Reading about the US Presidential horse race has been a past time of mine every four years since Bill Clinton took office in the 90’s. Being a keen follower of American presidential politics, the number to always keep in mind is 538. The US presidential election, like most forms of popular representative democratic governments, is not based on overall popular vote per se.
In essence, the US president is elected by the Electoral College. This institution consists of electors from each state of the US. The number of electors per state varies and is based on the size of the population of the said state.
This time around, Nate Silver, a blogger statistician had laid down his projections for an Obama second term with a stunning 92.2% certainty (as at 10:05pm, the night before the election).
Now, I like numbers. And I like statistics.
What makes Silver’s projections so interesting is that it is flying against most traditional polls that have indicated that the race is too close to call. For example, Gallup’s latest polls on the night before the elections gave Romney a single point lead of 49% to Obama’s 48% well within the margin of error. Most national polls indicate a slight lead by Obama but they are all within the margin of errors of each poll.
What then, gave Silver, the audacity to not only make a call in favor of Obama but to make it with such certitude? Silver’s projection is mainly based on a model that aggregates individual state polls, in the realm of thousands of polls, to determine who will win the most electors in the Electoral College. While traditional polls track likely voters in a cumulative national average, Silver takes thousands of state polls and calculates the probability of who is going to win each state- taking into account also the factors that had shown to have impacted election outcomes in the past. Based on that, he then calculates the probability distribution of the Electoral Collage results to forecast the next president of the United States.
I like the idea that with enough data points, election winners can be forecasted. I love creating mathematical models and love tinkering with such models to see the outcome. However, the field of statistical modeling is fairly new to me and I find it fascinating. Defending his forecast, Silver has laid it all on the table, risking his reputation should he be proven wrong.
Tomorrow, we shall find not only find out whether Obama retains a second term but whether complex statistical modeling will be a mainstay of modern politics.
Update: As at 11:29PM ET on Nov 5, Silver’s forecast of Obama winning has slid by 0.2% to 92%. I still wouldn’t bet against the president, though.
Update: As at 4:00AM ET on Nov 6, Silver’s forecast of Obama winning slid by another 0.4% to 91.6%. Intrade odds moved up 71.3% after hovering at the mid sixties over the past week.
Update: As at 10:10AM ET on Nov 6, Silver’s final forecast (after accounting for all of the state polls) is that there is a 90.9% chance that Obama will win a second term.