<p>
Well, it’s more than that. The NYT chart was never terribly accurate to begin with, and the study indicates why. Why read a 50 page study when you can look at a picture, though?</p>
<p>The Revealed Preferences study did not predict the results of one-on-one match-ups. Rather, it attempted to rank colleges based on desirability. Likewise, the RP study is not cross-admit data, though it does use it.</p>
<p>What the RP study attempts to provide is the comparative desirability of colleges based on the probable outcomes of hypothetical cross-admit battles as simulated by a model, the parameters of which are set by student info and in which factors like geography, legacy status, and financial aid have been equalized. The study freely admits that many colleges without significant overlapping applicant pools have been compared via intermediate colleges, and self-selecting colleges like BYU demonstrate a decidedly skewed result in the overall ranking. Essentially, the RP study is at best loosely correlated with actual cross-admit results. For example, in pitting Berkeley against Princeton in the study, a hypothetical student would be transformed into both (or neither) a resident of California and (nor) New Jersey, and financial aid/cost would be equal – obviously this is not actually the case. </p>
<p>The study cannot be trusted for reliable information about cross-admits. For example, the study indicates an 88% confidence that Yale wins against Stanford, and yet it has lost the majority of its cross-admits to Stanford since at least 2000. Even more egregious, it has Caltech beating out MIT, and yet we know that MIT collects more than 3 Caltech cross-admits for every 1 it loses to Caltech. You’ll notice that Caltech is conspicuously absent from the NYT chart, as it is a clear anomaly that would’ve tipped readers off that something was not quite right. </p>
<p>The RP study, unlike the NYT, did not attempt to pit colleges against one another directly (i.e. predict the outcomes of battles), and in fact the study outlined a number of problems about using its preference study as a substitute/source for/of actual cross-admit results:
[ul][<em>]Students ranked their preferences before they were admitted to college; the preference ranking fails to factor in how results affect the rankings of students.
[</em>]Students investigate colleges more extensively during senior year, and their preferences may change.
[li]Cross-admit results are based on actual matriculation results rather than students’ wishes.[/ul]</p>[/li]
<p>There is no doubt that Duke loses the majority of applicants also admitted to Harvard, Stanford, Yale, Princeton, and/or MIT. Exactly what that percentage is, however, is unknown. Of course, that will not stop me from speculating. I will use Stanford, since it’s the school with the most data available. We know two useful things about Stanford.</p>
<p>(1) 14% of Stanford admits were also admitted at Duke.</p>
<p>[Stanford</a> hopes to close financial-aid deficit in four to five years, Hennessy tells Stanford faculty | Stanford Daily](<a href=“http://www.stanforddaily.com/2010/10/08/financial-aid-deficit-may-persist-four-to-five-years-hennessy-tells-stanford-faculty/]Stanford”>Stanford hopes to close financial-aid deficit in four to five years, Hennessy tells Stanford faculty)</p>
<p>(2) 1% of Stanford’s lost admits matriculated at Duke.</p>
<p>[Faculty</a> Senate minutes - June 12, 2008 meeting](<a href=“You’ve requested a page that no longer exists | Stanford News”>You’ve requested a page that no longer exists | Stanford News)</p>
<p>Stanford admits: 2439
Duke-Stanford cross-admits: 341
Stanford losses: 717
Stanford losses to Duke: 7</p>
<p>Duke wins 2% of Duke-Stanford cross-admits.</p>
<p>For comparison purposes, here is how HYPM compare:
[ul][<em>]Harvard wins 55% (229) of 415 Harvard-Stanford cross-admits.
[</em>]Yale wins 31% (115) of 366 Yale-Stanford cross-admits.
[<em>]MIT wins 29% (93) of 317 MIT-Stanford cross-admits.
[</em>]Princeton wins 23% (100) of 439 Princeton-Stanford cross-admits.[/ul]
Unfortunately, that is where things get tricky. Although Duke wins only 2% of students also admitted at Stanford, that is with other colleges also in the mix. Most students do not apply to only two colleges, particularly such selective ones, and these percentages will shift if you take out colleges - for example, Yale’s success against Stanford may well rise if you take Harvard out of the mix, since the students looking to stay on the east coast may instead flock to Yale. </p>
<p>The only thing that we can really glean from this is (1) Stanford is increasingly gaining equal footing with Harvard and (2) YPM lose most of their cross-admits to Stanford, although many fewer than any other top universities. Based on the relative success of each against Stanford, I would guess the outcome of a one-on-one battle involving Duke would lead to a curb-stomping victory by HYPSM (90-95% won, 85% at the very least) and a relatively equal (40-60%) win rate for Duke against the “lower Ivies.”</p>
<p>Of course, what this has to do with Duke vs. Rice, I’m not quite sure. Rice fares no better than Duke against HYPSM, and it probably fares more poorly than Duke against the “lower Ivies.”</p>