High school grades in the U.S. aren’t standardized and aren’t reliable. A standardized test like the SAT is the strongest predictor of college success: https://www.reddit.com/r/Sat/comments/1alp6vh/the_evidence_i... (collecting evidence)

I took a grad marketing class once with a business professor who studied this sort of thing. GMATs rather than SATs but same idea. Basically GMATs mattered more than anything else especially metrics such as letters of recommendation that were basically worthless.

I knew the director of admissions somewhat at an elite school and he said that they basically put a couple of quantitative metrics (like SAT) on one axis and read essays and considered other metrics like interviews on the other axis for diversity before that term became popular.

The upper right more or less got in, the lower left didn't, and then they debated the middle ground.

In Quebec, grades are normalized using a statistical formula that factors in how well students from your high school tend to perform in university[0]. This means an average student at an "elite" school could end up with a similar score to a top student from a weaker school.

[0] https://en.wikipedia.org/wiki/R_score

There's a reason the military kept using the ASVAB even during the worst parts of COVID pandemic. ASVAB is a very solid predictor of success in training, and in Navy experience it's predictive value generally correlates with with how academic/technical the training pipeline is.

That paper is pretty misleading and flies in the face of most peer-reviewed research (I don't know that journal, for what it's worth).

My guess is because it was focused on those attending elite institutions:

"In their paper on admissions to highly selective colleges... students at each of the schools in this analysis... Students opting to not submit an SAT/ACT score achieve relatively lower college GPAs when they attend an Ivy-Plus college..."

My guess is the meaning of a high or low GPA versus standardized test changes quite a bit when you have groups very highly selected based on a wealth of other information.

The Dartmouth report has always frustrated because they, along with that other paper, selectively present conditional means rather than scatterplots, hiding the variability around points to make things look more predictive than they are. Means by predictor level are almost useless without knowing the conditional variance for each predictor level. They're basically deliberately pretending that there is no error variance in the prediction equation.

Meta-analyses suggest that both standardized test performance and GPA predict later performance. For example:

https://www.tandfonline.com/doi/abs/10.1080/10627197.2015.99...

In some literature, GPA is superior, and others, testing.

There are other studies from decades ago showing that when standardized tests are temporarily removed from admissions (e.g., due to a court ruling), it has almost no influence on outcomes of admitted students later, suggesting admissions committees are able to select comparable students without tests.

I'm not saying tests are horrible and should be omitted, I just think people really overstate their predictive utility and it causes a ton of problems down the road.

So? In many countries high school grades also aren't standardized and counts 100% for admission. The system still works reliably and not worse than in america.

The SAT isn't strong enough to predict anything. It can generally be answered by someone in their sophomore year at college or even their freshman, depending on what level of courses they are taking.

The problem finding a hard enough test with as little human intervention for assessments. Because human intervention brings with it subjectivity. This subjectivity was manageable when there weren't so many people applying for top schools (e.g. in the early 1900's). But right now its not.

SAT/ACT/GRE are no indicator of success. What this "study" is merely proving is that schools may have regressed in their rigor for grading hard courses.