Okay, so keep in mind that the NRC ranking survey I posted about a few days ago not only doesn't allow you to include faculty size as a criterion for department quality, but also presumably the measures of faculty quality used will be based on averages. Averages are only fair, right? Otherwise, small departments would be at a disadvantage, and that's unfair because, well, it's, well, um, unfair.
The trouble with averages in a measure of department quality is simple: to use averages is to imply that half the members of a department do not merely contribute less to the quality of a department than their more productive/cited/grants-getting colleagues, but that they actively harm it. Take any department, fire the lesser half according to whatever the quality measure is, and you've instantly created a better department. You can even repeat the step and create an still better department, until you have a one-faculty member graduate program.*
I'm not sure how many faculty members are actively bad for the programs that employ them, in the sense that the program would better if they immediately disappeared and took their hiring line with them.** I feel confident, though, in my sense that this number is quite a bit less than half at most places. If you are in academia, think about your own department, and don't just single out the worst person, but think of somebody you think would probably be a little below the dividing line. Do you really think your department would be better if that person just vanished? They really hurt more than they help? Maybe so: but, if that's really true for half your department, I hope you aren't surprised by my response that your department must suck.
Average measures also imply that if you take a department and add to it by hiring everyone's structural equivalent, you haven't improved the department at all. Think graduate students are better off having four people in their specialty area rather than two? Average-based measures of quality imply no. In sociology, faculty size has always seemed to me especially an issue to attend to insofar as the discipline is sparsely populated even at the largest departments and many graduate students change their mind about the area they want to work in after they begin (or come in not knowing what they want to study). The smaller the department, the more likely a graduate student is to suffer from a change of interest, a falling out with their advisor, or their advisor leaving.
Not that I think faculty size should be the be-all-end-all, especially as one would expect it to be correlated with what Fabio calls programs of "benign neglect". I would rather program ratings try to measure benign neglect more directly, however. Anyway, if faculty size ends up having no weight or very little weight in the way NRC ultimately does the "objective" part of its rankings, that's just not credible.
* Half per se assumes either we are talking average in terms of the median or a measure that is symmetrically distributed. You know what I mean. My guess is there are ~40 sociologists who would be have the highest "faculty quality" of any department in the country if only they could get themselves into a department with no colleagues.
** Not that they left and someone else was rehired; there is a difference between someone being not the best use of a faculty line and being actively bad, and averages treat half the department as actively bad.
Addendum: I actually wrote this post a couple days ago, and in the interim Dan has written his own reaction to my earlier post. Incidentally, I suppose one could think it convenient that I have this strong opinion about faculty size while being a member of one of the largest sociology departments in the US. Maybe so. But it just seems intellectually galling to me that you'd have a rating system based on the idea that if there were two great small departments out there and you could somehow merge them, the result would be no better. It's one thing to say a whole isn't greater than the sum of its parts, and another to say the sum is no better than the average of its parts.