Do you want to be able to identify the best of the best? That's the fundamental question here - if that's a useful thing for society, to have credentials that distinguish beyond "did well in high school," then you don't want grade inflation.

And if you don't - you could make a solid case that the extreme ends of those skills may not be that useful in general outside of academia - than do Harvard et al need nearly as much money as we throw at them today?

Peer group matters. Being bottom 10% at Harvard is quite different than being bottom 10% at Boston University (no offense).

The people who seem to not believe this are the people pushing for grade inflation, no? Grade inflation is all about making everyone who isn't top 10% have a prettier resume.

Ah so the us veriosn of the "gentleman's" degree - a third from Oxbridge in the UK

Identifying the best of the best requires complex statistical methods to remove several different types of bias.

I'd rather give students PRoPS, which is to say "percentile ranking of passing students". But that would require some means to normalize different professors teaching from different curricula, with variations in the capabilities of their students across different classes.

Automatically failing x% of your students every semester is as cruel and pointless as stack ranking. Clearly, if a student can demonstrate a minimum objective level of understanding, independent of the performance of their peers, they should not fail the course. Equally clearly, if a section is full of geniuses with perfect understanding, they should be graded more highly than a section full of normals that can barely get it after intensive study.

If you pile up enough testing data, year after year, it would be theoretically possible to construct a mathematical model to compare with some confidence the student achievement across disparate classes, professors, and institutions, as well as to say definitively who the good and bad professors are, and whether they write difficult or easy exams.

If you get a 100% on Dr. Aardvark's CHEM220 final at Example U., that may put you at PRoPS of 80 for "organic chemistry, milestone one", whereas if you get a 60% score on Dr. Zymurgy's CHEM235 final at Hypothetical Institute, that may actually have a PRoPS of 95 for "organic chemistry, milestone one". Dr. Z's exam was simply written to be impossible, and scored very harshly, while Dr. A's exam was built to be more forgiving, and maybe left out some elements that the Example U. chem faculty prefers to save for later courses.

This will, of course, never happen. Firstly, you would have to get academics to agree on the essential elements of a given subject at a particular checkpoint in a course of study. Secondly, institutions that rely on selective admissions and political connections for prestige would never participate, if there were even a shred of possibility that the objective evidence would show that they do not actually produce the quality of graduates that they would like to claim. Imagine if Ivy Limestone Woodpanels Law cranked out PRoPS for complete J.D. from 30 to 85, and Bigcity Urban Campus Law ranged from 50 to 99. The former is selective and does a lot of networking, but then passes anyone who still knows enough to not embarrass the school in a connected job, whereas the latter takes everyone, presents a huge challenge, and quickly flunks those who can't master the coursework and get hired through sheer merit. Those numbers would be a huge blow to the former's reputation.

Grade inflation is thus a natural by-product of institutions trying to be a "top tier school". It goes right along with publishing in the highest impact journals and wheedling more money out of your alumni. Schools that benefit from institutional prestige want to keep their reputations polished and shiny.