Pham, Thi T ang; Wong, Pui Hang
Wo king Pape
In e ne access and you h's men al heal h and well-being:
E idence om E hiopia
UNU-MERIT Wo king Pape s, No. 2025-007
P o ided in Coope a ion wi h:
Maas ich Economic and Social Resea ch Ins i u e on Inno a ion and Technology (UNU-MERIT),
Uni ed Na ions Uni e si y (UNU)
Sugges ed Ci a ion: Pham, Thi T ang; Wong, Pui Hang (2025) : In e ne access and you h's men al
heal h and well-being: E idence om E hiopia, UNU-MERIT Wo king Pape s, No. 2025-007, Uni ed
Na ions Uni e si y (UNU), Maas ich Economic and Social Resea ch Ins i u e on Inno a ion and
Technology (UNU-MERIT), Maas ich ,
h ps://doi.o g/10.53330/GYMA8538
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/326936
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-nc-sa/4.0/
#2025-007
In e ne Access and You h’s Men al Heal h and Well-being:
E idence om E hiopia
T ang Thi Pham and Pui-Hang Wong
Published 18 Feb ua y 2025
DOI: h ps://www.doi.o g/10.53330/GYMA8538
Maas ich Economic and social Resea ch ins i u e on Inno a ion and Technology (UNU-MERIT)
email: in o@me i .unu.edu | websi e: h p://www.me i .unu.edu
Boschs aa 24, 6211 AX Maas ich , The Ne he lands
Tel: (31) (43) 388 44 00
UNU-MERIT Wo king Pape s
ISSN 1871-9872
Maas ich Economic and social Resea ch Ins i u e on Inno a ion and Technology
UNU-MERIT
UNU-MERIT Wo king Pape s in end o dissemina e p elimina y esul s o esea ch ca ied
ou a UNU-MERIT o s imula e discussion on he issues aised.
h ps://c ea i ecommons.o g/licenses/by-nc-sa/4.0/
In e ne Access and You h’s Men al Heal h and
Well-being: E idence om E hiopia
T ang Thi Pham∗1,2 and Pui-Hang Wong1,2
1UNU-MERIT, Maas ich , he Ne he lands
2School o Business and Economics, Maas ich Uni e si y, Maas ich , he Ne he lands
Abs ac
This pape p o ides one o he i s obus e idence on he impac o in e ne access
on adolescen s’ well-being and men al heal h in a low-income coun y con ex . We
ind educed subjec i e well-being and inc eased measu es o men al heal h diso de s
among young people in E hiopia du ing 2020-2021 ollowing in e ne di usion. Ou
he e ogenei y analysis e eals ha he e ec s o in e ne access on men al heal h a e
unequal, wi h s onge nega i e impac s o adolescen s om lowe -weal h households.
The mechanism analysis sugges s ha passi e in e ne use, pa icula ly among you h
om less ad an aged socioeconomic backg ounds, migh d i e hese nega i e ou comes.
To add ess po en ial endogenei y, we employ ins umen al a iable echniques combined
wi h ixed e ec s. The ins umen is ele an based on ne wo k e ec a gumen s and
easonably exogenous condi ional on con ol a iables and ixed e ec s. Ou esul s o e
policy implica ions ega ding in e ne access and you h human capi al de elopmen in
he digi al age and highligh he signi icance o social causes in shaping men al heal h.
Keywo ds: In e ne , You h, Men al heal h, Well-being, Inequali y.
JEL Classi ica ions: I14, O33, J13.
∗We would like o hank Robin Cowan, B uno Ma o ano, Musa Hasen o many help ul commen s and
eedback and he Young Li es eam o da a access. This pape was also p esen ed a he Wo ld Bank’s
The Pulse o P og ess Ha nessing High- equency su ey da a o de elopmen esea ch con e ence in 2024
in Washing on, D.C.
Co esponding au ho : .pham@maas ich uni e si y.nl o pham[email p o ec ed]u.edu
1
1 In oduc ion
The in e ne ’s po en ial o d i ing de elopmen has ga ne ed signi ican en husiasm, spu ing
uni e sal connec i i y ini ia i es om go e nmen s and de elopmen dono s (Galpe in &
Viecens, 2017). Exis ing li e a u e widely highligh s he in e ne ’s posi i e impac s on de el-
opmen , including g ow h and income pe capi a p omo ion, and c ea ion o new employmen
oppo uni ies (Be schek e al., 2015). These posi i e impac s a e la ge in de eloping coun-
ies, especially in he se ice sec o and emo e egions, whe e mobile ne wo ks and b oad-
band in e ne enable elecommu ing and help o e come geog aphical cons ain s. Rega ding
wel a e and human de elopmen impac , he e a e bene i s in educa ion democ a iza ion and
imp o emen s in ce ain lea ning ou comes (Acemoglu e al., 2014; Ake e al., 2012), em-
ploymen ansi ion and labo mobili y (Ha je & H¨uble , 2017; Su i & Jack, 2016; Viollaz &
Winkle , 2022).1When ex ended o b oade socie al ou comes, mobile b oadband in e ne
access exposes co up ion in go e nmen , enhances ci izen pa icipa ion and en i onmen al
go e nance, and ‘libe a es’ coun ies by os e ing mass mobiliza ion (Bun aine e al., 2024;
Gu ie e al., 2021; Manaco da & Tesei, 2020).
A a he op imis ic pic u e o he impac s o b oadband di usion appea s o eme ge
om he exis ing li e a u e, pa icula ly o less de eloped egions and coun ies. The in e -
ne , howe e , also in oduces signi ican ad e se e ec s, especially conce ning well-being and
human capi al de elopmen . Ha ms including misin o ma ion, scams, digi al addic ion, and
men al heal h diso de s ha e a isen a a pace as e han indi idual o socie al awa eness can
keep up wi h (Fi h e al., 2019; Loh & Kanai, 2016). Mos o he exis ing li e a u e on he
in e ne ’s ad e se e ec s, howe e , p ima ily ocuses on de eloped coun ies. Fo example,
s udies om he US and I aly show ha in e ne use c owds ou quali y ime wi h amily
and iends and unde mines social capi al and ci ic engagemen in he UK (Allco e al.,
2020; Ge aci e al., 2022; Ro ondi e al., 2017). A he indi idual le el, sma phones and
social media a e known o be highly addic i e and can cause sel -con ol p oblems, nega i ely
a ec ing men al heal h, especially among young people in he US and Eu ope (Allco e al.,
2022; B aghie i e al., 2022; Dona i e al., 2022; Golin, 2022; McDool e al., 2020). Poli ically,
social media algo i hms c ea e social and poli ical bubbles and inc ease pola iza ion in he
US and UK (Ga azza e al., 2019; Le y, 2021). Fu he mo e, he in e ne has been linked
o inc eased con lic s in bo h cybe space and physical domain (Bha a i e al., 2024; de A i-
ma ´eia da C uz, 2014; Kos yuk & Ga zke, 2024) and is iewed as a ‘ ep ession echnology’
1Al hough some o hese pape s concen a e on ea u e phones a he han sma phones in he ea lie
pe iod be o e he widesp ead di usion o mobile in e ne , sma phones and 3/4G ne wo ks a e now widely
a ailable in A ica, including E hiopia. Consequen ly, u u e analyses o mobile phones will p edominan ly
ocus on sma phones.
1
a he han a libe a ion one (Rød & Weidmann, 2015).
Due o his o ically low in e ne pene a ion in some egions and limi a ions in exis ing
da ase s, empi ical e idence on he de elopmen al impac s o in e ne access in de eloping
coun ies is scan (Galpe in & Viecens, 2017), and wi hin his limi ed li e a u e, many a e
o en anecdo al. A 2024 New Yo k Times epo o e s some o he ea lies insigh s in o he
dange s posed by i s - ime wi eless in e ne access o an isola ed Amazonian ibe ollowing
he in oduc ion o sa elli e co e age (Nicas, 2024a,2024b).2While o e ing clea bene i s,
such as ideo cha s wi h dis an ela i es, abili y o call o help in eme gencies, and ele a ed
aspi a ions among young people o wo ld a el and p o essional ca ee s, nega i e ou comes
a e also e iden . In less han a yea a e gaining sa elli e in e ne , eenage s ha e become
a ached o hei phones and less mo i a ed o wo k, wi h g oup cha s illed wi h addic i e
social media, iolen ideo games, scams, and age-inapp op ia e con en .
Such signi ican nega i e e ec s on human capi al de elopmen , pa icula ly among
young people, unde sco e he u gen need o imely s udies on he de elopmen al impac s o
he in e ne in unde de eloped egions. Howe e , he p ima y ocus in he cu en de elop-
men li e a u e emains on add essing ba ie s o in e ne access and p omo ing g ea e us-
age. The lack o comp ehensi e s udies on nega i e e ec s, alongside he appa en ly one-sided
posi i e e idence in cu en li e a u e, p esen s ‘a bes , an incomple e pic u e’ (Goggin &
McLelland, 2017), o wo se, ‘op imis ic simplism’ discou se o in e ne connec i i y (F iede ici
e al., 2017). This emphasis o e looks he di e en ial socioeconomic and geog aphical con-
ex s, and po en ially p o ound ad e se e ec s on human de elopmen ou comes.
You h and adolescen s, in pa icula , ace heigh ened nega i e e ec s on well-being and
o he aspec s o human de elopmen due o hei s a us as bo h a ulne able demog aphic
(Pa el e al., 2016; Twenge & Campbell, 2019) and a ech-sa y g oup ha adop s digi al
echnologies and social media mo e han o he s (ITU, 2020,2023). Psychological and men al
diso de s a young ages can accumula e and ca y o e beyond adolescence and cons ain
human de elopmen by educing indi iduals’ capaci y and wo king abili y, con ibu ing o
in e gene a ional and wi hin-gene a ion social immobili y (Golbe s ein e al., 2019; Goodman
e al., 2011; UN, 2023; UNDP, 2022).
In e ne use can ha e a ied impac s on adolescen s’ men al heal h, wi h socioeco-
nomic dispa i ies wo sening nega i e ou comes. While some e idence sugges s s onge e ec s
among s uden s om lowe socioeconomic s a us (SES) backg ounds in de eloped coun ies
like he US and No way (Ab ahamsson, 2024; B aghie i e al., 2022), esea ch on hese in-
2S a link se ice, s a ing om mid-2023, despi e lacking elec ici y, phone sc eens ligh up a nigh , using
ba e y powe om sola panels.
2
equali ies emains limi ed. Few s udies ha e explo ed he mechanisms behind hese unequal
e ec s, such as B aghie i e al. (2022) on un a o able social compa isons. Resea ch also
shows ha passi e in e ne use, such as sc olling h ough au o- ecommended social media,
exposes use s o unheal hy compa isons and ha m ul con en (Ve duyn e al., 2015). This
disp opo iona ely a ec s disad an aged you h – hose wi h lowe educa ional le els, weal h,
o SES – compa ed o you h om ad an aged backg ounds, who ha e mo e oppo uni ies
o heal hie o line ac i i ies o help b eak he cycle o excessi e online ime and mindless
sc olling ha wo sens men al heal h (Geo ge e al., 2020).
Se agains his backd op, his pape b idges wo s ands o li e a u e: one ha pain s an
o e ly op imis ic pic u e o he impac o in e ne di usion on less de eloped coun ies, and
ano he ha highligh s he nega i e e ec s on well-being and human de elopmen , which
ha e hus a p ima ily been e idenced in de eloped coun ies. P o i ing om he Young
Li es longi udinal su ey, we p o ide ea ly empi ical e idence demons a ing he ad e se e -
ec s o in e ne di usion on men al heal h and well-being among adolescen s in E hiopia. To
alle ia e endogenei y conce ns by using obse a ional da a, we employ an ins umen al a i-
able (IV) app oach combined wi h indi idual and ime ixed e ec s (FE) in ou es ima ions.
Ou ins umen is he communi y’s a e age in e ne access a es, ollowing ex an li e a u e
(H¨uble & Ha je, 2016; Ma & Sheng, 2023; Ro ondi e al., 2017). The IV is ele an based on
ne wo k e ec a gumen s and is easonably exogenous condi ional o con ol a iables and
he inclusion o FEs. Ou esul s emain obus a e mul iple obus ness es s, including
addi ional communi y-le el ime- a ian con ols.
This pape con ibu es o he li e a u e in h ee ways. Fi s , ou s udy is, as a as we
know, among he i s o analyze he nega i e ou comes o in e ne access on adolescen s’
well-being and human capi al o he Global Sou h. The esul s add o he li e a u e by
highligh ing he nuanced e ec s o in e ne di usion in low-income coun ies. Second, we
con ibu e o he s ill la gely unde -explo ed li e a u e on he unequal impac s o in e ne use
on men al heal h and well-being, pa icula ly o adolescen s om di e en household weal h
le els – a no able inding in he leas de eloped coun ies con ex (OHRLLS, 2024). Thi d,
ou mechanism analysis e eals ha disad an aged you h end o engage mo e in passi e
in e ne use, a pa e n likely d i en by addic i e echnology algo i hms and exace ba ed by
sys emic inequali ies in suppo o add essing hese nega i e e ec s (Acemoglu & Johnson,
2023; Schee de e al., 2017). Ou indings ha e impo an implica ions o you h heal h and
human capi al de elopmen policies in de eloping coun ies, pa icula ly in he e a o mobile
digi aliza ion. They also shed ligh on how he digi al age is ampli ying he Ma hew e ec ,
widening exis ing inequali ies (K¨umpel, 2020; T ucano, 2013).
3
In he ollowing sec ions, ela ed li e a u e will be p esen ed in Sec ion 2; Sec ion 3
desc ibes da a sou ces and p o ides a desc ip i e analysis; Sec ion 4s a es ou empi ical
s a egy. Ou esea ch esul s, including he main indings, obus ness es s, and mechanisms
analyses, a e p esen ed in Sec ion 5. Sec ion 6concludes and p o ides discussion and policy
implica ions.
2 Ex an li e a u e and backg ound
He e, we syn hesize he ex an li e a u e on in e ne use and you h men al heal h and key
mechanisms behind he unequal e ec s on you h om di e en socioeconomic backg ounds.
2.1 You h’s men al heal h and he ole o in e ne use
Men al heal h among you hs has wo sened in he pas wo decades (Pa el e al., 2016), wi h
anxie y and dep ession a e he mos common mani es a ions o poo men al heal h (Huppe
& So, 2013; Ma ´ınez e al., 2020).3In E hiopia, Hunduma e al. (2024)’s desc ip i e s udy in
2020 shows ha among in-school adolescen s, he e a e high le els o in e nalizing p oblems,4
which p ima ily consis o dep ession and anxie y.
The in e ne ’s impac on well-being a ies ac oss demog aphic g oups, in luenced by pe -
sonal cha ac e is ics, capabili ies, and cul u al con ex s (Cas ellacci & T ei o, 2018). You h
and adolescen s, who ace unce ain ies du ing ansi ioning pe iod o adul hood5ha e lim-
i ed sel - egula ion and a e mo e suscep ible o pee p essu e, making hem ulne able as hey
na iga e and expe imen wi h social media (Achdu e al., 2021; Tanne & A ne , 2016).
Dona i e al. (2022) o ins ance, ind a signi ican posi i e e ec o b oadband di usion on
men al diso de s’ diagnoses and hospi aliza ions among he younge age g oups in I aly om
2001 o 2013, while no such e ec was de ec ed o olde people.
Nega i e men al heal h consequences esul ing om in e ne use can lead o a deple ion
o pe sonal, social, economic, o cul u al esou ces. Fo ins ance, indi iduals could lose con-
idence o in o mal ies ( an Dijk, 2019). In he long e m, hese e ec s can cons ain human
de elopmen by diminishing indi iduals’ capabili ies and p oduc i i y, he eby con ibu ing
3Dep ession and anxie y exhibi o e lapping symp oms, such as sudden mood swings and social wi h-
d awal, which can in ensi y eelings o isola ion and loneliness. Dep ession has he po en ial o escala e o
suicidal hough s o ac ions (WHO, 2021).
4In e nalizing p oblems in ol e inwa d- acing symp oms ha impac an indi idual’s in e nal emo ional
s a e; while ex e nalizing p oblems, such as conduc issues, hype sensi i i y, ina en i eness, impulsi i y, and
dis up i e diso de s, mani es as ou wa d beha io s ha impac an indi idual’s social en i onmen .
5Including iden i y o ma ion, educa ion, job-seeking, and ela ionship-building.
4
o social immobili y wi hin and ac oss gene a ions (Golbe s ein e al., 2019; Goodman e al.,
2011; UN, 2023; UNDP, 2022).
2.1.1 You h’s in e ne use sphe e and impac s on well-being
You h adop digi al echnologies and social media mo e han any o he g oups, i espec i e
o egion o coun y le el o de elopmen (ITU, 2020,2023). In E hiopia, mos use s access
he in e ne o social media, en e ainmen , and news (Adam e al., 2024).6Fo social
media use s in E hiopia, escapism (using digi al media o ill spa e ime), exchanging ideas
wi h iends, and knowing abou o he people, besides acqui ing in o ma ion, a e he op use
mo i a ions (Adam e al., 2024; Haile, 2024; In e news, 2023).7
Such use pa e ns can a ec you h’s men al heal h h ough mul iple mechanisms. Fi s ,
he uses o in e ne and sma phones educe social in e ac ions (Dwye e al., 2018; Ro ondi
e al., 2017). Ye , o young adul s who eel disad an aged in he digi al age, online so-
cial ne wo ks enhance social connec ions, which a e less accessible h ough o he means o
communica ion (Wohn e al., 2013). Achdu e al. (2021), howe e , show ha while online
ne wo king builds in o mal social capi al ha p o ec s well-being, i is also linked o highe
psychological dis ess, indica ing con as ing e ec s.
Wi h isen digi aliza ion, compe i ion among social media pla o ms and he adjus men
o ecommende algo i hms o boos use engagemen ha e inc eased social media’s addic-
i eness,8causing use s c a ing o check social media o g a i ica ion when idle (Hje land
e al., 2021; Ras e al., 2021). Desc ip i e e idence shows ha mos E hiopian you h use
social media o pass ime, o ‘escapism’, o eel like a was e o ime (Haile, 2024; Hussain
& Hussain, 2023; In e news, 2023). This beha io can be linked o ‘mindlessly sc olling’ o
passi e in e ne use, which has been shown in psychological and neu oscience li e a u e o
a ec b ain s uc u e and dynamics, cause inc eased dis ac ion and psychological dis ess
(A ness & Ollis, 2023; Fi h e al., 2024; Ras e al., 2021).
Such passi e in e ne use also inc eases use s’ exposu e o unheal hy con en 9and/o
cybe bullying10 (Benei o & Vicen e-Chi i ella, 2022; an Geel e al., 2014). The op isks on
695, 79, and 67 pe cen espec i ely; ollowed by banking – 31 pe cen , o he s, and online lea ning.
7Facebook is he main pla o m o in o ma ion sou cing, ollowed by Teleg am, TikTok, YouTube.
8Fo example, Facebook in 2018 announced ha he company’s algo i hm would be modi ied o p io i ize
pos s om o he use s, especially amily and iends, a he han news o ganiza ions and es ablished b ands
(Acemoglu & Johnson, 2023), o ecen inno a ions in sho - o m ideos wi h push no i ica ions, ecommen-
da ions, and au o-sc olling (Bake , 2023; Yang e al., 2021).
9Age-inapp op ia e con en , such as online po nog aphy, pee - o-pee abusi e beha io in ol ing sexually
sugges i e o hos ile commen s, p i acy b eaches, and he undue in luence o hi d pa ies such as ad e ise s.
10The use o digi al media o pos h ea ening messages, emba assing pho os, and umo s wi h he in en
o ha m o he s.
5
measu emen e o s and o he ime- a ian OVBs e.g., household o communi y weal h o
heal h in as uc u e imp o emen s o educ ions ha e downwa dly biased he OLS es ima es.
Table 2: In e ne and adolescen s’ men al heal h p oblems and subjec i e well-being
Subjec i e well-being GAD-7 PHQ-8
(1) (2) (3) (4) (5) (6)
FE FE-IV FE FE-IV FE FE-IV
In e ne 0.109 -1.723∗∗∗ 0.719∗∗∗ 9.310∗∗∗ 0.668∗∗∗ 8.338∗∗∗
(0.104) (0.557) (0.267) (1.603) (0.254) (1.486)
1s -s age Communi y in e ne 0.941∗∗∗
(0.132)
(Belie ed) in ec ed 0.176 0.474∗∗ 0.341∗2.731 3.200 0.108 -1.528
(0.263) (0.234) (0.204) (1.972) (3.188) (0.651) (1.604)
Income dec eases -0.057 -0.0818 -0.006 0.518∗∗∗ 0.937∗∗∗ 0.370∗∗ 0.744∗∗∗
(0.0905) (0.116) (0.0245) (0.176) (0.290) (0.167) (0.272)
Run ou o ood 0.302∗∗ 0.211 0.0470 0.110 0.740∗∗ 0.339 0.780∗∗
(0.149) (0.189) (0.039) (0.186) (0.373) (0.209) (0.377)
Cons an 4.463∗∗∗ 1.113∗∗∗ 1.129∗∗∗
(0.0713) (0.160) (0.153)
Obse a ions 2272 1846 1846 2262 1836 2266 1840
Indi idual FE Yes Yes Yes Yes Yes Yes Yes
Call FE Yes Yes Yes Yes Yes Yes Yes
R20.643 0.589 0.613
Adjus ed R20.284 0.175 0.223
F-s a is ic 50.59 50.59 50.16 50.60
No es: Each column epo s es ima ed e ec s o in e ne access on men al heal h p oblem indica o s:
GAD-7: Gene alized Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e-8. Well-being is subjec i e
well-being, wi h a ange be ween 1 (low well-being) and 9 (high well-being). Columns (1) and (4) a e OLS
esul s; columns (2) and (5) a e FE; (3) and (6) a e FE-IV. FE include indi idual and call ixed e ec s. The
ins umen is he p opo ion o households in a communi y ge ing access o he in e ne , excluding one’s
own alue. F-s a is ic is he Kleibe gen-Paap k Wald F s a is ic. The c i ical alue o he S ock-Yogo es
wi h 10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
12
Following B aghie i e al. (2022) o measu e he downs eam implica ions o poo men al
heal h, we also un addi ional analyses o a numbe o human capi al- ela ed a iables. YL
su ey in Call 2 had se e al i ems on ime spen du ing a ypical weekday,19 including ime
mo e spen o s udying.20 The e is also a ques ion on “Wha a e you mos looking o wa d o
do a e COVID-19 is o e ?”, and answe s include going back o educa ion.21 We conduc IV
es ima ions on he impac o in e ne use on hese a iables using linea p obabili y models
o bina y dependen a iables. As shown in Table 3below, in e ne use is associa ed wi h
less ime spen on s udying and less aspi a ion o go back o educa ion a e COVID-19. The
ull able wi h o he con ol a iables is p esen ed in Table B2 in he appendix.
Table 3: B oade impac s on o he human capi al- ela ed a iables
(1) (2)
(Pe cei ed) s udy ime Wan o go back o educa ion
In e ne -0.205∗∗∗ -0.225∗∗
(0.0750) (0.0927)
(Belie ed) in ec ed 0.287∗0.272∗∗∗
(0.166) (0.0506)
Income dec eases -0.0554∗∗ -0.109∗∗∗
(0.0258) (0.0256)
Run ou o ood -0.0298 -0.188∗∗∗
(0.0289) (0.0397)
Obse a ions 1370 1162
Con ol Yes Yes
F-s a is ic 184.21 121.15
No es: Samples include only child en in Call 2 due o da a a ailabili y. The ins umen o in e ne access
is he p opo ion o households in a communi y ge ing access o he in e ne . Each column epo s
es ima ed e ec s o in e ne access on he likelihood ha ing spen mo e ime on s udying o wan ing o go
back o school a e COVID-19. Co a ia es: pe cei ed weal h le els, household size, gende , (belie ed o be)
in ec ed by COVID-19, nega i e income shocks, un ou o ood, u ban. F-s a is ic is he Kleibe gen-Paap
k Wald F s a is ic. The c i ical alue o he S ock-Yogo es wi h 10% ole ance is 16.38. Robus s anda d
e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
19Ques ionai e i em: “Now I wan you o hink abou how you spen you ime du ing a ypical weekday
when he coun y was in ull lockdown. Do you ag ee, pa ially ag ee o disag ee wi h he ollowing s a emen :
1=Ag ee; 2=Pa ially ag ee 3=Disag ee.” We ecode hese a iables in o bina y o ma , wi h 1 equal Ag ee,
and 0 o he es .
20“I spen mo e ime s udying/ lea ning.”
21Op ions include 1-Going back o educa ion, 2-Going back o wo k/wo kplace, 3-Visi amily, 4-Visi
iends, 5-Go o he ci y cen e , 6-Go o he pa k, 7-Go ou o dinne , 8-Play a eam spo , 9-To go o
chu ch/ mosque/ emple/ ce emonies/ weddings, 10-None, 11-O he (speci y). We c ea e a new a iable ha
equals 1 i he answe alue is 1-Going back o educa ion and equals 0 i he answe alue is g ea e han 2,
hus excluding you h who a e al eady in he labo o ce/ going back o wo k.
13
5.2 Robus ness checks
In his sec ion, we p esen a numbe o es s showing he obus ness o ou esul s unde
some al e na i e es ima ions and speci ica ions.
Fi s ly, as communi y’s a e ages o in e ne access migh be co ela ed wi h o he ime-
a ian communi y con ols ha migh be ela ed o indi idual men al heal h, we calcula e
a numbe o a iables a communi y le el, including a e age a es o households wi h in-
come educ ions and ood sho age du ing COVID-19; and a e age weal h index (measu ed
in Round 5 su ey in 2016, hen in e ac ed wi h linea ime end). Ou FE-IV esul s inco -
po a ing hese a iables emain obus , as shown in Table B3 in he appendix, wi h sligh ly
s onge e ec sizes o bo h anxie y and dep ession sco es; bu no o subjec i e well-being,
hough e ec sign emains nega i e.
Second, we conduc a c oss-sec ional IV analysis using egional a ia ions o cell owe s
co e age om OpenCellID da a, a collabo a i e communi y p ojec un by Unwi ed Labs
(Pham & Calda ola, 2024; Viollaz & Winkle , 2022). We calcula e he numbe o 3G owe s
whose signals collec ed pe egion on GIS so wa e, using he in o ma ion on cell owe s’ lon-
gi ude and la i ude, and da e and ime signal collec ed. Since OpenCellID ecen ly changed
hei da a publica ions and we canno ob ain his o ical cell owe s da a, we conduc a c oss-
sec ional IV analysis by agg ega ing owe s coun s up o he yea 2020. Resul s a e shown in
Table B4 in he appendix, wi h s a is ical signi icance and e ec signs in line wi h heo ies,
and Kleibe gen-Paap k Wald F-s a is ic abo e he minimum o 10.
Thi d, we conduc a di e ence-in-di e ence (DiD) es ima ion o a subsample o ado-
lescen s who did no ha e in e ne access in Call 2, di iding hem in o wo g oups, wi h and
wi hou in e ne access in Call 5, o ea men and con ol g oups espec i ely, as a obus -
ness es . Fu he mo e, we also combine DiD wi h ke nel p opensi y sco e ma ching (PSM)
o adjus o selec ion bias using obse able cha ac e is ics a baseline (Li e al., 2021; Mu
& an de Walle, 2011; Villa, 2016). This is done by ma ching he compa ison indi iduals
in he DiD es ima e wi h he ea men g oup based on he p edic ed p obabili y o ge ing
in e ne access ( he p opensi y sco e) by a ke nel unc ion (Heckman e al., 1997; Mu &
an de Walle, 2011), using a p obi link unc ion. The impac es ima es a e hen cons uc ed
by compa ing he be o e and a e in ou come measu es o adolescen s wi h hose o he
ma ched compa ison obse a ions.
The base DiD es ima ion is as ollows:
Yi =β0+β1 ea edi +pe iodi +β3( ea edi ×pe iodi ) + β4Xi +ϵi (3)
14
whe e Yi is subjec i e well-being o men al heal h indica o o indi idual ia ime ;
ea edi is he ea men e ec . I equals 0 i ibelongs o he con ol g oup and equals 1
i ibelongs o he ea men g oup. pe iodi is ime e ec . I equals 0 i is o call 2 and
equals 1 i is call 5; Xi deno es indi idual, household, and communi y le el con ols; ϵi is
he e o e m.
The esul s a e epo ed in Table B7 in he appendix, showing signi ican posi i e im-
pac s on anxie y and dep ession sco es o bo h DiD and PSM-DiD es ima ions, wi h sligh ly
la ge e ec sizes o he la e . Table B5 shows ha co a ia es balancing has been imp o ed
a e ke nel ma ching. The o e all balance is sa is ied. This can also be seen by he weak
join signi icance o he ma ched co a ia es (dec easing LR chi2), and he educ ions in bias
sco es in Table B7 (Callais, 2022; Callais & Young, 2023; Cump on & P ince, 2018; Fan &
Zhang, 2021; Li e al., 2021).
Fou h, we conduc ano he es o a subsample o obse a ions who did no ha e
in e ne in Call 2; and un IV es ima es o c oss-sec ional da a in Call 5. To alle ia e sel -
selec ion bias o e e se causali y o household in e ne adop ion, i.e., households wi h highe
subjec i e well-being a e mo e likely o ge in e ne access, we also add Call 2’s men al heal h
alues as a con ol a iable in he c oss-sec ional model o Call 5. Resul s in Table B8 in he
appendix show signi ican e ec s o in e ne on well-being and men al diso de s indica o s in
models wi h and wi hou he lagged dependen a iables.
5.3 E ec he e ogenei y
Since he e ec s o in e ne use on indi iduals’ men al heal h can ha e signi ican he e ogene-
i y, mode a ed by weal h o socioeconomic s a uses as e idenced by Ab ahamsson (2024) and
Geo ge e al. (2020), we conduc e ec he e ogenei y analysis o compa e he e ec s be ween
wo ypes o households. Using weal h index de eloped by B iones (2017), we di ide he
sample in o wo g oups wi h low and high weal h le els. Table 4below shows ha , as ex-
pec ed, adolescen s om households wi h a lowe weal h le el ha e a signi ican ly highe isk
o men al heal h p oblems compa ed o adolescen s om households wi h a highe le el. The
e ec size is mo e han doubled and is s a is ically di e en om ze o a he 1% signi icance
le el o bo h anxie y (Columns (3) and (4)) and dep ession (Columns (5) and (6)) sco es.
In e es ingly, he e ec on subjec i e well-being o he lowe -weal h g oup is insigni i-
can (Column (1)), while o he highe -weal h g oup, i is s a is ically signi ican (Column
(2)), hus i seems ha he e is no signi ican impac o in e ne on subjec i e well-being
dimension o he less ad an aged g oup in E hiopia du ing COVID-19.
15
Table 4: He e ogenei y o e ec s o in e ne on men al heal h by SES
SWB GAD-7 PHQ-8
(1) (2) (3) (4) (5) (6)
Weal h index Lowe Highe Lowe Highe Lowe Highe
In e ne -1.583 -1.589∗∗∗ 14.070∗∗∗ 6.608∗∗∗ 12.860∗∗∗ 5.579∗∗∗
(1.058) (0.584) (4.203) (1.316) (3.818) (1.195)
(Belie ed) in ec ed 0.360∗∗∗ 0.543 6.737∗∗∗ 0.845 -0.197 -1.658
(0.119) (0.374) (1.189) (3.836) (1.185) (1.424)
Income dec eases -0.005 -0.173 1.100∗∗ 0.876∗∗ 0.882∗0.721∗∗
(0.163) (0.167) (0.554) (0.369) (0.507) (0.345)
Run ou o ood 0.497∗0.028 0.522 0.917∗0.377 1.000∗∗
(0.293) (0.246) (0.699) (0.477) (0.681) (0.466)
Obse a ions 902 868 898 862 902 868
Indi idual, Call FEs Yes Yes Yes Yes Yes Yes
Kleibe gen-Paap Wald F-s a is ic 15.19 38.24 16.38 37.77 15.19 38.24
No es: E ec s o in e ne access by weal h g oups. FE-IV es ima es wi h indi idual, communi y, and call
ixed e ec s. The ins umen is he p opo ion o households in a communi y ge ing access o he in e ne .
Each column epo s es ima ed e ec s o in e ne access on men al heal h p oblem indica o s: GAD-7:
Gene alized Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e-8. Well-being is subjec i e
well-being, wi h a ange be ween 1 (low well-being) and 9 (high well-being). The c i ical alue o he
S ock-Yogo es wi h 10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05,
∗∗∗ p < 0.01
Ins ead o dicho omizing he weal h index, we also employ models ha in e ac in e ne
use and he weal h index. Resul s a e shown in Table B9 in he appendix, wi h e ec
signs in line wi h heo y, hough a a weake signi icance le el. The e is a posi i e sign o
he in e ac ion e m o well-being, indica ing ha while in e ne dec eases subjec i e well-
being, household weal h alle ia es such e ec . Fo anxie y and dep ession indexes, he e a e
nega i e signs o he in e ac ion e ms, showing ha weal h alle ia es he posi i e in luence
o in e ne on you h’s men al heal h diso de s. The esul s sugges he mode a ing e ec o
weal h on he dynamics o in e ne and men al heal h ou comes.
Addi ionally, as a obus ness es o check whe he he he e ogeneous e ec s caused by
in e ne access a e no con amina ed by o he di e en ial impac s o he COVID-19 pandemic,
we add addi ional dicho omous a iables, including du ing he pandemic i young people’s
household had job loss, expe ienced inpu p ices and ood p ices inc eased, as con ol a iables
in he FE-IV analyses. E ec he e ogenei ies be ween he wo g oups emain, as shown in
Table B10 in he appendix.
Las ly, ollowing p e ious es s o he subsample o adolescen s who did no ha e in e -
ne in Call 2, we un IV es ima ions o c oss-sec ional da a in Call 5. Ou comes in Table
B11 p esen signi ican ly la ge e ec s o he g oup wi h lowe weal h index, co obo a ing
16
he e ogeneous e ec e idence o in e ne on you h’s men al heal h.
In e ms o gende he e ogenei y, al hough some s udies ind s onge impac s o women,
in ou da a, he e is no de ec ed e ec he e ogenei y be ween emales and males, which migh
be due o sel - epo ing issues and/o di e en con ex ual cha ac e is ics and mechanisms.
Resul s can be seen in Table B12 in he appendix.
We also y he e ogenei y analysis o u ban- u al egions. Howe e , he IV app oach
does no wo k o u ban a eas.22 Addi ionally, he e could be a ied e ec s among di e en
e hnic g oups, hough when we di ide e hnic g oups in o majo i y (include Amha a and
O omo, oge he accoun ing o 49.6% in YL sample) and mino i y ( he es ), he e is no
signi ican he e ogenei y de ec ed, and he IV does no wo k o he majo i y e hnic g oup.
While educa ion le el a ec s in e ne access, i s impac on men al heal h is inconclusi e
(Kim e al., 2020). Indeed, when we add use s’ highes g ade achie ed23 as a con ol a iable
in he FE-IV eg essions, he in e ne ’s impac on men al heal h emains unchanged, as can
be seen in he Table B13 in he appendix. Pa en s’ educa ion le els, howe e , a ec he way
child en con on wi h he nega i e impac s o in e ne as shown by Schee de e al. (2017),
wi h di e ences being o g oups wi h/wi hou e ia y le el. In ou da a, a ound 45% o
YL mo he s ha e ze o yea o schooling, and a ound 4.1% ha e pos -seconda y educa ion
( a iable dis ibu ion in Figu e A6 in he appendix, da a om YL ound 5 in 2016), hus we
could no conduc s a is ical he e ogenei y analysis o pa en al educa ional backg ound.
In he nex sec ion, we p esen addi ional eg ession esul s as sugges i e mechanisms
explaining he main ou comes and he e ec he e ogenei y by weal h index as de ec ed abo e.
5.4 Mechanism analysis
In his sec ion, we un linea p obabili y models es ima ions o se e al bina y dependen
a iables, which a e a ailable in Call 2 da a collec ion, ha may se e as mechanisms o he
e ec s de ec ed ea lie .
Fi s , as discussed in he li e a u e e iew sec ion, in e ne use can change he ime use
pa e ns o adolescen s bo h online and o line, wi h mos young people in E hiopia using
social media o ill spa e ime o ‘escapism’ (Haile, 2024; Hussain & Hussain, 2023; In e news,
2023). Such passi e in e ne use o mindless sc olling al e na e b ain s uc u e and dynamics,
causing inc eased dis ac ion and psychological dis ess by, o example, c a ing o check
22The communi y’s a e age in e ne is insigni ican in he i s -s age esul s, and he F-s a is ic is e y low.
23We c ea e a new a iable which equals highes g ade comple ed i he e is no missing alue, o equals g ade
cu en ly en olled minus one i o he wise. Du ing 2020-2021, highes g ade a ained inc eased by 0.42 yea s.
17
social media mo e o g a i ica ion when ha ing no hing o do (A ness & Ollis, 2023; Fi h
e al., 2024; Hje land e al., 2021; Ras e al., 2021). Using he su ey i ems on ime spen
du ing a ypical weekday,24 which include ime spen on playing o doing no hing,25 we
de ec signi ican he e ogenei y in he e ec sizes o he in e ne a iable. The esul s a e
p esen ed in Table 5below o bo h he a e age and he e ogeneous ou comes. Fo b e i y,
we only epo he main independen a iable, showing ha household in e ne access esul s
in adolescen s spen mo e ime on playing o doing no hing (Column (1)) while less likely
mee ing iends (Column (4)).26 The e is no signi ican e ec on ime o doing household
cho es (Column (7)).27 Full esul s including con ol a iables can be seen in Table B14 in
he appendix.
Table 5: In e ne access and sugges i e mechanisms
Time doing no hing Mee ing iends Time HH cho es
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Weal h All Low High All Low High All Low High
In e ne 0.148∗0.329∗∗ 0.147 -0.482∗∗∗ -0.566∗∗ -0.602∗∗ -0.0178 -0.0392 0.0783
(0.0777) (0.142) (0.151) (0.0951) (0.265) (0.251) (0.0823) (0.210) (0.189)
Obse a ions 1418 454 435 1354 431 421 1419 454 435
Con ols Yes Yes Yes Yes Yes Yes Yes Yes Yes
F s a . 179.55 38.79 41.58 186.38 40.34 34.74 179.55 38.79 41.58
No es: Samples include child en in Call 2 due o da a a ailabili y. IV es ima es, he ins umen is he
p opo ion o households in a communi y ge ing access o he in e ne . Each column epo s es ima ed
e ec s o in e ne access on he likelihood ha ing spen mo e ime on doing no hing. Co a ia es: pe cei ed
weal h le els, household size, gende , (belie ed o be) in ec ed by COVID-19, nega i e income shocks, un
ou o ood, u ban. F-s a is ic is he Kleibe gen-Paap k Wald F-s a is ic. The c i ical alue o he
S ock-Yogo es wi h 10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05,
∗∗∗ p < 0.01
Ou indings co obo a e e idence om a mechanism analysis by McDool e al. (2020),
who ound ha in e ne use c owds ou ime spen on o he bene icial ac i i ies o B i ish
adolescen s.28 These esul s e lec he e ec on mo e ‘mindlessly sc olling’ o passi e use o
he in e ne , hus less heal hy ac i i ies (A ness & Ollis, 2023; Fi h e al., 2024; MacLeod,
24Ques ionai e i em: “Now I wan you o hink abou how you spen you ime du ing a ypical weekday
when he coun y was in ull lockdown. Do you ag ee, pa ially ag ee o disag ee wi h he ollowing s a emen :
1=Ag ee; 2=Pa ially ag ee 3=Disag ee.” We ecode hese a iables in o bina y o ma , wi h 1 equal Ag ee,
and 0 o he es .
25“I spen mo e ime playing/doing no hing.”
26Su ey ques ion asking whe he esponden s had le hei houses du ing he pas se en days o ec e-
a ion, mee ing iends and amily. Ques ionnai e i em is “Did you lea e he house du ing he pas 7 days?
Rec ea ion, mee ing iends and amily.”.
27“I spen mo e ime on household cho es han be o e”, also in he ime spen du ing a ypical weekday i em.
28Including: playing spo s; ace- o- ace in e ac ion wi h iends and amily; going o you h clubs o o he
o ganized e en s; unde aking olun a y o communi y wo k; and a ending ou o school classes such as a ,
music e c.
18
2023) in he age o he a en ion economy (Bake , 2023; Sea e , 2019; Yang e al., 2021).
The e ec magni udes o in e ne access on ime spen doing no hing a e s onge o he
lowe weal h g oup (Column (2)) while he e ec o he highe weal h g oup is insigni ican
and less han hal in e ec size (Column (3)), p o iding sugges i e explana ions o he
e ec he e ogenei ies de ec ed in he sec ion 5.3. YL adolescen s om lowe weal h o SES
g oup who use in e ne a e signi ican ly (mo e) likely o pe cei e ha hey spend mo e
ime playing o doing no hing, compa ed o adolescen s om highe SES g oup who do no
(signi ican ly) seem o eel so. This esul sugges s ha you h in he poo e g oup spend
mo e unsupe ized sc een ime o mo e ime passi ely consuming media. Spending mo e
ime on mindless sc olling du ing he pandemic could ha e p omo ed mo e social compa ison
(Ve duyn e al., 2015; Yue e al., 2022), heigh ened he isk o exposu e o age-inapp op ia e
unheal hy con en (Hussain & Hussain, 2023), nega i e news o he pandemic, and ci il wa
(Ambelu e al., 2021; Chekol e al., 2023; Ta eke e al., 2023). These in e ne uses c owd
ou heal hie ac i i ies online and o line ha he iche g oup a e endowed wi h and hus
a ec ing b ain s uc u e and dynamics, causing inc eased psychological dis ess (A ness &
Ollis, 2023; Fi h e al., 2024; Ras e al., 2021), and as a esul , in ensi ying he gaps in he
consequences o in e ne ’s nega i e ou comes on men al heal h.
Pas s udies ha e shown ha poo e men al heal h and well-being ou comes could be
due o educed social in e ac ions (Dwye e al., 2018; P zybylski & Weins ein, 2013; Ro ondi
e al., 2017). We es his hypo hesis bu no signi ican he e ogenei y is de ec ed, as shown
in Columns (5) and (6) o Table 5. No unequal impac s a e ei he ound o ime spen on
doing household cho es, as in columns (8) and (9), ega ding he eplacemen hypo hesis o
household domes ic physical ac i i ies.
6 Conclusion
Ou s udy shows a obus ela ionship be ween in e ne use and you h men al heal h and
subjec i e well-being in E hiopia du ing 2020-2021, and addi ionally, on you h human capi al
ega ding less ime s udying and educed desi e o e u n o educa ion. No ably, he e a e
s onge e ec s o he poo e g oups wi h mo e han doubled e ec magni udes, indica ing
signi ican ly unequal impac s on he disad an aged g oups. As a as we know, his is among
he i s pape s o e ing obus e idence on he in e ne -you h men al heal h nexus, wi h
signi ican impac he e ogenei y in a leas de eloped coun y con ex .
Ou mechanisms analysis o e s sugges i e e idence suppo ing he e ec o inc eased
ime spen on playing o doing no hing, which migh indica e mindlessly sc olling o passi e
19
in e ne use, wi h a conside able e ec he e ogenei y be ween he poo e and iche g oups.
All in all, he highly unequal impac s on you h om less ad an aged backg ound aises an
ala m abou bo h he da k side o he in e ne and he g owing end o inequali y. This
shi highligh s he impo ance o no only in e ne access bu also he quali y o use and i s
ou comes, indica ing ha he in e ne may ein o ce exis ing social inequali ies, o in o he
wo ds, exace ba e he Ma hew e ec s in he digi al age.
As he in e ne becomes mo e widesp ead wo ldwide, especially ollowing COVID-19,
young people’s inc easing exposu e o online con en and ac i i ies can lead o wo sening men-
al heal h and a decline in human capi al, po en ially leading o a global epidemic o men al
diso de s among younge gene a ions (Øs e gaa d, 2017) i no in e en ions a e made. The
esponsibili y ul ima ely lies wi h amilies and policymake s o egula e and guide in e ne
use and online cul u e o os e a heal hy and p oduc i e gene a ion (PewResea chCen e ,
2018), since on he o he on , he business models o online con en a e jus ying o maxi-
mize p o i s (Abi-Jaoude e al., 2020; G anic e al., 2020; Laue , 2021).29 This is pa icula ly
impo an in ou echnologically ad anced and g owingly compe i i e socie y, whe e human
capi al in gene al and “emo ional heal h [in pa icula ] inc easingly is conside ed an impo -
an de e minan o ea nings in all pa s o he wo ld” (Becke , 1994). I also aligns wi h
global de elopmen goals, which aim o go beyond con en ional GDP and g ow h me ics30
o measu e people’s well-being (C´ena , 2020; Kanbu e al., 2018; S igli z e al., 2009).31
While policies banning sma phones in schools show some posi i e esul s (Ab ahamsson,
2024; Benei o & Vicen e-Chi i ella, 2022), adolescen s o en spend mo e ime online a home.
Banning sma phone use o ex ing a home may no wo k and could be coun e p oduc i e
(Abouk & Adams, 2013), especially o adolescen s and young adul s who ha e mo e con ol
o e hei pe sonal de ices ou side he school en i onmen . As said by an elde ly ibe
woman in he Amazon ibe men ioned ea lie “Bu please don’ ake ou in e ne away,”
and since he in e ne also p o ides use ul in o ma ion, policy and social in e en ions should
aim o p omo ing awa eness and use habi s o op imal bene i s. P ac ical policies may
include campaigns o aise awa eness abou he ha ms and dange s o he digi al space,32
e en be o e in e ne ’s each (Hosman, 2024); o manage he possible dele e ious e ec s, and
o implemen coping s a egies (Schee de e al., 2017). These should be p omo ed in bo h
29The policy space o de eloping coun ies migh be na owe compa ed o de eloped coun ies, o exam-
ple, la ely in he US, New Yo k s a e lawmake s enac ed legisla ion p ohibi ing social media pla o ms om
exposing use s unde 18 o ’addic i e’ algo i hmic con en wi hou pa en al consen (Singh & Dang, 2024).
30Tha ha e many limi a ions, e en when used as a measu e o ma ke ou pu .
31Which include bo h ma e ial li ing s anda ds like household income, consump ion, and subjec i e well-
being o social connec ions and pe cei ed quali y o li e.
32E.g., in he US ecen ly, he e has been a legisla i e call o a wa ning label o be added o social media
apps, like in obacco s udies, o emind he ha ms caused o young people (Singh & Dang, 2024).
20
o mal (school) and in o mal (household, communi y) lea ning en i onmen s (Cobb, 2023;
Malamud & Pop-Eleches, 2011).
Pa icula ly, policies should ocus on you h om less ad an aged backg ounds o b eak
he icious cycle be ween men al illnesses and po e y which aps many young people in
bo h socioeconomic and men al heal h disad an ages (Baue e al., 2021; Co nia e al., 2020;
Hausho e & Feh , 2014). Mobile digi al heal h ca e, o example, has he po en ial o add ess
men al heal h ca e inequali y by p o iding mo e a o dable access o esou ces and se ices.
Howe e , conce ns emain ega ding he abundance o un egula ed ma e ial (Bucci e al.,
2019) and he design o digi al p og ams in e ms o inclusi e ou each and a o dabili y (Ba-
na i e al., 2020). Mo e impo an ly, add essing he oo causes o socioeconomic inequali y
o child en and adolescen s is c ucial, as hese ac o s a e impo an d i e s o men al heal h
bu a e o en unde es ima ed in cu en men al heal h p ac ices (Bu ns, 2015; Ki maye &
Pede sen, 2014). These include no only educing po e y and ma e ial dep i a ion (Achdu
e al., 2021) bu also le eling he playing ield o you h in e ms o educa ion and p o iding
be e ICT pedagogical esou ces o ensu e mo e p oduc i e use o in e ne esou ces ( an
Deu sen & an Dijk, 2014; Va gas-Mon oya e al., 2023).
While ou analysis de ec s signi ican impac s o in e ne access on sel - epo ed men al
heal h, he ac ual incidence migh be biased i he e is sel - epo ing bias due o gende
o cul u al no ms su ounding men al heal h s igma (Cas ellacci & T ei o, 2018). Thus,
u he in o ma ion on heal hca e diagnoses o hospi aliza ion cases (Ab ahamsson, 2024;
Dona i e al., 2022) migh p o ide measu able economic impac s in e ms o heal hca e cos s
(Ama al-Ga cia e al., 2022), hough his is mo e common in de eloped coun ies con ex s.
New da a should also be collec ed o assess he long- un, li e-cycle achie emen s. Fu u e
da a a ailabili y can also be used o de ec explici ly mechanisms like he social compa ison
hypo hesis (B aghie i e al., 2022), inc eased isks o ha m ul con en and cybe bullying
causing PTSD (Benei o & Vicen e-Chi i ella, 2022; Cu ie & Mo gan, 2020; McHugh e al.,
2018), online class a igue (Fo d & F eund, 2022; Huckins e al., 2020), and nega i e news
dis ess (Ambelu e al., 2021; Huckins e al., 2020), wi h po en ially di e en ial e ec s be o e
and a e he pandemic.
21
K¨umpel, A. S. (2020). The ma hew e ec in social media news use: Assessing inequali ies in
news exposu e and news engagemen on social ne wo k si es (SNS). Jou nalism, 21,
1083–1098.
Laue , D. (2021). Facebook’s e hical ailu es a e no acciden al; hey a e pa o he business
model. AI and E hics, 1, 395–403.
Leu en, E. (n.d.). Ps es co a ia e imbalance es ing and g aphing. h p:// mwww.bc.edu/
epec/bocode/p/ps es .h ml
Le y, R. (2021). Social media, news consump ion, and pola iza ion: E idence om a ield
expe imen . Ame ican Economic Re iew, 111, 831–870.
Li, X., Guo, H., Jin, S., Ma, W., & Zeng, Y. (2021). Do a me s gain in e ne di idends om
e-comme ce adop ion? E idence om China. Food Policy, 101, 102024.
Loh, K. K., & Kanai, R. (2016). How has he in e ne eshaped human cogni ion? The
Neu oscien is , 22, 506–520.
Ma, J., & Sheng, L. (2023). In e ne use ime and men al heal h among u al adolescen s in
China: A longi udinal s udy. Jou nal o A ec i e Diso de s, 337, 18–26.
MacLeod, D. (2023, Feb ua y). Re iew o digi al madness: How social media is d i ing ou
men al heal h c isis – and how o es o e ou sani y by Nicholas Ka da as.
Malamud, O., & Pop-Eleches, C. (2011). Home compu e use and he de elopmen o human
capi al. The Qua e ly Jou nal o Economics, 126, 987–1027.
Manaco da, M., & Tesei, A. (2020). Libe a ion echnology: Mobile phones and poli ical mo-
biliza ion in A ica. Econome ica, 88, 533–567.
Ma ´ınez, L., Valencia, I., & T o imo , V. (2020). Subjec i e wellbeing and men al heal h
du ing he co id-19 pandemic: Da a om h ee popula ion g oups in Colombia. Da a
in B ie , 32, 106287.
McDool, E., Powell, P., Robe s, J., & Taylo , K. (2020). The in e ne and child en’s psy-
chological wellbeing. Jou nal o Heal h Economics, 69, 102274.
McHugh, B. C., Wisniewski, P., Rosson, M. B., & Ca oll, J. M. (2018). When social media
auma izes eens. In e ne Resea ch, 28, 1169–1188.
Mu, R., & an de Walle, D. (2011). Ru al oads and local ma ke de elopmen in Vie nam.
Jou nal o De elopmen S udies, 47, 709–734.
Nicas, J. (2024a, June). The in e ne ’s inal on ie : Remo e amazon ibes. Re ie ed June
15, 2024, om h ps://www.ny imes.com/2024/06/02/wo ld/ame icas/s a link-
in e ne -elon-musk-b azil-amazon.h ml
Nicas, J. (2024b, June). No, a emo e amazon ibe did no ge addic ed o po n. Re ie ed
June 15, 2024, om h ps://www.ny imes.com/2024/06/11/wo ld/ame icas/no-a-
emo e-amazon- ibe-did-no -ge -addic ed- o-po n.h ml
28
OHRLLS. (2024). Lis o LDCs. h ps://www.un.o g/oh lls/con en /lis -ldcs
O’Reilly, M., Dog a, N., Whi eman, N., Hughes, J., E uya , S., & Reilly, P. (2018). Is social
media bad o men al heal h and wellbeing? Explo ing he pe spec i es o adolescen s.
Clinical Child Psychology and Psychia y, 23, 601–613.
Øs e gaa d, S. D. (2017). Taking acebook a ace alue: Why he use o social media may
cause men al diso de . Ac a Psychia ica Scandina ica, 136, 439–440.
Pa el, V., Chisholm, D., Pa ikh, R., Cha lson, F. J., Degenha d , L., Dua, T., Fe a i, A. J.,
Hyman, S., Laxmina ayan, R., Le in, C., Lund, C., Mo a, M. E. M., Pe e sen, I.,
Sco , J., Shidhaye, R., Vijayakuma , L., Tho nic o , G., & Whi e o d, H. (2016).
Add essing he bu den o men al, neu ological, and subs ance use diso de s: Key mes-
sages om disease con ol p io i ies, 3 d edi ion. The Lance , 387, 1672–1685.
PewResea chCen e . (2018, Ap il). The u u e o well-being in a ech-sa u a ed wo ld. Re-
ie ed June 12, 2024, om h ps://www.pew esea ch.o g/in e ne /2018/04/17/ he-
u u e-o -well-being-in-a- ech-sa u a ed-wo ld/
Pham, T. T., & Calda ola, B. (2024). 3g in e ne di usion and seconda y educa ion a ain-
men : E idence o oppo uni y cos in Vie Nam.
P imack, B. A., Shensa, A., Sidani, J. E., Whai e, E. O., yi Lin, L., Rosen, D., Coldi z,
J. B., Rado ic, A., & Mille , E. (2017). Social media use and pe cei ed social isola ion
among young adul s in he U.S. Ame ican Jou nal o P e en i e Medicine, 53, 1–8.
P zybylski, A. K., & Weins ein, N. (2013). Can you connec wi h me now? How he p esence o
mobile communica ion echnology in luences ace- o- ace con e sa ion quali y. Jou nal
o Social and Pe sonal Rela ionships, 30, 237–246.
Ras , R., Coleman, J. T., & Simme s, C. S. (2021). The da kside o he like: The e ec s o
social media addic ion on digi al and in-pe son communica ion. The Jou nal o Social
Media in Socie y, 10, 175–201.
Ridley, M., Rao, G., Schilbach, F., & Pa el, V. (2020). Po e y, dep ession, and anxie y:
Causal e idence and mechanisms. Science, 370.
Rød, E. G., & Weidmann, N. B. (2015). Empowe ing ac i is s o au oc a s? The in e ne in
au ho i a ian egimes. Jou nal o Peace Resea ch, 52, 338–351.
Ro ondi, V., S anca, L., & Tomasuolo, M. (2017). Connec ing alone: Sma phone use, quali y
o social in e ac ions and well-being. Jou nal o Economic Psychology, 63, 17–26.
Sampasa-Kanyinga, H., Roumelio is, P., & Xu, H. (2014). Associa ions be ween cybe bullying
and school bullying ic imiza ion and suicidal idea ion, plans and a emp s among
Canadian schoolchild en. PLoS ONE, 9, e102145.
Sawye , S. M., Azzopa di, P. S., Wick ema a hne, D., & Pa on, G. C. (2018). The age o
adolescence. The Lance Child & Adolescen Heal h, 2, 223–228.
29
Schee de , A., an Deu sen, A., & an Dijk, J. (2017). De e minan s o in e ne skills, uses and
ou comes. A sys ema ic e iew o he second- and hi d-le el digi al di ide. Telema ics
and In o ma ics, 34, 1607–1624.
Sea e , N. (2019). Cap i a ing algo i hms: Recommende sys ems as aps. Jou nal o Ma-
e ial Cul u e, 24, 421–436.
Singh, K., & Dang, S. (2024, June). Us su geon gene al calls o social media wa ning labels
o p o ec adolescen s.
S igli z, J. E., Sen, A., & Fi oussi, J.-P. (2009). Repo by he commission on he measu emen
o economic pe o mance and social p og ess.
S ock, J., & Yogo, M. (2005). Iden i ica ion and in e ence o econome ic models. In D. W. K.
And ews & J. H. S ock (Eds.). Camb idge Uni e si y P ess.
S one, A. A., & K uege , A. B. (2018, No embe ). Unde s anding subjec i e well-being.
OECD.
Su i, T., & Jack, W. (2016). The long- un po e y and gende impac s o mobile money.
Science, 354, 1288–1292.
Ta e e, Y. (2014). Educa ion aspi a ions and ba ie s o achie emen o young people in
E hiopia.
Tanne , J. L., & A ne , J. J. (2016). The eme gence o eme ging adul hood: The new li e
s age be ween adolescence and young adul hood. Rou ledge.
Ta eke, S. A., Lelisho, M. E., Hassen, S. S., Seid, A. A., Jemal, S. S., Teshale, B. M., Wo ale,
T. W., & Pandey, B. K. (2023). The p e alence and p edic o s o dep essi e, anxi-
e y, and s ess symp oms among Tepi Town esiden s du ing he co id-19 pandemic
lockdown in E hiopia. Jou nal o Racial and E hnic Heal h Dispa i ies, 10, 43–55.
T ucano, M. (2013, June). The ma hew e ec in educa ional echnology. Re ie ed Oc o-
be 25, 2024, om h ps://blogs.wo ldbank.o g/en/educa ion/ma hew- e ec -
educa ional- echnology
Twenge, J. M., & Campbell, W. K. (2019). Media use is linked o lowe psychological well-
being: E idence om h ee da ase s. Psychia ic Qua e ly, 90, 311–331.
UN. (2023). Con ibu ion o he global digi al compac . Re ie ed July 15, 2024, om h ps:
//www.un.o g/you hen oy/wp-con en /uploads/2023/05/Con ibu ion- o-Global-
Digi al-Compac .pd
UNDP. (2022). Human de elopmen epo 2021/2022: Unce ain imes, unse led li es shap-
ing ou u u e in a ans o ming wo ld. Uni ed Na ions De elopmen P og amme,
Human De elopmen Repo O ice (HDRO).
an Deu sen, A. J. A. M., & an Dijk, J. A. G. M. (2014). The digi al di ide shi s o
di e ences in usage. New Media & Socie y, 16, 507–526.
30
an Dijk, J. (2019). The digi al di ide. Poli y P ess.
an Geel, M., Vedde , P., & Tanilon, J. (2014). Rela ionship be ween pee ic imiza ion,
cybe bullying, and suicide in child en and adolescen s. JAMA Pedia ics, 168, 435.
Va gas-Mon oya, L., Gimenez, G., & Fe n´andez-Gu i´e ez, M. (2023). ICT use o lea n-
ing and s uden s’ ou comes: Does he coun y’s de elopmen le el ma e ? Socio-
Economic Planning Sciences, 87, 101550.
Ve duyn, P., Lee, D. S., Pa k, J., Shablack, H., O ell, A., Baye , J., Yba a, O., Jonides, J.,
& K oss, E. (2015). Passi e acebook usage unde mines a ec i e well-being: Expe i-
men al and longi udinal e idence. Jou nal o Expe imen al Psychology: Gene al, 144,
480–488.
Villa, J. M. (2016). Di : Simpli ying he es ima ion o di e ence-in-di e ences ea men
e ec s. The S a a Jou nal, 16(1), 52–71.
Viollaz, M., & Winkle , H. (2022). Does he in e ne educe gende gaps? The case o Jo dan.
The Jou nal o De elopmen S udies, 58, 436–453.
Wassie, G., Wondie, Y., Anagaw, B., & Zemene, F. (2023). In e ne se ices ma u i y in
E hiopia: Impac s on he digi al economy and iden i ica ion o a eas o imp o emen
wi h policy ecommenda ions.
WHO. (2021, No embe ). Men al heal h o adolescen s. Re ie ed Sep embe 15, 2024, om
h ps://www.who.in /news- oom/ ac -shee s/de ail/adolescen -men al-heal h
Wohn, D. Y., Ellison, N. B., Khan, M. L., Fewins-Bliss, R., & G ay, R. (2013). The ole o
social media in shaping i s -gene a ion high school s uden s’ college aspi a ions: A
social capi al lens. Compu e s & Educa ion, 63, 424–436.
Yang, Z., G i i hs, M. D., Yan, Z., & Xu, W. (2021). Can wa ching online ideos be addic-
i e? A quali a i e explo a ion o online ideo wa ching among Chinese young adul s.
In e na ional Jou nal o En i onmen al Resea ch and Public Heal h, 18, 7247.
Young Li es. (2018, Janua y). Young li es su ey design and sampling ( ound 5) E hiopia.
Young Li es. (2023a). A guide o lis ening o young li es a wo k calls 1 o 5 cons uc ed
iles.
Young Li es. (2023b, Ma ch). A i ion epo : Fou h call lis ening o young li es a wo k:
Co id-19.
Yue, Z., Zhang, R., & Xiao, J. (2022). Passi e social media use and psychological well-being
du ing he co id-19 pandemic: The ole o social compa ison and emo ion egula ion.
Compu e s in Human Beha io , 127, 107050.
31
Appendices
Figu e A1: Weal h index dis ibu ion
Sou ce: Au ho s’, using YL da a
Table B1: Men al heal h symp oms ques ionnai es
Indica o s Ques ion i ems Scales
Gene alized
Anxie y
Diso de -7
(GAD-7)
In he pas 2 weeks, ha e you been? 1. Feeling ne ous, anxious o
on edge; 2. No being able o s op o con ol wo ying; 3. Wo ying
oo much abou di e en hings; 4. T ouble elaxing/ Can’ elax; 5.
Being so es less ha i ’s ha d o si s ill; 6. Becoming easily annoyed
o i i able; 7. Feeling a aid as i some hing aw ul migh happen.
(Like scales om no a all, se e al days, o nea ly e e y day).
Sco es be ween 5 -
9, be ween 10 - 14,
and abo e 15 ep-
esen mild, mode -
a e, and se e e anx-
ie y symp oms, e-
spec i ely
Pa ien
Heal h
Ques-
ionnai e
dep ession
scale-8
(PHQ-8)
In he pas 2 weeks, ha e you been? 1. Li le in e es o pleasu e in
doing hings; 2. Feeling down, dep essed o hopeless; 3. T ouble alling
o s aying asleep, o sleeping oo much; 4. Feeling i ed o ha ing li le
ene gy; 5. Poo appe i e o o e ea ing; 6. Feeling bad abou you sel
- o ha you a e a ailu e o ha e le you sel o you amily down;
7. T ouble concen a ing on hings, such as eading he newspape o
wa ching ele ision; 8. Mo ing o speaking so slowly ha o he people
could ha e no iced. O he opposi e - being so idge y o es less ha
you ha e been mo ing a ound a lo mo e han usual.
Sco e be ween 5 -
9 indica es mild, 10
- 14 mode a e, 15
- 19 mode a ely se-
e e, and sco es abo e
19 se e e, dep essi e
symp oms. (Like
scales om no a
all, se e al days, o
nea ly e e y day).
Sou ce: o ma ed om Young Li es (2023a).
32
Figu e A2: In e ne access and subjec i e wellbeing
Sou ce: Au ho s’ illus a ion based on YL da a
Figu e A3: In e ne access and anxie y sco es (GAD-7)
Sou ce: Au ho s’ illus a ion based on YL da a
33
Figu e A4: In e ne access and dep ession sco es (PHQ-8)
Sou ce: Au ho s’ illus a ion based on YL da a
Figu e A5: Dis ibu ion o p opensi y sco es
Sou ce: Au ho s’, using YL da a
34
Figu e A6: Mo he s’ educa ion a iable dis ibu ion
Sou ce: Au ho s’, using YL da a
Figu e A7: YL child en’s educa ion a iable dis ibu ion
Sou ce: Au ho s’, using YL da a
35
Table B2: B oade impac s on o he human capi al- ela ed a iables
(1) (2)
(Pe cei ed) ime spen s udying Wan o go back o educa ion
In e ne -0.205∗∗∗ -0.225∗∗
(0.0750) (0.0927)
Com o able 0.0123 0.0705
(0.0710) (0.0800)
S uggle -0.117 -0.0638
(0.0720) (0.0848)
Poo 0.0749 0.141
(0.0786) (0.0858)
Des i u e 0.171 -0.473∗∗
(0.225) (0.191)
Female 0.00573 -0.0481∗∗
(0.0238) (0.0243)
(Belie ed) in ec ed 0.287∗0.272∗∗∗
(0.166) (0.0506)
Income dec eases -0.0554∗∗ -0.109∗∗∗
(0.0258) (0.0256)
Run ou o ood -0.0298 -0.188∗∗∗
(0.0289) (0.0397)
U ban a ea 0.140∗∗∗ 0.227∗∗∗
(0.0437) (0.0624)
Cons an 0.332∗∗∗ 0.867∗∗∗
(0.0778) (0.0871)
Obse a ions 1370 1162
F-s a is ic 184.21 121.15
No es: Samples include child en in Call 2 due o da a a ailabili y. The ins umen o IV es ima es is he
p opo ion o households in a communi y ge ing access o he in e ne . Each column epo s es ima ed
e ec s o in e ne access on he likelihood ha ing spen mo e ime on s udying and wan o go back o
school a e COVID-19. The c i ical alue o he S ock-Yogo es wi h 10% ole ance is 16.38. Robus
s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
36
Table B3: Robus ness es wi h ime- a ian communi y con ols
(1) (2)
GAD-7 PHQ-8
In e ne 11.840∗∗∗ 9.613∗∗∗
(1.893) (1.609)
Communi y income educ ions 2.252∗∗∗ 2.189∗∗∗
(0.785) (0.645)
Communi y ood sho ages 0.869 0.169
(0.913) (0.862)
Communi y a e age weal h le el 0.651 -0.046
(0.577) (0.469)
O he con ol Yes Yes
Indi idual, Clus e , Call FEs Yes Yes
Obse a ions 1836 1840
Kleibe gen-Paap Wald F s a is ic 66.10 67.45
No es: Each column epo s es ima ed e ec s o in e ne access on men al heal h p oblem indica o s:
GAD-7: Gene alized Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e–8. Co a ia es: (belie ed o
be) in ec ed by COVID-19, nega i e income shocks, un ou o ood. The ins umen is he p opo ion o
households in a communi y ha ing access o he in e ne . The c i ical alue o he S ock-Yogo es wi h
10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
37
Table B11: He e ogenei y o e ec s o in e ne on men al heal h by SES in IV models using
Call 5 da a
Well-being GAD-7 PHQ-8
(1) (2) (3) (4) (5) (6)
Weal h index Lowe Highe Lowe Highe Lowe Highe
In e ne -4.358∗∗∗ -2.233∗∗∗ 12.10∗∗∗ 2.778∗∗∗ 9.581∗∗∗ 1.952∗∗∗
(1.399) (0.408) (3.197) (0.618) (2.838) (0.597)
Com o able -2.754∗∗∗ -1.489∗∗∗ 4.575∗1.183∗4.193∗∗ 1.364∗∗
(1.048) (0.442) (2.394) (0.669) (2.126) (0.647)
S uggle -2.827∗∗∗ -1.201∗∗∗ 4.492∗1.641∗∗ 4.336∗∗ 1.561∗∗
(1.032) (0.451) (2.355) (0.683) (2.092) (0.659)
Poo -4.672∗∗∗ -3.322∗∗∗ 5.486∗∗ 2.364∗∗∗ 4.274∗∗ 1.873∗∗
(1.070) (0.521) (2.442) (0.789) (2.170) (0.762)
Des i u e -7.186∗∗∗ 0 10.01∗∗ 0 12.29∗∗∗ 0
(1.982) (.) (4.524) (.) (4.019) (.)
HH size 0.0867 0.0711 -0.182 0.0495 -0.287∗∗ -0.0363
(0.0573) (0.0461) (0.131) (0.0699) (0.116) (0.0675)
Female 0.0147 -0.0202 -0.337 -0.0515 -0.356 -0.0342
(0.208) (0.166) (0.475) (0.251) (0.421) (0.242)
(Belie ed) in ec ed 0 -1.936 0 4.745∗∗ 0 3.452∗
(.) (1.343) (.) (2.033) (.) (1.964)
Income dec eases 0.143 0.110 0.523 0.156 0.451 -0.142
(0.259) (0.217) (0.592) (0.328) (0.526) (0.317)
Run ou o ood 0.579∗∗ 0.169 -0.445 -1.280∗∗∗ -0.714 -1.056∗∗∗
(0.260) (0.219) (0.594) (0.331) (0.527) (0.320)
U ban a ea 1.530∗∗ 1.724∗∗∗ -5.188∗∗∗ 0.535 -4.198∗∗∗ 0.346
(0.641) (0.463) (1.465) (0.701) (1.301) (0.677)
Cons an 7.072∗∗∗ 6.107∗∗∗ -3.130 -0.923 -1.498 -0.156
(1.091) (0.529) (2.492) (0.802) (2.213) (0.774)
Obse a ions 275 234 274 233 275 234
C agg-Donald Wald F s a is ic 17.274 105.032 17.181 104.308 17.274 105.032
No es: E ec s o in e ne access by weal h g oups o sample who did no ha e in e ne in Call 2. IV
es ima es. The ins umen is he p opo ion o households in a communi y ge ing access o he in e ne .
Each column epo s es ima ed e ec s o in e ne access on men al heal h indica o s: Well-being is
subjec i e well-being, wi h a ange be ween 1 (low well-being) and 9 (high well-being); GAD-7: Gene alized
Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e-8. Robus s anda d e o s in pa en heses. ∗
p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
44
Table B12: He e ogenei y o e ec s o in e ne on men al heal h by gende
SWB GAD-7 PHQ-8
(1) (2) (3) (4) (5) (6)
Female Male Female Male Female Male
In e ne -2.039∗∗ -1.296∗8.998∗∗∗ 9.633∗∗∗ 7.697∗∗∗ 8.947∗∗∗
(0.794) (0.779) (2.114) (2.405) (1.972) (2.198)
(Belie ed) in ec ed 1.023∗0.103 -0.745 7.248∗∗∗ -1.992 -0.713
(0.568) (0.110) (4.770) (1.062) (2.952) (0.516)
Income dec eases -0.0583 -0.0997 1.092∗∗ 0.869∗∗ 0.674∗0.810∗∗
(0.200) (0.137) (0.474) (0.369) (0.407) (0.372)
Run ou o ood 0.319 0.179 0.0902 1.382∗∗ 0.290 1.325∗∗
(0.293) (0.250) (0.462) (0.614) (0.439) (0.631)
Obse a ions 850 996 846 990 846 994
Indi idual, Call FEs Yes Yes Yes Yes Yes Yes
Kleibe gen-Paap k Wald F s a is ic 25.74 26.23 25.31 26.23 25.72 26.28
No es: E ec s o in e ne access by weal h g oups. FE-IV es ima es wi h indi idual and call ixed e ec s.
The ins umen is he p opo ion o households in a communi y ge ing access o he in e ne . Each column
epo s es ima ed e ec s o in e ne access on men al heal h p oblem indica o s: GAD-7: Gene alized
Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e-8. Well-being is subjec i e well-being, wi h a
ange be ween 1 (low well-being) and 9 (high well-being). The c i ical alue o he S ock-Yogo es wi h
10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
45
Table B13: E ec s o in e ne on men al heal h wi h educa ion as con ol
Well-being GAD-7 PHQ-8
(1) (2) (3) (4) (5) (6)
FE FE-IV FE FE-IV FE FE-IV
In e ne 0.0715 -1.969∗∗∗ 0.674∗∗ 9.456∗∗∗ 0.698∗∗∗ 8.765∗∗∗
(0.107) (0.604) (0.269) (1.679) (0.262) (1.583)
Highes educa ion 0.0644∗0.134∗∗∗ 0.110 -0.158 -0.0210 -0.282∗∗
(0.0372) (0.0494) (0.0672) (0.122) (0.0686) (0.126)
(Belie ed) in ec ed 0.177 0.510∗∗ 2.692 3.185 0.0890 -1.580
(0.276) (0.257) (1.969) (3.191) (0.651) (1.623)
Income dec eases -0.0741 -0.141 0.497∗∗∗ 1.021∗∗∗ 0.370∗∗ 0.874∗∗∗
(0.0922) (0.123) (0.181) (0.311) (0.173) (0.296)
Run ou o ood 0.327∗∗ 0.187 0.0612 0.854∗∗ 0.370∗0.956∗∗
(0.153) (0.199) (0.196) (0.403) (0.216) (0.412)
Cons an 3.924∗∗∗ 0.198 1.296∗∗
(0.316) (0.576) (0.576)
Obse a ions 2216 1804 2206 1794 2210 1798
Indi idual, Call FEs Yes Yes Yes Yes Yes Yes
Kleibe gen-Paap k Wald F s a is ic 47.62 47.87 47.67
No es: E ec s o in e ne access by weal h g oups. FE-IV es ima es wi h indi idual, communi y, and call
ixed e ec s. The ins umen is he p opo ion o households in a communi y ge ing access o he in e ne .
Each column epo s es ima ed e ec s o in e ne access on men al heal h p oblem indica o s: GAD-7:
Gene alized Anxie y Diso de -7; PHQ-8: Pa ien Heal h Ques ionnai e-8. Well-being is subjec i e
well-being, wi h a ange be ween 1 (low well-being) and 9 (high well-being). The c i ical alue o he
S ock-Yogo es wi h 10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05,
∗∗∗ p < 0.01
46
Table B14: In e ne access and sugges i e mechanisms, able showing con ol a iables
Time doing no hing Mee ing iends Time HH cho es
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Weal h All Low High All Low High All Low High
In e ne 0.148∗0.329∗∗ 0.147 -0.482∗∗∗ -0.566∗∗ -0.602∗∗ -0.0178 -0.0392 0.0783
(0.0777) (0.142) (0.151) (0.0951) (0.265) (0.251) (0.0823) (0.210) (0.189)
Com o . -0.168∗∗ -0.559∗∗ -0.183 -0.0936 0.0912 -0.118 0.151∗∗ 0.263 0.345∗∗∗
(0.0753) (0.229) (0.135) (0.0741) (0.219) (0.127) (0.0717) (0.208) (0.119)
S uggle -0.0932 -0.598∗∗∗ -0.117 -0.0905 0.0517 -0.0730 0.123∗0.162 0.301∗∗
(0.0777) (0.232) (0.141) (0.0783) (0.223) (0.139) (0.0747) (0.211) (0.127)
Poo -0.153∗-0.520∗∗ -0.137 -0.255∗∗∗ -0.00731 -0.195 0.251∗∗∗ 0.392∗0.431∗∗∗
(0.0811) (0.229) (0.153) (0.0817) (0.219) (0.150) (0.0792) (0.209) (0.141)
Des i u e 0.261 -0.0459 0.635∗∗∗ -0.393 0.103 -0.245 0.569∗∗∗ 0.537 0.967∗∗∗
(0.197) (0.305) (0.172) (0.264) (0.367) (0.555) (0.180) (0.340) (0.177)
Female 0.0183 0.0848∗∗ 0.0342 -0.130∗∗∗ -0.143∗∗ -0.141∗∗∗ 0.339∗∗∗ 0.235∗∗∗ 0.237∗∗∗
(0.0228) (0.0381) (0.0411) (0.0289) (0.0557) (0.0547) (0.0247) (0.0471) (0.0465)
In ec ed 0.232 0 -0.301∗∗∗ 0.286∗0 -0.310∗∗∗ -0.0639 0 -0.597∗∗∗
(0.154) (.) (0.0560) (0.149) (.) (0.0646) (0.136) (.) (0.0935)
Income 0.00766 0.124∗∗∗ 0.00908 -0.0373 -0.0320 -0.0656 -0.0959∗∗∗ -0.0697 -0.168∗∗∗
(0.0243) (0.0414) (0.0409) (0.0313) (0.0551) (0.0562) (0.0268) (0.0483) (0.0479)
Food 0.111∗∗∗ 0.156∗∗ 0.113 0.208∗∗∗ 0.0358 0.148 0.0532 -0.0107 0.126
(0.0330) (0.0737) (0.0738) (0.0374) (0.0901) (0.0929) (0.0327) (0.0834) (0.0838)
U ban 0.0131 -0.0746 -0.0104 0.254∗∗∗ 0.328∗0.413∗∗ -0.0382 -0.0804 -0.141
(0.0506) (0.103) (0.114) (0.0603) (0.191) (0.181) (0.0493) (0.141) (0.136)
Cons . 0.282∗∗∗ 0.507∗∗ 0.276∗0.784∗∗∗ 0.521∗∗ 0.769∗∗∗ 0.283∗∗∗ 0.226 0.184
(0.0806) (0.226) (0.152) (0.0800) (0.213) (0.168) (0.0794) (0.207) (0.148)
Obs. 1418 454 435 1354 431 421 1419 454 435
F s a . 179.55 38.79 41.58 186.38 40.34 34.74 179.55 38.79 41.58
No es: Samples include child en in Call 2 due o da a a ailabili y. IV es ima es, he ins umen is he
p opo ion o households in a communi y ge ing access o he in e ne . Each column epo s es ima ed
e ec s o in e ne access on he likelihood ha ing spen mo e ime on doing no hing. Co a ia es: pe cei ed
weal h le els, household size, gende , (belie ed o be) in ec ed by COVID-19, nega i e income shocks, un
ou o ood, u ban. F-s a is ic is he Kleibe gen-Paap k Wald F-s a is ic. The c i ical alue o he
S ock-Yogo es wi h 10% ole ance is 16.38. Robus s anda d e o s in pa en heses. ∗p < 0.1, ∗∗ p < 0.05,
∗∗∗ p < 0.01
47
The UNU-MERIT WORKING Pape Se ies
2025-01 De elopmen s a egies o he g een hyd ogen economy in eme ging economies
by Fabianna Bacil, An hony Black, Ma ina Domingues, Jun Jin, Rasmus Lema, Glen
Robbins and Sö en Schol in
2025-02 Do global alue chains and local capabili ies ma e o economic complexi y in EU
egions? by R. Boschma, E. He nández-Rod íguez, A. Mo ison and C. Pie obelli
2025-03 Inno a ion and p oduc i i y: he ecen empi ical li e a u e and he s a e o he a
by Jacques Mai esse, Pie e Mohnen and Ad No en
2025-04 Digi al Technologies, Hi ing, T aining, and Fi m Ou comes: E idence om AI and ICT
Adop ion in Indian i mss by Sneha Ma ydas, Nandi ha Ma hew, Gio dano De
Ma zo and Ca lo Pie obelli
2025-05 New echnologies and employmen : he s a e o he a by Ma co Vi a elli and
Guille mo A enas Díaz
2025-06 Fo mal-In o mal Supply Chain Linkages and Fi m P oduc i i y in Sub-Saha an A ica:
The Role o Human Capi al by Gbenoukpo Robe Djidonou, Neil Fos e -McG ego
and Nandi ha Ma hew
2025-07 In e ne Access and You h’s Men al Heal h and Well-being: E idence om E hiopia
by T ang Thi Pham and Pui-Hang Wong
