Varieties of Uncertainty

I’m not sure I have the right kind of vocabulary to describe what I’m trying to say here, but when thinking about elections it’s worth distinguishing between two kinds of uncertainty about the outcome. According to Pollster.com’s average, Al Franken has a two point lead over Norm Coleman with 18 percent of the vote going to a third party candidate. Leading is better than trailing, so you’d say Franken is favored. But this polling is perfectly consistent with Coleman winning the election. Sampling error or problems with the turnout model could cause the final result to deviate from the current projection, underlying views could plausibly shift the requisite, and preference for third party candidates tends to get unstable at the end of campaigns. We don’t know what’s going to happen.

The Obama-McCain race isn’t like that. Obama’s up by too much. Plus, the Electoral College seems to favor him. On top of that, any errors in likely voter models are overwhelmingly likely to favor Obama. And many people have already voted, which makes it all the more difficult for McCain to close the gap. We know what’s going to happen.

Except even when we do know, we don’t really know. But not because, as in the case of Franken-Coleman, the survey data is inconclusive. But because the course of events in the world is intrinsically uncertain. A meteor could crash into the planet earth. Or Russia could send tanks into Lithuania. Al-Qaeda could stage an attack designed at helping John McCain. And the public reaction to dramatic events like that would be hard to predict. Even though people talk a lot about “October surprises” it’s in the nature of such things that there’s very little historical data about them that can be evaluated rigorously.