Poll-Based System Predicts U.S. Election Results For President, Senate 519
An anonymous reader writes "Election Analytics is a website developed by Dr. Sheldon Jacobson at the University of Illinois designed to predict the outcomes of the U.S. presidential and senatorial elections, based on reported polling data. From the site: 'The mathematical model employs Bayesian estimators that use available state poll results (at present, this is being taken from Rasmussen, Survey USA, and Quinnipiac, among others) to determine the probability that each presidential candidate will win each of the states (or the probability that each political party will win the Senate race in each state). These state-by-state probabilities are then used in a dynamic programming algorithm to determine a probability distribution for the number of Electoral College votes that each candidate will win in the 2012 presidential election. In the case of the Senate races, the individual state probabilities are used to determine the number of seats that each party will control.'" You can tweak the site by selecting a skew toward the Republican or Democratic tickets, and whether it's mild or strong. Right now, this tool shows the odds favor another four years for Obama, even with a strong swing for the Republicans.
Must be true... (Score:4, Informative)
Then again there is this one http://dailycaller.com/2012/08/23/university-of-colorado-prediction-model-points-to-big-romney-win/ that has been correct every time since 1980 that says Romney will win big.
problematic Rasmussen (Score:5, Informative)
It's worth noting that this analysis includes data from Rasmussen, a pollster whose track record at predicting election outcomes is marred by a persistent, consistent bias. Not that they're faking the results (as some overtly partisan pollsters do), but their methodology appears to over-represent demographics that are more likely to vote Republican. According to one analysis, they overestimated votes for Republicans by 3.9%. Andrew Tanenbam's web site [electoral-vote.com] has a concise explanation of what's wrong with Rasmussen's numbers, and why he maintains a separate map that omits them from his own Electoral College projections. So if a system that includes Rasmussen data projects that a Democrat is going to win the presidency... that's a pretty strong indicator of which way the wind is blowing.
Re:Must be true... (Score:4, Informative)
Then again there is this one http://dailycaller.com/2012/08/23/university-of-colorado-prediction-model-points-to-big-romney-win/ that has been correct every time since 1980 that says Romney will win big.
That model is a joke. It didn't exist 6 months ago. Just look at it! PA for Romney? It's not even considered a battle-ground state anymore and the Romney campaign has pulled ALL advertising and has none scheduled.
Re:Not possible! (Score:3, Informative)
Fox news tells me that Romney will win 59 states and sweep Obama and his extreme socialism away forever.
Well, Paul Ryan ran a marathon in 43 seconds, so how could his ticket lose?
Re:The person who spent MORE? (Score:4, Informative)
five - thirty - eight (Score:2, Informative)
Nate Silver at five-thirty-eight uses a similar approach. While Silver is openly biased to the Democratic side, his poll weightings seem fair. Currently he is showing about 75% probability of an Obama win.
The Republican problem is that a very successful Southern strategy has hurt them badly in New England as well as among minority voters. I won't argue whether the auto bailout was wise or even successful, I do argue that the bailout is popular in Michigan and Ohio. In fact, Romney has apparently given up on Michigan but perhaps he can salvage Ohio.
Re:Must be true... (Score:5, Informative)
This is the model that passed peer review despite not passing the smell test, right?
Also the model that hasn't predicted anything yet? It simply correctly models past elections back to 1980. That's not quite the same thing.
A lot of it is also crap: The model for example claims a huge multiplier effect for unemployment ONLY for incumbent Dems, incumbent Reps are not affect by unemployment in this model.
Basically they threw bunch of crap of crap together , tweaked it until they got past elections correct... it also required them to be able to selectively dismiss 7 measurements from 8 elections
And not I didn't come up with this analysis, Nate Silver did. http://www.twitter.com/fivethirtyeight [twitter.com]
Re:Not possible! (Score:5, Informative)
Remember Michelle Bachman? Herman Cain? Newt Gingrich? Rick Santorum? John Huntsman?! I'm sure I'm forgetting half a dozen more.
I'm a atheist liberal utahn, some would even say socialist, and I'm here to tell you that John Huntsman does NOT belong with that group of whackos. I even think he'd make a really good president.
Re:How is that a response? (Score:5, Informative)
Nate Silver predicted Obama's win as far back as September in 2008 to within a few electoral votes. I am keeping an eye on him again this year....if he is accurate again then his model will have some strong validation.
Re:How is cutting anything being a Democrat? (Score:5, Informative)
Isn't remotely true. (Score:4, Informative)
Most statistics of this election predict an Obama win. [nytimes.com] If the race would be held today this is what it would look like [realclearpolitics.com] and if you look back, the math has been relatively stable. For Romney to win he'd have to pick up Florida, Ohio, Virginia and then another state besides that [nytimes.com]. Obama has had a fairly good if small lead in most Ohio polling [realclearpolitics.com] and has been slightly ahead in Florida [realclearpolitics.com] and Virginia [realclearpolitics.com]. Also, your model has Romney losing Pennsylvania, which I think is pretty much not going to happen [realclearpolitics.com]. The FiveThirtyEight model linked to gives Obama a 78.1% chance of winning currently and on Intrade Obama futures have given a roughly 60% chance. I think these are much more realistic models than your totally-base-the-election-on-one-thing model.
Every year since 1980 means the model has worked 8 times. In statistics 8 is a pretty lousy sample size.
Re:Well thats a relief. (Score:2, Informative)
I hear that a lot, but nobody seems to ever be able to prove it.
Define proof.
Would Lessadolla Sowers, and NAACP executive, being convicted of 10 counts of voter fraud (voted 11 times for Obama in 2008) and now serving a 5-year sentence be considered proof?
When people admit it [go.com], does that constitute proof?
How about when they plead guilty?" [foxnews.com]
This shit is why in 2008 the Supreme Court upheld Indiana's voter ID law, 6 to 3, and in their decision stated "Flagrant examples of such fraud have been documented throughout this Nation's history" - the fact that you choose to ignore all the proof, and then sit there asking for proof, is proof that you are a Democrat shill.