Polling Analysis and Election Forecasting

Final Estimates Tomorrow Morning

I entered nearly 50 new state polls in the most recent model update, posted earlier today. There have been over 30 additional polls released since then. I’ll wait a few more hours to see if any more come out, then run the model one more time, overnight. My final estimates will be ready in the morning.

In the meantime, you might have noticed that my EV forecast for Obama inched downward for the first time in weeks, from 332 to 326. That’s the median of my election simulations, but it doesn’t correspond to a particularly likely combination of state-level outcomes. Instead, it reflects the declining probability that Obama will win Florida (now essentially a 50-50 proposition), and Obama’s continuing deficit in North Carolina. I’ve updated the title on the chart in the banner to make this clear.

Depending on how things go with the final run, I’ll keep updating the chart as I have been, using the median. But I’ll also create a more true-to-life forecast, based on assigning each state to its most likely outcome. With Florida (and its 29 electoral votes) right on the knife edge, this will either be Obama 332-206 if the model projects an Obama victory there, or Obama 303-235 if the model shows Obama behind. I’ll also have all sorts of other tables and charts ready to go for comparing the election results as they’re announced.


  1. Paul

    I’d been wondering when/if the graph would ever move, given the state of Florida. I’d even wondered whether the graph was mode and not median. I really do wonder if your prior is exerting too much influence this late in the race…. That said, if Obama does win with 332, you will look quite prescient!

  2. Drew

    Paul: The mode may have actually been a more sensible choice, in hindsight. Anyway, I think Florida is just an anomaly. The prior has been adjusting the forecast upward from current estimates by less than half a percent for a while. Florida is the only state close enough for this small shift to make much difference. I’m curious to see what happens in the final run – whatever the model says, that’s what I’ll go with.

  3. Paul

    Drew: Thanks for the details about the effect of the prior. Good to know. If it is exerting influence over Florida and only Florida, that seems sensible: the state is essentially a coin flip at this point, and letting a prior tip the scales makes about as much sense as any other method.

  4. Allan Marlow

    We are witnessing an astonishing event, friends.

    While 538, this site, and others are creeping up to practically 100% for Obama, Intrade has stabilized at around 67% and is not budging. Intrade has over 150,000 contracts in open interest, these are very significant numbers.

    This is extraordinary. What do wagers know that models don’t know?

    I hope something will shed light on this little mystery.

  5. ML

    Intrade is a prediction site that is very easy to manipulate with a little money and programming nous. See the discussion on Intrade as to why the price seems to stuck below 6.80. The professional bookmakers have Obama at around 80%, e.g.

    The Intrade comments note that Intrade prices seems to be getting picked up in the mainstream media while the far larger betting world is ignored. It has been suggested that manipulating Intrade costs less than TV adds

  6. ML

    … costs less than TV ads that have diminishing value as the campaign wears on.

  7. whirlaway

    @Allan Marlow: There is no mystery! The reality is that Intrade is thinly traded. So, a few politically motivated gamblers with money to burn can place wagers that can manipulate the numbers. After all, the Romney campaign has spent a billion or so on ads etc. A few million on Intrade bets is no big deal.

  8. Jippy2031P@yahoo.com

    This looks like a nail biter election. None of the candiates seems to have broken free in the polls, but Obama seems to be the favorite. Read this article at cliffology.com which debunked the argument that Romney will win based on his lead among independents. http://www.cliffology.com/2012/11/how-true-are-the-polls-in-the-2012-presidential-race.html

  9. Paul

    @whirlaway Try about the thin trading, but I think even your explanation may be overcomplicated. My guess is that there’s a lot of money from both sides informed pure wishful thinking.

  10. Allan Marlow


    I wish it were that simple! Nothing keeps participants from buying and selling at the current price, go to Intrade and take a look. This isn’t a market being cornered or anything like that. During the time it took me to write this post, over 6,000 shares changed hands. That isn’t a manipulated market.

    Tomorrow it will be really interesting.

  11. Atrox

    All of the models based on polls assume that the polls do not have some kind of systematic bias. Romney fans have been arguing that they are biased in virtually every media venue where they have a voice. The Intrade market believes that the polls overstate Romney’s chances to a greater degree than this site, or 538, or Princeton. As Allan Marlow has cogently pointed out in another thread, your gut (yes, *your* gut) is more in line with Intrade than any of the prognostications from these sites.

  12. Commentor


    You are partially right. Intrade is thinly traded and therefore not necessarily efficient. But it has not been cornered.

    More likely, Intrade demonstrates that most people simply do not understand statistical analysis. Laplace invented the basic framework of Bayesian statistics in the 19th century, but most statisticians refused to believe it until the advent of computers forced them to face undeniable results in the 1990s.

    Watching CNN right now, they still speak in terms of a toss up, despite overwhelming evidence to the contrary. Even Intrade doesn’t predict a toss up.

  13. Alan Houston

    Although statistical models provide evidence that President Obama has a 90% to 98% chance of winning, I am very pleased that the mass media and TV reports that the election is a tie, and it could go either way.

    I saw some 80 year old black ladies wait four hours in line to early vote in Houston, Texas (Harris County limits the number of voting machines in black neighborhoods, the wait in wealthy white neighborhoods is 15 minutes to 30 minutes).

    Other black voters in Houston were forced to stand in line until 1 a.m., others were told they would be “killed by a bomb” if they did not leave the polling station. These folks were in line because “the election is a tie”.

    I did not have the heart to tell them that the President has zero chance of getting the electoral votes from Texas. I am hoping Obama supporters in Ohio are willing to stand in line three hours, or stay in line until 1 a.m., and refuse to run when tea party members stage a fake bombing of the polling center. If Democrats in Ohio believe that the election is a tie, they will vote, and if every Democrat in Ohio votes, the President will win another four year term.

  14. Allan Marlow

    You speak convincingly of what you believe is happening with Intrae, and you might be right. The question then arises why are you in the US right now (my assumption here), and not in Ireland making a bundle of money by acting on your beliefs? I would have gone there last week if I had felt anywhere near your level of certainty. In addition, Ireland is not only a gorgeous place, it is inhabited by some of the most sublimely beautiful young maidens you will find anywhere in the world.

    Therefore, unless you are so wealthy that making tens of thousands in two days is too meager a deal to interest you, I fail to see why you are not taking advantage of your understanding of the facts.

    This is especially the case if you think the market is being manipulated, because all that means is you can buy your shares at a better price over a longer horizon.. The eventual jump to parity is UNAVOIDABLE, that is how futures work.

  15. MarkS

    Allan, once again: I am not betting at intrade because of asymmetric utility and counterparty risk.

  16. Tobby

    Raining, and a bit chilly, here today along the Central Florida corridor. Democrats won’t go out in the rain. Probably worth another .5% bump for Romney. Yes, the .5% bump is a wild guess, but I’m afraid the Democrat’s fear of rain is not.

  17. MarkS

    We’re back up to 332! Good for all those (little, friendly!) bets I made! Sam and Nate both have 332 as the mode as well.

    Heading out to vote now!

  18. Finnegans Father

    A fascinating home stretch battle between the candidates, and an equally fascinating debate among the pollsters.

    My greatest concern, however, is the levels of corruption in the voting process. No polling can predict the results if substantial numbers of voters are pushed away from polls, if software misdirects results, or if fraudulent votes are counted.

    It seems to me that both sides have become so convinced that the other side will cheat, that almost any “counter-cheating” would be justifiable.

    And if both sides have pre-election polling that shows they will win, this complicates responses to even the most massive massive tampering.

    I’d like to believe this sort of thing only happens in third world countries, but I don’t like the direction of things, and polling may be playing a role in this downward slide.

  19. Steve

    Would you please explain to me why in terms that I can understand (if that is possible) why you think attempts to analyze poll data mathematically are not effective. I am an electrical engineer, but have been out of school for many years so my math is rusty. I think that I see pretty good agreement between slightly different approaches (pollster, 538, PEC) – wouldn’t this imply that a mathematical approach is valid? Also, what else do we have to try to understand the state of the race? Do I understand that you favor Intrade over other bettiing markets? Thanks.

  20. David

    @wheeler: Allan is proof that market agents are irrational. He won’t change his mind. It doesn’t matter that Intrade projected the Supreme Court rejecting ACA at an 80% chance. And if Obama wins, well Intrade got it right, kinda.

    But this election should silent pundits (both sides) who base their predictions upon gas in their intestines. The great Data Deluge and nerds are changing the game, and the old guys don’t like it. So they continue grasping at straws and seeing shapes in clouds (er, crowds), because they have not the slightest idea of probability or statistics done carefully.

  21. MattC

    Intrade is a really weird one. I can get 4.50 on Romney from my bookie here in the UK (William Hill), so that’s free money if you combine with Intrade on Obama. Romney was 5.50 a bit earlier, so there well be some of that going on. As a good indicator for Obama, William Hill had an under/over 300 EVs selection available for an hour or so. All the money went on 300+ and the selection suddenly disappeared as the 300+ odds dropped. I managed to get £120 on at 1.83. Feeling pretty safe on that one!!!

  22. peter thom

    I find it ironic that some of the discussion here involves the accuracy of a single (outlier?) betting establishment, InTrade, an exercise that would hold little credence for any of the statistical aggregators like Drew or Nate, or even PredictWise.

    First, the Nov. 6 odds at InTrade are 72.8% for Obama, not stuck at 67%. That might reflect a trend, or manipulation or a random walk. We can speculate, but who knows? However, this result is significantly different from other betting models. Betfair is at 79.7% Obama, and IEM (popular vote) is at 78.4% Obama. PredictWise, which aggregates two of these three betting sites, has shown a rise in Obama’s fortunes from about 65% to about 77% over the past 8 days.

  23. Paul

    wheeler’s cat: I agree that faith in markets can go much too far and overlook facts of human behavior, but…

    •Please• keep it civil.

    You can make your point without resorting to phrases like “pathetic” and “learn something.” Allan might be wrong, but he still deserves basic civility regardless.

  24. David

    @Allan:”This is extraordinary. What do wagers know that models don’t know?”

    A better question is “what do wagerers ignore that the models don’t?”

    Intrade failed in a major way on ACA, but to be fair, there was less data. And again, if you look at the housing bubble, the market agents ignored the warning of several prominent economists, to the tune of many trillions of dollars lost. So, market players need a reality check. The market is not as smart as they like to think, the accuracy of the market is inversely proportional to the amount of speculation on the position.

  25. Allan Marlow

    Intrade is starting to cautiously creep up. It is doing so as exit information is revealed. This suggests that the weak link all these models is last minute undecideds’ behavior.

    There is nothing wrong with having a piece of your model potentially faulty. What is wrong is to neglect the possibility that that piece may in fact be wrong. In the sites that report winning probabilities, their numbers should be multiplied by the probability that their faulty component functions properly.

    Of all the sites, PEC is the one that most egregiously violates this simple principle. I say this because even Sam Wang acknowledges that his assumption of zero net contribution by last minute deciders has been wrong in the past (when it said Kerry would win.)

    Therefore, I think the paradox is resolved, and it has nothing to do with preference curves, it has all to do with the models being wrong in that they don’t properly characterize undecideds’ behavior “during” election day.

    (Nervous girls needn’t get agitated about this.)

  26. Drew

    Just to jump in here: I unfortunately had to delete a comment. Please keep the tone friendly. Relax! Enjoy. The election is finally here.

  27. MarkS

    Allan, you have the facts wrong. (Why am I not surprised?) Sam at PEC predicted a Kerry win in 2004 because he assumed (incorrectly) that the undecideds would favor Kerry. Had he not made that extra assumption, his model’s prediction would have been correct.

  28. Michael

    Looking across the various polling analysts, it looks like it is either 303 or 332 depending on FL, which is very close. Given its history, I will assume FL finds some way to go red. It always seems to inexplicably slide a point or two in that direction at the last minute.

    Unfortunately, so does OH, but that one looks like a 3 point margin so should be OK.

    I believe the only wild card is a possible large miss on turnout estimates that skew one way or the other. Given the huge lines at early voting locations, the very high early voting totals, and reports of lines this morning, I would think that turnout will be high, which generally should favor Dems.

  29. Allan Marlow


    I don’t know how informed you are about asset pricing theory. It is a profound subject and it is the case that the market informs the math, not the other way around.

    Three people were awarded the Nobel Prize in economics precisely for clarifying this issue. I don’t want to be any more precise than this, but this isn’t something you can go and convince yourself of by a cursory reading of a couple of papers, anymore than you can change someone’s views on Quantum Electrodynamics when all that person has to argue in his favor are his weekend readings in Scientific American.

    If this subject interests you, and if you are willing to put in the effort, there are good sources, but be prepared to go far beyond the CAP model and simple bond and option pricing. Nobody wins the Nobel Prize for resolving a trivial issue.

  30. weichi


    “Of all the sites, PEC is the one that most egregiously violates this simple principle. I say this because even Sam Wang acknowledges that his assumption of zero net contribution by last minute deciders has been wrong in the past (when it said Kerry would win.)”

    ??? Isn’t this 100% backwards? Sam’s model (which, as you said, implicitly assumes undecideds break evenly) predicted a *Bush* win in 2004. But, as he has described, he was wanted a Kerry win so much that he managed to convince himself that the undecideds would break for Kerry, and this led him to manually add a fudge factor in Kerry’s favor.

  31. Allan Marlow


    Yes, I got the mechanics of Wang’s “mistake” backwards, but that isn’t the point, is it. The point is that he HAD to REVISIT his assumptions in the PAST, setting a precedent for his assumptions having a probability of being wrong. THAT probability should be part of his prediction TODAY, and it is NOT. That is why he gets an absurd 99.9%, and that is also why he is not making a bundle of money on that “knowledge”.

    I wish PEC were right! As I said, if I felt I believe those numbers, I would have flown to Ireland last week, and would be enriching myself as we speak.

  32. Trim


    What others have said, his 2004 model made the common belief that undecideds break for the challenger. When he eliminated that bias from the model, it found that Bush had won.

    Honestly, I think you’re just grumpy with Wang because he told you to “get off [his] lawn” after you repeatedly made some really obnoxious posts.

  33. Allan Marlow

    Trim, I have no idea what you are talking about. I went to PEC only once, the 99.9999% turned me off.

    It does seem that not many people are visiting PEC and similar sites, who put their money where their mouths are (it seems I am the only one here).

  34. weichi

    OK, alan we all get your point. No need to keep harping on it. Let’s talk about something more fun … like unskewedpolls.com!

    Personally I’m still gobsmacked by by unskewedpolls.com’s final prediction of 275 EV for Romney, with 50.67% of the popular vote. In some sense it isn’t surprising: it would be really tough to keep the con going past today if his final prediction was the usual Romney landslide. But I guess I didn’t really believe that he would do it; for some reason I thought he really was that dumb/ignorant, and discounted the possibility that he was just a grifter.

    Has anyone been following his “analysis”? What’s his excuse for the big shift? Didn’t he have Romney at like 380 EV and + 6-7% just a few weeks ago? It’s really hard to find old predictions on his site (another sign in how little interest he ever had in providing good information about the election to his readers).

  35. Allan Marlow

    I am naturally disinclined to read too much into right-wing “science”, but like anything, they add information, since it is impossible to subtract information (filtrations diffuse, for those into stochastics.) This time around, the unskewing guys might be right for the wrong reason (just like PEC is wrong for the right reason)

  36. Michael


    I agree the sudden shift to reasonableness for unskewedpolls cries out for an explanation. Prior to this change they were part of the Romney landslide crowd with Gingrich, Dick Morris, and George Will. Suddenly they have what is almost certainly the most likely Romney win scenario as their final prediction. Most interesting is that they gave up on NH, WI, and NV.

  37. Michael

    p.s. unskewedpolls had Romney at 311 just five days ago, including NH, MI, WI, and IA.


  38. weichi


    But the thing is, is there any “reason” at all going on at unskewed? If not, then it’s not telling us anything about the likely outcome of the election (unless perhaps indirectly, through revealing information about the state of mind of republicans … maybe we could argue that by latching on to such transparent BS, republicans revealed how low they really thought their chances in the election were).

  39. Scott

    Nailed the results weeks ahead…. Hats off!

    I must admit I sort of discounted this forecast in favor of Real Clear Politics as you were so often more favorable with Obama’s result and as one who did not want to assume such a margin and take it as a given win I was getting a bit concerned until Obama was up in the RCP national poll average. Next time I will pay much more attention to the votamatic forecast for sure! Great work!!!

Leave a Reply

Your email address will not be published. Required fields are marked *


Theme by Anders NorenUp ↑