After we had an election up here in Canada in the province of Alberta, people who don’t know very much about polling had lots to say about it. One of the things I fear, almost as much as the other things I fear, about a possible Romney victory is that it would lead people to discount polling numbers. Nate Silver of the New York Times “Five Thirty Eight” blog has made some confident predictions that Obama will win, mostly based on his polling information about the critical swing states.
In Alberta, there was recently a very interesting election. Conservatives probably wouldn’t have had much to say about it, given that the two major parties contesting that election are both right-of-centre – the Progressive Conservatives (moderate Right) and the Wildrose Alliance (conservative Right). Up until the last week of the election, pretty much everyone, including the pollsters, were confidently predicting a Wildrose victory. Over the weekend immediately preceding the vote, however, numerous people switched their votes, and the Progressive Conservatives not only won, but won big. Numerous people in English Canada took this as a cue to start ragging on pollsters. Pollsters, however, knew what was going on. As I noted at the time, Éric Grenier, the respected pollster for threehundredeight.com, caught the weekend shift as it was happening – it was just so lightning fast that it was hard to get a new poll out to reflect the change before the election.
But that’s the thing – even with the best polls, things can happen at the last minute. That, and not lack of respect for the poll analyst’s art, is why I doubt Nate Silver’s confident assertions that Obama has an 80% chance of winning.
Indeed, a lot of the same things that did happen in Alberta could happen here. The reason Alberta ended up electing the PCs is because people were afraid of a potential Wildrose government to the extent that they decided at the last minute to abandon the Liberals and the NDP and – gritting their teeth – vote PC to stop Wildrose. There are a few people – and yes, I’m looking at you, supporters of Jill Stein – who might feel a similar need to grit their teeth and vote Obama to stop an equally scary Republican government under Romney. Of course, the same might be true of the Libertarians, who might be motivated to abandon Gary Johnson and vote Romney to stop the Big O from getting four more. All of that could, also, happen over a weekend.
But, even if I do doubt Silver’s predictions, for the reasons I have just expressed, I will not stand by and watch while right-wing Know-Nothings slag scientific polling and try to replace it with numbers they and their friends have more or less made up. Nate Silver, if he does properly lay claim to the mantle of a scientific pollster, can demonstrate this by being right the percentage of the time he says he will be right. If he says a prediction has an 80% chance of being right and he’s wrong, it’s not because he’s an idiot, it’s because his 20% happened to come up that time. If his 20% comes up every time, of course, then it’s not really 20% – that’s why we care about what his predictions look like over the long haul. But getting one wrong is not a sign of weakness, it’s a sign that “margin of error” is a term that has an actual meaning. It means that occasionally, error exists.
Personally, I will love it, in more ways than one, if Silver turns out to be right in his current prediction. But if the fake-pollsters are right praying to voodoo gods to give them accurate numbers and Silver is wrong using reasonable methods? Well, I’ll just wait for the laws of probability to even that out.