Where CS faculty got their degrees

Appropos of discussions going on in my department about the quality of our graduate education, here is a list of where the tenure-track faculty in top CS departments got their PhDs. (Thanks to John Iacono for helping put this together!) The first several rows are taken from the top of the most recent US News rankings. Toronto is arguably the top CS department in Canada. We looked up Brooklyn Poly on a lark. The columns are mostly the same, except we removed Caltech and Georgia Tech (which had almost no PhDs in any department) and added Harvard (the most highly represented non-top-10 department). Dots means zero. We counted only full-time tenure-track faculty—no research faculty, adjuct faculty, lecturers, or emeriti.

Please do not take this data as gospel. We extracted it from whatever sources we could find on the web; I'm positive there are mistakes.

↓Job PhD→	MIT	UCB	Stan	CMU	UIUC	Cornl	UTex	UWash	Prin	Wisc	Toro	Hrvd	Other
MIT	10	8	7	2	.	1	.	1	.	.	1	2	3
Berkeley	8	4	5	5	.	.	.	1	1	3	1	3	9
Stanford	8	7	10	4	1	3	.	.	1	.	.	1	8
CMU	19	9	12	16	.	2	.	1	2	2	3	3	31
UIUC	2	2	4	3	4	1	.	1	1	5	.	1	26
Cornell	6	2	3	2	.	3	.	1	1	4	1	2	13
Texas	5	4	2	1	1	4	5	1	1	1	1	1	17
Wash	3	5	7	6	.	1	.	2	1	1	4	1	8
Princeton	4	3	6	2	1	1	.	2	1	.	.	1	6
Caltech	4	2	.	.	.	1	.	.	1	.	.	.	4
Wisconsin	.	6	2	2	1	3	1	.	.	1	.	.	12
GA Tech	5	3	.	8	.	.	.	2	.	3	.	2	34
Toronto	4	1	.	1	.	1	.	.	1	2	14	1	19
Brooklyn Poly	1	2	.	.	.	.	1	.	1	1	.	.	8

This table reflects baldy on my department in two ways. Looking at the UIUC row, we see that a slight majority of the faculty did not get PhDs from top-10 departments. More significantly, the UIUC column shows that we're historically abysmal at getting our PhDs into top-10 departments. We faculty at UIUC like to compare ourselves with MIT and Stanford, but Wisconsin does a significantly better job of placing students in top departments than UIUC.

We should do better!

Update (Dec 12): Let me add two major caveats about these numbers. First: Different departments define "computer science" differently. We chose to count tenure-track faculty in computer science departments only, regardless of the area in which they received their PhDs. So computer architechts with computer science degrees from UIUC who work in Cornell's ECE department are not counted, but Harvard-degreed psychologists working in the CS department at CMU are. Second: Faculty accumulate in computer science departments over a period of decades. As best, these numbers reflect years-long historical trends, not current departmental quality. Restricting attention to assistant professors might reveal a more accurate picture.

Comments

[Suresh Venkat]

Fascinating. Looks like CMU does the best at attracting top level Ph.Ds to their faculty, far more so than any other place.

MIT is often viewed as "inbreeding central": the chart though suggests that it is no worse than Stanford, and definitely not worse than CMU. Stanford suppposedly has a policy about not hiring its own grads, so I wonder how that number got so high (there is a loophole: if you go somewhere else for a while they don't mind so much)

[David Molnar]

Are the CMU numbers for the CS department alone, or for the entire School of Computer Science? I knew CMU was a large place, but I hadn't realized quite how large.

[John Iacono]

So here is a ranking the schools on this chart
based on the number of their graduates in schools
on the chart, normalized by the current size of their
faculty.

#1 (1.81) MIT
#2 (1.34) Stanford
#3 (1.21) Berkeley
#4 (0.70) Wisconsin
#5 (0.55) Cornell
#6 (0.52) CMU
#7 (0.48) Toronto
#8 (0.41) Princeton
#9 (0.31) Washington
#10 (0.16) UIUC
#11 (0.13) Texas

[Jeff Erickson]

David: The CMU are for the computer science department only.

One major caveat about these numbers: Different departments define "computer science" differently. We chose to count tenure-track faculty in computer science departments only, regardless of the area in which they received their PhDs. So computer architechts with computer science degrees from UIUC who work in Cornell's ECE department are not counted, but Harvard-degreed psychologists working in the CS department at CMU are.

[Mugizi Rwebangira]

I remember doing this informally when I was applying for grad schools a few years ago.

My impressions were:
(1)Berkeley and MIT excelled in placing grads at top places (compared to CMU and Stanford).
(2)A large fraction of Stanford faculty went to Berkeley.
(3)CMU faculty graduated from Berkeley or MIT.
(4)MIT faculty came from Berkeley or MIT.

Only (4) seems supported by your data.

My (subjective) conclusion was that Berkeley and MIT were slightly better in placement and MIT and Stanford were slightly better in recruitment.

As my contribution I divided the number of graduates placed at {MIT,Stanford,Berkeley,CMU} by the total number of graduates teaching at schools in the table.(Total graduates would be better but that is not available).

1.Stanford (34/58) = .59
2.MIT (45/79) = .56
3.CMU (27/52) = .52
4.Berkeley (28/58) = .48

Bah, this looks within the margin of error.

Turning to the faculty side here is the number of faculty from {MIT,Stanford,Berkeley,CMU} divided by total faculty size:

1.MIT (27/35) = .77
2.Stanford (22/40) = .67
3.CMU (56/100) = .56
4.Berkeley (29/43) = .55

This looks more significant. MIT and Stanford seem to hire a noticeably larger fraction from {MIT,Stanford,Berkeley,CMU} than CMU or Berkeley.

Obviously I wouldn't put too much weight on such analysis without a much more rigorous methodolgy. Still, very interesting.

Here is another indicator of student outcomes (ACM dissertation awards):

http://www-2.cs.cmu.edu/~rweba/doctoral.html

[David Molnar]

Thanks for the clarification on the numbers. I think your choice to count people in CS departments regardless of what their degree was in is sound. After all, people move interests and move fields.

[Victor Glazer]

Jeff, I think your numbers for Toronto are off. I don't know where everybody got their PhDs, but here are the ones I'm sure about:

MIT: Baecker, Rackoff, Marbach, Terzopoulos (4)
Stanford: Wortman (1)
Caltech: Jepson, Roweis (2)
Princeton: Almgren, Mendelzon, Mylopoulos (3)
Wisconsin: Kutulakos, Miller (2)

[Victor Glazer]

I just realized that this whole time I have been looking at the Toronto _column_ instead of the Toronto _row_. Sorry about the confusion. So your numbers for MIT and Wisconsin match mine, and Caltech doesn't count, which leaves Stanford and Princeton. Actually, CMU, Harvard and Washington don't seem quite right either:

CMU: Molloy, Demke-Brown (2)
Harvard: Cook, Hadzilacos (2)
Wash: Hertzmann (1)

Georgia Tech CS:

at least Prof. Milos Prvulovic got PhD from UIUC CS (advised by Josep Torrellas) http://www.cc.gatech.edu/~milos/

Georgia Tech ECE:

at least Prof. Raghupathy Sivakumar got PhD from UIUC CS (advised by Vaduvur Bharghavan)
http://users.ece.gatech.edu/~siva/

Cornell ECE:

at least the following professors are recent UIUC CS PhD holders:

José F. Martínez (PhD 02)
Sergio Servetto (PhD 99)

according to your statistics, only one UIUC PHD is on Wisconsin CS faculty, but actually there are two:

Prof Gurindar S. Sohi
Prof Jude W. Shavlik

Well, rankings should not count only the faculty placements, though this is an important factor. I have heard that UIUC has the third or fourth largest number of researchers in TJ Watson, which is arguably the best industrial research lab for CS grads. How do you account for that?

[James Landay]

This list posted by Eric is wrong:

He wrote:
"Turning to the faculty side here is the number of faculty from {MIT,Stanford,Berkeley,CMU} divided by total faculty size:

1.MIT (27/35) = .77
2.Stanford (22/40) = .67
3.CMU (56/100) = .56
4.Berkeley (29/43) = .55"

The Berkeley and Stanford %s seem to be swapped. Assuming the numbers are right, Berkeley is .67 and Stanford is .55.

[James Landay]

The list Eric posted seems to be wrong. He said:

"Turning to the faculty side here is the number of faculty from {MIT,Stanford,Berkeley,CMU} divided by total faculty size:

1.MIT (27/35) = .77
2.Stanford (22/40) = .67
3.CMU (56/100) = .56
4.Berkeley (29/43) = .55
"

Assuming the underlying numbers are correct, the Berkeley and Stanford %s are swapped. Berkeley should be #2 at .67 and Stanford #4 at .55

James

[sohbet]

Jeff, I think your numbers for Toronto are off. I don't know where everybody got their PhDs, but here are the ones I'm sure about:

[ruya]

www.ruya-tabirleri.com

[sohbet]

http://www.ruya-tabirleri.com/

[oyunlar]

Well, rankings should not count only the faculty placements, though this is an important factor.

[Rüya Tabirleri]

This ranking system is ver useful ..Thanks

http://www.ruya-tabirleri.org