http://www.buzzfeed.com/rubycramer/conservatives-embrace-alternate-polling-realitySome explanation/tl'dr: Of all the polling firms out there, one of the least reliable, most skewed polls tend to be Rasmussen. Almost consistently they've skewed far to the right when polling, for reasons that can't be explained by their use of landlines alone.
Nate Silver hates them, and a few aggregators exclude them from their calculations.
All polling data skew their data according to demographics like political affiliation, region, age, etc. Because pollsters usually only poll a few thousand at most, they will skew the data according to what they this is the makeup of the likely voting populace. All companies do it slightly differently, but Rasmussen has always tended to include more Republicans than Democrats.
This last part is important, because this is has led the right wing to accuse everyone else of bias towards the Democrats. One guy took it further and actually reworked all pollsters' data to fit Rasmussen's demographic and affiliation assumptions. So unsurprising for a makeup that is more heavily Republican, Romney wins.
Of course we already have solid evidence of how inaccurate Rasmussen is. In 2008 and 2010 they've performed worst of every pollster; there's a reason
this site has a Rasmussen-free map. This site says nope, the dozen or so other polls are wrong and this consistently-inaccurate company is right. Worst part is the gall of the title though. Unskewed? Really?