Dieckho , Ma ina; Giesecke, Johannes
A icle — Published Ve sion
G owing o declining penal ies? A c oss- empo al analysis
o unemploymen sca s in he Ge man labo ma ke
Social Science Resea ch
P o ided in Coope a ion wi h:
WZB Be lin Social Science Cen e
Sugges ed Ci a ion: Dieckho , Ma ina; Giesecke, Johannes (2024) : G owing o declining penal ies?
A c oss- empo al analysis o unemploymen sca s in he Ge man labo ma ke , Social Science
Resea ch, ISSN 1096-0317, Else ie , Ams e dam, Vol. 121, pp. 1-17,
h ps://doi.o g/10.1016/j.ss esea ch.2023.102960
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/310923
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/
Social Science Resea ch 121 (2024) 102960
A ailable online 31 May 2024
0049-089X/© 2023 Else ie Inc. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
G owing o declining penal ies? A c oss- empo al analysis o
unemploymen sca s in he Ge man labo ma ke
Ma ina Dieckho
a
,
b
,
*
, Johannes Giesecke
c
a
Uni e si y o Ros ock, Depa men o Sociology and Demog aphy, Ge many
b
WZB Be lin Social Science Cen e , Ge many
c
Humbold Uni e si y Be lin, Dep . o Social Sciences, Ge many
ARTICLE INFO
Keywo ds:
Unemploymen sca ing
C oss- empo al analysis
Ge man labo ma ke
Ins i u ional change
ABSTRACT
We know ha unemploymen lea es sca s. Unemploymen sca s a e he penal ies in e ms o
employmen ou comes ha wo ke s expe ience due o pas unemploymen . To da e we lack a
long- e m longi udinal accoun which examines how unemploymen sca ing has de eloped o e
ime. The aim o his a icle is o ill his gap. We d aw on longi udinal da a om he Ge man
Socio-Economic Panel spanning a pe iod o mo e han 30 yea s, om 1985 o 2020, and in es-
iga e long- e m ime ends o unemploymen sca ing. The Ge man labo ma ke has expe i-
enced p o ound s uc u al and ins i u ional change o e he pas decades. These changes ha e
been associa ed wi h inc eased inequali ies in he labo ma ke . We examine whe he he sub-
s an ial ans o ma ion o he Ge man labo ma ke also had epe cussions o he ex en o pos -
unemploymen penal ies. We ocus on employmen p obabili ies and wages, and conside bo h
sho - e m ( wo yea s a e he unemploymen incidence) and mid- e m ou comes ( ou yea s
a e he unemploymen incidence). Changes in he amoun o unemploymen sca ing o e ime
can also occu due o changes in he composi ion o he unemployed. Ou analyses he e o e do
no only in es iga e how mac o-economic and ins i u ional change a e associa ed wi h a ying
amoun s o unemploymen sca ing, bu also con ol o and examine he ole o composi ional
change.
1. In oduc ion
Few hings a e as damaging o an indi idual’s u u e labo ma ke ou comes as a spell o unemploymen . Fo some unemployed
wo ke s i becomes impossible o e-en e he labo ma ke , and o hose who manage o escape unemploymen , subs an ial disad-
an age pe sis s: pas unemploymen , on a e age, signi ican ly inc eases he isk o u u e unemploymen (e.g. A ulampalam e al.
2000), incu s subs an ial and endu ing ( ela i e o absolu e) wage losses
1
(Gangl 2006; G egg and Tominey 2005), and leads o a
educ ion in job quali y also when non-pecunia y p ope ies o jobs a e conce ned (B and 2006; Dieckho 2011). I has also been
* Co esponding au ho . Uni e si y o Ros ock, Dep . o Sociology and Demog aphy, Ulmens . 69, 18057 Ros ock, Ge many.
E-mail add esses: [email p o ec ed] (M. Dieckho ), [email p o ec ed] (J. Giesecke).
1
Wage sca s a e ypically concei ed as ela i e wages losses, meaning ha he a e age wage g ow h o wo ke s en e ing a spell o unemploymen
is lowe han he a e age wage g ow h o wo ke s no en e ing a spell o unemploymen . Howe e , spells o unemploymen can also be ela ed o
absolu e wage losses, meaning ha pos -unemploymen wages a e ac ually lowe han p e-unemploymen wages. In he ollowing, we e e o wage
sca s in he o me meaning and ope a ionalize hem acco dingly in ou analyses.
Con en s lis s a ailable a ScienceDi ec
Social Science Resea ch
jou nal homepage: www.else ie .com/loca e/ss esea ch
h ps://doi.o g/10.1016/j.ss esea ch.2023.102960
Recei ed 15 Ma ch 2022; Recei ed in e ised o m 6 No embe 2023; Accep ed 28 No embe 2023
Social Science Resea ch 121 (2024) 102960
2
shown ha unemploymen can spu downwa d social mobili y and es ic upwa d mobili y: middle-class indi iduals who expe ience
a spell o unemploymen ha e a highe p obabili y o descending o lowe class posi ions and a lowe p obabili y o mo ing o a highe
social class han hei con inuously employed coun e pa s (Gebel 2016). Unemploymen is hence unde s ood o be a c ucial
li e-cou se e en wi h o en ad e se epe cussions o a ec ed indi iduals’ u u e li es (DiP e e and McManus 2000; Gangl 2006) and
sociologis s emphasize he cen al ole unemploymen plays in p oducing s a i ica ion (B and 2006; Gangl 2004, 2006; Mooi-Reci and
Ganzeboom 2015; Dieckho 2011).
The phenomenon ha unemploymen does no only a ec indi iduals while hey a e unemployed, bu also exe s de imen al
e ec s a and a e labo ma ke e-en y is e e ed o as ‘unemploymen sca ing’. Exis ing wo k has also e ealed he ex en o
unemploymen sca ing o a y c oss-na ionally (Gangl 2006) and demons a ed ha ins i u ional di e ences play an impo an ole in
explaining his a ia ion. This compa a i e wo k analyzed wage sca s in Eu ope and he US (based on da a om he mid-1990s un il
2001). In his analysis he Ge man labo ma ke , which has used o se e as p ime exempla o a conse a i e wel a e s a e and a highly
coo dina ed labo ma ke , akes up an in e media e posi ion in e ms o sca ing in ensi y. I a es be e han coun ies classi ied as
libe al egimes such as he US o he UK, bu less well han he Scandina ian coun ies ep esen ing social-democ a ic egimes (ibid.:
1009).
O e he pas decades he Ge man labo ma ke has unde gone a p o ound es uc u ing p ocess, in e alia i has expe ienced
subs an ial labo ma ke and wel a e e o ms as well as impo an changes in indus ial ela ions. Ge many has expe ienced a mo e
owa ds a mo e libe al ma ke economy and wel a e s a e. Agains his backd op, his con ibu ion asks whe he hese libe aliza ion
p ocesses we e accompanied by inc eased unemploymen sca ing. So while exis ing c oss-na ionally compa a i e wo k has shown
ha coun ies ep esen ing libe al wel a e s a e egimes a e wo se in e ms o sca ing, we ake a c oss- empo al pe spec i e and ocus
on libe aliza ion p ocesses wi hin one coun y and hei implica ions.
To ou knowledge his is he i s s udy conce ned wi h long- e m ime ends in unemploymen sca ing in a Wes e n economy.
Ge many makes o a e y in e es ing case o such a s udy gi en he p o ound es uc u ing and e o ms i has expe ienced. We
in es iga e how sca e ec s in Ge many ha e e ol ed o e ime and aim o iden i y he ins i u ional, mac o-economic as well as
composi ional dynamics behind obse ed c oss- empo al sca ing pa e ns. Ou s udy add esses h ee cen al ques ions: 1.) Does he
size o he sca e ec o unemploymen a y o e ime? 2.) To which ex en is his a ec ed by he composi ion o he unemploymen
in low and ou low popula ion? 3.) Which ole do ins i u ional and s uc u al de elopmen s bu also economic cycle play in explaining
empo al change o luc ua ions in he se e i y o unemploymen sca s?
We add ess hese ques ions based on longi udinal da a om he Ge man Socio-Economic Panel (SOEP) spanning 36 yea s
(1985–2020). These da a come om a na ionally ep esen a i e longi udinal panel su ey (Wagne e al. 2007). We es ima e he
e ec s o an unemploymen expe ience (o a leas wo mon hs) on subsequen employmen ou comes. In pa icula , we ocus on he
e ec s o unemploymen on u u e employmen p obabili ies and (condi ional on being e-employed) on wages. The longi udinal
na u e o he da a a o ds us wi h he possibili y o examine he employmen ou comes o wo ke s who expe ience a spell o un-
employmen and compa e hem o hose o (o he wise simila ) wo ke s who did no expe ience unemploymen . We examine unem-
ploymen sca s sho ly a e he unemploymen spell ( wo yea s a e he unemploymen incidence) as well in he mid- e m ( ou yea s
a e he unemploymen incidence). We es ima e a e age ea men e ec s by combining a (s ill ela i ely) new ma ching me hod
called ‘Coa sened Exac Ma ching’ (he ea e : CEM) p oposed by Iacus e al. (2012) wi h he di e ence-in-di e ences app oach
(Heckman e al. 1997). This analy ical s a egy allows us o con ol o selec ion based on obse able and ( ime cons an ) unobse able
cha ac e is ics. The ole o composi ion in explaining change o e ime is in es iga ed by means o decomposi ion ia a mul i a ia e
eweigh ing me hod (Hainmuelle 2012). Finally, we use wo-s ep mul i-le el models (Lewis and Linze 2005) o es he associa ion
be ween he sca e ec s and mac o-economic as well as ins i u ional change.
2. Backg ound
2.1. Empi ical e idence on unemploymen sca ing
Exis ing wo k has shown epea edly ha unemploymen has de imen al e ec s o u u e labo ma ke ou comes o a ec ed
wo ke s in ad anced Wes e n economies. Much o his wo k has ocused on ea nings losses, which ha e been ound o be signi ican
and pe sis en (e.g. Addison and Po ugal 1989; Cha and Mo gan, 2010; Eliason and S o ie 2006; Gangl 2006; G ego y and Jukes
2001; M¨
olle and Umkeh e 2015; Ruhm 1991; Voßeme 2019a). Resea ch has also shown ha unemploymen has a subs an ial
nega i e e ec on u u e employmen p obabili y (e.g. A ulampalam e al. 2000; Biewen and S e es 2010; Gangl 2004, 2008; Voßeme
2019a). These e ec s a e pa ly d i en by a ec ed indi iduals’ di icul y o e-en e ing he labo ma ke , bu a e also due o he jobs
en e ed a e unemploymen being less s able (Gangl 2004) leading o a compa a i ely high isk o epea ed unemploymen spells.
Schmillen and Umkeh e (2017) show ha e en you h unemploymen has long- e m implica ions and inc eases unemploymen
olume o adul wo ke s. Compa ed o ea nings losses and e-employmen p obabili ies, non-pecunia y job-quali y ou comes (e.g.
ype o con ac , au ho i y, au onomy, subjec i e job secu i y) ha e ecei ed conside ably less a en ion. Bu hose who ocused on
hese pos -unemploymen ou comes ha e shown a subs an ial nega i e impac as well (e.g. B and 2006; Voßeme 2019b, Dieckho
2011).
2.2. Theo e ical explana ions o unemploymen sca s
How a e sca e ec s o unemploymen explained? Ea ly accoun s ying o explain why pas unemploymen p edic s u u e
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
3
unemploymen asked whe he unemploymen in i sel causes u u e unemploymen o whe he unobse ed wo ke cha ac e is ics
explain his phenomenon (Heckman and Bo jas 1980). In he mean ime, empi ical esea ch has p esen ed s ong e idence o genuine
s a e dependence (as e iewed in sec ion 2.1), i.e. e idence ha unemploymen i sel is ha m ul and lea es sca s. The ocus has
he e o e shi ed o discussing he unde lying mechanisms o sca e ec s. The e exis a ious heo e ical explana ions o unem-
ploymen sca ing. Some e e o mechanisms loca ed a he le el o he employe and o he s poin o mechanisms ha a e loca ed a
he le el o he unemployed jobseeke .
The explana ions ocusing a he employe le el pe ain o employe s’ pe cep ions o he unemployed: one explana ion is building
on conside a ions o human capi al heo y (Becke 1964). I employe s belie e ha human capi al and skills dep ecia e du ing un-
employmen - especially du ing long spells – his will cause sca ing. Ano he heo e ical mechanism is ha employe s deploy un-
employmen (and/o i s du a ion) mo e gene ally as a signal o unobse ed cha ac e is ics. Unemploymen is belie ed o signal
undesi able wo ke ai s such as low e o , commi men and mo i a ion (Lockwood 1991). This pe spec i e is also e e ed o as he
s igma explana ion (e.g. Viswana h 1989). A di e en a ian o he signaling heo y is he a ional he ding app oach (e.g. Küble and
Weizs¨
acke 2003): longe unemploymen spells indica e o po en ial u u e employe s ha o he ec ui men manage s ha e p e i-
ously al eady decided agains hi ing his applican . Following he judgemen o p io ec ui e s a he han sc u inizing he applican
hemsel es is conside ed o be a ional and e icien .
Some o he cen al mechanisms assumed o d i e sca ing a e loca ed a he le el o he unemployed jobseeke . One po en ial
mechanism is ac ual human capi al loss: p e iously unemployed wo ke s ha e los all i m-speci ic human capi al (Hame mesh 1987),
while hei occupa ion-speci ic o sec o -speci ic human capi al is los i hey do no ob ain e-employmen in hei p e ious occu-
pa ion/sec o . O he models ocus on he job sea ch p ocess. They emphasize ha unemployed wo ke s’ job sea ch akes place unde
inancial cons ain s (Bu de 1979). In addi ion, when ecei ing a job o e , he unemployed can ne e be su e i and when a be e job
o e will a i e in he u u e (Mo ensen 1970). The combina ion o inancial cons ain s and his unce ain y o ces he unemployed
o accep ‘low-quali y’ job o e s. These a e jobs ha do no ma ch hei p e-unemploymen quali ica ions, which o e signi ican ly
lowe wages han hei p e-unemploymen jobs, o which a e o a p eca ious, ixed- e m na u e (e.g. Addison and Blackbu n 2000;
Gangl 2004) and may hus lead o an inc eased isk o epea unemploymen .
2
The e hus exis s qui e a ange o possible explana ions o he phenomenon ha unemploymen lea es sca s. Howe e , di ec
empi ical es s o hese mechanisms a e di icul o cons uc . While some wo k has s a ed ying o de e mine which (demand-side)
mechanisms a e a wo k (e.g. Obe holze -Gee 2008; Van Belle e al. 2018), e idence is s ill sca ce and he ela i e impo ance o he
di e en mechanisms loca ed a he le el o he employe and a he le el o he unemployed job-seeke is s ill an open ques ion.
2.3. The ole o ins i u ions and mac o-economic con ex
While mos wo k on unemploymen sca ing has ocused on single coun ies, some ha e s udied pos -unemploymen ou comes in a
compa a i e pe spec i e and e ealed no able c oss-na ional di e ences in he ex en o unemploymen sca s (Gangl 2004, 2006;
Voßeme 2019b). Some o hese s udies p esen e idence clea ly sugges ing ha labo ma ke and wel a e s a e ins i u ions – spe-
ci ically unemploymen insu ance and employmen p o ec ion – play a cen al ole in mode a ing he de imen al e ec o unem-
ploymen (Gangl 2004, 2006). One accoun , by con as , inds su p isingly li le suppo o he mode a ing e ec o labo ma ke
policies (Voßeme 2019b).
Besides ins i u ional con ex , he ole o mac o-economic clima e has also been examined. Exis ing wo k add essing he ques ion o
how economic clima e a ec s he sca e ec o unemploymen a i ed a di e ging conclusions. Ea lie s udies sugges ha he ex en
o sca ing does no a y be ween imes o economic g ow h and imes o ecession (e.g. Fa be 2005), while mo e ecen analyses ind
sca ing o be mo e se e e du ing ecessiona y pe iods (e.g. Couch e al. 2011). This di e ging e idence may be a ibu able o
changing labo ma ke dynamics o di e en es ima ion s a egies, as conjec u ed by Gonalons-Pons and Gangl (2022: 172). Ve y
ecen wo k has also demons a ed ha ins i u ional se ing and mac o-economic condi ions in e ac in de e mining he ex en o
unemploymen sca ing (ibid).
Exis ing wo k has hus p esen ed impo an e idence ha ins i u ions and mac o-economic clima e ma e o unemploymen
sca ing. I has analyzed he in luence o mac o-le el ac o s by means o a c oss-na ionally compa a i e app oach. Agains his
backd op, ou con ibu ion akes a c oss- empo ally compa a i e app oach ocusing on long- e m de elopmen s and ins i u ional
change in one coun y. I seeks o examine whe he changes in he ins i u ional dimensions deemed ele an by ea lie c oss-na ionally
compa a i e esea ch ha e esul ed in di e en sca ing ou comes in Ge many, while also aking in o accoun mac o-economic
de elopmen . Gi en ha he Ge man labo ma ke has unde gone subs an ial change (as ou lined in de ail in he subsequen sec-
ion), i makes o a pa icula ly ui ul case o a such a long- e m c oss- empo al analysis.
2
Some au ho s ha e also ocused on he psychological e ec s o unemploymen and hese psychological sca s may hen o cou se also con ibu e
o u u e employmen and wage sca ing. Some au ho s ha e pu o wa d he possibili y ha he unemploymen spell may change indi iduals’
a i ude o wo k (Lynch 1989) o shown ha unemploymen can igge a loss o sel -es eem (e.g. Goldsmi h e al., 1996). O he s udies ha e shown
ha unemploymen educes well-being and men al heal h in a long-las ing manne p o iding e idence o a long- e m e ec e en a e
e-employmen (Cla k e al., 2001; Daly and Delaney 2013; Young 2012).
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
4
2.4. The Ge man labo ma ke in ans o ma ion
Fo a long ime, he Ge man labo ma ke se ed as he main exempla o a coo dina ed ma ke economy and a conse a i e wel a e
s a e wi h s ong unions, high le els o coo dina ion, and s ong employmen as well as unemploymen p o ec ion (Hall and Soskice
2001). Howe e , he model came inc easingly unde s ain wi h he beginning o he 1980s and his encou aged he i s a emp s o
de egula e he Ge man labo ma ke , which ook place in he mid-1980s. These i s e o ms ex ended he possibili ies o non-s anda d
employmen (Eichho s and Ma x 2011) and ma ked he beginning o labo ma ke dualiza ion in Ge many (e.g. B ady and Biege
2017; Eichho s e al. 2015). While he ini ial e o ms le he co e wo k o ce un ouched, he go e nmen ini ia ed mo e comp e-
hensi e labo ma ke e o ms ollowing he Ge man e-uni ica ion and he 1993 economic c isis (c . Fig. 1). These measu es included
i s ma ginal cu s o unemploymen insu ance and unemploymen assis ance bene i s (Ebbinghaus and Eichho s , 2009). The libe -
aliza ion p ocesses also included indus ial ela ions wi h unions losing much o hei powe (Eichho s and Ma x 2011).
In he second hal o he 1990s u he , mo e subs an ial, s eps owa ds he de egula ion o he unemploymen bene i sys em ook
place. A he same ime, he waning o union powe and he decline o collec i e ba gaining co e age con inued s eadily (e.g. Visse
2019) du ing he 1990s and he low-wage sec o s a ed g owing subs an ially om 1997 onwa ds (G abka and Sch ¨
ode 2019).
The peak o he libe aliza ion p ocess occu ed in he ea ly-mid-2000s. In he con ex o an economic down u n and a d as ic
upsu ge o unemploymen (see Fig. 1), he so-called Ha z e o ms we e in oduced (Eichho s and Ma x 2011). The di e en com-
ponen s o hese e y ex ensi e labo ma ke e o ms we e in oduced s ep-wise be ween he beginning o 2003 and 2005 (see e.g.
Kemme ling and B u el 2006 p. 93 . o an o e iew and de ailed depic ion o he Ha z legisla ion). The aim o he e o ms was o
educe (long- e m) unemploymen and encou age employmen g ow h.
The ini ial se o e o ms in e alia subs an ially de egula ed empo a y wo k agencies, igh ened eligibili y and job accep ance
egula ions o bene i eceip , eased i ms’ use o ixed- e m con ac s, and encou aged employmen in he low-wage sec o . The inal
e o m, Ha z IV, en ailed ha en i lemen pe iods o ea nings- ela ed unemploymen bene i s we e sho ened no ably. A e bene i
deple ion, unemployed wo ke s ecei e means- es ed minimum income suppo only (Kemme ling and B u el 2006). Receip o his
minimum income suppo is linked o s ic eligibili y c i e ia and inc eased condi ionali y: he long- e m unemployed a e obliged o
accep almos any job (ibid.). This ou h e o m, which d as ically weakened he social policy p inciple o s a us-p o ec ion is o en
cha ac e ized as he mos cen al depa u e om he p inciples o he p e-Ha z Ge man labo ma ke policy (e.g. M¨
olle 2015). Some
a gue ha he iew o he unemployed and unemploymen has changed because o he e o ms, wi h unemploymen inc easingly seen
as being he esul o indi idual beha io and sho -comings a he han a s uc u al economic p oblem (e.g. Bo h eld 2007).
Unemploymen ell d as ically in he yea s a e he inal e o m was implemen ed in 2005, and e en he g ea economic ecession
(2007–2009) wi h i s clea ly isible nega i e impac on GDP g ow h did no en ail a g ow h in unemploymen igu es (see Fig. 1). A e
he ecession unemploymen a es con inued o decline and we e below ou pe cen a he end o ou obse a ion window (c . Fig. 1).
The e is li le consensus, hough, as o whe he his was an e ec o he Ha z e o ms (e.g. M¨
olle 2015) o no (e.g. Dus mann e al.
2014). Union powe and impac con inued o decline subs an ially a e he e o ms (Eichho s and Ma x 2011).
3
The g ow h o he
low-wage sec o , which s a ed in 1997 con inued un il 2008, bu has s agna ed since a a le el o 25 pe cen (G abka and Sch ¨
ode
2019). Following he ecession, Ge man GDP g ow h since 2009 was posi i e each yea – excep in 2020 when i was nega i e (a a a e
o 5 pe cen ) because o he pandemic (c . Fig. 1).
3. Aims and expec a ions
Ou aim is o examine how unemploymen sca s ha e de eloped o e ime and in es iga e whe he o no he e is a ia ion o e
ime o e en a clea ime end owa ds educed o inc eased sca s. We a e also in e es ed in whe he any obse ed ends di e
depending on which ype o sca we look a , i.e. whe he we examine wage ou comes o employmen p obabili ies. I is likely ha he
ins i u ional and s uc u al changes he Ge man labo ma ke expe ienced ha e had epe cussions o he ex en o unemploymen
sca ing. I can also be expec ed ha he le el o sca ing would depend on labo ma ke condi ions and hence a y o e he business
cycle. While he e a e good easons o expec a ia ion o e ime, he di ec ion o he e ec , which s uc u al and ins i u ional change
as well economic clima e may ha e on he se e i y o sca ing is ha d o de e mine a p io i. Any p edic ion hinges on he ac ual causes
o unemploymen sca ing, i.e. on which o he po en ial mechanisms discussed ea lie a e ac ually a wo k o – assuming ha se e al
p ocesses a e a wo k simul aneously – on hei ela i e weigh in de e mining he labo ma ke disad an ages ollowing
unemploymen .
3.1. De egula ion and unemploymen sca ing
Which o he a o e discussed heo e ical mechanisms would p edic educed sca ing in he con ex o con inuous de egula ion and
which ones would ins ead poin o inc eased sca s? Wha would demand-side mechanisms, i.e. hose mechanisms loca ed a he le el o
he employe , p edic ? Assuming ha employe s’ conce n abou human capi al dep ecia ion is cen al in explaining sca e ec s, he
o e all end o de egula ion in he Ge man labo ma ke should ha e esul ed in educed sca s. The libe aliza ion o employmen
egula ion (speci ically also he de egula ion o empo a y wo k) acili a es labo ma ke e-en y o he unemployed wi h employe s
3
In ac , om he beginning un il almos he end o ou obse a ion window, collec i e ba gaining co e age has declined om 85 o 56 pe cen
(Visse 2019).
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
5
being gene ally less conse a i e in hei hi ing beha io due o lowe ed dismissal cos s. This makes i mo e likely ha he unemployed
a e hi ed despi e excep ed human capi al losses. Wi h inc eased de egula ion, he a e age unemployed job-applican should hus
spend less ime in unemploymen han be o ehand. Consequen ly, we would expec inc eased employmen p obabili ies o p e iously
unemployed indi iduals. Mo eo e , i employe s associa e sho e unemploymen spells wi h lowe le els o human capi al dep e-
cia ion, we would also p edic a ime end o educed wage sca s ollowing de egula ion.
Models, which conside he signaling e ec o unemploymen spells o be cen al in de e mining unemploymen sca ing, would
a i e a e y simila p edic ions. In his explana o y app oach, he isk a ached o he hi ing decision (e.g. hi ing wo ke s wi h low
p oduc i i y le els) is he cen al mechanism. Gi en he lowe isk associa ed wi h any hi e in less egula ed labo ma ke s, we would
expec a quicke e-en y o he unemployed ollowing de egula ion. I should also imply educed wage sca s as sho spells o un-
employmen a e a less nega i e signal han long spells o unemploymen . A ocus on he signaling app oach could also lead o he
con a y p edic ion, howe e . In he p ocess o de egula ion, and mos no ably upon he in oduc ion o he Ha z e o ms, he
pe cep ion o unemploymen may ha e changed. Unemploymen may be inc easingly unde s ood as being due o indi idual aul , as
has been a gued by some. I his is indeed he case, hen he nega i e signal a ached o unemploymen will likely ha e become mo e
pe inen o e ime ins ead, leading o he expec a ion o inc eased sca e ec s, bo h in e ms o employmen p obabili ies as well as in
e ms o wages. This would mean ha in e ms o he “signaling mechanism” posi i e and nega i e e o m e ec s could o -se each
o he (see also Voßeme 2019b o a simila a gumen wi h iew o di e en ins i u ional e ec s).
Wha p edic ions abou empo al ends would a ise based on he di e en supply-side explana ions, i.e. hose heo e ical mecha-
nisms loca ed a he le el o he unemployed indi idual? I human capi al explana ions loca ed a he le el o he unemployed job-seeke
ha e s ong explana o y powe , i.e. hose pe aining o ac ual sec o - and occupa ion-speci ic human capi al losses and de alua ion,
de egula ion and bene i e o ms will likely ha e esul ed in mo e p onounced wage sca ing. Inc eased condi ionali y and p essu es
su ounding job sea ch and bene i eceip make i mo e likely ha wo ke s (ha e o) accep jobs in ano he sec o o occupa ion which
means ha hei occupa ion- o sec o -speci ic human capi al los all i s alue and ha hei ( ela i e) wage losses a e likely mo e
p onounced and long-las ing han p io o de egula ion. Highe employmen a es o p e iously unemployed wo ke s hence come a
he p ice o ela i e o e en absolu e wage loss. Likewise, job sea ch models emphasizing he inancial cons ain s and he unce ain y
abou u u e job o e s unemployed job-seeke s a e con on ed wi h would p edic de egula ion, and speci ically he Ha z e o ms, o
ha e esul ed in lowe ed ese a ion wages and hence inc eased employmen a es, bu also inc eased ( ela i e) wage losses.
Mo e gene ally, and i espec i e o any po en ial mic o-le el mechanisms explaining pos -unemploymen ou comes, de egula ion
has also en ailed s uc u al change o e ime wi h inc eased dualiza ion (as de egula ion o employmen p o ec ion mainly and mos
subs an ially a ec ed non-s anda d employmen ), a la ge low-wage sec o (G abka and Sch ¨
ode 2019), highe ea nings dispa i ies
(Ca d e al. 2013; Dus mann e al. 2009) and inc eased employmen le els (OECD 2019). These de elopmen s should ha e made i
easie o e ime o e-en e he labo ma ke a e unemploymen , bu made i ha de o ob ain a job ha is o compa able wo h o he
one held p e-unemploymen , bo h in e ms o wages bu also in e ms o job s abili y. Especially he ise in non-s anda d employmen is
a c ucial ac o he e as i is known o be associa ed wi h lowe wages (Giesecke 2009).
Fig. 1. GDP g ow h and unemploymen a es in Ge many, 1985–2020.
Sou ce: S a is isches Bundesam (Fede al S a is ical O ice) 2022: Fachse ie 18 Reihe 1.5
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
6
3.2. Economic clima e and unemploymen sca ing
Aside om ins i u ional and s uc u al change, economic luc ua ions can also be expec ed o a ec he ex en o sca ing. Again,
any p edic ion abou he di ec ion o his e ec is di icul and hinges on which o he heo e ical mechanisms is ac ually a wo k (see
also Tumino 2015) o – i indeed se e al mechanisms a e a wo k – on hei ela i e weigh . And, as no ed ea lie , empi ical e idence is
mixed. Fo example, he e seems o be no mechanical ela ionship be ween agg ega e unemploymen and he employmen p obabili y
o he unemployed (o he wage ou comes o he p e iously unemployed). One cen al pe spec i e holds ha he nega i e signal o
unemploymen becomes less ele an when agg ega e le els o unemploymen inc ease (e.g. Lockwood 1991) leading o he expec-
a ion o educed sca s in imes when labo ma ke condi ions a e un a o able. A he same ime, ad e se labo ma ke condi ions
educe hi ing le els, which may inc ease he a e age leng h o unemploymen spells and hence esul in mo e se e e sca ing (bo h in
e ms o employmen p obabili ies and also wages). These wo mechanisms migh coun e balance each o he .
Tu ning o he a ionale and beha io o he unemployed job-seeke , in imes o slack demand s/he may be mo e likely o accep job
o e s ou side o hei p e ious sec o o occupa ion (Gangl 2006). This would mean ha his/he speci ic human capi al los all i s
alue. I human capi al explana ions a e cen al in explaining sca ing, job loss and job sea ch when labo ma ke condi ions a e
ad e se should inc ease sca s – especially wage sca s. Job sea ch models would also p edic lowe ese a ion wages in such a scena io
leading o he expec a ion o inc eased wage sca s.
3.3. Declining o Inc easing Sca s o e ime?
As he discussion so a has highligh ed, i is di icul o de elop hypo heses abou ime ends in he se e i y o unemploymen
sca s. The discussion has shown ha he e a e mechanisms whe eby we would expec de egula ion o ha e encou aged ime ends o
educed unemploymen sca s as well as mechanisms, which would lead us o expec he opposi e. The same ambi alence is ue o sca
e ec s in ecessiona y pe iods. Ra he han de eloping and es ing di ec ional hypo heses, ou s udy will p o ide a nuanced explo -
a i e accoun o he empo al de elopmen s o unemploymen sca s in he Ge man labo ma ke s and he unde lying mac o-le el
mechanisms. Mo eo e , he abo e discussion e e ed o mac o-le el de elopmen s leading o possible changes in he es ima ed
a e age sca e ec by a ec ing how dele e ious unemploymen is o u u e ou comes. Ye , mac o-le el de elopmen s could also a ec
he ex en o es ima ed sca s me ely by changing he composi ion o hose who become unemployed (o o hose who e-en e he labo
ma ke ) i he e a e g oup-speci ic sca e ec s. While wo k on g oup-speci ic e ec he e ogenei y is s ill a he sca ce, exis ing e i-
dence clea ly sugges s ha he ex en o sca s ends o a y g oup-speci ically (e.g. B and 2006; Mooi-Reci and Ganzeboom 2015), so
changes in composi ion would be enough o obse e changes in a e age ( ea men ) e ec s wi hou he se e i y o he sca s as such
ha ing changed. Ou s udy will be able o analy ically sepa a e “ ue” changes in sca e ec s om changes ha a e due o compo-
si ional change.
4. Da a and me hod
4.1. Da a and sample
We use he Ge man Socio-Economic Panel (SOEP) om 1984 o 2020, which p o ides na ionally ep esen a i e longi udinal da a
o Ge many (Goebel e . 2019). The ich and annually collec ed SOEP da a wi h mon hly calenda da a on economic ac i i y allow us o
iden i y indi iduals who a e employed a T and en e a spell o unemploymen be ween ime poin s T and T+1. The da a also p o ide
us wi h in o ma ion collec ed p io o ime poin T and subsequen o ime poin T+1. In pa icula , we use in o ma ion s emming om
he las SOEP wa e p io o wa e T (i.e. wa e T-1) and om SOEP wa es wo yea s and ou yea s a e ime poin T (i.e. wa e T+2 and
T+4, espec i ely). Acco dingly, ou sample is comp ised o esponden s, who a e obse ed a leas in ou (six) consecu i e wa es o
he SOEP.
Fu he mo e, ou sample is es ic ed o esponden s aged 25–55 (a ime poin T). This means ha ou s udy ocuses on indi iduals
o p ime wo king age, he eby excluding younge and olde pe sons, whose employmen si ua ions o en a e in luenced and shaped by
speci ic wo k a angemen s such as ea ly e i emen schemes, aining con ac s and ac i e labo ma ke p og ams, which would
equi e a sepa a e analysis. Mo eo e , we es ic ou sample o pe sons who a e in dependen employmen a ime poin T (i.e.,
excluding he sel -employed) and who did no pa icipa e in gene al o oca ional educa ion (e.g. app en ices o uni e si y s uden s).
4
To minimize he impac o s a is ical ou lie s, we also exclude esponden s epo ing hou ly wages lowe han h ee Eu o o highe han
200 Eu o (in cons an p ices o 2015, see below).
5
Finally, o add ess he speci ic economic si ua ion in Eas Ge many a e e-
uni ica ion in 1990, we include esponden s esiding in Eas Ge many only om wa e 1995 onwa ds. The ime be ween 1990 and
1995 was cha ac e ized by a majo es uc u ing o Eas Ge many’s economy and labo ma ke . I hus cons i u es a e y unique and
speci ic si ua ion when i comes o analyzing indi idual labo ma ke ajec o ies. The e o e, his speci ic pe iod would call o a
sepa a e analysis o unemploymen sca ing.
4
Resul s do no subs an ially change i we exclude hose esponden s, who pa icipa e in gene al o oca ional educa ion in T+2 o T+4,
espec i ely.
5
A e aged ac oss all wa es and applying ou sample es ic ions, he sha e o indi iduals wi h wages below h ee Eu o (in cons an p ices o
2015) is less han one pe cen ; he sha e o indi iduals wi h wages abo e 200 Eu o is below 0.1 pe cen .
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
7
We use su ey weigh s h oughou ou analyses o add ess he SOEP’s speci ic sampling s a egies (e.g., disp opo ional sampling o
ce ain social g oups) as well as panel a i ion. In pa icula , we use c oss-sec ional su ey weigh s a ime poin T and mul iply hem
wi h longi udinal weigh s ha cap u e he indi idual p obabili y o s aying in he sample (a T+1, T+2, e c.).
Ou wo ou come a iables a e employmen s a us and wages. These a e measu ed as ollows: Employmen S a us a T+2 and T+4 is
based on sel - epo s o he esponden s. Responden s gain ully employed on ull- ime o pa - ime basis as well as sel -employed
esponden s a e coded as employed, while esponden s a e coded as non-employed i hey epo ed o be ei he unemployed ( egis-
e ed wi h he Fede al Employmen Agency) o economically inac i e. Mo eo e , esponden s pa icipa ing in employmen -c ea ion
measu es o e ed by he Fede al Employmen Agency a e also coded as non-employed.
6
The a iable Wages a T, T+2 and T+4
measu es esponden s’ hou ly g oss wages by combining in o ma ion on g oss mon hly labo ea nings and weekly ac ual wo king
hou s. To keep wages compa able ac oss wa es we use he consume p ice index and calcula e wages in cons an 2015 p ices. Based on
his in o ma ion we calcula e an indi idual’s wage g ow h be ween T and T+2 as well as be ween T and T+4. In ou analyses, we use
log-wages – a common p ac ice when modelling wages as dependen a iable.
7
The key explana o y a iable cap u es in o ma ion o whe he o no a esponden indica ed o ha e en e ed a spell o unem-
ploymen be ween he in e iews in wa e T and in wa e T+1. This in o ma ion is based on e ospec i ely collec ed calenda da a
p o ided by esponden s’ sel - epo s. In his calenda da a, esponden s indica e on a mon hly basis whe he o no hey ha e been
unemployed a any poin om Janua y o Decembe in he yea p io o he cu en in e iew. Combining his in o ma ion ac oss all
a ailable wa es, we a e able o iden i y hose esponden s, who en e ed a spell o unemploymen be ween T and T+1. Ve y sho - e m
unemploymen spells o less han wo mon hs a e no coun ed as “becoming unemployed” as hese sho spells a e likely o e lec a
ansi ion phase om one job o ano he .
8
En e ing unemploymen is no a andom p ocess. We he e o e use an ex ended lis o indi idual cha ac e is ics ha a e known o
be de e minan s o bo h unemploymen and u u e ends in baseline ou comes and condi ion on hese in ou empi ical analyses. These
cha ac e is ics include gende , age, esidence in Eas /Wes Ge many, educa ion, enu e, p e ious unemploymen expe ience, wages,
heal h s a us – all measu ed a ime poin T.
9
Mo eo e , we use in o ma ion abou he indi idual wage g ow h be ween T-1 and T o
accoun o po en ially di e ging wage dynamics o hose becoming unemployed and hose no becoming unemployed ( he so-called
Ashen el e dip, see Ashen el e 1978). We also con ol o changes om un/non-employmen o employmen be ween T-1 and T.
Gi en wo o ou co e sample selec ion c i e ia - esponden s, who a e obse ed a leas in ou (six) consecu i e wa es o he SOEP;
esponden s aged 25–55 (a ime poin T) - he sha e o missing da a is e y low (less han one pe cen ). Agains his backg ound, we
e ain om using impu a ion models o missing da a and use lis wise dele ion ins ead.
10
4.2. Analy ic s a egy
To iden i y and es ima e (causal) e ec s o unemploymen o u u e employmen ou comes, we combine ma ching app oaches
wi h he analysis o di e ence-in-di e ences. In pa icula , we use coa sened exac ma ching (CEM, see Iacus e al. 2012). CEM is a
ma ching me hod which comes close o he ideal o exac ma ching. In con as o exac ma ching, which is o en no empi ically
easible, CEM does no use all a iables in hei o iginal o ma bu uses coa sened e sions when exac ma ching is no ealizable. By
using CEM we assu e ha ea ed (i.e. indi iduals becoming unemployed be ween T and T+1) and con ol (i.e. indi iduals no
becoming unemployed be ween T and T+1) cases a e s ic ly compa able wi h espec o gende and Eas /Wes Ge many esidency as
well as o coa sened p e- ea men in o ma ion (measu ed a T) on wages ( i e ca ego ies), age ( wo ca ego ies), educa ion ( h ee
ca ego ies), and enu e ( wo ca ego ies, see Appendix Table A.1.1 o mo e de ails). Combining all hese cha ac e is ics esul s in a
o al o 240 cells, wi hin which ea ed and con ol cases a e ma ched. I cells con ain only ea ed o only con ol cases, hen hese cells
and he co esponding cases a e disca ded.
11
To ensu e a su icien numbe o ea ed cases, in some ins ances we had o eso o
pooling pai s o adjacen wa es. This was mos ly necessa y in he wa es be o e he mid-1990s as he numbe o SOEP- esponden s (and
6
We decided o do so as we a e in e es ed in sca e ec s pe aining o egula employmen ; job c ea ion is pa o he policy, no he ou come.
Addi ional analyses coun ing hese indi iduals as employed did no change he esul s d as ically and a e a ailable upon eques .
7
Loga i hmiza ion ans o ms igh -skewed wage dispe sion in such a way ha he dependen a iable is app oxima ely no mally dis ibu ed,
which – in e alia – helps educe he e oscedas ici y p oblems. Th ough he ans o ma ion [exp(β
x
)-1]*100 he coe icien s o linea eg ession
models can be in e p e ed as pe cen age change in he (deloga i hmized) dependen a iable, which co esponds mo e o a concep o social
inequali y ha s esses ela i e a he han absolu e di e ences (Pe e sen 1989).
8
Res ic ing he minimum leng h o unemploymen spells e en u he (e.g. o a leas h ee o ou mon hs) esul s in a signi ican d op in he
numbe o cases expe iencing unemploymen . Thus, o main ain a su icien numbe o cases, we decided o apply he a he mild es ic ion o
conside ing spells o a leas wo mon hs.
9
These cha ac e is ics en e he analyses in ei he a coa sened o uncoa sened e sion. Fo mo e de ails see Appendix A1.
10
Only in case o wage in o ma ion we obse e a sligh ly inc eased sha e o missing da a (abou six pe cen ). Howe e , impu ing his wage in-
o ma ion does no seem o be use ul, as we a e employing his wage in o ma ion as dependen a iable (wage g ow h), a leas in one o you wo
model speci ica ions. As shown by on Hippel (2007), impu ing he dependen a iable and using hese impu ed alues in linea models induces
needless noise in he es ima es.
11
By using his app oach, CEM is able o gua an ee common suppo o he ma ched sample. Mo eo e , in ou applica ion, CEM achie ed a close-
o-pe ec balancing be ween ea ed and con ol cases (i.e. wi h espec o he coa sened e sions o he a iables en e ing he ma ching p ocedu e).
Acco ding o all imbalance measu es epo ed by CEM, balance has been achie ed in all SOEP wa es o all o ou ea men speci ica ions (i.e.
employmen and wages a T+2 and T+4, espec i ely).
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
8
o people obse ed as en e ing unemploymen ) was subs an ially lowe han in la e wa es (see Table A.2 o he numbe o ea ed by
yea as well as o in o ma ion on which wa es we e pooled).
Based on hese ma ched da a we hen un eg ession models, which ake emaining imbalances be ween ea ed and con ols in o
accoun . These models con ol o all ma ching a iables ou lined abo e (bu his ime in hei uncoa sened o m) plus in o ma ion on
p io unemploymen expe ience, heal h s a us and p e- ea men wage g ow h o es ima e u u e wages and employmen chances (see
Table A.1.2). In hese eg ession models we eg ess employmen s a us a T+2 and T+4 and (gi en employmen ) wage g ow h be ween
T and T+2/4 on he ea men indica o (ha ing become unemployed o no ) and he ex ended lis o con ol a iables. This se up
mi o s he well-known di e ence-in-di e ences (DID) es ima o (Heckman e al. 1997) ha compa es di e ences in he ou come
a iable be ween ea ed and con ol cases be o e and a e ea men . In ou case, p e- ea men di e ences a e ei he ze o by design
(all cases a e employed a T) o close o ze o by ma ching/condi ioning on p e- ea men le els o he ou come (i.e. p e- ea men
wages measu ed a T) and o he indi idual p e- ea men cha ac e is ics. Acco dingly, causal e ec s o unemploymen can be iden-
i ied and es ima ed by ou app oach i ou models su icien ly condi ion on obse ables (known as condi ional independence
assump ion) and/o we can assume ha – gi en condi ioning on con ol a iables – ends in employmen p obabili y and wages had
been he same o ea ed and con ol cases in he absence o ea men (known as common ends assump ion). Gi en he ex ended lis
o p e- ea men cha ac e is ics ha we condi ion on in ou models as well as ou da a s uc u e and esea ch design we a e con iden
ha a leas one o hese assump ions holds.
12
I is impo an o no e ha based on ou empi ical s a egy we es ima e a e age ea men e ec s on he ea ed (so called ATTs). To
he ex en ha ea men (i.e. sca ing) e ec s a e no homogenous ac oss all ea ed pe sons, changing ATTs ac oss ime migh pu ely
ep esen a changing composi ion o he ea ed popula ion (see also discussion in sub-sec ion Declining o Inc easing Sca s o e Time?).
I , o example, we expec sca e ec s o be di e en o di e en le els o educa ion, ATTs migh change ac oss ime simply due o he
ac ha he ea ed popula ion (i.e. hose who become unemployed) is changing wi h espec o i s educa ional composi ion. To
add ess compa abili y p oblems a ising om he e ogenous ea men e ec s and composi ional changes in he ea ed popula ion, we
ecalcula e and s anda dize ATTs by assuming he same composi ion o he ea ed a e e y ime poin o ou obse a ion pe iod. This
is done by using a mul i a ia e eweigh ing me hod ha was in oduced by Hainmuelle (2012). In pa icula , we calcula e he
“a e age composi ion” o he ea ed popula ion by a e aging dis ibu ions o all ma ching and u he con ol a iables ac oss all
wa es. Nex , we ecalcula e ATTs based on his “a e age composi ion” using he eweigh ing app oach o change he empi ical dis-
ibu ion o each ma ching and con ol a iable o i s a e age coun e pa . These eweigh ed ATTs allow o a be e analysis o ime
ends in unemploymen sca ing as hei calcula ion is based on a ime-cons an composi ion o he ea ed popula ion.
Finally, we di ec ly in es iga e ime ends in unemploymen sca ing and examine he ela ionship be ween mac o-le el change
and unemploymen sca ing. We use he es ima ed ATTs (and hei s anda dized coun e pa s) and eg ess hem on a linea ime end
as well as on an indica o o economic g ow h (g ow h a e o g oss domes ic p oduc a T) and an indica o o cu en labo ma ke
condi ions (unemploymen a e a T). This allows us o pu ge empo al change in unemploymen sca s o economic and labo ma ke
condi ions ha ha e subs an ially a ied o e he obse a ion pe iod (see Fig. 1 abo e) and ha migh be con ounded wi h long- e m
ends in unemploymen sca s. Mo eo e , o explici ly es o a s uc u al b eak in he ime se ies o unemploymen sca s, we include a
a iable ha dis inguishes pe iods be o e he implemen a ion o majo wel a e e o ms (be o e 2003) om pe iods he ea e (i.e.,
om 2003 onwa ds).
To accoun o he ac ha in hese models he dependen a iable (ATTs) con ains es ima ed en i ies we use he associa ed
s anda d e o s o each ATT and es ima e hese eg essions by a easible gene alized leas squa e (FGLS) as sugges ed by Lewis and
Linze (2005). This wo-s ep p ocedu e allows us o in eg a e es ima ion unce ain y a he i s s ep (i.e. es ima ion o ATTs) in o he
second analy ical s ep, i.e. es ima ion o models eg essing unemploymen sca s on ime ends and mac o con ols. Mo eo e , o allow
o a he e oscedas ic and au oco ela ed e o s uc u e we employ he a iance es ima o sugges ed by Newey and Wes (1987). This
es ima o p oduces consis en a iance es ima es when he e is au oco ela ion as well as he e oskedas ici y. Gi en he ime-se ies
na u e o ou second-s ep da a, i is likely ha bo h he e oskedas ici y and au oco ela ion a e p esen in he e o s uc u e. Fo
example, he e oskedas ici y (i.e. non-cons an e o a iance) may esul om he ac ha ou model i s he da a di e en ly well o e
some o e en all yea s o obse a ion. Au oco ela ion would, o example, esul i e o e ms in T (i.e. di e ence be ween measu ed
and p edic ed ATTs) a e co ela ed wi h e o e ms in T+1, T+2, and so o h. In ou es ima ion, we allow o an au oco ela ion o he
o de 1.
These wo-s ep models allow us o in es iga e ime ends mo e di ec ly and o ela e hem o ins i u ional and mac o-economic
change. Howe e , wi h he da a and esea ch design a hand, we canno es ima e he causal e ec o mac o-le el change and espe-
cially we canno p o ide a igo ous e alua ion o he labo ma ke e o ms ei he .
12
Bo h assump ions a e, un o una ely, empi ically un es able in a igo ous manne . Howe e , in he appendix, we p esen some esul s o es ing
o common ends in employmen ajec o ies be ween ea men and con ol g oup p io o T (see Figu e A2). Mo eo e , o check he obus ness o
ou esul s, we e- an all models using h ee al e na i e ma ching app oaches (mul i a ia e-dis ance ma ching, p opensi y-sco e ma ching, and
in e se p obabili y weigh ing). We a e able o show ha esul s based on hese al e na i e ma ching app oaches a e e y simila o he esul s ha
a e based on CEM and conclusions d awn om hese models would no be di e en om ou conclusions epo ed in he main ex o he pape .
De ailed esul s o he obus ness checks can be ound in Appendix B.
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
15
Howe e , i is impo an o no e, ha all o ou esul s on wage sca s (i.e. T+2 and T+4) clea ly speak agains he no ion o educed
wage sca s o e ime.
Finally, in line wi h ou analyses ocusing on employmen p obabili ies, he wage sca models sugges ha inc eased unemploy-
men a es seem o inc ease wage sca ing ( hough only in he models wi hou eweigh ing). Wage sca s a e es ima ed o inc ease by
abou one pe cen age poin i he unemploymen a e inc eases by one pe cen age poin . In addi ion, GDP g ow h is associa ed wi h
educed wage sca s, bu only a T+2 ( hough only in he models wi hou eweigh ing).
6. Discussion
Ou s udy o unemploymen sca ing in he Ge man labo ma ke complemen s and ex ends p e ious esea ch by ocusing on he
empo al a ia ion o unemploymen sca s om a mac o-le el pe spec i e. I examines he ole ins i u ional and s uc u al changes
may play in his a ia ion. We conclude his a icle by highligh ing and discussing h ee cen al indings o ou analyses as well as he
esul ing implica ions o heo y de elopmen , u u e esea ch and policy-making.
Fi s , ou esul s clea ly indica e ha sca s o unemploymen show empo al a ia ion. This is mos e iden in case o he e-
employmen chances o he unemployed ha ha e subs an ially imp o ed h oughou ou obse a ion window. Ou analysis o his
empo al a ia ion e eals ha , alongside a mode a e in luence o he gene al labo ma ke condi ions (i.e. unemploymen a e), he
s ong and las ing imp o emen o indi iduals’ e-employmen chances appea ed o be associa ed wi h he majo wel a e and labo
ma ke e o ms ha ook place in Ge many in he ea ly 2000s. Al hough ou s udy does no cons i u e a igo ous e alua ion o hese
e o ms and mo e gene ally is no able o es ima e he causal e ec o mac o-le el change, he empi ical e idence sugges s a subs an ial
associa ion be ween hese e o ms and he e-employmen chances o he unemployed. While imp o ed e-employmen chances migh
be conside ed a posi i e de elopmen , he e migh also be a downside o i : Pushing he unemployed back in o he labo ma ke mo e
quickly can inc ease he isk o job-misma ches and subs an ial wage losses upon e-employmen . Indeed, we ind no indica ion o
declining wage sca s. Thus, e en i he Ge man wel a e and labo ma ke e o ms seem o ha e imp o ed e-employmen chances o he
unemployed, spells o unemploymen a e s ill associa ed wi h indi idual wage losses. E en mo e, he e is some e idence ha wage
sca s ha e g own o e ime. These esul s migh be in e p e ed as mi o ing he p oblem o job sea ch unde (inc eased) inancial
cons ain s and he (inc eased) isk o subsequen lowe job quali y – a inding ha complemen s he esul s epo ed in p e ious
esea ch (Gangl 2006). Wi h iew o he heo e ical explana ions o unemploymen sca ing, his inding sugges s ha mechanisms
loca ed a he le el o he unemploymen job-seeke (speci ically hei job sea ch beha io and hei job o e accep ance beha io )
p obably ha e a lo o weigh in explaining sca e ec s – especially hose pe aining o wages.
Second, while ou indings indica e ha i migh be wo hwhile o conside empo al a ia ion in he ex en o unemploymen sca s,
we heo e ically discussed and empi ically demons a ed he impo ance o conside ing he (changing) composi ion o hose who
become unemployed as well as o hose who e-en e he labo ma ke whene e we wan o compa e unemploymen sca s o e ime
(o ac oss space). Because o po en ial g oup-speci ic ea men e ec s, no aking in o accoun changes in he composi ion o he
unemploymen in low and ou low popula ion could gene a e misleading in e p e a ions o ac ual ends in a e age unemploymen
sca s. Ou esul s showed ha in e ms o he in low popula ion and hei e-employmen chances, adjus ing he analyses o
composi ional change does no al e he esul s in any subs an ial way. Howe e , accoun ing o he changing composi ion o he
ou low popula ion u ned ou o be impo an o he analysis o ends in sca s on wages. The esul s indica e ha i he composi ion
o he unemployed who we e able o ind e-employmen had been s able o e he whole obse a ion pe iod, wage sca s would ha e
g own a he no ably o e ime. In u n his sugges s ha he composi ion o he unemployed who we e able o e u n o he labo
ma ke changed mo e subs an ially o e ime han he composi ion o hose becoming unemployed. Mo eo e , as composi ional
changes would no ma e i unemploymen sca s had he same magni ude o all indi iduals o he ou low popula ion, he esul s also
poin o he exis ence o g oup-speci ic sca e ec s o unemploymen . These indings unde line he impo ance o conside ing g oup-
speci ic e ec s (o mo e echnically: he e ogenous ea men e ec s) as well as composi ional changes in he analysis o empo al
a ia ion o unemploymen sca s. Fu he mo e, hese indings mo e gene ally unde sco e he impo ance o e ec he e ogenei ies o
ad ancing ou heo e ical unde s anding o unemploymen sca ing. Ad ancing ou insigh s in o he unde lying mechanisms ha
d i e unemploymen sca ing (i.e. in o he causes o obse ed e ec s) is only possible i hese e ec he e ogenei ies a e i mly in e-
g a ed in ou heo e ical models and heo e ical unde s anding.
Thi d, e en i e-employmen chances ha e imp o ed o e ime, he e s ill emains a subs an ial gap in he employmen p oba-
bili ies o people wi h and wi hou p io unemploymen . In ligh o his and also in ligh o he pe sis ing wage sca s, unemploymen
hence con inues o be a e y de imen al li e-cou se e en . Re o ms like hose o he ea ly 2000s in Ge many may end up being
ambi alen success s o ies i hey a e no complemen ed by a ge ed policies ha suppo speci ic g oups among he unemployed
popula ion. Such a ge ed policies would ha e o e lec empi ical e idence on he he e ogenei y o he unemployed popula ion as well
as on g oup-speci ic unemploymen sca s.
Un o una ely, esea ch o da e has no paid su icien a en ion o he he e ogenei y o he unemployed and he ques ion o how he
implica ions o unemploymen a y ac oss di e en socio-demog aphic g oups, li e cou se s ages and labo ma ke segmen s. Mos
wo k has ocused on es ima ing unemploymen ou comes o he o al popula ion o he unemployed. One eason o his sca ci y is he
ela i ely high equi emen s on da a s uc u e and quali y ha a e needed o in es iga e such g oup di e ences (longi udinal da a,
la ge numbe o cases e c.). Mos longi udinal su eys do no p o ide a su icien numbe o cases o allow o nuanced g oup-speci ic
analyses o he unemployed. Fu he mo e, in-dep h esea ch is needed o be e unde s and he dynamics and mechanisms unde lying
he p ocess o unemploymen sca ing. We ocused on pos -unemploymen labo ma ke ou comes measu ed a disc e e ime poin s. To
ully g asp he dynamics o unemploymen disad an age and i s implica ions o indi idual li e-cou ses, analyses ocusing on long e m
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
16
pos -unemploymen employmen and ca ee ajec o ies a e necessa y. This would also allow o a be e assessmen o he Ge man
wel a e e o ms. Ou analysis sugges s ha he e o ms a e associa ed wi h inc eased sho - and mid- e m e-employmen chances, bu
no wi h lowe ed wage penal ies upon e-employmen . In es iga ing long- e m wage and employmen dynamics would help us o
be e judge whe he hese e o ms can o e all be assessed as posi i e o nega i e om a wel a e pe spec i e. I such analyses ind ha
long- e m employmen ajec o ies a e mo e uns able han p e iously and wage losses a e pe sis en , hen om a wel a e pe spec i e
he e o ms should be assessed as nega i e.
Decla a ion o compe ing in e es
None.
Acknowledgemen s
We would like o hank he wo anonymous e iewe s as well as he edi o o Social Science Resea ch, S ephanie Molle , o hei
insigh ul eedback.
Appendix A & B. Supplemen a y da a
Supplemen a y da a o his a icle can be ound online a h ps://doi.o g/10.1016/j.ss esea ch.2023.102960.
Re e ences
Addison, John T., Blackbu n, Mc Kinley, L., 2000. The e ec s o unemploymen insu ance on pos unemploymen ea nings. Lab. Econ. 7 (1), 21–53.
Addison, John T., Po ugal, Ped o, 1989. Job displacemen , ela i e wage changes, and du a ion o unemploymen . J. Labo Econ. 7 (3), 281–302.
A ulampalam, Wiji, Boo h, Alison L., Taylo , Ma k P., 2000. Unemploymen pe sis ence. Ox . Econ. Pap. 52 (1), 24–50.
Ashen el e , O ley, 1978. Es ima ing he e ec o aining p og ams on ea nings. Re . Econ. S a . 60 (1), 47–57.
Becke , Ga y S., 1964. Human Capi al: A Theo e ical and Empi ical Analysis, wi h Special Re e ence o Educa ion. Na ional Bu eau o Economic Resea ch, New Yo k.
Biewen, Ma in, S e es, Susanne, 2010. Unemploymen pe sis ence: is he e e idence o s igma e ec s? Econ. Le . 106 (3), 188–190.
Bo h eld, Silke, 2007. Labou Ma ke Ins i u ions in Ge many: Cu en S a us and Ongoing Re o ms. WSI Discussion Pape 152 pp. 1–36.
B ady, Da id, Biege , Thomas, 2017. The ise o p eca ious employmen in Ge many. Res. Sociol. Wo k 31, 245–271.
B and, Jennie E., 2006. The e ec s o job displacemen on job quali y: indings om he Wisconsin Longi udinal S udy. Res. Soc. S a i . Mobil. 24 (3), 275–298.
Bu de , Kenne h, 1979. Unemploymen insu ance paymen s as a sea ch subsidy: a heo e ical analysis. Econ. Enqui y 17 (3), 333–343.
Ca d, Da id, Heining, J¨
o g, Kline, Pa ick, 2013. Wo kplace he e ogenei y and he ise o Wes Ge man wage inequali y. Q. J. Econ. 128 (3), 967–1015.
Cha, Youngjoo, Mo gan, S ephen L., 2010. S uc u al ea nings losses and be ween-indus y mobili y o displaced wo ke s, 2003–2008. Soc. Sci. Res. 39, 1137–1152.
Cla k, And ew E., Geo gellis, Yannis, San ey, Pe e , 2001. Sca ing: he psychological impac o pas unemploymen . Economica 68 (270), 221–241.
Couch, Kenne h A., Jolly, Nicholas A., Placzek, Dana W., 2011. Ea nings losses o displaced wo ke s and he business cycle: an analysis wi h adminis a i e da a. Econ.
Le . 111 (1), 16–19.
Daly, Michael, Delaney, Liam, 2013. The sca ing e ec o unemploymen h oughou adul hood on psychological dis ess a age 50: es ima es con olling o ea ly
adul hood dis ess and childhood psychological ac o s. Soc. Sci. Med. 80, 19–23.
Dieckho , Ma ina, 2011. The e ec o unemploymen on subsequen job quali y in Eu ope: A compa a i e s udy o ou coun ies. Ac a Sociol. 54 (3), 233–249.
DiP e e, Thomas A., McManus, Pa icia A., 2000. Family change, Employmen T ansi ions, and he wel a e s a e: household income dynamics in he Uni ed S a es and
Ge many. Am. Socio. Re . 65, 343–370.
Dus mann, Ch is ian, Luds eck, Johannes, Sch¨
onbe g, U a, 2009. Re isi ing he Ge man wage s uc u e. Q. J. Econ. 124 (2), 843–881.
Dus mann, Ch is ian, Fi zenbe ge , Be nd, Sch¨
onbe g, U a, Spi z-Oene , Alexand a, 2014. F om sick man o Eu ope o economic supe s a : Ge many‘s esu gen
economy. J. Econ. Pe spec . 28 (1), 167–188.
Ebbinghaus, Be nha d, Eichho s , We ne , 2009. Employmen egula ion and labo ma ke policy in Ge many, 1991-2005. In: de Bee , Paul, Schils, T udie (Eds.), The
Labou Ma ke T iangle Employmen P o ec ion, Unemploymen Compensa ion and Ac i a ion in Eu ope. Edwa d Elga , Chel enham, pp. 119–144.
Eichho s , We ne , Ma x, Paul, 2011. Re o ming Ge man labou ma ke ins i u ions: a dual pa h o lexibili y. J. Eu . Soc. Pol. 21 (1), 73–87.
Eichho s , We ne , Ma x, Paul, Tobsch, Ve ena, 2015. Non-s anda d employmen ac oss occupa ions in Ge many: he ole o eplaceabili y and labou ma ke
lexibili y. In: Eichho s , We ne , Ma x, Paul (Eds.), Non-S anda d Employmen in Pos -Indus ial Labou Ma ke s. An Occupa ional Pe spec i e. Edwa d Elga ,
Chel enham, pp. 29–51.
Eliason, Ma cus, S o ie, Donald, 2006. Las ing o la en sca s? Swedish e idence on he long- e m e ec s o job displacemen . J. Labo Econ. 24 (4), 831–856.
Fa be , Hen y S., 2005. Wha do we know abou job loss in he Uni ed S a es? E idence om he displaced wo ke s su ey, 1984-2004”. Econ. Pe spec . 29, 13–28.
Gangl, Ma kus, 2004. Wel a e s a es and he sca e ec s o unemploymen : a compa a i e analysis o he Uni ed S a es and Wes Ge many. Am. J. Sociol. 109 (6),
1319–1364.
Gangl, Ma kus, 2006. Sca E ec s o Unemploymen : an assessmen o ins i u ional complemen a i ies. Am. Socio. Re . 71 (6), 987–1013.
Gangl, Ma kus, 2008. Unemploymen and wo ke ca ee p ospec s: a c oss-na ional compa ison. In: Mu els, Ruud (Ed.), Flexibili y and Employmen Secu i y in
Eu ope: Labou Ma ke s in T ansi ion. Elga , Chel enham, pp. 169–194.
Gebel, Michael, 2016. Educa ion, employmen dynamics and he middle class in Ge many. Sociological P oblems 48, 23–45.
Giesecke, Johannes, 2009. Socio-economic isks o a ypical employmen ela ionships. E idence om he Ge man labou ma ke . Eu . Sociol. Re . 25 (6), 629–646.
Goebel, Jan, G abka, Ma kus M., Liebig, S e an, K oh, Ma in, Rich e , Da id, Sch ¨
ode , Ca s en, Schupp, Jü gen, 2019. “The Ge man socio-economic panel (SOEP).”.
J. Econ. S a is ics 239 (2), 345–360.
Goldsmi h, A hu H., Veum, Jona han R., Da i y J ., William, 1996. The psychological impac o unemploymen and joblessness. J. Soc. Econ. 25 (3), 333–358.
Gonalons-Pons, Pila , Gangl, Ma kus, 2022. Regula ed ea nings secu i y: he ela ionship be ween employmen p o ec ion and unemploymen sca ing o e he G ea
Recession. Soc. Econ. Re . 20 (1), 169–198.
G abka, Ma kus M., Sch ¨
ode , Ca s en, 2019. The low-wage sec o in Ge many is la ge han p e iously assumed. DIW Weekly Repo 14, 117–124.
G egg, Paul, Tominey, Emma, 2005. The wage sca om male you h unemploymen . Lab. Econ. 12, 487–509.
G ego y, Ma y, Jukes, Robe , 2001. Unemploymen and subsequen ea nings: es ima ing sca ing among B i ish men 1984-94. Econ. J. 111 (475), 607–625.
Hainmuelle , Jens, 2012. En opy balancing o causal e ec s: a mul i a ia e eweigh ing me hod o p oduce balanced samples in obse a ional s udies. Poli . Anal.
20, 25–46.
M. Dieckho and J. Giesecke
Social Science Resea ch 121 (2024) 102960
17
Hall, Pe e A., Soskice, Da id, 2001. An in oduc ion o a ie ies o capi alism. In: Hall, Pe e A., Da id Soskice (Eds.), Va ie ies o Capi alism. The Ins i u ional
Founda ions o Compa a i e Ad an age. Ox o d Uni e si y P ess, Ox o d, pp. 1–68.
Hame mesh, Daniel S., 1987. The cos s o wo ke displacemen . Q. J. Econ. 102 (1), 51–75.
Heckman, James J., Bo jas, Geo ge T., 1980. Does unemploymen cause u u e unemploymen ? De ini ions, ques ions and answe s om a con inuous ime model o
he e ogenei y and s a e dependence. Economica 47 (187), 247–283.
Heckman, James J., Ichimu a, Hidehiko, Todd, Pe a E., 1997. Ma ching as an econome ic e alua ion es ima o : e idence om e alua ing a job aining p og amme.
Re . Econ. S ud. 64 (4), 605–654.
Iacus, S e ano M., King, Ga y, Po o, Giuseppe, 2012. Causal in e ence wi hou balance checking: coa sened exac ma ching. Poli . Anal. 20 (1), 1–24.
Kemme ling, Achim, B u el, Oli e , 2006. ’New Poli ics’ in Ge man labou ma ke policy? The implica ions o he ecen Ha z e o ms o he Ge man wel a e s a e.
W. Eu . Poli . 29 (1), 90–112.
Küble , Do o hea, Weizs¨
acke , Geo g, 2003. In o ma ion cascades in he labou ma ke . J. Econ. 80, 211–229.
Lewis, Je ey B., Linze , D ew A., 2005. Es ima ing eg ession models in which he dependen a iable is based on es ima es. Poli . Anal. 13, 345–364.
Lockwood, Ben, 1991. In o ma ion ex e nali ies in he labou ma ke and he du a ion o unemploymen . Re . Econ. S ud. 58 (4), 733–753.
Lynch, Lisa M., 1989. The you h labo ma ke in he eigh ies: de e minan s o Re-employmen p obabili ies o Young men and women. Re . Econ. S a . 71, 37–45.
M¨
olle , Joachim, 2015. Did he Ge man model su i e he labou ma ke e o ms? J. Labou Ma ke Resea ch 48, 151–168.
M¨
olle , Joachim, Umkeh e , Ma hias, 2015. A e he e long- e m ea nings sca s om you h unemploymen in Ge many? J. Econ. S a is ics 235 (4+5), 474–498.
Mooi-Reci, I ma, Ganzeboom, Ha y B., 2015. Unemploymen sca ing by gende : human capi al dep ecia ion o s igma iza ion? Longi udinal e idence om The
Ne he lands 1980-2000. Soc. Sci. Res. 52, 642–658.
Mo ensen, Dale T., 1970. Job sea ch, he du a ion o unemploymen and he Phillips cu e. Am. Econ. Re . 60 (5), 847–862.
Nagl, Wol gang, Webe , Michael, 2014. Unemploymen compensa ion and unemploymen du a ion be o e and a e he Ge man Ha z IV e o m. I o Wo king Pape
No. 186.
Newey, Whi ney K., Wes , Kenne h D., 1987. A simple, posi i e semi-de ini e, he e oskedas ici y and au oco ela ion consis en co a iance ma ix. Econome ica 55,
703–708.
Obe holze -Gee, Felix, 2008. Nonemploymen s igma as a ional he ding: a ield expe imen . J. Econ. Beha . O gan. 65 (1), 30–40.
OECD, 2019. OECD Labou Fo ce S a is ics 2019. OECD Publishing, Pa is. h ps://doi.o g/10.1787/g2g9 b3e-en. (Accessed 10 Sep embe 2021).
Pe e sen, T ond, 1989. The ea nings unc ion in sociological s udies o ea nings inequali y: unc ional o m and hou s wo ked. Res. Soc. S a i . Mobil. 8, 221–250.
Ruhm, Ch is ophe J., 1991. A e wo ke s pe manen ly sca ed by job displacemen s? Am. Econ. Re . 81, 319–324.
Schmillen, Achim, Umkeh e , Ma hias, 2017. The sca s o you h: e ec s o ea ly-ca ee unemploymen on u u e unemploymen expe ience. In . Lab. Re . 156 (3–4),
465–494.
Tumino, Albe o, 2015. The sca ing e ec o unemploymen om he ea ly 90s o he g ea ecession. ISER Wo king Pape Se ies 05/2015. Ins i u e o Social and
Economic Resea ch.
Van Belle, E a, Di S asio, Valen ina, Cae s, Ral , De Couck, Ma ijke, Bae , S ijn, 2018. Why a e employe s pu o by long spells o unemploymen ? Eu . Socio Re . 34
(6), 694–710.
Visse , Jelle, 2019. ICTWSS Da abase. Ve sion 6.1. Ams e dam Ins i u e o Ad anced Labou S udies (AIAS): Uni e si y o Ams e dam, Ams e dam. h ps://www.
ic wss.o g/downloads. (Accessed 15 Oc obe 2022).
Viswana h, Ta a, 1989. Job sea ch, s igma e ec , and escape a e om unemploymen . J. Labo Econ. 7 (4), 487–502.
on Hippel, Paul T., 2007. Reg ession wi h missing ys: an imp o ed s a egy o analyzing mul iply impu ed da a. Socio. Me hodol. 37, 83–117.
Voßeme , Jonas, 2019a. Losing s anda d employmen in Ge many: he consequences o displacemen and dismissal o wo ke s’ subsequen ca ee s. Res. Soc. S a i .
Mobil. 63, 1–11.
Voßeme , Jonas, 2019b. The e ec s o unemploymen on non-mone a y job quali y in Eu ope: he mode a ing ole o economic si ua ion and labo ma ke policies.
Soc. Indica . Res. 144 (1), 379–401.
Wagne , Ge G., F ick, Joachim R., Schupp, Jü gen, 2007. The Ge man Socio-Economic Panel S udy (SOEP) −Scope, e olu ion and enhancemen s. Schmolle s Jah b.
127 (1), 139–169.
Young, C is obal, 2012. Losing a job: he non-pecunia y cos o unemploymen in he Uni ed S a es. Soc. Fo ces 91 (2), 609–634.
M. Dieckho and J. Giesecke
