- Ph.D. Production and Employment. Most CS Ph.D's ever (1189) last year. Nearly all of them found jobs.
- Bachelor and Master's Students. Enrollment continues to decline.
- Faculty Demographics. Faculty sizes continue to grow, though slowly, and are getting slightly more diverse.
- Research and Graduate Student Support. A drop in research expenditures likely due to fewer grants.
- Faculty Salaries. Truly useful when negotiating your pay. An interesting inversion where the schools ranked 13-24 in CS pay higher salaries than the schools ranked 1-12. And just who is being paid $400K?
As predicted last year, our field is producing Ph.D.s at a record rate, and the short-term forecast is for continued record production. While there is no evidence in our employment statistics that the increased production is resulting in an inability of Ph.D. graduates to find work, an increasing fraction of new Ph.D.s appear to be taking positions outside of North America. In the wake of accelerating globalization of the marketplace, this is not surprising.Three consecutive years of decreasing numbers of new Ph.D. students, and a sharply reduced pipeline at the Bachelor's level, will make it difficult to sustain this production rate in the longer term. Moreover, it is not yet clear when the decline in our undergraduate program enrollments will end. The double-digit percent decrease in bachelor's production observed this year is likely to continue for the next several years. Coupled with the declining representation of women in our undergraduate programs, our ability to produce a workforce that is sufficiently educated technically to meet the needs of the job market in computing is being severely challenged. The declining enrollments at the Bachelor's level also will increasingly challenge the ability of CS/CE departments to grow their faculty as they desire.
How are the rankings determined?
ReplyDeleteThe CRA uses the last (1995) NRC Rankings. You can find the list of schools in each category near the bottom of this page.
ReplyDelete... It is not yet clear when the decline in our undergraduate program enrollments will end.
ReplyDeleteFor the physics community, graduation rates went flat in the early 1970s and have stayed flat for the past thirty-five years. It is fair to say that no one in the physics community in the 1970s foresaw this generation-long "flattening" of the profession.
Could similar flattening happen in biomedicine and computer science? As Lance's statistics show, the flattening has already happened ... so the question properly is, how long will it persist?
The physics community has adapted to flattening, such that (most) physicists now perceive it as the normal state of affairs. The default future for biomedical researchers and computer scientists is to adapt in the same way as the physicists.
China has implemented a quite-different strategy that emphasizes system-level integration, as exemplified by China's excellent Journal of System Simulation.
"The Journal is the most important medium of the Chinese Association for System Simu1ation (CASS) in scientific and technological exchanges, and is also a window through which the recent progresses of the research and development of system simulation in China could be viewed from the outside."
"The scope of the Journal covers all important subjects related to system simulation, such as theory and methodology of modeling and simulation, credibility evaluation of simulation, VVA, simulation computer and software, artificial intelligence, simulators for training and system research and operation, DIS, simulation-based design, virtual prototype, virtua1 reality and visualization, and applications of simulation technology to industry, communication, aero-space, military, and national economics."
The resulting flow of Chinese system-level research literature is growing exponentially, and this literature is strongly coupled to similarly vigorous economic growth. As a result, it is already the case that the dominant language of system-level engineering research is no longer english, but mandarin.
President Hu Jintao of China is himself an engineer, and all nine members of the CCP Central Committee (the Poliburo) are engineers too. Because these leaders are all in their early sixties (which is young for Chinese leaders), it seems very likely that these trends will persist -- and even accelerate -- for at least the next two decades.
In China's strategic vision of the future, the consilience of the scientific literature is explicitly manifested in the system-level engineering literature. Shaping this literature is therefore a primary strategic and economic objective for China (as it is for any high-technology culture).
The point here is that there are alternatives to stagnation in computer science and biomedicine. But these alternatives are not decoupled from significant challenges relating to liberty, enterprise, democracy, the openness of the scientific literature, and the long-term integrity of the planetary ecology.
There are no simple answers on a planet that has six billion people on it.
The lack of student comment on this topic is amazing.
ReplyDeleteTo put it another way, what could/should senior physicists have done, back in 1965, to forestall the forty-year flattening of physics that is shown in this AIP graph.
I assure you, back in 1965, no one saw this coming! What could my generation have done differently?
And just who is being paid $400K?
ReplyDeleteForget that...what about the poor person being paid $43K who now realizes that he/she is the lowest paid prof in the survey?
Three consecutive years of decreasing numbers of new Ph.D. students, and a sharply reduced pipeline at the Bachelor's level, will make it difficult to sustain this production rate in the longer term.
ReplyDeleteWhat about graduate students from other countries who join for PhD programs in US-universities? As the craze and brand value of US-education is still high, I donpt think that it will take some time for the PhD number to flatten. (Also combine it with the increase in faculty number.)
to forestall the forty-year flattening of physics that is shown in this AIP graph. I assure you, back in 1965, no one saw this coming! What could my generation have done differently?
ReplyDeleteI think physics shot itself in the foot in a singularly bad way. First the time required to complete a PhD thesis moved up to eight years, second, it systematically sneered at any problem that had practical relevance and instead chose to study YAPs produced by high energy accelerators (YAP=Yet Another Particle). Third it continued to produce a large number of PhDs which would find no gainful employment. So they would go around telling anyone who would listen (including undergrads) how there were no jobs in physics when in reality they meant: "there are no academic jobs (or industry jobs) for people with PhDs in YAPs".
Computer science should try to avoid making the same (or similar) mistakes. We need to stay relevant, remain sensitive to overproduction of PhDs and keep the requirements for the PhD degree reasonable, among other things.
My impression is that studying CS as an undergrad is seen to most naturally lead to a job developing enterprise applications. I don't think that this appears to be a very fun job (it may actually be great fun -- I've never tried it). I think a lot of undergrads (at least those who do not desire to go to grad school) are put off by this, and would prefer to study something else that they are interested in, perhaps taking a few CS classes so they know their way around.
ReplyDeleteI'm not sure which came first, but it also seems a lot of CS departments have begun to cater to master's students who are looking to get better enterprise-app building jobs after graduation (my understanding is that master's students are a huge source of income for many CS depts). This causes even more of a slant in coursework toward "programming" and away from "CS".
Given this second issue, I think that many graduate advisors might prefer to have grad students come in with backgrounds in math or physics or statistics or something along those lines, because there seems to be a growing disparity between what is taught at the undergraduate CS level (how to program) and what is desired of a graduate student (how to think).
This is just my perception---a sample size of one---so it may be worthless. I also wasn't a CS undergrad, though I did attend an undergraduate university with a top-5 CS program.
Given this second issue, I think that many graduate advisors might prefer to have grad students come in with backgrounds in math or physics or statistics or something along those lines, because there seems to be a growing disparity between what is taught at the undergraduate CS level (how to program) and what is desired of a graduate student (how to think).
ReplyDeleteI definitely learned a lot about how to think as a CS undergrad. But then, I did go to CMU...
My impression is that studying CS as an undergrad is seen to most naturally lead to a job developing enterprise applications. I don't think that this appears to be a very fun job (it may actually be great fun -- I've never tried it). I think a lot of undergrads (at least those who do not desire to go to grad school) are put off by this, and would prefer to study something else that they are interested in, perhaps taking a few CS classes so they know their way around.
ReplyDeleteWell, I'm considering quitting my job developing enterprise applications to go to grad school. However, the opportunity cost is rather high. I currently make more than the overall median reported for tenure-track positions for new PhDs. The problem is worse if you consider that I've been out of school for long enough that I think I have to spend some time after quitting refreshing my memory on those things I studied long ago. I might decide to do it anyway, though.
many graduate advisors might prefer to have grad students come in with backgrounds in math or physics or statistics or something along those lines, because there seems to be a growing disparity between what is taught at the undergraduate CS level (how to program) and what is desired of a graduate student (how to think).
ReplyDeleteFor CS in general this is just plain false.
I think that many graduate advisors might prefer to have grad students come in with backgrounds in math or physics or statistics or something along those lines, because there seems to be a growing disparity between what is taught at the undergraduate CS level (how to program) and what is desired of a graduate student (how to think).
ReplyDeleteIm assuming you mean "think like a computer scientist". Is this speculation? Or do professors really feel this way?
In any case, I can see how this would be more than relevant for theory applicants. So does this make grad students with a strong tilt towards theory (in the eyes of the advisor) much dearer than the systems junta?
it continued to produce a large number of PhDs which would find no gainful employment. So they would go around telling anyone who would listen (including undergrads) how there were no jobs in physics
ReplyDeleteDuh... The similarity is striking: the last three years I heard CS grad students in France (and even in Canada recently) spreading "the word" that one should do a PhD for the fun of it, not in the hope of finding a good job at the end... That most probably the PhD would go back to the industry, recruited at the same level as a Master's student.
My perspective is biased as I am in Theory, but I remember a student studying processor design with the same arguments. There are bitter people everywhere (after all this is a selective system), but when the comment about the past of physicists show that sometime they reach the critical mass and explode ;)
The parallel with physics doesn't hold water.
ReplyDeletePhysics rode the tides of:
* the Apollo space program at NASA
* nuclear bombs at national labs
* the nuclear power industry
* the general increase in university participation by the populace
each of which created demand for physicists.
The first three shrank substantially at the same time that interest waned and the fourth was completed by then with most of the new jobs filled by young people. There was no sustained large industry that had high demand for physicists.
The situation that CS is in now shares only has some similarity in the last of the four points; we are unlikely to see as marked growth in CS faculty size as in the past. However, the demand for our graduates in industry is as strong as it has ever been and isn't likely to go away soon.
spreading "the word" that one should do a PhD for the fun of it, not in the hope of finding a good job at the end...
ReplyDeleteIgnoring the job market at the end is a sure recipe for disappointment.
... That most probably the PhD would go back to the industry, recruited at the same level as a Master's student.
This used to be the case for unemployed PhDs before the bubble. One of the lasting effects of the internet bubble is that now companies understand that a properly deployed PhD can be very beneficial, even if not in a research role per se.
Add to this Google, and you'll do even better. From what I've seen Google has a preference for PhDs over masters or undergrads, even though they do hire at all levels.
There was no sustained large industry that had high demand for physicists.
ReplyDeleteCorrection. There is no sustained large industry that has high demand for the type of physics that most PhDs specialize in: YAP, cosmology, relativity...
Get me as many PhD in physics specialized in friction as you want, and I'll give you a job for every single one of them.
In addition, in most subfields of physics a result is considered physics so long as it is abstract. If your result is particular and applicable, then it is almost surely called engineering (there are exceptions, such as superconductors, where it is considered kosher to study the properties of a specific superconductor).
No wonder physicists can't find employment: they pursue arcane fields of study with no industry behind them, shun applications and produce massive numbers of PhDs beyond actual need.
Could this happen to CS? For sure! there are times when theory has acted this way, shunning applications and pursuing arcane complexity related questions.
So have the "software engineers" which are doing formal methods with limited value in software engineering as done in practice. And so have programming language researchers who spend most of their time arguing about higher order logic and functional paradigms that have proven not to work in practice.
Other communities, such as networks, databases and graphics do seem a lot more driven by actual applications.
I got a problem with the report's summation:
ReplyDeleteCoupled with the declining representation of women in our undergraduate programs, our ability to produce a workforce that is sufficiently educated technically to meet the needs of the job market in computing is being severely challenged.
What does the declining representation of women have to do with the workforce's technical education level?
The previous anonymous wrote:
And so have programming language researchers who spend most of their time arguing about higher order logic and functional paradigms that have proven not to work in practice.
That is patently false. Lisp works in practice. O'Caml works in practice. Haskell works in practice. And they do without corporate backing (although that may be about to change if you keep an eye on Microsoft's F# and its influences on C# a la LINQ). If you speak of PhD's, those are the languages and paradigms they can take advantage of. Nothing's been proven not to work.
That is patently false. Lisp works in practice. O'Caml works in practice. Haskell works in practice.
ReplyDeleteThis would be funny if it weren't sad. In practice no one uses Lisp, O'Caml or Haskell. With a lot of effort you can find the odd example of someone using one of them. They are the exceptions that prove the rule. In the real world people use C, C++ or Java, plus PHP and Perl if it has anything to do with the web.
Nothing's been proven not to work.
To the contrary, 30 years of harping the functional paradigm with no widespread adoption is proof of failure. Contrast this with structured programming and object oriented programming which were rapidly adopted. Burying your head in the sand and denying it leads to the physics disaster. I'm sure you can find physicists today that will say there is nothing wrong about so many of their PhDs working on YAPs and that YAPs are quite useful in practice.
This would be funny if it weren't sad. In practice no one uses Lisp, O'Caml or Haskell. With a lot of effort you can find the odd example of someone using one of them. They are the exceptions that prove the rule. In the real world people use C, C++ or Java, plus PHP and Perl if it has anything to do with the web.
ReplyDeleteFirstly, what works is not determined by its popularity; perl is the best example of the million-flies argument. FP is never going to be popular just like not every soldier can be a sniper.
To the contrary, 30 years of harping the functional paradigm with no widespread adoption is proof of failure. Contrast this with structured programming and object oriented programming which were rapidly adopted.
Properly, a proof of failure is not done via a social exercise where you measure adoption rates. Those language researchers you criticise know a thing or two about true capabilities. Btw, OOP was 20 years old before it became popular as well.
Btw, OOP was 20 years old before it became popular as well.
ReplyDeleteYou are whistling through the graveyard, my friend.
I'm sure you can find YAP physicists making the same type of arguments ("YAPs have a long lead time for applicability; since when applicability of our results is a measure of quality") when all the while enrolment plummets, funding dries up, and unemployed PhDs accumulate.
In my lifetime I've seen more theory results make it into the real world that I've seen programming languages results from academia make it to the real world. That ought to raise all type of alarms.
It always amazes me when somebody claims that there aren't enough students going into physics, or that there aren't enough women, or enough Americans, or whatever. Isn't it obvious? As a professor of mine said, you should only do physics if you love it so much you can't bear to do anything else. For anyone who isn't in love with the field, there are far more rational career options.
ReplyDeleteEven for those who love physics, but also care about other things in their life, pursuing physics can be a tough choice. The handful at the top of the field do very well, but for every full tenured professor, there's perhaps a dozen postdocs trying to support a family on $30k a year with no health plan. This is the reason there aren't enough women or Americans or people in general in physics--the sacrifices involved simply outweigh the rewards, and some groups are more sensitive to these kinds of practical considerations than others.
People don't go into physics looking for an enormous salary or ironclad job security, but it would be be nice to know that a physics Ph.D. meant a better-than-50-50 chance of getting a job that pays well enough to comfortably support a family, and that they'd actually have time to spend with my family.
Right now, the field is so competitive that virtually all "successful" physics faculty can be classified into one of categories:
1) Neglects job responsibilities (teaching or students)
2) Does lousy research but hides it well
3) Neglects family
4) Requires less than 4 hours of sleep a night to function normally (a very small number of people have this ability)
Despite all this, a typical tenure-track job posting will attract an enormous number of applicants. There isn't a shortage--there's a surplus. When physics departments start handing out tenure-track positions with $5k signing bonuses to newly-minted Ph.D.s, then I'll believe there's a shortage.