Let me work out some logic here.
Lets assume you’re correct and that “affiliation should be based on last general election” is the standard by which Rasmussen determines the sample of his polls. If that’s the case then:
2012 polling should be based on 2008 results -D+6 or D+7 (Obama won by about 53-46).
and
2008 polling should have been based on 2004 results - R+2 ( Bush won by 51-49)
So, that being said
Rasmussen was accurate in 2008 - his polling predicted a 52-46 turnout and that’s within tenths of a point to what it was.
According to that logic, he’s saying:
He polled in 2008 using an R+2 sample
He came up with Obama winning 53-47 (or 52-46)
I am not a Rasmussen subscriber so I cannot go back and check the 2008 internals but I find that outcome - using an R+2 sample returned Obama winning by 4-6 points - very difficult to believe.
There is chicanery going on here. He’s varying his methodology for some reason (unexplained, to my knowledge)
It's nice to know I'm not alone.
However, I would think someone close to the industry lurks here and will step forward to explain.
Rasmussen does not use that standard. His basic standard is a running poll of 15,000 respondents.
Rasmussen Reports tracks this information based on telephone interviews with approximately 15,000 adults per month since November 2002. The margin of error for the full sample is less than one percentage point, with a 95% level of confidence.http://www.rasmussenreports.com/public_content/archive/mood_of_america_archive/partisan_trends/gop_edge_in_partisan_id_slips_to_2012_low