14 pollsters competed in the Florida Republican Primary. Only 2 pollsters released real-time poll results on how cell phone respondents and home phone respondents voted, so scholars could compare the two groups. Turns out: those 2 pollsters, Marist Institute and SurveyUSA, had the smallest error (the best performance) of the 14 competing pollsters.
Not all pollsters who excluded cell respondents had large errors, but the single largest error of the 14 pollsters came from a polling firm that did not include cell respondents. Chart below illustrates.
Quinnipiac University, which reports it dialed cell phones but did not release the results of cell respondents versus home phone respondents, also had an error of 0.25, using the Mosteller 5 scientific measure. Quinnipiac said Romney would win by 14 points.