Thursday, November 06, 2008

Football, presidential politics, and science

Frankfurt Galaxy Football Helmet
Image Credit:
NFL Europe

Monday, I watched the first part of the Monday Night Football game between the Pittsburgh Steelers and the Washington Redskins. Near the start of the broadcast, an interesting fact was mentioned: In 16 out of 17 instances, if the Redskins won their final game before a presidential election, then the incumbent party won the presidency. If the Redskins lost, the other party won. The one exception was 2004, when the Redskins lost but Bush won re-election. So, you can actually rephrase the fact to say that, in 17 out of 17 instances, when the Redskins won their final game before a presidential election, the party that won the popular vote in the previous election won the presidency in the new election, and if the Redskins lost, the party that lost the popular vote in the previous election won the presidency (this is because Bush lost the popular vote in 2000). And, since the Redskins lost on Monday and the Democrats won the election, the Redskins are now 18 for 18.

At first glance, this sounds like it can't be a coincidence. A quick calculation makes it look like the odds of this happening are 1-in-260,000. So, how can the presidential election possibly be influenced by the results of a football game?

The answer is: it's not. There's no relation. The Redskins football game does not influence the election. And while 1-in-260,000 sounds like long odds, with millions of people looking for silly ways of predicting the presidential election, some random occurrence with small odds is almost certain to appear! This doesn't mean that the two events are related.

There is a fancy Latin phrase for the belief that, because two things happen in sequence, there must be a link between them: post hoc ergo propter hoc. An example of this is the famous story of the rooster who noticed that the sun rose every day after he crowed, and so he assumed that his crowing caused the sun to rise. One day, the rooster overslept, yet the sun rose anyway.

There are indeed related events. If I aim a gun across the room and pull the trigger, and an instant later someone across the room falls to the ground with a gunshot wound, there's a pretty good reason to suspect that the two events are related. But there's a difference between the rooster crowing/sun story and the gun story: in the case of the gun, there's a physical and testable link between the two events. In the case of the rooster crowing, there's just no logical explanation for how a rooster crow could cause the sun to rise.

In astronomy, we are always in a battle to understand cause-and-effect, and to determine if observed trends and correlations are real, or just chance. For example, a supernova (exploding star) happens in a typical galaxy every 50 or 100 years. In 2007, a NASA satellite detected two supernova explosions from the same galaxy, only two weeks apart! If we knew nothing else about supernovae, we might think that, somehow, the explosions were linked. But they aren't -- they're thousands of light-years apart. If you calculate the odds, the chance of seeing a second supernova in a galaxy within two weeks of a first one is about 1-in-1000 or 1-in-2000. That sounds low. But, we are now looking for supernovae in thousands of galaxies, so the odds of seeing two closely-spaced supernovae are actually pretty high.

My point is (and I do have one), just because two things happen close together, and the odds seem like it cannot be chance, it still can be chance. If you want to think that one thing causes another, try and think of a reasonable explanation on how to connect the two. If there is no reasonable explanation, it's probably chance. Sorry, Redskins fans, but your team doesn't determine the outcome of the election.

No comments:

Post a Comment