8/12/2019 Science-2014-Xie-809-10

1/223 MAY 2014 VOL 344 ISSUE 6186 80SCIENCE sciencemag.org

In recent years, academic scholarshipand public discourse have become in-creasingly preoccupied with social andeconomic inequality, which has risenin many countries. It is surprising thatmore attention has not been paid to the

large, changing inequalities in the world ofscientific research. I suggest that althoughthe basic structure of inequalities in science

has remained unchanged,their intensities and mecha-nisms may have been altered

by recent forces of globalization and inter-net technology.

By inequalities, I mean differences inthree major domains: resources, researchoutcomes, and monetary or nonmonetaryrewards. At times, my discussion is concep-tual and somewhat speculative, as preciseand meaningful measurement of these out-comes is difficult. Even though I refer pri-marily to basic, natural science, most of mygeneral conclusions are also applicable to

the social and applied sciences.

AN INTRINSIC PROPERTY OF SCIENCE.Derek Price observed in 1963 that inequal-ity in science is inherently high and calledit undemocracy ( 1), meaning that a greatscientists value for science far exceeds thatof ordinary scientists. Scientific outputsand rewards are much more unequally dis-tributed than other well-being outcomes,such as education, earnings, or health ( 1, 2 ).

One source of inequality in science is what Robert Merton called the Mattheweffect ( 3 ), referring to Matthew 25:29 in

the Christian Bible: For to all those whohave, more will be given, and they will havean abundance; but from those who havenothing, even what they have will be takenaway. This rich get richer effect meansthat eminent scientists receive dispropor-tionately greater recognition and rewardsthan lesser-known scientists for compara-

ble contributions. As a result, a talented few

can parlay early successes into resources forfuture successes, accumulating advantagesover time.

Although science rewards all participantsthrough a skewed tier system with the mostsubstantial rewards going to the top per-formers in a tournament-like reward sys-tem, science has attributes that resemble awinner-takes-all market: high visibility oftop winners, a large contestant base, accu-mulation of advantages, absence of physicalor cultural boundaries, and intense com-petition ( 3 , 4 ). Thus, many scientists feelthat merely being good at their jobs is not

enough. Competition is all about priority, ascientists claim to be the first to make a bigdiscovery ( 5 , 6 ).

Whatever their cause, high inequality inscientific rewards is often defended on twogrounds. First, given the positive externali-ties of science ( 7 ), the more skewed rewardsare, the greater the incentive for outstand-ing scientific work that will ultimately

benefit all of humanity ( 8 , 9 ). Second, as aprofession, science is supposed to practice

what Merton called universalism ( 10 ), anorm that dictates that evaluation in sci-

ence be based solely on merit rather thanon functionally irrelevant factors such asgender, race, nationality, age, religion, andclass ( 2 , 10 ). This merit-based system makesinequality seem fair and acceptable.

Before the 19th century, science wasmainly a small-scale, personal pursuitenjoyed by a few leisure-class amateurs.Over the next two centuries, it expanded

enormously into an institution charac-terized by certain distinctive features: ahuge, well-paid, professional workforce;

large-scale government andindustrial support; relianceon the university as institu-tion; graduate student labor;and a peer-review systemof evaluation ( 7 ). Advancein Internet technology havefacilitated the rapid, broaddissemination of researchresults ( 11).

Although these features

have made scientific pro-duction faster and more

voluminous, they have alsorendered the evaluation ofscientists less substance-specific and more numbers-

based. Scientists are in-creasingly likely to be judged

by whatever numbers theycan generate in terms ofpublications, citations, re-search grants, prestigiousawards, research team size,and memberships in elite

academies than by their actual scientificcontributions ( 7 ). This tendency may have

been amplified by increasing specialization,such that scientists in one specialty areafind it difficult to understand content inanother. University administrators, faced

with uncertainties and competing demandsfor scarce resources, have strong incentivesto use externally generated and validatedindicators ( 12).

HAS INEQUALITY IN SCIENCE IN-CREASED? More empirical research isneeded to answer this question, but I be-

lieve that two trends have caused inequalityin science to rise over time. First, growth inhigh rewards in science has been limited,occurring much more slowly than the ex-pansion of science itself. It is well knownthat the number of Nobel Prizes is fixed,although many Nobel Prizes in natural sci-ence have been shared in recent decades[see supplementary materials (SM)]. Al-though the number of academically ap-pointed scientists with Ph.D.s increased by150% between 1973 and 2010 in the UnitedStates, the number of newly elected mem-

Undemocracy:inequalities in science

By Yu Xie 1,2

Inequality, an intrinsic feature of science,has trended upward in recent years

SOCIOLOGY OF SCIENCE

1Institute for Social Research, University of Michigan, Ann Arbor, MI 48104, USA. 2Center for Social Research, PekingUniversity, Beijing 100871, China. E-mail: yuxie@umich.edu

POLICY

I L L U S T R A T I O N : P E T E R A N D M A R I A H O E Y / W W W

. P E T E R H O E Y . C O M

Published by AAAS

8/12/2019 Science-2014-Xie-809-10

2/2

I NS IG HT S | PE R SP E CT I VE S

810 23 MAY 2014 VOL 344 ISSUE 6186 sciencemag.org SCIENCE

bers each year to the National Academy ofSciences covering all scientific fields stayedat 60 for a long time until it finally went to72 in 2001 (see SM).

Given sciences concentration of high re- wards to a select few, the large increase in thepopulation of scientists ( 7 ) means a higherconcentration over time. Furthermore, thetendency toward numbers-based evaluations

has made it possible for an outsidereither ascientist in a different field or a nonscientist,

who nonetheless may be in position to dis-tribute resourcesto pass judgment on anyscientist, thus providing an easy pathway forthe Matthew effect.

As with income inequality in the UnitedStates in general, inequality in academicssalaries has trended upward, both betweenprivate and public universities and amonguniversities of each type ( 13). For individualacademic scientists, salary inequality has in-creased substantially since the 1970s, acrossall ranks and diverse fields ( 14).

Here are two examples of intensified in-equality in science today. First, many newscience Ph.D. recipients from American uni-

versities in recent years have been unableto obtain regular academic positions andinstead are taking postdoctoral fellowshipsor other forms of nontenure-track employ-ment ( 7 ). One source of this problem may bethe large supply of well-trained foreign stu-dents and immigrant scientists ( 7 ). Of course,the system of postdocs and temporary em-ployees provides benefits to both senior and

junior (dependent) scientists. However, theformer benefit more, as they are given more

credit due to the Matthew effect.Second, Internet technology, a globalizedeconomy, inexpensive air transportation,and relatively peaceful world politics havecreated an unprecedentedly interconnected

world ( 15 ). In this new global environment,a successful scientist filled with ideas at aprestigious university in America or Europecan design studies and have them carriedout by dependent collaborators in less-developed countries, such as China, wherelabor-intensive scientific work can be con-ducted at lower costs. Such collaborationsare complementary and can lead to mutual

benefits ( 16 ); at the same time, they amplifyinequality across individual scientists. Al-though global collaboration likely benefitsall scientists, benefits vary with a scientistsposition in a collaboration network. More-successful scientists are much more likelythan less-successful ones to be centrally lo-cated in global collaborative networks.

CONTEXTUAL SOURCES OF INEQUALITY.The importance of institutional environ-ment to scientists is well documented ( 2 ,17 , 18). Scientists affiliated with prestigious

institutions are more productive and betterrewarded than those who are not. Hence,greater institution-level inequality serves tointensify the individual-level inequality ofscientists.

There is some evidence that institution-level inequality in resources has increasedover time. Institutional inequalities areshown in the chart by using the Gini coef-ficient, which ranges from 0 for absoluteequality, to 1 for absolute inequality (see

the chart). Despite already high levels ofinequality in three resource measures, theGini coefficients trended upward during theperiod. Part of the reason for the increases

was an expansion in institutions that par-ticipate in research. If we restrict our analy-sis to a limited set of universities that havealways been active in research, there is noclear trend (see SM).

Large country-level differences in scien-tific activities have long been noted ( 19).Historically, the world center of science hasshifted several times; in the past nine de-cades, America has dominated ( 7 ). However,

just as between-country income inequalityhas narrowed in the world, mostly because

of a rise of income in China ( 20 ), between-country differences in science have also nar-rowed, thanks to the globalization of science(7 , 16 ).

Globalization of science and the increas-ing use of Internet technology have con-flicting effects on inequality in science. Atthe levels of individual scientist and institu-tion, they have tended to intensify inequal-

ity. At the country level, they have narrowedit. However, the most significant conse-quences of these two forces for science have

been positive overall: the large expansionof science as a collective enterprise andthe resulting rapid progress of science on aglobal scale ( 16 ). In the long run, resourcesand rewards must be allocated so that in-equality, although incentivizing scientiststo make important scientific discoveries,is properly managed and controlled. Espe-cially important is the need to invest suf-ficient resources in young scientists beforethey gain recognition.

REFERENCES AND NOTES

1. D. J. Price, Little Science, Big Science(Columbia Univ.Press, New York, 1963) .

2. J. R. Cole, S. Cole,Social Stratification in Science(Univ.Chicago Press, Chicago, 1973).

3. R. K. Merton,Science 199 , 55 (1968). 4. R. H. Frank, P. J. Cook,The Winner-Take-All Society

(Penguin, New York, 1995). 5. H. Zuckerman,Scientific Elite: Nobel Laureates in the

United States (Free Press, New York, 1977). 6. R. K. Merton, Am. Sociol. Rev. 22 , 635 (1957).

7. Y. Xie, A. A. Killewald,Is American Science in Decline?(Harvard Univ. Press, Cambridge, MA, 2012).

8. E. P. Lazear, S. Rosen, J. Polit. Econ. 89 , 841 (1981).9. R. G. Ehrenberg, M. L. Bognanno, J. Polit. Econ. 98 , 1307

(1990).10. R. K. Merton, inThe Sociology of Science: Theoretical an

Empirical Investigations (Univ. Chicago Press, Chicago,1973), pp. 267278.

11. J. A. Evans,Science 321 , 395 (2008).12. O. Rodrguez-Ruiz, J. Educ. Adm. 47 , 250 (2009).13. R. G. Ehrenberg, J. Labor Econ. 21 , 267 (2003).14. P. Stephan, How Economics Shapes Science (Harvard

Univ. Press, Cambridge, MA, 2012), p. 41. 15. T. Friedman,The World Is Flat: A Brief History of the

Twenty-First Century (Farrar, Straus, and Giroux, New Y2005).

16. Y. Xie, Issues Sci. Technol. (Spring) (2014); http://issues.org/30-3/yu_xie/.

17. P. D. Allison, J. S. Long, Am. Sociol. Rev. 55 , 469 (1990).18. J. R. Cole,The Great American University: Its Rise to

Preeminence, Its Indispensable National Role, Why It MBe Protected (Public Affairs, New York, 2009).

19. D. A. King,Nature 430 , 311 (2004).

20. G. Firebaugh,The New Geography of Global IncomeInequality(Harvard Univ. Press, Cambridge, MA, 2003)

ACKNOWLEDGMENTS

Financial support provided by the Natural Science Foundatioof China (grant no. 71373012), Peking University, and thePopulation Studies Center at the University of Michigan, whireceives core support from the National Institute of ChildHealth and Human Development (R24HD041028). The authois grateful to C. Glovinsky, J. Hu, Q. Lai, and C. Zhang forresearch assistance.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/344/6186/809/suppl/DC1

0.83

Total research Endowment

Federal research0.81

0.79

0.77

0.75

0.73

0.71

0.691990 1995 2000 2005 2010

Inequalities in U.S. universitiesresearch fundingGini coe ffi cient

RESOURCE INEQUALITY. Gini for resource inequalityacross U.S. universities, 19002010. Data reflect totalresearch expenditure, federal research expenditure, andendowment for U.S. universities. Data from Center forMeasuring University Performance, which draws fromNational Science Foundation (research expenditures)and National Association of College and UniversityBusiness Officers (endowments). (See SM.) 10.1126/science.1252743

Scientific outputs and rewards are much moreunequally distributed than other well-being outcomes, such as education, earnings, or health.

Published by AAAS