Schaub, Max
A icle — Published Ve sion
Violen con lic and he demand o heal hca e: How
a med con lic educes us , ins ills ea , and inc eases
child mo ali y
Social Science & Medicine
P o ided in Coope a ion wi h:
WZB Be lin Social Science Cen e
Sugges ed Ci a ion: Schaub, Max (2024) : Violen con lic and he demand o heal hca e: How a med
con lic educes us , ins ills ea , and inc eases child mo ali y, Social Science & Medicine, ISSN
1873-5347, Else ie , Ams e dam, Vol. 359, pp. 1-10,
h ps://doi.o g/10.1016/j.socscimed.2024.117252
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/312201
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Con en s lis s a ailable a ScienceDi ec
Social Science & Medicine
jou nal homepage: www.else ie .com/loca e/ssm
Violen con lic and he demand o heal hca e: How a med con lic educes
us , ins ills ea , and inc eases child mo ali y✩
Max Schaub
Uni e si y o Hambu g and WZB Be lin Social Science Cen e , Ge many
ARTICLE INFO
Da ase link:h ps://doi.o g/10.7910/DVN/0B
UKZK
Keywo ds:
A med con lic
Child heal h
Heal hca e-seeking
Vaccina ion
Fea
T us
A ica
ABSTRACT
Wha a e he heal h e ec s o iolen con lic ? I is well known ha wa s kill ci ilians away om he ba le ield
and long a e he igh ing has s opped. Ye why his happens emains only pa ially unde s ood. While we ha e
good e idence ha ac o s such as he des uc ion o in as uc u e, poli ical neglec , and he ou -mig a ion o
heal h wo ke s – wha may be called supply-side ac o s – nega i ely a ec heal h ou comes, we know much
less abou how iolence shapes he a i udes and beha io owa ds heal hca e use among ci ilians exposed o
iolen con lic – wha may be called demand-side ac o s. He e, I heo ize ha exposu e o iolence supp esses
ci ilian demand o heal hca e h ough wo media ing channels – mis us o go e nmen ins i u ions and
ea o u u e iolence – wi h ad e se consequences o heal h ou comes, pa icula ly child heal h. To es
his heo y empi ically, I combine in o ma ion om o e 80,000 in e iews conduc ed in 22 con lic -a ec ed
coun ies in A ica wi h indi idual- and con ex -le el measu es o exposu e o iolen con lic . Exposu e o
iolence is associa ed wi h signi ican ly lowe le els o poli ical us and inc eased ea o u u e iolence,
which in u n p edic lowe heal hca e u iliza ion, lowe immuniza ion a es, and highe in an and child
mo ali y. To ully add ess he heal h consequences o a med con lic , i is essen ial ha we be e unde s and
he a i udinal and beha io al co ela es o exposu e o iolence.
In oduc ion
Wha a e he indi ec heal h e ec s o iolen con lic ? I is well
known ha wa s kill and maim ci ilians yea s a e he iolence has
ceased and in a eas a om he ba le ield (Ghoba ah e al.,2003,
2004;Iqbal,2006;Ga es e al.,2012;Kes e nich e al.,2014;Akbulu -
Yuksel,2017;Wagne e al.,2018). Howe e , why exac ly his is he
case emains only pa ially unde s ood. Schola s so a ha e ended
o ocus on s uc u al ac o s and he physical challenges o p o iding
heal hca e in he mids o con lic : Des oyed heal h in as uc u e
means ha popula ions a e poo ly se ed by essen ial medical se ices,
hampe ing disease con ol and pe ina al ca e (Iqbal and Zo n,2010;
Ki schne and Fina e ,2021). Landmines and unexploded o dnance
con inue o pose a h ea , especially o child en and ag icul u al wo k-
e s, yea s a e ac i e igh ing has ceased (Lin,2022). The exodus
o heal h wo ke s lea es con lic a eas unde supplied wi h essen ial
heal h se ices (McKay,1998;Chi e al.,2015;Chukwuma and Ekha o -
Mobayode,2019). Wha hese – clea ly impo an – ac o s ha e in
common is ha hey ocus on he supply-side o heal h. In his pape , I
✩Max Schaub is Assis an P o esso o Poli ical Science, esp. In e na ional Rela ions and Global Heal h a he Uni e si y o Hambu g, Allende-Pla z 1, 20146
Hambu g, Ge many. Fo commen s and eedback, I would like o hank Ca l M "ulle -C epon and semina pa icipan s a EPSA 2023, he Hambu g Cen e o
Heal h Economics, and he Uni e si y o Hambu g. I am also g a e ul o he e y cons uc i e commen s by wo anonymous e iewe s. Lenna Kasse a p o ided
ou s anding esea ch assis ance. This s udy ecei ed inancial suppo unde he Uni e si y o Hambu g’s Close- he-Gap p og am, g an numbe 23-CG-2021.
E-mail add ess: [email p o ec ed].
ins ead ocus on he a i udes and beha io o he ci ilians a ec ed by
iolence, and how hese shape hei demand o heal hca e se ices. I
a gue ha by inducing ea o u u e ic imiza ion and dis us o go -
e nmen ins i u ions, exposu e o iolence leads o beha io al changes.
People become eluc an o seek ou medical se ices o hemsel es o
hei child en, and o ha e hei child en accina ed, wi h de imen al
consequences o child heal h.
Schola s ha e shown ha he expe ience o iolence sys ema ically
goes along wi h lowe ed le els o in e pe sonal us and us in s a e
ins i u ions (Rohne e al.,2013;Hage e al.,2019;Conzo and Salus i,
2019). A he same ime, us is an essen ial p edic o o medical
ou comes. Dis us ul indi iduals a e less likely o seek and ollow
medical ad ice, isi hospi als, and ge hei child en accina ed (Vinck
e al.,2019;S oop e al.,2021;Lin,2022). The a gumen pu o wa d in
his pape ollows om he conca ena ion o hese a gumen s: Violen
con lic unde mines us and ins ills ea o u u e iolence, leading o
educed use o heal h se ices and lowe accina ion a es, which a e
associa ed wi h highe mo ali y, especially among he mos ulne able
h ps://doi.o g/10.1016/j.socscimed.2024.117252
Recei ed 31 Ma ch 2024; Recei ed in e ised o m 4 Augus 2024; Accep ed 13 Augus 2024
Social Science & Medicine 359 (2024) 117252
A ailable online 2 Sep embe 2024
0277-9536/© 2024 The Au ho (s). Published by Else ie L d. This is an open access a icle unde he CC BY license ( h p://c ea i ecommons.o g/licenses/by/4.0/ ).
M. Schaub
– young child en. Exposu e o iolence is also known o induce auma,
o en in he o m o pos - auma ic s ess synd ome (PTSD), which
has been linked wi h a oidance beha io (e.g., Obilom and Thache ,
2008;Neugebaue e al.,2009). Those exposed o en shy away om
socializing and public places, which, I a gue, may make hem eluc an
o seek heal hca e o hei child en, again esul ing in ele a ed a es
o in an and child mo ali y.
To demons a e his a gumen empi ically, I use in o ma ion om
o e 80,000 in e iews om 22 coun ies in A ica ha I analyze
by a a ie y o analy ical app oaches, including s uc u al equa ion
modeling, ixed e ec models, and media ion analysis. To es ablish
ha he ela ionship is plausibly causal, I es ima e wo-s age di e -
ences in di e ences models (Ga dne e al.,2023). Indi idual-le el
da a on expe iences o iolence, us , ea , and heal hca e-seeking
beha io come om A oba ome e (2018) su eys; con ex -le el da a
on iolence, accine up ake, and in an mo ali y om he Uppsala
Con lic Da a P og am (UCDP) and he Ins i u e o Heal h Me ics
and E alua ion (IHME), espec i ely. The esul s show ha exposu e
o iolence – bo h a he indi idual and con ex -le el – is associa ed
wi h signi ican ly educed le els o us and inc eased ea o iolence,
which hen p edic s lowe a es o DPT (diph he ia-pe ussis- e anus)
accina ion. Lowe accine up ake, in u n, di ec ly ansla es in o
highe a es o in an mo ali y. To he bes o my knowledge, his is one
o he i s s udies o ocus speci ically on he demand o heal hca e
among popula ions a ec ed by iolen con lic – and o di ec ly link
he expe ience o iolence o child heal h h ough he channels o
us and ea o u u e ic imiza ion. The indings sugges ha despi e
p og ess made in accoun ing o indi ec e ec s o iolen con lic s,
we may s ill be unde es ima ing he ue cos o wa . The indings
also sugges ha mid- and pos -con lic in e en ions should include
indi idual-le el componen s ha explici ly aim o es o e us and
add ess auma and ea . As will be demons a ed, hese measu es a e
no only impo an in hei own igh bu will also help imp o e child
wel a e.
1. Li e a u e and heo y
The de imen al consequences o iolen con lic on he heal h o
ci ilians no di ec ly in ol ed in he igh ing ha e long been no ed
by schola s (McKay,1998;Ghoba ah e al.,2003,2004;Plümpe and
Neumaye ,2006;Iqbal,2006;Kes e nich e al.,2014;Akbulu -Yuksel,
2017;Wagne e al.,2018,2019;Benda id e al.,2021). A pa icula
ocus has been on he heal h o young child en and women. Wagne
e al. (2018) p o ide nume ical es ima es. Thei s udy shows ha
be ween 1995 and 2015, he dea hs o 5 million child en unde he
age o 5 we e di ec ly o indi ec ly ela ed o a med con lic . This was
i e imes he o al numbe o di ec dea hs om a med con lic du ing
he same pe iod.
Mos schola ship o-da e has linked hese s agge ing igu es o he
des uc ion o c i ical in as uc u e and o he physical legacies o
iolen con lic . The des uc ion o non-main enance o sani a y in as-
uc u e acili a es he sp ead o disease, which may hen be ca ied
o o he a eas by e ugees (Iqbal,2006;G undy and Biggs,2018;
Hi sch eld e al.,2020;Ki schne and Fina e ,2021). Des oyed oads
complica e access o i al heal h se ices and he anspo o ood
and medicine (Ghoba ah e al.,2003;Plümpe and Neumaye ,2006;
Ki schne and Fina e ,2021). O he long- e m dange s esul om
unexploded o dnance, including clus e sub-ammuni ion and landmines
ha pose a h ea pa icula ly o child en and people wo king in
ag icul u e (Lin,2022). Ano he well-s udied phenomenon is he ou -
mig a ion o heal hca e wo ke s (McKay,1998;Sha a a and Kanj,
2014;P ice and Boha a,2013;Chi e al.,2015). Wo ied abou hei
own sa e y (and o en in high demand in mo e peace ul egions)
heal h wo ke s may be he i s o lea e con lic -s icken a eas. Di ec
a acks on heal h wo ke s, which ha e become mo e common in e-
cen yea s (Mahase,2022;Daniel,2023;Sau e ,2024), only spu his
p ocess on. As shown by Chi e al. (2015), he absence o heal hca e
wo ke s di ec ly ansla es in o wo se supply o pe ina al ca e and,
plausibly, in an mo ali y.
A cha ac e is ic sha ed by hese impo an s udies is ha hey ocus
on supply-side ac o s – ac o s ela ed o he heal h- ela ed in as uc-
u e and he heal h se ices a ailable o hose who seek hem. Violen
con lic comp omises hese supply-side ac o s, o en causing las ing
damage o he heal h o ci ilians. A dimension ha has seen much less
a en ion is he e ec o iolen con lic on heal h-seeking beha io (see
Adeyanju e al.,2024, o a e iew on his opic). In o he wo ds, we
know e y li le abou he ex en o which he demand o heal hca e
is a ec ed by iolen con lic .
This p oblem is exempli ied by a s udy by Meie ieks and Schaub
(2024), who es ima e ha each yea a ound 40,000 child en unde 5
die as a consequence o e o ism in A ica. This is despi e he ac
ha e o is a acks a ely a ge child en and do no cause he la ge-
scale physical des uc ion associa ed wi h o he ypes o a med con lic .
The au ho s a gue ha he nega i e heal h e ec s mus he e o e be
due o beha io al changes. Rela ed indings demons a e changes in
heal hca e-seeking beha io in esponse o con lic bu do no ypically
explo e mechanisms such as educed us o ea (P ice and Boha a
2013,Chi e al. 2015,Chukwuma and Ekha o -Mobayode 2019,D ue z
e al. 2020; see Adeyanju e al. 2024 o a e iew). Fo example, Chuk-
wuma and Ekha o -Mobayode (2019) show ha women exposed o he
Boko Ha am insu gency in Nige ia access ewe pe ina al se ices, a -
guably con ibu ing o excep ionally high neona al and in an mo ali y
a es in he no heas o Nige ia, he epicen e o he insu gency. How-
e e , he easons o his change in beha io emain unclea , leading
he au ho s o call o an explo a ion o mechanisms linking exposu e o
iolence o educed heal hca e-seeking. An impo an ad ance on his
ques ion is made by Tapsoba (2023), who sugges s ha wha people
espond o is he h ea o con lic and he ea o i . Simila ly, I he e
p opose wo mechanisms: a loss o us in s a e ins i u ions and ea o
u u e iolence.
Mis us in s a e ins i u ions. While a ich li e a u e shows how pa e ns
o in e pe sonal us change among ic ims o iolence, he na u e
o he change emains con es ed (Bellows and Miguel,2009;Voo s
e al.,2012;Cassa e al.,2013;Rohne e al.,2013;Baue e al.,
2016;Hage e al.,2019). While one s and o he li e a u e ends o
desc ibe a posi i e associa ion be ween expe iences o iolence and in-
e pe sonal us – a phenomenon explained wi h e e ence o so-called
pos auma ic g ow h (Tedeschi and Calhoun,2004) – mos ecen
wo ks end o ind a nega i e associa ion. T auma, mos schola ship
indica es, unde mines he belie ha o he people will no ha m you
i ulne able. Alongside in e pe sonal us , us in s a e ins i u ions
is unde mined. Resea ch consis en ly shows ha indi iduals who ha e
been ic imized ha e less us in he s a e (G osjean,2014;Voo s and
Bul e,2014;De Juan and Pie skalla,2016;Ga es and Jus esen,2020) –
an e ec ha can be ex emely las ing, some imes endu ing o e se e al
gene a ions (Nunn and Wan chekon,2011;Conzo and Salus i,2019).
The (implici ) explana ion is ha ic imiza ion in o ms indi iduals ha
he s a e ailed o ul ill i s ole as p o ec o , leading o a loss o us .
Impo an ly, he unde mined us seems o apply no only o secu i y
o ces, he police, and he cou s o law – i.e., ins i u ions di ec ly
asked wi h ci izens’ secu i y – bu also ex ends o s a e ins i u ions
such as he na ional p esiden and pa liamen , which bea no di ec
esponsibili y o indi idual episodes o iolence. I is he e o e a
leas plausible ha mis us in s a e ins i u ions also ex ends o heal h
se ices, such as go e nmen hospi als.
Mis us ma e s o heal h ou comes because i p e en s bo h di ec
con ac and belie upda ing. Dis us ul indi iduals a oid con ac wi h
medical p o essionals, which means ha hey may no seek medical
ca e o hemsel es o hei child en, o hey may do so a a la e s age
when a disease has al eady p og essed and he chances o success ul
ea men a e diminished (Alsan and Wanamake ,2018). Dis us o
Social Science & Medicine 359 (2024) 117252
2
M. Schaub
go e nmen ins i u ions may also mean ha o icial messages become
less e ec i e because ecipien s s op us ing he messenge . Impo an
o child en’s heal h, dis us in go e nmen ins i u ions can educe
p e en i e ca e, especially accina ion. Dis us o go e nmen is one o
he s onges p edic o s o compliance wi h s anda d childhood acci-
na ion ecommenda ions (S oop e al.,2021). Vaccina ion p o ides one
o he bes p o ec ions agains many se ious diseases ha con ibu e o
in an and child mo ali y. The e o e, he speed wi h which child en
a e p esen ed o heal h wo ke s is impo an (B eiman e al.,2004).
Vaccina ion is pa icula ly impo an in esou ce-poo se ings whe e
o he o ms o heal hca e a e no always eadily a ailable (Jones e al.,
2003). I people mis us go e nmen ins i u ions and o icials, hey
may shy away om con ac ing heal h p o essionals o ha e hei chil-
d en accina ed and/o s op lis ening o o icial go e nmen messages
abou he impo ance o accina ion.
Mis us can be pe sis en because i ep esen s a s able equilib ium.
As indi iduals do no seek help, hey also ha e li le oppo uni y o
posi i ely upda e hei (nega i e) p io s wi h ega d o s a e-indi idual
in e ac ions – simply because hey do no in e ac wi h he medi-
cal sys em a all. This is exempli ied by esea ch on he long- e m
heal h e ec s o he Tuskegee Syphilis S udy, whe e Black su e e s
o syphilis we e le un ea ed despi e he exis ence o a cu e (Alsan
and Wanamake ,2018). The au ho s show ha Black men who lea ned
abou his abuse consequen ly a oided using he medical sys em and
su e ed wo se medical ou comes han hose whose us was no
comp omised. In e es ingly, he same e ec was no ound among
women. The au ho s explain ha mos women come in o con ac
wi h he medical sys em when gi ing bi h. This ypically posi i e
expe ience gi es hem an oppo uni y o e ise hei nega i e p io
abou he medical sys em. As a esul , hei heal hca e u iliza ion is
no di e en om women no exposed o he Tuskegee s udy. This
example shows how ex e nal e en s can ha m us , and how his us
is only ebuil i indi iduals a e ‘ o ced’ in o a si ua ion whe e hey can
upda e hei p io s. In Nige ia, quali a i e wo k has singled ou us
in s a e ins i u ions as an impo an ac o o explaining accina ion-
seeking beha io . In he ea ly 2000s, a polio accina ion campaign
came o an ab up hal because umo s sp ead ha he Nige ian heal h
se ice and i s in e na ional pa ne s we e ying o s e ilize women
by means o he injec ion. The umo s only s opped a e eligious
au ho i ies in e ened (Obada e,2005;Jegede,2007). Simila ly, du ing
Ebola pandemics in cen al and wes e n A ica, compliance wi h con-
ainmen measu es was p ecipi a ed by us (Pe he ick,2015;Vinck
e al.,2019). Medical an h opologis s ha e poin ed ou ha in some
a eas, compliance was pa icula ly low because communi ies ea ed
he medical pe sonnel. In hei minds, hei beha io esembled ha
o a med g oups du ing he ecen ci il wa s in he egion (Wilkinson
and Fai head,2017).
T auma and ea . A second channel po en ially linking exposu e o
iolence o heal hca e-seeking beha io is auma and ea . As demon-
s a ed by nume ous s udies, ic ims o iolence a e egula ly le
auma ized by hei expe ience (Jong e al.,2001;Bleich e al.,2003;
Pham e al.,2004;Baye e al.,2007;Neugebaue e al.,2009). Typ-
ical symp oms o pos - auma ic s ess synd ome include nigh ma es,
lashbacks, di icul y concen a ing, dep ession, and ne ousness. The
ac ha auma ized people also egula ly exhibi a oidance beha -
io is pa icula ly ele an he e. Vic ims usually a oid places linked
o he auma ic e en , o , mo e gene ally, a oid public places and
social mingling. Au ho s ha e a gued ha , as a consequence, ic ims
will unde -u ilize heal hca e se ices (Nwo ah e al.,2014;Splin e
e al.,2021). Dep ession and gene alized anxie y, wo o he common
co ela es o expe iences o iolence, ha e also been shown o lead o
un a o able pa e ns o heal hca e-seeking o child en, such as delayed
checkups and missed ou ine accina ions (Minko i z e al.,2005).
T auma due o iolence can hus be expec ed o di ec ly and nega i ely
impac child wel a e.
Apa om hese clinical e ec s, exposu e o iolence may also
change pe cep ions o u u e h ea . In wo k p edic ing con lic e en s,
pas iolence is consis en ly one o he bes p edic o s o u u e i-
olence (Weidmann and Wa d,2010;Heg e e al.,2021). This holds
ue down o he local le el, meaning ha a eas ha ha e expe ienced
iolence in he pas ace a highe isk o expe iencing iolence again,
whe he due o s uc u al ac o s o social dynamics (Bazzi e al.,2022).
In o he wo ds, indi iduals exposed o iolence objec i ely ace an
ele a ed isk o becoming ic ims o u u e iolence, and his objec i e
shi is likely e lec ed in hei subjec i e h ea pe cep ion. No e ha
we would expec such changes in pe cep ion e en wi hou any psycho-
logical condi ion caused by exposu e o auma ic e en s. Indi iduals
will consequen ly end o adjus hei beha io o he newly pe cei ed
isk – a oiding la ge g oups o people, going ou less, and limi ing
e ands o only hose s ic ly necessa y (Malik e al.,2018;Sloan e al.,
2021).1I is highly plausible ha in his calculus, people will a ach
lowe impo ance o p e en i e measu es such as ha ing hei child en
accina ed o p esen ing hem o ou ine checkups – wi h po en ially
a al ou comes in cases whe e child en con ac accine-p e en able
diseases o a e p esen ed oo la e o cu e o he wise ea able condi ions.
My heo e ical expec a ions can be summed up in he ollowing hy-
po heses, which I will es using a ious da a sou ces in he emainde
o his pape :
H1: Exposu e o iolence leads o educed us in s a e ins i u ions
and inc eased ea o u u e iolence.
H2: Reduced us and inc eased ea lead o lowe heal hca e use,
lowe accina ion co e age, and highe in an and child mo ali y.
2. Da a and me hods
The challenge wi h es ing hese hypo heses is ha no single da ase
exis s ha combines all equi ed a iables in one place. I he e o e
combine da a om a ious sou ces, measu ed a bo h he indi idual-
and he con ex -le el.2A disad an age o his s a egy is ha o ce ain
concep s, impe ec p oxies ha e o be used. A he same ime, a clea
ad an age is scale. The da a sou ces used he e ely on long-s anding
da a collec ion e o s ha co e a la ge geog aphic a ea – ep esen ing
all majo egions o A ica – and a long ime span – om he ea ly 2000s
o a ound 2017. I use da a om all 22 coun ies ha we e a ec ed by
iolence acco ding o UCDP/PRIO’s A med Con lic Da ase (Gledi sch
e al.,2002;Da ies e al.,2023) – he mos widely used esou ce o
s udying a med con lic a ound he wo ld – while da a collec ion o he
A oba ome e was ongoing. These coun ies include Alge ia, Bu kina
Faso, Bu undi, Came oon, Co e d’I oi e, Egyp , Guinea, Kenya, Libe ia,
Mali, Mo occo, Mozambique, Nige , Nige ia, Senegal, Sie a Leone,
Sou h A ica, Sudan, Togo, Tunisia, Uganda, and Zimbabwe (see Fig. 1).
While he deg ee o iolence su e ed by he di e en coun ies a ies
widely, I op ed o his a he b oad selec ion o make he indings mo e
ep esen a i e. This selec ion is also wa an ed because o he con lic
da ase s, no ably he A med Con lic Loca ion and E en Da a P ojec
(ACLED), ano he commonly used esou ce (which is used he e o
obus ness checks), eco d a he high le els o exposu e o iolence
in he selec ed coun ies, e en when hey we e no in ou igh wa .
To u he add ess conce ns abou coun y selec ion, I eplica e co e
esul s o a mo e es ic i e sample o coun ies ha ha e seen 200
ba le- ela ed dea hs pe yea – ano he ai ly widely used h eshold
in oduced by Wea (1998). This selec ion includes Alge ia, Bu undi,
Came oon, Egyp , Mali, Nige , Nige ia, Sudan, and Uganda. As can be
seen in Table A9 in he Appendix, all o he co e esul s hold o his
sub-sample as well.
1Whe he he calculus is ac ually co ec , i.e., whe he he beha io al
adjus men s a e a ional in he sense ha hey p e en mo e ha m han hey
cause, is a ques ion o deba e (see, o example, he discussions in Becke and
Rubins ein,2004;Meie ieks and Schaub,2024).
2Figu e A1 in he Appendix p o ides a g aphical o e iew o he da a
s uc u e.
Social Science & Medicine 359 (2024) 117252
3
M. Schaub
Fig. 1. Map o con lic -a ec ed coun ies included in he s udy.
Indi idual-le el da a on demog aphics, us , and ea o u u e
iolence comes om he A oba ome e (2018), a la ge-scale da a
collec ion p ojec ha has been conduc ing na ionally ep esen a i e
public a i ude su eys on democ acy, go e nance, he economy, and
socie y in many A ican coun ies since 1999. I use da a om ounds
3 o 7 o he geocoded e sion o he A oba ome e , wi h geocoding
p o ided by BenYishay e al. (2017).
The main independen a iable o in e es is exposu e o iolence.
Bo h a con ex -le el and an indi idual-le el indica o a e used o
cap u e his concep . The main measu e o con ex ual exposu e o
iolen con lic is de i ed om he UCDP’s Geo e e enced E en Da ase
(GED) (Sundbe g and Melande ,2013;Da ies e al.,2023), which
p o ides con lic -e en da a, including p ecise coo dina es and he
numbe o casual ies. As wi h all con ex -le el in o ma ion, I agg ega e
he da a a he 0.5◦×0.5◦(∼55 ×55 km a he equa o ) g id-yea le el
using he PRIO-g id (Tolle sen e al.,2012).3Con ex -le el exposu e is
hen measu ed by a bina y indica o ha akes he alue o one i an
e en was eco ded in he UCDP GED da ase in a gi en yea and g id
cell, and ze o i no . The da ase only eco ds iolen e en s belonging
o a med con lic s esul ing in a leas 25 dea hs pe yea , meaning ha
only ai ly se ious iolen e en s end o be eco ded.
As a obus ness check, I eplica e all analyses wi h he
ACLED (Raleigh e al.,2010,2023) da ase . Simila o he UCDP GED,
ACLED p o ides in o ma ion on iolen con lic e en s, hei loca ion
and se e i y in e ms o indi iduals killed, bu does no equi e a
h eshold o se e i y o an e en o be eco ded. ACLED he e o e ends
o eco d a a highe numbe o e en s (as can be seen in he summa y
s a is ics in Table A1 in he Appendix), making i a mo e inclusi e, bu
a guably less p ecise (cp. Eck,2012) measu e o exposu e o iolence.
Responden s a e coded exposed o iolence i ACLED eco ded a iolen
e en in he g id cell and yea ha he in e iew ook place. Replica ing
all esul s using his indica o ins ead o he UCDP GED-based one gi es
e y simila esul s, as shown in Sec ion D o he Appendix.
Indi idual-le el exposu e o iolence is measu ed wi h he A o-
ba ome e su ey i em ‘‘Du ing he pas yea , ha e you o anyone in
you amily been physically a acked?’’, o which esponden s could
espond ‘‘No’’, ‘‘Once’’, ‘‘Twice’’, o ‘‘Th ee o mo e imes’’ (as wi h all
i ems, hey could also choose ‘‘Don’ know/Re use’’). All non-missing
alues o his a iable a e used. I is impo an o no e ha we do
no know whe he o no he iolence epo ed by a esponden is
3The combined da ase includes a ound 1700 dis inc g id cells, and a ound
4000 g id cell yea s.
ela ed o he b oade con lic cap u ed by he UCDP GED (o ACLED)
da a; ins ead, he indi idual-le el measu e should be unde s ood as an
al e na i e measu e o iolen ic imiza ion.
O pa icula in e es o his s udy a e he hypo hesized media o s
‘‘ us in he s a e’’ and ‘‘ ea o u u e iolence’’, and he in e media e
ou come ‘‘heal h-seeking beha io ’’. Measu es o hese a iables a e
d awn om he A oba ome e . Wi h espec o us in he s a e,
schola s wo king on sub-Saha an A ica end o emphasize he im-
po ance o us in he p esiden because mos poli ical egimes in
he egion a e cha ac e ized by s ong p esiden ial go e nmen s (e.g.
an C anenbu gh,2008;Aze edo-Ha man,2011;Ga es and Jus esen,
2020). Howe e , schola s wo king on poli ical us mo e gene ally
ypically op o some o m o index ha combines measu es o us in
di e en go e nmen ins i u ions and o ices (e.g. Ma es and Mo eno,
2018). O en, hese schola s u he dis inguish be ween poli ical us –
us in elec ed leade s such as he p esiden o he uling poli ical
pa y- and ins i u ional us – us in nonpa isan ‘o de ’ ins i u ions
such as he police o he cou s (e.g. Ro hs ein and S olle,2008;Ma ien,
2011).
Because o he impo ance o he execu i e in he A ican con ex ,
I he e ocus on poli ical us . In he A oba ome e , esponden s a e
asked, ‘‘How much do you us he ollowing?’’ ollowed by a lis o
s a e ins i u ions, including he p esiden , he uling pa y, he cou s,
and he police. Responden s can choose ‘‘no a all’’, ‘‘jus a li le’’,
‘‘somewha ’’, o ‘‘ us a lo ’’. Poli ical us is measu ed by a e aging
us in he p esiden and us in he uling pa y.4In Table A11
o he Appendix, I eplica e he main analyses wi h an indica o o
ins i u ional us , coded as he a e age o us in he police and us in
he cou s o law.5Un o una ely, he A oba ome e does no include
ques ions on us in he medical sys em, which would be o g ea
in e es in he con ex o his pape .
Fea o u u e iolence is measu ed wi h an i em asking esponden s
how o en hey o hei amily ea c ime in hei own home, o which
hey could espond ‘‘Ne e ’’, ‘‘Jus once o wice’’, ‘‘Se e al imes’’,
‘‘Many imes’’, o ‘‘Always’’. He e, he assump ion is ha he ea o
c ime encompasses he ea o being iolen ly a acked, pa icula ly in
con lic -p one con ex s, on which his s udy ocuses. While no exac ly
he same as ea o ic imiza ion, his is he only measu e a ailable
in he A oba ome e . The imp ecision in oduced by his less- han-
op imal measu e may explain he ela i ely weak explana o y powe
o he measu e in some o he analyses below. Heal h-seeking beha io
is cap u ed by an A oba ome e i em inqui ing whe he esponden s
o hei amily membe s ha e ‘‘gone wi hou medicines o medical
ea men ’’, o which hey again could espond ‘‘Ne e ’’, ‘‘Jus once o
wice’’, ‘‘Se e al imes’’, ‘‘Many imes’’, o ‘‘Always’’. I use his a iable
in i s e e sed o m so ha highe alues indica e mo e equen
heal hca e use. Apa om hese a iables o in e es , all models also
include a se o con ol a iables aken om he A oba ome e , no ably
a esponden ’s age, sex, educa ion le el, and employmen s a us.
Ul ima e ou comes o in e es a e in an mo ali y and accina ion
a es. In an mo ali y a es a e om Bu s ein e al. (2019), who p o ide
high- esolu ion es ima es co e ing he whole o mainland A ica o he
2000–2017 pe iod. In an mo ali y is measu ed as he p obabili y o a
child dying be o e eaching he age o one. These es ima es a e based
on na ional censuses and su ey da a, including he Demog aphic and
Heal h Su ey (DHS) and UNICEF Mul iple Indica o Clus e Su ey
(MICS) household su ey se ies, and a e he e used agg ega ed a he
PRIO g id-cell-yea le el. As a obus ness check, in Appendix Table
A10 I eplica e all co e esul s using da a on child mo ali y (also
om Bu s ein e al. 2019), de ined as he p obabili y o a child dying
be o e he age o i e.
4These wo dimensions o poli ical us ha e also been ound o inc ease
wi h imp o emen s in heal h se ice deli e y, see Chukwuma e al. (2019).
5The models show ha while mos ela ionships be ween he di e en
a iables a e simila o bo h poli ical and ins i u ional us , ins i u ional us
does no p edic accina ion a es, a guably due o he impo an messaging
ole o poli ical leade s.
Social Science & Medicine 359 (2024) 117252
4
M. Schaub
Fig. 2. T ends in co e a iables o e ime, condi ional on con ex -le el exposu e and indi idual ic imiza ion.
No e: The igu e shows ends o e ime o he indica ed heal h- ela ed ou comes and a i udinal and beha io al measu es (a) in con lic -a ec ed and non-a ec ed a eas, and (b)
among ic imized and non- ic imized indi iduals. Con ex -le el exposu e o iolence is measu ed as g id cells ha expe ienced one o mo e con lic e en s in a gi en yea , as
eco ded in he UCDP GED. Indi idual ic imiza ion is measu ed as esponden s in e iewed in a gi en yea indica ing ha hey o hei amily membe s we e physically a acked.
In an mo ali y is measu ed as he p obabili y ha a child will no su i e i s i s yea o li e. DPT co e age measu es he p opo ion o unde -5 yea old who ha e ecei ed he
ecommended h ee doses o he diph he ia-pe ussis- e anus accine. Fea o iolence measu es esponden s’ sel - epo ed conce n ha hey will expe ience c ime o some o m in
hei home. Poli ical us is an index ha combines esponden s’ sel - epo ed us in he p esiden and us in he uling pa y. F equency o heal hca e use eco ds he ease and
equency wi h which esponden s could access heal hca e se ices in he pas yea . Lines a e polynomial smoo hs wi h a bandwid h o h ee; shaded a eas a e 95% con idence
in e als.
Social Science & Medicine 359 (2024) 117252
5
M. Schaub
As he indica o o accina ion co e age, diph he ia-pe ussis- e -
anus (DPT) accine co e age is used, measu ed as he sha e o child en
unde i e who a e ully accina ed (i.e., who ha e ecei ed h ee sho s
o mo e). These es ima es we e p oduced by Mosse e al. (2019) using
da a om household su eys ha included dose-speci ic in o ma ion on
DPT co e age.6As a di e en ype o ou come and obus ness check, I
examine whe he simila esul s as o accina ion can be ep oduced
o o al ehyd a ion solu ion (ORS) use. ORS use is in e es ing because
i ep esen s a ype o heal h-seeking beha io ha is di e en om
accina ion. Howe e , simila o accina ion, i is a beha io ha
equi es con ac wi h and us in heal h p o essionals. This is because
al hough in p inciple ORS can be easily adminis e ed a home by
ca egi e s, in p ac ice, ‘‘ he in o ma ion ega ding i s adequa e usage is
es ic ed wi hin he heal hca e cen e s and p o essionals’’ (Aghsaei a d
e al.,2022). The indica o comes om Wiens e al. (2020) and mea-
su es he sha e o child en unde 5 wi h dia hea who ecei ed ORS.
The esul s o his obus ness check a e epo ed in Appendix Table A12
and con i m he esul s p esen ed below o accina ion co e age.
Con ex -le el con ol a iables include ac o s ha may in luence
bo h he onse o iolen e en s and heal h ou comes. Along wi h
longi ude and la i ude, hese a e: (i) popula ion size, as p o ided by
LandScan (B igh e al.,2018); (ii) nigh ligh in ensi y, a measu e o
economic de elopmen collec ed by he U.S. Ai Fo ce De ense Me eo-
ological Sa elli e P og am and made a ailable o esea ch by El idge
e al. (2021), Ghosh e al. (2021); (iii) ma ke access and s a e each,
measu ed as he a el ime o he nea es la ge ci y and o egional
and na ional capi als, espec i ely (Mülle -C epon,2021); (i ) emale
educa ion, measu ed as he a e age yea s o educa ion o 20-24-yea -
old women, om G ae z e al. (2018); and ( ) clima e ex emes, as
measu ed by he s anda dized p ecipi a ion e apo anspi a ion index
(SPEI), om Peng e al. (2019). Apa om he geog aphic coo dina es,
con ex -le el con ol a iables a e again agg ega ed a he g id-yea
le el using he PRIO-g id. Since some o he a iables a e only a ailable
up o 2015, I ex apola ed missing alues o 2016 in o de o no lose
obse a ions due o missingness in he con ol a iables.
Table A1 in he Appendix p o ides summa y s a is ics o
all indi idual-le el and con ex -le el a iables, and Fig. 2 shows ends
o e ime o he co e a iables o in e es . T ends a e displayed
sepa a ely o a eas exposed o con lic ( ha ha e seen a con lic e en
as de ined by UCDP) s. hose ha ha e no – Panel (a); and o
indi iduals ha ha e been ic imized (physically a acked) s. hose
ha ha e no – Panel (b). We can see ha in a eas no di ec ly exposed
o iolence, in an mo ali y a es ha e seen a s eep decline o e he
yea s unde obse a ion. This end does no hold in con lic -exposed
a eas, howe e . While hese a eas s a ed ou wi h ini ially lowe a es
o in an mo ali y (likely because con lic ends o concen a e in mo e
u ban a eas, which ypically also ha e lowe in an mo ali y a es),
in an mo ali y a es hen s agna ed. Gi en hese ends, nega i e
e ec s due o con lic exposu e a e bes in e p e ed as he absence o
imp o emen in in an mo ali y a he han absolu e inc eases. How-
e e , his is s ill conce ning because absolu e igu es emain ex emely
high (as shown in Table A1 in he Appendix): on a e age, 5.9 pe cen
o child en did no comple e hei i s yea o li e, wi h igu es as
high as 12.9 pe cen in some g id cells. E en mo e dispa a e ends a e
isible wi h ega d o accina ion a es. While hese ha e ended o
inc ease in non-con lic -a ec ed g id cells, he end is clea ly nega i e
in a eas exposed o con lic , likely con ibu ing o he s agna ing in an
mo ali y a es.
Simila , albei mo e complex, ends a e e iden in he a i udinal
and beha io al measu es. In e ms o ea o u u e iolence, while
o e all ends a e downwa d sloping – a guably e lec ing declines in
6These and o he heal h- ela ed, sub-na ional indica o s can be down-
loaded om he websi e o he Ins i u e o Heal h Me ics and E alua ion
(IHME), see: h ps://www.heal hda a.o g.
iolen c ime un ela ed o a med con lic ( an Dijk e al.,2022) –
con ex -le el exposu e o iolence is consis en ly associa ed wi h highe
le els o ea . No pa icula ends a e isible o indi idual ic im-
iza ion, hough, which is consis en ly associa ed wi h highe le els o
ea . As o poli ical us , again we see ha us is clea ly lowe
among esponden s exposed o iolence o ic imized. The gap in us
be ween hose exposed o iolence a he con ex le el and hose no
exposed seems o be na owing o e ime. I is possible ha coun ies in
A ica and he in e na ional communi y ha e become be e a dealing
wi h some o he consequences o iolen con lic , he eby educing he
nega i e impac o exposu e o iolence on poli ical us . Howe e , no
such end is e iden when looking a indi idual ic imiza ion, whe e
di e ences in us le els a e la ge and cons an o e ime.
A somewha pa adoxical end is obse ed wi h espec o access
o heal hca e. In a eas exposed o iolence, he p opo ion o people
no using heal hca e se ices dec eased o e ime and is lowe han in
a eas no a ec ed by iolence. I can only specula e as o why his is
he case. One possibili y is ha access has imp o ed. As poin ed ou
in a e iew a icle on heal h-seeking beha io in he mids o con-
lic (Adeyanju e al.,2024), access o heal hca e some imes imp o es
in con lic se ings due o he humani a ian esponse o na ional and
in e na ional NGOs. Howe e , he e is no such posi i e end when we
look a indi idual ic imiza ion. In all yea s, ic ims o iolence a e
signi ican ly less likely o ha e in e ac ed wi h medical p o essionals
han non- ic ims, and he disc epancy emains highly cons an o e
ime. I is possible ha access o and use o heal hca e is imp o ing o
some in con lic -a ec ed a eas, bu no o di ec ic ims o iolence.
Mo e impo an ly, he ends may be con ounded by hi d a iables
and hus may no uly cap u e he e ec o iolence on he epo ed
ou come. Fo his eason, we now u n o mo e igo ous analyses.
3. Analysis
S uc u al equa ion models. I begin he analysis by in es iga ing how
he di e en a iables jus in oduced ela e o each o he using s uc-
u al equa ion models. The models, g aphically p esen ed in Fig. 3,
show all ela ionships ha a e bo h heo e ically in e es ing and s a i-
cally signi ican . Again, I show esul s o bo h con ex - and indi idual-
le el exposu e o iolence. E ec s mani es hemsel es a he indi idual
le el (‘ ea o iolence’, ‘poli ical us ’, and ‘heal hca e use’) and
he con ex le el (‘DPT-3 co e age’, and ‘in an mo ali y’). Se e al
obse a ions s and ou : As hypo hesized, he e a e s a is ically signi -
ican co ela ions be ween exposu e o iolence, ea o iolence, and
lowe us in s a e ins i u ions. Highe poli ical us posi i ely p edic s
heal hca e use, while inc eased ea o ic imiza ion has he opposi e
e ec . Con a y o wha was hypo hesized, he e is no signi ican
ela ionship be ween ea o iolence and accina ion a es (a ow
omi ed). Howe e , again in line wi h he heo y, accina ion a es
a e s ongly ela ed o poli ical mis us : lowe us in s a e leade s
p edic s lowe accina ion a es. Heal hca e use and accina ion a es,
in u n, a e s ongly nega i ely associa ed wi h highe in an mo ali y.
Two-way ixed e ec s and mul ile el models. While he analyses p o ide
suppo i e e idence o he hypo hesized ela ionship, he models in
Fig. 3 shows me e co ela ions wi hou aking in o accoun po en ial
con ounding ac o s. I he e o e p oceed by p esen ing eg ession mod-
els, all o which include he ull se o con ol a iables in oduced
abo e. I es ima e he indi idual-con ex and he con ex -con ex links
using mul ile el models (allowing slopes o a y a he g id-le el), and
he indi idual-indi idual links using yea - and g id-cell ixed e ec s
models. The la e speci ica ion is pa icula ly s ingen because i
compa es indi iduals who we e in e iewed wi hin he same local
a ea, hence limi ing he po en ial con ounding e ec o con ex . The
inclusion o yea ixed e ec s ensu es ha ime ends canno in luence
he es ima es. Resul s a e shown in Table 1.
Exposu e o iolence a he g id cell le el is associa ed wi h a 0.14
poin dec ease in he poli ical us scale, and a 0.22 poin inc ease in
Social Science & Medicine 359 (2024) 117252
6
M. Schaub
Fig. 3. S uc u al equa ion models showing ela ionships be ween co e a iables.
No e: S uc u al equa ion models ela ing (a) con ex -le el exposu e o iolence, and (b) indi idual-le el ic imiza ion o a i udinal and beha io al measu es and s uc u al ou comes.
Es ima ed using maximum likelihood es ima ion, s anda d e o s clus e ed a he g id-cell le el, ∗𝑝 < 0.1,∗∗ 𝑝 < 0.05,∗∗∗ 𝑝 < 0.01. Resul s in abula o ma a e shown in Table
A4 in he Appendix.
Table 1
Two-way- ixed-e ec s (TWFE) and mul ile el models.
TWFE models Mul ile el models
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Pol. us Fea Pol. us Fea Heal hc. Heal hc. DPT-3 co . In . mo . In . mo .
Con ex . iol. exp. −0.143∗∗ 0.223∗∗∗
(0.056) (0.055)
Ind. ic imiza ion −0.091∗∗∗ 0.482∗∗∗
(0.008) (0.011)
Fea o iolence −0.337∗∗∗
(0.013)
Poli ical us 0.298∗∗∗ 0.124∗∗∗
(0.017) (0.032)
F eq o heal hca e use −0.062∗∗∗
(0.009)
DPT-3 co e age −0.644∗∗
(0.290)
N obse a ions 88,562 88,770 87,359 88,627 88,411 88,200 88,563 89,744 90,136
N ixed/ andom e ec s 1,785 1,783 1,782 1,783 1,782 1,784 1,786 1,786 1,786
Con ols Yes Yes Yes Yes Yes Yes Yes Yes Yes
No e: Two-way- ixed-e ec s (TWFE) and mul ile el models ela ing he indica ed a iables o in e es o each o he . TWFE models es ima ed using OLS; mul ile el models es ima ed
using maximum likelihood es ima ions. Obse a ions a y due o missing alues in indi idual dependen o con ol a iables. The abb e ia ed column and ow i les ansla e as
ollows: Pol. us ∼Poli ical us ; Fea ∼Fea o iolence; Heal hc.∼F equency o heal hca e use; DPT3-co .∼DPT-3 co e age; In . mo .∼In an mo ali y, p obabili y o a child dying
be o e he age o 1; Con ex . iol. exp.∼Exposu e o iolence a he con ex -le el; Ind. ic imiza ion∼Indi idual-le el ic imiza ion o he esponden o hei amily. S anda d e o s
clus e ed a he g id-cell le el in pa en heses, ∗𝑝 < 0.1,∗∗ 𝑝 < 0.05,∗∗∗ 𝑝 < 0.01. Full eg ession ou pu shown in Table A3 in he Appendix.
he measu e o ea o iolence. Simila e ec s a e seen o indi idual-
le el ic imiza ion. Vic imiza ion p edic s a 0.09 poin dec ease in
poli ical us and a 0.48 poin inc ease in ea o u u e iolence.
Fea , in u n, is associa ed wi h a 0.34 poin dec ease in heal hca e
use, while a 1 poin inc ease in poli ical us p edic s a 0.3 poin
inc ease in heal hca e use. The mul ile el models (Models 7 h ough
9) show ha he known ela ionships be ween us and accina ion
a es, be ween he use o medical se ices and in an mo ali y, and
be ween accina ion a es and in an mo ali y all hold. Taken oge he ,
he esul s hus con i m he obus ness o he co ela ional indings
abo e.
Two-s age di e ences in di e ences. Two-way ixed e ec s ha e aced
c i icism o no clea ly iden i ying he p ope coun e ac ual o ea ed
uni s due o ea men e ec he e ogenei y (Sun and Ab aham,2021;
Imai and Kim,2021). To ci cum en his p oblem, I adop he wo-s age
di e ences -in-di e ences app oach sugges ed by Ga dne e al. (2023)
and implemen ed by Bu s and Ga dne (2022). This app oach is only
applicable o he eg essions p obing o he e ec o con ex ual-le el
exposu e o iolence (i.e., Models 1 and 2), whe e a clea ime- a ying
ea men – being a ec ed by a UCDP- eco ded con lic e en – can
be iden i ied. The idea o Ga dne e al. (2023) is o i s eg ess he
dependen a iable o in e es on he uni and ime ixed e ec s plus
Table 2
Two-s age di e ences-in-di e ences model.
(1) (2)
Pol. us Fea
Con ex . iol. exp. −0.013∗∗∗ 0.029∗∗∗
(0.004) (0.005)
N obse a ions 88,563 88,770
Con ols Yes Yes
N ixed e ec s 1,785 1,783
No e: Two-s age di e ences-in-di e ences model (Ga dne e al.,2023) o poli ical
us and ea on con ex ual exposu e o iolence. The abb e ia ed column and ow
i les ansla e as ollows: Pol. us ∼Poli ical us ; Fea ∼Fea o iolence; Con ex . iol.
exp.∼Exposu e o iolence a he con ex -le el. S anda d e o s clus e ed a he g id-cell
le el in pa en heses, ∗𝑝 < 0.1,∗∗ 𝑝 < 0.05,∗∗∗ 𝑝 < 0.01.
con ols using only no p e iously ea ed uni s. In a second s ep, he
p edic ed esiduals om his i s eg ession a e hen eg essed on he
ea men indica o . In his way, he a e age ea men e ec on he
ea ed can be iden i ied in a consis en manne . Because he analysis
is pe o med on he esiduals only, he me hod ends o p oduce mo e
conse a i e es ima es. The esul s shown in Table 2 indica e ha
he indings a e obus o his p ocedu e. While he es ima ed e ec s
o con ex ual exposu e o iolence a e much educed in size, hey
Social Science & Medicine 359 (2024) 117252
7
M. Schaub
Table 3
Medi a ion analysis.
A. Con ex . exp. iol. →Heal hca e use
Media o : Fea Pol. us
Na u al indi ec e ec −0.021∗∗∗ −0.012∗∗∗
(0.006) (0.004)
To al e ec −0.031 −0.033
(0.040) (0.040)
P opo ion media ed 0.68 0.35
N 86,902 86,902
Con ols Yes Yes
B. Ind. ic imiza ion →Heal hca e use
Media o : Fea Pol. us
Na u al indi ec e ec −0.083∗∗∗ −0.012∗∗∗
(0.008) (0.002)
To al e ec −0.329∗∗∗ −0.327∗∗∗
(0.015) (0.015)
P opo ion media ed 0.25 0.04
N 86,773 86,773
Con ols Yes Yes
No e: Media ion analysis in es iga ing he ex en o which he ela ionship be ween
exposu e o iolence a he con ex - (A) and indi idual-le el (B) is media ed by ea o
iolence and ins i u ional (mis-) us . The abb e ia ed column and ow i les ansla e
as ollows: Fea ∼Fea o iolence; Pol. us ∼Poli ical us ; Con ex . iol. exp.∼Exposu e
o iolence a he con ex -le el; Ind. ic imiza ion∼Indi idual-le el ic imiza ion o he
esponden o hei amily. OLS eg ession on simula ed po en ial ou comes using S a a’s
causal media ion package; s anda d e o s clus e ed a he g id-cell le el in pa en heses,
∗𝑝 < 0.1,∗∗ 𝑝 < 0.05,∗∗∗ 𝑝 < 0.01.
emain highly s a is ically signi ican , inc easing ou con idence ha
he ela ionship we obse e is indeed causal.
Media ion analysis. How much is he e ec o exposu e o iolence on
heal hca e-seeking beha io media ed by poli ical mis us and ea o
iolence? To answe his ques ion, I conduc a media ion analysis. I
adop he causal media ion amewo k p oposed by Imai e al. (2010),
which also p o ides me hods o assessing he sequen ial igno abili y
assump ion in his ype o analysis. In cases such as he p esen one,
whe e he o al e ec is hypo hesized o un h ough se e al media o s,
i is ecommended o es ima e media ion e ec s in he same model,
o o con ol o o he po en ial media o s (Vande Weele and Vans ee-
land ,2014). Alongside he s anda d se o con ol a iables, I he e o e
include he non-conside ed media o as a con ol a iable, plus he
in e ac ion e m be ween he exposu e a iable and he media o .
As can be seen in Table 3, he educ ion in heal hca e seeking
beha io due o he exposu e o iolence is s ongly media ed by he
ea o iolence. This is pa icula ly ue o con ex ual exposu e o
iolence, whe e an es ima ed 68% o he e ec is media ed by ea
o u u e iolence, compa ed o 25% o indi idual ic imiza ion. In
con as , he media ing ole o poli ical mis us is less p onounced.
Fo con ex ual exposu e, an es ima ed 35% o he e ec is media ed
h ough lowe ed poli ical us , while o indi idual exposu e i is only
4%.
The media ion analysis hus p o ides clea suppo o he heo y
ha exposu e o iolence d i es down heal hca e-seeking beha io
h ough he channels o ea o u u e iolence, and, o a lesse ex en ,
mis us in he poli ical leade ship. Howe e , i can also be seen ha
only pa o he ela ionship be ween iolence and heal hca e-seeking
is explained by he p oposed media o s. This means ha he e mus
be o he ac o s media ing he e ec o exposu e o iolence ha
emain unaccoun ed o . In he heo y sec ion, I sugges ed ha one such
media o migh be psychological auma, which could no be es ed
he e because he e was no app op ia e measu e in he da a. Explo a ion
o his and o he media o s will ha e o be le o u u e esea ch.
Conclusion
This pape es ablishes a ela ionship be ween exposu e o iolence
and educed heal h-seeking beha io . This ela ionship is media ed by
poli ical mis us and he ea o u u e iolence. The knock-on e ec s
a e lowe ed accina ion a es, educed heal hca e use, and, ul ima ely,
highe le els o in an and child mo ali y. The s udy combines a la ge
numbe o da a sou ces om 22 con lic -a ec ed coun ies in A ica,
obse ed o e a ime pe iod o 12 yea s. This app oach is bo h a
s eng h and a weakness. On he one hand, he b ead h o he s udy
inc eases ou con idence ha he obse ed e ec s a e qui e gene al;
hei applicabili y beyond he A ican con inen is howe e a ques ion
o u he esea ch. A he same ime, se e al concep s could no
be op imally measu ed. This is ce ainly ue o us in poli ical and
s a e ins i u ions, which could only be ope a ionalized in a gene al
ashion. Fu u e esea ch should explici ly es o us in heal h- ela ed
messages, and should also measu e us in he medical sys em and
heal h wo ke s. Mo eo e , a he han elying on con ex -le el indica-
o s o in an mo ali y and accine co e age, i would be p e e able
o measu e hose ou comes wi hin he same ins umen whe e us
and ea o iolence a e assessed. Such esea ch should also explo e
addi ional channels, such as psychological auma. Conduc ing such
uni ied esea ch would be an impo an a ea o u u e explo a ion.
Ano he possible ex ension would be o igo ously in es iga e he long-
e m e ec s o he p oposed channels. Fu u e s udies could explo e how
las ing he e ec s o iolence on us and ea a e, and whe he ad e se
heal h e ec s o hese ac o s emain isible in he long un.
This pape con ibu es o he li e a u e on he heal h e ec s o
iolence by highligh ing he la gely o e looked demand-side e ec s
o iolen con lic on heal h. Exposu e o iolence changes he a i-
udes and beha io s o ci ilians, wi h nega i e consequences o heal h
ou comes, pa icula ly child heal h. These indings should be seen as
complemen a y o, a he han challenging, well-es ablished pa hways,
such as he des uc ion o c i ical in as uc u e o he ou -mig a ion
o heal h wo ke s, which p e ious esea ch has shown o se e ely and
nega i ely a ec heal h ou comes in con lic -a ec ed egions. Fu u e
wo k could explo e in e ac ions be ween supply- and demand-side
ac o s. Fo example, i is possible ha he decline in he quali y
o heal hca e due o he ou -mig a ion o heal h p o essionals om
con lic -a ec ed egions nega i ely a ec s us in he heal h sys em,
which in u n a ec s heal hca e-seeking beha io in he long un.
While addi ional esea ch is necessa y be o e de ini i e conclu-
sions can be d awn, he indings p esen ed he e seem clea ly ele an
o policy. They sugges ha in o de o mi iga e he de imen al
heal h e ec s o iolence, a i udinal and beha io al consequences
mus be explici ly add essed. Adjus ing exis ing p og ams could be
pa o his e o . Fo example, us -building measu es a e al eady a
well-es ablished componen o pos -con lic econs uc ion and peace-
building p og ams (Wong,2016;S ensson and B ounéus,2013). These
p og ams could be modi ied so o conside heal h as an explici a ea
o ac i i y. While us -building measu es would likely be one com-
ponen , he p o ision o ‘ha d’ secu i y also ma e s. As shown, he
ea o ic imiza ion is a majo d i e o de imen al adjus men s in
heal h-seeking beha io . Measu es ha con ibu e o p o iding physical
secu i y in con lic and pos -con lic se ings – be hey in he ealm o
secu i y-sec o e o m, o in he ealm o wel a e policy – can he e o e
also likely bene i popula ion heal h and educe child mo ali y.
CRediT au ho ship con ibu ion s a emen
Max Schaub: W i ing – o iginal d a , Visualiza ion, Valida ion,
Supe ision, So wa e, Resou ces, P ojec adminis a ion, Me hodol-
ogy, In es iga ion, Funding acquisi ion, Fo mal analysis, Da a cu a ion,
Concep ualiza ion.
Social Science & Medicine 359 (2024) 117252
8
