Connective financing: Chinese infrastructure projects and the diffusion of economic activity in developing countries

Bluhm, Richa d e al.
A icle — Published Ve sion
Connec i e inancing: Chinese in as uc u e p ojec s and
he di usion o economic ac i i y in de eloping coun ies
Jou nal o U ban Economics
P o ided in Coope a ion wi h:
Kiel Ins i u e o he Wo ld Economy – Leibniz Cen e o Resea ch on Global Economic Challenges
Sugges ed Ci a ion: Bluhm, Richa d e al. (2025) : Connec i e inancing: Chinese in as uc u e
p ojec s and he di usion o economic ac i i y in de eloping coun ies, Jou nal o U ban Economics,
ISSN 1095-9068, Else ie , Ams e dam, Vol. 145, pp. 1-22,
h ps://doi.o g/10.1016/j.jue.2024.103730
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/318203
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Con en s lis s a ailable a ScienceDi ec
Jou nal o U ban Economics
jou nal homepage: www.else ie .com/loca e/jue
Connec i e inancing: Chinese in as uc u e p ojec s and he di usion o
economic ac i i y in de eloping coun ies
Richa d Bluhma, Axel D ehe b,c,d,e, ,∗, And eas Fuchsg,h, B adley C. Pa ksi,j, Aus in M. S angek,
Michael J. Tie neyl
aIns i u e o Economics and Law, Uni e si y o S u ga , Ge many
bAl ed-Webe -Ins i u e o Economics, Heidelbe g Uni e si y, Ge many
cCe ES, I aly
dKOF Swiss Economic Ins i u e, Swi ze land
eCEPR, UK
CESi o, Ge many
gDepa men o Economics and Cen e o Mode n Eas Asian S udies, Uni e si y o Gö ingen, Ge many
hKiel Ins i u e o he Wo ld Economy, Ge many
iAidDa a, Global Resea ch Ins i u e, William & Ma y, Uni ed S a es o Ame ica
jCen e o Global De elopmen , Uni ed S a es o Ame ica
kDepa men o Poli ics and Public Adminis a ion, The Uni e si y o Hong Kong, Hong Kong, China
lDepa men o Go e nmen , William & Ma y, Uni ed S a es o Ame ica
ARTICLE INFO
JEL classi ica ion:
F35
R11
R12
P33
O18
O19
Keywo ds:
De elopmen inance
T anspo cos s
In as uc u e
Fo eign aid
Spa ial concen a ion
China
ABSTRACT
This pape s udies he causal e ec o anspo in as uc u e on he spa ial dis ibu ion o economic ac i i y
wi hin subna ional egions ac oss a la ge numbe o de eloping coun ies. To do so, we in oduce a new
global da ase o geoloca ed Chinese g an - and loan- inanced de elopmen p ojec s om 2000 o 2014 and
combine i wi h measu es o spa ial concen a ion based on emo ely sensed da a. We ind ha Chinese-
inanced anspo a ion p ojec s decen alize economic ac i i y wi hin egions, as measu ed by a spa ial Gini
coe icien , by 2.2 pe cen age poin s. The ea men e ec s a e pa icula ly s ong in egions ha a e less
de eloped, mo e u banized, and loca ed close o ci ies.
1. In oduc ion
In 2009, he Expo –Impo Bank o China (China Eximbank) ap-
p o ed a loan o he Kenyan go e nmen o widen and imp o e he
Nai obi–Thika Highway – a 50.4 km dual ca iageway ha ex ends
om he cen e o Nai obi o he own o Thika. The p ojec , locally
known as he ‘‘Thika Supe -Highway,’’ sough o educe conges ion and
a el imes be ween Nai obi and a se o sa elli e owns along a c i -
∗Co esponding au ho a : Al ed-Webe -Ins i u e o Economics, Heidelbe g Uni e si y, Ge many.
E-mail add esses: [email p o ec ed] (R. Bluhm), [email p o ec ed] (A. D ehe ), [email p o ec ed] (A. Fuchs),
[email p o ec ed] (B.C. Pa ks), [email p o ec ed] (A.M. S ange), [email p o ec ed] (M.J. Tie ney).
ically impo an anspo a ion co ido (A ican De elopmen Fund,
2007). Upon comple ion in 2012, a ic lows inc eased by 45 pe cen ,
jou ney speeds ose om 8 km pe hou o a leas 45 km pe hou
in sec ions wi h he highes egis e ed a ic, and a e age commu ing
imes om Thika o Nai obi ell om 2–3 h o 30–45 min (KARA
and CSUD,2012;A ican De elopmen Bank,2014a,b,2016,2019).
Economic ac i i y sp ead ou along he anspo co ido and became
subs an ially less concen a ed in he co e o Nai obi (see Fig. 1). The
case o he Nai obi-Thika Highway i s wi hin a b oade pa e n:
h ps://doi.o g/10.1016/j.jue.2024.103730
Recei ed 20 Feb ua y 2023; Recei ed in e ised o m 31 Oc obe 2024
Jou nal o U ban Economics 145 (2025) 103730
A ailable online 9 Decembe 2024
0094-1190/© 2024 The Au ho s. Published by Else ie Inc. This is an open access a icle unde he CC BY license (
h p://c ea i ecommons.o g/licenses/by/4.0/ ).
R. Bluhm e al.
S a ing wi h Baum-Snow (2007), a se ies o s udies in a a ie y o
coun ies show ha majo anspo in as uc u e in es men s can
decen alize economic ac i i y.1
The key con ibu ion o his s udy is o examine whe he and o
wha ex en Chinese- inanced in as uc u e p ojec s a e decen alizing
economic ac i i y wi hin de eloping coun ies. We conduc ou analysis
a he le el o i s -o de adminis a i e egions. Fi s -o de egions
a e one laye below he na ional le el and co espond o p o inces,
s a es, oblas s, go e no a es, o emi a es, depending on he adminis-
a i e di isions in a gi en coun y.2Ou analysis ocuses speci ically
on he p o ision o anspo in as uc u e inancing om Chinese
s a e-owned en i ies, which has assumed a dominan ole in he con-
s uc ion and ehabili a ion o anspo a ion in as uc u e a ound he
wo ld du ing he 21s cen u y. Mos o ou analysis cen e s on he
concen a ion o economic ac i i y wi hin egions,3bu we also p esen
esul s ela ing o in as uc u e inancing and concen a ion ac oss
egions. Ou s udy es s whe he he esul s om he exis ing, usually
coun y-speci ic, li e a u e can be gene alized ac oss a la ge sample o
de eloping coun ies ha hos p ojec s suppo ed by he wo ld’s la ges
p o ide o in as uc u e inancing.
Since 2000, China’s go e nmen has inanced many o he la ges
anspo in as uc u e p ojec s in he Global Sou h. The sho - and
long- un consequences o China’s in as uc u e inancing ac i i ies –
including hose along he US$1 illion Bel and Road Ini ia i e (BRI)
– a e he subjec o conside able deba e in he media and wi hin
policy ci cles. A g owing numbe o s udies ocus on he expec ed
impac o he BRI in di e en egions (e.g., Pe lez and Huang,2017;
Bandie a and Tsi opoulos,2020;Bi d e al.,2020;de Soy es e al.,
2020;Lall and Leb and,2020). Beijing’s c i ics claim ha i inances
poo ly designed and has ily execu ed p ojec s ha p o ide ew eco-
nomic bene i s, while Wes e n dono s and lende s ha e lea ned h ough
decades o expe ience o design and implemen in as uc u e p ojec s
in mo e ca e ul and sus ainable ways. In esponse o moun ing c i icism
ha i inances poli ically mo i a ed and economically unsus ainable
p ojec s, he Chinese go e nmen has doubled down on i s leade ship
ole in he ma ke o global in as uc u e inance.4
Many de eloping coun ies ha e unme in as uc u e inancing
needs, and he leade s o hese coun ies a e quick o poin ou ha
China is willing and able o swi ly inance and build oads, b idges,
ailways, and po s a a ime when Wes e n dono s and lende s a e
no (Swedlund,2017).5Fo example, du ing his enu e as he P esi-
den o Senegal, Abdoulaye Wade admonished adi ional dono s and
1See, o example, Baum-Snow e al. (2017) and Bane jee e al.
(2020) on China, Bayes (2007) on Bangladesh, Bi d and S aub (2014) on
B azil, Donaldson (2018) on India, Hende son and Kunco o (1996) on Indone-
sia, Ga cia-Lopez e al. (2015) on Spain, Gibbons e al. (2019) on he Uni ed
Kingdom, and Du an on and Tu ne (2012) on he Uni ed S a es. Redding and
Tu ne (2015) as well as Baum-Snow and Tu ne (2017) p o ide su eys o
his li e a u e.
2In ou sample, he a e age egion’s size is 37,644 squa e kilome e s,
which oughly co esponds o he land a ea o Sou h Ca olina. I has 2.1
million inhabi an s, which oughly co esponds o New Mexico.
3We use he e ms spa ial concen a ion and spa ial cen aliza ion in e -
changeably as hey bo h e e o changes in he dis ibu ion o people and
ou pu ac oss space.
4A he 2017 Bel and Road Fo um o In e na ional Coope a ion, P es-
iden Xi emphasized ha ‘‘[i]n as uc u e connec i i y is he ounda ion o
de elopmen h ough coope a ion. We should p omo e land, ma i ime, ai and
cybe space connec i i y, concen a e ou e o s on key passageways, ci ies and
p ojec s and connec ne wo ks o highways, ailways and sea po s [...]’’ (Xi,
2017).
5An impo an eason o hese in as uc u e inancing gaps ollows om
he ac ha ‘‘Wes e n dono s ha e by and la ge go en ou o ha d in as uc-
u e sec o s [...] and [ ]hey [ins ead] channel hei assis ance o e whelmingly
o social sec o s o o in as uc u e sec o s such as wa e supply and sani a ion
ha ha e di ec e ec s on household heal h’’ (Dolla ,2008).
c edi o s o hei cumbe some bu eauc a ic p ocedu es, no ing ha :
‘‘[w]i h di ec aid, c edi lines and easonable con ac s, China has
helped A ican na ions build in as uc u e p ojec s in eco d ime. [...]
I ha e ound ha a con ac ha would ake i e yea s o discuss,
nego ia e and sign wi h he Wo ld Bank akes h ee mon hs when we
ha e deal wi h Chinese au ho i ies’’ (Wade,2008).
We in oduce he i s global da ase o geo-loca ed Chinese
go e nmen - inanced p ojec s ha we e unde aken in de eloping coun-
ies be ween 2000 and 2014.6The da ase includes 3485 p ojec s in
6184 subna ional loca ions ac oss 138 coun ies du ing hese 15 yea s.
Fo ou analysis, we ocus on 269 Chinese go e nmen - inanced ans-
po a ion in as uc u e p ojec s unde aken in 1215 subna ional lo-
ca ions ac oss 86 coun ies. The lowe bound o he o al inancial
alue o hese p ojec s is US$64 billion. We es ima e he e ec s o
hese p ojec s on he spa ial concen a ion o economic ac i i y – bo h
wi hin and ac oss subna ional ju isdic ions – wi h sa elli e da a on he
geog aphical dispe sion o nigh ime ligh ou pu (simila o Hende son
e al.,2018).
We iden i y he causal e ec o Chinese go e nmen - inanced ans-
po p ojec s on he spa ial concen a ion o economic ac i i y ollow-
ing he in ui ion o a gene alized di e ence-in-di e ences design. Ou
ins umen al a iable elies on he a ailabili y o esou ces o he
cons uc ion o such p ojec s. This app oach has he ad an age ha
compa able da a a e a ailable o a la ge numbe o coun ies, and
plausibly exogenous ins umen s can be applied ac oss hese di e se
empi ical se ings.7Speci ically, we in oduce an ins umen al a iable
ha uses an exogenous supply push a iable in e ac ed wi h a local
exposu e e m: China’s domes ic p oduc ion o po en ial p ojec in-
pu s in e ac ed wi h each ecipien egion’s p obabili y o ecei ing
p ojec s. We use China’s annual p oduc ion o aluminum, cemen ,
glass, i on, s eel, and imbe o p oxy i s capaci y o p o ide phys-
ical p ojec inpu s.8The in ui ion behind his app oach is ha he
Chinese go e nmen has long conside ed hese p oduc ion ma e ials
as s a egic commodi ies and, he e o e, p oduced hem in excess o
domes ic demand. This policy esul s in la ge su pluses, some o which
China edi ec s o o e seas in as uc u e p ojec s.9We he e o e expec
China o be mo e lenien owa ds coun ies ha eques inancing o
anspo in as uc u e p ojec s in he yea s when such inpu s a e
abundan and less lenien in he yea s when such inpu s a e sca ce.
We also expec subna ional locali ies ha equen ly ecei e Chinese
go e nmen - inanced anspo p ojec s o be mo e hea ily a ec ed
by yea - o-yea luc ua ions in he supply o p ojec inpu s. We hus
compa e he e ec s o Chinese anspo p ojec s on spa ial concen a-
ion induced by changes in China’s domes ic p oduc ion o po en ial
p ojec inpu s ac oss wo g oups: egions ha a e egula and i egula
ecipien s o Chinese anspo in as uc u e inancing.
6Though he BRI was no o icially launched un il la e 2013, he Chinese
go e nmen had al eady begun p o iding a signi ican numbe o la ge-scale
inancing o anspo in as uc u e in de eloping coun ies by he u n o he
cen u y. These p e-2014 p ojec s sha e mos o he cha ac e is ics o anspo
in as uc u e p ojec s ha a e now o mally pa o he BRI.
7The li e a u e ypically uses his o ical anspo ne wo ks o o he
coun y-speci ic his o ical ci cums ances (such as minimum spanning ees
connec ing he la ges ci ies).
8Expo ing excess capaci y in a a ie y o ma e ials h ough in as uc u e
in es men s ab oad is one o he seconda y mo i es o en asc ibed o he
Chinese go e nmen ’s BRI ini ia i e. Fo example, he Economis w i es ‘‘M
Xi [...] hopes o [...] expo some o his coun y’s as excess capaci y
in cemen , s eel and o he me als’’ (see www.economis .com/ he-economis -
explains/2017/05/14/wha -is-chinas-bel -and- oad-ini ia i e). Ou app oach
ex ends he s a egy p oposed in D ehe e al. (2021a) which exclusi ely used
he le el o s eel p oduc ion.
9Chinese in as uc u e p ojec s usually equi e cons uc ion inpu s ha a e
o e supplied in China, and Chinese s a e-owned banks usually obliga e hei
bo owe s o impo hese inpu s on a p e e en ial basis (D ehe e al.,2021b).
Jou nal o U ban Economics 145 (2025) 103730
2
R. Bluhm e al.
Fig. 1. Nai obi–Thika Highway, change in nigh ime ligh s, 2008–2013. No es: The igu e illus a es he change in nigh ime ligh s om 2008 o 2013 along he ou e o he
Nai obi–Thika Highway in Kenya, which was cons uc ed om Janua y 2009 un il Oc obe 2012. Majo in e sec ions and poin s o in e es a e highligh ed along he highway. The
change in nigh ime ligh s is he di e ence be ween he F18 2013 image (in Digi al Numbe s om 0 o 63) and he F16 2008 image (in he same uni s). The di e ences ha e
a ange om −6 o 31 DN. The expansion o ligh a ound Nai obi is ela ed o o he in as uc u e p ojec s, many o which a e also Chinese- inanced bu no highligh ed he e.
Be ween 2008 and 2013, he geog aphical a eas wi hin a 4 km bu e o he highway expe ienced a 27 pe cen educ ion in he spa ial concen a ion o nigh ime ligh in ensi y.
Spa ial concen a ion, as measu ed by he Gini coe icien in oduced la e in his pape , ell om 0.425 in 2008 o 0.31 in 2013 in he 4 km bu e . A he same ime, land
alues doubled in Thika and ose e en as e in a eas close o Nai obi (like Kasa ani, whe e land alues inc eased om US$46,000 pe ac e o US$500,000 pe ac e), a mga e
p ices o dai y p oduc s and ho icul u e ose due o inc eased access o ma ke s, and ade and in es men alongside he oad co ido expanded (KARA and CSUD,2012;A ican
De elopmen Bank,2014b,2016,2019).
Ou esul s show ha egions which a e equen ecipien s o
p ojec s ecei e la ge amoun s o Chinese go e nmen inancing in
yea s o o e p oduc ion han egions ha in equen ly ecei e Chinese
go e nmen - inanced anspo p ojec s. This di e ence p esumably
occu s because exis ing local capaci y and ela ionships make i easie
o implemen addi ional p ojec s. This es ima e can be in e p e ed
as a di e ence-in-di e ences es ima e, simila o hose epo ed in
he ‘‘China shock’’ o aid and con lic li e a u es (e.g., Au o e al.,
2013;Nunn and Qian,2014). We essen ially compa e he e ec s o
Chinese anspo p ojec s induced by annual changes in he p oduc ion
o aw ma e ials in subna ional locali ies wi h a high p obabili y o
ecei ing such p ojec s and subna ional locali ies wi h a low p obabili y
o ecei ing such p ojec s.
We ind ha Chinese- inanced connec i e in as uc u e p ojec s
educe spa ial concen a ion wi hin i s -o de egions and accele a e
he di usion o economic ac i i y a ound ci ies (in line wi h Fig. 1
and Baum-Snow,2007). Speci ically, we ind ha he Gini coe icien
measu ing he spa ial concen a ion o economic ac i i y is educed by
abou 2.2 pe cen age poin s. Simila speci ica ions o concen a ion
be ween egions sugges e ec s o simila magni ude. Howe e , hese
e ec s a e less p ecisely es ima ed, sugges ing ha Chinese- inanced
in as uc u e canno be obus ly linked o changes in he concen a ion
o economic ac i i y ac oss egions. The absence o a ‘‘be ween e ec ’’
is in line wi h se e al ecen s udies ha emphasize ha anspo
in as uc u e has he e ogeneous and con ex -speci ic impac s on he
dis ibu ion o economic ac i i y ac oss egions (Baum-Snow e al.,
2020;He e al.,2020;Jedwab and S o eyga d,2022;Fajgelbaum and
Redding,2022). Ou main esul s hold unde a ious pe u ba ions,
such as he choice o con ol a iables o s eng hen iden i ica ion o
a ia ions o he ins umen .
Consis en wi h p edic ions om land use heo y, we ind ha
anspo p ojec s shi ac i i y om densely de eloped loca ions o less
densely de eloped ones, ha is, om he highes quin ile o he ligh
dis ibu ion o lowe quin iles. We ind no e idence ha in as uc u e
p ojec s inc ease he ac ion o illumina ed pixels in a ecipien egion
o ha a egion’s ligh pe capi a inc eases. Howe e , ou esul s show
a no able inc ease in o e all ligh in ensi y, de ined as he sum o
ligh emissions wi hin a egion ela i e o i s geog aphic a ea. This
sugges s ha egions expe ience posi i e popula ion g ow h in esponse
o Chinese- inanced in as uc u e p ojec s. The esul s also show ha
he impac o hese p ojec s on he concen a ion o ac i i y wi hin
egions is he e ogeneous. Chinese- inanced anspo in as uc u e e-
duces concen a ion mo e s ongly in egions wi h mo e u ban a eas,
low a el ime o ci ies, and highe oad densi y. We ake his as
indi ec e idence sugges ing ha ou esul s a e d i en by a eloca ion
o wo ke s o he ou ski s o ci ies a he han an inc ease in economic
ac i i y in pe iphe al ci ies o a egion. We also p o ide e idence
ha hese e ec s a e la ges in A ican coun ies and poo e egions
Jou nal o U ban Economics 145 (2025) 103730
3
R. Bluhm e al.
wi hin de eloping coun ies, which end o expe ience apid popula ion
g ow h and ha e a high demand o in as uc u e.
The emainde o he pape p oceeds as ollows. Sec ion 2b ie ly
discusses wha heo y and he exis ing empi ical li e a u e sugges
abou he ela ionship be ween anspo p ojec s and he spa ial con-
cen a ion o economic ac i i y wi hin and ac oss egions. Sec ion 3in-
oduces a subna ionally geo e e enced da ase o Chinese go e nmen -
inanced p ojec s a ound he wo ld and discusses he emo ely-sensed
measu e o spa ial concen a ion.10 Sec ion 4desc ibes he empi i-
cal s a egy. Sec ion 5p esen s and discusses he esul s. Sec ion 6
concludes.
2. T anspo in as uc u e and he concen a ion o economic
ac i i y
U ban land use heo y sugges s ha anspo in as uc u e should
educe spa ial concen a ion wi hin subna ional egions i hese ju is-
dic ions p ima ily consis o u ban a eas and hei su oundings. This
is a key p edic ion o he canonical monocen ic ci y model (Alonso
e al.,1964;Mills,1967;Mu h,1969), in which all wo ke s commu e
o a single loca ion in a cen al business dis ic (CBD). In his model,
agglome a ion bene i s and en s a e highes in he ci y cen e bu
decline wi h dis ance om he CBD. Ini ially, many people choose o
li e nea he cen e and pay highe en s o educe hei commu ing
imes. Subsequen in es men s in anspo a ion in as uc u e inc ease
anspo a ion speed, educe commu ing cos s, and inc ease he supply
o eadily accessible land, shi ing his g adien ou wa ds. T anspo a-
ion in as uc u e hus acili a es u ban sp awl – he low o people
ou o he ci y cen e – by u ning a ci y’s ag icul u al su oundings
in o aluable loca ions o li e in. The model also implies ha people
should sp ead ou along newly c ea ed highways (Baum-Snow,2007),
jus as we documen abo e o he case o Nai obi.11
Reali y is no as s ylized and mos ci ies a e no monocen ic bu
a he cha ac e ized by a CBD ha coexis s wi h o he employmen sub-
cen e s. Polycen ic ci ies a ise when he loca ion o employmen and
comme cial ac i i y a e de e mined endogenously wi hin he ci y ( o
ea ly con ibu ions, see Ogawa and Fuji a,1980;Fuji a and Ogawa,
1982;Hende son and Mi a,1996). In polycen ic ci ies, agglome a ion
economies a e smalle ela i e o he monocen ic benchma k, bu
commu ing cos s – hence, dispe sion o ces – a e also lowe . New
quan i a i e spa ial models allow o complex pa e ns o agglome a-
ion and dispe sion o ces, which can lead o a ious u ban o ms (see
Redding and Rossi-Hansbe g,2017, o a e iew). In hese models,
loca ions a e no independen bu connec ed ia ade, commu ing,
and mig a ion lows. Fi ms and esiden s make op imal choices based
on hese connec ions. The p edic ions o he amewo ks depend on
hese spa ial in e ac ions and, hence, he da a. In as uc u e imp o e-
men s can educe he concen a ion o ac i i y by making pe iphe al
a eas mo e accessible, bu hey can also inc ease concen a ion by
expanding he labo ma ke o cen al i ms. Fo example, ea ly in-
as uc u e imp o emen s, such as he s eam ailway, pe mi ed a
specializa ion o ci y cen e s in o wo kplaces and hei su oundings
in o esidences (Heblich e al.,2020).
10 We p o ide ou code, mos da a, and a desc ip ion o how o ob ain he
emaining da a equi ed o eplica e all exhibi s in ou pape in Bluhm e al.
(2024).
11 Fi ms ha e di e en incen i es han esiden s since hey ace a mo e
complex se o cos s when lea ing ci y cen e s. They ade agglome a ion
bene i s agains a ious cos s ( anspo a ion cos s o ou pu , in e ac ion
cos s, labo accessibili y, e c.). This gi es ise o a pa e n whe e i ms ha
bene i less om ace- o- ace in e ac ion and knowledge spillo e s, such as
manu ac u ing i ms, decen alize mo e han o he i ms (Rossi-Hansbe g e al.,
2009;Baum-Snow,2014).
A la ge and g owing body o empi ical e idence suppo s he basic
p edic ion o he monocen ic ci y model ha new o upg aded ans-
po a ion in as uc u e dispe ses economic ac i i y away om u ban
agglome a ions. Conside subu baniza ion in 20 h-cen u y Ame ica,
whe e esea che s ha e documen ed u ban sp awl and s ong popula-
ion g ow h in ci ies wi h mo e de elopable su oundings (Bu ch ield
e al.,2006;Saiz,2010). The cons uc ion o highways in he Uni ed
S a es d ama ically lowe ed commu ing imes and inc eased demand
o subu ban ela i e o u ban esiden ial space (Baum-Snow,2007).
Conges ion also plays an impo an ole, pa icula ly in con empo a y
se ings. Allen and A kolakis (2022) highligh ha s onge conges ion
no only educes he ma ke access o cen al loca ions and shi s
ac i i y away om he co e bu ha his e ec inc eases wi h ci y size.
The e is also e idence o simila p ocesses o di usion a ound
u ban a eas in de eloping coun ies (e.g., Bayes,2007;Zá a e,2022).
Baum-Snow e al. (2017) examine he e ec o oad and ailway in as-
uc u e on he spa ial dis ibu ion o economic ac i i y in China and
ind ha ing oad in es men s displaced 50 pe cen o indus ial GDP
om cen al ci ies o ou lying a eas. As Chinese- inanced in as uc u e
p ojec s in de eloping coun ies o en ep esen a subs an ial p opo -
ion o local in as uc u e in es men in a gi en yea , we simila ly
expec hem o decen alize economic ac i i y a ound u ban a eas. We
also expec hese p ojec s o spu he o ma ion o new sub-cen e s
as businesses, wo ke s, and o he economic ac o s eloca e in o he
pe iphe y.
T anspo p ojec s may also a ec he concen a ion o economic ac-
i i y ac oss egions, a cen al conce n o economic geog aphy esea ch.
The classic co e–pe iphe y model s esses he ole o inc easing e u ns
o scale when economic ac i i y s a s o concen a e in a pa icula
egion. When ade cos s a e high o p ohibi i e, i ms a e sp ead ou
e enly ac oss egions o loca e hemsel es close o consume demand.
When anspo p ojec s inc ease connec i i y be ween leading and
lagging egions, labo , and capi al should mo e om he pe iphe y o
he be e -connec ed co e, c ea ing a co e–pe iphe y spli un il he e
is almos comple e specializa ion (K ugman,1991). Howe e , some o
hese o ces e e se a high le els o concen a ion. Puga (1999), o
example, shows how a lack o mig a ion wi h low ade cos s implies
ha i ms will again loca e close o inal demand. This gi es ise o
a bell shape o egional inequali ies in ela ion o ade cos s. The
ad an age o being in a cen al loca ion well-connec ed wi h o he
ma ke s e odes when e y low ade cos s make i easy o each he
pe iphe y. While he bell-shaped cu e is a obus p edic ion, i equi es
addi ional he e ogenei y in ag icul u al ade cos s, u ban conges ion,
o mig a ion decisions ha make he o e all ela ionship di icul o
iden i y.
Empi ical e idence on how anspo cos s shape egional concen-
a ion e lec s his he e ogenei y. B ülha e al. (2020) show ha he
ad an ages o ma ke po en ial a e sh inking in he de eloped wo ld
bu emain an impo an de e minan o employmen g ow h in de el-
oping coun ies. Wi h espec o in as uc u e in es men s, Bi d and
S aub (2014) ind ha in es men s in B azil’s oad ne wo k inc eased
economic agglome a ion in he al eady p ospe ous popula ion cen e s
o he Sou h, while also acili a ing economic agglome a ion in less
de eloped a eas o he No h. On balance, hese in es men s educed
spa ial inequali y ac oss he coun y’s municipali ies.12 Howe e , Fabe
(2014) p o ides e idence ha China’s Na ional T unk Highway Sys em
– a majo in e - egional anspo a ion in as uc u e p ojec – educed
le els o economic ac i i y in newly connec ed pe iphe al egions el-
a i e o non-connec ed pe iphe al egions. Gi en hese mixed indings
and he c oss-na ional scope o ou s udy, we do no ha e s ong easons
12 In a ela ed s udy o A gen ina’s s eam ail oad ne wo k and he ag icul-
u al sec o , Fajgelbaum and Redding (2022) sugges ha lowe anspo cos s
can enable economic ac o s loca ed in emo e, in e io egions o pa icipa e
in s uc u al ans o ma ion.
Jou nal o U ban Economics 145 (2025) 103730
4

R. Bluhm e al.
o belie e ha Chinese- inanced anspo a ion p ojec s will uni o mly
inc ease o dec ease concen a ion be ween egions.
De eloping coun ies a e an impo an and use ul applica ion o
hese heo ies. Mos de eloping coun ies ace majo anspo a ion
in as uc u e gaps in bo h u ban and u al egions. In e nal anspo
cos s a e ou o i e imes highe wi hin E hiopia o Nige ia han
wi hin he Uni ed S a es (A kin and Donaldson,2015). The o al leng h
o he oad ne wo k pe 1000 people is oughly 10 imes lowe o
Sou h Asia, Eas Asia and he Paci ic, Sub-Saha an A ica, and he Mid-
dle Eas and No h A ica han o No h Ame ica (And és e al.,2014).
Many de eloping coun ies ha e bo h apidly expanding popula ions
and unde unded, poo ly designed anspo a ion sys ems (Ce e o,
2013). Majo in as uc u e inancing gaps make i di icul o de elop-
ing coun ies o o e come he spa ial bo lenecks c ea ed by high le els
o u ban concen a ion and u al neglec .
Wha is mo e, subna ional egions wi hin de eloping coun ies a e
o en de ined by dense cen al ci ies su ounded by unde de eloped
hin e lands.13 La ge ci ies in many A ican coun ies, o example,
end o be highly conges ed ela i e o o e all le els o in as uc u e,
indus y, and economic oppo uni y (Lall e al.,2017), while seconda y
ci ies end o be isola ed om wo ld ma ke s (Gollin e al.,2016). In
hese se ings, high le els o u ban conges ion in de eloping economies
a e a la en o ce o spa ial dispe sion as new anspo a ion op ions
become a ailable.
3. Da a
New geocoded da ase o Chinese go e nmen - inanced p ojec s
The Chinese go e nmen conside s he de ails o i s o e seas de-
elopmen p og am o be a ‘‘s a e sec e ’’ (B äu igam,2009, p. 2). I
does no publish a coun y-by-coun y b eakdown o i s aid expen-
di u es o ac i i ies. No does i sys ema ically publish p ojec -le el
da a on i s less concessional and mo e comme cially-o ien ed inan-
cial expendi u es and ac i i ies in de eloping coun ies. To o e come
his challenge, D ehe e al. (2021b) collabo a ed wi h AidDa a, a
esea ch lab a William & Ma y, o build a global da ase o Chinese
go e nmen - inanced p ojec s commi ed be ween 2000 and 2014. This
p ojec -le el da ase uses a publicly documen ed me hod called T ack-
ing Unde epo ed Financial Flows (TUFF) o acili a e he collec ion
o de ailed and comp ehensi e inancial, ope a ional, and loca ional in-
o ma ion abou Chinese go e nmen - inanced p ojec s (S ange e al.,
2017,2018). The TUFF me hod iangula es in o ma ion om ou
ypes o open sou ces – English, Chinese, and local-language news
epo s; o icial s a emen s om Chinese minis ies, embassies, and eco-
nomic and comme cial counselo o ices; he aid and deb in o ma ion
managemen sys ems o inance and planning minis ies in coun e pa
coun ies; and case s udy and ield esea ch unde aken by schola s
and non-go e nmen al o ganiza ions (NGOs) – in o de o minimize he
impac o incomple e o inaccu a e in o ma ion.14 Economis s, poli ical
scien is s, and compu a ional geog aphe s ha e o he mos pa used
hese da a o explain he na u e, alloca ion, and e ec s o Chinese
go e nmen - inanced p ojec s in A ica and in se e al coun y-speci ic
s udies ou side o A ica (e.g., He nandez,2017;D ehe e al.,2018;
Isaksson and Ko sadam,2018a,b;Anaxago ou e al.,2020;Isaksson,
13 As Baum-Snow e al. (2017) poin ou , he u ban dis ibu ion o eco-
nomic ac i i y in many de eloping coun ies oday la gely esembles ha o
ea ly 20 h cen u y Ame ica, in which indus y was ini ially o e whelmingly
concen a ed in u ban cen e s.
14 The me hod is o ganized in h ee s ages: wo s ages o p ima y da a
collec ion (p ojec iden i ica ion and sou ce iangula ion) and a hi d s age
o e iew and e ise indi idual p ojec eco ds (quali y assu ance). The TUFF
da a collec ion and quali y assu ance p ocedu es a e desc ibed a leng h
in S ange e al. (2017,2018).
2020;Ma o ano e al.,2020;D ehe e al.,2021b;Eichenaue e al.,
2021;Ho n e al.,2021;Geh ing e al.,2022;Baeh e al.,2023).
In his pape , we build on hese p ojec -le el da a o c ea e a i s -
o -i s-kind da ase o Chinese g an - and loan- inanced de elopmen
p ojec loca ions a ound he globe. In con as o p e ious e sions, ou
new da a enable subna ional analyses o Chinese- inanced p ojec s in
i e egions o he wo ld (A ica, he Middle Eas , Asia and he Paci ic,
La in Ame ica and he Ca ibbean, and Cen al and Eas e n Eu ope)
o e 15 yea s (2000–2014). Ou da ase akes all p ojec s om D ehe
e al. (2021b) ha secu ed inancial commi men s om China and
en e ed implemen a ion o eached comple ion as a s a ing poin .15
We hen subjec all o hese p ojec s o a double-blind geocoding p o-
cess (S andow e al.,2011), in which wo ained code s independen ly
employ a de ined hie a chy o geog aphic e ms and assign uni o m
la i ude and longi ude coo dina es and s anda dized place names o
each loca ion whe e he p ojec in ques ion was ac i e. Code s also
speci y a p ecision code o each loca ion. P ecision code 1 co esponds
o an exac loca ion; p ecision code 2 co esponds o loca ions wi hin
25 kilome e s o he exac p ojec si e; p ecision code 3 co esponds
o a second-o de egion; and p ecision code 4 co esponds o a i s -
o de egion.16 I he coo dina es and p ecision codes do no ma ch, a
senio ‘‘a bi a o ’’ iden i ies he sou ce o he disc epancy and assigns
a inal se o geocodes o all si es. This double-blind coding p ocess
aims o minimize he isk o missed o inco ec loca ions.17 In o al,
he esul ing da ase co e s 3485 p ojec s (wo h a leas US$273.6
billion in cons an 2014 dolla s) in 6184 disc e e loca ions ac oss 138
coun ies.18
To me ge hese geocoded p ojec da a wi h ou ou come measu es
o spa ial concen a ion wi hin and ac oss subna ional egions, we
agg ega e all p ojec s wi h p ecision codes 1–4 o i s -o de egions.
Fig. 2shows he loca ions o p ojec s ha can be placed wi hin i s -
o de egions o e he 2000–2014 pe iod. The esul ing subsample
includes 2140 Chinese go e nmen - inanced p ojec s a 4420 disc e e
loca ions (collec i ely wo h US$201 billion) ha we e comple ed o
being implemen ed in 883 i s -o de egions wi hin 129 coun ies
be ween 2000 and 2014.19 Ou da a can be disagg ega ed by inan-
cial low class and sec o . Wi h espec o he o me , we dis inguish
be ween O icial De elopmen Assis ance (ODA) and o he o ms o
concessional and non-concessional inancing om Chinese go e nmen
15 We build on ea lie geo e e enced da ase s ha co e A ica, he T opical
Andes, and he Mekong Del a o ewe yea s only (BenYishay e al.,2016;
D ehe e al.,2019a). No e ha we exclude all suspended and canceled p ojec s
as well as p ojec s ha eached he (non-binding) pledge s age o (binding)
o icial commi men s age bu ne e eached implemen a ion o comple ion
du ing he pe iod o s udy (2000–2014).
16 We exclude all p ojec s om he eg ession analysis below ha could no
be geocoded wi h su icien spa ial p ecision o be included in he egional-
le el da a (e.g., coun y-wide p ojec s, see h ps://docs.aidda a.o g/ad4/ iles/
geocoding-me hodology-upda ed-2017-06.pd o de ails).
17 No e ha he poin -based me hod used o geocode hese p ojec s is no
designed o measu e he exac linea pa h o anspo a ion in as uc u e. This
implies ha one canno ‘connec he do s’ and look o e ec s alongside he
oads, ailways, e c. Howe e , i is use ul o measu ing he e ec s wi hin
ea ed subna ional egions, as we do in he p esen pape .
18 Fo compa ison, he A ica-speci ic da a p o ided in D ehe e al. (2019a)
include 1650 p ojec s ac oss 2969 loca ions in he 2000–2012 pe iod. No e
ha , in con as o ou da ase , hey also co e p ojec s ha ha e no (ye )
eached implemen a ion s age.
19 We only ocus on low-income and middle-income coun ies. Mo e
p ecisely, we include coun ies ha he Wo ld Bank does no classi y as high-
income coun ies in a gi en yea (see h ps://da ahelpdesk.wo ldbank.o g/
knowledgebase/a icles/906519-wo ld-bank-coun y-and-lendingg oups, las
accessed Sep embe 13, 2023). We also exclude small s a es wi h a popula ion
size below a h eshold o 1,000,000 inhabi an s. Table A-1 in he Online
Appendix lis s all coun ies included in he analysis.
Jou nal o U ban Economics 145 (2025) 103730
5
R. Bluhm e al.
ins i u ions.20 Fo he pu poses o he la e , we use he OECD’s h ee-
digi sec o classi ica ion scheme, which ca ego izes p ojec s acco ding
o hei p ima y objec i es.21
269 p ojec s we e assigned o he anspo and s o age sec o
and implemen ed in 1215 loca ions. The combined alue o hese
p ojec s was a leas US$64.1 billion (when coun ing hose p ojec s
o which inancial alues a e a ailable) and amoun s o an a e age
o US$224.9 million pe p ojec .22 This is sizable in ligh o an a e age
GDP o abou US$4.4 billion pe i s -o de egion. The as majo i y
o hese p ojec s ocused on building anspo a ion in as uc u e, such
as oads, ailways, b idges, seapo s, and ai po s. Wi h 651 p ojec
loca ions, long-dis ance oads a e mos equen , ollowed by long-
dis ance ailways (245) and u ban oads (123) (see Table A-2 in he
Online Appendix).23
In ou da ase , he a e age inancial commi men om China o a
long-dis ance oad p ojec was US$231.2 million compa ed o
US$979.3 million o a long-dis ance ailway p ojec and US$148.3
million o an u ban oads p ojec . To be e unde s and he na u e
o hese p ojec s, conside h ee p ojec s om h ee di e en egions
ha a e b oadly indica i e o he ypes o ‘‘ ea men s’’ ha we analyze
in ou s a is ical analysis. Fi s , in Oc obe 2014, China Eximbank
p o ided a US$943.9 million loan o Mon eneg o’s Minis y o Fi-
nance o help inance he cons uc ion o a 169 km highway be ween
Ba ( he coun y’s main seapo in he sou h) and Bolja e on he
Mon eneg in-Se bian bo de in he coun y’s no h.24 The loan was
wo h app oxima ely a qua e o he coun y’s GDP a he ime ha i
was con ac ed. Upon comple ion, he highway was expec ed o educe
a el ime be ween he capi al o Podgo ica and he no he n ci y o
Kolašin om 90 min o 30 min and acili a e economic de elopmen
alongside he anspo co ido . Second, in May 2013, China Eximbank
issued a US$491.7 million loan o Djibou i’s Minis y o Finance o he
cons uc ion o a 100 km segmen o a 756 km ailway ha uns om
Addis Ababa, he capi al o E hiopia, o he Do aleh seapo in Djibou i.
A he ime o i s issuance, he loan ep esen ed app oxima ely 40% o
Djibou i’s GDP. Upon comple ion, he ailway was expec ed o educe
a el ime be ween Addis Ababa and Do aleh seapo om 7 days (on
oads) o 10 h.25 Thi d, in 2009, China Eximbank p o ided a US$44.2
million loan o he Go e nmen o Tonga o phase 3 o he Na ional
Road Imp o emen P ojec . The loan was wo h oughly 16% o GDP
20 Mo e p ecisely, we code all Chinese go e nmen - inanced p ojec s as
O icial De elopmen Assis ance (‘‘ODA-like’’), O he O icial Flows (‘‘OOF-
like’’), o ‘‘Vague O icial Finance.’’ Chinese ODA-like p ojec s a e inanced by
Chinese go e nmen ins i u ions wi h de elopmen in en and a minimum le el
o concessionali y (a 25 pe cen o highe g an elemen ). Chinese OOF p ojec s
a e inanced by Chinese go e nmen ins i u ions wi h comme cial o ep esen-
a ional in en and/o lack a g an elemen o 25 pe cen o mo e. P ojec s
assigned o he Vague O icial Finance ca ego y a e Chinese go e nmen -
inanced p ojec s whe e he e is insu icien in o ma ion in he public domain
abou concessionali y and/o in en o clea ly de e mine whe he he lows a e
mo e akin o ODA o OOF. To al Chinese O icial Finance (OF) is, he e o e, he
sum o all p ojec s coded as ODA-like, OOF-like, o Vague (O icial Finance).
Fo mo e de ailed discussion o he dis inc ion be ween hese ypes o Chinese
de elopmen inance, see D ehe e al. (2018).
21 The e a e 24 o hese OECD sec o codes (see www.oecd.o g/
de elopmen / inancing-sus ainable-de elopmen /de elopmen - inance-
s anda ds/pu posecodessec o classi ica ion.h m o de ails).
22 Many o hese p ojec s a e implemen ed in mul iple loca ions, o exam-
ple, when hey connec wo ci ies wi h a ailway line. The a e age amoun
commi ed pe p ojec loca ion is US$49.8 million.
23 T anspo p ojec s a e he ones we exploi o mos o ou analyses. We
also use a la ge sample o p ojec s ha suppo ed economic in as uc u e
and se ices, which includes oads, ailways, b idges, seapo s, and ai po s
bu also powe g ids, powe lines, cell phone owe s, and ibe op ic cable
lines (514 p ojec s a 1897 loca ions wi h a alue o abou US$165 billion).
24 See h ps://china.aidda a.o g/p ojec s/42330/ o de ails on he p ojec .
25 See h ps://china.aidda a.o g/p ojec s/46183/ o de ails.
a he ime ha i was con ac ed. The pu pose o he p ojec was o
imp o e he 1 km Alipa e Road on he island o Tonga apu, he 8 km
oad ha uns om Kolonga o Talasiu on he island o Tonga apu, and
he 4 km oad ha uns om Ha eluliku o La enga onga on he island
o Tonga apu.26
Fig. 2illus a es he global each o China’s o e seas de elop-
men p og am in he 21s cen u y. Consis en wi h ea lie pe iods o
Chinese aid gi ing (D ehe and Fuchs,2015), Chinese p ojec s co e
almos all de eloping coun ies (wi h coun ies ecognizing he Chinese
go e nmen in Taiwan as a no able excep ion).27 Chinese- inanced
de elopmen p ojec s a e densely concen a ed in A ican and Asian
coun ies. The igu e also illus a es ha many Chinese go e nmen -
inanced p ojec s a e si ua ed in coas al egions, including some o he
highes - alue anspo a ion p ojec s.
Measu ing concen a ion wi hin and ac oss subna ional egions
Reliably measu ing local economic ac i i y ac oss he globe wi h
o icial da a is di icul . Few coun ies collec and epo comp ehensi e
da a a he indi idual o plan /es ablishmen le el a egula in e als,
and subna ional GDP da a a e gene ally only a ailable in highly de el-
oped coun ies. To ci cum en his p oblem, we ollow p e ious wo k
ha uses nigh ime ligh in ensi y as a p oxy o local economic ac i -
i y (Hende son e al.,2012;Hodle and Raschky,2014;Michalopoulos
and Papaioannou,2014). While nigh ime ligh s we e ini ially p oposed
as a measu e o income o coun ies wi h weak s a is ical capaci y,
hey we e quickly adop ed mo e b oadly as a measu e o subna ional
economic ac i i y in de eloping coun ies. Subsequen s udies ha e
demons a ed ha changes in ligh emissions co ela e s ongly wi h
adi ional wel a e measu es down o he illage le el (Weidmann and
Schu e,2017;B uede le and Hodle ,2018).
We ollow Hende son e al. (2018), who use nigh ime ligh in en-
si y a he g id-cell le el as a measu e o agg ega e economic ac i i y
– i.e., he p oduc o popula ion and ligh ou pu pe capi a – and hen
calcula e a spa ial Gini coe icien based on he dis ibu ion o his
p oxy o o al GDP. While we a e p ima ily in e es ed in whe he and
o wha ex en in as uc u e in es men s eloca e economic ac i i y,
we also in es iga e below whe he such in es men s inc ease ou pu
pe capi a.
We ob ain da a on nigh ime ligh in ensi y om he De ense Me eo-
ological Sa elli e P og am’s (DMSP) Ope a ion Line Scan sa elli es. The
DMSP sa elli es ci cle he ea h in sun-synch onous o bi and eco d
e ening ligh s be ween 8:30 and 9:30 pm on a 6-bi scale anging
om 0 o 63. The Na ional Oceanic and A mosphe ic Adminis a ion
(NOAA) p ocesses hese da a, c ea es annual composi es o he daily
images a a esolu ion o 30 a c seconds and makes hem a ailable o
he gene al public. We use he so-called ‘‘s able ligh s’’ p oduc , which
il e s ou mos backg ound noise, o es i es, and s ay ligh s. E en
hough he e a e well-known issues in hese da a wi h bo om and op
coding (see Jean e al.,2016;Bluhm and K ause,2022), nigh ime
ligh s a e measu ed in a consis en manne a ound he globe and a oid
many o he measu emen e o s in ol ed in mo e adi ional su ey
da a.
We p oceed in ou s eps o calcula e ou measu e o spa ial concen-
a ion. Fi s , we di ide he en i e wo ld in o a g id o 6 a c minu e cells
(i.e., an a ea o abou 9.3 km by 9.3 km a he equa o ) and align he
g id wi h ligh s da a.28 Second, we in e sec his g id wi h he global
i s -o de adminis a i e bounda ies, which c ea es egula cells in he
26 Fo de ails, see h ps://china.aidda a.o g/p ojec s/39199/.
27 Fo ecen wo k ha s udies he alloca ion o China’s de elopmen inance
ac oss coun ies, see D ehe e al. (2022) o Hoe le and S e ck (2022).
28 Al hough he nominal esolu ion o he DMSP-OLS sys em is 30 a c
seconds, geoloca ion e o s and on-boa d p ocessing o ine- esolu ion pixels
lead o a ue g ound oo p in o 5 km by 5 km (El idge e al.,2013).
Taking abou wice his esolu ion educes he in luence o his mechanical
Jou nal o U ban Economics 145 (2025) 103730
6
R. Bluhm e al.
Fig. 2. Loca ions o Chinese- inanced p ojec s, anspo and non- anspo , 2000–2014. No es: The igu e iden i ies all Chinese- inanced anspo ( ed) and non- anspo p ojec s
(g ay) which we e commi ed and implemen ed in he pe iod om 2000 o 2014. I shows a o al o 2140 p ojec s in 4420 disc e e loca ions which ha e a p ecision accu acy o
(a leas ) a i s -o de adminis a i e di ision. 1345 p ojec s ha e a p ecision accu acy less han he i s -o de egion (no shown). Al hough he e a e ‘‘only’’ 269 anspo a ion
p ojec s, 1211 o he 4420 loca ions shown in he igu e ha e di ec ly ecei ed (some pa ) o a la ge anspo a ion p ojec . (Fo in e p e a ion o he e e ences o colo in his
igu e legend, he eade is e e ed o he web e sion o his a icle.)
in e io and ‘‘squiggly’’ cells along he egional bo de s.29 Thi d, o all
egula and squiggly cells in his g id and all yea s in he nigh ligh s
da a, we compu e he sum o ligh (𝑠𝑙), he land a ea o each cell in km2
(𝑎𝑙), and he ligh in ensi y in he cell (𝑥𝑙=𝑠𝑙∕𝑎𝑙).30 We a e age he
esul ing ligh in ensi ies whene e mo e han one sa elli e is a ailable
and u n o all pixels ha do no all on land be o e agg ega ing he
ligh s o he g id le el. Finally, we compu e he Gini coe icien o ligh
in ensi ies (on a 0–1 scale) o e all cells (including cells wi h ze o ligh
in ensi y) wi hin an adminis a i e egion as
Gini =∑𝑛
𝑙=1 𝑤𝑙∑𝑛
𝑚=1 𝑤𝑚|𝑥𝑙−𝑥𝑚|
2∑𝑛
𝑙=1 𝑤𝑙∑𝑛
𝑙=1 𝑤𝑙𝑥𝑙
,(1)
whe e 𝑤𝑙=𝑎𝑙
∑𝑛
𝑙=1 𝑎𝑙
is an a ea-based weigh and 𝑛is he o al numbe o
cells in a egion. We also cons uc Gini coe icien s o concen a ion
be ween i s -o de egions. The o mula emains he same, only ha i
is based on he a e age ligh in ensi y o a egion (swapping 𝑥𝑙 o 𝑥𝑙),
and 𝑤𝑙is hen de ined as he land a ea o he en i e egion.
Ou spa ial Gini coe icien can be in e p e ed as he a e age
(weigh ed) di e ence be ween he ligh in ensi ies o all possible pai s
o cells wi hin an adminis a i e egion. Geome ically, i is he a ea
unde he Lo enz cu e plo ing he cumula i e dis ibu ion o weigh ed
ligh in ensi ies agains he cumula i e dis ibu ion o cell a eas (in
km2). Including cells wi h ze o ligh in ensi y means he Lo enz cu e
will emain a ze o be o e sloping up o one, bu ensu es ha he
Gini coe icien is a p ope measu e o economic concen a ion, which
no only dec eases when he dis ibu ion o ligh becomes mo e equal
among al eady illumina ed cells bu also when new cells become
illumina ed. As can be seen om he long di e ences in he spa ial Gini
coe icien p esen ed in a wo ld map o i s -o de egions in Fig. 3, ou
dependen a iable shows conside able a ia ion o e he ime pe iod
unde analysis, bo h wi hin and ac oss coun ies (2000–2013).
spa ial au oco ela ion, educes he in luence o op coding and bo om coding,
and limi s he compu a ional bu den. The newe Visible In a ed Imaging
Radiome e Sui e (VIIRS) da a, which ha e supe io echnical p ope ies, do
no span a signi ican po ion o ou sample.
29 We ob ained he egional bo de s om he Da abase o Global Adminis-
a i e A eas (GADM) ec o da ase ( e sion 2.8). We used he same da a o
geocode he Chinese- inanced p ojec s.
30 Di iding by he land a ea adjus s o he ac ha 6 a c minu e cells do
no ha e a uni o m a ea ac oss he globe and may be co e ed by wa e . We
calcula e he land a ea o each cell using he G idded Popula ion o he Wo ld
( 4) land/wa e as e .
I is impo an o emphasize ha he Gini coe icien cap u es he
o e all dispe sion o economic ac i i y, which is a p oduc o he popu-
la ion dis ibu ion and he dis ibu ion o ligh pe capi a.31 Hende son
e al. (2018) show ha he c oss-sec ional a ia ion in popula ion
densi y ac oss adminis a i e egions is subs an ially la ge han he
a ia ion in income pe capi a. I his holds ac oss ime, hen a signi -
ican p opo ion o obse ed changes in he wi hin- egion dis ibu ion
o ligh in ensi ies should be a ibu able o shi s in he popula ion dis-
ibu ion a he han di e ences in pe -capi a income. This is p ecisely
he ype o a ia ion we a e in e es ed in and expec o be a ec ed by
anspo in as uc u e in es men s.
We p e e using nigh ime ligh s o e da a o popula ion densi y as
ou main ou come measu e. Popula ion da a a compa able esolu ions
– such as he Global Human Se lemen Laye , G idded Popula ion
o he Wo ld, o Landscan – a e based on a ely a ailable censuses,
which a e hen disagg ega ed in space and in e pola ed o e ime. They
would no allow us o exploi annual a ia ion in he commi men o
anspo p ojec s and changes in economic ac i i y, which a e he basis
o ou iden i ica ion s a egy. Census da a a e also less equen ly a ail-
able in poo e de eloping coun ies ha hos many Chinese- inanced
de elopmen p ojec s.
4. Empi ical s a egy
We a e in e es ed in changes in he spa ial concen a ion o eco-
nomic ac i i y caused by Chinese in as uc u e in es men s. Deno ing
i s -o de adminis a i e egions by 𝑗, coun ies by 𝑖,32 and yea s by 𝑡,
ou main equa ion ela es ou luminosi y-based measu e o spa ial con-
cen a ion, Gini𝑗 𝑖𝑡, o he o al numbe o yea s in which anspo a ion
p ojec s ha e been commi ed o a egion up o 𝑡− 2, deno ed 𝑁𝑗 𝑖,𝑡−2.
We chose 𝑁𝑗 𝑖,𝑡−2 as he baseline ea men because anspo a ion
p ojec s can a y widely in ype and size, anging om small b idges o
31 To see his, conside ha 𝑥𝑖is de ined as 𝑝𝑖
𝑎𝑖
×𝑠𝑖
𝑝𝑖
, whe e 𝑝𝑖
𝑎𝑖
is popula ion
densi y and 𝑠𝑖
𝑝𝑖
is ligh pe capi a in each cell.
32 Going below he i s -o de le el would change many sample cha ac-
e is ics. We would lose a signi ican sha e o p ojec s ha ha e only been
accu a ely coded o i s -o de egions. The exposu e a iable we in oduce
below would also be based on subs an ially ewe p ojec s pe egion. Wha
is mo e, in coun ies wi h a smalle land mass, second-o de egions o en
co espond o adminis a i e bounda ies o ci ies, which would e ec i ely
exclude hei su oundings om a wi hin-uni analysis.
Jou nal o U ban Economics 145 (2025) 103730
7
R. Bluhm e al.
Fig. 3. Long di e ences in spa ial concen a ion, wi hin i s -o de egions, 2000–2013. No es: The igu e illus a es he c oss- egional and empo al a ia ion in spa ial concen a ion.
I shows long di e ences in he Gini coe icien o spa ial concen a ion wi hin i s -o de egions, ha is, a egion’s alue in 2013 minus he alue in 2000. Only coun ies no
classi ied as high-income coun ies by he Wo ld Bank a e shown. Missing alues occu when he e a e oo ew li cells o compu e he Gini coe icien in he ini ial o inal pe iod.
(Fo in e p e a ion o he e e ences o colo in his igu e legend, he eade is e e ed o he web e sion o his a icle.)
(segmen s o ) ex ensi e mul i egional highways.33 To assess he impac
o any in as uc u e p ojec o bundle o p ojec s in a egion in a
gi en yea , his a iable agg ega es o e all yea s in which a leas one
p ojec was commi ed (see he discussion o he di e enced e sion
below). In obus ness checks, we use he numbe o p ojec loca ions
o inancial alues (whe e a ailable). By agg ega ing o e all yea s,
ou speci ica ion assumes ha he le el o concen a ion depends on
he en i e his o y o p ojec s. O cou se, he e ec o p ojec s could
ade o e ime so ha p ojec s commi ed in he dis an pas would no
longe a ec concen a ion. Since ou sample only co e s he pe iod
since 2000, and Chinese de elopmen inance was compa a i ely low
be o e ha ime, we ocus on he medium- e m e ec s o new p ojec s
a he han hose commi ed in he dis an pas . We lag his a iable by
wo yea s o accoun o he di e ence be ween he commi men da e
and he expec ed comple ion da e o a p ojec .34
We begin wi h a lexible speci ica ion ha allows he e ec s o
Chinese- inanced p ojec s o be a bi a ily co ela ed wi h egion-
speci ic ixed e ec s and egion-speci ic ime ends:
Gini𝑗 𝑖𝑡 =𝛽 𝑁𝑗 𝑖,𝑡−2 +𝜇𝑗 𝑖+𝜃𝑗 𝑖×𝑡+𝜆𝑖𝑡 +𝜖𝑗 𝑖𝑡,(2)
whe e 𝜇𝑗 𝑖a e egion- ixed e ec s, 𝜃𝑗 𝑖×𝑡a e egion-speci ic linea ime
ends, and 𝜆𝑖𝑡 a e coun y-yea ixed e ec s ha abso b a a ie y o
po en ial shocks o all egions o a coun y in a pa icula yea .
A mo e ac able and in ui i e e sion o his model can be es i-
ma ed in i s di e ences:
𝛥Gini𝑗 𝑖𝑡 =𝛽 𝛥𝑁𝑗 𝑖,𝑡−2 +𝜃𝑗 𝑖+𝜏𝑖𝑡 +𝛥𝜖𝑗 𝑖𝑡,(3)
whe e 𝜏𝑖𝑡 =𝜆𝑖𝑡 −𝜆𝑖,𝑡−1 ep esen s a new se o coun y-yea ixed e ec s
ob ained ia di e encing. 𝜃𝑗 𝑖now cap u es egion-speci ic ends in
33 Ano he echnical eason is ha p ojec s a e o en co-loca ed, e.g., di -
e en sec ions o a highway, bu may ha e dis inc p ojec IDs in ou da a
because hey a e inanced h ough di e en inancial anches. This does no
necessa ily cap u e he in ensi e ma gin o in as uc u e in es men s bu
e lec s he de ini ions adop ed du ing he geocoding p ocess.
34 S a and end da es a e a ailable o a subse o he p ojec s in ou da ase .
Ac oss hese app oxima ely 1100 Chinese go e nmen p ojec s, he a e age
ime om s a o comple ion is abou 2.1 yea s. His o ical da a on Chinese
de elopmen p ojec s also e eal a median o wo yea s be ween p ojec s a
and comple ion (D ehe e al.,2021b, based on da a om Ba ke,1989). The
wo-yea lag we use o he main analysis hus allows p ojec s o egis e e ec s
di ec ly a e (expec ed) comple ion. As we show below, he e ec s o Chinese
anspo p ojec s ma e ialize quickly. When longe lags a e used, es ima es
become smalle and less p ecisely es ima ed.
le els. In he di e enced o m, i becomes 𝜃𝑗 𝑖×𝑡−𝜃𝑗 𝑖× (𝑡− 1), which
ep esen s he change in he egion- ixed e ec o e ime.
The speci ica ion in di e ences shows ha 𝛽es ima es he pe sis en
e ec o a new anspo p ojec (o a bundle o anspo p ojec s
commi ed in he same yea ) on spa ial concen a ion wo yea s la e .35
The model nes s less lexible app oaches wi h a s ic pa allel- ends
assump ion since all 𝜃𝑗 𝑖could be ze o.
Ou p e e ed measu e o anspo a ion in as uc u e, 𝛥𝑁𝑗 𝑖,𝑡−2,
is hus a bina y a iable indica ing ha a leas one new p ojec is
commi ed o a pa icula egion in a yea . Since he size o he p ojec s
is no homogeneous ac oss loca ions, he e ec s o p ojec s on spa ial
concen a ion migh di e along he in ensi e ma gin. Un o una ely,
we lack in o ma ion on he inancial alues o mo e han a hi d o
hese p ojec s (see D ehe e al.,2021b), which is why we p e e he
bina y indica o and p esen addi ional esul s using ( he log o ) agg e-
ga e dolla alues o compa ison. Mo eo e , we de ine ou dependen
a iable based on commi men yea s a he han ac ual disbu semen
da es as comp ehensi e da a on disbu semen s a e no a ailable and
i ually impossible o es ima e h ough open-sou ce da a collec ion.
Wi h i s di e ences, a wo-yea lag, and a lack o nigh ime ligh s
da a a e 2013 ( om he DMSP-OLS sys em), ou sample e ec i ely
co e s he pe iod om 2002 un il 2013.36
We allow a wide ange o dependency s uc u es o occu in 𝛥𝜖𝑗 𝑖𝑡.
T anspo a ion in as uc u e p ojec s o en connec mo e han one
adminis a i e egion. Clus e ing s anda d e o s on he coun y le el
pe mi s a bi a y spa ial and empo al co ela ion among all egions
wi hin a coun y. To accoun o connec ions ac oss coun ies, we
also epo Conley e o s wi h a spa ial cu o o 500 km and a
he e oskedas ici y- and au oco ela ion-consis en (HAC) s uc u e wi h
a lag cu o o 1000 yea s in he ime-se ies dimension.37 While ou
35 The e en -s udy equi alen would be a pe manen change o size 𝛽
s a ing in 𝑡= 2 o a p ojec commi ed in 𝑡= 0.
36 The o iginal DMSP-OLS da a a e only a ailable h ough 2013. Nigh ime
obse a ions by NPP-VIIRS ha e eplaced his sys em, bu hey eco d ligh
in ensi ies wi h di e en senso s and on a di e en scale. Though esea che s
ha e aimed o ha monize he wo se ies, we only lose one yea in ou analysis
and he e o e p e e o use da a om a single, consis en sou ce.
37 Using a long lag cu o in he HAC pa o he e o s implies ha he
weigh o a ime-se ies shock is almos cons an , which is equi alen o
clus e ing on egions in he ime-se ies dimension. We also used highe spa ial
cu o s o es ima e Conley e o s bu ound no subs an i e changes beyond 500
km (and a e y small subse o s anda d e o s could no be compu ed beyond
500 km). Al e na i ely, we clus e ed a he le el o egions o a he le el o
egions and yea s. None o his a ec s ou quali a i e esul s.
Jou nal o U ban Economics 145 (2025) 103730
8
R. Bluhm e al.
Table 4
Iden i ica ion: Al e ing he ins umen , wi hin i s -o de egions, 2002–2013.
O e p od. All inpu s Bo h ac o s Lea e-one-ou sha es Cold Wa sha es A ica US s eel placebo De . s eel
(1) (2) (3) (4) (5) (6) (7)
Panel (a) 2SLS es ima es
P ojec s (𝛥𝑁𝑗 𝑖,𝑡−2)−0.0319 −0.0170 −0.0133 −0.0243 −0.0563 0.0952 −0.0262
(0.0118)*** (0.0109) (0.0067)** (0.0099)** (0.0288)* (0.1188) (0.0111)**
[0.0103]*** [0.0081]** [0.0066]** [0.0077]*** [0.0508] [0.1112] [0.0089]***
Panel (b) Fi s -s age es ima es
Ins umen 1 0.3361 – 0.4819 0.4881 0.0778 −0.0647 0.3495
(0.0597)*** (0.0853)*** (0.0860)*** (0.0350)** (0.0740) (0.0642)***
[0.0612]*** [0.0772]*** [0.0785]*** [0.0301]*** [0.0737] [0.0631]***
Ins umen 2 0.1748
(0.1026)*
[0.0778]**
Addi ions/Modi ica ions Modi ied shock All six ins umen s Modi ied shock Modi ied sha es Modi ied sha es Placebo ins umen Modi ied shock
J- es (𝑝- alue) – 0.34 0.09 – – – –
Fi s s age F-s a 31.73 36.63 17.38 32.20 4.94 0.76 29.63
Obse a ions 27,162 27,162 27,162 27,162 8401 27,162 27,162
Regions 2406 2406 2406 2406 729 2406 2406
Coun ies 122 122 122 122 48 122 122
No es: The able epo s eg essions esul s wi hin i s -o de adminis a i e egions. Panel a shows wo-s age leas -squa es ixed-e ec s eg essions whe e he dependen a iable is
he i s di e ence o he Gini coe icien o ligh in ensi y wi hin i s -o de adminis a i e egions. Panel b shows he co esponding i s -s age eg essions whe e he dependen
a iable is a bina y indica o o new p ojec commi men s (𝛥𝑁𝑗 𝑖,𝑡−2) in a egion. ‘O e p od.’ implies ha he ac o inpu s we e esidualized by unning a eg ession o each
inpu on he log o Chinese GDP a cons an local cu ency uni s be o e he i s ac o was ex ac ed. ‘De . s eel’ uses he linea ly de ended log o Chinese s eel p oduc ion as
he ime-se ies shock. ‘All inpu s’ uses in e ac ions wi h de ended Chinese aluminum, cemen , glass, i on, s eel, and imbe p oduc ion ime se ies as sepa a e ins umen s. ‘Bo h
ac o s’ uses he second p incipal ac o o he de ended inpu s as a second ins umen . ‘Lea e-one-ou sha es’ sub ac s he alue o 𝛥𝑁𝑖,𝑡−2 om he c oss-sec ional a e age in
each pe iod. The p obabili y is now ime- a ying, so we con ol o i in bo h s ages o he eg ession. ‘Cold Wa sha es A ica’ uses he p obabili y o ecei ing a p ojec wi h da a
om he Cold Wa pe iod om Ba ke (1989) geocoded o A ica by D ehe e al. (2021a). ‘US placebo’ uses a US s eel p oduc ion index om FRED (IPN3311A2BS) as pa
o a placebo ins umen . All speci ica ions include egion- ixed e ec s and coun y-yea ixed e ec s. S anda d e o s clus e ed a he coun y le el a e epo ed in pa en heses.
Conley e o s wi h a spa ial cu o o 500 km and a ime-se ies HAC wi h a lag cu o o 1000 yea s a e epo ed in b acke s. *** p<0.01, ** p<0.05, * p<0.1.
2000–2014 pe iod in which a Chinese non- anspo p ojec has been
commi ed (𝐹𝑡−3 ×
𝑀𝑗 𝑖). We ind ha ou esul s o anspo p ojec s
ha dly change once we con ol o non- anspo p ojec s. This implies
ha ou esul s do no e lec he e ec o all p ojec s bu hose o
anspo p ojec s speci ically.56
The inal column 5 o Table 3includes he h ee ime- a ying shocks
and he indica o o non- anspo p ojec s in conce . Again, ou
esul s emain unchanged. The coe icien o anspo p ojec s emains
signi ican and ha dly a ies in magni ude. While his does no ule
ou ha o he shocks ha co ela e wi h ou ins umen and spa ial
concen a ion bias ou coe icien , gi en ha he a guably mos ob ious
sou ces o bias ha dly change ou esul s, we expec such bias o be
negligible.
The nex se o es s ocus on he ins umen i sel . Ch is ian and
Ba e (2024) sugges es s o p obe he alidi y o he assump ions
unde lying ou ins umen al a iable app oach. In addi ion o isual
inspec ion o ends in Fig. 4and Figu e A-1, we conduc a andom-
iza ion in e ence es whe e we eassign he anspo p ojec indica o
and ins umen al a iable o di e en coun ies and yea s in he sam-
ple. As can be seen om he esul s o 999 Mon e Ca lo simula ions
shown in Figu e A-2, he esul ing coe icien es ima es cen e a ound
ze o. Acco ding o an exac Fishe es , he coe icien om ou main
es ima e (in oduced below and indica ed by he e ical dashed line)
signi ican ly di e s om he andomized coe icien s (𝑝- alue =0.019).
The same holds when we b eak he iming s uc u e equi ed o
iden i ica ion less adically and ins ead andomize (i) he en i e ime
se ies be ween egions, (ii) yea s wi hin egions, and (iii) egions wi hin
yea s (also shown in Figu e A-2). Thus, omi ed a iables a e unlikely
o co ela e wi h ou key a iables in a way ha spu iously b ings abou
ou main esul .
56 Fo b e i y, we only epo he i s -s age esul s o anspo p ojec s
bu no e ha he i s -s age F-s a is ic emains ela i ely high ( o an equa ion
wi h wo endogenous a iables), he es ima ed coe icien on ou p ima y
ins umen al a iable ha dly changes, while he ins umen o non- anspo
p ojec s only p edic s hose ypes o p ojec s.
Table 4p esen s es s o obus ness o modi ica ions in he ime-
se ies (‘shock’) and c oss-sec ional (‘sha e’) componen s o ou ins u-
men . Since ou iden i ica ion s a egy le e ages exogenous shocks,
we i s pe u b he ime se ies. Column 1 modi ies ou de ending
me hod and esidualizes each inpu se ies ia a eg ession on he log
o GDP a cons an local cu ency uni s. In ui i ely, his esul s in a
measu e o inpu g ow h ha is as e o slowe han GDP g ow h,
as opposed o de ia ing om a linea end. Column 2 (‘All inpu s’)
uses de ended Chinese aluminum, cemen , glass, i on, s eel, and imbe
p oduc ion a he han hei i s p incipal componen , each in e ac ed
wi h he same sha e componen as be o e, as sepa a e ins umen s.57
This modi ica ion deli e s es ima es ha a e one s anda d e o smalle
bu , mo e impo an ly, does no ejec a J- es ha es s whe he hese
ins umen s iden i y he same LATEs. In he same spi i , column 3
includes he second ac o we de i ed om ou ac o analysis o
China’s inpu ma e ials (in e ac ed wi h he o iginal sha e) as pa o a
second ins umen . The i s -s age esul sugges s some in o ma ion is
con ained in he second componen , bu he s eng h o he ela ionship
d ops and he con idence in e als o he 2SLS es ima e o e lap wi h
ha o ou baseline.
In columns 4 and 5, we es he obus ness o al e na i e measu es
o he sha e componen o ou ins umen – a egion’s p obabili y o
ecei ing a p ojec . While we ollow he app oach o Nunn and Qian
(2014), esea che s ypically use sha es calcula ed be o e he shock in
he b oade shi -sha e li e a u e. A conce n migh be ha compu ing
he p obabili y o ecei ing a de elopmen p ojec using all a ailable
yea s migh bias ou es ima es (e en hough he ixed e ec s con ol o
he endogenous p obabili y). We app oach his conce n in wo ways.
Fi s , in column 4, we e-es ima e ou baseline model bu exclude
con empo aneous p ojec s when we calcula e p obabili ies, excluding
𝛥𝑁𝑗 𝑖,𝑡−2 om he c oss-sec ional a e age, and deno e his he ‘lea e-
one-ou ’ ins umen .58 Second, column 5 eplaces he p obabili y o
57 We do no epo he six i s -s age coe icien s in he able o educe
clu e .
58 While his emo es he con empo aneous co ela ion be ween he ins u-
men and p ojec commi men s, he p obabili y is now ime- a ying and is
Jou nal o U ban Economics 145 (2025) 103730
15

R. Bluhm e al.
Table 5
Timing o e ec s, wi hin i s -o de egions, 2004–2013.
Lag s uc u e o 𝛥𝑁𝑖,𝑡−𝑠and 𝐹𝑡−𝑠−1
𝑠= 0𝑠= 1𝑠= 2𝑠= 3𝑠= 4
(1) (2) (3) (4) (5)
Panel (a) 2SLS es ima es, changes in 𝑠 o p ojec and IV
P ojec s (𝛥𝑁𝑖,𝑡−𝑠)−0.0223 −0.0188 −0.0181 −0.0045 0.0026
(0.0187) (0.0128) (0.0114) (0.0067) (0.0079)
[0.0167] [0.0110]* [0.0081]** [0.0073] [0.0088]
Panel (b) 2SLS es ima es, only changes in s o IV
P ojec s (𝛥𝑁𝑖,𝑡−2)−0.0563 −0.0195 −0.0181 −0.0043 0.0026
(0.0501) (0.0131) (0.0114) (0.0062) (0.0082)
[0.0568] [0.0113]* [0.0081]** [0.0069] [0.0089]
Fi s -s age F-s a panel a 6.08 6.74 16.56 20.43 25.79
Fi s -s age F-s a panel b 2.40 12.55 16.56 15.86 11.81
Obse a ions 22,445 22,445 22,445 22,445 22,445
Regions 2389 2389 2389 2389 2389
Coun ies 121 121 121 121 121
No es: The able epo s 2SLS eg essions whe e he dependen a iable is he i s di e ence o he Gini coe icien o ligh in ensi y wi hin
i s -o de adminis a i e egions. Bo h panels show wo-s age leas -squa es ixed-e ec s eg essions whe e he dependen a iable is he Gini
coe icien o ligh in ensi ies wi hin i s -o de adminis a i e egions. Panel a ixes he lag s uc u e o he i s di e ence o p ojec s and
he ins umen bu a ies yea s o commi men as indica ed in he column heade . Fo example, column 1 es ima es he e ec s o p ojec s
commi ed in 𝑡, ins umen ed wi h he in e ac ion including inpu s p oduced in 𝑡− 1. Panel b ixes he iming o p ojec s bu a ies he lag
be ween p ojec s and inpu p oduc ion. Fo example, column 1 epo s he e ec o p ojec s commi ed in 𝑡− 2, ins umen ed wi h he in e ac ion
including inpu s p oduced in 𝑡− 2. All speci ica ions include egion- ixed e ec s and coun y-yea ixed e ec s. S anda d e o s clus e ed a he
coun y le el a e epo ed in pa en heses. Conley e o s wi h a spa ial cu o o 500 km and a ime-se ies HAC wi h a lag cu o o 1000 yea s
a e epo ed in b acke s. *** p<0.01, ** p<0.05, * p<0.1.
ecei ing a p ojec wi h da a om he Cold Wa pe iod. We ake hese
da a om D ehe e al. (2021a), who p o ide geocoded in o ma ion on
688 Chinese p ojec loca ions comple ed in 47 A ican coun ies o e
he 1956–1987 pe iod. Column 4 o Table 4shows ha ou esul s
emain simila when we ins umen p ojec s wi h he ‘lea e-one-ou ’
ins umen . The esul s a e s ill b oadly in line wi h ou main indings
when we use Cold Wa p ojec s in a subsample o A ican coun ies
as he ‘sha e’ pa o ou ins umen (in column 5), bu he i s s age
is weake . This is no su p ising as p e-sample p obabili ies ha e a
less di ec ela ionship wi h he numbe o p ojec s in any gi en yea ,
making hem a less po en p edic o .
Column 6 epo s he esul s o a placebo eg ession, using (de-
ended) US s eel p oduc ion a he han Chinese p oduc ion alues.
This p esen s a alsi ica ion es as US s eel p oduc ion should be
un ela ed o Chinese- inanced in as uc u e p ojec s. To acili a e com-
pa ison, we show he analogous eg ession using Chinese s eel p o-
duc ion as pa o ou ins umen in column 7.59 As expec ed, he
i s s age o he placebo eg ession collapses and is ex emely weak,
wi h a Kleibe gen–Paap F-s a is ic o jus 0.76, while he 2SLS es ima e
is posi i e and imp ecisely es ima ed. In sho , US s eel p oduc ion
does no help o p edic he commi men o Chinese- inanced anspo
p ojec s. On he con a y, as can be seen om column 7, i s -s age and
second-s age es ima es a e simila o ou baseline esul s when we use
China’s s eel p oduc ion as shi componen o ou ins umen .
Taken oge he , hese es ima es sugges ha ou esul s do no hinge
on he speci ic choice o how we de ine Chinese p oduc ion shocks and
he p obabili y o ecei e Chinese p ojec s and ha spu ious ends do
no d i e he ime-se ies componen .
Table 5in es iga es he iming o he di usion e ec s o Chinese
anspo p ojec s in mo e de ail. Recall ha we ha e de e mined he
wo-yea lag du a ion o ou analysis based on D ehe e al. (2021b),
who p o ide s a and end da es o 300 p ojec s.60 We con inue o
hus no longe abso bed by he egion- ixed e ec s. Ins ead, we con ol o
i in bo h s ages o he eg ession. We ha e also calcula ed he p obabili y
based on ime pe iods be o e he espec i e commi men and using p ojec s
commi ed o neighbo ing egions a he han a egion i sel . In bo h cases,
i s -s age F-s a is ics a e low.
59 No e ha none o hese esul s con ol o he o he ‘‘China shocks’’ om
he p e ious able, bu he esul s ha dly change i we include hem.
ocus on i s di e ences bu a y he iming by which we allow p ojec s
o a ec spa ial concen a ion as well as he iming wi h which we
assume cons uc ion ma e ials o a ec p ojec s. Panel a ixes he one-
yea lag be ween he ins umen and Chinese anspo p ojec s bu
shi s bo h o wa d and backwa d in ime. Column 1 s a s wi h he
e ec s o p ojec s in he yea o commi men (𝑠= 0) ins umen ed by
he i s lag o p ojec inpu s. Column 5 ends wi h he e ec o p ojec s
commi ed ou yea s ea lie (𝑠= 4), ins umen ed by he i h lag
o p ojec inpu s. To allow o a meaning ul analysis o he e ec o
he lag s uc u e, we hold he sample cons an ac oss all eg essions,
esul ing in ewe obse a ions compa ed o he baseline eg ession.
Column 3 epo s esul s analogous o ou baseline speci ica ion. The
esul s a e somewha weake compa ed o hose epo ed abo e bu
emain quali a i ely simila . Es ima es become smalle o longe lags
(see columns 4 and 5) and a e imp ecisely es ima ed, e en hough he
i s -s age ela ionship be ween inpu s and commi men s is s ong o
deepe lags. Column 2 shows he e ec s on concen a ion one yea a e
commi men . The coe icien s a e simila o ou baseline eg ession in
magni ude bu a e es ima ed less p ecisely, and he i s -s age ela ion-
ship becomes weak. When we u n o con empo a y e ec s in he yea
o commi men in column 1, he i s s age u ns ou e en weake and
he es ima es a e no longe signi ican a con en ional le els.61 Taken
oge he , hese esul s suppo ou choice o a wo-yea lag and show
he obus ness o using a one-yea lag, in line wi h he no ion ha much
o he impac occu s ea ly.
Panel b o Table 5examines he lag be ween ou ins umen and
p ojec commi men s. We ocus on anspo p ojec s commi ed in
𝑡− 2, bu a y he lag o he ins umen be ween one and ou yea s
(co esponding o 𝑠= 0 o 𝑠= 4). We do no claim ha a ia ions
60 Mo e p ecisely, he a e age obse ed p ojec du a ion in his subsample
is 664 days. D ehe e al. (2021b) poin ou ha his lag choice aligns wi h he
p e ailing belie held by de elopmen p ac i ione s and go e nmen o icials
in hos coun ies ha Chinese de elopmen p ojec s a e execu ed swi ly.
61 We ha e also es ed whe he u u e p ojec s p edic pas concen a ion
(no epo ed). Signi ican es ima es o u u e p ojec s on pas concen a ion
would subs an ially weaken he c edibili y o ou es ima ion s a egy. As
expec ed, he ‘‘e ec ’’ o p ojec s one o wo yea s in he u u e on oday’s
concen a ion is es ima ed e y imp ecisely, wi h i s -s age F-s a is ics below
one.
Jou nal o U ban Economics 145 (2025) 103730
16
R. Bluhm e al.
Table 6
Ligh in ensi y and quin ile sha es, wi hin i s -o de egions, 2002–2013.
Momen s o spa ial concen a ion
Ligh Ligh Ex ensi e Quin ile sha es
densi y pe capi a ma gin 0%–20% 20%–40% 40%–60% 60%–80% 80%–100%
(1) (2) (3) (4) (5) (6) (7) (8)
Panel (a) 2SLS es ima es
P ojec s (𝛥𝑁𝑗 𝑖,𝑡−2) 0.1462 −0.0005 0.0122 0.0028 0.0032 0.0102 0.0191 −0.0353
(0.0476)*** (0.0028) (0.0173) (0.0015)* (0.0027) (0.0040)** (0.0075)** (0.0114)***
[0.0516]*** [0.0028] [0.0104] [0.0019] [0.0024] [0.0043]** [0.0078]** [0.0116]***
Panel (b) Fi s -s age es ima es
IV (𝐹𝑡−3 ×
𝑁𝑗 𝑖) 0.4505 0.4498 0.4505 0.4394 0.4394 0.4394 0.4394 0.4394
(0.0732)*** (0.0732)*** (0.0732)*** (0.0749)*** (0.0749)*** (0.0749)*** (0.0749)*** (0.0749)***
[0.0688]*** [0.0688]*** [0.0688]*** [0.0716]*** [0.0716]*** [0.0716]*** [0.0716]*** [0.0716]***
Fi s -s age F-s a 37.86 37.72 37.86 34.44 34.44 34.44 34.44 34.44
Obse a ions 28,037 28,025 28,037 26,877 26,877 26,877 26,877 26,877
Regions 2440 2439 2440 2379 2379 2379 2379 2379
Coun ies 122 122 122 122 122 122 122 122
No es: The able epo s 2SLS eg essions whe e he dependen a iable is he i s di e ence o he Gini coe icien o ligh in ensi y wi hin i s -o de adminis a i e egions. Panel
a shows wo-s age leas -squa es ixed-e ec s eg essions whe e he dependen a iable is indica ed in he column heade . Panel b shows he co esponding i s -s age eg essions
whe e he dependen a iable is a bina y indica o o new p ojec commi men s (𝛥𝑁𝑗 𝑖,𝑡−2) in a egion. We use he in e se hype bolic sine ans o ma ion in columns 1 and 2,
which is de ined as 𝑖ℎ𝑠(𝑧) =𝑙 𝑜𝑔(𝑧+√𝑧2+ 1), o ligh densi y and ligh pe capi a o include egions wi h ze o ligh and e ain an in e p e a ion simila o logs. The ex ensi e
ma gin in column 3 is de ined as he (un ans o med) ac ion o pixels wi h a non-ze o ligh densi y. All speci ica ions include egion- ixed e ec s and coun y-yea ixed e ec s.
S anda d e o s clus e ed a he coun y le el a e epo ed in pa en heses. Conley e o s wi h a spa ial cu o o 500 km and a ime-se ies HAC wi h a lag cu o o 1000 yea s
a e epo ed in b acke s. *** p<0.01, ** p<0.05, * p<0.1.
o ou ins umen al a iable a e excludable o di e en yea s bu
conside his as a p edic i e exe cise o ind he lag s uc u e o which
he ela ionship is s onges . Fo example, p ojec s commi ed in 2010
a e likely o be d i en by inpu ma e ials no only in 2009, bu also
in o he adjacen yea s. When we a y he lag be ween p ojec s and
inpu ma e ials, he i s -s age ela ionship is s onges o ou p e e ed
one-yea lag be ween inpu s and commi men s and declines o all
o he imings. The i s -s age ela ionship be ween inpu s and p ojec s
comple ely b eaks down when we p edic p ojec s wi h inpu s p oduced
one yea in he u u e (column 1). O e all, hese esul s suppo ou
choice o a one-yea lag be ween inpu s and commi men s.
Ex ensions
Ou inding ha Chinese- inanced anspo p ojec s educe he con-
cen a ion o economic ac i i y wi hin subna ional egions aises he
ques ion o whe e exac ly his di usion akes place. The monocen ic
ci y model implies ha we should obse e a shi in ac i i y om ci ies
o hei immedia e pe iphe y (and, hence, expec some he e ogenei y
wi h espec o he le el o u baniza ion). The model has li le o say
abou whe he his should also inc ease o e all ac i i y in a egion
o whe he his kind o de elopmen occu s by leap- ogging in o
unde eloped a eas o in eg a ing less densely de eloped a eas.
O he amewo ks o e some guidance. Heblich e al. (2020) s udy
how anspo a ion echnology a ec s he specializa ion o loca ions
wi hin a ci y in o a wo kplace and esidence in a la ge ange o
quan i a i e u ban models. A key inding is ha imp o emen s in
anspo echnology lead o as e g ow h o subu bs ela i e o he
cen al ci y by de eloping open ields and building ou p eexis ing
illages. Co e–pe iphe y models, in u n, o e p edic ions on o e all
ac i i y as egions become be e connec ed. Fo example, Fabe (2014)
shows ha he e ec o in as uc u e imp o emen s is nega i e o
pe iphe al egions bu he e ogeneous in he le el o p e-exis ing ade
in eg a ion and ela i e ma ke size be ween he co e and pe iphe y.
We ake hese esul s as mo i a ion o s udy he e ec s o Chinese
in as uc u e in es men o di e en momen s o he dis ibu ion o
nigh ime ligh s, di e en egions o he wo ld, and di e en le els o
u baniza ion and ade in eg a ion.
Table 6examines di e en momen s o he dis ibu ion. Columns 1
and 2 show a s ong e ec o anspo p ojec s on o e all economic
ac i i y bu no on ou p oxy o pe capi a incomes. A new anspo
p ojec inc eases he a e age ligh densi y in a egion by abou 15
pe cen (column 1), which is bo h economically and s a is ically signi -
ican . When we ins ead ocus on ligh pe capi a (column 2), we canno
ejec he null hypo hesis ha Chinese anspo p ojec s ha e no e ec
on changes in ligh pe capi a and es ima e a coe icien close o ze o.
This aligns wi h he esul s o D ehe e al. (2022), who also epo
a null e ec o Chinese de elopmen inance in gene al on ligh s pe
capi a in a global sample. The insigni ican coe icien o he wo ld
sample s ands in con as o esul s o he A ican con inen , whe e
p e ious wo k inds posi i e e ec s o aid on de elopmen (D ehe
e al.,2021a,2022). Since he pe capi a da a use in e pola ed pop-
ula ion da a in he denomina o , we canno ule ou ha he imp ecise
es ima e occu s due o added noise in he dependen a iable. Column
3 uses he ac ion o illumina ed pixels as he dependen a iable. The
insigni ican esul sugges s ha economic ac i i y does no seem o
expand p ima ily in o p e iously unde eloped a eas. To summa ize,
columns 1 o 3 show ha Chinese anspo p ojec s inc ease economic
ac i i y in he ecei ing egion. This may ep esen an inc ease in
popula ion a he han wel a e and appea s o be p ima ily occu ing
in al eady somewha de eloped a eas.
The emaining columns o Table 6 ocus on ela i e changes in
economic ac i i y ac oss quin iles o he ligh dis ibu ion. This allows
us o di ec ly ace which ype o changes educe he Gini coe icien .
The pa e n is consis en wi h p edic ions om u ban land use heo y.
We ind ha Chinese anspo p ojec s signi ican ly educe nigh ime
ligh s in he highes quin ile, while hey aise he sha e o ac i i y
aking place in he lowe quin iles ( hough es ima ed imp ecisely o
he second quin ile and wi h bo de line signi icance o he i s ).
I hus appea s ha Chinese anspo p ojec s g adually edis ibu e
ac i i y om he mos densely de eloped pa s o egions, ha is,
he ci y cen e s, o less densely de eloped places. The magni udes o
he es ima ed coe icien s sugges ha his p ocess bene i s he highe
quin iles mo e ela i e o he leas de eloped pa s o a egion.
Table 7p esen s a mo e di ec app oach o measu ing om whe e
o whe e he eloca ion o ac i i y akes place. We epo a se ies o
eg essions ha spli he sample along he median o se e al a i-
ables ypically linked wi h apid u ban g ow h. The esul s p o ide
u he suppo o he conjec u e ha hese e ec s occu a ound ci ies.
We ind a sizable decen aliza ion o ac i i y in egions wi h below-
median a el ime o ci ies, high u baniza ion a es, high oad densi y,
and abo e-median p oximi y o he coas . The es ima ed e ec s a e
subs an ial in hese sub-samples. By con as , hey a e imp ecisely
es ima ed and ypically o he opposi e sign o lowe in magni ude in
Jou nal o U ban Economics 145 (2025) 103730
17
R. Bluhm e al.
Table 7
Sample spli s, wi hin i s -o de egions, 2002–2013.
Spli ing a he median o …
T a el ime U baniza ion Road Dis ance Ligh
o ci ies a e densi y o coas pe capi a
(1) (2) (3) (4) (5)
Panel (a) Below median, 2SLS es ima es
P ojec s (𝛥𝑁𝑗 𝑖,𝑡−2)−0.0306 0.0044 −0.0101 −0.0207 −0.0276
(0.0147)** (0.0087) (0.0099) (0.0154) (0.0089)***
[0.0104]*** [0.0103] [0.0107] [0.0090]** [0.0056]***
Panel (b) Abo e median, 2SLS es ima es
P ojec s (𝛥𝑁𝑗 𝑖,𝑡−2) 0.0003 −0.0222 −0.0262 −0.0232 −0.0288
(0.0105) (0.0180) (0.0121)** (0.0096)** (0.0287)
[0.0097] [0.0129]* [0.0112]** [0.0098]** [0.0312]
Fi s -s age F-s a a 11.74 20.15 22.60 23.23 21.01
Fi s -s age F-s a b 39.68 12.11 16.90 16.33 13.76
Obse a ions (a) 13,016 13,642 13,527 13,509 13,142
Obse a ions (b) 13,875 13,254 13,451 13,545 13,672
No es: The able epo s 2SLS eg essions whe e he dependen a iable is he i s di e ence o he Gini coe icien o ligh in ensi y wi hin
i s -o de adminis a i e egions. Panel a shows wo-s age leas squa es ixed e ec s eg essions o i s -o de egions wi h below median alues
o he a iable indica ed in he column heade . Panel b shows wo-s age leas squa es ixed e ec s eg essions o i s -o de egions wi h abo e
median alues o he a iable indica ed in he column heade . ‘T a el ime o ci ies’ is measu ed as he a el ime o he nea es ci y o 50,000
o mo e people in he yea 2000 (Nelson,2008). The ‘u baniza ion a e’ is measu ed as he ac ion o land in he egion which is de ined
as an u ban clus e o u ban cen e in 2000 by he Global Human Se lemen Laye (Pesa esi e al.,2019). ‘Road densi y’ is measu ed as he
o al oad leng h o e he a ea o he egion whe e oad leng h is de i ed om he gROADS da a se (CIESIN and ITOS,2013). ‘Dis ance o
coas ’ is he a e age ‘‘as- he-c ow- lies’’ dis ance o he nea es coas line ( om Na u al Ea h). ‘Ligh pe capi a’ is he sum o ligh in a egion
di ided by i s popula ion in 2000 ( om he Global Human Se lemen Laye ). All speci ica ions include egion- ixed e ec s and coun y-yea
ixed e ec s. S anda d e o s clus e ed a he coun y le el a e epo ed in pa en heses. Conley e o s wi h a spa ial cu o o 500 km and a
ime-se ies HAC wi h a lag cu o o 1000 yea s a e epo ed in b acke s. *** p<0.01, ** p<0.05, * p<0.1.
he o he sub-samples ( he excep ion being below-median p oximi y o
he coas , wi h simila es ima ed magni udes and s anda d e o s in
bo h samples). Las bu no leas , we also ind ha he e ec seems o
be d i en by ela i ely poo egions as measu ed by below-median ligh
pe capi a. This is no su p ising, gi en ha some o he poo es egions
ha e some o he highes popula ion g ow h a es and a e home o
many o he as es -g owing ci ies. The e idence p esen ed he e aligns
wi h he li e a u e ocusing on indi idual coun ies o egions discussed
abo e. Fo example, in hei s udy o he expansion o China’s high-
way sys em, Baum-Snow e al. (2017) ind ha educ ions in spa ial
concen a ion we e la ge wi hin coas al and iche cen al egions.
Simila ly, al hough hey do no ocus on decen aliza ion wi hin egions
pe se, s udies ocusing on he spa ial impac o he Bel and Road
Ini ia i e (BRI) ypically es ima e ha coas al egions, bo de c ossings,
and u ban hubs will bene i mo e (Lall and Leb and,2020).62
Nex , we in es iga e majo wo ld egions sepa a ely. China’s global
in as uc u e oo p in is une en and mos o i s anspo a ion p ojec s
a e loca ed in A ica and Asia ( ecall Fig. 2). U ban popula ion g ow h
is apid and in as uc u e cons ain s a e mos se e e in hese e-
gions. As can be seen om he sub-samples in columns 1 o 3 o
Table 8, ou main indings a e d i en by egions in A ican coun ies,
whe e he e ec is la ge han ou baseline es ima es. The coe icien
on Chinese anspo p ojec s is insigni ican o subs an ially smalle
o Asia and he Ame icas, al hough he i s s age emains abou
equally powe ul in all h ee egions. This is no su p ising gi en ha
A ica lags behind he o he wo wo ld egions in e ms o in as-
uc u e de elopmen . I is also he egion whe e u ban p imacy is
mos p onounced and whe e de iciencies in u ban in as uc u e ha e
been linked o slowe economic g ow h a he na ional le el (Cas ells-
Quin ana,2017). Chinese- inanced p ojec s in A ica, he e o e, appea
o mi iga e conges ion, which, e en ually, could enable ci ies o eap
he bene i s o agglome a ion economies.
Column 4 es ic s he sample o coun ies classi ied by he Wo ld
Bank as low-income economies in 2000. I highligh s ha he di usion
62 This li e a u e sugges s ha BRI p ojec s will lead o an inc easing spe-
cializa ion among egions and hence mo e concen a ion o economic ac i i y
in egions wi h be e access o wo ld ma ke s bu does no conside he
dis ibu ion o ac i i y wi hin egions.
e ec s o Chinese anspo p ojec s also occu in he poo es coun ies
o he wo ld. Finally, in column 5, we es ic ou analysis o only
hose subna ional egions ha ha e ecei ed a leas one anspo
p ojec om China o e he en i e sample pe iod. This add esses one
las iden i ica ion challenge ha would a ise i egions ha ecei ed
any de elopmen - ela ed p ojec om China expe ience di e en non-
linea ends han hose which did no . Ou esul s become subs an ially
s onge .
Finally, Table A-6 in he Online Appendix explo es he issue o co-
loca ion wi h o he ypes o p ojec s. Ou esul s emain simila when
we con ol o he p esence o Wo ld Bank p ojec s in he anspo sec-
o (o in any sec o ) and o Chinese- inanced p ojec s in o he sec o s
(also ecall ha ou esul s a e obus o including any non- anspo
p ojec om China, as shown in column 4 o Table 3).
6. Conclusion
This a icle examined whe he and o wha ex en anspo in as-
uc u e p ojec s decen alize economic ac i i y in ecipien egions
ac oss he Global Sou h. We o e come he challenge o missing ge-
olocalized da a on compa able in as uc u e p ojec s ac oss coun ies
and how o es ima e hei causal e ec s by ocusing on in as uc u e
p ojec s inanced by he Chinese go e nmen – a single bu massi e
sou ce o in as uc u e inancing ac oss he de eloping wo ld. While
many schola s and policymake s a e skep ical abou he quali y and
e ec s o Chinese de elopmen p ojec s, China’s commi men o i-
nancing o e seas in as uc u e was unambiguous du ing he i s wo
decades o he 21s cen u y. T anspo p ojec s, such as oads, high-
ways, ailways, ha bo s, and ai po s, a e a he hea o his app oach,
and Chinese s a e-owned en i ies ha e inanced hund eds o hem in
de eloping coun ies since 2000 (D ehe e al.,2022).
One o ou key con ibu ions is o p o ide a new geocoded da ase
o China’s g owing de elopmen oo p in a ound he wo ld, much
o which comes in he o m o la ge-scale in as uc u e in es men s
bu ex ends ac oss mul iple sec o s. While ou da a co e he pe iod
om 2000 o 2014 and hus mos ly p ecede he BRI, he p ojec s we
ocus on sha e many cha ac e is ics wi h in as uc u e buil du ing he
i s decade o he BRI. Using hese da a, we es whe he in as uc-
u e p ojec s in luence he spa ial concen a ion o economic ac i i y
Jou nal o U ban Economics 145 (2025) 103730
18
R. Bluhm e al.
Table 8
Regional a ia ion, wi hin i s -o de egions, 2000–2013.
Regional subse s and ela ed sample pe u ba ions
A ica Asia Ame icas Low income 
𝑁all >0
(1) (2) (3) (4) (5)
Panel (a) 2SLS es ima es
P ojec s (𝛥𝑁𝑗 𝑖,𝑡−2)−0.0252 −0.0134 −0.0084 −0.0204 −0.0301
(0.0076)*** (0.0244) (0.0079) (0.0117)* (0.0073)***
[0.0100]** [0.0112] [0.0019]*** [0.0058]*** [0.0100]***
Panel (b) Fi s -s age es ima es
IV (𝐹𝑡−3 ×
𝑁𝑗 𝑖) 0.4413 0.4171 0.7981 0.4327 0.4348
(0.1123)*** (0.1020)*** (0.2285)*** (0.0897)*** (0.0753)***
[0.0997]*** [0.1096]*** [0.3355]** [0.0826]*** [0.0744]***
Fi s -s age F-S a 15.45 16.72 12.20 23.28 33.31
Obse a ions 8401 9191 4954 11,357 8639
Regions 729 791 430 982 735
Coun ies 48 34 22 60 92
No es: The able epo s 2SLS eg essions. Panel a shows wo-s age leas squa es ixed e ec s eg essions whe e he dependen a iable is he i s
di e ence o he Gini coe icien o ligh in ensi y wi hin i s -o de adminis a i e egions. Panel b shows leas squa es ixed e ec s eg essions
whe e he dependen a iable is a bina y indica o o new p ojec commi men s (𝛥𝑁𝑗 𝑖,𝑡−2) in a egion. Columns 1 o 3 epo egional subse s
as indica ed in he column heade . Column 4 uses only coun ies classi ied as low-income economies by he Wo ld Bank in 2000. Column 5 uses
only egions ha ha e ecei ed any anspo o non- anspo p ojec om China o e he en i e pe iod. All speci ica ions include egion- ixed
e ec s and coun y-yea ixed e ec s. S anda d e o s clus e ed a he coun y le el a e epo ed in pa en heses. Conley e o s wi h a spa ial
cu o o 500 km and a ime-se ies HAC wi h a lag cu o o 1000 yea s a e epo ed in b acke s. *** p<0.01, ** p<0.05, * p<0.1.
wi hin and be ween ecipien egions. Ou iden i ica ion s a egy elies
on commodi y inpu s p oduced in China ha a ec he a ailabili y
o p ojec s o e ime in andem wi h a a iable ha measu es he
likelihood ha coun ies ecei e a smalle o la ge sha e o China’s
p ojec s.
Ou esul s show ha Chinese go e nmen - inanced anspo a ion
p ojec s educe he concen a ion o economic ac i i y wi hin egions
in de eloping coun ies. While we ind simila e ec sizes o con-
cen a ion be ween egions, hese e ec s a e es ima ed less p ecisely.
Ou wi hin- egion esul s imply ha he Gini coe icien measu ing he
spa ial concen a ion o economic ac i i y is educed by 2.2 pe cen age
poin s wi hin i s -o de egions. These esul s a e obus in a la ge
numbe o di e en speci ica ions, o he choice o con ol a iables
and a ia ions o he ins umen al a iable. The e ec inc eases o
comple ed p ojec s, holds o p ojec s inancing economic in as uc-
u e mo e b oadly, and is la ges in poo egions and A ican coun ies,
which mos need in as uc u e inancing. In line wi h u ban land use
heo y, we ind ha ou esul s a e d i en by changes in economic
ac i i y in and a ound u ban a eas.
In inancing majo anspo p ojec s, he Chinese go e nmen ap-
pea s o be helping ci ies and egions in de eloping coun ies ans o m
om dense, c owded, and unp oduc i e places in o p oduc i e hubs.
While hese esul s a e encou aging, hey do no imply ha Chinese
go e nmen - inanced anspo in as uc u e p ojec s only ha e pos-
i i e e ec s. The e is g owing e idence ha Chinese de elopmen
p ojec s also p oduce nega i e ex e nali ies. Fo example, in ela ed
wo k, we ha e shown ha China’s ‘‘aid on demand’’ app oach is
ulne able o domes ic poli ical cap u e whe ein incumben go e n-
men leade s s ee Chinese de elopmen p ojec s owa ds hei home
egions, o en a he expense o poo e egions wi h g ea e ma e ial
need (D ehe e al.,2019a). The e a e many o he conce ns abou
he consequences o China’s de elopmen inance, anging om hei
impac on he en i onmen and deb sus ainabili y (Ho n e al.,2021;
Baeh e al.,2023). In sho , Chinese- inanced anspo a ion p ojec s
may help deal wi h conges ion in de eloping coun ies, bu ou s udy
should no be ead as a comp ehensi e assessmen o hei cos s and
bene i s. The e is conside able scope o u u e esea ch in his a ea.
CRediT au ho ship con ibu ion s a emen
Richa d Bluhm: W i ing – e iew & edi ing, W i ing – o iginal
d a , Me hodology, In es iga ion, Fo mal analysis, Da a cu a ion, Con-
cep ualiza ion. Axel D ehe : W i ing – e iew & edi ing, W i ing –
o iginal d a , Me hodology, In es iga ion, Funding acquisi ion, Fo mal
analysis, Da a cu a ion, Concep ualiza ion. And eas Fuchs: W i ing –
e iew & edi ing, W i ing – o iginal d a , Me hodology, In es iga ion,
Funding acquisi ion, Fo mal analysis, Da a cu a ion, Concep ualiza ion.
B adley C. Pa ks: W i ing – e iew & edi ing, W i ing – o iginal
d a , Me hodology, In es iga ion, Funding acquisi ion, Da a cu a ion,
Concep ualiza ion. Aus in M. S ange: W i ing – e iew & edi ing,
W i ing – o iginal d a , Me hodology, In es iga ion, Funding acqui-
si ion, Da a cu a ion, Concep ualiza ion. Michael J. Tie ney: P ojec
adminis a ion, Funding acquisi ion, Concep ualiza ion.
Acknowledgmen s
We owe a deb o g a i ude o he la ge eam o esea ch assis an s –
including Melanie Aguila -Rojas, Bilal Asad, Zach Bax e , Rachel Bena-
ides, Ellie Ben ley, Liliana Besosa, Allison Bowe s, A iel Cadby-Spice ,
Emma Cahoon, B ee Ca elino, Alex Chadwick, A a Cha in, Tina Chang,
Yuning Chen, Meng an Cheng, Ti anie Choi, Mi anda Cla ke, Ka e
Conno s, McKay Co be , G aeme C ans on-Cuebas, Ca he ine C owley,
Alex DeGala, Hannah Dempsey, Rohan Desai, Jus in DeShazo , Joseph
Dobbels, Isabel Docampo, Weiwei Du, Ash on Ebe , Caleb Ebe , Aili
Espigh, Clai e E he idge, Jo dan Fox, Robe F ancis, Ze Fu, Melanie
Gilbe , Sa a Gomez, Liz Hall, Thompson Hangen, Lau en Ha ison,
Michael Ha haway, Collin Henson, Jasmine He ndon, Elizabe h He i y,
Kei h Holle an, Weijue Huang, Daniel Hughes, To ey Jackson, Jiao ui
Jiang, Qi Jiang, Emmaleah Jones, Ama Kaki de, Rachel Kellogg, Con-
no Kennedy, Cie a Killen, Ian Ki kwood, Wa en Ki kwood, Emily
Koe ne , Dylan Kolho , Lidia Ko ace ic, Mi ian K eykes, Isabella K on,
Ka hik Kuma appan, Daniel Lan z, Ca oline Lebegue, Jade Li, Xiao
Liu, S e en Li ings on, Yaseen Lo i, Ad iane Lopez, Flynn Madden,
Dominick Ma gio a, Sa ah Ma in, Emily McLenigan, Ma ie Mullins,
Will Nelson, Qiuyan Ni, Jack Nicol, Alexand a Pancake, Ca ol Peng,
G ace Pe kins, Sophia Pe o i, Vic o Polanco, Lau a P eszle , Emily
Qiu, Kam an Rahman, Sa ah Reso, Da id Rice, Sa a Rock, Ann Roge s,
Elizabe h Saccoccia, Na alie San os, Dominic Sanzo a, Fai h Sa aiano,
Dominic Sce bo, Rebecca Schec man, Leigh Sei z, Ryan Sep on, Lu
Se ie , Kai lan Shaub, And ea Sole a, Lau en Su, Joanna Tan, Emily
Tanne , Na e Tanne , B i any Tennan , Rebecca Tho pe, Aus in T o a,
Anna Ums ead, Jessica Usjanauskas, Julia Va ou sos, Emily Walke ,
Yale Walle , Ka he ine Walsh, Xinyi Wang, Ma Wes o e , Tom Wes -
o e , James Willa d, (Jiacheng) Jason Xi, Hanyang Xu, Da ice Xue,
E ya Yang, An onio Tianze Ye, Jack Zhang, Yue Zhang, Echo Zhong,
Joana Zhu, and Jun ong Zhu—who helped o assemble and geo-loca e
Jou nal o U ban Economics 145 (2025) 103730
19
R. Bluhm e al.
he da ase o Chinese de elopmen p ojec s used in his s udy. We also
hank Se h Goodman and Mi anda L o hei spa ial da a quali y as-
su ance and in eg a ion assis ance, as well as Ge da Asmus o sha ing
code o agg ega ing he p ojec -le el da a.
This s udy was made possible wi h gene ous inancial suppo om
John D. and Ca he ine T. MacA hu Founda ion, Uni ed S a es, Hu-
mani y Uni ed, Uni ed S a es, he William and Flo a Hewle Foun-
da ion, Uni ed S a es, he Academic Resea ch Fund o Singapo e’s
Minis y o Educa ion, he Uni ed Na ions Uni e si y Wo ld Ins i u e o
De elopmen Economics Resea ch (UNU-WIDER), Finland, he Ge man
Resea ch Founda ion (DFG p ojec s DR 640/5-1&3 and FU 997/1-1&3),
and William & Ma y, Uni ed S a es. We also acknowledge ha his
s udy was indi ec ly made possible h ough a coope a i e ag eemen
(AID-OAA-A-12-00096) be ween USAID’s Global De elopmen Lab and
AidDa a a William & Ma y unde he Highe Educa ion Solu ions
Ne wo k (HESN) P og am, as i suppo ed he c ea ion o a spa ial
da a eposi o y and ex ac ion ool which we used o execu e ou da a
analysis. The iews exp essed he e do no necessa ily e lec he iews
o any o ou unde s.
Las bu no leas , his pape ecei ed many help ul commen s a
esea ch semina s, wo kshops, and con e ences. We hank B uce Bueno
de Mesqui a, Vi ien Fos e , E ik Haus ein, Robe Inklaa , Ma hilde
Leb and, Fe dinand Rauch, semina audiences a New Yo k Uni e si y
(New Yo k Ci y, USA, Sep embe 2018), he G adua e Ins i u e o In e -
na ional and De elopmen S udies (Gene a, Swi ze land, Ap il 2019),
he Uni e si y o Milano-Bicocca (Milan, I aly, May 2019), he Uni-
e si y o Hohenheim (S u ga , Ge many, July 2019), he Eu opean
S abili y Mechanism (Luxembou g, No embe 2019), he Uni e si y
College Dublin (Dublin, I eland, No embe 2019), he Uni e si y o
G oningen (G oningen, Ne he lands, Janua y 2020), and he Uni e si y
o B i ish Columbia (Vancou e , Canada, Feb ua y 2020); as well as
con e ence pa icipan s o he HSU-I W Wo kshop in De elopmen and
En i onmen al Economics (Hambu g, Ge many, No embe 2018), he
TEDE Wo kshop ‘‘Topics in De elopmen and En i onmen al Resea ch’’
a he Uni e si y o Bi mingham (Bi mingham, UK, Feb ua y 2019),
he Annual Mee ing o he Eu opean Public Choice Socie y a he
Heb ew Uni e si y (Je usalem, Is ael, Ap il 2019), he GeoDa a in
Economics Wo kshop a he Uni e si y o Hambu g (Hambu g, Ge -
many, May 2019), he Eu opean Mee ing o he U ban Economics
Associa ion a he V ije Uni e si ei (Ams e dam, Ne he lands, June
2019), he Annual Con e ence o he Ve ein ü Socialpoli ik Resea ch
Commi ee De elopmen Economics a he DIW (Be lin, Ge many, June
2019), he Annual Bank Con e ence on De elopmen Economics a he
Wo ld Bank (Washing on DC, USA, June 2019), he China Economics
Summe Ins i u e a Peking Uni e si y (Beijing, China, Augus 2019),
he Annual Economic Resea ch Sou he n A ica Wo kshop on ‘‘S uc-
u al Cons ain s on he Economy, G ow h and Poli ical Economy’’
a he Uni e si y o he Wi wa e s and (Johannesbu g, Sou h A ica,
Sep embe 2019), he Biennial Con e ence o he Economic Socie y o
Sou h A ica (Johannesbu g, Sou h A ica, Sep embe 2019), and he
Annual Mee ing o he In e na ional Poli ical Economy Socie y a he
Uni e si y o Cali o nia San Diego (San Diego, USA, No embe 2019)
o commen s on ea lie e sions o his pape .
Appendix A. Supplemen a y da a
Supplemen a y ma e ial ela ed o his a icle can be ound online
a h ps://doi.o g/10.1016/j.jue.2024.103730.
Re e ences
Adão, R., Kolesá , M., Mo ales, E., 2019. Shi -sha e designs: Theo y and in e ence. Q.
J. Econ. 134 (4), 1949–2010.
A ican De elopmen Bank, 2014a. Kenya: Nai obi-Thika Highway Imp o emen
P ojec . A ican De elopmen Bank, Abidjan, Cô e d’I oi e, A ailable
a h ps://www.a db.o g/ ileadmin/uploads/a db/Documen s/P ojec -and-
Ope a ions/P esiden ial_awa ds_2014_-_Kenya_-_Nai obi_-_Thika_Highway_
Imp o emen _P ojec .pd . (Accessed 3 Ap il 2020).
A ican De elopmen Bank, 2014b. S udy on Quali y o Bank Financed Road
P ojec s. A ican De elopmen Bank, Abidjan, Cô e d’I oi e, A ailable a
h ps://www.a db.o g/ ileadmin/uploads/a db/Documen s/E en s/ATF o um/
S udy_on_ he_quali y_o _Bank_Financed_Road_P ojec s_-_A DB.pd . (Accessed 3 Ap il
2020).
A ican De elopmen Bank, 2016. PCR E alua ion No e o Public Sec o Ope a ions:
Nai obi-Thika Highway Imp o emen P ojec . A ican De elopmen Bank, Abidjan,
Cô e d’I oi e, A ailable a h ps://e d.a db.o g/documen s/docs/EN_PN10706.pd .
(Accessed 3 Ap il 2020).
A ican De elopmen Bank, 2019. Nai obi-Thika Highway Imp o emen P ojec :
P ojec Comple ion Repo . A ican De elopmen Bank, Abidjan, Cô e d’I oi e,
A ailable a h ps://www.a db.o g/si es/de aul / iles/documen s/p ojec s-and-
ope a ions/kenya_-_ hika_highway_imp o emen _p ojec .pd . (Accessed 3 Ap il
2020).
A ican De elopmen Fund, 2007. App aisal Repo : Nai obi–Thika Highway
Imp o emen P ojec . A ican De elopmen Fund, Tunis, Tunisia, A ailable a h ps:
//www.a db.o g/ ileadmin/uploads/a db/Documen s/P ojec -and-Ope a ions/
Kenya_-_Nai obi-Thika_Highway_Imp o emen _P ojec _-_App aisal_Repo .PDF.
(Accessed Ap il 2020).
Allen, T., A kolakis, C., 2022. The wel a e e ec s o anspo a ion in as uc u e
imp o emen s. Re . Econ. S ud. 89 (6), 2911–2957.
Alonso, W., e al., 1964. Loca ion and Land Use: Towa d a Gene al Theo y o Land
Ren . Ha a d Uni e si y P ess, Camb idge, MA.
Anaxago ou, C., E hy oulou, G., Sa an ides, V., 2020. Elec o al mo i es and he
subna ional alloca ion o o eign aid in Sub-Saha an A ica. Eu . Econ. Re . 127,
103430.
And és, L., Bille , D., Dappe, M.H., 2014. In as uc u e Gap in Sou h Asia: In-
as uc u e Needs, P io i iza ion, and Financing. The Wo ld Bank, Washing on,
DC.
A kin, D., Donaldson, D., 2015. Who’s Ge ing Globalized? The Size and Implica ions
o In a-na ional T ade Cos s. NBER Wo king Pape 21439, Na ional Bu eau o
Economic Resea ch, Camb idge, MA.
Au o , D.H., Do n, D., Hanson, G.H., 2013. The China synd ome: Local labo ma ke
e ec s o impo compe i ion in he Uni ed S a es. Ame . Econ. Re . 103 (6),
2121–2168.
Au o , D.H., Do n, D., Hanson, G.H., 2016. The China shock: Lea ning om
labo -ma ke adjus men o la ge changes in ade. Annu. Re . Econ. 8, 205–240.
Baeh , C., BenYishay, A., Pa ks, B., 2023. Highway o he o es ? Land go e nance and
he si ing and en i onmen al impac s o Chinese go e nmen - unded oad building
in Cambodia. J. En i on. Econ. Manag. 122, 102898.
Bandie a, L., Tsi opoulos, V., 2020. A amewo k o assess deb sus ainabili y unde
he Bel and Road Ini ia i e. J. De . Econ. 146, 102495.
Bane jee, A., Du lo, E., Qian, N., 2020. On he oad: Access o anspo a ion
in as uc u e and economic g ow h in China. J. De . Econ. 145, 102442.
Ba ke, W., 1989. The Economic Aid o he PR China o De eloping and Socialis
Coun ies, second ed. K. G. Sau , Munich, Ge many.
Baum-Snow, N., 2007. Did highways cause subu baniza ion? Q. J. Econ. 122 (2),
775–805.
Baum-Snow, N., 2014. U ban T anspo Expansions, Employmen Decen aliza ion, and
he Spa ial Scope o Agglome a ion Economies. B own Uni e si y, P o idence,
Unpublished Manusc ip .
Baum-Snow, N., B and , L., Hende son, J.V., Tu ne , M., Zhang, Q., 2017. Roads,
ail oads, and decen aliza ion o Chinese ci ies. Re . Econ. S a . 99 (3), 435–448.
Baum-Snow, N., Hende son, J.V., Tu ne , M.A., Zhang, Q., B and , L., 2020. Does
in es men in na ional highways help o hu hin e land ci y g ow h? J. U ban
Econ. 115, 103–124.
Baum-Snow, N., Tu ne , M.A., 2017. T anspo in as uc u e and he decen aliza ion
o ci ies in he People’s Republic o China. Asian De . Re . 34 (2), 25–50.
Baum-Snow, N., e al., 2007. Subu baniza ion and anspo a ion in he monocen ic
model. J. U ban Econ. 62 (3), 405–423.
Bayes, A., 2007. Impac Assessmen o Jamuna Mul ipu pose B idge P ojec (JMBP) on
Po e y. Japan Bank o In e na ional Coope a ion, Dhaka, Bangladesh.
BenYishay, A., Pa ks, B., Run ola, D., T ichle , R., 2016. Fo es Co e Impac s o Chinese
De elopmen P ojec s in Ecologically Sensi i e A eas. AidDa a Wo king Pape 32,
AidDa a a William & Ma y, Williamsbu g, VA.
Be man, N., Cou enie , M., 2015. Ex e nal shocks, in e nal sho s: The geog aphy o
ci il con lic s. Re . Econ. S a . 97 (4), 758–776.
Bi d, J., Leb and, M., Venables, A.J., 2020. The Bel and Road Ini ia i e: Reshaping
economic geog aphy in Cen al Asia? J. De . Econ. 144, 102441.
Bi d, J., S aub, S., 2014. The B asilia Expe imen : Road Access and he Spa ial Pa e n
o Long-Te m Local De elopmen in B azil. Wo ld Bank Policy Resea ch Wo king
Pape 6964, The Wo ld Bank, Washing on, DC.
Bluhm, R., D ehe , A., Fuchs, A., Pa ks, B.C., S ange, A.M., Tie ney, M.J., 2024.
Replica ion Da a o : Connec i e Financing. Chinese In as uc u e P ojec s and
he Di usion o Economic Ac i i y in De eloping Coun ies. Mendeley Da a, V1,
h p://dx.doi.o g/10.17632/235 7ksk8y.1.
Bluhm, R., K ause, M., 2022. Top ligh s: B igh ci ies and hei con ibu ion o economic
de elopmen . J. De . Econ. 157, 102880.
Bo usyak, K., Hull, P., Ja a el, X., 2022. Quasi-expe imen al shi -sha e esea ch
designs. Re . Econ. S ud. 89 (1), 181–213.
Jou nal o U ban Economics 145 (2025) 103730
20

R. Bluhm e al.
B äu igam, D., 2009. The D agon’s Gi : The Real S o y o China in A ica. Ox o d
Uni e si y P ess, Ox o d, UK.
B azys, S., Elkink, J.A., Kelly, G., 2017. Bad neighbo s? How co-loca ed Chinese and
Wo ld Bank de elopmen p ojec s impac local co up ion in Tanzania. Re . In .
O gan. 12 (2), 227–253.
B azys, S., Vadlamanna i, K.C., 2021. Aid cu se wi h Chinese cha ac e is ics? Chinese
de elopmen lows and economic e o ms. Public Choice 188, 407–430.
B uede le, A., Hodle , R., 2018. Nigh ime ligh s as a p oxy o human de elopmen
a he local le el. PloS One 13 (9), e0202231.
B ülha , M., Desme , K., Klinke, G.-P., 2020. The sh inking ad an age o ma ke
po en ial. J. De . Econ. 147, 102529.
Bu ch ield, M., O e man, H.G., Puga, D., Tu ne , M.A., 2006. Causes o sp awl: A
po ai om space. Q. J. Econ. 121 (2), 587–633.
Cas ells-Quin ana, D., 2017. Mal hus li ing in a slum: U ban concen a ion,
in as uc u e and economic g ow h. J. U ban Econ. 98, 158–173.
Ce e o, R., 2013. Linking u ban anspo and land use in de eloping coun ies. J.
T ansp. Land Use 6 (1), 7–24.
Che, Y., He, X., Zhang, Y., 2021. Na u al esou ce expo s and A ican coun ies’ o ing
beha iou in he Uni ed Na ions: E idence om he economic ise o China. Can.
J. Econ./Re . Can. Écon. 54 (2), 712–759.
Ch is ian, P., Ba e , C.B., 2024. Spu ious eg essions and panel IV es ima ion:
Re isi ing he causes o con lic . Econom. J. 134 (659), 1069–1099.
CIESIN and ITOS, 2013. Global Roads Open Access Da a Se , Ve sion 1 (gROADS 1).
NASA Socioeconomic Da a and Applica ions Cen e (SEDAC).
de Chaisema in, C., d’Haul oeuille, X., 2020. Two-way ixed e ec s es ima o s wi h
he e ogeneous ea men e ec s. Ame . Econ. Re . 110 (9), 2964–2996.
de Soy es, F., Mulabdic, A., Ru a, M., 2020. Common anspo in as uc u e: A
quan i a i e model and es ima es om he Bel and Road Ini ia i e. J. De . Econ.
143, 102415.
Dolla , D., 2008. Supply Mee s Demand: Chinese In as uc u e Financing in A ica.
Wo ld Bank Blog, 10 July 2008. Accessed a h ps://blogs.wo ldbank.o g/
eas asiapaci ic/supply-mee s-demand-chinese-in as uc u e- inancing-in-a ica.
Donaldson, D., 2018. Rail oads o he Raj: Es ima ing he impac o anspo a ion
in as uc u e. Am. Econ. Re . 108 (4–5), 899–934.
D ehe , A., Fuchs, A., 2015. Rogue aid? An empi ical analysis o China’s aid alloca ion.
Can. J. Econ. 48 (3), 988–1023.
D ehe , A., Fuchs, A., Hodle , R., Pa ks, B.C., Raschky, P.A., Tie ney, M.J., 2019a.
A ican leade s and he geog aphy o China’s o eign assis ance. J. De . Econ. 140,
44–71.
D ehe , A., Fuchs, A., Hodle , R., Pa ks, B.C., Raschky, P.A., Tie ney, M.J., 2021a. Is
a o i ism a h ea o Chinese aid e ec i eness? A subna ional analysis o Chinese
de elopmen p ojec s. Wo ld De . 139, 105291.
D ehe , A., Fuchs, A., Langlo z, S., 2019b. The e ec s o o eign aid on e ugee lows.
Eu . Econ. Re . 112, 127–147.
D ehe , A., Fuchs, A., Pa ks, B.C., S ange, A., Tie ney, M.J., 2018. Apples and d agon
ui s: The de e minan s o aid and o he o ms o s a e inancing om China o
A ica. In . S ud. Q. 62, 182–194.
D ehe , A., Fuchs, A., Pa ks, B., S ange, A., Tie ney, M., 2021b. Aid, China, and g ow h:
E idence om a new global de elopmen inance da ase . Am. Econ. J.: Econ. Policy
13 (2), 135–174.
D ehe , A., Fuchs, A., Pa ks, B., S ange, A., Tie ney, M.J., 2022. Banking on Beijing:
The Aims and Impac s o China’s O e seas De elopmen P og am. Camb idge
Uni e si y P ess, Camb idge, MA.
D ehe , A., Langlo z, S., 2020. Aid and g ow h: New e idence using an excludable
ins umen . Can. J. Econ. 53 (3), 1162–1198.
Du an on, G., Tu ne , M.A., 2012. U ban g ow h and anspo a ion. Re . Econ. S ud.
79 (4), 1407–1440.
Eichenaue , V.Z., Fuchs, A., B ückne , L., 2021. The e ec s o ade, aid and in es men
on China’s image in La in Ame ica. J. Comp. Econ. 49 (2), 483–498.
El idge, C.D., Baugh, K.E., Zhizhin, M., Hsu, F.-C., 2013. Why VIIRS da a a e supe io
o DMSP o mapping nigh ime ligh s. P oc. Asia-Pac. Ad . Ne w. 35, 62–69.
Fabe , B., 2014. T ade in eg a ion, ma ke size, and indus ializa ion: E idence om
China’s Na ional T unk Highway Sys em. Re . Econ. S ud. 81 (3), 1046–1070.
Fajgelbaum, P., Redding, S.J., 2022. T ade, s uc u al ans o ma ion, and de elopmen :
E idence om A gen ina 1869–1914. J. Poli . Econ. 130 (5), 1249–1318.
Fuji a, M., Ogawa, H., 1982. Mul iple equilib ia and s uc u al ansi ion o
non-monocen ic u ban con igu a ions. Reg. Sci. U ban Econ. 12 (2), 161–196.
Ga cia-Lopez, M.-A., Holl, A., Viladecans-Ma sal, E., 2015. Subu baniza ion and high-
ways in Spain when he Romans and he Bou bons s ill shape i s ci ies. J. U ban
Econ. 85, 52–67.
Geh ing, K., Kaplan, L., Wong, M.H., 2022. China and he Wo ld Bank: How con as ing
de elopmen app oaches a ec he s abili y o A ican s a es. J. De . Econ. 158,
102902.
Gibbons, S., Lyy ikäinen, T., O e man, H.G., Sanchis-Gua ne , R., 2019. New oad
in as uc u e: The e ec s on i ms. J. U ban Econ. 110, 35–50.
Goldsmi h-Pinkham, P., So kin, I., Swi , H., 2020. Ba ik ins umen s: Wha , when,
why, and how. Am. Econ. Re . 110 (8), 2586–2624.
Gollin, D., Jedwab, R., Voll a h, D., 2016. U baniza ion wi h and wi hou
indus ializa ion. J. Econ. G ow h 21 (1), 35–70.
Guillon, M., Ma honna , J., 2020. Wha can we lea n on Chinese aid alloca ion
mo i a ions om a ailable da a? A sec o ial analysis o Chinese aid o A ican
coun ies. China Econ. Re . 60, 101265.
He, G., Xie, Y., Zhang, B., 2020. Exp essways, GDP, and he en i onmen : The case o
China. J. De . Econ. 145, 102485.
Heblich, S., Redding, S.J., S u m, D.M., 2020. The making o he mode n Me opolis:
E idence om London. Q. J. Econ. 135 (4), 2059–2133.
Hende son, J., Kunco o, A., 1996. Indus ial cen aliza ion in Indonesia. Wo ld Bank
Econ. Re . 10 (3), 513–540.
Hende son, V., Mi a, A., 1996. The new u ban landscape: De elope s and edge ci ies.
Reg. Sci. U ban Econ. 26 (6), 613–643.
Hende son, J., Squi es, T., S o eyga d, A., Weil, D., 2018. The global dis ibu ion o
economic ac i i y: Na u e, his o y, and he ole o ade. Q. J. Econ. 133, 357–406.
Hende son, J., S o eyga d, A., Weil, D., 2012. Measu ing economic g ow h om ou e
space. Am. Econ. Re . 102, 994–1028.
He nandez, D., 2017. A e ‘‘new’’ dono s challenging Wo ld Bank condi ionali y?. Wo ld
De . 96, 529–549.
Hodle , R., Raschky, P.A., 2014. Regional a o i ism. Q. J. Econ. 129 (2), 995–1033.
Hoe le , A., S e ck, O., 2022. Is Chinese aid di e en ? Wo ld De . 156, 105908.
Ho n, S., Reinha , C.M., T ebesch, C., 2021. China’s o e seas lending. J. In . Econ.
133, 103539.
Humph ey, C., Michaelowa, K., 2019. China in A ica: Compe i ion o adi ional
de elopmen inance ins i u ions? Wo ld De . 120 (C), 15–28.
Iacoella, F., Ma o ano, B., Me zge , L., San ilippo, M., 2021. Chinese o icial inance
and poli ical pa icipa ion in A ica. Eu . Econ. Re . 136, 103741.
Isaksson, A.-S., 2020. Chinese aid and local e hnic iden i ica ion. In . O gan. 74 (4),
833–852.
Isaksson, A.-S., Ko sadam, A., 2018a. Chinese aid and local co up ion. J. Public Econ.
159, 146–159.
Isaksson, A.-S., Ko sadam, A., 2018b. Racing o he bo om? Chinese de elopmen
p ojec s and ade union in ol emen in A ica. Wo ld De . 106, 284–298.
Jean, N., Bu ke, M., Xie, M., Da is, M., Lobell, D., E mon, S., 2016. Combining sa elli e
image y and machine lea ning o p edic po e y. Science 353, 790–794.
Jedwab, R., S o eyga d, A., 2022. The a e age and he e ogeneous e ec s o anspo a-
ion in es men s: E idence om Sub-Saha an A ica 1960–2010. J. Eu . Econom.
Assoc. 20 (1), 1–38.
KARA and CSUD, 2012. The Social/Communi y Componen o he Analysis o he
Thika Highway Imp o emen P ojec . Kenya Alliance o Residen Associa ions
(KARA) and he Cen e o Sus ainable U ban De elopmen (CSUD), Accessed a
h p://csud.ei.columbia.edu/ iles/2012/11/KARA- epo _FINAL.pd .
K ugman, P., 1991. Inc easing e u ns and economic geog aphy. J. Poli ical Econ. 99
(3), 483–499.
Lall, S.V., Hende son, J.V., Venables, A.J., 2017. A ica’s Ci ies: Opening Doo s o he
Wo ld. Wo ld Bank, Washing on, DC.
Lall, S.V., Leb and, M., 2020. Who wins, who loses? Unde s anding he spa ially
di e en ia ed e ec s o he Bel and Road Ini ia i e. J. De . Econ. 146, 102496.
Lang, V., 2021. The economics o he democ a ic de ici : The e ec o IMF p og ams
on inequali y. Re . In . O gan. 16 (3), 599–623.
Lessmann, C., Seidel, A., 2017. Regional inequali y, con e gence, and i s de e minan s
– A iew om ou e space. Eu . Econ. Re . 92, 110–132.
Lujala, P., Ke il Rod, J., Thieme, N., 2007. Figh ing o e oil: In oducing a new da ase .
Con l. Manag. Peace Sci. 24 (3), 239–256.
Ma chesi, S., Masi, T., Paul, S., 2024. P ojec aid and i m pe o mance. Econom.
De . Cul . Chang. URLs: Ma chesi: h ps://doi.o g/10.1086/730829, Wellne : h ps:
//doi.o g/10.1086/729539, o hcoming.
Ma o ano, B., Me zge , L., San ilippo, M., 2020. Chinese de elopmen assis ance and
household wel a e in Sub-Saha an A ica. Wo ld De . 129, 104909.
Michalopoulos, S., Papaioannou, E., 2014. Na ional ins i u ions and subna ional
de elopmen in A ica. Q. J. Econ. 129 (1), 151–213.
Mills, E.S., 1967. An agg ega i e model o esou ce alloca ion in a me opoli an a ea.
Am. Econ. Re . 57 (2), 197–210.
Mu h, R.F., 1969. Ci ies and Housing: The Spa ial Pa e n o U ban Residen ial Land
Use. G adua e School o Business, Uni e si y o Chicago, Chicago, IL.
Nelson, A., 2008. Es ima ed T a el Time o he Nea es Ci y o 50,000 o Mo e People
in Yea 2000. Eu opean Commission, Join Resea ch Cen e (JRC).
Nunn, N., Qian, N., 2014. U.S. ood aid and ci il con lic . Am. Econ. Re . 104 (6),
1630–1666.
Ogawa, H., Fuji a, M., 1980. Equilib ium land use pa e ns in a nonmonocen ic ci y.
J. Reg. Sci. 20 (4), 455–475.
Olea, J.L.M., P luege , C., 2013. A obus es o weak ins umen s. J. Bus. Econom.
S a is . 31 (3), 358–369.
Pe lez, J., Huang, Y., 2017. Behind China’s $1 illion plan o shake up he economic
o de . N. Y. Times.
Pesa esi, M., Flo czyk, A., Schia ina, M., Melchio i, M., Ma enini, L., 2019. GHS
Se lemen G id, Upda ed and Re ined REGIO Model 2014 in Applica ion o GHS-
BUILT R2018A and GHS-POP R2019A, Mul i empo al (1975–1990-2000–2015)
R2019A. Eu opean Commission, Join Resea ch Cen e (JRC).
Ping, S.-N., Wang, Y.-T., Chang, W.-Y., 2022. The e ec s o China’s de elopmen
p ojec s on poli ical accoun abili y. B . J. Poli ical Sci. 52 (1), 65–84.
Jou nal o U ban Economics 145 (2025) 103730
21
R. Bluhm e al.
Puga, D., 1999. The ise and all o egional inequali ies. Eu . Econ. Re . 43 (2),
303–334.
Redding, S.J., Rossi-Hansbe g, E., 2017. Quan i a i e spa ial economics. Annu. Re .
Econ. 9 (1), 21–58.
Redding, S.J., Tu ne , M.A., 2015. T anspo a ion cos s and he spa ial o ganiza ion o
economic ac i i y. In: Handbook o Regional and U ban Economics, ol. 5, Else ie ,
pp. 1339–1398.
Rossi-Hansbe g, E., Sa e, P.-D., Owens III, R., 2009. Fi m agmen a ion and u ban
pa e ns. In e na . Econom. Re . 50 (1), 143–186.
Saiz, A., 2010. The geog aphic de e minan s o housing supply. Q. J. Econ. 125 (3),
1253–1296.
S a e Council, 2011. Whi e Pape on China’s Fo eign Aid. Xinhua/In o ma ion O ice
o he S a e Council, People’s Republic o China, Beijing, China.
S andow, D., Findley, M., Nielson, D., Powell, J., 2011. The UCDP-AidDa a Codebook
on Geo-Re e encing Fo eign Aid. Ve sion 1.1. Uppsala Con lic Da a P og am,
Uppsala Uni e si y, Uppsala, Sweden.
S ange, A.M., D ehe , A., Fuchs, A., Pa ks, B., Tie ney, M.J., 2018. T acking unde -
epo ed inancial lows: China’s de elopmen inance and he aid–con lic nexus
e isi ed. J. Con l. Resolu . 61 (5), 935–963.
S ange, A.M., Ghose, S., Russel, B., Cheng, M., Pa ks, B., 2017. AidDa a’s Me hodology
o T acking Unde epo ed Financial Flows. Ve sion 1.3. AidDa a a William &
Ma y, Williamsbu g, VA.
Swedlund, H.J., 2017. The De elopmen Dance: How Dono s and Recipien s Nego ia e
he Deli e y o Fo eign Aid. Co nell Uni e si y P ess, I haca, NY.
Wade, A., 2008. Time o he Wes o P ac ise Wha I P eaches. Financial Times,
(Janua y 23).
Weidmann, N.B., Schu e, S., 2017. Using nigh ligh s o he p edic ion o local weal h.
J. Peace Res. 54 (2), 125–140.
Wellne , L., D ehe , A., Fuchs, A., Pa ks, B.C., S ange, A., 2024. Can aid buy o eign
public suppo ? E idence om Chinese de elopmen inance. Econom. De . Cul .
Chang. o hcoming.
Xi, J., 2017. Wo k oge he o build he silk oad economic bel and he 21s cen u y
ma i ime silk oad. A keyno e speech a he opening ce emony o he bel and oad
o um (BRF) o in e na ional coope a ion in Beijing, China.
Zá a e, R.D., 2022. Spa ial Misalloca ion, In o mali y, and T ansi Imp o emen s:
E idence om Mexico Ci y. Policy Resea ch Wo king Pape Se ies 9990, The Wo ld
Bank, Washing on, DC.
Zei z, A., 2021. Emula ion o di e en ia ion? China’s de elopmen inance and
adi ional dono aid in de eloping coun ies. Re . In . O gan. 16, 265–292.
Jou nal o U ban Economics 145 (2025) 103730
22