What am I missing?
There’s something peculiar about the election polls. The swing state in almost all the plausible scenarios is Ohio, where Obama has a 2.3% lead in the RCP average of state polls. Romney leads the RCP average of national polls by 1.0%, implying a massive 3.3% Democratic tilt in the electoral college. By comparison the EC tilted about 0.5% GOP in 2000 and about 0.3% Democratic in 2004.
Don’t get me wrong, I do understand how that sort of margin is mathematically possible; Romney could win lots of safe states by huge margins, and Obama could win even more states by narrower margins. But I’m not seeing state level polls that are consistent with such a big EC tilt to the Dems. Take the 4 big states; California, Texas, NY, and Florida. Obama will win two by big margins and Romney will win one by a big margin. Romney will probably win Florida by a narrow margin. The rest of the country has more states that Romney will win big, but I don’t see any evidence that the overall EC tilt is more than 1% Democratic. What am I missing?
And why does Nate Silver give Obama a 71% chance of winning the popular vote, when his own average of national polls show the race tied? Why does Silver think the national polls are wrong?
I predict that if someone took all the state polls, for all 50 states, the aggregate would be substantially more pro-Obama than the national polls. Is there a difference in methodology? If so, which are more reliable?
It seems to me that either the national polls are biased, or the state polls are biased, or else there’s a far bigger chance of Romney winning the popular vote and losing in the EC than people like Silver currently believe. In 2000 I said the Electoral College should be abolished, and my GOP friends were horrified. I wonder how they’ll feel in a couple weeks.
PS. I suppose one possibility is that if Romney were to rise to plus 3.0 in the national polls, the gains would come disproportionately in the swing states, where he is advertising heavily. But still . . .
Tags:
27. October 2012 at 12:06
Silver knows baseball, not politics ….
You need to read Michael Barone & Rasmussen, who actually know this stuff.
All of the art is in pretending to know Dem vs Rep identification & turnout, and the ground is moving like an earthquake on that over last 4 years and in last 3 weeks.
27. October 2012 at 12:08
It’s not going to be that close, Romney will win vote & EC handily.
27. October 2012 at 12:11
Siver gives Obama a 71% chance of winning because his model gives Obama about a 71% chance of winning Ohio. But then I repeat myself.
27. October 2012 at 12:13
If Obama wins California and New York by 15 points, and Romney wins Texas by 40 points, that could explain the disparity. It’s not just about there being lots of narrowly-Dem states and lots of overwhelmingly-Republican states, although that’s probably a factor. It’s that the safe Republican states might vote more overwhelmingly for Romney than safe Democratic states do for Obama.
27. October 2012 at 12:39
Theo,
But the polling doesn’t show Romney winning the big red states (Texas and Georgia) by more than Obama wins the big blue states (California, New York and Illinois). Quite the opposite, in fact.
Real Clear Politics does not even include Texas or Georgia among Romney’s most solid states, whereas California, New York and Illinois all make that grade for Obama.
27. October 2012 at 12:52
If you look at Gallup’s internals, Romney is trouncing Obama in the South but Obama is slightly ahead in all other regions — which suggests to me that there is a reasonable chance that Obama will win the electoral but not the popular vote.
Nate Silver’s view, I think, is that maybe the national polls are wrong, maybe the state polls are wrong, or maybe they’re both right, but we can’t tell which. Based on state polls alone, I think he would assign Obama a significantly higher probability of electoral college victory, but he allows for the possibility that the national polls are giving us a better picture of what’s going on in the swing states. The reason he doesn’t see a high probability that Obama will win the electoral college but not the popular vote is that he sees the state-national split as only one of 3 possible interpretations (the other 2 being that one or the other set of polls is wrong). And even if both state and national polls are right, the game isn’t over, and there is still (in Nate’s model, anyhow) a significant chance of either a Romney electoral victory or an Obama popular victory (especially the latter, since some national polls are showing Obama slightly ahead).
My guess is that there is a large tilt in the electoral college. It’s hard to see how Romney can win the electoral college now unless something weird happens, but he could easily win the popular vote (though my guess is he won’t, because I think we will see more of a shift toward Obama as the impact of the last debate shows up more thoroughly in the polls, and I think we may see more unlikely voters voting that the polls assume).
27. October 2012 at 13:02
Links for the Gallup regional split:
Analysis by Ezra Klein
http://www.gallup.com/poll/158048/romney-obama-among-likely-voters.aspx
These are outdated by now, but they strongly suggest where the split is coming from.
27. October 2012 at 13:03
I guess I got the HTML wrong for that second link:
Hopefully it works this time
27. October 2012 at 13:44
“It seems to me that either the national polls are biased, or the state polls are biased, or else there’s a far bigger chance of Romney winning the popular vote and losing in the EC than people like Silver currently believe.”
The last option is the most likely, in my opinion. I think there is a very large chance that Romney will win the popular vote and lose the electoral college. I think Silver’s state-by-state analysis is not too bad (because it favors his candidate), but his national vote odds are way off. Silver’s MO is to act as a propagandist until a few days before the election, where he changes his model and plays it straight up and gets it right. Mark my words: Silver will show a radical shift towards Romney a couple days before the election; I still think Romney will lose the electoral college, but Silver’s analysis will make more sense.
27. October 2012 at 13:51
It’s funny how the framing can change my impression of the election. If I think, “Romney wins the popular vote, but Obama wins the electoral college because he was ahead in Ohio and a few other important swing states, my next thought is, “So Obama wins on a technicality. Isn’t it strange that we let the people of Ohio choose a president for the whole country?” But if I think (see first paragraph of my first comment above), “Romney wins the popular vote, but Obama wins the electoral college because he is ahead in every region except the South,” then I think, “OK, Obama wins fair and square. Damned if we’re going to let Southerners choose the president just because they’re opinions are more lopsided than the rest of us!”
27. October 2012 at 13:53
Simon, You don’t understand what I’m talking about, read the post again.
Andy, That Gallup is way off–it has Obama up by 4% in the East. He’ll win the East by nearly as much as Romney wins the south. I’d guess Romney wins the South by 12% to 13% and Obama wins the East by 11% to 12%
Theo, You are insulting my intelligence. Texas by 40%???
Everyone, Come on, surely someone can answer my question–this is stuff that pundits are going over with a fine tooth comb. It’s a big mystery!
I defy anyone to come up with a plausible set of state results that has a 3.3% Dem tilt in the EC. It’s much harder than people seem to think.
Here’s one way for someone with lots of time on their hands. RCP has about 1/2 the states. For the other half adjust the 2008 margins by however much that group of states were seen as swinging toward toward the GOP. (McCain lost by 7.3%.) That will give you a ballpark estimate of the state results.
27. October 2012 at 13:55
Cedric, That makes sense.
Andy, Your first thought was best. Otherwise what to we do with the African- American vote?
27. October 2012 at 14:07
I don’t trust Gallup’s regional breakdown (provided by Andy Harless in his post at 13:03). It shows very little movement toward Romney, relative to McCain’s performance in 2008, outside of the south.
Yet Wisconsin, which Obama won by nearly fourteen points in 2008, is nearly even in the polls.
Obama is currently leading in the polls for Michigan by four points, which was a state he won by more than sixteen points in 2008.
Obama won Pennsylvania by more than ten points against McCain, but leads by less than five in the polls this year.
Colorado, which Obama won by nine points in 2008, is tied up this year.
Obama squeaked out a victory in Indiana in 2008, but trails in the state by more than a dozen this year.
All of the states I listed above are above average in population for a U.S. state. None of them are southern.
Meanwhile, some of the southern states haven’t moved as much toward Romney as one might expect. North Carolina has moved only four points. That state gave Obama a surprise victory by the smallest of margins in 2008, but is still listed as a swing state this year, with polling showing Romney with less than a four-point advantage. The polls for Virginia show that state has moved about seven points since the last presidential election. The same for Georgia. Florida? About four points.
Compare that to the nearly double-digit moves towards Romney in states like Michigan, Wisconsin, New Hampshire, Nevada, Connecticut, Oregon, and Colorado.
The only southern state I could find which shows a similarly large move toward Romney over the last four years is the border state Missouri, which appears ready to give the Republican candidate a double-digit victory after nearly going for Obama in 2008. But does Gallup even count Missouri as a southern state?
27. October 2012 at 14:11
Silver’s model adjusts for “house effects” of certain pollsters. For example, historically Rasmussen has overstated Republicans share of the vote when compared to exit polls, so his model discounts a 3 point lead by Rasmussen to perhaps 1 point.
27. October 2012 at 14:25
The specific internals of a particular poll are likely to be way off, just because of sampling error. But the overall pattern in the Gallup poll is suggestive. Here’s another poll, from Investors Business Daily, in which Obama wins the national popular vote but Romney is still way ahead in the South. Average that with Gallup, and and you get a narrow Romney popular vote lead because of the lopsided Southern vote, so I think the story works pretty well.
27. October 2012 at 14:35
Most likely the national polls are just wrong. It’s quite possible this has to do with simple statistical noise. Even when you combine all the national polls, I think the margin of error is still something like 2%. Obama’s lead in the swing states is consistent with a national lead of about 2%. Given that the average of the national polls has fluctuated around a tied race, you could probably chalk up the difference to noise.
It might also be that there are a couple of pollsters that are slightly skewing the average to Romney. I’m thinking Gallup and Rasmussen. Both were very bad in 2010. Rasmussen has questionable methodology: they don’t call cell phones, they weight by party ID, and when they conduct state polls they only do a single night of interviews. (The first and third save money but produces samples that may not be representative. The second assumes the outcome.) Gallup’s methodology is fine, but their track record when they diverge from the consensus is very bad. Aside from those two, everyone else has the race in the range of +2 to -2 for each side.
But let’s say, for the sake of argument, that the national polls really do show a slight Romney lead, thus suggesting a 3% split between the swing states and the popular vote. This wouldn’t be the first time this has happened. In 1996, state-by-state polling suggested a much narrower margin of victory for Clinton. In 2000, state-by-state polling suggested a more-or-less tied national race, while national polling had Bush with a big advantage. In both cases, the state polls turned out to be correct. (Some thought Gore might win the EC despite losing the PV. How ironic.)
Because of this, Nate Silver has constructed his model to be very skeptical of a PV-EC split, and so he essentially splits the difference between the national polls and the state polls. As a result, his popular vote estimate is more bullish for Obama, while his state-level estimates are more bearish for him, compared to simple averaging.
27. October 2012 at 14:36
If you want to know why Nate Silver’s predictions are what they are why don’t you read his posts? He explains his methodology pretty thoroughly on a regular basis. I guess it’s easier just to accept Cedric’s ridiculous explanation (“it’s the librul media, duh”) because it fits your partisan predisposition.
27. October 2012 at 14:40
“Otherwise what to we do with the African-American vote?
You mean that, by my logic, the lopsidedness of the African-American vote would seem like an unfair advantage for the Democrats? It’s not quite the same thing, though. If you divide the country by ethnic groups, you won’t find that all except the African-Americans support Romney. If that were the case, and if ethnic groups were comparable in size, I would be a little uncomfortable that one ethnic group alone was able to elect Obama.
In general, I don’t see why there’s anything magic about a popular majority. Democracy isn’t a perfect ideal; as Winston Churchill notes, it’s not a good form of government at all in an absolute sense. Hopefully it avoids the worst problems associated with other forms of government. I think you might avoid more problems by structuring it not to give too much power to the most cohesive groups. The electoral college system rewards people for living in places where there are a lot of people who disagree with them. Isn’t that a good thing?
27. October 2012 at 14:45
Professor Sumner, I’m surprised to see you relying on a single private forecaster rather than a prediction market – http://www.intrade.com/v4/markets/contract/?contractId=743474
27. October 2012 at 14:47
Scott, I don’t have a vote in US elections so I won’t delve into this. But when I put together these two points from the link below, I have the feeling that they go a long way to answering your question:
1) “Gallup’s seven-day tracking poll shows Mr. Obama winning the West (+6), the East (+4) and the Midwest (+4), while losing the South (-22), putting him behind in the popular vote, if not in the Electoral College.”
2) “Wyoming has roughly 189,000 people per electoral vote. California has about 679,000.”
http://takingnote.blogs.nytimes.com/2012/10/19/a-popularelectoral-split/
27. October 2012 at 14:52
As Greg said, it’s all about party ID. First of all Romney is winning independents by up to 8%. Second, state polls like Ohio generally sample about 5-6% more Democratic voters than Republican ones (See: http://www.nationalreview.com/corner/331046/ohio-closer-you-think-josh-jordan). On the other hand, national polls like Gallup and Rasmussen sample about equal numbers of both. If equal numbers of democrats and republicans vote, than Romney wins both the popular vote and electoral vote. If Democrats have a 5% turnout edge then Obama wins both. If the edge is somewhere in the middle than it will be a long night. See: http://baseballcrank.com/archives2/2012/10/politics_why_i_2.php
27. October 2012 at 15:08
Obama’s popularity has dropped 7 points in popularity in 3 days .. Romney is surging in every swing state, esp with independents & women.
The incompent Obama is well below 50 % in popularity & in voter preference.
It’s over — we have Carter vs Reagan all over again.
27. October 2012 at 15:16
Sorry if you felt I was insulting your intelligence, but there was a ‘Republicans have more safe states, Dems have more narrow states’ idea coming through in your post that I thought needed addressing. And obviously my example is spurious but I was just illustrating my (clearly unnecessary) point.
Anyway, it’s very difficult to know how big the discrepancy between state and national polls is, because safe states without competitive Senate races don’t get polled much. If we had enough polls to get a good idea of where the race is in New York or Texas, we’d have a much better idea of the size of the state-national discrepancy, and the residual would then be the EC split.
You can come up with plausible stories as to why the Dems would have a significant EC advantage this time around. The coalition that gives Dems their strength in safe states like NY – urban service workers, educated professionals, young women, college kids – is conceivably discouraged by poor economic performance and confused by Romney’s pivot to the centre. A fall in financial-sector support over Dodd-Frank is likely to cost Obama a disproportionate number of irrelevant Northeastern votes as well. The blue-collar workers who are on track to secure Obama Ohio, Wisconsin and Iowa are more amenable to the Obama’s campaign’s focus on the success of oil bailouts and assertive foreign policy. So I don’t think it’s totally implausible that there is a large EC advantage for Obama.
However, I agree that a 3.3% tilt is implausible. The state polls are overestimating Obama, or the national polls are overestimating Romney. I’m leaning towards the former explanation, as is Silver – he points out that state polls have historically done better than national ones when they’ve diverged. Perhaps Romney’s bounce helped to increase enthusiasm amongst safe-state Republicans, while months of ad bombardment had already worked to enthuse Ohio partisans, so there was less of an effect there. But equally it could be that all the polls of Ohio are skewing Democrat for some reason. And we’re not likely to know which polls are wrong until election night.
27. October 2012 at 15:17
I meant incumbent ….
Wake up folks — the electoral votes have moved South & West, except for California, which lost electorial votes for the first time ever.
Blue states continue to gain population & electorial votes, and Red state continue to lose them.
27. October 2012 at 15:18
BTW Scott, several of Nate Silver’s recent posts address the quesions you ask. See for example this, from 11th October (hence not topical but it relates to the problems you raise):
http://fivethirtyeight.blogs.nytimes.com/2012/10/11/oct-10-is-romney-leading-right-now/
27. October 2012 at 15:45
Kevin, Your argument cuts the wrong way.
Greg, Please tell me you aren’t that delusional. I thought you were a big believer in science and facts and all that good stuff.
Basil, I do relay on Intrade, but that has nothing to do with this post.
Andy, If Obama wins the national vote then he’ll blow Romney out of the water in the East and in California. And Churchill was wrong, democracy is far and away the best political system, nothing else comes close. The wisdom of crowds.
But the US Federal government is way too big–so that sort of fits with your argument. If we split into 50 countries, we might be better governed.
Steve, That’s a fair criticism. (Except I’m certainly not a Republican.)
Alexander, Those are good points, although I’d note that Rasmussen was pretty close in 2008, and Gallup was way too much for Obama. So I still think there’s a bit of a mystery. See my next post.
27. October 2012 at 15:58
Pincher Martin, Excellent. Oddly I wrote an almost identical post before I read your comment.
Theo, Good points. That’s a much better comment.
27. October 2012 at 16:00
Albert, That sounds plausible.
27. October 2012 at 16:43
Rasmussen is consistently more Republican than other pollsters.
If you go to the Real Clear Politics website and look at 2008 polls versus the actual outcome, Rasmussen consistently understated Obama’s lead.
For example, he had the only poll showing McCain actually beating Obama in Florida.
27. October 2012 at 23:15
Scott I think Nate’s 71% Obama advantage-it’s now almost 75%- refers to the electoral college not the popular vote.
He does have Obama up by about 1.4% in the popular vote-his model factors in the state polls a lot in determing a popular vote prediction
It’s certainly interesting that there is such a big gap between national and state polls.
There are a number of ways to explain it. One way is that the methodlogy of polls like Gallup are way off, particularly regarding their new likely voter poll.
In general I don;t think that national polls are very trusthworthy right now, least of all Gallup.
If you believe them Obama rose to a 54% job approval rating 3 days ago and is now at 46% There is so much volatility in it, to make it more or less worthless.
Gallup generally doesn’t distinguish itself when it becomes an outlier.
On reason teh national polls may be wrong is that many of them undercount cell phone voters and Latinos-who break for Obama.
If you’re for Romney your only hope is that somehow there is a systematic oversetimation of Obama’s state numbers through all these different polling groups
It’s possible but less likely than some of the national polls are askew.
27. October 2012 at 23:44
“Churchill was wrong, democracy is far and away the best political system, nothing else comes close. The wisdom of crowds.”
I would have thought Scott knew enough public choice not to make that argument…
28. October 2012 at 00:30
ssumner:
Both Churchill and Sumner are wrong, democracy is not even close to being the best political system. The crowd, if history is a judge, is rarely if ever right. Democracy is a watered down version of communism. Nobody’s property is safe. Nobody’s person is safe. If the majority want to enslave a minority race, if the majority wants to confiscate the wealth of the minority, then democracy condones it.
True wisdom comes from individuals, not crowds. The best political system by light years is the individual private property order. Not only can it attenuate whatever wisdom exists in “the crowd”, but it also protects the minority from the ignorance of “the crowd”.
The main reason why democracy appears to be the best is because we don’t have a single world democratic state. In each democratic country, the majority is constrained by the pseudo-anarchist world order, which prevents it from running total roughshod over the minority.
A world democratic state would be a nightmare. Former China and former India would have a very large size of vote, and they would almost certainly vote to redistribute wealth from the wealthy western former countries, to themselves. Would Sumner like it if half his wealth was voted away to some random people in the third world?
If any pro-democracy ideologue argues democracy works better at something less than 100% of the planet, then the same principle there is the same principle why democracy works better at something less than 50%, 40%, 30%, 20%, etc, all the way down to cities, neighborhoods, and finally… individual private property, the optimal.
Nobody truly believes in democracy because nobody believes that the majority has a legitimate right to murder the minority by vote.
28. October 2012 at 03:52
Speaking of the African-American vote:
http://www.politico.com/blogs/charlie-mahtesian/2012/10/georgia-black-turnout-on-record-pace-147409.html
Obama is probably not going to win Georgia, but if record turnout of Afican-Americans is a national trend it could be a much larger Obama win than anyone is predicting.
28. October 2012 at 06:11
Saturos, Which European country has been better governed than Switzerland over the past 100 years? Public choice theory (interpreted correctly) says real world “democracies” are nowhere near democratic enough.
Mike Sax, No, he has both percentages. I’m refering to his popular vote estimate.
28. October 2012 at 07:07
Polling is art.
The logic of statistics is logic.
Figurings out what meaning polling data have isn’t pure logic.
QED
28. October 2012 at 07:09
Ungrounded insults won’t change what happens in November.
28. October 2012 at 08:57
Minnesota is now in play ….
http://www.powerlineblog.com/archives/2012/10/minnesota-now-in-play.php
28. October 2012 at 10:17
Scott, I think it’s more the local scale of the Swiss government, with lots of devolution, combatting the irrational voter problem. Is that what you mean by “more democracy”? Still, political choice has the wrong marginal incentives, and I think things should be done with markets whenever possible. (Also I’m a sucker for non-coercion.)
28. October 2012 at 10:22
More than the low populations, there’s the possibity of switching cantons (i.e. Tiebout competition).
28. October 2012 at 10:24
So all the unnecessary “extra” functions can be devolved to the competitive parts, whilst the necessary order-keeping, communicating and cross-adjudicating functions only are kept ultimately at the Federal level. That’s the way it should be done. I agree that Switzerland is a model for us all. Krugman likes Sweden; well, I guess Switzerland is my Sweden.
28. October 2012 at 10:26
Saturos, Yes, democracy is local control. How could you claim the people rule if they don’t have local control? It also means direct referenda. Switzerland had more national referenda during the 20th century than all the other countries of the world combined. That’s why they have lower taxes than the other European countries. Who are you going to trust?
Greg, Yes, Minnesota is in play, and there really is a Santa Claus.
28. October 2012 at 10:32
MF interjects: You know what’s really local control? Everybody controlling themselves, and no one else.
And I have to say, I kind of agree…
What do you think is the optimal use of referenda in political decision making? Are you advocating Athenian democracy?
28. October 2012 at 10:39
What do you think prevents the Swiss from voting for more redistribution? Appreciation of the negative incentive effects? (Though I doubt they are at the Saezian optimum yet.) And are they just more enlightened than everyone else when it comes to, eg. the costs of warmaking, the benefits of free trade, free markets?
28. October 2012 at 10:54
Saturos, You are confusing procedures and outcomes. We might like to see a libertarian outcome, but that’s a different question from what sort of procedure we use to choose governments.
It’s makes no sense to say “I favor this political system because it’s most likely to implement my preferred policy.” Taken to the logical extreme, everyone should then want a system where they are dictator. But you don’t get to choose who is the dictator, which is the whole point of democracy. A non-democratic government might end up with Saturos enacting local control, or it might end up with Mao doing the Great Leap Forward. I’d rather take my chances with Swiss-style democracy, which has done better than the vast majority of non-democratic countries.
No one will ever dissuade me from favoring democracy by saying “suppose the voters pick a bad policy.” Yes, that could happen, but it’s far more likely to happen under a non-democratic regime.
28. October 2012 at 10:55
Saturos, The Swiss cantons each have their own income tax, determined by referenda. That’s a good system.
28. October 2012 at 13:25
“Taken to the logical extreme, everyone should then want a system where they are dictator.”
Congratulations, you’ve just figured out why governments exist.
28. October 2012 at 15:31
November.
And I though we were suppose to bow down before the headline numbers reported by journalists on the basis of the headline reports of polling firms.
No?
29. October 2012 at 04:57
I think Nate Silver is just following intrade.
29. October 2012 at 05:42
Obama below 40% on whether country is headed in the right direction, no enthusiasm among marginal Obama supporters, popular vote at R 52% , O 47%.
That’s the later Battleground Poll.
You can’t argue with science.
29. October 2012 at 06:41
Ohio — R 50, O 48 …
What’s Silver going to ‘predict’ today?
29. October 2012 at 07:01
So Seth MacFarlane is hosting the Oscars: http://www.nytimes.com/2012/10/29/movies/hollywood-seeks-to-slow-cultural-shift-to-tv.html
That’s it, America is doomed.
29. October 2012 at 20:11
“But here is what we do know: 220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008.”
29. October 2012 at 20:11
“But here is what we do know: 220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008.”
29. October 2012 at 20:11
“But here is what we do know: 220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008.”
29. October 2012 at 20:11
“But here is what we do know: 220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008.”
29. October 2012 at 20:12
“But here is what we do know: 220,000 fewer Democrats have voted early in Ohio compared with 2008. And 30,000 more Republicans have cast their ballots compared with four years ago. That is a 250,000-vote net increase for a state Obama won by 260,000 votes in 2008.”
2. November 2012 at 03:50
Greg you understand Romney’s chances of winning don’t increase by you convincing us that he’s supposedly already winning right?