Wednesday, August 1, 2012

Poll shows Obama with solid leads in three swing states, BUT....

A new Quinnipiac University/CBS/New York Times poll shows Barack Obama holding suspiciously large leads in Florida, Ohio, and Pennsylvania.

a. Florida: Obama 51% Romney 45%

b. Ohio: Obama 50% Romney 44%

c. Pennsylvania: Obama 53% Romney 42%

I say suspicious because Obama hasn't had that large a lead in Florida since March (also a Q poll) and only leads there by 1.4% in the RCP average.

He hasn't had that large a lead in Pennsylvania since May, and out of 18 polls conducted in the state this year, has only led by that much in 2.

When you look at the RCP averages and histories of the state-by-state polling, only Obama's big lead in Ohio seems reasonable (see average here).

One possible reason for these results? Sampling. The pollsters sampled Democrats at +9% in Florida, +8% in Ohio, and -- very weirdly -- only +6% in Pennsylvania.

How can you sample Democrats at only +6% in a state with a huge Democratic registration advantage (Pennsylvania), but go with +9% in Florida and +8% in Ohio?

To wit
: As of June, Democrats only had a 52%-48% registration advantage in Florida. So you're going to sample Democrats at twice that rate in an election where enthusiasm favors Republicans?

Meanwhile, Democrats have a 57%-43% registration advantage in Pennsylvania, and you're going to sample Dems at a lower rate than Florida?

As for Ohio, the state doesn't release partisan registration totals, so we're out of luck there, but again, it's hard to imagine why you'd sample Pennsylvania at +6% in favor of Democrats and +8% in Ohio, which is a much redder state.

Having said all that, these polls have some value.

You can dig into internal results to see how specific groups are voting etc., but it seems like a bridge too far and a mountain too high to accept the overall results.