The impact of internet access on COVID-19 spread in Indonesia

Kunz, Johannes; P oppe , Ca ol; T inh, T ong-Anh
Wo king Pape
The impac o in e ne access on COVID-19 sp ead in
Indonesia
ADB Economics Wo king Pape Se ies, No. 723
P o ided in Coope a ion wi h:
Asian De elopmen Bank (ADB), Manila
Sugges ed Ci a ion: Kunz, Johannes; P oppe , Ca ol; T inh, T ong-Anh (2024) : The impac o in e ne
access on COVID-19 sp ead in Indonesia, ADB Economics Wo king Pape Se ies, No. 723, Asian
De elopmen Bank (ADB), Manila,
h ps://doi.o g/10.22617/WPS240232-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/298169
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/3.0/igo/
ASIAN DEVELOPMENT BANK
ASIAN DEVELOPMENT BANK
6 ADB A enue, Mandaluyong Ci y
1550 Me o Manila, Philippines
www.adb.o g
THE IMPACT OF INTERNET
ACCESS ON COVID-19
SPREAD IN INDONESIA
Johannes S. Kunz, Ca ol P oppe , and T ong-Anh T inh
ADB ECONOMICS
WORKING PAPER SERIES
NO. 723
Ap il 2024
The Impac o In e ne Access on COVID-19 Sp ead in Indonesia
This s udy examines he impac s o 3G in e ne connec i i y on COVID-19 case a es ac oss dis ic s
in Indonesia. By analyzing geog aphical a ia ions in mobile in e ne access and employing ligh ning s ikes
as an ins umen al a iable, he s udy es ablishes a causal link be ween imp o ed in e ne connec i i y and
educed ansmission o COVID-19. The indings sugges ha in es men s in digi al in as uc u e migh be
a c ucial and e ec i e ool in pandemic p e en ion and esponse.
Abou he Asian De elopmen Bank
ADB is commi ed o achie ing a p ospe ous, inclusi e, esilien , and sus ainable Asia and he Paci ic,
while sus aining i s e o s o e adica e ex eme po e y. Es ablished in 1966, i is owned by 68 membe s
—49 om he egion. I s main ins umen s o helping i s de eloping membe coun ies a e policy dialogue,
loans, equi y in es men s, gua an ees, g an s, and echnical assis ance.
ASIAN DEVELOPMENT BANK
The ADB Economics Wo king Pape Se ies
p esen s esea ch in p og ess o elici commen s
and encou age deba e on de elopmen issues
in Asia and he Paci ic. The iews exp essed
a e hose o he au ho s and do no necessa ily
e lec he iews and policies o ADB o
i s Boa d o Go e no s o he go e nmen s
hey ep esen .
ADB Economics Wo king Pape Se ies
The Impac o In e ne Access on COVID-19 Sp ead
in Indonesia
Johannes S. Kunz, Ca ol P oppe ,
and T ong-Anh T inh
No. 723 | Ap il 2024
Johannes S. Kunz ([email p o ec ed]) and
T ong-Anh T inh ([email p o ec ed]) a e
esea ch ellows a he Cen e o Heal h Economics,
Monash Uni e si y. Ca ol P oppe (c.p oppe @
impe ial.ac.uk) is a p o esso o economics a Impe ial
College London.
C ea i e Commons A ibu ion 3.0 IGO license (CC BY 3.0 IGO)
© 2024 Asian De elopmen Bank
6 ADB A enue, Mandaluyong Ci y, 1550 Me o Manila, Philippines
Tel +63 2 8632 4444; Fax +63 2 8636 2444
www.adb.o g
Some igh s ese ed. Published in 2024.
ISSN 2313-6537 (p in ), 2313-6545 (PDF)
Publica ion S ock No. WPS240232-2
DOI: h p://dx.doi.o g/10.22617/WPS240232-2
The iews exp essed in his publica ion a e hose o he au ho s and do no necessa ily e lec he iews and policies
o  he Asian De elopmen Bank (ADB) o i s Boa d o Go e no s o he go e nmen s hey ep esen .
ADB does no gua an ee he accu acy o he da a included in his publica ion and accep s no esponsibili y o any
consequence o hei use. The men ion o speci ic companies o p oduc s o manu ac u e s does no imply ha hey
a e endo sed o ecommended by ADB in p e e ence o o he s o a simila na u e ha a e no men ioned.
By making any designa ion o o e e ence o a pa icula e i o y o geog aphic a ea in his documen , ADB does no
in end o make any judgmen s as o he legal o o he s a us o any e i o y o a ea.
This publica ion is a ailable unde he C ea i e Commons A ibu ion 3.0 IGO license (CC BY 3.0 IGO)
h ps://c ea i ecommons.o g/licenses/by/3.0/igo/. By using he con en o his publica ion, you ag ee o be bound
by he e ms o his license. Fo a ibu ion, ansla ions, adap a ions, and pe missions, please ead he p o isions
and e ms o use a h ps://www.adb.o g/ e ms-use#openaccess.
This CC license does no apply o non-ADB copy igh ma e ials in his publica ion. I he ma e ial is a ibu ed
oano he sou ce, please con ac he copy igh owne o publishe o ha sou ce o pe mission o ep oduce i .
ADB canno be held liable o any claims ha a ise as a esul o you use o he ma e ial.
Please con ac pubsma ke [email protected] g i you ha e ques ions o commen s wi h espec o con en , o i you wish
oob ain copy igh pe mission o you in ended use ha does no all wi hin hese e ms, o o pe mission o use
heADB logo.
Co igenda o ADB publica ions may be ound a h p://www.adb.o g/publica ions/co igenda.
No e:
In his publica ion, ADB ecognizes “China” as he People’s Republic o China and “Hong Kong” as Hong Kong, China.
ABSTRACT
The co ona i us disease (COVID-19) pandemic has highligh ed he c ucial ole o In e ne
access in pandemic p e en ion and esponse. In e ne access has acili a ed he apid
dissemina ion o i al in o ma ion, p o ided elemedicine se ices, and enabled emo e
wo k and educa ion. This s udy uses a wide ange o da a sou ces o in es iga e he
geog aphic a ia ion o In e ne access p oxied by 3G mobile b oadband du ing he
COVID-19 pandemic in Indonesia. We employ se e al app oaches o accoun o
po en ial con ounding ac o s, including using ligh ning s ikes as an ins umen al a iable,
o con i m he signi ican ole ha he In e ne played in he sp ead o COVID-19 cases.
Ou indings sugges ha inc easing In e ne access could posi i ely impac pandemic
p e en ion and esponse e o s, pa icula ly in egions wi h limi ed connec i i y.
The e o e, imp o ing In e ne in as uc u e in de eloping coun ies may be c ucial in
p e en ing u u e pandemics.
Keywo ds: heal h eme gencies, In e ne access, in o ma ion, COVID-19 sp ead
JEL codes: I12, I15, I31, O18, L96, H41
__________________
We hank A ie Ramayandi and Daniel C. Su yada ma o hei ou s anding suppo h oughou his p ojec .
We hank Ke in S aub, Gio anni an Empel, Paul Raschky and semina pa icipan s a SoDa Lab-Monash,
CHE-Monash, and ADB Con e ence 2023 (Manila) o hei use ul commen s. Financial suppo om he
TA-6989 REG: Asian De elopmen Ou look Upda e 2023 Theme Chap e : Asia’s P epa edness on Heal h
Eme gencies (55278-003) is g a e ully acknowledged. Nei he o he au ho s ha e any con lic o in e es
o decla e.

1 In oduc ion
The eme gence o he co ona i us disease (COVID-19) in Decembe 2019 and i s apid
sp ead in o a global pandemic has p o oundly impac ed socie ies wo ldwide, causing
widesp ead anxie y and conce n among he public. The need o con ol he sp ead o he
i us has necessi a ed s ic measu es such as na ionwide qua an ines and inc eased
sani a ion p o ocols, leading o social and economic dis up ion in a ious sec o s
(B odeu , G ay, Islam, and Bhuiyan 2021). In addi ion, ea and panic ha e ueled
p ejudice and disc imina ion agains ce ain g oups (Lu e al. 2021). A he same ime,
emo e wo king and eaching ha e become he no m o many, necessi a ing a majo shi
in wo king and educa ional p ac ices (Bloom, Da is, and Zhes ko a 2021). The pandemic
has also signi ican ly impac ed men al heal h, wi h he isola ion and anxie y caused by
he pandemic becoming a majo conce n (B odeu e al. 2021, Bu e wo h e al. 2022).
Mo eo e , he pandemic has exposed deep-sea ed po e y and heal h dispa i ies wi hin
many socie ies, highligh ing he u gen need o add ess hese issues (Ahmed e al. 2020).
As disease ou b eaks a e no likely o disappea in he nea u u e, i is c ucial o
unde s and p epa edness s a egies ha can educe he impac o u u e diseases.
The COVID-19 pandemic has b ough he impo ance o In e ne access in o sha pe
ocus. P io esea ch sugges s ha In e ne access may play a c ucial ole in heal h
eme gencies (Ba e o, Bloom, and Da is 2021; Alipou , Fadinge , and Schymik 2021;
Ba e o e al. 2021; Whi elaw e al. 2020). Wi h he apid sp ead o he i us, he e was
an u gen need o he quick and e ec i e dissemina ion o in o ma ion o he public. The
In e ne has enabled he apid sha ing o upda es on he i us h ough websi es, social
media, and messaging pla o ms. Addi ionally, he In e ne has p o ided heal h au ho i ies
2
a pla o m o communica e di ec ly wi h he public, sha ing guidance on p e en i e
measu es such as wea ing masks, washing hands, and social dis ancing. Telemedicine
has eme ged as a i al ool in he igh agains COVID-19, allowing heal hca e
p o essionals o p o ide i ual consul a ions, moni o pa ien s emo ely, and educe he
isk o exposu e o he i us in heal hca e acili ies. Howe e , he impac o In e ne access
on pandemic p e en ion may be pa ially posi i e, as i can also acili a e he sp ead o
misin o ma ion. Fo example, alse claims abou he e ec i eness o ce ain ea men s
o he sa e y o accines ha e been widely ci cula ed online, hampe ing e o s o con ain
and mi iga e he disease (Cinelli e al. 2020, Bu sz yn e al. 2020). The e o e, he o e all
e ec o In e ne access on pandemic p e en ion is an empi ical ques ion ha equi es
u he in es iga ion.
In de eloping coun ies whe e he heal h ca e sys em is o en unde de eloped,
In e ne access is c ucial in educing he sp ead o COVID-19. The Wo ld Bank epo s
ha only 19.1% o he popula ion in low-income coun ies has access o he In e ne ,
compa ed o 87.7% in high-income coun ies (Kelly and Rosso o 2011). This limi ed
access o he In e ne exace ba es he challenges aced by he heal h sys ems in hese
coun ies, making i di icul o manage he pandemic e ec i ely. Access o he In e ne is
essen ial o heal hca e deli e y, disease su eillance, and public heal h in e en ions. I
allows people o access eliable in o ma ion abou he i us, p e en i e measu es, and
elemedicine se ices ha educe he bu den on o e wo ked heal h sys ems.
Addi ionally, In e ne access enables people o wo k om home, educing he need
o physical con ac and minimizing he isk o sp eading he i us. The e o e, highe
access o he In e ne is likely associa ed wi h a educ ion in COVID-19 ansmission in
3
de eloping coun ies. Figu e 1 plo s he ela ionship be ween coun y-le el a e age
mobile In e ne speed and he numbe o COVID-19 cases o se e al economies in Asia
and he Paci ic coun ies, sugges ing ha s onge In e ne connec ions a e associa ed
wi h lowe COVID-19 ansmission a es and, consequen ly, dea hs.
This s udy p o ides empi ical e idence o he ole o In e ne access du ing he
COVID-19 pandemic by exploi ing subna ional da a on mobile b oadband, and COVID-
19 sp ead in Indonesia. Indonesia became he epicen e o he pandemic in Sou heas
Asia, wi h he highes numbe o con i med cases and dea hs in he egion by he end o
2022 (Figu e 1 and Appendix A.6). Se e al ac o s con ibu ed o he se e i y o he
si ua ion, including limi ed es ing and acing capaci y, inadequa e heal hca e
in as uc u e, and public skep icism owa ds he go e nmen ’s pandemic esponse
measu es. Mo eo e , he a ia ion in In e ne access ac oss egions and p o inces posed
a challenge in Indonesia, wi h implica ions o educa ion, heal hca e, and economic
oppo uni ies. Al hough some a eas had highe le els o In e ne access, pa icula ly in
u ban a eas and on he islands o Ja a and Bali, o he a eas, pa icula ly u al and emo e
ones, had limi ed In e ne access due o inadequa e in as uc u e and esou ces. This
limi ed access could exace ba e he isk o in ec ion in hose egions, especially du ing
he COVID-19 pandemic, whe e access o go e nmen in o ma ion abou he disease
became inc easingly c ucial. Figu e 2 shows ha a eas wi h highe In e ne access a e
associa ed wi h lowe disease ansmission a es, p o iding u he e idence o he
impo ance o In e ne access in con olling he sp ead o COVID-19 in Indonesia and
o he de eloping economies.
4
Ou empi ical s a egy builds on he a ia ion in 3G mobile b oadband access ac oss
dis ic s in Indonesia in 2019. While he p e-COVID pe iod is less likely o be ela ed o
he sp ead o diseases, conce ns emain abou he co ela ion be ween In e ne access
and o he ac o s ha could in luence he sp ead o he disease. The e o e, we adop an
ins umen al a iable app oach o add ess his issue, using he incidence o ligh ning
s ikes as an ins umen (Gu ie , Melniko , and Zhu a skaya 2021; Manaco da and Tesei
2020; Do, Gomez-Pa a, and Rijke s 2023). Ou s a egy assumes ha a eas wi h highe
incidences o ligh ning display slowe adop ion o mobile phone echnology. A he same
ime, i is easonable o assume ha ligh ning s ikes a e unco ela ed wi h he sp ead o
COVID-19. Howe e , his assump ion may no hold uncondi ionally as ligh ning s ikes
could be co ela ed wi h geog aphical and clima ic a iables o he a ailabili y o
in as uc u es o se ices ha migh ha e an independen e ec on COVID-19 ou comes.
Thus, we con ol o a wide ange o po en ial de e minan s o COVID-19, such as
demog aphic ac o s and heal h in as uc u e, o isola e he e ec o 3G access on he
sp ead o he disease.
We ind ha access o 3G In e ne plays a i al ole in educing he ansmission a e
o COVID-19 in Indonesia. Ou s udy shows ha he numbe o COVID-19 cases is
app oxima ely 25% lowe in a eas whe e 3G In e ne access is a ailable. Ou indings
hold e en when we con ol o o he ac o s ha could a ec he i us’s sp ead and
conside he po en ial endogenei y o In e ne access. One possible explana ion o his
esul is ha In e ne access p o ides indi iduals wi h easy access o accu a e in o ma ion
abou he i us, such as how o p e en in ec ion, whe e o ge es ed, and wha o do i
hey become ill. This in o ma ion can help people make in o med decisions abou hei
11
geog aphic da a. The OSM da a o Indonesia a e ins umen al in quan i ying he numbe
o heal hca e acili ies ac oss a ious geog aphic loca ions, speci ically clinics, and
hospi als. Clinics a e de ined as heal h acili ies whe e ou pa ien medical ca e is
p o ided, o en by gene al p ac i ione s o specialis s o amily o in e nal medicine.
Con e sely, hospi als a e mo e ex ensi e acili ies ha p o ide a b oade ange o
se ices, including inpa ien ca e, specialized su ge ies, and eme gency se ices. This
da ase o e s a comp ehensi e spa ial dis ibu ion o heal hca e in as uc u es ac oss
he na ion, hus enabling he iden i ica ion o po en ial dispa i ies in he p o ision o heal h
se ices. In conjunc ion wi h he OSM da a, we also inco po a e local heal hca e spending
da a a he egency (o ci y) le el, a ailable om 2017 o 2021.
Economic Indica o s
To es ima e local economic ac i i y in Indonesian dis ic s, we u ilize sa elli e nigh ligh
da a om he Visible In a ed Imaging Radiome e Sui e (VIIRS) adminis e ed by he
Na ional Oceanic and A mosphe ic Adminis a ion (NOAA). This me hodology has gained
popula i y in economic esea ch as a eliable p oxy o g ow h ou comes (Hodle and
Raschky 2014; Hende son, S o eyga d, and Weil 2012) and was ound o be supe io o
o he economic a iables in some ci cums ances (Ma inez 2022).2 We calcula e he
nigh ligh densi y o all Indonesian dis ic s by agg ega ing sa elli e images om daily
g ids o yea ly da a.3
2 In he con ex o Indonesia, Gibson e al. (2021) ind a posi i e ela ionship be ween dis ic -le el VIIRS
da a and g oss domes ic p oduc (GDP).
3 A mosphe ic condi ions may impac he abili y o sa elli e senso s o cap u e nigh ligh s. We ollow he
Cope nicus p og am’s ecommenda ion o exclude esul s om pixels wi h abo e 10% cloud ac ion by
pe o ming cloud masking o add ess his. Fo mo e de ails: Cope nicus. 2020. “Flawed Es ima es o he
E ec s o Lockdown Measu es on Ai Quali y De i ed om Sa elli e Obse a ions.” Ma ch 26.
h ps://a mosphe e.cope nicus.eu/ lawed-es ima es-e ec s-lockdown-measu es-ai -quali y-de i ed-
sa elli e-obse a ions?q= lawed-es ima es-e ec s-lockdown-measu es-ai -quali y-sa elli e-obse a ions.

12
We add he ecen ly made a ailable sub-na ional human de elopmen index compiled
o Indonesia in 2019 on he egency le el, cap u ing economic s a us, human well-being,
and lou ishing mo e b oadly.4
Wea he Da a
We ob ain wea he da a om ERA5, which is he i h gene a ion o eanalysis da ase
p oduced by he Eu opean Cen e o Medium-Range Wea he Fo ecas s (ECMWF).
Using eanalysis da a has se e al bene i s o ou s udy. Fi s ly, wea he s a ions’ spa ial
and empo al co e age is o en limi ed in many de eloping coun ies. In con as ,
eanalysis da a co e s a la ge geog aphical a ea and is a ailable o e a mo e ex ended
pe iod. Secondly, eanalysis da a may esol e issues ela ed o wea he da a, such as
endogenei y conce ns associa ed wi h wea he s a ions placemen and a ia ions in he
quali y and quan i y o da a collec ion, as no ed in p e ious s udies (Au hamme e al.
2013, Donaldson and S o eyga d 2016). ERA5 inco po a es in o ma ion om a ious
sou ces, including g ound s a ions, sa elli es, wea he balloons, and clima e models, o
p o ide wea he da a om 1979 onwa ds. The da a has a high spa ial esolu ion o 31
kilome e s (km) and has been s anda dized o a egula la i ude-longi ude g id o 0.25
deg ees. Fo his s udy, we use he yea ly empe a u e (measu ed in Celsius) and
p ecipi a ion (measu ed in millime e s) da a, which we agg ega e a he egency le el o
Indonesia using he in e se-dis ance weigh ing app oach (Deschenes and G eens one
2011).
As an ins umen o In e ne co e age, we employ da a on ligh ning s ikes sou ced
om he Wo ld Wide Ligh ning Loca ion Ne wo k (WWLLN) da ase (Kaplan and Lau
4 The sub-na ional indexes a e s ill ela i ely a e. An al e na i e app oach o o he con ex s, a ecen
machine lea ning app oach using sa elli e da a, ha e been p oposed (She man e al. 2023).
13
2021). This da ase p o ides de ailed g id-le el da a o ligh ning s ikes, wi h a p ecision
o 0.5◦ × 0.5◦ pe g id. We hen agg ega e hese da a om he g id le el up o he egency
le el o Indonesia. Ou measu e o ligh ning s ike is he mean s oke powe in megawa s
(MW), agg ega ed a he yea ly basis. This measu e p o ides a unique pe spec i e on
he in ensi y and ene gy o ligh ning ac i i y, which we hypo hesize o be a ele an p oxy
o assessing In e ne co e age in he egion.
3 Empi ical Model
We aim o in es iga e he e ec s o In e ne access, p oxied by p e-pandemic ( 0) 3G
a ailabili y (G3access , 0), on educing COVID-19 cases and dea hs a he egency le el
in Indonesia. We adop he me hodology p oposed by Kunz and P oppe (2022) o
assess he o e - ime associa ion be ween COVID-19 ou comes and p e-pandemic
access o he In e ne , such ha
𝐸𝐸�𝑦𝑦𝑟𝑟,𝑡𝑡�𝑋𝑋� =exp(𝛼𝛼𝑡𝑡+ 𝜏𝜏𝑡𝑡𝐺𝐺3𝑓𝑓𝑓𝑓𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎𝑟𝑟,𝑡𝑡0+ 𝑥𝑥𝑟𝑟,𝑡𝑡0
′𝛽𝛽𝑡𝑡+ 𝛿𝛿𝑝𝑝,𝑡𝑡 (1)
whe e 𝑦𝑦𝑟𝑟,𝑡𝑡 is he cumula i e cases (o dea hs) in egency (o ci y) in mon h , adjus ed
by 10,000 popula ion. We con ol o a se o co a ia es (𝑥𝑥𝑟𝑟,𝑡𝑡0
′), all measu ed on he
egency le el and be o e he pandemic’s s a . These include demog aphic ac o s, heal h
acili ies, economic ac i i ies, and wea he condi ions. Addi ionally, we include p o ince
ixed e ec s (δp) o con ol o unobse able ac o s a ec ing ou ou comes a he p o ince
le el ( he main le el o local policymaking). We employ obus s anda d e o s o accoun
o po en ial he e oskedas ici y when es ima ing epea ed c oss-sec ions and clus e ed a
he egency le el when es ima ing pooled models.
14
In ou p ima y analysis, we concen a e on he las da e in ou da ase (i.e., 18
Feb ua y 2023) and execu e a single c oss-sec ional eg ession. Howe e , we also
es ima e Equa ion (1) o each mon h sepa a ely o illus a e he ime ends in he
associa ion be ween In e ne access and COVID-19 ou comes. We u ilize a Poisson
eg ession model as ou p ima y speci ica ion, acknowledging he hea ily skewed na u e
o he cases (Figu e A.2). To ensu e he obus ness o ou indings, we e alua e
al e na i e model speci ica ions, including log- ans o med OLS and nega i e binomial
eg essions. We ul ima ely choose he Poisson model owing o i s bene icial p ope ies,
such as being pa o he linea , exponen ial amily, demons a ing obus ness o
misspeci ica ion, no necessi a ing ad-hoc ans o ma ion, and a oiding issues ela ed o
inciden al pa ame e p edic ion (Sil a and Ten ey o 2006 discuss his in de ail).
Like any o he in as uc u e, he expansion o mobile ne wo k co e age is likely
in luenced by endogenous ac o s. Fo example, p o ide s o ne wo k se ices end o
p io i ize a eas wi h high economic ac i i y and po en ial demand. Howe e , his end
may be linked o o he ac o s ha may, in u n, de e mine he sp ead o he pandemic.
As a esul , simple condi ional associa ions be ween mobile phone pene a ion and
COVID-19 sp ead migh no co espond o he causal impac . To add ess his issue, we
es ima e he ollowing i s s age o he Poisson-IV model o mi iga e he issue o
endogenei y:
𝐺𝐺3𝑓𝑓𝑓𝑓𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎𝑟𝑟,𝑡𝑡0=𝛽𝛽+𝛾𝛾1𝑓𝑓�𝐿𝐿𝑓𝑓𝐿𝐿ℎ𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝐿𝐿𝑟𝑟,𝑡𝑡0�+𝛾𝛾2𝑓𝑓𝑎𝑎𝑡𝑡𝑝𝑝𝑎𝑎𝑓𝑓𝑓𝑓𝑓𝑓𝑡𝑡𝑓𝑓𝑎𝑎𝑟𝑟,𝑡𝑡0+𝛾𝛾3𝑝𝑝𝑓𝑓𝑎𝑎𝑓𝑓𝑓𝑓𝑝𝑝𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑟𝑟,𝑡𝑡0+
𝑥𝑥𝑟𝑟,𝑡𝑡0
′𝛽𝛽+ 𝛿𝛿𝑝𝑝+ 𝜖𝜖𝑟𝑟,𝑡𝑡0 (2)
15
whe e he ins umen al a iable, 𝐿𝐿𝑓𝑓𝐿𝐿ℎ𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝐿𝐿𝑟𝑟, is measu ed by he a e age ligh ning s oke
powe a he egency le el. Ou empi ical s a egy is hus in line wi h a ecen s udy using
he Poisson-IV es ima o (G a Zi in e al. 2023).
We now discuss he alidi y o he ins umen employed in ou analysis. A good
ins umen al a iable should sa is y se e al condi ions, including exogenei y, ele ance,
and exclusion. Rega ding he exogenei y condi ion, ligh ning s ikes a e a na u al
phenomenon ha occu s independen ly o COVID-19 sp ead, making hem an exogenous
ac o in ela ion o ou ou come o in e es . Addi ionally, i is impo an o no e ha we
measu e ligh ning s ikes du ing he p e-pandemic pe iod, u he ensu ing he exogenei y
o ou ins umen .
In e ms o he ele ance condi ion, ou ins umen is s ongly co ela ed wi h In e ne
access, as suppo ed by p e ious s udies (Ande sen e al. 2012; Gu ie , Melniko , and
Zhu a skaya 2021; Manaco da and Tesei 2020; Do, Gomez-Pa a, and Rijke s 2023).
Speci ically, ligh ning s ikes can cause signi ican damage o digi al in as uc u e,
leading o highe cos s associa ed wi h digi al echnology di usion. This co ela ion is
pa icula ly ele an in a eas wi h a highe equency o ligh ning s ikes, whe e challenges
in se ing up and main aining In e ne in as uc u e may a ise. Fu he mo e, i is wo h
no ing ha his IV is ideally sui ed o ou con ex , as o coun ies a ound he equa o , as
much o he IV’s a ia ion is om his egion (Figu e B.1).
Howe e , he e is he possibili y o ligh ning s ikes a ec ing o he in as uc u e ha
could indi ec ly in luence he sp ead o COVID-19 du ing he pandemic. While he
exclusion condi ion is no di ec ly es able, we ha e aken measu es o con ol o
po en ial con ounding ac o s, such as wea he condi ions, humidi y, he numbe o cell
16
owe s, and economic ac i i ies in ou models. These addi ional con ols help o mi iga e
he po en ial indi ec e ec s o ligh ning s ikes on COVID-19 sp ead by cap u ing o he
ele an ac o s ha could media e his ela ionship.
Finally, we modi y Equa ion (1) by including in e ac ion e ms be ween In e ne access
and a ange o ac o s o examine he he e ogeneous e ec s o In e ne access on
COVID-19 ansmission.
𝐸𝐸�𝑦𝑦𝑟𝑟,𝑡𝑡�𝑋𝑋� =exp(𝛼𝛼𝑡𝑡+ 𝜏𝜏𝑡𝑡𝐺𝐺3𝑓𝑓𝑓𝑓𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎𝑟𝑟,𝑡𝑡0+ 𝜏𝜏𝑡𝑡
𝑥𝑥𝐺𝐺3𝑓𝑓𝑓𝑓𝑓𝑓𝑎𝑎𝑎𝑎𝑎𝑎𝑟𝑟,𝑡𝑡0× 1[𝑥𝑥 ≥ 𝑡𝑡𝑎𝑎𝑚𝑚𝑥𝑥]+ 𝑥𝑥𝑟𝑟,𝑡𝑡0
′𝛽𝛽𝑡𝑡+ 𝛿𝛿𝑝𝑝,𝑡𝑡
(3)
In his equa ion, 1[𝑥𝑥 ≥ 𝑡𝑡𝑎𝑎𝑚𝑚𝑥𝑥] ep esen s an indica o o whe he a iable 𝑥𝑥 is la ge o
equal o he median. We concen a e on he p ima y hypo hesis conce ning how In e ne
access migh a ec he sp ead o COVID-19 a he egency le el. Fo he he e ogenei y
analysis, we employ bo h he Poisson eg ession as in Equa ion (1) and he analogous
Poisson-IV eg essions, using 𝑓𝑓�𝐿𝐿𝑓𝑓𝐿𝐿ℎ𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝐿𝐿𝑟𝑟,𝑡𝑡0�× 1[𝑥𝑥 ≥ 𝑡𝑡𝑎𝑎𝑚𝑚𝑥𝑥] as an addi ional
ins umen . We ocus on a iables deemed impo an in he li e a u e ci ed he e, such as
educa ion le el, li ing a eas, heal h acili ies, economic ac i i ies, and labo o ce
cha ac e is ics.
4 Resul s
Main Findings
Table 1 p esen s esul s om Equa ion (1) o inc easing se s o co a ia es, p o iding a
comp ehensi e analysis o he ole o In e ne access in educing COVID-19
ansmission. In column 1, he unadjus ed associa ion is la ge, nega i e, and s a is ically
highly signi ican . This inding sugges s ha In e ne access, p oxied by 3G a ailabili y, is

17
indeed associa ed wi h a educ ion in COVID-19 cases a he egency le el in Indonesia.
Howe e , he associa ion sligh ly educes when including p o ince-le el ixed e ec s,
which accoun o policy-making di e ences (column 2).
Subsequen ly, we add co a ia es likely o a ec he sp ead o COVID-19. In doing so,
we aim o con ol o po en ial con ounding ac o s and isola e he speci ic e ec o In e ne
access on pandemic ou comes. Columns 3 o 7 esul s demons a e ha he associa ion
be ween In e ne access and COVID-19 ansmission emains obus e en when
accoun ing o a ious demog aphic, heal hca e, economic, e hnic, and labo o ce
cha ac e is ics. The mos subs an ial d op in he associa ion occu s when including
demog aphic a iables, a e which he associa ion emains ela i ely s able. The pseudo
R2 inc eases conside ably, pa icula ly when he analysis employs ixed e ec s and
demog aphic a iables.
A ew o hese a iables con ain missing alues, which educes he sample size. To
add ess his issue, we include hese a iables ei he by impu ing ze os along wi h
indica o s o missing alues (column 8) o using he mul iple impu a ion me hod p oposed
by Rubin (1996) wi h 99 andom d aws (column 9). Bo h app oaches yield nea ly iden ical
esul s, con i ming he obus ness o ou indings. Ou p e e ed model is column 8, and
we will ocus on his speci ica ion in he emainde o he ex . By accoun ing o missing
da a in ou analysis, we ensu e a mo e comp ehensi e assessmen o he ela ionship
be ween In e ne access and COVID-19 ansmission, u he enhancing he alidi y and
gene alizabili y o ou conclusions.
In e ms o in e p e a ion, we demons a e ha ha ing access o 3G In e ne is
associa ed wi h a subs an ial educ ion in COVID-19 cases by app oxima ely 25% using
18
ou p e e ed es ima ion in column 7 o Table 1. The magni ude o he e ec is
no ewo hy, pa icula ly in he con ex o a de eloping coun y such as Indonesia, whe e
esou ces o public heal h in e en ions may be limi ed. Fu he mo e, he magni ude o
he e ec obse ed in ou s udy is pa icula ly ele an compa ed o he e ec i eness o
o he non-pha maceu ical in e en ions (NPIs) in educing COVID-19 ansmission. While
he impac o speci ic NPIs a ies ac oss s udies, he 25% educ ion associa ed wi h
In e ne access is compa able o, o e en g ea e han, he e ec s epo ed o some
widely implemen ed measu es such as social dis ancing, mask-wea ing, and a el
es ic ions (VoPham e al. 2020, Leech e al. 2022, Kwok e al. 2021). Fo example,
VoPham e al. (2020) show ha highe social dis ancing was associa ed wi h a 29%
educ ion in COVID-19 incidence.
These esul s ha e impo an policy implica ions, especially o de eloping coun ies
ha ace challenges in con aining he sp ead o in ec ious diseases like COVID-19. Ou
indings sugges ha imp o ing In e ne access can educe disease ansmission, as i
acili a es be e dissemina ion o heal h- ela ed in o ma ion, enables emo e wo k and
educa ion, and p omo es elemedicine se ices, which collec i ely educe he need o
physical in e ac ion and help limi he sp ead o he i us.
Figu e 4 suppo s ou main indings by consis en ly demons a ing he e ec s o
In e ne access on COVID-19 ansmission h oughou he pandemic. Fu he mo e, he
s abili y o hese e ec s o e ime sugges s ha ac o s such as accine dis ibu ion and
po en ial da a inconsis encies a he pandemic’s beginning a e unlikely o signi ican ly
in luence he obse ed ela ionship be ween In e ne access and COVID-19 ou comes.
19
No ably, he s able es ima es o e ime indica e ha he bene i s o In e ne access in
mi iga ing he sp ead o he i us ha e pe sis ed h oughou he a ious s ages o he
pandemic. This consis ency ein o ces he impo ance o digi al connec i i y as a c ucial
ac o in public heal h s a egies, bo h du ing he ini ial esponse o an ou b eak and in
ongoing e o s o con ol and manage he sp ead o in ec ious diseases. As he pandemic
e ol es and new challenges a ise, such as new a ian s and changes in public heal h
guidelines, he pe sis en e ec s o In e ne access on educing COVID-19 ansmission
unde sco e i s endu ing signi icance in suppo ing e ec i e public heal h in e en ions.
Robus ness
Table 2 demons a es ha he e ec s p esen ed in ou main analysis a e highly obus o
a ious speci ica ions and al e na i e app oaches. These include Log-OLS (column 2),
Nega i e Binomial eg ession (column 3), al e na i e measu es o 3G access (impu ed
e sus no impu ed) (columns 4 and 5), no using popula ion weigh s (column 6), and
impu ing he ou comes o a eas ha did no epo any cases (columns 7 and 8).5 Ou
p e e ed speci ica ion is a guably he mos conse a i e among hese app oaches. Using
he log- ans o med OLS app oach is simila in e ms o he ma ginal e ec . The
obus ness o ou indings ac oss a ious speci ica ions and al e na i e me hods
s eng hens ou con idence in he highly s able associa ion be ween In e ne access and
he sp ead o COVID-19.
Speed o Access
Table 3 p esen s he esul s o di e en In e ne speeds, including 2G, 3G, and 4G. In
he con ex o he pandemic, hese di e ences in capabili ies and da a speeds may ha e
5 Columns 7 and 8 di e as o some, we obse e p o ince and popula ion. Fo he emainde , we impu e
popula ion in column 8.
20
signi ican implica ions o he e ec i eness o a ious non-pha maceu ical in e en ions,
such as emo e wo k, online lea ning, and he dissemina ion o public heal h in o ma ion.
While 2G p ima ily suppo s basic se ices such as SMS and oice calls, 3G enables
mobile In e ne access, and 4G o e s imp o ed da a speeds and enhanced mobile
In e ne expe iences. Access o highe -speed In e ne , such as 3G and 4G, allows
indi iduals and communi ies o be e adap o he challenges posed by he pandemic and
adhe e o public heal h guidelines while main aining social and economic ac i i ies. Ou
p e e ed measu e, 3G, exhibi s he mos subs an ial ela ionship wi h he pandemic
expe ience. When we include all h ee ypes o In e ne access simul aneously, 2G loses
i s signi icance, d opping close o ze o, and appea s less ele an ega ding i s impac on
COVID-19 ansmission. Con e sely, 4G demons a es a simila e ec o ou main
measu e, 3G, al hough i s associa ion wi h he pandemic expe ience is again weake .
Ano he aspec wo h conside ing is he po en ial non-linea e ec o In e ne access
on COVID-19 ansmission. This may be impo an because he ela ionship be ween
In e ne access and COVID-19 ou comes could be cha ac e ized by diminishing e u ns
o h eshold e ec s, whe e he impac o inc eased In e ne access on ansmission a es
pla eaus o e en e e ses beyond a ce ain poin . Fo ins ance, e y high le els o In e ne
access could exace ba e he sp ead o misin o ma ion o con ibu e o complacency in
ollowing public heal h guidelines, ul ima ely leading o ad e se e ec s on COVID-19
ansmission. Howe e , ou analysis does no ind clea e idence o non-linea e ec s
(Table A.1). Including a second-o de polynomial e m does no e eal a non-linea
ela ionship. This may be due o da a limi a ions, gi en ha we only ha e 454
obse a ions and conside able co a iance among he a iables. When we use an
27
ou indings unde sco e he po en ial ole o digi al in as uc u e in public heal h
esponses o pandemics and o he heal h c ises.
In conclusion, ou s udy unde sco es he i al ole o echnology in ackling he COVID-
19 pandemic and ad ancing public heal h in he digi al age. The impo ance o digi al
connec i i y in oday’s wo ld canno be o e s a ed, and ou esea ch calls o a enewed
ocus on his essen ial ace o mode n li e in u u e public heal h s a egies.

28
TABLES AND FIGURES
Table 1: Associa ions be ween G3 In e ne Exposu e and COVID-19 Cases:
La e Pandemic S age
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
Adding
co a ia es
Impu a ion
Raw
+P o .
FE
+Demo-
g aphics
+Heal h
acili ies
+Econ.
s a us
+E hnic
comp.
+Labo
o ce
ia:
Se 0 Mul iple
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8) (9)
In e ne exposu e -1.495
-1.125
-0.651
-
0.606 -0.611
-
0.690
-
0.787
-
0.630
-
0.630
(0.222)
(0.417)
(0.323)
(0.315)
(0.285)
(0.290)
(0.313)
(0.255) (0.255)
Semi-elas ici y (1sd)
-
0.32
-
0.28
-
0.20
-0.19
-0.19
-
0.21
-
0.23
-
0.20
N
454
454
454
345
323
321
321
454 454
Mean dep.
0.50
0.50
0.50
0.40
0.42
0.42
0.42
0.50 0.50
SD dep.
1.45
1.45
1.45
0.87
0.89
0.89
0.89
1.45
1.45
pR
2
0.05
0.41
0.51
0.37
0.38
0.38
0.38
0.53
P o ince FE
✓
✓
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
Labo o ce
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s, GSMA = Mobile Communica ions Associa ion,
SD = s anda d de ia ion.
No es: This able p esen s coe icien s es ima es om equa ion (1), o eg essions o cumula i e
COVID-19 cases pe 10,000 popula ion on he egency le el (18 Feb ua y 2023) and inc easing
a ious se s o co a ia es all measu e be o e he pandemic begins, o a desc ip ion o he co a ia es
(Table B.1). A he bo om, we p esen pR2 pseudo (ml) R2, he mean and s anda d de ia ion o he
ou come a iable, and he semi-elas ici y o 1 sd change in in e ne exposu e, ha is (exp(τ ) − 1)
∗
sd(In e ne exposu e). Column (1) depic s he aw associa ion, Column (2) adds p o ince (38) ixed
e ec s, h ee demog aphic a iables (popula ion densi y, popula ion, male- emale a io, he sha e o
he popula ion aged 65 and olde , he sha e o people wi hou any educa ion, he sha e o people
wi h g adua e deg ees, log household size, u al a ea), (4) adds heal h in as uc u e measu es
(numbe o clinics, numbe o hospi als, and log o heal h ca e spending), (5) adds economic
indica o s (second o de polynomial o nigh ligh , sub-na ional human de elopmen index - sco e),
(6) e hnici y (e hnic di e si y, and pola iza ion), and (7) labo o ce a iables (sha e o wo ke s able
o elewo k, he sha e o wo ke s wi h a long dis ance (>1hou ) commu e, and ha use public
anspo a ion, and he sha e wo king in ag icul u e). The inal wo columns eplace missing alues
in co a ia es wi h 0 and an indica o o missing alues (8), and (9) implemen s he mul iple
impu a ion app oach by Rubin (1996). All eg essions a e weigh ed by popula ion and use obus
s anda d e o s.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task
Fo ce, using measu emen s aken as o 18 Feb ua y 2023. Mobile in e ne da a was ob ained om
Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
29
Table 2: Tes ing Addi ional Robus ness
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
Log-cases Neg. Al . 3G measu e Wi hou Missing Impu e
Base
OLS
Bin.
Plain
Impu ed
weigh s
Ou comes
All
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
In e ne exposu e
-
0.630
-
0.365
-
0.630
-
1.005
-
0.627
-
0.627
(3G)
(0.255)
(0.200)
(0.255)
(0.226)
(0.255)
(0.255)
In e ne exposu e
-
1.090
(3G-
MCE
)
(0.491)
In e ne exposu e
-1.126
(3G-MCE)-impu ed
(0.496)
Semi-elas ici y (1sd)
-
0.20
-0.13
-
0.20
-0.13
-0.13
-
0.26
-
0.20
-0.19
N
454
454
454
365
454
454
465
510
Mean dep.
0.50
0.50
0.50
0.60
0.50
0.50
0.49
0.45
SD dep.
1.45
1.45
1.45
1.60
1.45
1.45
1.43
1.37
pR2
0.53
0.46
0.54
0.53
0.57
0.53
P o ince FEs
✓
✓
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
✓
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s, GSMA = Mobile Communica ions Associa ion,
OLS = o dina y leas squa es, SD = s anda d de ia ion.
No es: Table 1 column (8) p esen s he main model, see no es he ein, ep esen ed in column (1).
Column (2) p esen s an OLS eg ession on he logged ou come measu e— he ma ginal e ec is
adjus ed ia exp(τ ) − 1. Column (3) al e na i ely p esen s a nega i e binominal model, (4) uses he
al e na i e exposu e measu e and (Mobile Co e age Explo e - MCE) is sou ced di ec ly om he
ne wo k ope a o s and hus incu s gaps in co e age, which we impu e wi h 0 and add missing
indica o in (5). Column (6) d ops he popula ion-weigh ing and (7) and (8) impu e a eas wi h missing
obse a ion in he ou come (all a iables, espec i ely) wi h 0.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task
Fo ce, using measu emen s aken as o 18 Feb ua y 2023. Mobile in e ne da a was ob ained om
Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
30
Table 3: Associa ion be ween Di e en In e ne Speeds and Po en ial Non-Linea
E ec s
Dependen a iables: Cumula i e cases by 10T
popula ion 18 Feb 2023
Base
2G
4G
Join ly
(1)
(2)
(3)
(4)
In e ne exposu e
-
0.630
-
0.530
(3G)
(0.255)
(0.319)
In e ne exposu e
-
0.366
0.104
(2G)
(0.294)
(0.296)
In e ne exposu e
-
0.472
-
0.372
(4G)
(0.202)
(0.200)
Semi-elas ici y (1sd)
-
0.20
-
0.09
-0.13
[1em] N
454
454
436
436
Mean dep.
0.50
0.50
0.52
0.52
SD dep.
1.45
1.45
1.47
1.47
pR2
0.53
0.53
0.53
0.54
P o ince FEs
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s, GSMA = Mobile Communica ions Associa ion,
SD = s anda d de ia ion.
No es: The Table p esen s coe icien s analogous o Table 1-Column (8)— ep esen ed in Column
(1). Column (2) eplaces ou main measu e wi h 2G exposu e, and (3) wi h 4G, Column (4)
es ima es he model join ly.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-
19 Task Fo ce, using measu emen s aken as o Feb ua y 18 h, 2023. Mobile in e ne da a was
ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
31
Table 4: IV Resul s
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
IV
Reduced
5 km ad.
Main
o m
dec 2019
(1)
(2)
(3)
Panel A. Cases
In e ne exposu e
-
0.660
-
1.440
(3G)
(0.257)
(0.394)
Ligh ning s ike equency (5 km)
7.919
(2.429)
Semi-elas ici y (1sd)
-
0.20
-
0.32
Panel B. Fi s s age
Ligh ning (5 km, Dec 2019)
-
4.050
(0.518)
N
454
454
454
Di e ence pe million o 1sd inc ease
-1.66
-
2.50
Fs a
61.05
P o ince FEs
Demog aphics
Heal h acili ies
Economic s a us
E hnic composi ion
Labo o ce
Wea he con ols
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s, GSMA = Mobile Communica ions Associa ion,
km = kilome e , SD = s anda d de ia ion.
No es: The Table p esen s coe icien s es ima es om equa ion (1) using IV-Poisson wi h con ol
unc ion (2). Column (1) adjus s Table 1 Column (8) by adding wea he co a ia es. Column (2)
shows he educed o m o ligh ning s ikes in he 5 km adius in Decembe 2019,
con empo aneously wi h he 3G mobile exposu e. The bo om o he Table p edic s he di e ence
in cases o one s d. De ia ion in in e ne exposu e and he F-s a is ic o he i s s age.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-
19 Task Fo ce, using measu emen s aken as o 18 Feb ua y 2023. Mobile in e ne da a was
ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
32
Figu e 1: Mobile B oadband Speed and COVID-19 Cases,
Asia and he Paci ic
COVID-19 = co ona i us disease, Lao PDR = Lao People’s Democ a ic Republic, PRC = People’s Republic
o China.
No es: COVID-19 da a we e measu ed as o 1 Janua y 2023, and adjus ed o popula ion. Mobile
b oadband speed (in Mbps) was measu ed in 2020.
Sou ces: COVID-19 da a om he Indonesian COVID-19 Task Fo ce. Mobile b oadband speed a coun y-
le el om h ps://wo ldpopula ion e iew.com/coun y- ankings/in e ne -speeds-by-coun y

33
Figu e 2: Mobile B oadband Access and COVID-19 cases, Indonesia
COVID-19 = co ona i us disease, PRC = People’s Republic o China.
No es: COVID-19 da a we e measu ed as o Feb ua y 18 h, 2023, and adjus ed o popula ion. Mobile
b oadband speed (3G access) was measu ed in 2019.
Sou ces: COVID-19 da a om he Indonesian COVID-19 Task Fo ce. Mobile b oadband speed om
h ps://wo ldpopula ion e iew.com/.
34
Figu e 3: Numbe o COVID-19 Cases and Mobile Co e age, Indonesia
COVID-19 = co ona i us disease, GSMA = Mobile Communica ions Associa ion.
No es: This igu e shows cumula i e COVID-19 cases on 18 Feb ua y 2023 ( op) and 3G mobile ne wo k exposu e
(bo om) ac oss egencies in Indonesia.
The bounda ies, colo s, denomina ions, and any o he in o ma ion shown on his map do no imply, on he pa o he
Asian De elopmen Bank, any judgmen on he legal s a us o any e i o y, o any o he endo semen o accep ance
o such bounda ies, colo s, denomina ions, o in o ma ion.
Sou ces: COVID-19 da a om he Indonesian COVID-19 Task Fo ce. Mobile in e ne da a om Collins Ba holomew’s
GSMA Mobile Co e age Explo e da abase.
35
Figu e 4: Associa ion be ween Numbe o COVID-19 Cases and Dea hs, and
Mobile Co e age, Indonesia h oughou Pandemic
COVID-19 = co ona i us disease, GSMA = Mobile Communica ions Associa ion.
No es: The igu e p esen s eg ession coe icien s o eg essions (90-da k and 95-ligh le el con idence
in e als) p esen ed in Table 1 (see no es he ein) column 8, sepa a ely o a ious da es h oughou he
pandemic co e ing a ious wa es o new COVID-19 a ian s. Round ma ke s depic cases and squa es
dea hs.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task Fo ce.
Mobile in e ne da a was ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
36
Figu e 5: Placebo Tes —Local Heal h Ca e Spending by 10,000 Popula ion
GSMA = Mobile Communica ions Associa ion.
No es: This igu e p esen s analogous eg essions o Table 1 Column (8) using as ou come egional heal h
ca e spending ac oss yea s be o e and du ing he pandemic. Spending is acco dingly d opped om he se
o con ols.
Sou ces: Heal h ca e spending da a, adjus ed o popula ion o e he yea s. Mobile in e ne da a was
ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
43
Figu e A.7: Mobile B oadband Access and COVID-19 Dea hs, Indonesia
COVID-19 = co ona i us disease.
No es: COVID-19 da a we e measu ed as o 18 Feb ua y 2023, and adjus ed o popula ion. Mobile
b oadband speed (3G access) was measu ed in 2019.
Sou ces: COVID-19 da a om he Indonesian COVID-19 Task Fo ce. Mobile in e ne da a om Collins
Ba holomew’s GSMA Mobile Co e age Explo e da abase.

44
Figu e A.8: Numbe o COVID-19 Dea hs, Indonesia
COVID-19 = co ona i us disease.
No es: COVID-19 da a come om he Indonesian COVID-19 Task Fo ce, using measu emen s aken as o
18 Feb ua y 2023.
The bounda ies, colo s, denomina ions, and any o he in o ma ion shown on his map do no imply, on he
pa o he Asian De elopmen Bank, any judgmen on he legal s a us o any e i o y, o any o he
endo semen o accep ance o such bounda ies, colo s, denomina ions, o in o ma ion.
Sou ces: COVID-19 da a om he Indonesian COVID-19 Task Fo ce. Mobile in e ne da a om Collins
Ba holomew’s GSMA Mobile Co e age Explo e da abase.
45
Figu e B.1: Example o Wo ldwide Ligh ning S ikes
No e: The bounda ies, colo s, denomina ions, and any o he in o ma ion shown on his map do no imply,
on he pa o he Asian De elopmen Bank, any judgmen on he legal s a us o any e i o y, o any o he
endo semen o accep ance o such bounda ies, colo s, denomina ions, o in o ma ion.
Sou ce: Wo ld Wide Ligh ning Loca ion Ne wo k. h ps://wwlln.ne /.
46
Table A.1: Associa ion be ween Di e en In e ne Speeds and Po en ial
Non-linea E ec s
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
Replica ed om main ex
Base
2G
4G
Join ly
Polynomial
Indica o
(1)
(2)
(3)
(4)
(5)
(6)
In e ne exposu e (3G)
-0.630
-0.530
-1.138
(0.255)
(0.319) (0.712)
In e ne exposu e (2G)
-0.366
0.104
(0.294)
(0.296)
In e ne exposu e (4G)
-0.472
-0.372
(0.202)
(0.200)
In e ne exposu e2 (3G)
0.508
(0.580)
1[In e ne exposu e > p75] (3G)
-0.252
(0.159)
N
454
454
436
436
454
454
pR2
0.53
0.53
0.53
0.54
0.53
0.53
Mean dep.
0.50
0.50
0.52
0.52
0.50
0.50
SD dep.
1.45
1.45
1.47
1.47
1.45
1.45
Ma ginal e ec
-0.24
-0.14
-0.18
-0.43
P o ince FEs
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s.
No es: See Table 3’s no es.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task
Fo ce, using measu emen s aken as o 18 Feb ua y 2023. Mobile in e ne da a was ob ained om
Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
47
Table A.2: Role o In e ne o e Di e en Pandemic S ages
Dependen a iables: Cumula i e cases by 10T popula ion, di e en pandemic s ages
23 Aug
2020
23 Feb 2021
23 Aug 2021
23 Feb 2022
23 Aug 2022
18 Feb 2023
(1)
(2)
(3)
(4)
(5)
(6)
Panel A. Cases
In e ne exposu e
-1.160
-1.093
-
0.705
-
0.536
-
0.493
-
0.630
(0.303)
(0.274)
(0.271)
(0.247)
(0.245)
(0.255)
N
442
453
453
401
402
454
pR2
0.07
0.42
0.77
0.91
0.94
0.94
Mean dep.
0.01
0.10
0.32
0.42
0.48
0.50
SD dep.
0.04
0.29
0.80
1.19
1.41
1.45
Panel B. Dea hs
In e ne exposu e
-2.133
-1.387
-1.216
-
0.851
-
0.843
-
0.987
(0.466)
(0.321)
(0.316)
(0.300)
(0.299)
(0.296)
N
294
397
408
375
376
417
pR2
0.01
0.02
0.04
0.04
0.04
0.05
Mean dep.
0.00
0.00
0.01
0.01
0.01
0.01
SD dep.
0.00
0.01
0.02
0.02
0.02
0.03
P o ince FEs
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
✓
✓
COVID-19 = co ona i us disease, FE = Fixed e ec s, GSMA = Mobile Communica ions Associa ion,
km = kilome e , SD = s anda d de ia ion.
No es: See Table 3’s no es.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task
Fo ce. Mobile in e ne da a was ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e
da abase.
48
Table A.3: He e ogenei y: Poisson
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
In e ne exposu e
-0.518
-0.710
-0.509
-0.842
-0.750
-0.420
-0.203
-0.743
-0.565
-0.754
-0.675
-0.552
-0.446
-0.761
(0.284)
(0.301)
(0.251)
(0.268)
(0.248)
(0.308)
(0.270)
(0.289)
(0.256)
(0.260)
(0.267)
(0.286)
(0.297)
(0.271)
× Sha e o popula ion wi h no educa ion
-0.256
(0.198)
× Sha e o popula ion wi h g ad. deg ee
0.243
(0.207)
× Log a e age household size
-0.264
(0.197)
× Sha e o popula ion li ing in u al a eas
0.352
(0.180)
× Numbe o heal h si es
0.343
(0.161)
× Log heal h ca e spending
-0.367
(0.222)
× A e age nigh ligh
-0.499
(0.261)
× Sub-na ional HDI
0.104
(0.190)
× E hnic di e si y
-0.113
(0.184)
× E hnic pola iza ion
0.086
(0.195)
× Sha e o popula ion commu ing o e an hou
0.047
(0.161)
× Sha e o popula ion commu ing public anspo
-0.143
(0.199)
× Sha e o popula ion elewo k po en ial
-0.300
(0.183)
× Sha e o popula ion wo king ag icul u e
0.203
(0.195)
P o ince FEs
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Wea he con ols
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
FE = Fixed e ec s, HDI = Human De elopmen Index.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task Fo ce, using measu emen s aken as o 18
Feb ua y 2023. Mobile in e ne da a was ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.

49
Table A.4: He e ogenei y: Poisson IV
Dependen a iables: Cumula i e cases by 10T popula ion, 18 Feb 2023
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
In e ne exposu e
-2.371
-1.190
-0.627
-1.631
-1.472
-1.648
-1.460
-1.406
-0.924
-1.728
-2.133
-1.921
-0.876
-1.723
(0.596)
(0.468)
(0.715)
(0.412)
(0.398)
(0.439)
(0.376)
(0.695)
(0.517)
(0.485)
(0.625)
(0.585)
(0.516)
(0.486)
× Sha e o popula ion wi h no educa ion
1.006
(0.441)
× Sha e o popula ion wi h g ad. deg ee
-0.308
(0.207)
× Log a e age household size
-1.052
(0.882)
× Sha e o popula ion li ing in u al a eas
0.326
(0.280)
× Numbe o heal h si es
-0.121
(0.205)
× Log heal h ca e spending
0.242
(0.189)
× A e age nigh ligh
0.238
(0.513)
× Sub-na ional HDI
-0.034
(0.834)
× E hnic di e si y
-0.879
(0.617)
× E hnic pola iza ion
0.464
(0.315)
× Sha e o popula ion commu ing o e an hou
0.565
(0.352)
× Sha e o popula ion commu ing public anspo
0.561
(0.351)
× Sha e o popula ion elewo k po en ial
-0.740
(0.408)
× Sha e o popula ion wo king ag icul u e
0.520
(0.282)
P o ince FEs
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Demog aphics
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Heal h acili ies
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Economic s a us
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
E hnic composi ion
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Labo o ce
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Wea he con ols
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
FE = Fixed e ec s, HDI = Human De elopmen Index.
Sou ces: COVID-19 da a, adjus ed o popula ion, was sou ced om he Indonesian COVID-19 Task Fo ce, using measu emen s aken as o 18
Feb ua y 2023. Mobile in e ne da a was ob ained om Collins Ba holomew’s GSMA Mobile Co e age Explo e da abase.
50
Table B.1: Va iable O e iew
Va iable
N
Mean
SD
Desc ip ion
Sou ce
Cumula i e cases 18 Feb
2023
454
10942.60
31381.30
To al numbe o cases a he egency le el as o 18 Feb
2023
Indonesian COVID-19 Task Fo ce
Cumula i e dea hs 18 Feb
2023
417
305.66
504.09
To al numbe o dea hs a he egency le el as o 18 Feb
2023
Indonesian COVID-19 Task Fo ce
Mobile 3g OCI
500
0.37
0.45
Sha e o a eas a he egency le el ha ing access o 3G
as o Decembe 2019
Collins Ba holomew’s Mobile Co e age
Explo e
(OpenCellID)
Mobile 3g MCE
499
62.93
44.91
Sha e o a eas a he egency le el ha ing access o 3G
as o Decembe 2019
Collins Ba holomew’s Mobile Co e age
Explo e
Mobile 4g OCI
482
0.76
0.36
(Mobile Co e age Explo e )
Sha e o a eas a he egency le el ha ing access o 4G
as o Decembe 2019
Collins Ba holomew’s Mobile Co e age
Explo e
(OpenCellID)
Mobile 2g OCI
500
0.52
0.34
Sha e o a eas a he egency le el ha ing access o 2G
as o Decembe 2019
Collins Ba holomew’s Mobile Co e age
Explo e
(OpenCellID)
Popula ion densi y
465
0.89
2.43
Popula ion densi y a he egency le el
Indonesian COVID-19 Task Fo ce
Popula ion
465
2.85
2.14
To al popula ion a he egency le el
Indonesian COVID-19 Task Fo ce
Popula ion o e 65
510
0.08
0.03
Sha e o popula ion wi h a uni e si y deg ee o highe a
he egency le el
2010 Indonesian popula ion census
Popula ion wi h no
educa ion
510
0.19
0.11
Sha e o popula ion wi h a uni e si y deg ee o highe a
he egency le el
2010 Indonesian popula ion census
Popula ion wi h high
educa ion
510
0.09
0.05
Sha e o popula ion wi h a uni e si y deg ee o highe a
he egency le el
2010 Indonesian popula ion census
Household size
510
4.16
0.46
A e age household size a he egency le el
2010 Indonesian popula ion census
Ru al a eas
510
0.60
0.31
Sha e o popula ion li ing in u al a eas
2010 Indonesian popula ion census
Numbe o clinics
363
14.00
42.60
Numbe o clinics in 2019 a he egency le el
Open S ee Map (OSM)
Numbe o hospi als
363
8.00
14.03
Numbe o hospi als in 2019 a he egency le el
Open S ee Map (OSM)
Heal h spending
484
27.01
18.76
Amoun o ( ealized) heal h spending in 2019 a he
egency le el
Open S ee Map (OSM)
Con inued on he nex page
51
Va iable
N
Mean
SD
Desc ip ion
Sou ce
Nigh ime ligh
500
2.14
5.07
A e age nigh ligh (Visible In a ed Imaging Radiome e
Sui e - VIIRS) as
Na ional Oceanic and A mosphe ic
Adminis a ion (NOAA)
o Decembe 2019
Human de elopmen index
465
2.97
1.42
Human de elopmen index a he egency le el in 2019
(quin iles)
Badan Pusa S a is ik
E hnic ac ionaliza ion
510
0.46
0.31
P obabili y ha wo andomly selec ed people in a
egency belong o di e en e hnic g oups
2010 Indonesian popula ion census
E hnic pola iza ion
510
0.45
0.25
P obabili y ha a g oup o indi iduals in a egency is
di ided in o di e en
2010 Indonesian popula ion census
e hnic g oups
Long-dis ance wo k
510
0.05
0.03
Sha e o popula ion a eling ¿=1hou o wo k a he
egency le el
Na ional Labo Fo ce Su ey 2019
Public anspo wo k
510
0.06
0.05
Sha e o popula ion using public anspo o wo k a he
egency le el
Na ional Labo Fo ce Su ey 2019
Telewo kabili y
510
0.22
0.05
Telewo kabili y by indus y a he egency le el
Na ional Labo Fo ce Su ey 2019
Ag icul u e sha e
510
0.38
0.22
Sha e o popula ion wo king in ag icul u e
Na ional Labo Fo ce Su ey 2019
Tempe a u e
500
25.80
1.92
A e age mon hly empe a u e as o Decembe 2019
(Celsius deg ee)
ERA5 eanalysis da a
Rain all
500
0.21
0.08
A e age mon hly p ecipi a ion as o Decembe 2019
(millime e s)
ERA5 eanalysis da a
Ligh ning s ike
500
0.01
0.01
A e age ligh ning s oke powe as o Decembe 2019
(5-minu e esolu ion)
Wo ld Wide Ligh ning Loca ion Ne wo k
COVID-19 = co ona i us disease, MCE = Mobile Co e age Explo e , OCI = OpenCellID.
Sou ce: Au ho s’ compila ion.
52
REFERENCES
Ahmed, Faheem, Na’eem Ahmed, Ch is ophe Pissa ides, and Joseph S igli z. 2020.
“Why Inequali y Could Sp ead COVID-19.” The Lance Public Heal h, 5 (5): e240.
Ake , Jenny C., and Isaac M. Mbi i. 2010. “Mobile Phones and Economic De elopmen
in A ica.” Jou nal o Economic Pe spec i es, 24 (3): 207–32.
Aksoy, Ce a Gi ay, Jose Ma ia Ba e o, Nicholas Bloom, S e en J Da is, Ma hias
Dolls, and Pablo Za a e. 2022. “Wo king om Home A ound he Wo ld.” NBER
Wo king Pape No. 30446. Na ional Bu eau o Economic Resea ch (NBER).
Alesina, Albe o, and Eliana La Fe a a. 2005. “E hnic Di e si y and Economic
Pe o mance.” Jou nal o Economic Li e a u e, 43 (3): 762–800.
Alipou , Jean-Vic o , Ha ald Fadinge , and Jan Schymik. 2021. “My Home is My Cas le–
The Bene i s o Wo king om Home Du ing a Pandemic C isis.” Jou nal o Public
Economics, 196: 104373.
Ama al-Ga cia, So ia, Ma ia Na do o, Ca ol P oppe , and Tommaso Valle i. 2022.
“Mums Go Online: Is he In e ne Changing he Demand o Heal h Ca e?” Re iew o
Economics and S a is ics, 104 (6): 1157–73.
Ande sen, Thomas Ba nebeck, Jeane Ben zen, Ca l-Johan Dalgaa d, and Pablo
Selaya. 2012. “Ligh ning, IT Di usion, and Economic G ow h ac oss US S a es.”
Re iew o Economics and S a is ics, 94 (4): 903–24.
Asian De elopmen Bank (ADB). 2021. “COVID-19 and Labo Ma ke s in Sou heas
Asia: Impac s on Indonesia, Malaysia, he Philippines, Thailand, and Vie Nam.”
Manila.
Au hamme , Maximilian, Solomon M. Hsiang, Wol am Schlenke , and Adam Sobel.
2013. “Using Wea he Da a and Clima e Model Ou pu in Economic Analyses o
Clima e Change.” Re iew o En i onmen al Economics and Policy, 7 (2).
Ba gain, Oli ie , and Ulugbek Aminjono . 2020. “T us and Compliance o Public Heal h
Policies in Times o COVID-19.” Jou nal o Public Economics 192: 104316.
Ba e o, Jose Ma ia, Nicholas Bloom, and S e en J. Da is. 2021. “In e ne Access and
I s Implica ions o P oduc i i y, Inequali y, and Resilience.” NBER wo king Pape
No. 29102. NBER.
Ba e o, Jose Ma ia, Nicholas Bloom, S e en J. Da is, and B en H. Meye . 2021.
“COVID-19 is a Pe sis en Realloca ion Shock.” AEA Pape s and P oceedings, 111:
287–91.