Li, Xiang; Shen, Guihua; Zhang, Xiuwu
A icle
Can inno a i e ci y pilo policy educe income inequali y?
Jou nal o Inno a ion & Knowledge (JIK)
P o ided in Coope a ion wi h:
Else ie
Sugges ed Ci a ion: Li, Xiang; Shen, Guihua; Zhang, Xiuwu (2025) : Can inno a i e ci y pilo policy
educe income inequali y?, Jou nal o Inno a ion & Knowledge (JIK), ISSN 2444-569X, Else ie ,
Ams e dam, Vol. 10, Iss. 4, pp. 1-14,
h ps://doi.o g/10.1016/j.jik.2025.100749
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/327644
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-nc-nd/4.0/
Can inno a i e ci y pilo policy educe income inequali y?
Xiang Li
a,*
, Guihua Shen
b
, Xiuwu Zhang
c
a
School o Economics and Finance, Huaqiao Uni e si y, Quanzhou 362021, China
b
Jiangxi Voca ional Educa ion and Indus ial Resea ch Ins i u e, Jiangxi Science & Technology No mal Uni e si y, Nanchang 330000, China
c
Resea ch Cen e o Quan i a i e Economics, Huaqiao Uni e si y, Xiamen 361021, China
ARTICLE INFO
JEL classi ica ion:
C33
D31
Keywo ds:
Inno a i e ci y pilo policy
Income inequali y
Mul i-pe iod di e ence-in-di e ences model
ABSTRACT
Ensu ing ai ness in income dis ibu ion is a undamen al equi emen in achie ing common p ospe i y. This
s udy employed panel da a sou ced om 276 ci ies h oughou China, co e ing he ime ame om 2003 o
2022. I se s up a mul i-s age di e ence-in-di e ences (DID) model o explo e he impac s o he inno a i e ci y
pilo policy on income inequali y. Resul s indica e ha his policy signi ican ly educes income inequali y, a
conclusion ha emains obus ac oss a ious es s. Meanwhile, he e ec s a e mo e p onounced in cen al and
wes e n egions, non-bo de ci ies, Yang ze Ri e Economic Bel ci ies, non-old indus ial bases, e hnic-mino i y
a eas, low adminis a i e ci ies, low ini ial income inequali y and non- esou ce-based ci ies. The mechanism
analysis indica es ha hese policies mi iga e income inequali y la gely by os e ing labou esou ce agglom-
e a ion, s uc u e op imisa ion and inno a i e i ali y. Th ough an analysis o he impac s o inno a i e ci y pilo
ini ia i es, his s udy en iches ou comp ehension and p o ides signi ican pe spec i es o p omo ing income
equali y in he new epoch. In addi ion, i p o ides s a egic guidance o expanding and scaling hese pilo
policies o b oade con ex s.
In oduc ion
The apid p og ess o global economic in eg a ion and echnological
ad ancemen s has made income inequali y a majo challenge. I poses
no able isks o social cohesion and sus ainable economic de elopmen .
In China, he income gap has been u he exace ba ed by he u ban-
– u al dual s uc u e, cons aining esou ce e iciency and unde mining
social equi y. The g owing dispa i y in ensi ies social ensions and
h ea ens s abili y and socie al well-being. The e o e, de eloping
e ec i e policy measu es o add ess income inequali y has become a op
p io i y o policymake s and esea che s globally.
The pilo p og amme o inno a i e ci ies is a pi o al s a egy o
p omo e he de elopmen o an inno a ion-o ien ed na ion. I plays a
key ole in guiding he economic ansi ion o China om a phase o
ac o -d i en expansion o one cha ac e ised by inno a ion-led de el-
opmen . This policy aimed o s eng hen he capaci y o he ci ies o
independen inno a ion, op imise indus ial s uc u es and cul i a e an
inno a ion-conduci e en i onmen . I s goal was o d i e sus ainable,
high-quali y g ow h. Since he pilo wo k o building inno a i e ci ies
was launched in 2010, some ci ies ha e been included in he scope o he
pilo p og amme. In he de elopmen o inno a i e ci ies, local
go e nmen s ha e aken a leading ole. They ac i ely ha ness he
decisi e unc ion o ma ke mechanisms in alloca ing inno a ion e-
sou ces. This has esul ed in a dis inc i e model de ined by he syne gy
be ween go e nmen and ma ke o ces. The e is a syne gy be ween
go e nmen leade ship and ma ke dynamics. This syne gy has played a
i al ole in educing he isks linked o co po a e esea ch and de el-
opmen . Fo ins ance, he policy has implemen ed a ious measu es,
such as a ac ing high-calib e alen , inc easing inancial in es men s,
s eng hening in ellec ual p ope y p o ec ions and building inno a ion
in as uc u e. These measu es ha e e ec i ely imp o ed he inno a ion
ecosys em. These e o s ha e signi ican ly enhanced ci ies’ inno a ion
capaci y and a ac ed clus e s o high-end indus ies (Be one e al.,
2013). Howe e , whe he hese bene i s a e dis ibu ed equi ably
among all social g oups emains unclea , pa icula ly among u al and
low-income popula ions. Inno a ion ac i i ies can boos he demand o
highly skilled labou and inc ease hei income le els. Howe e , u al
and low-income g oups may be le behind i inno a ion emains
concen a ed in u ban cen es. Such exclusion isks u he exace ba e
he income gap. Consequen ly, how inno a i e ci y pilo policies a ec
income inequali y and hei e ec i eness emain i al conce ns ha
me i mo e ex ensi e esea ch.
* Co esponding au ho .
E-mail add ess: [email p o ec ed] (X. Li).
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.2025.100749
Recei ed 24 Feb ua y 2025; Accep ed 31 May 2025
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
A ailable online 18 June 2025
2444-569X/© 2025 The Au ho s. 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-NC-ND license (
h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/ ).
This s udy employed panel da a sou ced om 276 ci ies h oughou
China om 2003 o 2022. This amewo k es ablishes a mul i-s age
di e ence-in-di e ences (DID) amewo k. The aim o his s udy was
o explo e he impac s o he inno a i e u ban pilo policy on he income
inequali y si ua ion. The ou comes indica e ha he en o cemen o he
pilo policy signi ican ly educes income inequali y. This conclusion
holds ue a e mul iple obus ness checks. No ably, he e ec s o his
policy mani es mo e p ominen ly in speci ic geog aphical and admin-
is a i e con ex s. These include he cen al-wes e n egions, non-bo de
ci ies, ci ies posi ioned wi hin he ambi o he Yang ze Ri e Economic
Bel , ci ies wi h lowe adminis a i e hie a chies, locales ha ini ially
display a lowe le el o income dispa i y and non- esou ce-dependen
ci ies. Mechanism analysis indica es ha he policy p omo es ai in-
come dis ibu ion p ima ily h ough labou esou ce agglome a ion,
s uc u e op imisa ion and inno a i e i ali y.
The inno a ions o his s udy a e as ollows. Fi s , his s udy analysed
he impac o inno a ion policies on income dis ibu ion. I ills he gap
in social e ec analysis om he pe spec i e o inno a i e de elopmen .
In addi ion, i ound ha inno a i e u ban de elopmen helps na ow
income gaps and signi ican ly educes income inequali y. Mo eo e , his
s udy in es iga ed he dynamic e olu ion o such in luence, conside ing
policy ma ginal e ec s and ex e nal shocks. I p o ides decision-making
guidance o expanding he scope o inno a i e u ban pilo p og ammes
and also o e s a new pe spec i e on achie ing common p ospe i y.
Second, ega ding esea ch con en , his s udy dynamically demon-
s a es he long- e m impac o inno a ion policies on income inequali y.
In his p ocess, heo y and empi ical e idence a e e ec i ely in eg a ed
o conduc a de ailed demons a ion. By elabo a ing on he ins i u ional
backg ound, his s udy becomes mo e speci ic and eliable. In addi ion,
his s udy explo es mul iple in luence pa hs in dep h, highligh ing h ee
key pa hways: labou esou ce agglome a ion, s uc u e op imisa ion
and enhanced inno a i e i ali y. Mo eo e , his s udy add esses he
limi a ions o ex an esea ch on income dis ibu ion. In pa icula ,
p e ious s udies o en s uggle o isola e income inequali y om b oade
concep s such as common p ospe i y and end o o e only shallow
analyses o he unde lying mechanisms. This s udy cla i ies how he
policies mi iga e income inequali y and p o ides heo e ical suppo o
designing u he measu es o educe income gaps. Thi d, in e ms o
esea ch me hods, his s udy employed a mul i-s age DID model, which
is a igo ous and sophis ica ed econome ic app oach. Robus ness was
e i ied using p opensi y sco e ma ching DID (PSM-DID) and ins u-
men al a iable me hods, e ec i ely minimising es ima ion bias. By
analysing he e ogenei y, his s udy explo ed he complica ed implica-
ions o inno a ion policies on income inequali y. This enhances he
unde s anding o he social e ec s o inno a i e u ban de elopmen .
Li e a u e e iew
Inno a ion and income inequali y
Academia has ex ensi ely deba ed he impac o inno a ion on in-
come gaps, yielding wo con as ing pe spec i es. Some s udies sugges
ha echnological inno a ion ends o widen income dispa i ies. In
pa icula , income g ow h among he weal hies g oups is mo e p o-
nounced (Aghion e al., 2019). This e ec is p ima ily a ibu ed o he
unequal dis ibu ion o inno a ion e u ns, inc easing labou income
inequali y (Pe mana e al., 2018). In pa icula , skill-o ien ed echno-
logical ad ancemen s ha e spu ed a subs an ial inc ease in he demand
o highly skilled labou . Consequen ly, his has caused a signi ican
ele a ion in skill- ela ed p emiums and income inequali y (Acemoglu &
Res epo, 2018, 2019). In addi ion, echnological p og ess in China has
exhibi ed a no able skill bias. This skill bias is he p ima y cause o
widening egional wage dispa i ies (Wang e al., 2022). Mo eo e ,
echnological inno a ion shi s skill dis ibu ion in he labou ma ke
h ough a ‘sc eening e ec ’. This e ec bene i s high-skilled wo ke s;
howe e , i may sideline low-skilled wo ke s (Lee & Pose, 2013;
Michaels e al., 2014). This, in u n, in ensi ies income dispa i ies.
Va iances in subs i u ion elas ici y among echnological ad ancemen s
and di e se asks hold signi ican sway o e income inequali y.
High-skilled labou is ypically complemen a y o echnology, whe eas
low-skilled labou is ulne able o displacemen (Yu e al., 2021).
Some esea che s con end ha echnological inno a ion can educe
income dispa i ies ia mechanisms such as ‘knowledge spillo e s’ and
‘capi al conse a ion’. The ‘lea ning-by-doing’ phenomenon ela ed o
echnological inno a ion allows low-skilled wo ke s o upg ade hei
skills by lea ning, which os e s knowledge spillo e . These knowledge
spillo e s subsequen ly esul in a na owing o wage di e ences.
Technological inno a ion, cha ac e ised by capi al conse a ion, may
alle ia e income inequali y indi ec ly by educing en al expendi u es
(An onelli & Geh inge , 2017). Neu al echnological ad ancemen s can
p omo e g ow h in he a ailabili y o a skilled wo k o ce. The e o e, he
wage gap sepa a ing skilled om unskilled labou e s can be na owed
(Dong e al., 2014; Liu & Zhang 2017). Mo eo e , he combined in lu-
ence o u banisa ion and echnological inno a ion can mi iga e income
inequali y among esiden s (Zhao e al., 2018). Gene ally, wo ke s wi h
high human capi al a e di ec ly in ol ed in inno a ion and ob ain high
e u ns h ough esea ch o echnological complemen a i y (Aghion
e al., 2019). By con as , hose wi h low human capi al indi ec ly
pa icipa e in inno a ion. I wo ke s wi h low human capi al success-
ully acqui e new echnologies, hei income may ise. Howe e , i hey
ail o do so, hey may ace a isk o dec eased income o e en ma -
ginalisa ion. Consequen ly, as echnology di usion and skill upg ading
p oceed, he ad e se in luence o inno a ion on income inequali y may
p og essi ely weaken (Yan e al., 2023).
Economic e ec o inno a ion policies
Ci ies play a c ucial unc ion as he main d i e s o economic g ow h
in China, pa icula ly in e ms o p omo ing inno a ion (Da is & Dingel,
2019). Majo u ban agglome a ions in China hos ~90% o inno a ion
ac i i ies wi hin ~20% o i s land a ea (Zhou e al., 2021). Cu en ly, he
majo i y o esea ch e o s ha e been mainly ocused on economic
impac s. These impac s a e engende ed by execu ing he inno a ion
pilo ci y policy and a e one sided. Fo ins ance, esea ch has highligh ed
he in luence o os e ing he syne gy be ween pollu ion aba emen and
ca bon educ ion in u ban se ings. Inno a ion-d i en policies can p o-
mo e he coo dina ed imp o emen o pollu ion educ ion and ca bon
educ ion in ci ies (Yang & Xue, 2024). In addi ion, hese policies can
enhance u ban g een ecological e iciency. Meanwhile, in en ep e-
neu ial i ali y s udies, inno a i e ci y pilo policies s ongly p omo e
u ban g een en ep eneu ship (Yang & Liu, 2024). The impac o pilo
policies on u ban inno a ion ollows an asymme ic in e ed-V end,
ising hen alling (Li & Yang, 2019) and enhancing indus ial e iciency
and s uc u e ia ac o and echnology agglome a ion (Hu e al., 2020).
Howe e , he cons uc ion o inno a ion-o ien ed ci ies causes
changes in economic e ec s. In addi ion, i igge s al e a ions in social
e ec s, such as he pa e n o income dis ibu ion. Howe e , ex an
esea ch on his issue is compa a i ely limi ed. Mos s udies on he
impac o inno a ion in China on income inequali y ha e indi ec ly
examined i h ough ce ain angles such as echnological change. They
o e look he social impac s o he pilo policy, pa icula ly income
inequali y caused by changes in he inno a ion policy en i onmen .
Social e ec s o inno a ion policies
Li e a u e on he ela ionship be ween inno a ion policies and in-
come dis ibu ion is sca ce. Fo ins ance, Yang and Li (2023) examined
he impac o inno a ion policies on common p ospe i y. Thei s udy
cons uc ed a comp ehensi e index sys em o his pu pose. Howe e ,
he indings hea ily depend on index measu emen s wi hou a uni ied
s anda d. These esul s me ely e lec he o e all mac o-si ua ion and do
no ho oughly explo e income inequali y. Common p ospe i y, in
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
2
essence, is a comp ehensi e de elopmen objec i e, embodying ‘sha ed
de elopmen ’ and ‘equi able dis ibu ion o ou comes’. The da a and
me hodology issues a ise because esea ch using ‘common p ospe i y’
indica o s may no ully cap u e income inequali y. This cha ac e is ic
makes i impossible o sepa a e income dis ibu ions om he esea ch
amewo k. The e o e, i p ecludes in-dep h esea ch on income
inequali y. Al hough he e is a he e ogenei y analysis, an analysis o
in e nal di e ences among di e en egions and ci y ypes is no
de ailed enough. To add ess hese esea ch sho comings, his s udy
ocused on income inequali y. I educes biases using accu a e indica o
da a, applying ins umen al a iables and conduc ing mul iple obus -
ness es s. In addi ion, i s eng hens he e ogenei y analysis o enhance
he gene alisabili y o he esul s. Meanwhile, Xu and Zeng (2024)
s udied he impac o inno a i e ci y pilo policies on he income gap.
Howe e , by ocusing on only one mechanism, hei analysis is oo su-
pe icial, yielding in easible coun e measu es and sugges ions. This
s udy con ines i sel o he sho - e m ou comes o policies, neglec ing
he long- e m impac s o policy implemen a ion om a dynamic
pe spec i e. This s udy explo ed he impac o inno a ion policies on
income inequali y. Mo eo e , a mo e comp ehensi e analysis was con-
duc ed in h ee dimensions: labou agglome a ion, s uc u al op imisa-
ion and inno a ion i ali y.
Ins i u ional backg ound and heo e ical analysis
Ins i u ional backg ound
(1) The Inno a i e Ci y Pilo Policy
The inno a i e u ban pilo policy is a g adual e o m unde he
inno a ion-d i en s a egy in China. I ex ends inno a ion e o s om
indi idual ac o s o he ci y le el and in eg a es inno a ion ac i i ies
in o u ban go e nance. Ci ies a e he key spa ial pla o m o imple-
men ing his s a egy. The pilo policy is a unique Chinese policy ool. I s
di usion p inciple lies in he cen al go e nmen g an ing local go -
e nmen s ins i u ional space o ‘ea ly expe imen a ion’. This app oach
encou ages local policy inno a ion and selec s success ul p ac ices o
wide adop ion. In 2005, he S a e Council eleased a Na ional Medium-
and Long- e m Science and Technology De elopmen Plan
(2006–2020). This plan se s he s a egic goal o building an inno a ion-
d i en na ion. In 2008, Shenzhen was selec ed as he i s inno a i e ci y
pilo , ma king he o icial launch o an inno a i e u ban de elopmen
ini ia i e. In 2010, he Na ional De elopmen and Re o m Commission
(NDRC) and he Minis y o Science and Technology (MOST) join ly
app o ed 44 ci ies and dis ic s as inno a i e ci y pilo s. This app o al
u he ad anced he s a egy o building an inno a ion-d i en na ion.
By 2016, he NDRC and MOST consolida ed ea lie pilo p o-
g ammes, es ablishing 61 inno a i e ci y pilo s. This numbe g ew o 78
by 2018, encompassing na ional-le el inno a i e ci y pilo p og ammes
in ci ies and dis ic s. By 2022, MOST suppo ed an addi ional 25 ci ies,
including Baoding, o implemen inno a i e u ban de elopmen . The
o al numbe o inno a i e ci y pilo s app o ed by he wo minis ies
was 103. Among hem, he e we e 97 p e ec u e-le el ci ies, 4 dis ic s
in municipali ies di ec ly unde he cen al go e nmen and 2 coun y-
le el ci ies.
(2) Household Regis a ion Sys em and Income Gap
Du ing he economic ans o ma ion in China, ins i u ional b eak-
h oughs and policy inno a ion a e c ucial o ealising common p os-
pe i y. As he co ne s one o he u ban– u al dual sys em, he household
egis a ion sys em has long es ic ed labou mobili y. The adi ional
household egis a ion sys em c ea es mul iple ba ie s o u ban
employmen , social secu i y and public se ices o u al esiden s.
Consequen ly, his si ua ion educes he spa ial alloca ion e iciency o
he labou o ce. Despi e e o m- elaxed household egis a ion ules,
slow u al labou u banisa ion hinde s u ban– u al income gap educ-
ion. Fo ins ance, despi e he in lux o mig an wo ke s in o ci ies,
household egis a ion es ic s hei equal access o u ban educa ion
and heal hca e. This hinde s amily-le el mig a ion, unde mining long-
e m labou supply s abili y and impeding he na u al na owing o he
u ban– u al income gap h ough u banisa ion. Meanwhile, he Chinese
go e nmen -led inno a ion policies ha e deeply in e ened in he
ma ke mechanism h ough ins i u ional design and esou ce alloca ion.
The e o e, hese policies eshape he egional economic de elopmen
landscape. The go e nmen elies on policy ools such as inno a i e ci y
pilo s and de elopmen zone cons uc ion o b eak adminis a i e ba -
ie s and guide ac o s in ga he ing in key a eas. Th ough hese e o s, a
‘policy-d i en’ de elopmen model is o med (Hu e al., 2020). This
op–down policy in e en ion no only add esses ma ke ailu es bu
also eleases e o m di idends h ough ins i u ional inno a ion. This
app oach p o ides new pa hs o na owing income gaps.
Theo e ical analysis o ma ginal e ec s
P omo ing income dis ibu ion h ough echnological inno a ion is
an in eg al pa o achie ing common p ospe i y. The co e o income
equi y is ensu ing ha de elopmen bene i s a e ai ly and easonably
sha ed ac oss all social s a a. Mo eo e , i ocuses on os e ing eco-
nomic g ow h and inno a ion e iciency. Ins i u ional inno a ion and
policy op imisa ion enhance he inclusi eness and sha ing o socio-
economic de elopmen . The e o e, hey na ow income gaps and p o-
mo e common p ospe i y.
The hypo hesis is ha he economic wo k o ce is di ided in o highly
skilled (H) and low-skilled (L) wo ke s. Inno a ion policies ha p omo e
echnological p og ess (A) imp o e he p oduc i i y o low-skilled
wo ke s and hus dec ease he income gap. The p oduc ion unc ion
can be calcula ed as ollows:
Y=A(H
α
L1−
α
),0<
α
<1 (1)
whe e A is he g ow h in o al ac o p oduc i i y gene a ed by inno-
a i e policies and
α
is he ou pu elas ici y o highly skilled labou .
The incomes o he high-skilled and he low-skilled labou o ce can
be calcula ed as ollows:
WH=
α
⋅Y
H,WL= (1−
α
)⋅Y
L(2)
The indica o o income inequali y is as ollows:
G=WH
WL
=
α
1−
α
⋅L
H(3)
Assume ha inno a ion policy inpu I dec eases G h ough an in-
c ease in A. Fo ins ance, policies p o ide subsidies o low-skilled
wo ke s’ skills aining o imp o e p oduc i i y. Wi h an inc ease in I,
he g ow h o A declines, meaning ha
∂
A
∂
I>0,
∂
2A
∂
I2<0 (4)
This esul s in a educed pace o na owing he income gap G, i.e.
∂
G
∂
I<0,
∂
2G
∂
I2>0 (5)
Gi en he abo e analysis, we can in e ha he impac o inno a ion
policies on na owing income inequali y declines as he inpu o inno-
a ion policies inc eases.
Theo e ical mechanism analysis
The mechanism h ough which inno a ion-o ien ed policies a ec
income inequali y is p ima ily mani es ed in h ee espec s: labou
esou ce agglome a ion, s uc u e op imisa ion and s imula ed
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
3
inno a i e i ali y.
(1) The Agg ega ion Mechanism o Labou Resou ces
The pilo policy o inno a i e ci ies p omo es income equalisa ion
h ough he agglome a ion o labou esou ces. This p ocess is a c ucial
channel in educing income dispa i ies. The Chinese household egis-
a ion sys em causes he ‘semi-u banisa ion’ o he labou o ce.
Mig an wo ke s canno ully enjoy u ban esiden bene i s, a ec ing
amily mig a ion and labou o ce s abili y. F om he pe spec i e o he
spa ial agglome a ion o he popula ion, u banisa ion op imises he
u ban– u al spa ial layou , b eaks ba ie s, p omo es esou ce low and
deepens in eg a ion (Po no & Schwa z, 2009). This se ies o e ec s o
u banisa ion is undamen al o na owing he income dispa i y.
The pilo policy o inno a i e ci ies has been b oken down in o wo
aspec s. Fi s , inno a i e policies should upg ade in as uc u e and
p omo e public se ice equalisa ion (Zhao e al., 2023). This includes
building a o dable housing and imp o ing educa ion policies o
mig an wo ke s’ child en. These ac ions educe he cos o u al labou
mig a ion and weaken he cons ain s o he household egis a ion
sys em. Second, ‘ alen policies’ se e as a b eak h ough. Housing sub-
sidies, en ep eneu ship suppo and o he measu es a e implemen ed o
a ac highly skilled alen s. These ini ia i es os e an agglome a ion
e ec in which alen s a ac mo e alen s. This gene a es economies o
scale and knowledge spillo e . The e o e, mo e jobs a e c ea ed, and
wo ke s ea n mo e, he eby imp o ing income equali y (Dougal e al.,
2015; Shen e al., 2019). Fo ins ance, h ough alen policies, inno a-
i e pilo ci ies such as Shenzhen and Hangzhou ha e a ac ed high-end
elemen s, subs an ially educing he income gap wi h adi ional in-
dus ial ci ies (Sun e al., 2022).
(2) S uc u al Op imisa ion Mechanism
The ins i u ional en i onmen plays a shaping ole in egional
de elopmen and indus ial upg ading s a egies in China. Inno a i e
pilo ci ies can le e age echnological b eak h oughs o p omo e
balanced indus ial g ow h, d i e equipmen upg ades and boos e i-
ciency. In addi ion, hese ci ies can use such b eak h oughs o os e
inno a i e business models (Ba elsman e al., 2013; Uzunidis, 2016).
China is a a c i ical junc u e in i s shi owa ds high-quali y economic
de elopmen . Du ing his pe iod, indus ial upg ading no only en-
hances esou ce alloca ion bu also signi ican ly imp o es he ai ness o
income dis ibu ion (Wu e al., 2018). In his p og ession, labou
inc easingly mig a es o high- alue-added sec o s, c ea ing
wide- anging employmen oppo uni ies and enhancing household
ea nings (Deng & He, 2018). The e o e, his p ocess educes income
dispa i y. Fo ins ance, ‘Made in China 2025’ d i es he in elligen
ans o ma ion o he manu ac u ing indus y. This ini ia i e gene a es
nume ous high-skilled employmen oppo uni ies and s imula es in-
come g ow h. Indus ial s uc u e ans o ma ion is inc easingly linked
o e iciency gains, equi able dis ibu ion and he in e play be ween
p oduc ion and dis ibu ion (Guo & Luo, 2021). Rega ding he
employmen s uc u e, he go e nmen has sponso ed oca ional
aining and se ice imp o emen s. These e o s ha e inc eased he
alignmen be ween wo ke s’ human capi al and job oppo uni ies (Zhou
& Chen, 2021). In pa icula , he ‘Voca ional Skills Enhancemen
Ini ia i e’ o e s ailo ed aining o mig an wo ke s and o he g oups,
equipping hem wi h he necessa y capabili ies. This combina ion o
policy in e en ion and ma ke mechanisms ully le e ages he guiding
ole o he go e nmen . Mo eo e , i s imula es he esou ce-alloca ion
e iciency in he ma ke , he eby achie ing a balance be ween ‘e i-
ciency and ai ness’ in income dis ibu ion.
(3) Inno a i e Vi ali y S imula ion Mechanism
The s imula ion o en ep eneu ial i ali y is ano he c ucial aspec
h ough which inno a ion-d i en policies exe a posi i e impac on
income inequali y. The con e gence o en u e capi al and he expan-
sion o inancing channels ha e e ec i ely alle ia ed he inancial
p essu es aced by s a -ups (S igli z, 2015; Mulie & Sama in, 2021).
The agglome a ion o inno a i e alen accele a es he low o knowl-
edge and echnological inno a ion. The con inuous augmen a ion o
human esou ce endowmen holds he key o os e ing inno a ion
wi hin high- ech i ms (Huang e al., 2023). In addi ion, policies aimed
o e ine he business en i onmen and enhance go e nmen se ice
e iciency. They p o ide en ep eneu s wi h a mo e obus suppo
sys em and s eng hen hei abili y o wi hs and ma ke unce ain ies
(Ding e al., 2021; Juan e al., 2024). The lou ishing o en ep eneu ial
ac i i ies is a powe ul ca alys o he ise o no el indus ies and
inno a i e business pa adigms. Meanwhile, inc eased ma ke compe i-
ion and e icien esou ce in eg a ion concu en ly op imise income
dis ibu ion s uc u es (Zhao e al., 2020). A he u ban scale, en e-
p eneu ial ac i i ies dis up ma ke disequilib ia and os e a subs an ial
numbe o job oppo uni ies. Con e sely, a he coun y le el, hey play a
pi o al ole in augmen ing a me s’ incomes and educing he u ban-
– u al income gap (Ye e al., 2022). The concen a ion o alen accel-
e a es knowledge low and echnological inno a ion. Mo eo e , i
educes knowledge exchange cos s, p omo ing he apid di usion and
applica ion o new echnologies and ideas. The e o e, inno a ion pol-
icies ha e lowe ed en ep eneu ial ba ie s and enhanced suppo and
inancing (Bai e al., 2022). These policies ha e igni ed socie al en hu-
siasm o inno a ion and en ep eneu ship. This p omo es he g ow h o
mic o and small en e p ises, c ea es economic g ow h poin s and jobs
and inc eases income oppo uni ies. Consequen ly, i helps alle ia e
income inequali y.
Conside ing his, he subsequen hypo heses a e p oposed in his
s udy:
H1: The enac men o he inno a i e ci y pilo policy has a posi i e
impac on alle ia ing income inequali y. Howe e , wi h an inc eased
inpu o inno a ion policies, hei impac on na owing income
inequali y declines.
H2a: Income inequali y is educed h ough an inno a i e ci y pilo
policy ia labou esou ce agglome a ion.
H2b: Income inequali y is educed by he inno a i e ci y pilo policy
h ough he op imisa ion o indus ial and employmen s uc u es.
H2c: Income inequali y is educed h ough an inno a i e ci y pilo
policy ha enhances u ban inno a i e i ali y.
Model, a iables and da a
Model
The DID echnique is a equen ly used econome ic ool o e alua e
he in luence o policy enac men . The p oposed model allows he
analysis and quan i ica ion o policy impac s while minimising in e -
e ence om o he ac o s. The undamen al concep o DID is o ega d
policy implemen a ion and ins i u ional changes as exogenous ac o s.
In pa icula , hese ac o s a e conside ed ‘quasi-expe imen s’ o ‘na u-
al expe imen s’ wi hin an economic sys em. This app oach assumes
ha policy implemen a ion ollows a mechanism simila o andom
assignmen . This mechanism gua an ees ha he cha ac e is ics and
endencies o he ea men and con ol g oups a e compa able. This
me hodology examines he changes in ou comes wi hin he expe i-
men al and con ol g oups be o e and a e he policy is enac ed.
Th ough his examina ion, he di e ences ha can be a ibu ed o he
policy can be iden i ied. This app oach mi iga es endogenei y issues
a ising om ex e nal ac o s, allowing he es ima ion o he ne e ec o
he policy.
The pilo p ojec o inno a i e ci ies exhibi s he ai s ypical o a
‘quasi-na u al expe imen ’. D awing upon he ea u es o he DID model,
his app oach o e s wo key ad an ages. Fi s , i ha nesses he ime-
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
4
se ies da a ac oss mul iple pe iods. This allows he acking o dynamic
shi s in income inequali y a a ious in e als ollowing policy ollou .
Income inequali y is no ins an aneously a ec ed by he inno a i e ci y
pilo policy; a he , i un olds g adually o e ime. Second, he mul i-
pe iod DID me hod can e ec i ely dis inguish he ea men g oup
om he con ol g oup in he p e- and pos -policy implemen a ion pe-
iods. The e o e, i allows o a p ecise es ima ion o he causal
connec ion be ween he inno a i e ci y pilo policy and income
inequali y. This makes i sui able o e alua ion ia he DID me hod.
Howe e , as he policy was olled ou in se e al s ages, a mul i-pe iod
DID me hodology was adop ed in his esea ch o o mula e he model
(Beck & Le ko , 2010; Wang e al., 2023). To assess whe he he policy
e ec i ely educes income inequali y, he ollowing equa ion was
es ablished:
Ginii =β0+β1DIDi +β2∑Con oli +ui+λ +
ε
i (6)
whe e Gini
i
is he deg ee o income inequali y in ci y i in yea and DID
i
depic s whe he ci y i is designa ed as an inno a i e ial ci y in yea . I
ci y i belongs o he inno a i e ial ci ies, i s alue is se o 1. O he wise,
i is 0. u
i
and λ
a e he ci y- ixed and yea - ixed e ec s, espec i ely.
ε
i
is
he andom dis u bance e m, and Con ol
i
is he se o con ol a iables.
Va iables
(1) Dependen Va iable
In he empi ical analysis, he Gini coe icien is employed as a p oxy
o assess income inequali y. Acco ding o Fang and Meng (2024), he
Gini coe icien s o each ci y we e calcula ed. The de ailed calcula ion
o he Gini coe icien o ci y i in yea is p esen ed as ollows:
Ginii =∑
ni
k=1∑
ni
=1
|Lik −Li |
2n2
iLi
(7)
whe e ni is he o al numbe o u ban and ownship uni s in he i p e-
ec u e le el o abo e he ci y. Li is he a e age nigh ime ligh in ensi y
o he i p e ec u e le el o abo e he ci y in yea . Lik is he nigh ime
ligh in ensi y o u ban o ownship uni wi hin ci y i du ing he yea .
Li is he nigh ime ligh in ensi y o he u ban o ownship uni wi hin
ci y i du ing he yea .
(2) Co e Independen Va iable
A bina y a iable is c ea ed o mi o he pilo policy o inno a i e
ci ies, conside ing he ime and ange o i s implemen a ion. This a i-
able is deno ed as DID and is de i ed by mul iplying he a iable o ime
wi h ea . When a ci y is selec ed as an inno a i e ial ci y, i is
assigned o he ea men g oup in which he alue o ea is se o 1. In
cases whe e i is no , ea is se o 0. In case a ci y is iden i ied as
inno a i e in a ce ain yea , he alue o ime is se o 1 om ha yea
onwa ds. Fo all yea s be o e he designa ion and o ci ies no iden i ied
as inno a i e, he alue is se a 0. The ea men and con ol g oups
comp ise 97 and 179 inno a i e pilo and non-pilo ci ies, espec i ely.
(3) Con ol Va iables
Re e encing he exis ing body o li e a u e, he subsequen con ol
a iables we e chosen. The economic de elopmen le el (a gdp) is
measu ed by he in la ion-adjus ed pe capi a eal g oss domes ic
p oduc (GDP) o he ci y. This GDP was con e ed o cons an 2003
p ices. Subsequen ly, he esul ing alue o pe capi a eal GDP is log-
ans o med. Go e nmen in e en ion (go ) is calcula ed as he a io
o he budge a y spending o he local go e nmen in ela ion o he
egional GDP. A e age wage le el (income) is he log- ans o med
a e age wage o u ban employees. Fixed asse in es men (asse ) is
he log- ans o med o al ixed asse in es men . Deg ee o ade open-
ness (open) is calcula ed as he pe cen age o he combined alue o
impo s and expo s ep esen a i e o GDP. Popula ion densi y (popm) is
measu ed as he popula ion pe squa e kilome e.
Da a sou ces and desc ip i e s a is ics
Ci ies wi h subs an ial da a gaps we e excluded o main ain da a
comple eness. The da ase co e s 276 Chinese u ban a eas om 2003 o
2022. Among hem, 97 we e iden i ied as inno a ion pilo ci ies, and he
emaining 179 ci ies ac ed as non-pilo coun e pa s. The da a a e p i-
ma ily om a ious edi ions o he China U ban S a is ical Yea book and
s a is ical yea books o p o inces, ci ies and coun ies. Da a ega ding
he inno a i e ial ci ies we e sou ced om he documen i led
‘Guidelines o Es ablishing Inno a i e Ci ies’. Table 1 p esen s a syn-
opsis o he desc ip i e s a is ical igu es o he key a iables.
Empi ical analysis
Reg ession esul s
Table 2 p esen s he esul s o he impac s o he inno a i e ci y ial
ini ia i e on income dispa i y. Column 1 p esen s he ou comes wi hou
conside ing o he ac o s in o accoun . The eg ession analysis indica es
a s a is ically signi ican nega i e coe icien o he policy in e en ion
dummy (−0.0145), indica ing ha he inno a i e ci y pilo ini ia i e
e ec i ely mi iga es income inequali y. Column 2 p esen s con ol
a iables bu only accoun s o ci y- ixed e ec s. The policy- ela ed
coe icien is −0.0261, hough subjec o sligh modi ica ions, and
consis en ly e ains i s signi ican ly nega i e alue ac oss di e en
model speci ica ions. These indings indica e ha policy in e en ions
can educe economic dispa i y and p omo e sha ed p ospe i y. This
e ec i eness emains e en when unobse ed he e ogenei y a he ci y
le el is conside ed h ough ixed-e ec s es ima ion.
To mi iga e po en ial con ounding biases, he speci ica ion p esen ed
in column 3 inco po a es ime and ci y- ixed e ec s o assess he causal
ela ionship be ween he policy in e en ion and income dis ibu ion
ou comes. The empi ical analysis e eals a policy coe icien es ima e o
−0.0101 o he in e en ion indica o a iable. While he e is some
luc ua ion compa ed wi h he esul s wi hou he con ol a iables, he
nega i e e ec , s a is ically signi ican a he 1% signi icance le el, e-
mains e iden . Assuming o he ac o s a e held cons an , he policy
causes a educ ion o ~1.01% in he a e age Gini coe icien o pilo
ci ies in compa ison wi h he coe icien o non-pilo ci ies. The esul s
indica e ha he pilo ini ia i e o inno a i e ci ies, se ing as a
Table 1
De ini ions o a iables.
Va iable Symbols Obs Mean S d.
de .
Min Max
Gini
coe icien
Gini 5520 0.7481 0.1982 0.0388 0.9952
Policy dummy
a iable
DID 5520 0.1486 0.3557 0 1
Economic
de elopmen
a gdp 5520 6.7109 1.0606 3.2122 10.2859
Go e nmen
in e en ion
go 5520 0.1717 0.0945 0.0313 1.4852
A e age wage
le el
income 5520 10.5261 0.6947 2.2834 12.6780
Fixed asse
in es men s
asse 5520 15.7184 1.3228 10.2518 19.0834
T ade
openness
open 5520 0.2127 0.4026 0.0004 7.6201
Popula ion
densi y
popm 5520 5.8047 0.9593 1.5476 9.2350
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
5
co ne s one o he inno a ion-led g ow h s a egy in China, signi ican ly
enhances income dis ibu ion equali y. Thus, Hypo hesis 1 is con i med.
Pa allel end es
Sa is ying he assump ion ega ding pa allel ends is a equi emen
o he mul i-pe iod DID. I his condi ion is iola ed, he es ima ed
coe icien s canno accu a ely e lec he policy e ec . Owing o he
phased ollou o he in e en ion, he composi ion o ci y g oups
changes in each phase. To o e come his empi ical issue, his esea ch
employed he e en s udy me hodology, ollowing he p eceden se by
Jacobson e al. (1993). This me hodological amewo k simul aneously
alida es he pa allel end hypo hesis and assesses he empo al e o-
lu ion o policy impac s.
Conside ing he pos -policy sample size, he ime a iable anges
om −6 (6 yea s p e-policy) o 4 (4 yea s pos -policy). As shown in
Fig. 1, in he pe iod be o e he inno a i e ci y p og amme ook e ec ,
he in e en ion and con ol g oups had no signi ican end di e ences.
This e idence indica es he sa is ac ion o he pa allel end assump ion.
The pos -implemen a ion pe iod e eals s a is ically signi ican nega i e
coe icien s o income inequali y. This inding e eals a signi ican
di e ence be ween ci ies pa icipa ing in he pilo scheme and hose no
pa icipa ing. This di e ence e i ies ha he policy measu es a e
e ec i e in educing income inequali y. In conclusion, he obse ed
income inequali y educ ion is no a ibu ed o p e-policy ends.
Dynamic e ec analysis
As shown in Fig. 1, he dynamic e ec s o policy in luence a e
disce nible. No ably, a e he hi d pe iod, he e is a mino dec ease in
he posi i e in luence ha inno a i e ci y buildings ha e on income
inequali y. In essence, he e icacy o inno a ion policies in educing
income inequali y i s ises and hen alls. This phenomenon can be
comp ehensi ely in e p e ed om wo key pe spec i es: he heo y o
diminishing ma ginal e ec s and ex e nal shocks.
In line wi h he law o diminishing ma ginal e ec s, a he onse o
inno a ion policy implemen a ion, wo aspec s con ibu e o he ini ial
impac o he policy. Fi s , R&D subsidies and pa en incen i es uel
high-skilled indus y g ow h, d i e economic expansion and gene a e
nume ous high-paying jobs. In addi ion, hese policies upg ade adi-
ional indus ies, boos ing low-skilled wo ke s’ incomes and na owing
he income gap. Second, he in ensi ying agglome a ion o inno a ion
esou ces d aws alen s, unds and echnologies o ci ies. This gene a es
en ep eneu ial and in es men oppo uni ies, di e si ies income sou -
ces o a ious g oups and pa icula ly ele a es he ea nings o ac i e
inno a o s. Howe e , wi h he con inuous injec ion o policy esou ces,
sa u a ion o key inno a i e elemen s (such as alen and capi al) occu s.
Concu en ly, policy implemen a ion cos s a e ising, and he abili y o
a ac alen is weakening. These wo ac o s combine o unde mine he
e icacy o a policy in na owing income gaps, e en ually leading o
diminished policy ou comes.
This s udy akes ac ion o e i y he compliance o he ma ginal e -
ec s o inno a ion policies wi h he law o diminishing e u ns. The
model uses a dynamic panel model and a gene alised me hod o mo-
men s es ima ion o analyse he impac o inno a ion policy in ensi y on
income inequali y. He ein, he in ensi y o inno a ion policies is indi-
ca ed by he a io o R&D expendi u e ela i e o GDP (RD_ a io). Col-
umn 1 in Table 3 p esen s ha he coe icien o he R&D expendi u e
p opo ion is signi ican ly posi i e (−0.601), whe eas he coe icien o
i s squa ed e m is signi ican ly nega i e (0.7069). This indica es a U-
shaped ela ionship be ween inno a ion policy in ensi y and income
inequali y. The co ec ness o he heo e ical hypo hesis 1 was e i ied.
In addi ion, a dummy a iable DT is cons uc ed o he h ee-pe iod
Table 2
Impac o he inno a i e ci y pilo policy on income inequali y.
Va iables (1) (2) (3)
DID −0.0145** −0.0261*** −0.0101***
(0.0056) (0.0050) (0.0030)
a gdp −0.0709*** −0.0125*
(0.0081) (0.0068)
open 0.0073 0.0295***
(0.0055) (0.0034)
go 0.2658*** 0.0199
(0.0264) (0.0165)
income −0.0240*** 0.0053
(0.0058) (0.0040)
asse 0.0197*** −0.0005
(0.0028) (0.0018)
popm −0.0100*** 0.0087***
(0.0030) (0.0018)
Time FE Yes No Yes
Ci y FE Yes Yes Yes
R
2
0.704 0.149 0.710
N5520 5520 5520
No e: *, ** and *** ep esen s a is ical signi icance a he 10%, 5% and 1%
le els, espec i ely, wi h alues based on ci y-le el clus e ing p o ided in pa-
en heses. The same applies o he ollowing ables.
Fig 1. Pa allel end es .
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
6
ime ame a ound policy implemen a ion. A alue o 1 is assigned a e
he hi d-pe iod policy implemen a ion, and 0 is assigned be o e i . The
implemen a ion s eng h o alen policies is measu ed by ex ac ing
alen - ela ed e ms om go e nmen wo k epo s. These epo s co e
policy e alua ions, e ec i eness and u u e-o ien ed in ensi y. Column
2 in Table 3 p esen s he impac o he inno a ion policy dummy a i-
able DID and i s in e ac ion e m wi h DT on alen -in oduc ion in-
ensi y. The analysis indica es ha local policies ha e enhanced he
in ensi y o alen in oduc ion since he inno a ion policy ook e ec .
Howe e , his in ensi y wanes a e he hi d implemen a ion phase,
u he diminishing he posi i e in luence o he alen agglome a ion
e ec on income inequali y.
The s able implemen a ion o inno a ion policies can be dis up ed by
ex e nal e en s, causing a ia ions in hei in luence on income
inequali y. Among he a ious ex e nal shock ac o s, he home- e u n
en ep eneu ship policy is a quin essen ial example. Pilo ed in ce ain
egions in 2016 and 2017, his policy may dis up he labou o ce ag-
g ega ion pa e n ollowing he implemen a ion o u ban inno a ion
policies. A dummy a iable FX is cons uc ed by his pape o he policy
o e u ning o home own o en ep eneu ship. Tha is o say, when a
coun y/dis ic in a ci y en o ces he policy, i ge s a alue o 1; o he -
wise, i ge s 0. As shown in column 3, he coe icien o DID#FX is
signi ican ly posi i e wi h he inclusion o he e ec s o he home-
e u ning en ep eneu ship policy. This indica es ha he posi i e
impac o he inno a i e ci y pilo policy on income inequali y has
declined. The decline is pa icula ly no able a e he shock o he home-
e u ning en ep eneu ship policy. As s a ed o he wise, he policy o
e u ning o one’s home own o s a a business o se s some o he
e ec i eness o he inno a ion policy. Consequen ly, i has inc eased he
di icul y o na owing he income gap.
Robus ness es s
Resul s om he baseline eg ession e i y ha a ci y’s inclusion in
he inno a i e pilo p og amme no ably cu s down income inequali y.
To ensu e ha he conclusions a e no in luenced by con ounding ac-
o s, a se ies o obus ness es s we e conduc ed. These es s add ess
a ious dimensions, such as sample selec ion, exclusion o o he policy
in e e ences, PSM-DID analysis, non- andom sample selec ion and
ins umen al a iable eg ession.
(1) Sample Da a Fil e ing
To add ess he impac o ex eme alues, nume ical a iables we e
winso ised a he i s and i h pe cen iles, and he model was e-
e alua ed. In addi ion, ce ain special yea s in he sample may ha e
impac ed he accu acy o he esul s, p omp ing hei exclusion om he
analysis. Fo ins ance, he 2008 wo ldwide inancial u moil caused a
ma ked dec ease in he impo and expo ac i i ies wi hin China.
Despi e he in oduc ion o economic s imulus policies globally, he
c isis esul ed in ce ain challenges such as inancing di icul ies, ising
cos s, inc eased unemploymen and educed wages. Simila ly, he
COVID-19 ou b eak in 2020 caused widesp ead ci y and business shu -
downs, u he dis up ing economic ac i i y. To elimina e he in luence
o hese special yea s, da a om 2008 and 2020 we e excluded, and he
model was e-es ima ed using he emaining sample. As shown in col-
umns 1, 2 and 3 in Table 4, he coe icien s o he DID policy a e no ably
nega i e a he 10% signi icance le el (−0.0078, −0.0073 and −0.0060,
espec i ely). The consis ency o hese esul s wi h p e ious esea ch
bols e s he soundness o he indings.
(2) Excluding he In luence o O he Policies
To p ecisely e alua e he in luence o he inno a i e ci y pilo p o-
g amme on income inequali y, excluding he impac s o o he policy
measu es is c ucial. Th ough a e iew o ele an li e a u e and policy
documen s, he s udy iden i ied sma ci y policy as a po en ial con-
ounding ac o du ing he sample pe iod. The policy has acili a ed he
ad ancemen o in elligen echnologies, which a e used o ein o ce
u ban in as uc u e and p omo e economic de elopmen . In addi ion,
he policy has had a bene icial impac on u ban inno a ion. This may, in
u n, in luence income inequali y.
The sma ci y policy was implemen ed in h ee phases s a ing in
2010, and i coincided empo ally wi h he inno a i e ci y pilo policy.
To es o obus ness, an indica i e bina y a iable, SMA, o he sma
ci y ini ia i e was included. In pa icula , he SMA a iable is coded as 1
o ci ies pa icipa ing in he sma ci y ini ia i e and as 0 o hose ha
do no . Column 4 in Table 4 p esen s he ele an esul s.
The esul s indica e ha when he SMA a iable is included, he sign
o he coe icien ela ed o he DID policy a iable (−0.0077) does no
change. In addi ion, he DID policy a iable emained s a is ically sig-
ni ican . This unde sco es ha he esul s a e independen o he sma
ci y policy o o he po en ial con ounding ac o s. Mo eo e , he coe -
icien associa ed wi h he inno a i e ci y policy emained s able,
u he s eng hening he eliabili y o he s udy’s indings.
(3) PSM-DID Analysis
Owing o he ela i ely la ge size o China, ci ies signi ican ly a y in
e ms o economic de elopmen and policy en o cemen . The ea men
and con ol g oups may exhibi dis inc cha ac e is ics. Mo eo e , biases
could a ise om sample selec ion, e e se causali y o o he sou ces o
endogenei y. To add ess hese challenges, his s udy used he PSM-DID
me hod o alida ion. Table 5 p esen s he esul s o he p ocess
employing adius ma ching, ke nel ma ching and nea es -neighbou
ma ching echniques. The ob ained coe icien s we e signi ican unde
all h ee me hods. The esul s (−0.0108, −0.0080 and −0.0078) a i m
ha he pilo policy o inno a i e ci ies e ec i ely mi iga es income
inequali y. This con i ma ion s eng hens he obus ness o he
conclusions.
Table 3
The in ensi y o alen in oduc ion and home- e u ning inno a ion policy.
Va iables (1) Gini (2) The in ensi y o alen in oduc ion (3) Gini
RD_ a io −0.6010**
(0.2625)
RD_ a io2 0.7069**
(0.3418)
DID 0.0048** −0.0171***
(0.0022) (0.0038)
DID#DT −0.0071***
(0.0023)
DID#FX 0.0152***
(0.0053)
Con ols Yes Yes Yes
N5520 5520 5520
R
2
0.277 0.710
Wald es 1543.06
Table 4
Robus ness es s.
Va iable
(1) 1%
Winso isa ion
(2) 5%
Winso isa ion
(3) Excluding
special yea s
(4) Excluding
o he policies
DID −0.0078** −0.0073** −0.0060* −0.0077**
(0.0031) (0.0031) (0.0034) (0.0031)
SMA 0.0025
(0.0020)
Con ols Yes Yes Yes Yes
Time FE Yes Yes Yes Yes
Ci y FE Yes Yes Yes Yes
R
2
0.707 0.704 0.719 0.711
N5520 5520 4968 5520
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
7
(4) Non- andom Sample Selec ion
When selec ing he lis o pilo ci ies o inno a ion, go e nmen s
o en conside speci ic a ibu es such as geog aphic loca ion and eco-
nomic de elopmen le el. O e ime, hese a ibu es may ha e di e -
en ial impac s on income inequali y. Thus, i is c ucial o conside and
con ol hese ac o s o main ain he obus ness o he esul s. The DID
me hod adop ed in his s udy assumes a quasi-na u al expe imen in
which he ea men and con ol g oups a e ideally andomly selec ed.
The ac ual selec ion o inno a i e pilo ci ies, howe e , is in luenced by
a ious ac o s such as economic de elopmen , geog aphic loca ion and
social condi ions, which a e no en i ely andom.
This s udy aimed o add ess po en ial biases om ‘non- andom’ se-
lec ion and educe hei impac . To achie e his, i inco po a es in e -
ac ion e ms be ween baseline cha ac e is ics and linea ime ends in o
he baseline eg ession model (1). The upda ed model is exp essed as
ollows:
Ginii =γ0+γ1DIDi +γ2∑Con oli +γ3∑Dumc× end +ui+λ
+
ε
i
(8)
In his model, Dum
c
ep esen s a se o dummy a iables ha cap u e
speci ic ci y cha ac e is ics. These en ail aspec s such as whe he he ci y
belongs o he Yang ze Ri e Economic Bel (Dum
1
), whe he i holds he
s a us o a municipali y di ec ly unde he cen al go e nmen (Dum
2
)
and whe he i is designa ed as a special economic zone (Dum
3
). end
deno es he ime end e m. O he a iable de ini ions a e consis en
wi h hose in p e ious sec ions.
Table 6 p esen s he esul s. Columns 1–3 indi idually inco po a e
each in e ac ion e m, and column 4 simul aneously includes all h ee
in e ac ion e ms. The DID coe icien consis en ly and signi ican ly e-
mains nega i e ac oss all speci ica ions, wi h alues o −0.0066,
−0.0066, −0.0080 and −0.0054. This con i ms ha he inno a i e ci y
pilo policy signi ican ly in luences he educ ion o income inequali y.
Mo eo e , he esul s indica e ha , while ce ain ci y-speci ic ac o s
we e conside ed du ing he selec ion o pilo ci ies, he p ocess e ains
some deg ee o andomness.
(5) Ins umen al Va iable Reg ession
De e mining inno a i e pilo ci ies is no based on he p inciple o
andomness. Ins ead, i akes in o comp ehensi e conside a ion a ious
ac o s such as he egional posi ioning, inno a ion capabili ies and
economic de elopmen le els o he ci ies. Owing o he non- andom
selec ion, he ea men g oup is highly likely o be in e e ed wi h by
policy endogenei y. The e may also be po en ial endogenei y issues wi h
he policy a iable, ul ima ely leading o de ia ions in he esea ch e-
sul s. This s udy employed he ins umen al a iable me hod o es i-
ma ion o add ess he in e e ence o endogenei y issues in esea ch
esul s. I designa ed Na ional His o ical and Cul u al Ci ies as he
ins umen al a iable o policy. This choice is well- ounded. On he one
hand, he e is a simila i y in economic s a us be ween inno a i e ci ies
and Na ional His o ical and Cul u al Ci ies because hey a e economic
cen es. Inno a i e ci ies aim o c ea e inno a i e cen es wi h s ong
adia ing and d i ing e ec s and a e key o ces in p omo ing he eco-
nomic de elopmen o mode n socie y. Na ional His o ical and Cul u al
Ci ies we e mos ly impo an economic and poli ical a eas in his o y and
se ed as he economic co e egions in ancien socie y. The e o e, he e
is a s ong co ela ion be ween hem. On he o he hand, Na ional His-
o ical and Cul u al Ci ies canno di ec ly a ec he income inequali y o
cu en ci ies. These can only exe hei in luence h ough he es ab-
lishmen o inno a i e ci ies. In his way, hey mee he ‘exclusion e-
s ic ion’ and con o m o he equi emen o exogenei y.
As shown in Table 7, he ins umen al and policy a iables a e
signi ican ly and posi i ely co ela ed. In addi ion, he ele an es e-
sul s demons a e ha he ins umen al a iable mee s he weak iden-
i ica ion equi emen . The es ima ed coe icien o DID emains no ably
nega i e. This implies ha e en po en ial endogenei y issues a e u he
conside ed. I can s ill be concluded ha his inno a i e ci y pilo policy
will educe income inequali y.
He e ogenei y analysis
He e ogenei y by geog aphical loca ion
Municipali ies loca ed in a ious geog aphical egions exhibi sub-
s an ial dispa i ies in e ms o economic de elopmen le els and ap-
p oaches. These di e ences migh ha e an impac on he execu ion and
ou comes o he inno a i e ci y pilo ini ia i e. The e o e, i is c ucial o
analyse whe he he impac o a policy on income inequali y a ies by
geog aphical loca ion. Acco ding o he classi ica ion in he ‘China Na-
ional Economic and Social De elopmen S a is ics Bulle in’, ci ies a e
ca ego ised in o wo egions: eas e n and cen al wes e n. This s udy
u he classi ies ci ies based on hei geog aphical loca ions, using he
‘Hu Huanyong Line’ and ‘ he Yang ze Ri e Economic Bel ’ as bound-
a ies. Mo eo e , his s udy examined he di e ences be ween bo de
and non-bo de ci ies, as well as be ween ci ies along he Yang ze Ri e
and hose in o he egions.
The indings o his s udy p esen ed in Table 8 indica e ha he
inno a i e ci y pilo policy exe s a nega i e ye s a is ically insigni i-
can in luence in he eas e n egion. Con e sely, in he cen al and
wes e n egions, i s impac is ema kably nega i e, wi h a coe icien o
−0.0081. Such a di e ence migh be a ibu ed o he dis inc le els o
economic de elopmen p esen in hese di e en egions. By con as o
Table 5
PSM-DID es .
Va iable (1) (2) (3)
Radius Ke nel Nea es -neighbou
DID −0.0108** −0.0080*** −0.0078**
(−2.5656) (−2.6213) (−2.5398)
Con ols Yes Yes Yes
Time FE Yes Yes Yes
Ci y FE Yes Yes Yes
N 3463 5474 5520
R
2
0.7391 0.7126 0.7105
Table 6
Analysis o non- andom sample selec ion.
Va iable (1) (2) (3) (4)
DID −0.0066** −0.0066** −0.0080*** −0.0054*
(0.0030) (0.0031) (0.0030) (0.0030)
Con ols Yes Yes Yes Yes
Time FE Yes Yes Yes Yes
Ci y FE Yes Yes Yes Yes
R
2
0.717 0.715 0.719 0.731
N5520 5520 5520 5520
Table 7
Ins umen al a iable eg ession.
Va iable (1) The i s s age (2) The second s age
DID −0.0113**
(0.0043)
IV 0.8638**
(0.0119)
Con ols Yes Yes
Time FE Yes Yes
Ci y FE Yes Yes
R
2
0.810 0.943
N5520 5520
F- alue in he i s s age 10782.27
Kleibe gen-Paap LM s a is ic 726.06 (P alue =0.0000)
Kleibe gen-Paap Wald s a is ic 10782.27 (P alue =0.0000)
X. Li e al.
Jou nal o Inno a ion & Knowledge 10 (2025) 100749
8
