I agree with what you are saying, besides “reduce.”
And there is also some truth that in the more affluent towns the schools are taking it easy on the kids in an effort to keep their mental health in check. And the debate continues…
2 Likes
The following College Board research about discrepent (HS GPA and SAT score) applicants:
found that SAT discrepent applicants tended to be from higher income parents, while HS GPA discrepent applicants tended to be from lower income parents.
Even if HS GPA is something that advantages high SES (due to, for example, more grade inflation in higher SES schools), it has less of this effect than SAT scores.
1 Like
Thanks for posting that paper, I have seen many references to it.
Leaving aside the correlation of income with discrepacy, it is interesting that girls are sigificantly overrepresented in the group with high GPA but low SAT (that is, SAT lower than one standard deviation below that of GPA), in fact by a factor of more than 2 to 1.
It would seem that test-optional may therefore favor girls.
Also notable is one of the conclusions:
“… these findings suggest that students with a high HSGPA in the presence of low SAT scores will not do any better in college than students with lower HSGPA scores but higher SAT scores. Therefore, the SAT may be a more accurate predictor than HSGPA for these students.”
Perhaps the SAT has greater value for SAT discrepant students in college admissions than for the applicant pool as a whole.
I’m not sure about the relevance of this paper. First of all, it’s about whether HS GPA or SAT is relatively more predictable of college GPA, not whether one or the other is more of an impediment to college admissions. Secondly, the study is mostly about the effect of gender and race, not SES (of course, there’s a correlation). Last, but certainly not least, the high school discrepant group (the group with their HS GPAs significantly “better” than their SAT scores) were more likely to take easier and remedial classes than those in the SAT discrepant group (the group with their SAT scores significantly “better” than their HS GPAs) who were more likely to take harder courses “in mathematics, science, and arts, as well as advanced courses”. So their colleges GPAs weren’t even comparable.
2 Likes
Penn is extending test-optional through the 2021-2022 admissions cycle. The change is effective immediately for all transfer and first-year applicants.
Revised Testing Policy for the 2021-2022 Admissions Cycle
Pre-pandemic, I couldn’t imagine them being TO. Now, I believe it may last at least through the class of 2026.
3 Likes
Erm, I don’t get your reasoning, even with your explanation downthread.
You seem to be saying that every metric favors high-SES students, but that seems rather odd of a claim to me.
I’m using the hypotheticals to show that the oft-quoted studies on the effect of SAT do not demonstrate that it’s a bigger impediment to college admissions by lower SES students than any other admissions criterion, such as gpa or essays.
In our town they instituted a pass/fail system but parents complained so they added a pass plus metric. It seems that every student was getting a pass which left no objective way to determine the C students from the A students. In New York, they previously tried to institute a teacher testing system but so many teachers were found to be incompetent in the subject they taught that the unions complained and the testing was scrapped, leaving no objective way to determine competency. I’ll bet a good portion of those failing teachers received great reviews from the parents and students, especially if they were given top grades and letters of recommendation. If all schools and all teachers taught and graded the same way for every subject we would not be having this debate since GPA alone would be sufficient. Because that isn’t the case, the combination of GPA and some form of testing should be still used IMO, but I’m now having doubts that they will return once the pandemic subsides.
Yes, the paper suggests that test optional would most favor lower income, URM, females. While GPA optional would most favor White males with >$100k income.
Test optional favoring women (on average) is nothing new or surprising. Women average higher grades and graduation rates than men at every level from elementary school to college, yet women do not average higher (combined/composite) scores on average than men; so test optional tends to favor women.
So it should come as no surprise that women tend to be overrepresented among test optional admits and underepresented among test submitter admits.
For example, the previously linked Bates 25 years of test optional study found that non-submitters at Bates were 60% female, and test submitters were 48% female.
1 Like
It’s almost like you didn’t read the paper. The paper is titled “Students with Discrepant High School GPA and SAT® I Scores.” It lists groups that had a “discrepant” SAT with their HS GPA , meaning that their SAT was more than 1 SD away from expectations based on HS GPA. This discrepant grouping has nothing to do with college GPA. The first 3 categories that are analyzed are gender, race, and parents income. Parent’s income is absolutely a direct correlation with SES.
The paper found that kids with lower SAT than expected based on HS GPA were most overrepresented among female, URM, and <$40k income. It’s not a coincidence that this group is also overrpepresented among test optional admits. The paper found that kids with higher SAT than expected based on HS GPA were most overrepresented among male, white, and >$100k income. It’s not a coincidence that this group is also overrepresented among test submitter admits, at test optional colleges.
So? And you didn’t quote the last part of my comment. These types of studies are as useless as they come.
If you don’t like studies, you can look at the actual distribution of test optional and test submitter admits at existing test optional colleges, which match the discussed study conclusions.
Regarding the last part of your comment, the “discrepant” GPA/SAT was based on HS GPA, not college GPA. I wouldn’t assume that high SAT kids are more likely take advanced HS classes than high HS GPA kids within a particular HS, without evidence. For example, at the nearest HS to my location, getting approved for an AP class generally requires a high grade in the previous class (and recommendation from teacher). They don’t consider test scores. That said, yes, the study did not control for course rigor. However, selective college admissions does consider course rigor and shows the same pattern, as stated above.
1 Like
This makes me wonder whether you read the paper. It was talking about college classes, not HS classes. Considering how different (in difficulty, etc.) these courses are, does it make sense for the study to even compare their respective GPAs in these college courses?
I’ll assume you 2 have finished debating. This whole thread has really devolves into a back-and-forth between 4 users.
3 Likes
It’s a subtle point, and the paper could be much clearer (see under “Method”), but I am 99% certian that this is not what was being studied.
Apparently, the authors simply rank ordered self-reported HSGPA across all particpants’ high schools and derived a percentile rank for any individual participant through direct comparison with that “standardized” HSGPA curve. Of course, no adjustment was made for course rigor or, importantly, for the overall rigor of the high school at which any individual GPA was earned.
That percentile rank was then compared against a truly standardized curve, that of SAT test results, which despite all the acknowledged flaws of the underlying test itself, at least enjoy the benefit of single scoring system across all schools.
As the United States is highly segregated along SES lines (among other characteristics), the study really seems pretty irrelevant for SES effects on discrepancy. As an example, a 25th percentile GPA at a very rigorous and wealthy suburban high school is likely to demonstrate more academic achievement and aptitude than a 75th percentile GPA at a poorly-resourced lower quality school. To then go ahead and draw conclusions based on raw percentile correlations across widely different school systems seems a worthless exercise. For the middle-SES tiers, perhaps. But comparing raw data between schools at which >$100k per year students predominate against schools with predominately, say, <$30k students seems silly to me.
Yes, I listed a 12-word summary rather than going in to a lot of detail about methodology. The study essentially compared percentile rank between HS GPA and SAT, in isolation. So it does not control for differences in rigor of HS courses or different rigor of particular high schools. This no doubt contributes to why lower SES kids tend to have lower SAT scores than one might expect by looking at their HS GPA in isolation, without considering any other factors. Lower SES kids are generally more likely to attend poorer quality HSs.
Looking at HS GPA and SAT in isolation without control for HS course rigor or other factors is the normal methodology for College Scorecard SAT validity studies, or for the vast majority of studies that find that SAT offers a noteworthy benefit in predicting measures of college performance. When you consider differences in course rigor and other admissions factors used by highly selective colleges, the incremental benefit of the SAT beyond the combination of GPA + course rigor + everything else is generally quite small.
Evidence, please?
Seriously, there have been a lot of claims about GPA and rigor thrown about on this forum, and I would suggest that there are a lot of underlying assumptions about class and geography and race that are lurking beneath those claims. Fortunately, this claim has thrown them into relief.
So, please, any actual evidence for the claim?
1 Like
There have been multiple references about correlations between SES and HS GPA or SAT. However, information about correlations between SES and other areas of the application are lacking, as you mention. One can gain some insight about correlations between SES and other areas of the application by looking at the most simple model from the Harvard lawsuit, which only controls for demographic variables and planned major. It does not control for stats, or reader ratings of applicants.
For example, among unhooked, White, male applicants, the model predicts that an applicant who is flagged as SES “disadvantaged” by readers, has a fee waiver, and applied for FA is predicted to have the following differences from an unhooked, White, male applicant who does not have any of these 3 indicators associated with lower SES. Results are expressed in terms of number SD differences. -1.0 means the lower SES group is expected to average a large 1 SD difference from the not lower SES group.
Predicted Lower SES Difference in Mean Harvard Applicant Ratings
Academic: -0.7 SDs
Alumni Interview: -0.1 SDs
ECs: +0.1 SDs
GC LOR: +0.2 SDs
Teacher LOR #1 & #2: +0.2 SDs & +0.25 SDs
Personal: +0.45 SDs
After controlling for gender and race (latter has notable relationship with several of above), the only application rating category where lower SES Harvard applicants seem to be significantly weak compared to the overall pool is the academic rating. When also controlling for AI stats, the estimated mean academic rating lower SES difference drops from -0.7 SDs to -0.1 SDs, suggesting that this apparent lower SES academic weakness is almost entirely correlated with having lower AI stats. AI is computed as 1/3 GPA and 2/3 based on test scores, and previously referenced studies suggest scores are far more correlated with SES than GPA. So it stands to reason that test scores are likely one of the key weak areas among typical lower SES Harvard applicants, perhaps the largest barrier to admission among lower SES applicants on average*.
The Harvard OIR internal analysis at http://samv91khoyt2i553a2t1s05i-wpengine.netdna-ssl.com/wp-content/uploads/2018/06/Doc-421-112-May-1-2013-Memorandum.pdf shows a graph with more specific information about the degree of test score differences by income. Displayed percentage of applicants with 1500+ SAT scores by income from the graph is below:
Percentage of Harvard Applicants with 1500+ SAT Score by Income
$20k income – 15%
$50k income – 25%
$100k Income – 35%
$200k Income – 45%
*My comments above mentioned test scores being perhaps the largest barrier among “applicants.” However, lower SES kids often face many other barriers far before applying. The vast majority of high achieving lower SES kids do not apply to any selective colleges (see https://www.nber.org/papers/w18586 ), including Harvard. If you have admission by purely random lottery without considering application, the lower SES kids will still be dramatically underrepresented among admits compared to the overall population because lower SES kids are dramatically underrepresented among applicants.
The Harvard internal study link above actually found that lower SES kids who applied had approximately the same admit rate as other SES groups, much like would occur in a random lottery. This partially relates to Harvard favoring lower SES kids enough to make the admit rate approximately the same. This preference does not seem to directly influence the ratings except perhaps for personal rating, as suggested when adding controls for rest of application. Instead Harvard admits lower SES kids at a higher rate than expected based on their ratings, hook status, and various other modeled factors. Specific numbers are below from the previously linked Harvard internal study:
Harvard Admit Rate by Income Reported on FA (from several years ago)
$0-$40k – 11% admit rate (would be 6% without lower SES preference)
$40-$80k – 11% admit rate (would be 8% without lower SES preference)
$80-$120k – 9% admit rate
$120k-$160k – 10% admit rate
$160k-$200k – 10% admit rate (would be 11% without lower SES preference)
$200k+ – 12% admit rate (would be 13% without lower SES preference)
1 Like
Not to me.
Every single metric that you have to - and are able to - do something about (I am this vague because I am trying to avoid loaded words like effort, or hard work, or achievement, because sometimes it is just about having money, or connections) is easier for higher SES kids and their families, on aggregate, than for lower SES kids, completely independently of their personalities and inborn abilities or whatever. It’s even easier to get a job, or to volunteer.
Of course there are metrics explicitly defined to favour one SES group over another (legacy vs first gen, etc) and other metrics students and their families cannot control (ethnicity, geography). All of which questioned, and questionable, IMO, but that’s of course an outsider view.