[almeria]

And what about predictions based on GPA and test scores?

As has been in use at UC and just about everywhere else, up until now.

In math terms: just because predicts(A) > predicts(B), it does not follow that predicts(A) > predicts(A ∪ B).

[eesmith]

I take it you haven't consulted the references I listed second-hand? They are likely far more insightful than anything I can write here.

Before making my earlier comment, I read the start of the the Hiss and Franks paper is at https://web.archive.org/web/20140310113612/http://www.nacacn... to make sure the citation I gave wasn't misrepresenting the topic (it wasn't). Here's text from the abstract:

It "examines the outcomes of optional standardized testing policies in the Admissions offices at 33 public and private colleges and universities, based on cumulative GPA and graduation rates."

That is, UC isn't the first to do this, and we can look at real-world evidence from previous universities where test scores were optional, to gauge how useful test scores are and what effect they have.

It found: "Few significant differences between submitters and non-submitters of testing were observed in Cumulative GPAs and graduation rates, despite significant differences in SAT/ACT scores."

That is, SAT/ACT scores don't seem to affect metrics like Cumulative GPAs and graduation rates. (As I recall from elsewhere, they do correlate with first year grades, but that's a different and less important metric.)

Further: "Optional testing policies also help build broader access to higher education: non-submitters are more likely to be first-generation-to‐college students, minorities, Pell Grant recipients, women and students with Learning Differences."

That is, using SAT/ACT scores in the selection process appears to have measurable effect on the student population; reducing what is sometimes referred to as "diversity."

As to your correct observation, "it does not follow that predicts(A) > predicts(A ∪ B)", one of the issues is that college success is also correlated with other factors, including parental wealth. And success on the SAT/ACT is also correlated with parental wealth, who can afford special training on how to pass those tests.

We know this by looking at the early history of college boards, which emphasized the topics taught at prep schools (like Latin grammar) than public school, because college admissions preferred rich white male Protestants, who were likely to go to prep school.

To be clear, the SAT has done a lot of work to de-bias their tests, and I don't know enough about to topic to say anything what factors are actually involved.

But I don't need to, since the tests don't seem to be that effective in predicting college success.

[almeria]

Thanks for the detailed and nuanced response.

Which, if I can attempt to boil it down to one sentence, would seem to be: "Tests don't seem to be a good predictor of college success, even when taking GPA into account. But they do seem to promote diversity."

And to which I would like to add, solely from anecdotal observation[1]:

"They also provide a second chance for a significant number of people. Even with a bad year, or even two, in high school -- this doesn't necessarily mean you are doomed to a lifetime of minimum wage servitude. If you have the intrinsic cognitive skills, you can easily do moderately well on the SAT, even without prepping."

I would like to consider diversity and the possibility of a second chance to be not merely incidental, but essential benefits to be striven for in the admissions process. Instead of simply stack-ranking based on the most easily demonstrable predictors of success.

[1] Including observations of some of the smartest and most creative and inspiring people I have ever met -- but again, that's anecdotal.