You are using an outdated browser.
Please upgrade your browser
and improve your visit to our site.
Skip Navigation

The Anti-Democratic Origins of Voter Prediction

How the Simulmatics Corporation tried to engineer an election—and paved the way for the Cambridge Analytica scandal

I’m always a bit relieved when a tech company’s algorithm seems to fall short of mind reading. It’s somewhat reassuring when a so-called smart ad displays hideous shoes I would never buy just because I once bought a different pair from the same brand, or when Facebook suggests I befriend a good friend’s ex, having determined that he’s in my extended orbit, but lacking the sophistication to comprehend that I dislike him. In an age of unchecked data harvesting, in which corporations have free rein over the entirety of your internet history and personal information—not to mention, in some cases, your movements, thumbprint, face, and DNA—there’s a certain comfort in seeing where the machine still hasn’t quite mastered the quirks and nuances of human thought.

If Then: How the Simulmatics Corporation Invented the Future
by Jill Lepore
Liveright, 432 pp., $28.95

That sense of comfort, though, is almost certainly illusory. Tech companies’ surveillance and rampant data collection practices don’t need to replicate the exact intricacies of a person’s brain in order to invade the privacy of billions of people. There’s Google, for one, which quietly yet meticulously collects data on users’ searches, locations, and video-watching habits, and has even partnered with a private health care chain to obtain the medical information of millions of people. Then there’s also Facebook, which stores information on the pages its users have clicked on, the people they’ve searched for, and their ethnicities. These tech giants’ primary use of all this personal data so far seems to be to sell ads. But 2016’s Cambridge Analytica scandal also raised the possibility that these troves of information could be marshaled in an attempt to influence a national election, or change the course of democracy.

Though such predations may look like a uniquely twenty-first–century problem, Jill Lepore’s new book, If Then, demonstrates that this technological dystopia has been in the making since at least the Cold War era. Lepore traces the birth of one strain of predictive technology through the rise and fall of the mysterious Simulmatics Corporation, an advertising and political consulting company that produced a “People Machine,” or a computer program designed to profile and predict the behavior of voters. Among other ventures, Simulmatics consulted on John F. Kennedy’s 1960 campaign and (it claimed) helped him win, in part by persuading him not to attempt to obscure or downplay his Catholicism and to take a stronger stance in support of the burgeoning civil rights movement. Though the company was eventually shuttered and forgotten, it was also, Lepore writes, a shadowy and indirect participant in some of the most significant events of the decade, including the Vietnam War, the civil rights movement, and the construction of Lyndon Johnson’s Great Society.

Compared to today’s still-imperfect algorithms, Simulmatics’ People Machine was rudimentary and very often either missed the mark or produced rather commonsense results. And the company itself, though staffed by several brilliant minds, was at least as much the product of savvy hype and its founder’s well-placed connections as it was of groundbreaking technological innovation. What Lepore’s rich account unearths is the impetus behind the project, a set of attitudes that continue to drive psychographic microtargeting efforts today: For the stratum of professionals who developed voter prediction, politics was primarily a code to be cracked, rather than a means to a better life, let alone a matter of survival. The men who ran Simulmatics were archetypal Cold War liberals—committed, in a sense, to an idea of human progress and the prospect of a brighter world—but also completely in the thrall of the idea of engineering and controlling that world themselves.

The Simulmatics Corporation, which was in business between 1959 and 1970, was founded by Madison Avenue advertising executive and man-about-town Ed Greenfield. The company’s name, which now has the retro-futuristic feel of something out of The Jetsons (a show that first aired, as it happens, around the same time that Simulmatics was at its height), was a portmanteau of “simulation” and “automatic.” Greenfield hoped it would become a household phrase. “Instead, ‘artificial intelligence’ became that catchphrase,” Lepore writes. “Still, ‘artificial intelligence’ is pretty close to what Greenfield meant by ‘simulmatics.’”

Greenfield, a lifelong Democrat with an interest in civil rights, defeating communism, and presidential politics, had been stung by the loss of the Democratic Party candidate Adlai Stevenson to Dwight D. Eisenhower in the presidential election of 1952. That same year, he had also watched CBS unveil a massive computer on election night that tallied returns more quickly than ever before, and accurately forecast Eisenhower’s win. Convinced that such a tool could be harnessed for Democrats, Greenfield assembled a handpicked team of leading behavioral scientists over the next few years. They ultimately constructed what Greenfield called the People Machine, a computer program that analyzed voter data in order to predict election outcomes. The program they sought to create, according to Lepore, was “not, at an elementary level, different from what Cambridge Analytica sold as its services to the Trump and the Brexit ‘Leave’ campaigns in 2015 and 2016.” In the spring of 1959, Simulmatics fed 100,000 voter surveys from Gallup and Roper into the People Machine to create 480 discrete voter profiles—“Midwestern, rural, Protestant, lower income, female” was one example of a voter type—that could then be matched to election returns from past years, and used to make predictions about those voters’ future behavior.

It’s not entirely clear how well the technology ever worked. For example, an analysis produced by the machine suggested that Kennedy could appeal to black and Jewish voters by condemning the anti-Catholic prejudice he himself faced. He did so, and went on to win those groups. But, as Lepore observes, many of the recommendations that Simulmatics gave the Kennedy campaign largely echoed what was “fairly commonplace political wisdom among his close circle of advisers.” The machine was birthed during the height of America’s struggle for global ascendancy, amid a surge of interest from the political class in quantitative analysis. As Lepore notes, the primary preoccupation of social scientists in the United States during the Cold War era was, in essence, “How do voters in a democracy form their opinions? And how can the influence of communism be stopped?” A company like Simulmatics was perfectly positioned to take off—particularly with Greenfield’s personal connections—regardless of whether it truly delivered what it promised. (Or as Lepore puts it, “There’s a lot of puffery and nonsense in the archival trail left behind by flimflam men.”)

As a result, If Then is somewhat less about the People Machine than it is about the people who built the machine, who were all, in various ways, elites (if not outright masters of the universe) of their time. In fact, the Simulmatics Corporation itself doesn’t make its first real appearance in Lepore’s book until about 100 pages in, after a lively guided tour through the cast of characters that would come together to helm the company. Greenfield, the main force behind the company, was an outgoing, well-connected public relations mastermind who “collected people,” Lepore writes, “the way other men collect comic books or old stamps or vintage cars.” That particular talent led him to rub elbows with the academics, programmers, and politicians who would become enmeshed with Simulmatics. Greenfield recruited Eugene Burdick, a surfer, a political scientist, and the novelist who would go on to write The Ugly American as well as a popular, thinly veiled novel about Simulmatics called The 480 (a reference to the People Machine’s 480 voter types). There was Ithiel de Sola Pool, a onetime communist turned patriotic behavioral scientist who would involve Simulmatics in the Vietnam War and help found MIT’s political science department. And there was Bill McPhee, a moody, occasionally violent alcoholic who would come up with part of the design for the People Machine from the confines of a mental institution.

Lepore’s exceptional skill as storyteller and her sharp eye for seemingly quotidian details and small coincidences lend the Simulmatics world an intimate—and at times deliciously gossipy—feeling, which serves to underscore how tightly knit this particular echelon was. (The title of the first section of her book is “Small World.”) The men of Simulmatics summered together; their wives were friends who passed around copies of Peyton Place while their children romped on the beach. Those who worked out of the company’s Cambridge, Massachusetts, office lived down the street from esteemed figures such as the historian Arthur Schlesinger Jr., who was privy to the creation of the People Machine, and the politician Daniel Patrick Moynihan. Even public criticism of the Simulmatics project often came, so to speak, from inside the house. Writer Thomas Morgan, who would later publish an exposé in Harper’s of the company’s involvement in the Kennedy campaign, was himself a personal friend of Greenfield’s, and had even edited some reports for Simulmatics. (The Harper’s article led to a brief public outcry over the company’s attempt at electoral manipulation, but, as Lepore writes, Greenfield didn’t believe in bad publicity. Indeed, just a few months after the initial controversy, Simulmatics had a new slate of corporate clients, and Greenfield had hired Morgan to lead the company’s public relations efforts.)

It was Greenfield’s connections to members of the Democratic Party leadership that had enabled Simulmatics to consult on Kennedy’s 1960 campaign. The proximity of the company’s inner circle to power would enmesh Simulmatics in Vietnam a few years later. Academic studies by Pool and a few other staffers had provided an early blueprint for the U.S. Army’s Project Camelot, a counterinsurgency program meant to predict (and quash) political uprisings in foreign countries. On the eve of the Vietnam War, Pool was consulting frequently for the administration: “He’d sometimes bring his young son with him when he went to 1600 Pennsylvania Avenue; he’d park him in the Oval Office,” Lepore writes. “Adam would sit in Lyndon Johnson’s chair and pretend to be the president.” Pool’s relationship with Secretary of Defense Robert McNamara eventually helped Simulmatics land a Pentagon contract to assess the effectiveness of the military’s counterinsurgency efforts in Vietnam—the program of winning the “hearts and minds” of the Vietnamese, as McNamara put it—through interviews and drawing up psychological and behavioral profiles of the Vietnamese.

Though Vietnam would be Simulmatics’ most profitable venture, its work there produced virtually nothing—a metonymy, perhaps, for the entirety of U.S. involvement in Southeast Asia. None of the American staffers working in Saigon seemed willing or able to learn Vietnamese, and the company subsequently relied on local interpreters to help them conduct surveys of Vietnamese citizens, military personnel, and ex-Vietcong soldiers. Failing to explain nearly anything about the project or its goals to those same interpreters, the company’s interviews often floundered or ran off track. Furthermore, some of the managers tasked with overseeing operations spent more time on building lavish facilities than on the research itself, and in 1967, after “allegations of negligence, malfeasance, and even fraud” and little usable research from the Simulmatics’ Saigon office, an irate Department of Defense shuttered the project and sent the team home.

It was also a time, Lepore writes, “when a rising number of Americans had grown outraged by the application of behavioral science to a war many Americans did not support.” There’s an anecdote from the era that, while possibly apocryphal, captures both the government’s misguided reliance on data and the public’s growing skepticism toward it. That year, Lepore writes, Robert McNamara’s staff gathered all the data they had about Vietnam from their various sources—everything from the numbers of troops, deaths, and hamlets in the country to the price of rice and “the density of the peasant mind”—and fed it into a giant People Machine–like computer at the Pentagon. “When will we win in Vietnam?” they asked the computer. The machine crunched the numbers and returned an answer: “You won in 1965.”

And yet, when Simulmatics returned home from Saigon to a country defined by its own percolating unrest, the company received new offers to consult on a number of Great Society initiatives, including the now-famous Kerner Report. Daniel Patrick Moynihan—Pool’s neighbor in Cambridge and the author of 1965’s “The Negro Family: The Case for National Action,” better known as the Moynihan Report—also joined Simulmatics as a consultant and landed the company a job in Rochester, New York, where it was tasked with using its technology to predict upcoming race riots. “Simulmatics aimed to solve the ‘Negro Problem’ in American cities by way of a new simulation, not a People Machine, but a Riot Prediction Machine,” Lepore writes. In Rochester, a nascent version of that machine ended up producing “an unqualified prediction of violence to occur about 11 P.M. on Sunday night July 23.” (It’s not clear if that riot ever actually took place, though Simulmatics insisted it had, and that the presence of police had defused it.)

In the end, Simulmatics was, above all, a Cold War–era symbol of the wide expanse between the experts tasked with solving so-called social problems and the people living them. The Simulmatics men believed that sweeping changes like civil rights at home and democracy abroad could be achieved, at least in part, through psychological manipulations that they themselves had set into motion. But perhaps that isn’t terribly different from the hopes and motivations of a great many people who work to elect certain politicians or affect social change. “It would be easier, more comforting, less unsettling, if the scientists of Simulmatics were villains,” Lepore writes. “But they weren’t. They were midcentury white liberals in an era when white liberals were not expected to understand people who weren’t white or liberal.”

Simulmatics’ business dried up, and the company went bankrupt in the fall of 1970, a casualty, in a sense, of the post-1968 rupture in the country’s collective faith in authority. But if artificial intelligence technology since then has come a long way, the desires and aspirations of politicians and businesses today are still much the same as those articulated half a century earlier by Ed Greenfield’s company. Four years ago, in one of the most momentous elections of a generation, the tech behemoth Facebook found itself embroiled in a data-related scandal: The firm Cambridge Analytica, a kind of modern-day Simulmatics bankrolled by right-wing billionaire Robert Mercer, had accessed as many as 87 million Facebook users’ data in order to build targeted ads for the Trump campaign. As in the case of Simulmatics’ machine-generated recommendations to the Kennedy campaign, it’s not clear how successful Cambridge Analytica’s efforts were. Yet, it was, as Lepore put it, “a machine that applies the science of psychological warfare to the affairs of ordinary life.”

Cambridge Analytica may have been the most obvious data villain in 2016, but there was also a different kind of misguided faith in data that spread throughout the Hillary Clinton camp and may have had as much, if not more, of a hand in the outcome of the election. A damning report in Politico revealed that Clinton campaign director Robby Mook, besotted by a number of data models that predicted Clinton wins in Midwest battleground states, had all but scoffed at the idea of traditional door-knocking in states like Michigan that would go to Trump. In a similar vein, in the months leading up to the election, liberal-leaning data journalism sites like FiveThirtyEight, acting as People Machines, analyzed polling data and confidently forecast a Clinton landslide. And it was all topped off on election night by The New York Times’ predictive election needle—a twenty-first–century version of the computer that Greenfield first glimpsed on CBS in 1952—which started the evening as a sleek, confident symbol of Clinton’s assured victory and quickly became an anxiety-inducing weather vane that finally pointed to a Trump win when it was already clear to Americans that Clinton’s chances had evaporated.

Simulmatics ultimately sank into obscurity because it was a product of its time, which is to say unremarkable in so many ways. The company failed, after all, to permanently install Democrats in the White House, to change the course of the Vietnam War, or to end urban riots. Rather, its legacy lies in the political norms it inadvertently foreshadowed. As Lepore notes, after the 1960 Kennedy election, the idea that politicians might use advertising, psychological tricks, or even new technology in order to sway elections in their favor was still shocking to the public. But 60 years later, it’s such an accepted part of American political life that it takes a historian to excavate the moment in time where such notions began to cohere.

“Human behavior does not follow laws like the law of gravity, and to believe that it does is to take an oath to a new religion,” Lepore writes. “Predestination can be a dangerous gospel.” Segments of the American public are still irrational or idiosyncratic enough in their behaviors and political attitudes that their actions can’t be pinned down by an algorithm designed by elites—or perhaps those elites never actually sought to understand them in any significant way. That became clear after 2016, with disastrous results for the Democrats, and for the country. But over the long term, the alternative could very well be worse.