Return of the People Machine
Even a halfway-decent political campaign knows you better than you know yourself. A candidate’s army of number crunchers vacuums up any morsel of personal information that might affect the choice we make at the polls. In 2020, Donald Trump and the Republican Party compiled 3,000 data points on every single voter in America. In 2012, the data nerds helped Barack Obama parse the electorate to microtarget his door-knocking efforts toward the most-persuadable swing voters. And in 1960, John F. Kennedy had the People Machine. Using computers that were 250,000 times less powerful than a modern MacBook, Kennedy’s operatives built a simulation of the presidential election, modeling how 480 types of voters would respond to any conceivable twist in the campaign. If JFK made a civil-rights speech in the Deep South, the People Machine could, in the words of its creators, “predict the approximate small fraction of a percent difference that such a speech would make in each state and consequently … pinpoint the state where it could affect the electoral vote.”
But you don’t hear Nate Silver talking about the latest People Machine forecast, because it was, in fact, all bogus. The simulation—part hucksterism, part hubris—promised a. “The machine sputtered, sparks flying, smoke rising, and ground to a halt.” Instead, the best way we have to actually predict elections is still the jumbled mess that is polling. Because reaching people has become for pollsters, so has the job of figuring out who is going to vote, and for whom. If the polls had been spot-on, Trump would never have been president, and just hearing the phrase wouldn’t make any liberal’s skin crawl. Polling can still sometimes , but the problems are real: In 2020, presidential polls in 40 years, and what was predicted to be a quick win for Joe Biden turned into an .
