G übl, Dominik; Lackne , Ma io; Win e -Ebme , Rudol
Wo king Pape
In e gene a ional T ansmission o Unemploymen - Causal
E idence om Aus ia
IHS Wo king Pape , No. 14
P o ided in Coope a ion wi h:
Ins i u e o Ad anced S udies (IHS), Vienna
Sugges ed Ci a ion: G übl, Dominik; Lackne , Ma io; Win e -Ebme , Rudol (2020) : In e gene a ional
T ansmission o Unemploymen - Causal E idence om Aus ia, IHS Wo king Pape , No. 14, Ins i u
ü Höhe e S udien - Ins i u e o Ad anced S udies (IHS), Vienna
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/217041
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
IHS Wo king Pape 14
Ap il 2020
In e gene a ional T ansmission o
Unemploymen - Causal E idence
om Aus ia
Dominik G übl
Ma io Lackne
Rudol Win e -Ebme
Au ho (s)
Dominik G übl, Ma io Lackne , Rudol Win e -Ebme
Ti le
In e gene a ional T ansmission o Unemploymen - Causal E idence om Aus ia
Ins i u ü Höhe e S udien - Ins i u e o Ad anced S udies (IHS)
Jose s äd e S aße 39, A-1080 Wien
T +43 1 59991-0
F +43 1 59991-555
www.ihs.ac.a
ZVR: 066207973
License
„In e gene a ional T ansmission o Unemploymen - Causal E idence om Aus ia“ by
Dominik G übl, Ma io Lackne , Rudol Win e -Ebme is licensed unde he C ea i e
Commons: A ibu ion 4.0 License (h p://c ea i ecommons.o g/licenses/by/4.0/)
All con en s a e wi hou gua an ee. Any liabili y o he con ibu o s o he IHS om he
con en o his wo k is excluded.
All IHS Wo king Pape s a e a ailable online:
h ps://i ihs.ihs.ac.a / iew/ihs_se ies/se =5Fihswps.h ml
This pape is a ailable o download wi hou cha ge a : h ps://i ihs.ihs.ac.a /id/ep in /5286/
In e gene a ional T ansmission o Unemploymen – Causal
E idence om Aus ia
*
Dominik G ¨
ublb, Ma io Lackne b,c, and Rudol Win e -Ebme a,b,c
aIns i u e o Ad anced S udies (IHS), Vienna
bDepa men o Economics, Johannes Keple Uni e si y Linz (JKU)
cCh is ian Dopple Labo a o y “Aging, Heal h and he Labo Ma ke ”
Ap il 6, 2020
Abs ac
We es ima e he causal e ec o pa en s’ unemploymen on unemploymen among hei chil-
d en in hei own adul hood. We use adminis a i e da a o Aus ian child en bo n be ween
1974 and 1984 and apply an ins umen al a iables (IV) iden i ica ion s a egy using pa -
en s’ job loss du ing a mass layo as he ins umen . We ind e idence o unemploymen
inhe i ance in he nex gene a ion. An addi ional day o unemploymen du ing childhood
causally aises he a e age unemploymen days o he adul child by 1 o 2%. The g ea es
e ec s a e obse ed o unma ied pa en s, young child en, child en o low-educa ion pa en s,
and in amilies li ing in capi al ci ies. We also explo e a ious channels o in e gene a ional
unemploymen , such as educa ion, income, and job ma ching by pa en s.
Keywo ds: in e gene a ional ansmission, mass layo , unemploymen du a ion, ins umen al
a iable
JEL Classi ica ion: J62, J64
*
Co esponding au ho : Ma io Lackne , Depa men o Economics, Al enbe ge S . 69, 4040 Linz;
Tel.:+43/732/2468-7390; Email: ma io.lackne @jku.a . Thanks o commen s by R. An on B aun, Go don B.
Dahl, And eas Gulyas, Ma in Halla, Robe Kuns , Iacopo Mo chio, And ea Webe and semina pa icipan s in
Nu embe g and Vienna and o inancial suppo by he LIT and he Aus ian FWF. Rudol Win e -Ebme is also
a ilia ed wi h IZA, CEPR, CReAM and ROA.
1 In oduc ion
In e gene a ional mobili y and pe sis ence in social o labo ma ke ou comes a e impo an
ac o s o equal oppo uni y. Many s udies ha e been conduc ed on in e gene a ional pe sis ence
in a eas like income (Blanden,2019), educa ion, and heal h (Black and De e eux,2011), bu
s udies on in e gene a ional pe sis ence in unemploymen a e a e, o h ee easons: i) sui able
da a on pa en s and child en a e no eadily a ailable, ii) no e ec i e iden i ica ion s a egy has
been de eloped, and iii) he e is no clea -cu channel by which unemploymen in one gene a ion is
ans e ed o he nex . We add ess all hese h ee esea ch gaps using long- e m adminis a i e
da a on Aus ian wo ke s and ins umen al a iables s a egy based on mass layo s. Mo eo e ,
we p o ide an ex ensi e discussion on po en ial channels o in e gene a ional unemploymen .
Like in e gene a ional pe sis ence in incomes, pe sis ence in unemploymen is impo an o
public policies. While simple co ela ions o unemploymen be ween a he and son may say
no hing abou labo ma ke policies, causal ela ions can be impo an : A posi i e causal e ec o
pa en s’ unemploymen on child en’s unemploymen can mul iply he impac o a success ul labo
ma ke policy; a educ ion in he unemploymen o he a he can also educe unemploymen
among he sons. This e ec is independen om he ul ima e easons o his causal e ec . Such
a causal ela ionship may p oceed h ough di e en channels, which may allow a ying policy
in e en ions. Pa en al unemploymen may lead o highe unemploymen o he child en due o
income dep i a ion, educ ion in school access, loss o pa en al job-sea ch ne wo ks, o educed
wo k e hics. All o hese channels may call o labo ma ke policies designed o p e en pa en al
unemploymen . While we s ess he easons o a posi i e causal e ec , nega i e causal e ec s
a e also possible. Pa en al unemploymen may lead o h ea s o a s igma: The child en may
see unemploymen - ela ed p oblems mo e clea ly and may hus in es mo e in a oiding hese
p oblems hemsel es. The in e gene a ional ansmission o unemploymen is also ela ed o he
deba e abou he ansmission o he ”wel a e cul u e” (An el,1992), and we will discuss ou
esul s conce ning ha issue as well.
Ou iden i ica ion amewo k uses exogenous a ia ion in mass layo s, which can a ec he
employmen s a us o he pa en s du ing he childhood o he obse ed indi iduals. Being laid
o in he cou se o a mass layo se es as an ins umen o he pa en s’ deg ee o unemploymen ,
bu his is no di ec ly ela ed o he deg ee o unemploymen among he child en wo decades
la e .
Anonymized adminis a i e da a co e ing 1974 o 2016 allow us o obse e he indi idual labo
ma ke ou comes o each pa en -child pai in yea ly in e als. O e all, he esul s show posi i e
and signi ican causal es ima es o he ansmission o unemploymen om one gene a ion o
he o he . Ou main es ima es a y be ween 0.12 and 0.35 addi ional days o unemploymen pe
yea o adul child en i hei pa en had one addi ional day o yea ly unemploymen du ing
hei childhood. The e ec s a e s onge o unma ied pa en s, young child en, he child en o
low-educa ion pa en s and in amilies li ing in capi al ci ies.
We discuss wo po en ial iola ions o he ins umen ’s exclusion es ic ion: i) mass layo s may
be oo small o exogenously a ec employed pe sons, and ii) a mass layo in a egion/ illage
1

may ha e long-las ing impac s on he egional labo ma ke and a ec he u u e employmen
p ospec s o he child en. The esul s a e obus when we use a mo e conse a i e ins umen
ha conside s only la ge mass layo s. Conce ning he second po en ial iola ion, we show
ha ou esul s a e obus when we include ou -digi ZIP-codes o conside la ge ci ies only: in
hese ci cums ances, long-las ing labo ma ke e ec s should be accoun ed o o be unimpo an
ega ding capi al ci ies.
We iden i y child en’s educa ional ca ee as one plausible ansmission channel o in e gen-
e a ional unemploymen , bu i is no he only one: E en in si ua ions whe e he e a e no
educa ional consequences o pa en al unemploymen , child en s ill su e highe unemploymen
a es.
2 Con ibu ion o he li e a u e
The e is a la ge li e a u e on he in e gene a ional ansmission o labo ma ke ou comes. This
s udy concen a es only on unemploymen .
The unemploymen o a pa en and ha o a child may be co ela ed due o a ious ac o s, such
as gene ics, amily cul u e, abili y, o ambi ions. Pa en al unemploymen may no necessa ily
explain he child en’s labo ma ke ou come. Es ablishing e ec i e policies equi es ha a causal
link be iden i ied.
Se e al me hods o iden i ying causal e ec s ha e been sugges ed. One app oach is o use a
ixed e ec s model wi h siblings, whe ein one is exposed o pa en al unemploymen and he
o he is no , which will heo e ically minimize he con ounding ac o s sha ed wi hin a amily
(Ekhaugen,2009).1Ano he app oach is he Go schalk (1996) me hod, which can also be
applied o unemploymen . I includes pa en s’ u u e wel a e pa icipa ion as an explana o y
a iable in he eg ession o child en’s wel a e pa icipa ion on pa en s’ wel a e pa icipa ion. By
exploi ing he o de o e en s, he me hod aims o iden i y he e ec s o unobse ed he e ogenei y
in he amily. The emaining co ela ion be ween he pa en ’s and child’s ou comes can hen
be assumed o be causal. A modi ied e sion o his app oach uses pa en s’ p edic ed u u e
wel a e pa icipa ion. The mos common me hod is using an ins umen al a iable. O eopoulos
e al. (2008) analyze Canadian da a o iden i y he causal in e gene a ional e ec s o a he -son
pai s. They use plan closu es as exogenous shocks a ec ing he a he s’ displacemen du ing
hei sons’ childhood.
Maede e al. (2015) explo e he in e gene a ional ansmission o unemploymen om a he s
o sons in Ge many. They use he Ge man Socio-economic Panel (GSOEP) and ind posi i e
co ela ions bu no signi ican causal esul s using ei he he Go schalk me hod o an ins u-
men al a iable app oach. Howe e , he au ho s use annual indus y-speci ic unemploymen isk
o ins umen he a he s’ unemploymen (wi h ela i ely low explana o y powe ). In a ela ed
s udy, M¨
ulle e al. (2017) use he sibling ixed e ec s and he Go schalk me hod on da a om
he GSOEP and ind no e ec s o pa e nal unemploymen on he ou comes o sons, bu hey
1The a ia ion in exposu e be ween siblings is assessed by measu ing pa en al unemploymen a e he olde
sibling has al eady mo ed ou .
2
iden i y posi i e causal e ec s on he daugh e s’ wo klessness. In e es ing insigh s in o whe he
pa en s’ labo ma ke ou comes ha e a di ec e ec on child en’s labo ma ke ou comes o a e
channeled h ough educa ion a e p o ided by H´e aul and Kalb (2016) using Aus alian admin-
is a i e da a. They ind ha , al hough educa ion has an e ec on child en’s unemploymen , i s
e ec is independen o he di ec e ec , which is una ec ed by he inclusion o educa ion as an
endogenous o exogenous ac o . In con as o M¨
ulle e al. (2017), he au ho s u he show
ha a pe iod o six mon hs o pa en al unemploymen du ing childhood inc eases unemploymen
du a ion by 2.74 pe cen age poin s o sons and 0.44 pe cen age poin s o daugh e s be ween
20 and 54 yea s o age. O eopoulos e al. (2008) use i m closu es in Canada be ween 1980 and
1982 as ins umen s o he displacemen o a he s and ind a posi i e e ec on he unemploy-
men o child en be ween 25 o 33 yea s o age, d i en p ima ily by poo e amilies. Fu he
e idence on he posi i e in e gene a ional co ela ion be ween unemploymen and wo klessness
is p esen ed by Macmillan (2014) o he UK, albei wi h limi ed causal in e p e abili y. A
ecen su ey analysis on Eu opean coun ies by D oule ´y e al. (2019) e ealed ha pa en al
unemploymen when he child en a e 14 has a signi ican impac on he child en’s likelihood o
being unemployed when hey a e 18 o 35 yea s o age. No e ec s conce ning child en’s unem-
ploymen a e ound by G egg e al. (2012), who map indus y con ac ions in he UK in he
1980s on o he a he s’ displacemen . Howe e , hey men ion ha hey a e unable o di ec ly
a ibu e job displacemen o he a he s and ha hei esul s ha e low p ecision due o small
sample sizes. S udying No way, Ekhaugen (2009) also inds no signi ican causal ansmission
o unemploymen ac oss gene a ions, bu only o child en aged 24 o 26.
The in e gene a ional ansmission o unemploymen is a special case o in e gene a ional mobil-
i y. The e a e a leas wo o he ansmission channels wi h high in e gene a ional pe sis ence:
income (E iksson e al.,2005;Fan,2016;Le g en e al.,2012;Mazumde ,2005) and educa ion
(Ande sen,2011;Huang,2013;Wendelspiess Ch´a ez Ju´a ez,2015;M¨
ulle e al.,2017;Palomino
e al.,2019). These ac o s clea ly in e ac wi h one ano he , making i especially ha d o disen-
angle hem. Fo example, a sudden inc ease in pa en al unemploymen will induce an income
educ ion. E en ually, his o egone income migh be pa ly subs i u ed by unemploymen ben-
e i s o ese e asse s, bu la ge enough dec emen s could make educa ion mo e cos ly han he
ea ly labo ma ke en y o he child. Fu he mo e, expe ience wi h unemploymen migh lowe
he child’s inhibi ions agains being dependen on social bene i s (An el,1992;Dahl e al.,2014)
and he eby lowe e u ns o educa ion o he child. Con e sely, schooling e o s migh be
highe , ei he because unemployed pa en s migh ac as a de e ence o hei child en o be-
cause pa en al ime in es men in o hei child en is highe (Yum,2016). In any case, cong uen
o opposi e ansmission in hese ou comes ac oss gene a ions can ansla e in o a ia ions in
unemploymen among he child en; hose ansmission channels mus be kep in mind.
Amid he he e ogeneous esul s and he issues a ising om he a ious models used, we con-
ibu e o his li e a u e in se e al ways. Fi s , we o e come mos o he p oblems associa ed
wi h sample size issues and measu emen e o , by using comp ehensi e adminis a i e da a
aken om he Aus ian Social Secu i y Da a Base (ASSD). Second, using an ins umen al a i-
able es ima ion me hod enables us o mi iga e con ounding ac o s such as sha ed amily ac o s,
and allows us o ex ac he causal pa o he in e gene a ional ansmission e ec . Thi d, his
3
s udy is he i s o conduc his ype o analysis o Aus ia, whose social secu i y s uc u es di -
e om hose o he Uni ed S a es and he Uni ed Kingdom and a e much close o Ge many’s.
Finally, we explo e he channels o he posi i e in e gene a ional co ela ion in unemploymen .
3 Da a
The s udy’s da a a e d awn om he ASSD (Zweim¨
ulle e al.,2009), a comp ehensi e da abase,
co e ing all Aus ians since he 1970s and o e ing de ailed in o ma ion abou employmen spells
and wages collec ed by he Aus ian Social Secu i y Adminis a ion. We conside i s bo n
child en bo n be ween 1974 and 1984 and link yea ly pe sonal and employmen in o ma ion on
he pa en s o he child en when hey we e be ween 8 and 14 yea s o age. This se en-yea
pe iod will be e e ed o as ”pe iod X”. Addi ionally, he panel cap u es ele an indi idual and
labo ma ke in o ma ion abou he child en when hey we e 30 o 32 yea s old; his is deno ed
as ”pe iod Y”.
We o m yea ly a e ages o he pa en s’ obse a ional pe iod X and he child en’s inal ou come
pe iod Y (see Figu e 1). Fu he mo e, child en whose a he ’s o mo he ’s main occupa ion
du ing pe iod X was a seasonal job o public se ice a e d opped; hese cases would con ound
ou pa en al unemploymen measu e due o he excessi ely long pe iods o egula seasonal un-
employmen o unusually high employmen p o ec ion. A small numbe o child en wi h an
ob iously w ong (>365) sum o employmen and unemploymen days a e also elimina ed. Sim-
ila ly, child en who a e simul aneously unemployed and e i ed a e d opped om he sample.
The same logic is also applied o he pa en s. Missing in o ma ion on pa en al educa ion educes
he sample size u he by a small deg ee. The excluded obse a ions do no di e sys ema i-
cally om he es o he sample, no a e he e signi ican di e ences wi hin he son/daugh e
subsamples.
Obse ing pa en s’ unemploymen spells i is necessa y o impose he equi emen whe eby he
pa en s ha e o be o mally employed du ing he childhood pe iod o hei o sp ing o a pe iod
o mo e han 365 days.
3.1 Desc ip i es
Table 1 epo s he desc ip i es o he key a iables. In panel A, he desc ip i es e e o all
households exceeding 365 days o employmen o bo h pa en s, while panels B and C a e spli
in o a he s and mo he s, espec i ely. The o e all sample comp ises 154,011 indi idual child en.
The sha e o emales is 48.1%. The sha e o public o icials is qui e small, a a ound 0.6 %. The e
is also a small ac ion (9%) o seasonal wo ke s among child en, o which sons accoun o he
la ge po ion. The a e age pa en al age is 37 o a he s, and 33 o mo he s. The mo he ’s
yea ly income is a ound 9,450 eu os less han hal ha o he a he s, bu hese numbe s include
pe iods o non-wo k. The a he s end o wo k in sligh ly la ge i ms han he mo he s, while
mo he s end o be employed in whi e colla jobs mo e equen ly. As expec ed, pa en s do no
di e acco ding o he gende o hei child en.
4
The sample child en a e unemployed o 14 days and employed o 262 days pe yea on a e age,
while hey a e ou o he labo o ce o 59 days. Pa en al lea e is g an ed o 36 days pe yea ,
almos en i ely o emales. Pa en s a e on a e age unemployed 5.3 ( a he s) o 9.5 (mo he s) days
pe yea . The ewe pa en al unemploymen days compa ed o he child en e lec s he gene al
inc ease in unemploymen o e ime. Again, he e a e no sys ema ic di e ences in pa en al labo
ma ke ou comes be ween sons and daugh e s.
The c i e ia o iden i ying mass layo s adhe e o he manda o y epo ing guidelines o he
Public Employmen Se ice Aus ia (AMS). Fi s , we exclude i ms wi h ewe han 20 employ-
ees. Fo i ms wi h 20 o 100 employees, a mass layo is coded i a leas i e employees a e
dismissed; o i ms o up o 600 employees, a mass layo is coded i a leas i e pe cen o
he wo k o ce a e dismissed; o i ms wi h mo e han 600 employees, a mass layo is coded i
a leas 30 wo ke s a e laid o . To measu e he layo size o each i m, we use a wo ke low
app oach, which quan i ies ou going wo ke s om one qua ile o he nex . We exclude i ms
ha es uc u e, open new b anches, o a e in ol ed in M&A, akeo e s, o simila ac i i ies
ha canno be classi ied as ue mass layo s. To do so, we iden i y join mo e s, which a e
classi ied as a g oup o laid-o wo ke s who make up a leas 30 pe cen o all laid-o employees
in hei i m and mo e join ly o a new i m. Cases ha o mally exceed he h esholds o laid-o
wo ke s bu include join mo e s a e no classi ied as mass layo s. We p o ide mo e s ingen
measu es o mass layo s in sec ion 8.
Table 2summa izes he a e age likelihood o pa en s being dismissed in mass layo s (ML). The
numbe s show ha he e was a 11.3% chance o a a he being laid o in a i m ha had a
mass layo in he pe iod when he child was 8 o 14 yea s old. The isk o mo he s was 12.5%,
and he isk o ei he o he wo being laid o was 16.5%. In a subsequen analysis, we use a
simpli ied de ini ion o mass layo s o mula ed by Sulli an and on Wach e (2009), who de ine
a mass layo as a s a educ ion by mo e han 30% o he i m’s peak employmen du ing he
las six yea s. In ou da a, plan closu es a e ela i ely a e. Only 1,601 men and 717 women
los hei jobs due o a plan closu e; his amoun s o only 0.15% o pa en al pai s whe e a leas
one indi idual was laid o in a plan closu e. Consequen ly, we canno use plan closu es as an
exogenous shock on pa en al unemploymen .
We measu e he employmen ou comes o he child en when aged 30 o 32, o se e al easons.
Mos impo an ly, we wan o use a longe obse a ion pe iod o a oid andom e ec s and
compa e pa en s o child en o simila ages, which is impo an o mi iga ing li e-cycle biases
be ween he pa en s and child en. Child en o e 30 ha e likely inished hei educa ion (Black
and De e eux,2011). The s udy’s design is depic ed in Figu e 1. All a iables a e measu ed
as yea ly a e ages, unless s a ed di e en ly in he speci ica ions o he analysis and obus ness
checks.
5
6 E ec he e ogenei y
Ou discussion so a has been limi ed o o e all e ec s. In his sec ion we deepen ou unde -
s anding o he ansmission p ocess o he sample’s subg oups. Table 5spli s he sample in o
ma ied s. unma ied pa en s a he beginning o pe iod X. Unma ied pa en s can be ei he
ne e ma ied o di o ced. The able also epo s whe he he e ec s di e s a is ically om one
ano he . Unma ied pa en s ha e a signi ican ly s onge ansmission o unemploymen han
ma ied pa en s which may be due o a s onge (and mo e unique) ole model e ec . Ano he
eason migh be he ac ha loss o income sou ces is mo e se e e o a single pa en .
Columns (3) and (4) epo he e ec s o single on child en and child en wi h siblings. The
in e gene a ional ansmission o unemploymen is somewha s onge o child en who ha e
siblings a he beginning o pe iod X.
Table 5: E ec he e ogenei y - Family backg ound
Ma ied pa en s Single on
(1) (2) (3) (4)
no yes no yes
Pa en s’ UE daysa0.2291*** 0.1376*** 0.2067*** 0.1725***
(0.0335) (0.0310) (0.0199) (0.0299)
Di e ence c0.092 0.034
P- alued0.000 0.100
F-s a is icsb675.0 301.4 209.0 766.5
1s s age coe icien 17.8031*** 14.1023*** 14.5104*** 16.8498***
(0.6852) (0.8123) (1.0038) (0.6086)
Sample mean 15.58 12.6 13.14 14.43
OLS coe icien 0.0538*** 0.0435*** 0.0526*** 0.0477***
(0.0050) (0.0067) (0.0055) (0.0068)
N55,835 98,176 89,511 64,500
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Models (1) and (2)
a e a spli by he ma i al s a us a he beginning o pe iod X. Models (3) and (4) a e spli by whe he he child is a single on child up o he
end o pe iod X. All es ima ions include con ol a iables om he main speci ica ion (siblings excluded as con ol in column (3) and (4)) as in
Table3, as well as coho and egional ixed e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
cBoo s ap and pe mu a ion es o di e ence in coe icien s be ween g oups (Pe mu a ions=100).
dRepo ed p- alue o di e ence in coe icien s.
Nex , we explo e whe he child age (a he ime o pa en al unemploymen ) o pa en al educa ion
is ins umen al o he in e gene a ional co ela ion. Table 6 epo s he esul s o spli ing he
age ange o child en in wo g oups: ages 8 o 11 in column (1) and ages 12 o 14 in column
(2). The in e gene a ional co ela ion is signi ican ly g ea e i pa en al unemploymen happens
when he child is be ween 8 and 11 yea s old. This age ange coincides wi h he decision o send
he child o high-school o no . 3
3Aus ia has a school sys em wi h an ea ly acking schedule whe eby s uden s can p oceed o an academic
ack o mo e p ac ical s udies when hey u n 10.
12

Columns (3) and (4) e eal ha in e gene a ional ansmission is s onge o he mo e highly
educa ed pa en s. Table 7digs deepe in his issue by in e ac ing hese wo dimensions. We
show he e ec s o bo h age g oups sepa a ely o high and low pa en al educa ion. An age di -
e ence is obse ed only o low-educa ion pa en s (Cols. (1) and (2)). A la ge in e gene a ional
ansmission o unemploymen occu s o low-educa ion pa en s when he child is be ween 8 and
11. The e ec is only hal as s ong i pa en al unemploymen occu s when he child is be ween
12 and 14. This demons a es he g ea impo ance o school selec ion.
Table 6: E ec he e ogenei y - Age and pa en al educa ion
Age g oup Pa en al educa ion
(1) (2) (3) (4)
8-11 12-14 low high
Pa en s’ UE daysa0.1531*** 0.0739*** 0.1925*** 0.2157***
(0.0258) (0.0138) (0.0255) (0.0465)
Di e ence c-0.079 -0.023
P- alued0.000 0.000
F-s a is icsb289.4 221.2 534.3 170.1
1s s age coe icien 16.9943*** 26.4761*** 18.5205*** 10.6260***
(0.9989) (1.7803) (0.8012) (0.8147)
Sample mean 13.54 13.54 14.83 11.75
OLS coe icien 0.0345*** 0.0301*** 0.0559*** 0.0472***
(0.0038) (0.0030) (0.0050) (0.0098)
N151,638 151,638 96,466 57,545
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Models (1) and (2)
a e sepa a e eg essions de ined o age g oups o 8-11 and 12-14 yea s. Models (3) and (4) a e spli by whe he he highes achie ed educa ion
o he pa en s is abo e an equi alen o a g adua ion diploma (Ma u a). All es ima ions include con ol a iables om he main speci ica ion
(pa en s’ educa ion excluded as con ol in column (3) and (4)) as in Table3, as well as coho and egional ixed e ec s. Regional-le el clus e ed
s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
cBoo s ap and pe mu a ion es o di e ence in coe icien s be ween g oups (Pe mu a ions=100).
dRepo ed p- alue o di e ence in coe icien s.
7 T ansmission channels
Why a e he in e gene a ional ansmissions so la ge? As men ioned, educa ion is an ob ious
ac o . Pa en al unemploymen migh hinde he child en’s educa ional choices o po en ial
(Coelli,2011;Jones,1988;Rege e al.,2011). We explo e his possibili y in Tables 8and
9. A second mechanism migh in ol e amily ne wo ks and s uc u e. Unemploymen may
educe pa en al ne wo ks, which may educe he child en’s job- inding capaci y (Plug e al.,
2018). Mo eo e , pa en al unemploymen migh inc ease ension wi hin he amily, leading
o a highe chance o di o ce, which may in luence he adul child’s labo ma ke ou comes
(De-Goede e al.,2000;S ¨
om,2003). This possibili y is explo ed in Table 10. Finally, he
e ec s o income dep i a ion due o pa en al unemploymen a e in es iga ed in Table 11. O he ,
13
Table 7: Age g oups by pa en al educa ion
Lowe pa en al educ. Highe pa en al educ.
(1) (2) (3) (4)
8-11 12-14 8-11 12-14
Pa en s’ UE daysa0.2022*** 0.0942*** 0.1463*** 0.1365***
(0.0331) (0.0198) (0.0519) (0.0378)
Di e ence c-0.108 -0.011
P- alued0.000 0.330
F-s a is icsb348.0 776.3 245.3 87.7
1s s age coe icien 20.4185*** 32.7683*** 11.9756*** 20.2156***
(1.0946) (1.1760) (0.7647) (2.1588)
Sample mean 14.83 14.83 11.75 11.75
OLS coe icien 0.0392*** 0.0352*** 0.0355*** 0.0275***
(0.0040) (0.0043) (0.0091) (0.0059)
N96,466 96,466 57,545 57,545
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Samples a e
spli by he age o he obse ed child du ing pe iod X, as well as he highes achie ed educa ion o he pa en s is abo e o equal high-school
g adua ion (Ma u a). All es ima ions include con ol a iables om he main speci ica ion (excep pa en s’ educa ion) as in Table3, as well
as coho and egional ixed e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
po en ial ansmission channels a e ela ed o changes in he wo k e hics o he unemployed
pa en s; howe e , such channels canno be es ed using adminis a i e da a.
Table 8shows he ins umen al a iable es ima es o he e ec o pa en al unemploymen on
he educa ion o he o sp ing: yea s o educa ion, educa ion abo e he median, and e ia y
educa ion 4a e used as dependen a iables. While we ind no signi ican nega i e e ec s on
e ia y educa ion, 100 addi ional days o pa en al unemploymen educe educa ion by 0.72 yea s
and educe he p obabili y o being abo e he median by 0.18 pe cen age poin s. These esul s
s ess he impo ance o educa ion.
Table 9explo es he ole o educa ion u he by conside ing pa en al backg ound — which has
been ound o be a e y s ong p edic o o child en’s educa ion (Black e al.,2005) — and child
age as condi ioning ac o s. Table 9is simila o Table 7, bu , ins ead o looking a child en’s
unemploymen , Table 9conside s he p obabili y ha child en’s educa ion is g ea e han he
median. As in Table 7, we see la ge e ec s o pa en s wi h low educa ion. Among his g oup,
he g ea es e ec is o child en a he c i ical age o 10, when key educa ional decisions ha e
o be made. Howe e , he esul s o mo e highly educa ed pa en s a e subs an ially di e en .
We obse e no signi ican e ec o pa en al unemploymen on he p obabili y o child en aged 8
o 11 comple ing high-school. The e ec is signi ican bu small o child en aged 12 o 14.
4As se e al alues a e missing, we impu ed he educa ion a iable using he andom o es me hod. The esul s
a e obus o he use o gene ic educa ion only, which in ol es a smalle sample size.
14
Table 8: Educa ional choice o child en
(1) (2) (3)
yea s be e e ia y
Pa en s’ UE daysa-0.0072*** -0.0018*** -0.0005
(0.0019) (0.0004) (0.0003)
F-s a is icsb453.6 453.6 453.6
1s s age coe icien 15.6163*** 15.6163*** 15.6163***
(0.7332) (0.7332) (0.7332)
Sample mean 13.28 .49 .25
OLS coe icien -0.0024*** -0.0004*** -0.0002***
(0.0002) (0.0000) (0.0000)
N153,987 153,987 153,987
No es: IV esul s. The dependen a iable is yea s o educa ion o he child in model (1), a bina y a iable equal o 1 i he child’s highes
educa ion is abo e middle school in model (2), a bina y equal o 1 i he child has a uni e si y deg ee in model (3). All es ima ions include
con ol a iables om he main speci ica ion as in Table3, as well as coho and egional ixed e ec s. Regional-le el clus e ed s anda d e o s
in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
Table 9: Educa ional choice o child en by age g oup and pa en al educa ion – abo e o below he median
Lowe pa en al educ. Highe pa en al educ.
(1) (2) (3) (4)
8-11 12-14 8-11 12-14
Pa en s’ UE daysa-0.0030*** -0.0015*** 0.0002 -0.0007**
(0.0004) (0.0002) (0.0010) (0.0003)
F-s a is icsb347.8 776.5 244.9 87.6
1s s age coe icien 20.4234*** 32.7690*** 11.9766*** 20.2156***
(1.0951) (1.1760) (0.7654) (2.1593)
Sample mean .39 .39 .66 .66
OLS coe icien -0.0005*** -0.0004*** -0.0005*** -0.0004***
(0.0000) (0.0000) (0.0001) (0.0000)
N96,451 96,451 57,536 57,536
No es: IV esul s. The dependen a iable is 1 i he i he obse ed child’s educa ional a ainmen is abo e o equal high-school g adua ion
(Ma u a), 0 else. Samples a e spli by he age o he obse ed child du ing pe iod X, as well as he highes achie ed educa ion o he pa en s
is abo e o equal high-school g adua ion (Ma u a). All es ima ions include con ol a iables om he main speci ica ion (excep pa en s’
educa ion) as in Table3, as well as coho and egional ixed e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p <
0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
While he esul s o low-educa ion pa en s indica e an educa ional ansmission, he small
esul s o high-educa ion pa en s show ha educa ional ansmission canno be he only expla-
na ion o he in e gene a ional ansmission o unemploymen .
The second hypo hesis conce ns pa en al ne wo ks and amily s uc u e. Table 10 shows he
esul s o IV eg essions es ing whe he , a he beginning o pe iod Y he child wo ks in he
same sec o o in he same i m as one o he pa en s, which migh be due o a pa en al job
15
Table 10: Plausible ansmission channels: Family ne wo k and dis up ion
Same sec o Same i m
(1) (2) (3) (4) (5)
Fa he Mo he Fa he Mo he Di o ce
Pa en s’ UE daysa-0.0004* -0.0002* -0.0002*** 0.0001** 0.0010***
(0.0002) (0.0001) (0.0001) (0.0000) (0.0002)
F-s a is icsb454.1 454.1 454.1 454.1 454.1
1s s age coe icien 15.6168*** 15.6168*** 15.6168*** 15.6168*** 15.6168***
(0.7328) (0.7328) (0.7328) (0.7328) (0.7328)
Sample mean .16 .15 .01 .01 .06
OLS coe icien -0.0001*** -0.0001** -0.0000*** -0.0000** 0.0002***
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
N154,011 154,011 154,011 154,011 154,011
No es: IV esul s. The dependen a iable is a bina y equal o 1 i he child wo ks in he same sec o (NACE08) as he a he o mo he , o
models (1) and (2), espec i ely. The dependen a iable is a bina y equal o 1 i he child wo ks in he same i m as he a he o mo he ,
o models (3) and (4), espec i ely. The dependen a iable is a bina y equal o 1 i he pa en s di o ced be ween he pe iods X and Y in
model (5), 0 o he wise. All es ima ions include con ol a iables om he main speci ica ion as in Table3, as well as coho and egional ixed
e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
ne wo k, and whe he he pa en s ge di o ced. The esul s show ha bo h job ne wo ks and
amily dis up ions ma e : 100 addi ional days o he a he ’s (mo he ’s) unemploymen educe
he p obabili y o he child being in he same sec o by 0.04 (0.02) pe cen age poin s. The
nega i e e ec s o being in he same i m – an e en s onge indica ion o ne wo k e ec s – a e
a ound hal o ha . These a e ela i ely la ge e ec s. We also see ha he di o ce p obabili y
is inc eased.
In addi ion o less access o educa ion and educed pa en al ne wo k e ec i eness, income dep i-
a ion because o unemploymen is ano he po en ially impo an channel o in e gene a ional
unemploymen ansmission. Lowe amily income is ound o ha e nega i e e ec s on child
achie emen and labo ma ke ou comes (Dahl and Lochne ,2012). Un o una ely, ou da a
does no p o ide de ailed in o ma ion on wo king hou s and hou ly wages o p oxy po en ial
household income. Mo eo e , in o ma ion on income o pa en s in ou da a is pollu ed by bonus
paymen s, se e ance paymen s, as well as unemploymen spells wi hou de ailed in o ma ion on
unemploymen bene i s. Consequen ly, we cons uc a a iable o measu ing po en ial house-
hold income, which measu es he po en ial wage a household can expec gi en all obse ed
household cha ac e is ics.5
Table 11 p esen s he es ima ion esul s o ou main model o ou qua iles o he po en ial
household income. We es ima e a posi i e and signi ican e ec o pa en al unemploymen days
on child en’s unemploymen days o all qua iles o he po en ial income dis ibu ion. I income
5Po en ial income is calcula ed on a yea ly basis as [obse ed ac ual income/days employed]*calenda days. To
a oid un easonable esul s om possible e oac i e o bonus paymen s on a single day we es ic he po en ial
income compu a ion o wo ke s wi h a leas 31 days o employmen in a gi en yea . We impu e missing alues
wi h means by yea , sex, age, educa ion, and o eignness. I nei he ac ual income, no employmen days a e
obse ed, po en ial income is coded as missing. Figu e 2in he appendix p o ides ke nel densi y es ima ions o
ac ual and po en ial income in ou sample.
16
loss we e he p ima y sou ce o he in e gene a ional ansmission o unemploymen , amilies
in he op qua ile would show he smalles e ec , i any. Top-ea ning amilies a e expec ed o
ha e high inancial ese es and could expec o ind a job ela i ely quickly a e a mass layo .
Consequen ly, we conclude ha income loss canno be he main d i ing o ce o he obse ed
in e gene a ional co ela ion.
Table 11: A e age po en ial income qua iles o he households
(1) (2) (3) (4)
Q1 Q2 Q3 Q4
Pa en s’ UE daysa0.1295*** 0.2514*** 0.1361** 0.2096***
(0.0298) (0.0404) (0.0517) (0.0602)
F-s a is icsb510.4 509.1 235.9 127.1
1s s age coe icien 19.9112*** 17.6763*** 13.7042*** 9.8711***
(0.8813) (0.7834) (0.8923) (0.8755)
Sample mean 15.96 14.18 12.49 12.11
Uppe h eshold (income) 30,773.15 36,461.69 42,589.8 98,423.55
OLS coe icien 0.0492*** 0.0318*** 0.0658*** 0.0444***
(0.0060) (0.0053) (0.0107) (0.0120)
N38,495 38,495 38,494 38,494
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Columns (1) o
(4) p esen esul s o sepa a e samples in each po en ial income qua ile 1 o 4, espec i ely. The uppe h eshold ma ks he maximum alue
o po en ial income o each qua ile. All es ima ions include con ol a iables om he main speci ica ion as in Table3, as well as coho and
egional ixed e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
8 Robus ness checks
This sec ion explo es se e al p oblems wi h ou ins umen al a iables s a egy: i) mass layo s
could be selec i e and hi a e y speci ic g oup o wo ke s, and ii) he exclusion es ic ion could
be iola ed i a mass layo had an independen e ec on child en 20 yea s la e . Such an e ec
could happen i a (la ge) mass layo caused se ious long- e m dis o ions on a local o egional
labo ma ke (Foo e e al.,2019;Ga hmann e al.,2018). In such a si ua ion, he labo ma ke
would con inue o be weake due o his mass layo .
We y o cap u e he i s e ec by changing he s udy’s mass layo de ini ion. We i s de ine
a si ua ion as a mass layo only i i is abo e he median size (in e ms o he absolu e numbe
o dismissed employees) o all mass layo s. Second, and mo e conse a i ely, we use a de ini ion
d awn om Sulli an and on Wach e (2009), whe eby a layo a ec ing mo e han 30% o he
i ms’ peak employmen o e he las six yea s is de ined as a mass layo .6
6We de ia e om he o iginal de ini ion by also including i m his o ies sho e han six yea s i hey a e no
a ailable o he ull leng h. Fu he mo e, we se he minimum i m size o 20 ins ead o 50 employees and place
no es ic ions on employee enu e.
17

I we ede ine ou mass layo de ini ion, we in en ionally con amina e ou con ol sample, as
se e al mass layo s a e no longe coded as such. Cols (2) and (3) in Table 12 show hese
es ima es. Remo ing hese con amina ed con ol a iables (i.e. d opping hem om he sample)
educes ou sample (see Cols (4) and (5)).7Table 2desc ibes he new samples wi h sha es
e e ing o he o iginal sample size. Rede ining he ea men , educes he numbe o ea ed
elemen s by app oxima ely 50% o 90%, depending on he de ini ion.
The esul s a e epo ed in Table 12. The es ima ion esul s o he o iginal model a e epo ed
again in column (1) o compa ison. Using only la ge layo s (Cols. (2) and (4)) does no
change he esul s. Using he Sulli an and on Wach e (2009) de ini ion educes he es ima es
somewha ; he es ima es also lose s a is ical signi icance, pe haps due o he much smalle
ea men g oup used.
Table 12: Mass layo de ini ion and size
se ze o d op
(1) (2) (3) (4) (5)
O ig. ML de ini ion La ge ML ML Sulli an La ge ML ML Sulli an
Pa en s’ UE daysa0.1842*** 0.2027*** 0.1109 0.1995*** 0.1303
(0.0215) (0.0538) (0.0974) (0.0452) (0.0814)
F-s a is icsb454.1 135.9 188.2 161.7 263.4
1s s age coe icien 15.6168*** 11.8681*** 12.2577*** 13.6504*** 15.3014***
(0.7328) (1.0179) (0.8936) (1.0734) (0.9428)
Sample mean 13.68 13.68 13.68 13.29 12.99
OLS coe icien 0.0502*** 0.0502*** 0.0502*** 0.0502*** 0.0424***
(0.0035) (0.0035) (0.0035) (0.0048) (0.0057)
N154,011 154,011 154,011 141,856 130,881
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Model (1) is he
o iginal model, models (2) and (4) e-de ine mass layo s as a bina y equal o 1 i he size o he layo is la ge han he median o all layo s
in he sample. Models (3) and (5) use he layo de ini ion simila o Sulli an and on Wach e (2009). Fo he g oup “se ze o”, obse a ions
which whe e o me ly ea ed in he o iginal de ini ion a e shi ed o he con ol g oup. In he g oup “d op”, o iginally ea ed obse a ions
a e d opped i hey a e no ea ed acco ding o he new de ini ion. All es ima ions include con ol a iables om he main speci ica ion as in
Table3, as well as coho and egional ixed e ec s. Regional-le el clus e ed s anda d e o s in pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
Ano he conce n o add ess is he possibili y ha mass layo s had a pe manen impac on he
local labo ma ke . A la ge mass layo in a small local communi y may pe manen ly educe
he numbe o wo kplaces in and a ound ha communi y. Such an ou come would iola e
he exclusion es ic ion, because he labo ma ke o young wo ke s 20 yea s la e migh be
impac ed. To add ess his a gumen we p o ide ixed e ec s o local labo ma ke s. Table
13 epo s he o iginal esul s using wo-digi ZIP codes as ixed e ec s in Col. (1); we hen
inc ease he g anula i y (up o h ee- and ou -digi ZIP codes) in columns (2) and (3). The
esul s a e p ac ically unchanged. These esul s show ha e en wi hin e y small communi ies,
he a he ’s unemploymen ma e s o he child en. Ano he es conside s la ge communi ies,
whe e he long- e m e ec s o mass layo s will be dispe sed. Columns (4) and (5) epo he
7Using plan closu es as an addi ional obus ness check is no possible, because he i s -s age es ima ion is
oo weak due o he small sample size.
18
esul s o la ge communi ies wi h mo e han 10,000 inhabi an s and o Aus ian capi al ci ies,
espec i ely. The esul s a e unchanged.
Table 13: Regional con ols and size o he local labo ma ke
(1) (2) (3) (4) (5)
O iginal 3-digi ZIP 4-digi ZIP >10,000 Capi al ci ies
Pa en s’ UE daysa0.1842*** 0.1713*** 0.1684*** 0.1750*** 0.2526***
(0.0215) (0.0206) (0.0237) (0.0330) (0.0453)
Region FE le el 2-digi 3-digi 4-digi 4-digi 4-digi
F-s a is icsb454.1 801.3 807.3 434.9 289.9
1s s age coe icien 15.6168*** 15.7493*** 15.8231*** 15.3832*** 15.0781***
(0.7328) (0.5564) (0.5569) (0.7377) (0.8856)
OLS coe icien 0.0502*** 0.0496*** 0.0501*** 0.0557*** 0.0508***
(0.0035) (0.0050) (0.0051) (0.0074) (0.0088)
N154,011 154,008 153,910 88,293 45,477
No es: IV esul s. The dependen a iable is he a e age yea ly unemploymen days o he child du ing he age o 30 o 32. Model (1) is he
o iginal model, model(2) uses he i s h ee digi s o he ZIP code o ixed e ec s and clus e ing. Model (3) implemen s a ou digi ZIP code.
Model (4) includes only communi ies wi h mo e han 10,000 inhabi an s, while model (5) includes Aus ian capi al ci ies only. All es ima ions
include con ol a iables om he main speci ica ion as in Table3, as well as coho ixed e ec s. Regional-le el clus e ed s anda d e o s in
pa en heses, ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01.
aPa en s’ UE days ep esen s he a e age numbe o unemploymen days pe yea o bo h pa en s.
bKleibe gen and Paap (2006) s a is ics on he ins umen in he i s s age.
In he obus ness check, we include he i m co a ia es men ioned abo e ha ha e no been
included so a due o a subs an ial p opo ion o missing alues. Fi m co a ia es a e ga he ed
om he las a ailable eco d abou he main job o each pa en up o wo yea s be o e pe iod X.
Missing alues can o igina e om pa en s who we e no employed du ing hese yea s, which could
be e y selec i e. Fu he mo e, hese co a ia es we e measu ed a a single poin in ime leading
o snapsho alues, which migh also lead o bias i hey a e included. Ne e heless, he ea ed
and con ol g oups may di e sys ema ically in e ms o jobs in a manne unaccoun ed o . Table
A.2 epo s he desc ip i es o he a ailable i m co a ia es and he missing alues acco ding o
he a he and mo he samples. Since including hese co a ia es educes he sample size, we also
impu e missing alues by yea , sex, age, educa ion, and o eignness. Table A.3 lis s he means o
he ea ed and con ol g oups and o he o iginal and impu ed da a, espec i ely. The epo ed
p- alues om a es o di e ences in means show ha we can ejec he null hypo hesis ha
he means a e he same, excep o he loga i hmic ans o ma ion o he daily a he ’s wage.
We he e o e add all hese co a ia es as addi ional con ols in ou model and epo he esul s
in able A.4 o he a ailable da a and in able A.5 o he impu ed da a. Al hough he esul s
exhibi small quan i a i e di e ences o he main speci ica ion, he es ima es a e quali a i ely
compa able, especially o he impu ed sample. We a gue ha ou decision o exclude hese
co a ia es due o conce ns abou measu emen e o , selec ion issues, and educed sample size
is alid and ha hei inclusion does no p oduce sys ema ic di e ences.
19
9 Conclusion and discussion
E idence o a causal in e gene a ional ansmission o unemploymen would sugges a clea
social and economic policy op ion: Reducing he unemploymen o a pa en will ha e long- e m
posi i e consequences o he child. Using comp ehensi e Aus ian Social Secu i y Da a and
an ins umen al a iables app oach, we show ha 10 addi ional days o a e age yea ly pa en al
unemploymen du ing he childhood o he o sp ing (ages 8 o 14) inc ease he adul child’s
yea ly a e age days o unemploymen by 1.2 o 3.5 days, o 9 o 24 pe cen o he mean. The
ansmission seems o be s onges o unma ied pa en s, o sons, and young child en o low-
educa ion pa en s. Ou ins umen al a iables s a egy elies on mass layo s; his ins umen
is obus when only e y la ge layo s a e used. Due o ou highly-g anula use o communi y
ixed e ec s, we can also ule ou any di ec e ec s o he ins umen on child en’s job chances.
While his gene al policy conclusion is independen o he ac ual ansmission mechanism, i is
s ill impo an o dig deepe and de e mine wha channels migh cause his in e gene a ional
co ela ion. Among educa ion, income dep i a ion, he loss o amily ne wo ks, and changes in
pa en al wo k e hics, we explo e he i s h ee. We ind ha educa ion is an impo an channel:
Child en om pa en s wi h low educa ion le els ha e ewe yea s o schooling and a lowe
likelihood o comple ing high-school. Such child en ha e mo e di icul ies channeling hemsel es
in o he highe educa ional ack (a age 10), which is an impo an p e equisi e o success in
he Aus ian labo ma ke . While he same in e gene a ional co ela ion is obse ed o highly-
educa ed pa en s, hei child en’s schooling does no su e signi ican ly. Income dep i a ion is
unlikely o be a channel, bu he loss o pa en al job ne wo ks appea s o ha e an impo an
impac on child en.
20
Re e ences
Ande sen, Signe Hald, “Common Genes o Exogenous Shock? Disen angling he Causal E ec o Pa e nal Unem-
ploymen on Child en’s Schooling E o s,” Eu opean Sociological Re iew, 12 2011, 29 (3), 477–488.
An el, John J, “The In e gene a ional T ans e o Wel a e Dependency: Some S a is ical E idence,” The Re iew
o Economics and S a is ics, 1992, 74 (3), 467–73.
Black, Sand a E. and Paul J. De e eux, “Recen De elopmen s in In e gene a ional Mobili y,” in O. Ashen el e
and D. Ca d, eds., Handbook o Labo Economics, Vol. 4, Else ie , Decembe 2011, chap e 16, pp. 1487–1541.
Black, Sand a E, Paul J De e eux, and Kjell G Sal anes, “Why he apple doesn’ all a : Unde s anding in e gen-
e a ional ansmission o human capi al,” Ame ican Economic Re iew, 2005, 95 (1), 437–449.
Blanden, Joanne, “In e gene a ional income pe sis ence,” IZA Wo ld o Labo , 2019.
Coelli, Michael B., “Pa en al job loss and he educa ion en ollmen o you h,” Labou Economics, 2011, 18 (1), 25
– 35.
Dahl, Go don B. and Lance Lochne , “The impac o amily income on child achie emen : E idence om he
ea ned income ax c edi ,” Ame ican Economic Re iew, 2012, 102 (5), 1927–56.
, And eas Ra ndal Kos øl, and Magne Mogs ad,“Family Wel a e Cul u es,”The Qua e ly Jou nal o Economics,
08 2014, 129 (4), 1711–1752.
De-Goede, Ma ijn, Ed Sp uij , Co a Maas, and Vincen Duindam, “Family p oblems and you h unemploymen ,”
Adolescence, 2000, 35, 587–601.
DelBoca, Daniela, Ch is ophe Flinn, and Ma hew Wiswall, “Household choices and child de elopmen ,” Re iew
o Economic S udies, 2014, 81 (1), 137–185.
D oule ´y, Ondˇ ej, Ma in Lukeˇs, and Mihaela Vancea, “Indi idual-le el and amily backg ound de e minan s o
young adul s’ unemploymen in Eu ope,” Empi ica, Jan 2019.
Ekhaugen, Ty a, “Ex ac ing he causal componen om he in e gene a ional co ela ion in unemploymen ,”
Jou nal o Popula ion Economics, 2009, 22 (1), 97–113.
E iksson, To , Be n B a sbe g, and Oddbj¨
o n Raaum, “Ea nings pe sis ence ac oss gene a ions: T ansmission
h ough heal h?,” Oslo Uni e si y, Depa men o Economics, Memo andum, 01 2005.
Fan, Yi, “In e gene a ional income pe sis ence and ansmission mechanism: E idence om u ban China,” China
Economic Re iew, 2016, 41, 299 – 314.
Foo e, And ew, Michel G osz, and Ann S e ens, “Loca e you nea es exi : Mass layo s and local labo ma ke
esponse,” Indus ial and Labo Rela ions Re iew, 2019, 72 (1), 101–126.
Ga hmann, Ch is ina, Ines Helm, and U a Sch¨
onbe g, “Spillo e e ec s o mass layo s,” Jou nal o he Eu opean
Economic Associa ion, 12 2018, 18 (1), 427–468.
Go schalk, Pe e , “Is he co ela ion in wel a e pa icipa ion ac oss gene a ions spu ious?,” Jou nal o Public
Economics, 1996, 63 (1), 1–25.
G egg, Paul, Lindsey Macmillan, and Bilal Nasim, “The Impac o Fa he s’ Job Loss du ing he Recession o he
1980s on hei Child en’s Educa ional A ainmen and Labou Ma ke Ou comes*,” Fiscal S udies, 2012, 33
(2), 237–264.
H´e aul , Nicolas and Guyonne Kalb,“In e gene a ional co ela ion o labo ma ke ou comes,”Re iew o Economics
o he Household, Ma 2016, 14 (1), 231–249.
21