… I’m as geeky as they come, but I don’t see much point in arguing over which forecasting model is more accurate when we can just hold on for another 46 days, and then we’ll know for sure. The purpose of political science is not to give us a six-week edge on planning the next session of Congress. It’s not that forecasting is useless. It’s that fixating on one out-of-sample prediction is not what this sort of analysis should be for.
Now, this is where a bunch of grad students fly in to tell me that if we can’t predict, we’re not doing real science. OK, but we can predict. It’s just that predicting a single election is not the same thing as scientific prediction. Theory predicts that there is a relationship between various fundamentals and election outcomes. The model confirms this prediction. Out-of-sample predictions usually fall within the margin of error, but sometimes that margin is pretty big. CONT.
Hans Noel (Georgetown), The Mischiefs of Faction
First, they came for the statistical electoral models, and I said nothing because I got Cs in math and didn’t want to look stupid.
— pourmecoffee (@pourmecoffee) September 18, 2014