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

Advice on looking at national polls

October 16th, 2012, 8:00pm by Sam Wang

Briefly: for now, don’t. I know this is hard.

The problem is that the uncertainties are too large. In 4 polls where contacts began after the VP debate, Romney leads in two, Obama leads in two. Going back further, one finds a mix as well, with more Romney polls. But since it seems likely that conditions are changing, interpretation is hard.

Update: To clarify, I have done this myself. Yesterday I detected a shift back from the 10/4-5 peak. However, the variation in national polls makes it hard to get a good read. Better to watch the state level, but this takes patience. You may now resume the flogging.

State polls are far more numerous, and are concentrated where one wants them. If you had to watch anything, I’d say (1) Ohio polls and (2) the meta-analysis in the bar above.

I imagine a flood of polling after tonight’s debate. I’ll miss it – traveling. Comment below.

Tags: 2012 Election · President

14 Comments so far ↓

  • Somewhat skeptical

    So yesterday it was “In national polls, the race has swung back three points since the Presidential debate to a narrow Obama lead. This return has been steady over time, and so the role of the VP debate is unclear.” whereas today it’s “Briefly: for now, don’t. I know this is hard. The problem is that the uncertainties are too large.”

    I appreciate your model and this site, but it seems to me your commentaries are tinged with wishful thinking lately (e.g. how many times did you say Romney’s gain was over and then it kept going?).

    • Sam Wang

      I think this is a good criticism. However, what I am really pointing out is that I am also looking too hard at national polls. The 2 Obama, 2 Romney might be a move…might not. The state/Ohio calculation is more robust. Clear enough?

  • Somewhat skeptical

    Yes. I am relatively new to this site, but I got the impression early on that you favor sticking to state polls only (after all, the model uses only state polls), and that seems like a perfectly good approach. But I guess that, like you say, it can be hard in the face of temptation.

    • Sam Wang

      Indeed. This moment is exactly the time to let go of national polls, mainly because of Electoral College mechanisms. Unfortunately this can be very dissatisfying because of the lack of time resolution.

      National polls may help see what happens post-debate2…but only if the change rises from the noise. Like I said, keep flogging.

    • wheelers cat

      I have begun to favor sticking to RAND only.
      Because they have solved the non-responder problem.

      Hope you had some file gumbo Dr. Wang.

    • Some Body

      Cat – I like the RAND methodology too, but no one poll can replace an aggregate (or meta-analysis). Besides, the RAND poll has recently been giving Obama his best results, so I’m afraid that by sticking to it you’re engaging in wishful thinking. And I see people’s need to be reassured by the polls, but it’s not them that count in the end; it’s your votes.

    • wheelers cat

      I do not engage in magical thinking like a red phenotype.
      RAND has decoupled the enthusiasm/response problem with a clever experimental design. So I think RAND is giving a more accurate (and unique) picture of the sample population.
      RAND is not just another poll.
      ALL the other polls assume Gaussian on the sampled population. I do not think that is true. I believe in asymmetrical political behavior. For example, I think enthusiasm is more strongly correlated with LV for republicans than it is with democrats.
      If anything its my background talking. Game theory and evo bio.

  • Matt

    The suggestion seems to be that, inevitably, things are going to get better for Mr. Obama because A) an enthusiasm swing always has some sort of peak, B) Biden had a “draw” rather than “loss” against Ryan, C) The strongest polls for Governor Romney are set to expire with new state polls, D) those state polls seem to be waiting for the second national debate to grow, and E) That electoral math will continue to favor Mr. Obama down the line.

    I am honestly a bit shocked at how flat the line has come out since the end of the 1st debate peak, but I foresee, no matter what the result, Mr. Obama being declared a modest winner when the polls correct themselves. Since Republican enthusiasm is up in general from its dismal low, I simply cannot see how Obama can get back to the high he was once at.

    • wheelers cat

      Disagee Matt.
      Want to bet me?
      That was like Wesley Snipes vs. the Nosferatu in Blade.
      you know how visual I am. Romney looked old, pale and creepy.
      pop culture and cafe au lait complexions rule.

    • Froggy

      This debate marks the end of the flat line following the first debate peak. I predict a quick jump in the MM, certainly to over 1%, perhaps even overnight.

    • Matt McIrvin

      The PEC numbers are based on state polls, which have been really sluggish coming out over the past week. The Meta-Margin dropped almost instantly after the first debate because three Republican-leaning pollsters jumped out of the gate with snap polls immediately after the debate concluded. That didn’t happen after the Biden/Ryan debate and it’s probably not going to happen this time.

      There’s been an overnight change in Obama’s favor, but I don’t think it has anything to do with this debate, because there just hasn’t been time. It might be Biden’s win starting to work its way into the numbers.

    • Froggy

      Okay Matt, you got me — the overnight jump had nothing to do with the reaction to last night’s debate. It came from a bunch of YouGov polls taken 10/4-10/11, that all showed up on Pollster just before last night’s debate began. Given the polling dates, the bounce from these polls has nothing to do with the Biden-Ryan debate either, since the YouGov polls were completed prior to that debate.

  • pechmerle

    “The U.S. Supreme Court will not block early voting in Ohio, the Associated Press reported on Tuesday.
    The Court declined to take up a case pushed by Republicans to settle a dispute over early voting in the closely contested state.” So the Court of Appeals ruling that early voting goes forward in Ohio stands.

  • Ohio Voter

    Almost on cue, Gallup has R +6%LV.

    I’m ignoring them, I’m ignoring them, I’m ignoring them.

Leave a Reply to Matt (Cancel)