Missing data
This most recent salvo in the professorial politics wars seems to be missing something: students. (The first comment makes some valuable points about doing "[a]n ethnographic study of classroom practices.") I don't doubt the percentages of self-reporting faculty, nor do I think the faculty are lying or exaggerating. But I also don't believe that most of us, even the most self-reflexive of us, possess full self-awareness about our performance in the classroom. At the risk of making us all sound like the speakers in a dramatic monologue, our tics, crotchets, hobby-horses, and what-have-you all have a bad habit of manifesting themselves when we least expect it--and, often enough, without us knowing. I'm skeptical about Stanley Fish's neo-Arnoldian vision of the classroom--a utopia of disinterestedness--precisely because we may unwittingly betray ourselves every time we step in front of the class to lecture. Even the most controlled professor is still liable to slip now and then. Self-reporting tells us what faculty think they're doing, or what they think they ought to be doing, or what they think the interviewer thinks they ought to be doing. It does not, however, always tell us very much about what happens.
Ergo, students. But students aren't a straightforward resource, either. They may have partial information (e.g., their own grades, not necessarily anyone else's) or they may confuse a novel idea with a politically biased idea. For that matter, they will have their own political, cultural, and religious biases, all of which will affect how they recognize and interpret their instructors' biases. (And, of course, some of them may not really be going to class...) In fact, the inequality built into the faculty-student relationship may cloak some biases, even as it renders others visible; a freshman may not recognize a skewed syllabus, for example. Nevertheless, a broad representative sampling should allow an interviewer to control for most of these variables.
To complicate matters further, though, I think students need to be asked about politics in the classroom at least twice: once while they're students, and once several years down the road. Looking back at my undergraduate years, for example, I can think of two occasions when, at the time, I definitely thought that Someone Was Out to Indoctrinate Me. In retrospect, I still think that was true in one instance--an absolutely infuriating assignment that couldn't be completed satisfactorily unless you bought the politics involved, no matter if you had empirical evidence to the contrary--but not in the other, which was a case of my confusing new information with biased information. (Obviously, whatever else concerned me as an undergrad, political bias was low on my list--like most of my fellow undergrads, my complaints usually focused on professional lapses that had zilch to do with politics.)