Read on Twitter
If GRE gave little information relevant to preparation for grad school, how could it have significant correlation with undergraduate grades? https://news.ets.org/stories/gre-test-validity-putting-it-in-perspective/
The best way to decide whether the GRE is relevant to hiring grad students is to analyze the types of tasks they will be expected to do, assess the GRE questions, then assess whether thinking in that way is helpful to the expected tasks.
Check out these quant exercises. The student needs to consider the info given, then intuit the correct response.
https://www.ets.org/gre/institutions/about/general/quantitative_reasoning_sample_questions/
The GRE quant is highly G loaded, since it assesses quant skills that require lower level training even than SAT.
https://www.ets.org/gre/institutions/about/general/quantitative_reasoning_sample_questions/
The GRE quant is highly G loaded, since it assesses quant skills that require lower level training even than SAT.
By including exercises that require relatively little academic training, but higher level thinking, the GRE quant is a good quality intelligence test. Not a perfect one, though, as it's possible to be sufficiently intelligent to succeed in grad school and do below median.
Intelligence alone is a huge benefit to success in grad school, but intelligent people who lack patient focus (for example), may not be as well suited for grad school as a student somewhat less intelligent but more conscientious.
This is one reason why definitive cutoff GRE scores should never have been used to eliminate student applications. All applications should have been reviewed in their entirety. Great applicants who show their metal in other ways should never have been excluded.
Low GRE often co occurs with other signals of poor performance (such as poor grades in quant courses, lack of self-driven quality research experience, etc.). But when other indicators indicate success, the applicant should be considered anyway.
It is better to have many moderately good predictors of student success.
People who don't like the GRE frequently refer to research that suggests it poorly relates to graduate student success. These people frequently cite only the subset of research that appears to agree with their conclusions, while ignoring other research.
This is a sign that their rejection of the GRE is either ideologically motivated, or that they simply didn't expend the effort to look around.
Yet, whether supportive or critical of the GRE, research based on cohort studies of students who were admitted partially on the basis of the GRE (or correlated signals like grades) have the problem of truncated or censored distributions. http://www.atmos.albany.edu/facstaff/roundy/GREschem.pdf
People are placing uncritical trust in selected flawed statistical studies that happen to agree with the answer they hope to see, in studies that can't effectively demonstrate the conclusions they seek.
How the distributions are truncated can lead to both false positives and false negatives, so people who use these cohort studies in support of using GRE (including ITS), can also be misleading.
The problem doesn't even depend on using GRE itself: Since GRE is correlated with grades, if people collected GRE data but didn't use it for decisions, the GRE distributions will still be censored by using grades to admit students.
The most effective way to assess GRE effectiveness in a cohort study is to admit students entirely at random, then compare GRE with student success. But what kind of graduate program will admit students this way?
Another complaint about GRE is the cost to students. I agree on this complaint: Low income students need to be able to justify free assessment. Yet a much larger problem looms for GRADES! How much do undergraduate grades cost?
Finally, there are some mean group differences in GRE scores. Wealthier students do better than low income students, on average. Yet undergraduate academic success is a better predictor of high GRE than student wealth...
...and many low income students do well. GRE/SAT type standardized tests are harder for students to game than other components of applications. Students can get a lot of help from friends or hired tutors for grades. Coaching is much less effective on the GRE test.
Personal statements can be revised or written by others. Letters of recommendation are subject to subjective personal opinion.
It would be quite unlikely that low income students would on average do similarly well to high income students, even if there were zero benefit to paid coaching/training.
That's because higher income students tend to have greater intellectual stimulation, come from better school districts, & take more advanced courses, over many years of education.
If those things don't make students more academically qualified, what things would? If the GRE at all measures academic qualification, we would expect it to associate with things that mark great longterm preparation.
Since some groups of people have greater proportions of low incomes, it is therefore not surprising that some groups have lower GRE outcomes. Many people who oppose the GRE think that these differences mark cultural differences in the GRE.
Naturally, language IS a barrier to GRE success, but test writers have been removing other cultural barriers from the GRE for decades.
These opponents of the GRE are conflating poorer academic preparation (often associated with low income) with cultural differences.
Instead of fighting a relatively good, but moderately flawed indicator of student success on the basis of group differences in scores, people should instead show curiosity about why scores average lower for low income students...
...then work to solve the disparities in the education system, & other impediments to low income student success.
