If you asked statistician, sabermetrician, psephologist, and writer Nate Silver who the next president of United States would be, he’d tell you with confidence that right now Barack Obama has an 88 percent chance of winning. His number has nothing to do with personal bias and everything to do with simple math, statistics, and predictive modeling. And it would behoove you to trust Silver, in 2008, the 34-year-old statistician predicted the voting outcome of 49 of 50 states in the presidential race and correctly called all 35 senate races.
Silver has come under fire for his statistical projections he publishes at his blog, FiveThirtyEight, under the New York Times name. His predictions favor President Obama, which conservatives don’t like to see. But most of the anger and confusion about the percentages Silver publishes stem from the general lack of understanding for what statistics actually are.
People don’t like to see such slanted predictions when the perception of the race is so close. MSNBC’s Joe Scarborough said of Silver’s numbers, “Both sides understand that it is close, and it could go either way. And anybody that thinks that this race is anything but a tossup right now is such an ideologue, they should be kept away from typewriters, computers, laptops and microphones for the next 10 days, because they’re jokes.” Scarborough would be right if he were referring to statements of certainty made by some hack. But predictive models deal in uncertainty which is something far more useful, but also something harder to understand.
Statistical modeling is a very powerful tool that is used regularly, so to make the claim that it’s nonsense doesn’t stand up. Good models were able to predict Hurricane Sandy forming and hitting the east-coast days in advance, which gave inhabitants the chance to prepare and minimize the damage. Silver’s model does mostly the same thing using the same principles.
When Silver says that Obama has an 88 percent chance of winning, that means there is an 88 percent chance of that outcome, not that he will win with 88 percent of the votes (be they popular or electoral). It takes 270 electoral votes to win the presidency and there are many ways to get more than 270 votes. This means Obama (or Romney) could have an 80 percent chance of getting more than 270 votes and the vote can still be split nearly 50/50 in states like Ohio.
Silver’s model can’t predict what individuals will do, but it can look at the aggregate of voting and then run simulations and test the outcomes of each simulation. Silver looks at the polling numbers, accounts for some factors that contribute to biases, then runs a massive quantity of simulations. Each simulation can only have a couple of outcomes declaring a winner, but there are many ways the situation can unfold. If you run enough tests, Silver is saying that Obama wins in 88 percent of them. This doesn’t mean that Romney can’t win, it just means that Obama is favored by a certain amount.
To put it simply, the race is very close, but Obama is still more likely to win. These two pieces of information aren’t in tension even if it seems like it goes against common sense.
Part of this has to do with how the electoral college is set up. It only take 50.1 percent of the popular vote in a state to get all of its electoral votes. This mean that the winner takes all, and in a couple of states there are only small margins, a couple of points, between who will win all the votes. There are still a number of factors that could swing the final count in either direction. Get-out-the-vote efforts could push Romney ahead in key states. Bad weather could affect total voter turn-out. 80 percent confidence isn’t a guarantee of an outcome—that’s precisely why it’s 80 percent and not 100 percent.
Just because predictive models create probability distributions and not certainty doesn’t mean they should be cast aside and their creators be labeled “ideologues”—I can bet you’d like to take Silver with you to Vegas. It’s just another way of looking at the potential outcome of an event. Conservatives shouldn’t take this to mean the “liberal media” is making a claim of an early Obama victory. And Democrats shouldn’t fall victim to the hubris of assumed success. Silver’s predictions are just numbers—well-informed and justified numbers, but indicators of doubt nonetheless.
The Airspace reminds you to vote