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

A bad July for Romney

August 1st, 2012, 4:07pm by Sam Wang

I spoke too soon. July was not good for Romney. The trend is now clear:
and in units of Meta-Margin, it’s even clearer:

Disproportionate July effect of Bain / Romneyshambles / ?  in swing states The size of the change here is considerable, almost 3.0%, for a total of Obama +4.9%. That’s the amount of swing necessary for Romney to come back even.

The discrepancy between this and national polls is likely to stem from a combination of (a) biases in national polls and (b) real, focused change in key states such as Florida, Ohio, Virginia, and Pennsylvania. Whatever the campaigns are doing there is working to Obama’s advantage. (Update, August 10: another possibility is (c), time lags between national and state polls. At any given moment, national or state polls could be a leading indicator, depending on which are more recent.)

In the coming weeks I’ll show how to turn this snapshot into a real forecast, driven only by polling data. As I’ve written before, the odds of a November Obama victory are about 10-1. I’ll post some relatively transparent math to prove that point next week.

Tags: 2012 Election · President

14 Comments so far ↓

  • Bill N

    I noticed a discrepancy between your current state projections and those of Nate Silver at 538-dot-com. For example, you are currently showing Florida as a dark blue in favor of Obama, while Nate Silver shows Florida as a light pink (about a 50.7% lilelihood) for Romney. I assume that methodological differences explain these discrepancies. Can you explain the methodological differences and how they lead to such a variation in estimates of which candidate a state is favoring?

    Thanks in advance.

    • Sam Wang

      There’s something fundamentally different in what he and I mean by “probability.” A 50.7% likelihood doesn’t make any sense as a probability. Why would a probability derived from multiple polls be far less certain than a single poll? Fundamentally, Silver’s estimates are not real probabilities. He tends to count errors multiple times, mix together evidence about the present and the future, that kind of thing.

      As for the Florida thing, it’s pretty cut-and-dried. Obama’s currently ahead in five of seven Florida polls. He has a median-based lead of +1.0 +/- 1.2%, for a win-probability-today of about 80%. Probably Silver is using some funny business like GDP or unemployment reports. This is a conceptual error: it combines poorly-predictive information with precise information about today (i.e. polls), to give something that is neither a snapshot or a prediction. It’s only useful for a missing-data situation, as might occur in House or Senate races. I’ll write on that in the coming weeks.

      The bottom line is that Obama is unambiguously ahead at the moment. He’d win today, no question about it. As for November…stay tuned for my thoughts on that.

  • Sean

    Just as I expected.

    August might just be it for Romney. While I don’t think he can win the election this month, he can certainly lose it. It’s geared toward him – with the running-mate pick and convention. If he’s still unambiguously behind Obama on Sept. 1st, outside some external event that changes the dynamics of the race, I don’t know where he pulls ahead.

    We’ll see.

    • Sam Wang

      The conventions are unlikely to matter because their effects are transient. The only scheduled event that is likely to change the EV estimator in a lasting manner is the first debate.

  • Sean

    Probably right. I think there will be temporary movement for Romney, but it won’t last, especially with the Democratic Convention starting mere days after Romney’s convention speech.

    The debates will prove interesting. That first debate was when Kerry made his move and got back into the race from what I recall. It wasn’t enough to overcome Bush’s lead, but the debate really proved the difference between a very close election night and a comfortable Bush reelection.

    With that said, if Romney’s best chance is that first debate, you’ve got to wonder how good of a chance it really is. Obama is a good debater and not in the sense Kerry and Gore were – he’s very personable and likable in debates. Romney is a good debater too, but is prone to getting upset easily, which leads to gaffes, and he’s also so stiff and awkward that I don’t know how that’ll play next to the relaxed Obama.

  • Steven Clemens

    I simply cannot figure out why EV analyses such as on this site and at least 5 others I follow are not reported on by news networks / political analysts. The vast majority of people I speak with have no inkling as to how lopsided the EV map is at this point in the election cycle.

  • Matt McIrvin’s historical EV graphs:

    It appears Sean’s right about the debates. Kerry had had a narrow lead in the summer, Bush took the lead after the Republican convention, but Kerry recovered well enough around debate time to take it right down to the wire.

    I remember figuring out in hindsight that you could have called it from the last week of state polls, but only from the very last week.

  • Matt McIrvin

    …The idea that the PV/EV polling discrepancy is real, and down to swing-state phenomena, is fascinating.

    Has the split been this large before? In 2000, of course, both popular vote and Electoral College were nearly ties, which is how a discrepancy (and all the other craziness) became a real possibility.

    But one thing you can say about Obama’s campaign operations is that they’ve always understood how to play the map. That was particularly clear during the 2008 primary.

  • Hans

    Quick point, feel free to delete. Since it’s not a comment more a site question.

    On the front page, it still seems to be showing the 2008 charts, not the new 2012 ones. Is that intentional? Is it just me and my browser?

    • Sam Wang

      I thought only the House/Senate widget was old. Is there something else? We are slow this year, please be patient!

  • Brian

    @Steven Clemens – The reason you don’t hear about this stuff on TV is ratings. The media outlets utterly depend on a 6+ month presidential race that they can continually describe as “neck and neck”. They need to post individual polls that show the trailing candidate (in this case, Romney) “might be closing the gap”. Meta-analyses like these take most of the excitement and uncertainty out, which makes the viewers watch something else.

    As the election gets closer, you’ll see the cable networks line up to trot Nate Silver out for a segment in which he’ll predict that Obama will win, but it will be presented by the host as no different than that octopus that makes predictions, and they’ll close the segment with “well, we’ll see…”.

    • Sam Wang

      I think you are basically right. It is not in the NYT or Nate Silver’s interest to be more accurate.

      It’s too bad, because the real neck-and-neck race is who will control the Senate and House. This will have profound consequences for Obama’s second term. If this were emphasized, it might engage people more, because voters have far more influence in a district/state election than in a national election.

  • Slovo


    I wish I wish that this recent Quinnipiac poll is beginning of the big trend, but an article on:
    got me worried. Especially the end.

    • Sam Wang

      A major benefit of poll aggregation is that it spares you the trouble of fussing over individual polls. Chewing over the internals of a poll is exactly what I never do.

      You are also doing what these pundits want you to do – pay attention to them. A direct answer to Hewitt’s foolish questioning is that the Quinnipiac poll’s result is not an outlier: for example, Obama is well ahead in Ohio.

Leave a Comment