Tonight, Barack Obama continued the 44th US presidency, a second-term president; the first African-American to ever do so.
In one of the most contested elections in American history, a curiously rational voice stood above the punditry. Nate Silver of the New York Times’ FiveThirtyEight blog (which was drawing 20% of the New York Times’ total traffic during election night) tracked national polls, state polls, and numerous other mathematical markers throughout the election. With the aid of mathematical modeling based on demographics, averages, and voting records, Silver put the chances for election firmly in Obama’s corner (at one time reaching 92%).
Silver’s predictions were chided by conservatives all the way to the White House. They doubted his methods, his models, his math. They questioned his data. They decried his “bias.” But Nate Silver was right. Dead right. In 50 state-by-state predictions, Silver accurately predicted 50 of them.
Spencer Ackerman writes about the “nerdiest election” in Wired:
Nate Silver of The New York Times completely reshaped its coverage. Silver steadied the nerves of liberals and rattled the teeth of conservatives, all through a proprietary model of poll aggregation and weighting. Silver, who called the 2008 election with stunning accuracy, sought to do for politics what sabermetrics did for baseball: Factor out as many subjective judgments as possible, to determine who would win the race.
The pundits sensed their rapidly diminishing relevance. The Internet speculated that Silver’s techniques could be the result of witchcraft.
With social science, Silver demonstrated the predictive power of statistics. He almost single-handedly made pre-election punditry obsolete. And he did so with theories and methods sometimes hundreds of years old. He relegated the opinions that are often mistaken for something substantial on 24-hour news channels to a solitary confinement painted with superficiality.
The 2012 election woke people up to the idea that science could touch on politics. It was a surprise, but not to those who knew the basis for such a combination.
Via XKCD
The media (and social media) has largely focused on Silver as a lone figure, as though the man alone holds the power to predict the next most powerful person on Earth. Tweets of Silver’s “wizardry” flooded feeds and emboldened Obama supporters. But Silver is just that, a man. What he represents is far more important.
Silver is not an oracle, he is merely a rigorous statistician with good models. He’s not a “wizard,” he is a conduit for an evidence-based method. Even in the most “fuzzy” seeming data environments, science can sort the signal from the noise. Silver introduced rational methodology into a sphere polluted with wishful thinking and faith. Math cut through the obscuring mist of spin and speculation.
As an aside, if you are stupefied by how statisticians like Silver deal with all that data and still accept their analyses, you have another bridge to cross. The models and math that Silver used to accurately predict the president are tremendously less sophisticated and theory-based than other models in science. If you trust Silver on politics, you should support climate-scientists on human-caused global warming, for example (a point echoed here and here). This is not to say that complexity begets truth. But if rigorous statistics and models are your bar to vault, climate science leaps far higher than Silver’s predictions.
A distrust of science and an amazement at the predictive power of Silver are contradictory. If nothing else, Silver’s statistical victory smashes science denial of all kinds against a wall infused with the mortar of reason.
The tight correlations between the 2012 election predictions of Silver and reality proves that science-based thinking has a place in our public discourse. I think I like how Sean Carroll put it best:
Science isn’t the cold, disconnected lab coat, floating alone near a few bubbling beakers. It illuminates our lives, if only we let it.
Further Reading:
Science-Based Life 
You have to laugh – http://www.webcitation.org/6Bz2BnE8V
To tell you the truth I never really thought of Nate Silver’s “prediction” as a referendum on science and faith, but you are absolutely right. I actually sought solace in his numbers as the night went on. Thank you for this insightful article. Nerds Rule!
“The models and math that Silver used to accurately predict the president are tremendously less complex and theory-based than other models in science. If you trust Silver on politics, you should support climate-scientists on human-caused global warming, for example.”
It sounds like either this is a non-sequitur, or it’s falsely suggesting that tremendous complexity and dependence on theory are good reasons to trust a model.
Accurate, falsifiable, predictions lend credibility to Silver’s model. We can be confident of the validity of climate models to the extent that they can provide the same.
bitbutter makes some very good points. Nate Silver has a very different problem than client scientists. Mr. Silver takes a collection of historical proxy data (polls and surveys) and their correlated results (actual votes) and uses that build a model that predicts the result from current proxy data. Climate science uses large sets of proxy data (tree rings, ice cores) but limited sets of correlated results, and then attempts to model future state.
The fact that Nate Silver can solve his problem well does not in any way reflect, positively or negatively, on climate science.
I was not trying to equate the exact methods and bases of climate science and election polling.
I was trying to make the point that if you value the relatively simple averages and random sampling that Silver employs in his models, you would be astounded at how well-grounded climate science is. What Silver does finds its way into climate science in the statistics, but climate science then advances beyond that, far beyond that, theoretically and experimentally.
Accepting the power of statistics as shown by Silver, it’s only a hop, skip, and a jump to more rigorous science.
More rigorous? I would suggest that the models by Silver with their accurately calculated error bars are a hell of a lot more rigorous than the sloppy climate models which are forever being tweaked and bodged and shown to be based upon arbitrary assumptions and usually utterly devoid of any error bars.
He didn’t become the 44th US President last night. He already is the 44th US President.
Noted, corrected.
Its this sort of information that will ensure that the New York Times goes on for another 161 yeas.
It illuminates our lives lest we are blind.
Your statement that “The models and math that Silver used to accurately predict the president are tremendously less sophisticated and theory-based than other models in science. If you trust Silver on politics, you should support climate-scientists on human-caused global warming” troubles me. An increased level of sophistication tends to bring with it an increased level of complexity. Modern science and mathematics seem to be in general agreement that the more complex a system gets the less predictable it becomes. Predicting whether and how x number of people will answer an A or B question on a given date is massively simpler than predicting something as complex as climate change. To me the simplicty of Silver’s models versus climate change models is a reason for trusting him over climate change scientists. Not the other way around. (Personally, I think the scientists are probably right, but I’m trying to look at the question from a mathematical point of view here).
Pingback: Chi è Nate Silver, e perché parla male… di loro?
Reblogged this on Abetancort's Blog and commented:
Since when the mass was unpredictable? Modern economics has been in the business of modeling the actions of “rational” individuals acting randomly as units but with totally predictable outcomes as aggregate group of such quantity that they conform to what is known as a “Population”.
Pingback: Complexity != Predicitve Power | The Only Winning Move