It’s a simple but unfortunate fact that most humans don’t understand probability. It’s not our fault, exactly. Our primate brains are not designed for it. Take the Gambler’s Fallacy: If a coin flip comes up heads nine times in a row, you should bet on tails, right? It’s due, right? Well, no. If you’re educated you probably know the odds are still 50/50. But the multi-billion dollar gambling industry is built on how terrible our brains are at probability.
This is why it’s so infuriating to hear Nate Silver, patriarch and spokesman for data journalism, insist he didn’t swing the 2016 Presidential election.
The REAL Real story of 2016
Since Trump’s win, Silver has done innumerable interviews in which he disavows any responsibility. He’s prone to a judgmental tone about the way people treated polling data, and a blasé, even willfully obtuse response to suggestions that he had a role.
Yesterday, Nate Silver published a long analysis, titled “The Real Story of 2016,” in which he begins with a question: “Why, then, had so many people who covered the campaign been so confident of Clinton’s chances? This is the question I’ve spent the past two to three months thinking about.”
To his credit, Silver does acknowledge the role of polling data, and FiveThirtyEight specifically, but his analysis seats responsibility squarely with FiveThiryEight’s readers, and journalists who “lumped together” their model with other, differing forecasts. There is no recognition that the way Silver and his fellow data journalists represent their findings might be inherently misleading. No recognition that the winding path the election took might have been shaped by a certain hourglass-shaped graphic.
Data journalism changes how our elections work
Data journalism is young, and Silver’s model of compiling and analyzing poll data to produce a single projection has been with us for only three Presidential elections. Following his success in 2008, many (myself included) began treating Silver as some kind of wizard. Revisionist history will likely forget the absolute confidence with which many regarded FiveThirtyEight’s prediction pre-Trump.
In the months prior to the 2016 Election Hillary Clinton’s win seemed predestined. Journalists, talking heads, even candidates themselves tended to treat her as if she were already President. It is highly likely that strategic decisions were informed by that perspective, both within the Campaign and outside.
What might have been different, had people not assumed a Hillary win? Would the Obama White House have moved faster to let the nation know about Russian interference? Would FBI Director James Comey have sent his infamous letter to Congress? Would the Clinton Campaign have invested so much into states like Texas and Arizona, and so little in Michigan and Wisconsin? Would the mainstream press have focused so intensely on Clinton’s email “scandal,” or turned their attention to more intense scrutiny of Trump?
All of this is down to speculation, of course, but it stands to reason that some of the elements that shaped the outcome might have been different. In a few cases, dramatically so. Strategists live and die by polling data, and in 2016 Nate Silver was inarguably the forecast most people followed.
There is an argument that data journalism distorts our understanding of polling data. In October, following the final Presidential debate, polls showed Clinton with a four- to seven-point lead on Trump. That’s a sizable lead, but still a close race. FiveThirtyEight put Clinton’s odds at 88 percent. The Upshot said 89.
Silver’s front page projection is the core of the problem.
In interviews, Silver invariably defends himself by citing FiveThirtyEight’s disclaimers about polling errors. And that’s fair. It’s in the headline of this piece from November 4, and as a regular reader I will attest that this point was regularly included in their reporting. But direct your browser to FiveThirtyEight and you aren’t met by a nuanced analysis. You see this.
The 71 percent forecast comes from just before the election. By that point, it was too late for strategic shifts. A month earlier, Clinton’s odds were near 90 percent.The Upshot employs a similar presentation. Ignore the yellow bar here, that showed up only on Election Day.
“Who will win the Presidency?” asks FiveThirtyEight. “Who will be President?” says the Upshot. Then a simple percentage. No mention of polling errors, no disclaimer about margins. Not unless you click through to more detailed analysis. I can’t access internal traffic reports for either site, but I’m willing a small percentage did so. And that sets aside the way those numbers are reported across other outlets–though in fairness to Silver, that plays to his point about other journalists misrepresenting his work.
Nate Silver understands probability. He doesn’t understand people.
The fact is, the projection model pioneered by Nate Silver and FiveThirtyEight shaped the way Americans, including election strategists and the candidates themselves, understood the Election. Arguably, it was the single most influential factor. Yes, there are many other polls to take apart, and in past decades that’s what analysts did. But having a single projection is much neater, and easier for our primate brains to understand.
85 percent. Good. So Hillary will win.
Statisticians like Silver, who are better than most of us with probability, don’t see it that way. On Twitter, I’ve personally been chastised by people who say people should never have regarded Silver’s projection as “deterministic.” Which, again, is true, but it shows a failure on the part of data journalists to understand how bad most humans are with odds.
Personally, I still like Nate Silver, and I like data journalism. I don’t believe such projections should not exist. In the wake of what happened in 2016, however, I do think serious consideration should go toward how it’s presented. I do think a single projection, presented as a thermometer at the top of the front page, is harmful. It distorts the way people view the election and its likely outcome, and in doing so it shapes the election itself.
I’d like to see that change, but even more, I’d like Nate Silver to acknowledge his responsibility. It’s pretty obvious that when it comes to statistics, Nate Silver’s brain works better than most. The same does not appear to be true when it comes to human nature.