Stanford, Harvard, Dartmouth, Yale, Penn, Brown, CalTech, JHU, and UT-Austin to Require Standardized Testing for Admissions

But this is looking at it backwards. Given that we already had an extensive study that showed that the use of the SAT results in far better outcomes for low-income students, we don’t need another study to confirm that the lack of the SAT was the reason many more students are doing poorly. All that will do is result in more students floundering at UC who could have thrived elsewhere in California’s educational system.

I will pull directly from the UC Academic Senate report, which can be found here:

Let’s start with the predictive power of HS GPA only vs SAT only. Even back before COVID style grade inflation, the SAT was more predictive of UC freshman grades than HS GPA among lower income students ($60K and under).

Let’s remember the R^2 values of 0.12 and 0.13 for HS GPA only, and now let’s compare it to combining both HS GPA and SAT for this same group:

The R^2 values improve to 0.22 and 0.23 when we add the SAT, a substantial difference.

1 Like

Mods, how much leeway do we have to discuss UC’s use or non-use of the the SAT on this thread? I can address mtmind’s remaining points in detail, but only if it’s deemed on-topic.

Rather than repeating exaggerated claims and reproducing highly edited graphs, here is the actual study:

Despite the highly manipulated graph that is often shown to support the claim that “SAT is HIGHLY predictive of success”, this study actually demonstrates that all that they can say is that 17%-20% of the variance in first year GPA can be explained by SAT scores.

That is not really “highly predictive”, and furthermore, “first year GPA” is only a small part of what “success” means when looking at a college career.

The report actually demonstrated that “SAT has a moderate predictive ability for a single component of college success”. Exactly how this is translated to “SATs are HIGHLY predictive of college success” is a mystery to me.

1 Like

I am now reviewing the report again after a few years.

Upon seeing the actual numbers again, I will change my assessment from the SAT being “highly predictive” to just “far better than HS GPA alone”. Would you agree that’s accurate?

As for me reproducing “highly edited” graphs from the report, can you point to any one in the report that you feel is inaccurate?

As long as the posts remain focused on universities and standardized testing, the moderation practically takes care of itself.

1 Like

IMO, You aren’t really addressing my points at all, so what’s the point of any back and forth? All of the material in my previous comment came directly from the study. The study doesn’t indicated what you have represented.

I would agree with “better”.

Not you, and not the people here specifically.

I have seen these types of graphs repeatedly in the opinion pieces that support the use of SATs. They use graphs that include on the averages of first year college GPA versus averages of SAT scores without error bars, and generally title the first year GPA as “college GPA”.

An example of such a graph, though it is using a different data set than that used by the UC study, is this one from the NYT (the study that they cite specifically says that it is first year GPA). I am especially disappointed in the authors of the article that they cite for using this sort of graph and not including data like explained variance.

Just to be clear, I do see a place for testing in admissions. I don’t like the SATs specifically, and I don’t like how they used in admissions.

Rather than try to modify a test that was initially created to discriminate against minorities and to “prove” that minorities were inferior, the entire concept should be rethought from bottom up.

What the analysis here is ignoring is that the increased variance in SAT scores isn’t because higher SAT scores are being used cull admission to only high scoring kids, but rather because of the UC policy targeting kids with lower SAT scores.

“In other words, the increased variance is by design, an artifact of admissions policies which intentionally target lower-SAT applicants.” Getting rid of these low scoring kids would cut directly against the UCs Mission.

@MWolf,

I agree that this is graph you showed is highly deceptive, as it shows a far tighter relationship between SAT and test scores than actually exists. I really wish that everyone was required to take a statistics course in high school.

We both agree that the SAT and ACT have flaws. I think we both also agree it can be misused in a number of ways (e.g. to “prove” racial inferiority). And further agree that some colleges, particularly small liberal arts colleges, do a great job of selecting academically capable students without the SAT by delving deeply into the application (but of mostly wealthy students).

But the question here is whether the use of the SAT can help UC fulfill its mission of finding and educating capable students from all parts of California, and across all income groups. And I believe the evidence here is that yes, it does.

Backing up a bit, I feel many critics of the SAT (not you) are worried that colleges are simply going to rack and stack scores, thereby comparing the test score of a student from Palo Alto with parents at Google and Apple to a much lower income school in LA.

But that’s not what admissions committees do. Rather they look at scores in context. One of the advantages of a national test is that you can look at and adjust for various contexts. They could look at it within the context of just the high school itself, in which case it becomes an additional data point comparing a student to other students within just that school. They could attempt to look at it in terms of income, so that lower income students are considered relative to their opportunities, regardless of which school they attend. Other contexts are also possible for the purpose of understanding results, but may no longer be legal for the purpose of admissions (such as race after SFFA). This is the real power of a national test that many miss.

I don’t think it’s that simple. The relevant question is not whether SAT in isolation is better at predicting freshmen grades than HSGPA in isolation. It is instead how the combination of numerous factors used in admission predicts success at UC, and how use of SAT in that combination impacts both success at UC and demographics at UC. “Success” includes more than just freshman GPA. For example, I’d expect graduation rate to be more important to UC admins than freshman GPA.

Example stats for Berkeley are below. Looking at just the stats below, one might conclude that following the switch to test blind there was better URM and SES diversity, with little change in the listed measures of student performance. While I realize that there are other stats that show a decline, particularly at UCSD; my point is it’s a complex issue with advantages and disadvantages that warrants a review of actual outcome changes and related causes. For example, why is there apparently such a difference in math preparedness at UCSD and UCB, if both are test blind? Might it relate to things like UCSD dramatically increasing portion of students from of class from low income LCFF+ high schools, while UCB slightly decreasing portion, rather than just being about whether SAT/ACT is required?

Freshmen Retention (2019 → 2024): 96% → 97%
4-Year Graduation Rate (2015 → 2019 → 2021): 79% → 81% → 81%
Average Cumulative GPA among 4-Year Graduates (2019 → 2021): 3.5 → 3.5
Mean Grade Calculus I (2019 → 2024): B → B+
% Failing Calculus I (2019 → 2024): 3.5% (F) + 2% (NP) → 2%(F) + 1% (NP)
% Freshmen Math Starting in Pre-Calculus (2019 → 2024): 8% → 7%


Freshmen URM % (2019 → 2024): 20% → 27%
Freshmen Pell % (2019 → 2024): 27% → 33%

1 Like

Right. And because each of the UCs have a goal/mandate to include students from as many CA HSs as they can (that whole access thing), I’m not sure much will change if tests are required. For example, there will be students with a 1300 from one of the 62% or so Title 1 CA HSs being accepted to UCB or UCLA etc over a student from an affluent district with a 1520. Not sure that has solved anything.

It might prevent some students with very low math scores getting accepted (unless they are top 9% and are guaranteed a UC acceptance) but again, the in-state pool is the pool.

I wouldn’t be surprised that we see average test scores quite a bit lower than pre-pandemic if a high proportion of CA students start testing again. Perhaps that would spur some K-12 changes.

Deleted

I agree, and I was getting to that. Here is what the report said about that.

Here is the graph showing how graduation rates and graduation GPA were related to SAT scores and HS GPA bands. And @MWolf, while there are no error bars here for the GPAs, I believe that this is mitigated by the earlier disclosure of the R^2 values, plus it would make the graphs unreadable.

1 Like

I know some may disagree, but that’s less of a worry for me. A student with a 1300 can likely succeed at UCLA and UCB. I am much more worried about a student who would score 1000 being accepted to those colleges.

There’s a separate question of whether every UC campus should have the same admission standards in theory or in practice, but that’s a question for a different day.

1 Like

Not a worry for me either, but some people are really going to be disturbed by that, especially those from strong high schools with high test scores. (Although this was happening to some degree pre-pandemic.)

I’m not sure if you mean all the UCs or just Berkeley or UCLA. Historically though plenty of students with an SAT of 1000 have attended and graduated from a UC.

And we have no idea what the average of the pool would look like if more students test.

For reference, pre-pandemic 2019 CA SAT mean was 1065 (534 EBRW, 531 M).

Merced’s Class of 2023’s (2019 HS grads) 25th %ile SAT was 990. https://drive.google.com/file/d/16qnnBhH_89DFdkb_CEKWDahUtn94YrlC/view

Berkeley’s Class of 2023 25th%ile was 1330 https://drive.google.com/file/d/1Ph6AsUPP23emZRTwwHr2THYkCrb0HC-X/view

Regarding figure 3A-2

For any given high school GPA, a student admitted with a low SAT score is between two and five times more likely to drop out after one year, and up to three times less likely to complete their degree compared to a student with a high score

Refresh my memory if they controlled for family income. If not, of course the students with lower scores had lower retention rates…because data show many were likely from relatively lower income families. Many students from these families have difficulty paying for college even with federal and state assistance.

For sure. And if different standards by UC were attempted, I would expect to see Title VI lawsuits. :woman_shrugging:

One of the common critiques against the SAT or ACT is that scores are correlated with income, implying that means it is useless. But as I explained in a previous post, college admissions offices know how to adjust for that, making these tests very useful by identifying the student that overachieves relative to his or her peers. It’s no coincidence that MIT and the UC Academic Senate came to the same conclusion that these tests help identify talented low income or underrepresented students for their institutions.

But a related question should be, are there are enough low income students who get the scores that indicate they are likely to succeed at challenging colleges. And for this, I will ironically use a graph created by John Boeckenstedt, who is fiercely anti-testing, but doesn’t seem to accept that scores are used in context. Here is his graph showing how ACT test score are distributed by income, along with me showing the students from families making less than $60K who scored 28 or higher on the ACT (equivalent to 1300+ on the SAT).

My estimate is that students coming from families making $60K or less make up at least 15% of all students that score 28 or higher on the SAT. To me this means there is an ample pool for even the most selective UCs to find lower income students that could thrive on their campuses.

Again, some people will howl about what this will mean in terms of admission rates for different pools (higher income families will be admitted at lower rates), but given the existing UC Mission, they can fulfill that while using standardized tests.

3 Likes

Maybe, but we don’t know how many of those students are even planning to go to college. I don’t disagree that there are some low income students that would benefit from a relatively strong test score…but they have to want to go to college too, and be able to pay for it.

This is part of the reason why Jon B asks ‘is the juice worth the squeeze?’

Big picture, if the Regents do decide to require tests again, their biggest hurdle (besides potential lawsuits?) will likely be how to test an additional 200K or so students in their junior year, knowing that the pre-pandemic test administration structure has been completely dismantled in some areas of CA (and not unscathed anywhere).

I agree they have reason to be upset, but to me that’s a policy decision to be determined by the voters of CA through their elected officials, subject only to what is legal at the state and federal level. My point here is to show that UC can better implement its current policy by adding standardized tests to their admissions process, and by doing so, have a greater percentage of students that would succeed at that college.

I meant the most selective colleges like Berkeley and UCLA. I believe there is real value in having colleges at different levels of rigor.

This was shown in a later set of graphs, where they subdivided the results according to family income. The same patterns remain, although for lower income families both the GPAs and graduation rates are lower compared to the overall pool.

1 Like

Rigor mostly does not vary as much as admission selectivity. Open admission community colleges offer enough rigor that students transfer to UCs, Stanford, etc. without needing to retake all of the courses they completed at community college.

They often do not have SAT scores, but do have something more predictive of future college success than anything most high school students have: college courses and grades.

2 Likes

Yet if we look at actual freshmen retention or graduation rates there has been no significant change since switching from test required to test blind. Examples stats are below for CA resident freshmen across the full UC system . Unfortunately longer than 4-year graduation rate is not yet available for test blind class. Mean college GPA at graduation also seem to show little change among the few UCs that list public information.

Freshmen Retention: Entering 2019 → Entering 2024: 94% → 94%
4-Year Graduation: Entering 2019 → Entering 2021: 74% → 74%

Two relevant factors are the SFFA report indicates that SAT in isolation + HSGPA in isolation only explains ~2% of variance in retention and ~8% of variance in 4-year graduation rate (table 9). And UC admission considers more than just these 2 factors in isolation, in admission decision.

What has changed is a large increase in portion of students at UCSD (more so than other UCs) scoring at low levels on the math placement test. There was not a corresponding decline in freshmen retention or graduation rate at UCSD, suggesting that this group is generally not failing out of the college. However, I expect many of this group are switching majors to something that does not require calculus. The report indicated that by far the most common initial majors among this group were biology and psychology BS (psychology BS requires calculus, psychology BA does not).

2 Likes