Wednesday, June 29, 2016

Nate Silver's First Presidential Projection: Hillary 79% Chance Of Winning

No, Nate Silver’s Not ‘Predicting’ Who Will Win the Presidency

This hasn't been the best political year so far for pioneering data journalist Nate Silver. Partly because of the hype he earned in 2012 over the accuracy of his late projections, and partly because he shows up a lot of lazy and data-averse journalists all the time, Nate's taken a real pounding for underestimating, like virtually everybody else, Donald Trump's appeal. It could get worse for him if people keep misunderstanding what he is saying.
Today Nate's FiveThirtyEight site is releasing its first "projections" of the presidential general election. That's a regularly updated estimate of the odds of each candidate winning, mostly based on polling averages and state electoral histories. He set HRC's win probability at 79 percent, and Trump's at 20 percent, which seems reasonable at the moment. But then ABC News, a company under the same corporate umbrella as FiveThirtyEight, reported that projection as a "prediction": "FiveThirtyEight's Nate Silver Predicts Hillary Clinton Wins Election Against Donald Trump," read the headline. 
Now, let's say a lot of things go wrong for Hillary Clinton between now and November. Let's say the economy slows down even more; there are a couple of scary terrorist attacks; the former steelworkers of Pennsylvania become convinced that Trump can bring back the 1970s; Bernie Sanders is still nursing a grudge; Clinton's choice of Lincoln Chafee as a running mate doesn't go over real well ... Let's say she's down in many national and battleground state polls. I'm reasonably sure Nate Silver's presidential projection would then "predict" a different outcome. But in some circles he'll be saddled with that earlier "prediction," and maybe even accused of flip-flopping! That's what happens when people don't understand a data journalist's basic methodology.
I admit I'm prejudiced. I actually wrote for FiveThirtyEight back in its prehistoric days, before first the New York Times and then ESPN bought the site and gussied it up and Nate hired other data journalists instead of vaguely empirical scribblers like me. I was mainly useful as a guy who didn't mind staying up late on 2010 primary nights and was reasonably good at figuring out where votes were out and what that might mean. But even back then, a lot of people were jealous of Nate's growing success, and trying to take him down a peg.
Nate Silver can certainly defend himself, but periodic "projections" being interpreted as "predictions" is precisely the sort of thing that can inappropriately discredit data journalism. Let's don't rush to judgment, and for that matter, let's don't try to predict presidential elections this far out.

No comments:

Post a Comment