When in Danger, When In Doubt, Run In Circles, Scream and Shout!

I’m home after another long, long day of teaching and then another election talk, but I wanted to comment briefly on today’s Pew poll which has driven many Obama supporters to despair.  That poll has Mitt Romney up by 4%, 49%-45%, over Barack Obama among likely voters, and tied at 46% among registered voters.  This is a sharp turnaround from the last Pew poll in the field Sept. 12-16, which had Romney trailing Obama by 8%, 51%-43%.   The latest poll represents a 12% gain by Romney in less than a month – a turnaround that just rubs salt in the wound for Democrats already reeling from polls suggesting that Romney cleaned Obama’s clock in the most heavily watched first presidential debate since 1980, when incumbent Jimmy Carter squared off against Ronald Reagan in their only debate. (The figures for the Romney-Obama debate do not include those who viewed it on social media.)

Consistent with other polls, Pew found that Romney was viewed as doing a better job in the debate by 72-20%.  This included 78% of independents and even 45% of Democrats who thought Romney bested Obama.* As debate polls go, this is a rather significant drubbing; rarely do we see such a lop-sided verdict, particularly among the “loser’s” own partisan supporters.  If the polls are to be believed, Romney’s debate “victory” has led to a significant tightening of the race, both nationally and in the critical swing states.  For example, Gallup’s pre- and post-debate polls indicate that Romney has moved from a 5% deficit into a tie with Obama among registered voters.

In the RealClearPolitics composite poll, we see a similar result, with Obama’s 3.1% lead on the day before the debate dwindling to .5% tonight.

We see a similar effect in state-level polling in battleground states. In Michigan, the latest poll has Romney within 3% in a state considered out of reach just a week ago.   Colorado, Florida and Virginia are now essentially dead heats, and Romney has moved within 3% of Obama in the critical state of Ohio, a state in which he trailed by nearly 6% before the debate.

All this is a reminder of two points I have made repeatedly: first, national tides raise all of Romney’s state-based boats.  Too often pundits view states as having their own unique constituencies.  But the reality is that both candidates are fighting over the same type of undecided voters across all states, and if one candidate is able to win over these undecideds, it will boost his support across all the battleground states.  We see precisely this effect occurring after the debate.

Second – and in what some may view as a contradiction of my first point – we should not overestimate the impact of the first debate.   I have been arguing for some time now that the state-level battleground polls will gradually align with the national tracking polls.  At the same time, I have claimed that the economic fundamentals indicate that this will be a very close race (the mean prediction of the dozen or so political science forecast models has Obama winning slightly more than 50% of the two-party vote.)  If the mean forecast model is correct, Obama will win by a far smaller margin than what the national tracking polls were saying for most of September.  And if we factor in the uncertainty surrounding those forecasts, many political scientists believe this election is a dead heat. For that reason I was reasonably confident that the September polls indicating that Obama was running away with this race were overstating his support due, in part I believe, to how pollsters were constructing their likely voter screens.  What Wednesday’s debate did, I suspect, is to impact Pew’s likely voter screen in ways that increased the number of Republican respondents they included in their final poll relative to Democrats.  In other words, the debate didn’t switch votes so much as it increased Republicans’ enthusiasm for their candidate enough so that it affected Pew’s likely voter screen.

As evidence, consider the gender gap. One of the more surprising findings from the Pew poll is that Romney has apparently drawn even with Obama among women. Last month, for reasons that I discussed in a recent Economist post, Obama led Romney among women by 18 points, 56%-38%.  If the latest Pew poll is to be believed, Romney has now drawn even with the President with women, at 47%.  Did he really erase Obama’s lead among women in less than a month?  I suspect not.  Instead, I think this is probably a function of how Pew constructed their sample after Wednesday’s debate.

My bottom line is that Wednesday’s debate focused enough attention on the fundamentals to erase Obama’s polling lead which was largely based on his relative advantage in framing this race in a way that played to his strengths.  But we shouldn’t overreact and buy into Pew’s results which indicate that Romney has now established a substantial lead.  Instead, my read of the composite polls indicates that the race, as of today, stands almost exactly where I have been arguing it has stood for the last two months.  Obama is ahead, but by the slimmest of margins.

Game on!

*An earlier version of this post had those numbers slightly off – I’ve corrected them here.


  1. Hi Professer

    Thank you for the updates and analysis. I’ve found them very useful for making sense of where things are in the election.

    Is there any chance that you’ll be posting your election model? It would be interesting to see how you think this will end and compare it with the other models available.

  2. Ariel – Just to be clear, I don’t have an original model. Instead, I stand on the shoulders of those who do have models by, in effect, treating them as something like random draws from the underlying vote distribution, and then calculating my own forecast. The method has proved reliable in the past but, as they say in the investing business, one’s past record is no guarantee of future results! I should have had this out three weeks ago, but simply haven’t had the time as yet.

  3. Matt – My guess is you were remembering all my previous brilliant predictions: Rick Perry doing well in the Republican primary, Rick Santorum going nowhere, Scott Brown getting trounced by Martha Coakley, etc.

  4. Thanks for the response. Time permitting will you post a forecast or is it a bit late in the day?

  5. Pew’s numbers shifted even among registered voters (though they shifted more among LVs). So that leaves one of three possibilities, which are not mutually exclusive:

    1) The Pew poll is an outlier due to sampling error
    2) Dems became radically less likely to answer polls and Reps radically more
    3) There’s some real opinion change going on

    My kingdom for a panel study!

Leave a Reply

Your email address will not be published. Required fields are marked *