Höckel, Lisa So ie
A icle — Published Ve sion
Language lesson lea ned— o eign-o igin eache s and
hei e ec on s uden s’ language skills
Jou nal o Popula ion Economics
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Höckel, Lisa So ie (2024) : Language lesson lea ned— o eign-o igin eache s and
hei e ec on s uden s’ language skills, Jou nal o Popula ion Economics, ISSN 1432-1475, Sp inge ,
Be lin, Heidelbe g, Vol. 37, Iss. 2,
h ps://doi.o g/10.1007/s00148-024-01019-6
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/313837
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
h ps://doi.o g/10.1007/s00148-024-01019-6
ORIGINAL PAPER
Language lesson lea ned— o eign-o igin eache s
and hei e ec on s uden s’ language skills
Lisa So ie Höckel1
Recei ed: 17 Ap il 2023 / Accep ed: 21 Feb ua y 2024 / Published online: 15 Ap il 2024
© The Au ho (s) 2024
Abs ac
In e na ional mig a ion inc eases class oom di e si y a ound he wo ld, bu li le is
known abou he e ec o o eign-o igin eache s on s uden s’ academic achie emen .
This s udy in es iga es whe he o eign-o igin eache s causally a ec hei s uden s’
academic pe o mance. Exploi ing wi hin-s uden a ia ion in assignmen o eache s
in Ge many, I ind ha eache s who a e immig an s o descendan s o immig an s
signi ican ly inc ease he eading comp ehension o hei s uden s in seconda y school,
bu do no a ec hei ma h skills. This s udy is he i s o in es iga e bilingualism as
a po en ial mechanism and shows ha he e ec on eading comp ehension is d i en
by bilingual o eign-o igin eache s. Gi en hei own expe ience in language lea ning,
hey seem excep ionally well-equipped o each languages. This s udy con ibu es o
he scan e idence on he causal ela ionship be ween eache s’ o eign o igin and
s uden s’ academic achie emen in ligh o a la ge and pe sis en achie emen gap
be ween na i e and o eign-o igin s uden s.
Keywo ds Bilingualism ·Educa ion ·Language skills ·Mig a ion ·Role model
e ec ·Teache bias
1 In oduc ion
In e na ional mig a ion does no only a ec he labo o ce, i also a ec s i s ain-
ing g ound— he class oom. In OECD coun ies, mo e han 25% o 15–34-yea -olds
epo a o eigno igin,and hesha eo o eign-o igins uden scon inues o ise(OECD
2018). La ge and pe sis en achie emen gaps be ween na i e and o eign-o igin s u-
den s, meaning s uden s who a e ei he immig an s o descendan s o immig an s,
Responsible edi o : Milena Nikolo a
BLisa So ie Höckel
[email p o ec ed]
1RWI – Leibniz Ins i u e o Economic Resea ch, Hohenzolle ns asse 1-3, 45128 Essen, Ge many
123
Jou nal o Popula ion Economics (2024) 37:45
L.S. Höckel
p e ail inmos coun ies(e.g.,Schnep 2007;Algane al.2010;Giannelliand Rapallini
2016; Ruhose and Schwe d 2016; Alie a e al. 2018; OECD 2018), and he li e a u e
has iden i ied language as he single mos impo an de e minan associa ed wi h he
educa ional achie emen gap (e.g., Dus mann e al. 2010,2012; Geay e al. 2013;
Danze e al. 2022).
While he sha e o o eign-o igin s uden s con inues o inc ease, o eign-o igin
eache s emain se e ely unde ep esen ed in many OECD coun ies. This unde -
ep esen a ionis especially highin coun ies wi ha ela i elyla ge mig an popula ion.
In he US, unde 10% o all eache s a e Hispanic al hough Hispanic s uden s make
up abou one-qua e o all K-12 s uden s (U ban Ins i u e 2017). In Ge many, 33% o
unde -15-yea -old s uden s ha e a o eign o igin, bu his is only ue o 8% o p ima y
and seconda y school eache s (S a is isches Bundesam 2012,2018).Thesha eo
o eign-o igin eache s is also less han hal ha o he s uden popula ion in coun ies
such as he UK, I aly, o Denma k (Eu opean Commission 2016).
The Eu opean Commission ecommends i s membe s a es adop policies o
inc ease eache di e si y (Eu opean Commission 2016), bu he e is li le empi ical
e idence on he e ec o o eign-o igin eache s on s uden s’ academic achie emen s.
A e o eign-o igin eache s be e equipped o help o eign-o igin child en o e come
po en iallanguageba ie s and socio-economicdisad an ages?Andhow do heya ec
he achie emen o na i e s uden s? Fo he US, he li e a u e shows ha mino i y s u-
den s bene i om same- ace eache s (e.g., Dee 2004; Fai lie e al. 2014; Ge shenson
e al. 2016), bu i is unclea i hese indings can be ans e ed o he con ex o
mig a ion.
In his pape , I in es iga e whe he ha ing a o eign-o igin eache causally a ec s
s uden s’ eading comp ehension sco es in lowe -seconda y school, holding cons an
bo h obse ed and unobse ed ac o s ela ed o academic ou comes. I de ine o eign-
o igin eache s as o eign-bo n eache s, eache s wi h a leas one o eign-bo n pa en ,
o eache s who epo a mo he ongue o he han Ge man. Using da a om he
Ge man Na ional Educa ional Panel S udy (NEPS) wi h unique in o ma ion on eache
cha ac e is ics and exploi ing wi hin-s uden a ia ion in assignmen o eache s o e
ime, I ind ha o eign-o igin eache s inc ease objec i e eading comp ehension
sco es o hei s uden s bu do no a ec hei s uden s’ ma h es sco es. Impo an ly,
he posi i e e ec on language skills is d i en by o eign-o igin eache s who epo
a mo he ongue o he han Ge man. Ruling ou al e na i e explana ions, I a gue ha
bilingual eache s a e pa icula ly well-equipped in ansmi ing language skills.
This s udy con ibu es o he li e a u e in h ee ways. Fi s , i adds o he scan
e idence on he causal ela ionship be ween eache s’ o eign o igin and s uden s’
academic achie emen . Gi en ha second-gene a ion immig an s uden s a e on a e -
age he as es -g owing g oup ac oss OECD coun ies (OECD 2018), i is o pa icula
policy ele ance o s udy wha ole o eign-o igin eache s can play in hei school
in eg a ion. Second, his s udy is he i s o in es iga e eache s’ language skills due
o bilingualism as a po en ially mig a ion- ela ed channel. While he economic li e -
a u e has ocused mainly on ma ching e ec s h ough ole model and eache bias
e ec s, his pape no only in oduces a di e en mechanism bu also es s o all h ee
channels in he same se ing. Thi d, he panel da a employed allows o in es iga ing
s uden a ainmen wi hin he same subjec o e ime. Following he eache alue-
123
45 Page 2 o 32
Language lesson lea ned— o eign-o igin eache s
added li e a u e, I can es ima e he gain in s uden achie emen a e being augh by
a o eign-o igin eache using a ia ion in eache assignmen o e ime. This s a egy
adds o p e ious s udies (e.g., Seah 2018,2021), which ha e la gely elied on he
Na ional Educa ion Longi udinal S udy o 1988 (NELS) and used a ia ion in eache
assignmen ac oss subjec s.
Theo e ically, he e a e h ee easons why o eign-o igin eache s can a ec s u-
den s’ academic achie emen di e en ly han na i e eache s. Fi s , o eign-o igin
eache s can ha e di e en language skills compa ed o na i e eache s. On he one
hand, hey can be less p o icien in he language o ins uc ion. Fo eign-bo n each-
e s a e mo e likely o ha e accen s ha hampe s uden s’ unde s anding and hus
make he cou se con en less accessible. On he o he hand, hey can use hei own
expe ience in language lea ning o each hei s uden s. Acco ding o he “conscious
compe ence” lea ning model, o eign-o igin eache s who ha e consciously lea ned a
language could be be e a explaining i in compa ison o na i e eache s who use he
language la gely ins inc i ely (Robinson 1974).
Second, s uden - eache ma ching can in luence s uden s’ academic achie emen .
Ma ching e ec s comp ise wo complemen a y channels ha can make demog aphic
ma ching o s uden s and eache s pa icula ly ad an ageous o ma ched s uden s. A
ole model e ec desc ibes a posi i e eac ion o o eign-o igin s uden s o o eign-
o igin eache s. T igge ed by he eache ’s p esence, a he han an explici beha io ,
o eign-o igin s uden s’ belie s abou hei educa ional possibili ies can imp o e, mak-
ing hem mo e con iden and engaged in class. A eache bias e ec desc ibes eache s’
beha io . Fo eign-o igin eache s could ac i ely o unconsciously alloca e mo e class
ime o in e ac ing wi h s uden s o he same o igin o hey can be less likely o dis-
c imina e agains o eign-o igin s uden s in hei g ading (Dee 2007). A eache bias
e ec could also eme ge om di e en eaching cul u es (Fleishe e al. 2002)o a
lack o hos -coun y-speci ic human capi al o o eign-o igin eache s due o cul u al
di e ences. Fo eign-o igin eache s could hen be less able o p o ide cul u ally ele-
an examples o na i e s uden s bu se e as be e communica o s o o eign-o igin
s uden s.
Thi d, o eign-o igin and na i e eache s could a y in he e o and ime hey
alloca e owa ds hei eaching ac i i ies. I hey sys ema ically di e in e ms o hei
in insic mo i a ion o cul u al alues ega ding hei p o ession, one o he g oups
could, o example, wo k longe hou s o pu mo e e o in o hei eaching ac i i ies
(e.g., spend mo e ime men o ing hei s uden s).
Gi en hese mechanisms, he e ec o ha ing a o eign-o igin eache on s uden s’
academic achie emen is heo e ically ambiguous.
To in es iga e a o eign-o igin eache e ec , his s udy builds upon di e en li -
e a u e s ands. A small bu g owing s and o he li e a u e discusses he language
skills o eache s. Ea ly s udies analyze he e ec o o eign eaching assis an s on he
academic achie emen s o unde g adua e s uden s in uni e si y (Bo jas 2000; Asano
2008). The con adic o y e ec s ound by hese s udies can be explained by he non-
andom assignmen o eaching assis an s o s uden s. In he s udy mos closely ela ed
o his pape , Seah (2021) examines he e ec o ha ing a linguis ically simila eache
on he academic achie emen s o seconda y school s uden s in he Uni ed S a es. Using
da a om he NELS, he exploi s wi hin-s uden a ia ion in es sco es and he na i e
123
Page3o 32 45
L.S. Höckel
language o eache s ac oss wo subjec s. He inds no e ec o being assigned o a
linguis ically simila eache once he eache ’s e hnici y is con olled o .
A la ge li e a u e s and s udies ma ching e ec s wi h espec o demog aphic
cha ac e is ics, mos ly gende and ace (e.g., Dee 2004; Be inge and Long 2005;
Ho mann and O eopoulos 2009; an Ewijk 2011;Cho2012; Fai lie e al. 2014;
An ecol e al. 2015; G i i h and Main 2021). Dee (2004) examines es sco es om he
P ojec STAR class-size expe imen , which andomly ma ches s uden s and eache s
wi hin pa icipa ing schools. He shows ha assignmen o own- ace eache s signi -
ican ly inc eases Ma h and eading achie emen o bo h black and whi e s uden s.
Fai lie e al. (2014) and Egali e e al. (2015) con i m e hnici y and ace-ma ching
e ec s using la ge adminis a i e da a o US high schools and communi y colleges.
Mo e ecen ly, Seah (2018) in es iga es he e ec o immig an eache s on 8 h g ade s
in he US and inds no ad e se e ec o immig an eache s on he achie emen o
(na i e) s uden s.
Las ly, he pape builds upon he li e a u e showing ha immig an s and na i es
di e in e ms o hou s wo ked (Hame mesh and T ejo 2013;Va gas2016; Fe ig
2010). S udies by Hame mesh and T ejo (2013) and Coniglio e al. (2021) show ha
o eign-bo n indi iduals in es mo e ime in educa ional ac i i ies.
The emainde o he s udy is o ganized as ollows. Sec ion2gi es a sho ins i u-
ional o e iew o he Ge man school sys em and he ole o immig an eache s in i .
Sec ion3discusses he da a, Sec ion 4in oduces he empi ical s a egy, and Sec ion 5
p esen s he indings. Sec ion6concludes.
2 Ins i u ional backg ound
A key ea u e o he Ge man educa ion sys em is ha s uden s a e ypically acked
a e 4 yea s o elemen a y schooling.1S uden s a e so ed based on hei aca-
demic abili y and assigned o one o h ee seconda y school acks: lowe -seconda y
ack (Haup schule), middle-seconda y ack (Realschule), and uppe -seconda y ack
(Gymnasium).2Haup schule p o ides p ac ical educa ion and p epa es s uden s o
oca ional educa ion (un il g ade 9); Realschule has a b oade ange o emphasis o
in e media e s uden s (un il g ade 10); and Gymnasium quali ies s uden s o highe
educa ion (un il g ade 12 o 13).3Depending on he ede al s a e, he ack is de e -
mined by pa en al choice o a binding eache ecommenda ion based on he s uden s’
academic achie emen and abili y o wo k independen ly. Schooling is compulso y
o 9 o 10 yea s, depending on he ede al s a e. S uden s usually inish he ack hey
ha e been assigned o, bu swi ching acks is possible and became mo e common in
ecen yea s.4
1In he ede al s a es o Be lin and B andenbu g, s uden s a e acked a e 6 yea s o schooling.
2Addi ionally, he e a e comp ehensi e schools (Gesam schulen) ha combine all educa ion ypes and
amoun o 12% o all Ge man seconda y schools (Malecki e al. 2014).
3See Fig. A1 o an illus a ion o he Ge man educa ion sys em.
4School changes wi hin he same ack only accoun o app oxima ely 14% o school changes in Ge many
(Kliche and Täubig 2019).
123
45 Page 4 o 32
Language lesson lea ned— o eign-o igin eache s
A second ele an ea u e o he Ge man educa ion sys em is ha eaching is o ga-
nized in classes a he han cou ses. Mo e speci ically, a class e e s o a g oup o up
o 30 s uden s who a e alloca ed by hei school’s headmas e o he same class oom
upon en y in o seconda y school. All s uden s in one class sha e he same eache
o a gi en subjec . In con as o he US, s uden s do no ake di e en cou ses in he
same subjec based on hei p o iciency.5Ins ead, class composi ion may change when
s uden s choose di e en elec o al cou ses in g ade 7 o 9. Teache s do no specialize
in eaching a pa icula g ade bu he school’s headmas e assigns hem o classes on
a yea ly basis.
Mos Ge man eache s a e g adua es o a o mal eaching educa ion p og am
(Leh am s udium). Condi ional on ha ing ea ned a deg ee ha quali ies o e ia y
educa ion, eache aining o seconda y educa ion comp ises wo componen s. Fi s ,
eache candida es comple e 4 o 6 yea s o uni e si y cou ses co e ing he wo sub-
jec s hey la e wan o each in combina ion wi h pedagogical aining. A he end o
he i s phase, he candida es ake exams on pedagogic and heo e ical knowledge o
he subjec s s udied. In addi ion o he g ades ea ned a he uni e si y, hese exams
comp ise he i s s a e examina ion g ade. Second, candida es pa icipa e o 18–24
mon hs in a p ac ical p og am o eaching semina s (Re e enda ia ) a a eaching ain-
ing school. Du ing his phase, candida es ha e eaching posi ions, comple e a hesis,
and deli e demons a ion lessons a ed by head eache s. The combina ion o assess-
men s o he demons a ion lessons, he hesis g ade, and exams sum up o he second
s a e examina ion.
Thesecond s a e examina ioniscompulso y o en y in o ci ilse ice.6Ul ima ely,
he g ade o he second s a e examina ion, in combina ion wi h he local demand o
eache s, de e mines he school a eache is assigned o. Fede al s a es can hi e eaching
candida es wi hou he second s a e examina ion on egula sala ied posi ions wi hou
awa ding hem ci il se an s a us. Fo subjec s wi h eache sca ci y, e en eache s
wi hou a o mal eache aining a e eligible (Que eins eige ).
Immig an eache s o candida es who ha e s udied ab oad do no ha e o eside in
Ge many o a ce ain pe iod be o e becoming eache s bu hey ha e o exhibi a com-
pa able eaching deg ee o ob ain he second s a e examina ion in Ge many o become
ci il se an s. While he ecogni ion o o eign deg ees is easie o candida es om
wi hin he Eu opean Union, immig an eache s ypically ace signi ican obs acles
du ing he ecogni ion p ocess. Acco ding o GEW (2021), only abou 15% o o eign
eaching deg ees a e success ully g an ed wi hou u he equi emen s. A ound 68%
o ecogni ion eques s a e g an ed condi ional on addi ional aining and hus imply
high cos s o immig an candida es. The condi ioning is mainly due o he ac ha
he Ge man eaching deg ee consis s o wo di e en subjec s while in mos coun ies
one subjec is su icien . Ano he hu dle o becoming a ci il se an in Ge many is
he high Ge man language s anda d equi ed. While he ac ual p e equisi es di e by
ede al s a e, immig an s ypically ha e o speak Ge man a C1 le el i espec i e o
he subjec s udied. In compa ison o o he occupa ions, he ecogni ion p ocess o
5Ge man language and Ma h classes a e compulso y in Ge man schools. S uden s a e obliged o ake hem
and canno choose cou ses based on hei p e e ences o eache s.
6In mos ede al s a es, eache s become ci il se an s. Ne e heless, app oxima ely one in ou eache s
is hi ed as an employee a he han a ci il se an .
123
Page5o 32 45
L.S. Höckel
becoming a eache he e o e leads o a a he posi i ely selec ed g oup o immig an
eache s who ac ually end up wo king in hei job.
3 Da a and desc ip i e s a is ics
This s udy uses da a om he Ge man Na ional Educa ional Panel S udy (NEPS)
(NEPS Ne wo k 2021). The NEPS has a mul i-coho design and d aws om a ep-
esen a i e sampling ame o s uden s om six s a ing coho s. The NEPS ollows
s uden s as hey mo e h ough he educa ion sys em and con ains ex ensi e ques ion-
nai es answe ed by pe sons in he s uden s’ pe sonal en i onmen , such as pa en s,
eache s, and headmas e s.
Fo he empi ical analysis, I employ da a om s a ing coho 3 (SC3) as i p o ides
unique in o ma ion on eache s’ o igin o Ge man and Ma h eache s. SC3 ollows
s uden s om g ades 5 o 9, an age coho ha is pa icula sui able o he esea ch
ques ion, as he i s nine yea s o educa ion a e compulso y and c ucial o li elong
educa ion ou comes (Ang is and K uege 1991). The sampling popula ion o SC3
con ains all Ge man i h g ade s in schools o e ing lowe -seconda y educa ion in he
school yea 2010/2011. Fi s , schools a e andomly d awn om he popula ion o pub-
lic schools o be ep esen a i e by school ype.7Second, classes a e andomly selec ed
wi hin each school (see S einhaue and Zinn 2016 o sampling design). Pa icipa ion
is olun a y and implies ha s uden s comple e compe ency es s and answe a socio-
economic ques ionnai e. O e all, he pa icipa ion a e o s uden s is e y high wi h
94.5% in he i s wa e. The eache ques ionnai e con ains in o ma ion on eache s’
demog aphics and aspec s o hei ca ee choice and s udies. I dis ega d in e iew da a
on pa en s and headmas e s o minimize sample a i ion.
Ino de ode e mine hes uden s’ o eigno igin,Icombine h ee a iablesp o ided
by he NEPS da a se . (a) S uden s epo which ci izenship hey hold. I hey men ion
any (addi ional) na ionali y o he han Ge man, I code hem as o eign o igin. (b)
In e iewe s eco d s uden s’ coun y o o igin. I code non-Ge man o igins as o eign
o igin. (c) S uden s a e asked i hey ha e a Russian o Tu kish mig a ion backg ound.
I hey men ion one o he wo mig a ion backg ounds, I code hem as o eign o igin.8
To de e mine Ge man language eache s’ o eign o igin, I use wo a iables. (a)
Teache s a e asked abou hei mig a ion backg ound, namely i hey a e o eign bo n
o i hey ha e a leas one o eign-bo n pa en . (b) Teache s epo hei mo he ongue,
meaning he language hey lea ned as a child in he amily. I hey men ion a mo he
ongue addi ional o Ge man, o o he han Ge man, I code hem as o eign o igin. I
eache s do no epo a mig a ion backg ound o mo he ongue, I code hem as na i e
Ge mans.
Besides con aining unique in o ma ion on eache o igin, he NEPS has he ad an-
age o p o iding objec i e and unidimensional compe ence sco es o s uden s’
pe o mance in di e en subjec s conduc ed by NEPS in e iewe s. This ea u e is
7S uden s om schools wi h a p edominan o eign eaching language and s uden s who a e no able o
pa icipa e in he no mal es ing p ocedu e a e excluded.
8This ques ion does no bias he sample owa ds s uden s o Russian and Tu kish backg ound. I only
iden i ies an addi ional 6% o o eign-o igin s uden s.
123
45 Page 6 o 32
Language lesson lea ned— o eign-o igin eache s
c ucial in a se ing whe e eache bias e ec s migh be a wo k as Ge man language
eache s canno manipula e he NEPS compe ence sco es. Fo he empi ical analysis, I
ocuson eading comp ehensionas he mainou come a iable.Thesco eis designed o
measu e he abili y o unde s and and use w i en ex s, which a e impo an p econdi-
ions o de elop pe sonal skills and pa icipa e in social li e and he labo o ce (Geh e
e al. 2012). Reading comp ehension is assessed by mul iple-choice ques ionnai es,
which es he unde s anding o i e ex unc ions and associa ed ex ypes, namely
in o ma ional, commen ing, ins uc ional, ad e ising, and li e a y ex s. The eading
compe ence es las s 28min pe ex unc ion and i s he hema ic o ien a ion, lexical,
seman ic, and g amma ical p ope ies o he speci ic age coho (Geh e e al. 2012).
The answe s o he mul iple choice ques ions a e agg ega ed by a weigh ed maximum
likelihood es ima ion and cons ained o ha ing a mean o ze o in he i s wa e. This
s anda diza ion ensu es ha sco es a e compa able ac oss di e en su ey wa es. As
p omo ing eading comp ehension is one o he key objec i es o Ge man language
classes, I can a ibu e his skill o he domain o he Ge man language eache . O e all,
eading comp ehension skills can se e as a sui able p oxy o language skills (e.g., in
pa icula wi h espec o w i en language) as a subs an ial connec ion be ween ead-
ing comp ehension and bo h ecep i e and exp essi e ocabula y has been es ablished
in he li e a u e (Be endes e al. 2013).
I impose se e al es ic ions on he da a. F om he ini ially 6527 s uden s, I obse e
6485 in he yea s when he es s ake place. O hose, I d op 358 s uden s because hey
do no pa icipa e in he compe ence es . I can link 5758 s uden s o hei Ge man
language eache , and I ha e con ex da a o 718 Ge man language eache s o 4724
s uden s. This migh lead o a small bias owa ds be e (o ganized) schools.
Fu he , he da a do no allow o consis en acking o s uden s who change classes
and schools—a p ac ice which is a e in he Ge man school sys em.9I also canno
link s uden s i schools wi hd ew hei pa icipa ion om he NEPS, and i eache s
o s uden s do no answe he ques ionnai e.10 O he 4724 s uden s in my sample,
one- hi d o s uden s a e only obse ed once.
Toillus a e heda a, he i s wocolumnso Table1p esen desc ip i es a is ics o
all s uden s obse ed. The i s a iable desc ibes he main ou come a iable, i.e., he
objec i e eading comp ehension sco es. The a e age o 0.29 indica es ha he eading
comp ehension in he sample has inc eased o e ime.11 Ou o he 4724 s uden s in
he sample, 36% a e o o eign o igin. This sha e is in line wi h s a is ics o he Ge man
Mic ocensus, which es ima e a sha e o (na owly de ined) o eign-o igin s uden s o
33% in 2012 (S a is isches Bundesam 2012). The mos common coun ies o o igin in
he da a a e Tu key (16%), Russia (14%), and Poland (12%), which oge he accoun
9I pa en s wi h s ong an i-immig an sen imen s a e mo e likely o swi ch hei child en spon aneously
ou o class once hey a e in o med abou he o eign o igin o he eache a he beginning o he school
yea , and hei child en a e less likely o be ecep i e o he eaching con en o o eign-o igin eache s, he
o eign-o igin eache e ec migh be upwa ds biased. Gi en ha swi ches a e a e, I expec his conce n
o be o no s a is ical ele ance.
10 To a oid u he sample a i ion, I keep eache obse a ions wi h missing in o ma ion i he eache
answe s a leas wo ques ions used o he con ol a iables.
11 Fu he , he sample is sligh ly skewed owa ds highe achie ing s uden s as he a e age eading com-
p ehension in g ade 5 is 0.07 a he han 0.
123
Page7o 32 45
L.S. Höckel
Table 1 Desc ip i e s a is ics—s uden s
All Na i e Fo eign o igin
Mean S d. de . Mean S d. de . Mean S d. de . - es
Main ou come a iable
Reading comp ehension 0.285 1.29 0.411 1.29 0.058 1.27 ∗∗∗
S uden cha ac e is ics
Fo eign o igin 0.358 0.48
Female 0.494 0.50 0.486 0.50 0.509 0.50
Bi h yea 1999.502 0.61 1999.531 0.60 1999.449 0.64 ∗∗∗
G ade 6.647 1.57 6.685 1.57 6.578 1.56 ∗∗∗
School yea epea ed 0.025 0.16 0.023 0.15 0.028 0.16 ∗
Household size 4.439 1.38 4.392 1.31 4.523 1.49 ∗∗
O he ou come a iables
Ma hema ics sco e 0.304 1.23 0.453 1.21 0.036 1.22 ∗∗∗
Ge man g ade 4.375 0.83 4.445 0.83 4.247 0.82 ∗∗∗
Reading equency on a
school day
0.829 0.74 0.820 0.74 0.845 0.76
Teache expec s me o y
my e y bes
4.030 0.88 4.045 0.86 4.003 0.93 ∗
Sa is ac ion wi h school 7.276 2.39 7.334 2.33 7.173 2.48 ∗∗
Panel obse a ions 7351 4768 2583
S uden obse a ions 4724 3035 1689
No es: The desc ip i e s a is ics a e shown o he whole sample as well as o he wo subsamples o na i e
and o eign-o igin s uden s. Signi icance s a s indica e he esul o he espec i e - es o di e ences in
mean alues.
*p<0.10; **p<0.05; ***p<0.01
o a ound 42% o o eign-o igin s uden s. The sample is balanced wi h espec o
gende , and almos 95% o he s uden s a e bo n in 1999 o 2000. The sample con ains
obse a ions om g ades 5, 7, and 9, because eading comp ehension is es ed in hese
g ades.12 I obse e mos s uden s in i h g ade due o he 3% o s uden s who epea ed
a school yea and sample a i ion due o missing linkages. The a e age household size
is 4 and a ound 80% o s uden s li e in households o 3 o 5 people.
Besides eading comp ehension sco es and s uden cha ac e is ics, Table 1p o ides
in o ma ion on addi ional ou come a iables. I shows an a e age ma h sco e o 0.30,
indica ing ha he ma h es sco e also inc eases o e ime. The Ge man g ade co -
esponds o he subjec i e g ade gi en by he Ge man language eache a he end o
he school yea and i anges om one o six. He e, six co esponds o an ou s and-
ing achie emen while one (and wo) e e s o an insu icien achie emen . The mos
p e alen g ade is sa is ac o y (4), which 44% o s uden s ecei e. The NEPS da a also
con ain su ey i ems ha can be linked o he Ge man language eache ’s in luence.
12 The missing yea s a e no necessa ily a conce n, as eache s ypically change classes e e y wo yea s
implying ha a eache eaching a ce ain subjec in g ade 5 (7) is likely o each he same class in g ade 6
(8).
123
45 Page 8 o 32
Language lesson lea ned— o eign-o igin eache s
o o eign-o igin s uden s does no bias he esul s, I p o ide all esul s sepa a ely o
na i e and o eign-o igin s uden s. While he NEPS da ase con ains a la ge numbe
o eache cha ac e is ics, he empi ical s a egy does no allow o es o sys ema ical
di e ences in unobse ed cha ac e is ics be ween na i e and o eign-o igin eache s.
I he e o e add ess po en ial di e ences in e o as a mechanism in Sec ion 5.2.2.
5 Resul s
5.1 Fo eign-o igin eache e ec
I s a he empi ical analysis by es ing i o eign-o igin eache s a ec s uden s’ ead-
ing comp ehension. Panel A o Table 6displays he o eign-o igin eache e ec o
he ull sample. Panels B and C p o ide sepa a e es ima es o na i e and o eign-
o igin s uden s o in es iga e i he e a e di e en ial e ec s by s uden s’ o igin. As
we mo e along he columns, I inc easingly es ic he a ia ion used o iden i y he
pa ame e o in e es . The speci ica ion in column (1) con ols o s uden cha ac e is-
ics and class ixed e ec s. Following Eq. (1), i accoun s o pee e ec s and eache
alloca ion ac oss schools and classes. Column (2) includes eache cha ac e is ics, and
column (3) displays esul s om he p e e ed speci ica ion desc ibed in Eq. (2), which
con ains bo h s uden and class ixed e ec s.16
Fo he o e all sample, he e is a posi i e and signi ican co ela ion be ween ead-
ing comp ehension sco es and ha ing a o eign-o igin eache in column (1). The
o eign-o igin eache es ima e becomes la ge once i s e ec is disen angled om
eache cha ac e is ics in column (2). This inc ease in e ec size is mainly caused by
he inclusion o he eache g ade a he i s s a e exam con ol a iable as o eign-
o igin eache s pe o m signi ican ly wo se a his examina ion. In e es ingly, eache s
who s udied languages o he han Ge man also posi i ely a ec s uden s eading com-
p ehension (see Table A2). In column (3), he e ec is iden i ied by wi hin-s uden
a ia ion in eache alloca ion. This speci ica ion emo es co ela ions be ween being
augh by a o eign-o igin eache and changes in he class composi ion. Fu he , i con-
ols o di e ences in unobse ed s uden he e ogenei y. Consequen ly, less a ia ion
is employed and βis less p ecisely es ima ed. Ne e heless, he e ec s ays signi i-
can and he magni ude ises o 0.27. The coe icien o 0.27 amoun s o 0.2 s anda d
de ia ions in he eading comp ehension es sco e. The e ec size is compa able in
magni ude o he e ec ound by Seah (2018) o immig an eache s on science es
sco es o na i e s uden s.
Fo na i e s uden s, in panel B, he e ec size o ha ing a o eign-o igin eache is
simila in size in he speci ica ions including class and s uden ixed e ec s (columns
(2) and (3)). Fo o eign-o igin s uden s, in panel C, he co ela ion be ween s uden
unobse ed he e ogenei y and he eache ’s o igin ma e s. Wi h s uden ixed e ec s,
he e ec o ha ing a o eign-o igin eache doubles in size and becomes s a is ically
16 Fo ease o exposi ion, coe icien s o he con ol a iables a e p esen ed in Table A2. They a e in he
expec ed di ec ion.
123
Page15o 32 45
L.S. Höckel
Table 6 E ec o o eign-o igin eache on eading comp ehension
(1) (2) (3)
Panel A: all s uden s
Fo eign-o igin eache 0.191∗∗∗ 0.264∗∗∗ 0.270∗
(0.067) (0.073) (0.140)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.377 0.377 0.661
Obse a ions 7351 7351 7351
Panel B: na i e s uden s
Fo eign-o igin eache 0.286∗∗∗ 0.359∗∗∗ 0.342
(0.102) (0.107) (0.233)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.361 0.362 0.651
Obse a ions 4768 4768 4768
Panel C: o eign-o igin s uden s
Fo eign-o igin eache 0.116 0.151 0.366∗
(0.087) (0.095) (0.199)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.399 0.398 0.654
Obse a ions 2583 2583 2583
No es: Resul s a e ob ained om OLS eg essions. All epo ed s anda d e o s a e clus e ed a he eache
le el.
*p<0.10; **p<0.05; ***p<0.01
signi ican . This imp o emen in eading comp ehension can ha e impo an impli-
ca ions o li elong ou comes o o eign-o igin s uden s, such as p oduc i i y and
wages, because language luency acili a es he ans e and adap a ion o skills in
he job ma ke (Dus mann and an Soes 2001; Hay on 2001; Dus mann and Fabb i
2003; Cla ke and Sku e ud 2016; Lochmann e al. 2019; Foged e al. 2022; Lang
2022). O e all, in he p e e ed hi d speci ica ion, he o eign-o igin eache e ec is
la ge and mo e p ecisely es ima ed o o eign-o igin han na i e s uden s. Ne e he-
123
45 Page 16 o 32
Language lesson lea ned— o eign-o igin eache s
less, esul s ob ained om a ully in e ac ed model do no e eal signi ican ly di e en
e ec s o o eign-o igin eache s on s uden s wi h and wi hou a o eign o igin. This
inding is in line wi h Seah (2018), who shows a posi i e bu insigni ican co ela ion
be ween immig an eache s and bo h na i e and immig an s uden s o English es
sco es in he US. In con as o he li e a u e on ace-ma ching, he posi i e e ec
o o eign-o igin eache s is no caused by ma ching e ec s in he sense ha he e
a e posi i e e ec s on he ma ched g oup (he e: o eign-o igin s uden s) and ad e se
e ec s on he misma ched g oup (he e: na i e s uden s). Ins ead, Table 6indica es
posi i e e ec s o o eign-o igin s uden s and clea ly no nega i e e ec o na i e
s uden s. This esul is obus o he inclusion o s uden s’ ma h es sco es (see Table
A3). Ma h es sco es p oxy s uden -yea -speci ic pe o mance and acco dingly p o-
ide a lowe bound es ima e o he o eign-o igin eache e ec i he e a e posi i e
spillo e e ec s, i.e., i be e eading comp ehension inc eases ma h es sco es.
5.2 Wha explains he o eign-o igin eache e ec ?
5.2.1 Language-speci ic skills
To unde s and why he e is an e ec o eache s’ o igin, no only o o eign-o igin
s uden s bu also o na i e s uden s, his sec ion ocuses on an ob ious cha ac e is ic
in which na i e and o eign-o igin eache s di e : language skills. Na i e eache s—
by de ini ion—only ha e Ge man as a mo he ongue, while o eign-o igin eache s
o en g ew up lea ning di e en mo he ongues.
The economic li e a u e has es ablished he impo ance o languages in p e e ence
o ma ion (Ange e e al. 2016; Su e e al. 2018; Chen e al. 2019) and s udied he
ad an ageso bilingualeduca ion o bo h a ge edand un- a ge eds uden s(Chine al.
2013;Lle as-Muneyand She ze 2015;Cappella iand Di Paolo2018).17 Howe e ,we
know li le abou how bilingualism a ec s he way o eign-o igin eache s unde s and
and each languages.
The “conscious compe ence” lea ning model desc ibes he psychological p og ess
om incompe ence o compe ence in a skill. The model implies ha “conscious com-
pe ence” eache s (e.g., bilingual eache s who ha e consciously lea ned Ge man
language skills) a e be e eache s han “unconscious compe ence” eache s (e.g.,
na i e eache s), because “unconscious compe ence” eache s can ha e di icul ies
in explaining he skill ha has become la gely ins inc ual o hem (Robinson 1974).
Such an “awa eness” mechanism would be in line wi h ecen e idence indica ing ha
eache s o colo a e pa icula ly well-equipped in eaching s uden s in a cul u ally
ele an and engaging way in he US (Egali e and Kisida 2018).
This s udy is he i s o elici he e ec o a po en ial language-speci ic skill by es -
ing i o eign-o igin eache s who epo a mo he ongue o he han Ge man a ec
s uden s’ eading comp ehension di e en ly. In he da a a hand, he main mo he
17 Chin e al. (2013), o example, e alua e he e ec o a bilingual educa ion p og am on he achie emen
o limi ed English p o icien (LEP) s uden s and hei classma es. Employing a eg ession-discon inui y
design, hey ind no impac on he achie emen o s uden s o whom he p og am was designed (LEP
s uden s), bu es ima e a posi i e e ec o hei classma es.
123
Page17o 32 45
L.S. Höckel
ongues spoken by bilingual eache s a e Polish (19%), I alian (17%), and English
(17%) and, o e all, he dis ibu ion o di e en mo he ongues o bilingual eache s
is simila o he samples o o eign-o igin and na i e s uden s.18 Table 7spli s he
ca ego y o o eign-o igin eache in o wo and dis inguishes be ween bilingual each-
e s, i.e., eache s who epo a mo he ongue o he han Ge man, and o eign-o igin
eache s which do no epo o be bilingual. The able illus a es ha he posi i e e ec
o o eign-o igin eache s is d i en by bilingual o eign-o igin eache s. In compa ison
o he baseline eg ession in Table 6, he e ec o ha ing a bilingual eache is la ge
and mo e signi ican han he e ec o ha ing a o eign-o igin eache o bo h na i e
and o eign-o igin s uden s. Fo na i e s uden s, he eading comp ehension inc eases
by 0.53, i.e., 0.43 s anda d de ia ions, when hey ha e a bilingual eache . The posi i e
e ec o being augh by a non-bilingual o eign-o igin eache is smalle in size and
s a is ically insigni ican o bo h subsamples.
The in o ma ion on eache s’ mo he ongues u he allows o he analysis o
he linguis ic dis ance be ween he eache ’s mo he ongue and Ge man. In line wi h
he “conscious compe ence” lea ning model, we would expec ha bilingual eache s
wi h a la ge linguis ic dis ance o Ge man ha e a s onge awa eness o linguis ic
di icul ies in Ge man and hus a e mo e success ul in eaching Ge man. Table 8
displays he e ec o he linguis ic dis ance be ween bilingual eache s’ mo he ongue
and Ge man. I illus a es ha eache s wi h mo he ongues e y dissimila o Ge man
in luence s uden s’ eading comp ehension mo e posi i ely han eache s who ha e
Ge man o languages simila o Ge man as a mo he ongue. This posi i e e ec o
eache s wi h linguis ically di e en mo he ongues o Ge man adds aluable insides
o he li e a u e which has iden i ied nega i e e ec s o linguis ic dis ance on language
acquisi ion (Ispho ding and O en 2014; Ispho ding 2014) and employmen p ospec s
(Wong 2023) o immig an s.
As he majo i y o o eign-o igin eache s a e second-gene a ion immig an s, I
al e na i ely es i he e ec o bilingual eache s is caused by second-gene a ion
immig an s, who could be mo e likely o be bilingual. Howe e , only 33% o bilin-
gual eache s in he sample a e second-gene a ion immig an s, indica ing ha he
majo i y o bilingual eache s a e a he immig an eache s wi h s ong knowledge in
hei mo he ongue who ha e s udied Ge man long enough o be conside ed bilingual.
O e all, he e is no e ec o nei he i s no second-gene a ion immig an eache s
pe se (see Table A5) and di e ences in bilingualism a e mo e success ul in explain-
ing he posi i e e ec o o eign-o igin eache s’ e ec on hei s uden s’ eading
comp ehension.19
A second conce n is ha he e ec could be d i en by a pa icula ly well-equipped
immig an g oup, which is mo e likely o be bilingual. One example o such a g oup
couldbeGe manmino i iesinEas e n Eu ope who ypically lea nGe mana homeand
ha e easie access o pe manen esidence in Ge many h ough he Fede al Expellees
18 Table A4 displays desc ip i e s a is ics o s uden s wi h and wi hou bilingual eache s.
19 Al e na i ely, I es whe he i ma e s i bilingual eache s s a e ha hey ha e lea ned Ge man as hei
i s mo he ongue, hei second mo he ongue o no as a mo he ongue a all. While he o e all e ec
is less p onounced in compa ison o he e ec o bilingual eache s, Table A6 indica es ha o eign-o igin
s uden s bene i in pa icula om bilingual eache s who ha e lea ned Ge man as a second language o did
no men ion Ge man as a mo he ongue a all.
123
45 Page 18 o 32
Language lesson lea ned— o eign-o igin eache s
Table 7 E ec o bilingual eache on eading comp ehension
(1) (2) (3)
Panel A: all s uden s
Bilingual eache ( e .: na i e eache )
Fo eign-o igin eache 0.033 0.136 0.027
(0.091) (0.102) (0.138)
Bilingual o eign-o igin eache 0.292∗∗∗ 0.358∗∗∗ 0.440∗∗
(0.074) (0.103) (0.181)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.377 0.378 0.661
Obse a ions 7351 7351 7351
Panel B: na i e s uden s
Bilingual eache ( e .: na i e eache )
Fo eign-o igin eache 0.141 0.209 0.049
(0.167) (0.167) (0.249)
Bilingual o eign-o igin eache 0.365∗∗∗ 0.456∗∗∗ 0.528∗
(0.108) (0.134) (0.291)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.361 0.361 0.651
Obse a ions 4768 4768 4768
Panel C: o eign-o igin s uden s
Bilingual eache ( e .: na i e eache )
Fo eign-o igin eache −0.027 0.052 0.125
(0.107) (0.137) (0.291)
Bilingual o eign-o igin eache 0.231∗∗ 0.244∗0.575∗∗∗
(0.115) (0.133) (0.218)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.399 0.398 0.655
Obse a ions 2583 2583 2583
No es: Resul s a e ob ained om OLS eg essions. All epo ed s anda d e o s a e clus e ed a he eache
le el.
*p<0.10; **p<0.05; ***p<0.01
123
Page19o 32 45
L.S. Höckel
Table 8 E ec o linguis ic dis ance be ween eache ’s mo he ongue and Ge man on eading comp ehen-
sion
(1) (2) (3)
Panel A: all s uden s
Linguis ic dis ance 0.003∗∗∗ 0.004∗∗∗ 0.004∗∗
(0.001) (0.001) (0.002)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.382 0.382 0.669
Obse a ions 7206 7206 7206
Panel B: na i e s uden s
Linguis ic dis ance 0.004∗∗∗ 0.004∗∗∗ 0.005
(0.001) (0.002) (0.003)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.365 0.366 0.654
Obse a ions 4686 4686 4686
Panel C: o eign-o igin s uden s
Linguis ic dis ance 0.002 0.002 0.006∗∗
(0.001) (0.002) (0.002)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.406 0.405 0.672
Obse a ions 2520 2520 2520
No es: Resul s a e ob ained om OLS eg essions. All epo ed s anda d e o s a e clus e ed a he eache
le el.
*p<0.10; **p<0.05; ***p<0.01
Ac . While I do no obse e which o eign-o igin eache s belong o Ge man mino i ies
in Eas e n Eu ope, I can es o di e en ial e ec s ac oss language g oups. In doing
so, I ind posi i e and la ge poin es ima es o all language g oups (see Table A7).
Acco dingly, he esul s canno be explained by one pa icula ly mo i a ed o able
bilingual immig an g oup.
In summa y, he esul s in his sec ion show ha o eign-o igin eache s who a e
bilingual a e especially equipped o each eading comp ehension. They inc ease he
123
45 Page 20 o 32
Language lesson lea ned— o eign-o igin eache s
eading comp ehension o bo h na i e and o eign-o igin s uden s signi ican ly.20 Fu -
he mo e, bilingual eache s wi h linguis ically dis an mo he ongues o Ge man
a e pa icula ly success ul in imp o ing he eading comp ehension o hei s uden s.
Al hough he bilingual eache e ec is iden i ied by a a he small sample o bilin-
gual eache s, ela ed eache cha ac e is ics such as language amily o immig an
gene a ion do no p o ide be e explana ions o he posi i e o eign-o igin eache
e ec .
5.2.2 Al e na i e mechanisms
Besides language-speci ic skills o bilingual eache s, he posi i e e ec o o eign-
o igin eache s could be explained by o he mechanisms. Fi s , I es i o eign-o igin
s uden s bene i om eache bias and ole model e ec s in a simila way o mino i y
s uden s bene i ing om same- ace eache s. Second, I in es iga e i o eign-o igin
eache s pu mo e e o and ime in o hei wo k. Gi en he high equi emen s o
immig an eache s o p ac ice eaching in Ge many, o eign-o igin eache s who end
up wo king as eache s could be a pa icula ly posi i ely selec ed g oup o eache s.
Empi ically, ole model and eache bias e ec s a e di icul o disen angle. The e-
o e, some s udies di ec ly es o eache bias and disc imina ion e ec s (Pa edes
2014; Hinne ich e al. 2015; Ge shenson e al. 2016; Alan e al. 2018; Alesina e al.
2018). Dee (2005) exploi s s uden -speci ic e alua ions om eache s and shows ha
e hnic ma ching be ween s uden and eache has la ge e ec s on eache s’ pe cep-
ion o s uden pe o mance. Simila ly, Ge shenson e al. (2016) ind ha non-black
eache s ha e signi ican ly lowe educa ional expec a ions o black s uden s. E i-
dence om Ge many inds eache disc imina ion o essay g ades o s uden s wi h a
Tu kish-sounding i s name (Sp ie sma 2013) and g ade penal ies in p ima y school
o second-gene a ion immig an s (Kiss 2013). Fo he Ne he lands, an Ewijk (2011)
canno con i m a g ading bias bu lowe expec a ions and un a o able a i udes o
majo i y eache s owa ds mino i y s uden s. Employing Chilean da a o in es iga e
gende ma ching e ec s on academic achie emen , Pa edes (2014) uniquely es s o
ole model and eache bias e ec s in he same se ing. She inds ha gi ls bene i
om ha ing emale eache s and shows ha he e ec is only signi ican o subjec s
wi h lowe p opo ions o emale eache s and o gi ls wi h less educa ed mo he s.
Acco dingly, she in e p e s he esul s as a ole model a he han a eache bias e ec .
Following hese s udies, I i s es o a eache bias e ec , meaning an explici
posi i e beha io o eache s a ge ed owa ds ma ched s uden s. I analyze he g ading
o o eign-o igin eache s owa ds na i e and o eign-o igin s uden s. A eache bias
e ec would imply a posi i e e ec o o eign-o igin eache s on he g ade ewa ded
o s uden s whom hey demog aphically ma ch wi h, i.e., o eign-o igin s uden s and
a nega i e e ec on s uden s, hey demog aphically do no ma ch wi h, i.e., na i e
s uden s. In Ge many, ew cen alized exams a e conduc ed un il g ade 9, and he
g ading o s uden s is mos ly le o he disc e ion o he eache . Table 9shows he
20 One u he conce n could be ha s uden s who a e exposed o a change in bilingualism o he eache
signi ican ly di e om hose who do no . Table A8 shows ha o eign-o igin s uden s a e mo e likely o
be in schools wi h bilingual eache s and ha changes a he happen in highe g ades. Apa om ha he
s uden cha ac e is ics a e balanced.
123
Page21o 32 45
L.S. Höckel
Table 9 E ec o o eign-o igin eache on Ge man g ade
(1) (2) (3)
Panel A: na i e s uden s
Fo eign-o igin eache −0.032 −0.062 −0.081
(0.076) (0.050) (0.088)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.225 0.224 0.641
Obse a ions 5704 5704 5704
Panel B: o eign-o igin s uden s
Fo eign-o igin eache −0.174∗∗ −0.122 −0.075
(0.083) (0.101) (0.129)
Class FE Yes Yes Yes
S uden con ols Yes Yes Yes
Teache con ols No Yes Yes
S uden FE No No Yes
Adjus ed R20.174 0.176 0.630
Obse a ions 2864 2864 2864
No es: Resul s a e ob ained om OLS eg essions. All epo ed s anda d e o s a e clus e ed a he eache
le el.
*p<0.10; **p<0.05; ***p<0.01
e ec on he Ge man g ade ewa ded by he eache and does no e eal a eache
bias e ec in he sense ha o eign-o igin eache s a o o eign-o igin s uden s in
hei g ading.21 O e all, he esul s illus a e ha he o eign eache e ec is no
d i en by an ob ious eache bias leading o a be e g ading o o eign-o igin s uden s
by o eign-o igin eache s. Ins ead, he e is a nega i e bu insigni ican associa ion
be ween o eign-o igin eache s and s uden s’ Ge man g ade o bo h o eign-o igin
and na i e s uden s.
Second, I in es iga e a ole model e ec . While s udies in a uni e si y se ing
in e p e exposu e o emale acul y membe s o emale ins uc o s in ini ial cou ses as
emale olemodele ec s(e.g.,CanesandRosen1995;Be inge andLong2005), hese
s udies canno ule ou di ec eache in luence ia eache bias e ec s. To o e come
his p oblem, Dee (2007) compa es s uden s’ pe cep ion on he subjec augh by
ma ched and unma ched eache s o elici he ole model e ec mo e di ec ly. Su ey
ques ions included “Subjec is no use ul o my u u e” and “I am a aid o ask
ques ions in subjec class”. Likewise, I app oxima e ole mode e ec s by employing
su ey ques ions which e lec he po en ial in luence Ge man language eache s ha e
21 I ob ain simila esul s unning eg essions on a ha monized sample o s uden s aking he eading
comp ehension es .
123
45 Page 22 o 32
Language lesson lea ned— o eign-o igin eache s
on s uden s’ subjec in ol emen , hei eache pe cep ion and o e all happiness. Mo e
speci ically, I es i ha ing a o eign-o igin Ge man eache a ec s (i) he ime s uden s
spend eading ou side o school, (ii) s uden s’ pe cep ions i hei Ge man language
eache s expec s hem o y hei e y bes , and (iii) how sa is ied s uden s a e wi h
hei si ua ion a school. These h ee ques ions cap u e di e en dimensions in which
a ole model e ec can ope a e om a subjec -speci ic o a gene al way. I eache s’
o eign o igin can explain pa o he a ia ion in s uden s’ su ey answe s, his e ec
can be a ibu ed o a ole model e ec .22
Table 10 displays he esul s on a ole model e ec . Fo all h ee su ey i ems, he
coe icien s o o eign-o igin s uden s in panel B in compa ison o na i e s uden s in
panel A a e la ge bu no signi ican ly di e en om ze o in he p e e ed speci ica-
ions. The e ec sizes a e small and insigni ican o o eign-o igin s uden s’ eading
ime ou side o school, while he e ec s seem sizable bu insigni ican wi h espec o
eache pe cep ion and school sa is ac ion.23 This lack o a clea ole model e ec is
di e en om ea lie wo k om Egali e and Kisida (2018), who show ha s uden s
who ha e he same gende o /and acial cha ac e is ics as hei eache epo a mo e
posi i e pe cep ion o hei eache in e ms o eeling ca ed o and ins uc ional
cha ac e is ics ega ding he s uden - eache communica ion compa ed o unma ched
s uden s in he same class oom.24
Ins ead, he esul s in Table 10 can pa ly be explained by he “impe ec ” na u e o
he ma ches s udied. He e, a ole model e ec implies ha o eign-o igin s uden s can
be in luenced by any o eign-o igin eache equally and by a simila amoun . Howe e ,
hecoun ies o o igin di e widelybe ween o eign-o igin s uden s and eache s in he
sample, including combina ions in which ole model e ec s a e unlikely o eme ge.25
An al e na i e app oach is o allow o eign-o igin in e ac ion o ope a e only i a
s uden is ma ched wi h a eache which is a “close ma ch” wi h espec o he o eign
o igin. The e o e, I u he employ a language ma ch a iable as he main a iable o
in e es , which is equal o one i s uden and eache epo a mo he ongue o he
same language g oup. The es ima es o he language ma ch a iable a e insigni ican
o eading comp ehension (see TableA9) and g ading (seeTableA10). Fo he eading
comp ehension es ima ions o o eign-o igin s uden s, he poin es ima e o ha ing a
s uden - eache language ma ch is smalle han he e ec o ha ing a o eign-o igin
eache and hes anda de o sa ela ge .26 O e all, he posi i ee ec o o eign-o igin
22 In his con ex , I de ine a ole model e ec as a posi i e a i ude o he s uden i espec i e o his a i ude
being caused by an unde lying eache bias e ec o no .
23 The esul s o Table 10 a e compa able when con olling o s uden s es sco es.
24 Resul s on o he su ey i ems such as “I am con inced ha I can lea n a lo h ough eading”, “My
Ge man eache i s ies o unde s and my poin o iew, and hen ells me wha he/she would do.”, “My
Ge man eache encou ages me o ask ques ions”, and “How sa is ied a e you cu en ly and in gene al e ms,
wi h you li e?” do also no e eal a clea ole model e ec in he sense ha he e is a posi i e e ec on he
ma ched and a nega i e e ec on he unma ched s uden s. They a e a ailable upon eques .
25 E.g., a s uden om a cul u ally dis an coun y such as China would p obably no iden i y much mo e
wi h a Swiss han a Ge man eache .
26 Using a linguis ic dis ance measu e o p oxy he ma ch wi h espec o he disad an age in eading
comp ehension a ising om he linguis ic dis ance be ween s uden and eache (e.g., Ispho ding 2014)
yields in simila esul s (see Table A11).
123
Page23o 32 45
L.S. Höckel
Table 10 Role model e ec
Reading ime Teache expec a ion School sa is ac ion
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Panel A: na i e s uden s
Fo eign-o igin eache 0.050 0.034 −0.025 0.046 0.027 0.081 0.498∗0.420∗0.373
(0.073) (0.089) (0.143) (0.091) (0.088) (0.186) (0.300) (0.239) (0.429)
Class FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
S uden con ols Yes Yes Yes Yes Yes Yes Yes Yes Yes
Teache con ols No Yes Yes No Yes Yes No Yes Yes
S uden FE No No Yes No No Yes No No Yes
Adjus ed R20.113 0.111 0.447 0.072 0.070 0.174 0.089 0.092 0.343
Obse a ions 4601 4601 4601 4682 4682 4682 4708 4708 4708
Panel B: o eign-o igin s uden s
Fo eign-o igin eache −0.052 −0.076 0.056 0.072 0.059 0.285 0.083 0.393 0.619
(0.061) (0.078) (0.174) (0.090) (0.118) (0.261) (0.257) (0.289) (0.582)
Class FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
S uden con ols Yes Yes Yes Yes Yes Yes Yes Yes Yes
Teache con ols No Yes Yes No Yes Yes No Yes Yes
S uden FE No No Yes No No Yes No No Yes
Adjus ed R20.154 0.150 0.410 0.076 0.071 0.186 0.078 0.084 0.270
Obse a ions 2439 2439 2439 2521 2521 2521 2549 2549 2549
No es: Resul s a e ob ained om OLS eg essions. All epo ed s anda d e o s a e clus e ed a he eache le el.
*p<0.10; **p<0.05; ***p<0.01
123
45 Page 24 o 32
Language lesson lea ned— o eign-o igin eache s
Danze AM, Feue baum C, Piopiunik M, Woessmann L (2022) G owing up in e hnic encla es: language
p o iciency and educa ional a ainmen o immig an child en. J Popul Econ 35(3):1297–1344
Dee TS (2004) Teache s, ace, and s uden achie emen in a andomized expe imen . Re Econ S a
86(1):195–210
Dee TS (2005) A eache like me: does ace, e hnici y, o gende ma e ? Am Econ Re 2(95):158–165
Dee TS (2007) Teache s and he gende gaps in s uden achie emen . J Human Res 42(3):528–554
Dus mann C, Fabb i F (2003) Language p o iciency and labou ma ke pe o mance o immig an s in he
UK. Econ J 113(489):695–717
Dus mann C, F a ini T, Lanza a G (2012) Educa ional achie emen o second-gene a ion immig an s: an
in e na ional compa ison. Econ Policy 27(69):143–185
Dus mann C, Machin S, Schönbe g U (2010) E hnici y and educa ional achie emen in compulso y school-
ing. Econ J 120(546):F272–F297
Dus mann C, an Soes A (2001) Language luency and ea nings: es ima ion wi h misclassi ied language
indica o s. Re Econ S a 83(4):663–674
Egali e AJ, Kisida B (2018) The e ec s o eache ma ch on s uden s’ academic pe cep ions and a i udes.
Educ E al Policy Anal 40(1):59–81
Egali e AJ, Kisida B, Win e s MA (2015) Rep esen a ion in he class oom: he e ec o own- ace eache s
on s uden achie emen . Econ Educ Re 45:44–52
Eu opean Commission (2016) S udy on he di e si y wi hin he eaching p o ession wi h pa icula ocus
on mig an and/o mino i y backg ound: inal epo o DG educa ion and cul u e o he Eu opean
Commission. Luxembou g: Publica ions O ice o he Eu opean Union
Fai lie RW, Ho mann F, O eopoulos P (2014) A communi y college ins uc o like me: ace and e hnici y
in e ac ions in he class oom. Am Econ Re 104(8):2567–2591
Fe ig M (2010) The socie al in eg a ion o immig an s in Ge many. In: Eps ein GS, Gang IN (eds) Mig a ion
and cul u e. Eme ald G oup Publishing Limi ed
Fleishe B, Hashimo o M, Weinbe g BA (2002) Fo eign GTAs can be e ec i e eache s o economics. J
Econ Educ 33(4):299–325
Foged M, Hasage L, Pe i G, A end JN, Bol ig I (2022) Language aining and e ugees’ in eg a ion.
Re iew o Economics and S a is ics, pp. 1–41
Geay C, McNally S, Telhaj S (2013) Non-na i e speake s o English in he class oom: wha a e he e ec s
on pupil pe o mance? Econ J 123(570):F281–F307
Geh e K, Zimme mann S, A el C, Weine S (2012) The assessmen o eading compe ence (including
sample i ems o g ade 5 and 9): Scien i ic Use File 2012, Ve sion 1.0.0. Bambe g: Uni e si y o
Bambe g, Na ional Educa ional Panel S udy
Ge shenson S, Hol SB, Papageo ge NW (2016) Who belie es in me? The e ec o s uden - eache demo-
g aphic ma ch on eache expec a ions. Econ Educ Re 52:209–224
GEW (2021) Ve schenk e Chancen?! - Die Ane kennungs- und Beschä igungsp axis on mig ie en
Leh k ä en in den Bundeslände n. Handou zu P essekon e enz am 6. Sep embe 2021
Giannelli GC, Rapallini C (2016) Immig an s uden pe o mance in ma h: does i ma e whe e you come
om? Econ Educ Re 52:291–304
G i i h AL, Main JB (2021) The ole o he eaching assis an : emale ole models in he class oom. Econ
Educ Re 85:102179
Hame mesh DS, T ejo SJ (2013) How do immig an s spend hei ime? The p ocess o assimila ion. J Popul
Econ 26(2):507–530
Hay on JE (2001) Language aining, language p o iciency and ea nings o immig an s in No way. Appl
Econ 33(15):1971–1979
Hinne ich BT, Höglin E, Johannesson M (2015) Disc imina ion agains s uden s wi h o eign backg ounds:
e idence om g ading in Swedish public high schools. Educ Econ 23(6):660–676
Ho mann F, O eopoulos P (2009) A p o esso like me: he in luence o ins uc o gende on college
achie emen . J Human Res 44(2):479–494
Ispho ding IE (2014) Disad an ages o linguis ic o igin - e idence om immig an li e acy sco es. Econ
Le 123(2):236–239
Ispho ding IE, O en S (2014) Linguis ic ba ie s in he des ina ion language acquisi ion o immig an s. J
Econ Beha O gan 105:30–50
Kiss D (2013) A e immig an s and gi ls g aded wo se? Resul s o a ma ching app oach. Educ Econ
21(5):447–463
123
Page31o 32 45
L.S. Höckel
Kliche H, Täubig V (2019) Beglei ung schulische Übe gänge in Wohng uppen de Kinde -und Jugendhil e.
Soziale Passagen 11(1):47–63
Lang J (2022) Employmen e ec s o language aining o unemployed immig an s. J Popul Econ
35(2):719–754
Lle as-Muney A, She ze A (2015) Did he Ame icaniza ion mo emen succeed? An e alua ion o he
e ec o English-only and compulso y schooling laws on immig an s. Ame ican Economic Jou nal:
Economic Policy 7(3):258–290
Lochmann A, Rapopo H, Speciale B (2019) The e ec o language aining on immig an s’ economic
in eg a ion: empi ical e idence om F ance. Eu Econ Re 113:265–296
Malecki A, Schneide Ch is oph, Vogel Sil ia, Wol e s M (2014) Schulen au einen Blick: Ausgabe 2014.
S a is isches Bundesam , Wiesbaden
NEPS Ne wo k (2021) Na ional Educa ional Panel S udy, scien i ic use ile o s a ing coho g ade 5.
Leibniz Ins i u e o Educa ional T ajec o ies (LI Bi), Bambe g. h ps://doi.o g/10.5157/NEPS:SC3:
11.0.1
Niyubahwe A, Mukamu e a J, Ju as F (2013) P o essional in eg a ion o immig an eache s in he school
sys em: a li e a u e e iew. McGill Jou nal o Educa ion/Re ue des sciences de l’éduca ion de McGill
48(2):279–296
No eisch K (2006) School ca chmen a ea e asion: he case o Be lin. Ge many. J Educ Policy 22(1):69–90
OECD (2018) The esilience o s uden s wi h an immig an backg ound: ac o s ha shape well-being.
OECD Publishing, Pa is
Pa edes V (2014) A eache like me o a s uden like me? Role model e sus eache bias e ec . Econ Educ
Re 39:38–49
RobinsonWL(1974)Conscious compe ency- he ma k o a compe en ins uc o .Pe sonnelJou nal 53:538–
539
Ruhose J, Schwe d G (2016) Does ea ly educa ional acking inc ease mig an -na i e achie emen gaps?
Di e ences-in-di e ences e idence ac oss coun ies. Econ Educ Re 52:134–154
Schnep SV (2007) Immig an s’ educa ional disad an age: an examina ion ac oss en coun ies and h ee
su eys. J Popul Econ 20(3):527–545
Seah KKC (2018) Immig an educa o s and s uden s’ academic achie emen . Labou Econ 51:152–169
Seah KKC (2021) Do you speak my language? The e ec o sha ing a eache ’s na i e language on s uden
achie emen . Sou he n Economic Jou nal 88(1):245–273
Sp ie sma M (2013) Disc imina ion in g ading: expe imen al e idence om p ima y school eache s. Empi
Econ 45(1):523–538
S a is isches Bundesam (2012) Be ölke ung und E we bs ä igkei : Be ölke ung mi Mig a ionshin e g und
- E gebnisse des Mik ozensus 2012, Fachse ie 1, ol.2.2
S a is isches Bundesam (2018) Sonde auswe ung. Be ölke ung und E we bs ä igkei : Be ölke ung mi
Mig a ionshin e g und - E gebnisse des Mik ozensus 2012-2016. Wiesbaden
S a is isches Bundesam (2023) Jede zwöl e Leh k a an allgemeinbildenden Schulen wa im Schuljah
2021/2022 Que - ode Sei eneins eige /-in. P essemi eilung N . N053 om 4. Ok obe 2023
S einhaue HW, Zinn S (2016) NEPS echnical epo o weigh ing: weigh ing he sample o s a ing coho
3 o he Na ional Educa ional Panel S udy (Wa e 1 o 3). NEPS Wo king Pape No. 63
Su e M, Ange e S, Glä zle-Rü zle D, Le ge po e P (2018) Language g oup di e ences in ime p e e -
ences: e idence om p ima y school child en in a bilingual ci y. Eu Econ Re 106:21–34
U ban Ins i u e (2017). Di e si ying he class oom: examining he eache pipeline
an Ewijk R (2011) Same wo k, lowe g ade? S uden e hnici y and eache s’ subjec i e assessmen s.
Economics o Educa ion Re iew 30(5):1045–1058
Va gas AJ (2016) Assimila ion e ec s beyond he labo ma ke : ime alloca ions o Mexican immig an s o
he US. Re iew o Economics o he Household 14(3):625–668
Wong L (2023) The e ec o linguis ic p oximi y on he labou ma ke ou comes o he asylum popula ion.
J Popul Econ 36(2):609–652
Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps
and ins i u ional a ilia ions.
123
45 Page 32 o 32
