I found an interesting document on the assessment of Oberlin College graduates. From the school’s wikipedia page:
Oberlin College is a private liberal arts college in Oberlin, Ohio. It was founded in 1833 by Presbyterian ministers, and is home to theOberlin Conservatory of Music, making it the only top-ranked liberal arts college (#20 according to US News & World Report[2]) with a top-ranked conservatory. The school is noteworthy for its early admission of African-Americans (1834)[3] and women (1833)[4] into the academy. A study found that more 1999-2003 Oberlin College alumni receive doctorates than do alumni from any other liberal arts college in the country.[5]
If Oberlin College graduates earn more doctorates than alumni from any other liberal arts college (including the very best), then these graduates had to take the GRE’s. So how did they do? Here’s some numbers from the assessment:
Number Verbal Percentile Quantitative Percentile Analytical Percentile
1673 615 80.3% 633 64.1% 642 71.2%
But even more interesting is the racial breakdown:
Verbal Percentile Quantitative Percentile Analytical Percentile
African American 500 54.5% 524 40.8% 519 43.0%
Asian American 570 72.3% 646 66.8% 624 66.9%
Latino 540 64.1% 585 53.7% 565 52.5%
White/Other 624 83.5% 649 67.5% 666 76.3%
Foreign 544 63.4% 671 71.7% 598 62.8%
I’m not surprised that the Asian-Americans didn’t score as well as the Whites in verbal ability, but I’m surprised that they didn’t score better than the Whites in quantitative ability. The assessment notes the following:
The first thing to note in these tables is that science majors are over-represented in the population of test-takers. In other words, they are more likely to take the GRE than students majoring in other divisions. We also note that the biggest difference in scores and percentiles across major are in the quantitative and analytical sections with science majors out-performing other majors rather significantly. The differences in the verbal section are much more modest.
Here is one last nugget of information:
Differences across ethnicity are striking and similar to the differences we see in SAT scores. Indeed, GRE scores are highly correlated to SAT scores, much more so then they are to Oberlin GPA.
Correlations Verbal Quantitative Analytical Subject
SAT Verbal 0.65 0.31 0.41 .33
SAT Math 0.42 0.63 0.54 .36
OC GPA 0.35 0.23 0.27 .33
The Undiscovered Jew
3 March 2009
LW, thanks for the updated information!
Lover of Wisdom
3 March 2009
TUJ:
No problem—unfortunately, this information only concerns one particular liberal arts college. I still need to find information on the total GRE test-taking population.
Ryan
23 February 2010
But aren’t these GRE scores much lower than the avg SAT scores at Oberlin (particularly the verbal scores)? This is exactly opposite what we would expect to see if the tests if the scores had fairly similar ceilings. This might indicate that the GRE’s have a much higher iq ceiling than the SATs (not only a slightly higher one as we have postulated).
Ryan
26 February 2010
http://oberlin.edu/instres/irhome/
This strikes me as evidence for the proposition that the Verbal SAT iq ceiling is around 142. We see a 1999 average verbal sat of 681 for the Obies. In 2003 an average GRE of 615. Either a lot of drugs were taken between these dates by the grad school hopefuls or the SAT Verbal doesn’t select for high intelligence very well at all.
Ryan
7 March 2010
Maybe I am inferring too much, but over the years the verbal GRE averages have been consistently over 50 points lower than the verbal SAT averages at Oberlin. This is not just ‘noise’. It looks like the GRE is significantly more difficult than the SAT.
Ryan
18 March 2010
Any thoughts LOW?
Lover of Wisdom
18 March 2010
You’ve presented a good case. I can’t comment on it now, but give me a day or two. I do find your intuitions are right.
Ryan
19 March 2010
We might also note that the 75th percentile of LSAT scores at HLS is 176 while the 75th percentile SAT verbal is at HC is 800. I think it would make perfect sense for educational testing aimed at selecting a group that has already undergone a harsh selection process would have a higher ceiling.
Thanks LOW, I appreciate your response.
Also, if philosophy grad students at NYU are verbalists at the level at which say math PhD students at MIT are numerists… then we shouldn’t be surprised that their around 750 average puts them over the top 1 percent of Harvard verbalists.
No one would be surprised to learn that the math kids in a PhD program at MIT or Princeton are on average above the top 1 percent of Harvard kids in terms of quantitative ability. We shouldn’t be surprised by this in philosophy.
Ryan
19 March 2010
that was really poorly written… i swear I got a 6 on the essay section!
Ryan
19 March 2010
This should have serious implications for high iq societies that use the SAT in admissions for reasons of expediency and accessibility with the purpose of increasing their membership.
I would be inclined to say that it is the conventional SAT IQ scale used by HIQ societies (the scale that equates a 1520 or so with an iq at the 99.9 percentile) that needs to be re-tooled rather than your GRE scale. It isn’t that the GRE is selecting for stratospheric intelligence (with a ceiling at say 160) but that the SAT isnt selecting for very high intelligence at all. I received an 800 verbal on both tests but my IQ scores on the WAIS have been in line with your GRE IQ conversion scale.
Anecdotally, there were plenty of kids in my high school who were in the 1520 range (about 15 kids in a grade of 50) but I was still told that I had the highest verbal iq scores in the school (this comment was based on my being tested at 146 on the verbal portion of the WAIS as a HS freshman – this was not even my highest WAIS verbal score which was a 151 gotten at the end of my freshman yr of college.). This experience always made me skeptical of claims that the post 1995 SAT was particularly g-loaded at the high ends. Still, I don’t doubt that it selects reasonably well below the 99.5 or that it sets a hard ceiling which prevents those under the 90th percentile from getting above a 1200 or 1250 or whatever.
Ryan
19 March 2010
These considerations cohere well with a set of other facts: 1) College work generally requires lower (g) than grad school work -all else being equal regarding school caliber) 2) College admissions officers are less interested in (g) than grad school admissions officers because of both the lower (g) need for college work and their focus on creating a ‘full’ and ‘vibrant’ student body. 3) There is much more politically and economically at stake in undergrad admissions than grad admissions -thus it will be important not to have a test that really reveals the true range of abilities because it would be harder to justify taking certain kids for economic or political reasons over others if their full differences were revealed. If James got a 780 verbal and Yoni got an 800 verbals James’s dad could be a big donor it is easier to pick James over Yoni than if the SAT went out to 1000 and Yoni got a 995. Obviously this is connected to affirmative action too.
We should worry that there may be efforts (or that there have been efforts) to lower the g ceiling on the LSAT because law school is similar to college in both its relatively modest (g) requirements and the political and economic implications of its admissions policies.
Ryan
19 March 2010
Also, given the economic and social benefits of going to a good school there are many people who have an interest in preventing the SAT from distinguishing well above the 99.5 percentile these days. The upper middle class would like to be treated as a clump consisting of people in the 99 percentile as far as college admissions goes. Testing with a 142 ceiling allows them to distinguish themselves from their lower middle class peers intellectually while still preventing the brainiacs from outclassing them and stealing the spots which they were busy ‘earning’ via resume padding ECs and demonstrating their sincere interest in the denizens of East Timor.
Ryan
19 March 2010
LOW, obviously I don’t expect you to offer your thoughts on all of this -and certainly not any time soon. Still, I think they might be of interest to some.
Ryan
19 March 2010
It is the great clump of families with children in the 99th percentile of IQ but in the 99.99 percentile of busy-bodiness,ambition and resources available to devote to educational attainment that have the most interest in keeping the SAT ceiling low. They are the great beneficiaries of the SATs imprecision at the high end.
Lover of Wisdom
19 March 2010
You were busy today, Ryan. I’ll respond when I’ve digested your comments some more. I’ll post something soon.
Lover of Wisdom
25 March 2010
I promise I’ll respond soon, Ryan. I’ve been hunting down some data to judge against your thoughts.
Ryan
25 March 2010
Thanks LOW, I appreciate it.
Not to complicate things further but…
In the mean time I came across this post on the ‘new GRE’. http://www.gnxp.com/blog/2007/04/new-gre-cancelled-cost-of-attempted-gap.php
“Agnostic’ made points similar to mine, but brought in the gender angle which I definitely had not thought of.
I looked around and the ETS seems not to publish data on the gender distribution of really high end GRE scorers (though they do publish that data for the SAT (!).) I remember reading somewhere that for pre 1995 SAT 80 percent of the scorers of 750 or above on the verbal were men (way to go mom!), 91 percent over 780 and 99 percent of the 800s. I will try to find the source.
If we could find out the gender distribution of high GRE verbal scorers we might at least have some support for its having a considerably higher ceiling than the post 1995 SAT.
I think this issue is more sensitive for women because most could care less about their math or spatial abilities, but tell them that their ability for real verbal thinking breaks down at the high ends and their will be hell to pay. You can also cover up the verbal difference more easily by lowering the (g) loading of verbal tasks via reading comp as ‘agnostic’ explains.
Ryan
25 March 2010
Agnostic seems to think that a 145 iq is de riguer for high level grad school though… which i’m not so sure about at all
-though it’s not clear whether he thinks the GRE had a 145 ceiling or was trying to help top grad school’s select a crop with an avg of this score. He seems not to get clear on this during the article.
Ryan
25 March 2010
If a M>F gender gap at the high ends of the verbal scale exists then it would explain the relative lack of women in top philosophy programs (perennially bemoaned by the likes of brian leiter and puzzling to those of us weaned on the assumption that females enjoyed a verbal advantage).
Andy Shin of the Oriental Right
9 November 2010
Oberlin is an excellent liberal arts institution, but it is not well-known among Asian-Americans, who prefer big-name schools. I believe Oberlin favors Asian-Americans over whites in the admissions process, which would possibly lead to the results that you are seeing. In addition to the lower quality of the Asian applicant pool, you have to look at cross-admits and yield. Asians are more likely to choose a institution with greater prestige if admitted, giving Oberlin a lower-quality group of matriculating students. Most Asians are not interested in the liberal/SWPL/love crap espoused by a school like Oberlin. In sum, the intellectual ability of Asians attending Oberlin is not likely to be comparable to a similar school in terms of academic quality that attracts lots of Asians.
Oberlin has an excellent music education program. One of the best musicians I have ever encountered in my life went to Oberlin…to become a Cello teacher. Asian She was extremely gifted at composing and was adept at multiple instruments. Quite amazing.
OTOH, the other Asians I know that attend/attended Oberlin were not very smart, at all.
YMMV.
lanolac
23 December 2010
In response to Ryan’s comment about the discrepancy between Obie GRE’s and SAT’s (and sorry if this is redundant): the SAT is normed at 500 for the typical high school graduate, while the GRE targets the average college grad. The difference of 50 points would imply that there’s .5 sd difference (about 6-8 iq points, depending on the measure?) between the average high school grad and the average college grad. Seems reasonable to me.