Monsees, Daniel; Schmi z, Hend ik
A icle — Published Ve sion
The E ec o Compulso y Schooling on Vaccina ion Agains
COVID
Heal h Economics
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Monsees, Daniel; Schmi z, Hend ik (2025) : The E ec o Compulso y Schooling
on Vaccina ion Agains COVID, Heal h Economics, ISSN 1099-1050, Wiley, Hoboken, NJ, Vol. 34, Iss.
4, pp. 643-654,
h ps://doi.o g/10.1002/hec.4929
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/319348
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by-nc/4.0/
Heal h Economics
-
RESEARCH ARTICLE
OPEN ACCESS
The E ec o Compulso y Schooling on Vaccina ion
Agains COVID
Daniel Monsees
1
| Hend ik Schmi z
1,2
1
RWI Essen, Leibniz Science Campus Ruh , Essen, Ge many |
2
Depa men Economics, Pade bo n Uni e si y, Pade bo n, Ge many
Co espondence: Hend ik Schmi z ([email p o ec ed])
Recei ed: 16 Ap il 2024 | Re ised: 14 Oc obe 2024 | Accep ed: 11 Decembe 2024
Funding: This wo k was suppo by he Deu sche Fo schungsgemeinscha (P ojec : “COVID‐19 as ampli ie o social inequali y—SMall ARea Analyses wi h
Ge man neighbo hood Da a”, p ojec numbe 492351283).
Keywo ds: compulso y schooling | COVID | educa ion | accina ion
ABSTRACT
We s udy he e ec o educa ion on accina ion agains COVID in Ge many in a sample o indi iduals abo e he age o 60. In
o dina y leas squa es eg essions, we ind ha , in his age g oup, one mo e yea o educa ion goes along wi h a 0.7 pe cen age
poin inc ease in he likelihood o ge a COVID accina ion. In wo s age leas squa es eg essions whe e changes in compulso y
schooling laws a e used as exogenous a ia ion o educa ion, he e ec o an addi ional yea o educa ion is es ima ed o be
ze o. The esul s hold o he complie s o he policy change which a e olde indi iduals a he lowes ma gin o educa ion.
JEL Classi ica ion: I12, I26, J14
1
|
In oduc ion
Vaccina ion is conside ed an impo an —maybe e en he mos
impo an —s a egy o o e come he COVID c isis, wi h he
WHO aiming o a 70% accina ion co e age o he gene al
popula ion (WHO 2022). The i s accines we e de eloped
immedia ely a e he ou b eak o COVID and la ge accina ion
campaigns s a ed a ound he end o 2020 in he high income
coun ies. While, a ha ime, accina ion was mainly seen as a
way o s op in ec ions and o achie e he d immuni y, se e al
mu a ions o he i us made his goal ha dly achie able. How-
e e , he undamen al bene i o accina ion seems o be a
s ong educ ion o se e e illness and mo ali y a e a COVID
in ec ion (Nas een e al. 2022; No ds öm, Ballin, and No d-
s öm 2022). High accina ion a es and mu a ions o less le hal
a ian s o he i us a e conside ed he wo domina ing easons
ha mos socie ies la gely u ned back o no mal li e h oughou
he yea 2022 e en hough COVID in ec ions emained on a
high le el (Robe Koch‐Ins i u 2022b).
While in he i s hal o 2021 unde supply o accines was he
mos impo an p oblem, his was sol ed— o high‐income
coun ies—a ound mid o 2021. Then, howe e , a second
p oblem came up: oo low accina ion a es in o de o achie e
he d immuni y. Resea che s a ound he wo ld and ac oss ields
o s udy y o unde s and he de e minan s o accina ion
willingness and hesi ancy, which is necessa y o imp o e he
success and accep ance o accina ion campaigns. While his is
no necessa ily impo an anymo e o he cu en e sion o he
Co ona i us, his knowledge seems o be o high alue o
u u e pandemics bu also o endemic i uses such as in luenza
and po en ial mu a ions o he Co ona i us.
An impo an de e minan o accina ion many schola s can
ag ee on is educa ion. Many s udies epo posi i e associa ions
be ween COVID accina ion willingness and educa ional s a us
a ound he wo ld. See, o example Cascini e al. (2021) o a
gene al o e iew and Be gen e al. (2023), Bo ga e al. (2022),
G aebe , Schmid ‐Pe i, and Sch öde (2020), Mondal, Sinha oy,
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion‐NonComme cial License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal
wo k is p ope ly ci ed and is no used o comme cial pu poses.
© 2025 The Au ho (s). Heal h Economics published by John Wiley & Sons L d.
Heal h Economics, 2025; 34:643–654 643 o 812
h ps://doi.o g/10.1002/hec.4929
and Su (2021), Be gmann e al. (2021), Walkowiak and Walko-
wiak (2021), Huebene and Wagne (2021), Hume e al. (2023),
o a non‐exhaus i e lis o s udies. Some o heses s udies use
ac ual accina ion as ou comes, o he s use s a ed willingness.
Mo eo e , hese s udies di e in impo an aspec s such as
coun ies, how educa ion is measu ed, when he da a a e
collec ed (e.g. be o e/a e accina ion was a ailable, du ing/
a e local peaks o in ec ion a es). Ye , a ypical esul is he
posi i e— hough no necessa ily economically la ge—
associa ion o educa ion and accina ion willingness. This is
an impo an inding as his educa ion g adien can be in e -
p e ed as a p oblem o dis ibu ional jus ice ha , in some way,
would need o be add essed in socie ies ha aim a p e en ing
unjus inequali ies in heal h. I adds o he gene al inding ha
heal h ca e and p e en ion a e inequi ably dis ibu ed (Ca ie i
and Wuebke 2013; Van Doo slae , Koolman, and Jones 2004).
Ano he ques ion is whe he he ela ionship be ween educa-
ion and accina ion is causal. While he a o emen ioned
s udies—especially in hei sum—p o ide compelling e idence
o a co ela ion be ween educa ional a ainmen and COVID
accina ion, o he bes o ou knowledge, no s udy in es iga ed
whe he schooling is a causal ac o o accina ion willingness.
Ce ain cha ac e ai s, like ime o isk p e e ences, migh
d i e bo h schooling and accina ion willingness (Nuschele
and Roede 2016). This would be impo an o unde s and when
i comes o how his educa ion g adien in accina ion could be
ackled. The e exis s some e idence o he impac o educa ion
on gene al heal h beha io s. B unello e al. (2016) show ha
schooling has a nega i e impac on smoking, d inking, exe cise
and he BMI, Fle che and F is old (2009) show ha college
educa ion inc eases he use o p e en i e heal h ca e se ices,
such as physical examina ions, den al examina ions, choles e ol
es s, and lu sho s. In hei e iew, Galama, Lle as‐Muney, and
an Kippe sluis (2018) conclude ha OLS es ima es indica e a
s ong posi i e associa ion be ween schooling and heal h
beha io —in e ms o smoking and obesi y. Quasi‐expe imen al
me hods usually esul in la ge con idence bands, hey s ill
sugges posi i e e ec s. While ou s udy mainly speaks o he
e ec s o educa ion on heal h beha io , a po en ial link be ween
educa ion and heal h beha io is heal h s a us. Fo example,
indi iduals wi h ch onic condi ions we e p io i ized in he
COVID accina ion campaign in Ge many (Vygen‐Bonne
e al. 2021a). The li e a u e on causal e ec s o educa ion on
heal h is qui e la ge. o example, Schmi z and Tawiah (2023),
lis 22 s udies ha exploi exogenous a ia ion in educa ion.
Ne e heless, he indings a e no conclusi e. Acco ding o
Schmi z and Tawiah (2023) mo e han hal o he s udies do no
ind e ec s o educa ion on heal h while he e a e also s udies
ha do ind posi i e e ec s. All in all, o he bes o ou
knowledge, he e is no s udy ha claims o iden i y he causal
e ec o educa ion on accina ion, in pa icula accina ion
agains COVID. I is ou main con ibu ion o he li e a u e o
ill his gap and p o ide a i s s udy in his di ec ion.
In his pape , we s udy he e ec o educa ion on accina ion
agains COVID in Wes Ge many using in o ma ion om h ee
di e en su eys adminis e ed be ween mid o 2021 and begin-
ning o 2022. These su eys ela e ac ual accina ion s a us o
educa ional a ainmen o —in ou es ima ion sample—mo e
han 6000 indi iduals. Ou iden i ica ion s a egy makes use o
changes in compulso y schooling laws and allows us o es ima e
local a e age ea men e ec s o indi iduals be ween 59 and
95 yea s o age. The e o ms inc eased compulso y schooling om
eigh o 9 yea s in he 1940s–1960s in Ge many wi h some a i-
a ion in iming ac oss ede al s a es. Ou esul s only hold o a
speci ic subg oup o indi iduals bu , a guably, a highly impo an
one: olde indi iduals—usually deno ed he high isk g oup and
hus bene i ing mos om accina ions—a he lowes ma gin o
educa ion ( hose o ced o inc ease hei yea s o educa ion om
eigh o 9 yea s). As a esul , we do no ind a signi ican e ec o
an addi ional yea o compulso y schooling on accina ion
agains COVID. In se e al di e en speci ica ions o ins umen al
a iables es ima ions, he es ima ed e ec s luc ua e a ound ze o.
This also holds o an e en ‐s udy speci ica ion o he e ec o
he compulso y schooling e o m on accina ion ha accoun s
o po en ial p oblems o he e ogeneous e ec s in s agge ed‐
en y se ings by applying he es ima o p oposed by Sun and
Ab aham (2021). Finally, gi en ha we ha e a small sample and
imp ecise es ima es a e an issue, we ca y ou an ex‐pos powe
analysis as sugges ed by Black e al. (2022). Doing so, we iden i y
a minimum de ec able e ec size o he compulso y schooling
e o m o an 0.8% poin inc ease in COVID accina ion.
We do no claim ha ou main esul —no sizable local a e age
ea men e ec o educa ion on accina ion agains COVID—
has high ex e nal alidi y and can be gene alized o o he age
g oups o o he educa ion le els. S ill, we hink ha his is an
impo an piece o e idence and a s a o c ea e a pic u e on he
causal e ec o educa ion on COVID accina ion. Mo eo e ,
ge ing esul s ha only hold o speci ic subg oups o complie s
is inhe en o educed‐ o m ins umen al a iables eg essions
o all kind, no jus his s udy.
The emainde o his pape is s uc u ed as ollows. In Sec ion 2
we desc ibe he ins i u ional backg ound on COVID accina ion
and educa ion in Ge many. In Sec ion 3we p esen he da a and
he empi ical app oach. Resul s a e epo ed in Sec ion 4 ol-
lowed by he conclusion in Sec ion 5.
2
|
Ins i u ional Backg ound and COVID
Vaccina ion in Ge many
In his sec ion we p esen he ins i u ional se ing. This includes
he Ge man educa ional sys em, wi h a ocus on he schooling
e o ms ha we use as exogenous a ia ion, as well as he cou se
o he COVID pandemic in Ge many and he accina ion
campaign.
2.1
|
Educa ional Sys em and Schooling Re o ms
S a ing a he age o six, child en in Ge many spend hei i s
ou school yea s in p ima y school, be o e swi ching o one o
he h ee seconda y school acks. These acks can be di e -
en ia ed in o basic (Haup schule), in e media e (Realschule) and
high schools (Gymnasium). The basic ack (up o 8 h o 9 h
g ade) p epa es s uden s o app en iceship, he in e media e
ack (up o 10 h g ade) quali ies s uden s o app en iceship o
644 o 812 Heal h Economics, 2025
aining in whi e colla jobs, and he high school ce i ica e (up
o 12 h o 13 h) gi es access o academic educa ion in colleges
o uni e si ies. Wes Ge man basic ack schools used o co e
g ades i e o eigh , ill compulso y schooling yea s we e
inc eased o nine yea s in he pe iod be ween 1946 and 1969.
This e o m was implemen ed in di e en yea s by he a ious
s a es, as decisions and policies ega ding he educa ional sys-
em in Ge many a e made a he ede al s a e le el. Some s a es
in oduced a compulso y nin h g ade ea lie , while he majo i y
o he s a es only in oduced an addi ional yea o schooling due
o he Hambu g Acco d (Hambu ge Abkommen) in 1964
(Kamhö e and Schmi z 2016). The e exis di e en speci ica-
ions o when he e o ms occu ed and which coho s we e
a ec ed, howe e , we use he e o ms as desc ibed in Bege ow
and Jü ges (2022), as hei esea ch p ocess o iden i ying he
iming o e o ms seems o be he mos ho ough one. See
Table 1 o he e o m yea s. The e o m was in oduced due o a
sho age in labo ma ke oppo uni ies and app en iceships o
school lea e s, and o also inc ease he school lea ing age (see
Pischke and on Wach e 2008, o de ails).
Coinciding wi h hese ex ensions o compulso y schooling was
he in oduc ion o wo sho school yea s (SSY), in 1966 and
1967 in some s a es. The s a o he school yea mo ed om
Sp ing o Fall bu i was al eady in Fall o Ba a ia, see Pischke
and on Wach e (2008) o de ails. Howe e , i is commonly
ound ha hese only ha e li le explana o y powe o heal h
ou comes (Kemp ne , Jü ges, and Reinhold 2011), which is why
we only include hem as a obus ness check.
2.2
|
COVID Vaccina ion
The i s case o COVID in Ge many occu ed in Janua y 2020
(Ro he e al. 2020). The i s accine was app o ed by he Eu-
opean Medical Agency (EMA) a he end o Decembe 2020,
TABLE 1 |Re o m yea s and co esponding i s bi h coho s.
Fede al s a e Pi o al bi h coho Re o m yea
Schleswig Hols ein 1932 1947
Hambu g 1931 1946
Lowe Saxony 1947 1962
B emen 1944 1959
No h Rhine‐Wes phalia 1951 1966
Hesse 1951 1966
Rhineland Pala ina e 1952 1967
Baden‐Wue embe g 1952 1967
Ba a ia 1954 1969
Saa land 1943 1958
Sou ce: Bege ow and Jü ges (2022). Pi o al coho is he i s bi h coho he e o m applies o.
FIGURE 1 |Vaccina ion s a us and 7 day incidence in Ge many. L, Le ‐hand side axis; R, Righ ‐hand side axis. This igu e does no use he
es ima ion sample used in he pape bu shows o icial na ionwide numbe s. Sou ce: Bundesminis e ium ü Gesundhei (2022) and Robe Koch‐
Ins i u (2022a).
645 o 812
accina ions s a ed in Ge many sho ly a e (Die Bundes e-
gie ung 2020). While a i s he e was a p io i iza ion o he
accina ion o hose who we e a isk o se e e consequences o
an in ec ion (Vygen‐Bonne e al. 2021a), his was li ed in June
2021, when enough accines we e a ailable (Vygen‐Bonne
e al. 2021b).
1
Figu e 1shows he sha e o he accina ed pop-
ula ion oge he wi h he 7 day incidence be ween Janua y 2021
and Ap il 2022. The sha e o indi iduals ha ing ecei ed a leas
one accina ion (black long‐dashed line) is inc easing slowly a
i s due o a limi ed supply o accina ion doses. F om Ap il
2021 o July 2021 i inc eases sha ply om abou 12% o mo e
han 60% and hen con e ges o a ound 80%. The accina ion
a e o he second COVID accina ion (g ay sho ‐dashed line)
ollows a simila pa e n wi h a delay o 2 mon h. The hi d
accina ion was a ailable as o June 2021 (g ay dash‐do ) and
he sha e o accina ed con e ged o a lowe numbe o a ound
60%. In his pape , we es ic he analysis o he i s accina-
ion dose agains COVID.
3
|
Da a and Empi ical App oach
3.1
|
Sample Selec ion and Ou come Va iables
We combine h ee da a sou ces. The i s is he CASA moni o
da a se (hence o h called CASA da a), an online‐su ey pu
oge he by in as 360.
2
This ep esen a i e su ey consis s o
h ee c oss sec ions wi h each a ound 10,000 esponden s. The
i s wa e was conduc ed in Feb ua y and Ma ch 2021, he
second wa e in July 2021 and he hi d in Janua y 2022.
3
We
augmen he CASA da a wi h he Ge man pa o he Su ey o
Heal h, Aging and Re i emen in Eu ope (SHARE)
4
which— o
wa e 9 (also called SHARE Co ona su ey 2)—was in he ield
om June o Augus 2021, he eby coinciding wi h he second
CASA wa e. Las ly, we include he Ge man Socio‐Economic
Panel (SOEP)— he mos p ominen and long‐ unning Ge man
ep esen a i e household da a se (Goebel e al. 2019)—in ou
analysis. We use da a om he su ey in 2021, which was
collec ed h oughou he whole yea .
Figu e 2 epo s he ime pe iods when he da a se s we e
collec ed in combina ion wi h COVID accina ion a es in Ge -
many. Since only a mino i y o indi iduals had he possibili y o
ge a COVID accina ion in Feb ua y 2021, we do no make use o
wa e 1 o he CASA da a.
5
Simila ly, we exclude indi iduals om
he SOEP sample i hey we e su eyed be o e May 2021.
The ou come a iable o in e es is COVID accina ion s a us,
de ined as a bina y indica o equal o one i an indi idual has
ecei ed a leas one COVID accina ion. Figu e 3 epo s
accina ion a es by age and da a sou ce/wa e in ou sample.
As can be seen, he sha e o indi iduals wi h a leas one
COVID accina ion was abo e 75% o all age g oups in all
da a se s. This is sligh ly la ge han he numbe s in Figu es 1
and 2. No e, howe e , ha hose epo he numbe s om he
ull popula ion while only indi iduals olde han 50—who
ha e highe accina ion a es—en e Figu e 3. Fo ins ance,
he p io i iza ion o olde indi iduals and hose a isk o se-
e e in ec ion was li ed in June 2021, when enough accines
we e a ailable and 90% o indi iduals o e he age o 70 had
ecei ed a leas one accina ion. We obse e a small age end
in he i s COVID accina ion dose.
Since we exploi he Wes Ge man compulso y schooling e-
o ms, we es ic he analysis o Wes Ge many, ha is, all
ede al s a es lis ed in Table 1. Excep o he SOEP, we do no
ha e he in o ma ion in which s a e indi iduals li ed when
a ending school and he e o e assume in Casa and SHARE ha
hey a ended school in he same s a e hey cu en ly eside in.
While his seems o be a s ong assump ion, i is commonly used
in he li e a u e (see e.g. Kemp ne , Jü ges, and Reinhold 2011).
In hei online appendix, Bege ow and Jü ges (2022) show how
in only abou 5% o all cases in he SOEP he ins umen al
a iable would be inco ec ly assigned i i was based on he
ede al s a e indi iduals a e cu en ly li ing in ins ead o he
ede al s a e hey wen o school in. This sugges s ha mo ing
and s aying ou o he ede al s a e whe e educa ion ook place
is a a e e en in Ge many in his age g oup. This should hold
e en mo e o he g oup o complie s a he lowe educa ion
ma gin ha does no mo e ede al s a es o, o example, a end
a uni e si y.
In o de o ocus on indi iduals bo n a ound he e o m coho s,
we only include indi iduals bo n 7 yea s be o e o 7 yea s a e
he pi o al coho . In obus ness checks we make di e en
sample selec ions, such as i e o 10 yea s a ound he pi o al
coho s (Schneeweis, Ski bekk, and Win e ‐Ebme 2014) o
based on bi h yea s, speci ically he yea s 1945–1960 o 1940 o
1970. In ou main speci ica ion, we ha e a sample size o 6010.
Mo e desc ip i e s a is ics a e p esen ed in Table 2. The a e age
immuniza ion a e in ou sample mi o s he one om Figu e 3.
The bi h coho es ic ions imply ha ou es ima ion sample
only includes indi iduals be ween 59 and 95 yea s o age. While
he es ic ion o olde indi iduals limi s he gene alizabili y o
ou esul s, we a gue ha his ne e heless is a e y in e es ing
subsample as, in pa icula , indi iduals olde han 60 yea s a e
gene ally said o be he high isk g oup o COVID. We assign
indi iduals hei yea s o schooling based on hei highes deg ee
o schooling esul ing in an a e age o 10.41 yea s o schooling.
FIGURE 2 |Timing o he su eys. Sou ce: Bundesminis e ium ü
Gesundhei (2022).
646 o 812 Heal h Economics, 2025
3.2
|
Baseline Empi ical App oach
As a baseline model we s a wi h he ollowing linea eg es-
sion model:
Yics =β0+β1Sics +β2malei+γc+δs+α(ηs×c)+θsu ey
+εics
(1)
whe e Yics is a bina y indica o whe he an indi idual io bi h
coho cin ede al s a e shas ecei ed a COVID accina ion. Sics
a e yea s o schooling. maleiis a dummy a iable o being male.
γc,δsand θsu ey a e bi h coho , ede al s a e and su ey/wa e
ixed e ec s espec i ely. ηs×caccoun s o s a e‐speci ic linea
bi h coho ends as ypically done in his li e a u e (e.g.,
Pischke and on Wach e 2008; Kamhö e and Schmi z 2016;
Cygan‐Rehm 2022.). We a y he speci ica ion in obus ness
checks below. εics deno es he indi idual e o e m. S anda d
e o s a e clus e ed by s a e ×bi h coho s, as his is he le el
o ea men assignmen in he wo s age leas squa es e-
g essions (Pischke and on Wach e 2008).
In o de o de i e es ima es o he e ec o educa ion on
accina ion s a us, we use he changes in compulso y schooling
as an ins umen al a iable o yea s o schooling. This
app oach was i s used by Ang is and K uege (1991) o he
US and la e by Pischke and on Wach e (2008) o Ge many o
es ima e he e ec s o educa ion on wages. We use a wo s age
leas squa es (2SLS) app oach whe e he yea s o schooling a e
eg essed in he i s s age on he same a iables as be o e,
including Ziwhich is an indica o a iable equal o one i in-
di idual iwas bo n in o a s a e‐yea coho , o which
compulso y schooling yea s we e 9 yea s ins ead o eigh and
ze o else.
To de i e a causal e ec , he ins umen Zneeds o be alid and
ele an . To be conside ed ele an , an ins umen mus be
highly co ela ed wi h he explana o y a iable o in e es . We
show his in Sec ion 4.1. We a gue ha he ins umen is indeed
alid in ou se ing, meaning ha he ex ension o compulso y
schooling had no e ec on accina ion s a us o he han
h ough indi idual yea s o schooling. I may be ques ionable o
wha ex en he es ima ion esul s a e d i en by be ween s a e
di e ences since he ede al s a es a e solely esponsible o
educa ional e o ms, which—as also a gued by (Pischke and
on Wach e 2008)—also includes he con en o he cu icula
in he addi ional yea o educa ional a ainmen a e he e-
o m. We a gue ha his is a mino issue in ou analysis as he
a ia ion used o de i ing he causal es ima e p edomina ely
consis o wi hin s a e a ia ion. A simila a gumen could be
made conce ning local in ec ion dynamics, as all indi iduals
wi hin a egion a e exposed o he same in ec ion dynamics,
i espec i e o hei educa ional a ainmen . Ne e heless, we
con ol o local in ec ion dynamics in ou obus ness checks.
The ins umen o compulso y schooling in Ge many has been
used and discussed ex ensi ely in he li e a u e o di e en
FIGURE 3 |Dis ibu ion o COVID accina ion by age and da a sou ce. CASA, SHARE and SOEP da a. Da a o Wes Ge many only. No u he
sample es ic ions (e.g. ega ding bi h coho s). The igu es p esen uncondi ional sample means by age in yea s which a e smoo hed by LOWESS.
TABLE 2 |Desc ip i e s a is ics.
Mean S . de Min Max
Fi s co id accina ion 0.91 0 1
Yea s o schooling 10.41 1.81 8 13
Bi h yea 1952 4.79 1925 1961
Age 69.11 4.79 59 95
Male 0.53 0.5 0 1
Da a sou ce: Casa Wa e 2 0.27 0 1
Da a sou ce: Casa Wa e 3 0.26 0 1
Da a sou ce: SHARE Wa e 9 0.14 0 1
Da a sou ce: SOEP 2021 0.33 0 1
Obse a ions 6010
No es: CASA, SHARE and SOEP da a a e sample selec ion.
647 o 812
ou comes, such as wages (Cygan‐Rehm 2022; Kamhö e and
Schmi z 2016; Pischke and on Wach e 2008), heal h (Kemp -
ne , Jü ges, and Reinhold 2011; Bege ow and Jü ges 2022) o
e ili y (Cygan‐Rehm and Maede 2013), whe e hese au ho s
a gue ha use o his ins umen is a sui able way o deal wi h
endogenei y o schooling.
I ou assump ions hold, he es ima ed coe icien o ins u-
men ed yea s o schooling in he IV eg ession can be in e -
p e ed as he e ec o an addi ional yea o schooling on
accina ion s a us. Allowing o he e ogenous ea men e ec s
and addi ionally assuming mono onici y, we ge an es ima e o a
local a e age ea men e ec (Imbens and Ang is 1994), whe e
he complie subpopula ion is he g oup a he lowes ma gin o
educa ion: indi iduals who would ha e liked o ake 8 yea s o
schooling bu a e o ced o ake nine. In he se ing a hand,
mono onici y implies ha indi iduals do no educe hei
schooling yea s because o he inc ease o compulso y educa-
ion. We a gue ha his is he case, also since compulso y ed-
uca ion is he legal lowe bound o educa ional a ainmen .
The e o e we iden i y e ec s o he subpopula ion o complie s.
4
|
Resul s
4.1
|
Baseline Speci ica ion
Table 3 epo s he esul s om linea eg essions wi h he
speci ica ion desc ibed in Sec ion 3.2. Column (1) show he OLS
esul s, while column (2) shows esul s om ins umen al a -
iables eg essions.
S a ing wi h he OLS esul s, educa ional a ainmen seems o
be posi i ely ela ed o accina ion. An addi ional yea o
schooling goes along wi h a 0.7% poin highe likelihood o ge
COVID accina ion, which is s a is ically signi ican . This pos-
i i e co ela ion is in line wi h he indings om he li e a u e. I
is di icul o compa e he size o he coe icien o hose in he
p e ious li e a u e as mos s udies es ima e odds a ios o
di e en le els o educa ion. Among he ew examples ha use
linea p obabili y models a e Huebene and Wagne (2021),
showing o Ge many ha people wi h in e media e educa ion
a e 6% poin s mo e likely o ecei e a accina ion han people
wi h basic educa ion. Acco ding o Lindhol e al. (2021), wi hin
Eu ope, people wi h uni e si y educa ion a e 2% poin s mo e
willing o ecei e a accina ion. Bo ga e al. (2022) es ima e ha
in an EU‐wide sample, people wi h uni e si y educa ion a e
3.7% poin s less like o be accina ion hesi an . Ou OLS es i-
ma e o a 0.7% poin inc eased accina ion a e he e o e seems
qui e low in compa ison, howe e , i needs o be kep in mind,
ha i is based on yea s o schooling. Gi en ha Ge man in-
di iduals who quali y o a end a uni e si y ha e ou mo e
yea s o schooling han indi iduals wi h basic ack educa ion,
his would ansla e in o 0.7 imes 4 =2.8 mo e pe cen age
poin s o be accina ed. This is a he lowe end o he ci ed
numbe s bu no comple ely ou o ange.
The easonwhy ou es ima e is—in ou in e p e a ion— ela i ely
small is likely ha ou sample is es ic ed o olde indi iduals,
who a e concen a ed a he lowe ma gin o educa ional a ain-
men . Many we e bo n be o e he Ge man educa ional expansion
ook o in he 1950s and 60s and mos o he indi iduals in ou
sample ha e basic ack educa ion only. Thus, he po en ial
d i e s o a posi i e co ela ion—pa icula ly indi iduals wi h
e ia y educa ion a e mo e likely o ge accina ed—p obably
play a mino ole in ou da a se .
6
Nex , we u n o he ins umen al a iables es ima ion, p esen ed
in Column (2). The i s ow p esen s he es ima e o he e ec o
he ins umen on he yea s o schooling om he i s s age
eg ession ( he comple e i s s age esul s a e p o ided in
Table A2 in he Appendix). We use a bina y indica o ha is equal
o one i an indi idual is bo n in a s a e‐yea coho , o which
9 yea s o schooling we e compulso y. We ind a s ong posi i e
e ec o compulso y schooling on educa ional a ainmen . The
inc ease in compulso y schooling hus inc eased educa ional
a ainmen by 0.443 yea s. The es ima ed coe icien s in he p e-
ious li e a u e ange om 0.19 (Pischke and on Wach e 2008)
o mo e han 1 (Kamhö e and Schmi z 2016). Ou es ima e is
he e o e on he lowe end o i s s age coe icien s, mo e in line
wi h Pischke and on Wach e (2008), Kemp ne , Jü ges, and
Reinhold (2011), o Bege ow and Jü ges (2022) (wi h es ima es
anging om 0.5 o 0.69).
The lowe panel o Column (2) in Table 3 epo s he es ima es
o he second s age. We ind no signi ican e ec o he yea s o
schooling on accina ion s a us. While his is also a ma e o
inc eased s anda d e o s, he poin es ima e is ze o.
TABLE 3 |Reg essions esul s o he e ec o educa ion on COVID
accina ion.
OLS 2SLS
(1) (2)
Fi s s age coe icien o he ins umen
Pos e o m 0.443***
(0.072)
Second s age
Yea s o schooling 0.007*** 0.000
(0.002) (0.029)
Male 0.015** 0.018
(0.007) (0.012)
Casa W2 0.002 0.004
(0.010) (0.014)
Casa W3 0.056*** 0.058***
(0.009) (0.013)
Bi h coho FE Yes Yes
Fede al s a e FE Yes Yes
S a e‐speci ic linea coho ends Yes Yes
Obse a ions 6010
No e: S anda d e o s in pa en heses, clus e ed on bi h‐coho ×s a e le el.
The i s s age includes he same con ol a iables as he second s age.
Comple e i s s age esul s a e epo ed in Table A2 in he Appendix.
*p<0.1, **p<0.05, ***p<0.01.
648 o 812 Heal h Economics, 2025
4.2
|
Robus ness Checks
In ca ying ou he eg essions, he esea che has many deg ees
o eedom. This holds o he sample selec ion o ypes o
con ol a iables included. We un u he eg essions simila o
he one p esen ed in Sec ion 4.1 o es he obus ness o ou
es ima e o hese choices. The esul s a e p esen ed in Figu e 4.
In he g aphs, he do s ep esen he es ima e o he e ec o
yea s o schooling on accina ion s a us de i ed om he 2SLS
es ima ion, while he lines ep esen he 95% con idence in e -
al. We include he same co a ia es as in Sec ion 4.1 unless
o he wise speci ied.
We i s es o obus ness agains di e en sample selec ions.
We epea ou main speci ica ion using coho s ha a e bo n up
o 7 yea s be o e o a e he pi o al coho . This is hen
compa ed o changing he bandwid h o i e and 10 yea s
a ound he pi o al coho . Nex , we p esen esul s whe e he
same bi h coho s a e used o all s a es and include he bi h
coho s om 1945 o 1960 and bi h yea s 1940–1970. In ano he
obus ness check, we d op indi iduals who s a e o wo k in he
heal h ca e sec o . This in o ma ion is p o ided in he Casa
da a, he SOEP and pa ly ia he ISCO codes in SHARE. Gi en
ha mos o he indi iduals in he sample a e ou o he labo
o ce, his does no make a big di e ence.
We also y di e en speci ica ions. He e, we use ou main sample
selec ion, using he bi h coho s 7 yea s a ound he pi o al
coho . In he i s speci ica ion we include age ixed e ec s
ins ead o bi h coho ixed e ec s. In he ollowing ow, we
exclude he s a e‐speci ic linea bi h coho ends. We clus e
s anda d e o s on he s a e‐le el ins ead o he bi h‐coho ×
s a e le el in he hi d speci ica ion. Nex , we include an indica o
o coho s ha expe ienced a sho school yea . I migh be a
conce n ha local in ec ions dynamics in luence indi iduals'
decision o ecei e he accina ion. Fo his pu pose we include
he agg ega ed numbe o new COVID in ec ions pe s a e and in
he mon h o he in e iew in he nex speci ica ion. Las ly, we
include he poin es ima e o he educed o m, ha is, including a
dummy whe he an indi idual aced eigh o 9 yea s o compul-
so y schooling ins ead o he yea s o schooling in an OLS
es ima ion.
While he es ima ed coe icien s luc ua e a ound ze o o he
di e en speci ica ions, none o hem is close o being s a is i-
cally signi ican a he 5% le el. Taken oge he , he addi ional
eg essions lea e he imp ession o a obus inding o basically
no sizable e ec o an addi ional yea o compulso y schooling
on accina ion agains COVID. Ne e heless, a small sample
size and, hus, po en ial p oblems o s a is ical powe a e
impo an d awbacks o ou analysis. Thus, in he nex sec ion,
we add ess he issue o s a is ical powe o ou analysis.
4.3
|
An Ex‐Pos Powe Analysis
The poin es ima e in ou main speci ica ion is ze o bu he con-
idence in e als a e ai ly la ge. Gi en ha we ha e a compa ably
small sample, his aises he ques ion whe he he e migh be a
la ge e ec ha we jus ail o iden i y. To ge an es ima e o he
minimum de ec able e ec size in ou se ing we apply a simu-
la ed ex‐pos powe analysis de eloped by Black e al. (2022). The
main idea is o assign a andom ea men s a us wi h a known
ea men e ec o a sample o un ea ed indi iduals, un a
eg ession and hen see whe he we can ge a s a is ically signi -
ican es ima e o his e ec . Thesmalles imposed ea men e ec
de ec able a he 5% signi icance le el in 80% o he imes is called
he minimum de ec able e ec (MDE).
He e, we un he educed om eg ession. Tha is, Equa ion (1)
whe e yea s o educa ion a e eplaced by he ( andomized) in-
s umen . We do his because, a e simula ing he ins umen
(being a ec ed by he e o m), we do no need o make u he
assump ions on how indi iduals eac o he e o m, ha is,
FIGURE 4 |Robus ness checks and e ec he e ogenei y. CASA, SHARE and SOEP da a. Each do is a eg ession coe icien om 2SLS eg essions
as be o e wi h he same co a ia es as be o e. Lines indica e 95% con idence in e als.
649 o 812
whe he hey a e complie s o always ake s (Schmi z and
Tawiah 2023). To achie e he MDE, we apply he ollowing s eps
7
1. We base his analysis on ou main sample (N=6010).
F om his, we only ake he un ea ed indi iduals, ha is,
indi iduals bo n be o e he pi o al coho in hei espec-
i e s a e. We do his o p e en con usion o he imposed
ea men e ec wi h po en ial ac ual ea men e ec .
Gi en ha 55% o indi iduals in ou sample a e bo n in o
a e he pi o al coho , his lea es us wi h 2704
obse a ions.
2. Nex , we e ill he sample o ge 6010 obse a ions. We do
his by o e sampling un ea ed indi iduals such ha each
ede al s a e has he same numbe o obse a ions as
be o e.
3. We hen andomly assign a ea men s a us— ha is, be-
ing subjec o he compulso y schooling e o m— o 55% o
indi iduals and de ine a cons an ea men e ec o 0.001.
4. We es ima e he educed o m e ec in ou model as
desc ibed abo e.
5. We hen epea he en i e p ocedu e o e 40 ounds whe e
we s epwise inc ease he ea men e ec by 0.001 e e y
ound o a maximum o 0.04.
6. These s eps a e epea ed 1000 imes pe ound.
This lea es us wi h 40,000 es ima ions o which we only conside
whe he he poin es ima e o he ea men e ec is signi ican ly
di e en om ze o. The ques ion o each i e a ion is i we a e
able o (co ec ly) ejec he null‐hypo hesis ha he es ima ed
coe icien is ze o, using he h ee common alpha alues 10%, 5%
and 1%. The esul s a e depic ed in Figu e 5whe e he sha e o
ejec ed null‐hypo heses o each ue ea men e ec is depic-
ed. The MDE is—somewha a bi a ily bu ollowing Black
e al. (2022)—de ined by he x‐ alue o he poin , whe e he
alpha =5% line passes he ejec ion a e o 80%. In ou case his is
0.008. Thus, ou app oach should be able o de ec educed‐ o m
ea men e ec s o 0.8% poin s mos o he ime. O , combining i
wi h he i s s age, IV e ec s o abou 1.6% poin s. Thus, we
conclude ha ou da a would also allow o iden i y e ec s ha
a e no e y la ge.
4.4
|
E en ‐S udy Speci ica ion o Accoun o
Possible T ea men E ec He e ogenei y
The implemen a ion o compulso y schooling e o ms occu ed
a di e en poin s in ime and in di e en s a es. This aises
conce ns spelled ou in he ecen li e a u e on di e ence‐in‐
di e ences wi h s agge ed ea men iming whe e ea men
e ec he e ogenei y may cause biased es ima es (e.g., Goodman‐
Bacon 2021; Sun and Ab aham 2021). A possible p oblem in ou
applica ion is ha he ede al s a es a e esponsible o educa-
ion policies in Ge many, especially including he con en s o
he cu icula, which could be one o he main d i e s o po-
en ial ea men e ec s. To ge an idea whe he hese p oblems
migh be ele an in ou ca e, we u n o an e en s udy es i-
ma ion o he educed o m ( he e ec o he compulso y
schooling e o m on accina ion) and use he es ima o p o-
posed by Sun and Ab aham (2021) ha akes po en ial p oblems
in o accoun .
8
We use he ela i e bi h yea as he e en ime. I is de ined as
eics =Bi h yea ics −Pi o al coho ics. The bi h yea be o e he
pi o al coho (eics =−1)is used as he e e ence pe iod. The
esul ing e en s udy model akes he ollowing o m whe e he
con ol a iables in addi ion o he e en ime indica o s a e
hose also included in Equa ion (1):
Yics =∑
j≥−4;
j≠−1
δj1(eics =j)+δ−51(eics ≤ −5)+β2malei+γc+δs
+α(ηs×c)+θsu ey +εics
(2)
He e, δ−5is he combined es ima e o all e en imes smalle
han −4 in o de o sepa a e e en ime om calenda yea e -
ec s. The emaining δja e es ima es o he e ec o being
subjec o he compulso y schooling e o m j o he bi h coho
jyea s away om he pi o al coho . He e, we do no accoun
o yea s o schooling bu only o he age ela i e o he pi o al
coho . We es ima e he coe icien s o Equa ion (2) using he
p ocedu e sugges ed by Sun and Ab aham (2021).
The esul s a e p esen ed in Figu e 6. No p e‐ ends a e isible,
and he pos ‐ e o m e ec s do no show a con incing pa e n
FIGURE 5 |Powe analysis. Based on sample o un ea ed indi iduals.
650 o 812 Heal h Economics, 2025
