Wednesday, October 22, 2008

In the Presidential Race Tightening, Heading Toward a Landslide, or What?

Something quite odd has been going on with the presidential race polling data we’ve gotten over the past few days.

The RCP average of major national polls has moved a few points in Obama’s favor so far this week, with several surveys now putting his lead near or in double digits.  However, two polls that came out today show the race significantly closer, with the respected IBD/TIPP tracking poll showing an Obama lead of only 3.7 percent [note: see update below] and an AP poll showing a statistical dead heat.  Additionally, the latest result from the GWU/Battleground tracking poll gives Obama only a two point lead.

Moreover, things get even more interesting when we consider three state polls conducted by the highly regarded (by political pundits) Mason-Dixon polling agency.  Two NBC/Mason-Dixon surveys released today have Obama up by two in Virginia (well below his lead in other polls) and McCain—yes, McCain—up by one in Florida, respectively.  An NBC/Mason-Dixon poll released Sunday put McCain up one point in Ohio.

Finally, the political prediction markets currently see Obama as an exceptionally strong favorite.  On Intrade today McCain’s predicted probability of victory reached a new post-primary low of 12.6 percent.

So what the hell is going on?

My guess is that right now Obama’s “true” hypothetical lead* is about five points.  That’s roughly consistent with the Mason-Dixon state polling data (factoring in that from the  2004 election results one would probably expect a GOP presidential candidate to run somewhat better, to varying degrees, than the national average in Virginia, Florida, and Ohio).  Or at least more consistent than a double digit Obama blowout; I just don’t see Obama winning by ten points nationally but lagging that much in those states.  Moreover, a five point lead is closer to those national polls mentioned above showing a tight race. And last, a five point lead is exactly what today’s edition of the warhorse Gallup Daily Tracking Poll (with the traditional model of likely voter screening based on past voting behavior, not the new version that classifies someone as a likely voter just on the respondent’s bare assertion that they are going to vote) gives Obama. 

And as for the prediction markets, which proved prescient in the 2004 election?  Well, today I bought $25 worth of McCain contracts on Intrade at a 12.8 percent chance-of-victory price level.  To put it in ordinary market terms, in my view McCain’s chances have been way oversold.

However, if McCain’s predicted chances should rebound to say, 35 percent? 

Well, there’s a time for putting one’s money where one’s mouth is, and then there’s a time to take profits.


*Yes, it’s  incorrect to speak of any polling lead as a “true” lead, given that a poll is, at best, only a statistical sampling measure that takes an imperfect snapshot in time, and even more incorrect to say that about a guess as to what a perfectly valid poll would show.  Fortunately, this is a blog post, not a final exam answer in a poli sci statistics class.  


Update/Correction (10/24):  After reading a couple of items floating around the net in the past few days calling into question the IBD/TIPP poll’s current methodology and doing a bit of further review of the poll’s recent results , it now appears to me that the poll isn’t producing reliable or plausible results for this cycle.  I regret referencing the IBD/TIPP poll without doing some further research into its performance this year, but still stand by my guess that race is considerably closer than the majority of national polls are now showing, a guess that is supported by the other polling data I mentioned.

No comments: