Princeton Election Consortium

A first draft of electoral history. Since 2004

Ignoring Ludwig Wittgenstein

October 11th, 2012, 12:15pm by Sam Wang


Wittgenstein is useful today: Whereof one cannot speak, therefore thereof one must remain silent.* In modern parlance: STFU. Did I fall prey to the pressure to say something before its time? Still, Jon Stewart asked for nerds. And nerds always have something to say.

Examining the aggregates, it looks like Romney had a 5-point bounce – with a peak lasting a few days. Time will test this statement.

Today's electoral college map

Romney got a massive spike on 10/4 (why? see Richard Just, “How Twitter Has Done Us Wrong“). Is the pendulum swinging back? As of this morning, national polls show a tie, Obamney +0.0+/-0.9% (median +/- estimated SEM; n=8, Oct. 5-10), a bit of a return from Oct. 4-5, Romney +2.0+/-1.0% (n=3).

On October 2, I noted a discrepancy between national poll medians and the state poll-based meta-analysis. Both types of poll have very good track records when aggregated. The gap seems to come and go for two reasons:

  • National data are less precise because they have fewer respondents, but they react fast – in a few days. These have pegged Romney’s bounce at about 5 points.
  • State data are eventually extremely accurate at a national level thanks to their sheer volume, but take >10 days to fully adapt.

In other words, PEC gives the calm version of events. Does it feel calming?

Three states showing a large spike for Romney have subsided by about 5 points since October 4th:

  • In the critical state of Ohio, Obama +4.0+/-2.2% (n=5, Oct. 5-9). Last week it was at Romney +1.0+/-1.2% (n=3, Oct. 4-5).
  • Florida is less polled: two recent polls also show a movement of 5 points (O+1% and O+4% vs. R+2% and R+3% on Oct. 4).
  • Virginia’s even less polled but also shows a movement of 4-5 points.

Other states are less polled – and less clear.

It is certainly possible that Romney supporters will have more bright days ahead – but they should not waste time in living it up now. One indicator is that I am not hearing much talk of unskewing these days. Suddenly polls are great.

I’m traveling – at my field’s national conference. The Society for Neuroscience, in New Orleans. So not much original analysis around here for a while. Ludwig, take it away.

*Thanks to Winston for der fix.

Tags: 2012 Election · President

47 Comments so far ↓

  • steve in colorado

    I knew Jack Kennedy!
    Biden is on fire tonight

  • Michael Worley

    .72 … Will Ryan change the momentum? Or establish it as the new normal?

    • Sam Wang

      If things change, how will we know it was because of the VP debate?

    • wheelers cat

      we won’t…at least, not with certainty.
      “our tendency to perceive — to impose — narrativity and causality are symptoms of the same disease– dimension reduction.” — Nassim Nicholas Taleb

    • Michael Worley

      We won’t. correlation is not causation.

  • Arbitol

    With each passing day, I become more skeptical of using math to predict the behavior of actors with free will.

  • Lee

    I am proposing a change in how we compute the meta-margin. Instead of shifting all the states by a national shift percentage and then checking whether that makes an electoral college tie, we propose a national shift and adjust all the states by their beta times the national shift and check whether that makes an electoral college tie. If it does make a tie then we report that national shift value as the meta-margin.

  • Michael Worley

    Obama +0.96. We are almost to Bush-Kerry.

    • Sam Wang

      Indeed, though unlike 2004, it hasn’t crossed zero even once (yet).

      As per my previous comment (overall conditions Obama +3.0 +/- 2.2%), you’d think that 1/6 of the time this campaign it would have gone below Obama +0.8%, the get-out-of-bed threshold.

      I guess 1/6 of the campaign accurately describes the time from now to the election. Now there is an arresting thought.

    • Michael Worley

      I can’t disagree with you. However, I’d rather it cross 0 for 2 days at the end than four months beforehand.

    • Jack Rems

      Still, past coin tosses don’t affect future c0in tosses: Should we expect to cross zero and Romney to have the lead 1/6 of the remaining interval?

    • Sam Wang

      The answer to your question is to be found in the red strike zone.

      Given the past behavior of presidential races, we expect movement of up to 1.8% in 21 days. By itself, that assumption predicts that Romney would be ahead with probability 1/3 on November 6.

      However, your statement about coin tosses is probably not correct. The entire campaign gives us prior information about the likely range of outcomes. This is a “prior” that also informs the red strike zone. For this reason I believe Obama’s re-elect probability to be quite high.

    • Matt McIrvin

      “The entire campaign gives us prior information about the likely range of outcomes. This is a “prior” that also informs the red strike zone.”

      This is assuming that the campaign in its late stage isn’t basically a Markov process. This might be wrong.

  • wheelers cat

    Not many people discuss this, but this election is really about demographics. And the question is, can the lily white GOP elect a president?
    The electorate is evolving to have fewer white voters, and the GOP is trying to maintain an advantage through anti-democratic tactics like voter suppression and redistricting.
    I think the scales have already tipped, personally, and we can win if we just GOTV.
    Ron Brownstein

    “In the new survey, Romney leads Obama among non-college whites by 54 percent to 37 percent, almost exactly the same margin as McCain’s 18-percentage-point advantage over the president with those voters in 2008 (when they backed the Republican by 58 percent to 40 percent). The new poll shows Obama winning only 39 percent of non-college white men and 35 percent of non-college white women; but to overcome Obama’s other strengths, Romney will need to generate even larger margins with those voters. In fact, Obama’s performance with those working-class whites has slightly improved since the May survey.”

    This election isn’t really about the economy. Its about the demographic timer.
    see RAND.
    https://mmicdata.rand.org/alp/index.php?page=election#results-by-respondents-characteristics

    • Terry

      Very well put.

      For some states the ‘wave’ is still far from cresting. Unfortunately NC is in that category. Obama’s handlers have largely ignored it this year despite the squeaker in ’08.

      To make matters worse, my commute home in Asheville today is terribly delayed by Romney’s visit.

  • Olav Grinde

    Question: Which six topics have been selected for the 2012 Vice Presidential Debate?

    Apparently, these were announced several weeks ago, but I can’t find the list anywhere…

  • Lee

    In the current swing in the polls, I bet that some states’ polls have moved more than others. My intuition says, for instance, that highly partisan states probably had smaller percentage swings than swing states. Can this be quantified in a way that is useful for honing the predictions?

    That is, can you look at the data from this season and measure how much beta — see https://en.wikipedia.org/wiki/Beta_(finance) — each state’s poll has with respect to the benchmark of the national poll? Then you can compute the meta-margin not by assuming that each state’s polls move by the same amount, but by assuming that the national poll moves by some amount, and each state does its usual beta multiplier of that move.

    • Sam Wang

      That was true of post-DNC Senate race bounces, which were larger in pro-Democratic states. Your suggestion predicts that the last week’s change could be concentrated in pro-Republican states. This would be a potential cause of the disconnect between national and state polls.

  • Jack Rems

    Sam, isn’t it wrong to look at the meta-margin as having “momentum?” Isn’t it more of a derivative, or a “massless” construct of the superposition of a bunch of noisy signals?

    Romney, on the other hand, does seem to have “momentum”. Looking at RAND’s “shift between candidates” graph, Romney has been picking up a remarkably steady .25%/day since October 1. (I think that’s approximately one voter in the 500 surveyed that day. Have I got that right?)

    What big change happened October 1?
    Well, Romney stopped campaigning, to prepare for the debate. His campaign, up to then, had been like a pinata, a big gift to Obama– look at the RAND “shift” graph 9/7-9/30: I shouldn’t need calculus: to my eye the area between the lines looks like Obama had peeled off 40 or 50 voters out of 3500. (This was when Sam said “something is happening.”) Ten of them have switched back so far, and I expect more will.

    It’s an important election. Winning it shouldn’t be easy.

    • Sam Wang

      I don’t really think the Meta-Margin has momentum. I am more inclined to think of there being a sudden change caused by the debate. Then, because any aggregation introduces a delay, this shows up as “momentum” as made visible by a time-filtered trend. Then the chattering commentators come into the mix.

      I think many campaign events that we can identify (Michelle Obama speech, DNC, Palin-VP, Ryan-VP, Debate #1) lead to sudden jolts. In a separate category are slow-moving events: the Swift Boat campaign,the Bain campaign, and the 47%-video-tape slow bleed. Assigning causality to this second category is far harder. I have not made up my mind which category the 2008 Lehman Brothers collapse falls under.

      Regarding RAND, I am getting a little worried about overinterpreting it. It has the gift – and problem – of surveying things like voter intensity, then throwing that into the average. That topic requires drilling a bit deeper.

    • wheelers cat

      I love RAND but we should never forget that RAND samples a captive fixed population of respondents that are PAID $2 per survey response.
      So it is not going to model national polls accurately, where enthusiasm is likely the key variable in determining whether there is a response at all.

  • Michael Worley

    Wisconsin people will watch the debate. If Ryan wins, it could flip WI and put the MM down. Tonight is no black swan, but it could reinforce present trends.

  • cyntax

    Good analysis and a quote from Wittgenstein–what’s not to like?

  • Amitabh Lath

    My first thought on seeing such a steep cliff (Ryan VP, debate#1, and also an upward one in mid-June) is that it’s probably not (entirely) real. The system has more capacitance than that.

    Note the upward rise post conventions, including 47% remark. That rise looks more realistic.

    I do not know what this means. But if I was measuring say, potential across some circuit where I expected the time constant to be milliseconds and I saw large fluctuations on the order of nanoseconds, I would question my equipment first, before freaking out.

    I wonder if there is a correlation between people who are freaking out vs. not, and humanities majors vs. science & engineering majors.

    I am not saying they are wrong, just that I would like to know more about what is really going on. Then after this weekend probably I start freaking out. Or not.

    • wheelers cat

      Amitabh, here is my empirical observation. When red phenotypes freak out, they question the data (polls truthers, jobs truthers, unskewed, etc). When the blue phenotypes freak out we question ourselves and President Obama (witness Sully’s epic meltdown).
      I personally find Romney so loathsome that its inconceivable to me that sentients could vote for him. But im a blue phenotype. Dr. Wang has mentioned this before and its a big theme of Talebs as well– how do we remove our bias and see the real shape of the data?

  • Ms Jay Sheckley

    “PEC gives the calm version of events. Does it feel calming?” Yeah, it does. I like, crave and appreciate information- Detailed, high quality information. If PEC suddenly disappeared, that’d freak me out. Thanks for taking it along on vacation. For me, the more you post the better I understand, and the therefore the better I feel.
    As to then oft repeated question of when to panic, Is there any use to panic? I have enough trouble being effective without panic. With my knack for identifying real experts, I just come here and believe you. The tiny, shrinking meta-margin shows whoever cares about the election -whether or not you favor a Bork-selected Supreme Court- there’s work to be done.

    Gandhi said,”Let us do what we can.”

  • Alan Cobo-Lewis

    Unskew the skew. More kurtosis! (Wait a minute, that could mean black swan, which is maybe what the debate already was.)

    What’s the name for the moment after kurtosis? (I’m looking for analogues of velocity/acceleration/jerk/jounce.) Perhaps we can just borrow the 3rd derivative of position to label Jack Welch? :P

    • Sam Wang

      Third derivative humor? Oh my, the geeks are out in force today!

    • wheelers cat

      Dr. Wang already said no black swan because not enough sigma. I proposed we call it a “temporal-game-changer” or TGC.
      I am starting to believe that Mandelbrot is a better model for human population behavior than Gauss.
      After all, fractals are found in nature.

  • Winston Smith

    You’ve got a kind of typo in the Wittgenstein quote. It’s “Whereof one cannot speak, thereof one must remain silent.”

    Just letting you know, not nit-picking. I am very appreciative of your work here.

    • Sam Wang

      My old philosopher/chemist landlord told me wrong. Or I remembered wrong.

      I love my readers.

    • Some Body

      Well, it’s all a translation from German anyway, so there is no exact original in English. The German original, if anyone cares to know, is: “Wovon man nicht sprechen kann, darüber muss man schweigen.”.

    • Ms Jay Sheckley

      re Wittgenstein- the translation Sam used is without apparent distortion. LIKE.

    • Winston Smith

      Some and Ms,
      Yeah, of course it’s from the German, and there are a couple of ways to translate it, but the way Sam originally had it wouldn’t work. Also, the translation I pointed to is quasi-standard.

  • Olav Grinde

    That’s another sharp tick downward today. Since the debate, Obama’s drop has been precipitous. As far as I can tell, this morning the incumbent’s Meta-Margin reached its lowest point of the year.

    I suspect the reasoning Howie Weiner expressed in his earlier posts may well be right, and that President Obama’s non-verbal message in the debate was a catastrophe.

    I do wonder to what extent we actually do vote based, not on issues, but rather the candidate’s body language and self-confidence?

    For instance, if you didn’t have the sound on, George W Bush seemed more presidential in his debate with Al Gore. Same thing with Mitt Romney.

    In the October issue of Scientific American, Michael Shermer has an article called “Politically Irrational: Subliminal influences guide our voting preferences”. According to Shermer, decisive questions as: Who has the deepest and most resonant voice? Who looks more vigorous? Who looks more competent?

    If Michael Shermer is right, it may be very difficult for President Obama to undo his damaging impression in the first debate — and the issues matter far less than we might think.

    • Anbruch

      I don’t want to go too far off topic, but I believe the magnitude of the bounce here was greatly amplified by the poor post-debate strategy of the left pundits and bloggers who turned Obama rather than Romney into their target. Even after months of Romney making errors, we were only starting to see this kind of attack on Romney from the right after the conventions (which is arguably what was driving Obama’s numbers up). In that respect, the debate did reveal a huge vulnerability for Obama in terms of the fundamental softness of his support (and that the left side of the media do not in fact have his back). That, not the debate performance itself, may in fact be a game changer.

    • securecare

      I think Anbruch is correct since the variable that seems to be driving the polls over the last week is voter enthusiasm & not any real swings in voters from one side to the other side. Look at the RAND data.

      Also note that the non-swing states are not being polled much in comparison which leads one to ask “what is happening there ?” & how might more balanced polling change the numbers.

      Less “hair on fire” and more sanity please. Of course that isn`t what The Media desires so “We the people…” get a horse race instead.

    • Howie Weiner

      Yes there are certainly irrational and non-verbal forces at work in any political (as well as economic and all human endeavors) process and the sharpness of the decline in the MM certainly reflects a gut reaction on the part of many, many voters. I think those of us who are ardent Obama supporters can certainly testify to the sinking feeling inside that we felt as the debate ended. I watched the debate at the local Obama headquarters in my neighborhood. The crowd (around 200) was 90% Afro-American and even among that audience, there was a subdued quality as people listening to the President’s final statement. One reason that Reagan was so popular was that he exuded a kind of calm “father” image that said “everything will be fine.” And in difficult times people gravite to that sort of appeal. Having said all this I feel the MM will begin to rebound over the weekend or early next week (polls are moving that way) as people start to process what has happened and recover from the shock. In the end, the vast majority of people act in their own best interests.

  • Froggy

    A week ago, just after the debate, I predicted that the MM was due for a tumble. Today, it’s poised for a rebound. If you look at the six states where the oldest poll considered in the MM is at least partially post-debate, five of them would look better for Obama if you dropped that oldest poll (or polls, where there is a tie on polling dates). Here are the numbers:

    MI: O+3 -> O+4.5
    PA: O+3 -> O+4
    OH: O+1 -> O+1 (no change)
    CO: R+1 -> Tie
    VA: R+1 -> O+3
    FL: R+0.5 -> O+2.5

    NV is also poised to make a move for Obama. NH, IA, and NC, on the other hand, each have only one fully post-debate poll, and probably will move toward Romney when the polling catches up.

    • Froggy

      I overlooked WI, which also has three post-debate polls, and which would shift from O+2 to O+2.5 if the oldest post-debate poll was dropped.

    • Joel

      Froggy, I’m with you here. I sure would like to see the MM move up tomorrow.

  • Anbruch

    I would feel calm if the meta-margin stabilized and started reversing for a couple of days and if the EV red-zone also stopped drifting downward. Eyeballing the polls as they have been coming in via twitterfeed, I would have thought the meta-margin would have stabilized by Tuesday. At this point my gut (which tells me that the bounce had played out by Sunday) and the numbers (which continued to drive the meta-margin down through this morning’s 8am update) do not coincide. Clearly, I have to recalibrate my gut.

    • Froggy

      In the last thread, (at http://election.princeton.edu/2012/10/10/time-to-panic-rejoice-neither/#comment-15147 ) I showed that the MM is poised for an Obama rebound. Like bad chili dogs, the post-debate bounce takes some time to pass through the system.

    • Brian

      Froggy – it’s important to remember that Obama’s +6.2% MM wasn’t normal either, and that we probably won’t return to those levels unless Romney pulls another “47%” moment. Sam’s model shows 3% as being the “real normal”. All this to say: Obama’s DNC/47%/Libya bounce was *also* a bad chili dog — given the length/size of that bounce, a *particularly* bad chili dog. Okay now I’m starting to get grossed out.

    • Froggy

      Brian, get with the program — the DNC/47%/Libya bounce was a sugar high, not a chili dog. But I agree, it was a bubble just waiting to be popped, which it has been, decisively.

  • xian

    the unskewed guy says he shamed the pollsters into “correcting” their partisan samples, btw