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

Next campaign stop: Not New Jersey

October 14th, 2012, 10:00am by Sam Wang

I started doing poll analysis in 2004 because of an extremely close Presidential race, Bush v. Kerry, in which the candidates traded the lead several times. The effective margin was never greater than 3% in either direction. It seemed critical to get a good read on what all the state polls were telling us.

Compared with then, this year’s race is shifted toward President Obama. On average he has led by 3.0 points as measured in Popular Vote Meta-Margin. Obama v. Romney is a less volatile contest, and the lead has not switched, Ohio being a fairly considerable block to a switch.

However, at the moment things are quite close. The natural question is, where are your efforts most effective? The answer turns out to be the same whether you support Obama or Romney.

In 2004 I expressed it in terms of the value of an individual vote compared with my vote in New Jersey:

This question can be answered by calculating how much the Electoral College win probability is changed by one person’s vote. This affects where you should go because as an individual, you can only get out a finite number of votes….Here is a case study. If you are a New Jersey resident, your vote has some value, but it is low since the state is very likely to go Democratic by a substantial margin. In contrast, driving a voter to the polls in Pennsylvania is worth nearly 300 times as much. If you go to Ohio each vote is worth even more, over 500 “jerseyvotes.”

In 2008 I tweaked it a bit (explanation, August 8 2008). Now, voter influence is now calculated under the assumption that overall, state polls have shifted by a constant amount, one that makes the win probabilities 50-50. This change is equal and opposite to the Popular Meta-Margin. I have also normalized to the most powerful vote. Like the a developing country’s currency, jerseyvotes fluctuate a lot in value. The results are listed in the right sidebar as “The Power Of Your Vote.”

The application of the jerseyvotes calculation is to identify where local resource allocation (such as get-out-the-vote operations) will have the most effect. Utah residents should make travel plans to Nevada. Massachusetts activists should flood to New Hampshire. And so on. The calculation is similar for ad dollars, but not exactly.

And yes, everybody in the Midwest should go to Ohio to turn out people who have not already voted. Early voting has been big this year.

Tags: 2012 Election · President

62 Comments so far ↓

  • wheelers cat

    But what if a better tactic is phone canvassing on smartphones and Facebook canvassing?
    Exploiting social media?
    Consider cell phone useage in battleground states.

    • ML

      I worked for the Obama campaign in 08 and the field data guys always told us that it took something like 5 times as many phone convos and ten times as many read emails to have the effect of 1 convo on at the voter’s door. My personal experience bore that out, and after a year on the trail, I had a pretty good sample size!

  • Olav Grinde

    I have an additional suggestion:
    How about a swing-state by swing-state listing online newspapers and other websites that have debates? Well-phrased posts that respectfully address the concerns of voters in those swing-states might well have a positive effect. At least some effect…

    And as we know from Florida 2000, it may not take much.

  • Olav Grinde

    Yesterday, NC Obama Guy wrote a post suggesting that the very desire to participate (or not) in polling was dependent on enthusiasm. And that this might explain the striking differences seen between the Rand Poll and traditional poll.

    I think this makes a lot of sense.

    Dr Wang has already commented on the remarkable coincidence of right-leaning pollster flooding the news with polls immediately after the debate. I believe there is a very systematic effort to control and to drive the narrative — far more so in this election than previous ones.

    Polls have become as much a campaign tool as a bona fide metric. It seems very clear that Rasmussen, for one, is as interested in driving the narrative as actually measuring opinion. All you have to do to confirm this is read Scott Rasmussen’s own narratives.

    Thus, I believe, some pollsters have taken a leading roll in manipulating perception. Using questionable methods that consistently skew polls, is not necessarily just bad science — more importantly, it’s good politics!

    We know that there is a strong link between enthusiasm and the likelihood to vote.

    Thus, increasing the enthusiasm of voters on one side while decreasing the enthusiasm of voters on the other side, is good politics. This will certainly influence turnout — and it may well decide the 2012 election.

    • wheelers cat

      what if President Obama just told the truth?
      And that truth is that the demographic timer has already gone off and the election is entirely dependent on “enthusiasm” and not on “choice” or “referendum”?
      Not that he would do that, lol, because he is the president of “all the americans.”

    • Anbruch

      I’m glad that Sam let your point through, because I think it’s important. As I understand it, poll aggregation only works so long as its variation is randomly distributed. At what point would systematic distorted polling be sufficiently large to affect the aggregation? What would the statistical signature of such systematic distorted polling look like? Does the aggregation look for this signature?

    • Olav Grinde

      * leading role (apologies for spelling)

    • Ms. Jay Sheckley

      It’s irritating to read Rasmussen’s blog at least in part because his vision of what’s going conflicts with the one I follow [this] and he consults no polls but his own…
      However his text if not his forerunner goes down well and his general point of view seems sanguine.
      The “voter enthusiasm” issue is disturbing. If you think you might lose, you dont vote?? I get down too, believe me. But it’s just one little task every 4 years. I couldnt live with myself if I hadnt tried. Makes more sense not to vote if youre sure your guy has the race sewn up. Now I feel not only betrayed by my own, by like an Alien:” In booth, no one can hear your scream.”

  • A New Jersey Farmer

    So far, at least in Ohio and Iowa, early voting statistics show that Democrats are still very enthusiastic. A good sign.

  • Louis in NY


    In light of your recent analysis, I thought you would appreciate this recent piece in the WashPost on the five myths of polling, especially post debate:

    -Internal campaign polls are more accurate than public polls
    -Polls prove the first presidential debate radically changed the race
    -If a poll nails the election results, the pollster’s methodology must be good
    -A sample that doesn’t contain an equal number of Democrats and Republicans can’t be correct
    -News outlets are biased in favor of President Obama


  • Bender

    Professor Wang,

    I’m an editor, a word guy. Please take no offense, but when I see numbers and “meta-margin” this and “metric” that, my eyes start to glaze, and I get sleepy.

    In no way am I discounting the importance of math and numbers. In fact, I have a young son who’s mathematically oriented and wishes to be an architect.

    I think that’s awesome, and I feel the same way about your work and all the other numbers-rich sites I visit every day.

    Here’s my bottom line, though: I’m an avid supporter of President Obama. Do you still feel, as you stated recently on your site, that the chances of him getting re-elected stand at 93%?

    In other words, is his re-election at all in doubt, according to your calculations? Should I remain in a constant state of worry until Nov. 6, or should I go ahead and buy the bubbly?


    • Renee

      I’m not Sam, but the only advice to give all wobbly Dems is…Go VOTE!

    • Sam Wang


      Imagine a six-gun loaded with five blue cartridges and one red cartridge. The blue ones will pass through you harmlessly. The red one, not. You pick up the gun.

      Fade to black.

      You may glaze again now.

    • Olav Grinde

      Strangely enough, I find the thought of playing American Roulette less comforting than its Russian counterpart.

    • Prairie Pundit

      Having grown up in a place where folks are far more likely to combine guns and alcohol than gin and vermouth, I think what Professor Wang is telling you is that it’s totally o.k. to buy the bubbly now, but just be prepared for a 1-in-6 chance that the bubbly will be consumed at your own wake. In any event, your purchase will not go to waste.

    • Amitabh Lath

      Bender, I hear your pain. Probability is filled with counter-intuitive phenomena, and Sam actually does quite a good job not getting hung up on the math jargon.

      But there are some concepts that require the precision of the vocabulary related to the field.

      A “median” is kind of like an average, but not really. An “estimator” is not quite a guess, or even an educated guess, but something more specific.

      Might I suggest a book? It’s called “Innumeracy”, by John Allen Paulos. It is well written, funny, and demystifies everyday mathematics. I laughed out loud at some parts, and learned quite a bit.

    • Olav Grinde

      May I also suggest a book, by my wonderful teacher Sherman Stein — Mathematics: The Man-Made Universe

    • Olav Grinde

      Dr Wang, that is a striking image.
      Strangely enough, I find the thought of playing American Roulette less comforting than its Russian counterpart.

    • Albert Ericson

      So, if it is one in six, it is down to 83% now from a previous 93%? Any movement down sets off my anxiety!

      I have never posted before, but this site has always been my favorite. Thanks for all the great work!

    • Sam Wang

      Yes. I thought it was readable from the red strike zone. I’ll write on this.

    • Anbruch

      @Sam, it is apparent from the red zone, but you have to understand what the red zone (and yellow zone) mean in terms of percentages.

  • Richard Vance

    I’m still not convinced that Jersey Votes is a linear track with dollars contributed. The cost of advertising varies wildly between media markets thus the cost of moving a single vote is not 1:1 for each state. I broached this in the Kerry-Bush blog comment and got the answer that the “where do I put my $” question IS to use the Jersey Vote.

    But IF you spend all your money on Ohio and lose it you are screwed because you failed to carry the small states. So nail down the 270 EVs by using more bang for the buck advertising to seal NH, NV, and switch CO. Spend no more in FL or NC but at least visit VA and OH. I would saturate CO… Say again carry NV, NH, CO you can lose FL, OH, and VA and win. I’m an old coach and I’ll take a 270 EV win any day… stop swinging for the fences and manufacture runs.

    • wheelers cat

      You guys, Romney is absolutely not going to win Colorado.
      He is a Mormon.
      That is going to peel off some of the rural votes (that he desperately needs) on the western slope. They share a border with Utah and their ancestors fought the water wars.
      I think some poll responders wont say they arent voting mormon. Some false positives, and Romney cant aford to lose a single white vote.
      No way, Jose.

    • steve in colorado

      Best we can hope for is that the anti-mormon vote skips the election or goes 3rd party- I know some people here who live in Weld county that are absolutely convinced Obama is a Muslim. The choice for them is between a Muslim and a Mormon. Well, at least one of these people has switched to Romney (rather than nothing) since the debate.
      Colorado is going to be tough, I just hope we can get enough turnout in Boulder, Denver and Ft. Collins.

  • V

    Very interesting. I apologize if this has been asked elsewhere, but how do you account for the sudden flood of GOP affiliated polls that have gummed up the works on all the forecasting sites? I don’t mean to sound tinfoil hat conspiracy theorist, but when I looked at the polls coming out after the debate, the GOP affiliated/house effect pollsters were just FLOODING in, endlessly. It seemed very convenient to drive a narrative.

    Back in 2008, there were very few partisan or shady pollsters (mostly just Rasmussen, PPP, and Strategic Vision, which has since went out of business). Therefore, they didn’t really have too much of an effect on the forecasting sites. But this year you have:

    Rasmussen: Flooding the zone more than ever. I don’t think I need to detail their extensive GOP house effect.

    ARG: Nate Silver and RCP both saw them as a discredited firm during 2008 due to their awful 2008 primary polls, yet now they both use them as if nothing happened. They frequently have extremely rosy numbers for the GOP. They even showed Obama up 8 in West Virginia in 2008, which he ended up losing by 13 points.

    We Ask America: Connected to Illinois Republicans and a GOP lobbying group. They recently came out with a poll showing Joe Walsh leading Tammy Duckworth when every other poll shows her leading by double digits.

    Gravis Marketing: Appeared out of nowhere during the RNC and started cranking out pro-GOP outliers like there was no tomorrow. According to them, Dean Heller is up 17 points in Nevada and Romney is up 9 in NC, and only down 2 in MI.

    FMW/Baydoun: Had polls showing Romney up 15 in Florida, and Romney and Hoekstra up in Michigan. Had voters under 50 years old being about 10% (or some other similarly ridiculous number) of the electorate.

    McLaughlin: This is a GOP internal poll, yet many people include it in their forecasting for some reason anyway.

    And what Democratic polls are there to counter act these? Pretty much only PPP. I can’t help but feel like this broad discrepancy would destroy the median poll, and, for lack of a better word, “skew” the analysis.

  • Amitabh Lath

    Sam, if NJ were a swing state, then a couple of things would be happen.

    1) Traffic would be a mess every weekend. Imagine Rt 1 and Rt 27 clogged or even shut down every weekend.

    2) A lot more of your neighbors would be R-voters.

  • pechmerle

    “The second point is false, but thank you for the rest.”

    Sam, this comment by you to Louis i NY was a little too condensed. The WaPo article he cites says all five points are false.

    So, is your comment a double negative that says point two — the debate radically changed the race is true? Or did you think his other four points are true (which they are not)? Clarify, please.

  • Bill N

    I have posted here a couple of times that I have been shocked at how rapidly the meta-margin dropped the past ten days. I have thought about it a bit and have a theory that may be reasonable, but then again may be nothing more than WAT (“wild as* theorizing”). My area of specialty is not political science or polling or voting behavior. I am a Social Work researcher, (who happens to have an undergraduate in engineering science…a little knowledge is a dangerous thing), so bear with me. Assume that Obama got a double whammy bounce from the convention and then from the 47% tape. This bounce was just ending and the meta-margin just beginning to move back towards the mean (or median if you prefer). Just at that time the first debate occurred. The negative reactions to the debate served as a perturbation that pushed the meta-margin at a resonance point that accelerated it downward much faster than it would have without the impetus of the debate fallout. It was kind of like someone pushing a child on a swing just as the swing reaches the peak of its arc and is just beginning to swing back in the opposite direction. This perturbation hitting at a resonance point not only accelerated it downward, it also forced it further down than it would otherwise have gone.

    Maybe some of you who post on this site, and who have more knowledge in this area than I, can comment on this idea.

    • Richard Vance

      You have picked up on a law of nature. Momentum.
      Bodies at rest tend to stay at rest and bodies in motion tend to stay in motion. Barring infinite force application (the unit pulse function) changes in velocity takes time thus causations do not neatly correlate with the MM. Sam had previously discussed how the MM dv/dt had slowed ahead of the VP debate but now that it has stopped and reversed the VP debate will be provided the happy distinction of CAUSE. Joe can happily bask in the warmth..

  • dawolf

    Gravis Marketing is very likely fraudulent

    Sam – this is something I put together. I believe its solid, but would welcome your thoughts (and, of course, if it is fraudulent, it stops you using them :) )

    • Sam Wang

      I actually think they’re on the level. They could get their results by bad weighting.

    • Arbitol

      Gravis is run by a guy who graduated from college in 2008. It needn’t be fraudulent to be dismissed as worthless.

    • dawolf

      Sam – they can’t get their results by bad weighting. Look at the actual poll results. At one point they have other/unsure saying 0.05. At another, 0.06. At another, 0.07. You don’t weigh in 3 different ways according to the question! They also make absolutely NO mention of demographic weighting at any point.

      Then, you can weight one of their dodgy results, so you scale say 0.07 to 0.109, then check other results on the same line – they don’t make integers.

      Please check it again. Not with a thought of “probably ok”, but with a thought of “what does the evidence say?”

    • JR

      I have no doubt that Gravis is R-leaning, and is suspect enough because of their low-quality methods (i.e., robopolling) that we should consider dropping them from future election cycles. However, in this case, I don’t see chicanery. I use surveys for work (market research), and this seems like simple weighting to me. Respondents are likely weighted by demographics (e.g., race, age, gender, region). That’s why the weighted tables can have non-integers. As for why different questions have different numbers below 1 (e.g., .05, .06, .07), I suspect that that’s just a DIFFERENT single respondent each time, with his/her respective weight reflected. (That is, one time it’s a white woman from Florida, another time it’s a Latino man from Arizona, etc.)
      So, while I am all for investigating Gravis and deciding as a scientific community whether to reduce their impact or throw them out, in this particular example I don’t see any evidence of manipulation.

  • Amitabh Lath

    Bill N, I too have been wondering about the sudden drop in the meta-margin. It seems unphysical to me, and makes me wonder what exactly the polls are measuring.

    I too thought that the convention, 47% bumps were somewhat stale, and waiting for something to puncture the +O narrative.

    I am convinced that Obama will win.
    I am redirecting my (rather pathetic) donations to the Senate and House races.

  • CobaltinSF


    Just released, Gravis Marketing poll (which has appeared to favor Romney) of 2089 people in Colorado from 10/5-10/11 shows an OBAMA LEAD OF 48.4 to Romney 46%. Colorado has been red the past few days, first evidence it’s hueing blue again…

  • BillSct

    -Bill N.,

    My comment from yesterday pointed out that all of Pollster’s swing state graphs, except Iowa, that are linked to in the left hand margin, show an Obama bounce beginning in late August and reaching a maximum value between 21 and 24 September. The bounce was well on its way down before the debate. I speculated that Obama’s bounce was a combination no convention bounce for Romney and a big convention bounce for Obama. I had not considered the impact of the 47% tape but I think it is reasonable to conjecture that it was a factor too. I too had considered that what we might be seeing in the dramatic fall of the meta-margin was a combination of an already existing downward trajectory propelled past its natural minimum point due to extra energy being applied to the system (e.g. a weeks worth of the media having nothing else to talk about except Obama’s lousy debate performance.) The other analogy I have considered is what Oceanographers know as a “Rogue Wave” where several waves all with amplitudes of the same sign momentarily intersect to create a wave of enormous size compared to the surrounding waves. The puzzling thing in all of this is that all of the swing state graphs flatten out and/or turn in Obama’s favor at the time of the first debate and afterward. I also find it interesting that the first step in the falling meta-margin occurs roughly at the mid-point of where things have been since may (roughly 3.3%, by eyeball), the second step more or less coincides with Ryan Bounce trough, and the last low point, which is where are now is roughly the same amplitude as the previous peak, but of opposite sign. If we really are oscillating about a set point then we should see a rebound and then a series of waves decaying in amplitude. Of course, we are a day away from another debate, which should put some more energy into the system, followed by the media, which apparently function as a shock amplifier and not a shock absorber.

  • DaveM

    I like the wave metaphor. Romney’s riding the pipeline–will his board smack him upside the head when he wipes out? Will Obama’s beach of support still be there when the rough surf recedes?

    And it’s enjoyable to think of Joe Biden as breakwater…

    But seriously, the additive amplitudes of multiple waves coinciding is an image I’ve had as well, as the MM has dropped through more than one anticipated floor.

    • wheelers cat

      If you look at the MM its leveled off.
      Looks like the Red Wave just got turned to beach break.

  • wheelers cat

    Holy kurtosis, Sullivan is about commit seppuku.

    the guy is like a parody of himself.

    • CobaltinSF

      He’s become unhinged ….it appears to coincide with his recent move to New York….

    • Olav Grinde

      Pity President Obama can’t send his Secretary of Explaining Stuff in the ring. Clinton would strip away Emperor Romney’s new clothes.

  • BillSct

    Having been on insulin for over forty years I’m familiar with wave phenomena that have long time delays and behave in counter-intuitive ways.

    Somogyi Rebound


  • BillSct

    I also know, that if you ride it out, things come back…

    I am still intrigued by the fact that all of the falling state graphs either level out or turn in Obama’s direction at the time of the first debate. And I’m fascinated by the fact that Iowa never registered a bounce of any sort, but both Obama and Romney are trending up at about the same rate. Could it be that having been inundated by campaigning for almost a year that Iowan’s are saturated, have made up there minds, and that’s that?

  • BillSct

    Actually, now that I look again, the leveling off in most of the graphs occurs sometime after the first debate.

  • Ms. Jay Sheckley

    Marvelous modeling!
    First I’m shot through the heart, a pundit commits seppuku, we surf the curl, then lust after a Somogyi rebound.
    Which reminds me of what Woody Allen said:
    Her figure described a set of parabolas that would cause cardiac arrest in a yak.”

  • Ram

    My Take…. This is Romney’s best run and the current status to date as per RCP (apart from this site MM-Obama +0.64) is below

    Ohio-Obama +1.7
    Florida -Romney +2.8
    Virginia-Obama +0.4
    Iowa-Obama +3.2
    Colorado-Romney +0.7
    Nevada -Obama +1.6
    North Carolina -Romney +4.7
    New Hampshire-Obama +0.7
    Wisconsin-Obama +2.3
    It looks extremely difficult for Romney to achieve 270 unless some other trigger (??????) kicks in. The key states are Ohio, Virginia and Colorado.

    • wheelers cat

      I get the same vibe from Benghazi that I got from “Fast and Furious”– as desperately as Romney/Ryan and NRO are flogging the story, no one seems very interested, most importantly, the media seems uninterested.
      Plus Romney has to handle this delicately– he runs the risk of reminding the electorate that Obama got OBL and that Romney was handing out bibles in Paris as a mormon missionary during Viet Nam.
      Which goes back to the age thing.
      Obama is 14 years younger than Romney.

  • wheelers cat looks like Romney’s bounce is over having never flipped the curves.
    One sure sign is our conservative commenters seem to have disappeared…..and another is RAND.

    • Pat

      @wheelers cat
      Actually, there is something I can’t quite understand with the RAND poll. Today, these last few days, the curves are spreading apart once again, with Obama gaining ground, now reaching 49.9% and Romney losing ground, today reaching 44.9%.
      However, their “shift between candidates” seems contradictory, with MORE people shifting from Obama to Romney than the other way around.
      If so, how can Obama be gaining ground on Romney?
      I must be missing something…

    • wheelers cat

      Look again.
      The fractions of shift are very small, and the graph is very noisy. Nearly all shifts are within the MOE, and thus may be due to chance alone.
      And the scale is 1/2 of one percent of the respondents.
      With that much noise you cant really say anything definitive.

  • pechmerle

    Well, OK, but isn’t the RAND forecast graph running over two weeks behind real time?

    In which case, it is only just now beginning to reflect the time period of the first debate?

    • wheelers cat

      1/7 of the captive population responds each day for a week. So the effect persists for a week. Usually the response is immediately apparent and trends for the next six days.
      RAND posts at 1:00 am Pacific.

  • A New Jersey Farmer

    Looking at the RAND survey, why have both candidates lost almost 1.5% each in the “Willing to Vote” graph?

  • Olav Grinde

    Re: Breakdown of early votes

    Today, writes: “Early voting has started in over 40 states and 7% of the voters have already cast their ballots. These early votes have broken strongly for President Obama, 59% to 31%, according to a new Ipsos poll.”

    Am I reading those numbers correctly? Last time I checked, 59 + 31 = 90.

    Does this mean 10 % of the polled early voters have checked the box for a third-party candidate? Surely that exceeds all expectations!? And if true, it is a game-changer.

    Or am I totally misreading this?

    • Ohio Voter

      It’s also possible that people are casting ballots and not voting for President.

    • DaveM

      Perhaps it means that 10% of the poll respondents who were willing to acknowledge already having voted were unwilling to divulge for whom they voted.

  • pechmerle

    @wheelers cat: I see what was confusing me. On screen, RAND’s graph appears to stop at 10/1. Then I realized there are two 10/1’s — and the second one is really 10/14 although you have to hover over one of the end points on the blue/red curves to get the 4 to show.

    I feel much better now!

  • Ohio Voter

    Pretty positive ABC News/Washington Post poll for Obama today. +3LV

Leave a Comment