Thursday, July 12, 2007

Physics Jobs - Part 1 (Supply History)

Any discussion of the physics job market has to start with the facts, and the most basic facts are those of supply and demand. When I started grad school in the mid 70s, I was told by my major professor that "there are no jobs". Although this was a bit of an exaggeration, there were only a few tenure track jobs spread over thousands of recent PhD's. My university, which had hired several new faculty every year in the 1960s, only hired one new person during my time in grad school - and he came from overseas to replace someone who left his t-t position for a national lab.

It is not even close to being that bad right now. (More info on demand is in part 2 of this series.) Indeed, I would say the situation is pretty good for people currently in grad school, but the situation still has echoes of that past. I want to start with what you can learn from history about the supply of PhD physicists looking for academic jobs. Fluctuations in that supply, which look like the damped response of an oscillator hit with an impulse, likely reflect the hiring patterns in academia. Until equilibrium returns, job openings due to retirements will come in clusters about 30 to 40 years after a previous cluster was hired. (Filling new positions takes only a tiny fraction of the annual PhD production.)

How did we get where we are today? The following is my analysis of the data.

The figure below was assembled from information given in two figures (fig. 7 and fig. 9) produced by the AIP as part of regular reports on the job situation in physics. (I created my own version of these figures because web links are notoriously unstable, these figures from a 2004 report will be replaced in a few years, and so I could annotate them with some trend lines to make discussion easier.) I strongly recommend that graduate students look over those reports, since they contain far more information than I can write about it a few blog updates.

My figure shows the PhD production in the U.S. from 1900 through 2004 (actual numbers, from fig. 7) and an AIP projection of PhD production from 2005 to 2010 (from fig. 9). The error band on the projection is the light colored band. I won't say much about those projections (they will soon be replaced by real data), but they might be important to students considering the odds of getting certain kinds of jobs. I'll discuss the history reflected in these data in roughly 20 year increments. (Until I sat down to write this, I had not noticed that some underlying trends seem to change on a two-decade time scale.) Click this picture to enlarge it.

The first 20 years, from 1900 to 1920, show a fairly flat rate of PhD production. The start of research and graduate education was modest, producing a roughly constant number (20 to 30) PhD's per year from a small number of departments. Old articles in Physics Today reported that most of these students went into industry, which was a fairly easy transition because the academic research itself was mostly funded by the industries that hired the PhD students.

Things changed in 1920, when we start to see a steady increase in production. The slope (about 6.5 additional grads per year) corresponds to a large percentage growth rate, because the starting level was so small. PhD production more than doubled each decade. Even more interesting to me is that the Great Depression (starting in 1929) does not show up at all. I do not know what drove this growth, although demand for engineers (despite the Depression, this was a period that saw much civil construction) would require more physics faculty and hence more PhD programs across the country.

What does disrupt graduate education is World War II. Anyone who doubts what a big deal that war was need only look at these data. I have put a "line to misguide the eye" across the period from 1940 to 1960 to show that the missing war-time degrees were completed after the war. On average (following my line), the slope increases (to 20 extra grads per year) but the percentage growth rate slows somewhat.

Side comment:
The senior faculty during my time in grad school, including my major professor, came from this group. Many of them left undergrad or grad school to work on the research side of the war effort: radar, aerodynamics, nuclear weapons, computing, and code breaking. (My thesis adviser worked at Bletchley Park "breaking the Enigma" and I knew a person who watched the Hiroshima bomb go off from the observation plane.) Chemistry students also played major roles in these programs. A few PhD theses that were completed during that time were classified. After the war, they returned to college with skills not found in a typical undergrad or grad student today, and experience working in large research groups. They knew how to operate a large, federally-funded research enterprise when the 60s rolled around. They had no ties to industry.

Then comes Sputnik. I put a red arrow at 1957 to indicate that important event. Sputnik resulted in massive amounts of money in support of science and math education and research, from K to PhD. The increased funding for graduate fellowships, coupled with job demand from colleges and NASA, was a boon for physics and the graph reflects that change. Back then, and even in my time, you could save money (not borrow it) on a grad assistant salary if you did not have a family to support. Everyone got a job. Many faculty were hired directly from grad school, without even a post doc. Average was more than good enough, until about 1968 or so, and PhD production almost tripled in that decade. The peak at 1970 is the result.

It was the perfect storm. A decade of lush spending came to an end circa 1968. The NSF budget was cut to pay for Vietnam and to punish the colleges for harboring anti-war protesters. (Nixon was not known as a lover of intellectuals.) The end of the baby boom was in sight (high school grads crested around 1971), so universities did not need to add faculty because student growth was ending and enrollments would soon decline. NASA started to cancel projects and stretch out others, and no longer needed to hire physicists.

Side comment:
Have you ever seen a Saturn moon rocket when touring the Kennedy Space Center, or Huntsville, or the Smithsonian? Those are not models. They are actual Saturn rockets built to go to the moon. The Command Module and LEM at Kennedy are the real thing. Missions were canceled on the fly, after the rockets were built, because Nixon could not afford to launch them and fight a war and increase social welfare spending ... so they were just turned into museum pieces.

It was ugly. I know some really bitter people who got out in 1969 or 1970. Top notch thesis work meant nothing, while a few years earlier you could get a tenured job based on average work. The spike is really sharp because some people rushed their work to completion in hopes of grabbing the last job in 1970, while others simply dropped what they were doing and walked away. Why work two more years and graduate in 1972, if you knew there were no jobs? [One remarkable detail in the historical data is that the number of first year physics graduate students dropped just 1 year before the number of PhD's plummeted. The fraction completing a PhD after 6 years or so went from 40%, if you started in 1964, to 25%, if starting in 1970. See this figure from the AIP grad student report.]

People who went to grad school in the 70s knew (or eventually learned) what you were getting into. I knew some people who lingered and then left ABD, not bothering to write a dissertation when the reality sunk in. However, by 1980 there were a lot fewer people looking for PhD-based jobs. There were plenty of post docs, because the odds of turning one into an academic job were really bad. There were tech jobs to be had, particularly if you were a US citizen (and thus could get a security clearance), although no physics faculty had any clue about them. Faculty were now fully disconnected from industry, except in some areas of condensed matter research.

So the period from 1960 to 1980 was feast and famine. What happened from 1980 to 2000? Why a big peak in 1994? Part of the story is the discovery of quarks in 1974 and the death of the SSC circa 1993. The second peak you see is in 1994, right after the SSC was killed. This was also a time period when rumored demand for retirement replacements did not appear. (Universities shifted open positions from physics to new areas; biophysics is now hotter than nuclear physics and positions planned for the SSC went unfilled.) I think you are seeing a peak as people quickly wrap up their thesis so they can move on, and a drop as people quit. The fact that the AIP data also show a large (200 person) peak in the number of MS degrees exiting PhD institutions (solid yellow curve) in 1994 backs up this opinion. There is no peak from masters institutions. Those are PhD students who saw no point in finishing. I'll bet you would see the same thing in 1970 if those data went back far enough.

The other thing that happened from 1980 to 2000 was an increase in the fraction of foreign students in US universities. Faculty research programs had been built on huge numbers of graduate students and post docs, and there were also labs to be taught. American students would not take those jobs, so foreign students were recruited to take their place. (A graph of PhD's granted to US citizens would look very different from the one shown above.) This, along with immigration of PhD faculty from overseas, complicates the supply side of the academic job picture.

I have put a pair of red "lines to misguide the eye" in this area, reflecting two possible models for the supply situation. One tracks what a nuclear experimentalist would term the "background" below the peaks, while the higher one tries to split the difference and average the production over that period. The top line would be something like a long-term average of actual production, while the bottom line might be the level that could really be sustained. I find it interesting that the line from 1940 to 1960 roughly meets that bottom line around 1980, and that the top line meets the production curve around 1966 (when the faculty market started to saturate) and 1976 (when those who started just before the crash finally graduated). The bottom line might be a sustainable production rate, with the bottom line being the minimum graduation rate under any circumstances.

Demand will be taken up next, followed by issues related to preparing PhD students for the kinds of jobs that are likely to be available.

One consequence of the training and research experience of the current faculty is that major professors at a top R1 school with a 1/1 teaching load are (partially) training students for the only job they know about, which is not likely to be the one the student will take. This was a huge problem in the 1970s, and my impression from the blogs I follow (much like the impression on Usenet a decade earlier) is that it remains a problem today.

No comments: