I’m not saying “most” elite privates do it the way Harvard does. I haven’t surveyed most elite privates. I have looked at enough to be able to say that some do it the same way as Harvard.
As I have patiently tried to explain, there are two problems with the way Harvard calculates its s/f ratio. First, it squarely violates the instructions given by the Common Data Set–they’re DOING IT THE WRONG WAY, therefore they end up with a s/f ratio that is simply not comparable to the s/f ratios of schools that FOLLOW THE INSTRUCTIONS and make the correct calculation. The information put out by these schools is therefore highly misleading. Second, even apart from the comparability problem, I don’t think it’s the least bit reasonable for a school like Harvard that has more graduate students than undergrads to pretend that the graduate students don’t exist and don’t take up professors’ time in teaching, advising, mentoring, etc. That’s precisely why the Common Data Set instructs them to count the graduate students as “students” for purposes of calculating their s/f ratio, unless the graduate students are in “stand-alone” graduate/professional programs like law, medicine, divinity, etc., which have their own separate faculties who don’t deal with undergraduates (except possibly occasionally, at the margins). It’s a completely fictional world they’re creating in which roughly 1,000 Harvard faculty are attending to every need of 6,663 undergraduates. It’s a good bet most of those faculty spend half or more of their faculty-student interaction time–teaching, advising, supervising, mentoring–with graduate students, not undergraduates. And it’s completely misleading, and deeply unfair to a LAC like Amherst for Harvard to claim the same 7:1 s/f ratio as Amherst. Like other LACs, Amherst is an undergraduate-only institution. Its faculty spend all their teaching, advising, and mentoring time with undergrads. As a Harvard undergraduate you’re not going to get the same level and quality of faculty-student interaction that you would at Amherst, because the Harvard faculty are going to be spending half (my rough guess) of their faculty-student face time with graduate students. Yet the gullible will look at Harvard’s phony 7:1 s/f ratio and compare it to Amherst’s legitimate 7:1 s/f ratio and conclude they’re likely to have the same kinds and levels of interactions with faculty at Harvard as at Amherst. And that’s just pure hogwash. Shame on Harvard. Lord knows, they have plenty going for them; they don’t need to gin up a fictitious s/f ratio to attract students.
Not only Michigan but also UVA, Wisconsin, Georgia Tech, University of Texas-Austin, and every public university whose CDS I’ve looked at (admittedly a small and unscientific sample) FOLLOW THE INSTRUCTIONS and calculate their s/f ratios exactly the way the Common Data Set instructs them to, i.e., including graduate students (other than those in “stand-alone” programs) as “students” for purposes of calculating their s/f ratio. Perhaps because as public institutions they have a greater commitment to transparency and accountability. You may think that’s quaint or “goofy.” Some would call it honest. And yes, I do think the CDS way is better, because it gives you a truer picture of the level and quality of faculty-student interactions you can expect, insofar as it accounts for the time faculty spend teaching, advising, supervising, and mentoring graduate students, all of which reduces the time they have available for undergraduates.