Due to popular demand from the comment section, I spent some time this afternoon taking a look at the talks now posted from the KITP Snowmass on the Pacific conference held the past few days. This is part of an ongoing project for a US HEP Community Summer Study that will culminate at a meeting in Minneapolis later this summer. The US HEP community faces serious questions about what priorities for the future should be in an environment of flat-to-declining budgets, no energy frontier projects in the US, and discouraging news from the LHC about no evidence for BSM physics.
The KITP talks cover a wide range of topics, and I haven’t had a chance to look at very many of them. For theorists, one interesting session was the Wednesday panel on Structural Issues for Theorists, which featured presentations by and discussions with the people at DOE and NSF responsible for HEP theory grants (Simona Rolli and Keith Dienes). There’s a lot of information about the situation of US theory grants there, but I was a bit struck by the impression that despite the large problems faced by HEP in the US, for theorists things look much like they always have:
- Budgets are pretty flat. As salaries go up with inflation, and as new young people come into the system applying for grants, it’s harder and harder for people to get the grants and grant amounts they would like.
- The split between DOE and NSF funding of often exactly the same thing doesn’t make much sense and leaves people sometimes confused.
- There’s always a problem finding grant support for the number of students who want to do theory, leaving theorists trying to justify to their colleagues why their students should get more of the few TA positions available. The situation of funding about 180 theory Ph.D students in an environment where there are maybe 10 tenure track jobs/year in the country isn’t deemed even worthy of comment.
Some numbers from the presentations:
- The DOE spends roughly \$25 million/year on positions at government labs, another \$25 million/year on grants to universities (the bulk going to pay grad students/postdocs/summer salary). The NSF spends \$13-14 million/year on grants to universities, another $6 million on Frontier Centers, some of which have an HEP theory component.
- DOE funds 49 PIs at labs, 221 PIs at universities, NSF funds 186 PIs at universities.
The DOE split by field is 128 Phenomenologists, 73 “Formal” (often “string theory”), 42 Cosmology, 27 Lattice Gauge Theory. - DOE funds 123 postdocs, NSF funds 50 of them. DOE funds 130 grad students in theory, NSF 50 of them.
This stable system of government funding has been crucial in determining the structure of HEP theory in the US for the past few decades, and the academic system is built around it. I keep wondering what the effect will be as new sources of money come into the system from private sources, on scales approaching that of government funding. As a somewhat extreme example that is likely a one-time thing, last year the string theorists in Princeton got \$15 million in checks from Yuri Milner, a number somewhat larger than the entire NSF $13-14 million/year budget for HEP theory.
‘Snowmass on the Pacific’ … a preparation for ‘Snowmass on the Mississippi’ … I suppose it’s an inconvenient truth that one of the consequences of the flat/declining budgets is that ‘Snowmass at Snowmass’ is now too expensive?
`As salaries go up with inflation, …’
Is this a prediction, or has this phenomenon actually been observed?
In an equilibrium situation where the system was not growing and every PhD got an academic job, professors would have an average of one graduate student over the course of their entire careers.
In contrast, a typical professor has 2 or 3 graduate students… at any given time.
Food for thought.
That there are 20 times as many PhD students as tenure-track positions is not the problem; that people expect to stay in academia afterwards is. To be given the opportunity to complete a PhD and perhaps a post-doc in theoretical physics is really a wonderful gift from the taxpayers, for which one should be grateful. At least I am.
The information on funding by DOE and NSF is interesting but, in principle, it is not something not ‘already known’. What is there in these talks that really helps to ‘prepare’ for SOTM?
180 theory Ph.D students in an environment where there are maybe 10 tenure track jobs/year in the country
Is that particle theory only?
I wonder, what eventually happens to 170 – cooking meth can’t be really a career choice for theorists, besides, Dr.White already cornered that market…
If one wants to be hard-nosed about it, the optimum ratio of students to jobs is a balancing act between missing too many potentially top-notch scientists by taking in too few students, and spending unnecessary money on additional students without noticeably increasing the potential of an eventual prize catch. As far the fact that a number of students will have to go into alternative careers is a choice they can make for themselves, if one accepts that people are responsible for the choices they make.
Of course, if a professor only gets to have one student in his or her career, it will be harder to justify even 10 new faculty openings per year in the field, given that universities supposedely teach, but that’s a next-order problem. Perhaps a system could with more research positions and fewer faculty ones.
Peter, also interested in yours(and other’s opinions) regarding teh talks on supersymmetry and grand unified theories.
shantanu
There are probably 10 jobs/year advertised specifically as HEP theory research. However, people who get PhDs in HEP theory will do other things as well. Some will get jobs in teaching-oriented institutions; some of those jobs are advertised as HEP theory but some are advertised more broadly. They still use their training. A few will get various types of staff scientist jobs at government labs. A few will take their big physics brains and pedigrees and go into finance or consulting. A number of them will take their computational and mathematical skills into some sort of applied science work.
All of that is fine, and in my opinion society is well-served by the outcomes of their training. (Well, except maybe the consultants.) I don’t consider it a “waste” that they went into “alternative” careers.
The problem is that a non-trivial fraction of them will get stuck in various permadoc/soft-money stints, or adjunct stints. That’s fine when you’re starting off, but it’s a poor foundation for an adult life and career. Maybe this is their fault, because they didn’t prepare themselves to branch out. They stayed very narrowly-focused and never prepared themselves to leave. Or maybe it’s the fault of some advisors and “mentors” who actively or passively discouraged looking at anything outside of their academic niche. And there’s a system that enabled this.
I think that a lot of what NSF talks about with training and broader impact is fuzzy and silly, but I would not fault them if they started requiring that institutions getting funds for trainees in basic research also get some meaningful assistance in preparing for “alternative” careers. NSF should look favorably on departments that have seminar speakers from industry, that encourage their PhD students to take applied classes (e.g. computational methods, some engineering classes, materials science, other applied things related to physics) after finishing their qualifying exams, that have meaningful networking with alumni in industry (beyond something passive like a LinkedIn group), and generally have an atmosphere that says “You are doing fundamental research as a way to train your mind, now pay some attention to what you’ll do next.”
Actually, much of the preceding post applies quite generally across all of physics graduate research, not just HEP theory. It is a valid post, but not restricted to HEP theory.
Regarding the 10 jobs vs 180 students: many of the students are from outside the US, and many return to their home countries after PhD, or after one or two postdocs.
Re the 180 people and 10 jobs. This year the rumor mill lists 17 “faculty” positions — albeit some in Canada. Moreover, that 180 is presumably a stock, not a flow; given an average of 3 years in the “research phase” of grad school, that suggests supply outstrips demand by a factor of 3 or 4, not 18.
Some of those 180 people will fail to finish, others will finish and go to work for hedgefunds, or Google, or a start-up, or teach in a 4 year college, or take time out for family… I have friends, collaborators and former students who have done all of these things — the HET jobmarket is no bed of roses, the following statement
The situation of funding about 180 theory Ph.D students in an environment where there are maybe 10 tenure track jobs/year in the country isn’t deemed even worthy of comment.
is massively wrong; certainly by a factor of 5, and perhaps by a factor of 10.
Richard,
The statement is accurate, your interpretation of it is wrong: obviously the 180 students on NSF/DOE grants aren’t all graduating in one year. Your estimate of “3 or 4” for the ratio of phds/year to jobs I think is quite wrong. Some things to take into account:
1. Final numbers from the rumor mill for the past four years of faculty hires are 9,14,11,11, see here:
http://www.physics.utoronto.ca/~poppitz/Jobs94-08
This is post 2008-recession, and hard to know what the future holds, but I don’t see US universities expanding greatly in coming years. Looking at data back to 1994, the 1990s had numbers in the low teens, for a while in the early 2000s the best job market in decades got up to 20, over the last twenty years the average is probably 15 or so, with 11 the most recent number.
The 17 number you give includes not only Canadian universities, but all listed jobs, not all of which get filled, and some of which end up going to non-HEP candidates.
2. The 180 number is just those in the US on DOE or NSF HEP research grants. It doesn’t include theory students with some other kind of funding (TAs, other research fellowships). Depending on how you count, I’d guess around 300 Ph.D. theory students. There was a 1997 HEPAP census of US institutions on DOE/NSF grants, now very out of date, which found (with 99/120 institutions responding) 283 theory students (3rd year and above), and 78 theory Ph.Ds /year. See here
http://pdg.lbl.gov/hepap/hepapsurvey97.pdf
I’d be very curious to know if anyone has numbers on what the trends have been since the 1997 census. Speaking of Canada, one thing that didn’t exist in 1997 is Perimeter, where it seems that there are about 35 theory Ph.D. students now.
All in all, I think a current estimate of about 10 to 1 for theory Ph.Ds/job is reasonable, with this having been somewhat better during the early 2000s, but prospects for the future not better. Telling students that the ratio is 3 or 4 to one would be really irresponsible.
On the larger issue of whether this is a good thing, that has been debated here often in the past. I’ll just point out that with numbers like this it’s hard for HEP theorists to legitimately make a case to the NSF/DOE that there is too little theory graduate student funding. I’ll also point out that I’ve seen up close for many years the job market in mathematics (healthy) and in HEP theory (unhealthy), and the difference is huge. Why people in HEP want to defend living with such an unhealthy situation I don’t think I’ll ever understand.
A fair percentage of our graduate students in particle physics are from foreign countries. With US training, they often have good job opportunities in their home countries, which are not included in the 10 per year US jobs. There is also the military-industrial job market as well as the medical market. There is a projected shortage of big data analysts with “deep analytic talent” projected for 2018 of 140,000 to 190,000 people. These jobs currently start at over $100,000 for new graduates. The above discussion of about 80 theory Ph. D.s a year is totally swamped by the number of available positions in the technical job market.
Certainly, we will miss theorists that are highly trained in research directions that should be pursued, but as a training ground for technical people, our field is excellent.
Dennis,
I’d be wary of promising entering grad students that they’ll have no problem getting a $100K+ job data science job when they get their Ph.Ds or their postdocs end 5-10 years down the road, given that every major university is starting up graduate programs specifically in data science right about now.
But the point really isn’t that theory Ph.D.s will end up unemployed or working as baristas; these are smart, talented people who will find some sort of better than average employment doing something or other. The point is that there is a big difference in environment between a field where good young people see a future and the senior people working with them see them as future colleagues, and one where they are just cannon fodder, with everyone aware that unless they are both exceptional and lucky, they’re on their way out the door at one speed or another. I’ve seen both environments up close and I think describing one of them as healthy and the other as unhealthy is pretty accurate.
There are an infinite number of ways to slice this. But comparing the number of people in grad school (or a proxy for it) to the number of faculty slots is misleading.
Looking for some real data, found this
http://www.aip.org/statistics/trends/highlite/edphysgrad/figure7a.htm
which gives “particles and fields” as 208 students graduation / year in 2007 and 2008 (out of 1480). “Particles and fields” casts a fairly broad net and includes experimentalists while the Rumor Mill is incomplete in a variety of ways (people who take jobs internationally, or outside of the rumor mill’s scope but still in science) a reasonable guesstimate would be to say that 100 theory students compete for 20 jobs. Which is closer to the my “3 or 4” than it is to your (implicit) 18.
Richard,
There never was an “implicit” claim by me of a ratio of 18 people/job, that was just your (completely unreasonable) misinterpretation of what I wrote. The only estimate I’ve given of this ratio is about 10, you’ve now moved up from “3 or 4” to 5.
We seem to agree on the number of people (100/year). Yes, there are many ways of defining and counting what sort of jobs provide a research career in HEP theory. I still think the 11/year Rumor Mill count is a better estimate than your 20, note that this overcounts permanent positions (some people don’t get tenure or leave the field from tenure-track jobs). Yes, US educated students get research jobs outside the US, but students educated outside the US also get US jobs. In any case though, whether it’s 90% or 80% of theory Ph.Ds that have no future in the subject, that doesn’t much change the picture.
Just a quick comment on this. The chance of ultimately getting a faculty job also obviously depends a lot from institution to institution: students from Princeton or Harvard or Stanford have far better odds of ending up with a faculty job somewhere than those from Kansas State, and I think people know this from the outset. The former probably have odds a lot better than 1 in 3/4 (provided they are willing to stay in the system and do not self-restrict to top tier research universities), and the latter odds a lot worse than 1 in 10.
I think Erick Weinberg has roughly this data on his faculty jobs page.
Richard,
It is peculiar that you are promoting a rosy view, why? Things are in fact not rosy. Having numerous talented young people get Ph.Ds in HEP, and then sending them to goldman sachs or JP morgan sachs to develop high-frequency trading algorithms is not, in my view, an optimal outcome. I just don’t see how you could be stating this to be satisfactory. It’s not why people start PhDs in science, and if that was the final goal, there would be better training for it elsewhere than in physics.
I don’t think academia would be well-served by transitioning to a medical school model, where it would be virtually impossible to get into graduate school, but there would then be a guaranteed job for every who gets in. However, an intermediate solution, wheresay some ~50 of the 180 graduate student positions across the country are gradually replaced by 15-20 permanent staff scientist positions, should be contemplated as a positive step forward.
piscator,
I’m guessing you’re referring to the data gathered by Erich Poppitz (not Erick Weinberg) that’s in
http://www.physics.utoronto.ca/~poppitz/Jobs94-08
It shows most jobs going to people from a small number of places (Berkeley, Princeton, Harvard, Stanford), with those places placing about one person/year. I don’t have any numbers, but I’d guess those places produce something like 3 theory Ph.D.s/year. So, if you’re a student from such a place, a one in three chance of getting a Rumor Mill type job may be a good estimate.
I think there are too many PhD students – in my field (experimental HEP) the number of students in the country I come from has been doubled compared to say 10 years ago – yet there are no more faculty jobs than 10 years ago available. Furthermore I don’t see that a lot of these students are actually anything more than cannon fodder to do data crunching for PIs – the point of a PhD should be to develop your own idea in my opinion, and this is certainly what I was told to do when I did mine – but nowadays this no longer seems to be the case in a lot of cases.
I would reduce the number of students and instead use the money to provide a better career structure for those who do stay in the field post-PhD. Then you would have a more stable pool of experienced and competent people, and would in principal have a better pool of students coming into the system (I come across students regularly who I can honestly say have no place in a PhD program because they are just not capable – so if we can afford to fund them this is an argument we fund too many places). Furthermore with a better career structure you would be more likely to entice good people to stay on (many very good students in my group left academia straight after their PhD because of the poor career prospects and I don’t blame them for doing it).
I don’t think I am “promoting” an overly rosy view; I am genuinely interested in this question, and trying to explore the problem. There are certainly large uncertainties in the data if one can argue for prof:grad ratios that vary by a factor of 2 — and it is equally plausible that some people are motivated to make things look as dire as possible.
Piscator’s point that outcome depends very strongly on where you get your PhD is very well made, and it is clear that a handful of PhD programmes are producing a disproportionate fraction of the professoriate in the US. (My own PhD institution is one of the singletons on Poppitz’s “PH.D. INSTITUTION OF FIRST-TIME FACULTY HIRES” plot — one faculty member in North America in nearly two decades.)
As a passionate student considering a PhD in fundamental theoretical physics, between this and string theory’s dominance, it is quite demotivating. From my fairly uninformed point of view, I have a hard time seeing any progress beyond the standard model until our experimental capabilities greatly improve, and an easy time seeing people waffling like they are doing something useful. I hope I am wrong.
Unless you work on a project specified by someone else in a subject you aren’t interested in (or even believe in), you aren’t going to be able to afford food. Cheap food. Again I am probably wrong, but this is how it feels.
I would agree that taking on less PhD students would be wise. Certainly a lot of my colleagues simply should not be doing physics PhDs, to the point of cringing.
I really have no idea what to do!
Dudley: the standard model has several dozen unexplained features; there are numerous phenomena which require something beyond the standard model; for all these facts, there are numerous ideas about how they might be explained, but the field is still open for anyone with a new idea to voice it as well; and the mathematics and ontology of the subject also still needs work. So there’s no shortage of things to do.
Dear Peter,
As a (mathematically inclined) theoretical physics aspirant, I am extremely curious about your comments about the job scene in mathematics. In what way is the mathematics job scene much better ? Is that mainly an artefact of over-supply of particle theory graduates per permanent position available or is there a greater supply of mathematics permanent positions (again, why ? ) ? Does it have something to do with the style of work in each community (carefully substantiated and original in mathematics, versus speculative, flashy and “faddish” in high-energy theory) ?
Anonymous,
The main difference in the math academic job market is just that there is a much better ratio of Ph.Ds to jobs than in particle theory. Just about every higher educational institution does a lot of calculus and other math teaching, so needs a sizable math faculty. In contrast, most places that have physics departments typically have a fairly small number of theorists.
There’s a separate issue of faddishness, with the particle theory phenomenon of everyone jumping on the same hot topic and trying to write papers about it pretty much unknown in mathematics. This used to have some justification in particle theory, when experiments were producing unexpected results, pointing to what might be the most fruitful thing to work on. In recent decades, there has been little of this from the experimental side, but the theorist’s behavior is deeply ingrained culturally.
All in all, math research is a much healthier community. If you are interested in the boundary of math and physics and trying to decide which area to pursue graduate work in, math has a lot of advantages at this point.