Thursday, November 1, 2012

Nate Silver - An Example of Polluted Punditry

   Nate Silver, a baseball statistician, accurately predicted the 2008 election. He was correct in 49 out of 50 states. The one state where he was wrong was Indiana, which was the first time that state voted for a democratic President since 1964. So, it is fair to excuse him for that, but it is also important to remember that 2008 was the easiest election to project since 1984. This is a clear example of how polluted the analysis pool is in political punditry.

   Nate has made a prediction for 2012: Obama will take 300 electoral votes and that there is a 79% chance of him winning (NOTE: this measure on his site is active and may change). Nate is a very, very bright man, however, would you want Bobby Fischer to perform your bypass? The point here is that some skills do not apply across the board. Politico posted a good article about how Nate may be a one term celebrity, and even though Nate is going to be wrong in 2012, they probably will be too. The main stream media loves this guy, because he says what they want to hear. Thus, they will continue to promote him as the sage of their beliefs that he is.

   This is a premium example of an uninformed opinion. Has Nick ever ran a campaign? No. Has he ever conducted a poll? No. Does he understand statistical models, proper weighting, or trends? Probably not. So what does he bring to the table? He is well known in baseball for using Sabermetrics. Basically this is the practice of using "objective statistics" to measure the performance of baseball players, and now he is using similar methods in analyzing politics. Part of his system is feeding his database every poll possible. That system in and of itself is flawed. Bad numbers never help yield a more accurate number. Sabermetrics by its function assumes the law of averages over the season of a player. These principles are not applicable to an election that is essentially a one game season. Then there is the keyword "objective statistics." Objective is the rarest thing in politics. A batting average is objective, over a certain amount of attempts, a batter will hit the ball a certain number of times. In a game a batter may face a pitcher four or five times. In politics there is only one election day.

   The problem is not the tool being used (Nate Silver), it is the machine (main stream media) promoting the tool that is polluting what little scientific method there is in the field. Out of emotional self interest, the machine continues to promote the product validating its emotion, instead of producing the best product. This is something important and fun to watch, during the current election cycle, so that accurate assessments can be made in the next election cycle. Unless of course the goal is to be a one term accurate prognosticator in tandem with a one term President.

No comments:

Post a Comment