Pelli, Ma ino; Tschopp, Jeanne
Wo king Pape
S o ms, ea ly educa ion, and human capi al
ADB Economics Wo king Pape Se ies, No. 743
P o ided in Coope a ion wi h:
Asian De elopmen Bank (ADB), Manila
Sugges ed Ci a ion: Pelli, Ma ino; Tschopp, Jeanne (2024) : S o ms, ea ly educa ion, and human
capi al, ADB Economics Wo king Pape Se ies, No. 743, Asian De elopmen Bank (ADB), Manila,
h ps://doi.o g/10.22617/WPS240462-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/305430
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
STORMS, EARLY EDUCATION,
AND HUMAN CAPITAL
Ma ino Pelli and Jeanne Tschopp
ADB ECONOMICS
WORKING PAPER SERIES
NO. 743
Oc obe 2024
S o ms, Ea ly Educa ion, and Human Capi al
This pape examines he impac o school-age exposu e o s o ms on educa ion and employmen ou comes
in India. Using wind exposu e his o ies, he pape shows ha exposu e o an a e age s o m can cause a 2.4
pe cen age poin inc ease in educa ional delays, a 2 pe cen age poin d op in pos -seconda y educa ion
a ainmen , and a 1.6 pe cen age poin decline in egula sala ied employmen . The pape also highligh s he
ole o damaged school in as uc u e and declining household income in d i ing hese ou comes.
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
S o ms, Ea ly Educa ion, and Human Capi al
Ma ino Pelli and Jeanne Tschopp
No. 743 | Oc obe 2024
Ma ino Pelli ([email protected] g) is a senio economis
a he Economic Resea ch and De elopmen Impac
Depa men , Asian De elopmen Bank.
Jeanne Tschopp ([email p o ec ed]) is an
assis an p o esso a he Depa men o Economics,
Uni e si y o Be n.
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. WPS240462-2
DOI: h p://dx.doi.o g/10.22617/WPS240462-2
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
oano 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
oob 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
heADB logo.
Co igenda o ADB publica ions may be ound a h p://www.adb.o g/publica ions/co igenda.
ABSTRACT
This pape explo es how school-age exposu e o s o ms impac s he educa ion and
p ima y ac i i y s a us o young adul s in India. Using a c oss-sec ional coho s udy
based on wind exposu e his o ies, we ind e idence o a signi ican deskilling o a eas
ulne able o clima e change- ela ed isks. Speci ically, ou esul s show a 2.4
pe cen age poin inc ease in he p obabili y o acc uing educa ional delays, a 2
pe cen age poin decline in pos -seconda y educa ion achie emen , and a 1.6
pe cen age poin educ ion in ob aining egula sala ied jobs. Addi ionally, ou s udy
p o ides e idence ha deg aded school in as uc u e and declining household income
con ibu e o hese indings.
Keywo ds: clima e change, s o ms, educa ion, human capi al
JEL codes: I25, O12, Q54
We a e g a e ul o he Social Sciences and Humani ies Resea ch Council (SSHRC, g an numbe 039367),
and he Swiss Na ional Founda ion (SNF, g an numbe 100018_192553) o hei inancial suppo . We
hank, wi hou implica ing hem, Tee a Ga g, Blaise Melly, Ben Sand, E ic S obl. All emaining e o s a e
ou s.
1 In oduc ion
Low- and middle-income na ions con on an ino dina ely ele a ed isk o ca as ophes,
as hey a e bo h mo e exposed o clima e- ela ed haza ds and possess diminished
esilience (Dell e al., 2014).1This de iciency in esilience ende s child en pa icula ly
suscep ible, a conce ning ac gi en ha ex eme wea he e en s a e an icipa ed o
escala e due o clima e change (IPCC, 2023; Emanuel, 2021). Ou s udy del es in o he
endu ing ad e se consequences o school-age exposu e o opical s o ms and cyclones
on educa ion and pu sui s du ing ea ly adul hood in India. We sc u inize po en ial causal
pa hways spanning he school yea s, a c ucial phase demons a ed o shape li e ime
ea nings (e.g., O eopoulos, 2007; Ang is and K uege , 1991).
In his pape , we begin by e alua ing he ami ica ions o s o m exposu e du ing
school-age on educa ional ou comes in bo h he sho and long un, using a
c oss-sec ional coho s udy based on he 2018 elease o he Pe iodic Labou Fo ce
Su ey (PLFS). We assess educa ional ou comes by conside ing yea s o schooling and
he highes educa ional le el achie ed. To cap u e he cumula i e e ec s o s o ms
h oughou school yea s, we de ise a con inuous ea men ha agg ega es wind
exposu e his o ies o each dis ic and coho bo n be ween 1985-1995. Con as ing
wi h o he en i onmen al impac s udies (e.g., Ebens ein e al., 2016; Deuche and
Fel e, 2015), ou ocus cen e s on long- e m exposu e o s o ms, which enables us o
gauge he consequences o clima e change as opposed o me e wea he a iabili y. Ou
esul s indica e ha long- un s o m exposu e du ing school yea s causes educa ional
delays, and exe s signi ican las ing e ec s on educa ional a ainmen and ca ee
choices in ea ly adul hood.
An a e age s o m exposu e du ing school yea s yields an inc eased likelihood o
expe iencing an educa ional delay by 2.4 pe cen age poin s, which ansla es o a 7.25%
ise in he ac ion o delayed indi iduals. We also obse e a no able decline in he
numbe o indi iduals a aining pos -seconda y educa ion. An a e age exposu e leads o
a dec ease in his p obabili y by 2 pe cen age poin s, co esponding o a 7.35%
educ ion in he p opo ion o indi iduals wi h his le el o educa ion.
Fu he mo e, we iden i y de imen al e ec s o s o m exposu e on labo ma ke
ou comes. An a e age school-age exposu e esul s in a 1.6 pe cen age-poin dec ease
1De eloping coun ies’ educed esilience s ems om ac o s such as insu icien in as uc u e, weak social
sa e y ne s, ma ke ailu es like absen c edi and insu ance ma ke s, and an absence o e ec i e ea ly
wa ning sys ems and comp ehensi e disas e isk managemen . Re e o Hallega e e al. (2020) o a
comp ehensi e e iew o why economically disad an aged popula ions a e disp opo iona ely impac ed by
na u al haza ds and ela ed disas e s.
in he p opo ion o indi iduals employed as egula wo ke s, while concu en ly causing
a simila inc ease in he sha e o indi iduals occupied wi h domes ic du ies. These
s a is ics equa e o an 8% educ ion in he sha e o indi iduals employed as egula
wo ke s and a 4.8% ise in he sha e o indi iduals in ol ed in domes ic du ies.
Addi ionally, we obse e ha an a e age exposu e incu s a 3.9% educ ion in hou ly
wages. Se e e opical s o ms, once uncommon bu now p og essi ely p e alen due o
clima e change, ampli y all hese indings.
Ou esul s emain obus h ough a se ies o checks, encompassing a alsi ica ion es
and al e na i e speci ica ions o school-age exposu e o s o ms. We also demons a e ha
ou es ima es a e no in luenced by ea ly-li e exposu e o s o ms o o he en i onmen al
ac o s such as p ecipi a ion and empe a u e du ing ages 5-15.
These indings show ha p olonged exposu e o ex eme en i onmen al shocks
du ing school yea s may con ibu e o he p og essi e deskilling o egions mo e
suscep ible o clima e-change ela ed isks. This deg ada ion o skills will exace ba e
inequali ies, unde sco ing addi ional cos s o clima e change ha ha e no been
ex ensi ely examined hus a . To de ise app op ia e mi iga ing policies, we mus
comp ehend he unde lying pa hways ha d i e ou p ima y esul s.
In he second pa o he pape , we employ supplemen a y da ase s (Consume
Py amids and Dis ic In o ma ion Sys em o Educa ion) o p obe he sho - e m
mechanisms by which s o ms may in luence long- e m educa ional ou comes, ocusing
on hei impac on household income and school in as uc u e damage. Ou indings
e eal he p esence o bo h demand and supply shocks in he schooling sec o . Fi s ly,
using panel local p ojec ions, we ind ha ollowing an a e age s o m, household income
p og essi ely declines, eaching le els app oxima ely 8% below p e-disas e incomes 10
mon hs pos -shock. Secondly, we show ha school closu es signi ican ly escala e in he
a e ma h o an a e age s o m, wi h he p opo ion o closed schools su ging by 7.4%
wi hin wo yea s. Howe e , he impac o an a e age exposu e on he p opo ion o
well-main ained class ooms and eliable elec ici y a ailabili y a schools is modes ,
albei s a is ically signi ican . Las ly, we obse e a educ ion in p ima y school
a endance and a decline in academic pe o mance among middle school s uden s,
which is consis en wi h a nega i e income shock and a educ ion in schooling demand.
These indings o e indi ec e idence ha he endu ing consequences o s o ms on
educa ion ex end beyond he me e physical damages o schools, highligh ing he
signi icance o b oadening pos -disas e policies beyond econs uc ion e o s and
enhancing social sa e y ne s (see De yugina, 2017, o he impo ance o social sa e y
ne s in de eloped coun ies). Speci ically, ou esul s p opose ha inancial ans e s
2
ough o be pai ed wi h policies ad oca ing sus ained educa ion and pos -disas e school
en ollmen , po en ially by condi ioning cash ans e s on school a endance.
Fu he mo e, social policies, such as unemploymen insu ance, could p o e ins umen al
in os e ing esilience and isk managemen in u ban a eas, which a e obse ed o be
pa icula ly ulne able ollowing s o ms.
Ou pape enhances he body o esea ch examining he economic consequences o
en i onmen al dis u bances du ing childhood in de eloping na ions. Thus a , he
li e a u e on his subjec has concen a ed on h ee p ima y analy ical app oaches.
Fi s ly, a signi ican po ion o esea ch in es iga es he sho - e m e ec s o concu en
shocks (e.g., Spence e al., 2016; Bjö kman-Nyq is , 2013; Jensen, 2000).
Secondly, ano he b anch o li e a u e explo es he dele e ious epe cussions o
en i onmen al shocks expe ienced in u e o o ea ly li e (up o 4 yea s old). These shocks
ha e been linked o a ange o sho and long- e m ou comes, including a ious aspec s
o adul li e, such as heal h, educa ion, weal h, and o sp ing ou comes (e.g., Chang
e al., 2022; Hyland and Russ, 2019; Rosales-Rueda, 2018; Ak esh e al., 2017;
Dinkelman, 2017; Maccini and Yang, 2009).
The hi d analy ical app oach del es in o he long- e m consequences o sho - e m
inciden s occu ing la e in li e, beyond ages 0-4. This line o esea ch ypically ocuses
on ei he singula e en s, as illus a ed by Shidiqi e al. (2023), G oppo and K aehne
(2017), and Deuche and Fel e (2015), o on sho - e m occu ences coinciding wi h
c ucial momen s o indi iduals, such as high-s akes exam days, as seen in s udies by
Pa k (2022) and Ebens ein e al. (2016).
Ou pape makes wo main con ibu ions o his body o esea ch. Fi s , by examining
long- e m exposu e, ou indings indica e a po en ial p og essi e deskilling in a eas mo e
suscep ible o clima e change- ela ed isks – an addi ional cos associa ed wi h clima e
change ha has no ye been emphasized in he li e a u e. Second, we concen a e on
he impac o hese isks du ing he school-age yea s, highligh ing he signi icance o his
pe iod in shaping human capi al o ma ion. Recognizing he dis inc ion be ween in ancy
and school-age exposu e is c ucial, as he mechanisms h ough which ad e se shocks
a ec long- e m educa ion likely di e . While in u e o dis up ions a e known o in luence
human capi al ia child en’s heal h, disas e s du ing school-age yea s a e mo e likely o
impac long- e m educa ional a ainmen h ough changes in household income and
schooling in as uc u e.
Ou s udy is mos closely ela ed o Deuche and Fel e (2015) and Shidiqi e al.
(2023). Deuche and Fel e (2015) in es iga es he sho - and long- e m e ec s o Supe
Typhoon Mike on educa ional ou comes in Cebu Island, Philippines. The s udy e eals a
3
nega i e and endu ing impac on educa ion, alongside a g adual ealloca ion o unds
om educa ion o econs uc ion e o s. We build upon hese indings by concen a ing
on long- e m exposu e h ough a con inuous measu e a he han a bina y damage
indica o , o e ing aluable insigh s in o he labo ma ke ou comes o a ec ed child en in
ea ly adul hood and p esen ing an in-dep h analysis o he ac o s unde lying long- e m
educa ional delays.2In a simila ein, Shidiqi e al. (2023) examines he a e ma h o a
po en ea hquake in Yogyaka a, Indonesia, in 2006, obse ing i s epe cussions on
educa ional delays and a ainmen s. Like cyclones, he ea hquake ins iga ed
educa ional se backs and dec eased he likelihood o ul illing manda o y educa ion.
Howe e , con a y o cyclones, ea hquakes appea o hinde educa ion p edominan ly
h ough he des uc ion o school acili ies.
Las ly, ou s udy con ibu es o he li e a u e by conduc ing an ex ensi e examina ion o
he mechanisms connec ing school-age exposu e o s o ms wi h long- e m human capi al
deg ada ion, speci ically ocusing on income and schooling in as uc u e channels. To
ou knowledge, hese channels ha e p edominan ly been explo ed independen ly (see
Baez e al., 2010, o a e iew).
Ou indings on he income channel ela e o he body o esea ch ha e eals he
nega i e impac o economic ecessions on educa ion (e.g., S ua , 2022). The majo i y
o s udies add essing his channel ha e employed di e ence-in-di e ence es ima ions.
In con as , local p ojec ions o e ad an ages in cases o mul iple ea men s, making
hem a sui able al e na i e o he di e ence-in-di e ence app oach o add essing dynamic
ea men e ec s. To he bes o ou knowledge, ew pape s ha e u ilized his me hodology
o s udy en i onmen al shocks.3
Al hough i is undeniable ha disas e s lead o in as uc u e dis up ions and
damages (see e.g. Hallega e e al., 2019), he e a e limi ed s udies speci ically
es ima ing he e ec s o disas e s on educa ional acili ies. Ou esul s indi ec ly highligh
he impo ance o adequa e school in as uc u e o long- e m educa ion and labo
2In a ela ed s udy, Shah and S einbe g (2017) employs ain all as a p oxy o wages in u al
India, demons a ing ha highe ain alls, which co ela e wi h highe wages, in luence human capi al
accumula ion di e en ly depending on he child’s age. This esea ch is pa o a b oade li e a u e on
he impac o economic down u ns on educa ion (e.g., S ua , 2022). While he s udy also sc u inizes he
school-age pe iod, i s p ima y objec i e di e s om ou s, as i seeks o e alua e how a o able economic
condi ions al e he oppo uni y cos o schooling and, consequen ly, he incen i es o child en o a end
school.
3Local p ojec ions ha e been i s p oposed by Jo da (2005). They ha e been used widely in empi ical
mac oeconomics and mo e ecen ly in he con ex o en i onmen al economics (see e.g. Ba a ie i e al.,
2023; Ro h T an and Wilson, 2023; Naguib e al., 2022). As discussed by Dube e al. (2023), hey can be
used as an al e na i e me hodology o deal wi h he issue o dynamic ea men e ec s ha a ises wi h he
di e ence-in-di e ence app oach in he case o mul iple ea men s.
4
indica ing i he indi idual is a emale, a i s -bo n child and Hindu espec i ely.12 δdand δb
a e se s o dis ic FE and coho FE, espec i ely. Al hough ou da a a e c oss-sec ional,
we in oduce a dis ic -speci ic linea ela ionship ac oss coho s, τd, ha accoun s o
di e en ial ends in dis ic -le el educa ion policies and egional dispa i ies in economic
g ow h. Finally, iis he e o e m.13
Inco po a ing dis ic ixed e ec s in all ou model speci ica ions p o ides us wi h a
measu e o s o m exposu e ha is condi ionally exogenous. This allows us o accoun o
he possibili y ha ce ain a eas, such as coas al egions, may be mo e suscep ible o
haza ds. Ea lie s udies ha e demons a ed ha he occu ence o a cyclone does no
p o ide any in o ma ion on he p obabili y o obse ing a simila e en in he same
loca ion in he u u e (see e.g., Pielke e al., 2008; Elsne and Bossak, 2001). The e o e,
i is impossible o p edic he occu ence and exac pa h o s o ms, condi ional on
loca ion. By using dis ic FE, we a e le wi h andom ealiza ions o s o ms, pu ging any
co ela ion be ween loca ional economic decisions and he local dis ibu ion o s o m
exposu e.14 Fu he mo e, ou use o winds exclusi ely o cons uc exposu es can be
conside ed exogenous. Al hough o he s o ms’ haza ds include loods and su ges, hei
impac s a e a ec ed by land managemen and de o es a ion, which ha e been shown o
ac o in o people’s se ling decisions (Pe ko , 2022).
We use he PLFS su ey weigh s o weigh ou obse a ions, and we clus e
s anda d e o s a he s a e le el. Clus e ing a he s a e le el is app op ia e because
educa ion unding and p og ams a e p ima ily adminis e ed a he s a e le el.15
S a e-le el clus e ing also akes in o accoun spa ial co ela ions wi hin a s a e and ime
co ela ions in he exposu e index esul ing om he ac ha he same s o m a ec s
mul iple bi h-yea coho s simul aneously.16
12We include bi h o de as a con ol a iable in ou analysis, as ce ain s udies ha e indica ed ha bi h
o de may ha e an impac on pa en al in es men s in educa ion (see e.g., Black e al., 2005). We do no
include con ols o household headship, ma i al s a us, u al esidency, o household size, as each o
hese a iables could be a ec ed by school-age exposu e o s o ms and may lead o a bad-con ol issue
i included.
13I is wo h no ing ha ou ou comes a e obse ed only once o each indi idual in he c oss-sec ional PLFS
da ase , and hus ou speci ica ions a e nei he dynamic no s agge ed di e ence-in-di e ences models
despi e he inclusion o di e en coho s. Ins ead, we adop a c oss-sec ional coho s udy app oach,
whe e we e ospec i ely e alua e indi iduals’ exposu e his o ies be ween ages 5-15.
14E idence sugges s ha , e en wi h clima e change, any signal will appea in he dis ibu ion o s o m ac i i y
e y g adually (see e.g., Emanuel, 2011). The e o e, i is unlikely ha economic agen s a e awa e o
changes in he local dis ibu ion o s o m exposu e.
15h p://coun ys udies.us/india/37.h m
16We wo k wi h 35 clus e s, including 28 s a es and 7 union e i o ies. Tables D.4 and D.5 in he Online
Appendix show ha ou esul s emain consis en when using dis ic clus e ing o dis ic -coho clus e ing.
11
Resul s. In Table 1, Panels A and B p esen he esul s o equa ion (3). Panel A epo s
he esul s o educa ional delay measu ed as he di e ence be ween he epo ed yea s
o schooling and he expec ed numbe o yea s based on epo ed educa ional a ainmen .
Column (1) p esen s he baseline esul s, which sugges ha exposu e o s o ms
leads o a s a is ically signi ican delay in comple ing a gi en le el o educa ion. The
es ima ed delay o a child wi h uni exposu e is 0.43 yea s on a e age, which ansla es
o oughly a 5-mon h delay. Figu e E.2 o he Online Appendix shows ha while uni
alues in he exposu e index a e excep ional in ou sample pe iod, hey a e obse ed in
Odisha due o he 1999 BOB 06 supe cyclone, he mos se e e and des uc i e opical
cyclone eco ded in India om 1990 o 2000. Al hough ex emely se e e cyclonic s o ms
a e a e, ecen e en s such as s o ms Phailin and Fani in Odisha in 2013 and 2019
espec i ely, supe cyclone Amphan in Wes Bengal in 2020, and se e e cyclonic s o ms
Tauk ae and Yaas in Guja a and Wes Bengal and Odisha espec i ely in 2021, indica e
an inc easing equency o such e en s. The e o e, i is impo an o p o ide an
in e p e a ion o ou es ima es o la ge (uni ) alues o he exposu e index, as i in o ms
on he educa ional long- e m delays ha he cu en gene a ion o school-a ending kids
may ace. I we use he a e age exposu e in he sample o in e p e ou esul s, we ind
ha he educa ional delay is app oxima ely nine school days.17
We examine he obus ness o ou esul s by using di e en se s o ixed e ec s (FE) in
columns (3) and (4). In column (3), we include s a e-coho FE, which allows he economic
condi ions o a s a e a he ime o a coho ’s bi h o a ec long- e m educa ional delays.
Al hough he es ima e is less p ecise and smalle han he baseline, i emains quali a i ely
simila . In he las column, we add s a e-policy FE, which accoun o he in oduc ion o he
new Na ional Policy on Educa ion in 1986 and i s amendmen in 1992. The policy aimed
o p o ide compulso y educa ion o all child en up o he age o 14 and was e ec i ely
adop ed a he s a e le el. The in e ac ion e m co e s coho s bo n in 1985, hose bo n
be ween 1986 and 1991, and hose bo n a e 1991. Resul s a e simila o hose in column
(2) and emain s a is ically signi ican a he 5% le el.
In Panel B o Table 1, we es ima e a linea p obabili y model o examine he impac o
school-age exposu e o s o ms on he likelihood o expe iencing an educa ional delay o a
leas one yea . The esul s a e consis en ac oss speci ica ions and indica e ha exposu e
o s o ms du ing schooling yea s inc eases he likelihood o expe iencing an educa ional
delay. Speci ically, in he baseline speci ica ion (column 1), a uni exposu e (which is likely
o be d i en by se e e e en s) is associa ed wi h a 24 pe cen age poin inc ease in he
17Wi h an a e age s o m exposu e o 0.1in ou sample, 42 weeks pe yea , and a 5-day school week, his
numbe is compu ed as 0.43 ·0.1·42 ·5.
12
p obabili y o accumula ing an educa ional delay (i.e. epea ing a yea o d opping ou ). On
he o he hand, an a e age exposu e inc eases his p obabili y by 2.4 pe cen age poin s.
In ou sample, he p opo ion o indi iduals wi h an educa ional delay o a leas one yea is
0.331 (as shown in Panel B o Table D.3 o he Online Appendix). Based on he baseline
es ima e in Panel B, we can in e ha his p opo ion would inc ease by app oxima ely
72.5% ( o a sha e o 0.571) in he case o an ex eme cyclonic s o m exposu e, and by
7.25% ( o a sha e o 0.355) in he case o an a e age exposu e.
3.2 Educa ional A ainmen
In Panel C o Table 1, we in es iga e whe he s o ms no only cause educa ional delays,
bu also a ec he likelihood o comple ing a gi en le el o educa ion. We use an o de ed
logi model wi h a ca ego ical a iable ep esen ing epo ed educa ional a ainmen
(0=below p ima y, 1=p ima y school, 2=middle school, 3=seconda y educa ion,
4=abo e-seconda y educa ion), whe e ca ego y 0 includes indi iduals who ecei ed
some educa ion bu did no comple e p ima y school. The same se o a iables as in
equa ion (3) is included as con ols.
The i s column o he able displays he o de ed logi es ima es, while columns
(2)-(6) epo he ma ginal e ec s o school-age exposu e o s o ms o each ca ego y o
schooling. All o he es ima es a e s a is ically signi ican and ep esen he pe cen age
poin changes in he p obabili y o comple ing a ce ain le el o educa ion in he case o
uni school-age exposu e o s o ms. In gene al, posi i e exposu e o s o ms inc eases
he p obabili y o no comple ing p ima y school and comple ing a mos p ima y and
middle school, while dec easing he p obabili y o achie ing seconda y and
pos -seconda y educa ion. Speci ically, ou indings show ha uni exposu e educes he
likelihood o achie ing pos -seconda y educa ion by 20 pe cen age poin s. This
ansla es o a 2 pe cen age poin educ ion in he e en o an a e age exposu e.
To gi e a sense o he scale o ou indings, le us conside child en who expe ienced
he 1999 BOB 06 supe cyclone du ing hei schooling yea s (i.e., Cbd = 1). Based on he
educa ional a ainmen p opo ions in Table D.3 o he Online Appendix, he es ima es in
Panel C o Table 1 sugges ha he pe cen age o indi iduals who did no comple e
p ima y school (wi h a mos p ima y educa ion) would inc ease om 2.7% o 7.3% (o
9.8% o 20.8%), while he pe cen age o indi iduals who ob ained pos -seconda y
educa ion (wi h a mos seconda y educa ion) would decline om 27.2% o 7.2% (o
36.5% o 31%). The e o e, i can be in e ed ha exposu e o he 1999 supe cyclone
likely esul ed in a signi ican inc ease in indi iduals lacking basic educa ion. E en o an
13
a e age s o m exposu e, hese pe cen ages emain signi ican . Fo example, he
p opo ion o indi iduals wi h pos -seconda y educa ion would dec ease by 7.35%, while
he pe cen age o indi iduals wi h only p ima y school educa ion would inc ease by 11%.
In Figu e E.3 in he Online Appendix, we u ilize he es ima es ob ained om he
o de ed logi model o isualize he p edic ed p obabili ies o achie ing a pa icula le el
o educa ion ac oss he ange o s o m exposu es om 0 o 1, along wi h hei 95%
con idence in e als. The o e all indings om his analysis e eal ha s o ms esul in a
le wa d shi in he dis ibu ion o educa ional a ainmen , which is pa icula ly conce ning
o de eloping na ions such as India, whe e he dis ibu ion o skills is al eady hea ily
skewed o he le .
3.3 Type o Ac i i y
We expec ha he educa ional dis up ion caused by s o ms du ing compulso y schooling
would a ec he ype o labo ma ke ac i i ies indi iduals pe o m in ea ly adul hood, as
ce ain ypes o jobs equi e highe le els o educa ion o a leas basic eading, w i ing,
and compu ing skills. To in es iga e his issue, we es ima e a educed- o m speci ica ion
o school-age exposu e o s o ms on an indica o a iable o each ype o ac i i y in Panel
A o Table 2. Fo example, in column (1), he dependen a iable is a dummy a iable
equal o 1 i he main ac i i y o indi idual iis egula wo k. We include he same se o
con ols as in equa ion (3) o each ype o ac i i y.
Ou es ima es sugges ha indi iduals who we e exposed o s o ms du ing hei
schooling yea s a e less likely o wo k as egula sala ied wo ke s and mo e likely o
pe o m domes ic du ies. Howe e , we ind no s a is ically signi ican e ec on he
likelihood o being a casual wo ke , sel -employed, o an unpaid amily wo ke . To
illus a e he magni ude o ou esul s, le us conside he labo ma ke impac s
associa ed wi h he a e age posi i e exposu e in ou sample (i.e., Cbd = 0.1). The
es ima e in column (1) implies a 1.6 pe cen age poin educ ion in he p obabili y o being
a egula wo ke . Acco ding o Panel D o Table D.3 o he Online Appendix, 19.6% o
indi iduals in ou sample a e engaged in egula wo k. Thus, he es ima e in column (1)
implies an 8% dec ease in he p obabili y o being employed as a egula wo ke . The
es ima e in column (5) indica es a 1.6 pe cen age poin inc ease in he likelihood o
pe o ming domes ic du ies as he p ima y ac i i y in ea ly adul hood, which co esponds
o a 4.8% change when aking he sha e o indi iduals in ol ed in domes ic du ies (i.e., a
sha e o 0.33) as a baseline. These e ec s a e mo e p onounced o child en who
expe ienced mo e se e e exposu es. Fo example, wi h uni exposu es and aking he
14
same baseline sha es, he es ima es imply changes o app oxima ely 80% and 48% o
egula wo k and domes ic du ies, espec i ely.
In summa y, an a e age s o m inc eases he p obabili y o expe iencing a schooling
delay by 2.4 pe cen age poin s, while concu en ly educing he likelihood o comple ing
pos -seconda y educa ion by 2 pe cen age poin s. These indings a e cong uen wi h a
1.6 pe cen age-poin dec ease in he p obabili y o secu ing a egula sala ied posi ion,
sugges ing ha bo h delays in schooling and a skill educ ion a e likely con ibu ing ac o s.
In Panel B o Table 2, we in es iga e whe he posi i e exposu e o s o ms is
associa ed wi h lowe wages and longe hou s o wo k. In column (1), we es ic he
sample o wo ke s who ecei e a sala y, which explains he d op in sample size. We ind
no e idence ha , condi ional on being employed as a egula wo ke , school-age
exposu e o s o ms has a pe manen e ec on wages. Howe e , he subsample o
wo ke s wi h a posi i e sala y is a selec ed one since exposu e o s o ms educes he
p obabili y o being a egula wo ke (see column 1 o Panel A). To add ess his issue,
we un a Tobi es ima ion and epo he a e age ma ginal e ec (AME) on wages,
e alua ed a he means o he co a ia es, in column (2) o Panel B. The es ima e shows
a nega i e e ec on wages, which is s a is ically signi ican a he 10% le el. This
sugges s ha , on a e age, a supe s o m causes a 39% decline in hou ly wages, o
aking he a e age exposu e, a 3.9% wage d op. This esul is consis en wi h he ac
ha s o ms inc ease educa ional delays and educe he p obabili y o comple ing highe
educa ion.
Column (3) o Panel B shows he esul s on hou s o wo k, ocusing on indi iduals
epo ing posi i e hou s o wo k and a posi i e sala y, as in column (1). In column (4),
we epo he co esponding AME om a Tobi es ima ion, once again e alua ed a he
means o he co a ia es. We ind no e idence ha school-age exposu e o s o ms has a
pe manen e ec on hou s o wo k.
The dis up ion o educa ion caused by s o ms is likely o widen income and social
dispa i ies ac oss di e en dis ic s and age g oups in he long un. Ou indings sugges
ha his inc ease in inequali y is p ima ily d i en by changes in quali ica ions and ypes
o employmen , leading o less secu e and po en ially lowe -paying wo k. Addi ionally,
dispa i ies along he income dis ibu ion may u he inc ease as hose who expe ience
he la ges delays in educa ion o en come om ulne able social g oups.
15
3.4 Robus ness
In his sec ion, we p esen a se ies o obus ness checks. We ocus on he esul s ela ed
o educa ion and e e eade s o Online Appendix C o an analysis o he p ima y ac i i y
s a us o indi iduals.
Ea ly-li e Exposu e o S o ms. I is well-documen ed in he li e a u e ha shocks
expe ienced in ea ly li e can ha e long-las ing nega i e impac s on heal h, educa ion,
and labo ma ke ou comes (see e.g., Almond e al., 2018, o a ecen su ey o his
li e a u e). To e i y ha ou indings a e a ibu able o shocks du ing school-age yea s
a he han ea lie shocks, we augmen ou baseline speci ica ion by inco po a ing a
measu e o s o m exposu e du ing he ea ly yea s o li e (0-4 yea s old), a pe iod
conside ed c i ical o skill o ma ion in de elopmen al psychology, epidemiology, and
economics (see e.g., Duque e al., 2019; Heckman, 2008; Knudsen e al., 2006). The
measu e is simila o ou baseline index (see equa ion 1), bu speci ically ocuses on
ea ly li e.
Table 3 p esen s esul s om his exe cise. In column (1), we p esen ou baseline
speci ica ion. Including s o ms which ook place in ea ly yea s educes he sample o bi h-
yea coho s 1990-1995. Column (2) shows ha es ic ing he sample o indi iduals bo n
a e 1989 does no a ec ou baseline esul s. In column (3), we eplace school-age (Cbd)
wi h ea ly-li e exposu e. Es ima es a e s a is ically insigni ican o all ou comes, excep o
educa ional a ainmen , sugges ing ha o indi iduals who ac ually ecei ed some o mal
schooling, s o ms in ea ly li e ha e li le impac on educa ional delays. Howe e , his
does no necessa ily mean ha ea ly-li e exposu e o s o ms has no impac on educa ion;
i may s ill educe he p obabili y o ecei ing a o mal educa ion, which we a e unable
o assess. Finally, in column (4), we include bo h ea ly-li e and school-age exposu es
simul aneously. The inclusion o he ea ly-li e measu e does no impac he es ima es o
in e es , sugges ing ha ou esul s a e no d i en by s o m shocks ha occu ed in yea s
0-4, and ha school-age yea s a e c ucial o long- e m human capi al o ma ion.18
Falsi ica ion Tes . To con i m he alidi y o ou iden i ica ion s a egy, we conduc a
alsi ica ion es by andomly assigning s o ms o he sample. We shu le he measu e o
exposu e o s o ms ac oss he en i e sample and subs i u e his andomized a iable o
he ac ual exposu e measu e in he baseline speci ica ion. We an icipa e ha he esul s
18We also con ol o a e -school s o m exposu e by summing yea ly exposu es o e he a e -school pe iod
up o 2018 in he analysis o p ima y ac i i y s a us (see Table D.7 in he Online Appendix). Ou baseline
es ima es emain unchanged e en a e adding his con ol.
16
o his exe cise will yield mos ly s a is ically insigni ican es ima es o he a iable o
in e es , while lea ing he s a is ical signi icance o he es ima es o o he a iables
la gely unchanged. We epea his p ocess 1,000 imes and eco d he -s a is ics and
p- alues o each i e a ion.
We p esen he esul s o he alsi ica ion es in Figu e 1, which isually shows he
dis ibu ion o -s a is ics o he coe icien o in e es . Panel A ocuses on he eg ession
on yea s o educa ional delay, while Panel B conside s educa ional delay as a bina y
a iable. Panel C shows he esul s o educa ional a ainmen . The his og ams ep esen
he dis ibu ion o -s a is ics ac oss he 1,000 epe i ions o he alsi ica ion exe cise, wi h
he ed e ical line indica ing he -s a is ic o he baseline es ima es (3.01, 2.96, and -3.79,
espec i ely). We obse e ha he majo i y o he dis ibu ion alls wi hin he -1.96 and 1.96
bounda ies, indica ing ha mos o he coe icien s ob ained h ough he alsely-a ibu ed
s o ms a e s a is ically insigni ican .
Remo ing Ex eme Exposu es. Table 4 examines he sensi i i y o ou esul s o
ex eme alues o exposu e. The baseline esul s a e shown in column (1). In column
(2), we exclude indi iduals om Odisha, which has unusually high alues o exposu e
due o he 1999 supe cyclone BOB 06. The esul s ob ained om his subsample a e
simila o he baseline es ima es. Finally, in column (3), we exclude all winds wi h alues
abo e he 95 h pe cen ile o he wind speed dis ibu ion. As an icipa ed, his leads o
smalle e ec sizes and less p ecise es ima es.
Clima e Con ols. To accoun o he po en ial in luence o gene al clima e condi ions
du ing childhood on human capi al o ma ion and long- e m ou comes, we include con ols
o local clima e e ec s such as p ecipi a ion and empe a u e in columns (4) and (5) o
Table 4.
To his end, we augmen he baseline speci ica ion by in oducing a dis ic -speci ic
a iable measu ing he a e age annual p ecipi a ion (in millime e s) be ween ages 5 and
15. Addi ionally, we include con ols o he a e age empe a u e (in ◦C) and he numbe
o days ha child en in a pa icula dis ic we e exposed o di e en empe a u e anges
(0-10, 10-20, 20-30, and abo e 30◦C) du ing hei school-age yea s. We ob ain he aw
empe a u e and p ecipi a ion da a om he ERA5-Land a chi e, accessed h ough he
Google Ea h Engine, and hen agg ega e hem a he dis ic le el.19
19The ERA5-Land da a is gene a ed by esea che s a he Eu opean Cen e o Medium-Te m Wea he
Fo ecas ing (Muñoz Saba e e al., 2019). I is a clima e eanalysis da ase ha p o ides hou ly wea he
in o ma ion wi h a spa ial esolu ion o 0.1×0.1 deg ees, which is app oxima ely 10×10 kilome e s,
co e ing he pe iod om 1981 o he p esen .
17
Resul s a e shown in columns (4) and (5) o Table 4. Column (4) epo s he baseline
speci ica ion es ima ed on he subse o he sample o which clima e a iables a e
a ailable. In column (5), we include con ols o p ecipi a ion and empe a u e, as
desc ibed ea lie . Ac oss all panels, he coe icien o in e es emains p ecisely
es ima ed and quali a i ely simila o he baseline esul s, albei wi h a sligh ly smalle
size e ec .
Mig a ion. One po en ial sou ce o bias a ises om measu emen e o in s o m
exposu e a ibu ed o mig a ion. In ac , we de e mine s o m exposu e based on
indi iduals’ esiden ial loca ions a he ime o he su ey, as we lack in o ma ion abou
hei esidence du ing hei school yea s. I indi iduals ha e eloca ed o a di e en
dis ic by 2018, ou measu e may con ain e o s, pa icula ly i ei he he dis ic hey le
o he one hey mo ed o had expe ienced s o ms du ing hei compulso y schooling.
To add ess his conce n, we conduc ed an analysis using supplemen a y da a se s,
demons a ing ha pe manen ou -o -dis ic mig a ion is ela i ely a e and ha he
occu ence o a s o m ends o educe he p obabili y o mig a ion (see Online Appendix
A and Foo no e 7). As an addi ional es , ocusing on he PLFS sample, we p opose
excluding ma ied emales om he analysis (column 6 o Table 4). In India, ma iage is
a p ima y d i e o mig a ion, especially o women who ypically mo e o li e wi h hei
husband’s amily. Howe e , his es is admi edly app oxima e, as i is unclea whe he
all ma ied women ha e mig a ed ou o hei dis ic , and i is also unce ain whe he
unma ied women ha e no mig a ed. Simila ly, we obse e ha educa ed indi iduals
had o mo e mo e equen ly, o en o educa ional pu poses. Thus, we also sugges
excluding highly educa ed indi iduals om he sample (column 7 o Table 4), no ing ha
es ima ed delays will only pe ain o indi iduals wi h a mos seconda y educa ion. As
obse ed in columns (6) and (7) o Table 4, he key es ima es emain quali a i ely simila
o he baseline esul s, al hough hey a e somewha less p ecisely es ima ed in he
o me column, possibly due o he exclusion o app oxima ely 40% o he sample.
Educa ion Con ols. To accoun o he ac ha indi iduals wi hin educa ional
ca ego ies may sha e obse able cha ac e is ics o ha e simila abili ies ha make hem
mo e o less likely o expe ience educa ional delay, we p opose wo al e na i e
app oaches.20
20Since school-age exposu e o s o ms impac s bo h educa ional delay and educa ional a ainmen , using
educa ional ca ego ies ixed e ec s would cause a bad-con ol p oblem
18
The i s app oach in ol es using he p edic ed p obabili y o comple ing a epo ed
le el o educa ion, condi ional on obse able indi idual cha ac e is ics, as a p oxy o
educa ional a ainmen . Speci ically, we es ima e a linea p obabili y model on a se o
indi idual cha ac e is ics, such as he indi idual’s gende , yea o bi h, whe he hey a e
a i s -bo n child, and hei eligion (Hindu o non-Hindu) as p edic o s, along wi h
in e ac ions o hese a iables. To a oid a bad con ol issue in he inal eg ession, we
ocus on a subsample o s a es wi h ze o exposu e o s o ms be ween 1990 and 2010,
and es ic ou sel es o indi idual cha ac e is ics ha a e unlikely o be a ec ed by
s o ms. We use hese es ima es o p edic he p obabili y o comple ing each le el o
educa ion o each indi idual in he ull (baseline) sample and use he co esponding
p obabili y as a p oxy o educa ional a ainmen in he inal eg ession. The esul o his
exe cise is p esen ed in column (2) o Table 5 and is highly compa able o he baseline
es ima es p esen ed in column (1).
Second, we p opose including ixed e ec s cap u ing pa en al educa ion, which has
been shown o be a signi ican p edic o o child en’s educa ional achie emen s (Bjö klund
and Sal anes, 2011; Gu yan e al., 2008). Pa en al educa ion is less likely o be a ec ed
by child en’s exposu e o s o ms, al hough he e is a possibili y ha pa en s en olled in
uni e si y may ha e young child en a ending p ima y school. This may be pa icula ly ue
o ela i ely young pa en s. Howe e , i is implausible ha pa en al educa ion o e laps
wi h child en’s compulso y schooling a low le els o educa ion.
One po en ial d awback o his app oach is ha pa en al educa ion is only obse able
i bo h he indi idual and hei pa en s li e in he same household. Addi ionally, since
ma ied women o en mo e in wi h hei husbands’ amilies, he sample may include
ela i ely mo e males han in he baseline. We begin by eplica ing he baseline
app oach on he subse o he sample wi h a ailable da a on pa en al educa ion (column
3 o he able). Despi e he smalle sample size ( ep esen ing only 45% o he ini ial
sample), he coe icien s a e e y simila o he baseline es ima es. In column (4), we
include ixed e ec s o pa en al educa ion, and he es ima es emain nea ly iden ical
ac oss speci ica ions.
Al e na i e Measu es o School-age Exposu e o S o ms. In Table 6, we explo e
al e na i e speci ica ions o Cbd. The i s al e na i e (column 2) measu es exposu e by
summing o e he squa es o yea ly exposu es (i.e., P =b+15
=b+5 x2
d ), which assigns mo e
weigh o s onge exposu es. In con as , he baseline measu e simply sums o e
exposu es, ea ing a dis ic -coho exposed o mul iple small s o ms he same as a
dis ic -coho exposed o one iolen and po en ially des uc i e s o m. By summing
19
ac oss squa es, we can di e en ia e be ween s o m in ensi ies when agg ega ing o e
he yea s. Howe e , mul iple exposu es o e ime in a gi en dis ic a e a e (75.5% o
indi iduals wi h posi i e exposu e expe ienced only one s o m be ween ages 5 and 15),
so we do no expec his al e na i e speci ica ion o subs an ially al e ou esul s.
Second, we expe imen wi h a di e en unc ional o m o cap u e he ela ionship
be ween he o ce exe ed by winds on s uc u es and wind speed exposu e. Emanuel
(2011) no es ha he e a e physical easons o belie e ha damages o building
in as uc u e a e ela ed o wind speed exposu e in a cubic manne . To accoun o his,
we use a cubic speci ica ion in columns (3) and (5) o Table 6, whe e we eplace he
squa e in Equa ion 2 wi h a cube.
Thi d, we modi y he wind speed h eshold ha de ines a s o m. In he pape , we use
a benchma k h eshold o 50 kno s, based on Emanuel (2011), which is likely o cause
damages. Howe e , in columns (4) and (5) o Table 6, we aise he h eshold o 64 kno s,
equi alen o a ca ego y 1 cyclone on he Sa i -Simpson scale. In column (6) o Table
6, we elimina e he h eshold al oge he and include all winds ha occu du ing a cyclone
e en .
Finally, in column (7) o Table 6, we compu e he maximum wind speed hi ing each
dis ic using he HURRECON wind ield model (see Online Appendix B.2 o mo e de ails),
ollowing Boose e al. (2004) ins ead o Deppe mann (1947).
O e all, he es ima es in Table 6 la gely esemble he baseline esul s, excep o
column (6), whe e measu ing exposu e di ec ly wi h winds in oduces a downwa d bias.
This is likely due o he assignmen o non-ze o exposu e alues o dis ic s wi h only
mild wind speeds, which ha dly cause any damage.
4 Income and In as uc u e Channels
Ou s udy indica es ha o school-age child en, exposu e o s o ms leads o educa ional
delays and has nega i e long- e m consequences o bo h academic achie emen s and
ca ee p ospec s. This is conce ning, as a decline in human capi al o ma ion con ibu es
o a deskilling o he popula ion, po en ially impeding economic g ow h.
To de elop e ec i e policy ecommenda ions add essing he consequences o
s o ms, we nex explo e he ways in which educa ion may be impac ed. S o ms can
a ec human capi al h ough wo p ima y channels: shi s in schooling demand and
changes in schooling supply.21
21See Baez e al. (2010) o a comp ehensi e e iew o he li e a u e on he channels h ough which disas e s
damage human capi al.
20
middle school. The indings demons a e ha s o ms ha e a subs an ial impac on
a endance o p ima y school child en in le els C3 o C5, co esponding o ages 8 o 11.
The es ima ed educ ions in a endance a e sizable, wi h a d op o app oxima ely 15%
o a supe s o m and 0.75% o an a e age s o m. Howe e , no s a is ically signi ican
impac s on a endance a e obse ed o any o he le els.
O e all, ou indings align wi h p e ious esea ch indica ing ha wea he shocks and
disas e s igge ed by na u al haza ds lead o a d op in school en ollmen s in de eloping
coun ies (see, o ins ance, Jensen, 2000). Fu he mo e, ou indings lend suppo o
he idea ha a disas e -induced nega i e income shock p omp s pa en s o wi hd aw hei
child en om school. This could be due o inancial cons ain s, as amilies may no longe
ha e he means o a o d schooling, o because hei child en a e equi ed o wo k o
supplemen he household income. This hypo hesis ha disas e s inc ease he incidence
o wo king child en is well-suppo ed by he li e a u e (see, o ins ance, Baez e al., 2010;
De Jan y e al., 2006, o a e iew).
The imp ecise es ima es o le els C1 and C2 sugges ha child en a hese le els
may emain in school because hey a e ei he oo young o wo k o unable o con ibu e
o econs uc ion e o s due o physical limi a ions. Addi ionally, he lack o e ec s on
le els C6-C8 may be a ibu ed o he dedica ion o poo e pa en s whose child en ha e
ad anced o his le el o ensu e ha hei child en comple e middle school. Fu he mo e,
since weal hie child en a e mo e commonly ound in middle school, hey may be less
a ec ed by he income shock, as ou income shock s o y sugges s.
Despi e he lack o impac on a endance a he middle school le el, olde s uden s
may s ill be compelled o wo k o wo k mo e equen ly a e school and on weekends.
This may educe he amoun o ime a ailable o s udying, inc ease a igue, educe
concen a ion in school, and ul ima ely lead o a decline in academic pe o mance. We
explo e his possibili y in Panel B o Table 8, which examines examina ion esul s in he
inal yea o p ima y (C5) and middle (C8) school.29 Columns (1) and (2) ocus on he log
a e age numbe o s uden s who appea ed o he exam, while he subsequen wo
columns examine hose who passed he exam. Columns (6) and (7) show he e ec on
he log a e age numbe o s uden s who sco ed abo e 60
Ou esul s indica e ha he e is no e ec on exam appea ance o pe o mance o
s uden s a le el C5. Howe e , in acco dance wi h ou income shock na a i e, we
disco e ha middle school s uden s expe ience a decline in academic pe o mance.
Fewe s uden s appea o he exam, and e en ewe pass i . Addi ionally, he numbe o
s uden s who pass wi h a g ade abo e 60% also dec eases ollowing a s o m.
29Examina ion esul s o o he yea s o schooling a e no a ailable.
27
The es ima es a e consis en ac oss columns, and unlike he esul s o school
a endance, hey e eal subs an ial e ec s o abou 15% o an a e age s o m. This
implies ha e en a mode a e s o m can ha e signi ican impac s on educa ion. These
indings a e pa icula ly impo an gi en ha a decline in g ades may ul ima ely esul in
educa ional delays, al hough we canno o mally in es iga e his possibili y wi h ou
a ailable da a.
Taken oge he , ou esul s sugges ha he p ima y eason o he es ima ed
long- e m educa ional delays is he signi ican dec ease in income, esul ing in educed
a endance o p ima y school child en and dec eased academic pe o mance o middle
school child en. While he decline in a endance o an a e age s o m is ela i ely small,
he de e io a ion o academic pe o mance is subs an ial. Speci ically, he a e age s o m
esul s in a 15% educ ion in he numbe o child en who appea o he exam, pass he
exam, and ecei e a good g ade. These esul s align wi h p e ious esea ch ha has
demons a ed he ad e se con empo aneous e ec s o disas e s igge ed by na u al
haza ds and ex eme wea he shocks on educa ion in de eloping coun ies (see e.g.,
Deuche and Fel e, 2015; Spence e al., 2016).
5 Conclusion
In his s udy, we examine he impac o s o m exposu e du ing school yea s on long- e m
educa ional a ainmen and p ima y ac i i y s a us among young adul s in India. Ou
indings indica e ha indi iduals who expe ienced a s o m du ing hese c i ical yea s a e
mo e likely o expe ience educa ional delays and less likely o comple e highe
educa ion. Fu he mo e, we obse e a dec ease in he likelihood o secu ing egula
sala ied employmen and an inc ease in he p obabili y o engaging in domes ic du ies as
a p ima y ac i i y.
Ou esul s also p o ide indi ec e idence ha he endu ing e ec s o school-age
s o m exposu e can be a ibu ed o bo h he deg ada ion o educa ional in as uc u e
and a decline in household demand o schooling due o educed income. These indings
align wi h he no ion ha ad e se income shocks can lead o an inc ease in wo king
child en, mani es ed as dec eased school a endance o p ima y s uden s and
diminished academic pe o mance o middle schoole s.
O e all, ou s udy unde sco es he impo ance o obus social sa e y ne s and he
need o ex end pos -disas e policies beyond me e econs uc ion e o s. Such policies
should in eg a e inancial ans e s wi h educa ional ini ia i es, such as cash ans e s
condi ional on school a endance and enhanced school suppo . Local p ojec ions o
28
household income e eal ha social policies, including unemploymen insu ance, could
be ins umen al in building esilience and managing isk in u ban a eas, which a e
pa icula ly ulne able ollowing s o ms.
While ou esul s o e aluable insigh s, wo ca ea s wa an conside a ion,
sugges ing ha ou indings should be in e p e ed as lowe bounds o he ue e ec s.
Fi s , ou es ima ions do no accoun o he poo es indi iduals who a e likely no
en olled in school due o hei amilies’ low-income s a us. Including his segmen , which
ends o be disp opo iona ely a ec ed by disas e s, would likely yield la ge es ima es.
Second, ou da a only accoun s o indi iduals who su i ed he s o m and i s a e ma h
un il 2018. Al hough s o m- ela ed a ali ies ha e been ela i ely con ained in ecen
yea s, i is impo an o ecognize ha ou esul s may be subjec o su i o ship bias.
Howe e , we do no an icipa e his issue o signi ican ly a ec ou indings.
29
Tables and Figu es
Table 1: Educa ional Delay and Educa ional A ainmen
Educa ional delay
(1) (2) (3)
Panel A: # o yea s
School-age exposu e 0.43∗∗∗ 0.22∗0.28∗∗
(0.14) (0.13) (0.10)
Panel B: yes=1, no=0
School-age exposu e 0.24∗∗∗ 0.20∗∗ 0.18∗∗
(0.080) (0.084) (0.072)
Con ols Yes Yes Yes
S a e-coho FE No Yes No
S a e-policy FE No No Yes
Obse a ions 70,003 70,003 70,003
Panel A: Mean dep. a . 0.52 0.52 0.52
Panel B: Mean dep. a . 0.33 0.33 0.33
Educa ional a ainmen
Logi Below P ima y Middle Seconda y Abo e-seconda y
es ima es p ima y school school educa ion educa ion
(1) (2) (3) (4) (5) (6)
Panel C: Educ. a ainmen
School-age exposu e -1.18∗∗∗ 0.046∗∗∗ 0.11∗∗∗ 0.096∗∗∗ -0.055∗∗∗ -0.20∗∗∗
(0.31) (0.011) (0.029) (0.027) (0.015) (0.052)
Con ols Yes Yes Yes Yes Yes Yes
Obse a ions 70,003 70,003 70,003 70,003 70,003 70,003
Mean dep. a . 0.027 0.098 0.239 0.365 0.272
No es: Panel A and B show esul s on educa ional delay. In Panel A, educa ional delay is calcula ed as he di e ence be ween he
epo ed yea s o schooling and he minimum numbe o yea s equi ed in he schooling sys em o a ain he epo ed educa ional le el.
In Panel B, educa ional delay is measu ed using a dummy a iable ha akes a alue o one i he delay is a leas one yea . In Panel C,
educa ional a ainmen is a ca ego ical a iable indica ing he epo ed le el o educa ion (0=no o mal schooling, 1=p ima y school,
2=middle school, 3=seconda y educa ion, 4=abo e-seconda y educa ion), wi h ca ego y 0 including indi iduals who ecei ed some
educa ion bu did no comple e p ima y school. Column (1) shows he esul s om an o de ed logi es ima ion whe e he dependen
a iable is a ca ego ical a iable indica ing he epo ed educa ional a ainmen . Columns (2) o (6) epo he ma ginal e ec s o
childhood exposu e o s o ms o each ca ego y o schooling. Con ols include dis ic FE, coho FE, dis ic ends, and indi idual
con ols, including dummy a iables indica ing i he indi idual is emale, i s -bo n, and Hindu.Policy FE used in he in e ac ion e ms
include h ee FE co esponding o coho s bo n in 1985, hose bo n be ween 1986 and 1991, and hose bo n a e 1991. ∗p < 0.10,
∗∗ p < 0.05,∗∗∗ p < 0.01. S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
30
Table 2: Type o Ac i i y, Wages and Hou s Wo ked
Regula wo k Casual labo Sel -employed Unpaid amily wo k Domes ic du ies
(1) (2) (3) (4) (5)
Panel A: Type o ac i i y
School-age exposu e -0.16∗∗ -0.0093 -0.046 0.016 0.16∗∗∗
(0.076) (0.075) (0.059) (0.067) (0.056)
Con ols Yes Yes Yes Yes Yes
Obse a ions 70003 70003 70003 70003 70003
Mean dep. a 0.196 0.093 0.132 0.079 0.329
Log hou ly wages Hou s o wo k
Log hou ly wages AME Tobi Hou s o wo k AME Tobi
(1) (2) (3) (4)
Panel B: Wage & hou s wo ked
School-age exposu e 0.018 -0.393∗3.57 -4.33
(0.18) (0.230) (3.70) (4.684)
Con ols Yes Yes Yes Yes
Obse a ions 29,089 70,003 29,089 70,003
Mean dep. a 3.71 53.60
No es: In Panel A, he dependen a iable is a dummy a iable ha akes a alue o 1 i he main ac i i y o he indi idual is
o pe o m egula wo k (column 1), casual labo (column 2), sel -employmen (column 3), wo k as an unpaid amily wo ke
(column 4), o pe o m domes ic du ies (column 5), as desc ibed in he Sec ion 2. In Panel B, he dependen a iable is
he indi idual’s loga i hm o ( eal) hou ly wage in upees (columns 1 and 2) and hou s wo ked (columns 3 and 4). Columns
(1) and (3) es ima e he e ec o childhood exposu e o s o ms using a subsample o indi iduals who epo wages and
hou s wo ked. This subsample mainly consis s o indi iduals engaged in egula wo k and casual labo , and he numbe o
obse a ions may di e sligh ly om ha p esen ed in he summa y s a is ics due o single on obse a ions. In columns (2)
and (4), we epo he a e age ma ginal e ec s (AME) on wages and hou s wo ked, espec i ely, e alua ed a he means
o he co a ia es, using a Tobi es ima ion. Con ols include dis ic FE, coho FE, dis ic ends, and indi idual con ols,
including dummy a iables indica ing i he indi idual is emale, i s -bo n, and Hindu.∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
31
Table 3: Con olling o Ea ly-li e Exposu e (Educa ion)
Baseline Sub-sample Ea ly-li e School & ea ly-li e
(1) (2) (3) (4)
Panel A:
Educ. delay: # o yea s
School-age exposu e 0.43∗∗∗ 0.35∗∗ 0.35∗∗
(0.14) (0.15) (0.14)
Ea ly-li e exposu e -0.16 -0.16
(0.14) (0.14)
Panel B:
Educ. delay: yes=1, no=0
School-age exposu e 0.24∗∗∗ 0.38∗∗∗ 0.38∗∗∗
(0.080) (0.088) (0.087)
Ea ly-li e exposu e -0.086 -0.088
(0.080) (0.080)
Panel C:
Educ. a ainmen
School-age exposu e -1.18∗∗∗ -1.22∗∗∗ -1.22∗∗∗
(0.31) (0.44) (0.43)
Ea ly-li e exposu e 0.82∗∗∗ 0.83∗∗∗
(0.13) (0.15)
Con ols Yes Yes Yes Yes
Obse a ions 70,003 41,892 41,892 41,892
Panel A: Mean dep. a . 0.52
Panel B: Mean dep. a . 0.33
No es: The able p esen s esul s when con olling o ea ly-li e exposu e o s o ms. In Panel A, educa ional delay is calcula ed as
he di e ence be ween epo ed yea s o schooling and he minimum numbe o yea s equi ed in he schooling sys em o a ain
he epo ed educa ional le el. In Panel B, educa ional delay is measu ed using a dummy a iable ha akes a alue o 1 i he
delay is a leas one yea . In Panel C, educa ional a ainmen is a ca ego ical a iable indica ing he epo ed le el o educa ion
(0=no o mal schooling, 1=p ima y school, 2=middle school, 3=seconda y educa ion, 4=abo e-seconda y educa ion), wi h
ca ego y 0 including indi iduals who ecei ed some educa ion bu did no comple e p ima y school. Column (1) shows baseline
es ima es, while column (2) p esen s esul s o he baseline speci ica ion es ima ed on a subsample o indi iduals bo n a e
1989. In columns (3) and (4), he ocus is on he same subsample. Column (3) eplaces he school-age exposu e measu e wi h
he ea ly-li e exposu e index, and column (4) includes bo h measu es simul aneously. Con ols include dis ic FE, coho FE,
dis ic ends, and indi idual con ols, including dummy a iables indica ing i he indi idual is emale, i s -bo n, and Hindu. ∗
p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01. S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
32
Table 4: Addi ional Robus ness (Educa ion)
Excl. Excl. Sub-sample Clima e Excl. Excl.
Baseline O issa ex emes (clima e) con ols emales high educ.
(1) (2) (3) (4) (5) (6) (7)
Panel A:
Educ. delay:
# o yea s
School-age exposu e 0.43∗∗∗ 0.53∗∗ 0.29∗∗ 0.44∗∗∗ 0.22∗∗ 0.40∗0.52∗∗
(0.14) (0.22) (0.15) (0.16) (0.097) (0.20) (0.23)
Panel B:
Educ. delay:
yes=1, no=0
School-age exposu e 0.24∗∗∗ 0.23∗0.15∗0.24∗∗∗ 0.18∗∗ 0.24∗∗ 0.28∗∗
(0.080) (0.14) (0.077) (0.084) (0.071) (0.10) (0.11)
Panel C:
Educ. a ainmen
School-age exposu e -1.18∗∗∗ -0.74∗∗∗ -0.65#-1.30∗∗∗ -0.81∗∗ -0.86#-1.20∗∗∗
(0.31) (0.27) (0.41) (0.24) (0.39) (0.53) (0.41)
Con ols Yes Yes Yes Yes Yes Yes Yes
Clima e con ols No No No No Yes No No
Obse a ions 70,003 67,770 70,003 66,702 66,702 43,989 50,981
Panel A: Mean dep. a . 0.52
Panel B: Mean dep. a . 0.33
No es: The able p esen s esul s a e emo ing ex eme exposu es and con olling o clima e a iables. In Panel A, educa ional
delay is calcula ed as he di e ence be ween epo ed yea s o schooling and he minimum numbe o yea s equi ed in he
schooling sys em o a ain he epo ed educa ional le el. In Panel B, educa ional delay is measu ed using a dummy a iable
ha akes a alue o 1 i he delay is a leas one yea . In Panel C, educa ional a ainmen is a ca ego ical a iable indica ing
he epo ed le el o educa ion (0=no o mal schooling, 1=p ima y school, 2=middle school, 3=seconda y educa ion, 4=abo e-
seconda y educa ion), wi h ca ego y 0 including indi iduals who ecei ed some educa ion bu did no comple e p ima y school.
Column (1) shows baseline es ima es, while column (2) p esen s esul s o he baseline speci ica ion es ima ed on a subsample
o indi iduals loca ed ou side O issa. In column (3), we ecompu e he exposu e index by emo ing all winds wi h alues abo e
he 95 h pe cen ile o he wind dis ibu ion. Column (4) eplica es he baseline speci ica ion on a subsample o which clima e
a iables a e a ailable. In column (5), we include clima e con ols, such as a dis ic -speci ic measu e cap u ing he a e age
yea ly p ecipi a ion (in millime e s) expe ienced be ween ages 5-15. Addi ionally, we include he a e age empe a u e (in ◦C)
and he numbe o exposu e days wi hin empe a u e bins (0-10, 10-20, 20-30, and abo e 30◦C) o which child en o a gi en
dis ic we e exposed du ing school age. Column (6) excludes ma ied emales and column (7) emo es indi iduals wi h abo e-
seconda y educa ion om he sample. Con ols include dis ic FE, coho FE, dis ic ends, and indi idual con ols, including
dummy a iables indica ing i he indi idual is emale, i s -bo n, and Hindu.#p < 0.012,∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
33
Table 5: Educa ional Con ols (Educa ion)
P edic ed Pa en al
Baseline educ. a ainmen Sub-sample educa ion
(1) (2) (3) (4)
Panel A:
Educ. delay: # o yea s
School-age exposu e 0.43∗∗∗ 0.43∗∗∗ 0.37∗0.37∗∗
(0.14) (0.15) (0.19) (0.18)
Panel B:
Educ. delay: yes=1, no=0
School-age exposu e 0.24∗∗∗ 0.24∗∗∗ 0.24∗∗ 0.24∗∗
(0.080) (0.080) (0.11) (0.11)
Panel C:
Educ. a ainmen
School-age exposu e -1.18∗∗∗ -1.17∗∗∗ -0.79 -0.91∗
(0.31) (0.31) (0.59) (0.47)
Con ols Yes Yes Yes Yes
P edic ed educ. a ainmen No Yes No No
Pa en al educa ion No No No Yes
Obse a ions 70,003 70,003 31,243 31,243
Panel A: Mean dep. a . 0.52
Panel B: Mean dep. a . 0.33
No es: The able p esen s esul s on educa ional delay wi h he addi ion o educa ional con ols. In Panel A, educa ional delay
is calcula ed as he di e ence be ween epo ed yea s o schooling and he minimum numbe o yea s equi ed in he schooling
sys em o a ain he epo ed educa ional le el. In Panel B, educa ional delay is measu ed using a dummy a iable ha akes a
alue o 1 i he delay is a leas one yea . In Panel C, educa ional a ainmen is a ca ego ical a iable indica ing he epo ed
le el o educa ion (0=no o mal schooling, 1=p ima y school, 2=middle school, 3=seconda y educa ion, 4=abo e-seconda y
educa ion), wi h ca ego y 0 including indi iduals who ecei ed some educa ion bu did no comple e p ima y school. Column (1)
shows baseline es ima es, while column (2) con ols o he indi idual’s p edic ed p obabili y o comple ing he epo ed le el o
educa ion. Column (3) eplica es he baseline speci ica ion on a subsample o which pa en al educa ion is a ailable. In column
(4), he same sample as in column (3) is used, and pa en al educa ion is addi ionally con olled o . Con ols include dis ic FE,
coho FE, dis ic ends, and indi idual con ols, including dummy a iables indica ing i he indi idual is emale, i s -bo n, and
Hindu.∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01. S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
34
Table 6: Al e na i e Measu es o S o m Exposu e (Educa ion)
Baseline Sum o squa es 50, cubic 64, squa e 64, cubic All winds HURRECON
(1) (2) (3) (4) (5) (6) (7)
Panel A:
Educ. delay: # o yea s
School-age exposu e 0.43∗∗∗ 0.43∗∗ 0.48∗∗ 0.48∗∗ 0.42∗∗ 0.077∗∗∗ 0.39∗∗
(0.14) (0.19) (0.21) (0.18) (0.19) (0.024) (0.16)
Panel B:
Educ. delay: yes=1, no=0
School-age exposu e 0.24∗∗∗ 0.30∗∗∗ 0.30∗∗∗ 0.29∗∗∗ 0.30∗∗∗ 0.044∗∗∗ 0.27∗∗∗
(0.080) (0.077) (0.084) (0.078) (0.076) (0.014) (0.064)
Panel C:
Educ. a ainmen
School-age exposu e -1.18∗∗∗ -1.69∗∗∗ -1.65∗∗∗ -1.59∗∗∗ -1.68∗∗∗ 0.013 -1.24∗∗∗
(0.31) (0.27) (0.14) (0.11) (0.26) (0.085) (0.39)
Con ols Yes Yes Yes Yes Yes Yes Yes
Obse a ions 70,003 70,003 70,003 70,003 70,003 70,003 70,003
Panel A: Mean dep. a . 0.52 0.52 0.52 0.52 0.52 0.52 0.52
Panel B: Mean dep. a . 0.33 0.33 0.33 0.33 0.33 0.33 0.33
No es: The able p esen s esul s on educa ional delay wi h he use o al e na i e speci ica ions o he school-age exposu e o s o ms. In Panel A, educa ional delay is calcula ed as he
di e ence be ween epo ed yea s o schooling and he minimum numbe o yea s equi ed o a ain he epo ed le el o educa ion. In Panel B, educa ional delay is measu ed using a dummy
a iable ha akes a alue o 1 i he delay is a leas one yea . In Panel C, educa ional a ainmen is a ca ego ical a iable indica ing he epo ed le el o educa ion (0=no o mal schooling,
1=p ima y school, 2=middle school, 3=seconda y educa ion, 4=abo e-seconda y educa ion), wi h ca ego y 0 including indi iduals who ecei ed some educa ion bu did no comple e p ima y
school. Column (1) shows baseline es ima es, while columns (2)-(7) p esen esul s based on al e na i e speci ica ions o s o m exposu e. Speci ically, in column (2), s o m exposu e is
calcula ed using he sum o he squa es o yea ly exposu es. In column (3), s o m exposu e is calcula ed using a h eshold o 50 kno s and a cube. In column (4), exposu e is calcula ed
using a h eshold o 64 kno s and a squa e, and in column (5), exposu e is calcula ed using a h eshold o 64 kno s and a cube. Column (6) compu es exposu e using all winds, and inally,
in column (7), exposu e is compu ed using he HURRECON model, a h eshold o 50 kno s, and a squa e. Con ols include dis ic FE, coho FE, dis ic ends, and indi idual con ols,
including dummy a iables indica ing i he indi idual is emale, i s -bo n, and Hindu.∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01. S anda d e o s a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
.
35
Table 7: Damages o School Facili ies and School Des uc ion
Log a g. # o class ooms Sha e o schools
in good condi ions wi h elec ici y wi hou elec ici y wi h un eliable elec ici y
(1) (2) (3) (4)
Panel A:
Damages o school acili ies
S o m exposu e -0.101∗∗ -0.056∗∗∗ 0.045∗∗∗ 0.011∗∗
(0.047) (0.0082) (0.011) (0.0045)
Con ols Yes Yes Yes Yes
Obse a ions 153,789 153,789 153,789 153,789
Mean dep. a . 4.36 0.688 0.285 0.026
Exi sha e o schools Sha e o buildings unde cons uc ion
(1) (2) (3) (4) (5) (6)
Panel B:
School des uc ion
S o m exposu e -0.0068 -0.0046 0.0084 -0.019∗∗∗ -0.025∗∗∗ -0.029∗∗∗
(0.0089) (0.0091) (0.011) (0.0023) (0.0036) (0.0048)
S o m exposu e( −1) 0.016∗∗∗ 0.029∗∗∗ -0.053∗∗∗ -0.056∗∗∗
(0.0036) (0.0054) (0.0084) (0.0093)
S o m exposu e( −2) 0.067∗∗∗ -0.023∗∗∗
(0.019) (0.0056)
Con ols Yes Yes Yes Yes Yes Yes
Obse a ions 109,831 92,918 76,109 109,831 92,918 76,109
Mean dep. a . 0.026 0.008
No es: Panel A p esen s esul s on damages o school acili ies. In column (1), he dependen a iable is he log o he a e age numbe o
class ooms in good condi ions, wi h he a e age aken ac oss schools a he pincode-yea le el. In columns (2) and (3), he dependen a iable
is he sha e o schools wi h and wi hou elec ici y, espec i ely, in a pos al code-yea . In he las column, he dependen a iable is he sha e o
schools wi h un eliable elec ici y. Panel B shows esul s on school des uc ion. The dependen a iable is he sha e o exis ing schools and he
sha e o school buildings unde cons uc ion in columns (1)-(3) and (4)-(6), espec i ely, wi h sha es exp essed ela i e o he numbe o schools
in 2010 and aken wi hin a pos al code-yea . In bo h panels, s o m exposu e is compu ed om wind exposu es a he pos al code le el using a
quad a ic damage unc ion and a 50 kno s h eshold. In Panel B, he numbe o obse a ions di e s sligh ly om ha p esen ed in he summa y
s a is ics due o single on obse a ions ha a e d opped in he es ima ions. In column (1) o Panel A, he mean dependen a iable a he bo om
o he able is p esen ed wi hou logs. Con ols include pos al code FE and dis ic -yea FE. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01. S anda d e o s
a e clus e ed a he s a e le el.
Sou ce: Au ho s’ es ima es.
36
Hsiang, S. M. (2010). Tempe a u es and Cyclones S ongly Associa ed wi h Economic
P oduc ion in he Ca ibbean and Cen al Ame ica. P oceedings o he Na ional Academy
o Sciences, 107(35):15367–15372.
Hsu, S. and Zhongde, Y. (1998). A No e on he Radius o Maximum Wind o Hu icanes.
Jou nal o Coas al Resea ch, 14(2):667–668.
Hyland, M. and Russ, J. (2019). Wa e as Des iny – The Long- e m Impac s o D ough in
Sub-Saha an A ica. Wo ld De elopmen , 115:30–45.
IPCC (2023). Clima e Change 2023: Syn hesis Repo . Con ibu ion o Wo king G oups
I, II and III o he Six h Assessmen Repo o he In e go e nmen al Panel on Clima e
Change [co e w i ing eam, h. lee and j. ome o (eds.)]. IPCC, Gene a, Swi ze land,
184 pp.
Jensen, R. (2000). Ag icul u al Vola ili y and In es men s in Child en. Ame ican Economic
Re iew, 90(2):399–404.
Jo da, O. (2005). Es ima ion and In e ence o Impulse Responses by Local P ojec ions.
Ame ican Economic Re iew, 95:161–182.
Kee hi a ne, S. and Tol, R. S. J. (2018). Impac o Na u al Disas e s on Income Inequali y
in S i Lanka. Wo ld De elopmen , 105:217–230.
Knapp, K. R., K uk, M. C., Le inson, D. H., Diamond, H. J., and Neumann, C. J. (2010).
The In e na ional Bes T ack A chi e o Clima e S ewa dship (IBT ACS): Uni ying
T opical Cyclone Bes T ack Da a. Bulle in o he Ame ican Me eo ological Socie y,
91:363–376.
Knudsen, E. I., Heckman, J. J., Came on, J. L., and Shonko , J. P. (2006). Economic,
Neu obiological, and Beha io al Pe spec i es on Building Ame ica’s Fu u e Wo k o ce.
P oceedings o he Na ional Academy o Sciences, 103(27):10155–10162.
Maccini, S. and Yang, D. (2009). Unde he Wea he : Heal h, Schooling, and Economic
Consequences o Ea ly-Li e Rain all. Ame ican Economic Re iew, 99(3):1006–1026.
Muelle , V., G ay, C., and Kosec, K. (2014). Hea S ess Inc eases Long- e m Human
Mig a ion in Ru al Pakis an. Na u e Clima e Change, 4:182–185.
Muñoz Saba e , J. e al. (2019). ERA5-Land hou ly da a om 1981 o p esen . Cope nicus
Clima e Change Se ice (C3S) Clima e Da a S o e (CDS), 10.
Munshi, K. and Rosenzweig, M. (2016). Ne wo ks and Misalloca ion: Insu ance,
Mig a ion, and he Ru al-U ban Wage Gap. The Ame ican Economic Re iew,
106(1):46–98.
Naguib, C., Poi ie , D., Pelli, M., and Tschopp, J. (2022). The Impac o Cyclones on Local
Economic G ow h: E idence om Local P ojec ions. Economic Le e s, 220:110871.
Ne ia, Y., Nandi, A., and Galea, S. (2008). Pos -T auma ic S ess Diso de Following
Disas e s: A Sys ema ic Re iew. Psychological medicine, 38(4):467.
43
O eopoulos, P. (2007). Do D opou s D op ou oo Soon? Weal h, Heal h and Happiness
om Compulso y Schooling. Jou nal o Public Economics, 91(11):2213–2229.
Pa ag, M. and Yang, D. (2020). Taken by S o m: Hu icanes, Mig an Ne wo ks, and US
Immig a ion. Ame ican Economic Jou nal: Applied Economics, 12(2):250–77.
Pa k, R. J. (2022). Ho Tempe a u e and High-S akes Pe o mance. Jou nal o Human
Resou ces, 57(2):400–434.
Pelli, M. and Tschopp, J. (2017). Compa a i e Ad an age, Capi al Des uc ion, and
Hu icanes. Jou nal o In e na ional Economics, 108(C):315–337.
Pelli, M., Tschopp, J., Bezma e nykh, N., and Eklou, K. M. (2023). In he Eye o he S o m:
Fi ms and Capi al Des uc ion in India. Jou nal o U ban Economics, 134:103529.
Pe ko , I. (2022). Wea he Shocks, Popula ion, and Housing P ices: he Role o
Expec a ion Re isions. Economics o Disas e s and Clima e Change, 6(3):495–540.
Pielke, R., Landsea, C., May ield, M., La e , J., and Pasch, R. (2008). Hu icanes and
Global Wa ming. Ame ican Me eo ological Socie y, pages 1571–1575.
Rosales-Rueda, M. (2018). The Impac o Ea ly Li e Shocks on Human Capi al Fo ma ion:
E idence om El Niño Floods in Ecuado . Jou nal o Heal h Economics, 62:13–44.
Ro h T an, B. and Wilson, D. (2023). The Local Economic Impac o Na u al Disas e s.
Wo king Pape 2020-34, Fede al Rese e Bank o San F ancisco.
Shah, M. and S einbe g, B. M. (2017). D ough o Oppo uni ies: Con empo aneous and
Long-Te m Impac s o Rain all Shocks on Human Capi al. Jou nal o Poli ical Economy,
125(2):527–561.
Shakya, S., Basne , S., and Paudel, J. (2022). Na u al Disas e s and Labo Mig a ion:
E idence om Nepal’s Ea hquake. Wo ld De elopmen , 151:105748.
Sheldon, T. L. and Zhan, C. (2022). The Impac o Hu icanes and Floods on Domes ic
Mig a ion. Jou nal o En i onmen al Economics and Managemen , 115:102726.
Shidiqi, K.-A., Di Paolo, A., and Ál a o Choi (2023). Ea hquake exposu e and schooling:
Impac s and mechanisms. Economics o Educa ion Re iew, 94:102397.
Simpson, R. and Riehl, H. (1981). The Hu icane and I s Impac . Louisiana S a e
Uni e si y P ess.
Spence , N., Polachek, S., and S obl, E. (2016). How Do Hu icanes Impac Scholas ic
Achie emen ? A Ca ibbean Pe spec i e. Na u al Haza ds, 84:1437–1462.
S ua , B. A. (2022). The Long-Run E ec s o Recessions on Educa ion and Income.
Ame ican Economic Jou nal: Applied Economics, 14(1):42–74.
Topalo a, P. (2010). Fac o Immobili y and Regional Impac s o T ade Libe aliza ion:
E idence on Po e y om India. Ame ican Economic Jou nal: Applied Economics,
2(4):1–41.
44
ASIAN DEVELOPMENT BANK
ASIAN DEVELOPMENT BANK
6 ADB A enue, Mandaluyong Ci y
1550 Me o Manila, Philippines
www.adb.o g
STORMS, EARLY EDUCATION,
AND HUMAN CAPITAL
Ma ino Pelli and Jeanne Tschopp
ADB ECONOMICS
WORKING PAPER SERIES
NO. 743
Oc obe 2024
S o ms, Ea ly Educa ion, and Human Capi al
This pape examines he impac o school-age exposu e o s o ms on educa ion and employmen ou comes
in India. Using wind exposu e his o ies, he pape shows ha exposu e o an a e age s o m can cause a 2.4
pe cen age poin inc ease in educa ional delays, a 2 pe cen age poin d op in pos -seconda y educa ion
a ainmen , and a 1.6 pe cen age poin decline in egula sala ied employmen . The pape also highligh s he
ole o damaged school in as uc u e and declining household income in d i ing hese ou comes.
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 69 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.
