Politico.com is a site that I haven’t mentioned before in this blog, which is unlikely to surprise anybody, as their strong point is politics. Still an interesting read many a times if you follow political news. In a recent article Salon’s Alex Koppelman presents a model that can, rather accurately, predict the outcome of a US presidential election.
Now, I know that, given the interest in this (and previous) elections, all kinds of people have come with all types of models that could be use to predict the result of the elections. Many of them rely on polls (say, Gallup) some on the composition of the population of a given state and then extrapolation (it is said that Ohio is a microcosms of the US). Interestingly the one mentioned in the article depends on the performance of the Washington Redskins on the match prior to the election. So if the redskins win that would bode well for the party already in power, but if they lose, that’d be good news for the opposition.
I am writing this before the results of the election come out and I imagine that there are good chances that you are reading this after those results are computed and made public. The only prediction I could give is that I don’t see Ralph Nader winning this one but, according to our football model, the Redskins lost yesterday against the Pittsburgh Steelers so that would imply that Obama should win today.
The only hope I can give to McCain supporters reading this blog before the results are out (a rather unlikely possibility given time and the political bias among scientists) is that the model this prediction is based on is a non mechanistic model. This means that the only reason people found it interesting is that the coincidence of two events (redskins winning, incumbent staying) is extremely high. The capability of the model to predict the next election is as good as random chance. The connection between the game and the election is rather weak as anyone bribing the referee to change the outcome of the elections would eventually learn.
So why is it that, so far, the redskin model has been such a good predictor of electoral results? My guess is that humans have evolved to make as many connections as possible. If there is a statistical correlation between two events, we are likely to find that useful, even if there is no causation from one to the other.
Now, what makes the connection between the redskins and the elections special? probably nothing. There are gazillions of things that we could try to correlate with the outcome of the elections: the number of rainy days in Columbus, Ohio, the colour of the uniform of the winner of the superbowl…
Most of them will give a negative correlation but, of all the myriad of options there is a (statistically) good chance that you can find one that actually has correlated well in the past with whatever is it that you would like to predict.
Models, be it cancer or political are better if based on actual mechanisms, not only they are useful to shed light on how things work but also are more likely to withstand the test of time.
David, you have no faith in the football gods – I on the other hand had to watch the Steelers/Redskins game all the way to the end just to be sure that Obama would win. I couldn’t take the additional day of not knowing!
Interestingly, it used to be thought that this game determined who wins the election – however in 2000 it was discovered that in fact it only determines the winner of the overall majority vote, not of the electoral college.
Since this turned out to be predictive once again is the model now stronger?
Simon, I do have some faith in the Nacho Guacamole God (which would beat the Flying Spaguetti Monster any given day but Mon-Thu from 23-23:30). If I had to come with my own prediction mechanisms based on thin air I’d probably come with the probability of a certain Manchester team to win the premiership losing at home against Arsenal and then try to correlate that with some political event.
Now that the Redskin model has been vindicated one more time I think I will start playing the game in reverse. If in 2012 Obama is riding high on the polls then I will bet a lot of money on the Redskins winning their game just before the elections. Maybe there’s a psychological mechanisms working here!
I’m a bit worried about the Liverpool gods just now. Losing 2-1 a.e.t. the other day, having been 1 up and the better team. Last night, 1-0 down for almost all the game until a (dodgy? but making up for previous bad decision) penalty let them equalise.
Not that I am a football fan but life is so much more pleasant when they win!
Maxine, I feel your pain, I am not exactly a fan of Atletico de Madrid (who played agains Liverpool yesterday I believe). Still, just imagine if Liverpool losing the other day would have made a McCain/Palin victory inevitable. Luckily, there seems to be no correlation so far?
Good point, David ;-)