Scott Ellison at Election Projections and C.H. Truth at Coldheartedtruth have independently come to similiar conclusions. Some of the polling this cycle seems to have a Zogbyish odor.
I'd love to use only polls that I "feel comfortable" with but I know that doing so means being unwilling to face unpleasant truths (bye, bye Rick and Mike and Lincoln and maybe Conrad - I'll miss you, Rick). Nevertheless, Scott Ellison did a great job in calling races in '04 and if he is offended by the odor of Constituent Dynamics (shows as RT Strategies/CD at Real Clear Politics) then I'll be looking at their polling with a higher degree of suspicion.
I wonder if Zogby will get a little of their reputation back with this cycle? In order to so they are going to have to be scrupulously fair in their next round - this one appears to be loaded on the Rep side.
I have a question concerning sample size - the '02 turnout rate was 40% of the Voting Age Population. If you are searching for 500 people to survey who are truly likely to vote, how large should your initial random sample group be? It seems like the answer should be 1200 but no one uses a VAP group that large to begin with according to crosstabs. How do you get 450-500 LVs out of a sample of 800 VAP? Magic sauce?
UPDATE: Rasmussen has Tester closing to down 3 compared to down 6,7 and 9 in previous polls. Anomaly? Or perhaps some of the Republican electorate don't cotton to the sound of "Majority Leader Reid". That's a song that Republicans in Missouri and Tennessee will soon be learning to sing. It creates an interesting dynamic.