Rasmussen, Mason-Dixon on two different planets
Not only are the top lines nearly opposite -- Rasmussen has Scott leading Sink 50-44, while Mason-Dixon has Sink up 47-40 -- but the internal findings are just as divergent.
Take Rasmussen's finding that Scott holds a monster 21-point edge among no-party voters. In Mason-Dixon, Sink holds a 16-point edge among the same voters. Consider the favorable/unfavorable margin. In Rasmussen, Scott enjoys a 53-39 split, and Sink's almost even at 45-46. Mason-Dixon has Scott underwater in this area, to the tune of 30-47, while Sink is at 44-23.
Rasmussen has taken heat in the past from liberals for being biased toward Republicans. Polling guru Nate Silver explains that Rasmussen has a different model than other polls, predicting a much more conservative turnout on Election Day. Whatever the case, for those of us who drink coffee and care about polls (though we're certainly not polling experts), today's Rasmussen survey could be described as a "coffee-spitter."
The big differences: How they modeled, Mason-Dixon polled over two days (Sept. 20-22), while Rasmussen polled on one day (Sept. 22). Mason-Dixon uses live operators to poll. Rasmussen uses Interactive Voice Response (i.e., "robo" polls). The biggest difference between the two pollsters could be how they screened their likely voters.
As we saw during the just-ended primary, many polls (including Mason-Dixon) showed McCollum with a good lead with less than a week to go before the election, but he lost. We can't say how Rasmussen called the election because his polling outfit refused to post/poll the Florida Republican primary race -- something Scott supporters said was evidence of a pro-McCollum bias (something Rasmussen Reports denies). Of course, now the Scott people can't get enough of how great Rasmussen is.
-- Lee Logan and Marc Caputo