Academic Research Underlying Industrial Innovations: Sources, Characteristics, and Financing

Author(s): Edwin Mansfield

Source: The Review of Economics and Statistics, Vol. 77, No. 1 (Feb., 1995), pp. 55-65
Published by: The MIT Press
Stable URL: http://www.jstor.org/stable/2109992 .
Accessed: 30/09/2013 08:35

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of
Economics and Statistics.

http://www.jstor.org

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMIC RESEARCH UNDERLYING INDUSTRIAL
INNOVATIONS:SOURCES, CHARACTERISTICS,
AND FINANCING
Edwin Mansfield*
Abstract-There has been no systematicstudyof the charac- ers attemptingto increase the economic payoff
teristics of the universitiesand academic researchersthat
seem to have contributedmost to industrialinnovation.Nor from the nation's academicresearch.In this pa-
do we know how such academicresearchhas been funded. per I report the results of a study based on data
This paper, based on data obtained from 66 firms in seven obtained from 66 firms in seven majormanufac-
majormanufacturingindustriesand from over 200 academic
researchers,sheds new light on the sources, characteristics, turing industries and from over 200 academic
and financingof academicresearchunderlyingindustrialin- researchers.Although the findingsare subjectto
novation.The findingsshould be of interest to economists many limitations, they shed new light on the
concernedwith technologicalchange and to policy makers
attemptingto increasethe economicpayofffrom the nation's sources, characteristics,and financing of aca-
academicresearch. demic researchunderlyingindustrialinnovation.

I. Introduction
II. Academic Research and Industrial
IN recent years there has been a great deal of Innovation
interest in the process by which firms benefit To begin with, it is worthwhilereviewingsome
from externallyperformedresearchand develop-
earlier findings regarding the extent to which
ment (R&D), and the extent and importanceof
technological innovations in various industries
such spillovers.Researchby Acs, Audretsch,and have been dependent on recent academic re-
Feldman (1992), Jaffe (1989), Mansfield(1991a, search. Based on data obtained from 76 firms in
1991b, 1992), Nelson (1988), von Hippel (1988), the seven industrieslisted in table 1, about 11%
and others indicate that technologicalchange in of their new productsand about 9% of their new
important segments of the economy has been processescould not have been developed(without
based significantlyon academic research. How- substantialdelay) in the absence of recent aca-
ever, there has been no systematicstudy of the demic research (defined as academic research
characteristicsof the universities and academic occurringwithin 15 years of the commercializa-
researchersthat seem to have contributedmost tion of the innovation).1As shown in table 1, the
to industrialinnovation.Nor do we know where percentageof new productsand processesbased
such universities or academic researchers have in this way on recent academicresearchseems to
obtained fundingfor the relevant R&D projects be highest in the drug industryand lowest in the
or how big or small their projectshave been. petroleumindustry.2
Informationof this sort would be of interestto
economistsand other scholarsconcernedwith the
processof technologytransferand to policymak-
1By "substantialdelay,"we mean a delayof a yearor more,
accordingto roughestimatesmade by the firms.
Receivedfor publicationMarch1, 1993. Revisionaccepted 2 New productsand processessometimescould have been
for publicationFebruary7, 1994. developedwithoutthe findingsof recent academicresearch,
* Universityof Pennsylvania. but it would have been much more expensiveand time-con-
The researchon whichthis paperis basedwas supportedby sumingto do so. In table 1, such cases are designatedas ones
a grant from the National Science Foundation,which, of where developmentoccurredwith "verysubstantialaid from
course, is not responsiblefor the findings.I am indebtedto recent academicresearch."Approximately8% of these firms'
LeonardLedermanof the Foundationfor his encouragement new productsand approximately6% of their new processes
and advice.Thanksalso go to the manyfirms(76 in sectionII, during 1975-85 fell into this category.Often, while it was
and 66 in subsequentsections),as well as about220 academic technically possible for the firm to have developed them
researchersthat provideddata. This paperwas presentedat without the findingsof recent academicresearch,it seemed
the January1993annualmeetingsof the AmericanEconomic economicallyundesirableto have tried it. Consequently,in a
Association.A preliminaryversionwas presentedat the Na- practicalsense, many of these innovationscould not have
tional Science Foundation,LehighUniversity,and the Maine been developed(withoutsubstantialdelay) in the absenceof
Scienceand TechnologyCommission. recent academicresearch.

Copyright X 1995 [ 55 1

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
56 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 1.-PERCENTAGE OF NEW PRODUCTS AND PROCESSES

BASED ON RECENT ACADEMIC RESEARCH, SEVEN INDUSTRIES, UNITED STATES, 1975-85
Percentage that Could Not Have Percentage that Was
Been Developed (without sub- Developed with Very
stantial delay) in the Absence Substantial Aid from
of Recent Academic Research Recent Academic Researcha
Industry Products Processes Products Processes

Information processing 11 11 17 16
Electronics 6 3 3 4
Chemical 4 2 4 4
Instruments 16 2 5 1
Pharmaceuticals 27 29 17 8
Metals 13 12 9 9
Petroleum 1 1 1 1
Industry mean 11 9 8 6
Source: Mansfield (1991a).
aSee footnote 2.

To prevent confusion,it is worthwhileto note table 1, new products first commercializedin

that many of the innovationsbased on recent 1982-85 that could not have been developed
academicresearchwere not inventedat universi- (without substantialdelay) in the absence of re-
ties. Academicresearchoften providesnew theo- cent academicresearchaccountedfor about $24
retical and empiricalfindings and new types of billion of sales in 1985 alone. And in these indus-
instrumentationthat are essential for the devel- tries, new processes first commercialized in
opmentof a new productor process,but does not 1982-85 that could not have been developed
providethe specificinventionitself. Thus, to cite (without substantialdelay) in the absence of re-
an old and well-knowncase, academicstudies by cent academicresearchresulted in about $7 bil-
ProfessorsKippingand Staudingerprovidedbasic lion in savingsin 1985 alone. While these figures
informationconcerningorgano-siliconchemistry are rough, they certainlyindicate that industrial
which laid the groundworkfor industrial sili- innovationin these industrieshas been based to a
cones.3 substantialdegree on recent academic research,
For each firm's new products and processes and crudeestimatessuggestthat the social rate of
introducedin 1975-85 that, accordingto the firm, returnfrom academicresearchhas been high.4
could not have been developed(withoutsubstan-
tial delay) in the absence of recent academic III. Sources of Academic Research
research, informationwas obtained concerning Underlying Industrial Innovations
the mean time intervalbetween the relevantaca- Although the foregoing results indicate that
demic researchfinding and the first commercial recent academicresearch has made a significant
introductionof the product or process. (If more contribution to innovation in these industries,
than one such research finding was needed for they tell us nothing about the kinds of academic
the developmentof the innovation,this time in- researchthat the innovatingfirmsbelieve to have
tervalwas measuredfrom the year when the last been most importantin this regard.To help illu-
of these findingswas obtained.) The mean time minate this topic, we drew a randomsampleof 70
lag in these industries was about 7 years. In majorfirmsfrom these industries.Each firmwas
interpretingthis result,note once againthat these
data pertainonly to recent academicresearch. 4Mansfield (1991a). The problemsin allocatingthe social
Particularlyin industries like drugs, instru- returnsbetween academicand industrialresearchare obvi-
ous. However,crude estimatessuggestthat the social rate of
ments, and informationprocessing,the contribu- return from academic research remains substantialeven if
tion of academicresearchto industrialinnovation seemingly generous assumptionsare made concerningthe
has been considerable.In the seven industriesin social rate of returnfromindustrialresearchand other indus-
trial innovationcosts. See Mansfield(1992). Of course, as I
have stressed here and elsewhere, these estimatesof social
3Jewkes,Sawers,and Stillerman(1969).Also see von Hip- rates of return are only rough. See Mansfield(1977, 1980,
pel (1988). 1991b).

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMIC RESEARCH UNDERLYING INDUSTRIAL INNOVATIONS 57

asked to cite about five academic researchers nology of the industryin question. In the elec-
whose work in the 1970s and 1980s contributed tronics industry,over 60% of the cited academic
most importantlyto the firm'snew productsand researcherswere in electricalengineeringor me-
processesintroducedin the 1980s. chanical engineeringdepartments.In the chemi-
Although our initial requests for information cal industry, almost 70% were in chemistryor
and cooperationwere made to the firms' chair- chemical engineeringdepartments.In the phar-
men, the respondents generally were the top maceutical industry, the cited academic re-
R&D executives who based their responses in searchers seemed to be scattered over a wider
considerablepart on detailed data obtainedfrom variety of fields and departments than in the
people at lowerlevels of their organizations.Most electronics or chemical industries,but this may
of the firms went to a considerable amount of have been due to the fact that the pharmaceutical
trouble to provide these data. Writtenresponses industry,as defined here, includes some medical
often were supplemented with interviews with productsfirms.8
relevant company personnel. Eventually,usable
data were obtainedfrom 66 of the 70 firmsin the IV. Effects of Faculty Quality, Scale of
sample.5Since these firms account, on the aver- Research Effort, and Geographical
age, for about a third of the R&D expenditures Proximity on Perceived
in these industries,the sample seems quite ade- University Contribution
quate.
Taken as a whole, these 66 firms cited 321 As pointed out in the previous section, one
academicresearchers.6Table 2 lists the universi- factor that would be expected to influence how
ties and types of departments cited most fre- frequentlya particularuniversityis cited in this
quently by the firms in each industry.In most way is the qualityof the university'sfaculty.An-
industries,the most frequently-citeduniversities other factorthat is often stressedby policymakers
are world leaders in science and technology.For is the scale of a university'sR&D activitiesin the
example, MIT, Berkeley, Illinois, Stanford, and relevant area: a critical mass of researchersand
CMU are most frequently cited in electronics; equipment is often regarded as necessary to
and Harvard,UCSF, Stanford,and Yale are most achieve high productivityin particularaspects of
frequentlycited in pharmaceuticals.But not all of academicresearch.Still anotherfactor is the geo-
the most frequently-citeduniversitiesare world graphicalproximityof a universityto the firmsin
leadersin the relevantfields.Thus, neitherWash- the sample. Because there are obvious advan-
ington Universitynor the Universityof Utah are tages in firmsworkingwith, and keeping abreast
among the top dozen departmentsof chemistry, of developmentsat, local colleges and universi-
according to the assessments of the National ties, one might expect that colleges and universi-
Academyof Sciences.7 ties located near manyof the firmswould tend to
With regardto type of department,it appears be cited relativelyoften.9
that the bulk of the cited academicresearchtook To test whether these factors are useful in
place in departmentsclosely related to the tech- explainingthe differences among universitiesin
the numberof times they were cited, we assume
that
5 The industrialdistributionof the firmsin the samplewas
as follows:electronics,14; informationprocessing,16; phar-
maceuticals,8; chemicals, 13; petroleum 5; metals, 4; and Yi = ao +a1Qj + a2R + a3L + zi, (1)
instruments,6. An attemptwas made to allocate the sample
optimallyamong industries(that is, with sample size being where Yi is the number of citations received by
proportionalto the total numberof firms in each industry
times the relevantstandarddeviation).
6 The numbercited by each industrywas as follows:elec-
8 The framefor our sampleof firmswas the list of firmsin
tronics, 84; informationprocessing,64; pharmaceuticals,47; BusinessWeek'sannual R&D Scoreboard.Included in the
chemicals,51; petroleum,28; metals,25; and instruments,22. pharmaceuticalindustrywere some medicalproductsfirms.
Eighteen academicresearcherswere cited by more than one 9For some relevant discussion, see Jaffe (1989), Jaffe,
firm, so the numberof distinct researcherscited is 303, not Trajtenberg,and Henderson (1993), Acs, Aubretsch, and
321. In tables 2-3, we weight each researcherby the number Feldman (1992), Government-University-Industry Round-
of firmsthat cited him or her. table and IndustrialResearchInstitute(1991),and Petersand
7 See NationalAcademyof Sciences(1982). Fusfeld(1982).

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
58 THE REVIEWOF ECONOMICSAND STATISTICS
TABLE 2.-UNIVERSITIES AND DEPARTMENTS CONTAINING THE LARGEST PERCENTAGE OF ACADEMIC RESEARCHERS CITED
BY 66 MAJOR FIRMS (IN THE ELECTRONICS, INFORMATION PROCESSING, PHARMACEUTICAL, CHEMICAL, PETROLEUM,
METALS, AND INSTRUMENTS INDUSTRIES) AS CONTRIBUTING MOST IMPORTANTLY (DURING THE 1970s AND 1980s)
TO THE DEVELOPMENT OF THEIR NEW PRODUCTS AND PROCESSES INTRODUCED IN THE 1980s
Electronics Information Processing Pharmaceuticals
University Department University Department University Department

MIT (15%) Elect. Eng. (50%) MIT (9%) Comp. Sci. (38%) Harvard (13%) Biology (11%)
Berkeley (13) Mech. Eng. (11) Berkeley (8) Elect. Eng. (10) UCSF (6) Chemistrya (22)
Illinois (8) Illinois (6) Mech. Eng. (10) Stanford (6) Pharmacology (14)
Stanford (7) Minnesota (6) Yale (6)
CMU (7) Stanford (6)

Chemicals Petroleum Metals

University Department University Department University Department

Washington (12%) Chemistry (53%) Delaware (11%) Chem. Eng. (46%) Utah (16%) Mat. Sci. (32%)
MIT (8) Chem. Eng. (15) MIT (7) Chemistry (8) MIT (12) Civil Eng. (20)
Utah (6) Notre Dame (7) Ohio State (8) Chemistry (12)
Princeton (7) Mech. Eng. (8)
VPI (7)

Instruments
University Department
Yale (9%) Chemistry (26%)
Indiana (9) Physics (16)
Radiology (16)
a Includesbiochemistry.

the ith university,Ri is the amount spent by the neering, computer science, biochemistry,chem-
ilhuniversityon researchand developmentin the istry, and chemical engineering. These ratings
relevantarea10in 1980,11 L1 is the percentageof range from 0 (not sufficientfor doctoral educa-
firms in the sample that are located in the same tion) to 5 (distinguished).For the electronicsin-
state as the ith university,zi is a random error dustry,we assumethat the relevantdepartmentis
term, and the a's are parameters.Qi, a measure electricalengineering;for informationprocessing,
of the qualityof the ith university'sfacultyin the we assumeit is computerscience;for pharmaceu-
relevant department, comes from the National ticals, biochemistry;for chemicals,chemistry;and
Academyof Sciences(1982),which has published for petroleum,chemicalengineering.
facultyratingsfor departmentsof electricalengi- Since Yi must be non-negative,Tobit estimates
were made of the a's. Recognizingthat the ef-
10For the electronicsindustry,we assumethat the relevant fects of each of the independent variables may
departmentis electricalengineering;for informationprocess- differfrom one industryto another,the statistical
ing, we assume it is computerscience; for pharmaceuticals,
biochemistry;for chemicals, chemistry;and for petroleum, analysiswas carriedout separatelyin each of the
chemicalengineering.For each industry,the relevantarea of five industrieswhere the sample size is reason-
R&D is assumedto be R&D in this designateddepartment, ably large. The results, shown in table 3, suggest
except for pharmaceuticals,where it is R&D in life sciences.
(No data are availablefor pharmacologyalone.)As we saw in that all three of the independentvariablesgener-
table 2, these departmentsare responsiblefor much of the ally seem to be related directly to Y; in all but
cited academicresearch,but by no means all of it. Thus, our one case, the estimates of a1, a2, and a3 are
analysisobviouslyis crude. Nonetheless,given that no infor-
mationhas been availableheretoforeon this topic, the results positive, and in about half of the cases they are
should be of interest.The data come from NationalScience statisticallysignificant.However,there is consid-
Foundation(1985a). erablevariationin Y that is unexplainedby equa-
11In the case of chemicalengineeringand electricalengi-
neering,the R&D expendituredata pertainto 1983,not 1980. tion (1), which would be expected both because
See NationalScience Foundation(1985a). of obviousimperfectionsin the data and because

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMICRESEARCHUNDERLYINGINDUSTRIALINNOVATIONS 59
TABLE 3.-TOBIT ESTIMATES OF COEFFICIENTS IN EQUATION (1)

Independent Variables
Industry Intercept R Q L

Electronics -11.93 0.029 3.578c 0.030

(2.48) (0.093) (0.830) (0.035)
Information processing - 3.857c 0.60. 0.608 0.08
(1.384) (0.264) (0.520) (0.037)
Drugs - 5.300c 0.066c 0.374 0.083b
(1.322) (0.022) (0.439) (0.035)
Chemicals - 3.987a 0.549 - 0.171 0.096a
(2.419) (0.531) (0.953) (0.056)
Petroleum - 4.025c 0.701b 0.693b 0.014
(1.245) (0.300) (0.336) (0.024)
aSignificant at the 0.10 level.
b Significant at the 0.05 level.
Significant at the 0.01 level.

the independentvariables are by no means the a 1-in-5 chance that a firm in our sample would
only factors influencing Y."2 work with a universitydepartmentwith only a
To see the extent to which firmsare willing to "marginal"faculty,but if so, there was an over-
trade off faculty quality(as measuredby Q) for whelminglikelihoodthat this college or university
geographicalproximity,we chose a sample of would be located within 100 miles of the firm's
nine majorchemical,drug, and informationpro- R&D laboratories.14
cessingfirms,each of which estimatedthe proba-
bility that it would support research of a par-
V. Size of ResearchProjectsand Sources
ticular type at a university department with a
of FinancialSupport
specified value of Q and at various distances
from the firm's R&D laboratories.The results, While the foregoingfindingsindicatethat there
pertainingto 20 types of research,indicate that, tends to be a direct relationshipbetween the size
holding faculty quality constant, the probability of a university'sR&D expendituresand its per-
that a firm will support research at a college or ceived contributionto industrial innovation (in
universityless than 100 miles away tends to be the relevantarea), this sheds no light on the size
several times as great as the probabilitythat it of the researchprojectscarriedout by the cited
will supportthis research at a college or univer- academicresearchers.Based on data (in table 4)
sity 1,000 or more miles away. However, for re- we obtained from about 90% of these re-
search that (from the firm's vantage point and searchers,15their average annual academic re-
based on NSF's definitions13)is basic, geographi- search expendituresduring the 1970s and 1980s
cal proximityseems to play a smallerrole than for generallywere less than $250,000(about $425,000
applied R&D; that is, firms seem more likely to in 1992 dollars).16Outside the pharmaceutical
insist on high faculty quality and pay less atten-
tion to location in choosing universities to do 14
For furtherdiscussion,see Lee and Mansfield(1994).
15
basicresearch.For appliedR&D, therewas about Of the 303 distinctcitations,about 70 were to academic
researchersworking at foreign universitiesor to an entire
12
The numberof colleges and universitiesthat could be department.We correspondedwith each of the remaining
includedin this analysiswas electronics,75; informationpro- (roughly235) cited academicresearchersto obtain data con-
cessing,48; pharmaceuticals, 84; chemicals,93; andpetroleum, cerningtheir researchbudgets,sources of researchsupport,
67. If ordinaryleast squaresregressionsare run, the pattern and the influenceof users and funders on their choices of
of significantregressioncoefficientsis much like that in table projectsand directionstaken.Eventually,after telephoneand
3, and R2 is 0.30 to 0.50, exceptfor chemicals,whereit is only other follow-ups,these data were obtained(whollyor in part)
0.08. If Tobit estimatesare made, the normalscale parameter from over 90%of them. Thus, the responserate is very high
is about 3 in electronics and chemicals, 2 in information for a surveyof this sort. To obtain more detailed data, we
processingand pharmaceuticals, and 1.5 in petroleum. contacted a subsampleof these academicresearchers,and
13 The National Science Foundationdefines basic research discussedthese questionsat lengthwith them.
in industryas researchthat advancesscientificknowledgebut 16 The National Science Foundationand others often use
does not have specific commercialobjectives,althoughsuch the GNP (or GDP) deflatoras a price indexfor R&D inputs.
investigationsmaybe in fields of presentor potentialinterest The abovefiguresin 1992dollarsare based on this index.See
to the firm. Mansfield(1987).

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
60 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 4.-PERCENTAGE DISTRIBUTION OF CITED ACADEMIC RESEARCHERS BY THEIR AVERAGE ANNUAL

RESEARCH EXPENDITURES, 1970-89
Industry Citing the Academic Researcher
Average Annual Expenditure Information
of Researchera Electronics Processing Pharmaceuticals Chemicals Petroleum Metals Instruments

Less than $100,000 15 29 3 19 24 29 25

$100,000 to under $250,000 45 35 18 44 35 35 35
$250,000 to under $500,000 28 12 36 30 18 29 25
$500,000 to under $1 million 5 3 24 7 12 0 5
$1 million and over 8 21 21 0 12 6 10
Total 100 100 100 100 100 100 100
a These figures include overhead charged by the university.

industry,where the cited academic researchers portfolio of projects often are interrelated,and
tended to have comparativelylarge budgets,rela- because the firms often cited more than one
tivelyfew of them seemed to requiretotal annual project by a researcher,this seemed to be the
research budgets exceeding $500,000 (about best way to begin.
$850,000 in 1992 dollars).17The median annual The first point to note is that practicallyall of
research budget of the cited academic re- the cited academicresearchershad some govern-
searcherswas about 5 times as great as the R&D ment support for their research. In about two-
expenditureper academic scientist or engineer thirdsof the cases, it came, at least in part, from
(with a doctorate) in the relevant field, but the the National Science Foundation(NSF). The De-
bulk of these projects certainly fall under the partmentof Defense (DOD) was also important,
18
headingof "little science." particularlyin electronics,and the NationalInsti-
Besides obtainingdata from the cited academic tutes of Health (NIH) played a verymajorrole in
researchersconcerning the levels of their aca- supporting academic researchers cited by the
demic researchbudgets,we also got information health-relatedindustries,especiallypharmaceuti-
fromthem regardingthe sourcesof their financial cals. The Departmentof Energy(DOE) and the
support. Since our data pertain to all academic National Aeronautics and Space Administration
research carried out by each cited researcher (NASA) also providedsubstantial,but more lim-
during the 1970s and 1980s, they indicate the ited, support. In terms of dollar support, the
overall contours of a researcher'ssupport, not federal governmentprovidedabout two-thirdsof
just the supportof whateverparticularprojectthe the funding for the cited academic researchers
firmcited. Becausevariouspartsof a researcher's (table 5), which was somewhatless than the per-
centage of R&D expenditures in colleges and
17
Of course,the crudenessof these data shouldbe empha- universitiesfinancedby the federalgovernmentin
sized. Averages over such a long period of time are very the relevantfields (see note 10).19
rough, and the basic data sometimesare only approximate.
But the generalconclusionput forth in the text seems to be While governmentsupport was obviously im-
quite robust;it seems unlikelyto be affectedmuch by errors portantto the vast majorityof the cited academic
of this sort. researchers,this does not mean that industrydid
18 In each of the followingfields in 1983,the R&D expendi-
tures (in this field) at universitiesand colleges per doctoral not supportmanyof them as well. Overfour-fifths
scientist or engineer (in this field) employed by four-year of the cited academic researchersgot research
collegesor universitieswere the following:electricalengineer- funds from industry.However,industrygenerally
ing, $64,991;computerscience,$45,299;life sciences,$73,448;
chemistry,$24,153;chemical engineering,$51,656. See Na- supported a substantiallysmaller percentage of
tional Science Foundation(1985a,c). As pointed out in foot-
note 10, these are regardedas the "relevantfields"for the
electronics,informationprocessing,pharmaceutical, chemical, 19For the five industriesin table 5 for which a comparison
and petroleumindustries.Note that, in the pharmaceutical can be made, the federal governmentprovided,on the aver-
industry,the relevantfield is not the same as the "relevant age, about 66% of the funding for the cited academicre-
department"underlying Qi because of the nature of the searchers, as compared with about 68% of the academic
availabledata. R&D in the relevantfields.

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMICRESEARCHUNDERLYINGINDUSTRIALINNOVATIONS 61
TABLE5.-SOURCES OFFUNDINGFORCITEDACADEMICRESEARCHERSa
AND FOR ALLACADEMIC
R&D IN RELEVANT
FIELD
Percentage of Cited Mean Percentage of
Academic Researchers Research Budgets of
Where Research Was Cited Academic Percentage of Academic
Funded (wholly or in Researchers Funded R&D in Relevant
Citing the part) by byb Field Funded byc
Academic Federal Federal Federal
Researcher Government Firms Government Firms Government Private

Electronics 95 86 69 24 75 7
Information
processing 91 65 51 22 73 9
Pharmaceuticals 91 76 85 12 61 9
Chemicals 96 78 71 21 74 8
Petroleum 100 94 56 10 58 13
Metals 89 94 44 47
Instruments 100 80 72 18
Mean 95 82 64 22 68 9
Source: See the text.
a For those researcherswho were involved in academic research during only part of this period, the figures
pertain to only this part of the period. For those who were involvedin academicresearchduringthe entire period
and whose pattern of support was significantlydifferent during the 1980s than during the 1970s, we used the
pattern of support during the 1970s, since this was generally the period when the work occurred for which the
academicresearcherwas cited.
b For those researcherswhose pattern of support was significantlydifferent during the 1980s than during the
1970s, the figuresfor the federal governmentduringthe 1980s tend to be lower than those shown above, and the
figuresfor firmsfor the 1980s tend to be higher than those shown above.
c The "relevant field" for each industryis given in footnote 10. These figures pertain to 1983, and come from
National Science Foundation(1985a).

the total researchbudgets of the cited academic for example, the National Institutes of Health
researchers than did government (22% versus accountedfor an overwhelmingpercentageof the
64%). Only in the metals industry, where the governmentfundingof the academicresearchers
sample size is relativelysmall, did industrysup- cited by the pharmaceuticalindustry, and the
port exceed federal governmentsupport. None- Defense Departmentsupporteda largerpercent-
theless, as mightbe expected,firmsseemed to be age of the governmentfunding of academic re-
more important as sources of support for the searchers cited by the electronics industrythan
cited academicresearchersthan for all academic did any other agency (table 6). For a particular
researchersin the relevantfields. The percentage industry,the agencies that were the leading fun-
of R&D expendituresin colleges and universities ders of research in the relevant field tended to
financedby privatesourceswas about 10 percent- providethe biggest share of supportfor the cited
age points lower, on the average, than the per- academicresearchers,but they almostalwayswere
centage of the cited academic researchers'bud- less importantto the cited academicresearchers
gets financedby industry.20 than to researchersin the relevantfield because a
The relative importanceof particulargovern- substantialproportionof the cited academicre-
ment agencies varied from industryto industry; searcherswere outside the relevantfield. (Recall
table 2.) By the same token, agencies that were
20
For the five industriesin table 5 for which a comparison not the leading funders of research in the rele-
can be made, firmsprovided,on the average,about 18% of
the fundingfor the cited academicresearchers;on the other vant field tended to providea biggershare of the
hand, privatesourcesprovided,on the average,about 9% of support for the cited academic researchersthan
the fundingfor all academicR&D in the relevantfields. Of their share of research funding in the relevant
course, the fact that the cited academicresearchersreceived
more industrialfundingthan the typicalacademicresearcher field would indicate.
in the relevant field helps to explain why, as noted in the A substantial number of the cited academic
previous paragraphof the text, the proportionof support researchersreported that their sources of finan-
from the federal governmentis somewhatless for the cited
academicresearchersthan for all academicresearchersin the cial support shifted considerablyfrom the 1970s
relevantfield. to the 1980s. In the later decade, more of their

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
62 THE REVIEWOF ECONOMICSAND STATISTICS
TABLE6.-MAJOR SOURCES
OFFEDERAL
FUNDINGFORCITEDACADEMICRESEARCHERSa
R&D INTHERELEVANT
ANDFORALLACADEMIC FIELD
Industry Citing the Academic Researcher
Federal Information Pharma-
Agency Electronics Processing ceuticals Chemicals Petroleum Metals Instruments

Percentage of Cited Academic Researchers with Federal Funding Whose Research Was Financed
(wholly or in part) by Each Agency
NSF 71 65 49 61 88 69 70
DOD 77 48 20 46 41 44 45
NIH 13 20 84 39 24 12 50
DOE 8 16 13 27 24 25 25
NASA 28 7 0 11 12 7 20

Mean Percentage of Research Budgets of Cited Academic Researchers Financed

by Each Agency (as percent of mean percentage financed by all federal agencies)
NSF 32 27 6 24 57 30 36
DOD 43 31 6 14 20 36 21
NIH 7 12 65 31 11 2 18
DOE 3 6 6 17 4 9 7
NASA 7 6 0 6 2 0 6

Percentage of Federal Obligations for Academic Research in Relevant Field

by Each Agencyb
NSF 27 43 6 39 76 31
DOD 64 52 1 12 15 42
NIH 0 0 78 25 0 0
DOE 3 2 2 14 7 19
NASA 5 3 c 3 1 8
Source: See the text.
a See footnote a of table 5.
b
The "relevantfield" for each industryother than metals is given in footnote 10; for metals, it is metallurgyand materials.These figurespertain to 1981,
and come from National Science Foundation(1985b).
c Less than 0.5%.

funding came from industry, less from govern- VI. Complementarityof Government-Funded
ment. In some cases, the shift was large; for and Industry-FundedWorkof Cited
example, for five academic researcherscited by AcademicResearchers
the electronics industry,about 80% of their fi-
Practicallyall of the cited academicresearchers
nancial support came from the federal govern-
supported financially by both government and
ment (and 10% from firms) in the 1970s, as
industrybelieve that their industry-fundedwork
comparedwith about 20% from the federal gov-
complementedtheir government-funded work. In
ernment(and 80% from firms)in the 1980s.
the bulk of the cases, their government-funded
Most of the cited academic researchers re-
work preceded their industry-fundedwork, and
ceived support from multiple sources of funds.
very often their industry-fundedwork was aimed
The bulk of them were supportedby two or more
at extending,deepening,or furtheringthe results
of the followingfour sources:NSF, NIH, DOD,
of their previousgovernment-fundedwork (table
or firms. In electronics in particular,there fre-
7).21 On the average, more than 50% felt that
quentlywas supportfrom three or more of them.
their government-fundedwork was more funda-
Many of the cited academicresearchersseem to
have been entrepreneurialin outlook. Based on
21
discussions with a sample of them, there was The data underlyingtable 7 were collectedfrom a subset
of the academicresearchersincludedin tables 4-6 that had
considerableinteraction between them and po- obtainedfundingfrom both governmentand industry.In all,
tential sourcesof funding.As would be expected, data were gotten from 83 of the cited academicresearchers,
a number of them complained that it was very the industrybreakdownbeing: electronics, 35; information
processing,8; chemicals,14; pharmaceuticals,15; petroleum,
difficultto obtain support for projects that they 11. Smallersampleswere collected in the metals and instru-
regardedto be of prime importance. ments industries.

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMICRESEARCHUNDERLYINGINDUSTRIALINNOVATIONS 63
TABLE7.-RELATIONSHIPS
BETWEENGOVERNMENT-FUNDED
ANDINDUSTRY-FUNDED
WORK
OF CITEDACADEMIC WITHBOTHTYPEsOF FUNDINGa
RESEARCHERS
Information
Electronics Processing Drugs Chemicals Petroleum

(percentage of cited academic researchers)

Government-funded work
preceded industry-
funded work 85 62 67 86 64
Government-funded work
regarded as
more fundamental than
industry-funded work 76 50 57 50 36
Problems worked on in
academic research
frequently or
predominantly developed
out of industrial
consulting 58 62 43 57 73
Continuing consulting
relationships with
firms supporting
academic research 65 75 100 79 82
Students took jobs with
firms financing
academic research 77 87 73 64 91
Source: See footnote 21.
a The metals and instrumentsindustriesare omitted because the sample sizes are quite small.

mental than their industry-fundedwork, but in from ideas and problems they encountered in
the petroleum, chemical, and informationpro- industrialconsulting.23
cessingindustries,this figuredid not exceed 50%. Consequently, although their government-
In a large majority of the cases, the cited funded work tended to precede their industry-
academic researchershad continuing consulting fundedwork,the ideas and problemsthey worked
relationshipswith at least some of the firmssup- on (sometimesinitially,and certainlylater on) in
porting their academic research, and their stu- both their government-funded work and
dents have taken jobs with at least some of these industry-fundedworkoften were influencedin an
firms.22Moreover, in all industries other than importantway by their consultingand industrial
drugs,over half of the cited academicresearchers
reported that the problems they worked on in
their academic research frequentlyor predomi-
nantly developed out of their industrialconsult- 23
About two-thirdsof the cited academicresearchersin our
ing-and in many cases, the cited academic re- samplefelt that their own views were of primaryimportance
searchers' government-funded work stemmed in determiningthe nature and directionof the researchthey
carried out. Only about one-tenth of the cited academic
researcherssaid that potential funders or users of research
had the primaryinfluence over these matters in their case.
However,this does not mean that the cited researchersdid
22
One importantway that these researchersbecame famil- not take into accountthe views of potentialfundersand/or
iar with the availabilityof industrialfundingand the needs of usersof research.On the contrary,over half of them reported
industrialuserswas throughconsulting.Over90%of the cited that their choice of problems and direction of work were
academicresearchershave been consultantsto industry,the influenced considerably(sometimes primarily)by potential
medianamountof time devotedto industrialconsultingbeing fundersand/or users.The extentof this influenceseemed to
about 30 days per year. The percentagedevoting48 or more be greatest for researcherscited by the electronicsindustry
days to consulting was largest among those cited by the and for researcherssupported heavily by DOD. Also, it
electronicsindustry,and smallest among those cited by the seemed to be relativelygreat for researcherswhose support
pharmaceuticalindustry. was largelyfrom industry.

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
64 THE REVIEWOF ECONOMICSAND STATISTICS

experience.Whilemuchacademicresearchis car- researchsuch ties may be weaker and more spo-

ried out with little regard for industry'sneeds, radic. Also, the difference between highly-rated
this does not seem to have been true of the work and modestly-rateduniversitydepartmentsin ef-
of most of the cited academicresearchers. fectivenessand productivitymay well be greater
for basic researchthan for applied R&D.
With regard to governmentfunding, the Na-
VII. Summaryand Conclusions
tional Science Foundation, Department of De-
A substantialproportionof industrialinnova- fense, and National Institutesof Health seem to
tions in high-technologyindustrieslike drugs,in- play the predominantroles in financingacademic
struments,and informationprocessinghave been research cited by the seven industries studied
based directly on recent academic research, al- here. Whereas the National Institutes of Health
thoughin manycases the inventionitself did not are particularlyimportantin supportingacademic
stem from the universities.Based on the findings researcherscited by the health-relatedindustries
of this paper, the extent to which a universityis (drugs,chemicals,and instruments),the National
credited by firms in the electronics, information Science Foundationand the Departmentof De-
processing,drug, chemical,and petroleumindus- fense are major supporters of academic re-
tries with making major contributionsto these searcherscited acrossthe board.The leadingrole
firms'innovationstends to be related directlyto playedby the NationalScience Foundationwould
the qualityof the university'sfacultyin the rele- probablybe expected. The fact that the Depart-
vant department(accordingto the NAS ratings), ment of Defense is so importantin supporting
to the size of its R&D expendituresin relevant academicresearcherscited by such a wide swath
fields, and to the proportion of the industry's of industriesis noteworthy.The Department of
members located nearby. However, there is a Energy and the National Aeronauticsand Space
considerableamountof variationunexplainedby Administrationseem to play lesser roles, at least
these three factors, and in all industriesthe ef- in these industries.
fects of one or two of these factors, while they Finally, the findings regardingthe funding of
almostalwayshave the signs indicatedabove, are these projects seem to reveal a complex web of
not statisticallysignificant. financial and intellectual relationships among
The fact that a university'sfacultyrating(in the academic researchers,governmentagencies, and
relevantdepartment)tends to be related directly firms, Practicallyall of the cited academic re-
to its perceivedcontributionto industrialinnova- searchers were supported, at least in part, by
tion seems to contradictthe widely-heldview that federal funds. If governmentagencieswere trying
the highest-rankeduniversitydepartmentsfocus to supportwork of this sort, they seemed to miss
so heavilyon researchwith a relativelylong-term relativelyfew of the leading researchers,as indi-
payoff that they would be unlikely to show up cated by the data presented here. Industrytoo
well in citations of this sort. However, the rela- supportedthe work of most of them, althoughits
tionship between faculty rating and contribution contributionwas much smaller-and tended to
to industrialinnovationis so weak in several of come later-than that of the government.Once
these industries that it seems likely that many they achieved prominence,practicallyall of the
modestly-rankeddepartmentsplay as big a role in cited academicresearcherswere involvedin con-
this regardas some of the highest-rankeddepart- sulting relationshipswith firms,which helped to
ments. stimulatemany of the ideas and topics taken up
With regardto their researchsupportfor uni- subsequently in both their government-funded
versities,firmstend to trade off facultyquality(as and industry-fundedwork.
measured by the NAS ratings)for geographical
proximity, particularlyin the case of applied REFERENCES
R&D. For basic research,they seem to pay less
Acs, Zoltan,DavidAudretsch,and MaryannFeldman,"Real
attention to location in choosing universitiesto Effects of AcademicResearch:Comment,"American
workwith and support,perhapsbecause in many Economic Review 82 (1992), 363-367.
kinds of applied R&D, it is very useful for aca- Government-University-Industry Research Roundtableand
IndustrialResearchInstitute,"IndustrialPerspectives
demic and firm personnel to interact and work on Innovation and Interactions with Universities"
togetheron a face-to-facebasis, whereas in basic (Washington,D.C.: NationalAcademyPress, 1991).

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions
 ACADEMICRESEARCHUNDERLYINGINDUSTRIALINNOVATIONS 65
Jaffe,Adam,"Real Effectsof AcademicResearch,"American Industrial Innovations,"QuarterlyJournal of Eco-
EconomicReview79 (1989),957-970. nomics91 (1977),221-240.
Jaffe, Adam, Manuel Trajtenberg,and Rebecca Henderson, National Academyof Sciences, An Assessmentof Research-
"GeographicalLocalizationof Knowledge Spillovers DoctorateProgramsin the UnitedStates (Washington,
As Evidencedby Patent Citations,"QuarterlyJournal D.C.: NationalAcademyPress, 1982).
of Economics108 (1993),577-598. NationalScienceFoundation,University-Industry ResearchRe-
Jewkes, John, David Sawers, and Richard Stillerman,The lationships(Washington,D.C.: GovernmentPrinting
Sourcesof Invention,secondedition(New York:W. W. Office,1982).
Norton,1969). , Academic Science/Engineering: R&D Funds, 1983
Lee, J., and E. Mansfield,"IndustrialSupportof Academic (Washington, D.C.: Government Printing Office,
Research," Center for Economics and Technology, 1985a).
Universityof Pennsylvania,1994. , Federal Funds for Research and Development, 1981,
Mansfield,Edwin,"AcademicResearchand IndustrialInno-
vation,"ResearchPolicy 20 (1991a),1-12. 1982,and 1983 (Washington,D.C.: GovernmentPrint-
ing Office,1985b).
, "Academic Research and Industrial Innovation: A
FurtherNote," ResearchPolicy 21 (1992),295-296. , Characteristicsof Doctoral Scientists and Engineers in
, "Estimates of the Social Returns from Research and
the United States: 1983 (Washington,D.C.: Govern-
Development,"in M. Meredith, S. Nelson, and A. ment PrintingOffice,1985c).
Teich (eds.), Science and Technology Yearbook Nelson, Richard,"InstitutionsSupportingTechnicalAdvance
(Washington,D.C.: AmericanAssociationfor the Ad- in Industry,"AmericanEconomicReview 76 (1986),
vancementof Science, 1991b),313-320. 186-189.
, "PriceIndexesfor R&D Inputs,1969-83," Manage- Peters, Lois, and HerbertFusfeld,"CurrentU.S. University-
mentScience33 (1987), 124-129. Industry Research Connections," in University-In-
, "BasicResearchand ProductivityIncreasein Manu- dustryResearchRelationships(Washington,D.C.: Na-
facturing," American Economic Review 70 (1980), tional ScienceBoard,1982).
863-873. von Hippel, Eric, The Sources of Innovation(New York:
, et al., "Social and Private Rates of Return from OxfordUniversityPress, 1988).

This content downloaded from 128.111.121.42 on Mon, 30 Sep 2013 08:35:19 AM

All use subject to JSTOR Terms and Conditions