<p>afan:</p>
<p>I tried. I could not get the NSF database to give a list of where all the graduates of one college got their PhDs. It would only let me analyze by "baccalaureate granting college" or by "PhD granting university", but not by both simultaneously. I didn't spend a lot of time, but I'm pretty familiar with how to drive the NSF database. I don't think that report can be generated.</p>
<p>
[quote]
Actually, the Wall Street Journal methodology was so flawed that it is basically useless.
[/quote]
</p>
<p>The WSJ methodology is better than quoting schoolwide PhD rates, because it compares like to like in the outcomes, i.e. it does not equate graduate admission to Lower Podunk with admission to Yale. The WSJ does not compare like to like in the students, though you seem consider that OK (I don't: the only way to really compare the effect of school is to consider the relative chances of a given outcome for a given type of student).</p>
<p>
[quote]
They looked at too few grad schools, with too much northeast bias, and made the critical mistake of looking at a single year's data.
[/quote]
</p>
<p>A northeast bias does not favor Duke over Swarthmore or explain why choosing a like-to-like measure of elite outcomes changes the relative productivity by a factor of 2.8 (i.e., reverses a 2.5 to 1 advantage for Swarthmore into a slight advantage for Duke). If you try some calculations you will see that it is also not easy for annual variability or addition of more schools to explain an effect of that magnitude.</p>
<p>
[quote]
"Duke adjusted for major and type of student or even with just the athletes subtracted, will not be 41st" </p>
<p>So, what you are saying is that, if Duke had a different student body, it would produce a higher percentage of PhDs?
[/quote]
</p>
<p>No, not at all. What I said is that the question of Harvard vs LAC is about the "treatment effect" of applying one school's education rather than another's (Swat versus Duke) to a given category of student. It is no achievement for Harvard to admit the best and then monopolize the elite PhD admissions. It could even be that attending Harvard reduces one's chances compared to attending Princeton or Stanford, but that information is submerged in Harvard's stronger pool. Similar observations apply to Swarthmore. </p>
<p>In Duke's case, its numbers suffer the undertow of athletics (for example). Few of those athletes would have made it into Swarthmore. Of course, Swarthmore has students who are not at the level for admission to Duke, but self-selection factors (e.g., scholarly inclinations without high test scores) would still keep the PhD numbers of such students relatively high, just not at the top schools. What the WSJ numbers suggest is that there is higher variance within the PhD outcomes at a given top 10 research school than at a small LAC, which after all is more homogeous in its PhD affinity due to the self selection. One signature of that would be the LAC's having high or higher overall PhD rate but a weaker showing in elite PhD production.</p>
<p>The northeast bias was in the grad schools. Let me give you an example. Compare the cost of med school at your home state medical school to Harvard Med School. For many top students, planning to actually practice medicine, the home state public med school is the affordable option.</p>
<p>If you want to know the question, "where do really rich kids from the northeast go to grad school", the WSJ is an OK methodology. And, big surprise, they come from undergrad schools with large enrollments of really rich kids from the northeast! The logic is almost as circular as the WSJ's editorial page.</p>
<p>If they expanded to, say, the top 20 med schools, the top 20 law schools, and the top 20 biz schools, had some geographic balance, and looked at at least ten years of data, the methodology would have been fine. But, the Wall Street Journal isn't a research project. They are trying to sell newspapers. </p>
<p>
[quote]
In Duke's case, its numbers suffer the undertow of athletics (for example).
[/quote]
</p>
<p>Bullspit. This sentence tells me that you have no idea what you are talking about. Compare the percentage of varsity athletes at ANY top LAC to Duke. The impact of athletic recruiting at a school the size of Duke is small, only 12% of the men and 9% of the women are on varsity teams. At Swarthmore, the percentage is 22% of the men and 19% of the women. It's higher at LACs with a football team.</p>
<p>Many, or at least more, of those athletes at Amherst, Swarthmore etc have a shot at a PhD, if not a PhD at MIT. I don't think that's nearly as true of Duke.</p>
<p>I in no way suggest (or suggested) that athletics per se is cutting Duke's numbers in half or is a dominant force in these comparisons. Athletics, to the extent that it exerts a noticeable difference on a school's numbers, and that difference is not the same across schools, is an example of how the demographics and not the schooling are what drives the PhD results.</p>
<p>
[quote]
The northeast bias was in the grad schools. Let me give you an example. Compare the cost of med school at your home state medical school to Harvard Med School. For many top students, planning to actually practice medicine, the home state public med school is the affordable option.
[/quote]
</p>
<p>How would this bias affect Swarthmore but not Duke, and how could this or the other biases come close to creating an almost 3-to-1 reversal of the statistics? Try some calculations and see whether flawed methodology explains much of anything here. </p>
<p>Also, why is the WSJ like-to-like comparison flawed to the point of being "useless" while the apples-to-oranges PhD rates are a supremely meaningful data point?</p>
<p>
[quote]
A northeast bias does not favor Duke over Swarthmore
[/quote]
</p>
<p>No, that's the relative lack of interest in MBAs by Swarthmore students compared to Duke students.</p>
<p>The best way to get the information you want is to go straight to each college's career services department.</p>
<p>For example, 23% of Swarthmore grads go straight to grad school from college (that percentage increases to 65% over time, but most college students take time off before grad school these days).</p>
<p>For the last three graduating classes, that means the 250 Swarthmore graduates went straight to grad school -- approximately 122 to PhD programs, 67 to MA/MS/MPP programs, 32 to law school, 27 to med school.</p>
<p>Of those 250 who went to grad school immediately upon graduation, here are the 16 most popular grad schools, accounting for 94 of the 250 (38%). You may be a lot snootier than I am, but I would consider most of these top grad schools, although few of them were used in the WSJ sample.</p>
<p>13 University of Pennsylvania
12 Columbia
11 Harvard
9 Johns Hopkins
8 University of California - Berkeley
7 University of Michigan
5 Duke
5 University of Washington
3 Emory
3 Stanford
3 University of Cambridge
3 University of Rochester
3 University of Wisconsin
3 Vanderbilt
3 Washington University in St. Louis
3 Yale</p>
<p>
[quote]
changes the relative productivity by a factor of 2.8 (i.e., reverses a 2.5 to 1 advantage...)
[/quote]
</p>
<p>You seriously need to think about research design. You can't "reverse" a trend using ALL of the PhD degrees with a small, partial sample of MDs, LDs, and MBAs from each of the two schools! That's like saying that you "offset" a national Gallup pole by asking a few people in your neighborhood what they think.</p>
<p>I agree with you that the ideal data set would be the addition of all MDs, JDs, and MBAs to the database of PhDs. Alas, the MD and JD information is not publicly available (i.e. nobody has tracked it like the NSF has tracked PhDs since 1920). MBA data is nearly impossible to track, even for the colleges, because it is so frequently a later-life degree.</p>
<p>
[quote]
Granted, it is possible that a school at the top of future PhD production is sending its students to low-quality PhD programs
[/quote]
</p>
<p>Is there such thing as a low-quality doctorate program? I would think the faculty, even at a third-tier college, is dedicated, intelligent and enthusiastic about research. The level of intellectualism will probably surpass anything one subsequently does in the real world.</p>
<p>
[quote]
No, that's the relative lack of interest in MBAs by Swarthmore students compared to Duke students.
[/quote]
</p>
<p>If relative lack of interest explains Swarthmore having a lower JD+MBA+MD rate than Duke, then you would agree it can also explain Harvard, Duke etc having lower overall PhD rates than Swarthmore, n'est ce pas?</p>
<p>
[quote]
If relative lack of interest explains Swarthmore having a lower JD+MBA+MD rate than Duke
[/quote]
</p>
<p>I don't know that's the case. Swarthmore has a pretty high JD and MD rate. Always has. 9% of all Swarthmore graduates over the last 11 years have been accepted to med school. I don't have data going that far back, but over the last three years, the law school rate as been a bit higher than the med school rate. So figure 10%. That's on top of the 20% or so getting PhDs.</p>
<p>We simply have no data to make comparsions between colleges for the JD and MD part.</p>
<p>
[quote]
then you would agree it can also explain Harvard, Duke etc having lower overall PhD rates than Swarthmore, n'est ce pas?
[/quote]
</p>
<p>Of course. What did you think it was? That's pretty obvious isn't it?</p>
<p>
[quote]
I don't know that's the case. Swarthmore has a pretty high JD and MD rate.
[/quote]
</p>
<p>"Rate" above meant "rate at the elite schools measured by the Wall Street Journal", obviously. You appear to agree that this figure is a measurement of the students more than the school. </p>
<p>What then is the point of the PhD numbers as a metric of the school (not the students), and whether it might cause advantage or disadvantage to a given student? Measuring the students is really a different question.</p>
<p>It's entirely possible that if you take into account the profile of the incoming classes and the specific outcomes achieved, Swarthmore does worse than Ivy Leagure schools. I expect that the LACs in fact do slightly better (slightly worse in some fields), but your data don't point in either direction.</p>
<p>"Is there such thing as a low-quality doctorate program?"</p>
<p>Sure. I judge grad schools primarily by the quality of students coming in and by what they achieve coming out. There are many undistinguished doctoral programs by that standard. Merely having good professors who know the research well is not enough in my book. If that's the metric, then there are a lot of community colleges that ought to be considered fabulous.</p>
<p>
[quote]
"changes the relative productivity by a factor of 2.8 (i.e., reverses a 2.5 to 1 advantage...) "</p>
<p>You seriously need to think about research design. You can't "reverse" a trend using ALL of the PhD degrees with a small, partial sample of MDs, LDs, and MBAs from each of the two schools! That's like saying that you "offset" a national Gallup pole by asking a few people in your neighborhood what they think.
[/quote]
</p>
<p>No. It's like saying that one national poll had twice as many women as men answering Yes to some question ("ban handguns in 2009?"), and another national poll with a slight modification of that question showed only half as many women as men answering Yes. Such a reversal means that there is more than one thing being measured by each question, and that it would be naive to tout the results of the first survey as being all that meaningful without some way to separate the effects of the different things.</p>
<p>In the case of the reversal of Swarthmore vs Duke numbers based on the question asked, what one might be seeing is that at Duke the top students are (as a group) stronger than the ones at Swarthmore, but also that there is a long way down from the top, whereas at Swarthmore it's more compressed.</p>
<p>
[quote]
What then is the point of the PhD numbers as a metric of the school (not the students), and whether it might cause advantage or disadvantage to a given student?
[/quote]
</p>
<p>None of this -- not PhD production, not MD production, not JD production -- has one iota of predictive value to an individual high school senior. As I have pointed out many times, a gung-ho pre-med high school senior doesn't even know if he or she will flunk Orgo.</p>
<p>College statistics, all of them, are descriptive. Not predictive. The value of any statistic is that it describes the school and its students. (You can't separate the two). Put Swarthmore students at Duke and it's no longer Duke -- and vice versa.</p>
<p>
[quote]
..whereas at Swarthmore it's more compressed.
[/quote]
</p>
<p>Again, of course. What else could it be? A top student is a top student. One reason that the better LACs do so well in grad school placements is that, as a group from top to bottom, the entire student body is more self-selected for academics and more closely engaged in the academic learning process. Its certainly not that top LAC students are smarter than top students at any other top school or that the dumb students at Duke are dumber than those at a top LAC. It's simply that the LAC students self-selected a more engaged learning environment, which probably pays the biggest dividends for the average students. The academic engagement extends deeper in to the class.</p>
<p>
[quote]
None of this -- not PhD production, not MD production, not JD production -- has one iota of predictive value to an individual high school senior.
[/quote]
</p>
<p>It has some predictivity for populations of similar high school students. "How do students like me tend to do" is important even if there is a chance of being weeded out in Orgo. No different from perusing the course catalogue to answer the next question, "how many other options does it offer if I fail Orgo"; the number of courses in Anthropology says something even if it's not the last word on the quality of those other options.</p>
<p>Your stated point was that the PhD lists show there is no disadvantage to LAC over Harvard et al. I say that, apart from specific infrastructure to support grad/med/law school applications or other very direct factors, there is little effect (or at least, little evidence of any effect) of most schools on anything and these numbers are almost entirely about the students. The PhD's do show that it is far from impossible to get in from LACs, but we knew that already.</p>
<p>The numbers from Swarthmore Career Services (are the full data public?) that you posted look like a confirmation of the Wall St Journal method. The number attending grad school at Duke (not in the sample) is comparable to the number at Michigan (included in the sample). We know from the overall PhD rates that if you enlarge the definition of "top schools" enough, Swarthmore will catch up to and exceed the production of Harvard. But the issue was whether the distribution of PhD destinations (and admission chances) is different, so this would defeat the purpose. WSJ seems to have hit on something if it causes an almost 3-to-1 shift in the Duke to Swarthmore numbers. Law and business school are a reasonable approximate proxy for the strength of the humanities subjects and economics, and med school a reasonable proxy for science. Not perfect, but reasonable, and not imperfect enough to dismiss when explaining the difference between 3 and 1.</p>
<p>
[quote]
WSJ seems to have hit on something if it causes an almost 3-to-1 shift in the Duke to Swarthmore numbers.
[/quote]
</p>
<p>That's like saying apple production offsets orange production. They are two different things. Oh, and lets look at all apple production for the year, but only look at oranges harvested in five counties in the country.</p>
<p>Again, data for pre-professional admissions (if it were available) doesn't have any relationship to PhD admissions. They are different.</p>
<p>You can for example have a schools that:</p>
<p>a) have high biz school and high PhD rates</p>
<p>b) have high biz school and low PhD rates</p>
<p>c) have low biz school and high PhD rates</p>
<p>d) have low biz school and low PhD rates</p>
<p>I hardly think that exanding the defintion of "top grad schools" beyond five in each group is unreasonable, especially when you limit the top five to grad schools where you need a trust fund to pay for it. Particularly when looking at only one year already leads to serious issues of "small numbers" and the resulting margin of error problems. </p>
<p>Again, the numbers I provided show that at least 25% of the Swarthmore students who have gone to a grad school directly from college in the last three years have gone to an Ivy League grad school, a Stanford grad school, or one of the two top public university grad schools in the country. That's eight grad schools that, across the board, would be considered top programs. I hardly think that's diluting the sample to include Podunk U.</p>
<p>That's combined: med, law, PhD, masters.</p>
<p>
[quote]
"WSJ seems to have hit on something if it causes an almost 3-to-1 shift in the Duke to Swarthmore numbers."</p>
<p>That's like saying apple production offsets orange production. They are two different things.
[/quote]
</p>
<p>Talking about apples and oranges is not a substitute for calculations. I again invite you to display any numerical plausiblity to your claims as to how the Wall St Journal metric could have led to such a swing in the numbers as it relates to the question: what is the productivity of Swarthmore (vs Duke, Harvard, etc) for grad admissions to the very top schools. Not PhD or MD production in general. Differential admission to the top programs is not the same as saying that Swarthmore numbers are inflated by doctorates at Podunk, but it would certainly change the interpretation of the overall PhD rates.</p>
<p>I gave you the list of grad schools and the numbers enrolled at each for 94 of the 250 graduates who went to the 16 most popular grad schools immediately after college from the most recent three classes. (13 at Penn, 12 at Columbia, 11 at Harvard, and so forth.) I don't know what more data you are looking for.</p>
<p>This is just stupid. The only claim I made was in response to the suggestion that top LACs are at a disadvantage in grad school placement. That is clearly not the case. If you believe that Duke has higher grad school placement rate or grad school placement rates at whatever passes your threshold for a "top" program, that's fine with me. I have no opinion on that one way or another.</p>
<p>My personal belief is that a "top" student at any "top" undergrad program will be able to get into a "top" grad school program. I'll leave it to you to sort out the defintions of "very top". If you want to restrict it to just five programs in each category, fine. If you want to count Berkely as "very top" for law, but not "very top" for physics, that's OK. Heck, it's OK with me if you want to include the NBA draft in your numbers.</p>
<p>
[quote]
Law and business school are a reasonable approximate proxy for the strength of the humanities subjects and economics, and med school a reasonable proxy for science. Not perfect, but reasonable, and not imperfect enough to dismiss when explaining the difference between 3 and 1.
[/quote]
</p>
<p>No way. PhD programs require engagement in research and innovation, sometimes starting from scratch, as well as a sense that your ideas can change the way we see the world. It represents the ideal of learning for curiosity's sake.
Business, law and medicine are simply a different sort of beast. The grade grubbers usually people their ranks.</p>
<p>
[quote]
Business, law and medicine are simply a different sort of beast. The grade grubbers usually people their ranks
[/quote]
</p>
<p>That doesn't mitigate the proxy effect. </p>
<p>If school A gets its grade-grubbers into top med schools at a distinctly higher rate than school B, it is expected that it will also get more of its unworldly scientists into biochemistry PhD programs. In that sense med school is a proxy for science.</p>
<p>You would have to assume an unrealistically extreme simultaneous distaste for business, law and medical school at LACs to explain some of the Wall Street Journal findings, where the LAC-to-top10 ratio swings significantly when outcomes are measured at 15 of the top professional schools. It isn't a one-year sampling question, either; you could aggregate the top 3-4 LACS into one large Ivy-sized university, pool their Wall St Journal numbers and still get a large shift. Self-selection into the LACs does not include outright avoidance of professional schools, and these LACs are sending applicants to business/law/med at a not insubstantial rate compared to the Ivies.</p>