This morning, Clinton's internet director Peter Daou did a quick driveby to post a couple polls intended to contradict the recent wave of polling showing Clinton tied or falling to second place in Iowa polls. Daou pointedly didn't comment on when the polls were taken or their methodology.
Turns out both polls were sampled beginning the first week of November, conducted through mid-November, and finished well before the recent wave of polls showing a surge by Edwards and Obama. In the case of the Iowa State poll, the sample ended in mid-November.
Jonathan Stein of Mother Jones isn't having it.
As someone who uses statistical sampling and controlled communities daily in my professional life, this sort of stuff kills me. But forget that the laughable polling methodology, which depends on a small and scattered 2-3 week sample across a large, volatile population, is something that even a stats undergrad would groan about...yielding 6% margin of error.
It's the timing itself that really stinks:
Here's the catch. The AP-Pew poll was conducted November 7-25. Some of the results there are two to three weeks old. The Iowa State University poll was conducted November 6-18. All of the results there are two to three weeks old. They all predate the juvenation the Obama campaign has gotten going in Iowa these past few weeks.
To get a better sense of the current race, Stein is more than happy to provide:
More current numbers all show the race tied or with Obama leading slightly. An American Research Group poll conducted 11/26-11/29 has Obama 27%, Clinton 25%, Edwards 23%. A Des Moines Register poll conducted 11/25-11/28 has Obama 28%, Clinton 25%, Edwards 23%. A Rasmussen poll conducted 11/26-11/27 has Clinton 27%, Obama 25%, Edwards 24%.
The average, according to Real Clear Politics, is dead even: Obama 27.5%, Clinton 27.2%, Edwards 22.3%.
Planting polls in blog communities, sampled weeks ago, and spinning them as "new" results, is an awful lot like placing questioners in debate and event crowds and not mentioning their affiliations. Partial truth is not truth. The missing data is the critical part. It's not the issue date of the poll; it's the sample date. Conflating the two is fundamentally dishonest...a Rove-like gambit.
Someone is getting real nervous if the Internet Director himself dumped these dated polls in Kosland.
It's about the internals, folks. And the internals are clearly rumbling in a few campaign crews down the stretch.