Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Presidential prediction 2012 (Election Eve)

November 6th, 2012, 1:47am by Sam Wang

If state polls perform as well as they did in 2004 and 2008, most aggregators should get within +/-15 electoral votes and 48/51 races correct. The question is how to squeeze a bit more out of the data.

The topline listed just below the title of this website is not our prediction, but the automatically-generated snapshot. It will fluctuate as the last few polls trickle in. The last 3 updates are at 8:00am, 10:00am, and noon. Then we freeze it.

I make two electoral predictions.

ELECTORAL PREDICTION (mode): Barack Obama 332 EV, Mitt Romney 206 EV. The mode is the single most frequent value on the histogram. It corresponds to the map below, and has a chance of being exactly correct.

ELECTORAL PREDICTION (median): Obama 309 EV, Romney 229 EV, Popular Vote Meta-Margin Obama +2.34%. This is  the automatically-generated snapshot for November 5th 8:00pm. This prediction is almost guaranteed to be off, since 309 EV is not a common combination. (Note: I’ll edit this with the last few updates.)

ALL-STATE PREDICTION (binary outcomes):

Most outcomes arose clearly from the median of the last week of polling. The exception was Florida, for which I used polls with at least half of their respondents on Nov. 1 or after. This gave a median margin of Obama +0.5 +/- 0.8% (n=10), for a win probability of 72%. I will be unsurprised for it to go the other way (outcome Obama 303 EV, Romney 235 EV). In Florida, a recount is triggered by a margin of 0.5% or less (recount rules, Brennan Center for Justice). There is about a 50-50 chance that we will see that happen.

The next-closest states are Colorado, Virginia, and North Carolina, with margins of less than 2.0%. Excluding these three states and Florida still leaves Obama 281 EV, Romney 191 EV.

In addition to all-or-none outcomes, later today I will provide final polling medians in graphical form in the Geek’s Guide. I hope it will be useful as you follow the returns.

POPULAR VOTE. The average Meta-Margin for the last three days was Obama +2.7%. The median of national polls is Obama +1.0 +/- 0.5% (n=13 polls). The approach I described before for combining these measures gives

  • Final predicted popular-vote margin: Obama +2.2 +/- 1.0%.
  • Two-candidate vote share: Obama 51.1%, Romney 48.9%.
  • Allowing 1% for minor-party candidates: Obama 50.6%, Romney 48.4%.

Finally, here is a look at the histogram as it stands tonight. The EV histogram has resolved to just a few peaks because so few states are in play – the four states I mentioned before, plus maybe Iowa. I’ll update this with a cleaner version with longer-term-integrated probabilities in the morning.


I should point out that A note: Election Eve prediction is not the most impressive of feats. What we did in August (here, and here) was the interesting part. Those were true predictions, and were centered around Obama 315 EV, Romney 223 EV.

Put your predictions in comments.

Tags: 2012 Election · President

294 Comments so far ↓

  • Nathan Duke

    Ha! I just saw a tweet from Peggy Noonan saying that Romney had a massive lead in Ohio and that he was closing the gap in California. Right. And, of course, nothing to back these claims up.

    • Pat

      Haha seriously? Did she say Romney was also catching up in Vermont and Hawaii?

    • wheelers cat

      she’s drunk.

    • wheelers cat

      theres a reason they call her nooners.

      /thats what *she* said

    • mediaglyphic

      Why does anyone take Peggy Noonan seriously?

    • Sam Champion

      Peggy Noonan was castigated by and ostracized from the tribe back in September when she told the truth and called Romney and his campaign incompetents. She’s now working her way back into the tent by parroting the party line – Romney has the big ‘mo because there is just a “feeling out there” . She saw that she’s have no career as a pundit unless she spewed her particular brand of partisan guff.

    • Anne Boyd

      She saw Romney lawn signs in Florida! That’s better than data!!!

    • John

      Could not find the tweet you refer to, but she did say this:

      Romney’s “stealing in ‘like a thief with good tools.'”

      Perhaps not the best choice of words in the current political climate.

  • Marcus

    Romney 275
    Obama 263

    Popular vote 50-48 R.

    And then I wept myself to sleep.

  • Maurice Pinzon

    My prediction is: EV: 3o3 Pres. Obama
    235 Romney

    Reports of long lines in New York City. Hopefully, the turnout will be huge across the country. I’ll be checking the clock tonight for these poll closing times:

    Democracy is a beautiful thing! Looking forward to tonight. And bravo, Prof. Wang, for all your work.

  • Michael_NC

    O – 278
    R – 257 (Va. Fl. Co)

    • Freakshow

      I can say with 100% certainty this will not happen. Well, unless a state with 3 electoral votes magically disappears or is destroyed by a meteor strike…

    • Michael_NC

      So…you say I can not add… ;-)
      But I’ll stay with O – 278

  • Jeannius

    Obama 332, Robme 206 (Been saying this for a week now)

  • Neminem

    (sorry for cross-posting this from the FAQ)

    Hi can someone briefly expand my understanding of the polynomial, perhaps with a simple example? Struggling at the moment. What is “x” in the polynomial? What does the product of the terms in the polynomial respresent?

    • dgarbern

      As Dr. Sam refers to it (“a simple math trick”), this polynomial is a bit of a mathematical artifice. It is an elegant way of generating all the probabilities for the EV outcomes 0 to 538 by simple multiplication and the law of exponents.

      Once you do the math you generate a polynomial looking like:

      C538 * x^538 + C537 * x^537 + …+C1X + C0

      The coefficients (C538…C0) are the probabilities for each EV outcome (538 to 0). X is not assigned a value to yield any useful data, but it should be immediately apparent that if you make X=1 then the product of all the factors ((1 – P1) + P1 * x^EV1) * ((1 – P2) + P2 * x^EV2) * … * ((1 – P51) + P51 * x^EV51) is 1, the sum of all the probabilities.

      It may be helpful to calculate a simpler polynomial using an EV range of (0..2) and see if this makes sense.

    • dgarbern

      Actually, I meant to say a simpler polynomial using just 2 states, not an EV range of 0 to 2. Take Dr. Sam’s 2 state example of EV1 and EV2 and generate the coefficients for:

      ((1 – P1) + P1 * x^EV1) * ((1 – P2) + P2 * x^EV2)

      Pretty neat, huh?

  • Khan

    I do it so the rest of you don’t have to.

    Faux News headline story for the past 2 hours has been black panthers (a.k.a one black guy) standing outside a Philly polling station.

    They have even brought on special guests to discuss it.

    • David

      Have they yet said: “Oh my stars, a negro!”

      Because they might as well.

      Did you see that poll about how nearly all explicit racists are Republicans, though of course that doesn’t mean that all Republicans are racists. I think anti-black sentiments were roughly 55 Dem and 65 Repub. So they’ve got a strong base to play to.

  • David Mann

    I’m guessed 303-235 a few weeks ago elsewhere, so I’m sticking with that prediction.

  • SoleburyJim

    Voted already, was in line when our polling station opened. The line was healthy, but I don’t think it was quite as long as it was at about the same time in 2008. We live in a politically balanced but probably slightly Republican-tilting area, but am posting as an observation rather than to draw any conclusions.

  • badni

    I sure hope the poll workers in Ohio and Florida are better trained than the ones in Brooklyn. Chaos, lines down the block, ask 3 different poll workers get 3 different answers.

  • Buck the Liberal Dog

    Thank you, Sam and Andrew, for keeping me sane the past few months. PEC is the first thing I look at every morning and the last site I view each night (538 is second).

    You guys are national treasures! I can’t imagine an election without you!

    • Randall

      Anyone know anything about county micropolling?

    • Khan


      That article was posted yesterday. Given that Ohio does not report party identification of ballots there is no way they could know any of that data. They don’t even link a poll from where they gathered the “data.”

    • Ross C

      Even looking at their own data, that is not all that impressive — looking at past elections, the county seems to reliably vote Republican by a couple of points. Not sure this is any big news.

    • MarkfromLA

      Giant MOE of 5.65% on these Suffolk polls.

    • Randall

      I missed this yesterday. Never heard of this Suffolk outfit.

    • Michael_NC

      As a clarification to the above…
      This 11/5 polling data

    • Khan


      NVM, my phone’s browser cut off the bottom of that page.

      They only polled 300 people in a county with more than 200,000? The MOE is +/- 5.65%. That is horribly inaccurate.

      Romney could have anywhere from 52 to 42 percent support and it would be within the 95% confidence interval.

      I would ignore that poll.

  • mediaglyphic

    Can anyone figure out why Sean Trende is saying national polls are better?

    Its quite long and very difficult to follow, i usually find him to be coherent, but in all honesy i just don’t get it.

    • Khan

      Tried to understand it from a rational standpoint and simply came away with the impression that he is trying to cover his butt.

    • wheelers cat

      Sean has chosen the pundit [dark] side. At this point all he can do is throw chaff.
      I actually thought he might choose the side of the Poll Jedi, right up until he blocked me on twitter.
      hahaha, I guess the bromance with Dr. Linzer is over.

    • wheelers cat

      I shouldn’t be so mean. The poor guy is actually the best the right has got.
      He chose his paycheck is all.
      He has dependents.

    • BrianTH

      It appears to me that in a crucial sequence, he slips in the possibility that the state polls in all the necessary states would happen to have errors on the high end of the overall range of observed errors (crucially including 1992 in his dataset) AND he further hypothesizes that these errors systematically favored Romney, and then he combines that hypothetical possibility with the actual observed systematic bias in state polls over those elections to conclude the state polls are likely to be as subject to systematic bias as the national polls. But of course what he has done with all that hypothesizing is sneak in a much larger possibility of a very large systematic bias in Romney’s favor than the data set of states he is looking at would actually support.

      On the plus side, I think he is right that we can’t really just assume the historical track record of state polls in recent elections necessarily applies to the current state polls. Unfortunately, without that assumption we are left just not knowing what sort of possible distribution of systematic biases we should be using. Still, it is arbitrary to assume that the possibility of a large systematic bias in the state polls is as high or higher than the possibility of the same magnitude of systematic bias in the national polls.

    • BrianTH

      By the way, big picture wise, he is more or less insisting on the view that some one polling model has to be completely right, and all the other ones are wrong, and the only issue is we don’t know which is the completely right model (see his whole discussion about the capital of the United States). Of course we are talking about something very different from determining in the historical record the location of the capital of the United States, and it is in fact also possible that all the polling models are a little bit off. In fact, if you assume it is increasingly less likely that a good-faith model would be a lot off versus a little off, that second scenario (everyone a little off, versus some pollsters completely right and some pollsters a lot off) is probably more likely.

      In that sense, even if it were true there were distinct, separable modeling “camps” (and that is highly disputed), it could still easily be the case that the actual result will fall into a middle ground between the camps, notwithstanding the alleged relative lack of pollsters occupying that middle ground.

    • wheelers cat

      well there actually ARE distinct, separable modelling camps. There are the Poll Jedi or the Numerati as i like to call them, Wang, Linzer, Silver, Marshall, Blumenthal, etc. …and then there are the pundits and the Innumerati, people like Jay Cost who reject the Central Limit Theorem and people like Sean Trende that throw radar chaff to keep their paychecks…

  • Olav Grinde

    Here is an interesting analysis and election prediction from Prof. Michael McDonald of George Mason University.

    Dr McDonald is tallying early votes on his excellent website. Some very interesting numbers!

    • mediaglyphic

      Olav, thank you for posting this vimeo link. He is a very very knowledgeable guy and i love the amount of nuance in the video. I have been using his site for a while to track early voting, but this is a whole other level of information.

    • Michael_NC

      The D/R early votes in CO FL & PA look grim. Much higher R than what I expected.

    • mediaglyphic

      Michael, if you have time, watch Dr. Mcdonald’s Vimeo. The early vote is quite complicated and he explains the nuances very well, in a balanced manner and without a lot of hyperbole.

  • WingGirl

    O 303
    R 235

  • bks

    100.0% (Bayesian) chance of re-election? Yow!


  • ChrisD


    323 Obama
    215 Romney

    Romney pulls out CO because Johnson takes votes away from Obama, who narrowly wins FL but not NC.


    50.1% Obama
    48.3% Romney
    1.6% Other

    (Other candidates drew 1.4% of the national vote in 2008.)

  • Steven J. Wangsness

    O 51 (332EV)

    R 48 (206EV)

    Basically, Obama takes all the swing states except NC. I sense a deflation of Romney in the ether.

  • bsk

    Assuming no more FL polls come in, I’ll have to go with 332 as the mode. It’s more fun to be contrarian (and I certainly am among my friends!) but I think Sam is spot on, and taking the median of all the major poll aggregators points to the same outcome.

  • GBIllinois

    Wapo”s Michael Gerson’s hit piece on “pseudo-scientific” political analysis.

    ‘ Those who emphasize “objective” political facts at the expense of “subjective” values have strained out the soul and significance of politics. It is an approach, in the words of G.K. Chesterton, “that stores the sand and lets the gold go free.”

  • Stephanie

    Man I hope all you number crunchers are spot on! What’s the deal with PollTracker? That website tells a different story

  • GBIllinois

    My Prediction:

    Romney 440
    Obama 98

    Hey if Jim Cramer can do it, so can I. :-)

    Obama/Biden 2012

  • Stephanie

    Nate just recalculated check 538

    • Ross C

      Does Nate include early voting info in his secret sauce? Just wondering.

    • BrianTH

      No, he just uses polls as inputs at this point (even the economic factors are now pretty much irrelevant).

    • Ross C

      Ok thx. I’ ve been seeing some conflicting info on early vote totals. Or maybe disinformation.

    • mediaglyphic

      Michael Mcdonald of George Mason explains early voting in the video below from yesterday.

      He forecasts Turnout higher and an obama win probability. NC is almost for sure Romney, and NV/Iowa almost for sure Obama, he thinks Colorado and Ohio are Obama, but sees low but not insignificant probability that Rom gets these

  • Jonathan


    A simple thank you for being a voice of reason during the last few months.

    And I do hope your 100% prediction is correct.

  • Matt McIrvin

    Just voted. Today is a school holiday in Massachusetts; they get the kids out of the schools that are actually being used as polling places, and use the rest of the day for teacher training.

    So I took the day off to look after my kid, and had the opportunity to go in midmorning. There were no lines, and the voting population seemed mostly elderly as usual, but this was probably just because I waited until after the morning rush. There did seem to be two or three times as many people there as in off-year elections.

    I’m struck, hearing reports of long lines elsewhere, at how privileged I am at having the opportunity to vote so easily. On the one hand, this state unfortunately doesn’t have early voting at all. On the other hand, I personally had a really easy time voting today, partly because of my personal situation.

    • Khan


      I hear you. I live in a lakefront community and our polling lines are always so small. I cannot imagine waiting more than 10 minutes to vote–let alone 6 hours.

  • Uli Ohmes

    I am a very excited German Obama supporter and i want to thank Sam for his wonderful and competent analysis. I have been so worried about the President’s election chances but every time i looked at this site i felt a bit more relaxed. I wish you and me that our hopes for your President’s reelection-chances will come true.That would put America forward and the rest of the world, too. Good luck for you all and thanks again for your incredible work!

  • philly suburbs

    as a person who has spent her entire life avoiding math I just want to say thank you for showing me how down right fun, interesting, and engaging stats can be.

  • skmind

    Obama 303
    Romney 235

    Popular vote Obama 50.5% Romney 48.5%

    Dems + Independents Senate 53
    Rep 47

    Rep House 238
    Dems House 197

  • Dgarr

    Thanks Dr. Wang and Andrew :-) and to all the witty posters today. Finding this site has truly been a pleasure. I will be smiling all day as I think about how privileged we are, as Americans, to have this process. We will all collectively choose the direction of our country’s future today. This has been a dream campaign for this political junkie. You guys have made it all the better. VOTE!

  • Michael S

    Sam, I love you man, but with Florida you may be tipping the scale with hope a little too much like in 2004.

    I would suggest that scenarios 303/235 and 332/206 are a statistical tie as each has an 18% chance of happening, with the 303 chance having a slightly higher probability of happening. Looking at your own base numbers Barack Obama has a 73% chance of getting 303 electoral votes or more, a convincing win in media terms.
    The chance of a narrower victory is 26%

    Chance of a tie: 1 in 9000.

    Chance of an outright Romney win: 1 in 160.

  • Jack F.

    It’s not that all Republicans are racists. It’s that if you are a racist and you want to join a mainstream political party, which one would you choose? Bingo, Republican. Ergo, they have a lot of them.

  • abjr

    My prediction:

    O: 290 (CO; no Va or FL)
    R: 248

  • Mark Levine

    Obama 332
    Romney 206

    Popular Vote Obama 50.8% Romney 48.0%

    Senate: Dems + Independents = 54

  • piktor

    The Orange and black icon is back up at the address bar!

    Way to go, Princeton Elecshun Consorshum!!

  • Olav Grinde

    Meanwhile, an Ohio court is considering the legality of the unvetted software “patch” Secretary-of-state Jon Husted had installed on voting machines in 25 counties.

    The Christian Science Monitor reports:

  • E L

    Just for your entertainment: Romney punters are fleeing Intrade.

    • wheelers cat

      yeah, i just told *Allan Marlowe Phd* to go read a book at votamatic.
      That site is so infested with free market fantasists its pathetic.

    • Stephanie

      Intrade has Obama at 73

  • Melancholy aeon

    How can a Bayesian ever be at 100? Doesn’t that lose your prior? Don’t you have to keep it at 99.9?

  • NEPAblue

    Thank you Dr Wang and Andrew Ferguson for saving my sanity when I found your site a bit after the first debate.

    It has been a privilege to lurk here.

    I am still hoping for FL, but will go with the 303 O and 235 R scenario for now.

  • NM


    This is great work. I’m curious as to how your model will converge to a prediction as uncertainty becomes reduced tonight. As certain states become more probable or definitive towards a single candidate, that will reduce the total numbers of EV scenarios while increasing the accuracy of the prediction. It’d be great to visualize the iterations of the Median EV and the Meta Margin as the results come in incrementally, and how this corresponds to timelines of when states are officially called for a particular candidate.

    Great work.


  • Jay Bryant

    I continue to worry about the Republican party suppressing votes in Florida and tampering with the results in Ohio, so I won’t believe it until it’s official.

    I very much hope that Mr. Obama carries enough other states to render Florida and Ohio irrelevant.

    I could live with the country honestly picking Romney. I can’t live with theft.

    On a more positive note, I want to again thank Sam and Andrew for the best election site on the web. I look forward to its future incarnations.

  • Naphtle Small

    Historically when democrats run for reelection polls are off by 3-5 points in any direction.

    Truman, Carter, Clinton

  • Abraham

    I’m going with Obama 303, Romney 235.

    53 D/I in the Senate, 47 GOP.

    House is harder to call, but something like 180 D, 253, R.

    50.4 Obama, 48.6 Romney

  • Abraham

    Er, apparently I was still having my seizure with my House predictions. I meant to write ~195 D, 240 R.

  • Richard

    I think Obama wins Ohio, Wisconsin, Iowa, New Hampshire, Virginia, Florida and obviously wins re-election. My prediction is based largely on the polls but also a recognition that Team Obama has a superior ground game in all the battleground states. Thus, in Electoral Votes I predict Obama gets 332 to Romney’s 206, in popular vote Obama gets 50.9 to Romney’s 48.3%. Thanks to Dr. Wang and his people for a great website and great campaign year! Cheers!

  • swig

    Why can’t the rest of country follow Washington state and have all mail-in ballots. No long lines, no hassles.

  • Ross C

    FYI, in my little Northern Virginia town, which is pretty heavily Democratic, I just went to vote but the line was massive. I didn’t have much time, so I left, but I have this afternoon when I will go back for sure. I have the day off to do GOTV, so no problem there.

    I don’t remember this many people in ’08. I don’t remember this many people, ever.

    There were a couple of people from each party outside the polling place, and a guy asked if I wanted a Romney sticker. I said “absolutely not.” He said, “big mistake.” I replied “I’ll debate you on that.” He then said, somewhat apropos of nothing, “the debt.” I offered to debate him on that too.

    I remember being very upset when Bush was squandering our surplus for tax cuts and rebates. I wonder where this guy was? Or if he has any idea that Romney’s shadowy tax plan would either drastically raise middle class taxes, or massively run up the debt.

    That’s the thing that gets me about Romney supporters. When they say why they support him, it’s almost always with something demonstrably untrue. If you hate Obama and want to vote against him, just say so. But don’t talk about “the debt” and “unemployment” when Republicans have done everything they can to make things worse on both fronts.

    Or maybe the guy’s just rich and wants a tax cut. But at least be honest about it.

  • Venkat Ranganathan

    Sam, do you have an election day update on the house?


  • John

    Just voted in Poughkeepsie, NY. The poll worker I talked with said that turnout was roughly double 2008.

  • Till Sawala

    Dear Dr. Wang,

    I understand that you have been extremely busy during the election cycle, and I’m really thankful for all the updates and explanations you have provided. Once the election is over, I would very much appreciate more detailed discussions on how to deal with correlated errors in polls, and how to assess the performance of polls (or other prediction methods) – not in terms of the success in predicting the ultimate outcome of a (weighted) coin toss, obviously, but in terms of predicting the evolution of the prediction itself, given the stated margin of error. Looking at the EV estimator, for example, the fact that the prediction post Denver is outside the margin of error of the prediction pre Denver might suggest that the margin of error pre Denver was too conservative. By definition, all future predictions should remain within the margin of error of past predictions, if the error is designed to allow for drift – unless, of course, there was an event that is more significant and rare than could be reasonably expected during an election cycle. To do this in a rigorous manner would obviously require looking at the data in greater detail and computing autocorrelations and partial autocorrelations, but this might provide real insight, both about the nature of the prediction, and indeed about the significance and rarity of the event.

  • Andy

    Just talked to a neighbor on the schoolyard this morning here in central Seattle, who was amazed that an Obama campaign worker showed up at her door last night to ask how she would be sending in her ballot (we largely vote by mail in WA). If that is the level of organization in a safe-Obama state, the Obama GOTV effort is humming.

    • Steve16748

      We all vote by mail in Washington, (or drop it off at a collection point such as a city hall), and our turnout rate will be between 80 and 90 percent vs. 60 to 65 percent for the nation. There are a few polling places for people whose ballot was lost in the mail or otherwise and for people with special needs. Everyone can vote in the comfort of their own home and a paper trail is left behind. The paper trail has settled several close contests here. I cannot fathom how people accept voting systems without a back up of the votes.

Leave a Comment