Final election predictions

I have to leave for home soon and since I still do not have the landline and internet access at home, I will be reduced on election night to watching the returns the old-fashioned way, on TV.

It will be interesting to compare the final election outcomes with the various predictions, so here are the some predictions for the elections from the sites I follow. When it comes to electoral college votes, some give an average value, which is unlikely to be the actual outcome because of the fact that the electoral votes are awarded in clumps and is affected by the fact that Florida is truly a toss-up state whose 29 votes could go either way. Virginia, North Carolina, and Colorado are three other states that are too close to call with great confidence.

But given the state of the other states, the predictions are converging on an Obama win.

When it comes to averages of the national vote, Real Clear Politics has Obama 48.8%, Romney 48.1%.

The RAND survey has Obama at 49.5% and Romney at 46.2%. When it comes to thinking who will win, 54.1% pick Obama and 42.7% pick Romney.

When it comes to electoral votes, Sam Wang had Obama 303, Romney 235 as the most likely outcome with the next most likely outcome being Obama 332, Romney 206, the difference being Florida’s 29 electoral votes. For total votes he has Obama at 51.1% and Romney at 48.9%, taking into account just the votes for these two parties. He has a 99% probability of Obama winning.

Nate Silver has Obama at 313 and Romney at 225, with a 91% probability of Obama winning the election. He shows Obama getting 50.8% of the popular vote with Romney getting 48.3%.

Drew Linzer has Obama at 332, and Romney at 206.

At the betting sites, where the share price serves as a proxy for the probability of a win, Romney’s fortunes have taken a dive in recent days. Intrade has Obama at 71.3 and Romney at 28.8, while the Iowa Electronic Markets has Obama around 80 and Romney at 20. I think that these betting sites should no longer be considered independent indicators of the outcome as they once were. Given the past success of the poll aggregators like Silver and Wang, I think the bettors take their cues from them.

If Romney does win against all these odds, the entire polling enterprise will have to undergo some serious soul-searching to see how they could have got it so wrong.


  1. says

    To be fair, the Republican party isn’t working hard to keep people from being able to answer poll questions yet. How much of and effect their disenfranchisement tactics will have will be interesting to see.

  2. Alverant says

    Thanks. It’s a small comfort now to help me make it through for the time being. But I’m not going to hold my breath until Mittens gives his concession speech.

  3. brucemartin says

    You say: “If Romney does win against all these odds, the entire polling enterprise will have to undergo some serious soul-searching to see how they could have got it so wrong.”

    But another possibility could be that the advance polling is correct, but that the vote-counting is wrong, due to machines or tabulation programs possibly having been manipulated by private interests.

    So if there’s a disagreement, then both possibilities should be examined. So far, the public advance polling has been fairly open and transparent. But the software for the voting machines is generally not even allowed to be seen by the Secretary of State or the County Recorder of each county. So, evidence-based judgment must first question the election counting, more than the polling.

  4. ed says

    “If Romney does win against all these odds, the entire polling enterprise will have to undergo some serious soul-searching to see how they could have got it so wrong.”

    Sigh, no they won’t.

  5. CJO says

    They’re in the process of stealing Ohio right now, just as it was stolen in 2004. But (squeezes eyes shut and wishes, real hard) it’s not going to matter. Romney needs Ohio to have a shred of hope; Obama can still win without it on a number of probable scenarios.

  6. says

    A Romney win does not mean the projections were wrong! That’s like saying you are “wrong” if you roll a 6 after saying there was only a 1 in 6 chance. If the methodology is sound, the projections should be considered correct. 10% events are supposed to happen sometimes. This is a common mistake in interpreting statistical projections, but I’m disappointed whenever I see it from someone I respect such as Mano.

  7. Mano Singham says

    You make a good point. There is always the chance, however small, that the result just happened to be not the one that was the most likely and that there was nothing wrong with the methods used.

    But I don’t think that that is the first explanation that will be given for the surprising result, especially since people like Sam Wang were predicting 99% chance of an Obama win. When a 1% outcome occurs, people tend not to dismiss as a statistical outlier until other options are exhausted. I think that the search will begin to see if there was something wrong with the methodology used and whether new factors that inevitably crop up over time played a bigger and different role than what they traditionally did.

    One of the big problems that pollsters currently face is that of reaching people with cell phones. Another is that the response rates pollsters are getting are getting really low. Currently they have to make adjustments for these things. If Romney wins, I suspect that there will be a lot of research into these factors before they conclude that it was a statistical fluke.

    But your point is important and well taken.

  8. Corvus illustris says

    Your point is one that Nate Silver has constantly made on the NYTimes’ 538 blog. A more nuanced version of “the polls are wrong”–the one that I get from the last paragraph of Mano’s post–is that “something’s wrong with the aggregators’ methods.” More likely, though, would be the possibility that the estimators given by the various polls are consistently biased (in the technical statistical sense). That is a possibility that Silver has considered investigating once the dust settles. He has the resources (and the huge collection of data) with which to do it, so anything he says on the subject should be well-founded and interesting. One hopes he goes ahead, even though we now know that “the only poll that matters” turned out as expected.

  9. says

    > we now know that “the only poll that matters” turned out as expected.

    This is just a subtler form of the same misconception. The election did not “turn out as expected,” because no one specific result was expected. That is, unless you gave “303 to 206 with some undecided electoral votes” a probability of 100%!

    Also, I think Nate Silver already has protection against bias built into his system, and will automatically take this election’s result into consideration in the future. At least, that’s how it used to work when I looked at it back in 2010.

  10. says

    I certainly agree that even a single data point (like this election) can be used to help decide between the null hypothesis that “the model is sound” and the alternative that “the model is biased.” As it turns out, I guess the election is evidence that the model is reasonably sound.

  11. Corvus illustris says

    ‘This is just a subtler form of the same misconception. [etc., etc.]”

    You attribute to me a misconception that I do not misconceive. Given the date and the fact that I understand what a biased estimator is, your irony detector should have lit up on the use of “expected.” I may have expected too much.

    ” … I think Nate Silver already has protection against bias built into his system … “

    Oh sure, and his blog has discussed it; but that is different from detecting and quantitatively discussing systematic bias in particular pollsters–which Silver mentioned as a possible future project in a fairly recent post of his. (I apologize for not giving a specific reference.) He did get a lot of flak from the (biased) media for steadfastly disagreeing with the poll of the day about the odds and/or the popular-vote proportions.

  12. says

    Sorry my irony detector was deficient.

    I have not been reading everything on Silver’s blog, so I will take your word for what he said about his plans.

  13. Corvus illustris says

    Totally off-subject, but the link on your cognomen is currently giving advice worth taking seriously: it’s the only way to dislodge our current duopoly (though It wouldn’t have been a good idea in Ohio). We’re unlikely to get a repeat of the way the Republicans replaced the Whigs.

  14. says

    Thanks for taking a look at my blog, and feel free to take this conversation over there in the future! I don’t want to hijack Mano’s comment section.

Leave a Reply

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