How do we know this? Byron York reported that he stated:
“For the last 20 years, between 37 and 39 percent of voters on Election Day have been Democrats ... Republicans have ranged from 32 to 37 percent."
Where did these figures come from? Not from his own polling, but from exit polling data. Jay Cost listed the exit polling data from 1972:
Notice how the last 20 years (1993-2012) of exit poll data matches exactly Rasmussen's statement. Republican ID ranges from 32-37%, while Democrats range from 37-39%.
Rasmussen went on in the same interview to state:
"Right now, our sample looks like 36 percent Republican versus 39 percent Democrat.”
Given the long term Democratic bias in the exit polls, these splits are almost certainly wrong. The bias towards the Democrats seems to generally be somewhere between 2 and 4 points ... in other words, the entire advantage that Rasmussen is giving them right now.
In fact, Rasmussen's own polling shows a much more volatile range over the last 9 years. Republicans have moved between 30.8 and 37.6% while the Democrats have been between 32.4 and 41.7%. His most recent polling on the question, in September 2012, shows a 2.6 point advantage for the Republicans, 36.8 to 34.2.
So, rather than using the results of his own up-to-date polling, showing a small voter ID edge for the Republicans, he is instead weighting his results to a small Democratic edge on the basis of flawed historical exit polls. If this was, ahem, unskewed ... Rasmussen's current rolling average showing a 2 point edge for Obama would instead show a 2-3 point lead for Romney.
I am not sure how any of this might effect his state level polls. If he weights according to the state level exit polling of previous election years, that would make most of his state polls lean even further toward the Democrats. He may, however, treat the state polls differently.