He kenho , Philipp; K au heim, Sebas ian; Sem au, Finn Ole; S eglich, F auke
A icle — Published Ve sion
Co po a e Social Responsibili y along he global alue
chain
Jou nal o De elopmen 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: He kenho , Philipp; K au heim, Sebas ian; Sem au, Finn Ole; S eglich, F auke
(2024) : Co po a e Social Responsibili y along he global alue chain, Jou nal o De elopmen
Economics, ISSN 1872-6089, Else ie , Ams e dam, Vol. 167, pp. 1-31,
h ps://doi.o g/10.1016/j.jde eco.2023.103236
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/301393
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Jou nal o De elopmen Economics 167 (2024) 103236
A ailable online 9 Decembe 2023
0304-3878/© 2023 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY license (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 De elopmen Economics
jou nal homepage: www.else ie .com/loca e/de ec
Regula a icle
Co po a e Social Responsibili y along he global alue chain✩
Philipp He kenho a, Sebas ian K au heim b,c,∗, Finn Ole Sem aud, F auke S eglich d
aUni e si y o Mainz, Ge many
bUni e si y o Passau, Ge many
cCESi o, Ge many
dKiel Ins i u e o he Wo ld Economy, Ge many
ARTICLE INFO
Da ase link: h ps://gi hub.com/F aukeS eglic
h/CSR-along- he-GVC
JEL classi ica ion:
F12
F61
F63
L23
M14
O12
Keywo ds:
Co po a e Social Responsibili y
Global alue chains
Incomple e con ac s
Sus ainable de elopmen
E hical sou cing
India
ABSTRACT
Loca ing subs an ial pa s o he p oduc ion p ocess in de eloping and eme ging economies, many i ms
ace an inc easing demand by s akeholde s o Co po a e Social Responsibili y (CSR) along hei alue
chains. Con ac ual incomple eness be ween i ms and hei supplie s a di e en s ages o p oduc ion can
exace ba e he abili y o mee hese demands. We analyze a model o sequen ial p oduc ion wi h incomple e
con ac s whe e CSR by independen supplie s di e en ia es he inal p oduc in he eyes o ca ing consume s.
P og essing down he alue chain, ou model p edic s an inc easing CSR p o ile om ups eam supplie s wi h
low CSR o downs eam supplie s wi h highe CSR. We con i m his p edic ion using Indian i m-le el da a
– compu ing a i m’s alue chain posi ion by combining i s p oduc -le el sales in o ma ion wi h he Wo ld
Inpu –Ou pu Da abase. We ind ha mo e downs eam i ms epo highe CSR expendi u es as measu ed by
a combina ion o s a wel a e spending and social communi y spending.
1. In oduc ion
Pa icipa ion in global alue chains (GVCs) is o en conside ed
a pa h o de elopmen o low- and middle-income coun ies
(OECD/WTO,2013;Wo ld Bank,2020). Conce ns abou nega i e social
and en i onmen al impac s o GVCs in hose coun ies a e, howe e ,
widesp ead and boos ed by agedies such as he Rana Plaza collapse
in 2013. Consume s ha e esponded o hese conce ns by becoming
inc easingly ocal abou hei demand o e hical s anda ds along he
en i e alue chain. Wi h la ge pa s o oday’s alue chains loca ed in
de eloping and eme ging economies cha ac e ized by high leniency
✩We would like o hank Julia Cajal-G ossi, Ron Da ies, Ben Fabe , Gab iel Felbe may , Sou a el Gi ma, Robe Gold, Holge Gö g, Aoi e Hanley, Roza Khoban,
K isz ina Kis-Ka os, Ca ol Newman, Ma celo Ola eaga, Gianma co O a iano, Philip Sau é, Jan Schymik, Raine Thiele, Fa id Toubal, Gonzague Vannoo enbe ghe,
Thie y Ve die , Yo o Yo o , pa icipan s o he HSU-I W Wo kshop 2019, he PEGNe Con e ence 2019, he EGIT 2020, he launch e en o he Resea ch Ne wo k
Sus ainable Global Supply Chains 2021, he CESi o A ea Con e ence on Global Economy 2021, he Wo kshop ‘Shaping Globaliza ion’ a JGU Mainz 2021, he ETSG
2021, he Gö ingen Wo kshop on In e na ional Economics 2022, he TRISTAN Wo kshop a he Uni e si y o Bay eu h 2022, as well as semina pa icipan s
a T ini y College Dublin, Uni e si y o Passau, Uni e si y o No ingham, Uni e si y o Mainz, Kiel Ins i u e o he Wo ld Economy (I W), Kiel Cen e o
Globaliza ion (KCG), he Uni e si y o Gö ingen, and he Uni e si y o Pa is 1 Pan héon-So bonne. K au heim g a e ully acknowledges inancial suppo by he
Deu sche Fo schungsgemeinscha (DFG, Ge man Resea ch Founda ion) – g an no. 421074032; Sem au and S eglich acknowledge inancial suppo by he Leibniz
Associa ion – g an no. SAS-2016-I W-LWC.
∗Co esponding au ho a : Uni e si y o Passau, Ge many.
E-mail add ess: [email p o ec ed] (S. K au heim).
wi h ega d o en i onmen al o labo egula ion and/o en o cemen ,
his equi es ha supplie s in es in Co po a e Social Responsibili y
(CSR). CSR in es men s – by de ini ion – ex end beyond local legal
equi emen s (e.g., McWilliams and Siegel,2001) and a e he e o e
key o mee ing he demands o an e e g owing numbe o e hically
conce ned consume s and, mo e b oadly, o he gene al accep ance (o
no ) o globalized p oduc ion in he public deba e.
Pe sis en and sizable compliance issues conce ning labo and en-
i onmen al p ac ices in he alue chains o a la ge numbe o i ms
indica e, howe e , how di icul i can be o i ms o impose CSR
h ps://doi.o g/10.1016/j.jde eco.2023.103236
Recei ed 21 Sep embe 2022; Recei ed in e ised o m 1 Decembe 2023; Accep ed 2 Decembe 2023
Jou nal o De elopmen Economics 167 (2024) 103236
2
P. He kenho e al.
Fig. 1. Co ela ion be ween i ms’ CSR and GVC posi ion.
Co po a e Social Responsibili y (CSR) spending plo ed agains an ups eamness measu e compu ed a he i m-le el ollowing An às and Cho (2018). Own illus a ion based on
Indian i m-le el da a by P owess, p esen ed in Sec ion 3. 54,394 obse a ions summa ized in 100 bins. The nega i e slope o 0.14 is signi ican a he 1%-le el and sugges s ha
mo e ups eam i ms ha e lowe CSR expendi u es han mo e downs eam i ms.
le els on hei in e na ional supplie s, e.g., h ough legally binding
codes o conduc (e.g., Locke,2013b). In line wi h his obse a ion,
he li e a u e on CSR has iden i ied incomple e con ac s as a key ma ke
impe ec ion in he analysis o CSR (see C i o and Fo ge ,2015). No
ocusing on CSR, bu on he in e na ional o ganiza ion o p oduc-
ion, An às and Cho (2013) analyze he ole o incomple e con ac s
o he i m-supplie in e ac ion and ind ha he alue chain posi ion
o a supplie a ec s i s con ibu ion o he p oduc ion p ocess.
In his pape , we apply he in ui ion o An às and Cho (2013) o
he analysis o CSR in es men s along he global alue chain. Mo e
speci ically, we ask he ollowing ques ions: Which ole does he alue
chain posi ion o supplie s play in hei CSR in es men s? Wha a e
he esul ing pa e ns o CSR along a alue chain, om ups eam o
downs eam p oduc ion s ages?
Fig. 1 p o ides a i s glance a he ela ionship be ween CSR
spending and he alue chain posi ion, he e o Indian i ms. The igu e
highligh s a s a is ically signi ican uncondi ional co ela ion: mo e
ups eam i ms end o ha e lowe CSR in es men s han he mo e
downs eam i ms, sugges ing an inc easing CSR p o ile ‘‘along he
alue chain’’, i.e., om ups eam o downs eam (in Fig. 1, om igh
o le ).1
We make wo main con ibu ions. Fi s , we analyze a model o non-
con ac ible CSR in es men s by supplie s in a sequen ial p oduc ion
p ocess (a alue chain). In ou model, CSR in es men s o supplie s
in he Global Sou h p o ide ho izon al and e ical p oduc di e -
en ia ion when selling he inal p oduc o ca ing consume s in he
Global No h. We ind ha he equilib ium CSR in es men s a e low in
ups eam s ages, inc ease along he alue chain, and peak a he mos
downs eam s ages. Second, we es his p edic ion employing Indian
i m-le el da a. Combining de ailed in o ma ion on a i m’s p oduc mix
wi h he Wo ld Inpu –Ou pu Da abase (WIOD), we cons uc a measu e
o a i m’s GVC posi ion o empi ically in es iga e how i a ec s a
i m’s mone a y CSR spending. Ou empi ical analysis suppo s he
heo e ical p edic ion: CSR expendi u e inc eases along he alue chain
i.e., om ups eam o downs eam i ms, con i ming he uncondi ional
co ela ion in Fig. 1. We complemen hese esul s p o iding e idence
ha poin s a he ele ance o incomple e con ac s o his pa e n.
1De ails on he da a as well as de ini ions a e p o ided below.
In ou model, i ms and consume s a e loca ed in he Global No h,
while supplie s a e loca ed in he Global Sou h and ope a e unde
lenien egula ion. We in oduce a g oup o ‘ca ing’ consume s who
app ecia e a p oduc ’s ‘e hical quali y’ embodied in en i onmen al and
wel a e imp o emen s (CSR) implemen ed by supplie s along he alue
chain.2In line wi h a common app oach bo h in Economics and in
he Business li e a u e (see e.g., Ki zmulle and Shimshack,2012,C i o
and Fo ge ,2015, and Ding e al.,2022, espec i ely, and e e ences
he ein), we model CSR as p o iding p oduc di e en ia ion allowing
he i m o cha ge a (highe ) ma kup.3
A cen al p emise o ou analysis is ha supplie s make hei deci-
sions on CSR in an en i onmen cha ac e ized by incomple e con ac s:
he i m selling he inal good o consume s canno w i e binding
con ac s p esc ibing p oduc ion s anda ds in supplie ac o ies. We
a gue in de ail in Sec ion 2.1 ha o he case o CSR, con ac s
a e incomple e as he supplie ’s ac ion is ‘‘obse able, bu no e i-
iable’’ (Ha and Moo e,1999, p. 118).4Cen al o he a gumen is
he di e ence be ween ci cums an ial e idence on a supplie ’s in inge-
men s o e hical s anda ds (obse abili y) on he one hand and, on he
o he hand, e idence ha will hold in a cou o law ( e i iabili y).
While ci cums an ial e idence can be su icien o ca ing consume s
o e-assess hei alua ion o he p oduc and he e o e demand, i is
2Su eys (e.g., O’Rou ke,2005, and Lou ei o and Lo ade,2005) as well
as ield expe imen s wi h eal pu chasing decisions (e.g., Hiscox and Smy h,
2011, and Hainmuelle e al.,2015) sugges ha consume s do ca e abou he
e hical con en o hei consump ion and in ac ha e a highe willingness o
pay o ‘e hical’ p oduc s.
3No e ha his s anda d iew on CSR as p o iding p oduc di e en ia ion is
concep ually di e en o an app oach whe e CSR ac s like an insu ance agains
NGO campaigns and consume boyco s. The wo app oaches a e no mu ually
exclusi e and sha e some common ea u es. In ou iew, including massi e and
well-media ized NGO campaigns in o he analysis would also ha e implica ions
o he numbe o e hically conce ned consume s. In o de o keep he analysis
concise, we keep he ac ion o ca ing consume s cons an and ocus o CSR
as p o iding e hical p oduc di e en ia ion.
4While en o cemen by MNEs can imp o e compliance, moni o -
ing by MNEs is impe ec and canno be expec ed o ensu e ull
compliance (Boud eau,2022).
Jou nal o De elopmen Economics 167 (2024) 103236
3
P. He kenho e al.
insu icien in a cou o law, especially so in coun ies wi h weak legal
ins i u ions.5
We in oduce he abo e elemen s – ca ing consume s and incom-
ple e con ac abili y o CSR – in o a model o sequen ial p oduc ion
along he lines o An às and Cho (2013). We ind an inc easing p o ile
o he supplie s’ equilib ium CSR in es men s along he alue chain:
while CSR in es men s a ups eam s ages a e low, hey inc ease along
he alue chain owa ds mo e downs eam s ages. This pa e n a ises
because unde incomple e con ac s, he cumula i e CSR in es men
in p e ious s ages p o ides he supplie wi h an incen i e o choose
a highe CSR le el. We de i e he comple e indus y equilib ium wi h
ee en y and ind ha endogenizing he numbe o i ms has impo an
implica ions o iden i ying es able implica ions. While in he pa ial
equilib ium ( aking agg ega e a iables as gi en), CSR in es men s
inc ease in he o al expendi u e on e hically di e en ia ed goods, his
e ec anishes in he indus y equilib ium. This lea es us wi h a single
and clea es able implica ion o ou model: he inc easing p o ile o
CSR along he alue chain, which we ake o he da a in Sec ion 3.
Se e al policy implica ions di ec ly eme ge om ou heo y. Fi s ,
as highe CSR le els a ups eam s ages encou age CSR in es men s
a mo e downs eam s ages (sequen ial complemen a i y), policy in e -
en ions a ge ing CSR a ups eam s ages may bene i CSR along he
en i e alue chain. Second, o go e nmen s aiming a decen wo k and
economic g ow h (as speci ied in he UN’s Sus ainable De elopmen
Goal 8), he ac ha he mo e ups eam s ages o p oduc ion ea u e
he lowes CSR le els, p o ides an incen i e o s a egically ocus in-
dus ial policy owa ds specializa ion in mo e downs eam p oduc ion
s ages. Thi d, policy ini ia i es like he EU Di ec i e on Co po a e
Sus ainabili y Due Diligence, he F ench Du y o Vigilance Law o he
Ge man Due Diligence Law shi he esponsibili y o supplie ac ions
o i ms. Ou analysis hin s a he limi ed abili y o a i m o impose
e hical s anda ds on i s supplie s and highligh s ha his p oblem
is pa icula ly se e e o supplie s a ups eam s ages o p oduc ion.
Fu u e due diligence egula ion should ake his asymme y along he
alue chain in o accoun .
We es ou model’s p edic ion using he Indian i m-le el da ase
P owess p o ided by he Cen e o Moni o ing Indian Economy (CMIE),
which has been used by Goldbe g e al. (2010a), Goldbe g e al.
(2010b), De Loecke e al. (2016), Bos and Vannoo enbe ghe (2018), as
well as Ba ows and Olli ie (2021), among o he s. I is a special ea u e
o his da ase ha i con ains in o ma ion on he i ms’ s a wel a e
spending as well as social communi y spending. We use hese i ems o
cons uc a measu e o a i m’s CSR spending in mone a y uni s, which
we obse e o 15,512 medium-size and la ge i ms be ween 2000 and
2013 in India.
Ou app oach o compu ing a i m’s alue chain posi ion pa allels
he p ocedu e in Cho e al. (2021). Fi s , we use he Wo ld Inpu –
Ou pu Da abase (WIOD) and apply he me hodology om An às e al.
(2012) and An às and Cho (2018) o compu e he yea ly ups eamness
o each Indian indus y. Second, o each p oduc in a i m’s po olio,
we assign he alue o annual ups eamness o i s indus y and calcula e
he i m’s alue chain posi ion, i s ups eamness, as he sales-weigh ed
a e age o i s p oduc s’ ups eamness alues.
We es ima e he ela ionship be ween a i m’s alue chain posi ion
on i s CSR expendi u e. In ou analysis, we con ol o a la ge se o
5C i o and Fo ge (2015) su ey a li e a u e on CSR in managemen
sciences and indus ial o ganiza ion. They lis impe ec compe i ion and
con ac incomple eness as cen al d i e s o CSR (see, e.g., hei Figu e 1,
p. 114). Schille (2018) mo i a es his pu ely empi ical wo k alluding o he
same no ion o incomple e con ac s: he inabili y o a i m o ully con ol
CSR in es men s o i s supplie s. Mo eo e , he e is a ecen and quickly
expanding li e a u e on ela ional con ac s in de eloping coun ies su eyed
in Macchia ello (2022) and Boud eau e al. (2023), which lends suppo o
ou modeling on incomple e con ac s in alue chains in ol ing de eloping
and eme ging economies.
i m-le el de e minan s o CSR iden i ied in he li e a u e (e.g., New-
man e al.,2018;Gö g e al.,2018;Schille ,2018). These include a
measu e o in e na ionaliza ion (sha e o expo s in o al sales), expo-
su e o mo e demanding cus ome s (measu ed by he ac ion o expo s
o OECD coun ies), local embeddedness (sha e o domes ically sou ced
inpu s), i m size ( o al sales), age and i m owne ship (dummies o
s a e-owned and o eign-owned i ms). Impo an ly, we con ol o he
possibili y ha wage le els may a y sys ema ically along he alue
chain using a i m-speci ic wage measu e ha combines indus y-le el
in o ma ion on labo compensa ion wi h a i m’s p oduc mix.6
In ou es ima ions, we ely on wo di e en ypes o a ia ion. Fi s ,
in a c oss-sec ional speci ica ion, we con ol o se e al es ablished
i m-le el d i e s o CSR and indus y-yea as well as s a e ixed e ec s.
This app oach exploi s a ia ion in he alue chain posi ion o i ms
ha sha e he same main indus ies. Second, in a mo e demanding
speci ica ion, we exploi he ime dimension o ou da a. By including
i m ixed e ec s, we iden i y he e ec om changes in a i m’s
GVC posi ion o e ime. In bo h cases, we ind s ong suppo o ou
model’s p edic ion: mo e downs eam i ms ha e highe CSR le els.
Compa ing i ms a he 5 h and he 95 h pe cen ile, ou p e e ed
speci ica ion indica es ha p edic ed CSR spending is 55.31% highe
o he downs eam i m compa ed o he ups eam i m.
We complemen he main empi ical indings o ou pape by in-
es iga ing he ole o incomple e con ac s o hese esul s. We use
sub-na ional Wo ld Bank Doing Business da a on cou cos s o measu e
judicial ine iciency o 17 Indian s a es. As he da a a e only a ailable
o 2009, we a e limi ed o a c oss-sec ional analysis using one yea
o ou da a. Despi e he es ic ed sample, we ind ha , in line wi h
ou model, CSR spending inc eases wi h a s eepe slope along he
alue chain in egions wi h high judicial ine iciency, i.e., he e ec
o incomple e con ac s p edic ed by he model is mo e p onounced in
egions wi h lowe con ac en o cemen .
Ou esul s also hold in se e al obus ness checks. In one o hem,
we add ess he possible conce n ha mo e downs eam p oduce s
migh simply be mo e isible o inal consume s and he e o e ace
highe p essu e o in es in CSR. This would cons i u e a comple-
men a y channel also linking downs eamness o highe CSR spending.
We he e o e con ol o isibili y by adding ma ke ing expenses as an
addi ional con ol a iable and ind ha ou e ec is obus o i s inclu-
sion. In a second obus ness check, we show ha ou esul s a e obus
o using CSR no in le els bu ela i e o expendi u e as independen
a iable. Mo eo e , we show ha ou esul s a e obus o smoo hing
CSR spending o e wo yea s o con ol o he possibili y ha i ms’
CSR spending is concen a ed in a speci ic yea . Fu he , ou esul s
a e obus o applying in e se hype bolic sine (i.h.s.) ans o ma ion
as an al e na i e o Poisson Pseudo Maximum Likelihood (PPML) o
include obse a ions wi hou CSR spending. Finally, we accoun o he
possibili y ha i m-speci ic shocks may d i e bo h CSR spending and
downs eamness using an ins umen al a iable app oach. Also in his
case ou esul s a e con i med.
Ou wo k ela es o he ecen li e a u e s udying he ‘‘globaliza ion
backlash’’, he massi e skep icism economic globaliza ion encoun e s in
many coun ies (Pa cnik,2017), which is su eyed in Colan one e al.
(2022). We a e in e es ed in an addi ional and complemen a y sou ce
o skep icism owa ds globaliza ion: he ongoing iola ion o e hical
s anda ds in in e na ional p oduc ion — a ac ing c i icism om con-
sume s and ac o s in ci il socie y. Mo eo e , while he li e a u e abo e
6Con olling o i m-speci ic wages add esses he ollowing conce n: I
s a wel a e spending is a pa o he o e all compensa ion package, down-
s eam i ms migh o e sys ema ically highe pay due o highe skill in ensi y,
o example. A nega i e co ela ion be ween ups eamness and CSR spending
(as measu ed in pa by s a wel a e spending) could hen be in e p e ed
as ac ually e lec ing he nega i e ela ionship be ween ups eamness and
employee compensa ion.
Jou nal o De elopmen Economics 167 (2024) 103236
4
P. He kenho e al.
ocuses on public egula ion, we a e in e es ed in pa e ns o olun a y
sel - egula ion along in e na ional alue chains.
The e a e se e al s ands o li e a u e ha ocus on he (conse-
quences o ) he demand o e hical minimum s anda ds in in e na-
ional alue chains. In he absence o e ec i e in e na ional egula ion
(Ba aglini and Ha s ad,2020), his demand mani es s i sel as p essu e
on i ms o sel - egula e in e na ional p oduc ion, a p ocess Ba on
(2003) e ms p i a e poli ics.InBa on (2012), he a gues ha key ac o s
o p i a e poli ics a e ad ocacy NGOs which use bo h con on a ional
s a egies (like consume boyco s) as well as coope a i e app oaches
(like labels) o di ec ly in luence he ac ions o i ms (ins ead o a -
ec ing go e nmen policies in he case o public poli ics). Ou analysis
he e o e ela es o he li e a u e on labels, wi h he ai ade label
being he bes documen ed and mos esea ched case in poin .7The
o he s and o li e a u e is a ecen and expanding li e a u e on he ole
o in e na ional ad ocacy NGO ac i i y o in e na ional p oduc ion
and de elopmen .8The la e is mo e ocused on NGO campaigns and
consume boyco s. Ou objec o s udy, CSR along he alue chain,
is close o he o me , as i emphasizes he oppo uni y o i ms
o inc ease demand and ma kups, a he han he isk o acing a
de as a ing boyco . In ac , he p o ision o labels by NGOs can be seen
as a mechanism o s eng hen he posi i e e ec o CSR in es men s on
demand. In ou pape , we ake his posi i e e ec as a gi en and s udy
how i shapes CSR in es men s along he alue chain.9
The e is a ecen and quickly expanding li e a u e on ela ional
con ac s in de eloping coun ies su eyed in Macchia ello (2022)
and Boud eau e al. (2023), which suppo s ou no ion o he ele ance
o incomple e con ac s in sou cing ela ionships especially in ol ing
de eloping and eme ging economies.10 Mos in o ma i e o ou mod-
eling (see Sec ion 2.3.1), Cajal-G ossi e al. (2023) show ha supplie s
in he Bangladeshi ga men sec o ob ain highe p ices when selling
o ela ional buye s han when selling o spo buye s. While hese
pape s p o ide aluable insigh s in o he con ac ing ela ionships o
o e come con ac ual incomple eness in an en i onmen wi h weak
legal ins i u ions, hey do no analyze he ela ion o CSR in es men s.
Ou heo y is oo ed in he la ge li e a u e s udying he ole o
incomple e con ac s o he in e na ional bounda ies o he i m, based
on he seminal con ibu ions o An às (2003) and An às and Helpman
(2004), su eyed in An às and Yeaple (2014) and An às (2016). We
build ou modeling on (An às and Cho ,2013) whe e a i m’s alue
chain is modeled as a sequen ial p oduc ion p ocess. Like mos o
his li e a u e, hey ocus on he in eg a ion s. ou sou cing decision
along he alue chain. Cos ino e al. (2013) also model in e na ional
sou cing as a sequen ial p oduc ion p ocess, bu wi hou incomple e
con ac s. They ocus on he endogenous agmen a ion o he alue
7The ai ade label is s udied empi ically, among o he s, by Hainmuelle
e al. (2015), de Jan y e al. (2015), and D agusanu e al. (2022), and heo e -
ically by S ähle and Richa dson (2014) and Podho sky (2015). See D agusanu
e al. (2014) o a su ey. Fo s udies o labels mo e gene ally, see Fische and
Lyon (2014), Heyes and Ma in (2017), as well as Po e (2019).
8See, e.g., Aldashe and Ve die (2009,2010), Aldashe e al.
(2015), K au heim and Ve die (2016), Koenig e al. (2021), Fon agné and
Lima di (2023).
9Closes o ou model in e ms o he ype o policy conside ed is Al a o-
U eña e al. (2022) who de elop a quan i a i e gene al equilib ium model o
s udy he e ec s o esponsible sou cing in o igin coun ies based on da a on
a ilia es o mul ina ionals in Cos a Rica.
10 Macchia ello (2010) uses a model wi h incomple e con ac s o s udy he
e ec o e ical in eg a ion on he pe asi e use o subcon ac ing a ange-
men s among small i ms in de eloping coun ies. Casabu i and Macchia ello
(2019) show how impe ec con ac en o cemen gene a es ba ie s o en y in
he Kenyan dai y sec o . Machia ello and Mo ja ia (2021) highligh he impo -
ance o long- e m ela ional con ac s in he Rwandan co ee sec o . B ugués
(2020) analyzes he ole o impe ec con ac s in he ex ile, pha maceu ical
and cemen sec o s in Ecuado depending on ma ke powe .
chain ac oss coun ies. We abs ac om bo h he in e na ional make-
o -buy decision and he alloca ion o asks ac oss coun ies. Ou main
ocus, in con as , is on CSR in es men s along a sequen ial alue chain
cha ac e ized by incomple e con ac s. Ou modeling opens up bo h he
heo e ical and he empi ical oolki o he li e a u e on in e na ional
sou cing wi h incomple e con ac s o he analysis o social and en i-
onmen al issues in globalized p oduc ion ha ea u e p ominen ly in
he public deba e. Closes o ou pape in his espec is ecen wo k
by He kenho and K au heim (2022), who analyze a model ea u ing
ca ing consume s, consume boyco s and incomple e con ac s o s udy
he e ec s on he in e na ional make-o -buy decision in a se ing wi h
a single supplie . We, in con as , s udy he de e minan s o CSR
in es men s based on a model ha ea u es a con inuum o supplie s
along a sequen ial alue chain. This allows us o s udy he e ec o a
supplie ’s alue chain posi ion on i s CSR in es men s and he e o e he
p o ile o CSR along he alue chain.
The e is an ex ensi e empi ical s and o his li e a u e, wi h Al a o
e al. (2019) being a ecen example. These pape s di ec ly ela e o he
heo e ical unde pinnings, whe e incomple e con ac s lead o unde -
in es men in a ela ionship-speci ic non-con ac ible a iable (inpu s,
quali y-adjus ed inpu s, e c.). As hese non-con ac ible a iables a e
no o iously di icul o obse e in he da a, he empi ical s udies u n
o he obse able op imal esponse o he i m o he unde in es men :
he in eg a ion s. ou sou cing decision. An às and Cho (2013), o
example, in es iga e he ole o ups eamness o a supplie o i s –
empi ically unobse able – quali y-adjus ed in es men . To gene a e
es able implica ions, hey he e o e u n o he obse able op imal
esponse o he i m o he unde in es men : depending on he up-
s eamness o he supplie , ei he in eg a ion o ou sou cing deli e s
he igh incen i es minimizing he nega i e consequences o he unde -
in es men . Ou empi ical analysis complemen s he exis ing li e a u e
as we, in con as , di ec ly obse e he non-con ac ible a iable o
in e es : CSR spending. This allows us o in es iga e mo e di ec ly he
ole o he GVC posi ion o supplie in es men s unde incomple e
con ac s.
The e is a la ge li e a u e on CSR a he in e sec ion o he ields
o managemen science and indus ial o ganiza ion. Ki zmulle and
Shimshack (2012), C i o and Fo ge (2015), and Pisani e al. (2017)
p o ide su eys o his li e a u e. This includes a la ge body o empi i-
cal s udies on CSR which mainly ocuses on he e ec , a he han he
de e minan s, o CSR. We ake a di e en angle, con ibu ing o a be e
unde s anding o he de e minan s o CSR by analyzing he impac o
a i m’s alue chain posi ion. Mo eo e , mos o his li e a u e uses
indices combining quali a i e measu es ob ained by manda o y CSR
epo ing o lis ed i ms in de eloped a he han eme ging economies
(e.g., Ma ano and Kos o a,2016;Pisani e al.,2017). We, in con as ,
use da a om India, a key eme ging economy, which no only allows us
o cons uc a con inuous mone a y measu e o CSR, bu also o s udy
he beha io o non-lis ed medium-sized and la ge i ms.
Ou empi ical analysis complemen s a small numbe o empi ical
s udies assessing he ole o global alue chain ela ionships o CSR.
No ewo hy con ibu ions a e Schille (2018), Newman e al. (2018),
and Tanaka (2020). Schille (2018) uses da a es ic ed o la ge publicly
lis ed companies o show ha co po a e en i onmen al and social poli-
cies o i ms a e ansmi ed o supplie s in hei alue chain and s udies
he e ec o hese policies on i m pe o mance. He p o ides e idence
consis en wi h he in e p e a ion ha despi e incomple e con ac s,
i ms can s ill a ec hei supplie ’s CSR choices o some ex en . We,
in con as , a e in e es ed in how he limi ed con ol o a i m o e i s
supplie s shapes hei CSR choices.
Mos ela ed o ou empi ical analysis a e Newman e al. (2018)
and Tanaka (2020). Fo i ms in he ga men sec o in Myanma , Tanaka
(2020) shows ha expo ing imp o es wo king condi ions, like i e
sa e y, heal h managemen , and eedom o nego ia ion. Newman e al.
(2018) exploi CSR indica o s ha cap u e compliance wi h labo s an-
da ds, communi y and managemen - ela ed CSR p ac ices o analyze
Jou nal o De elopmen Economics 167 (2024) 103236
5
P. He kenho e al.
he ansmission o socially esponsible beha io h ough ade in
Vie namese i ms. They ind ha a i m’s CSR inc eases wi h expo ing
and impo ing, a ying ac oss expo des ina ions. The in e na ional-
iza ion o supply chain ela ionships he e o e seems o ma e o CSR
ac i i ies o i ms. Mo eo e , Newman e al. (2018) conjec u e ha
in addi ion o he in e na ionaliza ion o buye –selle ela ionships,
he supplie ’s posi ion along he alue chain may ma e and con ol
o i using i m ixed e ec s. We, in con as , use a ime- a ying,
i m-speci ic measu e o a i m’s alue chain posi ion ha allows us
o di ec ly measu e he GVC posi ion o a i m and o analyze i s
ela ionship o he i m’s CSR spending.
The emainde o he pape is s uc u ed as ollows. Sec ion 2
p esen s ou heo e ical analysis o CSR along he global alue chain.
Sec ion 3ou lines ou empi ical analysis. Sec ion 4concludes.
2. A model o CSR along he global alue chain
In his sec ion, we inco po a e consume alua ion o e hical p o-
duc ion p ac ices and CSR in es men in o a model o sequen ial p o-
duc ion wi h incomple e con ac s based on An às and Cho (2013).
We assume ha each i m in he Global No h has a alue chain wi h
uni measu e o equi ed inpu s, which a e all p o ided by independen
supplie s in he Global Sou h, whe e en i onmen al and labo egu-
la ion is insu icien om he pe spec i e o ca ing consume s in he
Global No h. We de i e he indus y equilib ium o he model and
show ha ou heo y implies an inc easing CSR p o ile along he alue
chain — a p edic ion which we ake o he da a in Sec ion 3.
2.1. Incomple e con ac ibili y o CSR
Be o e u ning o he ou line o ou model, we i s plead he
case o he ele ance (and he ype) o incomple e con ac s o CSR
along global alue chains. Assu ing su icien CSR le els by supplie s
is no o iously di icul o achie e. Nike’s esponse o he ‘‘swea shop
campaign’’ in he 1990s is p obably he bes documen ed and mos
esea ched case in poin . P o es s agains wo k condi ions in Indonesian
supplie ac o ies led Nike o implemen compulso y, legally binding
codes o conduc o all i s supplie s.11 Ha ison and Sco se (2010)
show ha he campaigns caused an inc ease in he eal wages in
Indonesian ac o ies manu ac u ing o Nike. Locke (2013a), howe e ,
shows ha mo e han a decade a e he in oduc ion o legally bind-
ing codes o conduc combined wi h an audi ing sys em, subs an ial
compliance p oblems pe sis ed in Nike’s supply chain. He analyzes da a
om ac o y audi s o wo king condi ions in mo e han 900 o Nike’s
supplie s loca ed ac oss i y coun ies and inds ha , despi e he ac
ha all supplie s a e obliged o sign he codes o conduc and despi e
la ge in es men s in o moni o ing, he e s ill is a subs an ial numbe o
supplie s ha a e anked as ‘‘noncomplian ’’. In Asia, he noncomplian
supplie s e en cons i u e he majo i y. While Nike is a pa icula ly well-
documen ed case in poin , hese p oblems a e by no means speci ic o
Nike o he oo wea and appa el sec o . Based on da a om one o he
wo ld’s la ges supply chain audi ing i ms con aining 16,795 audi s o
5,819 ac o ies in 13 indus ies ac oss 66 coun ies o e a pe iod o
six yea s, Sho e al. (2016,2020) documen widesp ead iola ions o
codes o conduc in a eas like child labo , o ced labo , wo king hou s,
occupa ional heal h and sa e y, he minimum wage and disciplina y
p ac ices.
In ou iew, his calls o a model o CSR along GVCs ha explici ly
accoun s o incomple e con ac s be ween a i m and i s supplie s. One
11 In May 1998, Nike CEO Phil Knigh amously made he ollowing s a e-
men : ‘‘The Nike p oduc has become synonymous wi h sla e wages, o ced
o e ime, and a bi a y abuse. I uly belie e he Ame ican consume does
no wan o buy p oduc s made unde abusi e condi ions.’’ (New Yo k Times,
1998).
s anda d way o mo i a ing con ac ual incomple eness is ha i is
p ohibi i ely cos ly – i possible a all – o co e all con ingencies in
a con ac . This does no seem o be a majo conce n in ou con ex :
codes o conduc and simila con ac s can speci y minimum social,
labo and en i onmen al s anda ds. In he case o CSR in es men s, he
con ac ual incomple eness s ems om ano he s anda d a gumen : he
ac ions o he supplie a e ‘‘obse able bu no e i iable’’ (e.g., Ha
and Moo e,1999). This means ha i is cos less (o a leas ela i ely
cheap) o de e mine wha is going on in a supplie ’s ac o y bu i is
p ohibi i ely cos ly o hold he supplie accoun able o he induced
losses in a cou o law.
The e a e qui e gene al and well-known p oblems wi h e i ying
labo and en i onmen al s anda ds implemen ed in p oduc ion plan s,
especially so in low- egula ion coun ies. Sa e y equipmen may be
dis ibu ed one day, bu no he o he , i e exi s may be accessible
one day, bu blocked he o he , wo ke s may ea he loss o hei
jobs, i hey epo o ced and excessi e o e ime, oxic was e may
be disposed in o a i e when no-one is wa ching.12 I is ne e heless
ela i ely cheap o , say, an ad ocacy NGO o alk o wo ke s, measu e
oxins in was e wa e o send an unde co e agen in o a ac o y o
ind ou abou he ac ual p oduc ion condi ions. Such ci cums an ial
e idence, possibly ob ained by b eaking domes ic laws, is insu icien
in a cou o law. Howe e , in many cases i is su icien o con ince
ca ing consume s o e-assess hei alua ion o he p oduc and o
educe demand.13 P oducing he ype o e idence ha would lead
o a con ic ion o he supplie – hough in p inciple possible – will
in many cases be p ohibi i ely cos ly o p oduce. I would equi e a
massi e and con inuous p esence o audi o s on si e and e en hen,
he e is no gua an ee ha hey will win he ca -and-mouse game wi h
local managemen . This highligh s ha he no ion in he incomple e
con ac s li e a u e ha he ac ion is ‘‘obse able bu no e i iable’’
(e.g., Ha and Moo e,1999) applies o CSR in es men s along GVCs.14
2.2. Se up: P e e ences and echnology
We now u n o he o mal desc ip ion and analysis o ou model
s a ing wi h p e e ences and echnology. P e e ences include he sec-
ond key ea u e o ou modeling — besides incomple e con ac s o
CSR discussed abo e: consume alua ion o – and p o ision o – CSR
along he alue chain.
2.2.1. P e e ences o ca ing consume s
Consume s de i e u ili y om he consump ion o di e en a ie ies
o a inal consump ion good. We assume ha all a ie ies sha e he
same physical p ope ies. The e a e wo ypes o consume s in he Global
12 Fo a i id desc ip ion o he ypical p oblems o audi o s in he ield,
see Locke (2013b), p. 35–37.
13 This is all he mo e so as NGOs a e epea edly epo ed o be he
mos us ed ins i u ions among consume s be o e go e nmen s and he
media. See, o example, he Edelman T us Ba ome e o 2020, a
h ps://cdn2.hubspo .ne /hub s/440941/T us %20Ba ome e %202020/
2020%20Edelman%20T us %20Ba ome e %20Global%20Repo -1.pd .
14 In ou modeling, we assume o simplici y ha he i m and consume s
can obse e he ac ual le el o CSR by he supplie — bu ha i is no
e i iable by a cou . In case ha some eade s ake issue wi h he no ion ha
he ac ual CSR le el is obse able bu non- e i iable, no e ha his assump ion
can be elaxed — u he b oadening he gap be ween he cos o obse ing
and he cos o e i ying. Fo he mechanics o ou model o unc ion, we do
no need he i m and consume s o obse e he ac ual CSR in es men s. The
i m only needs o know he CSR le el which consume s obse e o pe cei e
(howe e misguided hei pe cep ion may be) and which hey use o o m
hei opinion on he e hical quali y, 𝜀(𝜔). This obse able/pe cei ed CSR le el
is wha de e mines he su plus o he ma ch — bu i is ob iously impossible
o condi ion on i in a con ac , e.g., by speci ying a ine in esponse o
pe cep ions o consume o allega ions by NGOs.
Jou nal o De elopmen Economics 167 (2024) 103236
6
P. He kenho e al.
No h, which we label ca ing and non-ca ing (Da ies,2005;Besley and
Gha ak,2007).
The g oup o ca ing consume s alues he implemen a ion o high
labo and en i onmen al s anda ds in he p oduc ion p ocess. Fo hem,
highe s anda ds along he alue chain ansla e in o a highe pe cei ed
e hical quali y o a ie y 𝜔,𝜀(𝜔)≥0, which ca ing consume s alue
as well as he physical uni s o a ie y 𝜔 hey consume, 𝑞(𝜔).15 The
p e e ences o ca ing consume s a e gi en by:
𝑈=(∫𝜔∈𝛺
[𝜀(𝜔)𝑞(𝜔)]𝜌𝑑𝜔)1∕𝜌
,(1)
whe e 𝛺 ep esen s he se o a ailable a ie ies. This way o model-
ing quali y is s anda d in he ade li e a u e, see, e.g., Baldwin and
Ha igan (2011) and e e ences he ein. The pe cei ed e hical quali y
o a a ie y depends o CSR le els a he di e en p oduc ion s ages
along he alue chain. We discuss his in de ail in Sec ion 2.2.4 below.
In line wi h a common app oach in he li e a u e, we model CSR as
p o iding p oduc di e en ia ion (see e.g., Ki zmulle and Shimshack,
2012 as well as Ding e al.,2022 and e e ences he ein). The deg ee
o e hical di e en ia ion is go e ned by he pa ame e 𝜌∈(0,1∕2).16
Fo simplici y, we assume ha e hical quali y is he only means o
di e en ia ion ac oss a ie ies.17
2.2.2. P e e ences o non-ca ing consume s
Non-ca ing consume s a e indi e en abou e hical quali y and
he e o e only ca e abou he physical quan i ies hey consume. As
we assume ha all a ie ies a e iden ical in e ms o hei physical
p ope ies, in hei eyes, a ie ies a e homogeneous. Thei u ili y is
gi en by:
𝑈′=(∫𝜔∈𝛺
𝑞(𝜔)𝑑𝜔)𝜌
.(2)
The absence o he pe cei ed e hical quali y in hese p e e ences as well
as he absence o e hical di e en ia ion o a ie ies as in Eq. (1) implies
ha o non-ca ing consume s, all a ie ies a e pe ec subs i u es
i espec i e o hei alue o 𝜀(𝜔).
2.2.3. Compe i ion and ma ke segmen a ion
In o de o be able o sell o ca ing consume s, a i m has o incu
a ixed cos 𝑓o ma ke ing i s ou pu as an e hical p oduc o ca ing
15 The e is ample empi ical e idence ha consume s ca e abou he e hical
con en o hei consump ion and ha e a highe willingness o pay o ‘e hical’
p oduc s. The e idence s ems om su eys (O’Rou ke,2005, and Lou ei o
and Lo ade,2005) lab expe imen s (Ba ling e al.,2015) as well as ield
expe imen s wi h eal pu chasing decision (e.g., Hiscox and Smy h,2011,
and Hainmuelle e al.,2015). Basu and Tzanna os (2003) and Cone (2013)
p o ide e idence ha his awa eness is inc easing o e ime.
16 Res ic ing 𝜌 o be smalle 1∕2 may seem o be an unusual assump ion. The
need o his assump ion a ises because he e enue unc ion becomes con ex
in inal ou pu quan i y 𝑞(𝜔)when demand is oo elas ic (la ge 𝜌), see Eq. (16)
in Sec ion 2.4. This e ec a ises in ou model because choosing a highe
e hical quali y shi s he demand unc ion ou wa d wi hou aising he ma ginal
cos o physical p oduc ion, as is usually he case in he quali y li e a u e
(e.g., Baldwin and Ha igan,2011, and e e ences he ein). In Appendix A.3
we show how his assump ion can be easily elaxed by in oducing a pa ame e
𝜈∈ (0,1) ep esen ing he scope o (e hical) quali y di e en ia ion in he u ili y
unc ion used ecen ly in Fan e al. (2015) and Bas os e al. (2018) as well
as Aghion e al. (2020). In ou benchma k model, we ha e 𝜈= 1. The smalle
𝜈, he highe alues o 𝜌 > 1∕2 a e possible up o he usual uppe bound o
one. To keep he model as simple as possible, we choose o es ic he ange
o possible alues o 𝜌 a he han o in oduce an addi ional pa ame e .
17 P oduc di e en ia ion h ough CSR is one o he main explana ions
o CSR in es men s in he li e a u e, see e.g., Ding e al. (2022) and e e -
ences he ein; see Aghion e al. (2020) o a ela ed app oach o modeling
social- esponsibili y conce ns.
consume s.18 Mo eo e , in o de o sell o ca ing consume s, he i m
will ha e o assu e a posi i e le el o pe cei ed e hical quali y, as o
𝜀(𝜔) = 0, Eq. (1) di ec ly implies ha demand om ca ing consume s
will be ze o (we discuss he de e minan s o 𝜀(𝜔)below). Fi ms ha
payed he ixed cos 𝑓, e.g., o es ablish an e hical na a i e, and assu e
𝜀(𝜔)>0, can sell o ca ing consume s. Eq. (1) implies ha hey ope a e
unde monopolis ic compe i ion and can cha ge a posi i e ma k-up.
Fi ms ha do no incu he ixed cos 𝑓canno sell o ca ing
consume s. They se e only non-ca ing consume s and he pe ec
subs i u abili y o a ie ies in Eq. (2) implies ha hey ope a e un-
de pe ec compe i ion wi h hei p ice equal o ma ginal cos s (we
speci y he ma ginal cos s below and also show ha hese en ail –
qui e in ui i ely – ze o e hical quali y o goods sold o non-ca ing
consume s).
No e ha en e ing he ma ke o e hical goods p o ides a (disc e e)
e ical di e en ia ion om he ze o-CSR p oduce s se ing he non-
ca ing consume s. A he same ime, i allows he i m o en e a con ex
whe e i s a ie y is ho izon ally di e en ia ed om he o he e hically
p oduced a ie ies. Wi h hese obse a ions, i can al eady be seen a
his poin , ha simila o e.g., Ba on (2009)o Besley and Gha ak
(2007), a sepa a ing equilib ium will a ise wi h a g oup o e hically
di e en ia ed i ms selling o ca ing consume s a high p ices and a
g oup o pe ec ly compe i i e i ms selling o non-ca ing consume s a
low p ices. None o he wo ypes o i ms would wan o (could) se e
he o he g oup o consume s simul aneously. In ou analysis o he
indus y equilib ium o ou model in Sec ion 2.5.1, we will de e mine
he equilib ium measu e o e hically di e en ia ed i ms.
2.2.4. E hical quali y and CSR along he alue chain
Fo each in e media e 𝑗, he e is a la ge numbe o ex an e iden ical
po en ial supplie s in Sou h. The i m picks (is ma ched o) one supplie
o each s age.19 Fo simplici y, we assume ha all s ages a e iden ical
ac oss a ie ies.
The pe cei ed e hical quali y o a ie y 𝜔,𝜀(𝜔), is de e mined by
he en i onmen al and labo s anda ds implemen ed in he di e en
p oduc ion s ages. We deno e he implemen ed le el o he s anda d a
s age 𝑗by 𝑠(𝑗). We assume a sequen ial p oduc ion p ocess wi h 𝑗= 0
being he mos ups eam inpu and 𝑗= 1 he mos downs eam one.
The le el o CSR o he supplie a s age 𝑗is de ined as he di e ence
be ween 𝑠(𝑗)and he le el o egula ion in Sou h, which we no malize
o ze o. The e o e, 𝑠(𝑗)also ep esen s he le el o CSR implemen ed
by supplie 𝑗. We assume a cons an ma ginal cos o CSR in es men ,
𝑐𝑠, which is iden ical ac oss s ages. The CSR le els ac oss p oduc ion
s ages shape he pe cei ed e hical quali y o a ie y 𝜔:
𝜀(𝜔) = (∫1
0
𝑠(𝑗)𝛼𝑑𝑗)1∕𝛼
.(3)
Wi h he pa ame e 𝛼∈ (0,1), CSR le els ac oss p oduc ion s ages
a e g oss subs i u es. This implies ha pe cei ed e hical quali y is
ob ained by agg ega ing he CSR le els ac oss all p oduc ion s ages.
Because ou assump ion on 𝛼implies ha we main ain some deg ee
o complemen a i y be ween di e en s ages, CSR in es men s ha a e
homogeneously sp ead ou o e he en i e supply chain lead o a highe
le el o pe cei ed e hical quali y han concen a ing he in es men s in
a small ac ion o s ages. Bu how much complemen a i y is ac ually
wa an ed o CSR in es men s along he alue chain?
18 We hink o hese cos s as ma ke ing expenses simila o in es men s in
b and alue and epu a ion. In his case, one elemen o his b and epu a ion
may be a na a i e o he supe io e hical quali y o he a ie y compa ed o
a ie ies p oduced wi h inpu s sou ced on anonymous ma ke s a he lowes
possible p ice, a p ice ha in and o i sel is a odds wi h posi i e CSR le els
(see ou desc ip ion o he p oduc ion o goods o non-ca ing consume s
below).
19 In Foo no e 29, we show ha i is op imal o a i m se ing ca ing
consume s o pick only one supplie o each s age.
Jou nal o De elopmen Economics 167 (2024) 103236
7
P. He kenho e al.
Well-media ized bu a e e en s wi h s ong impac on demand like
he collapse o he Rana Plaza ac o y in 2013 o he Nike swea shop
campaign seem o sugges ha ex eme le els o complemen a i y may
be app op ia e when a single ‘‘bad apple’’ in he p oduc ion chain jeop-
a dizes o al sales. Ou aim, howe e , is no o model he e ec o such
e en s on he indi idual i ms in ol ed. Ou modeling a he conce ns
he impac o a gene al awa eness o and demand o e hical p oduc ion
(which may be spu ed by such excep ional e en s). Wi h CSR being
de ined as in es men s abo e and beyond local legal equi emen s, we
deem i plausible ha a le el o ze o in some s ages does no ha e
de as a ing e ec s on demand.
Howe e , we would like o main ain some le el o complemen a i y
be ween CSR a di e en p oduc ion s ages. Technically, his implies
ha we should hink o 𝛼as being ela i ely small, i.e., close o
ze o han o uni y. We will see below ha he alue o 𝛼 ela i e
o he alue o 𝜌ma e s in ou model. The la e also lies be ween
ze o and one (see Foo no e 16) and ep esen s he ease wi h which a
consume can subs i u e one e hically di e en ia ed a ie y o ano he
wi h he iden ical le el o e hical quali y. Gi en ha he only sou ce
o ho izon al di e en ia ion lies in he di e en e hical na a i es o
o he wise iden ical a ie ies, and gi en ou a gumen abo e ha 𝛼
should be ela i ely close o ze o, we ocus ou analysis on he case
whe e 𝛼 < 𝜌.20
2.2.5. P oduc ion o physical ou pu
Physical p oduc ion along he alue chain akes place acco ding o
a Leon ie p oduc ion unc ion. This implies ha he p oduc ion o one
uni o inal ou pu equi es a ixed quan i y o each in e media e inpu
𝑗. Fo simplici y, we assume symme y ac oss s ages and no malize all
he Leon ie coe icien s o 1. Physical ou pu o a ie y 𝜔is hen gi en
by
𝑞(𝜔) = min
𝑗{𝐼(𝑗)𝑥(𝑗)} (4)
whe e he minimum is aken o e all 𝑗∈ [0,1] and 𝑥(𝑗)is he quan i y
o inpu 𝑗used in he p oduc ion p ocess. The indica o unc ion 𝐼(𝑗)
equals one i he inpu is p o ided a he app op ia e s age and ze o
o he wise. This indica o unc ion in oduces sequen iali y in o he
p oduc ion p ocess.21
I ollows ha he quan i y o he inpu he i m sou ces a s age 𝑗
is gi en by 𝑥(𝑗) = 𝑞(𝜔).22 Fo simplici y, we assume iden ical ma ginal
p oduc ion cos s 𝑐𝑥 o all s ages. The o al p oduc ion cos o one uni
o he inal ou pu is he e o e gi en by:
∫1
0
𝑐𝑥𝑑𝑗 =𝑐𝑥.(5)
Wi h his Leon ie p oduc ion unc ion, we de ia e om An às and
Cho (2013) who assume g oss subs i u abili y ac oss s ages. Ou choice
20 In An às and Cho (2013), he pa ame e 𝛼go e ns he subs i u abili y
be ween quali y-adjus ed inpu quan i ies o he di e en supplie s (p oduc ion
s ages). Clea ly, he e i would be much mo e di icul o a gue why his –
echnical – subs i u abili y should be la ge o smalle han 𝜌.An às and
Cho (2013) hus conside bo h cases (sequen ial complemen s and sequen ial
subs i u es) and build hei empi ical analysis on his dis inc ion. We e ain
om doing so o wo easons. Fi s , we ha e a gued abo e why we hink
𝛼 < 𝜌 is app op ia e and second, o ou knowledge, he e has ye o be
ound a da ase and a me hod o empi ically de e mine he alue o alpha
( he subs i u abili y o CSR le els ac oss supplie s) o di e en sec o s.
21 This may seem like a e y agile p oduc ion p ocess: any iola ion o he
app op ia e o de ing o any ailu e o a supplie o con ibu e i s in e media e
leads o ze o ou pu . We assume, howe e , ha he i m can always buy any
inpu 𝑗on he inpu ma ke , whe e a gene ic (ze o-CSR) e sion o each inpu
is eadily a ailable a ma ginal p oduc ion cos s o supplie s. We desc ibe his
inpu ma ke in de ail in Sec ion 2.3.1. This implies ha he i m can always
assu e he igh sequencing and he equi ed quan i y o each inpu .
22 We pos pone he de ailed discussion o why his is he case o Foo no e
26, as he a gumen elies on he e ec s o he inpu ma ke in oduced below.
o p oduc ion echnology is in line, howe e , wi h o he pape s in he
li e a u e on he in e na ionaliza ion o p oduc ion like G ossman and
Rossi-Hansbe g (2008), Cos ino e al. (2013), o G ossman and Help-
man (2020). In ou iew, assuming ha inpu s along a alue chain a e
complemen s a he han subs i u es appea s plausible. Conside , o
example, he p oduc ion o a ca . Fou wheels and one s ee ing wheel
a e needed o an ope a ional ca and he e is no way o subs i u e one
ype o wheel o ano he . Ou choice is suppo ed by a ecen pape
by Boehm e al. (2019) who ind a le el o complemen a i y close o
Leon ie be ween impo ed and domes ic inpu s o Japanese a ilia es
in he U.S.23
2.3. Sou cing wi h sequen ial p oduc ion
We now desc ibe he op ions o he i m o sou ce inpu s om
independen supplie s in a sequen ial p oduc ion p ocess wi h incom-
ple e con ac s. We speci y demand by ca ing as well as non-ca ing
consume s, discuss he hold-up p oblem esul ing om incomple e
con ac s and speci y he ba gaining p ocess be ween he i m and i s
supplie s.
2.3.1. Rela ional s. Anonymous sou cing
Fi ms can choose o sou ce any inpu 𝑗ei he om a ma ched
supplie o om he inpu ma ke . While CSR in es men s o a ma ched
supplie con ibu e o he pe cei ed e hical quali y, 𝜀(𝜔), an inpu
sou ced om he inpu ma ke does no . We can hink o his as an
anonymous ma ke whe e he inpu canno be linked o a speci ic
supplie in a way ha would a ec he pe cei ed e hical quali y o
he inal p oduc . This can be he case when he iden i y o he
p oduce (s) and/o in o ma ion abou p oduc ion condi ions a e los o
no e i iable.24 The la ge numbe o po en ial supplie s o each inpu
𝑗in Sou h s ands eady o p oduce any quan i y o inpu s o he wo ld
ma ke o a p ice co e ing hei physical p oduc ion cos s.25
As CSR is cos ly and ze o-CSR inpu s a e eadily a ailable o
each p oduc ion s age om he inpu ma ke , i ms se ing non-ca ing
consume s will sou ce ze o-CSR inpu s only — i espec i e o whe he
hey sou ce om he inpu ma ke o hei ma ched supplie . Ma ginal
cos s o hese i ms he e o e amoun o he physical p oduc ion cos s
in Eq. (5), deli e ing he equilib ium p ice o ze o-CSR a ie ies o
𝑝ℎ=𝑐𝑥. Recall ha ma ke s o e hical a ie ies and ze o-CSR a ie ies
a e pe ec ly segmen ed. This he e o e concludes ou analysis o he
ma ke o ze o-CSR a ie ies sold o non-ca ing consume s. In he
emainde o he analysis, we ocus on e hically di e en ia ed a ie ies
sold o ca ing consume s. In his case, i ms need o assu e posi i e
CSR le els along hei alue chain and he e o e ha e an incen i e
23 The choice o pe ec complemen a i y has he addi ional ad an age ha
i allows us o s udy he decision o CSR in es men s in e hical quali y
independen ly o he decision on p oduc ion quan i ies. Wi hou pe ec com-
plemen a i y, ou model would imply ade o be ween quan i y and e hical
quali y in he eyes o ca ing consume s, an implica ion we nei he ind ealis ic
no appealing.
24 Fai ade co ee beans nicely illus a e he poin . The e is a ma ke o
Fai T ade co ee beans and a ma ke o ‘ egula ’ co ee beans. Due o limi ed
demand and o e -p oduc ion o he Fai T ade a me s, some o he Fai T ade
co ee beans a e no sold unde he Fai T ade label, bu a e sold as ‘ egula ’
co ee beans on he wo ld ma ke . As he p oduc ion condi ions canno be
in e ed om he co ee bean i sel , he in o ma ion is los and no buye can
use he Fai T ade label o ad e ise i s inal p oduc — e en hough he
co ee beans may ha e been p oduced espec ing he Fai T ade s anda ds. See,
o example, D agusanu e al. (2014) and de Jan y e al. (2015). Rela ed o
ou modeling o e hical quali y di e en ia ion h ough CSR, D agusanu e al.
(2022) show ha in he co ee ma ke ai ade ce i ica ion is linked o highe
sales.
25 In ecen wo k, Cajal-G ossi e al. (2023) ind pa e ns in he Bangladeshi
ga men s sec o ha a e consis en wi h ou modeling: supplie s ob ain highe
p ices when selling o ela ional buye s han when selling o spo buye s.
Jou nal o De elopmen Economics 167 (2024) 103236
8
P. He kenho e al.
o coope a e wi h hei ma ched supplie s who can choose o make a
ela ionship-speci ic CSR in es men a hei s age 𝑗o he p oduc ion
p ocess.
2.3.2. Demand and i m e enue
Maximizing u ili y o ca ing consume s in Eq. (1) subjec o he
budge cons ain gi es in e se demand as
𝑝(𝜔) = 𝑞(𝜔)−(1−𝜌)𝜀(𝜔)𝜌𝐴1−𝜌(6)
whe e 𝑝(𝜔)is he p ice o one physical uni o a ie y 𝜔and 𝐴=
𝐸∕𝑃−𝜌∕(1−𝜌)is a demand shi e consis ing o o al expendi u e 𝐸as
well as he p ice index 𝑃, which a e bo h aken as gi en by he i m.
Fi m e enue is 𝑟(𝜔) = 𝑝(𝜔)𝑞(𝜔). Inse ing (6) o 𝑝(𝜔)gi es 𝑟(𝜔) =
𝜀(𝜔)𝜌𝑞(𝜔)𝜌𝐴1−𝜌. Combined wi h (3), his implies
𝑟(𝜔) = 𝑞(𝜔)𝜌𝐴1−𝜌(∫1
0
𝑠(𝑗)𝛼𝑑𝑗)𝜌∕𝛼
.(7)
Re enues depend posi i ely on o al scale o p oduc ion, 𝑞(𝜔), he
gene al le el o demand, summa ized in 𝐴, as well as he pe cei ed
e hical quali y, which is composed o he CSR le els all along he alue
chain.
2.3.3. Hold-up and ba gaining
We main ain he assump ion on incomple e con ac s o he phys-
ical inpu 𝑥(𝑚) om he An às li e a u e. Howe e , due o ou as-
sump ion o a Leon ie p oduc ion echnology om Eq. (4), incomple e
con ac s do no lead o unde in es men in quan i ies and he ou come
is he e o e obse a ionally equi alen o a se ing o comple e con-
ac s.26 G ossman and Helpman (2020) also use a Leon ie echnology
and make a ela ed a gumen .
Because he CSR in es men makes he inpu p oduced by a supplie
a some s age 𝑚 ela ionship-speci ic, he i m and i s supplie ace a
s anda d hold-up p oblem and he need o sha e he su plus gene a ed
by he CSR in es men . In he cha ac e iza ion o he in e ac ion be-
ween he i m and he supplie , we closely ollow he se ing in An às
and Cho (2013). The i m only pays he supplie a e p oduc ion has
aken place, i.e., when he physical p oduc ion cos s a e sunk, he CSR
in es men is sunk and he i m obse es he implemen ed CSR le el.
As ou lined abo e, he i m can buy he necessa y amoun o inpu
𝑚on he inpu ma ke in case he supplie does no deli e . By he
same oken, he supplie can sell he inpu on he inpu ma ke i no
ag eemen wi h he i m is eached. This implies ha a he ba gaining
s age he supplie 𝑚has an ou side op ion o 𝑐𝑥𝑥(𝑚)and he i m has an
ou side op ion o −𝑐𝑥𝑥(𝑚). I is a s anda d esul ha in he equilib ium
o he Nash ba gaining game each pa y ecei es i s ou side op ion and
he emaining su plus is spli be ween he wo pa ies acco ding o hei
ba gaining powe , which we deno e by 𝛽 o he i m and 1 − 𝛽 o he
supplie . Gi en he symme ic ou side op ions, his implies ha a each
s age he i m pays he supplie i s p oduc ion cos s 𝑐𝑥𝑥(𝑚)as well as a
ac ion 1 − 𝛽o he emaining su plus. We ollow G ossman and Ha
(1986), who assume a i y- i y spli o he su plus, and An às (2003),
who assumes ha 𝛽 > 1∕2, and assume 𝛽≥1∕2.
26 Conce ning he op imal inpu quan i y 𝑥(𝑚)o a supplie unde incomple e
con ac s and a Leon ie p oduc ion unc ion, conside he ollowing. Fo any
gi en o de size 𝑞(𝜔)issued by he i m o all i s supplie s, a single supplie can
ei he p oduce mo e han, less han, o exac ly 𝑞(𝜔). P oducing mo e would
be ine icien , as he i m would no seek o buy i . Wi h he CSR in es men
ac ing like a ixed cos o he supplie , and he su plus o he ma ch, i.e., he
compensa ion o he supplie , inc eases in 𝑥(𝑚)un il 𝑥(𝑚) = 𝑞(𝜔)is eached.
This implies ha he supplie maximizes i s compensa ion o 𝑥(𝑚) = 𝑞(𝜔),
which is s ic ly p e e ed o ze o CSR, which deli e s ze o p o i s o any
𝑥(𝑚).
2.3.4. Inc emen al con ibu ion a s age 𝑚
As he i m is in ull con ol o he sequencing o he p oduc ion
s ages, we ha e 𝐼(𝑗) = 1 ∀𝑗 < 𝑚 when ba gaining a s age 𝑚 akes place.
Recall ha he i m can always comple e he emaining p oduc ion
s ages 𝑗∈ [𝑚, 1] wi h inpu s pu chased on he inpu ma ke , which
do no add o he pe cei ed e hical quali y o he inal p oduc . A
s age 𝑚, he i m can he e o e be ce ain o ob ain a leas he ollowing
e enues:
𝑟(𝑚) = 𝑞(𝜔)𝜌𝐴1−𝜌(∫𝑚
0
𝑠(𝑗)𝛼𝑑𝑗)𝜌∕𝛼
.(8)
These a e he e enues he i m has secu ed up o s age 𝑚when
nego ia ing wi h he supplie . The inc emen al con ibu ion o he
supplie a s age 𝑚can hen be compu ed using Leibniz’ ule:
𝑟′(𝑚) = 𝜕𝑟(𝑚)
𝜕𝑚 =𝜌
𝛼𝐴
𝛼(1−𝜌)
𝜌𝑞(𝜔)𝛼𝑠(𝑚)𝛼𝑟(𝑚)
𝜌−𝛼
𝜌.(9)
The inc emen al con ibu ion o e enue gene a ed a s age 𝑚in-
c eases in o al demand, as e lec ed by 𝐴, in he o al scale o p oduc-
ion, 𝑞(𝜔), and in he consume alua ion o he implemen ed s anda d,
𝑠(𝑚), as well as in all p e ious CSR in es men s included in 𝑟(𝑚). The
assump ion ha 𝛼∈(0, 𝜌)assu es ha he la e e ec is posi i e.27 We
ollow he baseline model in An às and Cho (2013) in assuming ha
his inc emen al con ibu ion o he supplie o he secu ed e enues
is he su plus o he ma ch and he e o e also he su plus ba gained
o e .28
2.3.5. Ba gaining along he alue chain
Based on he ba gaining be ween he i m and i s supplie a s age
𝑚, we can now speci y he game played by he i m and he con inuum
o supplie s along he alue chain. Ini ially, he i m selec s he physical
ou pu 𝑞(𝜔)i seeks o p oduce. The i m hen selec s one supplie o
each s age 𝑗 om a la ge pool o po en ial supplie s. Supplie s ha we e
no selec ed by a i m s and eady o sell inpu 𝑗on he inpu ma ke a
any p ice la ge o equal he p oduc ion cos 𝑐𝑥. A e his, p oduc ion
akes place sequen ially s a ing a 𝑗= 0 and he i m assu es ha he
op imal sequence o p oduc ion is espec ed. Based on he in o ma ion
con ained in Eq. (8), he supplie chooses i s CSR in es men 𝑐𝑠𝑠(𝑚). A
he end o he s age, i m and supplie ba gain o e he su plus o he
ma ch. The i m compensa es he supplie o he physical p oduc ion
cos s and pays i a ac ion (1−𝛽)o i s inc emen al con ibu ion o o al
e enues, 𝑟′(𝑚), gi en by Eq. (9). A e s age 𝑗= 1 is comple ed, he
inal consump ion good is p oduced by he i m and sold o consume s
gene a ing o al e enues gi en by Eq. (7).
2.4. CSR in es men s
We now u n o he CSR in es men s o supplie s. Fi s , we conside
an indi idual supplie and hen he en i e chain o supplie s aking
agg ega e a iables as gi en. We analyze he de e minan s o CSR
27 In his con ex , he assump ion o 𝛼∈(0, 𝜌)me i s some mo e discussion,
as i ela es o he key mechanism in An às and Cho (2013). Thei cen al
esul s s em om dis inguishing he cases o 𝛼 < 𝜌 (sequen ial complemen s)
and 𝛼 > 𝜌 (sequen ial subs i u es), whe e 𝛼shapes he subs i u abili y o inpu
quan i ies ac oss p oduc ion s ages. As ou lined in Sec ion 2.2.4, i seems o
be an eminen ly plausible assump ion ha subs i u abili y o CSR le els ac oss
p oduc ion s ages is qui e low so ha we conside he case o sequen ial
complemen a i y in ou baseline se ing.
28 This implies ha supplie 𝑚’s indi ec con ibu ion o inal e enue,
which goes h ough he e ec o supplie 𝑚’s in es men on he in es men
decisions o mo e downs eam supplie s, does no en e he ba gaining. This
simpli ying assump ion assu es ha he sequen ial ba gaining o he i m wi h
a con inuum o supplie s emains ac able. An às and Cho (2013) show
ha his assump ion can be elaxed and s udy an ex ension o he model ha
deli e s he iden ical esul s on he supplie ’s in es men s as in he baseline
model.
Jou nal o De elopmen Economics 167 (2024) 103236
15
P. He kenho e al.
Table 2
Summa y s a is ics.
Sou ce: CMIE P owess.
Va iable Obs Mean S d. De . Min Max
CSR (in USD) 81,749 141,268.6 1,260,149 0 142,000,000
log CSR 54,397 10.16 1.83 7.13 18.77
Ups eamness 81,749 2.05 0.54 1.17 4.08
OECD expo exposu e 81,749 0.31 0.22 0.00 0.86
log Wages 81,749 3.21 0.43 1.92 4.79
Expo sha e 81,500 0.11 0.25 0.00 1.00
Domes ic inpu s/ o al inpu s 81,749 0.35 0.44 0.00 1.00
S a e-owned 81,749 0.00 0.04 0.00 1.00
Fo eign-owned 81,014 0.06 0.25 0.00 1.00
log Sales 79,638 1.48 2.51 −6.68 11.07
log Age 81,736 2.78 0.74 0.00 5.01
numbe o p oduc s 74,809 2.95 2.31 1.00 51.00
numbe o 2-digi indus ies pe i m 74,809 1.52 0.80 1.00 10.00
No e: Only adable indus ies a e conside ed.
Table 3
Reg ession esul s.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
(ln) CSR CSR in USD CSR in USD (ln) CSR CSR in USD CSR in USD
OLS PPML PPML OLS PPML PPML
w/o ze os incl. ze os w/o ze os incl. ze os
Ups eamness −0.211*** −0.234** −0.225** −0.135*** −0.184*** −0.170***
(0.051) (0.104) (0.108) (0.039) (0.061) (0.060)
OECD expo exposu e 0.086 0.567** 0.651*** 0.288*** 0.533*** 0.567***
(0.105) (0.235) (0.231) (0.065) (0.118) (0.120)
log(Wages) 0.238*** −0.083 −0.053 0.223*** 0.009 0.010
(0.078) (0.190) (0.192) (0.039) (0.096) (0.097)
Expo sha e 0.379*** 0.149 0.200 0.143*** 0.222*** 0.265***
(0.046) (0.187) (0.186) (0.040) (0.083) (0.083)
Domes ic inpu s/ o al inpu s 0.228*** 0.245*** 0.323*** 0.041** 0.045 0.081**
(0.027) (0.078) (0.078) (0.018) (0.033) (0.034)
S a e-owned (D) 0.520* 0.139 0.161 0.148 0.049 0.050
(0.273) (0.172) (0.181) (0.169) (0.094) (0.094)
Fo eign-owned (D) 1.037*** 0.522*** 0.562*** 0.048 0.188*** 0.193***
(0.046) (0.093) (0.095) (0.049) (0.065) (0.067)
log(Sales) 0.582*** 0.857*** 0.881*** 0.315*** 0.463*** 0.472***
(0.008) (0.028) (0.028) (0.010) (0.025) (0.025)
log(Age) 0.219*** 0.197*** 0.221*** 0.141*** 0.302*** 0.323***
(0.015) (0.051) (0.051) (0.030) (0.063) (0.063)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m (D) No No No Yes Yes Yes
Yea (D) No No No Yes Yes Yes
Obse a ions 54,394 54,394 81,749 51,645 51,645 64,851
Numbe o i ms 11,291 11,291 15,512 8,533 8,533 9,596
Pseudo/Adjus ed R20.577 0.837 0.842 0.140a0.975 0.969
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
o he sample o i ms epo ing posi i e CSR spending only. Column
(3) shows he esul s o he PPML es ima ion including i ms wi h ze o
CSR spending, which allows us o use a subs an ially la ge sample.
In he c oss sec ional es ima ions, he un es ic ed sample includes
81,749 obse a ions (column (3)), while he exclusion o i ms e-
po ing ze o CSR spending educes he sample o 54,394 obse a ions
(columns (1) and (2)).
𝑈𝑝𝑠𝑡𝑟𝑒𝑎𝑚𝑛𝑒𝑠𝑠 is nega i e and signi ican in all h ee speci ica ions.
The coe icien s a e qui e s able ac oss speci ica ions, anging be ween
−0.21 and −0.23. Fi ms ha a e loca ed mo e ups eam ha e lowe
CSR spending compa ed o i ms ha a e mo e downs eam. We ake
his inding on he ela ionship be ween CSR spending and a i m’s GVC
posi ion as suppo o ou model’s p edic ion o highe CSR spending
in mo e downs eam i ms.
In he panel es ima ions in columns (4) o (6), we epo he esul s
om he speci ica ion ha exploi s ime a ia ion wi hin i ms. Excep
o he ixed e ec s s uc u e, hey mi o columns (1) o (3). Again,
we ind he expec ed nega i e e ec o ups eamness. I is signi ican
a he 1%-le el in all h ee speci ica ions. The OLS es ima ion and
he PPML es ima ion excluding ze o CSR spending exploi in o ma ion
om 8,533 i ms and 51,645 obse a ions in o al, espec i ely. The
speci ica ion o column (6) also includes i ms which epo ze o CSR
spending in some yea s. This inc eases he numbe o obse a ions
o 64,851 and he numbe o i ms o 9,596. Again, we ake he
signi ican indings o he panel es ima ions as empi ical suppo o
ou heo e ical p edic ion o inc easing CSR spending along he GVC.
Fi ms ha become mo e ups eam o e ime spend less on CSR on
a e age.
To analyze he economic ele ance o he indings, Fig. 2 illus a es
he p edic ed le el o CSR spending based on he i s ou columns,
while a ying he GVC posi ion and holding all o he a iables cons an
Jou nal o De elopmen Economics 167 (2024) 103236
16
P. He kenho e al.
Fig. 2. A i m’s GVC posi ioning and p edic ed CSR spending.
a hei means.48 In all ou columns, he nega i e ela ionship is isible
ac oss all pe cen iles. Tu ning o he g aph based on column (3), which
includes also ze o epo ed CSR spending, we ind a he i h pe cen ile
a p edic ed CSR spending o 164,130 USD compa ed o 105,679 USD a
he 95 h pe cen ile. Acco dingly, compa ing a e y ups eam posi ion
wi h 𝑈𝑓𝑡−1 equal o 3.29 o a e y downs eam posi ion wi h 𝑈𝑓𝑡−1
equal o 1.33 inc eases he p edic ed CSR spending by 55.31%, which
is a guably an economically ele an magni ude.
Ou empi ical es ima ions hus s ongly suppo ou main p edic ion
ha mo e downs eam i ms spend mo e on CSR. The empi ical inding
is bo h s a is ically signi ican and economically meaning ul.
Besides he e ec o he alue chain posi ion on CSR spending,
which is a no el y in he li e a u e on CSR de e minan s, also he
indings conce ning ou con ol a iables p o ide some in e es ing
insigh s o his li e a u e. In line wi h he pa ial equilib ium p e-
dic ions o ou model in Eq. (14), CSR spending inc eases wi h expo
exposu e o OECD coun ies. Only in he i s speci ica ion, applying
OLS and excluding i ms wi hou CSR spending, he e is no signi ican
ela ionship. Since he ela ionship is signi ican a he 1%-le el in he
emaining speci ica ions, he empi ical esul s s ongly imply ha i ms
selling o ma ke s wi h p esumably s onge demand o e hical goods
end o ha e highe CSR le els. As o wages, we ind only en a i e
suppo ha a i m’s CSR spending inc eases wi h he wage le el. The
ela ionship is only signi ican in columns (1) and (4). On he one
hand, his sugges s ha ocusing pu ely on wages paid by i ms does
no su icien ly cap u e i s goodwill owa ds employees and socie y.
Howe e , when in e p e ing ou esul s one has o bea in mind ha
we nei he ha e wage bill in o ma ion o he i ms no a spli o wages
by skill-le els. Mo eo e , we ind ha an inc ease in expo s ela i e
o o al sales signi ican ly inc eases a i m’s CSR spending. The la e
e ec is signi ican in all panel es ima ions and he OLS es ima ion
48 The ma ginal e ec s o linea models wi h ixed e ec s (like ou columns
1–4) can be compu ed and illus a ed as in Figu e 3. This is no possible in a
non-linea model wi h ixed e ec s (ou columns 5 and 6).
o he c oss sec ion. Acco dingly, we can conclude ha an inc ease in
expo ac i i y is associa ed wi h he le el o CSR spending. This esul
speaks o he ole o expo ma ke s in incen i izing supplie s’ social
ac i i ies and is in line wi h Newman e al. (2018). Mo eo e , domes ic
linkages a e a powe ul de e minan o CSR spending and a e signi ican
a he 1%-pe cen le el in all c oss sec ional speci ica ions and a he
5% le el in wo panel speci icaions. We ake his as empi ical suppo
o he claim ha i ms wi h s ong egional ies also commi o egional
sus ainable de elopmen , including spending on social and communi y
CSR (Newman e al.,2020). Finally, we u n o he ole o sha eholde s.
While s a e owne ship is signi ican only in one speci ica ion, o eign
owne ship is signi ican in i e ou o six speci ica ions. The la e migh
e lec he commi men o o eign owne s o in e na ional s anda ds in
labo condi ions, including CSR ac i i ies.
3.3.4. Con ac en o cemen
We complemen ou main esul s om abo e wi h an analysis o he
e ec o con ac ual impe ec ions. In ou model, incomple e con ac s
lead o an unde -in es men in CSR and – in combina ion wi h he
sequen ial ba gaining p ocess along he alue chain – o he inc eas-
ing CSR p o ile om ups eam o downs eam supplie s. We show in
Appendix A.2 ha wi h comple e con ac s, he model implies a la
CSR p o ile along he alue chain, which shows ha ou main esul is a
di ec consequence o con ac ual impe ec ions. In his sec ion, we will
analyze empi ically, i di e en deg ees o con ac ibili y (he e: ac oss
egions) imply di e en slopes o he CSR p o ile wi h a slope o ze o
ep esen ing he benchma k case o ze o con ac ual impe ec ions.
The e exis s subs an ial a ia ion in con ac en o cemen ac oss
Indian s a es e.g., Boehm and Obe ield,2020. He e, we exploi hese
di e ences ac oss s a es o analyze whe he s a es di e sys ema ically
in e ms o he ela ionship be ween CSR spending and he alue
chain posi ion. We expec he nega i e ela ionship e ealed in he
p e ious sec ion o be mo e p onounced in s a es wi h weake con ac
en o cemen .
Jou nal o De elopmen Economics 167 (2024) 103236
17
P. He kenho e al.
We measu e con ac en o cemen wi h in o ma ion on judicial
(in)e iciency using sub-na ional Wo ld Bank Doing Business da a.49
Simila o Sheng and Yang (2016), who analyze Chinese da a, we
employ a cou cos a iable. This a iable measu es he o icial cos s o
going h ough cou p ocedu es ela i e o he alue o he deb claims.
High cou cos s indica e ine iciencies in he legal sys em (Sheng and
Yang,2016). Fo India, he da a encompasses 17 s a es (see Table B.1).
While he co e age is no exhaus i e, i does include a subs an ial
po ion o he i ms in ou sample. Un o una ely, he measu e is only
a ailable o he yea 2009, limi ing us o a c oss-sec ional analysis
along he lines o Table 3 columns (1)–(3) o 2009.
We employ wo dis inc app oaches o de e mine whe he a i m
ope a es in a s a e wi h low judicial quali y (high-cos sha e). Fi s , he
bina y a iable Judicial ine iciency (a) akes a alue o one i he i m is
in a s a e wi h judicial ine iciency abo e he le el o judicial e iciency
he median i m is con on ed wi h in i s own s a e. Fo ou second
measu e, we ake he median s a e ins ead. While he i s app oach
spli s he g oup o i ms in he sample in o wo g oups o equal size,
he second spli s he s a es in wo g oups o equal size, deli e ing he
bina y a iable Judicial ine iciency (b) aking he alue o one i he
i m is in a s a e wi h abo e median judicial ine iciency. No ably, wi h
79%, he majo i y o i ms ope a e in hese s a es.
Table 4 p esen s he esul s, including an in e ac ion e m be ween
ou measu es o judicial ine iciency and ups eamness.50 Columns (1),
(4), and (7) p esen he esul s o he main speci ica ions limi ed o
he yea 2009 and o i ms loca ed in s a es whe e we can measu e
judicial ine iciency. As be o e, we use OLS and PPML wi h and wi hou
ze os. The subsequen columns in oduce he ele an in e ac ion e m
be ween judicial ine iciency and ups eamness. In all hese speci ica-
ions, s a e dummies a e included making he (s a e-speci ic) le el o
judicial ine iciency obsole e.
Al hough limi ed o only one yea , he main esul o ou pape –
ha CSR spending signi ican ly dec eases wi h ups eamness – emains
obus o all speci ica ions. All in e ac ion e ms a e s a is ically sig-
ni ican and ha e he expec ed nega i e sign. This implies ha i ms in
egions wi h low con ac ual en o cemen de ia e mo e s ongly and
in he way p edic ed by he model om he benchma k o comple e
con ac s (wi h a p edic ed slope o ze o).51
3.4. Robus ness
In his sec ion, we p esen se e al obus ness checks. Fi s , we con-
ol o i ms’ isibili y o inal consume s as an al e na i e mechanism
ha could explain mo e CSR spending in downs eam GVC posi ions.
Second, we measu e CSR in ela i e e ms by calcula ing CSR as sha e
o expendi u e. Thi d, we con ol o labo in ensi y as an al e na i e
o ou wage measu e. Fou h, we smoo h CSR spending o e wo yea s
o con ol o he possibili y ha i ms’ CSR spending concen a es
a ound a speci ic yea . Fi h, we apply OLS wi h i.h.s. ans o ma ion
as an al e na i e o he PPML es ima o . Finally, o add ess possible
conce ns ela ed o endogenei y, we con ol o i m-speci ic shocks o
ups eamness.52
49 Da a we e downloaded om h ps://subna ional.doingbusiness.o g/en/
epo s/subna ional- epo s/india on Augus 1, 2023.
50 The sample is limi ed o he yea 2009 bu simila ly o he main
speci ica ions, we use lagged a iables. The da a on judicial e iciency was
published in 2009 bu collec ed p io o ha yea .
51 In line wi h ou model, hese esul s highligh he ele ance o incomple e
con ac s o CSR in es men s along he alue chain. We in e p e his as s ong
suppo o ou model. I should be no ed, howe e , ha he e emains an
al e na i e mechanism elying on incomple e con ac s: i o some eason,
con ac s a e less incomple e in mo e downs eam p oduc ion s ages, incom-
ple e con ac s would also lead o an inc easing CSR p o ile along he alue
chain. As ou da a do no allow us o dis inguish be ween he wo app oaches,
we ha e o lea e his o u u e esea ch.
52 In Appendix B, we epo addi ional obus ness checks.
3.4.1. Fi ms’ isibili y o inal consume s
Mo e downs eam supplie s and hei ac ions may be mo e isible
o inal consume s, and he e o e, hei CSR may a ec demand o
he inal p oduc mo e s ongly. In his case, highe isibili y in mo e
downs eam s ages would p o ide an addi ional incen i e o CSR
in es men s and ou indings could be d i en by he isibili y o i ms
a he han hei GVC posi ion. We see his as a a he ad hoc bu plausi-
ble mechanism complemen a y o ou model’s mechanism: he wo a e
no mu ually exclusi e and ein o ce a he han o se one ano he .
In o de o con ol o highe isibili y/consume awa eness o mo e
downs eam sec o s, in he spi i o Se aes and Tamayo (2013), we use
a i m’s selling and dis ibu ion expenses, which include ma ke ing, ad-
e ising, and dis ibu ion spending as ou p oxy o isibili y. Table 5
shows he esul s when we con ol o ou p oxy o isibili y.
The sample size o ou b oades sample (column (3)) is educed
by 21% compa ed o ou main esul s in Table 3, as no all i ms
epo on ma ke ing expenses. Howe e , we s ill co e a la ge pa o
he sample. CSR and ma ke ing expenses a e indeed highly co ela ed,
bu he e ec o ups eamness on i ms’ CSR spending does emain
s a is ically signi ican and economically meaning ul. While isibili y
seems o play a ole o a i m’s CSR decision, he e is a sepa a e and
dis inc ole o ups eamness in de e mining CSR expendi u e, which is
in line wi h ou model’s p edic ion.
3.4.2. CSR as sha e o expendi u e
In ou main speci ica ion, ou dependen a iable is measu ed as
o al CSR spending and we con ol o i m size by including log(Sales)
as an independen a iable. In his obus ness check, we main ain
log(Sales) in he speci ica ion bu use CSR spending ela i e o a se o
o he expenses as he dependen a iable di ec ly accoun ing o size
e ec s on he side o expenses a he han sales.
The P owess da abase co e s 27 domains o expendi u e like spend-
ing on aw ma e ials, s o es and spa es bu also ou sou cing o jobs,
en and lease en cos s. The wage bill is also one o hem. Clea ly,
wages ha e a CSR componen : paying ‘‘li ing wages’’, o example, a e
CSR expendi u es. Bu wages a e also d i en by many o he ac o s.
We he e o e chose no o include wages in ou CSR measu e, see Sec-
ion 3.2.1. Due o hei CSR componen , howe e , we do no include he
wage bill in he denomina o o ou ela i e CSR measu e. To accoun
o he a e age wage le el a he alue chain posi ion o he i m, we
s ill con ol o he log(Wages) a iable desc ibed in Sec ion 3.3.1.
E en hough he da a quali y o P owess is e y high, he e a e some
un ealis ic sha es, e.g., u ning nega i e o la ge han one. Ins ead o
cleaning hese en ies by hand, which would be somehow a bi a y,
we cu o he highes pe cen ile and alues below ze o.53 Table 6
summa izes he esul s measu ing CSR as a sha e o expendi u e.
The esul s con i m ou indings om he main speci ica ion: also
wi h CSR ela i e o a b oad se o o he expendi u es o he i m
(accoun ing o size e ec s on he expendi u es-side) as he dependen
a iable, ou model’s p edic ions on he link be ween CSR and alue
chain posi ion a e con i med.
3.4.3. Labo in ensi y
Downs eam indus ies end o be mo e labo in ensi e. Ou CSR
measu e con ains elemen s o wo ke wel a e, which may be highe in
mo e labo in ensi e indus ies. In his sec ion we analyze whe he ou
esul s a e obus o he inclusion o labo in ensi y in ou eg essions.
As highligh ed abo e, i is a limi a ion o he P owess da a ha
in o ma ion on he numbe o employees is no compulso y o epo ,
which p ecludes us om compu ing labo in ensi y a he i m le el o
ou sample. We he e o e apply a simila p ocedu e o he cons uc ion
o he wage a iable (see Sec ion 3.3.1) o build a measu e o labo
53 We did no cu o he lowes pe cen ile because in his case, we would
lose all obse a ions wi h ze o CSR spending.
Jou nal o De elopmen Economics 167 (2024) 103236
18
P. He kenho e al.
Table 4
Reg ession esul s: Con olling o con ac en o cemen .
Va iables (1) (2) (3) (4) (5) (6) (7) (8) (9)
C oss sec ion
(ln) CSR CSR in USD
OLS PPML
w/o ze os incl. ze os
Ups eamness −0.303*** −0.260** −0.175 −0.644*** −0.473** −0.452** −0.667*** −0.477** −0.457*
(0.102) (0.106) (0.113) (0.222) (0.219) (0.230) (0.220) (0.220) (0.233)
Judicial ine iciency (a) −0.098* −0.389*** −0.396***
x Ups eamness (0.059) (0.149) (0.148)
Judicial ine iciency (b) −0.176** −0.287** −0.298**
x Ups eamness (0.070) (0.137) (0.137)
OECD expo exposu e 0.087 0.090 0.086 0.424 0.440 0.403 0.655* 0.662* 0.636
(0.195) (0.195) (0.195) (0.406) (0.404) (0.402) (0.395) (0.395) (0.391)
log(Wages) 0.276* 0.276* 0.277* 0.267 0.299 0.293 0.241 0.275 0.255
(0.156) (0.156) (0.155) (0.431) (0.420) (0.425) (0.415) (0.408) (0.409)
Expo sha e 0.413*** 0.412*** 0.416*** 0.434** 0.444*** 0.455*** 0.424** 0.436*** 0.446***
(0.074) (0.074) (0.074) (0.173) (0.169) (0.171) (0.172) (0.167) (0.169)
Domes ic inpu s/ 0.232*** 0.231*** 0.231*** 0.532*** 0.527*** 0.528*** 0.588*** 0.584*** 0.584***
o al inpu s (0.049) (0.049) (0.049) (0.138) (0.135) (0.137) (0.135) (0.132) (0.134)
S a e-owned (D) 0.714** 0.698** 0.724** 0.087 0.067 0.074 0.124 0.095 0.101
(0.298) (0.300) (0.299) (0.180) (0.177) (0.173) (0.171) (0.190) (0.176)
Fo eign-owned (D) 1.036*** 1.036*** 1.037*** 0.355** 0.348** 0.355** 0.405*** 0.398*** 0.406***
(0.067) (0.067) (0.067) (0.145) (0.142) (0.144) (0.145) (0.142) (0.144)
log(Sales) 0.562*** 0.561*** 0.561*** 0.901*** 0.895*** 0.897*** 0.932*** 0.926*** 0.928***
(0.013) (0.013) (0.013) (0.057) (0.054) (0.056) (0.055) (0.052) (0.054)
log(Age) 0.169*** 0.170*** 0.170*** 0.098 0.097 0.094 0.118 0.117 0.113
(0.024) (0.024) (0.024) (0.089) (0.087) (0.088) (0.090) (0.088) (0.089)
Indus y & S a e (D) Yes Yes Yes Yes Yes Yes Yes Yes Yes
Obse a ions 4,393 4,393 4,393 4,393 4,393 4,393 6,831 6,831 6,831
Pseudo/Adjus ed R20.574 0.574 0.574 0.849 0.850 0.849 0.853 0.854 0.854
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
Judicial ine iciency (a) equals 1 i he cos o en o cemen sha e is highe han ha o he median i m.
Judicial ine iciency (b) equals 1 i he cos o en o cemen sha e is highe han ha o he median s a e.
*** p <0.01, ** p <0.05, * p <0.1.
Table 5
Reg ession esul s: Con olling o ma ke ing expenses.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
(ln) CSR CSR in USD CSR in USD (ln) CSR CSR in USD CSR in USD
OLS PPML PPML OLS PPML PPML
w/o ze os incl. ze os w/o ze os incl. ze os
Ups eamness −0.222*** −0.251** −0.247** −0.089** −0.163** −0.149**
(0.050) (0.105) (0.108) (0.040) (0.065) (0.065)
OECD expo exposu e 0.121 0.386* 0.439** 0.330*** 0.552*** 0.600***
(0.103) (0.207) (0.205) (0.068) (0.127) (0.128)
log(Wages) 0.266*** −0.147 −0.113 0.173*** 0.013 0.003
(0.082) (0.195) (0.196) (0.040) (0.100) (0.100)
Expo sha e 0.189*** 0.261*** 0.269*** 0.101** 0.165* 0.163*
(0.048) (0.092) (0.093) (0.042) (0.085) (0.085)
Domes ic inpu s/ o al inpu s 0.187*** 0.177** 0.214*** 0.051*** 0.045 0.073**
(0.026) (0.073) (0.073) (0.019) (0.032) (0.033)
S a e-owned (D) 0.445* 0.125 0.147 0.166 0.053 0.053
(0.236) (0.152) (0.157) (0.173) (0.093) (0.093)
Fo eign-owned (D) 0.881*** 0.457*** 0.487*** 0.037 0.185** 0.189**
(0.044) (0.083) (0.085) (0.047) (0.073) (0.075)
log(Sales) 0.431*** 0.743*** 0.757*** 0.310*** 0.464*** 0.472***
(0.011) (0.041) (0.041) (0.012) (0.034) (0.034)
log(Age) 0.213*** 0.210*** 0.222*** 0.133*** 0.330*** 0.342***
(0.015) (0.046) (0.047) (0.031) (0.067) (0.066)
log(Ad Ma ke Exp) 0.226*** 0.110*** 0.117*** 0.083*** 0.029 0.029
(0.008) (0.023) (0.023) (0.006) (0.018) (0.018)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m (D) No No No Yes Yes Yes
Yea (D) No No No Yes Yes Yes
Obse a ions 49,400 49,400 64,379 47,064 47,064 55,498
Numbe o i ms 10,180 10,180 12,720 7,832 7,832 8,460
Pseudo/Adjus ed R20.616 0.844 0.845 0.162a0.975 0.971
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
Jou nal o De elopmen Economics 167 (2024) 103236
19
P. He kenho e al.
Table 6
Reg ession esul s: CSR as sha e o o al expendi u e.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
CSR/expenses CSR/expenses CSR/expenses CSR/expenses CSR/expenses CSR/expenses
OLS PPML PPML OLS PPML PPML
w/o ze os incl. ze os w/o ze os incl. ze os
Ups eamness −0.001*** −0.237*** −0.292*** −0.001** −0.110** −0.110*
(0.000) (0.058) (0.062) (0.000) (0.051) (0.058)
OECD expo exposu e 0.000 0.104 0.575*** 0.002*** 0.276*** 0.337***
(0.001) (0.118) (0.135) (0.001) (0.080) (0.098)
log(Wages) 0.000 0.071 0.053 0.001* 0.114** 0.072
(0.001) (0.088) (0.097) (0.000) (0.050) (0.059)
Expo sha e 0.003*** 0.356*** 0.501*** −0.000 −0.034 0.048
(0.000) (0.043) (0.048) (0.000) (0.050) (0.062)
Domes ic inpu s/ o al inpu s 0.000** 0.096*** 0.263*** 0.000 0.036 0.072**
(0.000) (0.033) (0.041) (0.000) (0.024) (0.028)
S a e-owned (D) −0.001 0.046 −0.018 0.001 0.163 0.179
(0.002) (0.193) (0.251) (0.001) (0.171) (0.207)
Fo eign-owned (D) 0.005*** 0.573*** 0.520*** 0.001 0.063 0.095*
(0.000) (0.044) (0.044) (0.000) (0.049) (0.051)
log(Sales) −0.002*** −0.201*** −0.007 −0.001*** −0.128*** −0.004
(0.000) (0.007) (0.007) (0.000) (0.011) (0.013)
log(Age) 0.002*** 0.258*** 0.286*** 0.001*** 0.146*** 0.048
(0.000) (0.019) (0.023) (0.000) (0.036) (0.042)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m dummy No No No Yes Yes Yes
Yea dummy No No No Yes Yes Yes
Obse a ions 53,502 53,502 80,857 50,783 50,783 63,763
Numbe o i ms 11,168 11,168 15,444 8,440 8,440 9,485
Pseudo/Adjus ed R20.298 0.0639 0.0583 0.0123a0.133 0.143
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
in ensi y using da a om he WIOD Socio Economic Accoun s. We p o-
ceed in wo s eps. Fi s , we calcula e indus y labo in ensi y as he a io
o employees o e capi al s ock and con i m ha mo e downs eam
i ms a e indeed mo e labo in ensi e in ou sample. Second, o make
he indus y-speci ic in o ma ion i m-speci ic, we weigh he indus y
labo in ensi y by i ms’ p oduc sales in he espec i e indus ies.
I u ns ou ha ou labo in ensi y a iable and ou wage a iable
a e s ongly co ela ed wi h a coe icien o −0.79. To a oid po en ial
p oblems o collinea i y, we e ain om including bo h a iables in
ou eg essions simul aneously.
Table 7 epo s he esul s when we include he labo in ensi y
a iable in ou eg essions ins ead o he wage a iable.
Labo in ensi y is posi i ely ela ed o CSR in es men s bu only
signi ican in he wo c oss sec ional PPML speci ica ions. I is p ecisely
in he c oss sec ion PPML case, whe e ups eamness is no signi ican ,
while i emains signi ican a he 5%-le el in all o he speci ica ions.
O e all, he esul s show ha ou main indings, especially on he panel
es ima ions, a e obus o he inclusion o labo in ensi y as a con ol.54
3.4.4. Smoo hed CSR spending
Like o he o ms o i m spending (e.g., in es men s), CSR migh
a y om yea o yea . Fo ins ance, a i m migh make subs an ial
in es men s in he communi y in as uc u e in one yea and e ain
om u he CSR spending in he ollowing yea s. To accoun o his,
we smoo h CSR spending by aking he mo ing a e age o wo yea s,
mo e p ecisely, yea s 𝑡and 𝑡−1, as he dependen a iable. We p esen
he esul s o his obus ness check in Table 8.
54 This in e p e a ion inds u he suppo in an addi ional un epo ed anal-
ysis (a ailable upon eques ), whe e we ind ha d opping he obse a ions
wi h he la ges 1% o CSR spending, ups eamness is again highly signi ican
ac oss all speci ica ions including he ones using PPML wi h c oss sec ional
da a.
The esul s con inue o suppo ou heo e ical p edic ion. The e ec
o ups eamness on CSR emains nega i e and s a is ically signi ican
when using smoo hed CSR spending. I is signi ican a he 5%-le el in
column (3) and a he 1%-le el in he emaining i e speci ica ions.
3.4.5. In e se hype bolic sine ans o ma ion
In ou main speci ica ion, we use PPML alongside OLS, as he
o me allows us o keep he ze os in he da a. An al e na i e app oach
is o use he in e se hype bolic sine (i.h.s.) ans o ma ion, allowing
us o es ima e OLS in na u al logs including he ze o obse a ions.
I is a ea u e o his me hod ha i is sensi i e o he uni choice
(c . Aihoun on and Henningsen,2020,De B auw and He skowi z,2021,
and Chen and Ro h,2023). We he e o e conside CRS spending bo h
in USD and in million USD. Fo he o me case, a highe weigh
is mechanically placed on he ex ensi e ma gin (swi ches om ze o
CSR o posi i e CSR), while in he la e case, he in ensi e ma gin
(c oss sec ional a ia ion ac oss i ms wi h posi i e CSR and a ia ion
in CSR spending wi hin he i m o e ime), has a s onge e ec . We
a e inclined owa ds measu ing CSR spending in million USD, as ou
es able implica ion in P oposi ion 1 conce ns he in ensi e a he han
he ex ensi e ma gin.55
55 Taking ou model a ace alue, a supplie has ze o CSR in equilib ium
only when i is ma ched o a i m se ing non-ca ing consume s. The only
eason o a supplie o ha e posi i e CSR is o be ma ched o a i m se ing
ca ing consume s. In e ms o ou model, he ex ensi e ma gin is he e o e
no e y in o ma i e: a swi ch om ze o o posi i e CSR would no be d i en
by changes in he ups eamness, bu by being ma ched o a di e en pa ne
(no ing, o cou se, ha ou model is s a ic so o mally, no swi ches occu ).
In equilib ium, all supplie s ma ched o a i m se ing ca ing consume s will
ha e posi i e CSR le els, bu , c ucially, hese CSR le els will inc ease along
he alue chain. This is he p edic ion we seek o es . The e o e, i is he
a ia ion in CSR le els ac oss i ms wi h posi i e CSR o he a ia ion o e
ime o CSR spending wi hin a i m ha speaks o ou heo y.
Jou nal o De elopmen Economics 167 (2024) 103236
20
P. He kenho e al.
Table 7
Reg ession esul s: Con olling o labo in ensi y.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
(ln) CSR CSR in USD CSR in USD (ln) CSR CSR in USD CSR in USD
OLS PPML PPML OLS PPML PPML
w/o ze os incl. ze os w/o ze os incl. ze os
Ups eamness −0.120** −0.072 −0.050 −0.137*** −0.200*** −0.185**
(0.054) (0.120) (0.121) (0.040) (0.073) (0.073)
OECD expo exposu e 0.255** 0.962*** 1.065*** 0.312*** 0.508*** 0.543***
(0.113) (0.241) (0.233) (0.070) (0.141) (0.143)
log(Labo in ensi y) 0.041 0.227** 0.233** −0.046* −0.035 −0.033
(0.039) (0.099) (0.098) (0.026) (0.069) (0.069)
Expo sha e 0.383*** 0.153 0.204 0.146*** 0.219*** 0.262***
(0.046) (0.187) (0.186) (0.040) (0.083) (0.082)
Domes ic inpu s/ o al inpu s 0.227*** 0.242*** 0.319*** 0.041** 0.046 0.082**
(0.027) (0.077) (0.078) (0.018) (0.033) (0.034)
S a e-owned (D) 0.525* 0.133 0.156 0.160 0.048 0.049
(0.272) (0.174) (0.184) (0.172) (0.092) (0.093)
Fo eign-owned (D) 1.041*** 0.522*** 0.562*** 0.045 0.184*** 0.189***
(0.046) (0.092) (0.094) (0.049) (0.064) (0.065)
log(Sales) 0.582*** 0.856*** 0.880*** 0.316*** 0.463*** 0.472***
(0.008) (0.028) (0.028) (0.010) (0.025) (0.025)
log(Age) 0.217*** 0.193*** 0.218*** 0.139*** 0.302*** 0.323***
(0.015) (0.051) (0.051) (0.030) (0.063) (0.063)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m (D) No No No Yes Yes Yes
Yea (D) No No No Yes Yes Yes
Obse a ions 54,394 54,394 81,749 51,645 51,645 64,851
Numbe o i ms 11,291 11,291 15,512 8,533 8,533 9,596
Pseudo/Adjus ed R20.577 0.838 0.843 0.138a0.975 0.969
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
Table 8
Robus ness check: Smoo hed CSR spending.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
(ln) CSR CSR in USD CSR in USD (ln) CSR CSR in USD CSR in USD
OLS PPML PPML OLS PPML PPML
w/o ze os incl. ze os w/o ze os incl. ze os
Ups eamness −0.211*** −0.277*** −0.274** −0.148*** −0.247*** −0.229***
(0.048) (0.103) (0.107) (0.035) (0.053) (0.053)
OECD expo exposu e 0.189* 0.617*** 0.698*** 0.286*** 0.529*** 0.573***
(0.097) (0.229) (0.225) (0.057) (0.125) (0.123)
log(Wages) 0.248*** −0.029 0.004 0.251*** 0.054 0.055
(0.071) (0.181) (0.183) (0.036) (0.089) (0.089)
Expo sha e 0.391*** 0.117 0.155 0.162*** 0.191** 0.208***
(0.045) (0.203) (0.202) (0.036) (0.077) (0.076)
Domes ic inpu s/ o al inpu s 0.236*** 0.262*** 0.322*** 0.045*** 0.076** 0.093***
(0.026) (0.078) (0.079) (0.016) (0.030) (0.030)
S a e-owned (D) 0.553* 0.168 0.185 0.127 −0.099** −0.096**
(0.284) (0.160) (0.165) (0.148) (0.042) (0.040)
Fo eign-owned (D) 1.097*** 0.561*** 0.592*** 0.056 0.101 0.106
(0.046) (0.090) (0.092) (0.054) (0.072) (0.072)
log(Sales) 0.543*** 0.852*** 0.870*** 0.275*** 0.501*** 0.509***
(0.007) (0.025) (0.026) (0.008) (0.022) (0.022)
log(Age) 0.265*** 0.257*** 0.275*** 0.223*** 0.435*** 0.459***
(0.014) (0.052) (0.053) (0.028) (0.062) (0.063)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m (D) No No No Yes Yes Yes
Yea (D) No No No Yes Yes Yes
Obse a ions 59,090 59,090 81,749 56,706 56,706 65,839
Numbe o i ms 11,660 11,660 15,512 9,269 9,269 9,823
Pseudo/Adjus ed R20.592 0.847 0.851 0.192a0.983 0.980
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
Jou nal o De elopmen Economics 167 (2024) 103236
21
P. He kenho e al.
Table 9
Reg ession esul s: Applying in e se hype bolic sine ans o ma ion.
Va iables (1) (2) (3) (4) (5) (6)
C oss sec ion Panel
CSR in USD asinh CSR asinh mCSR CSR in USD asinh CSR asinh mCSR
PPML OLS OLS PPML OLS OLS
incl. ze os
Ups eamness −0.225** −0.441*** −0.042*** −0.170*** 0.035 −0.035***
(0.108) (0.126) (0.008) (0.060) (0.130) (0.005)
OECD expo exposu e 0.651*** 1.301*** −0.058*** 0.567*** 1.386*** −0.006
(0.231) (0.255) (0.013) (0.120) (0.235) (0.009)
log(Wages) −0.053 0.449** 0.032*** 0.010 0.037 0.054***
(0.192) (0.176) (0.010) (0.097) (0.142) (0.010)
Expo sha e 0.200 1.149*** 0.033*** 0.265*** 0.296* 0.019***
(0.186) (0.125) (0.010) (0.083) (0.151) (0.006)
Domes ic inpu s/ o al inpu s 0.323*** 1.159*** 0.006 0.081** 0.418*** 0.009***
(0.078) (0.077) (0.005) (0.034) (0.073) (0.002)
S a e-owned (D) 0.161 0.818* 0.289* 0.050 −0.001 0.012
(0.181) (0.492) (0.170) (0.094) (0.443) (0.030)
Fo eign-owned (D) 0.562*** 1.591*** 0.305*** 0.193*** 0.268* 0.026
(0.095) (0.103) (0.023) (0.067) (0.157) (0.019)
log(Sales) 0.881*** 1.081*** 0.039*** 0.472*** 0.805*** 0.015***
(0.028) (0.012) (0.002) (0.025) (0.024) (0.001)
log(Age) 0.221*** 0.534*** 0.038*** 0.323*** −0.031 0.002
(0.051) (0.036) (0.003) (0.063) (0.104) (0.005)
Indus y ×Yea & S a e (D) Yes Yes Yes No No No
Fi m (D) No No No Yes Yes Yes
Yea (D) No No No Yes Yes Yes
Obse a ions 81,749 81,749 81,749 64,851 64,851 64,851
Numbe o i ms 15,512 15,512 15,512 9,596 9,596 9,596
Pseudo/Adjus ed R20.842 0.428 0.294 0.969 0.0669a0.0272a
Robus s anda d e o s in pa en heses clus e ed a he i m le el.
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
We he e o e un wo speci ica ions. The i s speci ica ion measu es
CSR spending in USD, 𝑎𝑠𝑖𝑛ℎ(𝐶𝑆𝑅), and he second measu es CSR
spending in million USD, 𝑎𝑠𝑖𝑛ℎ(𝐶𝑆𝑅𝑚𝑈𝑆𝐷). Bo h i.h.s. ans o med
dependen a iables a e de i ed as ollows56:
𝑎𝑠𝑖𝑛ℎ(𝑦) = 𝑙𝑛 (𝑦+(𝑦2+ 1)1∕2).(29)
Table 9 p esen s he empi ical esul s applying OLS i.h.s. ans-
o ma ion. We un he eg ession on he same sample as he PPML
es ima ion including ze os, which is he wides co e age o i ms in ou
empi ical se -up and makes i compa able o p io esul s. The PPML
esul s, which a e no sensi i e o he uni choice o he dependen
a iable, a e included in columns (1) and (4). Columns (2) and (3)
include he esul s o he c oss sec ion es ima ion applying OLS i.h.s.
ans o ma ion o CSR in USD and million USD, espec i ely. Columns
(5) and (6) include he esul s o he co esponding panel es ima ions.
The esul s show ha o e all ou main indings a e obus o apply-
ing he i.h.s. ans o ma ion. The c oss sec ion esul s o ups eamness
in columns (2) and (3) a e signi ican a he 1%-le el. The le els o
he es ima es di e conside ably also highligh ing he sensi i i y o
he app oach o he selec ed uni o CSR. In he panel es ima ion, he
ela ionship loses signi icance o he ans o ma ion o CSR spending
in USD bu emains signi ican a he 1%-le el o he ans o ma ion
o CSR in million USD. The di e ences in he panel es ima ions a e in
line wi h he p e iously desc ibed in ui ion ha wi hin i m a ia ion
in ups eamness is be e sui ed o explain he in ensi e ma gin, a he
han he ex ensi e ma gin.
3.4.6. Shocks o ups eamness
So a , we show obus e idence o he nega i e co ela ion be-
ween CSR in es men and ups eamness p edic ed by ou model. How-
e e , one migh be wo ied abou a i m-le el shock ha a ec s bo h
56 The espec i e dis ibu ions a e displayed in Fig. B.2 in Appendix B.
CSR spending and ups eamness. Recall ha ups eamness is de e -
mined by wo componen s, he i m’s p oduc po olio on he one
hand and he ups eamness o each p oduc ’s indus y on he o he .
The la e is a guably exogenous o he i m. As he p oduc po olio
is endogenous, a i m-speci ic shock could a ec i and CSR spending
simul aneously.
We he e o e apply a wo-s age leas squa es (2SLS) ins umen al
a iable (IV) app oach.57 To do so, we de i e an exogenous measu e
o he change o a i m’s alue chain posi ion om 𝑡− 2 o 𝑡− 1
in he spi i o shi -sha e (o Ba ik) ins umen s (e.g., Au o e al.,
2013 and Bo usyak e al.,2022). Mo e speci ically, o e e y i m 𝑖,
we conside i ms ha ing he same main i e-digi indus y 𝑣as i m 𝑖
be ween 𝑡−2 and 𝑡−1. In o de o exclude common s a e-speci ic shocks,
we only conside i ms loca ed in o he s a es han i ms 𝑖. Based on
his, we can calcula e he shi in he alue chain posi ion o i m 𝑖,
which is a guably exogenous o i .
Al hough ime a ia ion o ups eamness is ela i ely low, he ac-
o s o change a y be ween 0.547 and 1.657. Beyond hese high
alues, he 10%-pe cen ile wi h 0.960 and he 90%-pe cen ile wi h
1.038 u he indica e a ia ion which we can empi ically exploi .
To de i e he es ima ed ins umen al a iable o a i m’s alue chain
posi ion in 𝑡− 1, we mul iply a i m’s alue chain posi ion, 𝑈𝑓𝑡−2, by
he shi :
𝑍𝑓𝑡−1,𝑡−2 =𝑈𝑓𝑡−2 (𝑈𝑣𝑠𝑡−1 −𝑈𝑣𝑠𝑡−2
𝑈𝑣𝑠𝑡−2
+ 1),(30)
wi h 𝑍𝑓𝑡−1,𝑡−2 being he (lagged) ins umen o i m 𝑓a ime 𝑡using
he base-yea 𝑡−2. Mo eo e , 𝑈𝑠𝑣𝑡−1 is he a e age alue chain posi ion
o i ms ac i e in indus y 𝑣a ime 𝑡− 1 (excluding i ms in s a e 𝑠).
Eq. (31) and Eq. (32) summa ize he i s s ages o he ins u-
men al a iable app oach o he c oss sec ion and panel es ima ion,
57 The app oach is mos compa able o he OLS-speci ica ions in he main
speci ica ion (columns (1) and (4)).
Jou nal o De elopmen Economics 167 (2024) 103236
22
P. He kenho e al.
Table 10
2-SLS es ima ion esul s.
Va iables (1) (2) (3) (4) (5) (6) (7) (8)
C oss sec ion Panel C oss sec ion Panel
1s 2nd 1s 2nd 1s 2nd 1s 2nd
𝑍𝑓𝑡−1 0.772*** 0.494*** 0.772*** 0.494***
(0.011) (0.016) (0.018) (0.028)
𝑈𝑝𝑠𝑡𝑟𝑒𝑎𝑚𝑛𝑒𝑠𝑠 −0.220*** −0.183** −0.220** −0.183*
(0.065) (0.076) (0.096) (0.099)
OECD expo exposu e −0.073*** 0.091 −0.023 0.307*** −0.073** 0.091 −0.023 0.307***
(0.014) (0.117) (0.026) (0.075) (0.030) (0.172) (0.049) (0.089)
log(Wages) 0.174*** 0.269*** 0.132*** 0.211*** 0.174*** 0.269** 0.132*** 0.211***
(0.016) (0.087) (0.018) (0.045) (0.034) (0.137) (0.036) (0.064)
Expo sha e -0.004 0.386*** −0.003 0.162*** −0.004 0.386*** −0.003 0.162***
(0.003) (0.050) (0.009) (0.045) (0.004) (0.093) (0.007) (0.039)
Domes ic inpu s/ o al inpu s 0.004** 0.209*** 0.009** 0.019 0.004* 0.209*** 0.009** 0.019
(0.002) (0.029) (0.004) (0.020) (0.002) (0.029) (0.004) (0.017)
S a e-owned (D) 0.010 0.432 −0.009 0.184 0.010 0.432 −0.009 0.184
(0.015) (0.269) (0.019) (0.165) (0.015) (0.282) (0.017) (0.160)
Fo eign-owned (D) -0.003 1.011*** 0.008 0.036 -0.003 1.011*** 0.008 0.036
(0.002) (0.048) (0.011) (0.055) (0.002) (0.066) (0.010) (0.057)
log(Sales) 0.000 0.597*** −0.006*** 0.302*** 0.000 0.597*** −0.006*** 0.302***
(0.000) (0.008) (0.002) (0.011) (0.001) (0.030) (0.002) (0.018)
log(Age) −0.001 0.264*** 0.001 0.189*** −0.001 0.264*** 0.001 0.189***
(0.001) (0.017) (0.006) (0.040) (0.001) (0.022) (0.008) (0.047)
Cons an −0.067 7.808*** 0.656*** 8.499*** −0.067 7.808*** 0.656*** 8.499***
(0.041) (0.259) (0.065) (0.226) (0.084) (0.568) (0.101) (0.259)
Indus y ×Yea & S a e (D) Yes Yes No No Yes Yes No No
Fi m (D) No No Yes Yes No No Yes Yes
Yea (D) No No Yes Yes No No Yes Yes
Obse a ions 44,009 44,009 43,367 43,367 44,009 44,009 43,367 43,367
Numbe o clus e s 9,688 9,688 9,036 9,036 462 462 461 461
Adjus ed/pseudo R20.956 0.590 0.347a0.197a0.956 0.590 0.347a0.197a
F-s a is ic 1,030 22.48 979 22.48
Robus s anda d e o s in pa en heses clus e ed a he i m le el in columns (1) o (4) and i e-digi indus y-le el in columns (5) o (8).
*** p <0.01, ** p <0.05, * p <0.1.
aAdjus ed wi hin R2.
espec i ely.
𝑈𝑓𝑡−1 =𝛼0+𝛽1𝑍𝑓𝑡−1,𝑡−2 +𝛽2𝑋𝑓𝑡−1 +𝛾𝑣𝑡 +𝛿𝑠+𝜖𝑓𝑣𝑡,(31)
𝑈𝑓𝑡−1 =𝛼0+𝛽1𝑍𝑓𝑡−1,𝑡−2 +𝛽2𝑋𝑓𝑡−1 +𝜁𝑓+𝜏𝑡+𝜖𝑓𝑣𝑡.(32)
In he second s age, we include he p edic ed alues om he i s s age
(Eq. (33) and Eq. (34)):
𝑌𝑓𝑣𝑡 =𝛼0+𝛽1
𝑈𝑓𝑡−1 +𝛽2𝑋𝑓𝑡−1 +𝛾𝑣𝑡 +𝛿𝑠+𝜖𝑓𝑣𝑡,(33)
𝑌𝑓𝑣𝑡 =𝛼0+𝛽1
𝑈𝑓𝑡−1 +𝛽2𝑋𝑓𝑡−1 +𝜁𝑓+𝜏𝑡+𝜖𝑓𝑣𝑡,(34)
wi h 𝑌𝑓𝑣𝑡 being he na u al loga i hm o a i m’s CSR spending and
𝑈𝑓𝑡−1 being he p edic ed alue o he alue chain posi ion in 𝑡− 1.
Simila o he main speci ica ion, he e o e ms a e clus e ed a he
i m le el. Howe e , e o e ms migh be co ela ed be ween i ms
being ac i e in a simila main indus y. Also ollowing Bo usyak e al.
(2022) in he con ex o shi -sha e ins umen s, we conside simila
exposu e o changes in he alue chain posi ion in an al e na i e
speci ica ion by clus e ing e o e ms a he 5-digi indus y-le el.
Table 10 includes he esul s, including he i s s age.
The esul s o he i s s ages gi e insigh s on he ele ance (columns
(1), (3), (5), and (7)). In all speci ica ions, he ins umen al a iable is
signi ican a he 1%-le el and he epo ed F-s a is ics do no indica e
ha he speci ica ions su e om weak ins umen s. We he e o e
conclude ha ou ins umen s a e ele an .
The second s age esul s show ha ou main esul , he nega i e
ela ionship be ween ups eamness and CSR in es men , is signi ican
in all speci ica ions. Clus e ing he e o e m a he i e-digi indus y-
le el inc eases he magni ude o he e o e ms and wi h i dec eases
he le el o signi icance. This is in line wi h expec a ion since he
exogenous a ia ion o a i m’s alue chain posi ion is de i ed om
shi s in he alue chain posi ion a his le el. Ou esul s emain obus
also in hese mo e demanding speci ica ions: downs eam i ms ha e
highe CSR expendi u e.
Howe e , one migh conside he base-yea weigh 𝑈𝑓𝑡−2 as oo
close o 𝑡− 1. Fo he panel, his is no a conce n, as iden i ica ion in
he panel s ems om changes in he alue chain posi ion wi hin he
i m om one yea o he o he . Fo he c oss sec ion, we expe imen
wi h al e na i e base-yea s. Table 11 includes he esul s using he
base-yea s 𝑡− 3 o 𝑡− 7.58
The e ec o ups eamness on CSR spending is ema kably obus .
I only loses s a is ical signi icance a a lag o se en yea s, when mo e
han hal o he obse a ions a e los compa ed o he 𝑡−2 case epo ed
in Table 10.
4. Conclusion
Globalized alue chains in ol ing de eloping and eme ging
economies cha ac e ize mode n-day p oduc ion. Wi h di e en s ages
o p oduc ion being pe o med in di e en coun ies, hey a e also
pe o med in e y di e se egula o y en i onmen s. Wi h consume s
in he Global No h being mo e and mo e conce ned abou he en i-
onmen al and social oo p in s o hei consump ion choices, i ms a e
inc easingly unde p essu e o assu e ha ce ain minimum s anda ds
a e no only espec ed by he i m i sel , bu also by supplie s along i s
GVC.
58 The esul s o he i s s age a e epo ed in Appendix B. All IVs a e
signi ican a he 1%-le el. As o be expec ed, he coe icien size and F-
s a is ics dec ease wi h la ge lags. The smalles F- alue is o he base-yea
𝑡− 7. I akes he alue o 65.45, s ill indica ing a alid ins umen (c .
Table B.3).
Jou nal o De elopmen Economics 167 (2024) 103236
23
P. He kenho e al.
Table 11
2-SLS es ima ion esul s: C oss sec ion wi h lagged ups eamness.
Va iables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Base-yea s
-3 -3 -4 -4 -5 -5 -6 -6 -7 -7
𝑈𝑝𝑠𝑡𝑟𝑒𝑎𝑚𝑛𝑒𝑠𝑠 −0.236*** −0.236** −0.290*** −0.290** −0.335*** −0.335*** −0.277** −0.277* −0.222 −0.222
(0.079) (0.117) (0.092) (0.123) (0.106) (0.120) (0.121) (0.149) (0.149) (0.167)
OECD expo 0.121 0.121 0.118 0.118 0.129 0.129 0.130 0.130 0.187 0.187
exposu e (0.129) (0.186) (0.141) (0.181) (0.153) (0.187) (0.165) (0.198) (0.183) (0.219)
log(Wages) 0.281*** 0.281* 0.321*** 0.321** 0.376*** 0.376*** 0.375*** 0.375** 0.382*** 0.382**
(0.095) (0.148) (0.105) (0.142) (0.112) (0.136) (0.120) (0.155) (0.135) (0.168)
Expo sha e 0.410*** 0.410*** 0.413*** 0.413*** 0.422*** 0.422*** 0.451*** 0.451*** 0.494*** 0.494***
(0.054) (0.099) (0.057) (0.100) (0.060) (0.102) (0.063) (0.105) (0.067) (0.105)
Domes ic inpu s 0.197*** 0.197*** 0.202*** 0.202*** 0.199*** 0.199*** 0.199*** 0.199*** 0.208*** 0.208***
/ o al inpu s (0.031) (0.030) (0.033) (0.032) (0.034) (0.033) (0.036) (0.034) (0.038) (0.036)
S a e-owned (D) 0.383 0.383 0.384 0.384 0.399 0.399 0.396 0.396 0.327 0.327
(0.268) (0.280) (0.275) (0.287) (0.279) (0.294) (0.273) (0.287) (0.269) (0.281)
Fo eign-owned (D) 0.993*** 0.993*** 0.974*** 0.974*** 0.961*** 0.961*** 0.943*** 0.943*** 0.983*** 0.983***
(0.049) (0.066) (0.051) (0.066) (0.053) (0.067) (0.054) (0.068) (0.057) (0.072)
log(Sales) 0.610*** 0.610*** 0.620*** 0.620*** 0.630*** 0.630*** 0.643*** 0.643*** 0.641*** 0.641***
(0.009) (0.030) (0.010) (0.030) (0.010) (0.030) (0.010) (0.030) (0.011) (0.030)
log(Age) 0.310*** 0.310*** 0.353*** 0.353*** 0.397*** 0.397*** 0.444*** 0.444*** 0.457*** 0.457***
(0.020) (0.024) (0.023) (0.027) (0.026) (0.031) (0.029) (0.034) (0.033) (0.037)
Cons an 7.659*** 7.659*** 7.665*** 7.665*** 7.345*** 7.345*** 7.069*** 7.069*** 6.908*** 6.908***
(0.279) (0.579) (0.296) (0.601) (0.312) (0.576) (0.329) (0.564) (0.362) (0.616)
Indus y ×Yea
and S a e (D) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Fi m (D) No No No Yes No No No No No No
Numbe o clus e s 8,347 450 7,389 441 6,810 437 6,283 427 5,328 411
Obse a ions 37,201 37,201 32,068 32,068 27,669 27,669 23,586 23,586 19,146 19,146
Adjus ed R20.603 0.603 0.611 0.611 0.620 0.620 0.632 0.632 0.635 0.635
Robus s anda d e o s in pa en heses clus e ed a he i m le el in columns (1), (3), (5), (7), and (9) and i e-digi indus y le el in columns (2), (4), (6), (8), and (10). The
espec i e i s s age esul s a e shown in Table B.3.
*** p <0.01, ** p <0.05, * p <0.1.
In ou model, CSR along he alue chain p o ides e hical p oduc
di e en ia ion and allows he i m o se a p ice abo e ma ginal cos
o ca ing consume s. We ex end he common no ion o con ac ual
incomple eness in supply chain ela ionships o he p o ision o CSR
along he alue chain. Building on he model o sequen ial p oduc-
ion by An às and Cho (2013), we analyze CSR in es men s by a
con inuum o independen supplie s along he alue chain. We ind
an inc easing CSR p o ile wi h mo e downs eam supplie s ea u ing
highe CSR expendi u es.
We con i m his p edic ion using Indian i m-le el da a. Me ging
p oduc -le el in o ma ion wi h he Wo ld Inpu –Ou pu Da abase, we
cons uc a measu e o a i m’s GVC posi ion. Using combined s a
wel a e spending and social communi y spending as ou measu e o
CSR expendi u e, we ind s ong suppo o ou p edic ion. Mo eo e ,
we p o ide e idence ha he p edic ed pa e n is mo e p onounced in
egions o India wi h s onge con ac ual incomple eness, measu ed by
he ine iciency o he legal sys em. Ou indings a e con i med in a
subs an ial numbe o obus ness checks.
Ou main esul s a e highly ele an o policy make s. Mos im-
po an ly, ou analysis e eals ha he economic incen i es o CSR
in es men s a e lowes in ups eam indus ies. Conce ning laws on
co po a e sus ainabili y due diligence, his implies ha he obliga ion o
i ms o assu e e hical s anda ds along hei supply chain is pa icula ly
di icul o mee o ups eam supplie s. By he same oken, when he
aim o he policy make is o a ain high le els o CSR along he
en i e alue chain, ou esul s show ha e o s should be ocused
on supplie s in mo e ups eam indus ies. Finally, o go e nmen s
aiming a decen wo k and economic g ow h (as speci ied in he UN’s
Sus ainable De elopmen Goal 8) he ac ha mo e ups eam s ages
o p oduc ion a e cha ac e ized by low CSR in es men s, p o ides an
incen i e o s a egically ocus indus ial policy owa ds specializa ion
in mo e downs eam p oduc ion s ages.
In his pape , we highligh he ele ance o incomple e con ac s o
CSR in es men s depending on he alue chain posi ion o he supplie .
In oducing con ac ual incomple eness o CSR and e hical p oduc ion
in global alue chains opens up a whole ange o p omising a enues o
u u e esea ch. Especially so, as i opens up he ich heo e ical and
empi ical oolki o he li e a u e on he in e na ional o ganiza ion o
p oduc ion o he analysis o CSR and e hical p oduc ion in global alue
chains in ol ing de eloping and eme ging economies. In his con ex ,
u u e esea ch should u he in es iga e he ele ance o incomple e
con ac s o CSR in global p oduc ion. To his end, new heo ies could
be de eloped o nea ly link exis ing o no el empi ical measu es o
con ac ual incomple eness o i m’s CSR in es men s.
CRediT au ho ship con ibu ion s a emen
Philipp He kenho : Concep ualiza ion, Fo mal analysis, In es i-
ga ion, Me hodology, Valida ion, W i ing – o iginal d a . Sebas ian
K au heim: Concep ualiza ion, Fo mal analysis, Funding acquisi ion,
In es iga ion, Me hodology, P ojec adminis a ion, Supe ision, Val-
ida ion, W i ing – o iginal d a , W i ing – e iew & edi ing. Finn
Ole Sem au: Concep ualiza ion, Da a cu a ion, Fo mal analysis, In es-
iga ion, Me hodology, So wa e, Valida ion, Visualiza ion, W i ing –
o iginal d a , W i ing – e iew & edi ing. F auke S eglich: Concep u-
aliza ion, Da a cu a ion, Fo mal analysis, In es iga ion, Me hodology,
So wa e, Valida ion, Visualiza ion, W i ing – o iginal d a , W i ing –
e iew & edi ing.
Decla a ion o compe ing in e es
The au ho s decla e ha hey ha e no known compe ing inan-
cial in e es s o pe sonal ela ionships ha could ha e appea ed o
in luence he wo k epo ed in his pape .
Da a a ailabili y
The au ho s do no ha e pe mission o sha e da a. Replica ion
iles a e a ailable a h ps://gi hub.com/F aukeS eglich/CSR-along-
he-GVC.
Jou nal o De elopmen Economics 167 (2024) 103236
24
P. He kenho e al.
Appendix A. Theo y appendix
A.1. Indus y equilib ium wi h incomple e con ac s
We use Eq. (19) o calcula e he alues o 𝑞(𝜔)and 𝑠(𝑚)in indus y
equilib ium om Eqs. (14) and (17).
𝑞(𝜔) = 𝜌
𝑐𝑥
𝑓
1 − 2𝜌(A.1)
𝑠(𝑚) = 1 − 𝛼
𝛽
(1 − 𝛽)𝜌
𝑐𝑠
𝑓
1−2𝜌𝑚
𝜌−𝛼
𝛼(1−𝜌)(A.2)
Using he de ini ion o 𝜀(𝜔)we hen ge
𝜀(𝜔) = (∫1
0
𝑠(𝑚)𝛼𝑑𝑚)1
𝛼
=1 − 𝛼
𝛽
(1 − 𝛽)𝜌
𝑐𝑠
𝑓
1−2𝜌(1 − 𝜌
1 − 𝛼)1
𝛼.(A.3)
Plugging his, he solu ion o 𝑞(𝜔), as well as he exp ession o 𝑃in o
in e se demand gi es he p icing ule as
𝑝(𝜔) = 𝐴1−𝜌𝑞(𝜔)−(1−𝜌)𝜀(𝜔)𝜌=𝑐𝑥
𝜌
1 − 𝜌
𝛽.(A.4)
Using 𝜀(𝜔)and 𝑝(𝜔)in he de ini ion o he p ice index om demand,
𝑃−𝜌
1−𝜌=∫𝑛
0(𝜀(𝜔)
𝑝(𝜔))𝜌
1−𝜌𝑑𝜔, gi es Eq. (20), a second exp ession o he
p ice index as a unc ion o he numbe o i ms, 𝑛. Combining he wo
exp essions gi es he equilib ium numbe o i ms om Eq. (21),
𝑛=𝛽𝐸
1 − 𝜌
1−2𝜌
𝑓.(A.5)
A.2. Comple e con ac s
In his sec ion, we sol e he model o he case o comple e con-
ac s. Unde comple e con ac s, he headqua e can o e ully spec-
i ied con ac s o each supplie 𝑗and will app oach hem in he igh
o de . The headqua e o e s a con ac consis ing o a physical quan-
i y 𝑥(𝑗)and an acco ding paymen 𝑤(𝑗), as well as an implemen ed
p oduc ion s anda d 𝑠(𝑗)wi h he acco ding paymen 𝑡(𝑗). To make
compa ison wi h he baseline model easible, we deno e all endogenous
a iables in he comple e con ac s case wi h a ilde. The headqua e
hen maximizes i s p o i s wi h espec o he ou choice a iables
abo e, subjec o wo condi ions:
max
{𝑥(𝑗),𝑤(𝑗),𝑠(𝑗),𝑡(𝑗)}𝑗∈[0,1]
𝜋𝐹(𝜔) = 𝑞(𝜔)𝜌
𝐴1−𝜌(∫1
0
𝑠(𝑗)𝛼𝑑𝑗)𝜌
𝛼
−∫1
0
𝑤(𝑗)𝑑𝑗 −∫1
0
𝑡(𝑗)𝑑𝑗 −𝑓
s. . 𝑤(𝑗) − 𝑐𝑥𝑥(𝑗)≥0
𝑡(𝑗) − 𝑐𝑠𝑠(𝑗)≥0
Due o he la ge numbe o a ailable supplie s, he headqua e can
lowe he paymen s 𝑤(𝑗)and 𝑡(𝑗)so ha he cons ain s a e binding
and he supplie walks away wi h ze o p o i s. Because 𝑥(𝑗) = 𝑞(𝜔), we
ha e
∫1
0
𝑤(𝑗)𝑑𝑗 =𝑐𝑥∫1
0
𝑥(𝑗)𝑑𝑗 =𝑐𝑥𝑞(𝜔).(A.6)
and because o he binding cons ain s, ∫1
0𝑡(𝑗)𝑑𝑗 =𝑐𝑠∫1
0𝑠(𝑗)𝑑𝑗.
This implies ha we can es a e he headqua e ’s p oblem in e ms
o 𝑞(𝜔)and 𝑠(𝑗)as
max
𝑞(𝜔),{𝑠(𝑗)}𝑗∈[0,1]
𝜋𝐹(𝜔) = 𝑞(𝜔)𝜌
𝐴1−𝜌(∫1
0
𝑠(𝑗)𝛼𝑑𝑗)𝜌
𝛼
−𝑐𝑥𝑞(𝜔) − 𝑐𝑠∫1
0
𝑠(𝑗)𝑑𝑗 −𝑓(A.7)
Di e en ia ing wi h espec o he choice a iables gi es he i s
o de condi ions
𝑞(𝜔) = 𝜌𝑟(𝜔)
𝑐𝑥
𝑠(𝑗)1−𝛼=𝑟(𝜔)𝜌
𝑐𝑠(∫1
0
𝑠(𝑗)𝛼𝑑𝑗)−1
Exp essing (∫1
0…𝑑𝑗)in e ms o e enue 𝑟(𝜔)gi es 𝑠(𝑗)condi ional
on a le el o ou pu 𝑞(𝜔)as
𝑠(𝑗) = 𝑠 =𝑟(𝜔)
𝜌−𝛼
𝜌(1−𝛼)
𝐴
𝛼(1−𝜌)
𝜌(1−𝛼)𝑞(𝜔)
𝛼
1−𝛼(𝜌
𝑐𝑠)1
1−𝛼
∀𝑗
The di e ence o Eq. (11) is ha he e s age 𝑗s anda d depends on o al
e enue 𝑟(𝜔)ins ead o e enue up o ha s age 𝑟(𝑗).
Plugging his back in o he e enue exp ession om he maximiza-
ion p oblem gi es equilib ium e enue 𝑟(𝜔)as a unc ion o 𝑞(𝜔)as
𝑟(𝜔) =
𝐴[𝑞(𝜔)𝜌
𝑐𝑠]𝜌
1−𝜌(A.8)
Plugging he esul o 𝑟(𝜔)back in o 𝑠 om abo e gi es
𝑠 =
𝐴 𝑞(𝜔)
𝜌
1−𝜌(𝜌
𝑐𝑠)1
1−𝜌
∀𝑗(A.9)
This gi es a symme ic equilib ium in es men in he s anda d 𝑠 a each
s age 𝑗as a unc ion o exogenous pa ame e s and condi ional on he
scale o p oduc ion 𝑞(𝜔).
We know om he i s FOC ha
𝑞(𝜔) = 𝜌𝑟(𝜔)
𝑐𝑥
and using he esul o 𝑟(𝜔) om Eq. (A.8) gi es ha he op imal scale
o p oduc ion is gi en by
𝑞(𝜔) = [𝜌(
𝐴
𝑐𝑥)1−𝜌
𝑐−𝜌
𝑠]1
1−2𝜌
(A.10)
Combining he esul s o Eqs. (A.9) and (A.10) gi es equilib ium
CSR in es men a each s age 𝑗as
𝑠 =[𝜌(
𝐴
𝑐𝑠)1−𝜌
𝑐−𝜌
𝑥]1
1−2𝜌
∀𝑗(A.11)
Using he ac ha wi h ee en y, p o i s a e ze o o all (homoge-
neous) i ms, and applying his o Eq. (A.7) gi es he i s exp ession
o he equilib ium p ice index as
𝑃=𝜌−2𝑐𝑠𝑐𝑥(𝑓
1 − 2𝜌)1−2𝜌
𝜌𝐸−1−𝜌
𝜌(A.12)
Plugging his in o Eq. (A.11) o
𝐴=𝐸
𝑃
𝜌
1−𝜌gi es
𝑠 =𝜌
𝑐𝑠
𝑓
1−2𝜌.(A.13)
Le el o CSR wi h comple e con ac s and compa ison o incomple e con-
ac s. Unde comple e con ac s, he op imal CSR in es men is uni-
o m ac oss all p oduc ion s ages 𝑗and is gi en by 𝑠 =𝜌
𝑐𝑠
𝑓
1−2𝜌. The
uni o mi y esul a ises because all s ages in he p oduc ion p ocess
en e symme ically in o he p oduc ion o e hical quali y. The e o e, a
headqua e able o w i e comple e con ac s has no incen i e o assign
di e en le els o in es men s o supplie s a di e en s ages along he
alue chain.
I is ins uc i e o compa e his esul wi h ou main esul om
Eq. (24). In ac , i is possible o exp ess 𝑠(𝑚)as a unc ion o he
comple e con ac s esul 𝑠 as
𝑠(𝑚) = (1 − 𝛼)1 − 𝛽
𝛽𝑚
𝜌−𝛼
𝛼(1−𝜌)𝑠. (A.14)
Jou nal o De elopmen Economics 167 (2024) 103236
31
P. He kenho e al.
Fan, H., Li, Y.A., Yeaple, S.R., 2015. T ade libe aliza ion, quali y and expo p ices.
Re . Econ. S a . 97 (5), 1033–1051.
Fische , C., Lyon, T.P., 2014. Compe ing en i onmen al labels. J. Econ. Manag. S a egy
23 (3), 692–716.
Flamme , C., 2015. Does co po a e social esponsibili y lead o supe io inancial
pe o mance? A eg ession discon inui y app oach. Manag. Sci. 61 (11), 2549–2568.
Fon agné, L., Lima di, M., 2023. The gene alized sys em o p e e ences and NGO
ac i ism. J. De . Econ. 103207.
Goldbe g, P.K., Khandelwal, A.K., Pa cnik, N., Topalo a, P., 2010a. Impo ed in e me-
dia e inpu s and domes ic p oduc g ow h: E idence om India. Q. J. Econ. 125
(4), 1727–1767.
Goldbe g, P.K., Khandelwal, A.K., Pa cnik, N., Topalo a, P., 2010b. Mul ip oduc i ms
and p oduc u no e in he de eloping wo ld: E idence om India. Re . Econ.
S a . 92 (4), 1042–1049.
Gö g, H., Hanley, A., Se ic, A., 2018. Co po a e social esponsibili y in global supply
chains: Deeds no wo ds. Sus ainabili y 10 (10), 3675.
G ossman, S.J., Ha , O.D., 1986. The cos s and bene i s o owne ship: A heo y o
e ical and la e al in eg a ion. J. Poli . Econ. 94 (4), 691–719.
G ossman, G.M., Helpman, E., 2020. When a i s dis u b global alue chains. In: NBER
Wo king Pape 27722.
G ossman, G.M., Rossi-Hansbe g, E., 2008. T ading asks: A simple heo y o o sho ing.
Ame . Econ. Re . 98 (5), 1978–1997.
Hainmuelle , J., Hiscox, M.J., Sequei a, S., 2015. Consume demand o ai ade:
E idence om a mul is o e ield expe imen . Re . Econ. S a . 97.
Ha ison, A., Sco se, J., 2010. Mul ina ionals and an i-swea shop ac i ism. Ame . Econ.
Re . 100 (1), 247–273.
Ha , O., Moo e, J., 1999. Founda ions o incomple e con ac s. Re . Econom. S ud. 66
(1), 115–138.
He kenho , P., K au heim, S., 2022. The in e na ional o ganiza ion o p oduc ion in
he egula o y oid. J. In . Econ. 137 (103572).
Heyes, A., Ma in, S., 2017. Social labeling by compe ing NGOs: A model wi h mul iple
issues and en y. Manage. Sci. 63 (6), 1800–1813.
Hiscox, M.J., Smy h, N., 2011. Is he e consume demand o ai labo s anda ds?
E idence om a ield expe imen . In: SSRN Wo king Pape 1820642.
Ki zmulle , M., Shimshack, J., 2012. Economic pe spec i es on co po a e social
esponsibili y. J. Econ. Li . 50 (1), 51–84.
Koenig, P., K au heim, S., Löhne , C., Ve die , T., 2021. Local global wa chdogs: T ade,
sou cing and he in e na ionaliza ion o social ac i ism. In: CEPR Discussion Pape
no. 15878.
K au heim, S., Ve die , T., 2016. O sho ing wi h endogenous NGO ac i ism. J. In .
Econ. 101, 22–41.
K ugman, P., 1980. Scale economies, p oduc di e en ia ion, and he pa e n o ade.
Ame . Econ. Re . 70 (5), 950–959.
Locke, R.M., 2013a. Does p i a e compliance imp o e labo s anda ds? Lessons om
nike. In: The P omise and Limi s o P i a e Powe . P omo ing Labo S anda ds in
a Global Economy. Camb idge Uni e si y P ess, pp. 46–77.
Locke, R.M., 2013b. The p omise and pe ils o p i a e compliance p og ams. In: The
P omise and Limi s o P i a e Powe . P omo ing Labo S anda ds in a Global
Economy. Camb idge Uni e si y P ess, pp. 24–45.
Lou ei o, M., Lo ade, J., 2005. Do ai ade and eco-labels in co ee wake up he
consume conscience? Ecol. Econom. 53 (1), 129–138.
Macchia ello, R., 2010. Ve ical in eg a ion and in es o p o ec ion in de eloping
coun ies. J. De . Econ. 93 (2), 162–172.
Macchia ello, R., 2022. Rela ional con ac s and de elopmen . Annu. Re . Econ. 14
(1), 337–362.
Machia ello, R., Mo ja ia, A., 2021. Compe i ion and ela ional con ac s in he wanda
co ee chain. Q. J. Econ. 136 (2), 1089–1143.
Ma ano, V., Kos o a, T., 2016. Unpacking he ins i u ional complexi y in adop ion o
CSR p ac ices in mul ina ional en e p ises. J. Manag. S ud. 53 (1), 28–54.
McKenzie, D., 2023. In e p e ing ea men e ec s on an in e se hype bolic sine
ou come a iable and al e na i es. Wo ld Bank Blogs, h ps://blogs.wo ldbank.
o g/impac e alua ions/in e p e ing- ea men -e ec s-in e se-hype bolic-sine-
ou come- a iable-and. (Accessed 10 No embe 2023).
McWilliams, A., Siegel, D., 2001. Co po a e social esponsibili y: A heo y o he i m
pe spec i e. Acad. Manag. Re . 26 (1), 117–127.
New Yo k Times, 1998. Nike pledges o end child labo and apply U.S. ules
ab oad. h ps://www.ny imes.com/1998/05/13/business/in e na ional-business-
nike-pledges- o-end-child-labo -and-apply-us- ules-ab oad.h ml. (Accessed 19
Sep embe 2019).
Newman, C., Rand, J., Ta p, F., T i ko ic, N., 2018. The ansmission o socially
esponsible beha iou h ough in e na ional ade. Eu . Econ. Re . 101, 250–267.
Newman, C., Rand, J., Ta p, F., T i ko ic, N., 2020. Co po a e social esponsibili y in
a compe i i e business en i onmen . J. De . S ud. 56 (8), 1455–1472.
Nunn, N., 2007. Rela ionship-speci ici y, incomple e con ac s and he pa e n o ade.
Q. J. Econ. 122 (2), 569–600.
Nunn, N., T e le , D., 2008. The bounda ies o he mul ina ional i m: An empi ical
analysis. In: Helpman, E., Ma in, D., Ve die , T. (Eds.), The O ganiza ion o Fi ms
in a Global Economy. Ha a d Uni e si y P ess, pp. 55–83.
OECD/WTO, 2013. Aid o T ade A a Glance 2013: Connec ing o Value Chains. WTO,
Gene a/OECD Publishing, Pa is, h p://dx.doi.o g/10.1787/aid_glance-2013-en.
O’Rou ke, D., 2005. Ma ke mo emen s: Nongo e nmen al s a egies o in luence global
p oduc ion and consump ion. J. Ind. Ecol. 9 (1), 115–128.
Pa cnik, N., 2017. The impac o ade on inequali y in de eloping coun ies. In: NBER
Wo king Pape 23878.
Pisani, N., Kou ula, A., Kolk, A., Meije , R., 2017. How global is in e na ional CSR
esea ch? Insigh s and ecommenda ions om a sys ema ic e iew. J. Wo ld Bus.
52 (5), 591–614.
Podho sky, A., 2015. A posi i e analysis o ai ade ce i ica ion. J. De . Econ. 116,
169–185.
Po e , S., 2019. Label wa s: Compe i ion among NGOs as sus ainabili y s anda d se e s.
J. Econ. Beha . O gan. 160 (2), 1–18.
Schille , C., 2018. Global supply-chain ne wo ks and co po a e social esponsibili y. In:
13 h Annual Mid-A lan ic Resea ch Con e ence in Finance (MARC) Pape .
Se aes, H., Tamayo, A., 2013. The impac o co po a e social esponsibili y on i m
alue: The ole o cus ome awa eness. Manage. Sci. 59 (5), 1045–1061.
Sheng, L., Yang, D.T., 2016. Expanding expo a ie y: The ole o ins i u ional e o ms
in de eloping coun ies. J. De . Econ. 118, 45–58.
Shi odka , V., Beddewela, E., Rich e , U.H., 2018. Fi m-le el de e minan s o poli ical
CSR in eme ging economies: E idence om India. J. Bus. E hics 148 (3), 673–688.
Sho , J.L., To el, M.W., Hugill, A.R., 2016. Moni o ing global supply chains. S a eg.
Manag. J. 37, 1878–1897.
Sho , J.L., To el, M.W., Hugill, A.R., 2020. Imp o ing wo king condi ions in global
supply chains: The ole o ins i u ional en i onmen s and moni o ing p og am
design. ILR Re . 73 (4), 873–912.
Sil a, J.M.C.S., Ten ey o, S., 2006. The log o g a i y. Re . Econ. S a . 88 (4), 641–658.
S ähle , F., Richa dson, M., 2014. Fai ade. Econ. Rec. 90 (291), 447–461.
Tanaka, M., 2020. Expo ing swea shops? e idence om myanma . Re . Econ. S a .
102 (3), 442–456.
Timme , M.P., Die zenbache , E., Los, B., S eh e , R., V ies, G.J., 2015. An illus a ed
use guide o he wo ld inpu –ou pu da abase: he case o global au omo i e
p oduc ion. Re . In . Econ. 23 (3), 575–605.
Wo ld Bank, 2020. Wo ld De elopmen Repo 2020: T ading o De elopmen in he
Age o Global Value Chains. Wo ld Bank, Washing on, DC.
