The campaign of Mike McGinn just gave me permission to post an off-the-record e-mail that they blasted out to reporters this morning, questioning a Washington Poll that has McGinn trailing Mallahan by eight points (and also shows McGinn's change of position on the deep-bore tunnel causing him to lose votes).

From the e-mail:

The Washington Poll violates the cardinal rule of polling and the media should be very very skeptical:

They do not report the sample size of their poll.

And it's actually worse than that. They list a statewide sample size of 724 on the poll. But that is very misleading. They do not report the sample size for the poll in the mayor's race.

The population of washington state is 6.5 million. The population of Seattle is 582,000. Seattle represents about 9% of the state population.

Thus, if the statewide sample is 724, a proportional sample in Seattle would be a sample size of 65. Yes, that's right...65!

The pollster makes allusions to "oversampling" King County and the City of Seattle, but NEVER DISCLOSES what the extent of the oversample is. This is highly irregular in the scientific world of polling.

Low sample size also probably explains why they were so far off in their pre-primary polling:

Nickels 23%
Mallahan 11%
McGinn 8%

It's true, the new poll—which was conducted over 11 days—reached 724 registered voters throughout the state, and while the pollsters say they "oversampled" for Seattle, they don't provide the actual size of their Seattle sample.

It's not clear that the size of the sample was 65 voters, as the McGinn campaign speculates. That would indeed be a ridiculously small sample, if true. But the point is: we don't know the actual size of the Seattle sample.

I've sent an e-mail to the University of Washington pollsters asking for the sample size.


Matt Barreto, one of the pollsters, tells me via e-mail that the Seattle sample size in this poll was 400.

The Seattle oversample was an n of 400; and so was the King County oversample; The McGinn people are down in pretty much all polls, and I suspect they are just upset. As you probably know, our polls have pretty much predicted the outcome of every contest since we starting this effort in 2006. We have a team of public opinion scholars who are experts in polling and have PhDs, so we'll let the data speak for themselves.

The McGinn campaign, for its part, tells me that their internal tracking poll—which uses a Seattle sample of 1,000—has McGinn in a statistical dead heat with Mallahan.

Impossible to verify, of course, since it's an internal poll. But it's worth noting that the Seattle sample size the McGinn campaign is claiming to use is bigger than the sample size the Washington Poll used for the entire state.