Princeton Election Consortium

A first draft of electoral history. Since 2004

Meta-Margins for control: House D+2.8% Senate R+4.8% Find key elections near you!

NPR’s Science Friday – this week at 2:15pm ET

October 17th, 2012, 4:00pm by Sam Wang


This Friday at 2:15pm Eastern, I’ll be on National Public Radio’s “Science Friday” with host Flora Lichtman. Nate Silver and I will talk about poll aggregation and the state of the national race. We’ll take calls too. Tune in – or listen here.

Update: one topic that may arise is what’s been surprising about this year’s race. My nominees are (1) the stability/polarization of the race, (2) the lack of an RNC bounce, and (3) the giant debate bounce (and the apparent effect of the VP debate). What do you think?

Tags: 2012 Election · President

104 Comments so far ↓

  • Richard

    Very much looking forward to it. I think your approach and Nate’s approach complement one another and combined offer rich insights. Along with many others, you are my two favorite poll analysts/geeks.

  • Scott Crider

    Nice. Will listen in.

  • Obama 2012

    I’ve been telling everyone to stop getting worked up about one poll (whether good or bad) and to just stop by 538 and Sam Wang’s election site instead.

    I really think people end up way less informed by trying to keep track of all of these polls themselves rather than to just let you guys take it all in and make sense of it.

    Speaking of uninformed. Did anyone else notice that the CBS host said something like this following the debate: “the first debate moved the race from a tie to a small Romney lead.” (I’m paraphrasing, but that’s the gist of what he said) … WTF?! that’s so far off the mark of what’s actually happening… it’s just amazing. And this guy is a lead anchor for CBS News… could he possibly be that misinformed or is it something more nefarious?

    • Insidious Pall

      I cast my vote for b) something more nefarious. Because he’s only slightly off. The first debate started with a small O lead and after that debate there came a small R lead. I’m assuming he may have been a neophyte as regards polling which would indicate he was speaking of the national polls. The nefarious part is your interpretation. Gallup has a rolling average that is today up to Romney 51% and Obama 45%. I can’t reconcile that one.

  • steve in colorado

    Awesome! Too bad I have to work.
    Maybe I’ll head out to my car for lunch break

    • securecare

      You can listen to the archived version shortly after the live program is over. Go to the NPR website and look for Science Friday.

    • Olav Grinde

      @SecureCare: Thank you so much!

  • Elizabeth Duvert

    Great, Sam! One of my favorite programs. You deserve the spotlight. Will definitely listen. Continuing thanks for what you do here.

  • Mark F.

    I’m a Gary Johnsom supporter with no major party dog in this race.

    I agree that Obama has a slight advantage at the moment. but it needs to be remembered that Romney is likely to win VA, NC and FL . CO also seems like a good bet for Romney. So, if Romney can flip Ohio (and this may not be as hard as Mr. Wang thinks), he will almost certainly win the election.

    Bottom line: Democrats should be worried, but not in full panic mode.

    • Obama 2012

      You are giving no proof to back-up your statements.

      “it needs to be remembered that Romney is likely to win VA, NC and FL.”

      PPP & YouGov have Obama down by 1% point in Florida and Obama has also been very close in many recent NC polls (even Rasmussen only has Obama down by 3% points in NC.)

      And Virginia is a toss-up at worst. There’s no solid evidence that Romney is “likely” to win it. NYT/Q had Obama up by 5% there even after the first debate.

      As far as Ohio is concerned even during Romney’s peak (and I believe that’s what the past two weeks have been) he’s never been able to take a lead in a reputable poll and most Ohio polls have him down by 4 to 6% points… when adding in the huge advantage Obama has in early voting there… I don’t think you have much of an argument there either.

      And all of that was before President Obama’s magnificent ass kicking of that smarmy liar Romney last night.

    • Patrick

      Whenever I was down about the President’s chances, the one thing that has kept me confident are those polls in Ohio. If Obama can take Ohio, a Romney presidency is highly unlikely. I expect my blood pressure to spike in the hours after the Ohio polls close….

    • Matt McIrvin

      Romney wins without Ohio if he takes New Hampshire, Colorado and Wisconsin, on top of all the states where he’s currently leading. Some of the Romney-victory tail in Sam’s probability distribution is probably that.

      I suspect NH is going to flip back into Obama’s column, and that Obama will hold Wisconsin. But Ohio isn’t a leakproof firewall if Obama ends up having to defend it at the expense of the rest of the country.

    • Olav Grinde

      Well then, take heart once again. Survey USA has a new Ohio poll out, conducted 12–15 October. It shows Obama at +3%. And for the Senate Brown is still leading Mandel, at +5%.

    • Ohio Voter

      I’m a roster judge for my precinct in Ohio this year, I decided I had to actually go out and do something rather than fret about polls all day, haha.

    • Brian

      You forgot the other 2.3 quadrillion possibilities.

  • Than Hedman

    The real question:

    Is it going to be mildly awkward? You call out Nate a lot, and he sometimes obliquely references your criticisms without directly responding to them.

    Can’t wait!

    • Sam Wang

      I sincerely hope not. We do have the innumerate as our common foe!

    • mnpundit

      Is Nate even allowed to respond? I would assume the paper would frown mightily on a blog fight like that if it took up space from the regular poll stuff.

      I would like to see a fight between the two though, for my own amusement.

  • 538 Refugee

    So, Flora Lichtman will be moderating a debate between Nate and yourself? What format? Sounds like fun. ;)

    • Sam Wang

      not a debate, as far as I know. We answer questions.

    • 538 Refugee

      The “debate” reference was meant as a joke. Nate once referred to “competing” sites and I once considered asking if you felt like you were in a competition. I guess that is where that idea started.

  • Ms. Jay Sheckley

    Mark- Read more about this site, and watch the show this thread is about. We’re seeing some movement on most of the states you name. Come back often and you’ll see more than you do now.

  • Bill N

    2:15 Eastern time? Central? What Friday program? Science Friday? All Things Considered?

    • Ohio Voter

      I imagine it’s 2:15 Eastern during Talk of the Nation’s Science Friday.

  • Olav Grinde

    Dr Wang, I am very much looking forward to your discussion with Nate Silver and Flora Lichtman on NPR. Is there a website that streams this?

    PS. Does the audience get to submit questions to the moderator?

  • Tom

    Hello world. I have a question on statistics; I cannot quite figure out the answer despite significant research here on the Internet. It regards margin of error. As I understand it, if we have a poll of Obama 50 and Romney 45 with “a 3% margin of error” that means that if that poll were done 100 times, 95 of those times Obama would get between 47 and 53 and Romney would get between 42 and 48. Two questions:

    (1) Imagine a standalone poll–in a vacuum–with O 50 and R 45 with a MOE of 3%. Is it correct to say that, based on this poll, there are no grounds to believe that O is ahead?

    And (2) Imagine that now instead of just one poll there is a series of polls, spread over time. In each poll O is leading R (perhaps by different amounts in each poll) but ALWAYS within the MOE of that individual poll. Can we not ascribe some objective advantage to O even though this is not implied by any individual poll? Thanks.

    • Sam Wang

      (1) No. In that example, O is ahead. Reported margin of error is usually the 95% confidence band for a 50-50 race – for one candidate’s vote share. The candidate-candidate margin doubles the uncertainty. In this case the principle is: if A leads B by the margin of error, he/she really is ahead 84% of the time – in the sample that was surveyed.

    • Amitabh Lath

      Correct me if I’m wrong, but the margin of error most polls quote is simply the statistical uncertainty, 1/sqrt(N). The uncertainty on subsamples (by ethnicity, gender, age) is larger, but not often quoted.

      Also, they should quote uncertainties on likely voter screens, subsample weights, etc. and add those in quadrature, but don’t.

    • Matt McIrvin

      In the second case, the value of multiple polls depends on the type of series it is. In a RAND-style survey, where you re-ask the same sample of people over and over, asking them many times does nothing to reduce the possible error from randomly getting an unrepresentative sample of the population.f

      But if you’re re-sampling the whole population every time, a consistent signal you see ought to gain statistical significance over time. Of course, it could be real or it could still be some systematic error in your procedure.

      On the other hand, if the results are all within your MOE but Obama is never behind, at some point that’s a sign that something is wrong with your math. If you’re calculated the sampling error correctly, occasionally you should see some results that are outside the 95% confidence interval, and if the distribution is roughly symmetric, about 2.5% of the time the outlier should be in Romney’s direction.

  • Michael Worley

    Here’s my theory

    1. Obama won the debate due to the Libya fact check issue. It would have been a tie otherwise.
    2. Most people don’t care about Libya enough to change their vote, so very few minds were changed.
    3. As long as the YouGov polls stay in, we’re looking at O+1-O+1.5 in the MM
    4. Once Yougov rolls off (it’s a re-ask, so people are less likely to change their minds than never-asked, see RAND), we’re looking at an O+.5 to O+1.5 range.
    5. Either candidate could move the MM a little over the next 20 days, but we can’t predict that.
    6. If the MM is below the Kerry-Bush margin, Romney could win, as Kerry came close. He does have a potential undercount Hispanics issue, so I wouldn’t put 50-50 until the MM is at R+.5.
    7. If we’re at Kerry-Bush or above on election day, Obama will probably win. R+1.5<MM<O+.9= long election night.

    • RDT

      I think the “fact check” was bigger than the topic itself — people may not care about Libya, let alone on what day Obama said “terror” — but not having your facts straight when you make an attack that your advisors had been telling the world you planned to make just looks bad.

  • Amitabh Lath

    Please bug Nate about his lack of uncertainty bands.

    • Michael Worley

      He added them a few days ago.

    • Amitabh Lath

      Wow, I see. If you mouse over the red or blue lines, you get these faint bands, currently +- 68 EV. Does he mention what these are (1 or 2 sigma bands, or something else?)

      In the case of the Nov 6 forecast, how does he incorporate the uncertainty in the economic variables, and also correlation of the uncertainties of econ variables and polls?

    • wheelers cat

      Amitabh, I bet you $10,000 (channelling Mitt) that Nate’s variables all conform to assumed Gaussian distributions.
      ;)

  • wheelers cat

    Alas, its good and its bad. The more popular Dr. Wang gets, the more we are beset here with low information conservative trolls citing singleton outlier polls and whining about dem oversampling.
    And Im afraid some ginormous media conglomerate will buy him like Huffpo bought Blumenthal and the NYT bought Nate.
    /sadface

    • grandpa john

      Well I hope not.
      I was a Nate follower back when he ran his own site, and I am sad to say that I don’t see the same openness that I did when he was independent.

  • A New Jersey Farmer

    I will listen in on Friday. That’s a great program.

    RAND took quite a jump yesterday and now has Obama above 50%. Numbers are moving in the right direction.

    • wheelers cat

      Holy kurtosis a fractal!
      The Debate1 around 10/3 behavior is isomorphic and opposite to the Debate2 behavior around 10/16.
      Is that an artifact of the methodology?
      Respondents can turn in their surveys up to a week late….I wonder if some responders held onto their surveys until after the debate.
      I think RAND would tend to educate respondents about politics…maybe make them more aware.

  • Anbruch

    Fairly large jump in Rand this morning. Odd, given that PPP tweeted last night that they are seeing very little change from the polls they ran immediately before and after the debate.

    • Matt McIrvin

      If you look at the other charts on the RAND site, you can see that the jump wasn’t from anybody switching to Obama (the net motion is still toward Romney there). It was a jump in “likelihood of voting” among Obama supporters and a decline for Romney, which would be consistent with the theory that the main effect of the second debate will be on base turnout.

      It sounds as if PPP doesn’t do a lot of likely-voter screening in national polls besides telling people to hang up if they’re not going to vote. So they might miss some of that.

    • Olav Grinde

      Matt, I noticed this too.

      According to the Rand Poll, there is now less than 1.5% difference in likelihood to vote between Democrats and Republicans.

      Lowering that deficit, which was close to 5%, will be key to winning a sufficient number of swing states — and is perhaps even more critical to down-ticket races.

  • MAT

    Before anyone asks, it looks like this morning’s Meta Margin jump is mostly due to a Survey USA poll in Ohio and JZ Analytics/Newmax in Florida, both showing Obama leads. Both conducted prior to the Hosftra debate.

    • grandpa john

      Well Michael, just in your own mind add Nates 3-4% R bias for Ras and it fits in with the others

  • The Political Omnivore

    I have GOT to hear this! I’ll donate 10 bucks to the candidate of your choice if you offer to arm-wrestle Nate Silver on the air! (it’s hard to convey in text how much I love both sites so don’t take that too seriously).

  • Olav Grinde

    Questions to you and Nate Silver

    Dr Wang, I have several questions that I would be delighted to hear addressed in your discussion with Nate Silver.

    1) What assumptions lie at the core of you prediction model/method?

    2) Looking back at your work as an election predictor, what old assumptions have you found it necessary to correct / fine-tune, and why?

    3) Are there, in your opinion, substantive efforts to undermine the integrity of the American election and ballot counting process — and if so, how can you as an election predictor take this into account?

    4) How do you deal with the new phenomenon that some pollsters seem more concerned with driving the narrative in the news media, than in accurate polling? Or is this never the case…

    .
    Regarding 1, for instance: You have been extraordinarily open about your methods and assumptions. As far as I can see, one of your core assumptions is “the wisdom of the crowd” — and that the sum-total of polls in each state is unbiased. Is there any reason to question this?

    Nate Silver, however, often refers to “the Model” almost as though it is a third person, and he seems rather opaque as to his core assumptions. He seems to assume, though, that economic indicators are not already reflected in the polls and must thus be taken into account — whereas you’ve pointed out that this risks counting these factors twice.

    Regarding 2: You have quite openly discussed your mistaken assumption in past elections about how the undecided votes would break. But I’m sure you and Nate have other interesting examples…

    Regarding 3: You have, for example, recently analyzed the impact of re-districting on House races. And I believe you were quite astounded at what you found… (As I understand it, you have incorporated your findings into your House predictions.)

    • wheelers cat

      Heres my question for both you and Nate.
      Given that Team Romney has embraced the Sailer Strategy, what percentage of the non-hispanic caucasian vote will Romney need to capture to win?

    • Craigo

      @wheelers cat: In 2008, McCain would have needed 62% of the white vote to reach 50% of the total vote, if his proportions among all other groups held steady. (He won 55% in the end.) That’s your starting point – if you assume Obama’s 2008 proportions among nonwhite voters hold (probably not a good assumption) and that the white vote declines slightly (probably a good assumption), then Romney would need about two-thirds of the white vote to win.

    • wheelers cat

      Craigo, I know all that.
      Brownstein and Medved have both made analyses, and they have arrived at 63-65%.
      I was axing Nate and Dr. Wang their opinion.

  • BCC

    Sweet! Wang-Silver cage match!

    As live discussions/debates tend to favor the sizzle over the steak, Silver has the edge going in.

    I predict a slight Wang victory, on points.

    (Yes, I know it’s not a competition, but everything’s a horse race in October)

    • Sam Wang

      I see. In this comparison, am I debate#1 Barack Obama?

      On the other hand, I have been on NPR a lot…

    • wheelers cat

      Nah.
      In this non-debate you are Takeshi Kaneshiro and Silver is Tony Leung Chiu Wai allied in the epic battle against the innumerate.

  • Ms. Jay Sheckley

    Thanks for the nice link to the NPR show. I’ll use it and tell others! Is it true I can listen after just as easily? If so, what time will it end/post?

    I like the questions posed. Why not hand these or others in to the producer to look over? Better yet, say something quotable to affect voter enthusiasm, and get people out to the polls, or to inquire about absentee/mail voting right away.

    Does anyone think publicizing voting by mail can help fight voter intimidation? If so, please join me in spreading the blue link below the FAQ, upper left on the this page.

    In Berkeley CA, our mail in ballot is 4 oversize sheets of cardstock with 50+ candidates and 30+ measures covering all 8 sides. The included info says use $1.50 postage, but the Post Office advised $1.05, and of course there are drop off methods….

    Say, can you or anyone tell us anything about the new Meta-Margin dip? Is the VP debate not yet in the mix? Or is the MM trying to better resemble a fish hook? Sometimes these results just stick in my craw.

    • ChrisD

      I could be wrong, but I think the MM is very sensitive to new polls from very close states, e.g., OH, which had an O+3 poll last night and an O+1 poll late this morning.

      Btw, I read a few weeks ago that something like 400K absentee ballots were thrown out in OH in 2008 for various defects. If one party tends to use absentee ballots more than another, that might swing a very close election.

    • Olav Grinde

      “…400K absentee ballots were thrown out in OH in 2008 for various defects.”

      This should be a cause for great concern. Who processe absentee ballots? If it’s Ohio’s Secretary-of-State Jon Husted, then I would be very worried, indeed. That is the same guy who fought tooth-and-nail to prevent making early voting being made available on weekends for all Ohioans. He shamelessly pursued this all the way to the US Supreme Court, hoping to suppress the vote of Democratic-leaning demographics.

      I expect a lot Ohio absentee ballots to be “lost in the mail”.
      And if Husted & Co has a finger in the “processing” or “counting” process, it should not be trusted.

    • Ohio Voter

      Absentee ballots are processed by local boards of election, both in sending and receiving.

    • Olav Grinde

      Ohio Voter: Thanks for clarifying. What you say is comforting. I think…

  • Ms. Jay Sheckley

    Nix on the debate crash. I vote for stability: I love how stability stands the question on its head.

  • skmind

    To me it has been the astonishing neglect of the media to report the stunning impact of the SuperPACs.

    There is no way Obama gets that good of an early lead based on his record and the maligning that started on FOX “News” in Nov 2008.

    There is no way Romney recovers *in a debate* to reverse course so dramatically.

    The only fund-raising that gets press is what the campaigns raise.

    That is peanuts compared to what the SuperPACs are doing.

    It is the huge increase in spending that has shaped the election, and is likely to win it for Romney.

    http://bit.ly/OJyX8C

    Seriously does anyone really believe that what Mitt did in Debate #1 in “exposing” Obama changed the voter thinking process? Are the candidates’ positions, lack of details, weak points etc. so obscure that they were hidden till then?

  • Reason

    Did it go down 4 because Romney now has NH? If so, not good news.

  • Amitabh Lath

    One of the issues both you (SW) and Nate Silver have expressed concern over is the horribly low response rate for most pollsters nowadays (10% -ish).

    Was it always thus?

    If you were running an experiment where you accepted only 10% of the possible signal, you would worry that you had an extremely biased sample. And yes, there are weights etc. that pollsters apply to fix that problem to first order.

    But perhaps there is something else hinky with this sample. Maybe they are people who are more prone to suggestion than the general voting public? Maybe the probability they will pick up the phone changes with what they see on TV? This could explain the large sudden jumps in the polling, which seem counter-intuitive in a season when there are very few undecideds.

    • Amitabh Lath

      Just to clarify: In the hypothesis I am forming, this 10% (who still pick up the phone when caller-id shows someone unknown, in day and age!) they mirror the general voting public in their opinions and statistical makeup in all ways (after weighting).

      The only difference is that R-lean ones are less likely to answer in the aftermath of something like the 47% tape (and O-leans more likely); and O-lean ones less likely (R-lean more) after debate#1.

      So in the long run (integrated over ~month) you get the right answer. But the sudden peaks and troughs are artifacts due to the inconstancy of this sample of people.

    • wheelers cat

      well my hypoth, Amitabh cher, is that enthusiasm correlates with likelihood of voting/picking-up-the-phone at R^2 = 1 for republicans and R^2 < 1 for same correlation for democrats.
      Because of asymmetrical political behavior.
      I just dont think carbon-based systems behave as neatly Gaussian as classical physics systems.

      And btw the red robopoll houses only get a response rate of 2-3%.

    • Craigo

      I can tell you that when I got into the business in the early part of the last decades, response rates were much, much higher, about 25% on average. However, I will note that the two best presidential cycles for pollsters have been 2004 and 2008, which came after four straight cycles of poor predictions – despite much higher response rates in those eras.

      Now it’s true that methodology has probably improved overall during that time, and ideological cohesion has increased, which makes the job easier. so it’s impossible to say whether the decline response rates have negatively affected polling and just been overwhelmed by changes in other variables, or whether it had no effect at all.

    • Craigo

      I should note that 2010 was a pretty bad miss for pollsters as well. But I attribute that to elusiveness of the midterm likely voter and the lower volume of polling overall, not response rates.

    • Amitabh Lath

      wheeler’s cat: So if enthusiasm gooses R phone-picker-uppers more than D (and why this asymmetry?) then once we are two weeks or so past the event, things should get back to normal (ie, in agreement with the general voting public).

      Craigo: It’s interesting that you have firsthand experience of steep drops in response rates. Do you have any ideas why it happened (is happening)? And is there any un-correctable difference between the people who still respond to pollsters and the rest of the population?

    • Craigo

      Hi Amitabh:

      It’s a combination of factors: the rise of caller ID in the 1990s, followed closely by the rise of cell phones in the next decade, quickly followed by the rise of IVR polling (which has slightly lower response rates compared to live-interview).

      The biggest difference in response rate lies in the 18-29 (and now somewhat older) age range, who are much less likely to answer their phone or participate if they do.

      The second difference is those households where Spanish is the primary spoken language – lower response rates even when Spanish-language options are included.

      The good news is that A) These groups comprise a lower proportion of voters than they do of the population, so the potential error is relatively small, and 2) the error is correctable using demographic weighting.

      Some people worry that different partisan/ideological groups may have lower response rates, but I’ve seen no evidence for this claim. It was particularly widespread just after the 2004 election by pundits who confused the early exit polling (which predicted a Kerry victory) with pre-election opinion polling (which was exceptionally accurate that year).

    • wheelers cat

      @Craigo
      Ed Freeland was here last month and described fairly robust polling methods to capture the cell phone demographics; stratification adjustment and dual frame polling. The problem appears to be (for example the gap between traditional pollsters and robopollers) that Gallup and Rasmussen either dont use or misuse those methods. I suspect this is because of asymmetrical political behavior in red polling houses– of course I can’t prove it because their methdology is opaque.
      Which leads into Amitabh’s question. Red/blue genetics, neuropolitics, and asymmetrical ideology are all rather new domains of research.
      But the basic premise is that red phenotypes and blue phenotypes differ in significant and measurable ways, both in morphology and function, and these differences could lead to asymmetrical effects in behavior.

    • wheelers cat

      And by asymmetrical I do mean non-gaussian.
      I know, heretic!
      lol

    • Craigo

      @ wheelers cat:

      I’m sorry I missed that! I’m really happy that Sam was able to get an expert to do a guest-post, i was hoping for that. I favor dual-frame until further research indicates that stratification adjustment is as effective (that’s not to say that it’s flawed, however – it just has more uncertainty at this point). I admit that dual-frame is often cost-prohibitive.

      I have no doubt that there are neuropolitical differences (though it’s far, far from an area of expertise of mine). What I’m unsure of is that neurpolitics are producing different response rates among red and blue, and whether that potential difference is significant. I do find it quite interesting that Republican self-identification has dropped among conservatives this cycle.

    • wheelers cat

      @Craigo

      ‘zactly!
      its neuropolitics vs the CLT.
      But im a Talebian.
      Who. are. you?

    • buddhastalin

      One paper suggests that voter preferences normally should not undergo big swings in short periods of time. High volatility is due to the changing composition of the “likely” voter pool (as deemed by the pollster), which is affected by changes in enthusiasm.

      http://www.aapor.org/Content/aapor/Resources/PollampSurveyFAQ1/ElectionPollingResources/Erikson,_Panagopoulos_and_Wlezien%5B1%5D.pdf

    • buddhastalin

      Sorry, the link in my post does not appear to work. Just Google “Likely and unlikely voters and the assessment of campaign dynamics” by Erikson et al.

  • Reason

    Gallup has been getting it wrong since 1996. During the 2000 election, They first had Gore up 9 and then Bush up 8. Gore won the popular vote in the end. So I learned to never trust Gallup.

  • Rick in Miami

    As noted earlier, it’s interesting to consider the hypothesis that biased polls were released post-debate 1 to drive a narrative.

    Apparently the result was that Biden, then Obama, took subsequent debate prep very seriously indeed. I’m not sure that was the intended consequence!

  • Olav Grinde

    Dr Wang: If you and Nate Silver could request new polls at this juncture, what sort of polls would you request, how many, and where?

  • Reason

    I just looked over Gallup. So, let me get this straight. Obama now is behind only 1 point in registered voters. Day before, he was down 2. And his approval went up to 50%. Yet, he is down one among likely voters? This makes absolutely no sense whatsoever. So Gallup finds that more people approve of him. More people registered are voting for him, yet more are less likely to vote for him? This is why Gallup has gone downhill. I used to be addicted to them. But recent history shows they are horrible. Look at their polling for 2008. Even worse than 2010. I am glad I follow Dr Wang now. He is right. popular vote is meaningless in this democratic republic. It is the EC. My only concern was Dr Wang moving NH to the Romney column.

  • Reason

    *down seven among likely voters. Sorry for the typo.

    • Olav Grinde

      Look again, Loadar. That count ignores the swing states — and in most of those Obama has a small lead.

      This is the map and electoral count you should be looking at. As you can see, President Obama still has a significant lead:

      http://www.realclearpolitics.com/epolls/2012/president/2012_elections_electoral_college_map_no_toss_ups.html

    • Ms. Jay Sheckley

      This is meant as an addition to Olav’s words to Loader; Surely there’s naught I can say to Olav he doesnt know all about!
      I glance at other things, but if Sam work is right, it would mean it adds noise to go elsewhere, especially when PEC’s map is adjustable for this [I admit I had a friend travel quite a ways to get Java working just for PEC]
      Beneath the daily snapshot map [is _that_ the SINGLE snapshot, Sam? Or are all three models of the single snapshot in the ideal mind?] choose current map.

      Click on an unsafe colored safe and your map will firm up. That may tell you all you need to know. Read the “safe” EVs now. Interesting, eh?

      With white states, use the swingstate list in the right margin under the header POWER OF YOUR VOTE [Raised in Jersey I love “jerseyvotes”]. That will suggest the logical color to turn your white states if any. If a state remains you may as well play with it yourself. Click the states and they strobe through the choices, the EV’s self tabulating deliciously.

    • Ms. Jay Sheckley

      This is meant as an addition to Olav’s words to Loader; Surely there’s naught I can say to Olav he doesnt know all about!
      I glance at other things, but if Sam work is right, it would mean it adds noise to go elsewhere, especially when PEC’s map is adjustable for this [I admit I had a friend travel quite a ways to get Java working just for PEC]
      Beneath the daily snapshot map [is _that_ the SINGLE snapshot, Sam? Or are all three models of the single snapshot in the ideal mind?] choose current map.

      Click on an unsafe colored safe and your map will firm up. That may tell you all you need to know. Read the “safe” EVs now. Interesting, eh?

      With white states, use the swingstate list in the right margin under the header POWER OF YOUR VOTE [Raised in Jersey I love “jerseyvotes”]. That will suggest the logical color to turn your white states if any.

      Click the states and they strobe through the choices, the EV’s self tabulating deliciously.

      If a state or two remain/s white you may as well play with it yourself. Thanks to Sam, you know as much as anyone.

    • Matt McIrvin

      But still ahead in the “no toss-up states” count, which is actually pretty consistent with RCP.

  • Reason

    Even more confusing? A new Newsmax/Zogby poll, which is a conservative republican tracker. How gives O a +4 in Florida. That is +1 over the pre-debate 2 poll. Wow. This is either good or bad depending on how you look at it. This tracking poll skews republican usually. I would love to ask Dr Wang and NS tomorrow what their take on it is.

    • Sam Wang

      Commenting on single polls is anathema to me. I am often tempted to moderate comments on them. So: mu, which is Zen for “unask the question.”

      You could ask Nate…though is that really the *one*’question you want to ask???

  • CobaltinSF

    Dr. Wang, forgive me if you’ve explained this previously, but how can Gallup’s daily tracking poll be consistently such an outlier? i.e. today showing Romney ahead by 7%, yesterday by 6%. I don’t think anybody believes this.

    • Craigo

      Like he said, don’t bother looking at any one poll too closely – unless it has serious, obvious methodological problems.

      But that said – Random error plus likely voter screens can do funny things, and tracking polls are not meant to be a snapshot of the day they are published, but the mean of public opinion over their entire sample, which at this point lies almost entirely between the two debates when Romney hit his peak.

  • Reason

    Dr Wang, of course not! I have a ton of questions! But I would take up the entire time you and Nate are on the air. Heh. I was mainly replying to the one person’s freak out over Gallup, and pointing out the inconsistencies. I gave the Zogby poll as an example of why, as you state on your site, we need to track the EC, not the PV, which is what outfits like Gallup do. My concern is that the general public is a low information voter and they pay attention to name brands like Gallup etc, and not polls that actually track data in regards to the EC, which is really all that matters. 2000 showed us that. I cannot wait to hear the show tomorrow. Good luck!

  • Steven J. Wangsness

    Gallup has it R 52, O 47, but has Obama’s approval rating 50 approve, 44 disapprove. Maybe someone at Gallup is just transposing the numbers on one.

    I’m no polling expert, but I can’t see how Wisc. is considered a swing state, Paul Ryan or no. Wisc. hasn’t voted GOP in a presidential year since 1984.

    • Jen

      One explanation for how Gallup can have Romney increase his lead at the same time that Obama’s aproval rating is going up is that their approval polling is on a three-day cycle. Not 7 days like the horse race. So the post debate polling just won’t be felt as quickly.

      Rasmussen also had Romney expanding his lead. From 1 to two points. Likely because they just had a good Obama day roll off the cycle.

      I am training myself to ignore the national polling. It really is not as useful. A post-debate poll from Ohio/Nevada/Colorado/Iowa/Florida etc would be way better.

    • ChrisD

      I believe Gallup’s approval figures are drawn from U.S. adults, not registered voters, as well as being 3-day averages.

      About 5 days ago Gallup announced it was changing its methodology to make cellphone calls 50% of their daily sample (rather than 30%, I think). Since the change, Romney has steadily climbed among LVs. That seems counterintuitive given younger, and presumably more Obama-friendly, voters are being called. Although I assume they still adjust their sample’s responses to match underlying demographic assumptions, I wonder what influence, if any, the change has had on their poll results. (Caveat: I know very little about the inner workings of polling.)

  • Michael Worley

    What do people think about the idea that Romney wins the popular vote but not the election?

    • ChrisD

      The internals on Gallup’s R 51, O 45 poll two days ago showed this:

      East: O+4
      Midwest: O+4
      South: R+22
      West: O+6

      Though unlikely, a reversal of 2000 is in the cards.

    • Matt McIrvin

      I wouldn’t call it probable, but it could happen. It sounds as if opposition to Obama in the South is getting more intense than we’ve seen in a very long time. If Obama can hang onto the swing states elsewhere while losing the South as badly as Gallup says he will, we could have an EV/PV split. And a lot of drama, probably.

  • Phil

    Here’s the thing I’m wondering about, because while Gallup is an extreme example, they are not the only pollster to show Obama at 50% approval (or above), yet down to Romney. Look at the Pollster chart for Obama’s approval:
    http://elections.huffingtonpost.com/pollster/obama-job-approval
    And for the horse race:
    http://elections.huffingtonpost.com/pollster/2012-general-election-romney-vs-obama.

    There are no approval charts by individual states, but I keep seeing this pop-up, a gap of several points between approval and horserace for Obama and it strikes me as strange. Obama has frequently had a favorability gap because he is personally liked by people who don’t think he is doing a good job (enough of them anyway), but I don’t recall seeing this approval gap before.

    On the Pollster chart, at the height of Obama’s peak he was at 48.4% and basically an identical 48.8% approval. Since then he’s dropped on the horse-race chart (at 47% at the moment) but actually improved on the approval chart, to where is he at 49.4% today.

    I don’t have the time or ability to comb back through the polls to see if there is a true statistically significant difference between the correlation of approval rating and horse race number before and after the “fall” but if it’s there, I don’t see a rational explanation. According to the polls right now there is a 2%+ gap (nationally) of people saying they approve of the job Obama is doing but do not plan to vote for him for president. I just don’t get it.

  • Tim in CA

    Hi Dr. Wang,

    I have a question regarding how you calculate your median for the state polls. How many polls do you use to calculate the median? Is it a fixed number, or do you have a chronological cutoff (i.e., using all the polls published in the last ten days)?

    I’ve been trying to figure this out on my own, but I haven’t succeeded yet. Specifically, with regard to today’s data, I’m not sure how you arrive at R +2 in Florida. If I’m not mistaken, Romney was listed +1 there yesterday, and the new poll published in the last 24 hours has Obama +1. In this case, how did the median shift to Romney by a point?

    Thanks,
    Tim

    • Rick in Miami

      My understanding is that he uses (1) the most recent three polls, or (2) all the polls in the most recent 7 days, whichever is larger. Someone please correct me if I’m wrong.

  • Dgarr

    I love this site so much! Thank you Professor Wang for the NPR information. I have shared the link with many who are interested your analysis. I truly enjoyed your appearance on Sirius XM as well.

  • Matt

    I think the question of this election is how big of a lead does Romney need nationally to win Ohio. There seems to be a lot of data that point to a tied national popular vote. If you use Sam’s median method on the national polls you get an exact tie. With this tie, you have polls that show Obama +2 in Ohio with the same method. For all intents and purposes this is as close to a make or break state for either candidate. You could come up with plausible combinations for either that don’t include Ohio, but the most likely winning combinations for either do include it.

    There is a precident for a candidate winning the popular vote by 3%, getting over 50% and still losing the electoral college – Tilden in 1876. So even if Romney won the popular vote by up to 2% and lost Ohio and the election, it wouldn’t be the biggest miss between the popular vote and electoral college.

  • Ms. Jay Sheckley

    Youve probably thought all this, plus noticed I said metamargin when I meant EV estimator, which explains the sharpness, now, doesnt it? But not the down turn. When did.does the vp debate & this last debate show up in charts? Over the course of a week, after a week, or??

    I don’t figure to surprise you, but reading these comments, I think you and Silver could get get a bit of useful attention from collusion: are there procedures, questions or data you agree you need to see from polltakers ?
    Is there some soundbitable narrative etc you can agree on that will help get everyone out voting? Was the debate downturn an implosion caused by gish gallop and imploded superman syndrome?

    Seen as competitors, if you made surprising statement/s together, wouldnt it be hard for the mainstream media to resist? You may have real power here. Hide my comment if that helps. As IF!
    To this end [teaming up a moment for power! :D], you might also want in advance to skim the book Nate’s flogging or otherwise prepare a positive remark short of,
    If you only read one book this year, I think we found your problem.

  • Pastafarian

    Dr. Wang,

    My nominee for what’s most surprising would be your analysis of redistricting on the House. That seems to be one area where you surprised yourself with size of the advantage conferred to the Republicans by the 2010 redistricting exercise. I don’t think that item has gotten nearly enough mention in the national press.

  • wheelers cat

    So…I watched Silver on the Daily Show last night.
    And I noticed something that is pretty prevalent on this blog too– DSP analogy.
    What if DSP doesnt really work well for carbon-based systems? What if pattern recognition works better?
    Is the brain a machine, or something else?
    Is the underlying structure of reality Gaussian or Mandelbrotian?
    What if we are doing this all wrong?

Leave a Comment