Nate Silver Backlash

Around the time of the hurricane, there was apparently a big discussion about Nate Silver. First there was a Politico blog post about whether if Nate Silver is wrong will he lose all his popularity. There was also comments by Joe Scarborough saying Nate Silver doesn't know what he is talking about.

The responses from Silver supporters were great. Ezra Klein said that reporters don't like Silver because he makes them feel innumerate. Then they criticize him and prove him right. David Roberts and Matt O'Brien had similarly good tweets.

And I want to add my two cents in defense of Silver. In doing so, I need to say that I find the people who attack him to show a lack of understanding of high school probability and intro statistics.

A commenter on twitter said they don't understand what a probability means to a one-off event. That is high school probability. Here is an example: flipping a coin twice and getting two heads has a 25% probability. That doesn't mean it can't or won't happen. It just means it is unlikely to happen.

Here is another example: if the weather forecast says there is a 75% chance of rain, I would recommend you bring an umbrella. It might not rain. That doesn't necessarily mean the forecast was wrong. 

For a while, Nate Silver's blog had the probability of Obama winning as around 75% (now it is 90%). The same thing goes - according to Silver's model, Obama is likely to win.

The poltico story and others seemed to accuse Silver of being overly certain and yet hedging at the same time. Again, they are misunderstanding probability. Silver's model is not a prediction, it shows probability. And probabilities show the likelihood and the uncertainty of events. (After re-reading the Politico piece, it is amazing how stupid it is. Dylan Byers clearly doesn't understand probability at all.)

Romney winning doesn't necessarily mean that Silver was wrong. Silver's model says that based on the data he is using, he thinks there is a high chance Obama wins. But there is a small chance Romney wins. That small chance reflects how the data is unable to predict anything perfectly - that to a certain degree events are somewhat unpredictable.
 
But we should ask, and Silver will do this, what it does mean if Romney wins. Since Silver relies so heavily on polling, it could mean that the polling was wrong. Or it could mean that he was wrong to adjust the polling like he does. In other words, it could mean that his model was right or it could mean his model was wrong.

What makes Silver so popular is not his model so much but how he explains statistics. He explains to everyone what his assumptions are, and how he got to them. You can disagree with some of those assumptions and argue that his model should be different.

But that isn't what these critics are doing. They are saying he is wrong but from complete ignorance of what he is actually doing.

David Brooks has joined this chorus - and as much as I like Brooks on some issues, anything involving math and Brooks says it is witchcraft. He doesn't understand economics and so says it cannot predict anything. And he doesn't understand statistics and probability and so says it can't predict human behavior. With regard to statistics, history shows how much presidential election outcomes depend on fundamentals like the economy as well as how often outcomes are in line with long term polling averages.

All of this backlash is from media types that are threatened by Silver. Scarbarough, Brooks, and others have credibility not by evidence but by gut feelings and high level contacts. Their predictions are often wrong, but they don't lose credibility. That is because they are providing entertainment, not real news.

Silver's method asks us to calm down around so-called big events. The news media blows everything up to be game-changers. And they also say races are nail biters. Both provide entertainment and therefore viewers. Silver however says that the evidence shows these events rarely change polling and that the race isn't as close as the news wants to pretend. So he threatens their model. I don't read any polls or any poll coverage save for Nate Silver. If others did that, the news would have less viewers (and the world would have more informed people).

I'll finish by responding to the original question posed by the Politico story. If Romney wins, will I stop reading Nate Silver. The answer is no, because unlike political pundits who won't look at their predictions critically and ask why they were wrong, Silver will look at what happened and explain it. And that is why I read him, because he is smart and great at explaining what is really going on.