GRE subject test math... Is it useful?

[tm_member]

i have taken the test. i would say it depends much on your math ability. if you are good at math (i.e your mc is small), then do it. good score helps with admission of course. i took the test last fall and scored 92 perc. i wouldn't say it is easy for everyone. but... i only prepared for two weeks for the test, and i had done only three math courses: math'l analysis, linear algebra, and prob.

On the math subject test?

Anyways, even that score will not help for admissions. The process is 'relative', and standardized tests are desiged for that purpose. If no-one else has done the exam, it's not possible to compare your score to others.

[mindlessme]

With all the respect, how can you be 100% sure that GRE Subject Test has never helped anyone in the admissions process? Have you verified this with all the admissions committees from schools of various levels of competitiveness and selectivity? Is it absolutely not possible that all other things equal, adcoms will be more assured about a candidate's math ability because they've done well on the Math test? For example, if two candidates with not many math courses are on the table with all other credentials similar.
Especially if an assistant prof who's reviewing this particular application is familiar with the subject test (which is not at all unlikely).

A claim that it usually doesn't help much and thus is probably not worth the effort required is one thing. Claiming that "even that score will not help" is another thing. You simply can't know that with 100% certainty.

[OneArmedEcon]

imgur: the simple image sharer


Also, I never said that the GRE Math subject test has never helped a previous applicant, only that it is unlikely to help the OP and it's not inconceivable that it will hurt his application if the adcom misinterprets the score.

[tm_member]

With all the respect, how can you be 100% sure that GRE Subject Test has never helped anyone in the admissions process? Have you verified this with all the admissions committees from schools of various levels of competitiveness and selectivity? Is it absolutely not possible that all other things equal, adcoms will be more assured about a candidate's math ability because they've done well on the Math test? For example, if two candidates with not many math courses are on the table with all other credentials similar.
Especially if an assistant prof who's reviewing this particular application is familiar with the subject test (which is not at all unlikely).

A claim that it usually doesn't help much and thus is probably not worth the effort required is one thing. Claiming that "even that score will not help" is another thing. You simply can't know that with 100% certainty.

We seem to be going around and around on this topic and of course no one argument can cover every possible situation but for the vast majority of students at the vast majority of institutions and in the vast majority of situations the test is a waste of time and money. My rationale follows below...

Under the (I believe valid) assumption that admission to a program is based on relative merit (i.e. there is no absolute measure, you simply have to be in the top X of this year's candidate pool) the neccessary conditions for the math subject test to affect an application positively are that:

1. Adcoms are aware of what constitutes a 'good' score on the test (not at all clear that they would and what is a good score could even be interpreted badly)
2. The adcom believes that doing well on the test correlates with success in an Econ PhD program (reasonable assumption)
3. The adcom can infer from the exam something that they cannot infer from LORs or your transcript (not sure how this would work)

I spoke to a prior DGS at Pitt about this before and was informed that the adcom would be uncomfortable declining admission to a student who submitted the required documents based upon another student submitting scores for a test that was not required of all students. This is where the 'all else equal' fallacy comes into play... Suppose there is one open spot remaining, suppose there are also two candidates who are otherwise equal but one student has submitted a 'good' Math Subject Test score, what should an adcom do? Admitting the student who does well on the math test is punishing the one who did not submit a score as it was not required, and is extremely unfair - adcoms are humans and will feel they have done wrong. Moreover, what if the score is viewed as poor? Who should they admit in that case? It's easier to just form a rule of thumb to ignore the score.

The example above just gets at the point that adcoms cannot formally evaluate scores from the subject test. At the margin, it is not fair to consider a score on the test to break a tie, in either direction. For those not at the margin, the test should not be able to convey more information than the rest of the parts that make up an application packet.

Therefore spending time and money on the test is likely to be a waste of time. Now there is always going to be an exception to that rule, an adcom member somewhere who is a PhD in Math and a student's subject test score catches his eye despite an otherwise poor profile and he fights for you - but the conditions necessary for such a thing to predictably happen are extreme.

At the end of the day, if you have good math grades but want to differentiate yourself from the pack, the best way to do it is through LORs. Getting an A in Calc III is common, having a well published professor say you scored in the top 3 on the midterm and final exams and had one of the the top %ages in the class for the past five years will do more for you than any Math subject test.

[mindlessme]

That's a bit weird and possibly false logic...
What if Real Analysis is not listed as a "requirement" but an applicant has it on a transcript? Can't adcoms use this as a criterion to prefer this applicant over others? Sure they can.
What if an adcom member personally knows the recommender and thus trusts/weights their letter more than an equally strong letter for someone else from another professor? Is this unfair to admit such an applicant? It is definitely not a requirement...
There are very few strict rules in the admissions process. I think they actually end after the GRE/GPA cutoffs for auto rejections. After that point the thing that matters is your overall profile. GRE Subject Math is one small part of your profile, just like anything else. And completely disregarding it because "it is not required" doesn't happen. It can be a personal point of view of individual professors, but on the large scale this is not how the whole process works. Most of the things people do for signalling are not required.

I think we all agree here that GRE Math is not at all likely to be a deal breaker.
However if applicants want to take it and they think they can do well having spent little time on it then they have the full right to go ahead. All the people reading this forum should remind themselves as often as possible that noone on this forum knows what exactly will be passing through the minds of adcoms members when they'll be reading your file.

[tm_member]

That's a bit weird and possibly false logic...
What if Real Analysis is not listed as a "requirement" but an applicant has it on a transcript? Can't adcoms use this as a criterion to prefer this applicant over others? Sure they can.
What if an adcom member personally knows the recommender and thus trusts/weights their letter more than an equally strong letter for someone else from another professor? Is this unfair to admit such an applicant? It is definitely not a requirement...
There are very few strict rules in the admissions process. I think they actually end after the GRE/GPA cutoffs for auto rejections. After that point the thing that matters is your overall profile. GRE Subject Math is one small part of your profile, just like anything else. And completely disregarding it because "it is not required" doesn't happen. It can be a personal point of view of individual professors, but on the large scale this is not how the whole process works. Most of the things people do for signalling are not required.

I think we all agree here that GRE Math is not at all likely to be a deal breaker.
However if applicants want to take it and they think they can do well having spent little time on it then they have the full right to go ahead. All the people reading this forum should remind themselves as often as possible that noone on this forum knows what exactly will be passing through the minds of adcoms members when they'll be reading your file.

True story, but we'll have to agree to disagree on the subject test. I just don't see it as a similar aspect of a profile to a real analysis class that many many students have on their transcript or an letter of recommendation from a trusted professor.

[OneArmedEcon]

All right, let's briefly analyze this problem in terms of a simple application of economics of uncertainty...

Even if we make a monotonicity assumption about the effect that the signal of a GRE Math subject has on an applicant's expected admissions results (i.e., it can only help or be neutral), it's also seems obvious to me that the additional data introduces another element of variance into the process. Assuming applicants are risk averse and all error terms of the various signals are orthogonal to one another, then the question becomes whether this increased variance is adequately compensated for by an increase in expected results.

Since doing well on the GRE Math is almost certainly highly correlated with other aspects of the application, my argument is simply that when you condition on those aspects, there is little marginal increase from a strong score. On the other hand, there is introduction of variance from the additional signal. In other words, there is an increase cost (in terms of new variance) without adequate compensation in terms of expected value.

To abstract away for a moment to something simpler, which would you prefer: an investment that pays 5% with a standard deviation of 1% or an investment that pays 5% with a standard deviation of 2%? Assuming that you're risk averse, obviously the first investment. In principle, it's the same idea with constructing the portfolio that is your application with noisy signals. You're not getting sufficient compensation for the increased variation.

[eXcuvator]

So, to sum up: if you're a genius in mathematics and are able to get a "good" score without having had the classes (and therefore no other reference to mathematics in your CV), you can put it in. In the other case, the probability of adcoms using that information is low, the risk is high and your CV would signal the strength anyways.

[tm_member]

So, to sum up: if you're a genius in mathematics and are able to get a "good" score without having had the classes (and therefore no other reference to mathematics in your CV), you can put it in. In the other case, the probability of adcoms using that information is low, the risk is high and your CV would signal the strength anyways.

Yup, and in the first case, while it may help, it is not likely to help very much. As in, if you don't have the math classes, you can't just pass GO and collect $200 by taking the test.

[openwest]

i realize this is a controversial issue. i also struggled for some time before register for the test. my post was only intended to share some personal experience. my strategy was simple: solve all the problems on linear algebra and analysis (mostly just calculus) correctly and forgo those on abstract algebra and topology. the proprotion of last two topics is really limited. i remember i left 7 or 8 problems blank. in the two weeks before the test, i read "cracking gre math subject", all the chapters expect abstract algebra and topology. one last thing, i think guessing is not a good idea in subject test.