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

Will the remaining debates matter?

October 2nd, 2008, 10:54am by Sam Wang

The EV estimator has reached its highest level since July. The Meta-Margin is nearly 4%. Once a few stragglers come in – notably Missouri and New Hampshire – the picture will be reasonably current. The tactical benefits to McCain of the Celebrity ad (and other August attacks) and VP pick Sarah Palin have been fully reversed.

A CBS survey says that two-thirds of registered voters will watch the debate between Senator Joe Biden and Governor Sarah Palin. Palin’s future on the national stage may hinge on her performance. But is there anything that can happen that will affect the fate of the McCain/Palin campaign? Basically, no. (though see update below)

As I’ve written, game-shifting events seem to occur about once every 20-30 days (though it’s not always possible to identify the triggering event with confidence). For now let’s concentrate on the size of the events. First, let’s look at 2004. Andrew has applied the new rule (last 3 polls and last 7 days of polling) to that dataset to get a smoother result in October than what I’ve shown before.

There was a slow run-up from mid-June to mid-August. But the big game-changer was the Swift Boat ad campaign, which attacked Senator John Kerry’s wartime heroism and yielded 60 EV. The Republican convention yielded another 30 EV or so. Kerry’s strong performance in the first debate helped him (30 EV), but it wasn’t enough.

But here’s the important point: the remaining debates did not move the race. For those who recall, President Bush turned in a considerably stronger performance in the second debate than in the first. But evidently it didn’t matter.

What about tonight? Clearly this event has the potential to be more important than past vice-presidential debates. Palin’s convention speech was followed by an exceptionally large single-day bounce. But the bounce has dissipated in strong correlation with her unscripted television appearances. If she did well tonight, she might conceivably help McCain recover a little lost support. But let’s face it – however well she does tonight, she’s unlikely to persuade many people that she’s ready to be President, much less change their preference about the candidate himself. (Here’s some insightful analysis from

The remaining question is whether non-campaign events can change the picture. It’s hard to think of an October Surprise that would be large enough to change the November outcome. If there is, there’s an excellent chance that you won’t be spending much time on this website. Stockpiling canned goods maybe…

(Update: here are historical data from Gallup and From 1936 to 2004, 18 races, the only large swings in October not triggered by a debate were 1968, when Humphrey surged by 14 points against Nixon, and in 1992, when compiled data (but not Gallup) showed a 10-point swing for GHW Bush against Clinton. Neither was enough to carry the day. Based on these data the probability of a big late swing at about 2 out of 18 = 11%. Apologies for previous miscounts!)

Tags: 2008 Election

16 Comments so far ↓

  • Patrick

    Interesting to see the 2004 EV estimator. At the time I didn’t realize Kerry had been trailing for that long. (I was still getting my news from TV back then.) I would be very interesting in having a “this day in 2004” feature added for the histograms.

  • Bill B

    “Stockpiling canned goods maybe‚Ķ”

    Nice. :^)

  • Tony Z


    I wish I could share your optimism, but I worry. My model of “game-changing events” is that they’re drawn from a distribution with a long tail— it’s not too hard to imagine something (think Spitzer…) that would change the course of the election without leading us into Road Warrior territory.

    Let’s keep our fingers crossed.

  • Andrew Foland

    Iran, which a few months ago might have been considered a leading October surprise contender, now merits about 10% of a foreign policy debate.

  • Elizabeth Duvert

    Sam, It’s clear you’re working on a second career on late night comedy shows. Thanks for the post.

  • Ginny Mayer

    I have been contemplating whether or not something will occur that will shift the focus from the economy to national security, in which case presumably voters will turn to McCain as the more qualified candidate as polls have been showing him as strong in this area. Yet even so, I am not certain that the shift to a security focus would work in McCain’s favor at this point. Obama’s apparent win in the foreign policy debate suggested that McCain is not the undisputed champ in dealing with foreign threats, and his hectic and rash reactions to the economic crisis may have undermined his image a a strong, reliable leader (not to mention growing concern about putting “Palin” and “Commander in Chief” in the same sentence). If there is an October surprise on some national security-related issue, the surprise may ultimately be on the McCain camp (as the rest of us “go about stockpiling our canned goods.”:)).

  • Sam Wang

    Patrick, your suggestion is interesting one. Let me add it to the list. In the meantime, there’s a similar function available at, run by Andrew Tanenbaum.

  • Craig Z.

    Are enough comparable polling data available for previous elections? While it appears that the later debates didn’t have much effect in 2004, is that enough data to draw a general conclusion. If we had a similar meta analysis for a few more elections we might then be able to better assess whether the conclusion is warranted. I suspect that it may be hard to get everything you would need to do a complete meta analysis, but perhaps there is enough of a correlation between national poles and the EV estimate to shed some light on such questions (I’m assuming national pole data may be more readily available and easier to look at).

  • leah

    Hey Sam,

    I got turned onto your website by a friend of mine and it’s great! Since expectations are so low and so many excuses have already been proffered explaining her less than stellar performance in recent interviews, I think even if Sarah-intellectual-powerhouse-Palin melted into a babbling heap of tears on the floor during tonight’s debate, it would not change peoples’ opinions or the outcome of the race. Let’s face it, they are already suggesting that the debate is unfairly skewed in favor of the democrats based on choice of moderator. Unfortunate (and unfair) as this seems I think one slip up from Biden will affect the opinion (and polls) of voters.

  • leah


    I’m curious, what is the probability of her using “ing” at least once during the entire debate?

  • Andy

    Wasn’t that 8 point swing in 1980 the result of the first debate, which happened late, because Carter had tried to avoid debating Reagan. Well Debate #1 has already happened this year, so I think it might well take a canned-goods situation.

  • Eddie

    6% from the sample of 18 elections between 1936 and 2004, with the 1 case being an 8%(!) shift, and information is shared today at an incredible faster speed than 10, 20, 70 years ago.

  • Sam Wang

    Andy, good catch, thanks. I had forgotten how late the debate was in 1980. Now there’s a game-changing debate. The update is modified to take this into account.

    Leah – “ing”? Is this some mannerism of Palin’s?

  • Mark S

    One detail: I think the history at shows that Clinton dropped 10% in the last 3 weeks of 1992. He still won, though.

    A question: I’m surprised that there is such a close correspondence between the EV estimate and events. State polls are sporadic, and the 3-poll average should cause a time lag (which also varies from state to state).

    This is most noticeable right now. The Popular Meta-margin just took a huge jump (from +2.5 yesterday to +4.2 today). The national tracking polls seemed to move at the end of last week… so there appears to be a 1-week lag before it filters through the state polls and comes out here. Is that right?

  • Sam Wang

    Mark S, again thanks for this correction. That’s what I get for rushing to post.

    In regard to a time lag, a week is indeed about the right delay for post-Labor Day state polls. In addition, there will also be a delay of this length built into our averaging rule.

  • leah

    hey sam

    I’m be”ing” an articulation snob and should be thoroughly spanked down.

Leave a Comment