Structural change and gender sectoral segregation in sub‐Saharan African countries

Zuazu, Izaskun
A icle — Published Ve sion
S uc u al change and gende sec o al seg ega ion in sub‐
Saha an A ican coun ies
Jou nal o In e na ional De elopmen
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Zuazu, Izaskun (2024) : S uc u al change and gende sec o al seg ega ion
in sub‐Saha an A ican coun ies, Jou nal o In e na ional De elopmen , ISSN 1099-1328, Wiley,
Hoboken, NJ, Vol. 36, Iss. 6, pp. 2626-2654,
h ps://doi.o g/10.1002/jid.3925
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/306121
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/
RESEARCH ARTICLE
S uc u al change and gende sec o al seg ega ion
in sub-Saha an A ican coun ies
Izaskun Zuazu
Ins i u e o Socio-Economics, Uni e si y o
Duisbu g-Essen, Duisbu g, Ge many
Co espondence
Izaskun Zuazu, Ins i u e o Socio-Economics,
Uni e si y o Duisbu g-Essen, Duisbu g,
Ge many.
Email: izaskun.zuazu-be me[email p o ec ed]
Funding in o ma ion
Le y Economics Ins i u e; INET-YSI
Abs ac
S uc u al change has long been a he co e o economic
de elopmen deba es. Howe e , he gende implica ions o
s uc u al change a e s ill la gely unexplo ed. This pape
helps o ill his gap by analysing he ole o s uc u al
change in he gende dis ibu ion o sec o al employmen in
sub-Saha an A ican coun ies. I employ agg ega e and dis-
agg ega e measu es o gende sec o al seg ega ion in
employmen , which measu e he di e ence be ween he
gende dis ibu ion ac oss sec o s wi h espec o he o e -
all pa icipa ion o women and men in he labou ma ke . I
build a panel da abase consis ing o 10 sec o s and 11 coun-
ies du ing 1960–2010. Fixed e ec s and ins umen al a -
iables' eg ession models show a signi ican , nonlinea link
be ween labou p oduc i i y and gende seg ega ion.
Inc easing labou p oduc i i y dep esses gende seg ega ion
a ini ial phases o s uc u al change. Howe e , u he p o-
duc i i y gains beyond a ce ain h eshold o sec o al de el-
opmen inc eases gende seg ega ion. Coun y-indus y
panel da a models complemen he analysis showing ha
ela i e labou p oduc i i y has a nonlinea impac in gende
seg ega ion: Ini ial inc eases in ela i e p oduc i i y
inc eases eminiza ion bu u he ela i e p oduc i i y gains
os e he masculiniza ion o sec o s. The es ima es sugges
ha manu ac u ing, u ili ies, cons uc ion, business, and
go e nmen se ices a e key o co ec gende biases in
employmen along he p ocess o s uc u al change.
Recei ed: 13 Sep embe 2023 Re ised: 12 Ma ch 2024 Accep ed: 8 May 2024
DOI: 10.1002/jid.3925
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and
ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed.
© 2024 The Au ho (s). Jou nal o In e na ional De elopmen published by John Wiley & Sons L d.
2626 J. In . De . 2024;36:2626–2654.
wileyonlinelib a y.com/jou nal/jid
KEYWORDS
associa ion index, dissimila i y index, gende sec o al seg ega ion,
ins umen al a iables, s uc u al change
JEL CLASSIFICATION
E0, J1, Q5
1|INTRODUCTION
S uc u al change is he p ocess o shi ing p oduc ion om ag icul u e o manu ac u ing and se ice sec o s,
ollowed by a decline in manu ac u ing sha e and an inc ease in se ice sec o sha e in he o al economy. The p om-
inen deba es on economic de elopmen in sub-Saha an A ican (SSA) coun ies a e conce ned wi h s uc u al
change pa e ns ha depa om he canonical model ha depic s declining ag icul u e, hump-shaped manu ac u -
ing, and ising high-p oduc i e se ices (de V ies e al., 2015; T egenna, 2015). S uc u al he e ogenei y— he isola-
ion o highly p oduc i e ac i i ies om he es o he economy—and p ema u e deindus ializa ion—a p omp shi
in o a se ice economy wi hou p ope de elopmen o he indus ial sec o —a e among he pa hological phenomena
iden i ied in he li e a u e (Rod ik, 2016; T egenna, 2016). While he canonical wo ks on his opic (Kuzne s, 1966;
Lewis, 1965) ha e been complemen ed by esea ch on he impac s o gene al s uc u al change on economic de el-
opmen (de V ies e al., 2021; He endo e al., 2014; McMillan e al., 2017; McMillan & Rod ik, 2011), he gende
implica ions o such a ans o ma ion a e less unde s ood (Dinkelman & Ngai, 2022; Go lieb e al., 2022; Seguino &
We e, 2014).
This pape adds o he li e a u e in s uc u al change by analysing whe he and how labou p oduc i i y and gen-
de seg ega ion a e linked in sub-Saha an A ican coun ies. Gende sec o al seg ega ion e e s o he sec o al dis i-
bu ion o emale and male employmen , ha is, he p opo ion o women (men) in each sec o ela i e o he o al
emale (male) employmen in he economy. Gende sec o al/ho izon al seg ega ion di e s om occupa ional/ e ical
seg ega ion: The o me conside s he sec o al economic s uc u e, whe eas he la e conside s he occupa ional
economic s uc u e o a gi en coun y o egion.
1
In his pape , we look a how key phenomena ega ding s uc u al
change, such as labou p oduc i i y and sec o al p oduc i i y gaps, a ec he gende dis ibu ion o sec o s in bo h
o mal and in o mal employmen .
The key a gumen o his pape is ha labou p oduc i i y—compu ed as he a io be ween alue added and
employmen —migh ha e a nonlinea ela ionship o gende sec o al seg ega ion. Ini ial p oduc i i y gains de i ed
om he p ocess o s uc u al change can imply a lowe demand o physical equi emen s. As lowe physical
equi emen s a e ound o inc ease he demand o women in he paid wo k o ce (Rendall, 2013,2017), one migh
expec an inc easing pa icipa ion o women in all sec o s o he economy. Howe e , u he p oduc i i y gains
abo e ce ain le els can couple wi h gende s e eo ypes and disc imina ion o de e he en ance o women in spe-
ci ic sec o s, hus os e ing he c owding o emale employmen in o he speci ied sec o s (Be gmann, 1981;
Seguino & B auns ein, 2019).
This pape empi ically es s he link be ween labou p oduc i i y and gende seg ega ion using panel da a
models a bo h coun y le el and coun y-indus y le els. The e m “coun y-indus y”is employed he e o di e en-
ia e be ween seg ega ion measu es a coun y le el (which a y acco ding o coun ies and yea s) and indus y-le el
measu es o gende -sec o al seg ega ion (which a y acco ding o sec o s, coun ies, and yea s). These wo measu e-
men s e e o he same phenomenon, namely, gende -sec o al seg ega ion. I collec da a on sec o al ( o mal and
in o mal) employmen , disagg ega ed by gende , and sec o al alue added om he A ica Sec o Da abase (ASD) by
de V ies e al. (2015), and build a panel da abase consis ing o 10 indus ies ope a ing in 11 coun ies du ing 1960–
2010. Using his da abase allows a highe le el o da a disagg ega ion han p e ious ela ed wo ks (Bo owman &
Klasen, 2020). Desc ip i ely, I show ha gende seg ega ion has inc eased in ce ain coun ies (i.e., Senegal, E hiopia,
ZUAZU 2627
and Bo swana), bu i was educed in o he s (i.e., Zambia, Sou h A ica) du ing he pe iod conside ed. A he same
ime, I iden i y ha hose coun ies wi h educed gende seg ega ion had, a he same ime, highe le els o labou
p oduc i i y.
I me ge he ASD da abase wi h in o ma ion on emale labou o ce pa icipa ion and o he coun y-le el
co a ia es ha can play a ole in gende seg ega ion. As a p e iew o he econome ic analysis, I ind a nonlinea ela-
ionship be ween labou p oduc i i y and gende seg ega ion: p oduc i i y gains dep ess seg ega ion up o a ce ain
h eshold. Beyond ha h eshold, u he p oduc i i y gains inc ease gende seg ega ion by sec o s. This esul is
obus o al e na i e es ima ion echniques, such as ins umen al a iables, ha ci cum en endogenei y issues
ega ding he inclusion o emale labou o ce pa icipa ion in he se o independen a iables. Addi ionally, he main
esul holds when using al e na i e dependen a iables, such as agg ega e and disagg ega e measu es o gende sec-
o al seg ega ion, namely, he Dissimila i y index o Duncan and Duncan (1955), he so-called IP index o Ka mel and
MacLachlan (1988), and he Associa ion index o Cha les and G usky (1995).
The emainde o he pape goes as ollows. Sec ion 2 e iews he li e a u e. Sec ion 3p o ides he da a and
he measu emen s o agg ega e gende sec o al seg ega ion. Sec ion 4speci ies he econome ic models while
Sec ion 5shows he esul s and p o ides di e en obus ness checks. Sec ion 6concludes and discusses policy
implica ions.
2|LITERATURE REVIEW
S uc u al change a la ge, and he di e ences in agg ega e and sec o al labou p oduc i i y in pa icula , in e ac in
complex ways wi h gende ed labou ma ke s. These s uc u al changes come wi h p o ound demog aphic mo e-
men s and u baniza ion and allow o echnological di usion a bo h ma ke and home p oduc ion le els (Bose up
e al., 2013; Dinkelman & Ngai, 2022; McMillan e al., 2017; Ube i & Doua in, 2022). Fundamen ally, hese changes
migh ans o m he p e-exis ing gende dis ibu ion o paid and unpaid wo k. This heo e ical sec ion i s analyses
he gene al pa e ns o s uc u al change a la ge and con ex ualizes hem wi hin de eloping egions, speci ically
sub-Saha an A ican coun ies. Second, his sec ion e iews he s ylized ac s on he gende implica ions o s uc u al
change, oge he wi h he speci ic empi ically in o med ac o s o gende seg ega ion. Finally, he sec ion zooms in
on speci ic implica ions o s uc u al change in gende sec o al seg ega ion in he egion.
S uc u al ans o ma ion a ec s p oduc i i y and g ow h: as esou ce ealloca ions shi labou ou o low-
p oduc i e sec o s owa ds high-p oduc i e, mode n economic ac i i ies, he e is an agg ega e labou p oduc i i y
ise and an expansion o income (McMillan & Rod ik, 2011; Na eed & Ahmad, 2016; Van A k, 1995). As sugges ed
by McMillan and Rod ik (2011), when labou and o he esou ces mo e om less p oduc i e o mo e p oduc i e sec-
o s, o e all p oduc i i y g ows e en in he absence o wi hin-sec o p oduc i i y g ow h. The p ocess o s uc u al
change shows g ea c oss-coun y he e ogenei y, as some coun ies ansi ion as e om one phase o
s uc u al change o ano he , o ail o ully de elop a mode n manu ac u ing sec o be o e mo ing in o a se ice
economy (He endo e al., 2014; Rod ik, 2016). Thus, i is impo an o conside economy-wide labou p oduc i i y
and sec o al gaps in labou p oduc i i y as key phenomena ela ed o he p ocess o s uc u al change.
De eloping coun ies show la ge sec o al gaps in labou p oduc i i y and dual economies, wi h high sha es o
low-p oduc i i y, la gely u al ac i i ies and low sha es o high-p oduc i e (u ban) ac i i ies. Fo he speci ic case
o he sub-Saha an A ican egion, c oss-sec o al labou ealloca ion was p oduc i i y educing du ing 1990s bu i
p omo ed p oduc i i y g ow h in he 2000s, and his imp o ed pe o mance in he 2000s (McMillan e al., 2014).
While he exis ing li e a u e has iden i ied he ele ance o s uc u al change in he p oduc i i y le els, p oduc i i y
g ow h, and sec o al gaps in p oduc i i y in SSA, i s impac on labou ma ke ou comes a e a less well known
(Mensah e al., 2023), le alone he gende labou ma ke implica ions o s uc u al change.
In he sample o SSA coun ies he e conside ed, manu ac u ing expanded g ea ly om 1960 o 1975,
co esponding o he shi s om subsis ence ag icul u al socie ies owa ds mode n manu ac u ing (de V ies
2628 ZUAZU
e al., 2015). Howe e , a e 1970, he egion su e ed a poli ical and economic u moil, which coincided wi h s uc-
u al adjus men p og ammes. These p og ammes had c ucial, gende ed implica ions, which a e s ill el oday
(Elson, 1995). In he la e phases o he pe iod conside ed in his pape (1990–2010), SSA coun ies expanded
employmen sha es in se ice sec o whe eas manu ac u ing sha es emained low, which is coined as p ema u e
deindus ializa ion (Rod ik, 2016) o e en, deindus ializa ion wi hou indus ializa ion (T egenna, 2016).
Mensah e al. (2023) p o ide an analysis o he e ec s o s uc u al change in labou p oduc i i y g ow h in a se
o 18 SSA coun ies du ing 1960–2018, expanding he coun y and ime co e ages o he ASD, o iginally de eloped
by he G oningen G ow h and De elopmen Cen e (GGDC) by de V ies e al. (2015). Thei indings using his
upda ed da abase sugges ha s uc u al change in SSA has been unde es ima ed in p e ious s udies. Ag icul u al
employmen declined du ing he pe iod conside ed, leading o an inc ease o low-p oduc i e se ice sec o s. Thus,
hese changes had limi ed enhancing e ec s in agg ega e labou p oduc i i y. Mensah e al. (2023) poin a ypical
labou ma ke ins i u ions se ing o he SSA egion, such as he signi ican a ia ion in he deg ee o en o cemen o
laws and egula ion, as ba ie s o success ul expe iences o s uc u al change. The au ho s also highligh he link
be ween s uc u al change and labou mobili y, wi h labou shi s de ining he ype o s uc u al p ocess. S uc u al
change migh inc ease agg ega e labou p oduc i i y, bu simul aneously, i migh also os e empo a y unemploy-
men and wo ke ealloca ions ising low-paid jobs o in o mal employmen and employee unce ain y. Howe e ,
Mensah e al. (2023) do no conside gende dispa i ies in hese labou ma ke implica ions o p ocess o s uc u al
change. Hence, he cu en pape p o ides a di ec con ibu ion o hei wo k in pa icula and o he li e a u e on
s uc u al change a la ge.
The p ocess o s uc u al change can lead o domes ic dispa i ies, such as inc easing income inequali y
(Kuzne s, 1966; Lewis, 1965) and gende edis ibu ion o paid wo k and unpaid household p oduc ion (Dinkelman &
Ngai, 2022; Gaddis & Klasen, 2014; Ube i & Doua in, 2022). In his con ex , s uc u al change is linked o he eme -
gence o new ypes o paid wo k oppo uni ies o women (Dinkelman & Ngai, 2022). No wi hs anding some b oad
simila i ies in emale and male employmen shi s o e he s uc u al change pa h, he li e a u e iden i ies signi ican
gende dispa i ies. Fo ins ance, Dinkelman and Ngai (2022) use his o ical c oss-coun y da a o de eloped econo-
mies o ind ha women lea e he ag icul u e sec o and mo e in o he se ice sec o as e han men do and ha
manu ac u ing ises mo e s eeply o men han o women. Howe e , Dinkelman and Ngai's (2022) pape only
ocuses on a a he b oad sec o al pe spec i e. In he cu en pape , I complemen hei analysis by p o iding a mo e
nuanced sec o al analysis ha uses a g ea e le el o da a disagg ega ion by sec o . This allows me o iden i y how
s uc u al change is linked o gende dispa i ies as well as which sec o s a e d i ing hese dispa i ies. Economic
de elopmen and sec o al composi ion ha e a p o ound impac on he women's dis ibu ion o paid and unpaid wo k,
as p edic ed by he so-called eminiza ion U-shape. In a nu shell, his heo y was i s unco e ed using his o ical da a
o he Uni ed S a es in Goldin (1995), sugges ing ha ini ial inc easing le els o economic de elopmen a e associ-
a ed wi h dep essing women in he paid wo k o ce due o an income e ec . Fu he inc eases o economic de elop-
men a e go e ned by a subs i u ion e ec ha pushes women back o he labou ma ke .
A g owing body o esea ch ollowed up he U-shaped eminiza ion hypo hesis o Goldin (1995) and cas s
doub s in he ex e nal alidi y o he hypo hesis. Gaddis and Klasen (2014) sugges ha s uc u al change should be
included in ou unde s anding o he U-shaped co ela ion be ween economic de elopmen and emale labou ma -
ke pa icipa ion. They also ensu e ha he nonlinea link is inconsis en depending on he da a and quan i a i e
me hod employed. Ube i and Doua in (2022) ind ha he use o he plough ma e s o he eminiza ion U-shape,
as physical equi emen s o he plough can media e he bulk o women in paid and unpaid wo k. In ela ed wo ks,
Rendall (2013,2017) conside s he ole o s uc u al change in al e ing he composi ion o “b ain”and “b awn” asks
by sec o s: Lowe physical equi emen s migh lead o inc easing oppo uni ies o paid wo k o women, as women
ha e a compa a i e ad an age in b ain jobs. Beyond emale labou o ce pa icipa ion, wo king condi ions o emale
employmen can be dispa a e and di e subs an ially om hose o male employmen .
Ex an li e a u e has also ocused on he ole o s uc u al change in he gende wage gap in SSA coun ies. Van
den B oeck e al. (2023) use decomposi ion me hods o analyse how s uc u al change a ec ed gende wage
ZUAZU 2629

di e en ials in Malawi, Tanzania, and Nige ia o ind ha s uc u al ans o ma ion does no consis en ly help b idge
he gende pay gap. Addi ionally, hei analysis sugges s a u al–u ban di ide in he d i ing o ces behind gende pay
inequali y; while in u al a eas occupa ion is he mos ele an ac o , in u ban a eas bo h occupa ion and sec o a e
simila ly impo an . In his sense, he cu en pape complemen s he exis ing li e a u e by placing special a en ion
on he ole o sec o al seg ega ion in he p ocess o s uc u al change o a ec di e en ly he li elihoods o women
and men, con olling a he same ime o u baniza ion.
Gende seg ega ion in SSA coun ies is lowe in compa ison o o he egions in he wo ld. Bo owman and
Klasen (2020) use a da abase o 69 coun ies, o which 24 a e SSA coun ies, and ind ha gende seg ega ion is gen-
e ally lowe in his egion whe e women and men a e disp opo iona ely employed in ag icul u e. While hey ind a
limi ed and mos ly insigni ican ole o s uc u al change in gende seg ega ion, hei es ima es associa e emale
labou o ce pa icipa ion wi h lowe gende -sec o al seg ega ion.
2
O he ecen wo ks in he gende implica ions o
s uc u al change ind ha highe indus ial p oduc i i y is associa ed wi h lowe p esence o women in good jobs,
namely, indus y jobs in 15 La in Ame ican coun ies (A o a e al., 2023) using panel da a econome ic models du ing
1990–2018.
Yeboah e al. (2022) analyse he ole o emale labou o ce pa icipa ion in s uc u al change, bu hey do no
ocus on he implica ions o gende seg ega ion, bu a he , he ex en o which ising emale labou o ce pa icipa-
ion in SSA coun ies is linked o alue-added sha es in he ag icul u e, indus y, and se ice sec o s. Using dynamic
panel da a models on a balanced da ase o 33 SSA coun ies du ing 1990–2017, hey ind ha ising emale labou
o ce pa icipa ion leads o an inc eased sha e o se ices in o al alue added, bu hey do no ind a signi ican ole
in indus y o ag icul u e sec o s. This is a media ing e ec o in as uc u e—measu ed in e ms o ei he ixed ele-
phone subsc ip ions o g oss- ixed-capi al o ma ion as p opo ion o GDP—which magni ies his posi i e link
be ween women in he paid wo k o ce and he sha e o se ices.
The cu en pape d aws on he abo e-men ioned s ylized ac s o s uc u al change and gende o specula e
he ex en o which p oduc i i y ma e s o gende seg ega ion by sec o . This pape comes close o he wo ks o
Bo owman and Klasen (2020) in iden i ying he d i e s o gende -sec o al seg ega ion and combines i wi h he
sec o al-disagg ega ed pe spec i e in de V ies e al. (2015). Fu he , he cu en pape complemen s he a gumen in
Rendall (2013) in asse ing ha highe labou p oduc i i y can a ou emale employmen by conside ing ha p o-
duc i i y migh ha e a nonlinea ela ionship wi h emale employmen in ce ain sec o s. A su icien ly high le els o
labou p oduc i i y, u he gains can block he en ance o women media ed by gende disc imina ion and s e eo-
ypes in he compe i ion be ween women and men o newly c ea ed “good”jobs (Seguino & B auns ein, 2019).
3|EMPIRICAL ANALYSIS
3.1 |Da a
I collec da a om he ASD p oduced by de V ies e al. (2015). This da abase p o ides sec o al-le el disagg ega ed
da a (ISIC Re . 3.1) on alue added, employmen , and emale sha e o employmen o 10 sec o s ope a ing in
11 SSA coun ies, namely, Bo swana, E hiopia, Ghana, Kenya, Malawi, Mau i ius, Nige ia, Senegal, Sou h A ica,
Tanzania, and Zambia, du ing 1960 o 2010.
3
Using his da abase, I look a wo undamen al phenomena o s uc-
u al change, as discussed in Sec ion 2: agg ega e labou p oduc i i y and ela i e labou p oduc i i y. The ASD p o-
ides in o ma ion on o mal and in o mal employmen , as i de ines employmen as “all pe sons engaged,” hus
including all paid employees and sel -employed and amily wo ke s o 15 yea s and olde . An impo an ea u e o
he ASD is ha i p o ides sec o al pu chasing powe pa i ies (PPPs) o he yea 2005 (He endo e al., 2022). I
con e labou p oduc i i y le els measu ed in domes ic p ices o compa able measu es o labou p oduc i i y le els
measu ed in in e na ional p ices using he sec o al-le el PPPs.
2630 ZUAZU
Figu e 1shows he sec o al dis ibu ion o emale employmen and male employmen oge he wi h sec o al-
ela i e labou p oduc i i y.
4
Ag icul u e concen a es he bulk o bo h emale and male employmen , wi h espec-
i ely 65% and 60% sha es. To he con a y, he ag icul u al- ela i e labou p oduc i i y le el is a ound 0.5 du ing
he pe iod, meaning ha labou p oduc i i y in he sec o is hal ha o he o al economy. The ade se ices sec o
comp ises a highe pe cen age o emale employmen (12%) han male employmen (8%), and i s ela i e p oduc i i y
le el is sligh ly abo e he o al economy p oduc i i y le el. The pe sonal se ices sec o also employs a g ea e p o-
po ion o emales o males, al hough he labou p oduc i i y is lowe han he a e age (0.8). The anspo se ices,
cons uc ion, and mining sec o s concen a e low sha es o emale employmen , while main aining among he
highes labou -p oduc i i y sha es. The manu ac u ing sec o concen a es a simila pe cen age o emale and male
employmen , while i s p oduc i i y le els a e a ound wo imes ha o he o al economy. The sec o al pe spec i e
aken in Figu e 1should be complemen ed wi h coun y-le el measu es o seg ega ion ha accoun o di e ences
in s uc u al change pa e ns. To do so, he nex subsec ion p oposes he use o s anda d coun y-le el measu es o
seg ega ion ha a e compu ed based on c oss-coun y, ime-se ies, and sec o al-le el disagg ega ed da a.
3.2 |Measu ing gende seg ega ion
To measu e gende seg ega ion, I employ he Duncan index o dissimila i y (ID), which is a s anda d measu e o
ei he e ical o ho izon al seg ega ion (Cha les & G usky, 2005; Duncan & Duncan, 1955) a coun y le el. Fo he
pu poses o his pape , I ocus on sec o al seg ega ion, and use a 10-sec o le el o indus ial classi ica ion (ISIC Re .
3.1) (see Equa ion 1).
ID ¼1=2X
n
i¼1
Fi
FMi
M

100 ð1Þ
FIGURE 1 Sec o al sha es o employmen by gende and ela i e labou p oduc i i y. Sou ce: Own elabo a ion
using he ASD (de V ies e al., 2015).
ZUAZU 2631
i¼1,…,n½
F
i
is he numbe o women in sec o i, whe e Fis he o al o women employed in he economy, M
i
is he numbe
o men in sec o i,Mis he o al o men employed in he economy, and nequals 10 ( o al numbe o sec o s, as he
da abase he e employed he e a e 10 sec o s i). One o he bene i s o using he ID as a measu e o seg ega ion is i s
simple in e p e a ion: I p o ides he pe cen age o women who need o change sec o s in o de o b ing abou a
gende -equal dis ibu ion ac oss sec o s wi hin an economy. The highe he alue o he ID, he highe he seg ega-
ion in a coun y. In he sample, ID has inc eased om 13% in 1960 o 20% in 2010. I should be no ed ha his mea-
su e o gende seg ega ion compa es he sec o al dis ibu ion o male employmen wi h ha o women.
None heless, he ID is no exemp om limi a ions. As no ed in Bo owman and Klasen (2020), he o emos limi a-
ion o he ID is i s mechanical sensi i i y o c oss-coun y and empo al changes in he employmen sha e by sec o .
The index is hus in luenced by la ge sec o s, which can be wo isome in s udying s uc u al ans o ma ion and gen-
de seg ega ion in a panel-da a se ing. To alle ia e his di icul y, I employ he so-called “IP index”(IP) (Ka mel &
MacLachlan, 1988) as an al e na i e measu e o seg ega ion, which se es also as a obus ness check o he econo-
me ic models below. Figu e A2 in Appendix Ashows he e olu ion o ID and IP.
5
A common limi a ion o he ID and
IP is ha bo h depend g ea ly on he b ead h o he sec o al classi ica ion. Na ow classi ica ions yield highe le els
o ID han b oad classi ica ions; hence, hey can be manipula ed o o e highe o lowe le els o seg ega ion
(Nelson, 2017).
Table 1p o ides he e olu ion o ID in each coun y in he sample, oge he wi h ha o he loga i hm o o al
labou -p oduc i i y le els. In six ou o he 11 coun ies (Bo swana, E hiopia, Malawi Nige ia, Senegal, and Tanzania),
gende seg ega ion has inc eased o e he pe iod a sc u iny. The mos ema kable inc ease is shown by Senegal,
whe e ID inc eased om 8% in 1970 o 23% in 2010.
A he same ime, he g ow h a e o o al labou p oduc i i y be ween 1970 and 2010 is he lowes in he
sample (1.6%). To he con a y, gende seg ega ion dec eased ema kably in Zambia (38 pe cen age poin s, p.p.)
and, oge he wi h Ghana, is he coun y wi h he highes g ow h a e o o al labou p oduc i i y in he
pe iod (10%).
Figu e 2documen s co ela ions be ween ei he agg ega e labou p oduc i i y (log) o emale labou o ce pa -
icipa ion (FLFP, %) and gende seg ega ion (measu ed by ID). Figu e 2a ends o show ha a ini ial le els o labou
p oduc i i y, gende seg ega ion educes. Howe e , his cu e is nonlinea , as i sligh ly la ens a highe le els o
p oduc i i y. The co ela ion o agg ega e labou p oduc i i y wi h gende sec o al seg ega ion seems o a y by
coun y, whe e Zambia shows he g ea es change in bo h seg ega ion and p oduc i i y. Ghana and Tanzania seem
o ollow a clea e nonlinea pa e n be ween p oduc i i y and seg ega ion, whe eas Sou h A ica and Nige ia show
an in e ed pa e n. The s agna ion o FLFP in SSA coun ies sugges ed in Backhaus and Loichinge (2022) can be
seen in Figu e 2b, as he e is sca ce a ia ion o FLFP by coun y. The co ela ion be ween women in he paid wo k-
o ce, p oxied using FLFP (%) in x-axis, and gende sec o al seg ega ion, p oxied by ID (%) in y-axis, is a linea and
nega i e. Thus, we can conside p elimina y e idence o a nega i e, s ong associa ion be ween ising women in he
labou ma ke and he c owing o women in speci ic sec o s o he economy, as p e iously ound in Bo owman and
Klasen (2020).
The e a e limi s o wha can be disce ned om agg ega e c oss-coun y analyses in he con ex o s uc u al
change and gende ed impac s (Wamboye & Seguino, 2015). Bo h ID and IP indices a e coun y-le el measu es o
gende seg ega ion. While hey p o ide in o ma ion on he sha e o wo ke s who should change sec o s o inc ease
a gende -balanced dis ibu ion ac oss sec o s, hese indices a e no able o iden i y which p ecise sec o s should be
de eminized o demasculinized. To sol e o his, in his pape , I combine he use o coun y-le el measu es o seg e-
ga ion ( iz., ID and IP) wi h a measu e o he concen a ion o gende employmen , namely, he Associa ion Index
(A index) p oposed by Cha les and G usky (1995). The A index akes a log-linea app oach o ci cum en he limi a-
ions o bo h ID and IP indices, sol ing he e o e o he mechanical dependence o he la e on a ia ions in sec-
o al shi s o employmen and pa icipa ion o women in he labou ma ke . Addi ionally, he A iden i ies which
2632 ZUAZU
TABLE 1 Gende seg ega ion and labou p oduc i i y le els in sub-Saha an A ica.
Dissimila i y index (%) To al labou p oduc i i y
1970 1980 1990 2000 2010 Change (p.p.) 1970 1980 1990 2000 2010 Change (%)
Bo swana 8.8 18.5 25.1 21.9 14.1 5.3 3.5 5.7 7.3 8.4 9.5 5.9
E hiopia 6.3 5.7 5.5 7.3 16.0 9.7 4.4 4.7 4.7 5.4 6.9 2.4
Ghana 27.5 28.1 24.5 14.8 26.5 1.0 4.1 2.0 1.5 4.0 6.0 10.1
Kenya 22.0 23.1 20.1 20.3 19.9 2.1 6.1 7.2 8.1 8.9 9.6 3.5
Malawi 17.2 17.2 17.3 16.5 17.7 0.5 3.1 4.1 5.3 8.1 9.5 6.4
Mau i ius 27.4 26.2 29.4 31.0 25.0 2.4 6.4 7.8 9.0 10.0 10.8 4.4
Nige ia 28.5 28.4 23.7 20.1 30.2 1.7 3.9 5.4 7.1 9.7 11.1 7.2
Senegal 7.7 14.0 20.2 19.9 22.5 14.9 10.2 10.6 10.9 11.4 11.8 1.6
Sou h A ica 34.0 33.9 35.0 28.5 27.5 6.4 5.1 6.6 7.8 8.7 9.7 4.7
Tanzania 11.4 13.9 10.5 9.7 13.8 2.4 6.1 7.2 9.2 11.0 12.0 5.9
Zambia 49.2 35.3 23.2 14.6 11.6 37.7 4.6 5.6 8.6 12.7 14.6 10.0
Sou ce: Own elabo a ion using he ASD (de V ies e al., 2015).
ZUAZU 2633
FIGURE 4 Ma ginal e ec s o labou p oduc i i y in gende sec o al seg ega ion. Sou ce: Based on eg ession
esul s o model in column 3 in Table 2.
TABLE 3 To al labou p oduc i i y le els and gende seg ega ion (ins umen al a iables).
Dependen a iable:
(1) (2) (3) (4)
ID IP ID 5 yea s IP 5 yea s
Agg ega e labou p oduc i i y (log) 0.349*** 0.356*** 0.736*** 0.756***
(0.025) (0.024) (0.190) (0.195)
Agg ega e labou p oduc i i y (log) sq. 0.024*** 0.023*** 0.050*** 0.050***
(0.002) (0.001) (0.009) (0.009)
FLFP 0.154*** 0.154*** 0.297*** 0.303***
(0.008) (0.008) (0.056) (0.058)
Full se o con ols Yes Yes Yes Yes
Time ixed-e ec s Yes Yes Yes Yes
No. o obse a ions 60 60 46 46
No. o g oups 7 7 7 7
Wi hin R
2
0.971 0.972 0.962 0.961
Fi s s age Fs a 185.37 185.37 135.46 135.46
Unde iden i ica ion es (Kleibe gen–Paap k LM s a is ic) 18.500 18.500 32.659 32.659
p alue 0.000 0.000 0.000 0.000
Weak iden i ica ion es (Kleibe gen–Paap k Wald F
s a is ic)
26.589 26.589 18.666 18.666
No e: D iscoll and K aay s anda d e o s in pa en heses. All independen a iables one pe iod lagged (columns 1 and 2) o
5-yea pe iod lagged (column 3 and 4). Dissimila i y index and IP index in logs. Kleibe gen–Paap LM es null hypo hesis is
ha he ank condi ion ails in second s age equa ion, ha is, unde iden i ied (Kleibe gen & Paap, 2006). Kleibe gen–Paap k
Wald Fc i ical alues a ying be ween 5.53 and 16.38 (S ock & Yogo, 2005). Coe icien on he ins umen in he i s s age
was nega i e (0.68) and s a is ically signi ican a he 0.01 le el o columns 1 and 2 and 0.32 a 0.05 le el o he 5-yea
o e lapping o columns 3 and 4. Coun ies in sample: Bo swana, Ghana, Kenya, Nige ia, Senegal, Sou h A ica, and Zambia.
*p< 0.1.**p< 0.05.***p< 0.01.
2640 ZUAZU

simila esul s. I should be no ed ha , when using he ins umen al a iables app oach, he sample o coun ies is
educed as he e is no in o ma ion on he ins umen (size o he household) o Mau i ius and Tanzania.
Columns 3 and 4 eplica e he models using augmen ed, o e lapping i e-pe iod lags o he independen a iables.
The esul s emain simila : agg ega e labou p oduc i i y is associa ed wi h a nega i e e ec in seg ega ion, bu
i s quad a ic e m is associa ed wi h a seg ega ion-enhancing e ec . Rising women in he paid economy, measu ed
by FLFP, is associa ed wi h a nega i e coe icien ; hus, highe pa icipa ion o women in he labou ma ke is
linked o educing gende sec o al seg ega ion. The magni udes o hese coe icien s a e la ge when using 5-yea
lags models.
I pe o m pos es ima ion es s o he alidi y and weakness o he ins umen . Conc e ely, I conduc he
Kleibe gen–Paap es o he unde iden i ica ion o he i s -s age eg ession because he models a e compu ed
using he e oscedas ici y- obus s anda d e o s. Fu he , he Kleibe gen–Paap es is obus agains iola ions o he
i.i.d. assump ion (Kleibe gen & Paap, 2006). We ejec he null hypo hesis o unde iden i ica ion a he 1% le el o
signi icance. As o he weakness o he ins umen s, he Kleibe gen–Paap Fs a is ic is highe han he s anda d
ule-o - humb o 10 and is beyond he c i ical alues a any le el o accep able bias.
To del e deepe on his nonlinea ela ionship be ween agg ega e labou p oduc i i y and gende sec o al seg e-
ga ion, I eplica e he model in column 1 (Table 3) using pa i ions o he da abase on he basis o low-p oduc i i y
and high-p oduc i i y coun y-le el alues. The c i ical alue o agg ega e labou p oduc i i y ha se s he h eshold
o a posi i e link be ween p oduc i i y gains and gende seg ega ion is p o ided in Table A2 in Appendix A. This
alue is also shown in Figu e 5, which se s his alue a o 7 o he log- ans o med agg ega e labou p oduc i i y.
Using he subsample o low-p oduc i e obse a ions (a coun y-yea le el), he es ima es sugges ha ini ial gains
o agg ega e labou p oduc i i y inc ease he dissimila i y index ( he p oxy o gende sec o al seg ega ion), and
u he inc eases a e associa ed o lowe seg ega ion, hence, an in e ed-U-shape link be ween p oduc i i y and
seg ega ion. To he con a y, using he pa i ion o he high-p oduc i e obse a ions, he es ima es show again he
U-shaped nonlinea i y be ween agg ega e labou p oduc i i y and gende seg ega ion by indus ies. These pa i ions
p o ide u he le e age o he main indings o he pape (see he es ima es o his pa i ions in Table A3 in
Appendix A).
9
FIGURE 5 Ma ginal e ec s o agg ega e labou p oduc i i y in gende sec o al seg ega ion (IV es ima es). Sou ce:
Based on eg ession esul s o model in column 1 in Table 3.
ZUAZU 2641
5.3 |Coun y-indus y-le el panel da a es ima es
Aic ¼β0þβ1RLPic, 1þβ2RLP2
ic, 1þβ3FLFPc, 1þX0
c, 1βþ ic ð6Þ
ic ¼ωiþδcþγ þεic
i¼indus y;c¼coun y; ¼yea
Equa ion (6) uses as dependen a iable he coun y-indus y measu e o gende seg ega ion; namely, he associ-
a ion index (A
ic,
), (RLP
ic, 1
) is he ela i e labou p oduc i i y le el, a s anda d measu e o sec o al labou p oduc i -
i y ha is compu ed as he a io be ween labou p oduc i i y o sec o i, coun y c, and ime o agg ega e labou
p oduc i i y in coun y cand ime (de V ies e al., 2015). Rela i e labou p oduc i i y le el e e s o he sha e o
each sec o in o al labou p oduc i i y. The expec ed sign o he ela ionship be ween ela i e labou p oduc i i y in
gende seg ega ion a indus y-speci ic le els is inc easing, and his will con i m he esul s abo e. I inc easing labou
p oduc i i y by sec o s exe s a posi i e e ec in he associa ion index, i will imply ha his sec o employs a highe
p opo ion o women. To co espond o he p e ious esul s; howe e , he link be ween ela i e p oduc i i y and
he associa ion index should e e se a highe le els o he o me . Thus, he quad a ic e m in he model in
Equa ion (4) should be nega i e, implying ha a high le el o sec o al p oduc i i y c ea es a endency o employ
men. FLFP and he ull se o con ols, as explained in he p e ious sec ion, a e included in he coun y-indus y
panel da a model.
Table 4shows ixed-e ec s and IV es ima es o he coun y-indus y panel da a model. S uc u al change is
linked wi h an inc easing eminiza ion o sec o s. Howe e , he es ima es con i m a nonlinea ela ionship
be ween ela i e labou p oduc i i y and gende seg ega ion. A high le els o ela i e labou p oduc i i y, u he
inc eases o p oduc i i y educe he emale employmen in highly p oduc i e sec o s. I should be no ed ha he
explana o y powe o he coun y-indus y panel da a models is lowe ela i e o ha o coun y panel da a
models p esen ed abo e, as he e is limi ed a ailabili y o da a a coun y-indus y le el. None heless, some
insigh s om he esul s in Table 4can be d awn ela ed o e ili y a es and income inequali y. Fe ili y
dep esses he eminiza ion o sec o s, a esul ha can ep esen he highe unpaid ca e esponsibili ies ha
women shoulde wi h a ising numbe o child en and a limi a ion o join he paid wo k o ce (Bloom e al., 2009).
Con a y o he coun y panel da a models, income inequali y exe s a signi ican ole in he associa ion index,
educing he p esence o women in sec o s using ixed-e ec s models. This esul p o ides some empi ical
le e age o hose in Bo owman and Klasen (2020), who ind an inc easing e ec o income inequali y in sec o al
gende seg ega ion using he dissimila i y index. When using an ins umen al a iables model, he coe icien
associa ed wi h ela i e labou p oduc i i y is posi i e, and he coe icien o i s quad a ic e m is nega i e. This
sugges s again ha , o ini ial inc eases in ela i e p oduc i i y, sec o al eminiza ion is inc eased, while u he
inc eases in ela i e p oduc i i y dep ess he p esence o women. The e ec o FLFP is no signi ican in he
ins umen al a iables model.
Figu e 6shows IV es ima es o he ma ginal e ec o inc easing ela i e labou p oduc i i y on he associa ion
index. The con idence in e als a e epo ed in he g aph, showing ha , o ce ain le els o ela i e labou p oduc-
i i y, he e ec is no signi ican . Howe e , o low le els o ela i e labou p oduc i i y and high le els o ela i e
labou p oduc i i y, he e ec is signi ican . Fu he , I ind again a e e sal o he p oduc i i y link wi h seg ega ion.
As his las model uses he associa ion index as he dependen a iable, he in e p e a ion o he es ima es sugges s
ha ini ial escala ions in ela i e p oduc i i y inc ease he p esence o women. Howe e , u he inc eases a al eady
high le els o p oduc i i y a e ela ed o a dec ease in women in he sec o .
The las s ep in he empi ical analysis o he ole o s uc u al change in gende sec o al seg ega ion is o
conside each sec o sepa a ely. Hence, I eplica e he IV model in Equa ion (6) using one sec o a a ime. By doing
2642 ZUAZU
his, I am able o iden i y he ole o s uc u al change in each pa icula sec o and, a he same ime, conside he
gende domina ion o each. Since he associa ion index is no an agg ega e measu e o seg ega ion, bu a sec o -le el
one, in in e p e ing sepa a e models by sec o , one should conside he gene al gende label o each sec o .
TABLE 4 S uc u al change and indus y-speci ic gende seg ega ion (FE and IV es ima es).
Dependen a iable: Associa ion index
(1) (2) (3) (4) (5)
Reg ession model FE FE FE IV IV
Rela i e labou p oduc i i y 0.009*** 0.029*** 0.024** 0.004*0.032*
Rela i e labou p oduc i i y sq (0.002) (0.009) (0.009) (0.002) (0.016)
0.000** 0.000** 0.000*
(0.000) (0.000) (0.000)
Agg ega e labou p oduc i i y 0.084** 0.100*** 0.133*** 0.001 0.000
(0.035) (0.036) (0.044) (0.088) (0.009)
FLFP 0.009*0.002 0.003
(0.005) (0.046) (0.002)
GDP pc (logs) 0.024 0.037 0.007 0.004 0.030
(0.020) (0.022) (0.037) (0.153) (0.019)
T ade 0.000 0.001 0.000 0.000 0.000
(0.001) (0.001) (0.001) (0.002) (0.000)
FDI 0.001* 0.001 0.001 0.000 0.000
(0.001) (0.001) (0.001) (0.002) (0.000)
U ban pop. 0.003 0.000 0.001 0.004 0.001
(0.003) (0.003) (0.007) (0.028) (0.001)
Fe ili y (logs) 0.315** 0.254* 0.561*** 0.005 0.176**
(0.131) (0.141) (0.174) (0.832) (0.086)
Gini ne 0.012** 0.016** 0.032*** 0.002 0.000
(0.005) (0.006) (0.009) (0.073) (0.002)
Educa ion 0.000 0.001 0.000 0.001 0.002**
(0.002) (0.002) (0.003) (0.005) (0.001)
No. o obse a ions 1,456 1,456 1,027 790 790
No. o g oups 89 89 89 69 69
Wi hin R
2
0.028 0.040 0.063 0.010 0.024
Fi s s age Fs a 86.81 86.81
Unde iden i ica ion es (Kleibe gen–Paap k LM
s a is ic)
10.450 10.389
p alue 0.0012 0.0013
Weak iden i ica ion es (Kleibe gen–Paap k Wald F
s a is ic)
8.345 8.319
No e: D iscoll and K aay s anda d e o s in pa en heses (columns 1–3). Coun y-indus y clus e ed s anda d e o s in
pa en heses (columns 4 and 5). Kleibe gen–Paap LM es null hypo hesis is ha he ank condi ion ails in second s age
equa ion, ha is, unde iden i ied (Kleibe gen & Paap, 2006). Kleibe gen–Paap k Wald Fc i ical alues a ying be ween 5.53
and 16.38, and 6.66 a he 20% maximal IV size (S ock & Yogo, 2005). Coe icien on he ins umen in he i s s age was
nega i e (1.27) and s a is ically signi ican a he 0.01 le el.
*p< 0.1.**p< 0.05.***p< 0.01.
ZUAZU 2643
Table 5pools he 11 coun ies in he sample and ocuses on eg essions sepa a ely by each indus y. The able
p o ides in o ma ion on he gende label o each sec o , ha is, F o emale, M o male, and N o neu al ( hese
ca ego ies we e p o ided based on he a e age associa ion index o each sec o in he da abase). The only neu al
sec o is go e nmen se ices, whe e he es a e di ided in o F o M. To in e p e he sign o he es ima ed coe i-
cien s, one should conside whe he he sec o is emale domina ed, male domina ed, o gende neu al: a posi i e
(nega i e) coe icien would imply an inc ease in gende seg ega ion in a emale-domina ed (male-domina ed) sec o .
A posi i e (nega i e) coe icien o a neu al sec o will imply eminiza ion (masculiniza ion) o ha sec o .
The esul s p o ided in he eg ession models abo e a e d i en by ce ain indus ies such as manu ac u ing,
u ili ies, cons uc ion, inancial and business se ices, and go e nmen se ices, as hese a e he indus ies whe e
ela i e p oduc i i y gains a e ound o ha e a signi ican link o gende seg ega ion. Inc easing ela i e p oduc i i y
in manu ac u ing (column 3, Table 5), which is on a e age a emale-domina ed sec o , p obably due o he ex ile
subsec o , educes he gende seg ega ion o he sec o , making i a mo e gende -neu al sec o . Howe e , I do no
ind a nonlinea link in his associa ion, as he quad a ic e m o ela i e p oduc i i y is no s a is ically signi ican .
Rising ela i e labou p oduc i i y in u ili ies and cons uc ion (columns 4 and 5, Table 5) is associa ed wi h an
inc easing emale ep esen a ion in hose sec o s, bu his e e ses when his p oduc i i y gains occu in highe
le els o ela i e p oduc i i y. Thus, o low le els o sec o al labou p oduc i i y, inc easing p oduc i i y e e s he
male domina ion o u ili ies and cons uc ion. This goes in line wi h he indings o Rendall (2013,2017), sugges ing
ha echnological adop ion and subsequen ises in p oduc i i y educes he en y ba ie s o women in such sec-
o s. So long as ela i e labou p oduc i i y is conside ed a p oxy o echnological upg ading, as in Seguino and B a-
uns ein (2019) and Tejani and Kuce a (2021), we can dis il wo opposing mechanisms behind he indings in he
indus y-le el models. A b awn e sus b ain ade-o , as sugges ed by Rendall (2017), migh induce he en ance o
women in p e iously domina ed sec o s whe e physical demands a e pa icula ly impo an , as i is he case o u ili-
ies and cons uc ion. None heless, he es ima es o he quad a ic e m o ela i e labou p oduc i i y migh sugges
ha i depends on he le el o sec o al echnological adop ion. A ce ain le els o echnological adop ion, gende
s e eo ypes and disc imina o y hi ing p ac ices migh coun e ail he gende equali y gains o a p o-b ain e ec o
ising p oduc i i y and induce he c owding o women in gende adi ional sec o s and low-p oduc i e sec o s.
FIGURE 6 Ma ginal e ec s o ela i e labou p oduc i i y in associa ion index. Sou ce: Based on Reg ession
esul s o model in column 5, Table 4.
2644 ZUAZU
TABLE 5 Pooled indus y-le el eg essions (ins umen al a iables).
Dependen a iable:
Associa ion index
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Ag i. Mining Manu . U ili ies Cons uc ion T ade T ansp.
Fin. & Bus.
Se ices
Go e nmen
se ices
Pe sonal
se ices
Gende associa ion F M F M M F M F N F
Rela i e labou p oduc i i y 0.470 0.044 1.283** 0.140*** 1.010*1.132 0.415 0.148*** 0.304** 0.621
(3.492) (0.029) (0.520) (0.038) (0.548) (2.425) (0.332) (0.027) (0.127) (1.940)
Rela i e labou p oduc i i y
sq
0.103 0.000 0.268 0.002*** 0.409** 0.117 0.055 0.003*** 0.096*** 0.914
(3.176) (0.000) (0.165) (0.001) (0.200) (1.010) (0.033) (0.001) (0.025) (1.332)
FLFP 0.070** 0.192** 0.108** 0.104*0.027 0.134** 0.082** 0.068 0.064*0.001
(0.032) (0.081) (0.040) (0.061) (0.107) (0.059) (0.040) (0.042) (0.033) (0.035)
Agg ega e labou
p oduc i i y
0.096 0.429 0.032 0.352*0.280 0.323 0.465** 0.040 0.071 0.046
(0.175) (0.463) (0.105) (0.203) (0.441) (0.303) (0.221) (0.148) (0.123) (0.104)
Full se o con ols Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Time ixed-e ec s Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
No. o obse a ions 80 80 80 80 80 80 80 80 79 80
No. o g oups 7 7 7 7 7 7 7 7 6 7
Wi hin R
2
0.792 0.895 0.962 0.989 0.874 0.927 0.915 0.942 0.980 0.981
No e: Coun y-indus y clus e ed s anda d e o s. Models a e pooled by each o he 10 sec o s in he ASD da abase. Gende associa ion e e s o he sample a e age o each sec o : F
e e s o emale-domina ed, M e e s o male-domina ed, and N e e s o gende -neu al indus ies. No da a on go e nmen se ices o Zambia (column 9).
*p< 0.1.**p< 0.05.***p< 0.01.
ZUAZU 2645

Inc easing ela i e labou p oduc i i y o inancial and business se ices (column 8, Table 5) educes gende seg-
ega ion, bu again his is e e sed as ela i e p oduc i i y is la ge enough. The he e ogenei y o he subindus ies
conside ed in inancial and business se ices, which also accoun o eal es a e, should be conside ed when in e -
p e ing his model, as he limi ed access o A ican women in he inancial sec o (Mo sy, 2020) can be coun e ailed
by he p esence o women in small i ms in eal es a e se ices.
Finally, go e nmen se ices (column 9, Table 5), which include ca e, heal h, and educa ional se ices, a e he
only indus ies on a e age ha a e gende -neu al sec o s. Inc easing ela i e p oduc i i y a low le els o ha a i-
able il s he sec o owa ds males, bu u he gains o p oduc i i y a ou a mo e gende -balanced dis ibu ion in
he sec o . Gende equal access o educa ion and he access o women o highe educa ion in hose coun ies whe e
go e nmen p oduc i i y is high migh be a po en ial mechanism behind hese esul s.
6|CONCLUSION
De elopmen economics has long been di ec ed a he ole o s uc u al change in unde s anding egional dispa i ies
in economic g ow h. Indeed, he ole o s uc u al change in sub-Saha an A ica is a he co e o he economic de el-
opmen deba es in ecen li e a u e. Howe e , li le is known ega ding he gende ed impac s o his ans o ma ion.
S uc u al change p oduces composi ion shi s om low-p oduc i i y sec o s o high-p oduc i i y sec o s, bo h in
e ms o alue-added and employmen sha es. These shi s a e likely o a ec women's and men's employmen di -
e en ly, media ed by complex in e ac ions. This pape has documen ed a signi ican in e play be ween s uc u al
change and gende -sec o al seg ega ion in a sample o sub-Saha an A ican coun ies.
Using a da abase wi h in o ma ion on 10 sec o s ope a ing in 11 SSA coun ies du ing 1960–2010, I iden i y
ha hose coun ies whe e gende seg ega ion has dec eased show, simul aneously, an inc ease in labou p oduc i -
i y. To conside he causal ole o s uc u al change in he gende dis ibu ion o sec o al employmen , he pape
speci ies da a models a bo h agg ega e (coun y-yea ) and disagg ega e le els (coun y-indus y-yea ). Toge he
wi h ins umen al a iable app oaches, he esul s he e sugges a nonlinea co ela ion be ween labou p oduc i i y
and gende seg ega ion.
This pape inds a nonlinea ela ionship be ween ising labou p oduc i i y and gende sec o al seg ega ion. Ini-
ial gains in p oduc i i y inc ease emale employmen ac oss sec o al le els. None heless, u he p oduc i i y gains
imply inc easing sec o al seg ega ion by gende , possibly h ough highe ba ie s o women o en e highly p oduc-
i e sec o s, and a c owding e ec in he se ice sec o . Ano he impo an esul o his pape is in he ole o
emale labou o ce pa icipa ion in gende -sec o al seg ega ion. Rising emale labou o ce pa icipa ion appea s o
educe sec o al seg ega ion, p obably by changing cul u al no ms and e oding adi ional gende oles in paid and
unpaid wo k.
The main esul s o he pape emain when ci cum en ing endogenei y issues o emale labou o ce by using an
ins umen al a iables app oach. The es ima es s ongly suppo he iew ha e ili y decline can equalize he gen-
de dis ibu ion o sec o al employmen . U baniza ion and income inequali y a e gene ally associa ed wi h inc easing
seg ega ion. Finally, he esul s do no associa e economic g ow h wi h a signi ican ole in gende seg ega ion.
Coun y-indus y-le el panel da a es ima es u he allow iden i ica ion o which speci ic sec o s a e media ed by
s uc u al change in he eminiza ion o masculiniza ion o employmen . Speci ically, he e ec s o s uc u al change
in gende -sec o al seg ega ion in SSA seem o be media ed h ough manu ac u ing, u ili ies, cons uc ion, business,
and go e nmen se ices.
These indings add o he gene al li e a u e in s uc u al change and gende -awa e mac oeconomics. The e a e
impo an policy implica ions ha can be de i ed om he empi ical analysis he e. Fi s , he p ocess o s uc u al
change comes along wi h complex ans o ma ions o he p oduc ion o ma ke and nonma ke ac i i ies, o mal and
in o mal sec o s, as well as paid and unpaid wo k. A gende -sec o al pe spec i e is needed o ully unde s and he
implica ions o s uc u al change o he whole economy and he wo ke s. While emale labou o ce pa icipa ion is
2646 ZUAZU
ound o educe gende seg ega ion, o he ac o s o s uc u al change, such as employmen shi s in highly p oduc-
i e sec o s, can coun e ail hese gende equali y ends. The in e plays be ween he pa icipa ion o women in he
paid o ce and sec o al seg ega ion can be in e p e ed as e idence ha , as some gende inequali ies a e e oded,
o he , new ypes o inequali ies eme ge. Finally, declining e ili y appea s o be o i s -o de impo ance in p omo -
ing a gende -balanced dis ibu ion o sec o al employmen .
ACKNOWLEDGEMENTS
I app ecia e he commen s and sugges ions o Ipek Ilkka acan, Da id Kuce a, Luiza Nassi Pi es, Sheba Tejani, Fiona
T egenna, and Elissa B auns ein. I p esen ed his pape a he Gende Inequali ies and Economic Theo y and Policies o
a Pos -Pandemic Wo ld a he Le y Economics Ins i u e o Ba d College (NY, USA) in Sep embe 2022, o which I am
e y hank ul o Thomas Mas e son and Aji Zacha ias o he in i a ion, and he o he a endan s o discussions. I
am also g a e ul o he audiences o he Gende , Wo k and O ganiza ion (GWO) annual con e ence in June 2023
(S ellenbosch, Sou h A ica), he Ins i u e o New Economic Thinking (INET) –Young Schola Ini ia i e (YSI) 3 d
Young Schola s Con e ence on S uc u al Change and Indus ial Policy (Uni e si y o Johannesbu g) in June 2023 and
he Al e na i e App oaches o Inno a ion in a Dynamic Wo ld wo kshop (Complu ense Uni e si y o Mad id) in Ma ch
2024. I acknowledge inancial suppo in he o m o a el s ipends om he Le y Economics Ins i u e and
INET-YSI. This pape won he unne -up bes pape p ize a he INET-YSI 3 d Con e ence on S uc u al Change and
Indus ial Policy.
DATA AVAILABILITY STATEMENT
The da a and eplica ion ma e ials o his pape a e a ailable om he au ho upon easonable eques .
ORCID
Izaskun Zuazu h ps://o cid.o g/0000-0003-0638-2567
ENDNOTES
1
See Be io e al. (2009) o a u he insigh in he dis inc ion be ween e ical and ho izon al seg ega ion. Fo examples o
s uc u al analysis o gende e ical and ho izon al seg ega ion, see Bo owm and Klasen (2020) and A o a e al. (2023).
2
I should be no ed ha he measu e o s uc u al change in Bo owman and Klasen (2020) is exclusi ely based on sec o al
employmen , wi h is a odds wi h he sugges ion in T egenna (2015) on ha bo h employmen and alue added should be
conside ed in s udying he consequences o s uc u al change.
3
Some coun ies s a he ime se ies a di e en yea s (Bo swana: 1964; E hiopia: 1961; Kenya: 1969; Malawi: 1966;
Mau i ius:1970; Senegal: 1970; Tanzania: 1961; Zambia: 1965). Mensah and Szi mai (2018) upda ed he ASD al hough
did no p o ide in o ma ion on he emale sha e o employmen , and hus, he mos ecen yea his pape u ilizes is 2010.
See de V ies e al. (2015) o mo e de ails on he ASD.
4
The s a is ical capaci y o A ican coun ies su e ed om limi ed unding and hus de e io a ed he accu acy o es ima es
o in o mal economic ac i i ies. This should be conside ed when in e p e ing he desc ip i e and econome ic analysis
using ASD da abase (de V ies e al., 2013).
5
The IP index is gi en by he ollowing o mula: IP ¼1=TðÞ
P
n
i¼1
FiaM
iþFi
ðÞ
jj
, whe e Tis o al employmen and ais he
sha e o women in o al employmen . F
i
and M
i
co espond espec i ely o he numbe o women and numbe o men in
sec o i. See Wa s (1998) o mo e discussion on seg ega ion measu es.
6
See Figu e A4 in Appendix A o he e olu ion o A and ela i e labou p oduc i i y by sec o .
7
Models using o e lapping and nono e lapping 5 yea s pe iods, which a e a ailable upon eques , yield simila esul s as
exposed in he pape .
8
Augmen ing lagged pe iods is also employed in a ela ed wo k by Tejani and Kuce a (2021), o ensu ing he exogenei y o
he eg esso s.
9
I am hank ul o an anonymous e iewe o sugges ing his pa i ion, which se es as a obus ness check o he non-linea
link be ween labou p oduc i i y and gende sec o al seg ega ion.
ZUAZU 2647
REFERENCES
Aga wal, B. (1997). ‘Ba gaining’and gende ela ions: Wi hin and beyond he household. Feminis Economics,3(1), 1–51.
h ps://doi.o g/10.1080/135457097338799
A o a, D., B auns ein, E., & Seguino, S. (2023). A mac o analysis o gende seg ega ion and job quali y in La in Ame ica.
Wo ld De elopmen ,164, 106153. h ps://doi.o g/10.1016/j.wo ldde .2022.106153
Backhaus, A., & Loichinge , E. (2022). Female labo o ce pa icipa ion in sub-Saha an A ica: A coho analysis. Popula ion
and De elopmen Re iew,48(2), 379–411.
Ba en, J., de Haas, M., Kemp e , E., & Meie zu Selhausen, F. (2021). Educa ional gende inequali y in sub-Saha an A ica: A
long- e m pe spec i e. Popula ion and De elopmen Re iew,47(3), 813–849. h ps://doi.o g/10.1111/pad .12430
Bellema e, M. F., Masaki, T., & Pepinsky, T. B. (2017). Lagged explana o y a iables and he es ima ion o causal e ec . The
Jou nal o Poli ics,79(3), 949–963. h ps://doi.o g/10.1086/690946
Be gmann, B. R. (1981). The economic isks o being a housewi e. The Ame ican Economic Re iew,71(2), 81–86.
Be io, F., Ve ashchagina, A., Mai hube , I., & Kanjuo-M ˇ
cela, A. (2009). Gende seg ega ion in he labou ma ke : Roo causes,
implica ions and policy esponses in he EU. Publica ions O ice o he Eu opean Union.
Bloom, D. E., Canning, D., Fink, G., & Finlay, J. E. (2009). Fe ili y, emale labo o ce pa icipa ion, and he demog aphic di i-
dend. Jou nal o Economic G ow h,14(2), 79–101. h ps://doi.o g/10.1007/s10887-009-9039-9
Bo owman, M., & Klasen, S. (2020). D i e s o gende ed sec o al and occupa ional seg ega ion in de eloping coun ies.
Feminis Economics,26(2), 62–94. h ps://doi.o g/10.1080/13545701.2019.1649708
Bose up, E., Tan, S. F., & Toulmin, C. (2013). Woman's ole in economic de elopmen . Rou ledge. h ps://doi.o g/10.4324/
9781315065892
Cha les, M., & B adley, K. (2009). Indulging ou gende ed sel es? Sex seg ega ion by ield o s udy in 44 coun ies. Ame ican
Jou nal o Sociology,114(4), 924–976. h ps://doi.o g/10.1086/595942
Cha les, M., & G usky, D. B. (1995). Models o desc ibing he unde lying s uc u e o sex seg ega ion. Ame ican Jou nal o
Sociology,100(4), 931–971. h ps://doi.o g/10.1086/230605
Cha les, M., & G usky, D. B. (2005). Occupa ional ghe os: The wo ldwide seg ega ion o women and men (Vol. 200). S an o d
Uni e si y P ess. h ps://doi.o g/10.1515/9781503618183
de V ies, G., L. A el , D. D ees, M. Godemann, C. Hamil on, B. Jessen-Thiesen, and P. Wol je . (2021). “The economic
ans o ma ion da abase (e d): con en , sou ces, and me hods.”Technical No e, WIDER.
de V ies, G., Timme , M., & de V ies, K. (2015). S uc u al ans o ma ion in A ica: S a ic gains, dynamic losses. The Jou nal
o De elopmen S udies,51(6), 674–688. h ps://doi.o g/10.1080/00220388.2014.997222
de V ies, K., de V ies, G., Gouma, R., & Timme , M. (2013). A ica sec o da abase: Con en s, sou ces and me hods. G oningen
G ow h and De elopmen Cen e.
Dhana aj, S., & Mahamba e, V. (2019). Family s uc u e, educa ion and women's employmen in u al India. Wo ld
De elopmen ,115,17–29. h ps://doi.o g/10.1016/j.wo ldde .2018.11.004
Dinkelman, T., & Ngai, L. R. (2022). Time use and gende in A ica in imes o s uc u al ans o ma ion. Jou nal o Economic
Pe spec i es,36(1), 57–80. h ps://doi.o g/10.1257/jep.36.1.57
D iscoll, J. C., & K aay, A. C. (1998). Consis en co a iance ma ix es ima ion wi h spa ially dependen panel da a. Re iew o
Economics and S a is ics,80(4), 549–560. h ps://doi.o g/10.1162/003465398557825
Duncan, O. D., & Duncan, B. (1955). A me hodological analysis o seg ega ion indexes. Ame ican Sociological Re iew,20(2),
210–217. h ps://doi.o g/10.2307/2088328
Elson, D. (1995). Gende awa eness in modeling s uc u al adjus men . Wo ld De elopmen ,23(11), 1851–1868. h ps://doi.
o g/10.1016/0305-750X(95)00087-S
Gaddis, I., & Klasen, S. (2014). Economic de elopmen , s uc u al change, and women's labo o ce pa icipa ion. Jou nal o
Popula ion Economics,27(3), 639–681. h ps://doi.o g/10.1007/s00148-013-0488-2
Goldin, C. (1995). “The U-shaped emale labo o ce unc ion in economic de elopmen and economic his o y.”In es men
in Women's Human Capi al and Economic De elopmen .
Go lieb, C., D. Gollin, C. Doss, and M. Poschke. (2022). “Gende , wo k and s uc u al ans o ma ion.”
He endo , B., Roge son, R., & Valen inyi, A. (2014). G ow h and s uc u al ans o ma ion. In Handbook o economic g ow h
(Vol. 2) (pp. 855–941). Else ie . h ps://doi.o g/10.1016/B978-0-444-53540-5.00006-9
He endo , B., Roge son, R., & Valen inyi, A. (2022). New e idence on sec o al labo p oduc i i y: Implica ions o indus ializa-
ion and de elopmen (Tech. Rep.). Na ional Bu eau o Economic Resea ch.
Hoechle, D. (2007). Robus s anda d e o s o panel eg essions wi h c oss-sec ional dependence. The S a a Jou nal,7(3),
281–312. h ps://doi.o g/10.1177/1536867X0700700301
Ka mel, T., & MacLachlan, M. (1988). Occupa ional sex seg ega ion—Inc easing o dec easing? Economic Reco d,64(3),
187–195. h ps://doi.o g/10.1111/j.1475-4932.1988. b02057.x
Klasen, S., & Pie e s, J. (2015). Wha explains he s agna ion o emale labo o ce pa icipa ion in u ban India? The Wo ld
Bank Economic Re iew,29(3), 449–478. h ps://doi.o g/10.1093/wbe /lh 003
2648 ZUAZU
Kleibe gen, F., & Paap, R. (2006). Gene alized educed ank es s using he singula alue decomposi ion. Jou nal o
Econome ics,133(1), 97–126. h ps://doi.o g/10.1016/j.jeconom.2005.02.011
Kuzne s, S. (1966). Mode n economic g ow h: Ra e, s uc u e, and sp ead (Vol. 2). Yale Uni e si y P ess New Ha en.
Leszczensky, L., & Wolb ing, T. (2022). How o deal wi h e e se causali y using panel da a? ecommenda ions o
esea che s based on a simula ion s udy. Sociological Me hods & Resea ch,51(2), 837–865. h ps://doi.o g/10.1177/
0049124119882473
Lewis, W. A. (1965). A e iew o economic de elopmen . The Ame ican Economic Re iew,55(1/2), 1–16.
McMillan, M., D. Rod ik, and C. Sepul eda. (2017). “S uc u al change, undamen als, and g ow h: A amewo k and case
s udies.”Wo ld Bank Policy Resea ch Wo king Pape (8041).
McMillan, M., Rod ik, D., & Ve duzco-Gallo, Í. I. (2014). Globaliza ion, s uc u al change, and p oduc i i y g ow h, wi h an
upda e on A ica. Wo ld De elopmen ,63,11–32. h ps://doi.o g/10.1016/j.wo ldde .2013.10.012
McMillan, M. S., & Rod ik, D. (2011). Globaliza ion, s uc u al change and p oduc i i y g ow h (Tech. Rep.). Na ional Bu eau o
Economic Resea ch.
Mensah, E., Owusu, S., Fos e -McG ego , N., & Szi mai, A. (2023). S uc u al change, p oduc i i y g ow h and labou ma ke
u bulence in Sub-Saha an A ica. Jou nal o A ican Economies,32(3), 175–208. h ps://doi.o g/10.1093/jae/ejac010
Mensah, E. B., Szi mai, A., (2018). A ica sec o da abase (ASD): Expansion and upda e. UNU-MERIT Wo king Pape
(2018-020).
Milano ic, B. (2003) “Is inequali y in A ica eally di e en ?”A ailable a SSRN 636588.
Mo sy, H. (2020). Access o inance: Why a en' women leaning in? (pp. 52–53). IMF.
Na eed, A., & Ahmad, N. (2016). Labou p oduc i i y con e gence and s uc u al changes: Simul aneous analysis a coun y,
egional and indus y le els. Jou nal o Economic S uc u es,5,1–17. h ps://doi.o g/10.1186/s40008-016-0050-y
Nelson, J. A. (2017). Gende and isk- aking: Economics, e idence, and why he answe ma e s. Rou ledge. h ps://doi.o g/10.
4324/9781315269887
Ngai, L. R., & Pe ongolo, B. (2017). Gende gaps and he ise o he se ice economy. Ame ican Economic Jou nal:
Mac oeconomics,9(4), 1–44. h ps://doi.o g/10.1257/mac.20150253
Rendall, M. (2013). S uc u al change in de eloping coun ies: Has i dec eased gende inequali y? Wo ld De elopmen ,45,
1–16. h ps://doi.o g/10.1016/j.wo ldde .2012.10.005
Rendall, M. (2017). B ain e sus b awn: The ealiza ion o women's compa a i e ad an age. Uni e si y o Zu ich, Ins i u e o
Empi ical Resea ch in Economics. Wo king Pape (491)
Rios-A ila, F., Odu o, A., & Nassi Pi es, L. (2021). In ahousehold alloca ion o household p oduc ion: A compa a i e analysis o
sub-Saha an A ican coun ies. Le y Economics Ins i u e. Wo king Pape s Se ies
Rod ik, D. (2016). P ema u e deindus ializa ion. Jou nal o Economic G ow h,21(1), 1–33. h ps://doi.o g/10.1007/s10887-
015-9122-3
Seguino, S., & B auns ein, E. (2019). The cos s o exclusion: Gende job seg ega ion, s uc u al change and he labo sha e o
income. De elopmen and Change,50(4), 976–1008. h ps://doi.o g/10.1111/dech.12462
Seguino, S., & We e, M. (2014). Gende , de elopmen and economic g ow h in sub-Saha an A ica. Jou nal o A ican
Economies,23(suppl 1), i18–i61. h ps://doi.o g/10.1093/jae/ej 024
Sol , F. (2020). Measu ing income inequali y ac oss coun ies and o e ime: The s anda dized wo ld income inequali y
da abase. Social Science Qua e ly,101(3), 1183–1199. h ps://doi.o g/10.1111/ssqu.12795
Spie ings, N. (2014). The in luence o pa ia chal no ms, ins i u ions, and household composi ion on women's employmen
in wen y-eigh Muslim-majo i y coun ies. Feminis Economics,20(4), 87–112. h ps://doi.o g/10.1080/13545701.
2014.963136
S ock, J., & Yogo, M. (2005). Tes ing o weak ins umen s in linea IV eg ession. In And ews DWK iden i ica ion and
in e ence o econome ic models (Vol. 2005) (pp. 80–108). Camb idge Uni e si y P ess. h ps://doi.o g/10.1017/
CBO9780511614491.006
Tejani, S., & Kuce a, D. (2021). De eminiza ion, s uc u al ans o ma ion and echnological up- g ading in manu ac u ing.
De elopmen and Change,52(3), 533–573. h ps://doi.o g/10.1111/dech.12650
T egenna, F. (2015). Deindus ialisa ion: an issue o bo h de eloped and de eloping coun ies. In Rou ledge handbook o
indus y and de elopmen (pp. 111–129). Rou ledge. h ps://doi.o g/10.4324/9780203387061-13
T egenna, F. (2016). Deindus ializa ion and p ema u e deindus ializa ion. In Handbook o al e na i e heo ies o economic
de elopmen . Edwa d Elga Publishing. h ps://doi.o g/10.4337/9781782544685.00046
Ube i, L., & Doua in, E. (2022). The eminisa ion U, cul u al no ms, and he plough. Jou nal o Popula ion Economics.,36,
5–35. h ps://doi.o g/10.1007/s00148-022-00890-5
UN. (2019). Pa e ns and ends in household size and composi ion: E idence om a uni ed na ions da ase . Uni ed Na ions,
Depa men o Economic and Social A ai s, Popula ion Di ision.
Van A k, B. (1995). Sec o al g ow h accoun ing and s uc u al change in pos -wa . In Quan i a i e aspec s o pos -wa
Eu opean economic g ow h (Vol. 1). Camb idge Uni e si y P ess.
ZUAZU 2649