Nate Silver 538 Blog: Silver Projections are Right, So Get Over It


Nate Silver of the New York Times' FiveThirtyEight blog has been getting a lot of flak from pundits lately for its quantitative approach to election prediction. Granted, there is an over reliance on daily polls by most journalists, but what the back and forth between pro-Silver and anti-Silver pundits leave out is that political journalism is changing either way. Even if Silver's by the numbers election projection political writing is substantiated by an Obama win, this does not mean that narrative based journalism is doomed. In fact it may just give narrative political journalism just the push in the right direction it needs.

Silver's numbers are essentially a combination of national polls averaged together after being aggregated to how correct they were in the 2008 election. Him being correct in 49 out of 50 states last election gives him a certain amount of credibility as does the fact that he works for the Times. He has garnered some notoriety for consistently having good predictions for Obama, with him currently giving Obama an 80.9% chance of winning. MSNBC's Joe Scarborough got involved basically saying Silver's certainty was unfounded, even though technically every percentage less than 100 is a point towards uncertainty. There was then a series of tweets that ended with Scarborough and Silver betting $2,000 dollars on who wins the election, and Dylan Byers of Politico saying Silver could be a one-term celebrity if Obama loses.

Now there have been pieces written in defense of Silver's numbers and demographic focused predictive methods. Just as there have been good pieces on the follies of qualitative narrative journalism for predicting elections. While no prediction method is perfect in election politics, I think I can make a safe bet and say that Silver is probably more effective than predicting based off of the latest gaffe, debate, or supposed internal campaign strategy adjustment. But this does not mean pundit-based narrative journalism is useless.

The one thing largely missing from modern political journalism are in depth looks at candidate's policies. Admittedly, on the surface analyzing a candidate's tax, health care, education, and military plans doesn't seem like the ratings-pleaser that most pundits strive for. But in translating, researching, and espousing these plans there becomes a lot of material for qualitative pundits to make a narrative out of. Whether they are about how specific plans of a candidate will affect you, how taken all together a candidate's plans builds a utopia/dystopia, or why exactly candidates differ on specific parts, there are ample opportunities for qualitative analysis and narrative building out of candidate plans.

So perhaps the prediction market should be given to the statisticians. Qualitative pundits weren't very good at it anyway. A move into more discussion of actual policies would be good for the American decision-making process and having pundits build narratives about it would make it fun as well as educational. Also, this would mean that politicians would have more pressure to implement their policies if we had all grown familiar with them through the election process. Having qualitative pundits focusing on policies would encourage attention to politics to go beyond winning and losing, and toward being effective or ineffective about making society reflect a world we want to live in.