Languish Presidential Election!

[DGuller]

Powell seems to have destroyed his legacy on both sides by failing to stop the Iraq war on one hand and then endorsing Obama twice on the other. He should have quietly retired after the Army or run for President in 1996. Powell defeated Clinton 50-38 in a hypothetical match-up proposed to voters in the exit polls conducted on Election Day.

Winning hypothetical match-ups is easy.  Campaigning is hard.  Powell may be an honorable person, but he's no politician, as his stint with Bush's administration proved.

[OttoVonBismarck]

Something I've been noticing is a lot of people are ascribing a sort of scientific accuracy to polling averages and things like InTrade that aren't really historically that valid. InTrade is too new to know how it'll do over say, a century filled with occasional close elections or surprise results.

Averaging polls does not (mathematically) remove the margin of error from those polls, they are still just as error prone.

I say this only to point something out, because of how polling has happened in this election there is a genuine chance of say, Obama winning every toss up state and going 347-191.

There is also likely to be at least one or two states where something happens unexpectedly. As an example going by long term history, it isn't unrealistic a state like North Carolina where Romney's average lead puts him up +5 you could see Romney actually lose North Carolina. Or Michigan even +4 for Obama, historically the polls have been that wrong before and it could break for Romney.

Now my projected map and ultimate election results I think are a good guess based on the polling available to us now. (For Romney what especially hurts him in Ohio is he trails by over 10% in early voters who have been exit polled, so he actually would need a decent portion over 50% of election day voters to actually win the State.) But people have started to act like just because you average together a bunch of polls or use a market system you are guaranteed to know the election results. History shows polls and markets are both wrong all the time. Most prediction models that are popular are always right until they are wrong, then they get thrown in the dustbin of history and someone comes up with a new prediction model that they can massage to correctly predict all past elections. Then it works for a few election cycles until it fails and they start over again.

The reality is there are 7 swing states where Obama is up by less than 5: Ohio, New Hampshire, Wisconsin, Michigan, Pennsylvania, Iowa, Nevada.

There are three swing states where Romney is up by less than 5: Florida, Virginia, Colorado.

I'm basing my prediction off the best polling we have. But I'll be surprised if all those Obama States actually do go Obama, and all those Romney states actually do go Romney. Most likely we still get a close election but with some surprises.

It's sort of like picking the March Madness bracket, if you just go on ranking all the way you'll do terribly. Upsets will happen. The reason I still project an Obama victory is he can lose a lot of upsets and still win because he's got a nice basket of swing states he has had a lead in virtually the entire election.

[Fireblade]

So I voted yesterday. This shouldn't really be any big surprise, but I voted for Democrats all the way down. I also voted for Arkansas's medical marijuana bill. I never thought the day would come when I'd get a chance to vote for weed.

[MadImmortalMan]

I feel comfortable with these predictions: linky

Basically I think Romney is going to get those 261 for sure, then it comes down to Wisconsin, Iowa, and Ohio.

What I think will happen is sort of akin to how Truman won in 1948 in the electoral college, three states (for Truman, California, Illinois, and Ohio) went for Truman by less than a 1% margin (less than like 60,000 total votes) and won him the electoral college and another term.

I suspect Obama will basically have that happen, those three states will break for him in close elections and he wins 277-261. However I predict unlike Truman-Dewey, Obama does not drub Romney in the popular vote. That will be the closest it's been since 2000 and possibly we'll have the winner of the popular vote be Romney.

If that happens, all complaints about the electoral college will cease.   

[Phillip V]

Romentum can't be stopped. God save us all.

Take heart.  Nate Silver said a couple days ago that the Mittmentum had stopped.

Gallup has Romney back up at +5. In fact five of the six latest polls have Romney leading nationally by a few points. However, since Obama stubbornly clings to 2-point leads in pivotal Ohio, Iowa, Nevada, and Wisconsin, the fear increases that he will win the Electoral vote, but lose the popular vote. Romney still has a crapload of money and 10 days, which can be a long time in politics land.

http://www.realclearpolitics.com/epolls/2012/president/us/general_election_romney_vs_obama-1171.html

[garbon]

I feel comfortable with these predictions: linky

Basically I think Romney is going to get those 261 for sure, then it comes down to Wisconsin, Iowa, and Ohio.

What I think will happen is sort of akin to how Truman won in 1948 in the electoral college, three states (for Truman, California, Illinois, and Ohio) went for Truman by less than a 1% margin (less than like 60,000 total votes) and won him the electoral college and another term.

I suspect Obama will basically have that happen, those three states will break for him in close elections and he wins 277-261. However I predict unlike Truman-Dewey, Obama does not drub Romney in the popular vote. That will be the closest it's been since 2000 and possibly we'll have the winner of the popular vote be Romney.

If that happens, all complaints about the electoral college will cease.   

Wouldn't it just be Republicans who pick up the banner/torch/pitchfork?

[OttoVonBismarck]

I've always liked the electoral college. I'm obviously a Romney voter but I don't care if he won the popular vote by 5% points, if you can't carry the electoral college you shouldn't be President. Federalism is a good thing.

[CountDeMoney]

I've always liked the electoral college. I'm obviously a Romney voter but I don't care if he won the popular vote by 5% points, if you can't carry the electoral college you shouldn't be President. Federalism is a good thing.

Besides, if anything, removing it would actually increase voter participation in non-battlefield states.  We can't have that.

[OttoVonBismarck]

Where the EC is messed up is it represents small states more than it should. Any EC type system by design will do that, but there is no reason to do it as much as we do.

We have 435 House of Rep. members and 538 electoral votes (100 + 435 + 3 for DC.) While an aside, this applies to the House too. There is nothing in the constitution that restricts us to 535 EVs for the 50 states (+3 for DC) and and the 435 House Reps. We basically decided in the middle of the 20th century to lock in at a certain number of House Reps and electoral votes and then just re-divide the pie every 10 years.

In reality what we should do is just stipulate that the smallest state by population should get 3 votes (1 Senator + 2 Reps.) Every other State should get a number of votes equal to its population divided wholly by the population of the smallest state. That would make it a lot more appropriately distributed than it is now.

[MadImmortalMan]

Wouldn't it just be Republicans who pick up the banner/torch/pitchfork?

I doubt it. Bush is still recent enough memory.

[DGuller]

Averaging polls does not (mathematically) remove the margin of error from those polls, they are still just as error prone.

Not true.  Averaging does reduce the margin of error.  The problem is that margin of error is just a measurement of sample error, and that's not the only kind of error that polls have.

Most prediction models that are popular are always right until they are wrong, then they get thrown in the dustbin of history and someone comes up with a new prediction model that they can massage to correctly predict all past elections. Then it works for a few election cycles until it fails and they start over again.

Good statisticians are very much aware of the danger of over-fitting the past data.  It doesn't make them immune, of course, but if Nate Silver is as competent as he seems, then I'm sure he took precautions when fine-tuning the model.  There are ways to combat over-fitting.

[OttoVonBismarck]

Not true.  Averaging does reduce the margin of error.  The problem is that margin of error is just a measurement of sample error, and that's not the only kind of error that polls have.

I did not say it doesn't reduce it.

Good statisticians are very much aware of the danger of over-fitting the past data.  It doesn't make them immune, of course, but if Nate Silver is as competent as he seems, then I'm sure he took precautions when fine-tuning the model.  There are ways to combat over-fitting.

I'm sure they are, but I'll comfortably still say "every election prediction model is right until it is wrong." Many of the popular model makers actually release 3-4 competing models in an election cycle and obviously only publicize their successful model heavily. Nate Silver isn't anything new. Ray Fair has a model that has worked pretty well too.

I've noticed trends like this when it comes to election prediction/modeling and sabermetrics (interesting because there are actually overlaps between the professionals behind both things.) A real sabermetrician will tell you he's doing work that can help predict likely results or performance. An "advocate" then takes it and preaches it as gospel.

Note that all I said here was "don't be fooled into thinking these are perfect models." It's a lot easier relatively speaking to give an accurate % likelihood that a given candidate wins the whole shebang than it is to accurately predict who wins each individual state, and even harder still to accurately predict the final margin in the states. (The Fair Model gets close to that sometimes, but not always.)

I never said these models were wrong or useless, I said that people sometimes think they represent something they don't. I'll note that in response to me basically saying "keep in mind these models aren't perfect, and they break down on the margins sometimes" you come in and basically reject my argument (but not really, you refuted points I didn't make--I never said statisticians don't try to fine tune their models or that you can't reduce margin of error.) But what my argument actually was, that the models aren't perfect and even the guys who create them say that. So you either think the models are perfect and infallible in which case we disagree, or you agree they are not perfect and we have no disagreement here.

[Valmy]

Yeah ok so when are we supposed to do austerity?  During upturns...but nobody wants to rock the upturn...

So the plan is to never tackle the deficit?  It is not like we don't have plenty of bad examples of trying to spend your way out.

[CountDeMoney]

It's sort of like picking the March Madness bracket, if you just go on ranking all the way you'll do terribly. Upsets will happen.

Had a great exercise at the beginning of a statistical analysis course once;  everybody was handed out a racing form for a particular race, and told to best project the Win Place and Shows.
With race histories, jockey records, which horses were mudders and bleeders, track conditions, etc., everybody came up with reasoned, rational projections on which horses would likely have placed in the trifecta.

What the racing form didn't say is how the #4 horse collided with the #3 and #2 horses out of the gate, screwing up the pack, etc.

Who knows;  this whole election may be decided on whether it's raining or not in Maine's #2 district.

[Phillip V]

We will see a big Bradley effect this election where women and minorities fashionably support Obama in public, but abandon him in the voting booth.