Rasmussen uses a standardized system, so I think it increases the noise in his numbers. I believe he also uses likely voters more often than not, although I’m not sure that’s true in these daily tracking numbers.
The source of error rasmussen is particularly good at controlling is "party id". Although there have been endless discussions on FR about whether or not party id should be a controlled figure in polling, I think the way rasmussen does it has basically been proven to be highly accurate. What he does is, he polls for "party id" over long periods of time and uses an average over a huge amount of data. That means his party id figures are reasonably stable, but they do respond to demographic changes over time in a pretty accurate way.
Other pollsters either do not control for party id, meaning they frequently oversample democrats or republicans leading to more error, or they correct for party id using old figures from the prior major election which may not be accurate anymore.
To see how important party id is in controlling error, realize that democrats rate obama's performance positively at a rate of 77 to over 80% in recent polls, while republicans rate him favorably in the single digits. If you oversample one or the other group even by tiny amounts it nearly adds error at a 1 to 1 ratio to the oversampling.
Time and again Ras has been proven the most accurate. No polling is perfect, but there is good evidence that the way Ras does it is among the best available.
>>> Rasmussen uses a standardized system, so I think it increases the noise in his numbers. <<<
It probably has more to do with his normalizing for the number of R’s, D’s and I’s.
R’s and D’s are probably consistent groups.
But I’s are all over the map. So I would expect that his numbers fluctuate because of shifts in the different classes of I’s that he polls