Not Even Cambridge Analytica Believed Its Hype

If the consulting firm’s “psychometric” modeling was really the key to winning campaigns, why would it even flirt with sketchier skullduggery?

One can be forgiven for not being quite sure what to think about Channel 4’s expose on Cambridge Analytica, the political-consulting firm linked to Donald Trump and others. On Monday, the British news channel released a story based on hidden-camera videos they took during meetings with CA higher-ups, in which the officials discuss a range of skullduggery, including bribes and sexual entrapment.

That story followed paired scoops in The New York Times and the British Observer, which focused on how CA’s much-hyped system of targeting voters was premised on data that, according to a whistleblower, was obtained surreptitiously through Facebook. As the Times put it, CA had made big promises about what it could do, “but it did not have the data to make its new products work.” Instead, it allegedly obtained extensive data from a researcher who had gotten it from Facebook on the premise that the information was for academic purposes. The revelation has placed Facebook in a deeply uncomfortable position, especially since the company knew about the misuse of the data but did not acknowledge it. Facebook and CA are now under new regulatory scrutiny in both the U.S. and Britain.

It’s hard to know what to make of the specifics of the Channel 4 story. CA predictably denies any wrongdoing, and the video’s editing makes it hard to determine the context for the things that CEO Alexander Nix and others said. But even if one takes CA’s denials at face value, the conversations caught on tape creates the strong impression that the company cannot produce the “psychometric” alchemy it has promised to clients. In that respect, CA joins a long line of political consultants who have promised they have bottled a new, magical trick, only to be revealed to be selling old wine in new bottles.

As Bloomberg reported in 2015, CA made some bold claims:

Cambridge Analytica’s trophy product is “psychographic profiles” of every potential voter in the U.S. interwoven with more conventional political data. The emphasis on psychology helps to differentiate the Brits from other companies that specialized in “microtargeting,” a catch-all term typically used to describe any analysis that uses statistical modeling to predict voter intent at the individual level. Such models predicting an individual’s attitudes or behavior are typically situational—many voters’ likelihood of casting a ballot dropped off significantly from 2012 to 2014, after all, and their odds of supporting a Republican might change if the choice shifted from Mitt Romney to Scott Brown. Nix offered to layer atop those predictions of political behavior an assessment of innate attributes like extroversion that were unlikely to change with the electoral calendar.

As the author of that piece, Sasha Issenberg, wrote, “The firm promised to tell me things I might not even know about myself.”

But Issenberg, perhaps the most sophisticated journalist on such campaign tactics, came away skeptical. For one thing, he wrote, “In conference calls and pitch meetings, Cambridge executives and analysts have betrayed confusion, if not outright ignorance, about some basics of American campaigns—from the definition of precincts (the smallest unit at which voter data is collected) to the difference between turnout patterns in primaries and caucuses.” For another, he found that CA’s results for his own personality diverged widely from those he received on a test from the University of Cambridge Psychometrics Centre.

These sorts of glitches did not prevent CA from winning important contracts in the 2016 election. Among the customers were the campaigns of Ted Cruz and Ben Carson. But the Cruz campaign soon had reservations, my colleague Rosie Gray previously reported: Not only was it upset that CA was also working for Carson, but it found that CA’s products didn’t work very well, and complained that it was paying for a service that the company hadn’t yet built.

One reason that these campaigns would shell out for a little-known, unproven operator was that members of the Mercer family, important Republican donors, were major investors in CA. Cruz hired CA; the Mercers put big money in a Cruz super PAC. New York Times reporter Ken Vogel noted Monday:

BIGGEST SECRET ABOUT CAMBRIDGE ANALYTICA: It was (& is) an overpriced service that delivered little value to the TRUMP campaign, & the other campaigns & PACs that retained it — most of which hired the firm because it was seen as a prerequisite for receiving $$$ from the MERCERS.

— Kenneth P. Vogel (@kenvogel) March 19, 2018

In fact, the Mercers later switched allegiances to Donald Trump (long a favorite of Breitbart, another Mercer vehicle), and the Trump campaign hired CA. But CBS reports that the Trump team eventually dumped CA, too, finding it useless: “The Trump campaign never used the psychographic data at the heart of a [complaint from a] whistleblower who once worked to help acquire the data's reporting—principally because it was relatively new and of suspect quality and value.”

In the Channel 4 video, Nix, along with colleagues Mark Turnbull and Alex Tayler, make some eye-popping claims to a reporter posing as a wealthy Sri Lankan would-be client. Turnbull speaks of engaging companies run by ex-spies from the U.K. agencies MI5 and MI6 to do research. “They will find all the skeletons in the closet,” he says.

Nix describes sting operations: “We’ll have a wealthy developer coming in, someone posing as a wealthy developer. They will offer a large amount of money to a candidate to finance his campaign in exchange for land, for instance, we’ll have the whole thing recorded on cameras, we’ll blank out the face of our guy, and then post it on the internet. … Send some girls around to the candidate’s house. We have lots of history with these things.”

Are these things true, or are they just Nix and Turnbull trying to land a client? “I am aware how this looks,” Nix said in a statement. “I must emphatically state that Cambridge Analytica does not condone or engage in entrapment, bribes, or so-called ‘honeytraps,’ and nor does it use untrue material for any purpose.” In other words: Trust me, I was lying. Which in this case, isn’t necessarily all that hard to believe. Given that CA has overstated its powers in the past, why wouldn’t it do so here? Still, it’s hard to tell what to believe.

The more important takeaway is this: If Cambridge Analytica’s psychometrics were as effective as they have claimed, why would they need to even discuss engaging in this type of skullduggery? Forget the morality of engaging honeytraps or sting operations; a company that truly believed it could abseil into voters’ heads with sophisticated data and manipulate them that way would feel no need to make claims about other, less scientific methods. Nix and Turnbull could have told the ersatz Sri Lankan that stings and such were outmoded.

As for the actual stuff Nix and Turnbull discuss, it’s pretty well-worn. When Turnbull describes using ex-British intelligence officers to do research, he could be describing (or who knows, maybe even is describing) Christopher Steele, the ex-MI6 officer who compiled a dossier on Trump on behalf of Fusion GPS, which was in turn working for a law firm, which was in turn working for the Democratic National Committee. And Fusion had previously been hired by a GOP donor. That sort of work is hardly revolutionary.

The honeytraps and stings are a bit more exotic, at least in American politics; they are more the stuff of television shows and sometimes of politics in other countries, but in any case they are also not innovative. The fact that Cambridge Analytica is peddling such tactics, whether honestly or disingenuously, implies their psychometric tricks are a Potemkin village.

CA would hardly be the first offender. There’s a long line of false (or at least exaggerated) prophets in political consulting who have overhyped their own, messianic new campaign solutions. CA’s psychometrics pretended to supersede a previous idol, microtargeting. Barack Obama’s victories established sophisticated data modeling as an elixir, and Hillary Clinton’s campaign lapped it up, leading her campaign to discount warning signs in key Midwestern states. Data alone was not enough, either. In April 2016, Molly Ball visited the American Association of Political Consultants’s annual conference and found widespread concern that recent electoral results were showing that the pros had no idea what they were doing. Some political scientists have argued that campaign tactics make very little difference to the outcome of elections anyway.

New technological advances have, of course, changed campaigning, but it’s wise to be skeptical of anyone who argues that he and his company have alone discovered some new trick that no one else has—especially if that’s a vague and unproven scientific claim like the one CA makes.

“It’s no good fighting an election campaign on the facts, because actually it’s all about emotion,” Turnbull says in the Channel 4 video. Nix later says that there “are things that don’t necessarily need to be true as long as they’re believed.” These statements are deeply cynical, but they’re also among the most uncontroversial things that any Cambridge Analytica official makes on tape. This is a view that could be had from any pessimist on the street, to say nothing of many a political consultant. There’s no psychometric modeling required—which is a good thing, since there’s little evidence that what CA is peddling works.