Mazo odze, B ian Ta onga
A icle
Expo s, go e nmen deb and economic g ow h in sub-
Saha an A ica
Cogen Economics & Finance
P o ided in Coope a ion wi h:
Taylo & F ancis G oup
Sugges ed Ci a ion: Mazo odze, B ian Ta onga (2024) : Expo s, go e nmen deb and economic
g ow h in sub-Saha an A ica, Cogen Economics & Finance, ISSN 2332-2039, Taylo & F ancis,
Abingdon, Vol. 12, Iss. 1, pp. 1-25,
h ps://doi.o g/10.1080/23322039.2024.2426534
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/321664
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Cogen Economics & Finance
ISSN: 2332-2039 (Online) Jou nal homepage: www. and online.com/jou nals/oae 20
Expo s, go e nmen deb and economic g ow h
in sub-Saha an A ica
B ian Ta onga Mazo odze
To ci e his a icle: B ian Ta onga Mazo odze (2024) Expo s, go e nmen deb and
economic g ow h in sub-Saha an A ica, Cogen Economics & Finance, 12:1, 2426534, DOI:
10.1080/23322039.2024.2426534
To link o his a icle: h ps://doi.o g/10.1080/23322039.2024.2426534
© 2024 The Au ho (s). Published by In o ma
UK Limi ed, ading as Taylo & F ancis
G oup
View supplemen a y ma e ial
Published online: 11 No 2024.
Submi you a icle o his jou nal
A icle iews: 845
View ela ed a icles
View C ossma k da a
Full Te ms & Condi ions o access and use can be ound a
h ps://www. and online.com/ac ion/jou nalIn o ma ion?jou nalCode=oae 20
DEVELOPMENT ECONOMICS | RESEARCH ARTICLE
Expo s, go e nmen deb and economic g ow h in sub-Saha an
A ica
B ian Ta onga Mazo odze
The Depa men o Accoun ing and Economics, Sol Plaa je Uni e si y, Kimbe ley, Sou h A ica
ABSTRACT
Expo s in sub-Saha an A ica ha e s uggled o gene a e he kind o g ow h wi -
nessed in Sou heas Asia on he backg ound o ising go e nmen deb . This a icle
conside s he ex en o which go e nmen deb , which has ipled since 2008, may
ha e cons ained expo -led g ow h in he egion. Fu he examined is he ex en o
which deb educ ion is possible h ough e enue g ow h and expendi u e cu s. F om
a sample o 44 sub-Saha an coun ies obse ed be ween 2004 and 2023, he impac
o expo s on economic g ow h is ound o be s onge in coun ies wi h low o mod-
e a e le els o deb and weake in coun ies wi h high deb . To achie e obus eco-
nomic g ow h om expo s, sub-Saha an coun ies need o keep deb below 45% o
GDP. F om he esul s, his equi es cu ing expendi u e on GDP by 2–10 pe cen age
poin s o aising e enue (excluding g an s) by 2–12 pe cen age poin s. These es i-
ma es a e in line wi h he p ima y balance adjus men o a leas 2 pe cen age poin s
ecommended by he IMF. Es ima es u he indica e ha cu ing go e nmen expend-
i u e educes deb mo e han aising e enue. Coun e ac ual scena ios show ha cu -
ing go e nmen expendi u e on GDP by 15 pe cen age poin s would educe
go e nmen deb om 55 o 45% o GDP.
IMPACT STATEMENT
This pape add esses he ques ion o why ade has ha dly accele a ed g ow h in sub-
Saha an A ica and he impo ance o managing go e nmen deb as a mode a ing
ac o . The empi ical e idence om a panel o 44 coun ies demons a es he impo -
ance o p uden iscal policies and deb managemen s a egies as he g ow h e ec
o expo s is cons ained by high go e nmen deb . This e idence is consis en wi h
he IMF’s policy s ance on he need o u gen deb managemen s a egies in he
sub-Saha an egion.
ARTICLE HISTORY
Recei ed 10 June 2024
Re ised 24 Oc obe 2024
Accep ed 3 No embe 2024
KEYWORDS
Economic g ow h; expo s;
go e nmen deb ; Sou h
A ica; sub-Saha an A ica
JEL CODES
F10; F41; F62; F68
SUBJECTS
Mac oeconomics;
Economics; In e na ional
Economics
1. In oduc ion
The ela ionship be ween expo s and economic g ow h is one o he widely esea ched opics in de el-
opmen economics. A gene al lesson om his li e a u e is ha he impac o expo s on economic
g ow h depends on se e al ac o s such as he composi ion o expo s, egional con ex , model speci i-
ca ion and a iable measu emen (S deli
c&D
a ila-Fe n
andez, 2024; T o
s & Bojnec, 2016). The cu en
a icle con ibu es o his empi ical e idence by examining how he le el o go e nmen deb in luences
he impac o expo s on economic g ow h in sub-Saha an A ica. I explo es he possibili y o a non-
linea associa ion in which expo s impac economic g ow h s ongly and posi i ely a low le els o go -
e nmen deb and weakly a high le els o go e nmen deb .
The a e se e al easons why go e nmen deb can non-linea ly in luence he impac o expo s on eco-
nomic g ow h. These non-linea i ies can be explained ei he in he con ex o he gap illing heo y, he
likely he e ogenei y in he use o low and high deb , and he likelihood o o he g ow h d i e s being
CONTACT B ian Ta onga Mazo odze [email p o ec ed] The Depa men o Accoun ing and Economics, Sol Plaa je
Uni e si y, Kimbe ley, Sou h A ica.
Supplemen al da a o his a icle can be accessed online a h ps://doi.o g/10.1080/23322039.2024.2426534.
ß2024 The Au ho (s). Published by In o ma UK Limi ed, ading as Taylo & F ancis G oup
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0/), which
pe mi s un es ic ed 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. The e ms on which his a icle has been
published allow he pos ing o he Accep ed Manusc ip in a eposi o y by he au ho (s) o wi h hei consen .
COGENT ECONOMICS & FINANCE
2024, VOL. 12, NO. 1, 2426534
h ps://doi.o g/10.1080/23322039.2024.2426534
dependen on he le el o deb .
1
The mos compelling case ela es o deb in e es paymen s. A lowe le -
els o go e nmen deb , in e es paymen s a e ypically low hence go e nmen s can alloca e esou ces
owa ds he mo e e icien expo sec o hus aising agg ega e p oduc i i y (Eme y, 1967). A high le els o
go e nmen deb , deb se ice cos s ake away much o wha could o he wise be channelled owa ds
in es men . This is especially ele an in coun ies whe e go e nmen deb is o eign cu ency-denomi-
na ed and gua an eed by u u e expo ea nings. The expe iences o highly indeb ed coun ies such as
E i ea and Libe ia a e illus a i e. As Fede (1983) and Juh
asz e al. (2023) allude, i ms in hese coun ies
gene ally espond by ealloca ing esou ces om he p oduc i e adable sec o o he less p oduc i e non-
adable sec o , leading o a dec ease in agg ega e p oduc i i y. High deb se ice cos s also educe he
supply o o eign cu ency ela i e o he domes ic cu ency. This acili a es a cu ency dep ecia ion, which
ul ima ely slows down economic ac i i y by inc easing he cos o impo ing in e media es.
Few places p o ide a iche g ound o examining he in e play be ween go e nmen deb , expo s,
and economic g ow h han sub-Saha an A ica. Mos sub-Saha an coun ies ace high bo owing cos s
and con inue o g apple wi h ongoing deb ulne abili ies. Nine een ou o 35 low-income coun ies in
sub-Saha an A ica a e cu en ly in deb dis ess (IMF, 2024). In some o he egion’s mos deb -
dis essed economies, such as E i ea, Libe ia, S~
ao Tom
e and P
ıncipe, and Cabo Ve de, go e nmen deb
a e aged 224, 132, 109, and 102% o GDP, espec i ely, be ween 2004 and 2023. In addi ion, go e n-
men in e es paymen s o he egion as a sha e o o al e enue ha e wo isomely doubled in he las
decade (Figu e 1) and a e expec ed o inc ease o e he nex wo yea s (IMF, 2024). These dis u bing
ac s aise he ques ion o how go e nmen deb could be hinde ing expo -led g ow h as he egion
in eg a es in o global ade.
Compounding he egion’s wo sening deb p o ile is GDP g ow h ha has been slowing since 2004
(Figu e 2). A e age annual eal GDP g ow h declined om 6.3% be ween 2004 and 2008 o 2.7%
be ween 2018 and 2013. Simila ly, a e age annual eal pe capi a income g ow h ell om 3.8% be ween
2004 and 2008 o a me e 0.14% be ween 2018 and 2023. Wha could be unde mining he egion’s eco-
nomic g ow h momen um? In his analysis, he possibili y o high deb pulling down he egion’s abili y
o bene i om expo -d i en g ow h is conside ed.
Wi hin he li e a u e, his a icle combines wo s ands o li e a u e. The i s s and comp ises s udies
add essing he e ec o expo s on economic g ow h (Boza li e al., 2023; Fosu, 1990; Islam, 2023;
My odias, 2024; Rangasamy, 2009; Sunde, 2017). Me hodologically, hese s udies ei he eg ess g ow h
measu es on expo in ensi y o ely on G ange causali y es s. A unidi ec ional causali y unning om
expo s o g ow h is usually aken as e idence o expo -led g ow h. The second s and o li e a u e
comp ises s udies analyzing he non-linea e ec o go e nmen deb on economic g ow h. These s ud-
ies include Reinha and Rogo (2010), Ndo icimpa (2020), Liu and Lyu (2021), Law e al. (2021), and
Shah e al. (2024). An in e ed U-shaped ela ionship is con i med in hese s udies, sugges ing ha small
Figu e 1. Sub-Saha an A ica’s go e nmen deb and se ice cos s.
Sou ce: Own compu a ion using da a om he IMF.
2 B. T. MAZORODZE
o mode a e le els o go e nmen deb p omo e economic g ow h while high le els o deb could be
de imen al.
The cu en s udy di e s by es ima ing he impac o expo s on economic g ow h below and abo e
an es ima ed h eshold le el o go e nmen deb . Unlike Reinha and Rogo (2010), his analysis consid-
e s a di e en ype o non-linea i y be ween go e nmen deb and economic g ow h. Common
app oaches a ailable o achie ing his objec i e include he Panel Smoo h T ansi ion Reg ession (PSTR)
and he Hansen’s(1999) h eshold model. Al hough hese models ha e widely been applied in he li e a-
u e, hey equi e exogenei y in all co a ia es, including he h eshold a iable
2
(Seo & Shin, 2016). This
is a s ic assump ion, gi en he po en ial endogenei y o bo h expo s and go e nmen deb . I has
been a gued in he li e a u e ha deb is endogenous since low economic g ow h can a ac go e n-
men deb (Woo & Kuma , 2015). Some o he s udies ha e also a gued a e e se causa ion om g ow h
o expo s. The PSTR and he Hansen app oach a e addi ionally s a ic, which makes hem incapable o
cap u ing he dynamic p ocess o economic g ow h.
A e conside ing hese ca ea s, his analysis uses wo dynamic h eshold models p oposed by
K eme , Bick, and Nau z (2013) and Seo and Shin (2016). The o me accommoda es he endogenei y o
eg esso s, which is welcome gi en ha mos o he d i e s o economic g ow h a e endogenously
de e mined. The la e accommoda es he endogenei y o bo h he eg esso s and he h eshold a i-
able. Repo ing esul s om bo h models allows he analysis o e alua e he sensi i i y o he esul s o
he exogenei y and endogenei y assump ions o he h eshold a iable. Collec i ely, compa ed o p e i-
ous app oaches such as he Hansen (1999) and Cane and Hansen (2004) models, hese wo models
ha e he ad an age o simul aneously accommoda ing nonlinea asymme ic dynamics, endogenei y,
and c oss-sec ional he e ogenei y (Seo & Shin, 2016). The p e e ed dynamic h eshold models a e
applied o a mo e ecen da ase , which makes he analysis capable o cap u ing sub-Saha an A ica’s
ecen expo , g ow h, and go e nmen deb dynamics.
The esul s show ha expo s a ec economic g ow h posi i ely and s ongly when go e nmen deb
is 44–45% o GDP and below. Abo e his h eshold, he impac o expo s on economic g ow h is posi-
i e bu weake . This esul is obus o he use o eal GDP g ow h and eal GDP pe capi a g ow h, as
well as he inclusion o con ol a iables. In policy senses, he esul sugges s ha go e nmen deb in
sub-Saha an A ica, which closed 2023 si ing a 61% o GDP, could be cons aining he possibili y o
expo -led g ow h in he egion.
In he second pa o he a icle, an a emp is made o es ablish he ex en o which go e nmen
e enue g ow h and go e nmen expendi u e cu s can educe go e nmen deb in he egion. The aim
is o isola e he e ec o each policy op ion on go e nmen deb while holding he o he cons an . The
Figu e 2. G ow h dynamics in sub-Saha an A ica 2004–2023.
Sou ce: Own compu a ion using da a om he IMF.
COGENT ECONOMICS & FINANCE 3
IMF (2023) ecen ly showed ha mos ( hough no all) coun ies will need o adjus hei p ima y balan-
ces in he egion o 2 o 3% o GDP o e he nex i e yea s o he median coun y, and much la ge in
a ew cases. Using a a ie y o es ima o s, his analysis inds he need o simila policy adjus men s i an
a e age sub-Saha an coun y is o lowe i s go e nmen deb o below 45%. The e idence sugges s ha
an a e age sub-Saha an coun y ough o cu he sha e o go e nmen expendi u e on GDP in he
egion o 2 o 15 pe cen age poin s o aise he sha e o go e nmen e enue (excluding g an s) in he
egion o 2 o 12 pe cen age poin s.
These esul s can be used o guide iscal policy o mula ion and deb es uc u ing in sub-Saha an
coun ies wi h high le els o deb , such as E i ea, Libe ia, Cabo Ve de, S~
ao Tom
e and P
ıncipe,
Seychelles, Mozambique, and Bu undi. They also imp o e ou unde s anding o how iscal ou comes
indi ec ly in luence he possibili y o expo -led g ow h in sub-Saha an A ica.
The es o he a icle is o ganized as ollows. Sec ion 2 e iews heo e ical and empi ical li e a u e.
Sec ion 3 ou lines he me hodology (da a desc ip ion and model speci ica ion). Sec ion 4 p esen s he
eg ession esul s. Sec ion 5 summa izes he analysis, p o ides policy implica ions, and sugges s a eas
o u he esea ch.
2. Li e a u e e iew
In heo y, go e nmen deb can in luence g ow h in he sho and long un h ough a ious complica ed
channels. A use ul s a ing poin is he con en ional agg ega e demand in which agg ega e demand (GDP) is
exp essed as a unc ion o consume spending (C), in es men (I), go e nmen expendi u e (G) and ne
expo s (X-M), ha is, GDP ¼CþIþGþX–M. In he sho un, is a gued ha go e nmen deb can s imula e
agg ega e demand h ough he Keynesian mul iplie e ec o expendi u e componen s. The long un e ec
can ei he be posi i e o nega i e. The nega i e e ec is gene ally explained by he a endan inc ease in
in e es a e which c owds ou p i a e in es men as discussed in Mille and Modigliani (1961), Gale and
O szag (2003) and Baldacci and Kuma (2010). On he o he hand, he posi i e impac may occu in coun ies
whe e go e nmen deb is spen on p oduc i e sec o s o he economy and capi al accumula ion. The use o
go e nmen deb o imp o e oads, ai po s and ene gy capaci y among o he s o example can acili a e eco-
nomic g ow h in he long un h ough enhancing p oduc i i y g ow h (Elmendo & Mankiw, 1999).
The posi i e p oduc i i y e ec o go e nmen deb alloca ed o capi al accumula ion may no be
gua an eed, howe e (Adam and Be an, 2005; Aizenman e al., 2007; Sain -Paul e al., 1996). High go -
e nmen deb may hampe long un g ow h e en when spen in p oduc i e sec o s o he economy.
The a gumen is ha high s ock o go e nmen deb may induce dis o iona y axa ion o pay he deb ,
which consequen ly lowe s u u e po en ial economic g ow h. In addi ion, high go e nmen deb may
ake away he abili y o go e nmen s o implemen coun e cyclical iscal policies which leads o highe
ola ili y and lowe long un g ow h (Aghion & Kha oubi, 2007; Woo, 2009). O he s ha e shown ha
high go e nmen deb may culmina e in a so e eign deb c isis which can uel cu ency c ises (Bu nside
e al., 2001; Hemming e al., 2003).
One a ea ha , un il ecen ly, has ha dly ecei ed a ai sha e o schola ly a en ion is he mode a ing
ole o go e nmen deb on expo -led g ow h. In he main, i has been a gued ha go e nmen s ha
gua an ee deb epaymen s on u u e expo ea nings may cons ain long un g ow h. This channel is
especially ue in coun ies such as Zimbabwe whe e go e nmen s ha e in place expo su ende poli-
cies which compel expo e s including p i a e playe s o pay a po ion o hei o eign cu ency ea nings
o go e nmen . This a gumen can be econciled wi h li e a u e p oposing a non-linea associa ion
be ween go e nmen deb and economic g ow h. This li e a u e includes Reinha and Rogo (2010),
Ndo icimpa (2020), Liu and Lyu (2021), Law e al. (2021), and Shah e al. (2024). The a gumen aised in
hese s udies is ha low le els o deb can enhance g ow h while high le els o deb (abo e an es i-
ma ed h eshold poin ) may nega i ely a ec g ow h.
The non-linea associa ion be ween go e nmen deb and economic g ow h aises wo heo e ical
possibili ies. One is ha i go e nmen deb a ec s economic g ow h h ough in luencing o he compo-
nen s o agg ega e demand such as in es men , household consump ion and ne expo s, hen he
impac o hese agg ega e demand componen s on economic g ow h may di e depending on he le -
els o go e nmen deb . The esul ing implica ion (which is he second possibili y) is ha ei he
4 B. T. MAZORODZE
go e nmen deb may ha e an in e ac i e e ec wi h o he de e minan s o economic g ow h o o he
g ow h de e minan s may be egime dependen , ha is, hei impac on economic g ow h may depend
on whe he deb is below o abo e an es ima ed h eshold. Fo expo s in pa icula , he in ui ion could
be ha below a deb h eshold, expo ea nings can ha e a la ge mul iplie e ec and go e nmen s
can use expo ea nings o inance in es men s. Abo e a h eshold poin , in e es paymen s a e highe
and much o expo p oceeds may be used o epay deb which diminishes he posi i e impac o
expo s on economic g ow h.
Empi ically, he abo e heo e ical discussion comes wi h he challenge o iden i ica ion since bo h
go e nmen deb and expo s a e mos likely endogenous. In addi ion, i has been a gued in he li e a-
u e ha g ow h in sub-Saha an A ica is cons ained by many o he ac o s which may be co ela ed
wi h go e nmen deb . An app oach used o mi iga e his endogenei y p oblem is o ea bo h he
h eshold a iable (go e nmen deb ) and expo s as endogenous while con olling o ele an mac o-
economic a iables o be e iden i ica ion. Using a dynamic panel h eshold model ha add esses
endogenei y his way, Ndo icimpa (2020) inds e idence suppo ing he egime dependence o nea ly all
d i e s o economic g ow h. The e idence in Ndo icimpa (2020) pa icula ly show ha he impac o
domes ic in es men , ade openness, o eign di ec in es men and e ms o ade depends on whe he
go e nmen deb is below o abo e 61–65% o GDP.
Simila ly, and mo e ecen ly, Shah e al. (2024) use a dynamic panel h eshold model and con i m
ha in la ion nega i ely a ec s he public deb -g ow h ela ionship abo e he h eshold le els o 14.03%
(whole sample), 15.101% (lowe middle-income economies), and 9.628% (uppe -middle-income econo-
mies), espec i ely. This inding has a i ue o suppo ing he no ion ha he impac o g ow h de e mi-
nan s may di e depending on whe he go e nmen is below o abo e an es ima ed h eshold le el.
In line wi h he abo e easoning, he cu en a icle ocuses on he impac o expo s. I empi ically
es s he hypo hesis ha expo s may ha e a ma ginal e ec on economic g ow h once go e nmen
deb exceeds an es ima ed h eshold le el. I imp o es exis ing s udies such as Michaely (1977), Donges
and Riedel (1977), Balassa (1978), Tyle (1981), Eme y (1967), Donges and Riedel (1977), Wea e (1993),
Hul man (1967), Fosu (1990), T o
s and Bojnec (2016) and mo e ecen ly S deli
c, and D
a ila-Fe n
andez
(2024) by ea ing expo s as a egime dependen a iable whose e ec on g ow h depends on whe he
go e nmen deb is below o abo e an es ima ed h eshold.
3. Ma e ials and me hods
This sec ion p esen s he me hodological se up o he a icle. The sec ion has wo main subsec ions. I
desc ibes he da a in ol ed and speci ies he empi ical models.
3.1 Da a desc ip ion
Guided by da a a ailabili y, he analysis uses an unbalanced panel da ase comp ising 44 sub-Saha an coun-
ies obse ed annually be ween 2004 and 2023.
3
The o al sample con ains 868 annual obse a ions. The
igh -hand side a iables o in e es a e expo s and go e nmen deb . Bo h a iables a e pe cen ages o GDP.
Economic g ow h is he dependen a iable, and i is p oxied by eal GDP g ow h and eal pe capi a GDP
g ow h. Da a on all a iables a e sou ced om he In e na ional Mone a y Fund (IMF). The me hodology is di -
ided in o wo pa s. The i s pa , sec ion 3.1.1, examines how he impac o expo s on economic g ow h is
in luenced by he le el o go e nmen deb . The second pa , sec ion 3.2.1, de e mines he ex en o which
go e nmen e enue g ow h and go e nmen expendi u e cu s educe go e nmen deb in sub-Saha an
A ica. The a iables used in his la e sec ion a e he e o e go e nmen deb , go e nmen expendi u e and
go e nmen e enue excluding g an s. All a iables a e pe cen ages o GDP.
3.1.1 Model speci ica ion
The baseline model is he panel dynamic h eshold model p oposed by K eme e al. (2013). I imp o es
Hansen’s(1999) speci ica ion by elaxing he exogenei y o eg esso s including lags o he dependen
a iable. Applying he o wa d o hogonal de ia ions ans o ma ion o A ellano and Bo e (1995) o ci -
cum en endogenei y in a dynamic se up, K eme e al.’s(2013) model combines he ins umen al
COGENT ECONOMICS & FINANCE 5
a iable es ima ion o Cane and Hansen (2004) wi h he panel h eshold model o Hansen (1999). The
model akes he ollowing o m.
GROWTHi, ¼uiþh1expo si I DEBTi c
ðÞ
þh2I DEBTi c
ðÞ
þh3expo si I DEBTi >c
ðÞ
þh4xi þei (1)
i¼1, :::,44; ¼2004, :::, 2023
whe e GROWTHi, deno es eal GDP g ow h and eal GDP pe capi a g ow h o coun y iin pe iod ,
expo s a e o al expo s as a pe cen age o GDP, DEBT is go e nmen deb as a pe cen age o GDP, xis
a ec o o con ol a iables including he lagged dependen a iable, uiis unobse ed he e ogenei y,
h1,:::,h4a e unknown pa ame e s o be es ima ed, cis he h eshold poin and ei is he e o e m. In his
speci ica ion, DEBT is he h eshold a iable, and expo s a e he only egime dependen a iable. Speci ied
his way, h1cap u es he e ec o expo s on economic g ow h when go e nmen deb is a mos cwhile
h3cap u es he impac o expo s on economic g ow h when go e nmen deb is abo e c:Con ol a ia-
bles a e assumed o be egime independen . This assump ion is elaxed in he obus ness exe cise unde
he dynamic h eshold model o Seo and Shin (2016). Following A ellano and Bo e (1995), go e nmen
deb , expo s and he lagged dependen a iable a e ins umen ed using hei own lags.
To es he obus ness o he esul s, he al e na i e dynamic model p oposed by Seo and Shin (2016)
is applied. In his obus ness check, bo h expo s and he con ol a iables a e assumed o be egime
dependen . The es ima ed dynamic panel h eshold model akes he ollowing o m:
GROWTHi, ¼1, x0
i
h11DEBTi c
ðÞ
þ1, x0
i
h21DEBTi >c
ðÞ
þei (2)
ei ¼aiþ i ,
i¼1, :::,45; ¼2004, :::, 2023,
whe e GROWTHi, is ei he annual eal GDP g ow h o annual eal GDP pe capi a g ow h o coun y i
and pe iod ,x0
i is a k11 ec o o ime- a ying eg esso s including GROWTHi, −1and a measu e o
expo in ensi y, 1(∙) is an indica o unc ion, DEBTi is he ansi ion a iable, cis he h eshold pa am-
e e , h1and h2a e egime-dependen pa ame e s o be es ima ed, and ei is an e o e m de ined as
he sum o unobse ed coun y he e ogenei y aiand idiosync a ic shocks, i :Since he dynamic e m
GROWTHi, −1and he unobse ed coun y he e ogenei y in i e he Nickell bias (Nickell, 1981) by design,
he i s di e ence ans o ma ion o A ellano and Bond (1991) is applied o ci cum en his esul ing
endogenei y. This ans o ma ion leads o he ollowing speci ica ion.
DGROWTHi, ¼d0Dxi þk0X0
i 1i ðcÞþDei (3)
whe e Dis he i s di e ence ope a o , dki1¼ðh12,:::,h1,k1þ1Þ0,kðki1Þ1¼h2−h1, and
Xi
2ð1þk1Þ¼ð1, x0
i Þ
1, x0
i −1
!
&1
i c
ðÞ
21¼1DEBTi >c
ðÞ
−1DEBTi −1>c
ðÞ
!
The i s di e ence ans o ma ion c ea es an addi ional co ela ion be ween he ans o med eg es-
so s and he e o e m. To ci cum en his endogenei y, l1 ec o o ins umen s, (z0
i 0,:::,z0
iT ) o 2 <
0Ta e equi ed, such ha ei he ,
Ez
0
i 0Dei 0,:::,z0
iT ,:::,z0
iT DeiT
0¼0,
o , o each ¼ 0,:::,T,
EDei jzi
ðÞ
¼0
Con ained in za e in e nal ins umen s namely he lagged alues o xi and DEBTi :To es o linea -
i y, he ollowing null hypo hesis is conside ed.
H0:k¼0, o any c2C
Agains he al e na i e hypo hesis,
H0:k6¼ 0, o some c2C
6 B. T. MAZORODZE
The es s a is ic is,
supW¼sup
c2C
WnðcÞ
whe e WnðcÞis he s anda d Wald s a is ic o each ixed ci.e.,
Wnc
ðÞ¼n^
kðcÞ0^
Rkc
ðÞ
−1^
kðcÞ
whe e ^
kðcÞis he es ima ed alue o kgi en he h eshold pa ame e ob ained om he FD-GMM, and
^
RkðcÞis he consis en asymp o ic a iance es ima o o ^
kðcÞ:Exogenei y o go e nmen deb is es ed
using a Hausman ype es de eloped in Seo and Shin (2014). The null hypo hesis is ha he h eshold
a iable, go e nmen deb , is exogenous.
When es ima ing he impac o go e nmen deb and expo s on economic g ow h, i is necessa y o con-
ol o hi d a iables ha may join ly a ec hese a iables. In inding hese a iables, cau ion should be
exe cised no o include a iables ha may be possible channels h ough which he igh -hand side a iables
o in e es may a ec economic g ow h. This easoning means ha he inclusion o in es men and human
capi al s ock a iables as p edic ed by he augmen ed Solow-g ow h model in he model may be coun e in-
ui i e as hese a iables a e likely o be mechanisms h ough which go e nmen deb a ec s economic
g ow h in he long- un. The analysis he e o e con ols a iables ha a e less likely o se e as deb and
expo channels and hey include sa ings (as a pe cen age o GDP), a 3-yea olling s anda d de ia ion o
e ms o ade p oxying ex e nal shocks,
4
eal e ec i e exchange a e, in la ion, and ne o eign di ec in es -
men . Sa ings a e expec ed o a ec economic g ow h posi i ely h ough spu ing in es men (Cesa a o,
1999). Te ms o ade shocks a e expec ed o ha e a nega i e e ec on economic g ow h h ough inc easing
unce ain y, mac oeconomic ins abili y, and dis o ing go e nmen e enues (Hadass & Williamson, 2003;
Mendoza, 1997). An app ecia ion o he eal e ec i e exchange a e is expec ed o educe economic g ow h
h ough acili a ing a ealloca ion o esou ces om he mo e p oduc i e adable sec o o he less p oduc -
i e non- adable sec o (Eicheng een, 2007;Elbadawie al.,2012). Low le els o in la ion may enhance eco-
nomic g ow h h ough g easing he economy while high le els o in la ion may inc ease mac oeconomic
unce ain y which dis o s in es men (Sa el, 1996). I s coe icien is he e o e expec ed o be nega i e o
posi i e depending on whe he sub-Saha an coun ies expe ienced high o low a es o in la ion du ing he
sampling pe iod. Las ly, al hough he impac o ne o eign di ec in es men is deba able in he li e a u e,
mos accoun s (Bo ensz ein e al., 1998; Mouda sou, 2003) epo a posi i e e ec on economic g ow h ha
e lec s echnological spillo e s and job c ea ion.
In he second pa , he analysis de e mines he ex en o which expendi u e cu s and go e nmen
e enue g ow h educe go e nmen deb in sub-Saha an A ica. This line o inqui y ollows Kollias e al.
(2004), Small e al. (2020) and Reis (2022) and he IMF (2023). Some o he key me hodological ca ea s
d awn om his li e a u e is ha he ela ionship is likely dependen on he ime ho izon i.e., sho un
and long un, po en ial he e ogenei y in deb dynamics, ea men o endogenei y since go e nmen
deb na u ally in luences go e nmen e enue and expendi u e, and he accommoda ion o c oss sec-
ional dependence gi en he likely p esence o common ac o s such as he 2009 Global Financial C ises,
COVID-19 and global commodi y p ice shocks which commonly a ec sub-Saha an coun ies. Wa y o
hese ca ea s, his analysis employs a a ie y o es ima ion app oaches ha add ess hese challenges.
These include he pooled mean g oup es ima o
5
o Pesa an e al. (1999), he ins umen al dynamic
common co ela ed e ec s o Di zen (2018), he panel dynamic o dina y leas squa es es ima o
6
o
Saikkonen (1991) and S ock and Wa son (1993) ex ended o panel da a by Kao and Chang, and he sys-
em gene alized me hod o momen s a e applied o A ellano and Bo e (1995) and Blundell and Bond
(1998). The PMG es ima o builds om he ollowing dynamic model.
DEBTi ¼X
p
j¼1
kijDEBTi, −jþX
q
j¼0
d0
ijxi, −jþliþei (4)
i¼1, :::,44; ¼2004, :::, 2023
whe e kij is a scala , xis a ec o o wo a iables, go e nmen e enue (excluding g an s) and go e n-
men expendi u e, dis he co esponding ec o o slope coe icien s o be es ima ed, licap u es ime-
in a ian he e ogenei y and ei is he idiosync a ic e m and DEBT is as de ined be o e. Equa ion (3) can
COGENT ECONOMICS & FINANCE 7
insigni ican unde he ixed e ec s app oach and posi i e bu ma ginally signi ican unde he GMM
echnique.
The lagged e ms o he dependen a iable a e all s a is ically signi ican sugges ing ha modelling
economic g ow h using s a ic models isks model unde i ing and an omi ed a iable bias. The nega-
i e sign associa ed wi h he lagged dependen a iable indica es ha a ise in economic g ow h in he
cu en yea is accompanied by a slowdown in economic g ow h in he subsequen yea . This esul con-
i ms he u bulen na u e o economic g ow h pa e ns in sub-Saha an A ica.
Table 3 shows he esul s o a di e en dynamic h eshold model p oposed in Seo and Shin (2014). In
his able, he symbols b and d a e used o indica e he e ec o a iables below and abo e a speci ied
h eshold le el. In his al e na i e model, he h eshold le el o go e nmen deb is 44% o GDP in all
speci ica ions. When go e nmen deb is equal o o less han 44% o GDP, expo s a e posi i ely and
signi ican ly ela ed o economic g ow h. A one pe cen age poin inc ease in expo s as a pe cen age o
GDP is es ima ed o inc ease economic g ow h by 17 o 37 basis poin s when go e nmen deb is below
45%. When go e nmen deb su passes 44% o GDP, he impac o expo s is posi i e bu s a is ically
insigni ican in he i s wo speci ica ions and s a is ically nega i e in he las wo speci ica ions. The las
wo speci ica ions indica e ha expo s expansion in ac educes economic g ow h when go e nmen
deb su passes 44% o GDP.
The boo s ap p obabili y alue is lowe han he 1% le el o signi icance indica ing ha go e nmen
deb non-linea ly in luences he impac o expo s on economic g ow h.
The main esul is ha expo s ha e a la ge impac on economic g ow h when go e nmen deb is
below 46% o GDP. Abo e his limi , expo s ha e a weake e ec on economic g ow h a bes . In he
nex se o esul s, he coun ies we e sepa a ed in o h ee g oups. In o med by he a e age alues o
go e nmen deb displayed ea lie in Figu e 4, he i s g oup comp ises coun ies in he bo om 5.
These coun ies a e E i ea, S~
ao Tom
e and P
ıncipe, Guinea-Bissau, and Cabo Ve de, Libe ia. The second
g oup comp ises coun ies in he op 5 and hey a e Bo swana, Equa o ial Guinea, Eswa ini, Nige ia and
Como os. The hi d g oup comp ises coun ies wi h mode a e a e age le els o deb , and hey a e
Gabon, Leso ho, Kenya, Sou h A ica and Senegal.
Table 2. Go e nmen deb , expo s and economic g ow h.
Dependen a iable
(1) (2) (3) (4)
Real GDP
g ow h (%)
Real GDP
g ow h (%)
Real pe capi a
GDP g ow h (%)
Real pe capi a
GDP g ow h (%)
Real GDP g ow h (%) (−1) −0.0951 −0.116 ––
(0.0306) (0.0311)
Real pe capi a GDP (%) (−1) ––−0.0991 −0.126
(0.0305) (0.0312)
below_ h es_b 0.151 0.145 0.136 0.132
(0.0236) (0.0226) (0.0230) (0.0221)
abo e_ h es_b 0.0445 0.0730 0.04080.0704
(0.0224) (0.0226) (0.0220) (0.0221)
Deb (% o GDP) −0.0036 −0.0053 −0.0050 −0.00717
(0.00012) (0.00128) (0.00147) (0.00113)
ToT_sd3 –−0.0252–−0.0248
(0.0150) (0.0147)
Sa ings (% o GDP) –0.111 –0.106
(0.0265) (0.0259)
lnREER –−2.538–−1.882
(1.427) (1.393)
FDI (% o GDP) –0.0363 –0.0339
(0.0136) (0.0133)
In la ion (%) –−0.0591 –−0.0569
(0.0116) (0.0113)
Cons an 1.672 11.78−0.704 6.328
(0.737) (6.642) (0.725) (6.490)
Th eshold deb 45% 44% 44% 44%
SupWS a es (p- alue) 0.0000 0.0001 0.0000 0.0031
Obse a ions 824 824 824 824
Numbe o coun ies 44 44 44 44
S anda d e o s in pa en heses.
p<0.01; p<0.05; p<0.1.
14 B. T. MAZORODZE
Each se has 5 coun ies which ansla es o a panel ime se ies in which T is conside ably smalle
han N. To add ess unobse ed he e ogenei y, he Nickel-Bias, and he po en ial endogenei y a ising
om he eedback e ec , and omi ed ime- a ying ac o s nes ed in he e o e m, he analysis imple-
men s he Ande son and Hsiao es ima o wi h a bias app oxima ion sugges ed by Bun and Ki ie (2003).
In hese es ima ions, he con ol a iables a e simila o hose epo ed in Table 3. Go e nmen deb
does no appea in hese es ima ions since he ca ego iza ion i sel g oups coun ies in o high-deb ,
mode a e and low-deb coun ies. The in en ion is o in e i he e ec o expo s on economic g ow h
is he e ogeneous ac oss he h ee g oups.
Figu e 6 epo s he slope coe icien s o he h ee go e nmen deb ca ego ies. The es ima ed coe -
icien o expo is highe in low-indeb ed coun ies (Bo swana, Equa o ial Guinea, Eswa ini, Nige ia and
Como os) han in high-indeb ed coun ies (E i ea, S~
ao Tom
e and P
ıncipe, Guinea-Bissau and Cape
Ve de, Libe ia). The inclusion o con ols and ime e ec s changes he size o he coe icien s, bu no
he o iginal esul ha expo s ha e a weake impac on g ow h in coun ies wi h high go e nmen
deb .
To u he examine he obus ness o his esul , es ima es a e conside ed o he coun y wi h he
lowes go e nmen deb (Bo swana), he middle coun y (Gabon), and he coun y wi h he highes
(E i ea). These es ima ions use al e na i e da ase s om he espec i e cen al banks and he Wo ld
De elopmen Indica o s (WDI) co e ing he pe iod 1980–2023. This ime, he ou come a iable is eal
pe capi a income in loga i hm o m (i.e., base 10). The independen a iable o in e es is s ill he pe -
cen age o expo s o goods and se ices on GDP. Fo each speci ica ion, s anda d ime se ies p oce-
du es we e ollowed. Va iables we e subjec ed o uni oo es s using he Augmen ed Dickey Fulle and
Table 3. Go e nmen deb , expo s and economic g ow h.
Dependen a iable
(1) (2) (3) (4)
Real GDP
g ow h (%)
Real GDP
g ow h (%)
Real pe capi a GDP
G ow h (%)
Real pe capi a
GDP G ow h (%)
Lag_y_b 0.470 −0.2150.0515−0.123
(0.0892) (0.126) (0.0302) (0.0691)
ToT_sd3_b –0.0264 –0.0451
(0.0060) (0.0995)
Sa ings (% o GDP)_b –0.00230 –−0.00485
(0.0982) (0.0610)
lnREER_b –−0.0989 –−0.0444
(0.0485) (0.0158)
In la ion (%)_b –−0.0531 –−0.0239
(0.128) (0.0537)
FDI (% o GDP)_b –0.205 –−0.00464
(0.246) (0.0441)
Expo s (% o GDP)_b 0.170 0.2370.368 0.371
(0.0495) (0.121) (0.0230) (0.0484)
cons_d −6.908 −2.124 4.888 −2.919
(2.301) (8.530) (0.647) (7.486)
Lag_y_d −0.564 0.0318 −0.0832 0.0952
(0.0908) (0.143) (0.0370) (0.132)
ToT_sd3_d –0.0969 –0.948
(0.0789) (0.181)
Sa ings (% o GDP)_d –0.141 –0.573
(0.120) (0.0988)
lnREER_d –−0.0785 –−0.0503
(0.0152) (0.0166)
In la ion (%)_d –−0.205 –−0.225
(0.147) (0.163)
FDI (% o GDP)_d –0.169 –0.240
(0.233) (0.104)
Expo s (% o GDP)_d 0.0432 0.0924 −0.215 −0.108
(0.0553) (0.113) (0.0156) (0.0417)
44.55 42.86 42.34 43.58
(6.995) (18.14) (9.293) (6.795)
Boo s ap p- alue o linea i y es 0.0000 0.0000 0.0000 0.0000
Numbe o coun ies 44 44 44 44
S anda d e o s in pa en heses.
p<0.01; p<0.05; p<0.1.
COGENT ECONOMICS & FINANCE 15
he Phillips Pe on es s. The gene al conclusion was ha all a iables we e I (1) p ocesses whose linea
combina ion was I (0).
7
Figu e 7 displays he es ima ed slope coe icien s o he h ee coun ies. The main esul s ill shows
up in ime se ies es ima ions. All he app oaches ( he dynamic o dina y leas squa es, he ully modi ied
o dina y leas squa es, he au o eg essi e dis ibu ed lag model and he coin eg a ing ec o unde he
Johansen app oach) indica e ha expo s ha e a la ge e ec on economic g ow h ( he log o eal
income pe capi a) in he case o Bo swana ( he low-deb coun y), a la ge e ec in Gabon (a mode a e
deb coun y) and a small e ec on economic g ow h in he highly indeb ed E i ea. This esul is bo h
com o ing and eassu ing.
In he nex s ep, he analysis de e mines he ex en o which inc easing go e nmen e enue and
educing public expendi u e can educe go e nmen deb in sub-Saha an A ica. The s a ing poin
in ol ed es ing o panel uni oo es s. Go e nmen deb , go e nmen expendi u e and go e nmen
Figu e 7. Time se ies es ima ions.
No es: The DOLS uses 1 lead and 1 lag au oma ically selec ed by he AIC. The Johansen coin eg a ing ec o uses 2 lags au oma ically selec ed
by he AIC om an ini ially es ima ed VAR and VECM models. The ARDL uses 2 lags o he dependen a iable and 1 lag o he independen
a iable o in e es . The models passed s anda d diagnos ic es s i.e., se ial co ela ion, esidual no mali y, model speci ica ion and he e osce-
das ici y. The VAR and VECM we e dynamically s able.
Figu e 6. Es ima ions o high, mode a e and low deb coun ies.
16 B. T. MAZORODZE
e enue we e all non-s a iona y p ocesses.
8
The non-s a iona i y o go e nmen deb c udely sugges s
ha go e nmen deb is, on a e age, unsus ainable in sub-Saha an A ica. This is consis en wi h he
expe iences o mos sub-Saha an coun ies. In he ea ly 2000s o example, mos sub-Saha an coun ies
wen in o deb o e hang as he median go e nmen deb - a io peaked a abou 80% o GDP. Upon eal-
iza ion o his non-mean e e ing endency o go e nmen deb , he in e na ional communi y in e -
ened h ough mul iple deb elie p og ams which included he Hea ily Indeb ed Poo Coun ies (HIPC)
ini ia i e.
F om he Ped oni coin eg a ion es , se en ou o ele en es s a is ics ejec ed he null hypo hesis o
no coin eg a ion a he 5% le el. This indica ed he p esence o a long un ela ionship be ween go e n-
men deb , go e nmen expendi u e and go e nmen e enue. Simila ly, bo h null hypo heses o no
coin eg a ion o some panels and no coin eg a ion o all panels we e ejec ed by he Wes e lund
app oach a he 5% le el sugges ing ha go e nmen deb , go e nmen expendi u e and go e nmen
e enue can be modelled as a coin eg a ing ela ionship.
Gi en he indica ions o a coin eg a ing ela ionship, he analysis conside ed he PMG and he panel
dynamic o dina y leas squa es me hod. These esul s we e supplemen ed by addi ional non-coin eg a ing
es ima o s namely he sys em GMM and he ins umen al a iable DCCE. The es ima ed esul s a e p e-
sen ed in Table 4. In he long un, all es ima o s con i m ha expendi u e cu s ha e a la ge e ec on cu -
ing deb compa ed o aising e enue. A pe cen age poin cu in go e nmen expendi u e is es ima ed o
educe go e nmen deb by 1.1–5.8 pe cen age poin s holding cons an go e nmen e enue. To keep
deb below 45% o GDP he e o e, an a e age sub-Saha an coun y he e o e needed o cu go e nmen
expendi u e by abou 2–10 pe cen age poin s
9
(6 on a e age i.e., 2 þ10/2) holding go e nmen e enue
cons an . Raising go e nmen e enue by one pe cen age poin educes go e nmen deb on impac by
0.8–4.5 pe cen age poin s holding cons an go e nmen expendi u e. To keep deb below 45% o GDP
he e o e, an a e age sub-Saha an coun y needed o al e na i ely aise go e nmen e enue by abou 2–
12 pe cen age poin s (7 on a e age i.e., 2 þ12/2).
The es ima ed esul s p esen ed in Table 4 show ha go e nmen deb in sub-Saha an A ica is p i-
ma ily d i en by inc eases in go e nmen expendi u e compa ed o dec eases in go e nmen e enue.
They also sugges ha cu ing go e nmen expendi u e could be a mo e e ec i e way o educing go -
e nmen deb in sub-Saha an A ica compa ed o aising e enue (which in mos cases equi es go e n-
men s o impose dis o iona y axes).
The speed o adjus men unde he PMG column is −0.139 indica ing ha only abou 14% o he dis-
equilib ium is co ec ed each yea when he go e nmen deb de ia es om i s long un equilib ium
Table 4. Go e nmen expendi u e, e enue, and deb .
Dependen a iable
Go e nmen deb (% o GDP)
PMG (1, 2) IV-DCC S-GMM DOLS (1) 1)
Expendi u e 5.778
(0.397)
2.319
(1.262)
1.0918
(0.045)
3.144
(0.355)
Re enue −4.499
(0.529)
−0.8115
(0.465)
−1.070
(0.062)
−1.489
(0.441)
Coin eq01 −0.139
(0.029)
–––
Deb (−1) –0.554
(0.081)
0.896
(0.011)
–
D (G. Deb (−1)) 0.206
(0.039)
–––
D(Expendi u e) 0.526
(0.243)
–––
D(Re enue) −0.823
(0.253)
–––
Obse a ions 780 736 824 797
Numbe o coun ies 44 44 44 44
A-B AR (1) 0.0013
A-B AR (2) 0.6036
Sa gan 0.9868
CD-s a is ic 45.99 −0.64
S anda d e o s in pa en heses. The PMG (1, 2) was selec ed by he AIC.
p<0.01; p<0.05; p<0.1.
COGENT ECONOMICS & FINANCE 17
le el. Consequen ly, i akes up o 7 yea s o go e nmen deb o e u n o i s equilib ium alue in he
e en o a sho - un shock. This slow speed o adjus men is consis en wi h he deb dynamics o mos
sub-Saha an coun ies. I is well known o example ha mos sub-Saha an coun ies ha e challenges
honou ing so e eign deb . The lagged dependen a iable is, in bo h he ins umen al a iable
DCCE and he sys em GMM, posi i e and s a is ically signi ican con i ming he expe iences o mos
sub-Saha an coun ies in which bo owing in cu en pe iods is gene ally ollowed by addi ional bo ow-
ing in subsequen yea s. The sho - un coe icien s unde he PMG a e conside ably smalle indica ing
ha e o s o cu go e nmen expendi u e o aise go e nmen e enue o educe deb a e mo e
ewa ding in he long un compa ed o he sho un.
An al e na i e app oach, he sys em o equa ions app oach, was pu sued o obus ness pu poses.
Compa ed o he single equa ion app oach pu sued in he p e ious ables, a sys em o equa ions cap-
u es he possible in e dependencies among he a iables. The analysis speci ically es ima ed a h ee-
a iable panel ec o au o eg ession using he o wa d o hogonal de ia ion ans o ma ion p oposed by
A ellano and Bond. In o med by uni oo s esul s discussed ea lie , all a iables we e ans o med in o
i s di e ences o ensu e dynamic s abili y. In addi ion, i is assumed ha he coe icien s associa ed
wi h go e nmen deb , go e nmen expendi u e, and go e nmen e enue in he sys em a e common
ac oss he 44 coun ies and ha unobse ed he e ogenei y is abso bed, in pa , by he i s di e ence
ans o ma ion. Since he coe icien s hemsel es a e o li le impo ance, he analysis p oceeds o p e-
sen esul s om he impulse esponse unc ions.
The o de ing assumes ha shocks in go e nmen e enue and go e nmen expendi u e ha e a con-
empo aneous e ec on go e nmen deb . This is easonable since go e nmen deb in sub-Saha an is
mos ly sensi i e o budge a y dynamics. By implica ion, go e nmen e enue and go e nmen expend-
i u e a e placed be o e go e nmen deb . Go e nmen e enue is placed be o e go e nmen expend-
i u e since he o me usually de e mines how much go e nmen s can spend. Panel g ange causali y
esul s a ached in appendix suppo his o de ing. Causali y is ound unning om go e nmen e enue
o go e nmen expendi u e and no he o he way ound. The e idence u he con i ms ha go e n-
men e enue and go e nmen expendi u e g ange cause go e nmen deb .
Figu e 8 displays he impulse esponse unc ions based on he Cholesky o de ing explained abo e.
Consis en wi h he esul s obse ed ea lie , a shock in expendi u e is accompanied by an immedia e spu
Figu e 8. Impulse esponse unc ions.
18 B. T. MAZORODZE
in go e nmen deb which peaks in he second pe iod, decays and anishes a e 6–7 yea s. This pe iod o
adjus men is consis en wi h he e o co ec ion e m o −0.14 in Table 4 which implied 7 yea s o adjus -
men in esponse o a sho - un shock. A shock in e enue is also associa ed wi h a dec ease in deb . This is
expec ed since e enue g ow h and iscal wind alls can be pa ially used o se ice deb . A isual inspec ion
o he g aph addi ionally con i ms he p io esul ha go e nmen deb is mo e sensi i e o changes in
go e nmen expendi u e dynamics han changes in go e nmen e enue.
The a iance decomposi ion o deb is p esen ed in Figu e 9. The impo ance o deb ’s own shocks
is excluded since in e es is on expendi u e and e enue. A cu so y look a he g aph is eassu ing.
Expendi u e cu s play a bigge ole on deb educ ions han e enue g ow h.
The impulse esponse unc ions om a Bayesian VAR wi h Minneso a de e mined p io s p oduced a
simila pa e n (see Figu e 10) which is u he com o ing. Changes in go e nmen expendi u e ha e a
la ge e ec on go e nmen deb han changes in e enue.
Figu e 10. Impulse esponse unc ions om a Bayesian VAR.
Figu e 9. Va iance decomposi ion.
COGENT ECONOMICS & FINANCE 19
O e all, he e idence sugges s ha high le els o go e nmen diminish he posi i e impac o expo s
on economic g ow h. Go e nmen s s uck in high deb , de ined as deb o GDP a io o 46% and abo e
a e less likely o expe ience obus economic g ow h om expanding he expo sec o . The la ge e ec
o go e nmen expendi u e on deb compa ed o he e ec o e enue g ow h is consis en wi h he
p ac ice o mos sub-Saha an go e nmen s and he his o ical policy ad ice om he IMF and he Wo ld
Bank. The IMF has mos ly ad ised indeb ed go e nmen s o exe cise iscal discipline and educe was e ul
iscal expendi u e.
Coun e ac ual scena ios in Figu e 11 conside h ee al e na i e policy scena ios. The i s , second and
hi d scena ios conside cu ing go e nmen deb (as a pe cen age o GDP) by 5 pe cen age poin s, 10 pe -
cen age poin s and 15 pe cen age poin s, espec i ely. The baseline scena io is he a e age deb o 55% o
GDP and go e nmen expendi u e o 24% o GDP. In his exe cise, a simple s uc u al equa ion is consid-
e ed in which go e nmen expendi u e is en i ely inanced by go e nmen e enue (excluding g an s).
Re enue om g an s is omi ed due o da a una ailabili y. This omission is less likely o ha e s ong conse-
quences since we a e no p ima ily in e es ed in in e p e ing he ma ginal e ec o e enue on go e nmen
expendi u e. The aim is o con ol o wha is plausibly go e nmen ’s main sou ce o e enue.
In gene al, go e nmen s decide o cu o aise expendi u e o a speci ic eason. The la e may be
encou aged by e enue wind alls ( a o able commodi y p ices). The o me may be mo i a ed by a
decline in e enue. In each o hese cases, depending on whe he he decision is o cu o educe
spending, he smalle (la ge ) shocks in go e nmen e enue would ha e been passed along om he
auxilia y eg ession (which eg esses go e nmen expendi u e on go e nmen e enue), o shocks in
deb due o he co ela ion be ween go e nmen expendi u e and go e nmen e enue. The new le el
o deb would be he di ec e ec o a change in go e nmen expendi u e in he deb equa ion plus
he magni ude o he shock in go e nmen e enue.
By conside ing he a o e men ioned policy scena ios, he analysis is essen ially p obing he e ec on
go e nmen deb i go e nmen expendi u e as a pe cen age o GDP (which is conside ed endogenous)
was exogenously cu by 5 pe cen age poin s, 10 pe cen age poin s and 15 pe cen age poin s.
Me hodologically, we canno accu a ely de e mine how deb will espond o changes in go e nmen
expendi u e wi hou i s p obing abou changes in go e nmen expendi u e holding cons an shocks in
go e nmen e enue. This is he p ima y eason why coun e ac ual scena ios a e conside ed o his
exe cise.
Figu e 11 in e es ingly shows ha he h ee policy scena ios will educe go e nmen deb om he
baseline scena io o 55% o GDP o 52.7% o GDP, 48.8% o GDP and 45% o GDP. I is he hi d policy
scena io (a go e nmen expendi u e cu o 15 pe cen age poin s) which akes deb down o he
Figu e 11. Go e nmen deb le els o di e en expendi u e cu s.
20 B. T. MAZORODZE
h eshold alue o 45% es ima ed ea lie . The 15 pe cen age poin s cu imply cu ing go e nmen
expendi u e om he baseline scena io ( he obse ed a e age sha e) o 24% o GDP o 9% o GDP.
In summa y, he e idence sugges s ha high go e nmen deb hinde s he impac o expo s on eco-
nomic g ow h in sub-Saha an A ica and ha educing deb will equi e cu ing go e nmen expendi u e
on GDP by 2–15 pe cen age poin s. Agg essi e expendi u e cu s, ha is, hose close o 15 pe cen age
poin s a e mo e likely o b ing go e nmen deb down o 45% o GDP.
5. Conclusion
F om he esul s p esen ed, i can be concluded ha high go e nmen deb is an obs acle o expo -led
g ow h in sub-Saha an A ica. The mos indeb ed go e nmen s be ween 2004 and 2023, such as Guinea-
Bissau, Libe ia, S~
ao Tom
e and P
ıncipe and Cape Ve de, saw a smalle con ibu ion o expo s o eco-
nomic g ow h compa ed o coun ies wi h less deb . This esul shows he impo ance o go e nmen
deb managemen as a complemen a y measu e o expo o ien a ion. The IMF’s Economic Ou look p e-
dic s ha sub-Saha an A ica’s annual deb will inc ease om 61 o 59% o GDP in 2023, su passing he
es ima ed le el o 44 o 45%. In his con ex , combined wi h he main esul o his a icle, sub-Saha an
coun ies wi h high deb a e u ged o pu sue iscal consolida ion o unlock expo -led g ow h.
Addi ional e idence sugges s ha go e nmen expendi u e cu s a e mo e e ec i e han aising e enue
in educing go e nmen deb . Fiscal audi s may need o iden i y unnecessa y bloa ed expendi u es may
be necessa y in his ega d. Public wage bills can be educed in a manne ha s ikes a balance be ween
cu ing bloa ed expendi u e and main aining se ice deli e y. La ge public in es men s wi h limi ed eco-
nomic e u ns can be downscaled o elimina ed en i ely. Fo u he s udy, his wo k lea es oom o
imp o emen o u u e esea ch. Fi s ly, i did no explici ly decompose he e ec s o deb on all pos-
sible channels o economic g ow h. I can be a gued ha deb may in ac p omo e g ow h i chan-
nelled owa ds p oduc i e sec o s o he economy. Ra he han ea ing g ow h d i e s as egime
dependen a iables, i migh be in e es ing o eg ess go e nmen deb on he key d i e s o economic
g ow h. In addi ion, he sample o sub-Saha an coun ies is he e ogenous. In some o he coun ies, go -
e nmen deb has been well managed, and expo s may ha e had a signi ican con ibu ion on eco-
nomic g ow h. This aises a possible u u e need o using quan ile eg essions o es ablish how he
dis ibu ional e ec s o deb on g ow h be ween poo and be e managed economies. Las ly, he
‘expo s-g ow h-go e nmen deb ’may be expanded o unde s and he dynamics in capi alis econo-
mies whe e expo s a e p ima ily d i en by he p i a e sec o and go e nmen s deb epaymen may be
independen o expo ea nings.
No es
1. Acco ding o he gap heo y, go e nmen s p e e a speci ic s uc u e o hei deb ma u i y. Agg ega e supply
changes coupled wi h he in es o s’p e e ence o long- e m deb can change he ela i e p ice be ween sho -
and long- e m deb . To lowe he expec ed unding cos s, go e nmen s a e p epa ed o adjus hei long- e m
deb supply. Rega ding he he e ogenei y in he use o deb , one could a gue o ins ance ha low deb could
be easy o manage and u ilize e icien ly compa ed o la ge s ock o deb . As documen ed in he li e a u e,
la ge s ock o deb may also in i e incidences o co up ion and abuse o go e nmen esou ces.
2. Cane and Hansen (2004) accommoda e he endogenei y o co a ia es bu s ill assume exogenei y o he
h eshold a iable, deb .
3. The 44 coun ies included in he analysis a e Angola, Benin, Bo swana, Bu kina Faso, Bu undi, Cabo Ve de,
Came oon, Cen al A ican Republic, Chad, Como os, Congo, Dem. Rep. o he, Congo, Republic o C^
o e d’I oi e,
Equa o ial Guinea, E i ea, Eswa ini, E hiopia, Gabon, Gambia, The, Ghana, Guinea, Guinea-Bissau, Kenya, Leso ho,
Libe ia, Madagasca , Malawi, Mali, Mau i ius, Mozambique, Namibia, Nige , Nige ia, Rwanda, Senegal, Seychelles,
Sie a Leone, Sou h A ica, Sou h Sudan, Republic o S~
ao Tom
e and P
ıncipe, Tanzania, Togo, Uganda and
Zambia.
4. Use o a 5-yea olling s anda d de ia ion o e ms o ade did no signi ican ly al e he main esul s.
5. The Hausman speci ica ion es p e e ed he pooled mean g oup es ima o o e he mean g oup es ima o and
he dynamic ixed model.
6. Panel DOLS is ully pa ame ic and possesses compu a ional con eniency compa ed o he al e na i e panel
‘ ully modi ied’OLS es ima o o Phillips and Moon (1999).
7. Tables o uni oo oo s a e a ailable upon eques . They a e no epo ed he e o he sake o b e i y.
COGENT ECONOMICS & FINANCE 21
8. Tables o uni oo oo s a e a ailable upon eques . They a e no epo ed he e o he sake o b e i y.
9. The es ima ed h eshold is 45%. The a e age le el o deb be ween 2003 and 2023 was 55%. B idging his gap
he e o e equi ed dec easing deb by abou 10 pe cen age poin s.
Acknowledgemen s
The au ho wishes o hank he academic edi o , Goodness Aye, he acili a o s o a w i ing e ea acili a ed by
SPU om which his pape bene i ed signi ican ly, P o esso Mishack Gumbo o he Uni e si y o Sou h A ica, D
O ilia Chi amba o he Uni e si y o Johannesbu g, D Paul Iwuanyanwu o No h-Wes Uni e si y and D Vanessa
Adesuwa Vanessa Agbedahin o Sol Plaa je Uni e si y. I am also g a e ul o he anonymous e iewe s o his manu-
sc ip o p o iding c i ical bu cons uc i e commen s.
Disclosu e s a emen
The au ho has no compe ing in e es s o decla e.
Funding
This esea ch was unded by Sol Plaa je Uni e si y’s Resea ch O ice.
Abou he au ho
B ian Mazo odze is an economis wi h o e ou yea s o esea ch expe ience, a Lindau Alumni and a ounding
membe o he A ican Resea ch and Policy Ad ocacy whose esea ch co e s in e na ional ade and economic
de elopmen . He is a o me consul an o he Uni e si y o KwaZulu-Na al's Mac oeconomic Resea ch Uni (MRU)
and a membe o he Ins i u e o Risk Managemen in Sou h A ica. He cu en ly eaches unde g adua e cou ses in
s a is ics and econome ics a Sol Plaa je Uni e si y and is a Depu y Chai o he School's Resea ch Commi ee.
ORCID
B ian Ta onga Mazo odze h p://o cid.o g/0000-0002-7799-4627
Da a a ailabili y s a emen
The da a ha suppo he indings o his s udy a e a ailable om he co esponding au ho , [BTM], upon eason-
able eques .
Re e ence lis
Adam, C. S., & Be an, D. L. (2005). Fiscal de ici s and g ow h in de eloping coun ies. Jou nal o Public Economics,
89(4), 571–597. h ps://doi.o g/10.1016/j.jpubeco.2004.02.006
Adedeji, A. A., Oyinlola, M. A., & Adeniyi, O. (2023). Public deb , ax and economic g ow h in Sub-Saha an A ican
coun ies. Jou nal o Social and Economic De elopmen ,45,1–67. h ps://doi.o g/10.1007/s40847-023-00295-4
Aghion, P., & Kha oubi, E. (2007). Cyclical mac o policy and indus y g ow h: he e ec o coun e cyclical iscal policy
(pp. 1–29). WP Ha a d Uni e si y.
Aizenman, J., Kle ze , K., & Pin o, B. (2007). Economic g ow h wi h cons ain s on ax e enues and public deb : implica-
ions o iscal policy and c oss-coun y di e ences. Na ional Bu eau o Economic Resea ch. h ps://doi.o g/10.3386/
w12750
A ellano, M., & Bond, S. (1991). Some es s o speci ica ion o panel da a: Mon e Ca lo e idence and an applica ion
o employmen equa ions. The Re iew o Economic S udies,58(2), 277–297. h ps://doi.o g/10.2307/2297968
A ellano, M., & Bo e , O. (1995). Ano he look a he ins umen al a iable es ima ion o e o -componen s models.
Jou nal o Econome ics,68(1), 29–51. h ps://doi.o g/10.1016/0304-4076(94)01642-D
Balassa, B. (1978). Expo s and economic g ow h: u he e idence. Jou nal o De elopmen Economics,5(2), 181–189.
h ps://doi.o g/10.1016/0304-3878(78)90006-8
Baldacci, M. E., & Kuma , M. M. S. (2010). Fiscal de ici s, public deb , and so e eign bond yields. In e na ional Mone a y
Fund.
Beke e, B., & Be sisa, M. (2018). Impac o o eign di ec in es men on economic g ow h in Eas e n A ica. In:
De e minan s o economic g ow h in A ica (pp. 95–124). Palg a e Macmillan.
22 B. T. MAZORODZE
Blundell, R., & Bond, S. (1998). Ini ial condi ions and momen es ic ions in dynamic panel da a models. Jou nal o
Econome ics,87(1), 115–143. h ps://doi.o g/10.1016/S0304-4076(98)00009-8
Bo ensz ein, E., De G ego io, J., & Lee, J. W. (1998). How does o eign di ec in es men a ec economic g ow h?
Jou nal o In e na ional Economics,45(1), 115–135. h ps://doi.o g/10.1016/S0022-1996(97)00033-0
Boza li, O., Bal, H., & Albay ak, M. (2023). Tes ing he expo -led g ow h hypo hesis in Tu key: New e idence om
ime and equency domain causali y app oaches. The Jou nal o In e na ional T ade & Economic De elopmen ,
32(6), 835–853. h ps://doi.o g/10.1080/09638199.2022.2144932
Bun, M. J., & Ki ie , J. F. (2003). On he diminishing e u ns o highe -o de e ms in asymp o ic expansions o bias.
Economics Le e s,79(2), 145–152. h ps://doi.o g/10.1016/S0165-1765(02)00299-9
Bu nside, C., Eichenbaum, M., & Rebelo, S. (2001). Hedging and inancial agili y in ixed exchange a e egimes.
Eu opean Economic Re iew,45(7), 1151–1193. h ps://doi.o g/10.1016/S0014-2921(01)00090-3
Cane , M., & Hansen, B. E. (2004). Ins umen al a iable es ima ion o a h eshold model. Econome ic Theo y,20(05),
813–843. h ps://doi.o g/10.1017/S0266466604205011
Cesa a o, S. (1999). Sa ings and economic g ow h in neoclassical heo y. Camb idge Jou nal o Economics,23(6),
771–793. h ps://doi.o g/10.1093/cje/23.6.771
Choi, I. (2001). Uni oo es s o panel da a. Jou nal o In e na ional Money and Finance,20(2), 249–272. h ps://doi.
o g/10.1016/S0261-5606(00)00048-6
Chudik, A., & Pesa an, M. H. (2015). Common co ela ed e ec s es ima ion o he e ogeneous dynamic panel da a
models wi h weakly exogenous eg esso s. Jou nal o Econome ics,188(2), 393–420. h ps://doi.o g/10.1016/j.jeco-
nom.2015.03.007
Di zen, J. (2018). Es ima ing dynamic common-co ela ed e ec s in S a a. The S a a Jou nal: P omo ing
Communica ions on S a is ics and S a a,18(3), 585–617. h ps://doi.o g/10.1177/1536867X1801800306
Donges, J. B., & Riedel, J. (1977). The expansion o manu ac u ed expo s in de eloping coun ies: An empi ical
assessmen o supply and demand issues. Re iew o Wo ld Economics,113(1), 58–87. h ps://doi.o g/10.1007/
BF02696566
Eicheng een, B. (2007). The eal exchange a e and economic g ow h. Social and Economic S udies,56(4), 7–20.
Elbadawi, I. A., Kal ani, L., & So o, R. (2012). Aid, eal exchange a e misalignmen , and economic g ow h in Sub-
Saha an A ica. Wo ld De elopmen ,40(4), 681–700. h ps://doi.o g/10.1016/j.wo ldde .2011.09.012
Elmendo , D. W., & Mankiw, N. G. (1999). Go e nmen deb . In Handbook o mac oeconomics ( ol. 1, pp. 1615–1669).
Eme y, R. F. (1967). The ela ion o expo s and economic g ow h. Kyklos,20(4), 470–486. h ps://doi.o g/10.1111/j.
1467-6435.1967. b00859.x
E e ae , G., & De G oo e, T. (2016). Common co ela ed e ec s es ima ion o dynamic panels wi h c oss-sec ional
dependence. Econome ic Re iews,35(3), 428–463. h ps://doi.o g/10.1080/07474938.2014.966635
Fede , G. (1983). On expo s and economic g ow h. Jou nal o De elopmen Economics,12(1–2), 59–73. h ps://doi.
o g/10.1016/0304-3878(83)90031-7
Fosu, A. K. (1990). Expo s and economic g ow h: The A ican case. Wo ld De elopmen ,18(6), 831–835. h ps://doi.
o g/10.1016/0305-750X(90)90005-I
Gale, W. G., & O szag, P. R. (2003). Economic e ec s o sus ained budge de ici s. Na ional Tax Jou nal,56(3), 463–
485. 10.17310/n j.2003.3.02
Hadass, Y. S., & Williamson, J. G. (2003). Te ms-o - ade shocks and economic pe o mance, 1870–1940: P ebisch and
Singe e isi ed. Economic De elopmen and Cul u al Change,51(3), 629–656. h ps://doi.o g/10.1086/375259
Hansen, B. E. (1999). Th eshold e ec s in non-dynamic panels: Es ima ion, es ing, and in e ence. Jou nal o
Econome ics,93(2), 345–368. h ps://doi.o g/10.1016/S0304-4076(99)00025-1
Hemming, R., Kell, M., & Schimmelp ennig, A. (2003). Fiscal ulne abili y and inancial c ises in eme ging ma ke
economies. In Fiscal Vulne abili y and Financial C ises in Eme ging Ma ke Economies. In e na ional Mone a y Fund.
Hul man, C. W. (1967). Expo s and economic g ow h: a su ey. Land Economics,43(2), 148–157. h ps://doi.o g/10.
2307/3145238
Im, K. S., Pesa an, M. H., & Shin, Y. (2003). Tes ing o uni oo s in he e ogeneous panels. Jou nal o Econome ics,
115(1), 53–74. h ps://doi.o g/10.1016/S0304-4076(03)00092-7
In e na ional Mone a y Fund. (2023). Na iga ing Fiscal Challenges in Sub-Saha an A ica: Resilien S a egies and
C edible Ancho s in Tu bulen Wa e s. Depa men al Pape No 2023/007.
In e na ional Mone a y Fund. (2024). Economic Ou look. Sub Saha an A ica. A Tepid and P icey Reco e y.
Islam, M. S. (2023). Does an expo -led g ow h p oposi ion exis o Bangladesh’s eady-made ga men s sec o ? A
nonlinea ARDL app oach. Regional Science Policy & P ac ice,15(5), 939–956. h ps://doi.o g/10.1111/ sp3.12557
Juh
asz, R., Lane, N., & Rod ik, D. (2023). The new economics o indus ial policy. Annual Re iew o Economics,16,
213–242.
Kalai zi, A. S., & Chambe lain, T. W. (2020). Me chandise expo s and economic g ow h: Mul i a ia e ime se ies ana-
lysis o he Uni ed A ab Emi a es. Jou nal o Applied Economics,23(1), 163–182. h ps://doi.o g/10.1080/15140326.
2020.1722384
Kollias, C., Manolas, G., & Paleologouc, S. M. (2004). Mili a y expendi u e and go e nmen deb in G eece: Some p elimin-
a y empi ical indings. De ence and Peace Economics,15(2), 189–197. h ps://doi.o g/10.1080/1024269032000110559
COGENT ECONOMICS & FINANCE 23
