlawing such foreign spending on ads connected to domestic U.S. political topics, Xinhua News spent at least $16,600 on 51 undisclosed political ads, based on the NYU data.
“You want to figure out where the weak points are,” Edelson, the NYU researcher, said in reference to Facebook’s online systems. “But everything is a weak point.”
Nathaniel Gleicher, Facebook’s head of cybersecurity policy, told POLITICO that political actors had increasingly turned to not disclosing who was behind online political ads, in part because of the company’s clampdown on who could target voters.
The tech giant would now force any Facebook page that failed to publish who was buying ads on its behalf to go through an additional verification step, though Gleicher acknowledged it was a game of cat-and-mouse with those who did not want to play by the rules.
“There is an evolution of tactics,” he said. “We would expect actors to try other techniques, including running pages that fail to disclose who’s behind them and pretend that they’re independent.”
‘Inauthentic online communities’
Alongside the almost $40 million of political ads that did not disclose their backers, the NYU academics also found 16 clusters of Facebook pages which spent, collectively, $3.8 million on political ads, that had run the same paid-for messages while pretending to be different organizations.
That was a common deceptive tactic of the Internet Research Agency, an organization based in St. Petersburg, Russia, that used Facebook and other online platforms to try to influence U.S. voters on behalf of Moscow in 2016, according to U.S. intelligence agencies. The social network has repeatedly clamped down on groups that operate multiple Facebook and Instagram pages without disclosing their coordinated behavior.
In February, for instance, the company removed more than 100 Facebook and Instagram accounts with ties to Russia, Iran, Vietnam and Myanmar for what it called “inauthentic behavior,” in which online tricksters ran politically related social media pages without disclosing their true intentions.
“Political advertising is just part of the wider problem. Platforms like Facebook haven’t even started to address all these murky areas.” — Samantha Bradshaw, researcher at Oxford University
But analysts say these tactics have become commonplace across Facebook’s platforms, with often domestic and foreign-backed groups sharing content in increasingly sophisticated online operations that make it almost impossible to determine who is behind these digital campaigns.
“Actors who are trying to influence voters are always evolving,” said Chloe Colliver, who heads the digital research unit at the Institute of Strategic Dialogue, a London-based think tank that focuses on extreme speech who was not associated with the NYU research. “It’s clear that there is still large-scale inauthentic activity that’s getting through.”
In one series of political ads bought from May 7, 2018, to May, 31, 2019, the NYU academics discovered the same paid-for U.S. messages had been spread by 23 separate Facebook groups that promoted left-leaning causes or opposition to Trump without disclosing their coordinated online activity — a likely breach of Facebook’s code of practice.
In total, the groups spent more than $160,000 on these messages, mostly to target small audiences such as working moms or unionized workers in swing states like Pennsylvania and Ohio, according to a review of Facebook’s ad library.
Across those pages, whose followers totalled in the tens of thousands, partisan messages against Trump’s tax cuts and rollbacks of federal environment protections were interspersed with nonpolitical posts about being a working parent or support for local causes. POLITICO did not receive a response when it contacted the pages for comment on who was paying for the political messaging.
“These ads appear to be targeted to particularly small audiences, with an average spend of $23,” the researchers wrote.
The increasing sophistication in such online tactics is the latest evolution in groups’ efforts to target voters, according to misinformation analysts. That includes building such non-political Facebook communities over several years, only then to pepper voters with overtly partisan, paid-for messages in the buildup to national elections.
“Facebook hasn’t done enough,” said Samantha Bradshaw, a researcher at Oxford University’s Computational Propaganda Project, who published a recent report on social media manipulation and was not associated with the NYU research. “Political advertising is just part of the wider problem. Platforms like Facebook haven’t even started to address all these murky areas.”
Want more analysis from POLITICO? POLITICO Pro is our premium intelligence service for professionals. From financial services to trade, technology, cybersecurity and more, Pro delivers real time intelligence, deep insight and breaking scoops you need to keep one step ahead. Email firstname.lastname@example.org to request a complimentary trial.