Why We Should be Impressed by Nate Silver
It is now clear that the calculations Nate Silver made just before the election were a better guide to what was about to happen than Joe Scarborough’s gut or the voices in Peggy Noonan’s head. This is not a high bar to clear. Many other poll aggregators made similar predictions.
However, Silver’s performance this year has been amazing (again). It isn’t because the networks still haven’t decided if his expected Obama electoral votes forecast was too high or too low (they will when they call Florida) its that they haven’t yet concluded that the forecasts he made in July were noticeably different from the outcome (they will when they finally call Florida for Obama).
Silver stresses that he thinks it is important to make predictions about the future — that predicting on Monday what will happen Tuesday is of no use for people trying to decide what to do in July (say how aggressively to use tax shelters July 2012 or whether to invest in bottled water, canned food and ammunition July 2004).
The fact that he got it right the day before the end is, well better than cable babbler standard anyway. The fact that he got it right without having to constantly massively change his forecasts due to new information is much more impressive.
Sticking to a prediction is always possible if one is stubborn (or strategically boastful). Getting it right in the end is possible if tons of information are available at the end. To stick to a prediction of the distant future and get it right one must have precognition. To make predictions which change as little as Silver’s did and get it right one must be on to something.
I don’t think it’s an accident that one campaign trusted Silver’s analysis and used it to win and the other disregarded it.
It remains to be seen whether this is part of their general distrust of data and objective analysis (E.g. See nearby post on GOP rejection of CBO tax/growth analysis, AGW data etc.) or an election specific problem.
To be fair to the GOP I think it’s arguable they did believe the Silver analysis in some sense; it certainly explains the strategy of vote suppression and intimidation they pursued in multiple states.
Tweet from Chris Sacca:
“I’ve crunched all the numbers and it appears there’s a .0004% chance that Nate Silver is going to bed alone tonight.”
I’m pretty sure that many Republicans sincerely believe that there is no such thing as objective analysis — they start from the conclusions and work back to the assumptions and assume that everyone does that.
Many seem not to believe there is a difference between what one concludes and what one wanted to conclude (projection).
There is also the cynical strategic aspect that some people know that no matter how invalid their arguments Ballanced journalists will just report he said/she said, so insisting that something is true will convince people who only vaguely follow the news that it might be true.
But I think there is also a pathological faith in the triumph of the will over data. Last night Karl Rove got into a furious debate with the numbers geeks *employed by Fox News*. I don’t think this could have been strategic.
Nate Silver is such a mega nerd that the possibility that he, who is all grown up, sleeps alone comes to mind. Sour grapes as I am sleeping alone tonight (being married to someone travelling for work and returning tomorrow — yay).
Nate’s electoral analysis was 100% correct state by state. He even listed Florida as the only real tossup state (final odds 50-50 of a win by either candidate). The only real surprise of the night was how long it took the networks to call a number of states with 7+% final margins.
With respect to the cottage industry which sprung up to discredit Mr. Silver, as we know, “facts have a liberal bias”. Better to base your analysis on an apparently feasible set of hypotheses which support your position and ignore all contradictory data!
don’t forget Sam Wang
Mark Blumenthal (Pollster) was 100% correct on the electoral vote as well.
Sam Wang vs Nate Silver.
Near as I can tell the difference between Silver’s final 303 and Wang’s 323 was the allocation of Florida’s 29. Which given the 50/50 a full day later seems to be the equivalent of a pure coin flip. You couldn’t actually successfully attempt (based on reporting I’ve seen) to slide a razor blade between their actual numbers. Wang just chose to put Florida somewhere. And looks like he might get tiny bragging rights over Nate. Hopefully good for a beer or maybe a 24 year old Scotch
Don’t forget this spot-on model: http://votamatic.org/election-day-forecast-obama-332-romney-206/ I’ve been a big Nate Silver fan but many others do very good analysis. Most models that relied on academic, empirical political science studies and/or top-quality statistical analysis of polling numbers were accurate. That includes even the simple poli sci model that Ezra Klein used based on data available at the end of June. http://www.washingtonpost.com/blogs/ezra-klein/wp/2012/11/06/i-admit-defeat-wonkblogs-june-election-model-beat-me/