Why is this so? Along with Bloomberg, Rasmussen uses a "likely voter model" as opposed to a "registered voter model" in selecting the pool from which poll respondents will be chosen. It turns out Republicans are more likely to vote than Democrats so a likely voter model will contain more Republicans proportionally than will a registered voter model.
Bottom line, from a pool of registered voters the Republicans are more likely to vote than the Democrats and so, if you are trying to predict voting outcomes, you will weigh Republican responses slightly more heavily than Democrat responses.
Is Rasmussen correct or are the other pollsters who survey registered voters? Personally I like Rasmussen. Here's why: as a kid I remember my late father, a lifelong Southern Democrat, being mad that "those darned Republicans all vote." I suspect he was correct then, and now.
Republicans are rule-followers, it's how we were successful enough to become Republicans. We are supposed to vote so we vote. Democrats are more likely to question (or blow off) authority, including the authorities who tell them to vote.
