Belde bos, René; B ai o, Naza eno; Wang, Jian
A icle — Published Ve sion
He e ogeneous uni e si y esea ch and i m R&D loca ion
decisions: esea ch o ien a ion, academic quali y, and
in es men ype
The Jou nal o Technology T ans e
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Belde bos, René; B ai o, Naza eno; Wang, Jian (2024) : He e ogeneous uni e si y
esea ch and i m R&D loca ion decisions: esea ch o ien a ion, academic quali y, and in es men
ype, The Jou nal o Technology T ans e , ISSN 1573-7047, Sp inge US, New Yo k, NY, Vol. 49, Iss. 5,
pp. 1959-1989,
h ps://doi.o g/10.1007/s10961-024-10066-w
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/315302
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Vol.:(0123456789)
The Jou nal o Technology T ans e (2024) 49:1959–1989
h ps://doi.o g/10.1007/s10961-024-10066-w
1 3
He e ogeneous uni e si y esea ch and i m R&D loca ion
decisions: esea ch o ien a ion, academic quali y,
andin es men ype
RenéBelde bos1,2,3· Naza enoB ai o4· JianWang5
Accep ed: 19 Janua y 2024 / Published online: 2 Ma ch 2024
© The Au ho (s) 2024
Abs ac
Uni e si ies play an impo an ole in egional de elopmen and inno a ion and engage
wi h he indus y h ough a ious channels. In his pape , we examine he ole o he e oge-
neous cha ac e is ics o uni e si y esea ch, in pa icula uni e si ies’ o ien a ion owa ds
basic o applied esea ch and he quali y o his esea ch, in a ac ing i ms’ R&D in es -
men . We analyze he loca ion decisions in he Uni ed S a es by o eign mul ina ional
i ms a he le el o me opoli an a eas. We con as esea ch and de elopmen p ojec s
and explo e whe he hey a e d i en by di e en ac o s. We ind ha he d i e s o loca-
ion choice di e impo an ly as a consequence o he ype o he ocal R&D in es men o
he i m. Uni e si ies wi h an o ien a ion owa ds applied scien i ic esea ch and exhibi -
ing highe academic quali y o applied esea ch a ac mo e R&D in es men ocusing on
de elopmen ac i i ies. In con as , i ms’ in es men s in esea ch ac i i ies a e a ac ed
by he academic quali y o basic scien i ic esea ch o local uni e si ies. Hence, inc eased
uni e si y emphasis on academic engagemen and applied esea ch may ha e nega i e con-
sequences o indus ial esea ch in he egion.
Keywo ds Loca ion choice· R&D FDI· Indus y science links· Basic and applied
esea ch· Academic quali y
* Jian Wang
jian.wang@lancas e .ac.uk
René Belde bos
[email p o ec ed]
Naza eno B ai o
[email p o ec ed]
1 Depa men o Managemen , S a egy andInno a ion, Facul y o Economics andBusiness, KU
Leu en, Naamses aa 69, 3000Lou ain, Belgium
2 UNU-MERIT, Boschs aa 24, 6211AXMaas ich , TheNe he lands
3 School o Business andEconomics, Maas ich Uni e si y, Tonge ses aa 53,
6211LMMaas ich , TheNe he lands
4 Depa men o Managemen , S a egy andInno a ion, B ussels Campus, Facul y o Economics
andBusiness, KU Leu en, Wa moesbe g 26, 1000B ussel, Belgium
5 Lancas e Uni e si y Leipzig, Nikolais asse 10, 04109Leipzig, Ge many
1960
R.Belde bos e al.
1 3
JEL Classi ica ion F21· F23· O32
1 In oduc ion
Uni e si ies play an impo an ole in egional de elopmen and inno a ion, and a e
inc easingly expec ed o pe o m he ‘ hi d mission’ o engagemen in se ing he economy
and socie y, in addi ion o hei adi ional missions o eaching and esea ch (Bozeman,
2000; Bozeman e al., 2015; Pe kmann e al., 2013). The echnology ans e li e a u e has
ex ensi ely s udied uni e si y-indus y collabo a ion (Belde bos e al., 2021; B uneel e al.,
2010; Lehmann & Men e , 2016; Rybnicek & Königsg ube , 2019), uni e si y pa en ing
and licensing (Geuna & Nes a, 2006; Mazzoleni, 2006; Mowe y e al., 2001), uni e si y
spin-o s and en ep eneu ship (Ma hisen & Rasmussen, 2019; Wang e al., 2022; Zucke
e al., 1998), he ole o in e media ies such as echnology ans e o ices (Bolzani e al.,
2021; O’Kane e al., 2021; Siegel e al., 2003) and incuba o s a uni e si ies (Cado in e al.,
2021; Ro hae mel & Thu sby, 2005; You ie & Shapi a, 2008).
This expanding li e a u e has con ibu ed impo an insigh s in o how uni e si ies in e -
ac wi h incumben i ms o spawn new en u es. An impo an conclusion is ha uni e si-
ies di e impo an ly in hei esou ces and capabili ies, and hei academic engagemen
wi h indus y and in ol emen in en ep eneu ial ac i i ies (G imaldi e al., 2011; Pe k-
mann e al., 2013). Compe i i e and socie al p essu es and declining go e nmen suppo
ha e p o ided impe us o an inc eased ocus on academic en ep eneu ship on he pa o
uni e si ies, and g ea e in ol emen o indus y in unding uni e si y esea ch (Siegel
& W igh , 2015). An in luen ial policy ecommenda ion o ad ancing he hi d mission
o engagemen is o shi he ocus o uni e si y esea ch om basic esea ch ha is con-
ce ned wi h undamen al unde s anding o applied esea ch ha ocuses on p ac ical u il-
i y (Gibbons, 1994), al hough whe he uni e si ies a e d i en away om cu iosi y-d i en
basic esea ch owa d applied esea ch is s ill a subjec o deba e (Fini e al., 2021; Hen-
de son e al., 1998; Mowe y & Ziedonis, 2002; Pe kmann e al., 2021; Thu sby & Thu sby,
2011). A salien unanswe ed ques ion is whe he he e ogenei y in he cha ac e is ics o
uni e si y esea ch, in pa icula hei o ien a ion owa d basic o applied esea ch, has a
ma e ial impac on he ole uni e si ies play in economic de elopmen . In his pape , we
seek o answe his ques ion in a con ex o uni e si y esea ch a ac ing ( o eign) R&D
in es men s o hei egions.
A numbe o s udies ha e ound e idence o a posi i e associa ion be ween uni e -
si y esea ch and indus ial R&D in es men s, a he egional le el (Ab amo sky e al.,
2007; Au an -Be na d, 2001; Belde bos e al., 2014, 2017; Can well & Pisci ello, 2005).
Howe e , i emains unclea how his associa ion di e s depending on he cha ac e is ics
o academic esea ch he uni e si y is in ol ed in. In his pape , we ocus on wo impo -
an ea u es o academic esea ch: i s applied o basic na u e and i s academic quali y. A
pe inen ques ion is whe he a ocus on applied a he han basic scien i ic esea ch may
makes uni e si y esea ch mo e di ec ly ele an and accessible o i ms. On he one hand,
applied esea ch migh be mo e di ec ly use ul o he indus y and is p esc ibed as a key
o ad ancing he hi d mission (Gibbons, 1994; Nelson, 2003). On he o he hand, gi en
he obse a ion ha echnological inno a ion is inc easingly elying on science (Ma x &
Fuegi, 2020), while basic esea ch ac i i ies conduc ed in e nally by i ms a e declining
(A o a e al., 2018) and equi e a s ong in ellec ual p ope y p o ec ion egime (Sime h &
Ra o, 2013), i is concei able ha he complemen a y ole o basic esea ch a uni e si ies
1961
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
is impo an o i ms. A second ele an ques ion i is whe he science ha is pe cei ed
o be o high quali y by o he scien is s will also be mo e aluable o co po a e R&D,
gi en he con lic ing logics o science and echnology (Ali & Gi elman, 2016; Gi elman
& Kogu , 2003; Saue mann & S ephan, 2013). Unde s anding he alignmen o con lic
be ween he academic and indus y logics in e ms o quali y s anda ds is impo an o
in o ming policy and manage ial p ac ices suppo ing he “ hi d mission.”
We aim o p o ide a nuanced answe o hese ques ions by ecognizing ha i ms’ R&D
in es men s a e also he e ogeneous in na u e and objec i es. We examine o wha ex en
he in luence he e ogeneous uni e si y esea ch depends on he ype o R&D in es men s
unde aken by he i ms: i.e., whe he he R&D in es men ocuses on esea ch ac i i ies o
de elopmen ac i i ies. The mo i a ion o (mul ina ional i ms’) R&D in es men s ha e
gene ally been dis inguished be ween ma ke adap a ion (de elopmen ) and knowledge
sou cing and c ea ion ( esea ch) (Belde bos e al., 2009; Shimizu ani & Todo, 2008; on
Zed wi z & Gassmann, 2002). While de elopmen ac i i ies may be expec ed o bene i
mos om applied academic esea ch, i ms’ esea ch ac i i ies may be mo e likely o d aw
on basic scien i ic esea ch conduc ed a uni e si ies.
We examine he ole o such he e ogeneous academic esea ch in a ac ing he e ogene-
ous R&D in es men s by conside ing loca ion decisions o o eign mul ina ional i ms in
he Uni ed S a es a a ine-g ained egional le el o Me opoli an S a is ical A eas (MSAs),
which a e egions delinea ed in e ms o economic in eg a ion (Mowe y & Ziedonis, 2015).
Analyzing o eign i ms’ R&D in es men loca ion decisions has he ad an age ha hese
i ms a e ela i ely ee o choose a loca ion based on i s me i s, as hey ha e no home
egion in he US in which hey a e s ongly embedded ha may in luence such decisions.
We de elop geocoded academic publica ion da a based on Cla i a e’s Web o Science o
cha ac e ize he academic esea ch p o ile o uni e si ies in each MSA ac oss scien i ic
domains, including he academic quali y o hei publica ions (scien i ic ci a ions ecei ed),
and hei o ien a ion owa d basic o applied esea ch. We ake in o accoun he a ying
ele ance o egional academic esea ch (Hausman, 2020) o R&D in es men s ac oss
indus ies by u ilizing a conco dance be ween science ields, echnologies, and indus ies.
Ou analysis akes in o accoun o he channels h ough which uni e si y esea ch can in lu-
ence co po a e R&D such as he supply o doc o al g adua es, uni e si y pa en ing ac i i-
ies. We es ima e andom coe icien (mixed) logi models (Alcace & Chung, 2007, 2014;
Head e al., 1995) allowing o in es o he e ogenei y o analyze he loca ion decisions o
148 esea ch and 325 de elopmen p ojec s ac oss 354 MSAs du ing 2003–2012. We adap
a model ha iden i ies agglome a ion economies s emming om labo , supplie and cus-
ome specializa ion in a egion (Alcace & Chung, 2014; Glaese & Ke , 2009), ea ing
academic esea ch as an inpu o R&D a he i m le el.
We ind ha an applied scien i ic esea ch ocus and he academic quali y o his applied
esea ch o he uni e si ies in an MSA exe a posi i e in luence on he likelihood ha
he MSA is chosen o R&D in es men s in gene al and o de elopmen in es men s in
pa icula . In con as , esea ch in es men s a e d awn o egions wi h a highe academic
quali y o basic scien i ic esea ch, whe eas an applied esea ch ocus and he academic
quali y o applied esea ch play no ole he e. R&D in es men s a e u he mo e a ac ed
by MSAs wi h uni e si ies specialized in he science domains ele an o he in es ing
i m and deli e ing doc o al g adua es wi h a ele an doc o al deg ee.
Ou esea ch con ibu es o he li e a u e on uni e si y echnology ans e , by p o id-
ing impo an nuance o he deba e on he ela i e impo ance o basic o applied academic
esea ch o i m inno a ion and he deba e abou he con lic ing logics be ween science and
echnology (Cassiman e al., 2008; G imaldi e al., 2011; Hausman, 2020; Pe kmann e al.,
1962
R.Belde bos e al.
1 3
2013; Zah inge e al., 2017). We sugges ha undamen al academic esea ch s eng hs
emain impo an o p o ide an a ac i e en i onmen o mo e p o ound indus ial R&D
ac i i ies ocusing on esea ch a he han de elopmen . Ou pape also con ibu es new
insigh s o he li e a u e on R&D in es men loca ion decisions (Alcace & Chung, 2014;
Alcace & Delgado, 2016; Belde bos e al., 2014, 2017) ela ed o he impo ance o uni-
e si ies and hei cha ac e is ics on R&D loca ion choice.
2 Theo e ical backg ound andhypo heses
We e iew he li e a u e on he ela ionship be ween uni e si y esea ch and co po a e
R&D, basic e sus applied uni e si y esea ch, he ole o academic esea ch quali y, a e
which we de elop ou co e hypo heses ha he ole o basic and applied esea ch a uni-
e si ies and hei quali y depend on whe he i ms seek a loca ion o R&D ocusing on
esea ch a he han de elopmen .
2.1 Uni e si y esea ch andco po a e R&D
A la ge body o e idence suppo s he impo an ole o uni e si y esea ch in s imula ing
co po a e R&D and i m inno a i e pe o mance (Adams, 1990; Belde bos e al., 2012;
Fleming & So enson, 2004; Gamba della, 1992; Sal e & Ma in, 2001; Toole, 2012). Sci-
en i ic esea ch may yield use ul applica ions (B ooks, 1994; Ja e, 1989) and may ans-
o m he sea ch and p oblem-sol ing p ocess unde lying echnological inno a ion (Flem-
ing & So enson, 2004). Besides he gene al p o ision o academic esea ch, uni e si ies
may a ec i ms’ inno a ion ac i i ies h ough many o he channels (Cohen e al., 2002;
D’Es e & Pa el, 2007; Link & Siegel, 2005; Sal e & Ma in, 2001; Thu sby & Thu sby,
2002). They educa e scien is s and enginee s, who may cons i u e he u u e wo k o ce
o i ms, hey p o ide expe s and consul an s o help i ms sol e pa icula echnological
p oblems, hey se e as collabo a ion pa ne s on emb yonic and applied p ojec s, and hey
engage in knowledge ans e h ough pa en ing and licensing ac i i ies (Belde bos e al.,
2016; Cassiman e al., 2008; Hall e al., 2003; Pe kmann e al., 2013).
One ea u e o hese mechanisms h ough which uni e si ies con ibu e o i m inno-
a ion is he ole played by dis ance. Se e al s udies ha e unde lined he geog aphically
bounded na u e o uni e si y- i m spillo e s and he consequen necessi y o i ms o be
loca ed close o uni e si ies in o de o ully cap u e he ele an bene i s (Aud e sch &
Feldman, 2004; Mans ield, 1995, 1998). To success ully capi alize on uni e si y esea ch,
i ms o en need access o aci knowledge no con ained in con ac s o published wo k;
knowledge ha is di icul o be ansmi ed ac oss long dis ances (Mowe y & Ziedonis,
2015; Von Hippel, 1994). Scien i ic knowledge ela ed o esea ch can be complex and di -
icul o codi y (Von K ogh e al., 2000), which complica es e ec i e ans e a dis ance.
Se e al s udies ha e con i med he bounded na u e o knowledge spillo e s om uni e si-
ies (Belenzon & Schanke man, 2013; Ja e e al., 1993) and ha e documen ed he geo-
g aphically cons ained mobili y choices o uni e si y g adua es (Be y & Glaese , 2005;
Miguélez & Mo eno, 2012).
1963
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
2.2 Uni e si ies’ esea ch o ien a ion: basic e susapplied scien i ic esea ch
Scien i ic esea ch is he e ogeneous, and di e en ypes o esea ch may a y in how alu-
able hey a e o echnological inno a ion in i ms, as well as h ough which mechanism
hei con ibu ion un olds. In pa icula , he e is an impo an dis inc ion be ween basic
and applied scien i ic esea ch. Acco ding o he F asca i Manual (OECD, 2002), “basic
esea ch is expe imen al o heo e ical wo k unde aken p ima ily o acqui e new knowl-
edge o he unde lying ounda ion o phenomena and obse able ac s, wi hou any pa icu-
la applica ion o use in iew.” Applied esea ch is also conside ed as o iginal in es iga ion
o acqui e new knowledge. I is, howe e , “di ec ed p ima ily owa ds a speci ic p ac ical
aim o objec i e.”
The e s ill is conside able deba e ega ding wha he ela i e me i s a e o basic and
applied academic esea ch o i m inno a ion. On he one hand, we may expec a highe
added alue om basic esea ch. Fleming and So enson (2004) a gue ha basic esea ch
can p o ide a map o echnological inno a ion; heo e ical unde s anding o he p oblem
and solu ion space can ans o m p oblem-sol ing om a ela i ely haphaza d sea ch p o-
cess o a mo e di ec ed iden i ica ion o use ul new combina ions, leading o be e solu-
ions (Cassiman e al., 2008; Fleming & So enson, 2004). Al hough basic esea ch is less
likely o yield di ec p ac ical applica ions, i may lead o b oade , mo e adical, and unex-
pec ed applica ions, o en h ough a long se ies o ollow-on esea ch and de elopmen
(Bush, 1945). Fo example, basic esea ch on he CRISPR/Cas9 gene ic scisso s may lead
o cues o cance and inhe i ed diseases, and basic quan um esea ch may yield applica-
ions beyond quan um compu ing ha may e olu ionize many indus ies. P io s udy has
ound ha basic biomedical pape s ha e a highe chance o being ci ed by pa en s (Ke,
2020). Fu he mo e, esea che s wi h a basic esea ch o ien a ion o educa ion a e mo e
likely o deli e adical and aluable echnologies (G ube e al., 2013). S udies ha e also
obse ed ha i ms bene i om collabo a ing wi h “s a scien is s”, i.e., eli e scien is s
in hei scien i ic discipline, mos ly o ien ed owa ds basic esea ch (Higgins e al., 2011;
Pe kmann e al., 2011; Zucke e al., 1998, 2002).
On he o he hand, we migh expec applied esea ch o be mo e di ec ly aluable o
i m inno a ion, inhe en o i s goal owa ds p ac ical use (Nelson e al., 2011; Nigh ingale,
1998; Rosenbe g & Nelson, 1994). Applied esea ch ollows an epis emological logic ha
closely esembles echnological de elopmen p ocesses cha ac e izing R&D in i ms (Gi -
elman & Kogu , 2003; Nelson, 2003). One impo an poin o conce n abou uni e si y
engagemen wi h indus y is ha i may d i e uni e si ies away om cu iosi y-d i en basic
esea ch owa d applied esea ch di ec ly ele an o indus y (Fini e al., 2021; Hende son
e al., 1998; Mowe y & Ziedonis, 2002; Pe kmann e al., 2021; Thu sby & Thu sby, 2011).
While his is s ill deba ed, i e lec s a gene al belie e ha applied esea ch is mo e ele an
o he indus y. P io s udies sugges ha pa en s building on applied scien i ic esea ch
ha e a highe echnological and economic alue han pa en s building on basic esea ch
(Wang & Ve be ne, 2021). Fu he mo e, esea che s wi h an applied esea ch o ien a ion
o educa ion a e be e sui ed o helping i ms o c ea e alue om R&D (Ali & Gi elman,
2016; Baba e al., 2009; Ro hae mel & Hess, 2007; Sub amanian e al., 2013). Fi ms mo e
eadily collabo a e wi h uni e si ies on esea ch ha is applied in na u e (Godin & Gin-
g as, 2000; Hicks & Hamil on, 1999) and s udies ha e ound ha i is mo e ad an ageous
o i ms o wo k wi h b idging scien is s, in pa icula “Pas eu scien is ”, i.e., scien is s
wi h an o ien a ion owa ds applied esea ch (Baba e al., 2009; Ro hae mel & Hess, 2007;
1964
R.Belde bos e al.
1 3
Sub amanian e al., 2013). Bika d and Ma x (2020) ound ha geog aphic hubs acili a e
knowledge low om uni e si ies o indus y by acili a ing applied esea ch.
2.3 Uni e si ies’ academic quali y o scien i ic esea ch
The e is also an open ques ion ega ding whe he he quali y s anda ds o science and
echnology a e well aligned. Gi en hei di e en logics, science and echnology may also
di e in hei quali y s anda ds, so i is no s aigh o wa d ha esea ch wi h high aca-
demic quali y, i.e., being pe cei ed o be o high quali y by o he scien is s as e lec ed in
o wa d scien i ic ci a ions, will also ep esen high alue and ele ance o echnological
inno a ion. Empi ical e idence is mixed and inconclusi e. On he one hand, s udies ha e
shown ha highly ci ed scien i ic publica ions a e much mo e likely o be ci ed by pa en s
(Ahmadpoo & Jones, 2017; Hicks e al., 2000; Popp, 2017; Veugele s & Wang, 2019)
and ha e e ences in pa en s o highly ci ed publica ions ha e highe alue (Poege e al.,
2019). Highly ci ed academic publica ions au ho ed by in-house esea che s o i ms ha e
also been posi i ely associa ed wi h inno a ion ou comes (Sub amanian e al., 2013), and
he quali y o esea ch depa men s has been ound o s imula e colloca ed R&D (Ab a-
mo sky e al., 2007). On he o he hand, Gi elman and Kogu (2003) ound a nega i e
associa ion be ween impo an scien i ic pape s (i.e., scien i ic pape s ha a e highly ci ed
by o he scien i ic pape s) and high-impac echnological in en ions (i.e., pa en s ha a e
highly ci ed by o he pa en s), and Wang and Ve be ne (2021) ound an insigni ican asso-
cia ion be ween ci a ions o scien i ic esea ch and pa en alue.
Fu he mo e, Scandu a and Iamma ino (2022) in es iga ed UK uni e si ies and ound a
nega i e associa ion be ween academic esea ch quali y and he le el o engagemen wi h
indus y o depa men s in he basic sciences, bu a posi i e associa ion o hose in he
applied sciences. This sugges s he impo ance o di e en ia ing be ween he quali y o
basic and applied esea ch when examining mul ina ionals R&D loca ion decisions.
2.4 Hypo heses: esea ch e susde elopmen in es men s anduni e si y esea ch
The in luence o basic o applied academic esea ch and hei espec i e academic quali y
on indus ial R&D is likely o depend on whe he i ms engage in de elopmen o esea ch
ac i i ies. This ollows om he no ion ha R&D ac i i ies ca ied ou by i ms a e also
he e ogeneous in asks and objec i es (Belde bos e al., 2009; Sachwald, 2008; Suzuki
e al., 2017), wi h a salien dis inc ion be ween esea ch ac i i ies on he one hand and
de elopmen ac i i ies on he o he (Ba ge-Gil & López, 2014; Cza ni zki e al., 2010; on
Zed wi z & Gassmann, 2002).
The he e ogenei y among R&D ac i i ies is pa icula ly salien in he con ex o R&D
in e na ionaliza ion. R&D ac i i ies in o eign a ilia es o mul ina ional i ms can be ai-
lo ed o adap p oduc and echnologies o local consume p e e ences and suppo ing man-
u ac u ing ac i i ies in o eign coun ies, ocusing on de elopmen (Kuemme le, 1997).
They can also be mo i a ed by he sou cing o eign echnologies augmen ing he knowl-
edge base a home (Almeida, 1996; Can well & Mudambi, 2005; Flo ida, 1997; Zan ei,
2000) and ocus on esea ch. Shimizu ani and Todo (2008) and Belde bos e al. (2009)
ound ini ial e idence ha he dis inc ion be ween esea ch and de elopmen in es men s
ma e s o loca ion decisions. They obse ed ha ma ke size was mos closely associa ed
wi h de elopmen loca ions, while local esea ch in ensi y was mo e closely associa ed
wi h esea ch in es men s.
1965
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
Ou conjec u e is ha in es men s in esea ch ac i i ies a e mo e likely o seek bene i s
ela ed o high quali y basic scien i ic esea ch a uni e si ies, while i ms’ in es men s in
de elopmen ac i i ies a e mo e likely o seek bene i s om high quali y applied scien-
i ic esea ch. Fi ms conduc ing esea ch can seek ad an ages in capi alizing on he basic
esea ch pe o med in academia, expanding he knowledge base on which hey can d aw
o hei own inno a ion ac i i ies (Cockbu n & Hende son, 1998; Cohen & Le in hal,
1989, 1990; Kle o ick e al., 1995). Fi ms ha a e awa e o new ad ances in ele an
scien i ic ields a e in a be e posi ion o iden i y p omising esea ch pa hs ha can hen
ansla e in o new in en ions. This may gi e a i s mo e ad an age o he in oduc ion o
new p oduc s and p ocesses (A o a e al., 2021; Fab izio, 2009; Rosenbe g, 1990). Knowl-
edge o basic esea ch is also impo an because i p o ides i ms wi h a be e unde s and-
ing o he o e all echnological landscape, which can help i ms o mo e e ec i ely sea ch
o new in en ions and a oid was e ul expe imen a ions (Fleming & So enson, 2004). In
con as , i ms’ de elopmen ac i i ies a e mos likely o bene i om applied scien i ic
esea ch, gi en he close connec ion o applied scien i ic esea ch o de elopmen and
comme cializa ion and he absence o such a di ec connec ion o basic scien i ic esea ch
(Balconi e al., 2010). Hence, we o mula e he ollowing hypo heses:
Hypo hesis 1 Fi m in es men s in esea ch a e a ac ed o loca ions wi h uni e si ies’
basic esea ch, while i m in es men s in de eloped a e a ac ed o loca ion wi h uni e si-
ies’ applied esea ch.
Hypo hesis 2 Fi m in es men s in esea ch a e a ac ed o loca ions wi h a high quali y o
uni e si ies’ basic esea ch, while i m in es men s in de eloped a e a ac ed o loca ion
wi h a high quali y o uni e si ies’ applied esea ch.
3 Da a, a iables andempi ical model
3.1 Da a
We cons uc a da ase on he cha ac e is ics o uni e si y esea ch a he MSA le el and
ma ch i wi h he loca ion decisions o o eign mul ina ional i ms’ R&D in es men s in
he Uni ed S a es (2003–2012) ob ained om he Di Ma ke s da abase o he Financial
Times L d. The Di Ma ke s da abase is conside ed o be one o he mos comp ehensi e
sou ces o in o ma ion on c oss-bo de g een ield in es men s, co e ing in es men s made
by mul ina ional i ms ope a ing ac oss indus ies and coun ies. I is based on mo e han
8000 news and p op ie a y sou ces and eco ded mo e han 120,000 wo ldwide c oss-bo -
de g een ield in es men s du ing he pe iod. In es men s a e classi ied in o indus ies ha
can be mapped in o a co esponding 3-digi NAICS sec o . In es men s a e also ca ego-
ized in o di e en alue chain ac i i ies: manu ac u ing, dis ibu ion, logis ics, R&D, e c.
The da abase has equen ly been used in p io esea ch (Cas ellani e al., 2013; C escenzi
e al., 2014; D’Agos ino e al., 2013), and i s alidi y and eliabili y ha e been con i med
independen ly by di e en esea che s (Cas ellani e al., 2013; C escenzi e al., 2014). We
es ic ou analysis o R&D in es men s made in he Uni ed S a es by i ms ope a ing in
manu ac u ing indus ies. The main eason o his ocus is ha he use o pa en s is el-
a i ely a e in he se ice sec o , such ha conco dances be ween se ice sec o s, ech-
nologies, and science ields canno be es ablished well. The da ase con ains 473 o eign
1966
R.Belde bos e al.
1 3
R&D in es men s unde aken by 328 i ms based in a a ie y o coun ies. Among he 473
R&D in es men s, 148 could be classi ied as esea ch in es men s based on and he ex
desc ip ion accompanying each R&D in es men in da abase. In es men s a e classi ied as
esea ch i he desc ip ion o he p ojec e e s o (basic o undamen al) esea ch, while
desc ip ions o de elopmen in es men s e e o adap a ion, solu ions, and de elopmen .
Fi ms based in Ge many a e esponsible o he la ges sha e o R&D in es men s (17.5%),
ollowed by i ms based in Japan (17.3%), he U.K (9.5%), F ance (5.5%) and Sou h Ko ea
(5.3%). Mos R&D in es men s ake place in he pha maceu ical and chemical indus y
(27.1%), ollowed by he compu e s and elec onics indus y (24.1%) and he anspo
equipmen indus y (16.5%).
We a e in e es ed in he ole played by academic esea ch and i s cha ac e is ics in
a ac ing R&D in es men s. Wi h indus y science linkages and in luences o uni e si y
esea ch on co po a e R&D mos salien in geog aphic p oximi y, we need o de ine an
app op ia e geog aphical uni o analysis o ou s udy. Such a sui able geog aphic uni
is he Me opoli an S a is ical A ea (MSA) de ined by he Uni ed S a es O ice o Man-
agemen and Budge (OMB) and used by se e al ede al go e nmen agencies o s a is i-
cal pu poses (Nussle, 2008). Each MSA con ains a co e u ban a ea wi h a leas 50,000
inhabi an s. I consis s o one cen al coun y plus adjacen coun ies wi h a high deg ee o
economic in eg a ion wi h he cen al coun y, as measu ed h ough wo ke commu ing ies.
A e each decennial census ealized by he Census Bu eau, he OMB e ises he lis o
cu en MSAs o e lec changes in he demog aphic composi ion o such a eas. Gi en ha
in es men s con ained in ou da abase we e pe o med be ween 2003 and 2012, we use he
lis o MSAs eleased by he OMB in 2003 ollowing he 2000 decennial census.
We iden i y he ela ionships be ween uni e si y esea ch and R&D in es men s om
a ia ion in he olume, ype and quali y o publica ions au ho ed by uni e si y a ilia ed
esea che s ac oss MSAs o e ime. We posi ha an indi idual i m deciding on a loca ion
o a speci ic R&D p ojec in a gi en yea ega ds he exis ing s a e o uni e si y esea ch
in MSA egions as gi en. Whe eas (la ge) domes ic incumben i ms may ha e had an
in luence on uni e si y esea ch h ough i m-uni e si y R&D collabo a ions and o he
in e ac ions (Hausman, 2020), his ea u e will ypically be no be p esen o o eign i ms
es ablishing an R&D uni in a egion.
3.2 Uni e si y esea ch
We use publica ions o cons uc indica o s o uni e si y esea ch. We assign each aca-
demic publica ion e ie ed om Cla i a e’s Web o Science (WoS) published by a leas
one au ho esiden in he Uni ed S a es o an MSA. Fo each publica ion wi hin WoS, he
add esses o he au ho s a e epo ed, which may include he s a e, he ci y and he i s i e
digi s o he zip code. We ma ched he zip codes o he co esponding MSA using he con-
co dance able p o ided by he Uni ed S a es Census Bu eau. Fo hose WoS add esses ha
do no include zip codes, we ma ched on ci y and s a e names. The sha e o publica ions
by esiden au ho s in he Uni ed S a es ha ha e a leas one au ho add ess in an MSA
is 97.1%. We ake a ac ional coun o publica ions ac oss au ho s in case he e a e co-
au ho s based in loca ions o he han he ocal MSA o in mul iple MSAs. We subsequen ly
dis inguish academic publica ions om publica ions by i ms and esea ch ins i u es using
keywo d lis s (college, uni e si y) and manual alida ion.
To ake in o accoun he ele ance o scien i ic esea ch o in es ing i ms in di e en
manu ac u ing indus ies, we use conco dance ables o ma ch publica ions o echnology
1973
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
Educa ional A ainmen , which is he sha e o he MSA popula ion wi h a mas e ’s deg ee.
Da a we e e ie ed om he Uni ed S a es Census Bu eau. As da a be o e 2005 we e no
a ailable, we impu ed missing alues o he yea s 2003 and 2004 based on he se ies o
uni e si y publica ions.2 We con ol o he le el o co po a e axes by including he a i-
able Tax, which measu es he s a e le el co po a e ax a e ( om ax ounda ion.o g). When
an MSA spans mul iple s a es, he a e age o he ele an s a es’ co po a e ax le els is
used. We also con ol o he le el o R&D ax c edi s, employing da a a he s a e le el
om Wilson (2009) and Fala o and Sim (2014). Popula ion Densi y is e ie ed om he
Uni ed S a es Census Bu eau, as popula ion pe squa e mile, scaled by 1000 o ease o
in e p e a ion. The e ec o Popula ion Densi y migh be nonlinea , as a densely popula ed
loca ion may allow o g ea e knowledge spillo e s bu a high le el o densi y may also
lead o conges ion. The e o e, we include bo h linea and squa ed e ms. To con ol o
in a- i m co-loca ion e ec s (Alcace & Delgado, 2016; Cas ellani and La o a o i, 2020),
we cons uc he a iable P e ious In es men as a dummy a iable aking he alue 1 i he
MSA al eady hos s an exis ing subsidia y o he i m, and 0 o he wise. In cons uc ing his
a iable, we ely on he ORBIS da abase on i ms’ global a ilia es, as well as he in o ma-
ion on p e ious in es men by he ocal i m con ained in he Di Ma ke s da abase.
All he explana o y a iables a e one yea lagged wi h espec o he yea o he o -
eign R&D in es men decision o allow a esponse ime by he in es ing i m. All a iables
excep o bina y a iables and Popula ion Densi y a e aken in na u al loga i hms, which
allows in e p e a ion o he coe icien s in condi ional and mixed logi models in e ms o
a e age elas ici ies (Head e al., 1995).3 The co ela ions and desc ip i e s a is ics o he
a iables a e p esen ed in Table3. The co ela ions do no indica e mul icollinea i y issues.
3.5 Empi ical model
In o de o model he loca ion choices, whe e each i m chooses one MSA among he se o
354 MSAs, we employ andom coe icien condi ional logi models. The condi ional logi
model is widely used in loca ion choice s udies (e.g., Head e al., 1995). Based on a u ili y
maximiza ion amewo k, McFadden (1974) p oposed modeling expec ed u ili y in e ms
o choices’ a ibu es a he han cha ac e is ics o agen s making he decision. Fi m cha -
ac e is ics ha do no a y by loca ion, such ha i m o indus y e ec s canno be included
in condi ional logi models, as hei alue would be iden ical ac oss choice such ha hey
would d op ou o he equa ion. This ea u e o he condi ional logi model has led i o be
ega ded as inhe en ly con olling o ime-in a ian i m ai s such as indus y (Alcace &
Chung, 2014; Li e al., 2023). Suppose in es ing i m makes a loca ion decision Y among
L al e na i es. Le U l be he expec ed u ili y o he l h choice o he i m. U l is an inde-
penden andom a iable wi h a sys ema ic componen x
′
l𝛽
, whe e x l ep esen s a ec o o
cha ac e is ics o he l h choice. Then he expec ed u ili y o i m’s R&D loca ion choice
is modeled in e ms o he obse able a ibu es o he choice (i.e., loca ion; MSA) and an
unobse able e o e m:
2 Unlike da a on doc o a es, he Na ional Science Founda ion does no epo he numbe o mas e ’s
deg ee ecipien s by academic ields.
3 The a e age elas ici y o he p obabili y o loca ion choice wi h espec o a loga i hmic ans o med a i-
able can be calcula ed as (L−1)/L imes he coe icien o he a iable, whe e L equals he o al numbe o
loca ion choices.
1974
R.Belde bos e al.
1 3
McFadden (1974) showed ha i he L al e na i es a e independen and iden ically dis-
ibu ed wi h Type I ex eme- alue dis ibu ion, he p obabili y ha i m chooses o in es
in MSA l is gi en by he ollowing o mula:
The condi ional logi model elies on he independence om i ele an al e na i es
(IIA): he odds a io be ween wo al e na i es is independen o changes in any o he al e -
na i es. This is an assump ion ha may no hold. A andom coe icien mixed logi model
gene alizes he condi ional logi , elaxes he IIA assump ion, and allows o gene al unob-
se ed he e ogenei y in in es o p e e ences (McFadden & T ain, 2000). Because we ha e
no p io i expec a ions abou whe he ce ain coe icien s ha e a andom componen o no ,
we allow all coe icien s o be andom (Basile e al., 2008; Chung & Alcace , 2002; Re el
& T ain, 1998; T ain, 2009).
The mixed logi p obabili y is a weigh ed a e age o he condi ional logi o mula e alu-
a ed, wi h he weigh s p o ided by he densi y unc ion o he andom pa o he pa am-
e e s: g(λ ). The loca ional choice p obabili y has o be calcula ed o e all possible alues
o λ . The mixed logi p obabili y is he e o e ob ained by aking he in eg al o he mul-
iplica ion o he condi ional p obabili y wi h he densi y unc ions desc ibing he andom
na u e o he coe icien s. We ollow he mos gene al app oach by allowing a no mal dis-
ibu ion unc ion (Basile e al., 2008; Belde bos e al., 2014; Chung & Alcace , 2002);
es ima es a e based on 100 simula ion d aws (Re el & T ain, 1998; T ain, 2009), and we
clus e e o e ms by i m. Since one o ou esea ch ques ions is i he ole o local uni-
e si ies’ esea ch in i ms’ loca ion decisions is di e en o esea ch o o de elopmen
in es men s, we es ima e sepa a e models o he wo ypes o R&D in es men s (Hoe ke ,
2007).4
We no e ha he mixed logi model is ac ually a mo e gene al speci ica ion han an
al e na i e b anch o andom coe icien models, he La en Class Random Pa ame e
model (Paci ico & Yoo, 2013; Rasciu e & Downwa d, 2017). In he LCRP models he
andom na u e o he in luence is modelled a he class le el only and he esea che has
o p ede e mine which cha ac e is ics would de e mine class membe ship. In he mixed
logi model, on he o he hand, andom in luences a e modelled and andom pa ame e s
a e es ima ed o each indi idual i m and he e is no equi emen o se p ede e mined
i m cha ac e is ics ha could cause p e e ence he e ogenei y. In addi ion, LRCP models
equi e ha a iables en e ing he class membe ship model a e cons an ac oss al e na-
i es o he same agen (Paci ico & Yoo, 2013, p. 628), which does no hold o ou ocal
in luence: whe he a i m in es s in esea ch o de elopmen can di e o he same i m
U
l
=
x
�
l
𝛽
+
𝜀
l
P
�Y =l�=
exp
�
x�
l𝛽+𝜀 l
�
∑
L
k=1exp
�
x�
k𝛽+𝜀 k
�
P
�Y =l�=∫exp �x�
l𝛽+x�
l𝜆 �
∑
L
k=1exp
�
x�
k𝛽+x k�𝜆
�
g�𝜆 �d�𝜆
�
4 We no e ha a ying esidual a ia ions, compounded by he andom componen speci ica ion, p e en
he di ec compa ison o coe icien s o di e en mixed logi models (Allison, 2009; Hoe ke , 2007).
1975
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
Table 3 Desc ip i e s a is ics and co ela ions
Mean Sd 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Loca ion
Choice
(DV) (1)
0.0028 0.0533
GDP Pe
Capi a (2)
36,279.63 10,484.78 0.07
Wage Cos s
(3)
69,809.93 9745.94 0.04 0.20
Educa ional
A ain-
men (4)
9.29% 3.90% 0.06 0.48 0.18
Popula ion
Densi y
(5)
0.29 0.47 0.11 0.33 0.13 0.26
Popula ion
densi y
Squa ed
(6)
0.31 2.69 0.07 0.18 0.09 0.13 0.86
Tax (7) 6.57 2.66 0.00 0.02 −0.13 0.10 0.02 0.03
R&D Tax
C edi (8)
4.90 4.57 0.01 0.04 0.10 −0.10 0.11 0.09 0.30
P e ious
In es -
men (9)
0.01 0.10 0.16 0.11 0.04 0.08 0.15 0.09 0.01 0.03
Indus y
Es ablish-
men s
(10)
27.41 135.55 0.03 0.04 −0.03 0.07 0.05 0.01 0.01 −0.02 0.03
Technology
Fi (11)
0.99 1.05 0.10 0.41 0.09 0.31 0.43 0.18 0.02 0.05 0.15 0.18
Doc o a es
Fi (12)
0.14 0.30 0.04 0.21 0.07 0.38 0.19 0.08 0.01 −0.02 0.06 0.07 0.25
1976
R.Belde bos e al.
1 3
Table 3 (con inued)
Mean Sd 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Academic
Resea ch
Fi (13)
1.04 1.29 0.01 0.02 0.05 0.07 0.01 0.00 −0.02 −0.02 0.00 0.10 0.00 0.13
Ra io non-
Uni e si y
Pub. (14)
37.78% 32.60% 0.00 −0.03 0.01 −0.23 −0.04 0.00 −0.05 −0.05 −0.02 −0.02 −0.13 −0.32 0.08
Uni e si y
Pa en s Fi
(15)
0.42 3.58 0.01 0.03 −0.01 0.05 0.03 0.01 0.02 0.00 0.01 0.05 0.03 0.08 0.02 −0.02
Applied
Sha e (16)
53.91% 31.62% 0.00 −0.04 −0.06 −0.16 −0.03 −0.03 −0.02 −0.02 0.00 0.00 −0.08 −0.12 0.26 0.34 0.01
Academic
Quali y:
Applied
(17)
0.61 0.63 0.04 0.26 0.20 0.43 0.22 0.11 0.02 0.05 0.06 0.02 0.27 0.29 0.07 −0.25 0.02 −0.23
Academic
Quali y:
Basic (18)
0.64 0.50 0.04 0.27 0.18 0.40 0.21 0.10 0.01 0.04 0.06 0.03 0.27 0.31 0.10 −0.24 0.03 −0.12 0.47
Means and s anda d de ia ions epo ed be o e loga i hmic ans o ma ion
1977
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
pe in es men p ojec . We conclude ha he LCRP model is less sui able o ou esea ch
endea o s.
4 Resul s
4.1 Uni e si y esea ch andR&D loca ion choice
Resul s o i e mixed logi models a e epo ed in Table4. In he i s column, we epo
he model es ima ed on he ull sample o 473 in es men s wi h only he con ol a iables.
In he second column, we add a se o uni e si y cha ac e is ics excep o he ocal a i-
ables, i.e., Applied Sha e, Academic Quali y: Basic and Academic Quali y: Applied. In col-
umn h ee, we add hese ocal a iables. Finally, in column ou and i e we epo models
sepa a ing esea ch and de elopmen in es men s.
In model 1, all he con ol a iables ha e he expec ed sign excep o Wage Cos s,
which displays a posi i e e ec (β = 2.172, p = 0.004). As poin ed ou by C escenzi e al.
(2014), wages may also p oxy o he a ailabili y o skilled wo ke s, and hus highe wages
may be posi i ely associa ed wi h loca ion choice o high alue-added unc ions such as
R&D. Coe icien s o GDP Pe Capi a, P e ious In es men and R&D Tax C edi s a e
all posi i e, while Tax does no seem o ha e an e ec (β = −0.121, p = 0.141). The posi-
i e coe icien o Popula ion Densi y and he nega i e coe icien o Popula ion Densi y
Squa ed sugges ha i ms a e a ac ed o dense loca ions up o a ce ain poin , a e which
conges ion e ec s may ende highe popula ion densi y less a ac i e. The u ning poin
o his ela ion is a he 95 h pe cen ile o he dis ibu ion o popula ion densi y (abou
3500 inhabi an s pe squa e mile). The a iable Indus y Es ablishmen s, as a gene al indi-
ca o o he le el o agglome a ion in he MSA, displays a posi i e and sizable coe icien
(β = 0.721, p < 0.001), as expec ed. Simila ly, he p esence o agglome a ion economies
s emming om he p esence o ele an echnological knowledge and R&D agglome a ion
(Technology Fi ) also exhibi s a s ong posi i e e ec (β = 0.993, p < 0.001).
In model 2 uni e si y ela ed agglome a ion economies s emming om labo supply
(Doc o a es Fi , β = 0.965, p = 0.016) and he supply o academic knowledge (Academic
Resea ch Fi , β = 0.891, p < 0.001) bo h display posi i e e ec s. Pa en ing ac i i ies by
uni e si ies (Uni e si y Pa en Fi , β = 0.0786, p = 0.356) and academic esea ch by o he
non-uni e si y ac o s (Ra io Non-Uni e si y Publica ions, β = 0.395, p = 0.328) ha e no
addi ional signi ican in luence. Resul s o model 3 show ha a ocus on applied academic
esea ch a ac s R&D in es men s, as indica ed by he sizable and signi ican coe icien o
Applied Sha e (β = 2.089, p = 0.004). Academic Quali y: Applied also posi i ely a ec s he
loca ion o R&D in es men s (β = 1.265, p < 0.001), bu Academic Quali y: Basic does no
ha e an e ec (β = 0.917, p = 0.075).
4.2 Resea ch e susde elopmen in es men s
We now examine di e ences in he loca ional d i e s be ween esea ch and de elopmen
in es men s. The subsample models ocusing on ei he esea ch in es men s o de elop-
men in es men s a e p esen ed in Table4, columns 4 and 5. Resul s sugges a majo he -
e ogenei y in he ole o uni e si y esea ch depending on he ype o R&D in es men .
Fo esea ch in es men s Applied Sha e (β = 0.959, p = 0.322) is insigni ican , while o
1978
R.Belde bos e al.
1 3
Table 4 Mixed logi es ima es o he de e minan s o R&D in es men loca ion choices among MSAs
Es ima ion esul s o mixed logi models. Robus s anda d e o s clus e ed by pa en i m, and co espond-
ing p alues in pa en heses
Con ol Basic Full Resea ch De elopmen
GDP Pe Capi a 1.105 1.193 0.948 1.705 0.656
(0.000) (0.000) (0.005) (0.001) (0.055)
Wage Cos s 2.172 1.975 1.827 1.520 1.784
(0.004) (0.008) (0.017) (0.128) (0.037)
Educa ional A ainmen 1.269 1.174 1.133 1.318 1.066
(0.000) (0.000) (0.000) (0.001) (0.001)
Popula ion Densi y 0.530 0.542 0.534 0.921 0.451
(0.029) (0.031) (0.059) (0.077) (0.160)
Popula ion Densi y Squa ed −0.0706 −0.0718 −0.0804 −0.240 −0.0798
(0.030) (0.033) (0.196) (0.066) (0.098)
Tax −0.121 −0.117 −0.0892 −0.213 −0.00510
(0.141) (0.150) (0.299) (0.284) (0.958)
R&D Tax C edi 0.124 0.123 0.115 0.199 0.0511
(0.037) (0.038) (0.054) (0.059) (0.481)
P e ious In es men 2.147 2.129 2.103 2.252 2.024
(0.000) (0.000) (0.000) (0.000) (0.000)
Indus y Es ablishmen s 0.721 0.708 0.658 0.487 0.796
(0.000) (0.000) (0.000) (0.000) (0.000)
Technology Fi 0.993 0.951 0.971 1.223 0.899
(0.000) (0.000) (0.000) (0.000) (0.000)
Doc o a es Fi 0.965 0.911 1.136 0.898
(0.016) (0.032) (0.122) (0.065)
Academic Resea ch Fi 0.891 0.862 1.010 0.825
(0.000) (0.001) (0.029) (0.011)
Ra io Non-Uni e si y Pubs 0.395 0.531 0.303 0.701
(0.328) (0.263) (0.674) (0.165)
Uni e si y Pa en s Fi 0.0786 0.0729 −0.767 0.164
(0.356) (0.395) (0.269) (0.063)
Applied Sha e 2.089 0.959 2.561
(0.004) (0.322) (0.004)
Academic Quali y: Applied 1.265 0.677 1.518
(0.000) (0.241) (0.000)
Academic Quali y: Basic 0.917 2.227 0.266
(0.075) (0.001) (0.687)
# In es men s 473 473 473 148 325
# Al e na i e Choices 354 354 354 354 354
Wald Chi21086.61 1039.28 1175.32 364.09 739.25
(0.000) (0.000) (0.000) (0.000) (0.000)
Signi ican andom componen s
P e ious In es men 1.079 1.103 1.195
(0.000) (0.000) (0.000)
1979
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
de elopmen in es men s Applied Sha e has a la ge and s ongly signi ican coe icien
(β = 2.561, p = 0.004), in suppo o Hypo hesis 1. Fo esea ch in es men s, Academic
Quali y: Basic has a subs an ial posi i e associa ion wi h loca ion decisions (β = 2.227,
p = 0.001) bu his is no obse ed o Academic Quali y: Applied (β = 0.677, p = 0.241).
Simila ly, o de elopmen in es men s, Academic Quali y: Applied (β = 1.518, p < 0.001)
has a posi i e and signi ican coe icien , while Academic Quali y: Basic has no signi ican
in luence (β = 0.266, p = 0.687). These indings suppo Hypo hesis 2. We also obse e ha
Academic Resea ch Fi has oughly simila coe icien s ac oss he wo models. The a i-
able Doc o a es Fi loses signi icance in e ms o p- alue in bo h models, which is pe haps
due o he educed numbe o obse a ions.
The e ec sizes o he ea u es o uni e si y esea ch a e economically ele an : he
a e age elas ici y o he p obabili y o ecei ing R&D in es men s wi h espec o Aca-
demic Quali y: Basic (in he esea ch model) and Academic Quali y: Applied (in he de el-
opmen model) a e 2.2 and 1.5, espec i ely. The implied a e age elas ici y o Applied
Sha e in he de elopmen model is 2.5, and he es ima ed a e age elas ici y wi h espec o
Academic Resea ch Fi a e 1.0 and 0.8 o esea ch and de elopmen in es men s, espec-
i ely. These a e in he same o de o magni ude o exceed he elas ici ies o Technology Fi
o Educa ional A ainmen .
4.3 Science‑based e suso he indus ies
The ole o uni e si ies in a ac ing R&D in es men s may be con ingen on he indus y
o he in es ing i ms. Pa i (1984) classi ied indus ies in o ou ca ego ies: supplie dom-
ina ed, p oduc ion in ensi e, and science based, whe e p oduc ion in ensi e indus ies a e
u he sepa a ed in o scale in ensi e indus ies and specialized supplie indus ies. Pa i ’s
axonomy has been widely used and p o en aluable o inno a ion esea ch (A chibugi,
2001; Bogliacino & Pian a, 2016). In he con ex o ou esea ch ques ions on he ole o
uni e si y esea ch o i m inno a ion and R&D loca ion decisions, he mos impo an
dis inc ion is be ween science-based indus ies and he o he ype o indus ies. We exam-
ine whe he he ole o academic esea ch is mo e p onounced in science-based indus ies,
which include he chemicals and pha maceu icals indus y and he compu e s and elec-
onics indus y. Resul s o models dis inguishing be ween hese wo g oups o indus ies
a e epo ed in Table5. Resul s indica e ha , in line wi h expec a ions, o science-based
indus ies, he Academic Resea ch Fi and he quali y o uni e si y esea ch (bo h Aca-
demic Quali y: Basic and Academic Quali y: Applied) a e c ucial, while o he o he
indus ies a ocus on applied esea ch (i.e., Applied Sha e) is mo e impo an .
4.4 Supplemen a y analysis
We conduc ed a numbe o supplemen a y analyses o examine he obus ness o ou
empi ical esul s, esul s o which a e elega ed o he elec onic supplemen a y ma e ial.
We es ic ed es ima ion o o eign i ms ha es ablished hei i s R&D in es men in
he MSA, such ha p io R&D ac i i ies could no po en ially ha e in luenced uni e si y
esea ch cha ac e is ics. Gene ally, no p onounced di e ences wi h he indings epo ed
in Table4 we e ound. We also examined whe he he size o he in es men in luences
empi ical esul s, by es ima ing he mixed logi model wi h obse a ions weigh ed by an
indica o o in es men size. Fo size we use an es ima e o he dolla alue o he p ojec
1980
R.Belde bos e al.
1 3
p o ided by he Di ma ke s da abase. This deli e ed e y simila esul s. Simila esul s
we e also ob ained when es ima ing models wi h s a e ixed e ec s included.
Finally, we examined he obus ness o esul s o he po en ial p esence o spa ial au o-
co ela ion. We may expec his o be a lesse conce n in he con ex o ou esea ch o
wo easons. Fi s , he mixed logi models allow o andom a ia ions in p e e ences, un e-
s ic ed subs i u ion pa e ns ac oss loca ions, and co ela ions in u ili y (p e e ences) due
o co ela ion be ween unobse ed ac o s (McFadden & T ain, 2000, p. 649). Hence, he
es ima es a e obus o po en ial co ela ions in he e o e ms ac oss loca ional choices
due o hese ea u es. Second, only a mino i y o MSAs a e loca ed adjacen o each o he ,
which mi iga es spa ial co ela ion. We examined he sensi i i y o he esul s o he po en-
ial p esence o spa ial au oco ela ion by examining models omi ing MSAs whe e such
co ela ion is mos likely o occu : geog aphically adjacen o p oxima e MSAs. Omi ing
29 MSAs wi h a neighbo ing MSA wi hin 150 miles, esul s appea ed obus . In addi ion,
when we added spa ial lags o a numbe o a iables, hese lags we e insigni ican while
he es ima es o he ocal a iable emained obus .
5 Conclusion
This pape examined he ole o he e ogeneous academic esea ch in a ac ing indus ial
R&D in es men s, dis inguishing be ween esea ch in es men s and de elopmen in es -
men s. Ou indings, which ocused on in es men s by o eign mul ina ional i ms in me -
opoli an a eas in he U.S., con i med ha uni e si ies play an impo an ole in a ac -
ing R&D in es men s. The specializa ion o academic esea ch in domains ele an o he
ocal R&D in es men and he supply o doc o al s uden s wi h ele an specializa ion bo h
ha e a posi i e associa ion wi h i ms’ R&D loca ion decisions. We ound suppo o ou
hypo heses ha he ole o uni e si y cha ac e is ics di e s depending on whe he i ms
in es in esea ch o de elopmen . While an applied esea ch o ien a ion is gene ally asso-
cia ed wi h a g ea e a ac i eness o he MSA o R&D in es men s, such a ac i eness
is no p esen in he case o esea ch in es men s. Resea ch ac i i ies a e a ac ed by he
academic quali y o basic esea ch, whils de elopmen in es men s a e a ac ed by he
o ien a ion owa ds applied esea ch and he academic quali y o applied esea ch. We con-
clude ha , in o de o unde s and he ole o uni e si y esea ch in R&D loca ion choices
o i ms, i is c ucial o ake in o accoun bo h he he e ogenei y in academic esea ch and
he he e ogenei y in i ms’ R&D in es men s.
Ou esea ch con ibu es o he li e a u e on echnology ans e li e a u e and indus-
y-science linkages, in pa icula he li e a u e on he e ec s o academic esea ch on
co po a e inno a ion (Cassiman e al., 2008; G imaldi e al., 2011; Hausman, 2020;
Pe kmann e al., 2013; Zah inge e al., 2017), by p o iding no el insigh s o he ole o
basic e sus applied uni e si y esea ch. Ou inding ha he he e ogenei y o academic
esea ch wi h espec o academic quali y, specializa ion, and basic s. applied o ien a-
ion a ac s di e en ypes o R&D in es men s sugges s a mo e nuanced pe spec i e
on he ole o uni e si ies as a posi i e o ce in i ms’ R&D in es men s, adding o p e-
ious e idence sugges ing a p edominan impo ance o applied uni e si y esea ch (Ali
& Gi elman, 2016; Baba e al., 2009; Ro hae mel & Hess, 2007; Sub amanian e al.,
2013). Ou inding on he posi i e ole o he academic quali y o uni e si y esea ch—
including basic scien i ic esea ch— o co po a e R&D sugges ha he logics in
1981
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
Table 5 Mixed logi es ima es o he de e minan s o R&D in es men loca ion choices among MSAs: science based indus ies e sus o he indus ies
Full Resea ch De elopmen
Science based
indus y
O he indus ies Science based
indus y
O he indus ies Science based
indus y
O he indus ies
GDP Pe Capi a 1.223 0.826 1.218 2.547 1.228 0.382
(0.003) (0.051) (0.067) (0.013) (0.018) (0.404)
Wage Cos s 0.883 3.015 0.972 1.993 0.833 2.443
(0.291) (0.002) (0.484) (0.225) (0.442) (0.030)
Educa ional A ainmen 0.806 1.204 0.943 1.506 0.881 1.180
(0.014) (0.001) (0.037) (0.015) (0.053) (0.017)
Popula ion Densi y 0.510 0.630 1.270 8.362 −0.132 0.759
(0.127) (0.168) (0.038) (0.000) (0.738) (0.160)
Popula ion Densi y Squa ed −0.115 −0.132 −0.186 −6.012 0.00984 −0.128
(0.047) (0.146) (0.039) (0.000) (0.895) (0.142)
Tax −0.0189 −0.0763 −0.0268 −0.390 −0.0716 0.0372
(0.895) (0.444) (0.932) (0.049) (0.642) (0.764)
R&D Tax C edi 0.107 0.0915 0.177 0.244 0.0865 0.0262
(0.203) (0.266) (0.211) (0.144) (0.431) (0.775)
P e iousIn es men 2.011 2.493 2.455 2.446 1.316 2.152
(0.000) (0.000) (0.002) (0.001) (0.047) (0.000)
Indus y Es ablishmen s 0.657 0.759 0.279 1.026 0.910 0.736
(0.000) (0.000) (0.127) (0.000) (0.000) (0.001)
Technology Fi 0.617 0.985 0.765 1.318 0.714 0.920
(0.044) (0.000) (0.069) (0.002) (0.067) (0.000)
Doc o a es Fi 1.250 0.876 1.638 1.235 0.539 1.033
(0.031) (0.085) (0.045) (0.346) (0.446) (0.121)
Academic Resea ch Fi 1.589 0.933 1.631 1.150 1.392 0.680
(0.026) (0.002) (0.176) (0.039) (0.187) (0.161)
1982
R.Belde bos e al.
1 3
Es ima ion esul s o mixed logi models. Robus s anda d e o s clus e ed by pa en i m, and co esponding p alues in pa en heses
Table 5 (con inued)
Full Resea ch De elopmen
Science based
indus y
O he indus ies Science based
indus y
O he indus ies Science based
indus y
O he indus ies
Ra io Non-Uni e si y Pubs 0.904 0.275 1.403 −2.212 0.460 0.954
(0.150) (0.600) (0.148) (0.058) (0.574) (0.124)
Uni e si y Pa en s Fi −2.421 0.0168 −6.172 −1.612 −1.826 0.0786
(0.110) (0.900) (0.055) (0.026) (0.250) (0.528)
Applied Sha e −0.248 3.051 −0.904 3.249 0.142 3.292
(0.824) (0.005) (0.556) (0.016) (0.930) (0.007)
Academic Quali y: Applied 1.538 1.250 2.280 0.234 0.577 1.651
(0.001) (0.001) (0.000) (0.828) (0.449) (0.000)
Academic Quali y: Basic 2.073 −0.0555 2.489 1.495 1.762 −0.443
(0.002) (0.940) (0.015) (0.252) (0.047) (0.579)
# R&D In es men s 232 241 86 62 146 179
# Al e na i e Choices 354 354 354 354 354 354
Wald Chi2 565.89 639.75 292.76 231.52 498.96 468.37
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Signi ican Random Componen s
Wage Cos s 3.476
(0.000)
Popula ion Densi y 6.607 0.254
(0.000) (0.047)
P e ious In es men 1.088 1.988 2.113 1.224
(0.035) (0.000) (0.003) (0.024)
1989
He e ogeneous uni e si y esea ch and i m R&D loca ion…
1 3
Scandu a, A., & Iamma ino, S. (2022). Academic engagemen wi h indus y: The ole o esea ch quali y and
expe ience. The Jou nal o Technology T ans e , 47(4), 1000–1036.
Schmoch, U., La ille, F., Pa el, P., & F ie sch, R. (2003). Linking echnology a eas o indus ial sec o s. Final
Repo o he Eu opean Commission, DG Resea ch, 1, 100.
Shimizu ani, S., & Todo, Y. (2008). Wha de e mines o e seas R&D ac i i ies? The case o Japanese mul ina-
ional i ms. Resea ch Policy, 37(3), 530–544.
Siegel, D. S., Waldman, D., & Link, A. (2003). Assessing he impac o o ganiza ional p ac ices on he ela-
i e p oduc i i y o uni e si y echnology ans e o ices: An explo a o y s udy. Resea ch Policy, 32(1),
27–48.
Siegel, D. S., & W igh , M. (2015). Academic en ep eneu ship: Time o a e hink? B i ish Jou nal o Man-
agemen , 26(4), 582–595. h ps:// doi. o g/ 10. 1111/ 1467- 8551. 12116
Sil e man, B. S. (1999). Technological esou ces and he di ec ion o co po a e di e si ica ion: Towa d an
in eg a ion o he esou ce-based iew and ansac ion cos economics. Managemen Science, 45(8),
1109–1124.
Sime h, M., & Ra o, J. D. (2013). Wha makes companies pu sue an open science s a egy? Resea ch Policy,
42(9), 1531–1543.
Sub amanian, A. M., Lim, K., & Soh, P.-H. (2013). When bi ds o a ea he don’ lock oge he : Di e en
scien is s and he oles hey play in bio ech R&D alliances. Resea ch Policy, 42(3), 595–612. h ps:// doi.
o g/ 10. 1016/j. espol. 2012. 12. 002
Suzuki, S., Belde bos, R., & Kwon, H. U. (2017). The loca ion o mul ina ional i ms’ R&D ac i i ies ab oad:
Hos coun y uni e si y esea ch, uni e si y–indus y collabo a ion, and R&D he e ogenei y. Ad ances
in S a egic Managemen , 36, 125–159.
Thu sby, J. G., & Thu sby, M. C. (2002). Who is selling he i o y owe ? Sou ces o g ow h in uni e si y
licensing. Managemen Science, 48(1), 90–104. h ps:// doi. o g/ 10. 1287/ mnsc. 48.1. 90. 14271
Thu sby, J. G., & Thu sby, M. C. (2011). Has he Bayh–Dole ac comp omised basic esea ch? Resea ch Pol-
icy, 40(8), 1077–1083. h ps:// doi. o g/ 10. 1016/j. espol. 2011. 05. 009
Toole, A. A. (2012). The impac o public basic esea ch on indus ial inno a ion: E idence om he pha ma-
ceu ical indus y. Resea ch Policy, 41(1), 1–12. h ps:// doi. o g/ 10. 1016/j. espol. 2011. 06. 004
T ain, K. E. (2009). Disc e e choice me hods wi h simula ion. Camb idge: Camb idge Uni e si y P ess.
Veugele s, R., & Wang, J. (2019). Scien i ic no el y and echnological impac . Resea ch Policy, 48(6), 1362–
1372. h ps:// doi. o g/ 10. 1016/j. espol. 2019. 01. 019
Von Hippel, E. (1994). S icky in o ma ion and he locus o p oblem-sol ing—Implica ions o inno a ion.
Managemen Science, 40(4), 429–439. h ps:// doi. o g/ 10. 1287/ mnsc. 40.4. 429
on Zed wi z, M., & Gassmann, O. (2002). Ma ke e sus echnology d i e in R&D in e na ionaliza ion: ou
di e en pa e ns o managing esea ch and de elopmen . Resea ch Policy, 31(4), 569–588. h ps:// doi.
o g/ 10. 1016/ s0048- 7333(01) 00125-1
Wang, M., Soe an o, D., Cai, J., & Muni , H. (2022). Scien is o En ep eneu ? Iden i y cen ali y, uni e si y
en ep eneu ial mission, and academic en ep eneu ial in en ion. The Jou nal o Technology T ans e ,
47, 119–146.
Wang, J., & Ve be ne, S. (2021). Two ales o science echnology linkage: Pa en in- ex e sus on -page e -
e ences. a Xi p ep in a Xi : 2103. 08931.
Wilson, D. J. (2009). Begga hy neighbo ? The in-s a e, ou -o -s a e, and agg ega e e ec s o R&D ax c edi s.
The Re iew o Economics and S a is ics, 91(2), 431–436.
You ie, J., & Shapi a, P. (2008). Building an inno a ion hub: A case s udy o he ans o ma ion o uni e si y
oles in egional echnological and economic de elopmen . Resea ch Policy, 37(8), 1188–1204.
Zah inge , K., Kolympi is, C., & Kalai zandonakes, N. (2017). Academic knowledge quali y di e en ials and
he quali y o i m inno a ion. Indus ial and Co po a e Change, 26(5), 821–844. h ps:// doi. o g/ 10.
1093/ icc/ d w050
Zan ei, A. (2000). T ansna ional i ms and he changing o ganisa ion o inno a i e ac i i ies. Camb idge Jou -
nal o Economics, 24(5), 515–542. h ps:// doi. o g/ 10. 1093/ cje/ 24.5. 515
Zucke , L. G., Da by, M. R., & A ms ong, J. S. (2002). Comme cializing knowledge: Uni e si y science,
knowledge cap u e, and i m pe o mance in bio echnology. Managemen Science, 48(1), 138–153.
h ps:// doi. o g/ 10. 1287/ mnsc. 48.1. 138. 14274
Zucke , L. G., Da by, M. R., & B ewe , M. B. (1998). In ellec ual human capi al and he bi h o U.S. bio ech-
nology en e p ises. Ame ican Economic Re iew, 88(1), 290–306. h ps:// doi. o g/ 10. 2307/ 116831
Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps and
ins i u ional a ilia ions.
