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

An under-the-radar runaway for Obama?

November 2nd, 2008, 10:00am by Sam Wang

By current standards the McCain campaign is not spending that much on polling – only $1.1 million since July. During that time the Obama campaign spent almost four times as much. Maybe that’s why on Friday, McCain campaign manager Rick Davis felt able to write a memo containing plainly absurd statements such as “Our numbers in Iowa have seen a tremendous surge in the past 10 days.” Obama is up by 11% in Iowa. With little enough data, Davis could convince himself of all kinds of swings.

He needs to. What’s amazing about the Presidential race is how little change has occurred in October in national polls and in meta-analysis of the Electoral College. By these measures, the campaign ended after the first debate. However, there are still a few surprises lurking in the data. Battleground states have presented a steep hill for McCain to climb. In the last four weeks, the hill has grown into a mountain.

The height of the mountain can be measured using the Popular Meta-Margin. But first…

Let’s look at national surveys since so many of you fret over them. These have been at a near-standstill for weeks:

As you can see from this graph custom-generated from (have you tried that? it’s fantastic!), it’s hard to see any movement. Here’s a list of values:

Date range
Obama-McCain national margin
10/3-7 5.0 +/- 1.1%
10/8-11 9.0 +/- 0.7%
10/12-17 5.0 +/- 1.2%
10/17-21 7.0 +/- 1.1%
10/28-31 7.0 +/- 0.9%

On the other hand, in the last comment thread Frank pointed out that eight tracker polls have shown a widening of 1.5 +/- 0.5%. If true, that should show up in the Meta-Analysis on Monday morning.

The Median EV Estimator. In terms of stability, it’s like 2004 all over again, when the race stabilized after the first debate. In the last few weeks the short-term fluctuation has increased because of increases in polling frequency.
History of electoral votes for Obama since April 1
For the last 3 weeks it’s been bouncing around Obama 360 EV, McCain 178 EV.

The Popular Meta-Margin. Here’s the surprise. This is the amount of across-the-board shift in Obama-McCain margins that would be needed to make the race a perfect 269-269 EV toss-up. Here’s a graph of it over time:

Since mid-September, when Sarah Palin spoke unscripted and Lehman Brothers collapsed, the Meta-Margin has been increasing steadily. It’s now twice as large as it was at any point during the summer.

The increase in the Meta-Margin over the last four weeks is quite striking. Despite the fact that national surveys and the Electoral College have been solid during October, the effective gap between the candidates has been steadily widening in battleground states. The cementing of Obama’s advantage is a hidden story of the campaign endgame.

Tags: 2008 Election

59 Comments so far ↓

  • E. Duvert

    Thanks for doing this post. Very reassuring.

  • MM

    Yes, thank you. We are all antsy to read something not just speculative and to see at least something under the stasis, which is almost as nerve-wracking as see-sawing would be.

  • D. Raphael

    I’ve copied and sent to several of my friends this site and that of to calm their attacks of nervousness. Many of them, after reading one or another poll, are given to shrieks and fainting spells. I find your polling practices and that of the aforementioned site reassuring. I concur with E. Duvert. I thank you for your reassuring post. Not just because it bolsters my hope regarding the election outcome, but more importantly, we need sound thinking to replace soundbites. We need reason to replace rashness. We need hope to replace hobgoblins. Thank you again.

  • The Liberal Crab

    Your stats and the fact that Barr/Nader isn’t included in many of these polls – definitely are going to contribute to the difference.

  • Sam Wang

    D. Raphael – My original hope in 2004 was that this kind of analysis could replace all the bloviating about individual polls, thereby making room for discussion of other types of campaign news.

    This hasn’t happened yet – the advent of FiveThirtyEight is a blessing for poll geeks, but also a curse because of the celebritization of the process. In the meantime, keep reading!

  • jay berk

    Truman went to bed thinking he had lost…what a surprise….the Bradley is out and about…socialism isn’t welcome…a steady experienced hand counts….not rhetoricky…Mac is back….I know many who smile and say they haven’t made up their minds…they have!

  • Sam Wang

    jay berk – I used to screen purely political posts without analysis-related commentary. But lately I’ve weakened. Also, these comment threads have gotten ideologically rather one-sided. Please come back Wednesday for the postgame commentary. You betcha!

  • LB

    Hi Sam,

    While the GOP is hoping for the Bradley effect, I think what you’re outlining here is the Coriolis effect-that is McCain is circling the drain. Thanks for all your work; holding the facts above the fray.

  • Wagster

    One of your best posts, Sam. I’ve heard this point before but your last graph makes it very stark and very striking. I would guess that this effect is due to Obama’s big advantage in TV advertising. His decision to skip public financing might well have been worth 30-40 EVs.

  • Mike T

    That graph of the meta-margin looks eerily similar to a graph of global temperatures in the last 120 years… I mean, STRANGELY similar.

    Ok, that was totally irrelevant, but I’m a climate scientist, and it really stood out. :)

    As for the polls, I agree that people should stop freaking out about one individual poll.

  • a

    Tell your friends, and show them the proof.

    Adolf Hitler was well-known to have used hypnosis on crowds to gain power in Germany. (Just Google it if you doubt it.) So why cant this happen in America? It is happening.

    Now, this document, An Examination of Obama’s Use of Hidden Hypnosis Techniques in His Speeches is really spreading. Many people, including young people, are starting to wake up.

    Young people have come up with a saying, “Don’t drink the kool-aid Obama is pushing on you.”

    Obama has no accomplishments, we don’t know him, he wrote a memoir, he has racist connections, he is using hypnosis, and he thinks he’s the Messiah and that he should be in control of the world. We’ve seen this story before in Germany in the 1930’s, haven’t we? We know what happened there.

    Obama’s use of hypnosis in speeches is why:
    -nothing sticks to him
    -the huge crowds
    -the huge money contributions
    -the mesmerized effect, especially in young people, who, because of their imaginations are more susceptible to hypnosis (read the document)
    -the exorbitant election fraud, rule-breaking and bias in the media
    -he gets away with changing every position
    -people are calling him the “greatest leader of a generation” with no accomplishments

    You have to read this document to understand how his hypnosis works. Logically and rationally, why not get to know this guy a little more before handing the world over to him? Why not wait?

    An Examination of Obama’s use of Hidden Hypnosis Techniques in His Speeches can be found at

    The media must feel compelled to cover this information and decide to let informed people decide.

    PS – “Dont drink Obama’s Kool-Aid”

  • Snowball

    “the advent of FiveThirtyEight is a blessing for poll geeks, but also a curse because of the celebritization of the process.”

    Well said. This is exactly why I was skeptical of of Nate Silver’s site. Sure, it’s entertaining and a ‘guilty pleasure’ at times, but my time as a researcher has taught me to (try to) avoid the ‘narrative fallacy’ and focus on the numbers.

    “I used to screen purely political posts without analysis-related commentary. But lately I’ve weakened. Also, these comment threads have gotten ideologically rather one-sided. Please come back Wednesday for the postgame commentary. You betcha!”

    The dark side beckons… Looking forward to it ;)

  • Snowball

    The troll’s post is a sign that this site is going …MAINSTREAM (not just for elitist geeks)! Sounds like something out of :D

  • Paul

    The contrast between Sam’s post and comment #11 is a striking study in the different models of thought and rhetoric one can use in studying an election.

    There, I said that as nicely as I could.

  • Mike

    Hi, Sam! Thanks for all your work on this. As others have said, it’s very nice to see, and is a great stress reliever!

    Would it be possible to show a graph over time of the data in the heading of the page: “As of November 2, 8:02AM EST Obama: 364 McCain: 174”. I’d like to see how the EV numbers changed with both candidates. I see Obama’s chart here, but would it add anything to see what McCain’s numbers did over time?

    Thanks again!

  • Michael

    Irrespective of the size of his budget, what exactly would we expect McCain’s chief pollster to be saying at this point? On the Sunday before the 1980 election, Pat Cadell told Jimmy Carter that he was going to get crushed on Tuesday. But he didn’t go on Meet the Press to tell the world that it was over.

  • Sam Wang

    blair alef – your assertion that FiveThirtyEight was “better” a few months ago has no empirical support. The assumptions used there are not egregious. But their net effect is to reduce the precision of the snapshot without adding any predictive value.

    The only advance predictions with any general election track record are the political scientists’ models.

    Mark S – I am in favor of putting in uncertainties where they exist. However, a 3% probability of a McCain win is so high that it’s in the category of keeping away from >99% in order to avoid incredulity. I think it’s better to avoid reporting that particular quantity. For this site I won’t even tell you what it is.

    The Meta-Analysis is a data snapshot and therefore has as its goal maximum precision. The Meta-margin graph has fluctuations of only a few tenths of a percentage point.

    Mike – The history plot is on the right sidebar. Click on it. We thought about giving an option to plot the McCain number, but it slipped away from us. Note that EV_McCain = 538 – EV_Obama. You could print a hard copy of the history, then hold it upside down with the printed side away from you.

  • lanny

    i’ve been lurking for a long time and finally wanted to say… thanks!

    it is a shame that with increased popularity this site has begun to attract “mouth breathing, dinosaurs walked the earth beside humans” trolls who don’t even know the meaning of socialism…


  • Displaced Canuck

    I enjoy your commentary and
    depth of understanding of polling and stats.
    Two quick comments:
    I love’s plot customization option as well. I think I made almost exactly the same plot last night. It shows the direction of the internet to some degree – every site will be customized by each viewer in the future, probably automactically by the viewer’s preset preferences.
    I enjoy 538 reporting than their predictions. The in person coverage of the ground game across the swing states has been outstanding.

  • gary

    I have two theories about undecided voters. One is that some who say they are undecided are not really, but they just don’t want to tell their preference. The second is that those who are undecided are less likely to vote. If either is true, there is less room for movement or last minute influence than candidates might think. Have you any thoughts on this, or have there ever been studies of the undecided group?

  • gprimos1

    Dr Wang,

    If I understand correctly, the reason the meta-margin has not shifted the median EV estimate is that the shift is all in states where Obama was already leading. However, shouldn’t an increasing meta-margin make it make it less likely that Obama would get a particular low EV total? Wouldn’t that mean the CI should be decreasing instead of staying constant as it has?

  • Ginny Mayer

    I keep reading the reassuring words and I grasp their meaning, but they’re not sinking in. Like many Democrats who are terrified of the prospect of a McCain-Palin presidency, I cannot relax until Obama actually wins. Tuesday night can’t get here fast enough.

  • Sam Wang

    Gary – I have quite a lot of thoughts about undecided voters. Here are more thoughts.

    gprimos1 – There are always some states in the toss-up group. But they have a trail of safe states behind them with varying margins.

  • Glenn

    I find Sam Wang’s analysis — and commentary on that analysis — far superior to any other aggregate state polling site.

  • Raghav

    Thanks for releasing a graph of the popular meta-margin! I had been generating my own from your data, mostly for myself and a few interested friends, but as you can see, my scripting skills (not to mention my web design skills) are pretty rubbish. I hope you don’t mind.

  • Frank


    I didn’t know we were getting into the neuroscience of hypnosis.

    Re tracking polls:

    I’m sure that graphs of margin by date exist wherein data points that are supplied by the same pollster are connected. I was happy to see that recently started doing this for the nation and selected states and selected pollsters, but these are not combined into a single picture.

    Not that you incorporate this in your numerical analysis, but I wonder whether you (or Mr Silver, if he’s reading) have a sense about how much party bias remains in the second order, namely whether Democratic- or Republican-leaning polls can still be seen to lean when trends rather than cross-sections are examined.


    I hope you are right, I just remember going to bed in 2004 with Kerry way ahead in Ohio, and waking up to “surprising differences between exit polls and the electronic voting machines”. How many of the swing states votes will be cast non traceable/electronically? The impact of that on the margin? That is a cross analysis that I haven’t seen.

  • George Smiley

    Jay Berk: thank you for confirming that the concerns about “socialism” and the (so-called) Bradley effect are refer to exactly the same thing: the impossibility of a McCain win without reliance on racial hatred.

    The next step: re-name the Republican party The American National Front.

  • George Smiley

    “a” splutters:

    “We’ve seen this story before in Germany in the 1930’s, haven’t we?”

    Yup. Dude’s name was Jesse Owens.

    Thanks for playing!

  • MattF

    I wonder if the meta-margin depends more on the spread of the distributions than the median estimator– After all, all the medians could stay the same, but if the spreads of the various polling distributions narrow (due to, I don’t know… more polling?), then the meta-margin would increase… no?

  • Dan F.


    Thank you for the great work on this site. It’s a pleasure to peruse.

    Given the daunting task Sen. McCain faces at this moment–being so far behind in the polls both at the state and national levels–can you foresee any potential major external event that could materially affect the trajectory of this race? A flood of Reverend Wright ads? Obama’s half-aunt?

    Democratic palpitations of fear notwithstanding, at this late stage, do you perceive any real potential for a McCain surge?

  • George Smiley

    Dan F. —

    At this point Obama could be caught in bed with a live boy, a dead girl, and Sarah Palin, and he’d still win. Martial law in Philly wouldn’t be enough. Massive tampering with electronic voting machines probably wouldn’t do it. Some combination of those things might be enough to put McCain in the White House.

    Too bad for Republicans that they don’t believe in the Central Limit Theorem.

  • Independent

    “The increase in the Meta-Margin over the last four weeks is quite striking. Despite the fact that national surveys and the Electoral College have been solid during October, the effective gap between the candidates has been steadily widening in battleground states.”

    Could you elaborate on those statements? Is the meta-margin increasing even while the median EV estimator is essentially holding steady? If so, why is that happening, and how is it related to Obama’s growing strength in the battleground states? Thanks.

  • Vincent Ferrera

    Call me crazy, but I think it’s going to be a lot closer than the current projections indicate, either here or 538, RCP, intrade, etc.

    If you start with the standings (364/171), then give McCain ND, MO, CO, NC, OH and FL, it comes out 282/256.

    Give Obama OH, 302/236

    Give Obama FL, 309/229.

    Give Obama FL and OH, 329/209.

    My guess is Obama 302-309, McCain 229-236.

  • George Smiley

    Okay, Vincent, you’re crazy!

    Seriously, though, I think there is about a 25-35% chance that we’ll see the numbers you suggest.

    But those numbers do not put McCain close to the White House. This is not horseshoes, and it is not hand grenades. Almost does not count.

  • Vincent Ferrera

    Not hatin’, just sayin’

    As a native Chicagoan, you can guess who I’m backing. But I think it’s going to be closer to 1960 than 1972, or even 1992.

    I’d like to see the stats with different percentages of undecideds breaking for McCain. Like what if they go 60 or 70% McCain? Because I think that is possible if not likely.

  • Independent


    Assuming that undecideds are 5% of voters, a 60% or 70% break for McCain do not constitute more than 1% or 1.5% shift in the final tally.

  • Emily

    I feel confident that more people will vote for Obama than McCain, but I wish someone could predict what effect tamper-friendly electronic voting machines will have on the outcome. I’ve been reading about these DRE systems and giving myself nightmares. The only hope I have right now is that Obama’s lead is significant enough that it would be very difficult for the GOP to cheat without raising serious suspicion in the media. Of course, the media has already prepared a convenient explanation for any major discrepancies between preliminary polls and the actual outcome: the supposed Bradley effect! As Rachel Maddow would say, “Can anybody talk me down?”

  • Mike L

    Excellent commentary again today. Looking keenly ahead to more of the same in the final hours before Tuesday night.

    Alas, I still am unclear whether your Mets-Analysis % is based on all states or just battlegrounds. If all states, is the same % indicated for the battlegrounds?

    Your scenario is certainly viable, except how could you “give McCain” CO. Isn’t it safer for Obama than Nevada?

  • Vincent Ferrera

    I meant on a per-state basis.

    E.g. has OH for Obama by +2.1, with a margin of error of 3.2. This seems to suggest that McCain could win OH if he gets enough undecideds. So, how do the overall numbers change if you vary the proportion of undecideds in every state?

    NV, MO, NC, OH, FL are all within the margin of error. If they all break for McCain, then it’s 286/252. How likely is that?

  • Bill S

    My favorite absurdity form the McCain Memo:

    “Expanding the Field: Obama is running out of states if you follow out a traditional model. Today, he expanded his buy into North Dakota, Georgia and Arizona in an attempt to widen the playing field and find his 270 Electoral Votes. This is a very tall order and trying to expand into new states in the final hours shows he doesn’t have the votes to win. ”

    Seems to me expanding into new states (i.e. red states) shows that Obama feels he’s in pretty good shape in the traditionally blue states and the traditional battle ground states.

    The most intersting stuff to me on FiveThiryEight is the reporting from the field about all the busy Obama campaign offices and the empty McCain offices which contradicts assertions in the McCain memo about their ground game.

    My impression is that the Obama campaign understands the Powell Doctrine and has applied it to elections.

    Does anyone think Palin has ever heard of the Central Limit Theorem?

  • Observer

    Vincent, you are ignoring a factor that Nate Silver has several times elucidated:

    When the polls are within the margin of error, but they are mostly in the same direction, the likelihood that the result is the lower in the polls wins is still very, very low.

    This is — I think, I’m no statistician — another way of stating Sam’s metamargin analysis: you aggregate all those polls and the MOE is actually very small.

    Stop worrying (but make GOTC calls if you want)! Sam declined to state the probability of McCain winning now, based on Sam’s analysis, because he didn’t want to come out say, ‘It’s zero as now!’

    (Got my champagne in the fridge for Tuesday night)

  • Sam Wang

    Mike L – For Mets analysis, Silver’s your man. But seriously…this analysis considers all 2^51 possibilities, 2.3 quadrillion in all.

    Vincent Ferrera – The confidence interval is far smaller when one averages polls. This is the point of aggregating polls for a state. See the FAQ.

    Bill S – If Sarah Palin does not know that fruit fly research is a cornerstone of biology, and therefore helping special needs children, then forget about the Central Limit Theorem. Then again, a family that size might have a pretty good margin of error.

  • Mike L


    As I said, got no trouble with your basic premise, but please stop asserting dubious information:

    Nevada and Ohio are NOT within the margin of error, and both are trending in Obama’s favor.


  • Cliff

    Emily, most states have made some improvements in vote security. Few (if any) are perfect, of course, but even if they fail to prevent theft they make it harder to hide. The only real bleeding sore is Pennsylvania, and if McCain is declared the winner there we won’t need to prove anything, it will be objectively fraudulent.

    The major differences between 2000 and 2008 is that (1) the Democrats control Congress, which counts the votes and certifies the election, and (2) the experience of 2000. The congressional Democrats can reject fraudulent results, if they have the necessary backbone.

  • Mike L

    Sam, That was a Freaky Freudian typo of mine—but yeah, Mets (0r Rays or Phils) analysis would be Silver’s territory. lol

    Also I total concur with your comments about the new celebrity status attached because the MSM giveth and the MSM taketh away (see Allen, Woody, “Celebrity” (1998), Sweetland Films). Doggone it!

  • Michael

    Since hackable electronic voting machines were brought up, I thought I’d mention that I recently read that Maryland and Virginia will begin transitioning back to good old-fashioned paper ballots after this election cycle. Not punch cards; not optical scanners; not machines that produce hard copy printouts. But paper and pencil and count ’em by hand. Ditching electronic voting machines that they haven’t even finished paying for in some places.

  • Vincent Ferrera

    Just to be clear, has a per-state breakdown. For each state, there is a line labeled “Projection.” I assume that the numbers in bold are some kind of running average and that the MOE is either the s.d. or s.e. of this mean. In any case, it’s the only line with error bars and the numbers are in bold type, so I assume it’s based on some kind of pooled estimate of the mean and variance.

    In any case, for OH, the avg. is +2.1 for Obama with MOE of +/-3.2. For NV, the avg. is +2.6 for Obama with MOE of +/-3.2. In both cases, the mean difference appears to be smaller than the MOE.

    Is this dubious information? Am I looking at the wrong numbers? Are the numbers themselves wrong? Just asking.

  • Frank

    Sam, Looks like there’s something wrong with the simple exercise that I did yesterday, namely taking national polls ending 10/31 and comparing each margin (Obama minus McCain) with the same pollster’s previous margin. Yesterday the median change was 1.5 (Obama’s running away). Today it’s -0.5 (McCain’s catching up). So I’m confused.

  • Michael


    You’re reading the numbers right, sort of. Those bottom lines for each state are his projections, not the poll aggregations. I believe that Silver’s model assigns the undecideds. He projects Ohio, for instance, at 50.5 – 48.5. The other 1% are third party voters, not undecideds. You are correct that the OH and NV projections are within the MOE. But again, that’s the MOE on the projection, not the MOE of the aggregated or averaged polls.

Leave a Comment