[drawnwren]

Keep in mind that top 5 US schools are taking a much smaller percentage of a much larger population. They simply can't differentiate on academic performance alone. They generally admit somewhere around 5% of their applicants. Keeping in mind that those applicants are self-selecting, to some degree, thanks to the previously known admit rates and high application fees -- they have to use something else. A professor of mine at Stanford told me that Stanford starts by disqualifying all students that they believe won't make it at Stanford. This leaves them with about 60% of the applications. They then take the academic prodigies out. You know, researched mouse cancer at 14. That's about 1%. Stanford's admit rate was 4.something% last year. Getting that other 3% is a pain for US institutions.

[arkh]

In France, the most sought after enginnering schools use an entrance exam. No need to check for old academic performances as everyone is tested on the same things.

They usually also have alternate admittance system but that's for 2% or 3% of students.

[melkiaur]

In France, Polytechnique is probably the most sought after. There's an oral exam to get in, and guess what ? Only 25% of student from Paris are eliminated, versus 40% of kids from la province.

7% of girls are eliminated in the Computer Science branch (MP-Info), versus 27% of boys.

More info, in french: https://www.lemonde.fr/campus/article/2014/11/25/l-ecole-pol...

And besides, the real filtering is actually done after the Baccalauréat: not everyone gets to be admitted to the big Parisian Lycées (Henry-IV, Louis-Le-Grand, etc.).

[Amezarak]

By 'academic performance' he is almost certainly including their ACT and SAT test scores, which serve essentially as standardized university entrance exams in the US. The problem is for schools like Harvard or Stanford, almost all the applicants have near-perfect scores, so they're not a good differentiator - people with lower scores will not apply, since they know they will not be accepted.

[thecleaner]

I guess thats a problem with the SAT. I saw a couple of questions and a lot of them are very easy for college entrance exams. I think university entrances should be a bit on the tough end (China's standardized tests and JEE from India come to mind). Of course schools could look at a lot more but if so many people get perfect scores then that clearly is a problem with the tests themselves.

[Amezarak]

0.06% of ACT takers scored perfectly in 2013. A smaller percentage get a perfect score on the SAT. I don't think the problem is the test is too easy.

The problem is there are 400 million people in the USA, and the 1-2% of entering freshmen across the country having high scores mostly apply to the same few schools.

[georgecmu]

The challenge of SAT is not in the difficulty of the questions. The challenge is that there are 40 questions and you have 1.25 minutes to do each one.

[Ntrails]

When i was at uni they described the intention as:

    40% for book learning (memorisation)
    30% for variations on things you've seen before
    30% for applying concepts in new ways, requiring "deeper" understanding

From that perspective, if I want good students I don't care if you can answer a standard question quickly and accurately, I care whether you can reason something about a harder question given time to think.

[thecleaner]

I see. That is tight. But that simply tests thinking under time pressure. Fewer problems with higher difficulty level may help here (like the Olympiads or Putnam but both being extreme examples). But the challenge would then be to come up with large number of such problems.

[jstoeffler]

The French system he’s talking about is more a contest than an exam. They are intentionally too hard/long for anybody to score perfectly. This way it’s possible to differentiate between two excellent students. Not saying it’s perfect, but at least it’s sort of fair.

[extra88]

Don't the families who are well off spend a lot of money on test prep? The bright students who don't have the time or money for that are at a disadvantage. There's also the unlucky ones who have something happen and can't do their best on the exam. High-stakes testing has its own problems.

[cycrutchfield]

If you want to see how putting the entire admissions process on a single entrance exam will play out, look no further than the dystopian cram schools of China and India.

Unless we want to train a generation of rote-memorizers and elite test takers, this is not a direction that should be pursued.

[home_boi]

Those admissions processes are actually seen as very fair processes in those countries. IIRC, there is very high economic diversity in the top colleges in China and India which can't be said about the Ivy League schools.

[cycrutchfield]

Those admissions processes are actually well-known for producing excellent students of rote memorization and standardized tests. They are not well-known for producing well-rounded students that are capable of thinking outside of the box.

[baud147258]

In particular, these written part of the entrance exam usually cover multiple schools, which allow students to pass two or three written examination, whose result will examined by multiple schools.

[Spooky23]

Just do a random lottery, assign points for objective achievements, whether academic or extracurricular. Pick some low (say 5%) percentage of admissions subject to discretion.

The legacy/z-list stuff demonstrates that the curation/engineering of the student body aspect of the process is bullshit. Kids who are marginally qualified and have rich parents seem to be able to make it.

Introducing a lottery system would eliminate the injustice and resentment where some admission officer is applying a "cool filter" to determine whether chinese/korean/indian people are worthy, or creating the assumption that other minorities were somehow unworthy of their opportunity.

I'm just an idiot who went to a SUNY school. In the 90s, if you met an SAT threshold and were a B-ish student you were in. We used to joke that we were the 13th grade of Massapequa high school -- there was no curation of the social skills of the student body. That said, we had a fine social life, and in rigorous subjects the dopiest idiots were filtered out. Computer Science 201 was a lecture hall with 800 students, and about 40-50% of the population was culled each year. About 40-50 students graduated. It was no Harvard, but we seemed to muddle though.

[SiempreViernes]

Just do a unbiased lottery draw for the rest, that's about as fair you can do until you go down the really dystopian full facebook micro surveillance.