New study about undergrad quality's effect on grad school





Click here to go to the NEW College Discussion Forum

Discus: Parents Forum: 2004 Archive - Part 2: New study about undergrad quality's effect on grad school
By Anthony (Anthony) on Tuesday, August 17, 2004 - 10:34 pm: Edit

A new study has been released studying the effect of undergraduate institutional quality/prestige on the quality of graduate school attended.

Controlling for undergraduate GPA, undergraduate major, race, gender, and various other demographic characteristics, undergraduate institution attended has a significant effect on graduate school enrollment.

All else held equal, graduates of high quality undergraduate institutions are 50% more likely to enroll in high quality research universities.

http://www.ilr.cornell.edu/cheri/wp/cheri_wp51.pdf

By Marlgirl (Marlgirl) on Tuesday, August 17, 2004 - 11:36 pm: Edit

Maybe it's just that really bright students are 50% more likely to go to high quality undergraduate institutions? I think it's more an issue of who would want to go to the high quality research university to begin with than where they went to college. That said, I do think that a college that doesn't offer enough classes in what you want to major in could put you at a big disadvantage. If many students do graduate work and you take fewer classes than the average college student in your field, you would be underprepared so getting into top programs would be more difficult.

By Anthony (Anthony) on Wednesday, August 18, 2004 - 05:06 am: Edit

Nope, there is no self selection involved here -- the regression controls for academic factors such as GPA and major, as well as demographic factors like income, race, gender, and age.

In other words, a white 25 year old male whose family makes $30k a year who majored in sociology at Cornell and got a 4.0 GPA is 50% (or whatever) more likely to enroll in a high quality research university for graduate school than a white 25 year old male whose family makes $30k a year who majored in sociology at Michigan State and got a 4.0 GPA.

Not to mention that quite a few graduate school adcomms have outright stated that undergraduate institution matters (though for whatever reason some people here have outright dismissed such statements, or have claimed that they only apply to a few select programs).

By Backhandgrip (Backhandgrip) on Wednesday, August 18, 2004 - 07:54 am: Edit

Maybe the Michigan State grad got a job.

By Emeraldkity4 (Emeraldkity4) on Wednesday, August 18, 2004 - 09:44 am: Edit

Some schools are frankly prep schools for grad school. Its good to know what you want at the outset, but if you take a few years off before grad school I think that is to your advantage. Of course some people never want to leave the "bubble" but keep acculmulating degrees.
http://web.reed.edu/ir/phd.html

By Northstarmom (Northstarmom) on Wednesday, August 18, 2004 - 11:39 am: Edit

"n other words, a white 25 year old male whose family makes $30k a year who majored in sociology at Cornell and got a 4.0 GPA is 50% (or whatever) more likely to enroll in a high quality research university for graduate school than a white 25 year old male whose family makes $30k a year who majored in sociology at Michigan State and got a 4.0 GPA. "

Of course, and in my opinion, it's self selection. The Ivy student who ends up at the high quality grad school may have ended up at an Ivy because even when in high school, the student aspired to going to a highly ranked school.

Equally qualified students who didn't want an Ivy challenge may have not even bothered to have applied to highly ranked schools.

When it comes to grad school, the students may treat the process the same way they addressed applying to undergrad institutions. For the Iviy student, prestige may be most important. The other student may value staying near home, being in a comfortable environment, being the top student in their graduate program more.

Just think about the personality and preference differences between the students with high stats who make a site like this their home as they try to go Ivy and the students with similar stats who happily apply only to State U, where they know they will be automatically admitted.

By Xiggi (Xiggi) on Wednesday, August 18, 2004 - 02:26 pm: Edit

Yawn!

I am not surprised at the results and conclusions of the research paper. However, despite introducing every control element feasible, the "study" cannot account for the most important part: the individual merit.

As I said before, you do not need to compare the the 4.0 students at Michigan State to the Cornell 4.00 students. Considering the relative difficulty of the grading at Cornell, an equally gifted student might earn a 4.00 at MSU while tolling in the middle or bottom of the pack at Cornell.

Lastly, I have no idea how this study -or any other for that matter- could account for the vast differences in the individual quality of applications, especially in essays and work experience.

What would be interesting is to see a study on the differences in success among the Ivies, but I doubt that such a study would ever see the daylight at Cornell!

By Massdad (Massdad) on Wednesday, August 18, 2004 - 02:59 pm: Edit

Anthony, did you actually read the article? It does NOT say "All else held equal, graduates of high quality undergraduate institutions are 50% more likely to enroll in high quality research universities. "

Look at table 5. It looks at what influences enrollment in comprehensive, doctoral or research universities.

By Dmd77 (Dmd77) on Wednesday, August 18, 2004 - 08:19 pm: Edit

I read this article. Frankly, it looked to me like the statistics were kind of nebulous--but I don't understand stats all that well.... The standard deviations looked very large relative to the values.

Could someone explain to me what the t-ratio of a set of data is? I can't seem to find a coherent link explaining its use in the social sciences.

I was mainly concerned that the study was on a very limited data set (1993 graduates) and did not define what how it decided what tier a school was in (at least that I could find).

By Anthony (Anthony) on Wednesday, August 18, 2004 - 11:43 pm: Edit

Northstarmom:

"When it comes to grad school, the students may treat the process the same way they addressed applying to undergrad institutions. For the Iviy student, prestige may be most important. The other student may value staying near home, being in a comfortable environment, being the top student in their graduate program more."

Very good point... However, considering that several adcomms have outright stated that they give preference to certain undergrad institutions, I doubt this sort of self selection is the only cause.

Xiggi:

"As I said before, you do not need to compare the the 4.0 students at Michigan State to the Cornell 4.00 students. Considering the relative difficulty of the grading at Cornell, an equally gifted student might earn a 4.00 at MSU while tolling in the middle or bottom of the pack at Cornell."

Do you have anything to back that claim up besides your own speculation?

"What would be interesting is to see a study on the differences in success among the Ivies, but I doubt that such a study would ever see the daylight at Cornell!"

It probably would, if top grad schools were willing to release that sort of information. It would be particularly easy to do for law school admissions, since that process is very heavily numbers-based... However it's impossible to well unless several top law schools are willing to provide access to their admissions data to a researcher.

By Familyguy24 (Familyguy24) on Wednesday, August 18, 2004 - 11:52 pm: Edit

I don't really understand this study. It seemed that overall they said that if you attend a better undergraduate school, over like a public college that isn't so great, your more likely to attend graduate schools, and do better. Well isn't that obvious? The research doesn't prove that going to a top private school is better than going to a public schoool, it just says that if you attend a top private school, you are more likely to attend graduate school, which I think is obvious given the kind of student that are enrolled in those schools. They are the type of students who would seek out higher education, and will probably do better. The study doesn't say, that if those students were put in the public school instead of the top private school, they would not do as well, which would mean that the top private school is better. The debate is whether or not a capable student will do just as well in a private school as opposed to a public. Not whether people from the top schools do better than those from the not so top, because that is just obvious. On the whole, people from the top schools are probably more learned and have more drive to suceed then tthose from the not so top. What they should have compared is the acceptance rates of people applying to the same graduate school, but from different universities (top and bottom). If you could find people with similar high school statistics, but one having gone to a lesser school, and then them both applying to the same graduate school, or same caliber grad schools, then This would tell you if one allowed for a better opportunity than the other. Maybe I missed something though, read it very quickly.

By Dstark (Dstark) on Wednesday, August 18, 2004 - 11:54 pm: Edit

Didn't Princeton do a study and found that private school kids do better at Princeton the first year, then public school kids catch up?

By Originaloog (Originaloog) on Thursday, August 19, 2004 - 12:30 am: Edit

Xiggi-I like and agree with your comment regarding individual merit. However you then make an uninformed comment comparing a student with a 4.0 gpa at MSU with one at Cornell, stating that the MSU student would be an average student at Cornell.

As I have mentioned before, I am a graduate of both OSU and Cornell and currently teach engineering at the college level. In my case, it was easier to get an A at Cornell and my gpa refleted that fact. I found the graduate level course work at Cornell to be comparable to the senior level course work at OSU. What I did find as a TA at Cornell for several years was that perhaps 25-35% of the undergrad students at Cornell would receive an A in an intro engineering class while at OSU perhaps 15-25% would receive an A. The difference in the %age was largely due to the calibre of student at Cornell. A student with a 3.5 gpa at OSU would probably have a comparable gpa at Cornell. And whereas perhaps 25% of the engineering class at OSU would have a gpa of 2.5 or less, a far lower %age of Cornellians would have this low a gpa, again beause of the general calibre of the students.

I have found this to be true of grad students in our department. Unless you know where a student has done his undergraduate work, its almost impossible to pick out the Penn grad from the Purdue grad.

By Mini (Mini) on Thursday, August 19, 2004 - 02:14 am: Edit

"Very good point... However, considering that several adcomms have outright stated that they give preference to certain undergrad institutions, I doubt this sort of self selection is the only cause."

Excuse me? One adcom said (except he wasn't an adcom) at one NON-graduate school (it was a professional school).

The question is always whether there is value-added to equation by the school itself. We actually have information (though not a study) about issues of this kind, in the NSF data on grad admissions. Even without correcting for SAT scores, there are several schools -- notably Hope, Kalamazaoo, St. Olaf's, and Earlham -- where students go on to earn a largely disproportionate share of Ph.D.s; correcting for entrance SAT scores or "selectivity", the numbers wouldn't even be close (these and several others have by far the greatest "value added" to graduate school admissions.) Of course, among decent colleges, high graduate school admission can also be a sign of the relative lack of income of students: poorer students are less likely to take on additional debt in even applying to high-flying professional schools in law, medicine, or business.

Xiggi's speculation would be more interesting if there wasn't such great grade inflation at top schools (Ivies, LACs, etc.).

By Xiggi (Xiggi) on Thursday, August 19, 2004 - 11:06 am: Edit

"As I said before, you do not need to compare the the 4.0 students at Michigan State to the Cornell 4.00 students. Considering the relative difficulty of the grading at Cornell, an equally gifted student might earn a 4.00 at MSU while tolling in the middle or bottom of the pack at Cornell."

1. In the past, we have seen seen several posts on grade inflation at Ivies. Most reports concerning Cornell indicated that the school was known for DEFLATING grades in the undergraduate program and did not follow the same practices than, for instance, Harvard. This issue came up for the engineering and pre-med programs. Just use google and define your search on "CC+Cornell grading" and you'll see dozens of pages.

2. Grading at the graduate school level is not really relevant to this discussion.

3. I listed MSU to remain consistent with Anthony's example.

By Massdad (Massdad) on Thursday, August 19, 2004 - 01:42 pm: Edit

This study is an interesting one, but lay folks have a great risk of over-interpretation. Familyguy24 had it right when he said " The study doesn't say, that if those students were put in the public school instead of the top private school, they would not do as well, which would mean that the top private school is better. "

To help you readers understand what the paper REALLY says, I'll go through the tables one by one. Background: in each line of a table, the author gives two numbers. the first, the coefficient, can be interpreted like this. 0.2000 means someone with that characteristic is 20% more likely to "do" the outcome of that table (model) than the average member of the dataset. Negative numbers mean less. The second number is the t-ratio, which is a measure of significance of the coefficient. If the t-ratio is above about 2, that means there is about a 2% chance the coefficient is meaningless. More to the point, lower t-ratios indicate coefficients that are probably really 0.

Consider also that low coefficients, whether or not statistically significant, are in a practical sense meaningless. Why worry about something that only influences an outcome by a few percent?

So, on to the tables.

Table 1 and 2 are descriptive, so we'll skip.

Table 3 looks at how the variables predict graduate school enrolment. Note that only 4 have much predictive power. High quality public or private increases odds by about 17%, not a huge bump up, but among the biggest in the model. GPA has the biggest positive impact, 22% more. No surprise here. Business majors are 22% LESS likely to go to grad school. Chasing the $$?

Table 4 must be looked at carefully. It looks at the probability of enrolling in a PhD program OVER a MA program. All datapoints were students who went to grad school, so it is a subset. Curiously, none of the coefficients are very big here. The largest + is being a bio or math major, with 11% greater odds. High quality public (but not private), are only 7% more likely. Bus majors are 7% less likely. Big deal here - no surprises.

Table 5 is the most interesting, but again, do not overinterpret. Like 4, it looks at those who actually enrolled in grad programs. It is unclear why the sample size for table 5 is smaller than 4. Perhaps there was a good deal of missing data. Anyway, this model tries to predict attendance at a comprehensive, doctoral or research university, using the Carnegie classifications. The Carnegie never intended that the classifications be used as "quality" indicators, like the auther may have done here, and even changed the classifications recently to obfuscate this use.

At any rate, Table 5 contains some of the widely quoted and misundestood results of the study, such as the 50% increased probability of attending a research U if you went to a high quality public or private U. But it also showed that middle quality publics were 27& more likely, and engineering, math and bio majors 25-34% more likely. So what?

Table 6 looks at what affects completion of a degree. No surprises here, nothing seems to make much difference. But keep in mind previous results. This only looks at those that got in.

Table 7 is similar to table 5, but looks at completion. The results have a lot in common with Table 5.

So, let's look at interpretation. None of this work says anything about institutions. The work is focused on those that attend - the students. Given selection effects, is it any surprise at all that a kid that choses (or is chosen) to attend a highly selective college will be more likely to pursue, and complete, a graduate degree? Is it any surprise that a kid that choses to attend an easy school may have lower aspirations that his more motivated colleagues? No. To me, the real surprise is that none of these school differences matter more than they do.

To me, these results are consistent with a view that American higher education allows an INDIVIDUAL student great opportunity to change direction and move to widely varying options. It really says the OPPOSITE of what the OP thought!

By Candi1657 (Candi1657) on Thursday, August 19, 2004 - 02:06 pm: Edit

Thanks, Massdad, for helping to parse the data for average folk like me that have a difficult time interpreting statistical models.:)

By Dmd77 (Dmd77) on Thursday, August 19, 2004 - 02:49 pm: Edit

Ditto--thanks Massdad for the careful look at the tables. Can you please explain to me how the t-ratio is calculated? I've run into it before (as a measure of data reliability) and don't understand it. Thanks.

By Massdad (Massdad) on Thursday, August 19, 2004 - 03:59 pm: Edit

DMD77,

Technically, the T-ratio is calculated dividing the difference in means by the standard error of measurement. In other words, if you have measurements with a lot of variability, you need bigger differences between to groups to say the difference is significant. The T statistic quantifies this.

By Dmd77 (Dmd77) on Thursday, August 19, 2004 - 04:08 pm: Edit

Massdad--okay, I get the terms--but:

I assume by "the difference in means" you mean the difference between the means of the two groups being compared?

So if we have a group of students with a mean grade of 3.0 and another group with a mean grade of 3.5, with a measurement error of 0.1, would the t-ratio be 0.5?

Thanks for your continuing patience.

By Massdad (Massdad) on Thursday, August 19, 2004 - 04:46 pm: Edit

DMD77, right track, but calculation error: the T ratio would be 5.

For just about any sample size (technically degrees of freedom) this would be a significant result.

By Pafather (Pafather) on Thursday, August 19, 2004 - 05:37 pm: Edit

I completely agree with MassDad.

"Is it any surprise that a kid that chooses to attend an easy school may have lower aspirations than his more motivated colleagues? No."

I was an undergrad in engineering at University of Florida. When I graduated (20 years ago), I knew very few of my friends who intended to go to grad school, and only one or two who were applying to top level schools. I went to grad school at MIT, and was a graduate resident (like an RA) in an undergrad dorm there for 3 1/2 years. I believe that the top undergraduate students at Florida were academically comparable to the top students at MIT (in terms of how well they had mastered fundamentals of engineering courses). However, the ATTITUDES at MIT were very different than Florida. Whereas few of my friends at Florida were considering grad school and fewer still were considering highly competitive grad schools, the undergrads on my floor at MIT applied to Cornell and Michigan as backup, safety schools. At the premier universities, expectations are much higher. However, a well-motivated undergrad from a lesser college can definitely get an excellent education and excel at a top-flight grad school.

By Anthony (Anthony) on Thursday, August 19, 2004 - 06:39 pm: Edit

Mini:

"Excuse me? One adcom said (except he wasn't an adcom) at one NON-graduate school (it was a professional school)."

That was just ONE example... Many other examples have been posted on the primary board that I post on whose link is not allowed to be posted here.

Xiggi:

"In the past, we have seen seen several posts on grade inflation at Ivies. Most reports concerning Cornell indicated that the school was known for DEFLATING grades in the undergraduate program and did not follow the same practices than, for instance, Harvard. This issue came up for the engineering and pre-med programs. Just use google and define your search on "CC+Cornell grading" and you'll see dozens of pages."

Sorry but that's a crock of bull... I graduated from Cornell and it's certainly as grade-inflated as other Ivies (with the possible exception of Brown, due to Brown's very liberal pass/fail and add/drop policies). The only reason Cornell appears grade deflated is that it has a very large engineering program which drags the average GPA down.

Don't believe me? Check out the median grade reports for yourself:

http://www.sws.cornell.edu/our/student/mediangrade.html

Massdad:
You have a plausible argument, however the researcher (who I emailed) does not believe that there is much self selection at work here.

By Mini (Mini) on Thursday, August 19, 2004 - 06:45 pm: Edit

"Excuse me? One adcom said (except he wasn't an adcom) at one NON-graduate school (it was a professional school)."

That was just ONE example... Many other examples have been posted on the primary board that I post on whose link is not allowed to be posted here."

Except it wasn't even an example. It wasn't an adcom, and it wasn't about graduate schools. All it was an example of was a failure on your part to correctly characterize what you read.

By Massdad (Massdad) on Thursday, August 19, 2004 - 07:41 pm: Edit

Anthony, so the author doesn't think self selection is going on? Sorry. That doesn't count for much, as someone (I'm not sure who) is trying tom make a data driven statistical argument here.

What about the author's misuse of the Carnegie classifications? What about lumping together all kinds of graduate education? Remember that a research university (even Harvard) has terminal MA programs, professional degrees as well as Ph.D. degrees. Lumping all these together is a bit "rough" .

What about using GPA as a common yardstick?

Keep in mind that this is typical social science research. The experimental design, to the degree one exists, was driven by the data available, not by the ideas (hypotheses) to be explored. As a result, it is impossible to control for confounding variables (other than "the author thinks"). This in turn raises a strong risk of overinterpretation of the results.

So, I don't think anyone here would disagree with the hypothesis that grads of differing kinds of higher ed institutions will pursue different career paths. This hypothesis is supported by the analysis. The analysis does NOT support any conclusion, however, that it is due to institutional differences. That would call for a different kind of study.

By Anthony (Anthony) on Thursday, August 19, 2004 - 08:48 pm: Edit

Mini:

"Except it wasn't even an example. It wasn't an adcom,"

If you have the ability to accept or reject applicants, you're an adcomm.

"and it wasn't about graduate schools."

Law school is graduate school. If you want to sub-label law school, business school, etc. as "professional school," then go ahead, but they're still graduate schools.

"All it was an example of was a failure on your part to correctly characterize what you read."

Nope, all I see here is a difference of opinion regarding the terms "adcomm" and "graduate school" -- you're obviously using a much narrower definition than I am.

Massdad: well stated.


Report an offensive message on this page    E-mail this page to a friend
Posting is currently disabled in this topic. Contact your discussion moderator for more information.

Administrator's Control Panel -- Board Moderators Only
Administer Page