Identifying potential pitfalls in the quantitative
appraisal system for scientific careers
Experiencing Ethics Seminar
December 3, 2012
Alexander M. Petersen
IMT Institute for Advanced Studies, Lucca Italy
Ethics in the appraisal of Scientific Careers
• Competition (“fairness”):
•
•
•
•
strategizing / extreme behavior, e.g. scientific fraud
CED (cognitive enhancing drugs)
free-riding + “tragedy of the commons”
Careers: predicting future career achievement using
incomplete information and poorly understood/
designed achievement measures
• Funding:
•
•
financial incentives & who should subsidize early
career risk
how to attribute / appraise / reward achievement,
especially in the case of extremely large team
projects
Outline
•
•
•
Science: growth and emergent complexity
Citations: proxy measure for scientific impact:
•
•
comparison across discipline
comparison across time / role of output inflation (baseball analogy!)
Careers: measured by appraisal of the publication/
impact portfolio:
•
•
•
h-index + empirical regularities + criticisms
•
New results: complexity of career predictability and co-evolution
of the scientific production function
career longevity
Competition: cognizant enhancing drugs (CED) + Is academia
becoming more like a professional sport?
Evolution of Science
I
imtlucca.it
addressed within these three levels.
Downloaded from stm.sciencemag.org on Septe
e
d
s.
e
al
ll
.
s
s
nt
n
of
e
e
d
l.
s
m
of
d
a
y
rn
).
e
g,
MOVING FORWARD WITH SCITS
An “atomic” We
view
ofwithScience
asthea Multi-level system
conclude
a description of
more general challenges and opportunities
surrounding SciTS. First, research relevant
to SciTS is conducted in a variety of settings—academic and commercial, technology development, and government sector.
As such, the variety of research results published, approaches and tools applied, and
data produced is impressive. We identified
more than 180 core papers and reports
that convey key results in team science research. Of those papers, 17 were published
between 1944 and 2000, with the remainder being published since 2001, showcasing a surge of activity on SciTS. Many of
the reported studies use proprietary publication data sets (such as Web of Science
by Thomson Reuters or Scopus by Elsevier)
and most tools are commercial, making it
difficult to replicate results. Data such as
journal publications,
conference
proceed*
Michael
Stuart
Brown
ings, and book chapters, but also patents
Joseph
L.comprehensiveGoldstein
and grant *
awards,
are not
ly collected across the sciences. The data
of the 1985
Nobel Prize in
studied areRecipients
typically published
in English,
Physiology
or Medicine
describing the
although science
is international
andfor
multilingual. Furthermore,
the unification
of
regulation of cholesterol
metabolism.
data records (such as the identification of
all papers by one scholar as stored in differFig. 1. Multi-level, mixed-methods approach ent databases) and the interlinkage of colK.SciTS.
Börner,
et science
al. A multi-level
systems
to
Team
can be studied
at differ- lections of data (such as the retrieval of all
perspective
the science
of team
science.papers that
supported
by one funding
* were
Marilyn
Kozak
(also cell biologist)
ent
levels usingfor
different
approaches.
Together,
Sci.insights
Transl.
Med.from
2, 49cm24
(2010).
the
derived
these studies
are worth award) proves difficult because no unique
Navailable.
= 70, Nsolo = 59
more than the sum of their parts.
identifiers are
Interactions mediated by social “forces”:
⤷
•
•
•
Collaboration (attractive)
Competition (repulsive)
Knowledge (an “exchange particle”)
Watson-Crick strategy:
451
publications
434
(95%)
Solo-artist strategy:
458
publications
Evolution of Science: “In the beginning...”
Social networks in science:
serve as the backbone for
reputation signaling used to
overcome the asymmetric
information problem
NATURE PHYSICS DOI:10.1038/NPHYS2180
reputation tournaments
Chain-like NON
Star-like NON
Tree-like NON
INSIGHT | PROGR
Pi (0) = Pi (1) = 0, that is, for a networ
singly connected nodes97 .
(3) In the case of a loop-like N
one direction) of n Erdős–Rényi net
unidirectional, and the no-feedback co
initial attack on each network is the sam
Galileo (15)
Galilei
ki = k, using equations
and (16) we
Noble patron
−kP∞
Figure 6 | Three types of loopless NON composed of five coupled (king, wealthy aristocrat,PPope)
)(qP
∞ = p(1 − e
networks. All have the same percolation threshold and the same giant
component. The dark node represents the origin network on which failures
Note that if q = 1 equation (19) ha
initially occur.
P∞ = 0, whereas for q = 0 it yields th
of a singlereputation,
network, and
equation (12),
Paul A. David. The Historical Origins of ‘Open Science’: An essay on patronage,
NON, (2) a tree-like random regular fully dependent NON, (3) a numerical solutions of equation (19)
common
agencyErdős–Rényi
contractingpartially
in the scientific
Capitalism
Society
3(2): Article 5 (2008).
loop-like
dependentrevolution.
NON and (4)
a randomand Fig.
7b. Interestingly, whereas for q =
regular network of partially dependent Erdős–Rényi networks.
equations (17) and (18) depend on n, fo
László Barabási2 & Tamás Vicsek1,3
1–7
ns between individuals in society
unity structure, capturing highly conmilies or professional cliques in a social
quent changes in the activity and comviduals, the associated social and combject to constant evolution7,11–16. Our
ms governing the underlying communYSICS
DOI:10.1038/NPHYS2180
is essential
for a deeper understanding
-optimization of society as a whole17–22.
orithm based on clique percolation23,24
e the time dependence of overlapping
e, and thus uncover basic relationships
evolution. Our focus is on networks
n between scientists and the calls beWe find that large groups persist for
Star-like NON
Tree-like NON
of dynamically altering their members of loopless
NON composed
of five
coupled
ility
to change
the group
composition
e
same
percolation
threshold
and
the
same giant
ty. The behaviour of small groups
disnode represents the origin network on which failures
cy—the condition for stability is that
unchanged. We also show that knowent
of members
a given
community
ke random
regular to
fully
dependent
NON, (3) a
the partially
community’s
lifetime.
These
ényi
dependent
NON and
(4) a findrandom
offundamental
partially dependent
Erdős–Rényi
networks.
differences
between
the
different
generalizations
of
percolation
theory
and large institutions.
ork.
In all
we apply
are (1)
theexamples
monthlyexcept
list of (3)
articles
in thethe
tion.
e-print96 condensed
xplicitly
the case of amatter
tree-like(cond-mat)
NON (Fig. 6)
25
92–94 authors , and (2)
hs,
with
over
30,000
ős–Rényi networks
with the same average
tween
the
a mobile
phone
i �=customers
1 and qij = 1of
(fully
interdependent).
i = 1 for
5)
and (16) we obtain
exact expression perfor the
(accumulated
overantwo-week-long
he
size
of
the
mutual
giant
component
for
all p, k
mmunication patterns of over 4 million
ration events (a new article or a phone
n
nceP∞of= p[1
social
interaction
between the
− exp(−kP
(17)
∞ )]
s), and can be represented as (timeeralizesofknown
results for link
n = 1,2.
For n from
= 1, we
ction
the changing
weights
esult pc = 1/k, equation (11), of an Erdős–Rényi
bed
in Supplementary Information. In
pc ) = 0, which corresponds to a continuous
ale transition.
structureSubstituting
at a givenntime
in the
= 2 instep
equation
(17)
ity ofofref.a 73.
randomly chosen individual
ults
The
repreuationcommunities
(17) are shown(social
in Fig. 7agroups
for several
values
se
n
=
1
is
the
known
Erdős–Rényi
second-order
erconnected parts within a network of
emergence of small-world collaboration networks with the increasing
b Phone call
a Co-authorship
role of team-workINSIGHT
in science
| PROGRESS ARTICLE
200+ years
•
•
Pi (0) = Pi (1) = 0, that is, for a network without disconnected or
G. Palla, A.-L. Barabasi, T.Vicsek. Quantifying social group
singly connected nodes97 .
(3) In the case of a loop-like NON (for dependences in evolution. Nature 446, 664-667 (2007)
one direction) of n Erdős–Rényi networks96 , all the links are
unidirectional, and the no-feedback condition is irrelevant. If the
initial attack on each network is the same, 1 − p, qi−1i = qn1S.=
q and B. F. Jones, B. Uzzi. The increasing dominance
Wuchty,
ki = k, using equations (15) and (16) we obtain that P∞ satisfies
of teams in production of knowledge. Science 316, 1036-9 (2007)
P∞ = p(1 − e−kP∞ )(qP∞ − q + 1)
(19)
cNote that if q = 1 equation (19) has
d 0.6only a trivial solution
14 = 0, whereas for q = 0 it yields the known giant component
P∞
Zip-code (12), as0.5expected. We present
Zip-code
of12 a single network, equation
Age
Age
numerical
solutions of equation (19) for
10
0.4 two values of q in
Fig.8 7b. Interestingly, whereas for q = 10.3and tree-like structures
equations
(17) and (18) depend on n, for loop-like NON structures
6
0.2
equation
(19) is independent of n.
4
0.1is dependent on exactly
2(4) For NONs where each ER network
m 0other
Erdős–Rényi
networks
(the
case
0 of a random regular
40
60
80
100
120
0
20
40
60
80
100
0
20
network of Erdős–Rényi
networks),
we
assume
that the initialsattack
s
Chait
RP, ed.isThe
of Tenure.
(Harvard
on
each network
1 − p,Questions
and each partially
dependent pair
has
ethe Growth
ContractionThe nf equations of equation (15)
same q in both directions.
University
Press,
Cambridge
2002).
are
exactly the same
owing
to symmetries, USA,
and hence
P∞ can be
obtained
t
t analytically,
+1
t
t+1
organizational shifts in the business structure of research universities
⟨nreal⟩ / s
•
•
dominated by single links, whereas the co-authorship data have many
dense, highly connected neighbourhoods. Furthermore, the links in
the phone network correspond to instant communication events, capturing a relationship as it happens. In contrast, the co-authorship data
⟨nreal⟩ / ⟨nrand⟩
•
Emerging trends in Science
shifts away from tenure towards shorter-term contracts + bottle
neck in the number of tenure-track positions available
120
�
redefining the role of
teaching
-vsresearch faculty
p
)[1 − q + (1 − q) + 4qP ]
(20)
P = (1 − e
Merging
∞
Splitting
2m
t
−kP∞
t
t+1
t
t+1
Death
t
t+1
t
t+1
2
t∪t+1
∞
m
t+1
shifts in the competitive aspects of science, universities, and scientists:
1
reputation tournaments
in
omnipresent
competition
arenas
p =
(21)
Figure 1 | Structure and schematic
dynamics
of the two networks
k(1 −
q)
fromBirth
which we obtain
c
m
considered. a, The co-authorship network. The figure shows the local
Again, as instructure
case (3), at
it is
surprising
that in
both
critical
community
a given
time step
thethe
vicinity
of threshold
a randomly selected
are developing fast enough to use all the PhDs
features of a changing
landscape.
lot of pent-up
good science going on
graduates who
science and edu
sell them off w
companies aro
each to take on
they can crank out, and more — but the quality
that
we
are unable to fund.” TilghTENURE’S DECLINE
of the graduates is not consistent. Only a few
workforce
advisory
At most Northman’s
American
institutions,
tenure commitnations,
including
Germany, are successfully
is typical for senior
appointments
such
tee faculty
will tackling
try
to determine
size
the
problemthe
by redefining
the PhD as
as professors and
associate
professors.
AchievAcademia's Crooked Money Trail - Science Careers - Biotech, ...
and
composition
of
the
biomedical
ing tenure generally requires a strong record
workforce
the
7 that
6 | N
A TNIH
Uwork
R Ecan
| Vsupport.
OL 472 | 21 APRIL 2011
of published research
and2administrative
soft money positions.” Stephan also©
wants
more
attention
2
8
0
|
N
A
T
U
R
E
|
V
O
L
4
7
2
|
2
1
A
P
R
I
L
2
0
1
1
2011
Macmillan
(seethe
‘HowNIH
to needs to
ent thoughts about how one might go including
Not committee
everyonemembership
thinks that
systems that support researchers over time, as the Howar
tenure’). Most
tenure
systems
allow junior
this,” she says. “We want to have data and get
reconsider
how
grant
funding
is apportioned.than for specific, short-term projects. Importantly, she note
recommendations that lack teeth.”
tenure-track faculty members a period of sevThe world is producing more
and information before we resort to any eral
Several
scientist
bloggers
believe
that
Noonan’s
PhDs than ever before.
years to establish such a record. In addition
Is it
time to stop?
The short space at my
disposal
allows me to present just a
engineering of the workforce.”
comments
imply
that
scientists
should
have
to job security, academic tenure aims to proto what extent does patenting enrich some faculty membe
tect
academic
freedom;
faculty members
can of ability
import increasing numbers of foreign students and postdo
as for
reformisextend
unhappy
access
to NIH
resources
regardless
ble,
tenure
not asbeyond
prevalent
as
Academia's Crooked Money Trail - Science Careers - Biotech, growing
...
http://sciencecareers.sciencemag.org/career_magazine/previous_...
surplus
of native-born scientists struggle to find jo
disagree
with
popular
opinion,
express
negaidual investigators. The 12,000-member or outcomes; they counter that meritocracycontinue to attract students into graduate programs despit
tive views about their institution, or research
dican
thatSociety
may for
notBiochemistry
be a bad thing.
supporting scientists over time, rather than individual gran
and Molec- unpopular
should rule.
topics.One contributor wrote that with
These and many other apparent quandaries yield to Steph
iology (ASBMB) in Bethesda, Maryland, budgets
shrinking,
Nevertheless,
tenure researchers
is receding
inreally
the should be
http://sciencecareers.sciencemag.org
conveys her findings in clear, comprehensible prose. If you
Butsome
Carazo recommendations,
Salas, a group leader at thesays
Uni- United
States, where
tight budgets havethe value of
roposed
concentrating
on communicating
academic science today, here’s my advice: Read this enlig
versity of Cambridge, UK, isn’t lying awake at prompted universities to hire more adjunct
min Corb, the society’s director of public research to the public.
ave night trying to dream up ways to manoeuvre faculty members. In 1970, roughly three-quar- Academia's
Money
Beryl Lieff Crooked
Benderly writes from
Washington, D.C.
s.ueThe
ASBMB
seeks aorcap
on theresearch
fund- ters of
Rockey
advises
mid-career
scientists
facing
himself
into a tenured
tenure-track
all faculty
members
were in the tenure
Trail
on. to
position.
Funded
by a portable
five-year grant
in the
United States,
accordingatono-cost
figoing
any one
person,
and suggests
that stream
an R01
renewal
to consider
exten-http://sciencecareers.sciencemag.org
10.1126/science.caredit.a1200001
“EVERYONE TENDS TO LOOK AT
from the European
he is The
ures
amassed
by the American
Association
of funds) for
changing
yhebe redirected
to the Research
R01 poolCouncil,
from large
sion
(stretching
out existing
grant
THE FUTURE OF THE PHD LABOUR
ure pleased with what he calls a high level of scien- University Professors (AAUP). By 1975, that
MARKET
VERY PESSIMISTICALLY.”
Undergraduates
also carry an
tives
thatindependence
have not made
medical
breakanother
to to
gather
data
or publish results;
ofyear
tenure
ed tific
conferred
by the grant,
even face
number
had
dropped
56%, and
it continued
increasing share
of the load, she
Academia's
Croo
Cancel adds: Their tuition, often paid with
ghs,though
such he’s
as well
theaware
Genome-Wide
Asso- Although
to
try 42%
making
that he has no guarantee
toor
fall.
Only
received
tenure
in 1995,
still
highly
desirable,
tenurecontingency
is not
as prevalent
as plans, such as
student loans, Trail
rises as more funds
in some places — and that may not be a bad thing.
n Studies programme and the Protein it wasseeking
bridge funding from their institutions.
go to research. Their teachers,
4 NOV E M B E R 2 0 1 0 | VO L 4 6 8 | NAT U R E | 1 2 3
meanwhile, increasingly are
ture©Initiative.
The
ASBMB
proposes
“Your reviewer is taking into account what
2010 Macmillan
Publishers
Limited.also
All rights
reserved
cut-rate adjuncts rather than the
famous professors the recruiting Undergradu
he NIH adopt a sliding scale, to partially Byou have already accomplished, so be sure to
increasing s
brochures boast about.
adds: Their
lower-scoring but meritorious grants. highlight how well your research is going and
student loan
CAREERS
go to resear
ocietyFixpresented
itsitrecommendations
to the strengths of your research team,” she says.
it, overhaul it or skip
completely — institutions
meanwhile,
individuals are taking innovative approaches to
stituteanddirectors
and
Lawrence
Tabak,
the
Rockey’s
top
recommendation,
she
says,
is
to
cut-rate adju
postgraduate science training.
famous prof
deputy director, on 14 March.
seek advice from the relevant grant-review
brochures b
Chait
RP,
ed.
The
Questions
of
e idea of capping the number of awards to programme officer. ■
EUROPE Transferable pensions for
researchers may soon be a reality p.125
“M
Q&A ETH Zurich’s president explains the
appeal of Switzerland p.126
NATUREJOBS For the latest career
listings and advice www.naturejobs.com
of a continuing position at Cambridge at the
end of the next four years.
“Everyone would like to have job security,”
says Carazo Salas, who moved this year from
ETH Zurich in Switzerland after his partner
secured a Cambridge post. But Carazo Salas
is fine with his current position. He may not
have job security in perpetuity, but he has
autonomy, few administrative duties, and no
teaching obligations. “If I secure funds to continue paying my own salary, I can conceivably
stay here as long as I want,” he says.
Although most academics strive for tenure, experiences such as Carazo Salas’s suggest that it is not the only satisfying career
course. Early-career academic researchers in
the United States, the European Union (EU)
and elsewhere are wrestling with major shifts
in tenure’s definition, availability and value.
Seen for decades as the only route to long-term
job security and academic freedom, its longstanding symbol as the ultimate prize for academic researchers has been eroding on many
fronts. Tenured and tenure-track positions,
already hard to secure, have become rarer in
some areas because of budget concerns. Other
regions are seeing increased
interest,
as govcientists who
attain a PhD
are rightly training for high-level positions in careers out- unemployed postdoctoral students (one of
proud —
theyto
have
gained
entry to Money
side academia.
Here,-Nature
examines
graduate- - Biotech,
several initiatives
ernments and institutions
try
attract
top
Academia's
Crooked
Trail
Science
Careers
... that have been introduced
http://sciencecare
Issues
&
Perspectives
an
academic
elite.
But
it
not as elite education systems in various states of health.
to improve the situation). “It’s just hard to find
talent. Tenure is no longer what it onceis was,
it once was. The number of science
a match” between postdoc and company, says
and young scientists as
might
want
to
survey
the
doctorates earned each year grew by JAPAN: A SYSTEM IN CRISIS
Koichi Kitazawa, the head of the Japan Science
features of a changing
landscape.
nearly
40% between 1998 and 2008, Of all the countries in which to graduate with a and Technology Agency.
IMAGES.COM/CORBIS
nging
enure
... articles on the
academic labor market
CAREERS
THE PHD FACTORY
S
to some 34,000, in countries that are members
science PhD, Japan is arguably one of the worst.
This means there are few jobs for the current
In the 1990s, the government set a policy to crop of PhDs. Of the 1,350 people awarded
tion
and
Development
(OECD).
The
growth
triple
the
number
of
postdocs
to
10,000,
and
doctorates
in natural sciences in 2010, just over
At most North American institutions, tenure
shows no sign of slowing: most countries are stepped up PhD recruitment to meet that goal. half (746) had full-time posts lined up by the
is typical for senior
such
buildingfaculty
up theirappointments
higher-education
systems The policy was meant to bring Japan’s science time they graduated. But only 162 were in the
as professorsbecause
and associate
Achievthey seeprofessors.
educated workers
as a key
academic sciences or technological services,; of
to economic
growth
‘The rise
of doctorthe rest, 250 took industry positions, 256 went
ing tenure generally
requires
a strong
record
“Follow
the (see
money!”
ates’). But in
much
of the world, science
PhD
into education and 38 got government jobs.
of published research
and
administrative
work
According
togetthe
film toAll
graduates
may never
a chance
takethe
full
With such dismal prospects, the number
including committee
membership
(see ‘How to
advantage
of their qualifications.
entering PhD programmes has dropped off
President's
Men
get tenure’). Most
tenure
systems
allow
junior
In some countries, including the United
(see ‘Patterns of PhD production’). “Everyone
Issues & Perspectives
(http://www.imdb.com/title
andmembers
Japan, people
who have
trained at
tends to look at the future of the PhD labour
tenure-track
faculty
a period
of sevCREDIT: Kelly Krause,
AAAS States
great/tt0074119/quotes
length
and expense
to be
eral years to establish
such
a record.
Inresearchers
addition con- capacity up to match that of the West — but market very pessimistically,” says Kobayashi
front a dwindling number of academic jobs, and is now much criticized because, although it Shinichi, a specialist in science and technoljob security,
academic
tenure
aims
to
pro%E2%80%9D)) , thistoadvice
from
the
shadowy
informant
an industrial sector unable to take up the slack. quickly succeeded, it gave little thought to ogy workforce issues at the Research Center
tectguided
academic
freedom;
faculty
members
known as Deep Throat
Washington
Postdemand
reporters
Bob where all those postdocs were going to end up. for University Studies at Tsukuba University.
Supply
has outstripped
and,can
although
disagree with
popular
opinion,
express
negafew
PhD holders
end up
unemployed,
it is not
Academia doesn’t want them: the number
Woodward and Carl Bernstein
in
cracking
the
Watergate
clear their
that spending
years securing
this high- of 18-year-olds entering higher education has CHINA: QUANTITY OUTWEIGHS QUALITY?
tive views about
institution,
or research
conspiracy.
level qualification is worth it for a job as, for been dropping, so universities don’t need the The number of PhD holders in China is going
unpopular topics.
example, a high-school teacher. In other coun- staff. Neither does Japanese industry, which has through the roof, with some 50,000 people
Nevertheless,
tenure
is receding
ineconomies
the
tries, such
as China
and India, the
traditionally preferred young, fresh bachelor’s graduating with doctorates across all disci“Follow the money!”
But Carazo Salas, a group
at thealso
Uni-serves
United
States,
whereUniversity
tight
budgets
have
are developing
fast enough
toeconomist
use all
the PhDs graduates who can be trained on the job. The plines in 2009 — and by some counts it now
Theleader
strategy
Georgia
State
According
to the all
film
All
the The main probthey can crankto
out,
and more
but the quality science and education ministry couldn’t
even surpasses
other
countries.
versity of Cambridge, UK,
isn’t lying
awake (http://www2.gsu.edu/~ecopes/)
at prompted universities
hire
more—
adjunct
Paula
Stephan
extremely
of the In
graduates
is not consistent.
Only a few sell them off when, in 2009, it started
offering lem Men
is the low quality of many graduates.
night trying to dream up ways to manoeuvre faculty members.
1970, roughly
three-quarPresident's
well in her illuminating and accessible
new book,
How
nations, including
Germany,
are successfully companies around ¥4 million (US$47,000)
Yongdi Zhou, a cognitive neuroscientist at
himself into a tenured or tenure-track research ters of all faculty
members
were
in
the
tenure
(http://www.imdb.com/title
tackling the problem by redefining the PhD
as each
to take
on some
of the country’s 18,000 the East China Normal University in Shanghai,
CREDIT:
Kelly
Krause,
AAAS
Economics Shapes Science (http://www.hup.harvard.edu
of the Organisation for Economic Co-operaTENURE’S DECLINE
ACA D E M I A
RETHINKING PHDS
BY KAREN KAPLAN
iologist Rafael Carazo Salas doesn’t have
tenure — nor is he expecting to pursue
the tenure-track system any time soon.
As a faculty member at a UK institution, he
doesn’t have that option — academic tenure
per se in the United Kingdom was abolished
more than 20 years ago.
position. Funded by a portable five-year grant stream in the United States, according to fig/tt0074119/quotes
/catalog.php?recid=31302)
. A2 7by
leading
on
6 |the
N A American
T U Rexpert
E | V O LAssociation
4 7 2 the
| 2 1 A P Rof
IL 2011
from the European Research
Council, he is ures amassed
© 2011 Macmillan
Publishers Limited.
All the
rights reserved
%E2%80%9D))
, this advice
from
shadowy informant
scientific
labor
market,
Stephan
isn’t looking
to sniff
out
pleased with what he calls
a high level
of scienUniversity
Professors
(AAUP).
By 1975,
that
known as Deep Throat guided Washington Post reporters Bob
tific independence conferred
by the grant,
even number
had dropped
to 56%,
andthe
it continued
high-level
government
corruption.
Rather,
using
“tool bag”
Woodward and Carl Bernstein in cracking the Watergate
though he’s well aware that
he has no guarantee
to fall.
Only 42%the
received
tenure in 1995,
economics
provides for
“analyzing
relationships
between
conspiracy.
Education bubble!
incentives and 4costs,”
she penetrates the financial structure of
NOV E M B E R 2 0 1 0 | VO L 4 6 8 | NAT U R E | 1 2 3
explaining the motivation and behavior of everyone from august university presidents and
The strategy
also serves Georgia State University economist
professors to powerless and impecunious graduate students
and postdocs.
Paula Stephan (http://www2.gsu.edu/~ecopes/) extremely
well in her illuminating and accessible new book, How
It's a remarkably revealing approach. Most of what the public hears about the arrangements that govern research comes
Economics Shapes Science (http://www.hup.harvard.edu
from reports by blue-ribbon commissions, prestigious panels, and university-oriented advocacy organizations. Such reports
/catalog.php?recid=31302) . A leading expert on the
rarely use hard-headed economic analysis; rather, the groups writing them tend to consist of top administrators at leading
scientific labor market, Stephan isn’t looking to sniff out
universities, eminent faculty members in major science and engineering departments, and high executives of large
high-level government corruption. Rather, using the “tool bag”
corporations -- “not,” Stephan pointedly notes, “students and postdocs who could not find jobs.”
economics provides for “analyzing the relationships between
incentives and costs,” she penetrates the financial structure of
The documents that result from those high-end studies
lean toward self-congratulatory
invocations
of science’s
role in of everyone from a
university-based
science, explaining
the motivation
and behavior
advancing human welfare. Their suggestions generally
favor solving
what ails and
universities
by giving
them more
of what
they
professors
to powerless
impecunious
graduate
students
and postdocs.
already have: funding, grants, graduate students, and postdocs. But, warns Stephan with an astringency that she infuses
around 15–25%.
“If
you
lose
that
one
grantindividuals
has
received
some
attention.
Sevthroughout the book, when “assessing recommendations,
should be
leery ofapproach.
those coming
groups
have
a about the arrang
It's a one
remarkably
revealing
Mostfrom
of what
thewho
public
hears
renewal opportunity,
it’s hard
to recover
in thisthe way
eralitinvestigators
receive multiple
NIH grants:
vested interest
in keeping
the system
is.” from reports
by blue-ribbon commissions, prestigious panels, and university-oriented
university-based
© 2010 Macmillan Publishers Limited.
All rights reservedscience,
UNITED STATES
Tenure. (Harvard University
Press, Cambridge USA, 2002).
Macmillan Publishers Limited. All rights reserved
Mid-career crunch
Some senior scientists feel neglected by the National Institutes of Health’s grant formula.
B Y1 K7
E N D AM
L L PA
O WR
E L LC
H 2 0 1 1 | VO L 4 7 1 | NAT U R E | 3 9 9
istrative and managerial inefficiencies should
not conflict with the core activities of research
and teaching — ‘academic freedom’. Actually,
the reverse is true. With my colleagues at the
consultancy firm Berinfor, which advises
on the management of research and highereducation institutions, we have found that
reducing bureaucracy can increase scholars’
time for research and teaching.
Clean up
the waste
REDUCE AUTONOMY
institutions to learn to manage themselves
better, by adapting concepts from private
taxpayers do not get a return on their investbusiness. There are differences, however:
ment. Neither does the postdoc: someone
whereas unstructured time is anathema
who did not go to graduate school and
in business, it is key in research, enabling
faculty members to develop new ideas.
entered
the labour market in 2001 was earnFunding uncertainty strands
Spain's
young scientists
: Nature
N...
http://www.nature.com/news/funding-uncertainty-strands-spain-..
Better
management
is not about
telling
ing about US$58,000 in 2008; a first-year
professors how to teach and researchers how
to run experiments. Cutting back on adminpostdoc who started graduate school in the
istrative and managerial inefficiencies should
NATURE | NEWS
United States in 2001 was making around
not conflict with the core activities of research
and teaching — ‘academic freedom’. Actually,
$37,000 in 2008 on graduation3. During
the reverse is true. With my colleagues at the
a three-year postdoc position, a scientist
consultancy firm Berinfor, which advises
gives up more than $60,000 on average in
on the management
of research and collaborations.
higherDelayed decisions
disrupt international
education institutions, we have found that
return for highly uncertain job prospects.
reducing bureaucracy can increase scholars’
Lucas Laursen
And many postdocs will not get a research
time for research and teaching.
... on inefficiencies...
Funding uncertainty strands Spain's young scientists
Higher education relies heavily on the autonomous,inefficiencies
expert work of brilliant
minds.institutions
But
Fixing
at academic
sometimes,
that
autonomy
can
be
taken
tooteaching
will strengthen — not jeopardize —
far. Some
academics have
tendency Marty.
to
and research,
saysaThomas
Perverse
incentives
5 APRIL 2012 | VOL
job. There are few faculty openings, and
limited numbers of research positions in
Higher education relies heavily on the autonMADRID omous, expert work of brilliant minds. But
government and industry. So even if indisometimes, that autonomy can be taken too
vidual postdocs cost less, from a societal
far. Some academics have a tendency to
4 8 4 | NAT U R E | 2 7
Spanish researchers are feeling the budget squeeze — until
now restricted
to creditors
Spain's regional
perspective
they
can beofexpensive.
5 A P—
R Ias
L 2the
0 1 2country
| V O L 4scrambles
8 4 | N A T Uto
R Enegotiate
| 27
governments
a 2012Equally
budget. harmful are rules that encourage
© 2012 Macmillan Publishers Limited. All rights reserved
scientists to support graduate students on a
Last November, Diego de la Fuente, a graduate student inresearch
astronomyassistantship
at the National Aerospace
Technical
(RA) rather
thanInstitute
on a
in Madrid, made a bet. He would gamble travel costs and two months' living expenses of his own money to visit
training grant, despite evidence that the
06 March 2012
REDUCE AUTONOMY
“Consider the financial calculations
that encourage universities
hire
series
ofyear to work with astronomer Donald F. Figer at the Rochester Institute of
theto
United
States inaMarch
and April this
Counterproductive
financial incentives
divert
time
Technology in New York. At the time, the bet seemed a safe one: de la Fuente’s name was on a provisional list of
postdocs
rather
thanscientific
staff scientists.
Postdocs
earn
mobility-grant
winners
under
the Research
Personnel Training programme run by the Ministry of Economy and
and
resources
from the
enterprise.
We
should
SUMMARY
Competitiveness.
around
half
to
two-thirds
of
a
staff
scientist’s
● Science is full of incentives that
spend the money more wisely, says Paula Stephan.
By the
first week
neitherwhen
de la Fuente nor any of encourage bad financial choices, such
salary..... and are temporary, so
can
beof January
let go
as expanding labs and hiring too many
the roughly 1,200 other provisional winners had received
cientists
may portray
themselvesBut,
as not in reality,
Forconfirmation
instance,
incentives
adopted
by
temporary scientists.
budgets
decline.
postdocs
are
not
ofcash
their grants,
according
to Pilar Navasbeing motivated by money, but they
countries
as China,
Korea
and
Parejo,such
a graduate
student South
in geology
at the University
of ● These incentives hurt both individual
cheap:
substantial
resources
both
their
own
and
and the institutions where they work
Turkey
encourage
local scientists
to sub-of Young scientists and society as a whole, which
Granada
and a spokeswoman
for the Federation
respond in spades to financial opportunities.
mit papers
to
high-end
journals
despite
Investigators (FJI)/Precarios advocacy group.the
Provisional gets minimal return on its investment
society’s
have
been
invested
in
training
them.
If
a
Incentives that encourage people to make
low probability
success.
winners of the of
previous
year These
had theirpayments
funding confirmed by when someone is trained for a field with
end of December
— although
one decision
instead of doesn’t
another for monetary
havethe
achieved
little
more
thanpayments
overloadPerverse
postdoc
get a [permanent]
research
job,typically arrived no career prospects.
later.
reasons play an important part in science.
ing reviewers, taking them away from their
● The way forward is to fix incentives that
notareget
a return
onhave
their
investment.
This istaxpayers
good news if thedo
incentives
right.
work, and
increased
submissions by
are damaging
the system, by considering
incentives
Worse,
de
la
Fuente's
January
paycheck,
managed
by
the
But if they are not, they can cause considerable
the three countries to the journal Science by
theirCounterproductive
true social and financial
personalincentives
cost. divert time
Neither
does
the
postdoc....”
National
Research
Council,
another
central-government
damage to the scientific enterprise.
46% in recent years, with no corresponding
and resources from the scientific enterprise. We should
S
agency, was nowhere to be found. "At that time I had to ask
for money urgently from my family," he recalls.
© 2012 Macmillan Publishers Limited. All rights reserved
The Spanish government has left many young
researchers waiting months for their grant money to
the money more wisely, says Paula Stephan.
bespend
paid out.
5 A P RINTERNATIONAL
I L 2 0 1 2 | PHOTOBANK
V O L 4 8 /4ALAMY
| NAT U R E | 2 9
cientists may portray themselves as not
being motivated by money, but they
For instance, cash incentives adopted by
countries such as China, South Korea and
Downloaded f
... increasing team size & the reward system in science
CREDIT (TOP TO BOTTOM): ATLAS; DENIS BALIBOUSE/REUTERS/LANDOV
Online
Even the famous particle, the Higgs boson,
doesn’t quite live up to its legendary status.
Often portrayed as the engine that drives the
1286
C I TAT I O N I M PA C T
Name recognition. Peter Higgs was one of six theorists to have the same idea.
9 DECEMBER 2011
14 SEPTEMBER 2012
quarks into protons and neutrons, and the
weak nuclear force, which produces a kind
of radioactivity. (The standard model does
VOL 334 SCIENCE
VOL 337 SCIENCE www.sciencemag.org
Published by AAAS
Published by AAAS
Two Saudi institutions are aggressively acquiring the affiliations of overseas scientists
with an eye to gaining visibility in research journals
4 years largely through initiatives specifically
targeted toward attaching KSU’s name to
research publications, regardless of whether
the work involved any meaningful collabora-
10
and they want to build something out of it.”
Another KAU affiliate, astronomer Gerry
-4
Gilmore of the University of Cambridge in
the United Kingdom, notes that “universities
buy people’s reputations all the time. In principle, this is no different from Harvard hiring
a prominent researcher.”
-5 did not respond to
Officials at KAU
Science’s request for 0
an interview. But
Surender Jain, a retired mathematics professor from Ohio University in Athens who
is an adviser to KAU and has helped recruit
several of the adjuncts, provided a list of 61
10
10
CREDIT: AMMAR SHAKER/WIKIMEDIA COMMONS
CELL
NEJM
Nature
PNAS
of mathematics at -3
Ohio State University in
Columbus
who
has
signed
on,
says
he
has
10
PRL
www.sciencemag.org
no concerns about the offer. “It’s just capiScience
talism,” he says. “They have the capital
Saudi Universities Offer Cash
In Exchange for Academic Prestige
At first glance, Robert Kirshner took the
e-mail message for a scam. An astronomer
at King Abdulaziz University (KAU) in Jeddah, Saudi Arabia, was offering him a con-
10
cemag.org on December 12, 2011
“50-way tie for the Nobel Prize”
10
CDFCDF(a
(authors
> x)≥ x)
Downloaded from www.sciencemag.org on September 14, 2012
Who Invented the
Higgs Boson?
CREDIT (TOP TO BOTTOM): ATLAS; DENIS BALIBOUSE/REUTERS/LANDOV
4
Downloaded from www.sciencemag.org on December 1
Saudi Universities Offer Cash
In Exchange for Academic Prestige
Another KAU
affiliate,
dreamed had
leagues
a Nobel team
Prize, of
opinions
vary.grew
what
sayseveral
they did.”
s,ingdom,
so its members
to fimerits
nd rooms
physicists
topeople
include
astronomers.
But the
groupastronomer Gerry
Gilmore of the University of Cambridge in
he origins of mass. “Certainly, yes, I think it is at least as prof the gate,”
says
a thatstill
had
a tough
time persuading
review committees
of telescope
facilithe United
Kingdom,
notes that “universities
standard
model
of Andrew
found asFruchter,
other things
have
been
given the
The problem
ties
to
grant
them
observing
time.
buyalbeit
people’s
reputations all the time. In prind forces. Experi- Nobel in the past,” says Frank Close, a theo- What Higgs and company did,
unwitthe accelerating
While
theUnited
SCP was
led by
interested
in
as from
a Harvard hiring
ciple,
thisastronomy
is no
escovery
world’soflargest
rist at the universe,
University of Oxford
in the
tingly,
wasphysicists
to dynamite
a huge
boulder
thatdifferent
Two the
Saudi
are aggressively
the affiliations
of overseas
a prominent
researcher.”
Hadron
Collider
was
blocking
progress
been
firstinstitutions
to start. Founded
in the acquiring
tool to understand
the universe,
thescientists
High-z
collaboration
grew
out of a
Officials
at
KAU did not respond to
rticle
on the
model
with
an
eye labto gaining
in research
journals
er
andphysics
Richard
Muller,visibility
both physiteam
of astronomers who realized that Type
1astandard
supernova
explosions
Science’s
request
for of
an interview. But
erland, have now
of particle
physics.
could help them answer
a
fundamental
physics
question:
the
fate
sciencemag.org
first p.
glance,
initiatives specifically
Surender
Jain,isa retired mathematics pro, At
13 July,
141). Robert Kirshner took the 4 years largely through
The standard
model
the cosmos. These astronomers—including
Mario
Hamuy,
Nicholas
Podcast interview
e-mail message for a scam. An astronomer targeted toward attaching
KSU’s nameato
fessor
from
Ohio University in Athens who
quantum
field
theory,
with writer
Adrian
Suntzeff,
Mark
Phillips,
and
others—had
been
studying
nearby
at King
Abdulaziz
of whether
is an adviser
to KAUType
and has helped recruit
what
happened.
In University (KAU) in Jed- research publications,
so it focuses
on quanChoregardless
(http://scim.ag/
1a
supernovae
for
years
before
they
began
the
search
for
distant
Type
dah,
Saudi
Arabia, was offering him a con- the work involved any
meaningful collaboraseveral
adjuncts,
provided a list of 61
ade
a fairly
narrow
tum waves
or of
“fithe
elds”
pod_6100).
the universe is expanding,
far-off
supernovae
tract forNEWSFOCUS
anphysadjunct professorship that would1a supernovae.
tion with KSUBecause
researchers.
academics
who
have signed contracts simiNEWSFOCUS
athematical
that describe
the prob$72,000
a year. Kirshner, an astrophysi-recede Academics
both
inside
and
Saudi that
lartheir
to the
oneand
sent
to Kirshner.
The financial
tspay
made
the same
of fioutside
nding
various
particles
here
from Earth
at ability
such great
velocities
light
reaches
us
cist atintellectual
Harvard University, would be expectedstretched
Arabiain
warn
that such
practices
could
arrangements
Their
there.
It contains
elds
dozen
types in
of the contracts vary, Jain says:
wavelengths
toward
thefidetract
red for
endtheof
the electromagnetic
to supervise
a research group at KAU and from the genuine efforts
that
Saudi Arabia’s electrons,
For instance,
evelopment
of the
matter
particles—including
thesome
up
spectrum—a “redshift”
represented
by the letter
z. That’s
whyadjuncts
these will receive their
week or two a year on KAU’s cam- universities are making
to and
transform
them- that
compensation
hespend
most aelaborate
quarks
down quarks
make up pro-not as salary but as part of a
areinto
known
as tons
high-redshift
or high-z
Unlike
Perlpus, butBut
thattheir
requirement was flexible, theobjects
selves
world-class
research
centers.
Forthesupernovae.
research
grant
provided
by KAU.
science.
and neutrons,
and
heavier analogs
group,
the High-z
team
was
a flbilat
organization.
Even though
making
instance,
the Saudi
government
has
spent
acknowledges
that a primary goal of
n person
the most
impor-the offer wrote in the e-mail.mutter’s
of those
particles
that
emerge
inJain
high-energy
technically
leader,
the team
a collaboration
What
Kirshner
lions ofwas
dollars
to build
thethe
new
King Abdulthewas
program—funded
by Saudi Arabia’s Minped
to solve.
Other would be required to do,Schmidt
particle
collisions.
was add King Abdulaziz Univer-among
lah University
of Science
andmembers
Technology
in istry
of Higher
to “improve
equals, with
different
getting
primary
authorship
on
uthowever,
their contribuThese
particles
interact
through
three Education—is
sityin
as1979)
a second
Thuwal,
which
boasts
state-of-the-art
labs
the
visibility
and
ranking
of
King
Abdulaziz
urels
stillaffiliation to his name on thepapers
forces:
the
electromagnetic
force
that
binds
that they individually led about different aspects of the work.
and
dozens
of
prominent
researchers
as
fullUniversity.”
But
he
says
KAU
also
hopes the
ofInstitute
all mass.for Scientific Information’s (ISI’s) In
the
atom,
the
strong
nuclear
force
that
binds
1993, the year before the team began taking those high-redshift
of highly
cited researchers.
time faculty members
(Science,
16 October
foreign academics
will help it kick-start indige,r list
the
Higgs
boson,
quarks
into
protons
and
neutrons,
and
the
(third) ceded SCP to Perlmutter (second).
observations
from the Cerro Tololo Inter-American
Observatory
in
“I thought
it wasName
a joke,”
says Kirshner,
2009,
enous research
legendary
status.
a kindprograms. “We’re not just givrecognition.
Peter Higgs was
onep.
of354).
six theo- weak nuclear force, which produces
who
forwarded
the
e-mail
to
his
department
But
the
initiatives
at
KSU
and
KAU
are
ing
away
money,”
he
says.
Most recruits will
ine that drives the rists to have the same idea.
of radioactivity. (The standard model does
Five living theorists have claims to having dreamed up the most famous Kingdom. Others question whether the
chair, noting
in
jest
that
the
money
was
a
lot
aimed
at
getting
speedier
results.
“They
are
be
expected
to
visit
for
a
total
of 4 weeks in a
advance
was a big
enough step beyond
presubatomic particle in physics. But what did they really do?
vious work to merit science’s biggest prize.
more attractive than the 2% annual raise pro- simply buying
names,”
saysa theorist
Mohammed Al- year to “give crash courses”; they will also be
In any case,
says Chris Quigg,
14NOW
SEPTEMBER
THAT THE HIGGS 2012
BOSON—OR VOL
standard337
model, the SCIENCE
Higgs boson itself is in awww.sciencemag.org
at Fermi National Accelerator Laboratory
fessors typically
get.
Then
discovered
that
Qunaibet,
a professor
of historians
agricultural eco- expected to supervise dissertations and help
something much
like it—is
in the he
bag, the
way the byproduct
of more important
under- (Fermilab)
in Batavia, Illinois,
Published
byof AAAS
question on many people’s minds is who lying
physics. Instead
the boiler on a steam and prize committees must be careful to give
a highly gets
cited
colleague
at another
U.S.it’sinstinomics
KSU,
KAU’s full-time
faculty members develop
the Nobel
Prize for the discovery.
locomotive,
more like the whistle:
Its toot at
credit
precisely who
where andrecently
for what it is due:criticized the
0
NEWSFOCUS
If you go by the pop history, the answer proves that the boiler is there and working, “If these people receive their rewards, either
tution hadis obvious.
accepted
anorarticle
hebewrote
for the leading research proposals. Even the “shadows” of
In 1964,KAU’s
Peter Higgs,offer,
a mild- adding
but it doesn’t KAU
turn the wheels. programsinin
heaven
before, it would
nice if it was
mannered theorist from the University of
As for whether the work of Higgs and col- for something they actually did and not for
as a second
affiinliation
on ISIhighlycited.com.
newspaper,
Aldid.”Hayat. Teddi Fishman, such eminent scholars would inspire local stuEdinburgh
the United Kingdom,
dreamed leagues merits a Nobel Prize, Saudi
opinions vary.
what people say they
up the particle to explain the origins of mass. “Certainly, yes, I think it is at least as proShiny. Kingdents
Abdulaziz
steps to gain
Kirshner’s
colleague
not
alone.
Sci-that havedirector
the Center for Academic IntegandUniversity’s
faculty members,
he says.
He completed
physicists’ standardis
model
of found
as other things
been given the of
The problem
particles and forces. Experi- Nobel in the past,” says Frank Close, a theo- What Higgs and company did, albeit unwitvisibility are controversial.
ence hasfundamental
learned
ofwithmore
than
rity
at
Clemson
in South
Carolina,
The
recruits Science
spoke to say they
menters
the world’s
largest
risttop-ranked
at the University of Oxford
in the
United
tingly, was toUniversity
dynamite
a huge boulder
that
Diercks
A.working
Filippenko
P.60
Garnavich
R.
Gilliland
C. Hogan
S. Jha
R. Kirshner
B. Leibundgut
atom smasher, the Large Hadron Collider
was blocking progress
-1
researchers
from
different
disci- says the programs deliberately
create “a false have a genuine
interest in promoting research
(LHC) at
the European
particle physicsscientific
labon the standard model
Online
tutions. All have changed their affiliation on
oratory, CERN, in Switzerland, have now
of particle physics.
sciencemag.org
plines—all
on
ISI’s
highly
cited list—who impression
that
these
universities
at
KAU,
even
though
none
of them knew how
seen that
particle
(Science,
13 July, p. 141).
The
standard model is are producISI’s highly cited list—as required by
KAU’s
Podcast interview
So Higgs gets the glory.
a quantum field theory,
with writer Adrian
have recently
signed
awhat
part-time
ing great Cho
research.”
theirsome
individual
research
plans
Only that’s
not exactly
happened. In employment
so it focuses on quancontract—and
have added
KAU as
an would match
(http://scim.ag/
fact, theorists say, Higgs made a fairly narrow
waves
or “fields”
pod_6100).6 APRIL
www.sciencemag.org
SCIENCE VOL
336
2012
27
arrangement
with
the
university
Academics
whotum
have
accepted
KAU’s
upresearch
with the
interests
abilities
of KAU’s
affiliation on
papers.
Otherand
requireand esoteric
advance
in mathematical
phys- that is structhat
describe the
probics. Several other physicists made the same Published by AAAS
ability of finding various particles here and
ments in the
contract
include and
devoting
“the Ray Carlberg,
tured along
lines
offer represent
wide
variety
of faculty
faculty
members
students.
advancethe
at the same
time. of
Their what
intellectualKirshner was
there. It containsafields
for the dozen
types of
was essential to the development of the
particles—including electrons, the up
whole of your
time,-2
attention,
and abili- of Toronto in
offered. leap
Meanwhile,
a most
bigger,
from elitematter
institutions
inmake
the
United States,
an astronomer
at skill
the University
standard
model, perhaps the
elaborate more promiquarks
and down quarks that
up proties
to
the
performance
your duties”
and precise theory in all of science. But their
tons and neutrons, and the heavier analogs
nent Saudi
Canada, ofEurope,
Asia,
Australia. All Canada whoofaccepted
the and
offer, says he had
papersinstitution—King
didn’t even mention the most impor-Saud Univerthose particles that
emerge in and
high-energy
doing “work equivalent to a total of 4 months
tant problem their work helped to solve. Other
particle collisions.
sity in Riyadh—has
climbed
several
hundred
are
men.
Some
are
emeritus
professors
who
to
Google
the
university
after he received the
scientists did that later—but their contribuThese particles interact through three
per
contract
period.”
(which won Nobel laurels in 1979) still
forces: the electromagnetic force that binds
places intion
international
rankings in the past have recently
retired from their home instie-mail. Hea admits
that
he was initially condoesn’t explain the origins of all mass.
the atom, the strong nuclear force that binds
Neil Robertson,
professor
emeritus
1958-2008
10
10
1
10
2
Number of authors per paper, x
10
3
Measures for Scientific Impact
II
.... a short baseball interlude...
Accounting for Inflation
× 20
Just as the price
of a candy bar
has increased by
a factor of ~ 20
over the last 100
years (roughly
3% inflation rate),
s demonstrates the utility of the detrending method. Annual
r pitchers), (C) home run, and (D) detrended home runs (for
t over the 90-year “modern era” period 1920–2009, however
e. This residual trend in the strikeout average may result
nd the increased role in the bullpen relievers, which affects
en season. This follows from the definition of the detrended
innings pitched per game might remove this residual trend
81 and 1994–1995
correspondfactors
to player strikes resulting
Institutional
Just asinthe price
y(t)" for these seasons.
of a candy bar
Accounting for Inflation
matter!
Home Run Prowess <P(t)>
has increased by
a factor of ~ 20
here P is the average of "P (t)# over the entire period,
over the last 100
years (roughly
2009
0.04
"mound (’62)
1 the
3% inflation rate),
Raising
P
≡
"P
(t)#.
(8) run
0.035
the
home
110 t=1900
0.03
hitting ability
PED
0.025
is players
arbi- has
he choice
of normalizing with respecterato P of
also increased by
rary, and
0.02 we could just as well normalize with respect to
a significant
(2000),
placing
all
values
in
terms
of
current
“2000
US
0.015
factor over the
ollars”, as is typically done
in
economics.
Lowering the mound (’69)
0.01
same period
x3
In Figure 4 we compare the seasonal average of "x(t)#
end of dead-ball era, emergence
of
0.005
D
o the prowess-weighted
average
"x
(t)#,
for strikeouts per
“Ruthian” power hitters
0
layer and
per player.
We
1900home
1920runs1940
1980 define
2000 "x(t)# as
1960
Year, t
A 900
B
800
700
600
500
400
300
200
1950 1960 1970 1980 1990 2000 2010
800
12
10
8
6
4
2
1940
1960
Year
≈ 600 pts / season
700
600
500
400
300
200
1950 1960 1970 1980 1990 2000 2010
Detrended league average < HRD >
League average < HR >
Baseball
D
14
1920
900
statistical baseline
Year
C
Detrended league average < PtsD >
League average < Pts >
Basketball
Accounting for socio-technological factors that underly achievement
1980
2000
Year
14
12
10
≈ 7 HRs / season
8
6
4
2
1920
1940
applying
detrending method
1960
1980
Quantitative measures
for success are
important for comparing
both individual and group
accomplishments, often
achieved in different time
periods.
However, the
evolutionary nature of
competition results in a
non-stationary rate of
success, that makes
comparing
accomplishments across
time statistically biased.
2000
Year
A. M. Petersen, O. Penner, H. E. Stanley.
Methods for detrending success metrics to account for inflationary and deflationary factors
Eur. Phys. J. B 79, 67-78 (2011).
Pre-print title: Detrending career statistics in professional Baseball: accounting for the Steroids Era and beyond
20 1960
1970
1980
1990
1960
2000
2
1970
1980
1990
2000
1960
20
Fighting output inflation in Science:
towards bibliometrics that account for
time-dependent productivity/impact factors
0
1980
1985
1990
1980
1995
2000
2005
1960
1970
1980
Year
1990
1990
0
1960
2000
1970
1980
1990
2000
Year
Year
PRL
1960
2000
Year
0
1975
1970
Year
Year
PNAS
3500
3000
Ave # citations Ave # coauthors
# papers
per paper
per paper
3000
2500
Np 2000
productivity inflation and
population growth
2000
Np
1500
1000
1000
500
0
0
1960
1970
1980
1990
2000
1960
1970
Year
1980
1990
2000
Year
6
15
5
80
4
60
Ave_NA 10
growth
of team science
3
4
SD_NA 40
Ave_NA
3
SD_NA
2
2
5
20
1
1
0
0
1960
1970
1980
1990
2000
0
19601960
0
19701970
Year
1980 1980
1990 1990
2000 2000
1960
1970
Year Year
1980
1990
2000
Year
250
80
150
1500
200
60
100 150
Ave_C
SD_C
Ave_C
40
SD_C
100
50
20
500
50
0
0
1960
1970
1980
1990
2000
0
19601960
Year of publication
19801980
19901990
1960
20002000
25
1500
1970
1980
Year
35
30
30
0
19701970
YearYear
Year
40
time-dependent citation rate
depends strongly on the (sub)
discipline, the era, and the
age of the paper
1000
1200
1000
1990
2000
1970
20 1960
1970
1980
1990
1960
2000
2
1970
1980
1990
2000
1960
20
Fighting output inflation in Science:
towards bibliometrics that account for
time-dependent productivity/impact factors
0
1980
1985
1990
1980
1995
2000
2005
1960
1970
1980
Year
1990
1990
0
1960
2000
1970
1980
1990
2000
Year
Year
PRL
1960
2000
Year
0
1975
1970
Year
Year
PNAS
3500
×4
Ave # citations Ave # coauthors
# papers
per paper
per paper
3000
Np 2000
×4
3000
2500
2000
Np
productivity inflation and
population growth
1500
1000
1000
500
0
0
1960
1970
1980
1990
2000
1960
1970
Year
1980
1990
2000
Year
6
15
5
80
Open Access Journals
4
60
Ave_NA 10
growth
of team science
3
4
“[Acceleration of scientific progress
via fast peer-review/publication]”
SD_NA 40
Ave_NA
3
SD_NA
2
2
5
20
1
1
0
0
1960
1970
1980
1990
2000
Year
0
19601960
0
19701970
1980 1980
1990 1990
2000 2000
PLoS One:
~ 6,700 articles in 2010 and ~ 14,000 in 2011
× 2 growth in one year alone!
... who is reading/refereeing all these papers??
1960
1970
Year Year
1980
1990
2000
Year
250
80
150
1500
200
60
100 150
Ave_C
SD_C
Ave_C
40
SD_C
1000
100
50
20
500
50
0
0
1960
1970
1980
1990
2000
0
19601960
Year of publication
19901990
1960
20002000
25
1500
1970
1980
Year
35
30
30
19801980
YearYear
Year
40
0
19701970
1200
1000
1990
2000
1970
ogy, business and management, research of collaboration. Micro-level studies can be cess), and why (mod
development, biomedical and life sciences, quantitative and, if considering network 1 presents key insigh
and physical sciences. The increasing inter- analyses, involve many attributes for nodes ing these differing le
est in professional gatherings centered on and linkages. Other methods include indiEach level of team
SciTS combined with recent progress in vidual-level analysis of researchers partici- of challenges. Macr
SciTS research and practice suggest
that pating
within
teams in which members
are dress organizational
K. Börner,
et al.
A multi-level
systems
this community is coalescing into its own queried about their experiences as team ing culture that eith
perspective
for the science of team science.
area of inquiry.
members (13, 14).
collaboration and in
Sci. Transl.Each
Med.
2, 49cm24
(2010).
of these
levels addresses
different lenges at the mesoMULTI-LEVEL, MIXED-METHODS
issues that can be roughly classified into ing the group dynam
APPROACH FOR SCITS
science as well as ho
The burgeoning field of SciTS can serve as a
and train teamwork
transformative melting pot of existing thethe micro-level (the
ories and scientific techniques. We propose
tightly intertwined
a multi-level, mixed-methods approach
meso-level issues, a
that can serve as a framework capable of
how individual scie
organizing the diverse forms of inquiry and
in the scientific aspe
interlink research on individual scientists,
process of innovati
teams, and populations of teams (Fig. 1).
in communication a
Researchers working at different levels
Table 2 lists key cha
study different facets of the team science
addressed within th
ecology, contribute different theories and
techniques, and generate diverse findings.
MOVING FORWAR
Each level might analyze different data; use
We conclude with
multiple approaches, techniques, and visual
more general challen
representations; and provide different insurrounding SciTS.
sights. The combination of insights from all
to SciTS is conduct
levels is considerably larger than their sum.
tings—academic and
First, “macro-level” research examines
ogy development, a
teams at the population level and leads
As such, the variety
to insights about patterns of collaboralished, approaches a
tion that are broad in both their amount
data produced is im
and their form, and that provide input on
more than 180 cor
how to measure the growth and effect of
that convey key resu
knowledge. Macro-level studies might use
search. Of those pap
terabytes of data that require large-scale
between 1944 and 2
computing infrastructures to process and
der being published
communicate results. Recent work coming a surge of activ
bines computational, behavioral, organizathe reported studies
tional, and other methodological approachlication data sets (s
es to derive new insights at this broad level.
by Thomson Reuters
Second, “meso-level” research increases
and most tools are
our understanding at the group level, exdifficult to replicate
amining, for example, how interaction patjournal publications
terns, the nature and amount of intra-team
ings, and book cha
communications, and the composition of
and grant awards, a
the team contribute to team process and
ly collected across
outcomes. Such approaches can use netstudied are typically
work analysis—the representation of data
although science is
as nodes and their interlinkages—to study
tilingual. Furtherm
the evolution and impact of (social) netdata records (such
work structures at varied time scales or anall papers by one sch
alyze the specific quality and type of interent databases) and t
Question: how to measure
scientific output and impact
at various scales while accounting
for systemic heterogeneity
* Country
* Institution
* Lab / Team
* Paper
/SCIENCE TRANSLATIONAL MEDICINE
* Individual
Measures for Scientific Careers
Quantifying impact and productivity in science
“Math-letes”
III
American Physical Society
Using “big-data” to better understand
academic careers
Log in | Create Account (what's this?)
RSS Feeds | Email Alerts
Home
Browse
Search
Subscriptions
Vol.
Citation Search: Phys. Rev. Lett.
Help
Page/Article
APS » Journals » Phys. Rev. Lett. » Volume 42 » Issue < Previous Article | Next Article
10
>
Phys. Rev. Lett. 42, 673–676 (1979)
Scaling Theory of Localization: Absence
of Quantum Diffusion in Two
Dimensions
Abstract
References
Download: PDF (622 kB) Buy this article
Citing Articles (2,099)
Physics - spotlighting exceptional research
For example, P. W. Anderson:
n = 64 articles
published in PRL over this
51-year period
METHODS FOR MEASURING THE CITATIONS AND…
Read the latest from Physics:
Viewpoint: In pursuit of a nameless metal
Viewpoint: Spin Hall effect goes electrical
Trends: The quest for dilute ferromagnetism in
Page Images
semiconductors: Guides and misguides by theory
Export: BibTeX or EndNote (RIS)
TABLE I. Summary of data set size for each journal. Total number N of unique !but possibly degenerate" name identifications.
E. Abrahams
Serin Physics Laboratory, Rutgers University, Piscataway, New Jersey 08854
*
P. W. Anderson , D. C. Licciardello, and T. V. Ramakrishnan
†
Joseph Henry Laboratories of Physics, Princeton University, Princeton, New Jersey
08540
Received 7 December 1978; published in the issue dated 5 March 1979
Arguments are presented that the T=0 conductance G of a disordered electronic system
depends on its length scale L in a universal manner. Asymptotic forms are obtained for
the scaling function !(G)=dlnG/dlnL, valid for both G≪Gc!e2/
and G≫Gc. In three
Journal
Years
Articles
Authors, N
CELL
NEJM
Nature
PNAS
PRL
Science
1974–2008
1958–2008
1958–2008
1958–2008
1958–2008
1958–2008
53290
17088
65709
84520
85316
48169
31918
66834
130596
182761
112660
109519
dimensions, Gc is an unstable fixed point. In two dimensions, there is no true metallic
behavior; the conductance crosses
over smoothly from logarithmic or slower to
exponential decrease with L.
© 1979 The American Physical Society
URL: http://link.aps.org/doi/10.1103/PhysRevLett.42.673
DOI: 10.1103/PhysRevLett.42.673
*Also at Bell Laboratories, Murray Hill, N. J. 07974.
†
On leave from Indian Institute of Technology, Kanpur, India.
< Previous Article | Next Article >
Our data collection procedure begins with downloading
all“Methods
“articles” for
formeasuring
each journalthe
forcitations
year y from
and ISI Web of
Knowledge.
From
the set of N!y"
articles
each
particular
productivity
of scientists
across
timeforand
discipline”
journal and year, we calculate #c!y"$, the average number of
A. M. Petersen, F. Wang, H. E. Stanley. Phys. Rev.
citations per article at the date of data extraction !May 2009".
E, 81
(2010)
036114
Each
article
summary
includes a field for a contributing author’s name identification, which consists of a last name and
pe
de
th
la
th
pu
a) Paper arena
Log-normal citation distribution w/in journal (subfield)
accounting for (paper) age cohorts
Radicchi F, Fortunato S, Castellano C. Proc Natl Acad Sci USA 105, 17268–17272 (2008)
gle author within a
account
in number
of coautho
theory the
and variability
empirical data
for career
longevit
b) Career
“The
1%”
ievements combined
botharena:
within
across
Fig.art
3
For
eachand
author,
we discipline
combine all!see
his/her
PETERSEN, heavy-tailed
WANG,
STANLEY
REVIEW
E 81,we
036114propose
"2010#
define
n asANDthe
total citation
To remedy
thesePHYSICAL
problems,
distribution
w/in
career
journal.
Specifically,
a publication
career
in at
a given journal over
metric
citation
shares,
which
normali
to thetermed
lifetime
achievements
of ain single
a
Each
author
has
n articles
a
e traditional citation~ 1%ci!y"
of paper
i by
#c!y"$,
the
average
number
β = 0.5
single
journal,
and
not
the
lifetime
achievem
given
journal
j.
cess/impact
CELL within a
papers
a given
journal analyzed.
in year y,We
anddefine
divid
amongin
the
six journals
NEJM
CELLi, published in year y,
Each
article
NEJM
PNAS
received
by
ations ci Nature
Nature
c
!y"
/
#c!y"$
into
equally
distributed
shares
am
number
of
papers
for a given
author
in a giv
i
PRL
PNAS
can
be
quantified
by
the
number
Science
PRL
Science
authors.
Dividing
theInshares
equally
obv
the
50-year
period.
analogy
with will
the trad
Completed Careers 1958-2008
of of
citations
at con
Careersc
1958-2008
i it has
tally,
one
can
thereceived
career
success/im
the
value
theAllcalculate
efforts
made
by greater
the
time
ofthe
data
(a)
Rank,together
r extraction.
Total citations, C
given the
journal
by of
adding
the citations
raising
value
efforts
made
by less
!1"
2009)
FIG. 5. "Color online#(May,
The Zipf
plot emphasizes the stellar cathe n papers,
Without
more
accurate
reporting
schemes
o
reers corresponding to large C , the total number of citation
shares
~ 1%
Two
possible
toEq.measure
within
a particular
journalways
defined in
"2#, and showscitations:
a signifieach authors’
contributions
!as
is
now imple
cant scaling regime corresponding to the top-ranking
“champions”
n
α
of each journal. For comparison, we list the top 20 careers within
ζ
≈ 2.5
CELL
Nature
and
PNAS",
dividing
shares
equa
eer productivity
of
a
(i)journals
Total
citations:
the
CELL,
NEJM, and
PRL in Table III.the
The total
number
NEJM
c
.
C
=
Nature
i journal
of career citation shares for a particular author '
in a given
reasonable
method
given
the
available
s the totalPNAS
number
P
serves as a proxy for the career success of the scientist.
PRL
i=1 The statis- data.
Science
tical regularity in the rank ordering of scientific achievement exculate the
career
citation
shares a
A main point
raised in
tends normalized
over four orders of magnitude.
The similarity
in scaling exCompleted Careers 1958-2008
ponent
among
the
journals
analyzed
possibly
suggests
that
there
are
(ii) Totalone
citations
``shares”:
Furthermore,
can
calculate
the career pr
fundamental
forces
governing
success
in competitive arenas such as
tation metrics which
high-impact journals. For visual clarity, wenplot the power law with
Total citation shares, C
(b)
given
author
within a specific1journal
ci!y" as the
scaling exponent % ( 0.5.
tion of citation accuCs = '
FIG. 4. "Color online# We estimate the career successof
of a papers
scipublished within
the journal. .A main
larger number of stellar careers than would be
expected
from
entist within a given journal using the citation shares metric C
a
#c!y"$
i
i=1
defined in Eq. "2#, which accounts for both the number ofthis
authors paper
the number
papers.
is oftostellar
discount
the value of citation
eandmore
citations
the age of the
paper. "a# PDF ofthan
total raw citations C according
Another illustrative method for comparing the distribution
10
-4
10
-6
10
10
3
10
-2
total citation shares, Cs
P(C)
10
-8
1
10
0
10
-1
10
-2
10
-3
10
-4
-10
0
10
10
2
10
1
10
2
10
3
4
10
10
10
5
10
0
10
1
10
10
2
3
10
2
s
10
0
P( C s )
-2
10
-4
10
-6
10
-8
10
-2
10
-1
10
0
10
1
2
10
10
3
10
s
s
4
10
5
10
6
10
< τ(n) > / < τ(1) >
served in the pdf of the number of authors per paper "Fig. 3#.
an arbitrary combination of stellar papers and cons
Hence,
the
decay
in
the
curve
for
PRL
which
approaches
cess, as demonstrated
III. In all, the statis
Empirical evidence for the Matthew “rich-get-richer”
effectininTable
Science
zero might be due to the project leaders at large experimental
larity found in the distributions for both citation
Gospel of St. Matthew: “For to all those who have, more will be
given.”
Still lend
true 2000
years
paper
shares
naturally
to later!!
methods based o
1
statistics in order to distinguishing stellar care
methods have been developed for Hall of Fame ca
0.8
baseball %16,34&, where statistical benchmarks
lished using the distribution of success.
Statistical physicists have long been interested i
0.6
interacting systems, and are beginning to succeed
CELL
NEJM
ing social dynamics using models that were develo
0.4
Nature
context of concrete physical systems %35&. This s
PNAS
PRL
spired by the long term goal of using quantitativ
0.2
Science
from statistical physics to answer traditional questi
Completed Careers 1958-2008
in social science %36&, such as the nature of co
0 0
1
2
success, productivity, and the universal features
10
10
10
th
Author’s n paper
activity. Many studies begin as empirical descrip
as the studies of common mobility patterns %37&
FIG. 7. "Color online# A decreasing waiting time !"n# between
%38,39&, and financial fluctuations %40&, and lead
publications in a given journal suggests that a longer publication
understanding of the underlying mechanics. It is po
career "larger n# facilitates future publications, as predicted by the
the empirical laws reported here will motivate usef
Matthew effect. We plot !!"n#$ / !!"1#$, the average waiting time
tive theories of success and productivity in comp
!!"n#$ between paper n and paper n + 1, rescaled by the average
vironments.
waiting time between the first and second publication, !!"1#$. The
values of !!"1#$ are 2.2 "CELL, PRL#, 3.0 "Nature, PNAS, Science#,
and 3.5 "NEJM# years. Physical Review Letters exhibits a more
rapid decline in !"n#, (1)
reflecting the
rapidity
of (4)
successive publica(2)
(3)
(n)
tions "often by large high-energy experiment collaborations#, which
is possible in this high-impact letters journal.
τ
1
τ
2
τ
3
…τ
τ
4
5
…n
•
For a given journal:
the waiting time τ(n)
is the number of years
between an author’s
paper n and paper n+1
•
A decreasing〈τ(n)〉
indicates that it
becomes “easier” to
publish in a journal
with each successive
ACKNOWLEDGMENTS
publication
We thank S. Miyazima and S. Redner for he
ments and the NSF for financial support.
036114-8
career age, t
Persistence vs Uncertainty
Can a quantitative picture of career dynamics shed
light on the saying: “publish or perish” ?
Longitudinal career data:
Set A: 100 most-cited physicists, average h-index〈h〉= 61 ± 21
Set B: 100 additional highly-prolific physicists,〈h〉 = 44 ± 15
Set C: 100 current assistant professors from 50 US physics depts.,〈h〉 = 15 ± 7
Set D: 100 most-cited cell biologists,〈h〉 = 98 ± 35
Set E: 50 highly-cited mathematicians,〈h〉 = 20 ± 10
ball AssociationWe
(NBA)
careers
during
the
63-yea
model the
career aspublicatio
an aggre
BA) careers
during
the
63-year
period
The career trajectory in science:
1946–2008.
portunities which arrive at
theadva
vari
tive
a tale ofWe
knowledge,
collaboration,
and
reputation
spillovers
model the
the
career
as
an
aggregation
of
out
reputation
of
aopscientist
is
typica
Annual
production
of
individual
i
a as
reer
an
aggregation
of
output
tional
me
portunities which
arrive
at
the
variable
rate
n
(t
i
resentation of number
his/her
contributions
!
of
publications
in �
year t e
rrive at
variableofrate
ni (t). isSince
thethe
reputation
a scientist
typicallyinternal
a cumulat
t
mulative
production
N
(t)
≡
� =1
i
t
Cumulative
production,
a
proxy
for
career
reputation
resentation
of his/her
contributions,
we
considerd
cientist
is typically
a cumulative
repinternal
�t Fig. 1 �shows th
career
achievement.
!=1
cent
er contributions,
we
consider
mulative production
Ni (t)the
≡ cu) ascase
ap
� =1 ni (t
t
tion� Ni (t) of six notable careers wh
�t Physics
Fig.
1 shows
the
cumulative
for many
winners
h
α
n Ni (t)career
≡ achievement.
n
(t
)
as
a
proxy
for
i
�
i
relation
N
(t)
≈
A
t
.
However,
th
t =1
prolific careers!
i
i
tion
N
(t)
of
six
notable
careers
which
display
a
i
year,
t
publicatio
Fig.
1 shows the cumulative
producNi (t), see
Fig.
S1, which
b relation
reputation,
and do not
αi knowledge,
Nα i (t) ≈ Ai t .collaboration
However,
there
aretrigge
also
spillovers
contribute
can
table careers which
adisplaying
scaling marked non-s
ity,display
instead
to the increasing
returns
across such
Ni (t), see Fig. S1, which
do not
exhibit
αi
cascading
!>1
α =are
1
linearity
arising
from
significant
exo
t . However,
there
also
cases
of
the academic
career
!=1
ity, instead displaying
marked non-stationarity a
Collabo
We justify
this
2-parameter
model
which
do
not
exhibit
such
regularCumulative
advantage:
Successful
linearity arising
from significant exogenous career
Dataset A
leaders
become
“attractors”
of fluctu
scaling
methods
and
data
collapse
(
Dataset
B
ing
ng marked
non-stationarity
and
nonDataset
C
We justify this
2-parameter
model in the SI tex
new opportunities
Math
to
show
that
most
N
(t)
can
be
mod
i
the
abilit
m significant
exogenous
career
shocks.
scaling year,
methods
and data collapse (see Figs. S2
year, tt
functional
form.
Careers
with
α
lated
to
arameter
model
the NSI(t)text
to show
thatinmost
can using
be modeled by thisi c
10
Anderson, PW
Barabasi, AL
Herrmann, HJ
Leggett, AJ
Polkovnikov, A
Wilczek, F
2
i
Ni (t)
10
3
10
10
1
0
10
10
< NiN
(t)i (t)
/ <ni> >
10
0
10
1
10
2
2
3
Wiles, A : 0.91(3)
Erdos, P : 1.66(2)
i
1
10
2
10
0
101
10
-1
0
10
10
0
10
10
0
10
110
1
2
102
10
“Persistence and Uncertainty in the Academic Career,”
A. M. Petersen, M. Riccaboni, H. E. Stanley, F. Pammolli.
Proc. Natl. Acad. Sci. USA 109, 5213-5218 (2012).
Publication trajectory, ⟨N!(t)⟩
the characteristi
N (t)
c(r,put
t). in
InSI.
order to extracting
thetogether
characteris
andr=1
[E]
and
Are there characteristic career growth patterns? the
jectory of
C(t),
we
factor
out
the
scale
�c
�
w
i
_
each
dataset,
�
Ñ
_
Collaboration
2
10
c
!
or
each of 350 scientists,
put(normalized)
in networkand averageα the
considerably
across
scientists,
[A/B] 1.30(1) 1.5
�
production �N (t)� ~
= t�Ñ (t
[C]
1.15(2)
ons
to
other
factors
trajectory
[D] 1.55(1)
jectories
C̃
(t)
≡
C
(t)/�c
�
for
the
A
scientis
i
i
i
[E] 1.01(1)
�A
This regularit
10
each dataset, �C̃(t)�
≡ (normalized)
1
i_�. F
αii(t)/A�c
> 1 correspo
i=1 C
ζ
cumulative
�
a scaling
citation �C
strates
the
super-linear
(t)�
=
celeration
is�C̃(
con
~
t
10
10
10
trajectory
_
DUCTION
i knowledge and p
10
"
3
1
0
0
1
Citation trajectory, ⟨C!(t)⟩
4
3
10
ζ > α >1 ⇒ increasing returns
[A/B] 2.52(1)
[C] 2.42(4)
[D] 2.65(1)
[E] 1.39(3)
nal 10career dataset covering 350
How strong are the reputation
3,693
papers
and
7,577,084
ci10
1
spillovers which manifest in a
b tipping point whereby careers
aper10years.
become “attractors” of new
2
Open questions:
1
0
100
101
career age, t
Citation network
opportunities
instead of “pursuers”
B.
Co-Evolution of the Scientific Production Function
IV
Collaboration
i
Publication/
Citation
Knowledge
A
Complexity
- coevolutionary system
- behavioral components
The holy grail:
a comprehensive disambiguated career portal for the entire scientific labor force
Online user-input repositories
non-proprietary: orcid.org
integrated
proprietary: researcherid.com
- also integrates grant/funding info
Potential problems:
- honor system
- require constant updating
- might only serve to reinforce the “rich-get-richer” effect in science
- How to quantify the other productivity outputs associated with an
academic career (teaching, community outreach, press/online
coverage
, etc.)
Physiological/Behavioral components of games
High competition levels can make careers vulnerable to early career negative production
shocks (ie stress, burn-out, productivity lulls, etc.)
Achievement-oriented systems: incentives for cut-throat “zero-sum” behavior, i.e. use of
performance/cognitive enhancing drugs, blatant cheating and falsification
Sudden career termination in
science due to ethical scandals
Jan Hendrik Schön Scandal (2001)
On October 31, 2002, Science withdrew eight papers written by Schön
On December 20, 2002, Physical Review withdrew six papers
On March 5, 2003, Nature withdrew seven papers
Diederik Alexander Stapel Scandal (2011)
Social psychologist made up data for at least 30 publications according
to preliminary investigation, which is still ongoing.
Hisashi Moriguchi Scandal (2012)
“Transplant of induced pluripotent stem cells to treat heart failure
probably never happened.... He is affiliated with University of Tokyo
but not with Massachusetts General Hospital nor with Harvard
Medical School. The study did not receive Institutional Review
Board approval.” nature.com
Vol 450|20/27 December 2007
CognizantCOMMENTARY
Enhancement Drugs (CED)
The use of cognitive-enhancing drugs by both ill and healthy individuals raises ethical questions that
NATURE|Vol 452|10 April 2008
NEWS
should not be ignored, argue Barbara Sahakian and Sharon Morein-Zamir.
T
oday there are several
drugs on the December 2007
NATURE|Vol
450|20/27
market that improve memory, concen-
tration, planning and reduce impulsive
behaviour and risky decision-making, and
many more are being developed. Doctors
already prescribe these drugs to treat cognitive disabilities and improve quality of life
for patients with neuropsychiatric disorders
and brain injury. The prescription use of such
drugs
is beingNature
extended
to other conditions,
In January,
launched
an informal survey into readers’ use of cognition-enhancing drugs. Brendan
including shift-workers. Meanwhile, off-label
Maher
has
waded
through
the
results and found large-scale use and a mix of attitudes towards the drugs.
and non-prescription use by the general public
is becoming
increasingly
commonplace.
he US National Institutes of Health is to prescribed for cardiac arrhythmia that also behind ‘other’ which received a few interesting
Although
the appeal
of pharmaceutical
cog-have an anti-anxiety effect. Respondents who reasons, such as “party”, “house cleaning” and
crack down
on scientists
‘brain doping’
nitivewith
enhancers
— to help one study
longer,had not taken these drugs, or who had taken “to actually see if there was any validity to the
performance-enhancing
drugs such
work
more
effectively
or better
manage
as Provigil
and
Ritalin, a press
release
declaredeve-them for a diagnosed medical condition were afore-mentioned article”.
last week.
The release,
brainchild of evolutionOur question on frequency of use, for those
ryday
stresses
— is understandable,
potentialdirected straight to a simple questionnaire
ary biologist
Jonathan
of themust
University
users,
both healthy
andEisen
diseased,
considerabout general attitudes. Those who revealed who took drugs for non-medical purposes,
of California,
turned
to be an
the
pros andDavis,
cons of
theirout
choices.
ToApril
enablethat they had taken these drugs, or others, for revealed an even split between those who took
Fools’scientists,
prank. Anddoctors
the Worldand
Anti-Brain
Dop- non-medical, cognition-enhancing purposes them daily, weekly, monthly, or no more than
this,
policy-makers
ing Authority
website
that itto
linked
to was likeonce a year. Roughly half reported unpleasant
should
provide
easy access
information
about
wiseadvantages
fake. But with
a number
side effects, and some discontinued use because
DRUG SOURCES
the
and
dangersofofco-conspirators
using cognitivespreading rumours about receiving anti-doping
of them. Some might expect that negative side
Answered
question:
201
enhancing drugs and set out clear guidelines for
Vol 456|11
2008
affidavits with their first R01 research grants,
effects would correlate positively
with aDecember
low
Skipped question: 1,227
their
future
Togave
trigger
broader
discussion
frequency of use, but that doesn’t seem to be
the ruse
no use.
doubt
pause
to a few
of the of
these
issues we
offer thesurvey
following
questions, to
respondents
to Nature’s
on readers’
the case in our sample (see bar graph, below).
which
readers can responddrugs.
in an online forum.
use of cognition-enhancing
Reported side effects included headaches, jitPrescription
Now,
to thewas
questions.
Morning
pick-me-up: will drugs
that help you stay
alert become
widely
acceptable as coffee?
Internet
Theon
survey
triggered by a Comteriness,
anxietyas
and
sleeplessness.
(52%)
NATURE|Vol 452|10 April 2008
Poll results: look who’s doping
“ Is it cheating to use cognitive-enhancing
drugs?.... How would you react if you knew
your colleagues — or your students — were
taking cognitive enhancers?... we know that
a number of our scientific colleagues ...
already use modafinil [Modiodal, Provigil]
to counteract the effects of jetlag, to enhance
productivity or mental energy, or to deal
with demanding and important intellectual
challenges...”
NIKOS/ZEFA/CORBIS
Professor’s little helper
PHOTOTAKE/ALAMY
encouraged, such as by air-traffic conrs, surgeons and nurses who work long
. One can even imagine situations where
enhancing-drug-taking would be recomed, such as for airport-security screeners,
soldiers in active combat.
are
“One inBut
fivethere
respondents
said they had used
raightforward answers
fruitful reasons to stimulate
drugsand
forany
non-medical
e must address eachtheir
situation
inconcentration
turn.
focus,
or memory. Use did
not differ greatly across age-groups..., which
“...one survey estimated that almost 7%
would you react if will
you knew
your
surprise
some. “
of students in US universities have used
mentary by—
behavioural
neuroscientists
Neuroscientist Anjan Chatterjee of the
agues
or
your
students
—
were
Barbara adults
Sahakian
andsevere
Sharonmemory
Morein- and
University
Pennsylvania
inneuropsychiatric
Philadelphia
additional neuro-protective effects.
drug. of
For
patients with
disorShould
with
prescription stimulants [Adderall and
Zamir
of the University
of Cambridge,
predicts
a rise
the useofofthe
these
drugs
andbe weighed
problems
from neuropsyCountries with ageing populations are seeing
ders,
thein
benefits
drugs
must
g concentration
cognitive
enhancers?
UK, whodisorders
had surveyed
their colleagues
other neuroenhancing
products
and procechiatric
be given
cognitivea surge in the number of people with Alzheimagainst the potential
short-term
and long-term
Ritalin] in this way, and that on some
on the use of drugs?
drugs that purportedly enhance
dures asside
theyeffects,
becomeand
available
Chatterjee
enhancing
er’s. The personalof
andour
social costs are staggerthese(A.
factors
should be discing
drugs.
Brendan
ademia,
we(Nature
know
thatyes.a number
focusbelieve
and attention
1157–1159;
Cam. Q.cussed
Healthc.
Ethics
16, 129–137;
2007).to ascertain
We
the answer
is 450,
a resounding
ing and in the United Kingdom, economic
with
the individual’s
doctor
2007).
Indebilitating
the article, aspect
the twoofscientists
asked
Like thethe
riselevel
in cosmetic
surgery,risk
usein
of each
cogni-case.
campuses, up to 25% of students had
A
large
manythe
neuropsycosts associated
with dementias
are estimated
of acceptable
tific
colleagues
in
States
and
readers
whether they would consider
“boost-United
were asked several
additional
questions about tive enhancers is likely to increase as bioethical
chiatric
disorders
is
cognitive
impairment.
to
rise
to
£10.9
billion
(US$22
billion)
by
2031.
des
towards
the
drugs.
ing their brain power” with drugs. Spurred by their use. Here’s what they had to say:
and psychological concerns are overcome (see
used them in the past year. These
Thus,
cognitive-enhancing
drugs
useful One
According
to a reportsaid
commissioned
the If drugs
canand
beas
shown
to havegain
mild side
nited
Kingdom
already
use
the tremendous
response, Nature
ranare
itsaown
in fivemodafinil
respondents
they had used by ‘Worrying
words’)
the products
therapy
option
for
several
disorders,
includAlzheimer’s
Research
Trust
in
Cambridge,
UK,
should
they be prescribed
informal survey. 1,400 people from 60 coun- drugs for non-medical reasons to stimulate culturaleffects,
acceptance.
One difference,
Chatterjee more
widely
forcognitive
other psychiatric
disorders?
ing
andpoll.
Attention
that
wouldorreduce
severe
cognitive
students are early adopters of a trend
triesAlzheimer’s
respondedthe
todisease
the online
theirtreatment
focus,to
concentration
memory.
Use did
says, is that
use of
enhancers doesn’t
unteract
effects
ofDeficit
jetlag,
enhance
Hyperactivity
Disorder
(ADHD).
impairment
in
older
people
by
just
1%
a
year
We
believe
that
cognitive-enhancing
We
asked
specifically
about
three
drugs:
not
differ
greatly
across
age-groups
(see
line
rely
on
training
of
medical
specialists
such
which
received
a
few
interesting
Society must respond to the growing demand for cognitive enhancement. That response must startasby drugs
Alzheimer’s
is
a energy,
has surprise
been
estimated
to cancel
with
minimal
side effects
would
also benefit
methylphenidate
(Ritalin),
a stimulant
nor- graph,
left),
which
will
some. Nora
surgeons.
Internet
availability
will also
greatly
that is likely to grow, and indications
uctivity
ortreatdisease
mental
deal
with
“The
chief or
concern
rejecting
idea
that
is director
ato
dirty
Henry
colleagues.
mally
used to the
attention-deficit
hyper- Volkow,
ofword,
the National
Institute
on Greely
accelerate
use,
he
disease
of ‘enhancement’
out
all argue
predicted
increases
in and
many
of says.
the patients with schizophrenia, a
asneurodegenerative
“party”,
“house
cleaning”
and
cautioning
against
activity
disorder
butcharacterwell-known on
college Drug
Abusethe
(NIDA)long-term
in Bethesda,
Our poll
found that
one-third
thenot
drugs
the
ageing
mind
careMaryland,
costs due to the
condition
for which
theyofare
yet routinely
nding
and
important
intellectual
chalsuggest that they’re not alone.”
being used
for non-medical
purposes
were affects
campuses
a ‘studyinaid’;
modafinil (Provigil),
says that household
surveys
that stimubyoday,
a as
decline
cognitive
ageing suggest
population
.
prescribed.
Currently,
the disorder
use
of
medications
is
on was
university
campuses
around
e ized
if
there
any
validity
to
the
prescribed
to treatfunctioning,
sleep disorders but also lant use is highest in people
aged
18–25 years,
purchased
over24
themillion
Internetpeople
(see pieworldwide.
chart). The
and
behavioural
For
all
medications,
the
chief
about
thegraphic
world,
students
are striking
deals
toandModafinil
s (see
opposite).
and
adverse
effects.”
usedin
off-label
to combat
general
fatigue
or side
in
students.
rest were obtained
pharmacies
on pre- and social
NATURE|Vol
456|11
December
As with from
Alzheimer’s,
theor
personal
and
particular
learning
and drugs
concern cautioning against
buyjet
and
sell
prescription
such as For those2008
ned
article”.
overcome
lag;
and
beta
blockers,
drugs
who
choose
to
use,effects
methylpheniscription.
It isare
unclear
whether
theeconomic
prescribedcosts in the
memory.
There
are,
at
present,
no
treatments
their
use
is
adverse
side
that
affect
the
costs
immense,
with
Adderallare
and Ritalin
— not to get high,
but to date
drugs
available
online,
but
their
nonwas the
most popular:
62% of users
for
Alzheimer’s
disease
that can
stop
orover
reverse
individual’s health and well being. These may United States estimated in the tens of billions of
get
higher
grades,
to provide
an
edge
their
reported
taking it. 44% reported taking
on
on
frequency
of
use,
for
those
TRENDS
IN
USE
OF
NEUROENHANCERS
EFFECT OF DOSE ON SIDE EFFECTS
the
decline
in
brain
function,
range
from
mild,
temporary
fellow
students
or to
increasebut
in cholinestesome measand
15% said
they hadphysical
taken symp- dollars . It is common knowledge that people
ription
and
long-term
usemodafinil,
has
not
been
T
g COMMENTARY
(34%)
Pharmacy
(14%)
Towards responsible use of cognitiveenhancing drugs by the healthy
1
C. GALLAGHER/SPL
T
1
2
B
search volume
Deconstructing the h-index
Google Trends
100
“h-index”
80
60
40
20
0
12/2005
12/2006
12/2007
12/2008
12/2010
12/2011
12/2012
J. E. Hirsch,
“An index to quantify an individualʼs scientific research output”.
Proc. Natl. Acad. Sci. USA 102, 16569- 16572 (2005).
Google Citations = 2,653
｛
* ~ Nov. 2005
12/2009
citations, c(r)
Quantifying scientific achievement:
productivity -vs- impact
“A scientist has index h if h of his or her Np papers
have at least h citations each and the other (Np-h)
papers have ≤ h citations each.”
J. E. Hirsch, “An index to quantify an individualʼs scientific research output”. PNAS 102, 16569- 16572 (2005).
Prof. Ioannis Pavlidis, U. Houston
Google Scholar Citations profile
β
Ch ~ h2
140
120
y=x
h = 25
100
102
80
101
60
40
y=x
β = 0.54 ± .03
Np = 104
h = 25
20
0
100
0
20
40
60
paper rank, r
80
100
100
101
paper rank, r
102
Regularities in the rank-citation profile ci(r)
4
10
Top-cited physicists
citations, c(r)
y = px
10
hp-index line
ci(r) is the rank-ordered
h1-index line
3
(Zipf) citation distribution
of the N papers published
by individual i in his/her
entire career
2
10
10
1
H1 : y = x
Witten, E (121)
Gossard, AC (108)
Cohen, ML (107)
Anderson, PW (103)
0
10 0
10
1
10
paper rank, r
10
2
A. M. Petersen, H. E. Stanley, S. Succi. “Statistical
regularities in the rank-citation profile of scientists.”
Scientific Reports 1, 181 (2011).
3
10
Interestingly, even the very
top scientists have a
significant number of papers
that go relatively un-cited.
Accounting for the time-dependence of ci(r)
FEINBERG, A, VOGELSTEIN, B.
A TECHNIQUE FOR RADIOLABELING DNA RESTRICTION ENDONUCLEASE FRAGMENTS TO
HIGH SPECIFIC ACTIVITY. Anal. Biochem. (1983) Citations [ISI] = 21,270 in 2010 (ave. rate 760 /year)
citations, c(r)
10
10
P.W. Anderson
B. Vogelstein
4
10
10
5
H80
3
2
h-index line
1
10
10
3
τi > 9 years
(h = 103)
(h = 176)
citation rate, c(r) / τ
10
⤷
10
2
10
10
1
P.W. Anderson
B. Vogelstein
0
H1
0
10
0
τi
10
1
paper rank, r
10
2
3
10
10
-1
10
0
10
1
paper rank, r
10
2
is the lifetime of the paper at the time of download
·
Even the average citation rate ci(r) ≡ ci(r)/τ
is heavily tailed!
model for the
ci(r)8
elation h ≤ h for p > q. The value p ≡ 1 recovers
proposed by Hirsch so that h = h . We will use the generr
a
ndex to establish quantitative indicators of scale Scaling
invariance
z ofzthe h index.
⤸ z mobility
ation profiles, as well as the
relation
between
β
c
C, h, and β
or c(r)
1+β
0
C
~
h
scientist
find that c2i (r) can be approximated
10 0 i analyzed, we
1
10
10
10
crete generalized
beta
(DGBD) [5, 6], Hence,
paperdistribution
rank, r
*
r
knowing both
,b
Fortu- 4 [5] Guimera, R., Uzzi,−β
B., Spiro, J., Amaral, L. A.γN. Team assemthe
10
c(r)
≡ Ar
(N
+
1
−
r)
.
[3]
m shifts
bly mechanisms
determine
collaboration
network
structure
and
*
c(r ) hScience 308, 697–702 (2005).
h-index and C is
3
team performance.
H1
10 [6] Malmgren, R. D., Ottino,
1
59, 56–
J. M., Amaral,
L. A. N. The role ≈ redundant
= #and
of publications
meters2 A, Nofβ,
γ protégé
and performance.
N are Nature
each463,defined
i mentorship
in
622 – 626 for a given
10
mulative
(2010).
β
scaling
slope of top
papers i, however we suppress
i =an
esponding
to
individual
scientists
SIS 79, 1 [7] Azoulay, P., Zivin, J. S. G., & Wang, J. Superstar Extinction. Q.
γi z
= truncation scaling
of less-cited
papers
10
z
z
J.
of
Econ.
125
(2):
549–589
(2010).
i in equations
to
keep
the
notation
concise.
We
estimate
x
+
all papers
i = total
Illinois, 0 [8] C
Radicchi,
F.,citations
Fortunato, from
S. & Castellano,
C. Universality of ci10 parameters
caling
β
and
γ
using
multiple
linear
regression
tation distributions:
Toward
an
objective
measure of scientific
c 0
citations, c(z)
citations, c(r)
hically
the derivation
of the characteristic
ci (r) crossover(DGBD)
values that
Theillustrate
Discrete
Generalized
Beta Distribution
∗
p
q
ished “peak” paper regime corresponding to paper ranks r ≤ r (shaded region).
ex reinforcement relation between the impact of a scientist’s
1 most famous papers
*
on log-log
4 axes with N = 278, β = 0.83 and γ = 0.67, corresponding to the
genta curve
10 is the H1 (z) line on the log-linear scale with corresponding h-index
us. (b) We3 plot on log-linear axes the centered citation profile ci (z) (solid black
- the peakxpaper regime,
+ but fails
10[7]. This representation better highlights
0 in Eq.
onding logarithmic derivative χ(z)* of c(z) (solid black curve), which represents
2
o −χ,10
where χ is the average value of χ(z) given by Eq. [12]. The values of z ± ,
ntersection of χ with χ(z)
by Eq. [13]. The regime z < z − corresponds to
Hgiven
1 −
1 crossover between the β h
sponds
and
1 γ scaling regimes.
10to zx which marks the
hich
10
H
c(hp ) = php
top-100 “champions” of
ca- A comparison of ci(r) for the
10
10
10
10
10
striPhysical
Review
Letters
paper
rank,The
r
with
the
relation
h
≤
h
for
p
>
q.
value
p
≡
p
q
a
b
beta10
h-index
Average c(r) proposed by Hirsch so that h = h1 . We will
ows10
10
alized h-index to establish quantitative indicators of s
mon
! = 0.92
10
11/r’ of the h
in
the
citation
profiles, as well as the mobility
g exci (r)10 H
10
scal-10 H
Model for c(r)
s of10
10
10
10
10
10
10
om- 10
For 10each
scientist
i analyzed,
we
find
that
c
(r)
can
be
paper rank,
r
i
scaled paper rank, r' = r
applying
b Set A: average h-index <h> = 61± 21
tion
scaling transformation
by the discrete generalized
beta distribution (DGBD)
e an10
−β
γ
the Discrete Generalized
c(r) ≡ Ar (N + 1 − r) .
1
Beta
Distribution(DGBD):
star
), β,10
The
parameters
andmodel
γ and
N are each defin
Average
values ofA,
the β,
DGBD
parameters:
essci (r)
<β>corresponding
= 0.83 ± 0.23 to
andan individual
<γ> = 0.67scientists
± 0.19 i, howev
also
the 10index i 10in equations
to keep the notation concise
10
dic- 10
1
2
0
0
1
2
3
0
1
2
3
"
5
c(r') = c(r)/A (N+1-r)
citations, c(r)
4
3
2
80
1
2
0
-2
0
0
1
2
3
c(r') = c(r)/A (N+1-r)
"
!
0
-2
0
1
2
3
“Don’t throw the important career data out with the bathwater”
towards comprehensive publication, impact, and collaboration profiles
Andrew Wiles
Albert-Laszlo Barabasi
! = 2.3
paper
rank
101
100
1
2
5
10
Career age, Τ
career age, t
20
18
9
1
2
5
10
Career
age,
Τ
career age, t
Β � 0.66 , Γ � 1.48
2
101
t=5
Monday, July 23, 12
paper
rank,
r r
paper
rank,
1
10
! = 4.6
101
100
1.0
2
170
160
150
140
130
120
110
100
90
80
70
60
50
40
30
20
10
1
102
1.5 2.0
3.0
5.0
7.0
10.0
Career age,
career
age, t
15.0 20.0
Τ
100
1.0
C� r, T �
10
10
103
1.5
2.0
3.0
growth of the impact portfolio
t = 25
0
cites
Cum. cites
Cum. papers
papers
Cum. coauthors
coauthors
Cum. new
coauth.
new
coauth.
Cum. unique coauth.
unique
coauth.
104
paper
rank
101
20
103
100
103
10
101
100
1
C� r, Tprofile
�
rank citation
c(r,t)
10
Longitudinal career measures
10
10
2
4
Cumulativecitations
citations
cumulative
2
cites
Cum. cites
Cum. papers
papers
Cum. coauthors
coauthors
Cum. new
coauth.
new
coauth.
Cum. unique coauth.
unique
coauth.
103
R. Albert, H. Jeong, A.-L. Barabási,
Diameter of the world wide web.
Nature 401,AL
130-131
(1999).
Barabasi
rank citation profile c(r,t)
cumulative
Cumulativecitations
citations
103
Longitudinal career measures
Wiles, Andrew, Modular elliptic curves
and Fermat's Last Theorem. Annals of
Mathematics 141, 443–551
A Wiles (1995).
104
5.0
7.0
Career
age,
Τ
career age, t
10.0
15.0 20.0
Β � 1.06 , Γ � 0.66
t = 22
103
102
101
t=5
100
100
emergence of the DGBD profile
101
paperrank,
rank,
paper
r
r
102
C
Forecasting careers?
2nd period h-index, h(t = 3 | {T2})
Difficulty of predicting impact of future papers
15
[B]
33
19
10
16
10
14
25
15
10-6
p<
R2 = 0.32
12
10
12
9
10
5
[A]
p < 10-10
R2 = 0.24
5
6
6
5
4
1
4
10
0
15
29
[C]
16
p = 0.01
R2 = 0.05
10
7
8
0
0
8
5
9
0
10
14
5
[D]
26
p < 10-5
R2 = 0.14
15
11
12
9
11
10
5
8
8
5
8
8
6
7
5
5
2
0
0
0
5
10
0
5
1st period h-index, h(t = 3 | {T1})
1
3
6
career age
9
t
10
s on the
Starting with neuroscientists, we
attempted to predict the h-index of each
predictscientist 5 years ahead — a timescale relpeers to
evant for tenure decisions — using a linear
esearch
regression with elastic
ould be
NATURE.COM
net regularization6 (see
For more on science Supplementary InforDaniel
E.
Acuna,see:
Stefano Allesina
and Konrad
P.
s informetrics,
mation
at go.nature.
Kording present
a formula to estimate
the future
ations,
go.nature.com/nj2xqk
com/mtvuzr).
The
of about 34,800 neuroscientists, 2,000 scientists studying the fruitfly Drosophila and 1,300
evolutionary researchers. We matched these
authors to records in Scopus, an online database of academic papers and citation data. We
restricted our analysis to authors who had
accrued an h-index greater than 4 (to exclude
inactive scientists); to publications after 1995
(because
records are sparse
before
These electronic
are approximate
equations
for
precise for life scientists, but likely to be less
then); to authors who had published their
predicting
h-index
neuroscientists
in
meaningful for the other sciences. Try it for
first
manuscriptthe
in the
past 5–12ofyears;
and
to
authors
who were
identifiable
in Scopus.reasonably
the
future.
They
are probably
yourself online at go.nature.com/z4rroc.
That left us with 3,085 neuroscientists,
57 Drosophila researchers and 151 evolu● Predicting
(R 2 = 0.92):
tionary
scientists fornext
whomyear
we constructed
a
history of publication, citation and funding.
h+1 = 0.76 + 0.37√―
n + 0.97h − 0.07y + 0.02j + 0.03q
For each year since the first article
published by a given scientist, we used the features
that
were available at5the
time tointo
forecast
their
● Predicting
years
the
future (R 2 = 0.67):
h-index a number of years into the future.
h+5the
=CV
4 + 1.58√―
n + 0.86h − 0.35y + 0.06j + 0.2q
For example, we reconstructed how
features of a scientist looked five years after
publishing his or her first article, and found
2
Predicting
10those
years
intoand
the
a●relationship
between
features
the future (R = 0.48):
―
reconstructed h-index five years
on. = 8.73 + 1.33√n + 0.48h − 0.41y + 0.52j + 0.82q
h+10
Starting with neuroscientists, we
attempted to predict the h-index of each
scientist
years ahead
a timescale
relKey: n,5number
of—articles
written;
h, current h-index; y, years since publishing first article;
evant for tenure decisions — using a linear
j, number of distinct
journals
published
in; q, number of articles in Nature, Science, Nature
regression with elastic
6
NATURE.COM
net regularizationof(see
Neuroscience,
Proceedings
the National Academy of Sciences and Neuron.
For more on science Supplementary Informetrics, see:
mation at go.nature.
go.nature.com/nj2xqk
com/mtvuzr). The
METRICS
Predict your future h-index
Predicting
scientific success
｛
h-index of life scientists.
e research scientists often worry
s reservedabout the future of our careers. Is
63 and 53, respectively. According to Hirsch,
an h-index of 12 for a physicist — meaning
12 papers with at least 12 citations each —
could qualify him or her for tenure at a major
university.
However, the h-index3 and similar metrics4
can capture only past accomplishments, not
future achievements5. Here we attempt to
predict the future h-index of scientists on the
basis of features found in most CVs.
We maintain that the best way of predicting a scientist’s future success is for peers to
evaluate scientific contributions and research
depth, but think that our methods could be
valuable complementary tools.
The typical research CV contains information on the number of publications,
Major Flaws!
our research an exciting path or
a dead end that will end our careers prematurely? Predicting scientific trajectories is
a daily task for hiring committees, funding
agencies and department heads who probe
CVs searching for signs of scientific potential.
One popular measure of success is physicist Jorge Hirsch’s h-index1, which captures
the quality (citations) and quantity (number) of papers, thus representing scientific
achievements better than either factor alone.
A scientist has an h-index of n if he or she has
published n articles receiving at least n citations each2. Einstein, Darwin and Feynman,
for example, have impressive h-indices of 96,
1) aggregating across different
career ages
2) h-index is non-decreasing
R2 will be artificially large
hi(t+∆t)
early
1
late
mid
t
∆t
career age
The accuracy of future h-index prediction decreases over time, but the Acuna et al. formula predicts
future h-index better than does current h-index alone (left). The contribution of each factor to the
formula accuracy also changes over time (right). Shading indicates 95% confidence error bars.
Percentage variance explained (R 2)
hi(t)
PATHS TO SUCCESS
1 3 S E P T E M B E R 2 0 1 2 | VO L 4 8 9 | NAT U R E | 2 0 1
© 2012 Macmillan Publishers Limited. All rights reserved
t
1
Acuna et al.
h-index
0.5
6
Standardized coefficient
W
1 3 S E P T E M B E R 2 0 1 2 | VO L 4 8 9 | NAT U R E | 2 0 1
In top journals
Distinct journals
√—
# articles
4
h-index
Career
length
2
0
–2
0
1
5
Years ahead
10
1
5
Years ahead
10
Acuna Model R2(t, ∆t)
“prediction power”
hi(t+∆t | t): predicting growth conditional
on (early) career age t is much more difficult
1
1
Acuna model, t = All
Acuna model, t = 3
Acuna model, t = 2
Acuna model, t = 1
100 Assistant
Professors in Physics
0.8
0.8
0.6
0.6
0.4
0.4
Acuna model, t = All
Acuna model, t = 3
Acuna model, t = 2
Acuna model, t = 1
0.2
0.2
0
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
future prediction horizon, ∆t (years)
D
Career longevity:
uncertainty associated with career hazards
Difficulty in predicting career longevity
(survival probability)
0
10
P(x)
10
F
-1
10
-2
-2
10
10
-3
Nature
PNAS
Science
10
-4
10
-3
CELL
NEJM
PRL
10
P(x)
E
-1
-4
10
-5
10
0
10
-5
0
10
1
10
10
0
10
1
10
210
x (career longevity in years)
high-impact journals are competitive arenas
•
Each author i has n articles in a given journal j. As a proxy
for career longevity in academia, we define the journal
longevity x as the number of years separating his/her first
and last publication in journal j :
xi,j = yi,j (f) - yi,j (0) +1
Empirical longevity distributions in sports and academica
0
10
1
10
2
3
10
10
-2
10
α = 0.77
4
1
10
10
2
3
10
4
10
A
Major League Baseball
5
10
10
B
•
-2
10
α = 0.63
-4
-4
Baseball (fielders)
Basketball (NBA)
AB (Korea, KPB)
AB (USA, MLB)
-6
P(x)
10
10
10
-1
10
α = 0.71
C
α = 0.55
D
Baseball (pitchers)
-5
10
-4
Soccer (Premier League)
IPO (Korea, KPB)
IPO (USA, MLB)
10
Games
-7
-6
0
10
1
10
2
3
10
10
4
10
0
10
1
2
10
10
10
3
10
x (career longevity in opportunities)
0
0
10
10
E
F
-2
-2
10
10
-3
Nature
PNAS
Science
10
-4
10
-3
CELL
NEJM
PRL
10
-4
10
-5
0
10
1
10
10
0
10
1
•
-1
10
10
-5
210
x (career longevity in years)
“Quantitative and empirical demonstration of the Matthew effect in a study of
career longevity.“ A. M. Petersen, W.-S. Jung, J.-S. Yang, H. E. Stanley.
Proc. Natl. Acad. Sci. USA 108, 18-23 (2011).
P(x)
α ≈ 0.40
-1
10
P(x)
3% of all fielders finish their
career with ONE at-bat!
•
3% of all pitchers finish their
career with less than one
inning pitched!
-2
10
10
•
10
-3
10
``One-hit wonders”
-6
Minutes
P(x)
10
130+ years of player
statistics, ~ 15,000 careers
•
•
``Iron horses”
Lou Gehrig (the Iron Horse): NY Yankees
(1923-1939)
Played in 2,130 consecutive games in 15
seasons! 8001 career at-bats!
Career & life stunted by the fatal
neuromuscular disease,
amyotrophic lateral sclerosis (ALS),
aka Lou Gehrig’s Disease
institution. Overall, men and women are retained and
ences (8). Between
density function shows that women are much more
overofthe
run.time
Those
wholeftarethe
hired
first
10 years.
In the first
years, departure
rates ments
time
hirelong
to the
they
university.
retention
of women
in 3science
and engineering.
likely
to
leave
very
soon
after
hiring
than
are
tend
to
stay,
and
the
number
hired
is
increasing.
are
somewhat
lower,
whereas
in
the
next
3
years,
ofthan
women
in all ld
move
to
positions
outside
of
academia,
whereas
.S.
universities
are
concerned
about
fachowever,
faculty
leave
significantly
earlier
other
In the disciplines included in this study, the rate All data were obtained from publicly available
men.
This results women
in a survival
curve in which
change
in department
departure rates are high. The survival function The time constant for theulty
are significantly
more likely to be un- for life sciences a
retention
in science and
engineering
than persist
men,to4.45
yearsmark.
compared
with 7.33 years.
very fewleaves
women employed
the
very along.
shows a rather steeper decline in faculty at early gender composition is, however,
or 20-year
to exit the
tenure track for adjunct physical science i
(1–4). When
faculty member
Fig. 1. Nonparametric
large-scale
analyses
of the retention
Although there
are no significant
times and a more moderate descent at later times.
positions
(3). Women
with Ph.D.s in science, tech- of having a tenur
prematurely,
they suffer differdisruptions inPrevious
teaching
survival
curve
for
faculty
of academic
faculty
in theengineering,
United States
bymentoring
gender foras
thewell
entire
popIt is apparent that posttenure faculty leave at a ences in retention
nology,
andhave
mathematics (STEM) Ph.D. is significa
and
as
significant
economic
who entered
between
1990 faculty. The hazard ulation, there are some disciplinary variations. relied on aggregate data.S.
universities
are concerned
about fac- m
or surrogate
variables
to
lower
rate than
pretenure
disciplines
have
also
been
found
to be less likely life sciences but ab
losses
(1).
Start-up
costs
in
engineering
and
natuand
2002
by
gender.
IQR,
track to
year-to-year
turnover
STEM
disciplines.
median
time can
to exit
for men
function tells a similar story. This is the rate of Table 2 lists the ral
ical and engineerw
menin to
be employed
time, although
sciences
range
fromand
$110,000
nearly than
ulty
retention
in full
science
and engineering
interquartile
range.point in a faculty career, and women by discipline.
The
large-scale
analysis
reported
here,
which
A
log-rank
and
Wilcoxon
attrition
at a given
ical and life scie
if they apply (11).
$1.5 million (3), and it may take up to 10 years equally likely to succeed
(1–4).
When
ashown
faculty
member leaves e
individual
across
difference
two (4).
sur- tracked the retention of
it peaks at about 6 years. The differences be- test for significant
Women
havefaculty
also been
to have greater heavily represent
to recoup
thisbetween
investment
confirmed
some
of thetoearlier
but disruptions
vival curves (20) shows
no differences
except
tween men and women are small.
prematurely,
they
suffer
chance of gettingp
intentions
leaveresults
the
STEM
disciplines (12),in teaching
Retention
rates for faculty
in thetime,
United
States
Table 1 gives the median time to departure for in the case of mathematics, where P = 0.0522 points to new areas of concern. Although the
not academia
as a whole (13), and to ure, or getting pro
have been consistent. From 1971 through 1989,and
fa- although
n
mentoring
as aswell
catalogs
and bulletins
a dataas significant economic
cohorts 1 to 3 by gender. Half of all entering fac- and 0.008, respectively. Because of the choice use of college
culty members were retained at rates of 90 to 92% leave for different reasons. Whereas salary is the terms of retention
work has
madein engineering and natuulty have departed by 10.9 years. There is no of weighting function, the log-rank test puts source is time-intensive,
(1).previous
Start-up
costs
one
reason for
men, women cite more and minority facd
for associate and full professors and 84 to 86% losses
for number
statistically significant difference between men and more emphasis on differences at long times, and the case for the value of using publicly available
tentions
in the put
interpersonal
and family
15). Delays to
assistant
professorson
(5).
In 1996–1997
and 2001–
ral sciences
range
from(14,$110,000
nearly
sources
to monitor
institutionalcan
change
(21).reasons
more emphasis
differences
women. The 95% confidence intervals (CIs) on the the Wilcoxon puts
tenureonresulting
in lower
salaries could account well as in the pur
the retention
ratesmodfor associate
An profesearly$1.5
studyinmillion
based
American
AssociaCox proportional
hazards
median for men and women are nearly coincident. at short times. A2002,
(3),
and
it may
takebutupde-to 10
the years
applied sciene
for
women
leaving
before
tenure (16),
sors
were
again
in
the
range
of
90
to
92%
(6).
Another perspective on faculty retention is el was used to detect differences among disciplines tion of University Professors data (5) found averpartment
climate
is aprofessors
primary reason
with the
retention
in scitoratesrecoup
this
investment
(4).why women gineering fields (
age attrition
of
0.15 for
assistant
Two disciplines
were
signif- of women
available by examining promotion rates. The per- regardless of gender.Problems
In this study,
are
less
satisfied
and
more
likely
to quit (4).
ence
and
engineering
in
the
United
States
have
professors.
The values
for
centage of faculty in the first two cohorts who icantly different. In mechanical engineering, facul- and 0.08 for associate
i
Retention
rates
for
faculty
in
the
United
States
Significant
disciplinary
variations exist in the engineering facu
been well documented. Like a leaky
pipeline,
assistant
professors are
much higher
than our
havevalues,
beenindicating
consistent.
From
1971
through 1989,
time of fahire to tha
retention
of women
science and
engineering.
each career stage in engineering and
the natural
hazard
function
that in
retention
All
data
were obl
In
the
disciplines
included
in
this
study,
the
rate
sciences shows the retention of women
lower
of assistantculty
professors
may have improved
since
members
were retained
at rates of 90 to 92%
than
the
stage
before
it
(3,
7).
In
particular,
al1990. In a more recent study (6), the attrition
ARY 2012 VOL 335 SCIENCE www.sciencemag.org
associate
and
full professors
and 84 to 86% for n
though women are increasingly represented
Fig.professors
1. Nonparametric
rates foramong
allfor
associate
averaged
0.077.
curve
for faculty
those with earned doctorates, they From
lag behind
in thesurvival
our data,
hazard
function
at (5).
10 years
assistant
professors
In 1996–1997 and 2001– i
entered between
1990 was
(representative
ofwho
associate
professors)
representation in the academic faculties
(8).
i
the
retention
2002
by gender.
IQR,rates for associate profes0.0709,
and2002,
at 8.5and
years
it was
0.073, consistent
The problem appears to lie in
differential
REPORTS
interquartile
range.
with
the
two
earlier
studies.
application rates. Once women apply for sors
or were again in the range of 90 to 92% (6).
f
Acknowledgments:
We thank Y. S.
Chooimportance
for critical commentsof Institute
(A.L.O.). A.L.O. is a paid consultant for and holds stock
Suppor
Our
work
confirms
the
the
are in consideration for a career move,
they
are
on the manuscript, H. Tien and D. Marciano for automated
options in CBLI which has developed the flagellin derivative
p
the retention
ofcountermeasure.
women
in8,007,812
sci- www.scie
period
period
oftechnical
critical
risk
crystal screening,
X.Problems
Dai and as
M.-A.aElsliger
forwith
expert
CBLB502 into a radiation
U.S. patent
Materials
equally likely to succeed, but theylate
are pretenure
often
not
assistance, and K. Namba and K. Yonekura for generously sharing
for pharmacologically optimized FliC derivative CBLB502 is
SOM Tex
in thecoordinates
retention
of faculty
indiffraction
STEM.
Like earlier
in the pool (3, 9–11). Men have been
found
toflagellar
of
filament.
X-ray
data sets
licensed
to CBLI by
Cleveland Clinic. Reagents
are available
under
Figs. S1a
ence
and11-1
engineering
in the
United
States
have
were collected
at SSRL beamline
and APS beamlines
a materials transfer agreement. This is Scripps Research Institute
Tables S
analyses,
we
find
that
posttenure
faculty
membe significantly more likely to receive tenure
23ID-B and or
23ID-D. Supported by NIH grants AI042266 (I.A.W.),
manuscript 21369. The data presented in this paper are tabulated
Referenc
been
well
documented.
Like
a
leaky
pipeline,
bers are
overall
less
likely
to (A.V.G.);
depart
R01 AI080446
(A.V.G.),
and RC2
AI087616
the than prein the main paper and the supplementary material. Structure
Institute for Chemical Biology (I.A.W.); and research
factors and coordinates for TLR5-N12 and TLR5-N14 /FliC-DD0
tenureSkaggs
faculty.
Overall,
the tochances
that any
1
r
each
career
natural
20 Octob
are deposited in the Protein and
Data Bank the
under accession
codes
grants from
Cleveland BioLabs
Inc. (CBLI) stage
Roswell Parkin engineering
U
Survival (longevity) analysis of tenure track assistant professors
U
50% of faculty leave by year 10
high-risk period peaks
around pre-tenure year 6
Department of Mechanical, Aerospace, and Nuclear En3V44 and 3V47, respectively.
Cancer Institute (A.V.G.) and Sanford-Burnham Medical Research
gineering, Rensselaer Polytechnic Institute, Troy, NY 12180,
2
USA. Faculty of Communication, Art, and Technology,
Table 1. Simon
Median times to exit the tenure track by
Fraser University, Harbour Centre 7410, 515
West Hastings
gender,
cohorts 1 to 3.
Street, Vancouver, BC V6B 5K3, Canada.
VLR
VLR
10.1126
sciences shows the retention of women lower I
grow
than theAnalysis
stage beforeofitFaculty
(3, 7). In Retention
particular, al- ofsentati
Survival
in
though
women are increasingly represented among tion
F
*To whom correspondence should be addressed. E-mail:
Median time
marked
in
Science
and
Engineering
by
Gender
95% Normal CI
kamind@rpi.edu
women
to exit, years
s
those
with earned doctorates, they lag behind in the
nin
Deborah Kaminski * and Cheryl Geisler
Lower
Upper
percentw
representation
in thefaculty)
academic
faculties (8).
assistant
(a total
hired in science and engineering since 1990
for, an
All 17Individual
10.94professors
10.3of 2966
11.61
864
FEBRUARY
2012
VOL
335
SCIENCE
www.sciencemag.org
at 14 United States universities were tracked from time of hire to time of departure by using publicly
percena
The
problem
appears
to
lie
in
differential
Men available catalogs
11.05
10.34of survival11.81
and bulletins. Results
analysis showed that the chance that any given faculty pool fo
Fig. 2. Survival analysis of faculty who entered between 1990 and 2002 by gender. LSXY, least squares;
Women
11.89
member will 10.40
be retained over time 9.094
is less than 50%;
the median time to departure is 10.9 years. Of
i
application
rates.
Once
women apply for or physic
F, number that left; C, number still remaining.
all those who enter as assistant professors, 64.2% were promoted to associate professor at the same
than in
institution. Overall, men and women are retained and promoted at the same rate. In mathematics,
are in consideration for a career move, they are resenta
however, faculty leave significantly earlier than other disciplines, and women leave significantly sooner
Wo
1
2
E
A “real-world” example of the tenure decision:
Italian Abilitazione
courtesy of F. Radicchi
A practical example
the Italian National Scientific Qualification
1) Number of papers:
2) Number of citations:
3) Contemporary h-index:
Sidiropoulos A et al. Scientometrics 72, 253 (2007)
calculated over a population of 1400 Italian physicists
Np total number of publications
AA academic age
Nc total number of citations
C(i,ti ,t) citations accumulated up to year t
by paper i published in year t
R. Manella and P. Rossi, arXiv:1207.3499 (2012)
source: ”Abilitazione Scientifica Nazionale – La normalizzazione degli indicatori per l'eta' accademica”, ANVUR, Jul. 2012
courtesy of F. Radicchi
A practical example
the Italian National Scientific Qualification
associate professor
Chemistry Biology
Physics Mathematics
norm. pub.s norm. citations
full professor
h-c index
norm. pub.s norm. citations
h-c index
01/A1
5
1.74
2
4
1.37
2
01/A2
8
1.65
2
9
3.23
3
01/A3
10
4.34
4
14
8
5
02/A1
59.5
104.08
18
78
105.03
22
02/B2
37.5
40.08
11
47.5
75.94
14
05/A2
14
24.45
8.5
20
37.47
10
05/C1
21.5
15.77
8
26
18.63
9
03/A1
26
29.47
9
41
53.81
12
03/B1
31
47.05
11
49.5
62.38
13
source: ”Mediane per candidati allʼabilitazione scientifica nazionale a professore associato/ordinario”, ANVUR, Jul. 2012
General take-home messages
•
Complex career dynamics: Knowledge, reputation, and collaboration spillovers are major factors leading
•
Science as an evolving institution: An institutional setting that neglects specific features of academic
•
Nano-sociology: A data-centric (“big data”) understanding of the production function of individual
•
Competition and Reward: There are many analogies between the superstars in science and the
to increasing returns along the scientific career trajectory
career trajectories (increasing returns from knowledge spillovers and cumulative advantage, collaboration
factors, career uncertainty) is likely inefficient and unfair.
scientists can improve academic policies aimed at increasing career sustainability and decreasing career risk
superstars in professional sports, possibly arising from the generic aspects of competition. Currently, the
contract length, compensation, and appraisal timescale in these two professions are VERY different.
Is Science becoming more like professional sports?
I ) “Methods for measuring the citations and productivity of scientists across time and discipline,”
A. M. Petersen, F. Wang, H. E. Stanley. Phys. Rev. E 81, 036114 (2010).
II) “Quantitative and empirical demonstration of the Matthew effect in a study of career longevity,”
A. M. Petersen, W.-S. Jung, J.-S. Yang, H. E. Stanley. Proc. Natl. Acad. Sci. USA 108, 18-23 (2011).
III) “On the distribution of career longevity and the evolution of home run prowess in professional baseball,”
A. M. Petersen, W.-S. Jung, H. E. Stanley. Europhysics Letters 83, 50010 (2008).
IV) “ʼMethods for detrending success metrics to account for inflationary and deflationary factors,”
A. M. Petersen, O. Penner, H. E. Stanley. Eur. Phys. J. B 79, 67-78 (2011).
V) “Statistical regularities in the rank-citation profile of scientists,”
A. M. Petersen, H. E. Stanley, S. Succi. Scientific Reports 1, 181 (2011).
VI) “Persistence and Uncertainty in the Academic Career,”
A. M. Petersen, M. Riccaboni, H. E. Stanley, F. Pammolli. Proc. Natl. Acad. Sci. USA 109, 5213-5218 (2012).
Thank You!
A special thanks to my collaborators:
Santo Fortunato, Woo-Sung Jung,
Fabio Pammolli, Raj Pan, Orion
Penner, Massimo Riccaboni, Gene
Stanley, Sauro Succi, Fengzhong
Wang, and Jae-Sook Yang
http://physics.bu.edu/~amp17/
Title: Identifying potential pitfalls in the quantitative appraisal
system for scientific careers
Abstract:
Quantitative measures are becoming increasingly prevalent in
the scientific appraisal of countries, universities, departments,
and notably, individuals. In this talk I will discuss the potential
pitfalls arising from the appraisal of individual careers based on
citation metrics, a proceeding which is likely to occur at several
stages of an academic career, from postdoctoral and faculty
appointments to career achievement awards. Using longitudinal
career data for 450 scientists, ranging from assistant professors
to Nobel laureates and Fields medal winners, I will demonstrate
a graphically intuitive method for visualizing an individual's
publication profile. While much ado has been made about the hindex, a metric intended to measure simultaneously the
productivity and impact of a scientist, I will argue for the
careful use of this and related quantitative measures. With the
remaining time, I will illustrate the complex dichotomy of
competition and collaboration in science.
β
The β -vs- h parameter space
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Physics (A)
Physics (B)
Physics (C)
Biology (D)
Math (E)
Matthias Ernzerhof (U. Montreal)
c (1) = 16,314
“Generalized gradient approximation made simple”
Perdew, JP; Burke, K; Ernzerhof, M
PRL 1996
Bert Vogelstein
P. W. Anderson
20 40 60 80 100 120 140 160 180 200
A.-L. Barabasi
h1
For a given h, a large β value corresponds to a larger total citations,
Ci ~ h1+ β ,
which is a proxy for career publication impact
ball AssociationWe
(NBA)
careers
during
the
63-yea
model the
career aspublicatio
an aggre
BA) careers
during
the
63-year
period
The career trajectory in science:
1946–2008.
portunities which arrive at
theadva
vari
tive
a tale ofWe
knowledge,
collaboration,
and
reputation
spillovers
model the
the
career
as
an
aggregation
of
out
reputation
of
aopscientist
is
typica
Annual
production
of
individual
i
reer asportunities
an aggregation
of
output
tional
me
which
arrive
at
the
variable
rate
n
(t
i
αi resentation of his/her contributions
of publications
in �
year t e
rrive at
variableofrate
ni (t).number
internal
thethe
reputation
a scientist
isSince
typically
a cumulat
t
mulative
production
N
(t)
≡
� =1
α=1
i
t
Cumulative
production,
a
proxy
for
career
reputation
resentation
of his/her
contributions,
we
considerd
cientist
is typically
a cumulative
repinternal
�t Fig. 1 �shows th
career achievement.
cent
er contributions,
we
consider
mulative production
Ni (t)the
≡ cu) ascase
ap
� =1 ni (t
t
tion� Ni (t) of six notable careers wh
�t
Fig.
1
shows
the
cumulative
for many
winners
h
α
n Ni (t)career
≡ achievement.
n
(t
)
as
a
proxy
for
i
�
i
Biology
relation
N
(t)
≈
A
t
.
However,
th
t =1
prolific careers!
i
i
tion
N
(t)
of
six
notable
careers
which
display
a
i
publicatio
Fig. 1 shows
the cumulative
producNi (t), see
year, t
αi Fig. S1, which do not
relation Ni (t) ≈ Ai t . However, therecan
aretrigge
also
table careers which
scaling marked non-s
ity,display
instead adisplaying
Ni (t), see Fig. S1, which
Careers aredo
also not
subjectexhibit
to “shocks”: such
αi
cascading
linearity
arising
from
significant
exo
t . However,
there
are
also
cases
of
ity, instead displayingsometimes
marked
non-stationarity
a
good:
(Nobel
Prize,
promoted
Collabo
We justify
this
2-parameter
model
which
do
not
exhibit
such
regularlab director) exogenous career
linearity arising from tosignificant
scaling
methods
and
data
collapse
(
ing
fluctu
ng marked
non-stationarity
and
nonWe justify this 2-parameter
model
the fraud
SI tex
sometimes bad:
accused ofin
scientific
to
show
that
most
N
(t)
can
be
mod
i
Physics
the
abilit
m significant
exogenous
career
shocks.
scaling methods and data collapse (see Figs. S2
functional
form.
Careers
with
α
lated
to
year,
t
arameter
model
the NSI(t)text
to show
thatinmost
can using
be modeled by thisi c
Ni (t)
10
10
3
2
10
10
Vogelstein, B: (h = 185)
Baltimore, D: Nobel at age 37
1
0
0
10
10
1
10
3
Gossard, AC
Parrinello, M
Pfieffer, LN
Weinberg, S
Glashow, SL
2
Ni (t)
10
10
10
10
1
0
10
0
10
1
2
Ave. annual output, < ni >
i
Alternatively,
is va
al
Collaborationacademic
Radius and career
team it
efficiency
A
the relation between the
th
Individuals
are
constantly
entering
and e
ntitative relations between
career
growth,
career
Towards
a micro-level
production
function:
inr
and
the with
collaboration
professional
market,
birth
and death rat
aboration efficiency.
The
fluctuations
in
producψ
ing
on
complex
economic
and
institutional
fac
ca
he unpredictable horizonlation
of “career
shocks”
which
�ni � and
∼ S
i quant
to competition, decisions
performance
at
siz
e ability of a scientists
tothe
access
new
creative
opB
stages
of
career
can
have
long
lasting
con
with
ψ
=
0.74
±
0.04
a
(A) Relation between
average
annual
average
number
of production
Si iscareer
median number
[16, 28].
To
better
understand
uncert
publications
per year
of coauthors
peror
year
ψ
=
0.25±0.04
(R =
0.3
aboration radius S
≡
M
ed[k
]
shows
asaying
decreasing
Ou
trayed
by
the
common
“publish
pe
i
i
we analyze
the outcome
fluctuation
put per collaborator
asrelation
demonstrated
by sublinsiz
between
sequen
terestingly, dataset [A] scientists have on average
sig
r
(t)
≡
n
(t)
−
n
(t
−
∆t)
n
(t
+
1)
is
relatively
sm
i
i
i
i
ut-to-input efficiency. (B) The production flucthere is aof careerexponents
calculated
fo
i
in
year
t
over
the
time
interval
∆t
σNot
(r)surprisingly,
ismarginal
a quantitative
measure
for
uncertainty
idecreasing
returns with
m
ψ/2
Fig.
2(A)
and
(B)
show
the
unconditional
pdf
increasing
collaboration
radius,
likely
careers,
with scaling relation
σi (r)
∼ Si .This
(C) resu
mately
equal.
ca
attributable to team management
which are leptokurtic but remarkably symme
inefficiencies,
however inefficiencies
here
is an increasing
returns
evident
infrequencies
the annual
trating
the
endogenous
of positive
as
the
convolution
ofψan
tu
aggregate sub-linearally,ψ < 1C
Management,
coordination,
and arise
i (t) since α > 1. tive
output growth.
Output fluctuations
of
tion P (n) for each scien
Dataset A
Dataset B
1
! = 0.74(4)
10
0
Std. Dev. annual change, !i (r )
10
1
10
Dataset A
Dataset B
" / 2 = 0.40(3)
Output change (“growth fluctuation”),
0
10
0
1
10
2
10
10
Number of publications, ni(t)
Median # of coauthors per year, Si = Med [ki]
A: ! = 0.68(1)
B: ! = 0.52(1)
C: ! = 0.51(2)
1
10
std. deviation of publication change
!
team efficiency
parameter
0
10
0
10
1
10