A word about some of the battle ground states, including Florida, Ohio, North Carolina, Virginia and Pennsylvania.
As you know, McCain has been claiming that the “gap is closing” in many toss up states. This is an expected exaggeration, but it is not pure fantasy, and it is a bit worrisome. There are three important factors here.
First, the McCain campaign has less money than the Obama Campaign (because fewer people like him, and all the rich people are busy losing all their cash on the economy that Bush/McCain has destroyed). Because of this, you would expect the McCain campaign to hold back on some of their cash until the last several days because later advertising is more relevant to votes than earlier advertising (up to a point). You should adjust any polling you see by about 1 percentage point in favor of McCain because of this.
Second, the Republicans have developed, over the years, a “72 hour strategy.” This has to do with how the on the ground organization of the campaign is carried out. You can expect ALL of the results to be shifted towards the republican number by about two percent off of something like the Real Clear Politics average because of this.
Third, there is the combination of Bradley effect, the Bubba vote, the unreliability of the Youth Vote, the African American Vote, and all the other factions that don’t vote much, often say they are going to vote, but then at the last second, manage to not show up. We have no idea how much these factors will play out. The Bradley effect is largely a myth, and I think the bubba vote vs. the black/youth vote pullback may cancel each other out. To be save, I’m giving this one percentage point for McCain (which I hope is not a serious underestimate).
So in total, whatever numbers you see in the polls, you should shift four points from Obama and to McCain. So, in my view, the best point estimate of the status of the candidates in the state by state popular vote is the Real Clear Poltics average of the last few polls (with some adjustments, where a longer term average is more appropriate) adjusted by four points in favor of McCain.
Now, a look at some of these toss up states:
Florida: Obama has a small, statistically insignificant lead. Contrary to reports by the McCain camp, the gap is widening or staying stable, possibly oscillating back and forth. There is no evidence of a closing of the gap in this state.
Nothing dramatic has changed in Florida. The two candidates have been oscillating back and forth since October 1st. I therefore think the best estimate of their difference is the average of all polls since October 1st. From this, I calculate using RCP data 3.33, which differs very little from the RCP average (for only more recent polls) of 3.5. Making the adjustment recommended above, this puts McCain ahead by a half a point in Florida. My best estimate is therefore that McCain will take Florida.
In Ohio, again, there is no evidence of a narrowing of the gap, and Obama is just above a statistically significant lead. However, Ohio is a large state known for its crooked politics, so this is not by any means in the bag. The polls have been similar with what looks like random variation since October 22. Tossing out one absurdly large number (in Obama’s favor) and using RCP data from 22 October to the present, I calculate an average of 5.31, which is close to RCP’s estimate from recent polls of 5.8. Applying my adjustment, Obama is ahead in Ohio by between 1 and 2 points. I am reluctant to estimate that Obama will take Ohio because the Dirty Politics Factor may be bigger there, but hoping that the Dems are watching more closely than ever, and for the sake of consistency, I think I’ll go ahead and take the plunge. Ohio = Obama.
North Carolina has one of the most anomalous runs of data of any state, with an overwhelming surge of support for McCain arising in September and disappearing in the same month. The difference between the two candidates has remained statistically insignificant for weeks. The RCP spread is 2.6, and any alternative estimate based on a longer term will be smaller. There is no question in my mind that McCain will take North Carolina.
The spread in Virginia has been increasing in favor of Obama, and is at a statistically significant level, but just barely. Going back to the 15th of October, we see relatively stable but oscillating poll results. Using all RCP data from that point, I calculate a spread of 7.14 points, in contract to RCP’s estimate of 6.5. Virgina will be sending its electoral delegation to vote for Obama.
A gap-closing trend IS happening in Pennsylvania, possibly due to the intensive efforts of McCain in that state. Calculating from Oct 10th, after which the RCP data is relatively stable, I get a spread of 10.33. The RCP spread using only the most recent polls is 9.8. The Bradley Effect, the Bubba Factor, and the RNC surge are all strongest in this state, so I’m going to adjust this estimate by six rather than four points, and use the RCP estimate as the baseline. The spread will be between 3 and 4 points, with Obama still taking Pennsylvania. This is subject to change as more polling data comes in.
Pennsylvania is tricky. The last two polls are a partisan poll and a Mason Dixon poll (which tends to be conservative), at 4/5 points. From this, we may guess that the spread is currently larger, like 6-8 points. On the other hands, it might not be larger (the MD poll underestimates the spread). The good news is that as this spread decreases, so does the undecided estimate. In other words, both Obama and McCain are picking up likely voters, at the expense of “U,” as opposed to Obama bleeding votes off to McCain.
Now, off to the electoral map for some more extensive number crunching …