Sunday, November 4, 2012

Polarization in Prediction


The ability to predict the outcome of future events is a well storied pop culture trope. And it's understandable why precognition is such a popular storytelling element, surfacing in history from MacBeth to That's So Raven. The world is an unpredictable, frightening place and knowing what will happen next gives order to it. This base level  human desire to see the future and use this knowledge to your own advantage is endlessly appealing and is a major factor in the rise of statistician and predictive political polling in the US. The use of polls to gauge political leanings is certainly not a revelation but the increased focus on formulas and demographic data has changed the way races are predicted. The sweet science of predicting who the next president will be has become much more of a personalized numbers game. In theory this would lend itself to more accurate predictions among the various prediction services. In actuality polling and prediction has become as mired in political polarization as the races themselves.  

A Number's Game

"Numbers never lie." This phrase, repeated as an endless mantra by my middle school Pre-Algebra teacher, should be fact. Numbers can never lie. But numbers can be skewed, manipulated and distorted for political motives. Leading the charge for stat based prediction is the New York Time's Nate Silver. Silver is not a pollster in the typical sense, he doesn't call potential voters or even interact with voters at all. All of Silver's predictions are a result of a complex formula that crunches millions of numbers to find the most likely outcome of an event. This method of prediction first gained attention after Silver used it to correctly predict the winner in all 35 of the 2008 Senatorial races.

This approach is far from perfect. Although Silver uses the results of dozens of polling sites and the stats of past races at the end of the day, he controls the formula. Silver can choose to lend more weight to a particular poll or factor in greater state polls if he chooses to do so. The best way to illustrate this point is by underlining the disagreement between Nate Silver and Scott Rasmussen. 

Pundits and Polarization

Scott Rasmussen, cofounder of ESPN and President of Rasmussen Reports, is one of the most highly regarded men in the political prediction industry. So when Nate Silver criticized his results as having an inaccurate and  conservative bias, people took note. Silver accused Rasmussen of using shoddy data and ineffective methods to garner results. Mr. Rasmussen predicts races in a much more traditional sense, he utilizes an army of employees who make constant polling calls, but his methods are still incredibly number reliant. America is becoming more diverse than ever and polls must adapt to account for this. After years of gross inactivity the youth voting demographic is slowly becoming more politically involved and the Hispanic and non-english speaking voting demos are also gaining in numbers. To show a true picture of America, a pollster must account for these changes and factor their numbers slightly to reflect this. 

Two Men, Two Far Different Results

One example of how two different men can produce such different results can be seen in the November 4th prediction results for Rasmussen and Silver. Rasmussen predicts a perfectly tied race while Silver predicts an 85% chance of an Obama victory. These are vastly differing numbers, not even in the same ballpark yet both can be backed by detailed and exact methods. And the battle between the two men is far from over, partisan lines have been drawn in the sand. NBC pundit Joe Scarborough recently called out Nate Silver saying that “[a]nybody that thinks that this race is anything but a tossup right now is such an ideologue…they're jokes.” Both men have come under fire from the other side and their own personal allegiances have only made it more intense. Nate Silver's blog, FiveThirtyEight is hosted by the New York Times, a paper often accused of liberal leanings and Scott Rasmussen is a consistent contributor to Fox News. This begs the question, do people seek out polls with a history of accuracy or polls that will only confirm their own preconceived biases? The disputes in accuracy between stat based polls and traditional  polls will only increase as time goes on and political punditry becomes even more of a big business. The age of easy access to voters and rise in number crunching prediction should have eased the variances between polls but instead has done the opposites. Numbers and polls, previously trusted to be infallible, have been corrupted with the stain of political bias. November 6th may mark the end of a strenuous campaign cycle but it'll only mark the beginning of another stress cycle for those who stake their credibility and livelihood on the accuracy of prediction numbers.,0,1129852.story

No comments:

Post a Comment