[xzins]

Applied retrospectively, the model predicts the correct winner in every presidential contest going back to 1980,

Not to throw cold water on those encouraged by this, but I think that sentence just said that when the election is over, this model is very good at picking the right outcome.

I can do that too.

[Charles Martel]

Not to throw cold water on those encouraged by this, but I think that sentence just said that when the election is over, this model is very good at picking the right outcome.

The statement obviously means that the economic data of the past election years has been fed through the model as though those elections had not yet occurred.

Social attitudes change, but the way that voters react to a crappy economy and high unemployment numbers seems to be very consistent through the years.

[paudio]

Economic models employ various independent variables to predict the outcome. If the model, without any adjustment beyond year dummy, is applied to the data for those exact variables and give similar outcome as the actual (allowing some statistical errors) then I’d say it’s a good model. I’d still like to know how far their predicted values from the actual values are.

[citizen]

That’s not at all what it says.

Go throw cold water on yourself :) Otherwise, you’re just wasting your cold water.

[newzjunkey]

They built the model and tested it against actual elections. Has it be predictive as well? That's unclear but the release suggests it has been used that way in prior elections since the model's been altered since Obama got into office.

Here's the University of Colorado press release