Silver's model proves science of election

November 10, 2012|By TIM ROWLAND

This was a problematic election for people of my ilk, because the art of punditry might just have experienced a John Henry moment.

Worse, many pundits doubled down, proclaiming that machine could never beat man in the art of tea-leaf reading, taking specific aim at the steam-powered hammer of Nate Silver and his 538 forecasting model, which digests polling and economic statistics and spits out projected election results.

Before the election, Silver was subject to intense criticism from those bosom buddies, the mainstream media and the Republican Party. Both (correctly, it turns out) saw the scientific model as a threat to their job security.

Pundits had hoped that there was still a level of skill, or gut, or instinct in prognostication that a machine simply couldn’t intuit. “An election is not a mathematical equation,” sniffed columnist Michael Gerson, calling Silver’s model “a trend taken to an absurd extreme.”

Conservatives were so offended by what turned out to be the truth that they initiated their own model, which purported to “unskew” the polls, but in reality was just one more parasol for shielding themselves from the sunlight of science.

“Lamestream media” news desks across the country followed suit, peddling a trendy conventional wisdom that Obama would win the electoral college but lose the popular vote.

By now, of course, we all know what happened: Silver’s model correctly forecast all 50 states and the exact electoral count. He predicted Obama would win 50.8 percent of the popular vote, Romney 48.3 — both within four-tenths of 1 percent of the actual outcome. In the battleground state of Virginia, where the individual polls fluctuated wildly in the final week of campaigning, Silver predicted Obama 50.7, Romney 48.7. Final count: Obama 50.8, Romney 47.8.

This isn’t the Farmer’s Almanac hitting three out of five winter storms; this is the laser of science cutting clean lines in hardened sheet metal.

And everything we said going in — that computer models can’t account for the human element — turned out to be true, but we followed that argument down the wrong road. In fact, computer models are more accurate because they ELIMINATE the human element.

They subtract out instincts and hunches, they pay no mind to all those supposed lessons and accumulated knowledge that we wise old political vets claim as our stock in trade. Ethereal things such as “momentum” and other straws that wishful-thinking campaigns grasp at are summarily ignored.

In the business, these buffeting winds are culled as meaningless “noise.”

Following Silver’s model through the summer and fall made one thing particularly clear: This race was decided early, and all those events that we thought were so pivotal — the conventions, the “47 percent” tape, the debates, Hurricane Sandy — were really just so much background noise. They moved the needle for a few days and then the race settled back to where it had been all along.

(Nearly as impressive was a Moody’s projection of an Obama win — with 303 electoral votes — made back in February, using no polling data at all. Instead, Moody’s model was based on an improving economy that has been more evident to the men and women on the street than to institutional America.)

Even attempts by the campaigns to use math against math proved worse than useless. A couple of weeks before the election, Intrade, a stock-market-like website that allows investors to bet on social and political outcomes, suddenly saw a huge bet placed on Romney, and consequently, the “odds” of a Romney win spiked — an event duly reported in the press. It now appears that Romney partisans themselves sank a large chunk of cash into Intrade to, as they would say, “skew” reality.

An interesting and emerging story with similar overtones is that an enormous number of robotically generated tweets during the first debate trumpeted a Romney win. Again, the GOP’s partners in crime, the mainstream media, used Twitter (as opposed to the debate itself) as a basis for its reports of a stunning Romney win.

This coverage came full-circle, and no one fell victim to these artificial news stories more than the GOP itself, which interpreted the self-planted buzz as a sign of momentum and a genuine shift of public opinion toward Romney. It was almost exactly like the scene in “Hunt for Red October” where the Soviet sub is killed by its own torpedo.

For Republicans and media alike, what should be evident is that science is ignored at one’s peril. This is just as applicapable for climate change and trickle-down economics as it is for political campaigns. Gut instinct, ideology, fact-manipulation and wishful thinking are no substitute for the truth.

Tim Rowland is a Herald-Mail columnist. His email address is

The Herald-Mail Articles