11.03.2012: Nate Silver and Statistical Science

It must be frustrating to be a statistician. I’m no expert in statistics, but I understand the science well enough to understand that statistics are often counter-intuitive*. Unless you’re surrounded by fellow statisticians, the bulk of your time talking about work must be explaining to lay people who wrong they are. Right now, to be Nate Silver, America’s most high-profile and controversial¬†statistician, must be absolutely infuriating.

So it’s understandable that he might act of that frustration sometimes – as he did this week when he challenged Joe Scarborough to a $2,000 charity bet [side note: $2,000? If Nate Silver is worth 1/5 what Mitt Romney’s worth, then the New York Times is doing better than anyone thinks!]. This warranted a scathing critique from Margaret Sullivan, Public Editor at the New York Times, who apparently finds such things as charity bets to be unbecoming a blogger affiliated with the Gray Lady.

Sullivan’s criticism is not entirely unwarranted, though my own feeling is that Silver does himself a disservice by allowing Scarborough to get under his skin. My own totem pole of journalism, at least regarding established outlets, flows Print > Blog > Television, and stooping to Scarborough’s level is beneath Silver. Of course, few in the established press agree with me on that – including, apparently, Margaret Sullivan.

To me, the most interesting thing about her critique is how boldly she comes across as an anti-Silver partisan. Like many election wonks, I’ve followed Five Thirty Eight for quite some time, and what’s most won me over about Nate Silver is how rigorously he sticks to the science. One of the most common criticisms directed at him, and one to which Sullivan lends weight, is that Silver is partisan. It’s true, he has stated repeatedly that he is personally an Obama supporter, but the reason no one should mind that he does so is that his work has nothing whatsoever to do with his personal political preferences. The reason he’s consistently predicted an Obama victory is because that is the outcome to which his statistical models have consistently pointed. Period.

Nate Silver isn’t working voodoo, he’s working statistics – statistics that have been demonstrated accurate, again and again. Any regular reader of Five Thirty Eight can tell you that Nate Silver isn’t one tenth as in love with politics as he is with numbers, data, and statistics. He may personally be partisan, but his work is as non-partisan as it can get – which, of course, is why it’s called science.

It’s tempting here to get into a diatribe about how Republicans don’t like Nate Silver because they don’t like science, and how because Republicans approach everything from a lawyer’s perspective (decide on a result first, and then cherry-pick the evidence that supports your result) they assume this is how scientists must operate. The trouble is, I also read the first few comments on Margaret Sullivan’s critique, which come from pro-Silver partisans, and after reading them I don’t feel it’s fair to pick on Republicans for being illogical. Poor logic, sadly, is an entirely bipartisan enterprise.


* The¬†simplest¬†illustration is to try explaining to the average person that a coin that has been flipped 99 times and come up heads every time still has a 50/50 chance of coming up tails. Try it sometime, and see how enraged many people get at your ‘lack of common sense.’ But then, if statistics were intuitive, I suppose casinos wouldn’t be profitable.


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.