Political Predictions: Kicking The Nate Silver Habit

In the weeks leading up to the 2016 presidential election, I, like many of my fellow obsessed Washingtonians, clicked over to Real Clear Politics several times a day. I would scour all the latest polls, checking for a trend in a swing state, deciding which Senate races were worth following. And the last thing I would always do before logging off was glance at the betting odds on the race for the White House. I would reassure myself with the knowledge that Hillary Clinton’s chances of winning during that run-up never dipped below 80%.

You could have gotten very rich betting on Leicester City and Donald Trump in 2016.

But not many of us get rich betting, and therein ought to lie a lesson.

I want to talk briefly about the media and the way it covers elections. But it’s not an easy thing to do. When James Fallows, a media insider, tried to do this twenty years ago in his book “Breaking the News,” the backlash was intense. Fallows’ basic point, which to me seems incontrovertible, is that media coverage of the electoral process has evolved into the equivalent of sports reporting. The balance has shifted away from what a policy would do to how that policy would affect the election. The reporting seems more concerned with telling us who will win than with analyzing the policy impact of a particular appointment. In Fallows’ words, the media now more than ever covers the “horse race.”

I had this mini-epiphany a few years ago while attending a panel discussion about the political press. Some heavy hitters were present. HuffPost’s own Howard Fineman was there. Josh Earnest, shortly before being promoted to the post of White House Press Secretary, was peppered with questions. The New York Times was moderating. One of the main topics of discussion was the stunning recent defeat of Congressman Eric Cantor in his reelection bid in Virginia’s 7th district. Cantor was the House Majority Leader at the time, and he lost to a mostly unknown economics professor. Everyone was flabbergasted.

No one saw this coming, and most of the panel took themselves to task for not predicting the outcome. One panelist pointed out the local Richmond newspaperman actually did write about a possible Cantor loss, and the conclusion that everyone reached was that we needed more local reporting. The panel moved on to another topic.

I wanted to ask a question at this point but time ran out. It is a question that has bothered me ever since.

Why is it the press’s job to predict anything?

Of course, we all enjoy playing the prediction game. Psychologist Daniel Gilbert labeled humans “anticipation machines.” Our brains are hard-wired to make predictions about what will or won’t happen. We plan our lives around it. But as Gilbert points out in his seminal book “Stumbling On Happiness,” despite our best efforts to predict, we are pretty bad at it. We use faulty logic, assuming that the world as we know it today will be the same in our predicted future.

Again, precious few get rich by gambling.

So, aside from providing amusement, what benefit derives from the any media organization predicting what will happen? A friend of mine explained that if he saw an opposing politician predicted to win, it might inspire him to become more active in the campaign of his preferred candidate. But isn’t it just as likely to have the opposite effect? Who’s to say whether such a prediction might demoralize the trailing candidate’s base? Or perhaps, as many Clinton supporters now wonder, did her predicted easy victory cause her fans to grow complacent. The point is we don’t know the impact of such prediction reporting. And since it is wrong so often, why do we even bother?

Predicting is different from polling, but the two are obviously intertwined. There is so much polling now that Fallows’ caution twenty years ago about the primacy of the “horse race” seems more prescient than ever. The best pollsters take great pains to remind us that polls are not predictive – they are merely a snapshot of opinion at a given point. They are always subject to change.

Pollsters have always been valuable to political campaigns and as internal departments, they justifiably provide analysis to their clients. The new model of media polling has tried to provide similar analysis to their consumers. The poster child for this type of reporting has been Nate Silver. This is the first line of Silver’s current Wikipedia entry:

Nathaniel Read “Nate” Silver is an American statistician and writer who analyzes baseball and elections.

Does this strike anyone as odd? Many of my liberal friends swear by Silver’s analytical ability. But I stopped listening to him after the second Republican debate. Silver, using his analytics, had predicted that Wisconsin governor Scott Walker was best positioned to win the nomination. That second debate was the first time I had ever seen Walker in person and he was a train wreck. He was so clearly outmatched by his stronger rivals that, to go back to baseball for a moment, he looked like a Double A prospect who was helpless against a big league curve ball. I have no problem with Silver being wrong about Walker. Back then, I thought Chris Christie had a chance. But a day later, in another panel discussion, Silver double downed on the Walker pick, saying that he was a great bet precisely because people were down on him after the debate. He was a good betting value. Walker dropped out of the race later that week.

These attempted predictions add nothing to the debate. Years ago, when a newspaper or magazine would have a fun little contests amongst pundits to predict elections, those contests would serve as a passing amusement. But they have grown in scope and number, and now they threaten to choke off actual political reporting.

I’ll stick to losing my money in fantasy football.