> Such logic leaves you with no idea if the top X% is qualified or not, just that they are the top X%. That's what MBA schools want. They want to know who are

> Such logic leaves you with no idea if the top X% is qualified or not, just that they are the top X%. That's what MBA schools want. They want to know who are the top students on the GMAT: how does XYZ candidate compare to their peers? If you don't have a normal bell curve distribution and the results are skewed hard to the left (very difficult test) or hard to the right (very easy test), then you lose the ability to discern the relative aptitude of the test takers. > You need to enumerate requirements and test against them. That's beside the point. The GMAT is just one of many criteria that allow schools to determine who is a qualified candidate for admission into their MBA program. > Saying "wow, everyone passed calculus, that can't be right" and making the test harder serves no one. You're talking about something completely different altogether. Sure, for classrooms (i.e., a small sample size) there's a valid argument to be made against curving calculus tests or any other kind of test. After all, in a classroom the goal is to evaluate who understands the course material -- not to necessarily to rank the students. However, the GMAT is a whole different ball game. The purpose is to provide a relative ranking of the test takers. Additionally, hundreds of thousands of people take the test every year -- a massive sample size. You should expect to see a normal distribution if your test effectively serves it's intended goal of ranking the aptitude of the test takers.