In a few questions/answers on here (e.g., here), passing reference has been made to the difficulty of determining a good/acceptable/healthy dropout rate for a PhD program. However, I haven't seen any questions directly asking how this question is or can be addressed. In practice, it seems this question must come up. If a lot of students drop out of a program, the faculty (not to mention the other students) must begin to wonder if something is awry.
The difficulty, of course, is that an ideal dropout rate is probably neither 0% nor 100%. (If no one drops out, the program is likely to be too easy; if everyone drops out, the program is probably too challenging.) How do departments/schools deal with this question in practice? How do they distinguish between "good" dropouts (where the faculty and student agree dropping out is best for all concerned and there is no ill will) and "bad" dropouts (where the dropout is due to burnout, a harsh environment, lack of funding, etc.), and track those outcomes over time? How do they avoid self-protective rationalizations (e.g., "our program can't be too/hard easy" based on review of the program content rather than the dropout rates)? Are there established ways of looking at the dropout pattern and evaluating it?
As a sort of side note, I'd be interested in hard data on dropout rates. Although there are numerous popular-press articles and blogs lamenting the dropout rates, the only concrete data I've been able to find is this slideshow and a few other things that seem to be drawing on the same data. (It's rather surprising that there is so much discussion of the "problem" with so little discussion of how to decide whether it's actually a problem or not. There seem to be many articles saying "the dropout rate is so horribly high!" and very few saying why it's considered high, and what would be considered low or normal.)
Just to clarify: I recognize that the dropout rate is not the only meaningful assessment of a program's health or quality. My question is not "How do you decide everythig about a program based just on the dropout rate?" My question is "How do you decide what the dropout rate is telling you about the health and quality of the program?" You will of course combine that information with other kinds of assessments of the program.