The following essay is still somewhat in a "draft" form, but is otherwise
substantially cleaned up compared to the posts I made in early 1997 of
most of the "parts" below.
====
The Short Career Half-life of Scientists
by Arthur E. Sowers, PhD (Dec, 1997)
This essay is devoted to the question of how long the typical
biomedical science career lasts. In the middle of the essay,
I studied faculty job turnover at two medical schools and at
the end reviewed career half-life for physicians in a clinical
setting and lawyers. Parts of this essay originally appeared
in early 1997 as separate posts on the internet newsgroup
<sci.research.careers>
I defined a functioning scientist as a person who would
publish at least one journal paper per year and in a journal
covered by Current Contents, which is a leading scientific
literature indexing service. The version of CCOD I used was
the 1200 journal "Life Sciences" edition, which covers the top
1200 high impact biology/medicine journals. I picked journal
author names in an earlier year and then used the multi-issue
search function, to search all of the CCoD issues for all of
one year several years later. For the three journal author
searches, in the CCOD database, below, I took outputs that
were appended into one file (about 20 pages of journal paper
references containing roughly 150 papers). Each author
attrition study required about 10-15 minutes of personal
computer time. I then took the 15-20 page print outs, and
compared, by inspection, the author lists (which should have
located any paper that had one of the original 51 names on it
that was published in any of the 1200 covered journals). Each
author attrition study took about an additional hour, or a
little more, of my time. This is stated to give anyone who
might want to reproduce this kind of study some idea of how
much time it takes to do one of these studies.
1. Attrition of authors who published in the journal Science
I took an issue of the journal _Science_ in 1989 and copied
down all the biology article author names from a number of
papers. I had a total of 41. This was my sample population.
I found, in 1996, that 12 names could NOT be found on any
author list of any journal paper covered by the CCOD Life
Sciences 1200 journal edition. That is an absolute 29%
attrition in 6 years.
The author list publication pattern indicated that 13 of the
41 names were _probably_ PIs and therefore were: i) THE BOSS,
ii) had a secure/safe position, or iii) tenure, or, something
else. This is because some of the names appeared as sole
authors (eg. on review articles) or when they appeared on
several papers with few authors, the other author names
changed (as if the others were grad students or postdocs who
came and left or were on unrelated subprojects in a lab).
Subtracting the 13 from the 41, we have 28 probable graduate
or postdoctoral students. With 12 out of 28 leaving, after 6
years, the pool of publishing scientists, we have an attrition
rate of about 43% per six years.
How about the publication patterns. One of the 41 names had 60
publications. The second most prolific had 50 publications.
Yes, that is publications per one year (1996)! Two more had 17
and 22 publications in 1996. Seven names had only one
publication in that year. The rest had between 2-3 and up to
7-10 per year.
2. Attrition of biomedical authors that published in PNAS
I xeroxed the table of contents of the Proc. Natl. Academy of
Science, the last issue of 1991 and looked under the subjects
Biochemistry, Immunology, Cell Biology, and Medical Science.
These represent as much as possible "hot topics." I also
xeroxed the author index for the whole year. I went back to
the table of contents and started going down the author lists
for the articles and circled each author name which was
totally unique (i.e. there were no other identical names in
the year long author list). I stopped after I had 51 names
(about 10-15 from each of the four categories).
In the end, there were 20 names on the original list of 51
which could NOT be found as authors of 1996 journal papers.
This is close to a 40% attrition after 5 years as based on
a measure of a research endproduct, journal papers.
Some comments are in order. First, people who get papers into
high impact journals such as PNAS are starting off with an
excellent credential on their CV record. Usually, the
institutional affiliations are "top" names. It should be noted
that some of the names were surely PIs, who would be in a
better position to have "job security" by virtue of tenure,
name recognition reputation, etc. In such research labs, most
of the names on these papers were probably postdocs and maybe
research assistant professors and maybe a few graduate
students. Probably 4-5 of the names were indeed PIs, meaning
that the attrition figure is closer to 50% per 5 years than
40% (51 minus 5 = 46). Second, the question which is harder
to answer is how many of those who "left" science did so
voluntarily (eg. "Man, I'm tired of the 60-70 hour weeks" or
"This is just not what I want to do") and how many are due to
involuntary terminations (denial of tenure [on political
and/or grant problem grounds], non-renewal of appointments by
chairs, etc.).
3. Attrition of authors at pharmaceutical and biotech
companies
For the "biotech" career question, I started with the 11 Dec
1989 issue of CCOD (1200 titles, Life Science edition) and
looked at the most applied 35 journal titles and scanned
through each article in each issue for those articles where
all authors were affiliated only with a pharmaceutical or a
biotech company (i.e. I skipped universities, colleges,
institutes, and hospitals) as the institution of affiliation.
Less than 5% of all papers published were from companies. This
shows that the vast majority of science jobs (leading to
publications in the open literature) are in academia. Even in
pure biotech journals (eg. Nature/biotech, Agr. Biol.Chem,
Bioch. Pharm, Biotech Let., Drugs Dev Res., J Pharmacy &
Pharmacol., Nature, PNAS, Pharm Forum, Pharma Bulletin) the
articles were primarily from non company entities (i.e.
universities, hospitals, and institutes/centers). From those
articles I did somehthing a little different than with the
academic study. Because, in some of the cases the address
given for correspondence was the first author rather than the
last author (as is widespread in academia), I took the name
given with the correspondence address. This is for two
reasons. First, in some organizations the head of the lab
might place his/her name as first author, regardless of who
does the work, compared to academia where the PI usually puts
his/her name at the end. Second, I noticed for many of the
journal papers in this population, the correspondence name was
NOT the same as the last name on the author name list. This
might mess up the comparability with my prior studies, but
then finding out for sure who is who would mean lots of
expensive phone calls and I couldn't afford that. So, I
decided to see what we found out and then worry about how
comparable the results would be. After all, this was meant to
be a study I could finish in a couple of hours rather than
several months. All of this took about 30-45 minutes. I had
a total of 38 key names out of a total of about 200 authors.
Plugging those 38 names into the "search" lines for all of the
1995 CCOD issues (again matching to all authors in all journal
titles covered in the top 1200 life science journals) I found
that out of the 38 names I started with from 1989, a total of
15 names could not be found in the 1995 top journals in the
life science literature.
I can't rule out that there were some different people with
the same names (eg. as in common names like Smith and Jones)
and so the its possible that a few of the remaining 18 names
were different people, but that would require additional time
and effort that I could not justify at this time. A loss of
15 out of 38 is about a 39% attrition in 5 years, which is
also close, in round numbers, to what I found for the papers
from "academic" environments (see below) and should certainly
be larger if we looked at a real ten year interval instead of
a five year interval. Dont forget, the idea here is that even
if a scientist leaves one company and goes to another company
to do science and still gets to publish, then that guy should
still be publishing. The explanation may lay in the fact that
many companies hire PhDs (and maybe MDs from time to time) on
one and two year postdoctoral-type postions which are
sometimes not renewable past the second or third year (as part
of company policy). Hence, the temporary help "moves on," for
whatever reason, and, apparently, never publishes again. The
"regular" (read: "old permanent") staff, which may include
managers and section heads, stay around, luckily, for the
longer term. This is where a lot of you folks out there might
want to target. Incidentally, here is a list of most of the
actual author affiliations: Merck, Nicolet, P & G, Squib,
DuPont, Beckton-Dickinson, Baxter, Immunex, Hoff-Laroche,
Cytel, Upjohn, Exxon, Synergen, Rohrer, Hazelton, Lilly,
Schering-Plough, Janssen, Parke- Davis, Sandoz, Dow, Squib,
Syntex. Most of these are NOT biotech companies but
pharmaceutical companies. The biotech published literature
back in 1989 was an even smaller fraction of the total
non-academic open publication in the journals than for the
commercial affiliations compared to the academic affiliations.
4. American Society for Cell Biology Membership turnover
The source data came from a comparison of members in my 1997
and 1996 directory of members of the American Society for Cell
Biology. I compared all last names starting with "A" in the
1996 directory with the appearance of the same name in the
1997 directory. Bear in mind, that in an ideal world, a 30
year old graduate student or postdoc, would join a scientific
society for life long professional reasons (about 90% of the
actual membership is academic) and retire at, say, 60. Thus we
would expect 1/30, or about 3%, of the population to retire
each year (also making the other assumption that the "input"
into the membership pool was stable with the "output" from the
pool). In the 1996 directory, there were 233 last names
starting with "A", but in the 1997 directory, I could not find
32 of those names. Thats over 13% attrition in one year! For
the record, in the 1997 directory, there were 236 last names
starting with "A," which indicates the inflow of new gradaute
students and postdocs is roughly in equilibrium with the
outflow. Some of the 13% attrition could be due to scientists
changing their specialization focus, but I doubt that could
account for much of it. There is very high pressure to get
into these societies and stay in them so that the membership
becomes a line on a CV and, in some cases I know, gives the
membership a "ticket" to submit an abstract for presentation.
Reduced registration fees for meeting attendance covers much
of the membership fee.
The following two sections are indicators faculty job half-
life rather than career half-life
5. Faculty turnover at UMAB
The sources of data were a 1992 and a 1996 staff telephone
book for UMAB. These telephone books included the professional
school campuses (of which 90% of the staff is the school of
medicine) and the hospital system (so there are a lot of MDs,
BS, MS, and non-degreed personnel). I chose to find, in the
1992 telephone book, all the PhD-only (i.e. no MDs, no MD/PhD)
assistant professors (regardless of prefix, except that I did
not include visitings & adjuncts), for the letters A - G. For
each one that I found in the 1992 telephone book, I looked
that name up in the 1996 telephone book. I found 49 PhDs in
that range of letters in the 1992 telephone book and out of
those names, 21 were not in the 1996 telephone book.
That is a 40% loss in 4 years.
I was not counting, but I did notice in the 1996 phone book
that about 10-15% of the names became assoc. profs, or became
heads of some operation (i.e. a professorial rank was not in
that "field" of the phone book). There is no specific data on
who had tenure except that traditionally "research" prefixes
were non-tenure track but in fact faculty with pure asst
profs, assoc, and full profs names for their appointments had
no absolute relationship to whether they had tenure. A memo
came out which said that 50% of the basic science faculty had
tenure, and 30% of the clinical science faculty had tenure. A
fact: in nearly all cases, tenured faculty had to get
substantial salary from their grants and if they lost their
grants, then their salary dropped to a base level which could
be as low as 5-10% of their max salary.
There are only two ways PhD assistant profs leave an
institution: voluntarily, or involuntarily. I did not count
MDs at all because many of them can easily move on,
laterally, horizontally, vertically, obliquely, etc. or
after a timed "program" ends, but PhDs seem to me to be
unlikely to move voluntarily in such high numbers because
they are back into the 300/1 applicant pool and unless the
credentials are really outstanding, they can't be finding
better offers that easily. As I recall, UMAB has about 10%
of its "doctors" as PhDs. The other 90% are MDs. The other
reasons that come to mind are: i) grants that they are "on"
don't get renewed, or their grants don't get funded (thus
leading to denial of tenure), ii) "hostile political
conditions" exist in the department (someone gives them notice
to "move on" for whatever reason, or they are mistreated, or
they detect that their future there is in serious doubt and
they got themselves out, by themselves). Most of the time,
asst profs don't have the time to be moving around. They can't
afford disruption (moves usually cost 6-12 months in research
and publication momentum, not to mention family upheavals, the
need to sell and buy a house, etc). So, you can do it once
maybe, but its very hard (and it looks very bad on a CV) to
have too many moves per ten years. If you are on a campus,
particularly one with a medical school or a health science
center, and have telephone books separated in time by at least
four years, consider doing the above excercise (or your own
variation of it) and post the results here on SRC, or email
them to me privately and I will post them anonymously.
Also, consider asking at least one person in your department
(or faculty senate) if they have any idea what the tenure
denial rates have been in the past.
6. Medical school faculty turnover at Washington University
(St. Louis)
I used Washington University (St. Louis) SoM campus telephone
books from the years 1991 and 1996 to see how long PhD faculty
last there. I found that about 30% of those who were in the
1991 book could not be found in the 1996 book. One thing,
compared to the UMAB study above, is that this medical school
did not list ranks along with their degrees with consistency
within a book, nor were the listing patterns consistent
between books. As a comparison, I followed MDs as well. For
them, the turnover was even higher: 50% Turnover for MDs
might be expected to be higher since they can generally find
jobs more easily. Compared to the National Academy of Science
study, and the general experience of PhDs, the PhD route is
fraught with a much shorter career half-life than many other
professions that require advanced training. The medical
profession pipeline seems, as far as success is concerned, way
out in front of the grad school pipeline. When I was there, at
a SoM council meeting in 1995, the dean was reviewing what
they call the "matching program" results. Graduating MDs
apply for their next position, usually a internship or
residency, and then join staffs at clinics, private practices,
or HMOs or whatever, complete with god-like respect from most
eveyone (except, these days, insurance companies, malpractice
lawyers, etc.), and a very respectable salary to start really
getting professional satisfaction (after paying off large med
student tuition loans). The MD students apply to their next
station in life and then report to the dean what they "got".
As I recall, the vast majority of the graduating class got at
least one of their top three choices of where to go next. This
is a far cry from the hundreds of applications a typical PhD
makes in the transition going off of postdocs and onto faculty
appointments or "regular" or "permanent" jobs in industry.
Except for some problems with specialists (because of pressure
from the insurance companies), most of these guys go out and
have a career with a half-life that must take them into
retirement. I also remember noticing that the school had class
reunions throughout the year for classes graduating way back
in the 20s and 30s (yes, the pictures showed some 80 & 90 year
old guys), and I noticed for those graduating two decades ago,
the class reunion participation rate looked quite high. Now
that has to be because these guys are still doing medicine and
surely are not unhappy about it. Another indication: Medical
schools graduate about 16,000 MDs per year now. Another 8,000
come in, apparently, from outside the country. So we have
about 24,000 new doctors/yr adding to an existing pool of
about 575,000 doctors. If that pool is at about equilibrium,
then about 24,000 old doctors are retiring (or exiting or
whatever) every year, and dividing 575 by 24 gives about 20
years of full time active career doctoring for 100% of those
entering the active pool. The National Academy of Science
study says that half of the PhDs never get into the career
they trained for, and my little survey below suggests that a
career half-life for PhDs, after they do their first research
paper (just before or just after they get their degree), could
be about a half decade. The MD? I'm sure that at least 80-90%
of them, as indicated above, are still going strong after two
decades. With new pressures developing on MDs the last few
years (and more to come in the near future), this may change
in the future for the worse.
Now lets look at two careers with longer career half-lives
7. Physician turnover in community private practices
Using "Yellow Pages" telephone books for a western region on
the Delmarva peninsula (what is locally called "the Eastern
Shore" [i.e. east of the Chesapeake Bay, which is east of Wash
D.C.]). I looked at the entries for physicians and started
at the beginning of the alphabet and looked for individual
names. Then in the "2 year later" phone book I looked to see
if that name was there or not there. I didn't count all of
them, but did go through the A through C names and stopped
after 37. So... here is the big result: Out of 37 names in
the first book, I found 35 were still there after two years.
That means that the community lost 2 out of 37 MDs in two
years. On a percentage basis, and comparing with the analyses
I gave below, the "attrition rate" is about 10% per 4 years.
This is to be compared with a roughly 40% attrition for PhDs.
I would be interested in the nation-wide attrition rate for
practicing physicians.
8. Lawyer career success rate
The figures I got were from an small sidebar article in
Science, vol. 271, page 1795 (29 March 1996). it is titled "A
matter of degrees." It compares scientists with lawyers. It
says a survey by the American Bar Association in 1994 found
that 89% of the working 770,000 law degree holders are
"...pursuing their profession either in private practice,
industry, or the courts. It goes on to say, "In contrast, the
latest figures from... [NSF] on the country's 462,000 employed
PhD scientists and engineers show that 30% hold jobs outside
science." Another breakdown of the same data showed that 41%
were in research, 22% in teaching, 18% in management, sales,
or administrative duties, 4% in computers, and 15% "do 'other'
things." Compare "89%" with "41%."
Future studies and questions
It would be interesting to look at attrition from industrial
scientist populations (such as industrial microbiology), and
other fields (eg. physics, chemistry). To sample the work of
nonjournal publishing scientists, the patent literature, for
example, might be a valid equivalent of the open journal
literature.
=====
Art Sowers
-------------------------------------------------------
Written in the public interest, the essays on
"Contemporary Problems in Science Jobs" are located at:
http://www.access.digex.net/~arthures/homepage.htm
hit stats: http://www.access.digex.net/~arthures/.stats
Snail mail adr to me: P.O.Box 489, Georgetown, DE 19947
Email: arthures at access.digex.net
My "home" newsgroup: sci.research.careers
-------------------------------------------------------
