re: #36 JamesWI

LOL

The models aren’t all over the map. There’s A model, which simply looks at the economy and says “Hey, the economy isn’t great under this President, people probably won’t vote for him.” And then there’s all the other models, which actually look at what voters are actually saying.

Another bit is that the University of Colorado’s model is sold on a false premise, namely that it’s correctly “predicted” 8 elections. No, what they did was build a model, use it to correctly predict one election, then apply the model to past elections, taking correct “predictions” as vindication of the model.

By contrast, Silver’s model has been used in two elections and been 80-90% accurate. And it’s built on actually looking at past polls and running the numbers, instead of trying to fit a model to the numbers.