The most exciting point of Panos Toullis’s work involved a late night of caffeine-powered data-hacking, when Google executive chairman Eric Schmidt sauntered into the room, nudged him on the shoulder and said something to the effect of, “You know, you guys are doing something new and innovative and something that people will be repeating over and over again from now on.” To Toullis, a Ph.D. student in Harvard’s statistics department, this was like an aspiring pianist meeting Mozart. He recounted to the HPR, “The company is Google, and this dude sees so many technologies and integrations and he’s saying that this was something innovative! Surprising, right?”
The innovation to which Toullis was referring is the new world of political analytics that emerged out of Barack Obama’s 2012 campaign, known as Obama for America or OFA. OFA, initially designed for the 2008 race, was widely regarded as revolutionary when it started using data in earnest in 2012. Having hired Facebook and Google veterans, high-energy particle physicists, and executives from various up-and-coming companies in the Bay Area, OFA’s team was 10 times the size of Mitt Romney’s Republican equivalent. From conducting experiments to optimizing fundraising email wordology, the analytics team completely changed senior political operatives’ approaches to campaigning. It was the first to spend over $100 million on online advertising; it was also the first campaign to simulate database failures to practice handling end-of-the-world level disasters. It was, after all, the year 2012.
As 2016 approaches, the focus on data and analytics will continue to grow. Civis Analytics and BlueLabs, both founded by members of OFA’s analytics team, have expanded from the campaign world into the private sector as data consulting firms for progressive non-profit organizations and political campaigns. Hillary Clinton has already hired a key subset of Obama’s top digital and analytics personnel. And on the Republican side, the Koch brothers have invested $50 million into i360, a Republican equivalent to Civis and BlueLabs. Analytics, it seems, is here to stay.
It looks like the politicians will wage their battles with armies of data for the years to come. As campaigns gather more information about individual voters and continue to target at finer granularities, what is the ultimate affect of analytics on our democracy?
The 2012 Obama campaign relied heavily on analytics to plan advertising and voter outreach.
Making Democracy Weird
One of the major functions of analytics is to allocate resources to swing voters. If done correctly, these are the voters who should get targeted ads, the voters who will see TV commercials every night before the elections, the voters living in Henrico, Virginia and Jefferson, Colorado and Washoe, Nevada who get to shake your candidates’ hands when they walk off their planes.
But while targeting makes campaigns more efficient, it also induces a series of concerns about privacy, Orwellian governmental controls, and data mining. Matthew Rattigan, who masterminded OFA’s Facebook mobilization algorithm, acknowledged to the HPR that these concerns are valid. “This is a question a lot of us ask ourselves every day. There definitely is a data arms race. … Is micro-targeting something we do instead of leadership? Are we promoting democracy, or are we making democracy weird?”
Joe Trippi, who managed Howard Dean’s 2004 presidential campaign, argued that while the infusion of digital analytics into politics promises to give the people a voice, it has actually reinforced the power of the political elite. “Ten years ago, it was more about empowering people,” he explained to the HPR. Then, the growth of the Internet enabled 50 million voters to coalesce online around a single, shared cause—this was what OFA did in 2008 when Obama overcame Hillary Clinton for the Democratic nomination through his grassroots operation.
But after the Supreme Court’s 2010 Citizens United ruling—which opened the door for more political spending by outside organizations—and the resulting influx of massive amounts of money into politics, campaigns are at a point where, as Trippi puts it, “politicians are speed-dating billionaires.” Data is now being used to buy the best TV commercials and the most efficient online ads in order to manipulate people, rather than to bind them together. “It’s turning out to not be a great equalizer and great democratizer,” Trippi said. As far as he is concerned, data has been used more as a tool for the powerful to model and control the behaviors of individual voters.
A larger concern is not that political candidates have suddenly found some holy grail of mass-manipulation. Instead, the worry is something akin to what Eli Pariser, the CEO of Upworthy, terms the “Filter Bubble,” in which the content that people see online is increasingly tailored to what they want to hear. While we get to decide our own votes, the path to that decision is riddled with targeted ads that may be a reflection less of the candidates and more of ourselves.
A liberal-leaning college student might get a message from Rand Paul on same-sex marriage that “people ought to be treated fairly under the law.” On the other hand, a conservative grandfather on the same website might see Rand Paul’s comment to a group of pastors that it is a “moral crisis.” Rand Paul has actually said both of these statements, and both the student and the grandfather may be led to like him for incompatible reasons. Even worse, if data scientists have identified you as a “decided” voter who will not change your mind regardless of advertisements and TV spots, then you may theoretically only receive any communications from one side of the campaign.
This, then, is the risk of data. While the decision is up to the voters, we perhaps are receiving minimal exposure to contrasting opinions and basing our decisions on biased information, tailored to minimize discomfort and disagreement.
According to Trippi, in the short-term, data harbors more dystopian possibilities: “If you thought we could manipulate you with a TV ad, you haven’t seen anything yet.” From measuring cellphone usage to recording website visits to tracking location, campaigns are looking to expand their use of data. For the 2016 presidential election, a major focus will be matching online identities with offline ones: Facebook profiles with public profiles and web history with voter history. The increasingly holistic integration between phones, TVs, and computers, and the sharing of identity across all of those platforms will continue making data a more powerful tool in the coming election cycles. From Trippi’s perspective, data has made democracy weirder.
The use of data to target small groups of voters harkens back to the small-town stump speeches of the nineteenth century.
Social (Media) Capital
For some, these concerns present a naïve view of democracy. “Campaigns have always been targeting and trying to do it at the smallest unit at which they can profile or deliver communication,” Sasha Issenberg, author of The Victory Lab: The Secret Science of Winning Campaigns, told the HPR. “Starting in the sixties we were computerizing that and sorting through a more automated way.” While perhaps we can target and predict individual votes better than we could before, Issenberg argued, the distinction is purely one of degree, not of a qualitative difference. Campaigns have always at least tried to target voters at scale; who can hold it against the data scientists today for doing what they do well?
Issenberg also disputes the idea of a growing separation in our electorate: “What makes someone persuadable for Obama is exactly what makes them persuadable for Romney. If [the campaigns] are doing their jobs correctly, they should all be targeting the same people.” This means that both Democratic and Republican campaigns are really trying to attract the same sliver of people. Instead of pushing moderate candidates to the extremes of the left and the right by the filtered media, campaigns can now spend more their budgets trying to convince a set of undecided Ohioan stay-at-home-moms and Floridian retirees who are bombarded with a set of materials from both sides of the aisle.
Trippi’s concerns about the new “Filter Bubble” were perhaps best manifested in Mitt Romney’s 47 percent remark back in 2012. To a room of millionaire fundraisers, the then-candidate declared, “There are 47 percent … who are dependent upon government. … My job is not to worry about those people. I’ll never convince them that they should take personal responsibility and care for their lives.” The secret behind the 47 percent remark is that it really was the perfect message for a specific group of people.
However, as former OFA analyst and current BlueLabs strategist Harrison Kreisberg pointed out, “You can make the same argument in the 1800s. A politician can say something in Illinois and then, if they go to South Carolina, there’s no way of knowing what they’re saying is consistent.” As politicians travelled from city to city, they had the advantage of both targeting people in specific, smaller groups and having relative impunity in crafting messages only that local community would want to hear.
Kreisberg is right that today’s microtargeting is not so different than what politicians were doing in the 1800s. Issenberg is right that they have been trying to target voters at scale since the 1960s. Analytics betroths big data at a highly specialized, personal level to mediums well suited for large-scale fast distribution. Politicians can now target messages to specific groups of 30 people at a time while still reaching the 65 million Americans they need to come out on Election Day. And increasingly, they can do it without ever contacting the 47 percent who would never vote for them anyway.
But the issue with Issenberg’s counter-argument is that our ability to successfully target voters at an alarmingly individualized level is the qualitative difference between the sixties and today. Campaigns can now almost guarantee that a message is heard only by voters who they want to hear it . There are millions of dollars of funding and thousands of the most talented data-cowboys and cowgirls busy making sure we’re getting better and better at doing just that.
Ultimately, the long-term effect of analytics remains unclear. Whether the dystopian concerns become an issue or are merely overstatements will only become clear years down the line. For now, as evidenced by candidates stockpiling data and talent for 2016, analytics will continue to grow in influence. As many see it, politics is undergoing an irreversible transformation, like baseball did when introduced to a world of sabermetrics. We’re seeing a “money-balling of politics,” says OFA’s Rattigan.
But unlike making America’s favorite past time a little more efficient, this may change the very fabric of our democracy.
Image Credits: Maryland GovPics / Flickr; Obama for America / Flickr.