The polls are insane today, and why McCain is still fighting in Pennsylvania

by JS O’Brien

As usual, I checked out the early polls this morning, spilled hot tea all over myself, and have spent the rest of the day trying to figure out what the @#$% is going on.  As the day has gone on, the polls have only gotten squirrellier.

First, the overall picture is that Obama has gained a bit.  The Real Clear Politics average is back up to a 6.9-point lead for Obama after dropping down to nearly four.  That’s the average.  But the spread on the most recent polls is stunning.

  • The GWU/Battleground Poll has Obama with a one-point lead, 48% to 47%
  • The Pew poll has Obama with a 14-point lead; 53% to 39%
  • Gallup, which has taken the cowardly way out by publishing three poll results (“all registered voters,” “likely voters 1” and (expanded) “likely voters 2″), puts Obama up by seven points in Ilikely voters 1”, and 10 points in “likely voters 2”
  • Rasmussen, a high-quality and steady poll, maintains the Obama lead at four points
  • Zogby, which tends to lean Republican because it weights its likely voters based on turnout in 2004, gives Obama an eight-point lead, 50% to 42%.

There are a number of other polls yielding various results, but the message is the same:  Obama appears to be gaining, but the standard deviation on these polls is bizarrely high and begs the question: What’s going on?  After spending much time going over as much data as are available on methodology for these polls, my answer is a simple “I don’t know.”

By all rights, Zogby should be showing a closer race than it is because it probably underestimates the percentage of Democratic voters this year.  Instead, it’s actually a bit higher than the median.  Apparently, GWU/Battleground recently changed its polling technique because it had been undercounting young voters, who tend to lean toward Obama.  So, one would expect this poll to actually be at least in the middle.  In addition, Battleground is reporting that the number of undecideds remains largely unchanged, so the implication of its results is that a large number of voters who had already decided for Obama switched to McCain quite rapidly – enough to cut Obama’s lead to one point.  That seems highly unlikely

Go figure.

I’m just going to assume that both Battleground’s one-point Obama lead and Pew’s 14- point-lead are anomalies.  The real Obama advantage is probably back up to where it has been for weeks:  a six- to eight-point lead.

The state races, which are what really count, show a tightening in Florida, North Carolina, and Ohio.  West Virginia seems to be solidly back in the McCain camp.  All these states voted for George W. Bush, so it’s not surprising that they would respond to McCain’s attack ads and begin to “come home.”  Virginia still appears to have around a six- to ten-point Obama lead, and while Virginia has been reliably Republican for a very long time, liberal Northern Virginia has expanded to the point that the state has elected a Democratic senators and two Democratic governor recently, and the Democratic candidate for the state’s other senatorial seat is expected to win easily this year.  Virginia is still an Obama lean.

State results tend to lag the tracking polls a bit, so the next few days could be very interesting

McCain still fighting in Pennsylvania

Ordinarily, a candidate trailing in the polls by ten points or more in a state this close to the election would abandon that state and focus on places where he has a greater chance to win.  That’s why so many people are wondering what the heck McCain is doing spending so much time and money in Pennsylvania, which has been considered solid for Obama for at least three weeks.

The answer is in the electoral college numbers and the probabilities.

To reiterate an earlier post, if McCain takes all the states that are solid or leaning towards him, AND he takes all the toss-up states, he will still be 18 electoral votes short of the 270 he needs to win.  That means he must take back one or more of the states currently solid or leaning towards Obama.

Currently, the only states leaning towards Obama that are not considered “solid” for him are Virginia (13 EVs), Colorado (9 EVs), and New Mexico (5 EVs).  If McCain has only the path through these three states open to him, then he simply has to flip Virginia and either Colorado or New Mexico.  If his odds of flipping Virginia are 35%, and Colorado and New Mexico’s odds are 40% each, then the probability of flipping Virginia and one of the others (which he must do to win) is only 14%.

Another path to victory is to lose Virginia, Colorado, and New Mexico but take Pennsylvania and its 21 EVs.  If the odds of flipping Pennsylvania are only 20% … well … those are better odds than flipping both Virginia and one other Obama lean, aren’t they?

James Carville has said about Pennsylvania that it’s “Pittsburgh on one end, Philadelphia on the other, and Alabama in between.”  McCain probably hopes that he can win Pennsylvania if turnout in the cities and inner suburbs is low and turnout in the “Alabama” part of the state is high.  Who knows?  It might work.  Regardless, McCain’s odds are still very long, and unless there is a substantial swing in his direction soon, election night’s only suspense will center around tracking the number of Senate and House seats the GOP loses.

4 replies »

  1. Nate Silver on Five Thirty Eight did a long, long piece about the different strengths and weaknesses of the polls. He agrees on Rasmussen being the most dependable.

    I forget where I read this, but someone mentioned how North Carolina had served its purpose because Obama didn’t need to win it so much, just turn it into a toss-up so McCain would be forced to spend money there. Seeing as he’s concentrating so fervently on switching Pennsylvania, I’m wondering if that was money well spent. Pollster has been showing NC going more steadily towards Obama despite being toss-up.

    Either way, the arithmetic is not working out well for the Republicans.