Obama's Slim Yet Consistent Lead Will Carry Him to Victory, Unless Battleground State Polls Are Statistically Biased

According to the Real Clear Politics (RCP) Battleground State poll averages, President Obama leads in six of the eight most contested races.

According to the Real Clear Politics (RCP) Battleground State poll averages, President Obama leads in six of the eight most contested races. This matters, as Nate Silver of the New York Times explains because poll averages tend to accurately predict which candidate wins a battleground state, even if the margins are off slightly.  However, there is a real possibility that the polls have systematically overestimated Democratic turnout in several of these key battleground states. For instance, the CBS/NYT/Quinnipiac poll found Obama with a five-point lead in Ohio, but their likely voter model assumes a +8 Democratic advantage in 2012, which happens to be exactly what it was in 2008. Although possible, it seems implausible that the Obama campaign could galvanize enthusiastic support in 2012 similar to the unprecedented level of support it had in 2008. In other words, if battleground state likely voter models in polls have systematically overestimated Democratic turnout, this could indicate a Romney win. In fact, as Nate Silver writes, the only way for Romney to win is if state polls are systematically biased.

Based on the polling data, here are a few potential electoral map scenarios. If Nate Silver's forecasting model is correct, we'd expect Obama to essentially sweep the battleground states, except for North Carolina, resulting in a landslide for Obama with 332 electoral votes to Romney's 206.

 

However, based on the polling data there is a very real possibility that Romney could take Colorado, Iowa, and New Hampshire, Virginia, and very likely Florida. Iowa Even the poll averages for Iowa overestimated Obama's margin in 2008 by a share larger than his poll average lead today. And with only a two point Obama lead in Iowa, it is entirely possible Romney could in fact win the state. Florida Although recent polls give Obama an ever-so-slight advantage in Florida, the RCP average has Romney winning the state.  Virginia RCP indicates Obama's tenuous lead in Virginia with only a +0.3 percentage lead, suggesting Romney could very realistically take the state. New Hampshire polls pretty consistently find Obama the winner, except for an ARG poll which found the candidates tied. However, the Romney campaign's headquarters are just a few short hours away in Boston, MA and could arguably mount a formidable GOTV effort in New Hampshire on Election Day. Colorado is a right-leaning state, but that swung for Clinton in 1992 and Obama in 2008. Moreover, polls have shown vacillation between the two candiates.  Nevada It is unlikely that Romney will win Nevada, despite it's swing state status and previous Republican presidential voting history. In 2008, poll averages underestimated Obama's actual vote margin over McCain by nearly five percentage points and Obama currently has about a 2.8 point lead.

Nevertheless, even Romney winning these additional swing states still ensures an Obama victory 271 to 267 electoral votes.

Ohio If Romney were to win the electoral college he would have to win one additional prized swing state, Ohio. RCP has Obama with an average 2.9 point lead suggesting Romney may fall short of winning the presidency. However, if the Ohio state polls' likely voter models fail to reflect actual turnout, Romney could win the state.

An Ohio win for Romney also makes winning other smaller battleground states less crucial. For instance, Romney could lose Iowa, New Hampshire, Nevada, and Wisconsin and manage an electoral victory of 275 electoral votes.

If Romney loses Ohio, it most likely spells out a win for Obama. However, there are several other paths for a Romney victory. For instance, if Romney sweeps nearly all the other swing states, except Wisconsin, he could lose Ohio and pull ahead with 273 electoral votes. However, a Nevada win seems incredibly unlikely.

There are a few game changers; one would be a Wisconsin win. Romney's vice presidential pick Paul Ryan heralds from the Badger state. Ground game infrastructure is also still intact six months after Republican Gov. Scott Walker withstood a fierce recall election after he tried to reform public employee unions in the state. However, Obama continues to lead in the state with a 4-point RCP average advantage, while only Rasmussen has found the race tied. Also, Wisconsin rebounded faster for Obama after his Denver-debate debacle. Romney would also have to pick up New Hampshire or Iowa in combination with Wisconsin to withstand an Ohio loss.

Another remote possibility, but certainly a game changer would be a Pennsylvania win for Romney. Pennsylvania, like Ohio, is electorally rich. However, Pennsylvania hasn't voted for the Republican presidential candidate since 1988. Clinton effectively turned the state blue, and Republicans haven't been able to flip it since. However, a Tribune-Review/Susquehanna poll of 800 likely voters found the race tied and RCP has Obama's average at +3.8 points. We've also seen the Romney campaign ramp up advertising in the state. If there were nothing for the Obama campaign to worry about, they wouldn't have sent out campaign surrogate Bill Clinton to the state in response. But as David Axelrod has said he'll shave off his mustache if Obama loses Pennsylvania, Michigan, or Minnesota on Election Day.

 

Comments

[John]

Obama's Slim Yet Consistent Lead Will Carry Him to Victory, Unless Battleground State Polls Are Statistically Biased

That is really the whole question isn't it? The state polls don't jive with the party identification polls or the polls that give Romney a big lead among independents and virtual lock on Republican votes. It is really pretty simple math. Are there enough Democrats to make up for Romney's lead among independents? If you believe the sttate polls, there are. If you believe every other poll, there are not. It is really that simple.

[robc]

And it gets really tricky if the answer is somewhere in between.

Thats where we get the win the popular vote, lost the EC type results.

[John]

Before the 2000 election, the thinking was that Bush was going to win the popular vote and lose the EC. It turned out the opposite. Wouldn't it be funny if that happened again?

[Joe M]

Jibe. The word is jibe.

[Raven Nation]

Thank you.

[Enough About Palin]

I would contend that either works works as state polls don't jive.

http://www.youtube.com/watch?v=rxvQ6yNGIes

[Emmerson Biggins]

racist

[Omni]

Unless you believe that many R's have switch to I's. This is the explanation that was pushed on me as to why the D+ gap is so big and why Romney has a lead among independents. It's basically an illusion, less R's leads to a larger D+ sample with the same group, and the still vote republican anyways, so there's the lead amongst independents.

[bendover]

Man this election is nuts - couple of thoughts:

I noticed all the national polls that were 4-5% for Romney went back to even. Is this a sign that pollsters are just trying to cover their asses - either way the vote goes, they could explain it away as being "in the margin of error"

How can the independent vote in Ohio be double digits for Romney, and yet some polls have a 5% Obama lead. Hasn't Ohio always gone the way the independents break?

[bendover]

Continued ...

Finally, going back to those polls that showed Romney ahead 5% nationally - but behind in Ohio. I also thought that voting patterns in Ohio mirrored the national vote. If so, then why the 5-8% spread in voting patterns?

[Mo]

One theory I saw was that there was a big jump in Independent identification in 2010 that was concurrent with a sharp drop in Republican identification. Since it's highly unlikely that the electorate moved to the left that year, the theory is that a bunch of Tea Party Republicans just started calling themselves Independents without changing the makeup of the electorate.

[BarryD]

That may be totally true.

But what does it mean for 2012?

[Mo]

If it is true, then it means the focus on who is winning independents (not to mention partisan splits) is not a very useful metric.

[BarryD]

Except...

If one assumes that nearly all R's vote R, nearly all D's vote D, then it really matters who wins the I's. It matters none what the I's were 10 years ago.

Of course, it also matters whether R's bother to vote, D's bother to vote, and I's bother to vote (and which I's).

Bottom line? You can't tell what will happen, from polls.

Polls are useful to campaigns, but not so useful to the rest of us.

[A Serious Man]

Nate Silver's hubris would be icing on the salty-ham tear cake of liberal pundits if Obama loses.

[John]

It is a case of garbage in garbage out. Silver assumes that the state polls are right and that no polarized race can ever shift on election day. If those assumptions are true, he dingbat model works. If they are not, it doesn't.

[Auric Demonocles]

Did you know if there are 1.4 Democrats for every 1 Republican, the Democrats win the election? It's amazing!

[BarryD]

Given that Democrats would vote for someone caught on film raping babies, as long as he told them they would get free shit if they voted for him...

[protefeed]

Yeah, Nate Silver's website would be toast if Romney wins it after Silver projected a 91% chance of an Obama win, with Obama projected to win almost every swing state.

He's gonna look like a goddamn fool if Obama doesn't win, and win big.

[John]

But it won't matter to him personally. He has a cult of followers. Saying "Nate Silver" has become a way for stupid people to appear smart. If Romney wins 300 electoral votes, Silver's idiot followers will still be saying how he nailed the election this time.

[sloopyinca]

I doubt it. More likely they will be calling for investigations into mass voter fraud because their polling just has to be right. It's the same as the climate alarmists. Their "adjusted" data must be right, even though they hide their methodology and shun any peer-review that scrutinizes it's findings or uses raw (read: real) data and comes up with totally different results.

Remember the polls that did not ask party affiliation and came up with strong Romney leads? Those were true polls, not massaged to reflect past turnouts. And they should have been used, because if you randomly call 1000 people, you get a truer representation of the populus of an area than if you called 1000 people and disregarded many of them because of how other people acted 4 or 8 or 12 years ago.

Either way, I can't call a winner and won't even try. It's simply too close to call.

[John]

The other thing is that acurate polls are really expensive and hard to do. The media doesn't have the money to pay for good ones so they get bargain basement ones that are really nothing more than a SWAG. They are not totally off the wall. But they are not anywhere acurate enough to predict anything other than a blowout.

[sloopyinca]

Nate Silver missed his true calling. He needs to contact the IPCC, stat!

[John]

I am always skeptical of people who are supposed to be "geniuses" in a particular subject but don't work in that subject. Silver got his start doing baseball analysis. Now if someone could actually predict how a given player is going to play in the coming year, something Silver claims, you would think he would be working for a baseball team making millions. That knowledge is literally worth tens of millions of dollars to a team like the Yankees or Red Sox. Yet, instead Nate is scraping for a living writing about election results. That tells you all you need to know about his ability to analyze baseball or really anything else.

[Mo]

Unless you're a GM, which involves a lot more than predicting player outcomes, you're not going make millions as a SABRE stat geek for any team. Even GMs only make between $500K - $3M.

An assistant GM dealing with just the stats side and none of the negotiations probably pulls in $150-200K tops. My guess is his gig with the Times pays similarly for less effort.

[FD]

I agree with your re the "geniuses" paradox.
However, I'd also caution that swinging the bat is different than pulling the lever.
Voters are not up against an opposing pitcher, which in baseball adds the variability to the equation. Polls are quite scientific today, admittedly they are snapshots, but by yesterday the choices were made. I would guess they will not defer much at all from the predictions. The Repubs are clinging to sudden enthusiasm for Mitt and some unlikely disregard of many, many state polls. The Dems are clinging to the numbers.
The numbers almost always win the day.

[Brett L]

Yeah. Its the same sort of problem. Having enough relevant data to train an internally consistent model, then do an external test on a different dataset is pretty much impossible. You can't run 2 independent datasets over the same conditions like in traditional physical sciences experiments. If I run two separate sets of experiments over the same parameters, I can use one to train the model to internal consistency and then test its external consistency on the other. You can't do that with election results. Or temperature data. There's only one election (or temperature) for a given time period and you can't use it for internal and external validation.

People refuse to recognize that this is an inherent problem with applying the traditional scientific standards to these problems.

[Mo]

But those same methods are used all of the time daily. It's the foundation of things like predicting the weather and it worked really well in forecasting the path of Sandy, for example. It's also what sports books do all the time.

[Brett L]

Eh. Given a decent pressure ans surface temperature map I would say storm tracking is better than 75% right now. Because there is a large enough dataset to be predictive. You realize we've had between 16 and 28 named storms a year since 2000, even if only a dozen made landfall. That's a big dataset compared to 18 elections since we went to 538 electors. Also, given that those electors are redistributed every 10 years, it means we've had even fewer predictive elections. Add to the fact that this is the first election of a new allocation.... etc. I think you're markedly conflating the amount of data generated and its consistency.

[R C Dean]

It's the foundation of things like predicting the weather and it worked really well in forecasting the path of Sandy, for example.

Various models produced a variety of tracks for Sandy. It happened to follow the European model, which was also the model that showed the most impact and thus got the most publicity.

[T o n y]

Romney has a 9% chance of winning. That's roughly the same chance of rolling a 10 with two dice. People roll 10s all the time. So it can happen. Romney just has to hope that given 100 elections in 100 identical universes, this one contains one of the 9 elections in which he wins. Silver's analysis does depend on polling, so the claim made here is true: Romney wins pretty much only if there is something systematically wrong with the polling. Which is possible.

The real test of Silver's analysis will be how far outside his confidence interval the results fall. Since lots of polls are pretty close, there are opportunities for things to go against the prediction.

[John]

As I said above, Silver is a way for stupid people to think they are smart. Nothing will get people like Tony to give that up no matter how wrong Silver is.

[T o n y]

Are you gonna apologize to the people you're now calling stupid if Silver ends up being about right?

Or are you going to do what you always do, change the subject and pretend you never said the bullshit you said?

At least admit you are far too emotionally invested in this election to be a better judge of the odds.

[Ken Shultz]

Are you aware that stupid people are sometimes right for stupid reasons?

[T o n y]

I'm aware that conservatives always have an excuse.

[Ken Shultz]

An excuse for what, Tony?

For pointing out that you'll never change your mind about anything because you're oblivious to the facts?

John's right, Tony. You wouldn't let a thing like being wrong on the facts get in the way believing whatever you want to believe.

And you're oblivious to the idea that you could be right for the wrong reasons--just like you're oblivious to the idea that you could be wrong for the right reasons.

You're oblivious. That's what being a card carrying member in Obama's personality cult is all about!

[T o n y]

An excuse for everything. It's you guys who live in a fact-optional universe, not I.

If I'm wrong about something the reason is usually pretty granular: I just don't know enough. I strive to be as right as possible--that is, to understand as much about the world as I can. And there are far more interesting questions in this universe than "which American political party should I support?" I settled that a few years back once I got over my Republican phase (age 0 to 8th grade), and my libertarian phase (9th grade), and settled on siding with the people who believe in science, because the other guys don't, making them very dangerous.

[BarryD]

I thought Silver is notoriously wrong.

[Lord Humungus]

http://www.bobkrumm.com/blog/?p=2507

[Lord Humungus]

http://www.bobkrumm.com/blog/?p=2507

[Brett L]

Nate Silver: Whoops. Time to give myself an out if it turns out that I'm totally wrong on this. Although, 300 +/- 58 is a pretty big frickin margin. Nate could still be "right" if Romney gets to 296. Although, I was having this fight the other day about whether or not there was a normal distribution in electoral votes such that one could apply standard error bars. I vote nay on building an internally and externally consistent model based on standard statistical tests.

[Mo]

I think this is his out:

[I]n poker, making an inside straight requires you to catch one of 4 cards out of 48 remaining in the deck, the chances of which are about 8 percent. Those are now about Mr. Romney's chances of winning the Electoral College, according to the FiveThirtyEight forecast.

As any poker player knows, those 8 percent chances do come up once in a while. If it happens this year, then a lot of polling firms will have to re-examine their assumptions ? and we will have to re-examine ours about how trustworthy the polls are. But the odds are that Mr. Obama will win another term.

Granted, at least he's self-aware to know that he has a GIGO problem. Silver is actually pretty upfront about the potential flaws in his models. His superfans tend to overstate it. I give him points for putting his money where his mouth is. Do you think George Will would predict a Romney win in PA if he had ten large riding on it?

[Ken Shultz]

I see all the momentum on Romney's side with formerly undecided voters; Obama's advantage is just incumbency. But that's a big advantage!

It means the question isn't whether to elect Romney. The question is whether to reelect Obama. Most people think Obama hasn't done anything to disqualify him from office. So, I expect Obama to win.

Hope I'm wrong.

[sloopyinca]

It means the question isn't whether to elect Romney. The question is whether to reelect Obama.

Yes, because in America we are legislated to only have two options in each presidential election.

Most people think Obama hasn't done anything to disqualify him from office. So, I expect Obama to win.

Perhaps 25% of the voting age population will vote for Obama. I would hardly say that means most people don't think he has done anything to disqualify him from office. It's just that most people realize their vote either a. doesn't matter because the parties are the same or b. will vote for one of the two major party morons because they think they are as different as night and day. Of course, there are a few of us that believe in c. there has to be a better way than the road either Obamney or Robama want to take us down.

[Ken Shultz]

"Yes, because in America we are legislated to only have two options in each presidential election."

Single Member Districts.

Winner Takes All.

One of two people will win this election.

That's not a misread of the way thing are. That's reality.

Voters are choosing between two candidates.

"I would hardly say that means most people don't think he has done anything to disqualify him from office. It's just that most people realize their vote either a. doesn't matter because the parties are the same or b. will vote for one of the two major party morons because they think they are as different as night and day."

Personally, I think Obama has done more than enough to disqualify him from office. ...but a huge number of people don't. We libertarians need to change those people's minds, so that they see things our way. But that's not the way they see things now.

They don't think Obama has done anything horribly wrong. Not just in elections, but with everything--it's harder to get people to change direction than it is to convince them to stay on the same course.

That's what I'm alluding to here.

That's the advantage of incumbency. I'm not saying that's the way it should be. I'm saying that's the way it is.

[Timrek]

You may be right. And that's why Romney was such a poor choice. He really doesn't differ from Obama on any major issue. Thus he has no footing from which to call Obama out for all of his many failings as president. While Santorum was a horses ass and probably a horrible nation wrecking choice for president; he at least would have been different.

Best case long term scenario...Romney losses, Republican Party splits and the christian right begins to go their own way towards a Santorum led Jesus fest allowing Goldwater style Republican to retake control of the GOP. Rand Paul win in 2016.

[Ken Shultz]

I think he differs from Obama on some important issues--I think he chose not to emphasize those differences.

...and I think that's why.

If convincing swing voters to change course is more of a challenge, then what he should do is try to convince swing voters that he's really not a change of course.

You see the strategy there?

[BarryD]

Libertarians tend not to understand politics or what it takes to win elections.

Chicken? Egg? Hard to say.

Ronald Bailey wrote a great article about it. Today, I believe.

[T o n y]

John I'm no statistician but you really don't know what the hell you're talking about. Nate Silver calculates odds. He is not a clairvoyant. He got into the election business after losing a lot of money in poker. Even the best poker players sometimes lose a lot of money playing poker.

I'll grant you that this kind of amounts to "Nate Silver can't possibly be wrong," and that's fair criticism. But the implicit claim that he's doing partisan work is very suspect. He should be much more incentivized to try to be right (and maintain his cred) than to try to get his guy elected by affecting perceptions or whatever.

[Randian]

If you predict something with a 92% confidence rate, you are as close to absolutely wrong as it comes if that thing doesn't occur.

At some point it crosses over from "oddsmaking" to "a wrong projection".

[BarryD]

Exactly.

92% confidence that an event has occurred is something we can't always have even after we've observed the event occur...

Not only that, someone who can predict events with that level of accuracy ought to be very wealthy. He'd also be smarter than to bet a lot of money at poker.

[anon]

Meh, kinda tired of these articles. Realistically, we'll know who won when/if PA or OH are called for Romney. If PA goes Romney (Unlikely, but possible), it's over.

[FD]

I was trying to reply to Tony, but the refresh kept failing.

"He should be much more incentivized to try to be right (and maintain his cred) than to try to get his guy elected by affecting perceptions or whatever."

I agree with this. And probably inadvertently, it is quite free-marketish! A prognosticator wants to advertise his prescience for future customers.
Which brings me to the hacks on Fox. Dick Morris? Come on, man! How are they making these nutty predictions? Well, they do it with professional smiles. It's more than I can say for the hacks at MSNBC during W's term-2 pre-election polling. Back then Matthews and company were not reaching for mathematical hope. They were just flailing deniers. Until Kerry conceded.
Tomorrow the spin begins again. For both sides.

[Mo]

I think the world would be a better place if all political prognosticators had to put half of their earnings in an escrow account. If they are right, they get paid out, if they are wrong the money goes to charity as a deduction for the employer.

[protefeed]

The real test of Silver's analysis will be how far outside his confidence interval the results fall.

Silver made some mighty bold predictions. For example, saying Obama has a 96.7% chance of winning Wisconsin will make him look like a damn fool who doesn't know what he's doing if Romney wins there.

Maybe Silver is right and a statistical genius, maybe he'll look like an idiot who disregarded the "Garbage In, Garbage Out" disclaimer for poll aggregation.

We'll know by tomorrow.

[HazelMeade]

I do think it is totally plausible that the polls could systematically underestimate republican turnout and overestimate democratic turnout this year.

After all, polls also repeatedly tell us that Republicans are more enthusiastic about voting them Democrats this year. And everyone, anecdotally knows that people just are not enthused about Obama anymore, and that the Republicans are very very enthused about voting Obama out.

I also think that the Tea Party retains a lot more influence than people credit it with. All those people who turned out at the streets at angry rallies havn't changes their minds and they will quietly show up and vote for Romney.

The real question is whether all those young people who turned out in droves for Obama in 2008 will quietly show up and vote for him again.
Has Obama's cult evaporated, or has it just quieted down like the Tea Party?

[Ken Shultz]

I think the real question is whether all those independent minded Tea Party types will stay home, break for Johnson, or vote for Romney.

The daily tracking polls in swing states suggest they're breaking for Romney.

It went from 46 for Obama and 46 for Romney in August to 46 for Obama and 50 for Romney in the latest daily tracking poll for voters in swing state. That means half the undecideds have already broken for Romney in swing states--and the other half still can't bring themselves to commit to the guy that's been in office for the last four years.

[BarryD]

Tea Party types are going to vote Romney.

Ron Paul had some traction there, but Johnson and the LP never have.

The Tea Party may use symbols like the Gadsden and Founding Father costumes, but in the end, it's a rather pragmatic group. I am not sure most know a thing about Johnson.

[HazelMeade]

That said, how do pollsters determine who is a "likely" voter? Is it by asking them, or by doing some sort of modelling of turnbout?

[BarryD]

Both. Sometimes not at all. It depends on the pollster.

Rasmussen has by far the best track record with this, but any pollster is always just one election away from demonstrating his fallibility. 🙂

[Auric Demonocles]

Ekins, who do you let your consistent opposition to alt-text carry you to ruin?

[Tulpa (LAOL-PA)]

Silver is either a liar or doesn't know what the fuck he's talking about. The margin of error doesn't disappear when you average polls.

If the polls are independently conducted, if they are polling from the same sample population, if the questions they're asking are the same, then the margin of error of the average of the poll results will be smaller than the margin of error of an individual poll. (and I could come up with quite a few reasons to doubt these assumptions) But it doesn't go to 0, it goes down by a factor of the square root of the number of averaged polls (weighted for their own individual margins of error). For example, if you average 4 polls with the same margin of error, the margin of error of the average is half that of one of the polls individually...NOT ZERO.

[Syd Henderson]

When did he claim that? I think you're misunderstanding something.

[highnumber]

Good morning! Just strolling by to see if anybody has anything to say about Silver today.

No?