The Silver Standard

It took two days to wear thin the cliché that Nate Silver was the real winner on election night. Starting in September, a wave of speculation challenged the pollmonger's methods, motives, and numbers when he responded to the multiplying reports of a dead heat with an almost insolent certainty in Obama’s chances for reelection. FiveThirtyEight's famous probability of victory percentage, a complex amalgam of various statistical acrobatics, surpassed 91% for Obama by Election Day. His predictions, of course, largely turned out to be correct. Silver foresaw Obama’s easy electoral victory by getting 49 out of 49 states correct—not including his especially correct prediction that Florida was a tossup before eventually going blue—and Obama’s two-point victory in the popular vote.

The buzz about Silver had reached a fever pitch before the election. FiveThirtyEight seductively promised to distill the overabundant election squawk in the mode of what the Google search bar did for the internet. His byzantine analyses may confuse most of us, but that sidebar is all of it jelled together: two simple numbers, one greater than the other. Silver’s ‘vindication’ comes over no particular enemy; his success validated his legion of supporters who based their assumptions on a ridiculously simple data point, and turned out right to do so.

FiveThirtyEight’s hyper-accurate prediction says volumes about the nature of elections. If we ever liked to pretend that an election is a solemn civic engagement, that curtain has been removed. Elections, now as ever, are as objective a numbers game as any Sudoku. The parties have long understood this, but FiveThirtyEight makes it accessible to us. There was talk that the ever-rising percentage of Obama victory was going to keep supporters from both sides at home. It didn’t seem to have that effect, despite a lower turnout than 2008. But can we reasonably expect the public to stay engaged if they think they know how the election is supposed to go?

First of all, it’s worth noting that Silver is not unique by the fact that he built a predictive model; he’s unique because its predictive success justified his outsize profile. To quote National Review’s Jonah Goldberg, who takes an articulate step back from the whole thing: “The truth is that any statistician can build a model. They do it all the time…I’m not saying Silver’s just lucky or shoveling garbage. He’s a serious numbers guy. But so are the folks at the University of Colorado’s political-science department, whose own model…predicted Romney would win, as did many other models.” In other words, the proverbial start-up costs of becoming the next Nate Silver are low: an understanding of statistical manipulation and electoral trends. I predict that in two years, and especially in 2016, the number of oracular quants will have multiplied into a much more crowded population.

What’s not easily available is the soft knowledge that led Silver to build his model the way he did. When the wave of copycats arrives, the first thing that we’ll notice is that they don’t all weight things the same way he does. In an attempt to define their own niches, different people will include some polls that FiveThirtyEight rejects (it won’t be Silver himself; he’s announced that his model will be run by someone else after 2012) and tinker differently with the same numbers. The result will be a baffling cavalcade of opposing predictions, a poli-sci D-Day storming the beach with sooth-sayers. With such a lack of consensus, there’s no way that expectations will be united enough to deprive the electorate of its perceived—and essential—agency.

Even now, we don’t even know how correct Silver was. Confirming a probability with a single data point doesn’t mean much. Was Silver wrong to give winning South Dakota Senate candidate Heidi Heitkamp an 8% chance to do so? All that his model projected was 920 defeats in a simulation of 1,000 elections. Based on the tiny sample size of the one actual election, the outside chance prevailed. You’re still smart to bet against the royal flush. And he’d be quick to tell us that. So the chances of even Silver repeating his success in 2014, while not unlikely, are certainly lower than his new throngs of followers would take for granted. Factor in the generation’s worth of new statisticians that enter the scene, and the likeliest scenario would appear to be complete inundation in rock-solid opinions pointing right at each other. It may very well drown itself out like so many campaign ads.

I also expect Silver’s newfound importance to usher in the next generation of polling. The antiquated methods used by most polling firms—they often involve landlines—require results to be manipulated before being considered usable. This is due for an overhaul. It’s conceivable that a good number of the impending Silver copycats will develop proprietary surveys and deployment that make full use of our digital space, which they’ll hopefully know how to translate into a vastly clearer picture than is available now. The optimistic would expect that a better understanding of the political mood will aid our officials toward better governance. Maybe, though, a more-defined look at ourselves will do what HDTV did to Larry King. Either way, it’s coming.

If Nate Silver is the fortune teller we want him to be, then it should mean something that he’s made the choice to leave the world of electoral prediction. True, he is a man of many pursuits, but politics has been a windfall for the former KPMG consultant-turned-Sabrmatician. Perhaps he knows that it would be best to remove his direct involvement while the removing’s good; perhaps he also sees a future of drowning in a sea of his own acolytes. In any event, we should hope that even if electoral prediction becomes as robust a science as economics, that we find a way to separate or minimize the forecasts from the ideology that is supposed to drive elections.