Khez i, Mohsen
A icle
Assessing en ep eneu ial ecosys ems' in luence on g een
echnology inno a ion: A c oss-coun y analysis
Jou nal o Inno a ion & Knowledge (JIK)
P o ided in Coope a ion wi h:
Else ie
Sugges ed Ci a ion: Khez i, Mohsen (2025) : Assessing en ep eneu ial ecosys ems' in luence on
g een echnology inno a ion: A c oss-coun y analysis, Jou nal o Inno a ion & Knowledge (JIK),
ISSN 2444-569X, Else ie , Ams e dam, Vol. 10, Iss. 4, pp. 1-15,
h ps://doi.o g/10.1016/j.jik.2025.100738
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/327633
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 ps://c ea i ecommons.o g/licenses/by/4.0/
Assessing en ep eneu ial ecosys ems’ in luence on g een echnology
inno a ion: A c oss-coun y analysis
Mohsen Khez i
Visi ing Senio Fellow. Depa men o Geog aphy and En i onmen , London School o Economics and Poli ical Science (LSE), London, UK
ARTICLE INFO
JEL classi ica ion:
Q55
Q58
L26
O31
C23
Keywo ds:
En i onmen al egula ions
G een Technology Inno a ion, GTI
panel smoo h h eshold eg ession, PSTR
En ep eneu ial indica o s
ABSTRACT
This s udy explo es he impac s o 11 di e se en ep eneu ship indica o s on g een echnology inno a ion (GTI)
o de e mine he op imal en i onmen al egula o y amewo k ha os e s g een en ep eneu ship. Addi ionally,
he s udy in es iga es he impac s o en i onmen al egula ions on GTI by u ilizing nonlinea panel smoo h
h eshold eg ession (PSTR) models on da a collec ed om 18 coun ies om 2002 o 2020. By iden i ying a
c i ical egula o y h eshold o 1.89, he esea ch e eals how a ying le els o en i onmen al egula ions
signi ican ly in luence GTI dynamics. The es ima ion esul s emphasize ha GDP pe capi a and inancial
de elopmen a e c i ical in os e ing GTI. Howe e , s ingen en i onmen al egula ions can coun e ac hese
posi i e e ec s. U baniza ion and ade openness also posi i ely in luence GTI, wi h en i onmen al egula ions
complemen ing hei impac s. The ansi ion o a se ice-o ien ed indus ial s uc u e posi i ely a ec s GTI. The
esul s unde sco e he nega i e impac o en ep eneu ship indica o s, po en ially di e ing esou ces away om
GTI. None heless, en i onmen al egula ions wi h s ingen en o cemen mechanisms can coun e balance he
nega i e impac s o speci ic en ep eneu ship me ics. Among he en ep eneu ship indica o s analyzed,
inancing o en ep eneu s, go e nmen al suppo and policies, and go e nmen al p og ams exhibi an in e ed
U-shaped impac pa e n, peaking a speci ic le els o en i onmen al egula ion.
In oduc ion
The con empo a y challenge o he clima e c isis looms la ge,
necessi a ing immedia e a en ion and collec i e ac ion (Reckien e al.,
2018). A shi om unsus ainable g ow h models owa d sus ainable
de elopmen pa hways is necessa y o ackle his u gen issue. This shi
equi es ealloca ing capi al and p omp ly adop ing clean p oduc ion
echnologies (Huang e al., 2021; G. Luo e al., 2023). A he co e o his
ansi ion lies he concep o G een Technology Inno a ion (GTI). I
plays a pi o al ole in ha monizing en i onmen al p o ec ion and eco-
nomic de elopmen , os e ing a u u e cha ac e ized by g een-o ien ed,
inno a ion-d i en g ow h (J. L. Du e al., 2019; Fei e al., 2016; Guo
e al., 2020; M. Wang e al., 2021; Xu e al., 2023). GTI se es as a d i e
o a shi owa d a mo e sus ainable u u e (S ucki & Woe e , 2017).
En ep eneu s’ inclina ion o adop en i onmen ally iendly in-
no a ions is signi ican ly in luenced by en i onmen al conce ns
(Hobman & F ede iks, 2014; Polas e al., 2023; Xie & Zhao, 2018). I
also shapes hei p e e ence o eco- iendly ad ancemen s (K aus e al.,
2020). Ne e heless, he commi men o businesses o ecological sus-
ainabili y emains he subjec o ongoing deba e. Many i ms hesi a e o
in es in long- e m ini ia i es ha do no yield immedia e e u ns,
emphasizing he need o policy-d i en in e en ions (Melande &
A idsson, 2022; Zhang e al., 2020b). G een echnology de elopmen
in ol es p olonged esea ch and de elopmen (R&D) cycles, subs an ial
inancial commi men s, inhe en isks, and complexi y ha demand
inc eased co po a e dedica ion o inno a ion (Shao e al., 2020; H. Yu
e al., 2023). Policy in e en ions a e pi o al in o e coming hese
challenges and d i ing g een inno a ion o wa d (Neme , 2012; Popp &
Newell, 2012; Rogge & Schleich, 2018).
En i onmen al egula ions can impac GTI di e en ly due o he
inno a ion compensa ion e ec and he cos compliance e ec . These
egula ions a e cons ain s and incen i es (Tian e al., 2021). While hey
may inc ease p oduc ion cos s (Zhang & Dong, 2022), en e p ises mus
weigh he ade-o s be ween hese e ec s. When he bene i s o inno-
a ion compensa ion ou weigh compliance cos s, i p o ides a s ong
incen i e o co po a e GTI. The in ensi y o en i onmen al egula ion
implemen a ion di ec ly in luences he cos compliance e ec in com-
panies; howe e , i s con ibu ion o he inno a ion compensa ion e ec
emains subjec o heo e ical and empi ical con lic s.
The Po e hypo hesis sugges s ha i ms a e mo i a ed o inno a e
E-mail add ess: [email p o ec ed].
Con en s lis s a ailable a ScienceDi ec
Jou nal o Inno a ion & Knowledge
jou nal homepage: www.else ie .com/loca e/jik
h ps://doi.o g/10.1016/j.jik.2025.100738
Recei ed 22 Oc obe 2024; Accep ed 19 May 2025
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
A ailable online 28 May 2025
2444-569X/C own Copy igh © 2025 Published by Else ie España, S.L.U. on behal o Jou nal o Inno a ion & Knowledge. This is an open access a icle unde he
CC BY license ( h p://c ea i ecommons.o g/licenses/by/4.0/ ).
o enhance hei long- e m compe i i eness in esponse o en i on-
men al egula ions (Po e & Van De Linde, 2017). Con e sely, a
cos - ocused pe spec i e a gues ha en i onmen al egula ions migh
escala e aba emen cos s, po en ially discou aging in es men s in
inno a ion (Ja e e al., 2000). Nume ous s udies ha e del ed in o he
ela ionship be ween en i onmen al egula ion and g een inno a ion,
e ealing a complex pic u e. Some empi ical s udies epo posi i e as-
socia ions (X. Cai e al., 2020; Chen e al., 2022; Yi e al., 2020), while
o he s unco e nega i e o nonlinea ela ionships (Yi e al., 2019;
Zhang e al., 2020a; Zhang e al., 2022). These di e gen indings
highligh he adap abili y o he c owding-ou heo y and he Po e
hypo hesis ac oss di e en con ex s.
The e ec i eness o he Po e hypo hesis is closely linked wi h he
local con ex s in which i is applied, which can a y signi ican ly in
space and ime. P io esea ch emphasizes he ole o local con ex s in
he Po e hypo hesis, ocusing on empo al and spa ial a ia ions in he
in e play be ween en i onmen al egula ion and p oduc i i y o inno-
a ion (G. L. Zhao e al., 2022; Zhu e al., 2019). Thus, a complex,
nonlinea pa e n eme ges in how en i onmen al egula ions a ec GTI
o e ime and ac oss a ious samples, an aspec o en o e looked in
empi ical s udies.
En ep eneu ship is in ica ely linked wi h knowledge, echnological
ad ancemen , and inno a ion, holding he po en ial o s imula e eco-
nomic g ow h (Bendig e al., 2022; W. M. Cohen & Le in hal, 1989).
Howe e , en ep eneu ial ac i i ies a ely un old in isola ion; a he ,
hey occu wi hin en ep eneu ial ecosys ems consis ing o in e -
connec ed ac o s, ins i u ions, and esou ces (Isenbe g, 2010; Spigel,
2017). Such ecosys ems—which include in es o s, incuba o s, sup-
po i e policies, and echnology ans e in as uc u es—can shape
en ep eneu s’ decisions o pu sue g een inno a ions by p o iding
essen ial esou ces, men o ing, and collabo a i e pla o ms (Al edalen
& Boschma, 2017; Elia e al., 2020). Indeed, when he ecosys em is
obus and well-coo dina ed, en ep eneu s encoun e lowe ba ie s o
en e ing ma ke s wi h eco- iendly p oduc s, making i mo e easible o
align business oppo uni ies wi h sus ainabili y goals (S am & an de
Ven, 2021). Consequen ly, g een en ep eneu ship eme ges as a
powe ul a enue o sus ainable de elopmen as en ep eneu s le e age
ecosys em suppo o de ise inno a i e p oduc ion p ocesses and
p oduc s capable o mi iga ing en i onmen al deg ada ion (B. Cohen
e al., 2008; Coulibaly e al., 2018; Yo k & Venka a aman, 2010). The
impo ance o he en ep eneu ial ecosys em is u he highligh ed by
ecen s udies showing ha es ablished i ms a e inc easingly pa ne ing
wi h bo n g een s a ups— h ough acquisi ions, in es men s, and s a-
egic alliances— o access high- ech en i onmen al knowledge (Demi el
e al., 2019). This syne gis ic ela ionship illus a es how en ep e-
neu ial ecosys ems no only os e collabo a ion bu also ca alyze he
adop ion o GTI, guiding bo h s a ups and es ablished en e p ises o-
wa d mo e sus ainable ajec o ies.
Exis ing esea ch p esen s con lic ing iews on whe he en i on-
men al egula ions spu o hinde inno a ion (Ja e e al., 2000; Po e &
Van De Linde, 2017), sugges ing a need o de e mine whe he such
policies exhibi h eshold e ec s. Likewise, despi e a g owing body o
e idence linking well-de eloped en ep eneu ial ecosys ems wi h
s onge inno a ion capabili ies (Iqbal e al., 2020; Yo k & Venka a a-
man, 2010), exis ing esea ch p o ides limi ed insigh s in o how speci ic
ecosys em ac o s in e ac wi h en i onmen al egula ion egimes o
os e GTI. Mo eo e , while schola s acknowledge ha en ep eneu ial
ecosys ems can supply he inancial, social, and ins i u ional suppo
equi ed o acili a e g een s a ups and eco- iendly p ocesses, he e
emains a dea h o s udies examining h eshold o nonlinea mecha-
nisms h ough which hese ecosys ems migh ampli y—o diminish— he
e icacy o en i onmen al policies. Acco dingly, his s udy add esses
hese gaps by in es iga ing how en ep eneu ship indica o s in eg a e
wi h di e en in ensi ies o en i onmen al egula ion o shape GTI
ou comes ac oss coun ies. This s udy makes wo p ima y con ibu ions:
(1) he panel smoo h h eshold eg ession (PSTR) model (Colle az &
Hu lin, 2006; Fok e al., 2005) is adop ed o o e come he
c oss-sec ional he e ogenei y c i ique o en posed agains con en ional
panel da a app oaches (Hsiao, 2014), hus allowing o cap u ing he
nonlinea and con ex -speci ic impac s o en i onmen al egula ion on
GTI; and (2) he e ec s o 11 di e se en ep eneu ship indica-
o s— e lec ing he mul i ace ed na u e o he en ep eneu ial ecosys-
em—on GTI a e e alua ed, demons a ing how hese indica o s
in e sec wi h a ying le els o en i onmen al egula ion. In line wi h
hese con ibu ions, h ee esea ch ques ions a e posed: (1) How does
he in ensi y o en i onmen al egula ion, iewed h ough di e en
egula o y egimes, shape he nonlinea ela ionship be ween egula ion
and GTI? (2) How do di e se en ep eneu ship indica o s in luence GTI
unde a ying in ensi ies o en i onmen al egula ion? and (3) Is he e
an op imal alignmen be ween egula ion in ensi y and en ep eneu -
ship de elopmen o maximize GTI? F om hese ques ions, i is hy-
po hesized ha :
H1: En i onmen al egula ion exe s h eshold e ec s on GTI;
H2: En ep eneu ship indica o s—key componen s o he en ep e-
neu ial ecosys em—posi i ely in luence GTI, albei o di e en de-
g ees depending on he egula o y egime; and,
H3: S ic e egula ions ampli y hese posi i e en ep eneu ial
ecosys em impac s on GTI, signaling a syne gis ic in e ac ion be-
ween policy en o cemen and en ep eneu ship.
These hypo heses a e es ed h ough he PSTR amewo k o o e
esh insigh s in o how en ep eneu ial ecosys ems and en i onmen al
egula ions join ly shape g een inno a ion ajec o ies, hus enhancing
he unde s anding o —and policy ecommenda ions o —sus ainable
de elopmen .
The emainde o his pape is o ganized as ollows: Sec ion 2 p o-
ides an in-dep h explo a ion o he li e a u e; Sec ion 3 del es in o he
da a sou ces and in oduces he p oposed econome ic models; Sec ion 4
ad ances he discussion by add essing empi ical indings and hei in-
e p e a ions; and, Sec ion 5 concludes he pape and o e s and po-
en ial policy implica ions.
Li e a u e e iew
G een echnology inno a ion (GTI)
Engaging wi h GTI is in ica ely linked wi h he p inciples o
ecological mode niza ion heo y, which con ibu e o he cu ailmen o
pollu ion and d i e he ans o ma ion o di e se indus ial sec o s
(Bu el, 2000; W. Cai & Li, 2018; Z. Li e al., 2023). In his espec , i
sa egua ds aluable esou ces, enhances he en i onmen , and ad ances
economic p og ess, o ging a ha monious syn hesis be ween ecology and
he economy (Ba bie i e al., 2020; F. Dong e al., 2022; Miao e al.,
2017; Shan e al., 2022). This in eg a ion o g een inno a ion encom-
passes a wide spec um o ecologically sus ainable c ea i e endea o s,
spanning om eco- iendly echnologies o en i onmen ally-conscious
p oduc s and se ices. This p esen s a comp ehensi e bluep in o a
mo e sus ainable u u e (Ma ínez-Ros & Kunapa a awong, 2019).
The signi icance o g een inno a ion su passes me e economic
expansion; i is c ucial in d i ing a b oade ecological ans o ma ion
and mi iga ing en i onmen al damage by applying no el echnologies
(Flamme e al., 2019; Ka imi Takalo e al., 2021). In his con ex , GTI
ex ends angible business ad an ages and pa es he pa h o sus ainable
de elopmen (Deng e al., 2019; K. Du e al., 2021). Sus ainable ene gy
s a egies pionee his me amo phosis, encompassing aspec s such as
g een ene gy adop ion, ene gy e iciency enhancemen s, and p o ec i e
measu es a ge ing clima e change mi iga ion and o e all ene gy se ice
imp o emen (Appiah e al., 2022; Islam e al., 2012). Na ions wo ld-
wide a e uni ed in hei commi men o champion g een inno a ion as a
condui o ealizing sus ainable de elopmen , acknowledging i s po-
en ial o ackle he p essing challenges o he cu en e a (A. Wang
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
2
e al., 2023). This collec i e dedica ion unde sco es he immediacy o
sha ed esponsibili y o ha ness he o ce o inno a ion o pu sue a
g eene and mo e sus ainable u u e.
Po e hypo hesis
Con en ional economic wisdom has held o a conside able ime ha
en i onmen al egula ions place inancial bu dens on businesses, and
his po en ially hampe s hei p o i abili y (Palme e al., 2018). How-
e e , Po e (1991) in oduced a g oundb eaking pe spec i e, u he
expanded by Po e and an de Linde (1995), which challenges his
con en ional belie . Known as he Po e hypo hesis, i s a es ha
en i onmen al egula ions can spa k inno a ion wi hin companies,
which po en ially leads o imp o ed p oduc i i y and cos sa ings.
This hypo hesis mani es s in wo dis inc e sions; he weake
e sion sugges s ha egula ions can s imula e inno a ion, ul ima ely
gene a ing bene i s ha ou weigh associa ed cos s, while he s onge
e sion p oposes ha inno a ion d i en by egula o y demands can
e ec i ely o se compliance expenses. Ne e heless, in es iga ions in o
hese hypo heses ha e p oduced mixed ou comes (M. A. Cohen & Tubb,
2018; Eli & Bui, 2001). The s ong o m o he Po e hypo hesis has
emained a opic o con en ion and examina ion, con ibu ing o
inconclusi e indings a ibu ed o a ious ac o s. Howe e , ecen
esea ch leans owa d mo e posi i e ou comes han ea lie s udies
(Ambec e al., 2013; M. A. Cohen & Tubb, 2018).
Cohen and Tubb (2018) conduc ed an ex ensi e me a-analysis o
e alua e he impac s o en i onmen al egula ions on p oduc i i y and
compe i i eness and ound ha hese egula ions o en ha e a mo e
a o able impac a b oade geog aphic scales. These indings align wi h
he co e p inciples o he s ong e sion o he Po e hypo hesis.
Empi ical esea ch has also s a ed o conside egional dispa i ies in he
implica ions o he Po e hypo hesis (Y. Luo e al., 2021). The e ec-
i eness o en i onmen al egula ion in p omo ing sus ainable ans-
o ma ions depends on a numbe o ac o s, including local indus ial
s uc u es, ypes o policies, le els o economic de elopmen , and
i m-speci ic dis inc ions (Cos an ini e al., 2017; J. L. Du e al., 2019;
F anco & Ma in, 2017; Liu e al., 2022)
En i onmen al egula ion and GTI
The in ica e in e play be ween en i onmen al egula ions and he
ad ancemen o GTI has been he subjec o ex ensi e schola ly in es-
iga ion. Howe e , achie ing a consensus ega ding his complex ela-
ionship emains a o midable challenge. Wi hin his in ica e web o
esea ch, se e al s udies ha e shed ligh on a posi i e connec ion be-
ween en i onmen al egula ions and he p oli e a ion o g een pa en s.
Fo example, Cai e al. (2020) and Fang e al. (2021) independen ly
p esen ed empi ical e idence suppo ing a posi i e co ela ion be ween
GTI and en i onmen al egula ions, pa icula ly in hea ily pollu ing
indus ies. Thei indings unde sco e he nuanced na u e o his
connec ion.
F om a empo al pe spec i e, i ms’ a i udes owa d g een inno a-
ion exhibi e ol ing dynamics o e ime. Aghion e al. (2016) and
S ucki and Woe e (2017) demons a ed ha en i onmen al egula-
ions exe a wo-s age impac on inno a ion, ansi ioning om an
ini ial wai -and-see app oach o a p oac i e s ance on g een inno a ion
o e ime. The s ingency o en i onmen al egula ions signi ican ly
in luences i ms’ p e e ences o g een inno a ion. As egula ions
become mo e s ingen , he cos s o aba emen ise, po en ially leading
o a c owding-ou e ec . Howe e , a g owing awa eness o he bene i s
associa ed wi h g een inno a ion d i es a heigh ened demand o such
ini ia i es. I is an icipa ed ha he e will be a pi o al junc u e a which
he inno a ion o se supe sedes he c owding-ou e ec
(Dechezlep ˆ
e e & Sa o, 2017).
F om a spa ial pe spec i e, he du a ion o he c owding ou o
inno a ion o se s ages a ies conside ably ac oss di e en egions.
Some ci ies may ind hemsel es linge ing in he ini ial s age due o
limi a ions in hei inno a ion capaci ies and high ans o ma ion cos s,
while o he s apidly p og ess o he second s age. The magni ude o
hese e ec s also a ies among ci ies, as demons a ed by Balland and
Rigby (2017))) and Hidalgo e al. (2018). Zhang e al. (2022) ound ha
ca bon emission ading can po en ially s i le g een inno a ion, espe-
cially in eas e n China wi h low emission in ensi y. Fu he mo e, local
and adjacen en i onmen al egula ions can a ec g een p oduc i i y
h ough mechanisms ela ed o g een inno a ion and pollu ion ans e
(Peng, 2020). The le el o economic de elopmen wi hin ci ies can
signi ican ly shape he dynamics o his ela ionship, as epo ed by Du
e al. (2021). The openness o local ma ke s can p o ide i al inancial
access and echnological suppo o g een inno a ions (Feng e al.,
2018).
Fu he mo e, Li and Du (2021) and Dong e al. (2020) un eiled a
U-shaped cu e ela ionship be ween hese a iables, emphasizing he
p esence o spa ial spillo e e ec s, which add ano he laye o in icacy
o his mul i ace ed issue. Fu he mo e, compe i ion among local go -
e nmen s is pi o al in shaping he dynamics o he GTI en e p ise (Deng
e al., 2019). This compe i i e landscape can gi e ise o a complex,
in e ed U-shaped ela ionship wi h GTI. As he inancial s uc u e
wi hin hese local go e nmen s s eng hens, i ends o p omo e GTI.
Howe e , he scale and e iciency o inancial ope a ions may hinde he
e y inno a ion in ended o be os e ed (L e al., 2021).
C ucial ac o s ha in luence he pu sui o g een inno a ion unde
en i onmen al egula ion include local indus ial a ibu es, economic
s uc u e, and inno a ion capaci ies. Ci ies wi h obus iscal capaci ies
and well-es ablished esea ch ins i u ions ha e a no able ad an age in
p omo ing g een inno a ion. While p e ious s udies explo ed he im-
pac s o en i onmen al egula ion in di e en con ex s, he spa io em-
po al non-s a iona i y in he ela ionship calls o inc eased a en ion.
Mos exis ing s udies ely on egional dummies o con en ional econo-
me ic echniques, which u nish global a e age es ima es ye o en ail
o cap u e he in ica e spa io empo al pa e ns ha unde lie he in e -
ac ion be ween en i onmen al egula ion and inno a ion.
Empi ical model and da a
This in es iga ion del es in o annual da a om 2002 o 2020 o
un eil he ac o s shaping GTI ac oss 18 coun ies. A cen al cons ain
o his s udy lies in he sca ci y o a ailable da a ega ding en ep e-
neu ial me ics ac oss di e en empo al and geog aphical con ex s.
Relying on insigh s gleaned om di e se schola ly in es iga ions, i
becomes appa en ha a numbe o pi o al de e minan s signi ican ly
in luence GTI. Acco ding o he li e a u e (J. Li e al., 2022; B. Lin & Ma,
2022; Yang e al., 2021; H. Yu e al., 2023), hese ac o s encompass he
loga i hm o GDP pe capi a (lnGDPP), a ma ke o a na ion’s economic
de elopmen s age, he ese oi o human capi al (lnHC), he ex en o
u baniza ion (lnURB), he deg ee o ade openness (lnOPE), and he
composi ion o indus ial s uc u e (lnIS). Nume ous in es iga ions ha e
s udied he ela ionship be ween indus ial s uc u e and GTI, e ealing
ha he composi ion o indus ies, speci ically he a io o he e ia y
sec o o he seconda y sec o , exe s a posi i e impac on GTI (Shen
e al., 2021; Zhao e al., 2022b). These e ec s can be a ibu ed o he
alignmen o e ia y indus y de elopmen wi h he ad ancemen o
g een echnology (K. Du e al., 2021).
Fu he mo e, he inancial sys em plays a key ole in he concen-
a ion and alloca ion o unds (C. H. Yu e al., 2021). A less de eloped
inancial sys em can impede en e p ises’ access o c edi inancing,
esul ing in inadequa e in es men in GTI (Ande sen, 2017). In a
b oade con ex , inancial de elopmen , as indica ed by lnFD, signi i-
can ly in luences he p opensi y o en e p ises o engage in echnolog-
ical inno a ion (Noailly & Smee s, 2022). As a esul , hese elemen s a e
in eg a ed as con ol ac o s wi hin he analy ical amewo k o his
s udy, which is elucida ed in he ollowing model, e e ed o as Model
A:
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
3
lnGTIi =β0+β1lnGDPPi +β2lnHCi +β3lnURBi +β4lnOPEi +β5lnISi
+β6LnFD
(1)
Table 1 and 2.
This analysis in ol es a loga i hmic ans o ma ion o he a iables,
yielding coe icien s ha signi y elas ici ies. Table 3 epo s he indings
esul ing om he es ima ion o his model using a ange o panel da a
echniques. Fu he mo e, in conjunc ion wi h he inclusion o con ol
a iables, his s udy places pa icula emphasis on wo pi o al explan-
a o y elemen s: he na u al loga i hm o en ep eneu ship indica o s
(lnENT) and en i onmen al egula ions (REG). Inco po a ing hese
a iables in o he model esul s in he o mula ion o Model B, as illus-
a ed below. The ou comes o es ima ing he model a e p esen ed in
Table 4.
lnGTIi =β0+β1lnGDPPi +β2lnHCi +β3lnURBi +β4lnOPEi +β5lnISi
+β6LnFD +β7lnENTi +β8REG
(2)
Panel smoo h h eshold eg ession (PSTR)
A panel smoo h h eshold eg ession (PSTR) model is a p ac ical
app oach o add essing nonlinea i y wi hin he model. PSTR has dual
in e p e a ions: i s , as a egime-swi ching model wi h a ew ex eme
egimes linked wi h he ex eme alues o a ansi ion unc ion, an-
si ioning smoo hly; second, as a model allowing a con inuum o egimes,
each wi h dis inc ansi ion unc ion alues. PSTR p o es help ul in his
con ex by accoun ing o c oss-coun y he e ogenei y and ime ins a-
bili y o elas ici ies wi hou equi ing p e-de ined classi ica ions. Addi-
ionally, he use o he PSTR model can enhance es ima e eliabili y
ega ding non-s a iona i y. In con as o ime se ies, non-s a iona i y
e ec s in linea panel models a e di e en . Pooling c oss-sec ion and
ime se ies obse a ions can mi iga e esidual impac s while e aining
explana o y a iable s eng h, yielding consis en long- un eg ession
coe icien es ima es (Phillips & Moon, 1999). Focusing on a basic sce-
na io wi h wo independen a iables (x1i and x2i ), wo dis inc e-
gimes, and a soli a y ansi ion unc ion, he esul an PSTR model is
o mula ed as:
yi =
μ
i+
α
0x1i +β0x2i + [
α
1x1i +β1x2i ]h(qi ;γ,c) +
ε
i (3)
whe e qi is he h eshold a iable. The e o e m
ε
i is conside ed in-
dependen and iden ically dis ibu ed wi h a mean o 0 and a a iance o
σ
2. The unc ion go e ning he ansi ion, deno ed as h(qi ;γ,c), emains
limi ed and con inuous conce ning he h eshold a iable qi . Building
upon ea lie esea ch by G ange and Te ¨
as i a (1993) on STAR models
in ime se ies, Gonz´
alez e al. (2004) p oposed a ansi ion unc ion as:
h(qi ;γ,c) = [1+exp(−γ∏
m
z=1
(qi −cz))]−1
,γ>0,c1≤.. ≤cm
whe e c= (c1, .., cm)ʹ ep esen s a mul i-dimensional ec o deno ing
loca ion pa ame e s. He e, γ is esponsible o de e mining he s eepness
o he ansi ion unc ion. Ma hema ically, his model can be e o mu-
la ed in o:
yi =
μ
i+Ψʹ
0Wi +Ψʹ
1Wi h(qi ;γ,c) +
ε
i
whe e Ψj=(
α
jβj)ʹ
o j= (0,1),and Wi = [ x1i x2i ]ʹ. Addi ionally,
Wi is de ined as [x1i x2i ]ʹ, encapsula ing he a iables x1i and x2i o
coun y i a ime . Gonz´
alez e al. (2004) p oposed an ex ension in o-
ducing +1 ex eme egimes. This ex ension, e med he gene al ad-
di i e PSTR model, is de ined as:
yi =&
μ
i+
α
0x1i +β0x2i +∑
j=1[
α
jx1i +βjx2i ]hj(qi ;γj,cj)+
ε
i (4)
o equi alen ly,
yi =
μ
i+Ψʹ
0Wi +∑
j=1
Ψʹ
jWi hj(qi ;γj,cj)+
ε
i
The ansi ion unc ion hj(qi ;γj,cj)is in luenced by bo h he slope
pa ame e s γj and a se o m loca ion pa ame e s cj. In his b oade
concep ualiza ion, he o al impac o x2i on yi , wi hin he con ex o
coun y i a ime , is a icula ed as he weigh ed mean o he +1
coe icien s βj acqui ed om he +1 dis inc ex eme egimes.
∂
yi
∂
x2i
=β0+∑
j=1
βjhj(qi ;γj,cj)∀i,∀ (5)
Table 5 and 6.
To es ima e he nonlinea i y o Eq. (4) o he p esen esea ch,
model C is de ined in Eq. (6). The es ima ion esul s o Eq. (6) a e e-
po ed in Table 7. He e, en i onmen al egula ion (REG) is de ined as
h eshold a iables, and depending on hei alues in di e en coun ies
and o e ime, hey can in luence he e ec s o independen a iables in
he o m o a ying en ep eneu ship egimes.
lnGTIi =β0+β10lnGDPPi +β20lnHCi +β30lnURBi +β40lnOPEi
+β50lnISi +β60LnFD +β70lnENTi +∑
j=1
[β11lnGDPPi
+β21lnHCi +β31lnURBi +β41lnOPEi +β51lnISi +β61LnFD
+β71lnENTi ]hj(REGi ;γj,cj)+
ε
i
(6)
D awing upon he es ima ion esul s o Eq. (6), Eq. (5) can be
calcula ed o all independen a iables o he model. The e ec s o o al
Table 1
De ini ions o a iables.
Va iable Va iable cons uc ed Sou ce
lnGTI =log(GTI);GTI =In en ions pe capi a in en i onmen -
ela ed echnologies
OECD
REG =Tax e enue ( % o GDP) on o al en i onmen OECD
lnGDPP =log(GDPP);GDPP=GDP pe capi a (cons an 2015 US$) WDI
lnURB =log(URB);URB=U ban popula ion ( % o he o al
popula ion)
WDI
lnOPE =log(OPE);OPE=T ade Openness ( % o GDP( WDI
lnIS =log(SE)
MA);SE=Se ices, alue added (cons an 2015 US
$));;MA=Manu ac u ing, alue added (cons an 2015 US$)
WDI
lnFD =log(FD);FD =Financial de elopmen index IMF
lnFE =log(GS);GS =Go e nmen al suppo and policies GEM
lnGS =log(GS);GS =Go e nmen al suppo and policies GEM
lnTB =log(TB);TB =Taxes and bu eauc acy GEM
lnGP =log(GP);GP =Go e nmen al p og ams GEM
lnBE =log(BE);BE =Basic school en ep eneu ial educa ion and
aining
GEM
lnPE =log(PE);PE =Pos school en ep eneu ial educa ion and
aining
GEM
lnRD =log(RD);RD =R&D ans e GEM
lnCP =log(CP);CP =Comme cial and p o essional in as uc u e GEM
lnMD =log(MD);MD =In e nal ma ke dynamics GEM
lnMO =log(MO);MO =In e nal ma ke openness GEM
lnPS =log(PS);PS =Physical and se ices in as uc u e GEM
WDI: Wo ld De elopmen Indica o ; h ps://da aca alog.wo ldbank.o g/da ase
/wo ld-de elopmen -indica o s.
GEM: Global En ep eneu ship Moni o ; h ps://www.gemconso ium.
o g/da a.
OECD: O ganiza ion o Economic Coope a ion and De elopmen ; h ps://s a s.
oecd.o g/.
IMF: In e na ional Mone a y Fund; h ps://da a.im .o g/?sk= 8032e80-b3
6c-43b1-ac26-493c5b1cd33b.
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
4
g oss domes ic p oduc (GDP) pe capi a on GTI a e w i en as:
∂
lnGTIi
∂
lnGDPPi
=β10 +∑
j=1
β1jhj(REGi ;γj,cj)∀i,∀ (7)
The analysis o he impac s o a ious ac o s on GTI included es i-
ma ing 21 dis inc models. These models inco po a ed consis en con ol
a iables in all cases, including he loga i hm o GDP pe capi a, human
capi al, u baniza ion, ade openness, and indus ial s uc u e. To
ensu e he independence o he assessmen and o mi iga e po en ial
collinea i y challenges among he 11 en ep eneu ship indices, sepa a e
es ima ion models we e used o each. To s eamline he p esen a ion o
he indings in Table 5, he abb e ia ion "lnENT" was used o ep esen
all en ep eneu ship indica o s. Fo example, in Models 3 and 7, lnENT
deno es he loga i hms o go e nmen al p og ams and comme cial and
p o essional in as uc u e, espec i ely. Table 1 shows which en e-
p eneu ship indica o co esponds o a gi en es ima ion model.
Da a
While ea lie in es iga ions ha e p ima ily d awn upon a ange o
me ics, including R&D in es men , pa en s, o he expe ise o ech-
nology p o essionals, o e alua e a co po a ion’s capaci y o inno a ion
(Jiang e al., 2018; Zhan e al., 2023), he e is a g owing end o use
g een pa en s as he p e e ed benchma k o quan i ying a company’s
inno a i e ou pu . Indeed, Fang e al. (2021) asse ed ha g een pa en s
can e ec i ely se e as a su oga e measu e o GTI, unde sco ing hei
undamen al ole in his pa icula con ex . The p e e ence o g een
pa en applica ions o e g een pa en g an s p ima ily a ises om he
empo al lag associa ed wi h he pa en -g an ing p ocess, which o en
ex ends om one o wo yea s ollowing he ini ial applica ion (H. Lin &
Long, 2021). Consequen ly, pa en applica ions p esen a mo e sui able
e lec ion o GTI ac i i ies.
Table 2 p o ides a comp ehensi e o e iew o he da a om 2002 o
2020. The analysis o he s anda d de ia ions conce ning he mean
ac oss a ious a iables sugges s ha he da a se con ained no ou lie s.
Fu he mo e, he consis en ly lowe s anda d de ia ions han he means
indica e ema kable s abili y and limi ed ola ili y wi hin he a iables
o he model o e he ex ensi e ime ame unde conside a ion. This
s udy employed he PSTR model, which inhe en ly allows o unbal-
anced panel da a. This enables he model o u ilize all a ailable obse -
a ions, e en i some a iables ha e ewe da a poin s han o he s. By
con as , o con en ional linea panel models—whe e a balanced da a
se is ypically assumed— he sample was es ic ed o hose obse a-
ions ha we e simul aneously a ailable ac oss all a iables. This e-
s ic ion, while necessa y o consis ency wi hin he linea amewo k,
na u ally led o a smalle numbe o obse a ions han in he PSTR
model.
Resul s
Con en ional panel models
This s udy c a ed es ima ions o he p oposed equa ions by using a
se ies o diagnos ic assessmen s aimed a de e mining he mos sui able
panel da a model. This s udy le e aged an ex ensi e a ay o panel da a
models and o mula ed Eq. (1), as shown in Table 3, which exclusi ely
inco po a es con ol models. The p ocess was ini ia ed by me iculously
examining he po en ial inclusion o bo h empo al and geog aphic ixed
e ec s. I en ailed a comp ehensi e compa ison be ween models
encompassing simul aneous ime and spa ial ixed e ec s and models
ea u ing dis inc ime and spa ial ixed e ec s. Two sepa a e likelihood
a io (LR) es s we e conduc ed o e alua e he es ima ed models, wi h
he associa ed p alues p o ided wi hin pa en heses. A low p alue
Table 2
S a is ics summa y (2002–2020).
Mean S d. De . Maximum Minimum Median Obse a ions
lnGTI 2.283 1.688 4.418 −2.996 2.889 360
lnREG 2.121 0.802 4.010 0.650 2.170 360
lnGDPP 10.403 0.616 11.375 8.817 10.529 360
lnURB 4.388 0.100 4.528 4.084 4.397 360
lnOPE 4.124 0.532 5.530 2.973 4.142 360
lnIS 1.564 0.339 2.180 0.361 1.554 360
lnFD −0.383 0.281 −0.003 −1.327 −0.305 360
lnFE 1.479 0.175 1.920 0.742 1.497 299
lnGS 1.461 0.195 1.844 0.824 1.470 299
lnTB 1.369 0.260 1.820 0.708 1.430 299
lnGP 1.525 0.181 1.828 0.863 1.541 299
lnBE 1.217 0.192 1.766 0.802 1.209 299
lnPE 1.547 0.122 1.869 1.147 1.535 299
lnRD 1.452 0.146 1.828 1.040 1.459 299
lnCP 1.655 0.129 1.949 1.172 1.673 299
lnMD 1.551 0.150 1.950 1.115 1.554 299
lnMO 1.499 0.135 1.828 1.131 1.504 299
lnPS 1.854 0.126 2.083 1.526 1.875 299
Table 3
Es ima ion o a ious panel models o con ol a iables.
Pooled
OLS
Spa ial
ixed
e ec s
Time ixed
e ec s
Spa ial
and ime
ixed
e ec s
Panel
EGLS
(C oss-
sec ion
andom
e ec s)
cons an −37.924 −25.880 −36.480 −18.806 −25.117
0.000 (0.000) (0.000) (0.000) (0.000)
lnGDPP 1.678 1.712 1.687 1.438 1.731
(0.000) (0.000) (0.000) (0.000) (0.000)
lnURB 6.048 1.578 5.780 0.908 1.689
(0.000) (0.047) (0.000) (0.284) (0.023)
lnOPE −0.401 0.545 −0.479 0.245 0.227
(0.000) (0.000) (0.000) (0.050) (0.038)
lnIS −1.422 0.962 −1.503 0.747 0.649
(0.000) (0.000) (0.000) (0.000) (0.000)
lnFD 2.873 1.138 2.813 0.418 0.592
(0.000) (0.000) (0.000) (0.012) (0.001)
LnREG 0.563 0.054 0.595 0.061 0.120
(0.000) (0.319) (0.000) (0.185) (0.007)
Log −lik −284.386 30.943 −268.862 103.889 473.420
R20.900 0.983 0.908 0.988 0.913
LR − es 145.892 745.502
(0.000) (0.000)
Hausman Tes 39.55
(0.000)
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
5
indica es he ejec ion o he null hypo hesis. As shown in Table 3, he
esul s o hese es s s ongly suppo he ejec ion o he null hypo hesis,
ad oca ing he inco po a ion o bo h ime and geog aphic ixed e ec s
in o he model.
Addi ionally, he ixed e ec s (FE) model and andom e ec s (RE)
model we e compa ed using he Hausman es . The Hausman es is a
aluable ool o dis inguishing be ween ixed e ec s and andom e ec s
models in panel analysis. The ixed e ec s model was selec ed unde he
al e na i e hypo hesis, while he andom e ec s model aligns wi h he
null hypo hesis. The indings o he Hausman es unequi ocally dismiss
he null hypo hesis, solidi ying he ixed e ec s model as he op imal
choice o he analysis.
Based on he indings, a 1 % inc ease in GDP g ow h pe capi a
signi ican ly leads o a 1.731 % inc ease in GTI. These a o able impac s
ex end o o he con ol a iables, al hough i is wo h no ing ha he
coe icien s o u baniza ion and en i onmen al egula ion a iables do
Table 4
Tes s o nonlinea i y.
Wald Tes s (LM) Fishe Tes s (LMF) LRT Tes s (LRT) ∗
Model B1 H0: =0 s H1: =1 13.077 (0.219) 1.251 (0.257) 13.321 (0.206) 1
H0: =1 s H1: =2 16.111 (0.097) 1.509 (0.135) 16.482 (0.087)
Model B2 H0: =0 s H1: =1 10.176 (0.601) 0.790 (0.661) 10.353 (0.585) 1
H0: =1 s H1: =2 20.227 (0.063) 1.554 (0.106) 20.944 (0.051)
Model B3 H0: =0 s H1: =1 10.082 (0.609) 0.782 (0.669) 10.255 (0.594) 1
H0: =1 s H1: =2 21.365 (0.045) 1.648 (0.079) 22.167 (0.036)
Model B4 H0: =0 s H1: =1 12.976 (0.371) 1.017 (0.433) 13.266 (0.350) 1
H0: =1 s H1: =2 17.368 (0.136) 1.321 (0.207) 17.892 (0.119)
Model B5 H0: =0 s H1: =1 9.353 (0.672) 0.724 (0.728) 9.503 (0.659) 1
H0: =1 s H1: =2 24.878 (0.015) 1.944 (0.030) 25.975 (0.011)
Model B6 H0: =0 s H1: =1 18.630 (0.098) 1.490 (0.128) 19.235 (0.083) 1
H0: =1 s H1: =2 16.908 (0.153) 1.284 (0.228) 17.405 (0.135)
Model B7 H0: =0 s H1: =1 13.038 (0.366) 1.018 (0.432) 13.347 (0.344) 1
H0: =1 s H1: =2 14.434 (0.274) 1.080 (0.378) 14.813 (0.252)
Model B8 H0: =0 s H1: =1 13.059 (0.365) 1.024 (0.427) 13.353 (0.344) 1
H0: =1 s H1: =2 5.838 (0.924) 0.426 (0.952) 5.895 (0.921)
Model B9 H0: =0 s H1: =1 13.382 (0.342) 1.050 (0.403) 13.690 (0.321) 1
H0: =1 s H1: =2 12.188 (0.431) 0.910 (0.537) 12.444 (0.411)
Model B10 H0: =0 s H1: =1 11.217 (0.510) 0.874 (0.574) 11.433 (0.492) 1
H0: =1 s H1: =2 16.458 (0.171) 1.248 (0.251) 16.928 (0.152)
Model B11 H0: =0 s H1: =1 12.711 (0.390) 0.995 (0.454) 12.989 (0.370) 1
H0: =1 s H1: =2 9.889 (0.626) 0.733 (0.719) 10.056 (0.611)
Model B12 H0: =0 s H1: =1 8.807 (0.719) 0.680 (0.770) 8.939 (0.708) 1
H0: =1 s H1: =2 14.359 (0.278) 1.080 (0.377) 14.715 (0.257)
Model B13 H0: =0 s H1: =1 12.650 (0.395) 0.990 (0.458) 12.926 (0.374) 1
H0: =1 s H1: =2 15.368 (0.222) 1.160 (0.312) 15.777 (0.202)
Table 5
Pa ame e es ima es o he PSTR models.
Model B1 Model B2 Model B3 Model B4 Model B5 Model B6
Pa ame e B10 1.109 1.062 2.320 1.203 2.293 1.091
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Pa ame e B20 1.330 1.648 2.275 1.810 2.573 1.419
(0.078) (0.054) (0.068) (0.061) (0.042) (0.136)
Pa ame e B30 1.178 1.046 −0.885 1.201 −0.817 1.162
(0.000) (0.000) (0.002) (0.000) (0.025) (0.000)
Pa ame e B40 0.941 0.931 1.545 0.917 1.462 0.953
(0.000) (0.000) (0.031) (0.000) (0.045) (0.000)
Pa ame e B50 1.115 1.034 1.447 1.002 1.784 1.029
(0.000) (0.000) (0.141) (0.001) (0.100) (0.001)
Pa ame e B60 −0.310 1.260 −0.213 0.571 0.041
0.002 (0.184) (0.183) (0.603) (0.805)
Pa ame e B11 0.808 0.814 −0.972 0.851 −0.855 0.791
(0.000) (0.001) (0.055) (0.001) (0.100) (0.002)
Pa ame e B21 −1.008 −1.285 1.550 −0.931 1.061 −0.881
(0.040) (0.024) (0.174) (0.095) (0.378) (0.115)
Pa ame e B31 −0.890 −0.785 1.514 −0.933 1.422 −0.818
(0.000) (0.001) (0.000) (0.000) (0.000) (0.001)
Pa ame e B41 0.032 0.138 −0.997 0.013 −0.886 0.021
(0.877) (0.536) (0.200) (0.955) (0.262) (0.927)
Pa ame e B51 0.361 0.524 −0.023 0.681 −0.454 0.423
(0.197) (0.113) (0.983) (0.057) (0.697) (0.194)
Pa ame e B61 0.397 −1.507 −0.350 −0.747 −0.593
(0.077) (0.134) (0.119) (0.522) (0.011)
Fi s T ansi ion Func ion [3.77, 1.83] [1.83, 3.77] [1.55, 1.55] [3.77, 1.83] [1.57, 1.57] [1.83, 3.77]
Slope Pa ame e y1 1242.976 1273.997 20.990 2428.660 23.674 2315.730
loca ion pa ame e s (m)m =2m =1m =2m =2m =2m =2
AIC o m =1−2.995 −3.116 −3.085 −3.134 −3.084 −3.083
Schwa z o m =1−2.790 −2.954 −2.848 −2.896 −2.933 −2.932
AIC o m =2−3.001 −3.123 −3.115 −3.123 −3.104 −3.116
Schwa z o m =2−2.861 −2.979 −2.954 −2.961 −2.942 −2.954
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
6
no exhibi s a is ical signi icance. No ably, a 1 % inc ease in ade
openness co esponds o a 0.245 % inc ease in GTI. A he same ime, he
inc ease in he indus ial s uc u e a iable, wi h a coe icien o 0.747,
ep esen s a c i ical impac on GTI. Mo eo e , a 1 % inc ease in inancial
de elopmen unde pins a 0.418 % ise in GTI. O pa icula signi icance
among hese a iables is he en i onmen al egula ions ac o (REG).
In e es ingly, i s e ec s p o e nonsigni ican in he spa ial ixed e -
ec s model, ye he es ima ed coe icien ca ies signi icance wi hin he
ime ixed e ec s model. Consequen ly, he gene al e ec s o he spa ial
and empo al ixed e ec s models appea inconclusi e. While he spa ial
ixed e ec s model add esses c oss-sec ional he e ogenei y, he ime
ixed e ec s model ocuses on he e ogenei ies o e ime. Thus, he e-
sul s indica e ha he a iable o en i onmen al egula ions may no
e ec i ely explain changes in GTI o e ime, e en hough i plays a
subs an ial ole in delinea ing di e ences in GTI among dis inc na ions.
As men ioned, he in ica e in e play o posi i e and nega i e aspec s
wi hin en i onmen al egula ions can be mul i ace ed. Thus, he lack o
s a is ical signi icance o hese e ec s could sugges a neu al balance
be ween hei di e se impac s. None heless, his e iew necessi a es a
mo e in-dep h analysis, which is p o ided la e .
PSTR esul s
This s udy in ol ed an in-dep h examina ion o he esul s by
applying a PSTR model. This analy ical p ocess encompasses a se ies o
pi o al s ages. Ini ially, i iden i ies he op imal numbe o loca ion
pa ame e s ep esen ed by m wi hin ansi ion unc ions, u ilizing bo h
he Schwa z and Akaike c i e ia. The ou come o his de e mina ion,
along wi h he ideal m alues, is p o ided in Table 5.
Subsequen ly, he s udy con as s he log-linea con igu a ion o he
GTI model wi h an al e na i e speci ica ion ha inco po a es h eshold
e ec s, aking in o accoun he p e iously es ablished m alue. In cases
whe e he log-linea assump ion alls sho o alidi y, he analysis
p oceeds o asce ain he numbe o ansi ion unc ions equi ed o
encapsula e nonlinea i y o he e ogenei y in GTI model pa ame e s. In
Table 6
Pa ame e es ima es o he PSTR models.
Model B7 Model B8 Model B9 Model B10 Model B11 Model B12
Pa ame e B10 0.957 1.434 1.146 1.044 1.243 1.094
(0.001) (0.000) (0.000) (0.000) (0.000) (0.000)
Pa ame e B20 1.091 1.867 1.774 1.469 1.770 1.438
(0.284) (0.033) (0.026) (0.072) (0.033) (0.098)
Pa ame e B30 1.184 1.059 1.111 1.199 1.135 1.165
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Pa ame e B40 0.878 0.978 1.029 0.909 0.973 0.957
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Pa ame e B50 1.138 1.190 0.988 1.020 0.927 1.052
(0.000) (0.000) (0.002) (0.002) (0.004) (0.000)
Pa ame e B60 0.203 −0.020 −0.082 0.208 −0.314 0.212
(0.278) (0.912) (0.800) (0.103) (0.186) (0.424)
Pa ame e B11 0.865 0.722 0.489 0.935 0.597 0.766
(0.001) (0.000) (0.065) (0.000) (0.022) (0.003)
Pa ame e B21 −0.838 −0.955 0.211 −1.000 −0.337 −0.656
(0.165) (0.007) (0.739) (0.071) (0.574) (0.304)
Pa ame e B31 −0.882 −0.648 −0.838 −0.954 −0.826 −0.887
(0.000) (0.001) (0.000) (0.000) (0.000) (0.000)
Pa ame e B41 0.038 −0.092 −0.448 0.086 −0.261 0.002
(0.870) (0.538) (0.075) (0.683) (0.285) (0.993)
Pa ame e B51 0.352 −0.192 0.514 0.489 0.526 0.430
(0.308) (0.572) (0.130) (0.194) (0.133) (0.184)
Pa ame e B61 −0.926 −0.499 −0.925 −0.746 −0.412 −0.581
(0.002) (0.087) (0.031) (0.001) (0.226) (0.169)
Fi s T ansi ion Func ion [3.77, 1.85] [2.11, 3.77] [1.91, 3.77] [3.77, 1.86] [1.92, 3.77] [1.83, 3.78]
Slope Pa ame e y1 1090.519 22.650 1165.309 15.066 1237.513 1257.877
loca ion pa ame e s (m)m =2m =2m =2m =2m =2m =2
AIC o m =1−3.191 −3.106 −3.121 −3.122 −3.131 −3.076
Schwa z o m =1−2.953 −2.869 −2.884 −2.884 −2.894 −2.925
AIC o m =2−3.195 −3.071 −3.131 −3.104 −3.126 −3.092
Schwa z o m =2−3.033 −2.909 −2.969 −2.942 −2.964 −2.930
Table 7
A e age es ima ed pa ame e s o indi idual PSTR o Model B1.
lnGDPP lnURB lnOPE lnIS lnFD
A gen ina 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
B azil 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
Chile 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
Finland 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
F ance 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
Ge many 1.271 1.129 1.000 0.948 1.187
(0.332) (0.414) (0.365) (0.013) (0.148)
G eece 1.299 1.094 0.969 0.949 1.200
(0.341) (0.425) (0.376) (0.013) (0.152)
I eland 1.271 1.129 1.000 0.948 1.187
(0.332) (0.414) (0.365) (0.013) (0.148)
Is ael 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
I aly 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
Japan 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
Ne he lands 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
No way 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
Spain 1.466 0.886 0.785 0.956 1.274
(0.406) (0.506) (0.447) (0.016) (0.181)
Sweden 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
Swi ze land 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
Uni ed Kingdom 1.109 1.330 1.178 0.941 1.115
(0.000) (0.000) (0.000) (0.000) (0.000)
Uni ed S a es 1.918 0.323 0.287 0.973 1.476
(0.000) (0.000) (0.000) (0.000) (0.000)
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
7
his endea o , he wo k o Gonz´
alez e al., (2004) plays a pi o al ole by
o e ing a obus es ing me hodology o e alua ing linea i y and he
numbe o ansi ion unc ions ( e e ed o as ∗). The es ing p ocess
in ol es a con inuous compa ison be ween he null hypo hesis (H0:
= ∗) and he al e na i e hypo hesis (H1: = ∗+1), inc emen ally
inc easing ∗un il he null hypo hesis gains accep ance. The analysis is
con ined o he PSTR models wi h a maximum o ou ansi ion unc-
ions. The ou comes o hese es ing p ocedu es a e documen ed in
Table 4, deli e ing a comp ehensi e e alua ion o he model cha ac-
e is ics and esul s. Mo eo e , he s udy employed a a ie y o diag-
nos ic es s, including Wald es s, Fishe es s, and likelihood a io es s,
o o e a ho ough assessmen o he model a ibu es and ou comes.
This ensu es a comp ehensi e and nuanced e alua ion o he PSTR
model.
Wi hin he amewo k o a PSTR model, nonlinea cha ac e is ics a e
e ec i ely ep esen ed by a cons ained se o excep ional pa e ns.
These pa e ns co espond o a ia ions in pa ame e s o independen
a iables obse ed ac oss di e en coun ies and o e ime. Acco ding
o Table 4, he en i onmen al egula ions ac o exhibi s a conside able
capaci y o explain hese a ia ions. The assessmen o he absence o
pe sis en nonlinea i y is e lec ed in speci ica ions p ima ily inco po-
a ing single ansi ion unc ions o all models in Table 4. The pa am-
e e es ima es o he inal PSTR models a e epo ed in Tables 5 and 6.
The ou comes o he es ima ion unde sco e he signi icance o he ma-
jo i y o he es ima ed coe icien s ac oss a ious models. This sugges s
ha he nonlinea analysis o GTI emphasizes he impo ance o all
con ol a iables employed in his esea ch.
D awing upon he pa ame e es ima es ob ained om he PSTR
models, i becomes iable o measu e he e ol ing impac s o indepen-
den a iables ac oss di e en ime poin s o e e y coun y wi hin he
sample. These dynamic e ec s a e encapsula ed in Eq. (5). I is c ucial o
emphasize ha he in e ed pa ame e s p o ide indi ec in e p e abili y
p ima ily h ough hei signs. The implica ions o hese pa ame e signs,
pa icula ly wi hin di e se con ex ual amewo ks, p o ide pa ial in-
sigh s. Howe e , a comp ehensi e g asp o he in ica e nonlinea e ec s
o independen esea ch a iables on GTI necessi a es he u iliza ion o
Eq. (5). I enables he de i a ion o weigh ed coe icien s o each a -
iable. The esul s o es ima ing Eq. (5) using he coe icien s de i ed
om Model B1 in Table 5 a e as ollows. No ably, hese calcula ions
should also be ex ended o he o he models. This ho ough explo a ion
is i al o obus ly comp ehending he complex ela ionships be ween
he speci ied a iables and hei impac s on GTI emissions unde a ying
en i onmen al egula ions.
∂
GTIi
∂
lnGDPPi
=1.109 +0.808 ×hj(REGi ;1242.976,[3.77,1.83]) (8)
∂
GTIi
∂
lnURBi
=1.330 −1.008 ×hj(REGi ;1242.976,[3.77,1.83]) (9)
∂
GTIi
∂
lnOPEi
=1.178 −0.890 ×hj(REGi ;1242.976,[3.77,1.83]) (10)
∂
GTIi
∂
lnISi
=0.941 +0.032 ×hj(REGi ;1242.976,[3.77,1.83]) (11)
∂
GTIi
∂
lnFDi
=1.115 +0.361 ×hj(REGi ;1242.976,[3.77,1.83]) (12)
The indings allow o a mo e e ec i e unde s anding o he di ec
impac s by examining β0 in Eqs. (4) and 5. Fu he mo e, Eq. (5) p o ides
insigh in o he indi ec e ec s, exp essed as he sum o hj(qi ;γj,cj). Fo
ins ance, in Eq. (8), he di ec e ec s a e quan i ied a 1.109, while he
associa ed indi ec e ec s a e 0.808 ×hj(REGi ;1242.976,[3.77,1.83]).
The dynamic na u e o he ansi ion unc ion hj(REGi ;1242.976,
[3.77,1.83]) is no ewo hy, which inhe en ly in oduces empo al and
c oss-na ional a ia ions in o hese indi ec e ec s.
Sc u inizing he s a is ical signi icance o he coe icien s indica es
ha bo h di ec and indi ec e ec s in Model B1 a e signi ican ac oss all
a iables, excep o he indi ec e ec s o lnIS and lnFD. Gi en he
complexi y o de ailing he p ojec ed e ec s encompassing Eqs. (8) o 12
o e e y indi idual model, i is necessa y o p o ide a concise summa y
o he es ima ion esul s. These a e aged ou comes, as indica ed in Eqs.
(8) o 12, a e me iculously p esen ed in Table 7. I is wo h highligh ing
ha hese alues ep esen he coun y a e age o indi idual e ec s o
hese a iables. The alues enclosed in pa en heses signi y he s anda d
de ia ion cha ac e izing he es ima ed coe icien s wi hin each na ion.
The consis en ly low s anda d de ia ions dis inc ly unde sco e he
obus ness and s abili y o he coe icien s a he na ional le el.
Despi e o e ing aluable insigh s in o how model a iables impac
GTI on a na ional le el, Table 7 does no en i ely enable a comp ehen-
si e unde s anding o he complex nonlinea e ec s o hese a iables on
GTI, no does i e eal he unde lying explana ions o he changes in he
es ima ed pa ame e s ac oss coun ies and o e ime, pa icula ly in
esponse o a ious en i onmen al egula ion egimes. To explo e hese
unique e ec s, i is c ucial o isually ep esen he es ima ed co-
e icien s conce ning di e en le els o he na u al loga i hm o en i-
onmen al egula ion (lnREG). Mean coe icien s a e calcula ed on
mul iple scales. A a na ional le el, as shown in Fig. 1, hese coe icien s
ep esen a e age pa ame e s ac oss a ious ime ames wi hin speci ic
coun ies. Addi ionally, on a empo al scale, as isualized in Fig. 2, hey
e lec a e age pa ame e s ac oss di e en coun ies a pa icula poin s
in ime.
The e ical axis wi hin he g aphical ep esen a ions illus a es he
mean alues o he es ima ed pa ame e s o each model a iable, while
he ho izon al axis ep esen s he a e age lnREG le els ac oss di e en
coun ies. No ably, he a iables demons a e consis en posi i e o
nega i e e ec s in bo h diag ams. Fu he mo e, Figs. 1 and 2 i idly
show ha he a e age es ima ed coe icien s, bo h a he empo al and
na ional le els, shi simila ly as he lnREG le els inc ease. The only
excep ion is he educed dispe sion seen in he empo al a e age. This
di e ence in a iance could be a ibu ed o he ela i ely sho s udy
pe iod, limi ing signi ican empo al a ia ion. None heless, hese dis-
pa i ies a e less p onounced in compa ison o he dis inc ions in he
es ima ed coe icien s o he linea spa ial and ime e ec s models, as
shown in Table 5, pa icula ly in e ms o he signs o he coe icien s.
This sugges s ha he use o he nonlinea o mula ion e ec i ely
add essed such dispa i ies.
As shown in Fig. 1, all con ol a iables wi hin he model exhibi ed
posi i e impac s on GTI. Howe e , an escala ion in en i onmen al eg-
ula ions appea s o diminish he posi i e impac o bo h GDP pe capi a
and indus ial s uc u e. Con e sely, a highe deg ee o en i onmen al
egula ion co ela es wi h ampli ying he impac s o u baniza ion and
ade openness. These indings we e de i ed om Model 1. The
emaining igu es pe ain o he analysis o 11 en ep eneu ship in-
dica o s, each co esponding o he es ima ed coe icien s om Models
2–12. The esul s e eal ha he majo i y o hese en ep eneu ship
indica o s exe a nega i e impac on GTI. In e es ingly, an inc ease in
en i onmen al egula ions g adually mi iga es he ad e se impac s and
e en ans o ms se e al indica o s in o posi i e con ibu o s. No ably,
pos -school en ep eneu ial educa ion and aining, basic school en e-
p eneu ial educa ion and aining, in e nal ma ke dynamics, and
physical and se ices in as uc u e all in o his ca ego y.
Among he 11 indica o s, inancing o en ep eneu s, go e nmen al
p og ams, and go e nmen al suppo and policies exhibi dis inc i e
impac pa e ns. These h ee a iables ini ially ha e nega i e impac s on
GTI a e y low le els o en i onmen al egula ion. Howe e , as en i-
onmen al egula ions in ensi y, hese nega i e impac s g adually
become posi i e. A en i onmen al egula ion le els a ound 1.69 o
go e nmen al suppo and policies and go e nmen al p og ams and
1.49 o inancing o en ep eneu s, posi i e impac s peak, while a
u he inc ease in en i onmen al egula ions diminishes he posi i e
impac s, ul ima ely making hem nega i e. The esul s unde sco e a
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
8
o Xi’an. Sus ainable Ci ies and Socie y, 57. h ps://doi.o g/10.1016/j.
scs.2020.102123
Zhang, W., Li, G., & Guo, F. (2022). Does ca bon emissions ading p omo e g een
echnology inno a ion in China? Applied Ene gy, 315. h ps://doi.o g/10.1016/j.
apene gy.2022.119012
Zhang, X., & Dong, F. (2022). De e minan s and egional con ibu ions o indus ial CO2
emissions inequali y: A consump ion-based pe spec i e. Sus ainable Ene gy
Technologies and Assessmen s, 52. h ps://doi.o g/10.1016/j.se a.2022.102270
Zhang, Y., Sun, J., Yang, Z., & Wang, Y. (2020b). C i ical success ac o s o g een
inno a ion: Technology, o ganiza ion and en i onmen eadiness. Jou nal o Cleane
P oduc ion, 264. h ps://doi.o g/10.1016/j.jclep o.2020.121701
Zhao, G., Zhou, P., & Wen, W. (2022a). Wha cause egional inequali y o echnology
inno a ion in enewable ene gy? E idence om China. Applied Ene gy, 310. h ps://
doi.o g/10.1016/j.apene gy.2021.118464
Zhao, L., Zhang, L., Sun, J., & He, P. (2022b). Can public pa icipa ion cons ain s
p omo e g een echnological inno a ion o Chinese en e p ises? The mode a ing ole
o go e nmen en i onmen al egula o y en o cemen . Technological Fo ecas ing and
Social Change, 174. h ps://doi.o g/10.1016/j. ech o e.2021.121198
Zhu, J., Fan, Y., Deng, X., & Xue, L. (2019). Low-ca bon inno a ion induced by emissions
ading in China. Na u e Communica ions, 10(1). h ps://doi.o g/10.1038/s41467-
019-12213-6
M. Khez i
Jou nal o Inno a ion & Knowledge 10 (2025) 100738
15
