Prompted by Ed's excellent piece, I looked at the NBER working paper that undoubtedly will be used to undermine tenure at a variety of universities. It is worth noting that one of the authors of the piece is the current President of Northwestern University; undoubtedly this helped with gaining access to the data. But it also makes one wonder if, perhaps, there wasn't a directed search in the data. After all, one can get a bit too close to the subject studied. It would be nice if some independent statistician can obtain access to the raw data.
Either way, one question worth asking is this: why did the study not control for class size? For one plausible scenario that could account for the results is that (long-term) non-tenure-track faculty teach relatively small introductory sections, while 'star' faculty teach large survey introductory courses assisted, say, by an army of PhD-student graders. (I am familiar with such pattern of teaching at a variety of elite institutions not unlike Northwestern.) Of course, if my hypothesis were on target, it might offer a good argument to abandon large survey courses.