Why 2016 made a mockery of Nate Silver
The 2008 election seemed like a major political transformation at the time. The economy was collapsing, the nation was blearily realizing that George W. Bush was a historical disaster, and in the race to replace him, a young black man named Barack Obama was beating Hillary Clinton — the biggest name in Democratic Party politics. So when an unknown blogger named Nate Silver correctly predicted both the primary and then the general election, he rode the ensuing acclaim to a perch at The New York Times, and from there to his own "data journalism" site, FiveThirtyEight.
How basically ordinary those days seem now, when a cartoonishly racist businessman with absolutely zero political experience has captured the Republican nomination over the howling objections of the GOP donor class, and the very same Hillary Clinton has had a devil of a time beating back an aged no-name socialist from the second-smallest state in the country. And when it comes to political prediction, this time Nate Silver has fallen completely flat on his face. It's hilarious — but also raises the question: What is the point of data journalism anyway?
Perhaps the best demonstration of how badly Silver's vaunted predictive model has performed is by comparison with literal parodies. Carl Diggler, a character created by CAFE's Felix Biederman and Virgil Texas, began as a parody of DC pundit idiocy and groupthink (among other antics, he was briefly captured by Russian intelligence), but ended up repeatedly challenging Silver to a predict-off. As Texas notes, "The Dig" completely blew Silver out of the water, calling 89 percent of races correctly — as accurate as the official FiveThirtyEight predictions, but including 31 more primaries that Silver didn't venture to call as they were too close, and outright besting him on the big Democratic upset in Indiana.
Another group of Twitter jokesters, the Matt Bruenig Election Team (or MBET), took aim at a more specific failure of the political quants: Donald Trump. While Silver (and his descendants like Nate Cohn at The New York Times) were dismissing Trump out of hand for most all of 2015, the MBET was by late September calling the primary outright for Trump.
Among both these parodies, there is a heavy dose of simply wanting to beat the self-appointed Politics Knowers at their own game, but they also both come from a perspective of genuine concern. What began as a jokey parody of the political journalism format became a devastating critique of a profession that was failing at its most elementary task: simply describing what was happening. Trump was consistently leading in the polls from only slightly after he entered the race, but the data-first crowd — along with the usual innumerate media chuckleheads — refused for months to accept the bleeding obvious.
People interested in politics like to talk about who is going to win. It's fun and interesting, gets the clicks and views, and can segue nicely into more important topics. And as a political writer myself, I can't always help speculating about who will win what contest, particularly when it bears on the prospects for some substantive issue.
However, I have always regarded purely predictive analysis as junk food journalism at best and thinly-disguised propaganda at worst. This election cycle could not possibly have confirmed that belief any more strongly. If all one cares about is the outcome of an election, it's hard to imagine a more worthless journalistic task than simply predicting who will win — first, a simple average of the polls is usually a pretty safe bet, and second, the outcome is going to be loudly announced after the votes are counted. It's like predicting what moment a ship is going to come into port. You could expend enormous effort on timetables, fuel efficiency, weather forecasts, and barnacle fluid dynamics, or you could just wait until the dang thing arrives.
There are a great many ways for journalism to inform and improve politics. I prefer normatively-grounded, opinionated journalism, with an explicit political perspective, but there's room for traditional beat reporting, fine-grained data analysis (though not very much of only that), and even simple hunches based on previous experience. (For instance, the MBET gurus based their Trump prediction on bizarrely extensive knowledge of right-wing talk radio.)
But all but the most stone simple journalism requires engagement with normative questions. I strongly suspect that Silver, understandably horrified by Trump, ended up doing the political equivalent of talking his book, predicting things that he wanted to happen rather than thinking with a clear head. This is the exact same problem with most gut-check political reporters, except they usually camouflage their own views by putting them in the mouth of a taxi driver or some other service employee. It's a constant pitfall of attempting to purge ideology from your reporting — it has a tendency to sneak back in when you aren't looking.