Who is going to win?

[DinoDoc]

While we're evaluating election predictors, what's your opinion of the University of Colorado model?

[Ben Kenobi]

This is because he doesn't account for covariance in the states and just multiplies out the polynomial as if each state's outcome is statistically independent from the next. Commenters have been yelling at him for this since 2008. His FAQ answer on the subject, written in 2008, shows that he didn't really understand the issue. It wasn't a big deal in 2008, because Obama had a 100% chance of victory. It's a big deal in any close election.

It's also why Obama is undervalued on Intrade. If you look at say, the 90 percent confidence interval and go until you hit 50 percent, you get a baseline of about 240 for Obama vs about 190 for Romney. Cutoff is Michigan for Obama, and has been for quite some time now.

This is why Romney won't win this election, this is why Romney wasn't going to win this election back in February. Battleground polls mean very little when your baseline is 50 down, and Obama has 240 EVs locked down.

[Tupac Shakur]

Jaguar

A couple excellent posts since I last viewed the thread.

[Uncle Sparky]

You have to admit though that Wang Approval Points has a better ring to it.

Wang Approval Points (or WAP).

[Wezil]

WAD would be better.

Just sayin'.

[Jaguar]

While we're evaluating election predictors, what's your opinion of the University of Colorado model?

(1) It puts itself at a monstrous disadvantage because it uses no current polling data whatsoever.
(2) Economics-based models tend to work worse on future data (out-of-sample data) than on past data (in-sample data.) That's because they're particularly vulnerable to overfits and other statistical trolls.

Get enough monkeys on keyboards and a few will appear to type coherent words. Get enough economic variables to peruse, and a few will a appear to successfully predict elections.

There might be some usefulness in taking that idea, cleaning up statistical trolls if necessary, and seeing whether a state's individual economic circumstances are actually relevant to incumbent party vote or not. But if you just want to know who will win the election, look elsewhere.

[DinoDoc]

Wang Approval Points (or WAP).

It is the statistical measurement of the future. Much more reliable than JAP.

[-Jrabbit]

This thread has taken a bizarre turn toward relevance since last I looked.

[Asher]

It is the statistical measurement of the future. Much more reliable than JAP.

JAPs haven't matter since the 80s.

[DaShi]

This thread has taken a bizarre turn toward relevance since last I looked.

Horrible, isn't it.

[Tupac Shakur]

You're welcome.

[Egbert]

As a sidenote, HC, what the **** is with this comment?


When did I even make such a comment? I don't know or care what US voters care for.

My arguments for gay marriage are grounded in humanity and rationality, not US voting demographics and historical trends.

Your reading comprehension skills rival your compassion.

Your arguments for gay marriage are grounded in your inherent gayness, actually.

[Kidlicious]

Your arguments for gay marriage are grounded in your inherent gayness, actually.

Check!

[Dinner]

While we're evaluating election predictors, what's your opinion of the University of Colorado model?

Unless you are a drooling idiot, which I believe you to be, then you already know that "model" is no such thing. In fact it has not predicted anything because it is an entirely new model created just this year. It is, in fact, a model which has been proven wrong in absolutely EVERYTHING it has been used on since it has been created. In short, no one with half a brain gives a **** about that model.

[Tupac Shakur]

Jesus Oerdin, you're such a ****ing ******.