Place-based Policies and Household Wealth in Africa

Abagna, Ma hew Amali inga; Ho nok, Cecília; Mulyuko a, Alina
A icle — Published Ve sion
Place-based Policies and Household Weal h in A ica
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: Abagna, Ma hew Amali inga; Ho nok, Cecília; Mulyuko a, Alina (2025) : Place-
based Policies and Household Weal h in A ica, Jou nal o De elopmen Economics, ISSN 1872-6089,
Else ie , Ams e dam, Vol. 176, pp. 1-18,
h ps://doi.o g/10.1016/j.jde eco.2025.103482
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/318396
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Con en s lis s a ailable a ScienceDi ec
Jou nal o De elopmen Economics
jou nal homepage: www.else ie .com/loca e/de ec
Regula a icle
Place-based Policies and Household Weal h in A icaI
Ma hew Amali inga Abagnab,c, Cecília Ho noka,∗, Alina Mulyuko a a
aKiel Ins i u e o he Wo ld Economy, Kiel, Ge many
bTax Jus ice Ne wo k, B is ol, England, Uni ed Kingdom
cUni e si y College Dublin, Dublin, I eland
A R T I C L E I N F O
JEL classi ica ion:
F6
F21
O15
O25
Keywo ds:
Special economic zone
Place-based policy
Household weal h
A ica
A B S T R A C T
This pape p o ides no el e idence on he impac o a p ominen place-based policy – Special Economic Zones
(SEZs) – on he economic well-being o A ican households. Exploi ing ime a ia ion in SEZ es ablishmen
on a da ase o epea ed c oss-sec ions o households in 10 A ican coun ies du ing 1990-2020, we show
ha households li ing nea SEZs become weal hie ela i e o he na ional a e age a e SEZ es ablishmen .
The e ec acc ues mos ly wi hin 10 km o SEZs, is no d i en by selec i e mig a ion, and is accompanied by
imp o ed access o household u ili ies, highe consump ion o du able goods, inc eased educa ional a ainmen
and a shi away om ag icul u al ac i i ies.
1. In oduc ion
Place-based p og ams – go e nmen al policy ools used o p o-
mo e local economic de elopmen and educe egional dispa i ies –
ha e gained signi ican p ominence in de eloping coun ies in ecen
decades (e.g., Du an on and Venables, 2021). These policies p ima ily
aim o a ac in es men s, c ea e jobs, and imp o e he wel a e o
local esiden s. Despi e hei g owing popula i y, empi ical e idence
on hei e ec i eness in de eloping coun ies, pa icula ly in A ica,
emains limi ed. P io esea ch om de eloped coun ies (Glaese and
Go lieb, 2008; Busso e al., 2013; Neuma k and Kolko, 2010; Kline,
2010; Ham e al., 2011; Kline and Mo e i, 2014a; Neuma k and Simp-
son, 2015; Blou i and Eh lich, 2020) canno be di ec ly ex apola ed
o he de eloping wo ld. Meanwhile, exis ing s udies on de eloping
coun ies ha e la gely ocus on adi ional i m-le el ou comes, such
as p oduc i i y, wage, and employmen (e.g., Schminke and Van Biese-
b oeck, 2013; Wang, 2013; Lu e al., 2019; Gö g and Mulyuko a, 2024),
IAcknowledgemen : The au ho s a e g a e ul o Jan Da id Bakke , Is án Boza, Ri am Chau ey, Ron Da ies, Josh Dean, Rachel Glenne s e , Vic o Gimenez-
Pe ales, Robe Gold, Holge Gö g, Hannes Schwand , No a S ecke , wo anonymous e e ees, and pa icipan s a he ollowing wo kshops and semina s o hei
commen s: 24 h Gö inge Wo kshop, 13 h Eu opean Mee ing o he U ban Economics Associa ion, ETSG 2024 Annual Con e ence, and in e nal semina s a he
Kiel Ins i u e, Uni e si y College Dublin, Uni e si y o Chicago and No hwes e n Uni e si y. Funding: This wo k was unded by he Riksbankens Jubileums ond
o Sweden as pa o he esea ch p ojec "Special Economic Zones: A Fo ce o Good o Reduce Inequali y?" (G an ID: GI19-1501:1). Ho nok and Mulyuko a
g a e ully acknowledge he inancial suppo . Ho nok and Mulyuko a also hank he Ga es Founda ion o pa ially unding his esea ch (G an ID: INV-048832).
The indings and conclusions con ained wi hin a e hose o he au ho s and do no necessa ily e lec posi ions o policies o he Bill & Melinda Ga es Founda ion.
∗Co esponding au ho .
E-mail add esses: [email p o ec ed] (M.A. Abagna), [email p o ec ed] (C. Ho nok), [email p o ec ed] (A. Mulyuko a).
1In o ma ion on popula ion, GDP, and income s a us a e om he Wo ld Bank, Wo ld De elopmen Indica o s 2023 and he Kiel Ins i u e A ica
Moni o (K an z, 2023).
wi h li le a en ion paid o he wel a e and dis ibu ional e ec s on
households (e.g., Pica elli, 2016).
This pape p o ides no el e idence on he e ec s o Special Eco-
nomic Zones (SEZs) – one o A ica’s mos widesp ead place-based
p og ams – on household economic well-being. The e m SEZ is used in
his pape as an umb ella e m ha encompasses a ious ypes o zones,
such as indus ial pa ks, expo p ocessing zones, o echnology pa ks.
SEZs a e designa ed geog aphical a eas ha p o ide speci ic incen i es
and egula ions o os e economic g ow h, inno a ion, and job c ea ion
in he a ge a ea. Acco ding o AEZO (2021), he numbe o SEZs in
A ica has isen om 20 in he ea ly 1990s o mo e han 200 in 2021.
While SEZs ha e become an impo an de elopmen ool, he ques ion
o whe he SEZs ha e con ibu ed o imp o ed wel a e o households
emains unclea .
To es ima e he local impac o SEZs, we compile a no el da ase o
epea ed c oss-sec ions o households esiding in a ious non-
o e lapping dis ance bands a ound SEZs in 10 A ican coun ies o e
h ps://doi.o g/10.1016/j.jde eco.2025.103482
Recei ed 10 Ma ch 2024; Recei ed in e ised o m 24 Feb ua y 2025; Accep ed 25 Feb ua y 2025
Jou nal o De elopmen Economics 176 (2025) 103482
A ailable online 1 Ap il 2025
0304-3878/© 2025 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/ ).
M.A. Abagna e al.
he pe iod o 1990 o 2020 using he Demog aphic and Heal h Su eys
(DHS) and geo e e enced SEZs da a. The 10 coun ies – Egyp , E hiopia,
Ghana, Kenya, Mali, Mozambique, Nige ia, Tanzania, Uganda, Zambia
– a e a ai e lec ion o he policy in he A ican con inen , as hey
hos mo e han hal o he SEZs and accoun o nea ly hal o he
con inen ’s popula ion as well as i s G oss Domes ic P oduc (GDP). In
e ms o economic de elopmen , hey ange om lowe -middle-income
coun ies o low-income economies, wi h some s ill hea ily elian on
ag icul u e and mining (e.g. Mozambique, Zambia), while o he s ha e
mo e di e si ied economies (e.g. Kenya).1
We measu e he o e all economic well-being o a household wi h
he DHS weal h index, a composi e indica o based on household su ey
esponses ega ding asse owne ship and access o household se ices.
To iden i y he e ec o SEZ policies on household weal h, we employ
a di e ence-in-di e ences (DiD) amewo k wi h s agge ed ea men
iming, le e aging ime a ia ion in he es ablishmen o SEZs. Focusing
on ul ima ely ea ed loca ions, ou es ima e compa es weal h ajec-
o ies o households wi hin he same dis ance band o ea ly- e sus
la e- ea ed SEZ si es, wi h ea men de ined as he es ablishmen o
a nea by SEZ. The es ima ed weal h e ec e lec s changes ela i e o
he na ional a e age, as he weal h index is no malized by cons uc ion.
We complemen his esul by examining he weal h componen s, such
as e ige a o owne ship and elec ici y access, o illus a e absolu e
weal h changes.
Es ima ing he causal e ec o SEZs on households is no i ial
due o he non- andom assignmen o SEZs in space. I , o example,
go e nmen s use SEZs o a ac o eign in es o s, hen i is mo e
likely ha ela i ely de eloped a eas ge ea ed, which may lead o
an upwa d bias in he es ima es. We add ess his issue in ou base-
line es ima ion bo h by concen a ing on loca ions ha a e ul ima ely
ea ed and by accoun ing o SEZ ixed e ec o abso b any ime-
in a ian di e ences be ween he loca ions, such as di e ences in
ini ial condi ions. Howe e , ou es ima es may s ill be biased i , in he
absence o SEZ, household weal h is ending di e en ly in ea ly han in
la e- ea ed loca ions. We p o ide e en s udies and obus ness checks
o show ha a common end assump ion is no iola ed. A complemen-
a y e en s udy, using nigh ligh in ensi y as he ou come a iable,
p o ides addi ional con i ma ion o pa allel ends in p e- ea men
economic ac i i y. We also pe o m al e na i e es ima ions wi h house-
holds loca ed a he om he SEZs as a con ol g oup, combined
wi h p opensi y sco e ma ching, and ind simila es ima es. Finally, o
add ess he issue o po en ial biases s emming om ea men e ec
he e ogenei y ac oss ime and loca ion, we p o ide es ima es using
he es ima o o Callaway and San ’Anna (2021) and he Ex ended
Two-Way Fixed-E ec s me hod p oposed by Woold idge (2021).2
Ou baseline es ima e sugges s ha households esiding wi hin 10
km o an SEZ expe ience an inc ease o 0.25 s anda d de ia ion o he
weal h index ela i e o he coun y a e age a e he es ablishmen
o he SEZ. This e ec is economically subs an ial and compa able o
owning a compu e in Nige ia in 2008. The e ec diminishes apidly
wi h dis ance, bo h in e ms o magni ude and s a is ical signi icance,
and disappea s beyond 20 km. Mo ing beyond he a e age e ec , we
also show ha he weal h e ec is dis ibu ed b oadly ac oss weal h
classes, as shown by Quan ile T ea men E ec (QTE) es ima ion a
he deciles o he weal h dis ibu ion (Fi po, 2007; Fi po and Pin o,
2016). The la ges poin es ima es occu nea he middle o he local
weal h dis ibu ion, sugges ing ha SEZ policies ha e con ibu ed o
s eng hening A ica’s middle class.
Place-based policies impac popula ion mig a ion in wo ways: new
economic oppo uni ies d i e in-mig a ion, while inc easing local de-
mand and ising p ices may igge ou -mig a ion (Eh lich and O e -
man, 2020; Kline and Mo e i, 2014b; Glaese and Go lieb, 2008). We
2Fo u he discussion see Goodman-Bacon (2021), de Chaisema in and
D’Haul œuille (2020), Callaway and San ’Anna (2021), and Woold idge
(2021).
a gue ha he es ima ed weal h e ec is no me ely a byp oduc o
changes in he local popula ion caused by he in-mig a ion o weal hie
and he ou -mig a ion o poo e households. On a sub-sample whe e
he mig a ion his o y o households is obse ed, we ind ha SEZ
es ablishmen does no lead o any no iceable ou -mig a ion o na i e
households, al hough i does inc ease he in-mig a ion o ela i ely
young households. Fu he mo e, he weal h e ec is ai ly e enly dis-
ibu ed ac oss households wi h di e en mig a ion backg ounds, wi h
na i es bene i ing a leas as much as he a e age weal h gain.
An examina ion o he componen s o he weal h index e eals
ha he es ablishmen o SEZs leads o imp o ed access o household
u ili ies such as elec ici y and sani a ion, enhanced housing quali y,
and inc eased household consump ion o du able goods such as ele-
isions, e ige a o s, and au omobiles. While hese imp o emen s can
be seen as mani es a ions o highe household incomes, be e access
o elec ici y and sani a ion acili ies may also esul om he physical
in as uc u e upg ades ha o en accompany SEZs de elopmen s (e.g.,
Fa ole and Akinci, 2011).
An impo an channel h ough which SEZs imp o e he economic
well-being o households is he c ea ion o new, be e jobs (Neuma k
and Simpson, 2015; Mo e i, 2024). Se e al s udies ha e documen ed
ha place-based policies ha e inc eased employmen in su ounding
a eas (Ham e al., 2011; Gobillon e al., 2012; Kline and Mo e i, 2014a;
Chau ey, 2017; C iscuolo e al., 2019; Gibbons e al., 2019; Blakeslee
e al., 2022). Looking a SEZ p og ams, Zheng e al. (2017) and Lu
e al. (2019) ind ha he es ablishmen o SEZs in China posi i ely
impac s bo h employmen le els and employmen g ow h in designa ed
a eas. This e ec is u he accompanied by wage inc eases o local
wo ke s (Wang, 2013) and a sec o al shi owa ds non-ag icul u al
ac i i ies (Zhao and Qu, 2024).
The DHS da a o e limi ed scope o s udy he employmen channel.
We obse e employmen ou comes o a selec i e sample o household
membe s, namely emales in ep oduc i e age wi h child en and hei
male pa ne s, and ind li le o no e idence ha SEZ es ablishmen
inc eases he p obabili y o employmen in his g oup. Howe e , con-
sis en wi h p io li e a u e, we documen a shi om ag icul u e
o non-ag icul u al occupa ions among household membe s ollowing
SEZ es ablishmen . We in e p e his esul as e idence o an u -
baniza ion end, as locals ansi ion om ag icul u al occupa ions o
non-ag icul u al jobs.
Fu he , place-based policies can a ec local economic g ow h
h ough hei impac on in es men in human capi al. Technology-
o ien ed policies ha e been shown o lead o highe educa ional a -
ainmen as hey inc ease he e u ns o educa ion (e.g., Lu e al.
(2023) on Chinese SEZs o Os e and S einbe g (2013) on a new
IT cen e in India). In con as , expo -o ien ed manu ac u ing zones
can discou age s uden s o inish hei educa ion because o inc eased
oppo uni y cos s (A kin, 2016). This s udy demons a es ha SEZ
policies in A ica a e associa ed wi h imp o ed educa ional ou comes,
as young household membe s a e mo e likely o comple e seconda y
educa ion ollowing he es ablishmen o a nea by SEZ. The posi i e
educa ional esponse may be d i en by ising household incomes o
inc eased skilled wage p emiums.
O he channels h ough which place-based policies can a ec lo-
cal economic de elopmen include consump ion spillo e s and de-
mand mul iplie s. Employmen and wage inc eases may gene a e lo-
cal demand o housing, e ail and o he ameni ies, as demons a ed,
e.g., by Zheng e al. (2017) in he con ex o China’s indus ial pa ks,
which in u n os e s addi ional job c ea ion h ough a local mul iplie
e ec . Mo e i (2010) es ima es ha each addi ional manu ac u ing job
in a ci y gene a es 1.6 jobs in he non- adable sec o o he same ci y,
wi h signi ican ly la ge e ec s o skilled jobs. While such de elop-
men s end o lead o highe local li ing cos s (Busso e al., 2013),
hese a e o en o se by inc eased ea nings, esul ing in highe wo ke
pu chasing powe despi e ising expenses (Wang, 2013; Ho nbeck and
Mo e i, 2024). Place-based policies can also in luence local economic
Jou nal o De elopmen Economics 176 (2025) 103482
2
M.A. Abagna e al.
de elopmen by gene a ing en i onmen al ex e nali ies, pa icula ly
when hey a e implemen ed alongside s ic e adhe ence o sus ain-
able en i onmen al p ac ices, con ibu ing o imp o ed en i onmen al
quali y and highe li ing s anda ds (Hua e al., 2023). Fu he , ising
house p ices can also d i e he upg ading o he local housing s ock,
ul ima ely imp o ing housing condi ions (Kos e and Van Omme en,
2019). While simila mechanisms a e likely con ibu ing o he es i-
ma ed weal h e ec o SEZs in A ica, we canno es hei p esence
due o da a limi a ions, as disagg ega ed da a on wages, consump ion,
e ail ac i i y, o housing p ices a e no a ailable a he equi ed le el
o g anula i y.
Th ee pape s a e mos closely ela ed o ou wo k. The i s is
he pape by Pica elli (2016), which examines he e ec o Expo
P ocessing Zones (EPZs) on he le el o pe capi a expendi u e in
Nica aguan municipali ies. She inds ha , on a e age, he consump ion
le el inc eased by 12% in he ea ed municipali ies. Second, using
household su eys om Cambodia, B usse ich (2024) inds ha SEZs
inc ease employmen and dec ease income inequali y a he dis ic
le el. In con as o hese s udies, which use adminis a i e a eas as
he uni o analysis, we use geoloca ed household in o ma ion and
assign ea men o DHS clus e s, which co espond o a illage o
u ban neighbo hood and a e hus mo e g anula han dis ic s o mu-
nicipali ies. This leads o mo e p ecise es ima es and educes conce ns
abou measu emen e o while a he same ime allows us o look a
households as he uni o analysis. The hi d closely ela ed wo k is
by Shenoy (2018), who inds ha in as uc u e and in es men subsi-
dies inc eased he a ailabili y o public goods and imp o ed household
wel a e. In con as o Shenoy (2018), who looks a one s a e and a
single ax ans e p og am in India, we examine 10 A ican coun ies
and all SEZs incen i es, which allows us o p o ide e idence-based
policy ad ice a he agg ega e le el.
P e ious s udies on SEZs ha e p ima ily used i m-le el da a o
analyze he e ec s o he policy. Wang (2013) and Lu e al. (2019),
o ins ance, show ha he es ablishmen o SEZs in China inc eases
o eign di ec in es men s, p oduc i i y, employmen and wages o
manu ac u ing i ms. The inc ease in wages is highe han he in-
c ease in he local cos o li ing, gene a ing ne bene i s o wo k-
e s. Schminke and Van Bieseb oeck (2013) show ha p e e en ial e-
gional policies p omo e expo ing ac i i y among manu ac u ing i ms
bo h in e ms o olume and he numbe o des ina ion coun ies. In
India, Gö g and Mulyuko a (2024) show ha he es ablishmen o SEZs
does no ha e any disce nible e ec on he p oduc i i y g ow h o
i ms, whe eas Alkon (2018) documen s ha he p og am did no b ing
any local socio-economic de elopmen . O he ela ed wo k e alua ing
place-based ax incen i e p og ams is by Chau ey (2017), Hasan e al.
(2021) and Blakeslee e al. (2022) who ind ha indus ial policies
signi ican ly inc ease i m en y and employmen .
The li e a u e assessing he impac o SEZs in A ica is sca ce and,
i any hing, p edominan ly desc ip i e in na u e, using indi idual SEZs
as case s udies.3 Mo eo e , qui e a ew o hese s udies ocus on he
employmen gene a ion o wage e ec s o SEZs, o en wi h an emphasis
on emale empowe men (see, e.g. Glick and Roubaud (2006) o
Madagasca and Obeng e al. (2015) o Ghana). On he con a y, we
le e age de ailed household-le el da a o explo e he a e age e ec o
SEZs and how he policy e ec s a y along he weal h dis ibu ion. Ou
da a, mo eo e , allow us o decompose he agg ega e e ec and look
a he componen s o he weal h index sepa a ely. To he bes o ou
knowledge, his is he i s pape o examine he e ec s o A ican SEZs
om hese pe spec i es.
Mo e b oadly, ou pape ela es o he li e a u e on he e ec s o
place-based policies. P io esea ch has ocused p ima ily on e alua ing
3See, o ins ance, Thompson (2019) o Sou h A ica, Phi i e al. (2020)
o Sou h A ica and Zambia, Xu and Wang (2020) o E hiopia and Zambia,
o Adunbi (2019) o Nige ia.
spa ially a ge ed policies in de eloped coun ies, see, e.g. Glaese
and Go lieb (2008), Busso e al. (2013), Kline (2010), Neuma k and
Simpson (2015). Kline and Mo e i (2014a) ind ha he Tennessee
Valley Au ho i y p og am inc eased manu ac u ing employmen and
gene a ed agglome a ion economies. Neuma k and Kolko (2010), on
he con a y, ind no employmen gains ollowing Cali o nia’s en e -
p ise zone p og am. Ham e al. (2011) show ha S a e En e p ise Zones
ha e a la ge posi i e impac on he local labo ma ke . Blou i and
Eh lich (2020) documen ha egional policies ha e con ibu ed o
wel a e gains. While hese s udies gene ally sugges ha such policies
yield imp o emen s in he well-being o he in ended bene icia ies, he
esul s canno be di ec ly ex ended o de eloping coun ies due o he
di e ences in he p og ams and ins i u ions.
The emainde o he pape is s uc u ed as ollows. Sec ion 2 o e s
a b ie o e iew o SEZ policies in A ica, while Sec ion 3 p o ides a
de ailed desc ip ion o he da a used in he empi ical analysis. In Sec-
ion 4, we ou line he es ima ion amewo k and add ess key challenges
ela ed o iden i ica ion. Sec ion 5 p esen s he main esul s, including
obus ness checks, an analysis o mig a ion pa e ns, and an explo a ion
o e ec he e ogenei ies and dis ibu ional impac s. Sec ion 6 del es
in o he ole o indi idual weal h componen s and examines he policy’s
e ec s on employmen and educa ion. Finally, Sec ion 7 concludes.
2. Ins i u ional backg ound
A ican coun ies a e ela i e la ecome s in applying SEZ policies.
The es ablishmen o SEZs has only sp ead ac oss he con inen in
he las wo decades. Howe e , by 2022, mos coun ies had adop ed
ac i e SEZ p og ams, esul ing in app oxima ely 203 ope a ional SEZs
and nea ly 100 mo e unde cons uc ion (Rod íguez-Pose e al., 2022).
Fig. 1 illus a es he geog aphical dis ibu ion o all ope a ional SEZs
in A ica, wi h ed do s indica ing hose in he en coun ies ana-
lyzed. This g owing end e lec s he adop ion o SEZs by policy-
make s o align wi h he goals o he A ican Union’s Agenda 2063
and he UN Sus ainable De elopmen Goal 9, which unde sco es he
ole o indus ializa ion in gene a ing employmen and ele a ing li ing
s anda ds.4
Two main pa e ns accoun o he ecen inc ease in SEZs in
A ica. Fi s ly, coun ies wi h es ablished SEZ p og ams such as Ghana,
Egyp , Nige ia and Kenya a e ac i ely expanding and di e si ying hei
exis ing SEZ po olios. Thei goal is o d i e s uc u al ans o ma-
ion, enhance pa icipa ion in Global Value Chains (GVC), and c ea e
employmen oppo uni ies h ough es ablished p oduc ion ne wo ks.
Secondly, eme ging playe s a e es ablishing new SEZs o a ac Fo eign
Di ec In es men (FDI), boos expo s, and gene a e employmen , pa -
icula ly o low-skill labo e s (UNCTAD, 2021). These ac i e ini ia i es
highligh he e ol ing signi icance o SEZs in shaping he economic
landscape o coun ies in he con inen .
The placemen o SEZs is a s a egic decision. This mul i ace ed
p ocess is in luenced by a ious ac o s such as he in ended pu pose,
owne ship s uc u e, and he economic de elopmen le el o he hos
coun y (UNECA, 2022). SEZs ini ia ed by he s a e ypically align
wi h local de elopmen p io i ies and esou ce a ailabili y, o en con-
cen a ing in less u ban a eas wi h h i ing ag icul u e and na u al
esou ce ex ac ion. In con as , SEZs p oposed by p i a e in es o s,
subjec o s a e app o al, ypically ope a e in he manu ac u ing o
se ices sec o s and a e s a egically loca ed in u ban and semi-u ban
4Agenda 2063 is a s a egic amewo k ha aims o achie e inclusi e
and sus ainable g ow h, e lec ing he Pan-A ican d i e owa ds uni y, sel -
de e mina ion, eedom, p og ess, and collec i e p ospe i y as exp essed in
he ideals o Pan-A icanism and he A ican Renaissance. Goals 1 and 4 o
he agenda aim o imp o e he s anda ds o li ing and p omo e well-being
o all by c ea ing employmen oppo uni ies and ans o ming economies
h ough indus ializa ion. Fo he UN Sus ainable De elopmen Goal 9, see
h ps://sdgs.un.o g/goals/goal9.
Jou nal o De elopmen Economics 176 (2025) 103482
3
M.A. Abagna e al.
Fig. 1. Map o SEZs in A ica.
a eas, o en nea ai po s and seapo s. The s a egic posi ioning se es
a dual pu pose: le e aging anspo a ion hubs and na u al esou ces
o a ac in es o s while os e ing he agglome a ion o i ms capa-
ble o p o iding employmen oppo uni ies o bo h u ban and u al
popula ions (UNCTAD, 2021). While he loca ion choices o SEZs in
some sample coun ies a e no explici ly s a ed, i is wo h no ing
ha poli ical conside a ions, in addi ion o economic ac o s, may also
in luence hese decisions.
While go e nmen s o p i a e in es o s can ini ia e SEZs, a s a e
en i y ends o o e see hei p og ess. This en i y can be an independen
SEZ au ho i y, a ele an minis y like he Minis y o Indus y, o a
na ional In es men P omo ion Agency (IPA).5 These au ho i ies ha e a
ange o unc ions, including app o ing applica ions o es ablish p i a e
SEZs and egula ing p i a e SEZs. The designa ed SEZ ins i u ions also
play a ole in de eloping, au ho izing i m ope a ions and managing
s a e-owned SEZs (Fa ole and Mobe g, 2017).
Aligned wi h he con en ional SEZ incen i e s uc u e, A ican SEZs
o e a ious o ms o iscal and non- iscal bene i s such as educ ions
o exemp ions om co po a e and local axes, alongside wai e s o
impo du ies on machine y and p oduc ion inpu s and ou pu s, as well
as in as uc u al upg ades (UNCTAD, 2019). Some coun ies also o e
ax deduc ions o skill de elopmen p og ams sponso ed by SEZ-based
i ms ha a ge local wo ke s, hi e local wo ke s, use local con en
o mee designa ed expo a ge s. Addi ionally, ew coun ies p o ide
one-s op se ice cen e s wi hin he SEZs o deli e go e nmen se ices
o he SEZs i ms (UNCTAD, 2021). Impo an ly, when essen ial in-
as uc u e is lacking, ei he he s a e o p i a e de elope s ake he
lead in p o iding he necessa y ameni ies such as oads, elec ici y,
5One example o an independen SEZ au ho i y is he Ghanaian F ee
Zones Au ho i y, while an example o an IPA is he E hiopian In es men
Commission.
d ainage, and esiden ial accommoda ions be o e he cons uc ion and
commissioning o he SEZ. No able examples o such e o s include he
Kigali SEZ in Rwanda (s udied in S eenbe gen and Ja o cik (2017)) and
he Mombasa SEZ in Kenya.6
The majo i y o he SEZs in ou s udy a e o ien ed owa ds indus ial
ac i i ies, pa icula ly manu ac u ing and assembling, ag o-p ocessing,
and na u al esou ce-in ensi e ac i i ies (Appendix Table A2). SEZs
specialized solely in se ices cons i u e only sligh ly o e 10% o ou
sample SEZs, while a signi ican numbe o SEZs a e engaged in mixed
(i.e. bo h indus ial and se ice) ac i i ies. The se ice sec o SEZs
o en unc ion as logis ics hubs p o iding comme cial and wa ehousing
se ices nea anspo a ion hubs. O e all, he sec o dis ibu ion o he
SEZs poin s o a lack o specializa ion in mos A ican zones.
3. Da a
One o he main challenges in e alua ing he impac o SEZs on
household weal h in de eloping coun ies, pa icula ly in A ica, is he
limi ed a ailabili y o da a on households loca ed nea he zones. To
o e come his limi a ion, we employ a no el app oach by geocoding
SEZ loca ions and spa ially joining hem wi h geo e e enced household
da a, which o ou knowledge has no been used in he li e a u e on
place-based policies be o e.
3.1. Households
We de i e ou household da a om he Demog aphic and Heal h
Su eys (DHS) P og am adminis e ed by he Uni ed S a es Agency o
In e na ional De elopmen (USAID). The DHS is a comp ehensi e and
na ionally ep esen a i e su ey ha collec s da a on a ious aspec s o
popula ion, heal h, nu i ion, demog aphics, and socioeconomic cha -
ac e is ics in de eloping coun ies. Conduc ed pe iodically, ypically
e e y i e yea s, i p o ides s anda dized and compa able in o ma ion
ac oss coun ies. The da a is used widely by esea che s, policymake s,
and in e na ional o ganiza ions o in o m policy decisions and guide
in e en ion s a egies.7
The selec ion o coun ies o ou analysis is p ima ily guided by
da a a ailabili y. We begin by le e aging ou SEZs da a (discussed in
he nex sec ion) o iden i y all A ican coun ies wi h a leas one
ac i e SEZ. Subsequen ly, we na ow ou ocus o coun ies ha ha e
geocoded DHS wi h in o ma ion on household asse weal h bo h be o e
and a e he es ablishmen o a leas one o hei SEZs. We collec da a
om bo h s anda d DHS and In e im DHS (I-DHS) ounds. Addi ionally,
we inco po a e da a om he Mala ia Indica o Su eys (MIS) and
he AIDS Indica o Su eys (AIS) o he DHS P og am, p o ided hey
o e in o ma ion on households’ weal h and geo e e enced loca ion.
Al hough he di e en su ey ypes di e in hei opical ocus and
sample size (wi h he S anda d DHS being he la ges ), hey a e all
based on a ep esen a i e sample o a coun y’s popula ion.
Ou inal sample comp ises 10 A ican coun ies, spanning a o al
o 57 DHS su ey ounds. Table 1 p o ides a comp ehensi e lis o he
coun ies in ou sample and he co esponding pe iods co e ed by ou
da a. Excep o Mozambique and Zambia, coun ies in ou sample ha e
mo e han i e ounds o DHS, a o ding us ex ensi e empo al co e age
o analyze he impac o SEZs on household weal h.
Gi en he su ey’s p ima y ocus on popula ion, heal h and nu-
i ion ou comes, he e a e limi ed socio-economic indica o s a he
household le el, such as employmen , wages, o consump ion expen-
di u e. Consequen ly, we p ima ily ely on he household asse weal h
index ( e e ed o as he weal h index) o gauge he economic s a us
6The in o ma ion is om h ps://openjica epo .jica.go.jp/pd /12338448.
pd . Accessed on: 22.01.2024.
7Fo a de ailed desc ip ion o he da a isi : h ps://dhsp og am.com/
Me hodology/Su ey-Types/DHS.c m.
Jou nal o De elopmen Economics 176 (2025) 103482
4

M.A. Abagna e al.
Table 1
DHS su ey ounds by coun y.
Coun y Su ey ounds
Egyp 1995, 2000, 2003 (I-DHS), 2005, 2008, 2014
E hiopia 2000, 2005, 2011, 2016, 2019 (I-DHS)
Ghana 1993, 1998, 2003, 2008, 2014, 2016 (MIS), 2019 (MIS)
Kenya 2003, 2008–09, 2014, 2015 (MIS), 2020 (MIS)
Mali 1995–96, 2001, 2006, 2012–13, 2015 (MIS), 2018
Mozambique 2009, 2011, 2015, 2018 (MIS)
Nige ia 1990, 2003, 2008, 2010 (MIS), 2013, 2015 (MIS), 2018
Tanzania 1999, 2003–04 (AIS), 2007–08 (AIS), 2010, 2011–2012 (AIS), 2015–16, 2017 (MIS)
Uganda 2000–01, 2006, 2009 (MIS), 2011, 2014–15 (MIS), 2016, 2018–19 (MIS)
Zambia 2007, 2013–14, 2018
No e: S anda d DHS su eys unless o he wise no ed. I-DHS: In e im DHS, MIS: Mala ia Indica o Su ey, AIS: AIDS
Indica o Su ey. I a su ey was conduc ed in wo consecu i e yea s, we assign he da a o he i s yea , unless
mo e han wo hi ds o he in e iews ook place in he second yea , in which case we assign he da a o he second
yea .
o households. The DHS weal h index has been used ex ensi ely o
measu e household economic well-being in he de elopmen economics
li e a u e (e.g., on de Gol z and Ba nwal, 2019; Lowes and Mon e o,
2021) and is pa icula ly aluable in coun ies wi h no eliable income
o expendi u e da a. I se es as a composi e measu e depic ing he
cumula i e li ing s anda ds o a household, as i is cons uc ed using
P incipal Componen Analysis based on household esponses ega ding
owne ship o selec ed asse s and access o se ices.8 The DHS weal h
index is s anda dized by design o ha e a mean alue o ze o and
a s anda d de ia ion o one o each su ey, p o iding a measu e o
a household’s weal h posi ion ela i e o he na ional a e age in he
su ey yea .9
3.2. SEZs
In cons uc ing ou SEZs da ase , we d aw in o ma ion om wo
p ima y sou ces; he Open Zone Map and he A ica Economic Zones
O ganiza ion (AEZO), supplemen ed by ou independen da a collec-
ion.10 Ou lis o SEZs co e s a wide ange o zone ypes, including
indus ial pa ks, expo p ocessing zones, echnology pa ks, and o he s,
all o which a e g ouped unde he umb ella e m SEZ. Howe e ,
single-company zones a e excluded om he analysis.11
We ob ain in o ma ion on he loca ion, managemen ype, land
size, and ope a ional s a us o SEZs om he Open Zone Map and
he AEZO. Due o inconsis encies in da a ela ed o he es ablishmen
da es o some SEZs om hese sou ces, we manually collec he yea
o es ablishmen om a ious web-based sou ces, including he o icial
websi es o he SEZs and coun y-le el ins i u ions esponsible o
managing he SEZs. Simila ly, we collec in o ma ion on he sec o al
specializa ion o indi idual SEZs and de ine ou b oad ca ego ies:
Indus y, Se ices, Mixed ac i i ies, and No iden i ied o hose SEZs
whose sec o al in o ma ion is no a ailable.12
8These se ices and asse s include he sou ce o d inking wa e , he ype
and p i acy le el o oile acili ies, he ma e ial composi ion o he main
loo , walls, and oo , and ype o windows in he house. Addi ionally, i
inco po a es he ype o cooking uel, household se ices and possessions
(such as elec ici y, TV, adio, wa ch, and ehicles), ag icul u al land size
and ype o owne ship, he numbe o owned animals, and he p esence
o a bank accoun . Fo mo e de ails o he weal h index cons uc ion, see
h ps://dhsp og am.com/ opics/weal h-index/.
9Fo a isual ep esen a ion o he weal h index sco e dis ibu ion among
sample households in he i s and las DHS by coun y, please e e o
Appendix Figu e A1.
10 Open Zone Map p o ides a comp ehensi e mapping o SEZs wo ldwide.
The Ad ianople G oup main ains i and can be accessed a : h ps://www.
openzonemap.com/. The AEZO A las is a ailable upon subsc ip ion a : h ps:
//www.a icaeconomiczones.com/aezo-a las/.
11 Single- i m zones di e om he zones we co e as hey do no e e o
a speci ic geog aphical a ea. They also end o be small and employ ela i ely
ew wo ke s (UNCTAD, 2021).
A o al o 114 SEZs a e p esen ac oss he en coun ies o which
he a o emen ioned da a is a ailable. This is educed o 51 SEZs in
he inal es ima ion sample, de e mined by he a ailabili y o DHS
da a. The dis ibu ion o SEZs by coun y and ime o es ablishmen
is shown in Appendix Table A1 while he dis ibu ion by managemen
ype (p i a e, public, p i a e-public pa ne ship), sec o al ac i i y, and
land a ea size is de ailed in Appendix Table A2, o bo h he comple e
se o SEZs and hose in he es ima ion sample. The es ablishmen da es
indica e ha SEZ policy is a ela i ely ecen phenomenon in A ica,
wi h mo e han hal o he SEZs being es ablished a e 2010, and
an addi ional 20% du ing he 2000s. The ela i e majo i y o SEZs
a e publicly managed, specialized in indus ial ac i i ies, and a e o
medium in land size, anging be ween 100 and 1000 hec a es. The
numbe o SEZs ha a e la ge in land size o specialized only in se ices
is ela i ely low.
SEZs a e impo an playe s in he local economies. Al hough com-
p ehensi e da a on he economic impo ance o SEZs in A ica a e no
a ailable, we ha e in o ma ion on he numbe o i ms and jobs o 15
o he 51 SEZs in ou es ima ion sample. The 15 SEZs, which a e all
ei he small o medium-sized in e ms o land a ea, hos an a e age o
42 businesses and gene a e an a e age o 12,300 jobs.13 This indica es
ha SEZs ep esen a conside able economic o ce.
3.3. Combining DHS and SEZ da a
We spa ially join he household and SEZ da a by u ilizing he
GPS coo dina es o bo h he SEZs and DHS clus e s. Fig. 2 isually
demons a es ha he DHS clus e s in ou sample coun ies align closely
wi h he loca ions o SEZs. Since he DHS is a na ionally ep esen a i e
su ey, his spa ial o e lap indica es ha he SEZs in ou sample
coun ies a e si ua ed in egions whe e mos o he su eyed popula ion
li e. As a esul , a signi ican numbe o households a e exposed o he
po en ial impac o SEZs, enhancing he ele ance o ou analysis.
To cons uc ou es ima ion sample, we adop he ollowing me hod-
ology. Ini ially, we de ine he inne a ea o each SEZ by d awing a
ci cle a ound he cen oid o he SEZ wi h a adius p opo ional o he
SEZ o icial a ea size. Then, we c ea e dis ance bands by inc emen ally
ex ending he adius by 10 km.14 By doing so, we es ablish non-
o e lapping dis ance bands a ound each SEZ and assign households
o hese bands based on he geog aphical coo dina es o he su ey
clus e s ( illages o u ban neighbo hoods) whe e he households a e
loca ed. In cases whe e a household’s neighbo hood is wi hin he
12 Indus y includes manu ac u ing, ag o-p ocessing and ene gy. Se ices
include anspo , logis ics, R&D, ICT, medical and inancial se ices.
13 The au ho s would like o hank he AEZO o p o iding access o his
in o ma ion.
14 This me hod akes in o accoun he di e ences in he size o he SEZ a eas.
SEZs wi h la ge inne a eas can each g ea e dis ances.
Jou nal o De elopmen Economics 176 (2025) 103482
5
M.A. Abagna e al.
Fig. 2. The loca ion o DHS clus e s and SEZs in ou sample.
icini y o mul iple SEZs, he household is assigned o he SEZ ha was
es ablished i s . This ensu es ha he su ounding a eas o di e en
SEZs do no o e lap and ha households a e ea ed as soon as he
i s SEZ is es ablished in hei neighbo hood.
In ou baseline analysis, we include dis ance bands up o a max-
imum o 30 km and di e en ia e be ween households li ing inside
o no a he han 10 km om he bounda y o he SEZ inne a ea
(wi hin 10 km), households li ing 10 o 20 km and households li ing
20 o 30 km om he bounda y o he SEZ inne a ea. We op ed
agains using na owe dis ance bands (less han 10 km) o minimize
he andom measu emen e o a ising om he geog aphic masking
o DHS clus e coo dina es.15 The a ionale o no going beyond 30
km is o app oxima e commu ing zones wi hin which households a e
likely o be a ec ed by he p esence o SEZs. This app oach aligns
wi h he exis ing li e a u e demons a ing ha he e ec o place-based
policies apidly decays wi h dis ance (F ick e al., 2019). Fu he , in a
obus ness check whe e he maximum dis ance is ex ended o 120 km,
we show ha u he dis ance bands a e no ele an .
Ou inal sample comp ises epea ed c oss-sec ions o households
obse ed wi hin he 30-km- adius ci cles su ounding each SEZ es ab-
lished om 1990 o 2020. SEZs es ablished be o e he i s wa e o he
coun y’s DHS, i.e., ou side he sample pe iod, and households esiding
15 DHS andomly displaces he GPS la i ude/longi ude posi ions o all
su eys in o de o ensu e ha esponden con iden iali y is main ained. The
displacemen is andomly ca ied ou so ha : (i) u ban clus e s con ain a
minimum o 0 and a maximum o 2 kilome e s o e o ; (ii) u al clus e s
con ain a minimum o 0 and a maximum o 5 kilome e s o posi ional e o
wi h a u he andomly-selec ed 1% o he u al clus e s displaced a minimum
o 0 and a maximum o 10 kilome e s (h ps://dhsp og am.com/pubs/pd /
SAR7/SAR7.pd , accessed on 07.09.2023). This geog aphic masking in oduces
classical measu emen e o and should no bias ou es ima ion esul s.
nea hem a e comple ely excluded om he analysis. The es ima ion
sample encompasses a o al o app oxima ely 90,000 household-yea
obse a ions in h ee non-o e lapping dis ance bands a ound 51 unique
SEZs.16
3.4. Cha ac e is ics o SEZ loca ions
SEZs a e no andomly loca ed in space, and he e o e hei loca-
ions ha e dis inc i e cha ac e is ics e en be o e he es ablishmen o
he SEZs. A ailable in o ma ion sugges s ha A ican SEZs end o be
loca ed in populous, u banized places. Abou 80% o he SEZs in ou
sample a e no mo e han 10 km om a populous ci y and oughly a
hi d o hem a e loca ed nea a majo ai po o seapo (Appendix
Figu e A2).
This sec ion examines he cha ac e is ics o SEZ loca ions wi h
ega d o hei p e- ea men le els o de elopmen and u baniza ion.
Table 2 epo s p e- ea men means and s anda d de ia ions o he
household weal h index and h ee a iables cap u ing u baniza ion in
DHS clus e s alling in o he a ious dis ance bands. U ban esidence is
a bina y a iable, which is 1 o u ban and 0 o u al su ey clus e s.
The buil -up index quan i ies he p esence o buil s uc u es in a su ey
clus e on a scale o 0 o 1 and e e s o he yea 1990. Popula ion
densi y is he numbe o pe sons pe squa e kilome e in a su ey
clus e , e e ing o he yea 2000.17
The uppe panel o Table 2 shows ha loca ions chosen o hos
SEZs and hei 10 km immedia e neighbo hoods end o be iche ,
mo e u banized and mo e densely popula ed han he mo e dis an
neighbo hood al eady be o e he es ablishmen o he SEZs. The mean
o he weal h index o households li ing wi hin 10 km o SEZs is 0.78,
indica ing ha hei weal h is 0.78 s anda d de ia ion highe han he
coun y a e age. In con as , households in he 10–20 km band a e
abou he same a luen (−0.05), while households in he 20–30 km
band a e poo e (−0.14) han he coun y a e age. Simila dec easing
pa e ns a e obse ed o he deg ee o u baniza ion and popula ion
densi y. The SEZs a e hus loca ed close o u ban cen e s.
The middle and lowe panels o Table 2 illus a e how he a o e-
men ioned loca ion pa e ns a y by he da e o es ablishmen . We
dis inguish be ween "ea ly" SEZs, de ined as SEZs es ablished be o e
2015, and ‘‘la e’’ SEZs es ablished in 2015 o la e . We obse ed
ha SEZs se up ea lie a e assigned o la ge u ban a eas han SEZs
es ablished la e . The neighbo hoods o ea ly SEZs exhibi a highe
buil -up p esence and popula ion densi y, which decline less sha ply
wi h dis ance om he SEZs. This aligns wi h he obse a ion ha ea ly
SEZs a e, on a e age, close o majo ai po s han la e SEZs (Appendix
Table A3). The abo e cha ac e is ic is obus o using o he cu -o
yea s han 2015 o classi y SEZs in o ea ly and la e g oups. Mo eo e ,
he pa e n ha la e SEZs a e sys ema ically placed in less a o able
loca ions is in line wi h he li e a u e (e.g., Lu e al., 2019).
Guided by his desc ip i e e idence, we ollow an iden i ica ion
s a egy ha compa es households wi hin he same dis ance bands o
SEZs while aking in o accoun di e ences in he ini ial cha ac e is ics
o SEZ loca ions.
4. Empi ical amewo k
We de ine household exposu e o SEZs based on he geog aphical
p oximi y o an ac i e SEZ. Since ou sample consis s only o SEZs
and hei su ounding a eas, all loca ions (and hei households) a e
16 None o he SEZ su ounding a eas in ou baseline sample happen o
ex end o neighbo ing coun ies. This does no , howe e , apply o samples
ex ending beyond 30 km.
17 The u ban esidence a iable is sou ced om he DHS household su eys.
The buil -up index and popula ion densi y a e aken om he DHS Geospa ial
Co a ia e Da ase , h ps://spa ialda a.dhsp og am.com/co a ia es/.
Jou nal o De elopmen Economics 176 (2025) 103482
6
M.A. Abagna e al.
Table 2
P e- ea men loca ion cha ac e is ics o SEZs.
(wi hin 10 km) (10–20 km) (20–30 km)
N Mean S d. De . N Mean S d. De . N Mean S d. De .
All SEZ Loca ions
Weal h index 20,706 0.78 1.09 10,663 −0.05 0.91 12,773 −0.14 0.83
U ban esidence 20,706 0.76 0.43 10,663 0.29 0.45 12,773 0.21 0.41
Buil -up p esence index 20,441 0.21 0.24 10,368 0.07 0.17 12,447 0.09 0.25
Log popula ion densi y 20,531 7.05 1.66 10,628 6.00 1.30 12,738 5.69 1.43
Ea ly SEZ Loca ions (Es ablishmen be o e 2015)
Weal h index 5113 1.10 1.28 1989 0.50 1.20 3421 0.34 1.08
U ban esidence 5113 0.83 0.37 1989 0.54 0.50 3421 0.40 0.49
Buil -up p esence index 5113 0.42 0.33 1989 0.24 0.28 3421 0.28 0.40
Log popula ion densi y 5113 7.25 1.83 1989 6.44 1.84 3421 6.02 2.01
La e SEZ Loca ions (Es ablishmen in 2015 o la e )
Weal h index 15,593 0.68 1.00 8674 −0.18 0.78 9352 −0.31 0.63
U ban esidence 15,593 0.74 0.44 8674 0.23 0.42 9352 0.14 0.35
Buil -up p esence index 15,328 0.14 0.15 8379 0.03 0.09 9026 0.02 0.06
Log popula ion densi y 15,418 6.98 1.60 8639 5.90 1.12 9317 5.57 1.12
No e: Summa y s a is ics o p e- ea men household-yea obse a ions. The weal h index is speci ic o households, he o he h ee indica o s a e speci ic o DHS su ey clus e s.
U ban esidence is a bina y a iable aking alue 1 o u ban and 0 o u al loca ions. The buil -up p esence index anges on a scale o 0 o 1, wi h a highe alue indica ing
mo e buil s uc u es. Popula ion densi y is exp essed as log numbe o pe sons pe squa e kilome e s. The la e wo a iables e e o a single yea , 1990 and 2000 espec i ely.
1990 is a p e- ea men yea o all SEZs, 2000 o all bu 5 SEZs in ou sample. Wi h a cu -o yea o 2015, he e a e 22 ea ly and 29 la e SEZs in ou es ima ion sample.
e en ually exposed o an SEZ, albei a di e en poin s in ime. We
exploi his ime a ia ion in SEZ es ablishmen o es ima e he weal h
e ec o SEZs and ollow a s agge ed ea men di e ence-in-di e ences
app oach.
Ou empi ical model desc ibes he ela ionship be ween household
weal h and SEZs as ollows.
𝑊ℎ𝑧𝑐𝑡 =𝛽SEZ𝑧𝑡 +𝛾𝑋ℎ𝑧𝑐𝑡 +𝛼𝑧+𝛼𝑐𝑡 +𝜀ℎ𝑧𝑐𝑡 (1)
The weal h index o household ℎ li ing in he p oximi y o SEZ 𝑧 in
coun y 𝑐 and obse ed in yea 𝑡 depends on whe he he SEZ has
al eady been es ablished in 𝑡, which is cap u ed by he ime- a ying
bina y ea men a iable SEZ𝑧𝑡. The ea men a iable swi ches om
0 o 1 in he yea when he SEZ s a s o ope a e and emains 1
h oughou he sample pe iod.
In addi ion, Eq. (1) con ols o a se o household-speci ic a iables,
𝑋ℎ𝑧𝑐𝑡, as well as SEZ-speci ic and coun y-yea speci ic ixed e ec s,
𝛼𝑧 and 𝛼𝑐𝑡, espec i ely. The household-speci ic a iables a e indica-
o s o basic cha ac e is ics o households ha cap u e he changing
composi ion o households be ween su eys and ac oss dis ance bands.
They include bina y a iables o abo e-median household size, abo e-
median age o household head, and emale household head. Summa y
s a is ics o he household a iables a e epo ed in Appendix Table A4.
We chose no o con ol o he u ban s a us o he su ey clus e o he
household o a oid o e con olling o mechanisms ha may unde lie
he ea men e ec . Agglome a ion o ces d i en by SEZ policy o en
mani es as popula ion concen a ion in u ban clus e s nea SEZs, a
dynamic cap u ed by he ime- a ying u ban indica o . Ne e heless,
as shown in Appendix Table A9, ou main es ima e emains s a is ically
signi ican , al hough i s magni ude dec eases, when he u ban indica o
is included as a con ol.
The SEZ-speci ic ixed e ec s abso b di e ences be ween SEZ lo-
ca ions in hei ini ial cha ac e is ics. As shown in Sec ion 3.4, SEZs
es ablished la e a e sys ema ically placed in poo e loca ions han SEZs
es ablished ea lie . Because ou ea men a iable a ies by SEZ and
yea , he inclusion o SEZ ixed e ec ensu es ha he ea men e ec
is iden i ied om he ime a ia ion in households’ ela i e weal h
posi ion (compa ed o hei espec i e coun y a e age). Speci ically,
ou coe icien o in e es , 𝛽, cap u es how he ela i e weal h posi ion
o households nea an SEZ changes a e i s es ablishmen , compa ed
o simila households nea SEZs wi h di e en opening da es.
Eq. (1) also con ols o coun y-yea ixed e ec s, which pa ial ou
coun y-speci ic ends in he ela i e weal h posi ions o he sample
households as well as any e ec s speci ic o he su ey ounds. Al-
hough he weal h index as p o ided by he DHS p og am is pu ged
o coun y ends due o s anda diza ion, households in ou sample
may s ill exhibi ends in ela i e weal h independen o SEZ, jus i ying
he inclusion o coun y-yea e ec s. Howe e , we ind ha ou esul s
emain obus e en when con olling only o common yea e ec s.18
We es ima e Eq. (1) sepa a ely o each dis ance band wi h O dina y
Leas Squa es (OLS) and clus e s anda d e o s a he le el o he SEZ.19
Following on de Gol z and Ba nwal (2019), we es ima e wi hou DHS
sampling weigh s bu demons a e ha he esul s emain obus when
weigh s a e applied.
Ou empi ical app oach add esses he challenge o endogenous
ea men assignmen in wo ways. Fi s , ou es ima ion only conside s
loca ions ha e en ually become ea ed. Speci ically, i compa es he
weal h ajec o ies o ea lie e sus la e - ea ed households li ing in
he same dis ance band (wi hin 10 km, be ween 10–20 km o be ween
20–30 km) a ound hei espec i e SEZ. Second, he po en ial sys em-
a ic di e ences be ween ea lie and la e ea ed loca ions (and hei
households) a e add essed by he inclusion o SEZ-speci ic ixed e ec s,
which abso b any ime-cons an di e ences be ween SEZ loca ions.
Howe e , ou es ima es may s ill be biased i , in he absence o
SEZ, household weal h is ending di e en ly in ea ly han in la e-
ea ed loca ions.20 SEZ loca ions wi h a ying g ow h dynamics may
also a y in hei esponsi eness o he policy. To add ess his conce n,
we conduc an e en s udy analysis o con i m he absence o p e-
ea men ends and pe o m h ee ypes o obus ness checks. Fi s ,
we show ha ou baseline esul s a e obus o he inclusion o SEZ-
speci ic linea ends in he es ima ion. Second, we allow o di e en ial
ends based on he ini ial le el o de elopmen o he SEZ loca ions.
Thi d, we conduc placebo es s by shi ing SEZ es ablishmen da es
back o pe iods when no SEZs we e ac i e a he si es. A s a is ically
insigni ican 𝛽 es ima e in he placebo es s would u he alida e he
obus ness o ou indings.
18 These esul s a e a ailable upon eques .
19 To minimize he impac o measu emen e o , we exclude om he
es ima ion sample su ey clus e s wi h less han 15 households. Fu he , we
winso ize he lowe and uppe 1% o he weal h index dis ibu ion in e e y
su ey o elimina e po en ial ou lie s in ou dependen a iable.
20 Di e en ial ends ha pu ely a ise om di e ging coun y-speci ic ends
a e abso bed by coun y-yea e ec s.
Jou nal o De elopmen Economics 176 (2025) 103482
7
M.A. Abagna e al.
In he e en s udy analysis, we es ima e he ollowing e sion o Eq.
(1),
𝑊ℎ𝑧𝑐𝑡 =
−1
∑
𝑘=−𝐾
𝛽𝑘SEZ𝑝𝑟𝑒(𝑘)
𝑧𝑡 +
𝐿
∑
𝑘=1
𝛽𝑘SEZ𝑝𝑜𝑠𝑡(𝑘)
𝑧𝑡 +𝛾𝑋ℎ𝑧𝑐𝑡 +𝛼𝑧+𝛼𝑐𝑡 +𝜀ℎ𝑧𝑐𝑡,(2)
which es ima es ea men e ec s o a ious e en pe iods 𝑘 p eceding
and ollowing SEZ es ablishmen , wi h he coe icien o he base
pe iod 𝑘=0 se o ze o. Gi en he spa se na u e o ou da a bo h ac oss
coun ies and o e ime, we de ine e en pe iods as mul i-yea in e als
o ensu e a su icien numbe o independen e en s unde lying each
es ima e. We choose he pe iod immedia ely p eceding he yea o SEZ
es ablishmen ( ea men yea ) as he base pe iod. In de e mining he
leng h o he base pe iod, we accoun o he ac ha he es ablishmen
o SEZs is o en p eceded by planning and cons uc ion phases, which
can span om less han a yea o se e al yea s and po en ially in luence
local economic ac i i y p io o o mal implemen a ion. To in o m ou
choice, we ga he ed supplemen a y in o ma ion on he du a ion o his
‘‘cons uc ion’’ phase o he sample SEZs, inding ha i can las up o
six yea s.21 Consequen ly, we de ine he base pe iod as encompassing
he six yea s p eceding he ea men yea , i.e., yea s −6 o −1.
Fo ou main e en s udy, we u he di ide he imeline in o
wo p e- ea men pe iods and wo pos - ea men pe iods. Speci ically,
pe iod 𝑘 = −1 includes yea s −12 o −7, and pe iod 𝑘 = −2 cap u es
yea s −13 and ea lie . Fo he pos - ea men imeline, pe iod 𝑘 = 1
spans yea s 0 o 9, while pe iod 𝑘 = 2 co e s yea 10 and beyond, which
allow us o dis inguish be ween medium- e m and long- e m e ec s.
The selec ion o pe iod leng hs is p ima ily guided by da a a ail-
abili y, wi h he aim o achie ing a oughly balanced dis ibu ion o
obse a ions and SEZs ac oss e en pe iods.22 A pe sis en limi a ion,
howe e , is ha he longe - e m ea men e ec is iden i ied om a
ela i ely small numbe o ea ly-es ablished SEZs. Speci ically, only 12
SEZs con ibu e o he es ima ion o he long- e m coe icien unde he
baseline speci ica ion, inc easing o 17 when he wo pos - ea men
e en pe iods a e spli a yea 5 ins ead o yea 9. Ne e heless, he
o e all indings o he e en s udy emain quali a i ely consis en
ac oss al e na i e pos - ea men pe iod spli s.
Ou empi ical model (1) co esponds o a wo-way ixed-e ec s
(TWFE) model a he le el o he SEZ, e en hough he analysis is
conduc ed a he household le el. An eme ging li e a u e subjec s
he TWFE model in s agge ed ea men se ings unde sc u iny and
a gues ha , unless he ea men e ec is homogeneous ac oss ime
and uni s, he es ima ed a e age ea men e ec is biased (Ro h e al.,
2023).23 Se e al al e na i e es ima ion me hods ha e been p oposed
o accoun o his p oblem, bu only ew can accommoda e epea ed
c oss-sec ional da a. A u he challenge in applying some o hese
me hods in ou se ing is he spa si y o he DHS da a. We, he e-
o e, op o pe o m ou baseline es ima ion wi h OLS, which we
hen complemen wi h obus ness checks using he es ima ion me hod
21 Examples o SEZs wi h cons uc ion pe iods spanning mul iple yea s
include he Manga-Mungassa SEZ in Bei a, Mozambique, whe e cons uc ion
began in 2012, h ee yea s be o e i s o icial es ablishmen ; he Kano F ee
T ade Zone in Nige ia, whe e cons uc ion s a ed six yea s p io o i s
es ablishmen in 1998; and he Tema Expo P ocessing Zone in Ghana, which
unde wen a ou -yea cons uc ion phase om 1995 o 1999.
22 The pe iod de ini ions may be adjus ed in e en s udies conduc ed on
smalle samples, o ins ance, when analyzing al e na i e ou come a iables
o employing di e en es ima ion me hods.
23 In a s agge ed ea men se ing, he TWFE es imand is a weigh ed
a e age o he di e en wo-by- wo di e ence-in-di e ences es ima es com-
pa ing he indi idual g oups ea ed a di e en imes. The sou ce o he
iden i ica ion issue is ha some o hese compa isons a e inco ec , such as
using an ea lie - ea ed g oup as a con ol o a la e - ea ed g oup. See,
among o he s, Goodman-Bacon (2021), de Chaisema in and D’Haul œuille
(2020), Callaway and San ’Anna (2021), and Woold idge (2021).
o Callaway and San ’Anna (2021) and he Ex ended TWFE es ima ion
p oposed by Woold idge (2021).
Ano he app oach used in he li e a u e o deal wi h endogenei y
conce ns is o use p opensi y sco e ma ching. Zheng e al. (2017)
and Lu e al. (2019) compa e he a eas su ounding SEZs o he
ma ched a eas loca ed u he away. We implemen his s a egy as
a obus ness check complemen a y o ou baseline esul s. Thus, we
ma ch hose households loca ed wi hin 10 km o an SEZ o he house-
holds loca ed be ween 30 and 50 km. The a ionale behind choosing
his dis ance band is ha he con ol g oup should be loca ed u he
away o no be a ec ed by he spillo e e ec , bu no oo a away o
be compa able o he ea ed g oup. T ea men p obabili ies a e es i-
ma ed using co a ia e balancing p opensi y sco e (CBPS) due o Imai
and Ra ko ic (2014), which a e hen ans o med in o weigh s.24 Fi-
nally, we apply hese weigh s in he weigh ed e sion o ou baseline
eg ession.
Besides he a e age e ec o he policy, we a e also in e es ed in
he e ec s along he weal h dis ibu ion and examine how household
weal h a di e en quan iles o he local weal h dis ibu ion changes
ollowing he es ablishmen o an SEZ. Relying on empi ical model (1),
we pe o m he Quan ile T ea men E ec (QTE) es ima ion me hod
p oposed by Fi po (2007) and Fi po and Pin o (2016) o es ima e he
ea men e ec a each decile o he weal h dis ibu ion.25 The e a e
wo ca ea s conce ning QTE es ima ion in he cu en con ex . Fi s ly,
i only allows us o measu e changes in he weal h dis ibu ion a
speci ic deciles bu no changes in he weal h posi ion o indi idual
households a hese deciles. Secondly, he esul s e e o he local
dis ibu ion, i.e. he dis ibu ion o weal h in he speci ic dis ance band
o he SEZs, and no o he weal h dis ibu ion o he coun y as a
whole. Ne e heless, he analysis can be o policy signi icance because
i p o ides insigh in o how b oadly he bene i s o SEZs a e dis ibu ed
ac oss he local popula ion and wha ypes o households bene i he
mos .
5. Resul s
5.1. Main esul s
Resul s om es ima ing Eq. (1) o he mu ually exclusi e dis ance
bands – wi hin 10 km, 10–20 km, and 20–30 km – a e p esen ed in
Table 3. The es ima es indica e ha SEZ policy has a posi i e o e all
impac on he ela i e weal h posi ion o households. Following he
es ablishmen o an SEZ, he weal h o households li ing wi hin 10
km ises by 0.25 s anda d de ia ion o he weal h index (signi ican a
1%) ela i e o he coun y’s a e age. Con e ed in o eal e ms, his
inc ease is subs an ial, oughly equi alen , ce e is pa ibus, o owning
a compu e in Nige ia in 2008 o ha ing a inished loo made o
ce amic iles in Kenya in 2014.26 The weal h gain is also sizeable when
exp essed in mone a y e ms: in 2008, he a e age p ice o a pe sonal
compu e was USD 692, which was abou 60% o Nige ia’s annual
pe -capi a g oss na ional income (GNI) in ha yea .27
24 The main ad an age o CBPS is ha i exploi s he dual cha ac e is ics o
he p opensi y sco e by es ima ing condi ional ea men p obabili ies while a
he same ime op imizing co a ia e balance. This alle ia es he need o check
he balance o he co a ia es. T ea men p obabili ies a e ans o med in o
weigh s such ha he ea men g oup ecei es a weigh o 1
𝑃 𝑟(𝐷𝑖=1|𝑋0
𝑖) and he
con ol g oup is weigh ed by 1
1−𝑃 𝑟(𝐷𝑖=1|𝑋0
𝑖), whe e 𝑋0
𝑖 is he se o p e- ea men
co a ia es.
25 We use he S a a unc ion i hd eg, which is a wo-s ep es ima ion
p ocedu e elying on ecen e ed in luence unc ions and which allows o
high-dimensional ixed e ec s (Rios-A ila, 2020).
26 These compa isons a e d awn om de ailed in o ma ion on he con-
s uc ion o he weal h index in each DHS su ey, p o ided by he
Jou nal o De elopmen Economics 176 (2025) 103482
8
M.A. Abagna e al.
Table 6
Resul s o employmen and occupa ion wi hin 10 km.
(1) (2) (3)
Wo ked in he las 12 mon hs Occupa ion in ag icul u e Occupa ion in non-ag icul u e
Female Ou come
SEZ −0.0222 −0.0573∗∗∗ 0.0311
(0.0323) (0.0195) (0.0361)
R-squa ed 0.209 0.217 0.168
N 23,744 23,744 23,744
Numbe o clus e s 50 50 50
Husband/Pa ne Ou come
SEZ 0.0158∗−0.0196 0.0353
(0.0092) (0.0285) (0.0279)
R-squa ed 0.858 0.118 0.464
N 19,317 19,317 19,317
Numbe o clus e s 49 49 49
No e: Resul s om OLS es ima ion on indi idual-le el emale and male samples o e sions o Eq. (1). All eg essions include SEZ ixed-e ec s,
coun y-yea dummies, household-le el con ol a iables and emale-speci ic con ol a iables. S anda d e o s (in pa en heses) a e clus e ed a
he SEZs le el. * 𝑝 < 0.1, ** 𝑝 < 0.05, *** 𝑝 < 0.01.
se o emale-speci ic cha ac e is ics as con ol a iables.47 The main
esul s a e summa ized in Table 6, wi h es ima es o emales shown
in he uppe panel and males in he lowe panel. He e ogeneous e -
ec s based on skill le el and mig a ion backg ound a e p esen ed in
Appendix Table A11, whe e ha ing high skill is de ined as ha ing
comple ed seconda y educa ion.
O e all, we ind li le o no e idence ha SEZ es ablishmen in-
c eases he p obabili y o employmen among sampled indi iduals
wi hin a 10 km adius, as shown by he es ima es in Column (1) o
Table 6. The poin es ima e o emales is s a is ically insigni ican ,
while he es ima e o males is posi i e bu only ma ginally signi ican ,
e lec ing a modes 1.58 pe cen age-poin inc ease (signi ican a he
10% le el). This inding la gely pe sis s when he es ima es a e dis-
agg ega ed by skill o mig a ion s a us, as shown in Column (1) o
Appendix Table A11.
Howe e , he abo e indings should be in e p e ed wi h cau ion
due o he selec i e na u e o he indi idual-le el sample. No ably, he
sample includes only emales who a e mo he s, po en ially o e looking
employmen e ec s on o he demog aphic g oups, such as young single
women. P io esea ch highligh s ha his g oup o en cons i u es key
bene icia ies o SEZ- ela ed job c ea ion in de eloping coun ies (Glick
and Roubaud, 2006; B usse ich, 2024).
A mo e obus inding eme ges ega ding sec o al shi s among
emale household membe s. Speci ically, SEZ es ablishmen is asso-
cia ed wi h a signi ican mo e away om ag icul u al occupa ions,
as e idenced by he nega i e and signi ican es ima e in Column (2)
o Table 6 and he e en s udy g aph in Appendix Figu e A9. The
es ima e indica es a 6 pe cen age-poin dec ease in he p obabili y o
emales wo king in ag icul u e ollowing SEZ es ablishmen . Decom-
posing his es ima e by educa ional a ainmen e eals ha he e ec
is s a is ically signi ican only among unskilled emales, unde sco ing
he limi ed ole o ag icul u e in skilled employmen . Addi ionally,
he e ogeneous es ima es by mig a ion backg ound indica e ha his
sec o al shi is obse ed among bo h ne e -mo e and mig an emales,
sugges ing ha he e ec is no solely d i en by he in-mig a ion o
non-ag icul u al wo ke s.
Fo unskilled males, a simila sec o al shi appea s o occu , ma ked
by a signi ican decline in ag icul u al occupa ions a he 5% le el,
alongside a signi ican inc ease in non-ag icul u al employmen a he
1% le el (Columns (2) and (3) in Appendix Table A11). Howe e , e en
s udy analyses o hese ou comes sugges ha hese ends may ha e
al eady been p esen among unskilled males p io o SEZ es ablishmen ,
47 These include age, age squa ed, ma i al s a us, p egnancy s a us, and he
numbe o child en unde 5 yea s old. Howe e , he esul s emain obus e en
wi hou he inclusion o indi idual-speci ic con ols.
aising ques ions abou he ex en o which he obse ed shi can be
a ibu ed o SEZs.48
Taken oge he , hese indings p o ide no s ong e idence ha SEZ
es ablishmen leads o ne job c ea ion in ou sample o mo he s and
hei male pa ne s. Ne e heless, he esul s highligh an impo an
sec o al shi away om ag icul u al employmen , which may signal
imp o emen s in bo h pay and job quali y as households ansi ion o
non-ag icul u al occupa ions.
6.3. Educa ional implica ions
Acco ding o Lu e al. (2023), SEZs can in luence educa ional ou -
comes h ough h ee p ima y channels. The i s is he income channel,
whe eby inc eased household income om SEZ- ela ed oppo uni ies
may encou age pa en s o in es mo e in hei child en’s educa ion.
The second is he job oppo uni y channel, sugges ing ha access o
low-skill jobs may educe high school en ollmen , while he p ospec
o high-skill jobs could incen i ize con inued educa ion. The hi d is
he wage p emium channel, which posi s ha SEZs c ea e highe wage
p emiums o high-skill jobs, mo i a ing indi iduals o pu sue u he
educa ion.
Educa ion decisions depend on weighing he long- e m bene i s
o schooling agains i s sho - e m cos s. SEZs ocused on low-skill
indus ies may educe incen i es o pu sue u he educa ion, as indi-
iduals a e d awn o immedia e employmen oppo uni ies. Con e sely,
echnology-o ien ed SEZs o en gene a e highe e u ns o educa ed
wo ke s, encou aging longe s ays in educa ion. Fo example, Os e
and S einbe g (2013) ind ha he es ablishmen o an In o ma ion
Technology Cen e in India led o a 5% inc ease in p ima y school
en ollmen .
To assess whe he he weal h gains associa ed wi h SEZ es ab-
lishmen ansla e in o highe educa ional a ainmen among nea by
esiden s, we examine he educa ional ou comes o emale household
membe s and hei male pa ne s li ing wi hin 10 km o SEZs. Ou
p ima y ou come a iable o educa ional a ainmen is a bina y indi-
ca o o comple ing a leas seconda y educa ion. Desc ip i e s a is ics
e eal ha emales in ou sample a e, on a e age, less likely han hei
male pa ne s o ha e comple ed seconda y educa ion (37% s. 48%,
as shown in Appendix Table A6).
As de ailed in Sec ion 5.3, SEZ es ablishmen o en induces in-
mig a ion, wi h ecen ly a i ed households ha ing, on a e age, a
mo e highly educa ed head han na i e households (Appendix Table
A5). To disen angle he e ec s o skilled in-mig a ion om changes
48 The espec i e e en s udy esul s a e a ailable upon eques .
Jou nal o De elopmen Economics 176 (2025) 103482
15

M.A. Abagna e al.
Table 7
Female educa ional a ainmen wi hin 10 km.
(1) (2) (3)
Dep. a : A leas seconda y educa ion
SEZ 0.0848∗∗∗
(0.0254)
SEZ × Ne e -mo e 0.0789∗∗
(0.0314)
SEZ × Mig an 0.0896∗∗∗
(0.0300)
SEZ × Ne e -mo e × Age < 25 0.1717∗∗∗
(0.0329)
SEZ × Ne e -mo e × Age ≥ 25 0.0491
(0.0352)
SEZ × Mig an × Age < 25 0.1344∗∗∗
(0.0310)
SEZ × Mig an × Age ≥ 25 0.0717∗
(0.0362)
R-squa ed 0.199 0.201 0.202
N 18,222 18,222 18,222
Numbe o clus e s 46 46 46
No e: Resul s om OLS es ima ion on he indi idual-le el emale subsample, whe e
mig a ion in o ma ion is obse ed. Ve sions o Eq. (1) in e ac ed wi h ne e -mo e
s a us and a bina y indica o o being less han 25 yea s old. All eg essions include
SEZ ixed-e ec s, coun y-yea dummies, household-le el con ol a iables and emale-
speci ic con ol a iables. S anda d e o s (in pa en heses) a e clus e ed a he SEZs
le el. * 𝑝 < 0.1, ** 𝑝 < 0.05, *** 𝑝 < 0.01.
in locals’ human capi al accumula ion, we ocus ou analysis on he
DHS subsample whe e mig a ion in o ma ion is a ailable and es ima e
sepa a e ea men e ec s o indi iduals classi ied as ne e -mo e s and
mig an s. Females and hei male pa ne s a e de ined as ne e -mo e s
i hey belong o ne e -mo e households.49 All o he indi iduals a e
classi ied as mig an s.
Since younge indi iduals a e mo e likely o adjus hei educa ional
decisions in esponse o policy changes, we u he es ima e ea men
e ec s sepa a ely o young and old age g oups. Gende -speci ic cu o
ages (25 o emales and 30 o males) a e used, e lec ing he en-
yea a e age age gap be ween male pa ne s and emales (Appendix
Table A6) and o a oid small sample sizes in he young male subg oup.
No ably, he male esul s emain obus when using smalle cu o
ages.50
The es ima ion esul s a e p esen ed in Tables 7and 8 o emales
and hei husbands/pa ne s, espec i ely. Complemen a y esul s o
he yea s o educa ion o husbands/pa ne s (an ou come a iable no
a ailable o emales) a e p o ided in Appendix Table A12. We ind ha
SEZ es ablishmen is associa ed wi h a s a is ically signi ican imp o e-
men in he educa ional a ainmen o bo h emales and males esiding
wi hin 10 km o an SEZ. The coe icien es ima es in Column (1)
indica e inc eases in he p obabili y o comple ing seconda y educa ion
by 8.5 pe cen age poin s o emales and 7.4 pe cen age poin s o
males, bo h signi ican a he 1% le el.
These esul s emain consis en in magni ude and signi icance o
indi iduals in bo h ne e -mo e and mig an households (Column (2)),
which sugges s ha , beyond he e ec s o skilled in-mig a ion, he e is
e idence o human capi al accumula ion among he local popula ion.
When he coe icien s a e u he disagg ega ed by age g oup in
Column (3) o Table 7, he esul s indica e ha he posi i e e ec
obse ed o ne e -mo e emales is d i en p edominan ly by younge
emales unde he age o 25, while he coe icien o olde ne e -
mo e emales is no s a is ically signi ican . In con as , o mig an
49 Recall ha a household is classi ied as a ne e -mo e i all i s in e iewed
emale membe s esponded ‘‘always’’ o he ques ion abou how long hey had
li ed con inuously in hei cu en place o esidence.
50 These obus ness es ima es a e a ailable upon eques .
Table 8
Husband/pa ne educa ional a ainmen wi hin 10 km.
(1) (2) (3)
Dep. a : A leas seconda y educa ion
SEZ 0.0743∗∗∗
(0.0213)
SEZ × Ne e -mo e 0.0892∗∗∗
(0.0312)
SEZ × Mig an 0.0685∗∗
(0.0261)
SEZ × Ne e -mo e × Age < 30 0.1349∗∗
(0.0562)
SEZ × Ne e -mo e × Age ≥ 30 0.1006∗∗∗
(0.0337)
SEZ × Mig an × Age < 30 0.0946∗∗
(0.0417)
SEZ × Mig an × Age ≥ 30 0.0823∗∗∗
(0.0291)
R-squa ed 0.180 0.182 0.187
N 14,923 14,923 13,517
Numbe o clus e s 46 46 46
No e: Resul s om OLS es ima ion on he indi idual-le el husband/pa ne subsample,
whe e mig a ion in o ma ion is obse ed. Ve sions o Eq. (1) in e ac ed wi h ne e -
mo e s a us and a bina y indica o o being less han 30 yea s old. All eg essions
include SEZ ixed-e ec s, coun y-yea dummies, household-le el con ol a iables and
emale-speci ic con ol a iables. S anda d e o s (in pa en heses) a e clus e ed a he
SEZs le el. * 𝑝 < 0.1, ** 𝑝 < 0.05, *** 𝑝 < 0.01.
emales, we obse e a ma ginally signi ican posi i e coe icien o
he olde age g oup as well. This aligns wi h ou expec a ion ha
educa ional imp o emen s in olde coho s a e p ima ily a ibu able
o in-mig a ion.
In con as , he decomposed es ima es o males in Column (3) o
Table 8 a e posi i e and s a is ically signi ican ac oss all age g oups
and mig a ion backg ounds, including olde males in ne e -mo e
households. This aises conce ns abou po en ial upwa d selec ion bias
in he male esul s, as mo e highly educa ed young emales may p e -
e en ially selec highly educa ed (and pe haps olde ) males as pa ne s.
Consequen ly, we conside he indings o male pa ne s less sui able
o isola ing he e ec s o he human capi al accumula ion channel.51
O e all, he indings on educa ion p o ide e idence ha SEZ poli-
cies in A ica con ibu e o imp o ing educa ional a ainmen in SEZ
neighbo hoods. These imp o emen s a e d i en no only by skilled
in-mig a ion bu also by enabling and incen i izing local esiden s –
pa icula ly young emales – o achie e highe le els o educa ion.
7. Conclusion
Many coun ies ac oss A ica ha e implemen ed a ied SEZs o e
he las wo decades o p omo e expo di e si ica ion, GVC pa icipa-
ion, and local economic de elopmen by using incen i e packages o
a ac i ms in o he desi ed loca ion. Despi e he wide sp ead o SEZs
in he con inen , he e is limi ed empi ical e idence on he economic
implica ions o he policy, pa icula ly a he mic o le el, la gely due
o da a una ailabili y.
Using geocoded DHS da a, we demons a e ha he es ablishmen o
SEZs in A ica con ibu es o he g ow h o asse weal h o households
li ing wi hin 10 km o he SEZs. While he bene i s o SEZs a e b oadly
dis ibu ed ac oss households, he p ima y bene icia ies appea o be
51 Reg essions using he household head’s comple ion o seconda y edu-
ca ion as he ou come a iable, ega dless o whe he he head is male o
emale, p oduce esul s simila o hose obse ed in he male eg essions.
These esul s a e a ailable upon eques . Howe e , an addi ional selec ion
issue may a ec hese es ima es: in labo ma ke s whe e quali ica ions a e
mo e highly ewa ded, highly educa ed indi iduals a e be e posi ioned o
become household heads.
Jou nal o De elopmen Economics 176 (2025) 103482
16
M.A. Abagna e al.
membe s o he (lowe ) middle class. We u he obse e ha inc eases
in household asse weal h a e associa ed wi h inc eased access o house-
hold u ili ies, g ea e consump ion o du able goods, imp o ed dwelling
quali y, highe educa ional le els o household membe s, and a shi
away om ag icul u al employmen . Impo an ly, he weal h gains
a e e iden among bo h na i e and immig an households, coun e ing
conce ns ha place-based policies, such as SEZs, disp opo iona ely ail
o bene i local popula ions. Collec i ely, he indings sugges ha SEZ
policies ac as a ca alys o u baniza ion in he a ge a eas.
Ne e heless, ou analysis has wo limi a ions ha call o cau ious
policy in e p e a ion and sugges di ec ions o u u e esea ch. The
i s s ems om he non- andom placemen o SEZs, which in A ica
a e ypically loca ed nea u ban cen e s. As a esul , ou indings
may no necessa ily apply o SEZs es ablished in less de eloped u al
a eas. Likewise, ou dis ibu ional esul s may no cap u e he poo es
households in he coun ies, as hey a e mo e likely o eside in emo e
u al loca ions.
Second, he employmen da a a ailable om DHS p ima ily ocus
on a speci ic demog aphic g oup, namely emales o ep oduc i e age
wi h child en and hei male pa ne s, po en ially o e looking b oade
employmen e ec s ac oss o he popula ion segmen s. Mo eo e , he
absence o ne employmen gains among his demog aphic g oup o
emales con as s wi h p io li e a u e, which iden i ies young emales
as key bene icia ies o SEZ policies in de eloping coun ies. This dis-
c epancy aises impo an ques ions abou he ole o mo he hood as a
po en ial cons ain on emale pa icipa ion in SEZ-d i en employmen
oppo uni ies.
Finally, ou esul s highligh he policy implica ion ha SEZ policies
can se e as e ec i e ools o policymake s in de eloping coun ies o
s imula e u baniza ion and imp o e he wel a e o esiden s in a ge ed
loca ions. This is pa icula ly ele an o A ica, whe e signi ican
egional dispa i ies in e ms o economic oppo uni ies pe sis . By
add essing hese inequali ies, SEZs ha e he po en ial o con ibu e o
mo e inclusi e economic g ow h ac oss he con inen .
CRediT au ho ship con ibu ion s a emen
Ma hew Amali inga Abagna: W i ing – e iew & edi ing,
W i ing – o iginal d a , Resou ces, In es iga ion, Concep ualiza ion.
Cecília Ho nok: W i ing – e iew & edi ing, W i ing – o iginal
d a , Visualiza ion, Supe ision, Me hodology, In es iga ion, Fo mal
analysis, Da a cu a ion, Concep ualiza ion. Alina Mulyuko a:
W i ing – e iew & edi ing, W i ing – o iginal d a , Visualiza ion,
Me hodology, Fo mal analysis, Da a cu a ion, Concep ualiza ion.
Decla a ion o compe ing in e es
The au ho s decla e ha hey ha e no known compe ing inancial
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 .
Appendix A. Supplemen a y ma e ial
Appendix Figu es A1–A9 and Tables A1–A12 can be ound online a
h ps://doi.o g/10.1016/j.jde eco.2025.103482.
Da a a ailabili y
The au ho s do no ha e pe mission o sha e da a.
Re e ences
Abebe ese, A.B., 2017. Elec ici y cos and i m pe o mance: E idence om India. Re .
Econ. S a . 99 (5), 839–852.
Adunbi, O., 2019. (Re)in en ing de elopmen : China, in as uc u e, sus ainabili y and
special economic zones in Nige ia. A ica 89 (4), 662–679.
AEZO, 2021. A ican Economic Zones Ou look. AEZO.
A DB, 2011. The middle o he py amid: Dynamics o he middle class in A ica. A .
De . Bank Ma k. B ie .
Agga wal, A., Kokko, A., 2022. SEZs and po e y educ ion: E idence om Andh a
P adesh, India. In . J. Eme g. Ma k. 17 (8), 1793–1814.
Alkon, M., 2018. Do special economic zones induce de elopmen al spillo e s? E idence
om India’s s a es. Wo ld De . 107, 396–409.
A kin, D., 2016. Endogenous skill acquisi ion and expo manu ac u ing in Mexico. Am.
Econ. Re . 106 (8), 2046–2085.
Bellema e, M.F., Wichman, C.J., 2020. Elas ici ies and he in e se hype bolic sine
ans o ma ion. Ox . Bull. Econ. S a . 82 (1), 50–61.
Blakeslee, D., Chau ey, R., Fishman, R., Malik, S., 2022. Land ezoning and s uc u al
ans o ma ion in u al India: E idence om he indus ial a eas p og am. Wo ld
Bank Econ. Re . 36 (2), 488–513.
Blou i, Y., Eh lich, M.V., 2020. On he op imal design o place-based policies: A
s uc u al e alua ion o EU egional ans e s. J. In . Econ. 125, 103319.
B usse ich, M., 2024. The socioeconomic impac o special economic zones: E idence
om Cambodia. Wo ld Econ. 47 (1), 362–387.
Busso, M., G ego y, J., Kline, P., 2013. Assessing he incidence and e iciency o a
p ominen place based policy. Am. Econ. Re . 103 (2), 897–947.
Callaway, B., San ’Anna, P.H.C., 2021. Di e ence-in-di e ences wi h mul iple ime
pe iods. J. Econome ics 225 (2), 200–230.
Chau ey, R., 2017. Loca ion-based ax incen i es: E idence om India. J. Public Econ.
156, 101–120.
C iscuolo, C., Ma in, R., O e man, H.G., Van Reenen, J., 2019. Some causal e ec s o
an indus ial policy. Am. Econ. Re . 109 (1), 48–85.
de Chaisema in, C., D’Haul œuille, X., 2020. Two-way ixed e ec s es ima o s wi h
he e ogeneous ea men e ec s. Am. Econ. Re . 110 (9), 2964–2996.
Du an on, G., Venables, A.J., 2021. Place-based policies: P inciples and de eloping
coun y applica ions. In: Fische , M.M., Nijkamp, P. (Eds.), Handbook o Regional
Science. Sp inge Be lin Heidelbe g, Be lin, Heidelbe g, pp. 1009–1030.
Eh lich, M. ., O e man, H.G., 2020. Place-based policies and spa ial dispa i ies ac oss
Eu opean ci ies. J. Econ. Pe spec . 34 (3), 128–149.
Fa ole, T., Akinci, G., 2011. Special Economic Zones. P og ess, Eme ging Challenges,
and Fu u e Di ec ions. Technical Repo , The In e na ional Bank o Recons uc ion
and De elopmen /The Wo ld Bank.
Fa ole, T., Mobe g, L., 2017. Special economic zones in A ica: Poli ical economy
challenges and solu ions. P . Ind. Policy: Go e nmen - Bus. Coo d. A . Eas Asia
234–254.
Fi po, S., 2007. E icien semipa ame ic es ima ion o quan ile ea men e ec s.
Econome ica 75 (1), 259–276.
Fi po, S., Pin o, C., 2016. Iden i ica ion and es ima ion o dis ibu ional impac s o
in e en ions using changes in inequali y measu es. J. Appl. Econome ics 31,
457–486.
F ick, S.A., Rod íguez-Pose, A., Wong, M.D., 2019. Towa d economically dynamic
special economic zones in eme ging coun ies. Econ. Geog . 95 (1), 30–64.
Gibbons, S., Lyy ikäinen, T., O e man, H.G., Sanchis-Gua ne , R., 2019. New oad
in as uc u e: he e ec s on i ms. J. U ban Econ. 110, 35–50.
Gibson, J., Oli ia, S., Boe-Gibson, G., Li, C., 2021. Which nigh ligh s da a should we
use in economics, and whe e? J. De . Econ. 149, 102602.
Glaese , E.L., Go lieb, J.D., 2008. The Economics o Place-making Policies. Technical
Repo , Na ional Bu eau o Economic Resea ch.
Glick, P., Roubaud, F., 2006. Expo p ocessing zone expansion in Madagasca : Wha
a e he labou ma ke and gende impac s?. J. A . Econ. 15 (4), 722–756.
Gobillon, L., Magnac, T., Selod, H., 2012. Do unemployed wo ke s bene i om
en e p ise zones? The F ench expe ience. J. Public Econ. 96 (9–10), 881–892.
Goodman-Bacon, A., 2021. Di e ence-in-di e ences wi h a ia ion in ea men iming.
J. Econome ics 225 (2), 254–277.
Gö g, H., Mulyuko a, A., 2024. Place-based policies and i m pe o mance: E idence
om special economic zones in India. Eu . Econ. Re . 165, 104752.
Ham, J.C., Swenson, C., İm oho oğlu, A., Song, H., 2011. Go e nmen p og ams
can imp o e local labo ma ke s: E idence om s a e en e p ise zones, ede al
empowe men zones and ede al en e p ise communi y. J. Public Econ. 95 (7–8),
779–797.
Hasan, R., Jiang, Y., Ra ols, R.M., 2021. Place-based p e e en ial ax policy and
indus ial de elopmen : E idence om India’s p og am on indus ially backwa d
dis ic s. J. De . Econ. 150, 102621.
Ho nbeck, R., Mo e i, E., 2024. Es ima ing who bene i s om p oduc i i y g ow h:
local and dis an e ec s o ci y p oduc i i y g ow h on wages, en s, and inequali y.
Re . Econ. S a . 106 (3), 587–607.
Hu, Y., Yao, J., 2022. Illumina ing economic g ow h. J. Econome ics 228 (2), 359–378.
Hua, Y., Pa idge, M., Sun, W., 2023. Pollu ion e ec s o place-based policy: E idence
om China’s de elopmen -zone p og am. J. Reg. Sci. 63 (3), 703–727.
Jou nal o De elopmen Economics 176 (2025) 103482
17
M.A. Abagna e al.
Imai, K., Ra ko ic, M., 2014. Co a ia e balancing p opensi y sco e. J. R. S a . Soc. Se .
B S a . Me hodol. 76 (1), 243–263.
Kline, P., 2010. Place based policies, he e ogenei y, and agglome a ion. Am. Econ. Re .
100 (2), 383–387.
Kline, P., Mo e i, E., 2014a. Local economic de elopmen , agglome a ion economies,
and he big push: 100 yea s o e idence om he Tennessee alley au ho i y. Q.
J. Econ. 129 (1), 275–331.
Kline, P., Mo e i, E., 2014b. People, places, and public policy: Some simple wel a e
economics o local economic de elopmen p og ams. Annu. Re . Econ. 6 (1),
629–662.
Kos e , H.R., Van Omme en, J., 2019. Place-based policies and he housing ma ke .
Re . Econ. S a . 101 (3), 400–414.
K an z, S., 2023. The Kiel Ins i u e A ica Moni o . Kiel Ins i u e o he Wo ld
Economy.
Le ien, M., 2013. Regimes o dispossession: F om s eel owns o special economic zones:
Regimes o dispossession. De . Chang. 44 (2), 381–407.
Li, X., Zhou, Y., Zhao, M., Zhao, X., 2020. Ha moniza ion o DMSP and VIIRS nigh ime
ligh da a om 1992–2021 a he global scale. Figsha e. Sci. Da a 7, 168.
Lowes, S., Mon e o, E., 2021. Concessions, iolence, and indi ec ule: E idence om
he congo ee s a e. Q. J. Econ. 136 (4), 2047–2091.
Lu, F., Sun, W., Wu, J., 2023. Special economic zones and human capi al in es men :
30 yea s o e idence om China. Am. Econ. Jou nal: Econ. Policy 15 (3), 35–64.
Lu, Y., Wang, J., Zhu, L., 2019. Place-based policies, c ea ion, and agglome a ion
economies: E idence om China’s economic zone p og am. Am. Econ. Jou nal:
Econ. Policy 11 (3), 325–360.
Mo e i, E., 2010. Local mul iplie s. Am. Econ. Re . 100 (2), 373–377.
Mo e i, E., 2024. Place-based policies and geog aphical inequali ies. Ox . Open Econ.
3 (Supplemen _1), i625–i633.
Neuma k, D., Kolko, J., 2010. Do en e p ise zones c ea e jobs? E idence om
Cali o nia’s en e p ise zone p og am. J. U ban Econ. 68 (1), 1–19.
Neuma k, D., Simpson, H., 2015. Place-based policies. In: Handbook o Regional and
U ban Economics, ol. 5, Else ie , pp. 1197–1287.
Obeng, F.A., W igley-Asan e, C., Teye, J.K., 2015. Wo king condi ions in Ghana’s expo
p ocessing zone and women’s empowe men . Wo k. O gan. Labou Glob. 9 (2).
Os e , E., S einbe g, B.M., 2013. Do IT se ice cen e s p omo e school en ollmen ?
E idence om India. J. De . Econ. 104, 123–135.
O chia, C., Asongu, S., 2021. Indus ial g ow h in sub-Saha an A ica: e idence om
machine lea ning wi h insigh s om nigh ligh sa elli e images. J. Econ. S ud. 48
(8), 1421–1441, Publishe : Eme ald Publishing Limi ed.
Phi i, M., Manchishi, S., UNU-WIDER, 2020. Special Economic Zones in Sou he n A ica:
Whi e Elephan s o La en D i e s o G ow h and Employmen ? The case o Zambia
and Sou h A ica, hund ed and six ee h ed. In: WIDER Wo king Pape , ol. 2020,
UNU-WIDER.
Pica elli, N., 2016. Who eally bene i s om expo p ocessing zones? E idence om
Nica aguan municipali ies. Labou Econ. 41, 318–332.
Reynolds, C.L., Rohlin, S.M., 2015. The e ec s o loca ion-based ax policies on he
dis ibu ion o household income: E idence om he ede al empowe men zone
p og am. J. U ban Econ. 88, 1–15.
Rios-A ila, F., 2020. Recen e ed in luence unc ions (RIFs) in S a a: RIF eg ession and
RIF decomposi ion. S a a J. 20 (1), 51–94.
Rod íguez-Pose, A., Ba alucci, F., F ick, S.A., San os-Paulino, A.U., Bolwijn, R., 2022.
The challenge o de eloping special economic zones in A ica: E idence and lessons
lea n . Reg. Sci. Policy & P . 14 (2), 456–481.
Ro h, J., San ’Anna, P.H.C., Bilinski, A., Poe, J., 2023. Wha ’s ending in di e ence-
in-di e ences? A syn hesis o he ecen econome ics li e a u e. J. Econome ics
235 (2), 2218–2244.
San ’Anna, P.H., Zhao, J., 2020. Doubly obus di e ence-in-di e ences es ima o s. J.
Econome ics 219 (1), 101–122.
Schminke, A., Van Bieseb oeck, J., 2013. Using expo ma ke pe o mance o e alua e
egional p e e en ial policies in China. Re . Wo ld Econ. 149, 343–367.
Shenoy, A., 2018. Regional de elopmen h ough place-based policies: E idence om
a spa ial discon inui y. J. De . Econ. 130, 173–189.
Shimeles, A., Ncube, M., 2015. The making o he middle-class in A ica: E idence om
DHS da a. J. De . S ud. 51 (2), 178–193.
S eenbe gen, V., Ja o cik, B., 2017. Analyzing he impac o he Kigali special economic
zone on i m beha io . In . G ow h Cen . Policy Wo k. Pap.
Thompson, L., 2019. Al e na i e sou h–sou h de elopmen collabo a ion? he ole o
China in he coega special economic zone in sou h A ica. Public Adm. De . 39
(4–5), 193–202.
Tus ing, L.S., Bo omley, C., Gibson, H., Kleinschmid , I., Ta em, A.J., Lindsay, S.W.,
Ge hing, P.W., 2017. Housing imp o emen s and Mala ia isk in Sub-Saha an
A ica: A mul i-coun y analysis o su ey da a. PLoS Med. 14 (2), e1002234.
UNCTAD, 2019. Wo ld In es men Repo 2019: Special Economic Zones. UN.
UNCTAD, 2021. Handbook on Special Economic Zones in A ica. Technical Repo , p.
221.
UNECA, 2022. Ha nessing he Po en ial o Special Economic Zones o P i a e Sec-
o De elopmen and Inclusi e Indus ializa ion in Sou he n A ica. Economic
Commission o A ica.
on de Gol z, J., Ba nwal, P., 2019. Mines: The local weal h and heal h e ec s o
mine al mining in de eloping coun ies. J. De . Econ. 139, 1–16.
Wang, J., 2013. The economic impac o special economic zones: E idence om Chinese
municipali ies. J. De . Econ. 101, 133–147.
Woold idge, J.M., 2021. Two-way ixed e ec s, he wo-way mundlak eg ession, and
di e ence-in-di e ences es ima o s.
Xu, J., Wang, X., 2020. Re e sing uncon olled and unp o i able u ban expansion in
A ica h ough special economic zones: An e alua ion o E hiopian and Zambian
cases. Sus ainabili y 12 (21), 9246.
Zeng, D.Z., 2015. Global expe iences wi h special economic zones: Focus on China and
A ica. Wo ld Bank Policy Res. Wo k. Pap. (7240).
Zhao, C., Qu, X., 2024. Place-based policies, u al employmen , and in a-household
esou ces alloca ion: E idence om China’s economic zones. J. De . Econ. 167,
103210.
Zheng, S., Sun, W., Wu, J., Kahn, M.E., 2017. The bi h o edge ci ies in China:
Measu ing he e ec s o indus ial pa ks policy. J. U ban Econ. 100, 80–103.
Jou nal o De elopmen Economics 176 (2025) 103482
18