New article on MSN.com-"Is the Ivy League Worth It?"

<p>Sakky - I complain about Gourman, I think that its a graduate ranking pretending to be undergrad. Same with SJT. I think both are grad rankings.</p>

<p>However, RP is based off of hypothetical model based off of a small sample of students - in contrast, US News and similar rankings take into account real data. I think the data from the RP is sometimes thought to be actual data - and I dislike that. </p>

<p>Is a hypothetical model, RP doesn't stand a test of reason - if 2/3 of students choose Penn over Duke, why are the students at Duke as strong (or slightly stronger in some categories)? Same with Columbia Dartmouth and Brown. Of course, it being half/half makes sense since the students are similar. </p>

<p>I don't use US News either. Its most valuable characteristic is that it gathers a great deal of data. I just use holistic numbers such as SAT scores, professional feeder rates, National Merit Scholars, etc. when I ever do a side-by-side comparison of how strong students at a college are.</p>

<p>
[quote]
However, RP is based off of hypothetical model based off of a small sample of students - in contrast, US News and similar rankings take into account real data. I think the data from the RP is sometimes thought to be actual data - and I dislike that.

[/quote]
</p>

<p>Come on now, I think you're romanticizing the value of USNews. You say that it takes into account "real data". But the question is, what "real data", and does that "real data" really translate into what the rankings say they translate to? For example, while the faculty resources data of USNews is, I agree, "real data", where is it written that a final ranking of the "Best Colleges" has to be derived from a methodology that weighs faculty resources by 20%? What is the theoretical justification for that? Like I said before, at least with the RP, at least a theoretical justification for the methodology does exist. Yes, the sample size is small (although not distressingly so - as they also break down the data by region and obtain only minor differences). Yes, we can quibble about the model. But it's still better than whatever else is available. At the very least, we should give credit to RP for doing that. </p>

<p>
[quote]
Is a hypothetical model, RP doesn't stand a test of reason - if 2/3 of students choose Penn over Duke, why are the students at Duke as strong (or slightly stronger in some categories)? Same with Columbia Dartmouth and Brown. Of course, it being half/half makes sense since the students are similar.

[/quote]
</p>

<p>Uh, what do you mean it doesn't stand a test of reason? The RP model measures * preferences *. Nothing more, nothing less. It doesn't measure anything about student strength, nor does it pretend to. It only measures * preferences *.</p>

<p>Let me give you a rival example. The overall strongest average student body of any school in the country is almost certainly at Caltech. Caltech has a stronger student body than Harvard does. Yet does that mean that Caltech is preferred to Harvard? Hardly so. I think that everybody - even a Caltech fanatic like Ben Golub - will agree that the vast majority of people will prefer Harvard to Caltech, even though Caltech's student body is ostensibly stronger. The point is, just because a school has an equivalent or better student body to another school doesn't mean that the former is truly "more preferred".</p>

<p>Sakky,</p>

<p>I'm saying the US News uses real data. It uses actual SAT scores, actual alumni giving rates, actual resources, class sizes, etc.</p>

<p>Its still real information...which, IMO, has an edge over hypothetical info.</p>

<p>listen, in the north east, Penn is simply a more desirable school for various reasons. This doesn't mean it has a stronger student body. But from my experiences, people who choose between Penn and Duke usually pick Penn.</p>

<p>Similarly, in California, a lot of people choose Stanford over Harvard. This does not mean Stanford has a better student body. In fact, Harvard has a stronger student body. However, does this mean that Stanford is not as good as Harvard. Absolutely not.</p>

<p>"Revealed Preference":</p>

<p>Lets say they their data showed some Mormons admitted to BYU and Northwestern, and picking BYU.</p>

<p>And then maybe they observe a bunch of non-Mormons not applying to BYU at all, admitted to Northwestern and U Chicago, and maybe a couple more of them picked Northwestern.</p>

<p>So based on these results the study would (and in fact did) produce a "Revealed Preference ranking of:</p>

<p>21 BYU
23 Northwestern
28 U Chicago</p>

<p>But is this really the relative preference for these schools among the applicant pool as a whole? In reality it's probably not even the preference order of even the very students that were surveyed. The people in that second bunch, which is a larger number of individuals, did not even think enough of BYU to even apply to it, and if they had been admitted to it they would likely have preferred both Northwestern and U Chicago to BYU.</p>

<p>If the opinions of the whole applicant pool, or even just those surveyed, had been properly captured their real relative preference for these schools overall might have, dare I say almost certainly would have, shown BYU least preferred of these three. But this wasn't "revealed" because the last pool didn't apply to BYU, hence didn't get to reject it. They rejected it at an earlier stage in the process.</p>

<p>The method would be extrapolable if each sample of applicants were unbiased when compared to the underlying population of all applicants; ie if it were a random sample. But in fact each sample of applicants is itself a biased sample, not an unbiased sample representative of the underlying population, and that's where this procedure fails.</p>

<p>And fail it does- unless you believe people really prefer BYU to Northwestern and chicago. As I indicated above I don't beleive even the students who were surveyed hold this view. Do you???</p>

<p>I just picked BYU as, to me anyway, an obvious example. But each other case may also be biased in some other, less obvious way.</p>

<p>The "accepted methodologies with a theoretical justification" may exist in other social science contexts probably where each sample could be represented to be unbiased. Not the case here.</p>

<p>The obvious anomolies merely allude to the less obvious anomolies that are undoubtedly also manfested in these results. Go BYU.</p>

<p>
[quote]
I'm saying the US News uses real data. It uses actual SAT scores, actual alumni giving rates, actual resources, class sizes, etc.

[/quote]
</p>

<p>And like I said, the whole point of the RP study is to deal with gaps in the "real" data, particularly in the preference data. After all, you can't use "real" data about choices that never actually existed. For example, the only "real" data that exists regarding preferences between Duke and School X is from those students who applied to and were admitted to both schools. What about those students who only applied to one or none of those schools? What about those students who did apply to both, but didn't get into both? Obviously no "real" preference data can exist. That's why you have to infer from modeled data. </p>

<p>This is akin to a marketing study. You can't realistically offer every customer every possible product choice, because there are just too many product combinations out there. So you have to infer through modeling what you think customers would prefer if they were given certain choices. Yes, the models are not completely accurate. Yes, customer behavior is somewhat unpredictable. But you have to do * something * because the real data does not exist. It's a copout to just throw your hands up and say that you don't know just because the real data does not exist. For example, if you know that a customer enjoys gory horror movies, then it is quite likely that that same customer also probably enjoys gory video games. Not guaranteed of course. But probably. </p>

<p>The problem is simple. Any examination of just cross-admit data simply ignores the fact that you are, by definition, sampling on the dependent variable. Anybody who applies to your school has, by definition, some preference for your school (otherwise, why are they even applying?). Therefore, those students who apply to Duke are obviously a deeply skewed sample. As I'm sure you'd agree, there are plenty of other people out there who never even apply to Duke at all. Yet we have to find a way to count these people. By definition, no "real" cross-admit Duke data can exist for these people, yet these people must still somehow be included in a final aggregate preference ranking. {And yes, there are some people who apply to Duke but not School X, so we have to somehow include them also}. That's where the modeling comes into play.</p>

<p>know when I sent my reply card to Duke I had to write in 8 other schools I applied to and what the decision was.</p>

<p>This is what Duke is relying on. There is no way for Duke to systematically tell if these kids are telling the truth or if they are leaving out details.</p>

<p>For instance, person x may enroll at Duke. May have been wait listed at Cornell, Penn, and Brown.</p>

<p>On the sheet. That person may say, I am turning down Cornell, Penn, and Brown, as if they got accepted.</p>

<p>
[quote]
Lets say they their data showed some Mormons admitted to BYU and Northwestern, and picking BYU.</p>

<p>And then maybe they observe a bunch of non-Mormons not applying to BYU at all, admitted to Northwestern and U Chicago, and maybe a couple more of them picked Northwestern.</p>

<p>So based on these results the study would (and in fact did) produce a "Revealed Preference ranking of:</p>

<p>21 BYU
23 Northwestern
28 U Chicago
So based on these results the study would (and in fact did) produce a "Revealed Preference ranking of:</p>

<p>21 BYU
23 Northwestern
28 U Chicago</p>

<p>But is this really the relative preference for these schools among the applicant pool as a whole? In reality it's probably not even the preference order of even the very students that were surveyed. The people in that second bunch, which is a larger number of individuals, did not even think enough of BYU to even apply to it, and if they had been admitted to it they would preferred both Northwestern and U Chicago to BYU.</p>

<p>If the opinions of the whole applicant pool had been properly captured their real relative prefence for these schools overall might have shown BYU least preferred of these three. But this wasn't "revealed" because the last pool didn't apply to BYU, hence didn't get to reject it. They rejected it at an earlier stage in the process.</p>

<p>The method would be extrapolable if each sample of applicants were unbiased when compared to the underlying population of all applicants; ie if it were a random sample. But in fact each sample of applicants is itself a biased sample, not an unbiased sample representative of the underlying population, and that's where this procedure fails.</p>

<p>I just picked BYU as, to me anyway, an obvious example. But each other case may also be biased in some other, less obvious way.

[/quote]
</p>

<p>Uh, no, there it is again. Come on, once again, you have shown that you didn't even bother to read the study. Please read it. The WHOLE POINT of the RP study is PRECISELY to attempt to measure the preferences of people who don't even bother to apply to certain schools. That's the WHOLE POINT of the study. They MODELED what would happen if people were given certain theoretical choices. These aren't actual choices that they had in the real world, and that's precisely were the value-add of the study is. </p>

<p>In other words, monydad, your complaint is not a reason that detracts from the RP study. In fact, it is precisely the opposite - it is PRECISELY the reason that the RP study is so valuable. What you have done by your complaints is actually REINFORCED the strength of the study. </p>

<p>I've said it before to you, monydad, I'll say it again. * Please READ the study before you complain about it *. You will find that the study deals with EXACTLY what you are complaining about. In fact, that's the whole theoretical justification of the paper.</p>

<p>Look, the bottom line is this. Nobody is arguing that the RP is a perfect study. Of course it is not. But at least it has a theoretical underpinning, and an accompanying mainstream model to generate its results. * The other rankings out there don't even have that *. For those of you who incessantly complain about RP, I have one simple question for you - what's better? What's the alternative?</p>

<p>NUGrad:From my experience, it's the reverse! AND I live in the northeast!
Unless its Wharton..then they pick Penn.</p>

<p>As for that NYT study.. Duke, I will admit, wasn't having the best of years last year due to the lacrosse scandal.. but that survey still seems to be somewhat skewed and it's kinda dumb if people are actually using that to decide what college to go to!</p>

<p>"The WHOLE POINT of the RP study is PRECISELY to attempt to measure the preferences of people who don't even bother to apply to certain schools. That's the WHOLE POINT of the study."</p>

<p>Yeah and they did it wrong. It didn;t work, progbably fot the reason I gave above.</p>

<p>Here's the proof:</p>

<p>21 BYU
23 Northwestern
28 U Chicago</p>

<p>What's better? Anything; an opinion poll. This thing is trash.</p>

<p>
[quote]
21 BYU
23 Northwestern
28 U Chicago</p>

<p>But is this really the relative preference for these schools among the applicant pool as a whole?

[/quote]
</p>

<p>I've been griping about this elsewhere; I may as well start here, too.</p>

<p>That study was not meant to be used in this way. The authors were setting out to show their colleagues how revealed preference methodology (a conceptual framework that has been used in other areas), might be applied to college choice behavior. The authors grabbed a small sample, enough to show how such a methodology might be used. They did not intend to produce a usable ranking from their paper, so they didn't worry about sample size or representativeness. Why should they--it wasn't the point of their paper. They did show how things shook out in their sample, but that's all it was--an example, just an illustration. Who knows what they'd find if they used a larger sample? </p>

<p>The problem comes because people have been posting excerpts, and it's very tempting (and perhaps understandable) to see the results and start talking about them like they are a research finding. They're not. The study authors have paved the way for other authors to produce such a finding using other data, if they want, but the authors' point was to try out the methodology and walk other interested scholars through that process.</p>

<p>It's possible to get a counterintuitive result when you're using a small sample. It would be an interesting thing if they were trying to produce a real ranking--or a condemnation of their methodology, if they were making claims about what they've "found." But since they used just a small sample to demonstrate the methodology, it shouldn't be too surprising that some of the results are odd.</p>

<p>
[quote]
For those of you who incessantly complain about RP, I have one simple question for you - what's better? What's the alternative?

[/quote]
</p>

<p>I am one who points out flaws in the RP data input, but not sure it rises to the level of 'complaining," and it is certainly not "incessant" which I would define as a certain supporter of RP. :)</p>

<p>
[quote]
Yet we have to find a way to count these people.

[/quote]
</p>

<p>Why? If they don't appy to Duke and Penn, to use two examples, they don't apply to Duke and Penn. </p>

<p>But, IMO, who NEEDS an alternative? I find PR, Fiske, USNews, etc., fun reading. RP is also fun.</p>

<p>My one comment is that if more people are choosing Penn over Duke..
Duke must get more people with high SAT scores!
US News seems to confirm that.. :D</p>

<p>As I said before, I do not know what Duke's real SATs will be this year. As of last year in US News....Penn 1340-1520. Duke 1360-1540.</p>

<p>Penn's acceptance rate was lower and Penn had more in the top 10 percent of their class. Penn has a larger class to fill. Duke offers MERIT MONEY and Penn does not.</p>

<p>hoedown: my guess is in the other areas where this methodology has been more appropriately utilized, each "contest" data point could be more reasonably characterized as a random sample drawn from the underlying population, not a biased sample.</p>

<p>I don't believe this methodology can achieve completely reliable results even if larger sample sizes are utilized, because each cross-admit sample is a biased sample of the underlying population, not a random sample.</p>

<p>In my hypotherical above, as long as Mormons truly prefer BYU to Harvard et al, and relatively few people other than mormons will apply to BYU so their preferences will not be adequately "revealed", the same general ranking will be "revealed" even when a large sample is used.</p>

<p>NUGrad, </p>

<p>As usual, your personal experience in the Northeast means nothing - thats just one school district out of several. I already pointed out that in my area Penn admissions are easier than Duke (in Northeast). But my school district represents a tiny fraction of the entire world of applicants. </p>

<p>Refer again to my post (I think it was 48#) if you want a thorough side-by-side comparison between Penn and Duke's student body. I'm not sure how referring to statistics like SATs helps your argument about how Penn is superior to Duke...since all of those data favor Duke (slightly.) Again, I think the schools are completely equal in overall strength.</p>

<p>For the Class of 2010, Duke has about 30-40 points on SATs, around same 10%, more national merit scholars (proportionally and absolute numbers), better Wall Street Journal feeder rates into top Ivy league professional schools, near identical resources per student, etc.</p>

<p>Duke has 30 merit scholars enroll per year - that is 2% of the student body that receives tuition. When Duke's 25/75% for SAT's is about 30-40 points higher, I don't see why the 2% would make much of a difference. Either way, the students are still enrolling at Duke. I guess you need to win the discussion somehow even after seeing a thorough break down of the two schools proving most of your perceptions of Duke are off base.</p>

<p>thoughtprocess, you can think whatever you want. I am in the real world. I am a corporate lawyer. I have a pretty firm grasp on what schools carry weight in the north east. I didn't say Duke didn't carry its weight. However, I still argue that Penn is more prestigious in the northeast. If Prestige was here, he would back me up on that.</p>

<p>NUGrad - I guess since the facts don't match up, its again time to refer to personal experience.</p>

<p>I'm not sure how your position as a lawyer gives you a unique perspective on how Penn and Duke undergraduate schools stack up against one another. Do you and your coworkers regularly discuss which undergrads you value more highly? Apparently the view shared by your corporate lawyer friends is one that isn't uberly well-informed. Didn't you claim to go to Duke for law school - you aren't a very proud alumni.</p>

<p>After all, doesn't Duke send more students to top Ivy league law schools? And also, the THES corporate survey (survey of several global corporations) ranks Duke as a top 10 school to recruit from, whereas Penn is slightly lower. Not saying any of these things make Duke better, but that overall the schools are similar.</p>

<p>You: "At the end of the day, Penn is an ivy in the north-east whereas Duke is known more for its sports and scandals than anything else, as its yield shows. Sorry,that is the harsh reality."</p>

<p>Hm. Yeah, Duke has had an interesting sports year - lacrosse championship tomorrow, Women's Golf championship two days ago...and in terms of scandal, having all of its students declared innocent. But you are right about Penn - it is an Ivy. I guess that automatically makes it superior to Duke. </p>

<p>And yes, having a lower yield, but still getting (slightly) stronger students to enroll is definitely a bad thing.</p>

<p>thoughtprocess, Penn has a Wharton school, engineering, and nursing school.</p>

<p>If you compare Penn CAS to Duke CAS, you will find they are pretty equal in size.</p>

<p>Now, let's look at harvard law</p>

<p><a href="http://www.law.harvard.edu/admissions/jd/colleges.php%5B/url%5D"&gt;http://www.law.harvard.edu/admissions/jd/colleges.php&lt;/a&gt;&lt;/p>

<p>Duke University 41</p>

<p>University of Pennsylvania 57</p>

<p>You are the one that brought up law school numbers. Care to explain that discrepancy....</p>