B üggmann, Daniel; K eyen eld, Michaela
A icle — Published Ve sion
Ea nings T ajec o ies A e Di o ce: The Legacies o he
Ea ne Model Du ing Ma iage
Popula ion Resea ch and Policy Re iew
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: B üggmann, Daniel; K eyen eld, Michaela (2023) : Ea nings T ajec o ies A e
Di o ce: The Legacies o he Ea ne Model Du ing Ma iage, Popula ion Resea ch and Policy Re iew,
ISSN 1573-7829, Sp inge Ne he lands, Do d ech , Vol. 42, Iss. 2,
h ps://doi.o g/10.1007/s11113-023-09756-4
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/311024
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/
Vol.:(0123456789)
Popula ion Resea ch and Policy Re iew (2023) 42:23
h ps://doi.o g/10.1007/s11113-023-09756-4
1 3
ORIGINAL RESEARCH
Ea nings T ajec o ies A e Di o ce: The Legacies
o  heEa ne Model Du ing Ma iage
DanielB üggmann1· MichaelaK eyen eld2
Recei ed: 17 Janua y 2022 / Accep ed: 11 No embe 2022 / Published online: 2 Ma ch 2023
© The Au ho (s) 2023
Abs ac
Di o ce ma ks he legal endpoin o a ma i al union. While di o ce is inc easingly
seen as a ‘clean b eak’, he pas ma i al his o y o he couple may ne e heless shape
hei p esen condi ions. In pa icula , he e may be a legacy o a highly gende ed
di ision o labou du ing ma iage ha may a ec he ex-spouses’ ea ning ajec-
o ies beyond he da e o di o ce. Using egis e da a om he Ge man Pension
Fund, we examine he ea ning ajec o ies o he e osexual couples who iled o a
di o ce in 2013 (24,616 men and 24,616 women). Using ixed-e ec s and ma ching
echniques, we compa e he ea ning ajec o ies o di o cees wi h hose o a con ol
g oup o ma ied pe sons in he pe iod spanning wo yea s be o e and wo yea s a e
di o ce. In pa icula , we examine how he ea ne models di o cees ollowed du ing
ma iage shaped hei u u e ea ning ajec o ies. Ou esul s show ha , on a e age,
he ea nings o a di o ced woman in a male b eadwinne cons ella ion inc eased
a e di o ce, while he ea nings o he male ex-spouse declined. Ne e heless, la ge
gende di e ences in ea nings pe sis ed: 2yea s a e sepa a ion, a di o ced woman
who had been in a male b eadwinne cons ella ion was, on a e age, ea ning 72%
less han he ex-spouse. We discuss ou indings agains he backg ound o ecen
policy e o ms in Ge many, which assume ha ex-pa ne s should be economically
‘sel - elian ’ a e di o ce.
Keywo ds Di o ce· Employmen · Fixed-e ec s· Ea nings· Ge many· Ma ching
* Michaela K eyen eld
k eyen eld@he ie-school.o g
1 Ge man Pension Fund, Be lin, Ge many
2 He ie School, Be lin, Ge many
D.B üggmann, M.K eyen eld
1 3
23 Page 2 o 34
In oduc ion
A di o ce o sepa a ion is a u ning poin in a pe son’s li e cou se. As well as ma k-
ing he endpoin o a oman ic ela ionship, a di o ce gene ally means ha he cou-
ple will no longe be li ing in a join household (Thomas e al., 2017). The pa -
ne s’ esou ces a e no longe pooled, and hei asse s a e gene ally spli (Boe ien &
Le sch, 2021). In ecen yea s, go e nmen s a e inc easingly adop ing he concep
o a ‘clean b eak’, in which he pa ne s’ inancial obliga ions and claims e mina e
upon di o ce (Ma iny, 2012; Miles & Sche pe, 2020). Ge many is also mo ing in
his di ec ion. In 2008, he Ge man go e nmen enac ed a ‘main enance e o m’,
which emphasised he economic independence o bo h pa ies a e di o ce. The
e o m bill e e s o he p inciple o sel - eliance (‘Eigen e an wo ung’), which
implies ha he di o cees should ea n hei own income, and should no longe ely
on alimony paymen s om an ex-spouse o subsis ence. While ex-spousal suppo
awa ds we e ai ly gene ous p io o his e o m, he cou s now g an suppo in
excep ional cases only (such as when a cus odial di o cee has child en unde age
h ee). This e o m was mo i a ed by he belie ha in Ge many, he compa ibili y
o child ea ing and employmen has imp o ed in ecen yea s, and he di ision o
labou has become mo e equal wi hin couple households.1
Howe e , despi e he e minal na u e o a di o ce, he ex-pa ne s sha e a join
pas ha may con inue o be ele an o e ime. The ex-spouses’ pas decisions—
such as he decision ha one o he pa ne s would educe his/he labou ma ke
pa icipa ion o ocus on aking ca e o he couple’s child en—canno be undone,
bu hey con inue o in luence he ex-pa ne s’ employmen and ca ee op ions a e
di o ce. This pape uses da a om he Ge man pension egis e s o examine how he
ea ne models di o cees ollowed du ing ma iage a ec ed hei employmen ajec-
o ies a e di o ce. We analyse he ea nings and ma i al his o ies o 24,616 he e o-
sexual ex-couples (24,616 men and 24,616 women), all o whom iled o di o ce
in he yea 2013. We ha e selec ed ha pa icula yea because i is ecen enough
o allow us o s udy he beha iou o he membe s o a di o ce coho who we e
sepa a ed a e he enac men o he abo emen ioned e o m, which assumes ha
bo h pa ies can be economically independen a e di o ce. We did no selec a la e
di o ce coho because he obse a ion window o examining hei pos -di o ce
beha iou would ha e been mo e limi ed.
Ou s udy adds o he la ge body o longi udinal esea ch ha has in es iga ed
he economic ami ica ions o di o ce and sepa a ion (Bonne e al., 2021; Hogen-
doo n, 2022; Bu khause e  al., 1991; Damme e  al., 2009; Duncan & Ho man,
1985; Kalmijn, 2005; Boe ien & Le sch, 2021; McManus & DiP e e, 2001; Raz-
Yu o ich, 2013; Thielemans & Mo elmans, 2019; Weiss, 1984). These s udies ha e
shown ha a e a di o ce o sepa a ion, women end o expand hei employmen
ac i i ies, and o en ea n mo e han hey did while ma ied, pa icula ly in coun-
ies such as Ge many (B öckel & And eß, 2015; Damme e al., 2009). Howe e , he
1 The d a law speaks o ‘geände e Rollen e eilung inne halb de Ehe’ (changing di ision o labou
wi hin ma iage) (h ps:// dse e . bunde s ag. de/ b d/ 16/ 018/ 16018 30. pd ).
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 3 o 34 23
esea ch indings on he e ec s o di o ce o sepa a ion on men’s employmen and
ea nings ha e been less conclusi e han hose o women (Co izzi, 2008; Kalmijn,
2005; McManus & DiP e e, 2001). Mo eo e , only a ew s udies ha e examined
how he ea ne models di o cees ollowed du ing ma iage a ec s hei subsequen
li e cou ses (see, howe e , Bonne e al., 2021; Hogendoo n, 2022).
This pape con ibu es o he li e a u e in se e al ways. Fi s , we use la ge-scale
egis e da a o examine how specialisa ion wi hin ma iage a ec ed he ex-spouses’
ea ning ajec o ies a e di o ce. Second, we con ibu e o he spa se li e a u e on
men’s beha iou a e di o ce by compa ing he ea nings o women and men who
we e p e iously ma ied. Thi d, while mos p io s udies on his opic ha e been
based on sepa a e samples o women and men, we go beyond p e ious esea ch by
using couple da a, i.e. we ollow ex-couples o e ime. Thus, he da a p o ide us
wi h a p ecise accoun o he gende pay gap by compa ing a di o ced woman’s
ea nings wi h ha o he ex-husband a he same momen in ime.
Al hough he use o egis e da a has many ad an ages, i also has se e al disad-
an ages ha should be men ioned up on . While we ha e p ecise longi udinal ea n-
ings eco ds ha allow us o compa e ea nings be o e and a e di o ce on he couple
le el, we do no ha e any in o ma ion on whe he he ea nings emana ed om pa -
ime o ull- ime employmen . Mo eo e , while we ha e p ecise ma i al his o ies o
he di o cees, we do no ha e his in o ma ion o he ‘ne e di o ced’. This may
be ega ded as a mino conce n, as he analysis uses a ixed-e ec s app oach ha
elies on he wi hin a ia ion o indi iduals who unde wen a di o ce. As we will
demons a e la e in he pape , he lack o a sui able ‘con ol g oup’ could s ill se i-
ously bias he ixed-e ec s models. Fo ha eason, we de o e conside able a en ion
o selec ing a sui able con ol o his in es iga ion. We gene a e i h ough he use
o ma ching echniques. Thus, he empi ical analysis is based on a combina ion o
ixed-e ec s and ma ching echniques ( o a simila s a egy, see A khangelsky &
Imbens, 2019; Hogendoo n, 2022; Jones & Lewis2015).
P io esea ch
Di o ce, Sepa a ion, andHousehold Income
In ecen decades, he body o schola ly li e a u e on he economic consequences o
di o ce and sepa a ion has g own subs an ially. I was, in pa icula , he inc easing
a ailabili y o longi udinal su eys since he 1980s ha boos ed esea ch in his a ea
(Bu khause e al., 1991; Duncan & Ho man, 1985; Weiss, 1984). Mos o hese
ea lie s udies we e conduc ed by schola s wi h a s ong in e es in po e y esea ch.
As a esul , hese au ho s we e less conce ned wi h how di o ce a ec s indi idual
employmen ca ee s and ea nings, and we e mo e in e es ed in examining he ex en
o which household income and go e nmen ans e s amelio a e some o he nega-
i e consequences o di o ce and sepa a ion. A consis en inding o his body o
esea ch is ha women ypically expe ience a sha p decline in equi alen household
income a e union dissolu ion (Bu khause e al., 1991). While he obse a ion ha
di o ce has ad e se e ec s on women’s economic s anding is di icul o dispu e, he
D.B üggmann, M.K eyen eld
1 3
23 Page 4 o 34
e idence on he e ec s o di o ce on men’s economic well-being is mo e ambi -
alen . Schola s ha e ound ha in some coun ies, including in Ge many, men’s
equalised household income ends o inc ease subs an ially a e di o ce (And eß
e al., 2006).
Conce ns ha e been aised ha some o hese indings may be sensi i e o he use
o equi alen scales ha s anda dise household income (Boll & Schülle , 2021; Bon-
ne e al., 2021). A e a di o ce, household size shi s because he ex-spouses spli
in o di e en households. As he couple’s child en usually eside wi h hei mo he
a e a union dissolu ion, he obse ed gende di e ences a e sensi i e o he use o
an equi alen scale, and hus o he weigh s ha a e used o accoun o he p esence
o child en in he household (Boll & Schülle , 2021, p. 19; Bonne e al., 2021).
The indings o s udies ha ha e examined he subjec i e economic well-being o
di o cees do no suppo he assump ion ha di o ce boos s he economic s anding
o male di o cees. These s udies gene ally epo ed ha bo h women and men expe-
ience a decline in subjec i e economic well-being ollowing a di o ce o union dis-
solu ion, al hough he e ec ends o be sligh ly mo e p onounced o women han
o men (Leopold, 2018). In a ecen pape , Bonne e al. (2021) used F ench egis e
da a o examine he household income o women and men a e di o ce, and ound
ha abou one-hal o all men and h ee-qua e s o all women expe ience a decline
in hei equalised income (a e ax and ans e s). The au ho s also obse ed s a k
di e ences be ween ex-couples depending on he ea ne models hey p e iously ol-
lowed, which we e ope a ionalised o e he sha e each pa ne con ibu ed o he
household income du ing he ma iage. While he ‘seconda y ea ne ’ ( ega dless
o gende ) gene ally expe ienced a decline in his/he equalised household income
a e union dissolu ion, he ‘main p o ide ’ (who con ibu ed mo e han 80% o he
household income) usually saw an inc ease in his/he equalised household income.
Di o ce, Sepa a ion, andIndi idual Ea nings
A sepa a e, bu ela ed s and o li e a u e has examined how di o ce and sepa a ion
a ec indi idual employmen and ea ning ca ee s. Mos o hese s udies ound ha
di o ce usually leads o changes in emale employmen and ea nings, pa icula ly in
coun ies whe e ma ied women end o be less a ached o he labou ma ke . Fo
example, based on su ey da a om he U.S., Tambo ini e al. (2015) epo ed ha
women’s ea nings inc ease subs an ially a e di o ce. Simila ly, using da a om he
Di o ce in Flande s Su ey, Thielemans and Mo elmans (2019) ound ha in he
yea s immedia ely a e di o ce, women’s employmen a es ise sha ply. In a c oss-
na ional s udy, Van Damme e al. (2009) obse ed mo e modes employmen e ec s
o some coun ies, bu la ge posi i e employmen e ec s o Wes Ge many.
By con as , he esul s o men sugges ha union dis up ion nega i ely a ec s
hei ea nings and employmen a es. Howe e , he magni ude o he e ec s ound
di e s g ea ly ac oss s udies and ou come a iables. In an analysis o egis e da a
om Is ael, Raz-Yu o ich (2013) epo ed ha while men’s employmen s abil-
i y (de ined as con inuous employmen h oughou a gi en yea ) ends o de e io-
a e a e a union dissolu ion, di o ce does no a ec men’s ea nings o hei sala y

1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 5 o 34 23
g ow h a es. In a s udy based on Du ch su ey da a, Kalmijn (2005) ound ha
men’s isk o unemploymen inc eases a e di o ce. These esul s a e co obo a ed
by he indings o Co izzi (2008), who, using da a om he Swiss Household Panel,
ound ha union dissolu ion has a mo e pe asi e e ec on men’s han on women’s
likelihood o being unemployed, and ha poo heal h is an impo an in e ening
a iable (see also Bio eau e al., 2019; Couch e al., 2015).
Ins i u ional Backg ound andResea ch Ques ion
Family Policies andMain enance Regula ions
Whe he and, i so, how di o ce a ec s he ex-spouses’ employmen and ea ning
ca ee s depends on he coun y con ex , he p e ailing gende ed employmen pa -
e ns, and he policies ha egula e pos -sepa a ion beha iou . The Ge man go e n-
men has in oduced impo an policy e o ms in ecen yea s, including he expan-
sion o day ca e o child en unde age h ee as well as he implemen a ion o an
ea nings- ela ed pa en al lea e bene i sys em in 2007. Va ious s udies ha e shown
ha hese e o ms ha e had a sizeable impac on mo he s’ ull- ime employmen
a es (Geye e al., 2015), and on a he s’ up ake o pa en al lea e (Geisle & K ey-
en eld, 2019).
Despi e hese de elopmen s, ma e nal and pa e nal employmen pa e ns in Ge -
many s ill di e g ea ly. While mos a he s wo k ull ime, mos mo he s subs an-
ially educe hei wo king hou s a e hei i s child is bo n. Simila ly, he di ision
o household labou has emained s ongly gende ed, al hough he e a e signs ha
a he s a e s a ing o pe o m mo e childca e (Köppen & T appe, 2019). While he
ecen e o ms signal he Ge man go e nmen ’s clea commi men o he dual-ea ne
model, o he policies ha e emained in place ha suppo he ma ied b eadwinne
amily model. Among hese policies is he op ion o ‘income spli ing’, which allows
ma ied couples o ile hei axes join ly. Due o he p og essi e ax schedule in
Ge many, his sys em p o ides subs an ial bene i s o ma ied couples, especially
when he ea nings o he wo pa ne s di e (BMFSFJ, 2021). A u he peculia i y
o he Ge man sys em is he ea men o ‘ma ginal employmen ’, a ca ego y ha
includes jobs ha gene a e ea nings o €450 o less pe mon h. Because ma ginal
employmen is exemp om axa ion and om social secu i y con ibu ions, i does
no p o ide wo ke s wi h public heal h o unemploymen insu ance co e age.2 As
a ma ied pe son can be included in he public heal h insu ance co e age o his/
he spouse, ma ginal employmen is mainly a ac i e o ma ied people. In sum,
he e a e se e al bene i s ha acc ue o ma ied couples, and pa icula ly o hose in
male (and emale) b eadwinne cons ella ions. As hese bene i s a e o ei ed upon
di o ce, he employmen pa e ns o couples in b eadwinne cons ella ions may be
expec ed o change he mos a e di o ce.
2 They a e exemp om con ibu ions o heal h, old-age ca e, and unemploymen insu ance. Un il 2013,
hey we e also exemp om pension con ibu ions.
D.B üggmann, M.K eyen eld
1 3
23 Page 6 o 34
Some o he ad e se e ec s o di o ce a e bu e ed by he possibili y o collec -
ing ex-spousal main enance paymen s. In line wi h he logic o a conse a i e male
b eadwinne egime, he Ge man egula ions shielded he ‘weake pa ’ in a ma -
iage om he loss o he suppo o he male b eadwinne . Un il ecen ly, his sys-
em was ( om he pe spec i e o he claiman ) among he mos gene ous in Eu ope.
Ex-spousal suppo paymen s we e g an ed based on he logic ha ‘ma i al solida -
i y’ ex ends beyond he b eakdown o a union, and ha ex-spousal main enance
paymen s should e lec he p io s anda d o li ing o he couple. The ‘ca egi ing
spouse’ was no obliged o be employed be o e he couple’s younges child u ned
eigh yea s old, and was expec ed o wo k on a pa - ime basis only when he child
was be ween ages eigh and 15. Thus, he esiden pa en was no expec ed o be in
ull- ime employmen un il he younges child was 15yea s old, and i was only a
ha poin ha he e usal o ake up ull- ime wo k could be used as a eason o cu b
ex-spousal suppo paymen s.
In 2008, a majo e o m came in o o ce. Ins ead o ‘spousal solida i y’, he
e o m emphasised he p inciple o he sel - eliance (‘Eigen e an wo ung’) o he
ex-pa ne s a e di o ce. Whe eas p e iously he esiden pa en was no obliged o
be employed un il he couple’s younges child was eigh yea s old (o o be ull- ime
employed un il he child was 15yea s old), he h eshold a which he esiden pa -
en had o employed was lowe ed o he younges child’s hi d bi hday. Unde he
e o m, he amoun and pa icula ly he du a ion o ex-spousal suppo we e sha ply
educed, wi h suppo gene ally being p o ided only in he pe iod immedia ely a e
he di o ce (Peschel-Gu zei , 2008; Rö hel, 2009). Se e al obse e s a gued ha
he e o m would be disad an ageous o women; ha i was oo adical; and ha i
igno ed ‘ ac ual socie al condi ions’, i.e. ha mo he s a e s ill he main p o ide s o
ca e (Lenze & Funcke, 2016). Thus, he assump ion unde pinning he e o m ha
he esiden pa en would become ‘economically sel - elian ’ did no seem o ma ch
he eali ies in Ge many, whe e mos ma ied couples s ill p ac iced a a he adi-
ional di ision o labou . In 2013, he egula ions we e adjus ed, wi h a clause being
added o he law ha allowed o ex-spousal suppo o be g an ed in ha dship cases,
such as in cases in which he couple had been in a e y long ma iage wi h a e y
unequal di ision o labou .
The e is li le exis ing empi ical e idence on how he main enance e o m
a ec ed di o cing couples in p ac ise. Al hough he law wen in o e ec na ion-
wide in Janua y 2008, i seems likely ha he e we e di e ences be ween a eas o
ju isdic ion in how igo ously he new egula ions we e applied. Based on a cu so y
assessmen o he ex-spousal alimony cases ha wen o appeal, i appea s ha no
all judges immedia ely emb aced he new egula ions (Willenbache , 2010, p. 381).
Un o una ely, social science su eys do no p o ide comp ehensi e insigh s in o
alimony paymen s. Es ima es based on he Ge man Family Panel sugges ha he
sha e o emale di o cees wi h child en who ecei ed ex-spousal suppo declined
om a ound 10% in 2009/10 o 3% by 2018/19.3 Mo eo e , he o icial cou s a-
is ics indica e ha he e has been a apid decline in cases ela ed o ex-spousal
3 Own es ima es based on he Ge man Family Panel (Huinink e al., 2011).
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 7 o 34 23
main enance since he e o m was enac ed. Fo example, he sha e o amily cou
p oceedings ha deal wi h ex-spousal suppo dec eased om 12% in 2006 o 4% in
2019 (S a is isches Bundesam , 2020).4
Apa om p o iding ex-spousal suppo , he non- esiden pa en is obliged o
pay child main enance, wi h he amoun a ying depending on he income o he
non- esiden pa en . Cu en ly, he ques ion o whe he main enance egula ions
should be e accoun o sha ed physical cus ody a angemen s is being deba ed.
Howe e , only a e y small sha e o di o ced pa en s in Ge many p ac ice sha ed
physical cus ody (Walpe e al., 2021). In he o e whelming majo i y o cases, he
couple’s child en li e wi h hei mo he a e di o ce, while he o he pa en has a
‘ igh o con ac ’ (‘Umgangs ech ’). Like in o he coun ies, a la ge ac ion ( oughly
40%) o non- esiden pa en s in Ge many pay none o only a po ion o he child
main enance hey owe (Hube e al., 2020), usually because he non- esiden pa -
en is unable o pay o he esiden pa en wan s o a oid con lic o con ac wi h he
non- esiden pa en (And eß e al., 2003).
Resea ch Ques ion andExpec a ions
The e a e con lic ing ac o s ha de e mine he ea nings ajec o ies o di o cees.
On he one hand, di o ce usually en ails he loss o he economies o scale o li -
ing in a join household. This may p omp di o cees o en e he labou ma ke o o
expand hei wo king hou s. On he o he hand, he heal h o di o cees may de e io-
a e due o he s ess hey expe ience as a esul o he union dissolu ion. This may
help o explain why he ea ning ajec o ies o di o cees ha e been ound o be less
s eep han hose o a con ol g oup o people who did no unde go a di o ce. Thus,
he loss o economies o scale and he isk o a de e io a ion in heal h a e among he
ac o s ha a ec he ea nings and employmen choices o all di o cees. In addi ion,
he e a e o he ac o s ha a e speci ic o he ea ne models ex-couples ollowed
while ma ied. These ac o s a e lis ed in a s ylised ashion in Table1 and desc ibed
in mo e de ail below.
P ima y Ea ne inaSingle‑Ea ne Model
We assume ha he ea nings o he p ima y ea ne in a single-b eadwinne cons el-
la ion will decline a e di o ce (Hypo hesis 1). The economic incen i es o indi-
iduals in single-b eadwinne cons ella ions change ma kedly upon di o ce. Fi s ,
he p essu e on he p ime ea ne o gua an ee he economic secu i y o he house-
hold eases a e di o ce. As a esul , s/he may be mo e likely han be o e o u n
down less a ac i e oppo uni ies o ad ance in his/he ca ee . Mos impo an ly, he
p ime ea ne loses he p i ileges s/he had unde he ax and ans e sys em. Du -
ing ma iage, he main ea ne is no ully axed, as s/he ecei es a ax c edi i his/
4 These p oceedings also include pending cases as well as e-nego ia ions o ex-spousal suppo o pas
di o ces.
D.B üggmann, M.K eyen eld
1 3
23 Page 8 o 34
Table 1 Fac o s ha lead o a dec ease/inc ease in ea nings/employmen a e di o ce in Ge many
Fac o s ha esul in a dec ease in ea nings/employmen Fac o s ha esul in an
inc ease in ea nings/employ-
men
P ima y ea ne in single-ea ne model Taxa ion (shi om join o indi idual axa ion)
Alimony paymen s
Loss o economies o scale
Seconda y ea ne in single-ea ne model Dep ecia ed human capi al
Ca e obliga ions
Loss o economies o scale
Loss o ee heal h insu ance
co e age
Taxa ion (shi om join o
indi idual axa ion)
Dual-ea ne couple Alimony paymen s (non- esiden pa en ) Loss o economies o scale
Low-income couple Alimony paymen s (non- esiden pa en )
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 15 o 34 23
in unemploymen is likely a ibu able o mo e women becoming eligible o unem-
ploymen bene i s a e di o ce, as hese bene i s a e means- es ed o indi iduals
who ha e no been p e iously wo king o who a e unemployed o a longe pe iod
o ime. Many ma ied women do no ha e access o hese bene i s because he ea n-
ings o hei pa ne a e included in he calcula ion o hei eligibili y. Mo eo e , by
egis e ing o unemploymen , some o he di o ced women in ou sample who we e
p e iously co e ed by hei ex-spouse’s heal h insu ance, and who los his co e age
upon di o ce, may ha e egained access o heal h insu ance.
Finally, he igu e in he lowe - igh co ne o Fig.2 shows ha he up ake o
disabili y lea e inc eased mo e apidly o women han o men a ound he ime
o di o ce. This may indica e ha women’s heal h was mo e s ongly a ec ed by
di o ce han men’s heal h. Howe e , cau ion is wa an ed in in e p e ing his ind-
ing, because pa en s may ake disabili y lea e o ca e o sick child en as well as
o hemsel es, and mo he s a e mo e likely han a he s o ake lea e o his ea-
son. Thus, he inc ease in he up ake o disabili y lea e among women may ha e
been due o lone mo he s ha ing inc eased childca e obliga ions a e sepa a ion,
and needing addi ional days o lea e o deal wi h hei child en’s bou s o sickness.
I should also be no ed ha lone mo he s a e eligible o mo e days o lea e o ca e
o sick child en han ma ied mo he s.
Analy ical S a egy: Combining Fixed‑E ec s andMa ching
In o de o es ima e he e ec o sepa a ion on indi idual ea nings, we use a ixed-
e ec s app oach. We model he ea ning ajec o y by employing a dummy impac
unc ion ( o de ails, see B üde l & Ludwig, 2015; Ludwig & B üde l, 2021). The
eg ession can be o malised as ollows:
yi
a e he indi idual ea nings ha a y by ime = 2011, …, 2015.
Dk
i
is a se
o dummy a iables ha deno es whe he he pe son was di o ced. No e ha we
assume ha people an icipa e hei di o ce o some ex en . Thus, we include no
only ea men dummies o he ime a e di o ce, bu also ea men dummies o
he wo yea s p io o di o ce.
X′
i 𝛿
includes he ime- a ying co a ia es. We a e,
howe e , e y limi ed in he ime- a ying co a ia es we can include. Ideally, we
would ha e liked o include in o ma ion on, o example, e-pa ne ing, alimony
paymen s, physical cus ody egula ions, o he bi ho a child. As we do no ha e
such in o ma ion, and he sole ime- a ying a iable ha we use in ou modelling
app oach is age (dummy o single ages),
𝛼i
ep esen s he ime-cons an indi idual
ai s and
𝜀i
is he idiosync a ic e o e m.
The ixed-e ec s modelling d aws on de-meaned da a. Tha en ails ha he
di o ce e ec is gene a ed om he indi idual mean o he ea nings du ing he
obse a ions window. Figu e3 displays a ic ious case o illus a e how he model
ope a es. A emale di o cee ea ned 0.5 ea ning poin s in he pe iod 2011 o 2015.
He ea nings we e 0.1 poin s below he a e age ea nings be o e he di o ce, and
y
i =
k
∑
k=0
𝛽kDk
i +X�
i 𝛿+𝛼i+𝜀i

D.B üggmann, M.K eyen eld
1 3
23 Page 16 o 34
we e 0.1 poin s abo e he indi idual a e age ea nings a e he di o ce, which sug-
ges s ha he di o ce caused he ea nings o inc ease by 0.2 poin s on a e age. The
igu e illus a es ha an analysis ha solely elies on he wi hin a ia ion o di o -
cees may lead o biased esul s i he e a e s ong ‘secula ’ ends. In his exam-
ple, ea nings inc eased o all g oups s a ing in 2013. In o de o assess he ‘ ue’
impac o di o ce on ea nings, we need o compa e di o cees wi h a sui able con-
ol g oup o people who did no di o ce in he gi en ime window. This is o i al
impo ance in his con ex , as di o ce ends o occu a ages when mos indi id-
uals a e s ill expe iencing inc eases in hei ea nings. Fu he mo e, Ge many has
implemen ed majo amily policy e o ms since 2005, which has likely con ibu ed
o an inc ease in he labou ma ke pa icipa ion o women wi h child en. We ha e
employed ma ching echniques o add ess his conce n. Thus, we ha e cons uc ed a
sui able con ol g oup, which accoun s o he ac ha di o ced women in Ge many
would ha e expe ienced inc eases in ea nings and employmen pa icipa ion o e
ime e en i hey had no di o ced.
Ma ching onFu u e Di o cees
We ha e cons uc ed a con ol g oup by d awing a sample o people who had no ye
been di o ced in ou s udy pe iod, bu who di o ced in 2018, 2019, o 2020. He e,
we capi alise on he ac ha ou obse a ion window spans he pe iod om 2011 o
2015, bu we ha e in o ma ion on di o ce un il 2020. We ollow an app oach ha is
equen ly used in he e alua ion o labou ma ke p og ams, in which he con ol
g oup consis s o people who ha e no been ea ed ye (Sianesi, 2004). The ad an-
age o his app oach is ha we can use an a ay o ma iage-speci ic a iables o
achie e a good i (du a ion o ma iage, age a i s ma iage, ea ne model du ing
ma iage) in he ma ching p ocedu e ( o de ails, see he appendix).
Mos ea lie ma ching algo i hms used logis ic eg ession o calcula e he p open-
si y sco e. The ques ion o how hese models should be speci ied has ye o be ully
esol ed, e en hough nume ous co a ia e selec ion algo i hms ha e been w i en
o add ess his issue (Dehejia & Wahba, 1999; Hi ano & Imbens, 2001; McCa ey
Fig. 3 Impac o di o ce on
ea nings (2 ic ious cases)
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 17 o 34 23
e al., 2004; Wes e al., 2000, pp.69–70). Resea ch has shown ha simple logis ic
models a e o en no su icien o balance co a ia es (Pi acchio e al. 2014; Wes e-
ich e al., 2010; Lee e al., 2010). Machine lea ning (ML) echniques a e new, lex-
ible, non-pa ame ic ways o add essing he co a ia e selec ion p oblem. We used
he R package ‘ wang’, which uses he ML echnique o achie e a ma ch ( o de ails,
see Ce alu e al., 2021; McCa ey e al., 2004).
Robus ness Check: Ma ching on heTo al Popula ion
A disad an age o he abo emen ioned ma ching app oach is ha he con ols a e
d awn om u u e di o cees. Al hough ha he di o ce happened in he dis an
u u e, we canno ule ou he possibili y ha some o he indi iduals an icipa ed
hei own sepa a ion mo e han ou yea be o e i occu ed. In a obus ness check,
we cons uc ed a con ol g oup om he en i e sample. We used a ‘simple’ p open-
si y sco e ma ching app oach o gene a e a con ol g oup (Rosenbaum & Rubin,
1985), because he massi e sample size (mo e han 30 million pe son-yea s) did
no allow us o employ mo e sophis ica ed ma ching echniques. We can, howe e ,
assume ha he ma ch is easonably good because we ha e a la ge sample size, as
well as de ailed pas employmen , egional, and ea nings in o ma ion ( o de ails,
see he appendix).
Sample S a is ics
Table 2 p o ides sample s a is ics o he ea ed indi iduals and he con ols.
I shows ha he a e age du a ion o ma iage (up o he da e he di o ce pe i-
ion was iled) was 14.9yea s, and he a e age age a sepa a ion was 43.2 o
men and 40.5 o women. A majo i y o he di o ced couples ollowed he male
Table 2 Sample s a is ics
Di o ced Con ols: u u e
di o cees
Con ols: o al sample
( obus ness check)
Men Women Men Women Men Women
Cha ac e is ics (mean)
Ea ningpoin s in 2011 1.15 0.41 1.16 0.41 1.15 0.41
Age in 2013 43.20 40.54 43.20 40.66 43.20 40.54
Ma iage du a ion in 2013 14.88 14.88 14.80 14.80 – –
Ea ne model in 2010/11 (in %)
Male b eadwinne 13,952 (57%) 16,517*) – –
Female b eadwinne 963 (4%) 543*) – –
Dual ea ne 7,105 (29%) 7,212*) – –
Bo h low income 2,596 (11%) 1,853*) – –
Case numbe
Pe sons 24,616 24,616 33,869 33,869 7,129,087 9,872,161
Pe son-yea s (2010 o 2015) 147,696 147,696 203,214 203,214 42,774,522 59,232,966
D.B üggmann, M.K eyen eld
1 3
23 Page 18 o 34
b eadwinne model (57%), while a smalle sha e ollowed he dual-ea ne model
(29%). Only 4% o he couples p ac iced he emale b eadwinne model. In 11%
o he cases, nei he o he spouses ecei ed mo e han 0.5 ea ning poin s, mean-
ing ha hei ea nings we e 50% below a e age. The able also shows ha he
di o ced g oup aligned well wi h he weigh ed ea ed g oup in e ms o hei
ages and ea nings a baseline (in 2011). This was pa icula ly ue o he sample
ha ma ched on u u e di o cees.
Figu e4 displays he a e age ea nings o he di o cees as well as he ea n-
ings o he con ol g oup. The le panel p o ides he esul s o he analysis ha
used he u u e di o cees as a con ol g oup, while he igh panel shows he
esul s om he obus ness check. Bo h igu es display he same pa e n. These
indings suppo ou assump ion ha he ea nings o he con ol g oup inc eased
o e ime. Thus, he ixed-e ec s analyses would be biasedi hey we e based
solely on he ea ed g oup (di o cees). The igu e also shows ha he bias an
in he opposi e di ec ion o women andmen. Fo women, we would ha e o e -
es ima ed he di o ce e ec , as he ea nings o di o ced women inc eased o e
ime, while he e was a simila ‘secula ’ end in he con ol g oup. Fo men, he
pa e ns we e di e en . We see a mode a e decline in ea nings in he g oup o
di o cees. Howe e , he e ec appea s o be much s onge i we conside ha
he con ol g oup expe ienced an inc ease in ea nings o e he same ime pe iod.
*As he con ol g oup is based on weigh ing echniques, p o iding a sample
size o each ea ne model is no s aigh o wa d. The sample sizes a e calcu-
la ed by he e ec i e sample size (ESS) (
ESS
=(∑i∈Cwi)
2
∑i∈C
w2
i
, whe e w is he indi-
idual weigh ha is d awn om he weigh ing app oach ( o de ails, see Ridge-
way e al., 2021).
Fig. 4 A e age ea nings a ound he ime o sepa a ion by gende . Wes Ge man women and men who
iled o di o ce in 2013 and a con ol g oup
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 19 o 34 23
Resul s
The esul s om he ixed-e ec s models a e displayed in Table 3. We ha e es i-
ma ed sepa a e models by ea ne model. The only u he con ol a iable in he
model is age, which is included in single ages. Fi s , we ocus on women and men
who p ac iced he male ea ne model. Ou expec a ion ha we would obse e la ge
changes in beha iou in his g oup was indeed suppo ed by ou analysis. Men’s
ea nings declined by 0.058 ea ningpoin s om wo yea s be o e sepa a ion o wo
yea s a e sepa a ion (a dec ease o 4.5% in ela i e e ms). This is equi alen o a
decline in g oss annual ea nings o €2,050.9 The ea nings o women in male b ead-
winne cons ella ions inc eased by 0.083 ea ning poin s (o by 41.2% in ela i e
e ms o €2,940 pe annum).
While ou indings o he male b eadwinne model i ou expec a ions, ou
hypo hesis ha di o ce would no lead o any majo changes o dual-ea ne couples
was no ully con i med. Al hough he pa e ns we e less a enua ed han hey we e
in he male b eadwinne cons ella ion, we also ind o couples in he dual-ea ne
cons ella ions, ha women’s ea nings inc eased while men’s ea nings declined a e
di o ce. Fo men, we wi ness a decline in ea nings in he pe iod wo yea s p io
o wo yea s a e sepa a ion o 0.052 ea ning poin s (o 3.8%). Fo women, he
inc ease in hese cons ella ions amoun s o 0.058 ea ningpoin s (o 6.7%).
Fo he small g oup o women and men in o he cons ella ions ( emale b ead-
winne o bo h low income), we did no see any la ge changes in beha iou a ound
he ime o di o ce. I is, howe e , no ewo hy ha while he absolu e changes a e
Table 3 Resul s om ixed-e ec s models, be a coe icien s. Ou come: ea ning poin s
*p < 0.05, ** p < 0.01, *** p < 0.001. Fu he con ol a iable is age (in single ages).
Time since sepa a ion
−2 −1 0 + 1 + 2
Male ea ne
Female Re 0.039*** 0.076*** 0.084*** 0.083***
Male Re −0.029*** −0.048*** −0.059*** −0.058***
Dual ea ne
Female Re 0.023*** 0.052*** 0.057*** 0.058***
Male Re −0.026*** −0.040*** −0.047*** −0.052***
Female ea ne
Female Re 0.002 −0.001 −0.015 −0.024
Male Re 0.023** 0.022 −0.011 −0.022
Bo h low/no ea ne s
Female Re 0.020*** 0.024*** 0.026*** 0.027***
Male Re 0.024*** −0.001 −0.022* −0.035**
9 Ea nings a e calcula ed by mul iplying he ea ning poin s wi h he a e age g oss income (€ 35,363)
om 2015 (Anlage 1 SGB VI: h ps:// www. gese ze- im- in e ne . de/ sgb_6/ anlage_ 1. h ml).
D.B üggmann, M.K eyen eld
1 3
23 Page 20 o 34
indeed small, ela i e changes a e subs an ial in some ins ances. Fo example, when
bo h spouses had low o no ea nings be o e he di o ce, emale ea nings inc eased
by 0.027 ea ning poin s, which is an inc ease o 18.6% in ela i e e ms. Male ea n-
ings declined by 0.035 ea ning poin s. In ela i e e ms ha is a dec ease o 18.5%
wi h espec o he e e ence yea ( -2).
To ge u he insigh in o he magni ude o he ea nings o he di o cees in he
di e en ea ne cons ella ions, we ha e calcula ed p edic ed alues o he main
ea ne models (dual-ea ne and male b eadwinne models). Figu e5 shows ha e en
in he dual-ea ne model, women and men s a ed o a di e en base le els. The
igu es also indica e ha while women’s ea nings inc eased s ongly o e he obse -
a ion window, la ge gende di e ences pe sis ed in bo h he male b eadwinne
model and he dual-ea ne model. Howe e , gende di e ences a e adically mo e
p onounced in he male b eadwinne model. Two yea s a e sepa a ion, women
in male b eadwinne cons ella ions we e ea ning 66% less han he a e age wage.
Based on his analysis, we can conclude ha mos women who had been in a male
b eadwinne cons ella ion we e no able o achie e ea nings ha would quali y hem
as ‘sel - elian ’ in he wo yea s a e sepa a ion.
Conclusion
This pape examined he ea nings o di o cees in wes e n Ge many. We sough
o answe he ques ion o whe he and, i so, how he ea ne models men and
women ollowed du ing hei ma iage a ec ed hei ea nings ajec o ies a e a
union dissolu ion. We a gued ha Ge man law inc easingly demands ha he p e-
iously ma ied pa ne s achie e economic independence, pa icula ly since he
implemen a ion o he main enance e o m o 2008. This means ha ollowing a
Fig. 5 P edic ed ea nings by du a ion since sepa a ion (see Table3 o be a coe icien s)

1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 21 o 34 23
di o ce, he seconda y ea ne in a single-ea ne cons ella ion expe iences imme-
dia e p essu e o en e he labou ma ke o o inc ease his/he wo king hou s. In
he Ge man con ex , his p essu e can be especially acu e because single-ea ne
couples ha e p i ileges ha hey lose when hey di o ce, such as he ee co-
insu ance o he non-ea ning o ma ginally employed spouse in he heal h ca e
sys em. We he e o e a gued ha women who had been in a male b eadwinne
cons ella ion ace s ong p essu e o be economically independen a e a di o ce.
Howe e , whe he hese women a e indeed able o achie e economic independ-
ence was an empi ical ques ion ha had ye o be answe ed.
To p o ide e idence o help answe his esea ch ques ion, we used da a om
he Ge man pension egis e s. We selec ed a homogenous coho o couples, all
o whom sepa a ed in he yea 2013. Thus, we examined couples who expe i-
enced a di o ce well a e he implemen a ion o he abo emen ioned main e-
nance e o m, which equi es bo h pa ne s o achie e economic independence.
Mo eo e , by selec ing di o cees om only one di o ce coho , we made su e
ha he esul s we e no d i en by changes in beha iou ac oss coho s o ime
pe iods. An addi ional ad an age o using hese da a was ha hey we e on he
couple le el, which enabled us o di ec ly compa e he ea nings o he ex-pa ne s
a e di o ce. Ou desc ip i e in es iga ion showed ha he employmen a es o
male and emale di o cees con e ged in he p ocess o sepa a ion. This con e -
gence was he esul o a signi ican sha e o women ansi ioning om ma ginal
o egula employmen a ound he ime o hei di o ce. Howe e , women’s a es
o egis e ed unemploymen also ose sha ply a ound he ime o hei di o ce.
While a e age emale ea nings inc eased in he p ocess o di o ce, subs an ial
gende di e ences pe sis ed. These di e ences we e la ges among di o cees who
we e p e iously in a male b eadwinne cons ella ion. In his g oup, he women
ea ned oughly 66% less han he a e age wage, while he men ea ned 21% mo e
han he a e age wage wo yea s a e hei di o ce.
An in iguing inding om ou in es iga ion was ha men’s ea nings declined in
he p ocess o di o ce. We adop ed a igo ous causal app oach o iden i ying he
size o his ‘di o ce penal y’. He e, we employed a ixed-e ec s app oach ha we
combined wi h ma ching echniques. Such an app oach seemed wa an ed because a
simple compa ison o he ea nings o di o cees be o e and a e di o ce may unde -
s a e he ue e ec s o di o ce, as ea nings gene ally inc ease ac oss he li e cou se.
While we iden i ied a la ge and nega i e di o ce penal y o men, explaining ha
pa e n is mo e di icul . The main b eadwinne ’s indi idual ea nings a e mo e hea -
ily axed ollowing a di o ce, which educes his/he incen i es o engage in ca ee
ad ancemen . We he e o e assumed ha we would see a decline in he ea nings o
p ime ea ne s in single-ea ne cons ella ions, while he ea nings ajec o ies o dual-
ea ne couples would be less a ec ed by di o ce. Indeed, male ea nings in male-
b ead-winne cons ella ions decline mo e s ongly han in dual-ea ne cons ella ions.
O e all, howe e , one mus conclude ha he e is a subs an ial and signi ican male
di o ce penal y o men in bo h dual-ea ne and in male b eadwinne cons ella ions.
D.B üggmann, M.K eyen eld
1 3
23 Page 22 o 34
Limi a ions
This pape adop ed a causal app oach by combining ixed-e ec s models wi h
ma ching echniques. While we may ha e been able o p oduce ai ly unbi-
ased‘di o ce es ima es’, we ha e no been able o unco e he mechanisms a
play. We ound ha men’s ea nings declined apidly in he p ocess o di o ce.
We a ibu e his decline o he economic incen i es ha we discussed in he
heo e ical pa o his pape . Howe e , as we also ound a simila pa e n in he
dual-ea ne and male b eadwinne cons ella ions, i appea s ha o he mecha-
nisms con ibu ed o hese ou comes as well. I may be he case ha men’s
heal h was s ongly a ec ed by di o ce, which esul ed in a decline in hei ea n-
ings. I is also possible ha men we e mo e engaged in childca e esponsibili ies
a e hei di o ce han hey we e be o e, which may ha e inhibi ed hei ca ee
ad ancemen . Al hough sha ed physical cus ody is a e in Ge many, we canno
ule ou he possibili y ha his mechanism was also a play.
Beyond heal h, main enance paymen s, and physical cus ody, we we e no
able o con ol o pa e ns o e-pa ne ing, and hus o whe he he di o cees
en e ed a new ea ne model. Ano he impo an limi a ion o ou s udy was ha
we had no in o ma ion on he numbe o child en, and we e he e o e unable
o in es iga e whe he child en mode a ed he ela ionship be ween di o ce and
ea nings. In addi ion, ha ing access o mo e e ined in o ma ion on he di o -
cees’ occupa ional s a us and wo king hou s would ha e been desi able, as such
da a would ha e helped us be e unde s and why he di o ced women in ou
sample had ea nings ha we e so a below a e age. Beyond he lack o addi-
ional a iables, we also mus no e ha he s udy popula ion does no co e he
o al esiden popula ion in Ge many, bu only he popula ion ha has a pension
accoun ( oughly 90% o he popula ion). Fu he , no all di o ces a e cap u ed
in he pension accoun s (Keck e al., 2019). Finally, ou ocus has been on a e -
age changes in ea nings. Unde which condi ions people see and inc ease, s ag-
na ion, o a decline in hei ea nings a e di o ce would ha e been an impo an
addi ional pe spec i e.
Despi e hese many limi a ions, ou in es iga ions p o ide clea and ala ming
e idence o he gende di e ences in ea nings a e di o ce. By examining p e-
iously ma ied couples, we p o ided e idence o he di e ging a es o women
and men in male b eadwinne cons ella ions. These couples s ill make up he
majo i y o he eshly di o ced couples in Ge many. Al hough he women who
had been in a male b eadwinne cons ella ion expanded hei employmen ac i i-
ies a e hei di o ce, hei ea nings did no , on a e age, come close hose o
hei ex-spouse. The main enance e o m ha was enac ed in Ge many in 2008
was mo i a ed by he belie ha women would be able o achie e economic inde-
pendence a e di o ce. Ins ead, ou indings show ha he majo i y o emale
di o cees in Ge many a e a om ha ing he le el o economic independence
lawmake s assumed hey would be able o a ain.
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 23 o 34 23
Appendix I: Ma ching andweigh ing amewo k. Con ols de i ed
om‘ u u e di o cees’
In he ollowing, we add ess he main p ocedu es and assump ions o he ma ching
and weigh ing amewo k.
TWANG
The R package “TWANG” con ains a p ocedu e o suppo causal modelling o
obse a ional da a ia he es ima ion and e alua ion o p opensi y sco es and associ-
a ed weigh s. P opensi y sco es a e es ima ed by g adien boos ed models. G adien
boos ed models (GBM) a e machine lea ning app oaches ha ake in o accoun non-
linea i ies and in e ac ions among he co a ia es ( o de ails, see Ridgeway e al.,
2021).
Condi ional Independence Assump ion (CIA)
The CIA add esses he exis ence o unobse ed co a ia es ha simul aneously a ec
he ea men assignmen and he ou come. In an obse a ional s udy, we can con ol
o obse ed co a ia es, bu he e may be unobse ed ac o s ha selec people in o
ea men . The CIA holds only i all unobse ed co a ia es ha e no impac on ea n-
ings o he selec ion in o ea men . I is only hen ha he es ima ed ea men e ec
equals he ue unbiasede ec . We canno ule ou he possibili y ha he e we e
unobse ed co a ia es ha biasedou es ima ion.
S able Uni T ea men Value Assump ion (SUTVA)
The SUTVA add esses he possibili y ha he ea ed indi iduals will compe e on
he labou ma ke wi h pe sons om he con ol g oup. The di o cees migh ha e
s ong incen i es o en e he labou ma ke o o inc ease hei wo king hou s (i
hey a e al eady employed) due o he inancial bu den o di o ce. As a consequence,
employe s migh subs i u e indi iduals om he con ol g oup wi h di o cees who
a e willing o accep lowe wages o wo se employmen condi ions. As a esul , ou
es ima ed ea men e ec migh be dis o ed because he ou come o he con ols
se es as he coun e ac ual ou come o he di o cees. While his scena io is in p in-
ciple possible, we belie e ha his e ec would no be la ge, gi en ha only a ound
200,000 di o ces occu in Ge many each yea . The numbe o di o cees seems oo
low o hei beha iou o a ec he labou ma ke op ions o ma ied people. Thus,
we assume he SUTVA holds.
Common Suppo Requi emen
The common suppo is a equi emen in he ma ching amewo k. Howe e , i is
less clea whe he i is ele an in he weigh ing amewo k. Ridgeway e al., (2021:
D.B üggmann, M.K eyen eld
1 3
23 Page 24 o 34
p. 9), o example, a gued ha excellen co a ia e balance can be achie ed, e en
when he p opensi y sco es es ima ed o he ea ed and he con ol g oup o e lap
only a li le. In ou sample, he dis ibu ion o he p opensi y sco es do no o e lap
pe ec ly, bu hey o e lap o a la ge ex en . Thus, in ligh o he a gumen by Ridge-
way e al. (2021) and he ai ly good o e lap in ou sample, we decided agains im-
ming, and es ima ed he ea men e ec wi hou he common suppo equi emen .
Balance Resul s
The balance esul s a e impo an o assessing whe he weigh ing emo ed sys-
ema ic di e ences be ween he ea ed g oup and he con ol g oup. Ou balance
s a is ics a e shown in Tables4 and 5. Be o e weigh ing (Table4), he e a e s a is i-
cally signi ican di e ences be ween he g oups o many a iables. Howe e , a e
weigh ing, he di e ence is g ea ly educed. Only some mino ou lie s exceed 20%
o a s anda d de ia ion (see Table4 (column SD max and SD mean) and A2 (col-
umns SD)). In pa icula , he small g oups o “bo h low ea nings” and he “ emale
b eadwinne ” we e di icul o weigh , judged by he balance esul s.
Appendix II: Ma ching andWeigh ing F amewo k. Con ols De i ed
om‘To al Sample’
In he ollowing, we p o ide he balance s a is ics o he analysis whe e he con ols
we e de i ed om he ‘ o al sample’ (Fig.4, igh panel). We did no use he e he
TWANG amewo k as he compu a ional bu den would be oo high due o he eno -
mous sample size. Ins ead, we calcula ed he p opensi y sco es ia a simple logi
es ima ion. We ha e chosen simila co a ia es as be o e (see Table6), howe e , we
Table 4 Balance s a is ics o subsamples be o e and a e weigh ing (con ols de i ed om ‘ u u e
di o cees’)
SD s anda dised di e ence s a is ic, ESS e ec i e sample size ( o de ails, see Ridgeway e al., 2021)
N ea N con ol ESS ea ESS con ol SD max SD mean
Bo h low ea nings Unweigh ed 2,596 33,869 2,596 33,869 4.9008 0.0927
Weigh ed 1,852 0.7998 0.0235
Male b eadwinne Unweigh ed 13,952 33,869 13,952 33,869 1.7058 0.0386
Weigh ed 16,517 0.1223 0.0082
Dual ea ne Unweigh ed 7,105 33,869 7,105 33,869 5.6042 0.1092
Weigh ed 7,212 0.1441 0.0108
Female b eadwinne Unweigh ed 963 33,869 963 33,869 14.8989 0.2039
Weigh ed 543 4.5671 0.0476
All Unweigh ed 24,616 33,869 24,616 33,869 0.2996 0.0165
Weigh ed 25,943 0.0436 0.0055
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 31 o 34 23
did no include any in e ac ion e ms. The p edic ed p opensi y sco es we e hen
ans o med in o odd-weigh s.10 Balance esul s in Table7 a e based on hese odd-
weigh s. We also included he e he p alues om a - es o equal means be ween he
ea ed and con ols. Despi e he ac ha we only use a simple ma ching app oach,
he balance esul s a e ai ly good.
Acknowledgemen s Fo de ailed eedback on an ea lie e sion o his pape , we wish o hank Ch is ina
Boll (Ge man You h Ins i u e) and Mi iam Beblo (Uni e si y o Hambu g). Fo language edi ing, we
hank Mi iam Hils. All emaining e o s a e ou s. Funding o he “Fö de onds Wissenscha in Be lin”
is acknowledged.
Funding Open Access unding enabled and o ganized by P ojek DEAL.
Table 7 (con inued)
Female sample Male sample
T ea ed p alue SD T ea ed p alue SD
Days o unemploymen I 5.55 0.94 0.00 5.42 0.94 0.00
Unemploymen I bene i s 213.46 0.95 0.00 362.29 0.94 0.00
Days o unemploymen II 19.36 0.92 0.00 12.17 0.94 0.00
Days o wo k disabili y 3.37 0.95 0.00 5.32 0.93 0.00
Bene i s while in wo k disabili y 160.17 0.95 0.00 415.11 0.93 0.00
Days o con ibu ion o ca ing 4.25 0.96 0.00 0.16 0.99 0.00
Bene i s o ca e ac i i ies 33.38 0.97 0.00 1.19 0.99 0.00
Days o olun a y con ibu ions 0.18 0.99 0.00 1.43 0.98 0.00
Days o childca e 2.07 0.97 0.00 0.03 1.00 0.00
Educa ion
No in o ma ion 0.26 0.99 0.00 0.27 1.00 0.00
Wi hou school diploma 0.01 1.00 0.00 0.01 1.00 0.00
Elemen a y school diploma 0.21 0.99 0.00 0.30 1.00 0.00
Seconda y school diploma 0.36 0.99 0.00 0.25 1.00 0.00
High school diploma 0.16 1.00 0.00 0.17 1.00 0.00
Highe educa ion
No in o ma ion 0.19 1.00 0.00 0.16 1.00 0.00
Wi hou oca ional aining 0.08 0.99 0.00 0.06 0.99 0.00
Voca ional aining 0.66 1.00 0.00 0.60 1.00 0.00
Bachelo diploma 0.02 1.00 0.00 0.09 1.00 0.00
Mas e diploma 0.00 1.00 0.00 0.01 1.00 0.00
Mas e / echnician college quali ica ion 0.05 1.00 0.00 0.08 1.00 0.00
Doc o a e 0.00 1.00 0.00 0.00 1.00 0.00
Con ols de i ed om ‘ o al sample’
SD s anda dised di e ence. p alues a e om a - es o equal means. Resul s om 357 egion dummies
a e no displayed
10
ODD =TREAT +(1−TREAT)∗ps(1−ps)

D.B üggmann, M.K eyen eld
1 3
23 Page 32 o 34
Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License,
which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as
you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Com-
mons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his a icle
a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line o he
ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use is
no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission
di ec ly om he copy igh holde . To iew a copy o his licence, isi h p:// c ea i eco mmons. o g/ licen
ses/ by/4. 0/.
Re e ences
Adle , D. & S. Kelly, T. (2021). ioplo : iolin plo . R package e sion 0.3.6. h ps:// gi hub. com/ TomKe
llyGe ne ics/ iopl o
And eß, H.-J., Bo gloh, B., B ockel, M., Giesselmann, M., & Hummelsheim, D. (2006). The economic
consequences o pa ne ship dissolu ion: A Compa a i e analysis o panel s udies om Belgium,
Ge many, G ea B i ain, I aly, and Sweden. Eu opean Sociological Re iew, 22(5), 533–560.
And eß, H.-J., Bo gloh, B., Güllne , M., & Wilking, K. (2003). Wenn aus Liebe o e Zahlen we den:
Übe die wi scha lichen Folgen on T ennung und Scheidung. Wes deu sche Ve lag.
A khangelsky, D., & Imbens, G. W. (2019). The ole o he p opensi y sco e in ixed e ec models. Wo k-
ing Pape s wp2019_1905, CEMFI.
Bio eau, A.-L., Ca ole, B., & Cambois, E. (2019). Risk o majo dep essi e episodes a e sepa a ion:
The gende -speci ic con ibu ion o he income and suppo los h ough union dissolu ion. Eu o-
pean Jou nal o Popula ion, 35(3), 519–542.
BMFSFJ. (2021). Neun e Familienbe ich . h ps:// www. bm s j. de/ esou ce/ blob/ 174094/ 93093 98370
4d614 85814 1b8 1 44012 44/ neun e - amil ienbe ich - lang assung- da a. pd
Boe ien, D., & Le sch, P. (2021). Gende and changes in household weal h a e he Dissolu ion o ma -
iage and cohabi a ion in Ge many. Jou nal o Ma iage and Family, 83(1), 228–242.
Boll, C., & Schülle , S. (2021). Sha ed pa en ing and pa en s’ income e olu ion a e sepa a ion.
New explo a i e insigh s om Ge many. SOEPpape s on Mul idisciplina y Panel Da a Resea ch,
1131/2021.
Bonne , C., Ga bin i, B., & Solaz, A. (2021). The lip side o ma i al specializa ion: The gende ed e ec
o di o ce on li ing s anda ds and labo supply. Jou nal o Popula ion Economics, 34, 515–573.
h ps:// doi. o g/ 10. 1007/ s00148- 020- 00786-2
B öckel, M., & And eß, H.-J. (2015). The economic consequences o di o ce in Ge many: Wha has
changed since he u n o he millennium? Compa a i e Popula ion S udies, 40(3), 277–312.
B üde l, J., & Ludwig, V. (2015). Fixed-e ec s panel eg ession. In H. Bes & C. Wol (Eds.), The
SAGE handbook o eg ession analysis and causal in e ence (pp. 327–357). Sage.
B üggmann, D. (2020). Women’s employmen , income and di o ce in Wes Ge many: A
causal app oach. Jou nal o Labou Ma ke Resea ch, 54, 10. h ps:// doi. o g/ 10. 1186/
s12651- 020- 00270-0
Bu khause , R. V., Duncan, G. J., & Hause , R. (1991). Wi e o au, women do wo se: A compa ison
o men and women in he Uni ed S a es and Ge many a e ma i al dissolu ion. Demog aphy, 28,
353–360.
Ce alu, M., Ridgeway, G., McCa ey, D., Mo al, A., G i in, B. A., & Bu ge e, L. (2021). TWANG.
h ps:// c an. - p oje c . o g/ web/ packa ges/ wang/ index. h ml
Cohen, J. (1988). S a is ical powe analysis o he beha io al sciences (2nd ed.). Law ence E lbaum
Associa es, Publishe s.
Couch, K. A., Tambo ini, C. R., & Reznik, G. L. (2015). The long- e m heal h implica ions o ma i al
dis up ion: Di o ce, wo k limi s, and social secu i y disabili y bene i s among men. Demog a-
phy, 52(2), 1487–1512.
Co izzi, I. (2008). Does union dissolu ion lead o unemploymen ? A longi udinal s udy o heal h
and isk o unemploymen o women and men unde going sepa a ion. Eu opean Sociological
Re iew, 24(3), 347–361.
1 3
Ea nings T ajec o ies A e Di o ce: The Legacies o  heEa ne … Page 33 o 34 23
Damme, M. V., Kalmijn, M., & Uunk, W. (2009). The employmen o sepa a ed women in Eu ope:
Indi idual and ins i u ional de e minan s. Eu opean Sociological Re iew, 25(2), 183–197.
Dehejia, R., & Wahba, S. (1999). Causal e ec s in nonexpe imen al s udies: Ree alua ing he e alua-
ion o aining p og ams. Jou nal o he Ame ican S a is ical Associa ion, 94(448), 1053–1062.
h ps:// doi. o g/ 10. 2307/ 26699 19
Duncan, G. J., & Ho man, S. D. (1985). A econside a ion o he economic consequences o ma i al
dissolu ion. Demog aphy, 22(4), 485–497.
Geisle , E., & K eyen eld, M. (2019). Policy e o m and a he s’ use o pa en al lea e in Ge many:
The ole o educa ion and wo kplace cha ac e is ics. Jou nal o Eu opean Social Policy, 29(2),
273–291. h ps:// doi. o g/ 10. 1177/ 09589 28718 765638
Geye , J., Haan, P., & W ohlich, K. (2015). The e ec s o amily policy on ma e nal labo supply:
Combining e idence om a s uc u al model and a quasi-expe imen al app oach. Labou Eco-
nomics, 36, 84–98.
Gue zgen, N., & Hank, K. (2018). Ma e ni y lea e and mo he s’ long- e m sickness absence: E idence
om Wes Ge many. Demog aphy, 55, 587–615. h ps:// doi. o g/ 10. 1007/ s13524- 018- 0654-y
Hi ano, K., & Imbens, G. W. (2001). Es ima ion o causal e ec s using p opensi y sco e weigh ing: An
applica ion o da a on igh hea ca he e iza ion. Heal h Se ices & Ou comes Resea ch Me hodol-
ogy, 2, 259–278. h ps:// doi. o g/ 10. 1023/A: 10203 71312 283
Ho mann, B., K eyen eld, M., & Uhlendo , A. (2017). Job displacemen and i s bi h o e he business
cycle. Demog aphy, 54, 933–959.
Hogendoo n, B. (2022). Why do socioeconomic di e ences in women’s li ing s anda ds con e ge a e
uniondissolu ion? Eu opean Jou nal o Popula ion, 38, 577–622.
Hube , S., Neube ge , F., & Somme , M. (2020). Alleine ziehend, alleinbezahlend? Kindesun e hal ,
Un e hal s o schuss und G ünde ü den Un e hal saus all. Zei sch i ü Soziologie de E ziehung
und Sozialisa ion, 40, 19–38.
Huinink, J., B üde l, J., Nauck, B., Walpe , S., Cas iglioni, L., & Feldhaus, M. (2011). Panel analysis o
in ima e ela ionships and amily dynamics (pai am): F amewo k and design o pai am. Zei sch i
ü Familien o schung, 23(1), 77–101.
Jones, K. W., & Lewis, D. J. (2015). Es ima ing he coun e ac ual impac o conse a ion p og ams on
land co e ou comes: The ole o ma ching and panel eg ession echniques. PLoS ONE, 10(10),
e0141380. h ps:// doi. o g/ 10. 1371/ jou n al. pone. 01413 80
Kalmijn, M. (2005). The e ec s o di o ce on men’s employmen and social secu i y his o ies. Eu opean
Jou nal o Popula ion, 21(4), 347–366.
Keck, W., Radenacke , A., B üggmann, D., K eyen eld, M., & Mika, T. (2019). S a u o y pension insu -
ance accoun s and di o ce: A new Scien i ic Use File. Jah büche ü Na ionalökonomie und S a is-
ik, 240, 825–835.
Konie zka, D., & K eyen eld, M. (2002). Women’s employmen and non-ma i al childbea ing: A Com-
pa ison be ween Eas and Wes Ge many in he 1990s. Popula ion - E, 57(2), 331–357.
Köppen, K., & T appe, H. (2019). The gende ed di ision o labo and i s pe cei ed ai ness: Implica ions
o childbea ing in Ge many. Demog aphic Resea ch. h ps:// doi. o g/ 10. 4054/ DemRes. 2019. 40. 48
Lee, B. K., Lessle , J., & S ua , E. A. (2010). Imp o ing p opensi y sco e weigh ing using machine
lea ning. S a is ics in Medicine, 29(3), 337–346. h ps:// doi. o g/ 10. 1002/ sim. 3782
Lenze, A., & Funcke, A. (2016). Alleine ziehende un e D uck: Rech liche Rahmenbedingungen, inan-
zielle Lage und Re o mbeda . Be elsmann S i ung.
Leopold, T. (2018). Gende di e ences in he consequences o di o ce: A mul iple-ou come compa ison
o o me spouses. Demog aphy, 55, 769–797.
Ludwig, V., & B üde l, J. (2021). Wha you need o know when es ima ing impac unc ions wi h panel
da a o demog aphic esea ch. Compa a i e Popula ion S udies, 46, 453–486.
Ma iny, D. (2012). Cu en de elopmen s in he na ional laws o main enance: A compa a i e analysis.
Eu opean Jou nal o Law Re o m, 65, 65–85.
McCa ey, D. F., Ridgeway, G., & Mo al, A. R. (2004). P opensi y sco e es ima ion wi h boos ed
eg ession o e alua ing causal e ec s in obse a ional s udies. Psychological Me hods, 9(4), 403–
425. h ps:// doi. o g/ 10. 1037/ 1082- 989X.9. 4. 403
McManus, P. A., & DiP e e, T. A. (2001). Lose s and winne s: The inancial consequences o sepa a ion
and di o ce o men. Ame ican Sociological Re iew, 66(2), 246–268.
Miles, J., & Sche pe, J. M. (2020). The legal consequences o dissolu ion: P ope y and inancial sup-
po be ween spouses. In J. Eekelaa & R. Geo ge (Eds.), Rou ledge Handbook o Family Law and
Policy. 2nd edi ion. London: Rou ledge.
D.B üggmann, M.K eyen eld
1 3
23 Page 34 o 34
Pane, J., Bu ge e, L., G i in, B. A., & McCa ey, D. (2021). OV ool. h ps:// c an. - p oje c . o g/ web/
packa ges/ OV ool/ index. h ml
Peschel-Gu zei , L. M. (2008). Un e hal s ech ak uell. Die Auswi kungen de Un e hal s e o m au die
Be a ungsp axis: Nomos.
Pi acchio, R., Pe e sen, M. L., & an de Laan, M. (2015). Imp o ing p opensi y sco e es ima o s’
obus ness o model misspeci ica ion using supe lea ne . Ame ican Jou nal o Epidemiology,
181(2), 108–119. h ps:// doi. o g/ 10. 1093/ aje/ kwu253
Raz-Yu o ich, L. (2013). Di o ce penal y o di o ce p emium? A longi udinal analysis o he conse-
quences o di o ce o men’s and women’s economic ac i i y. Eu opean Sociological Re iew, 29(2),
373–385. h ps:// doi. o g/ 10. 1093/ es / jc 073
Ridgeway, G., McCa ey, D., Mo al, A., Ce alu, M., Bu ge e, L., Pane, J., & G i in, B. A. (2021).
Toolki o weigh ing and analysis o nonequi alen g oups: A guide o he wang package. RAND
Co po a ion.
Rosenbaum, P. R., & Rubin, D. (1985). Cons uc ing a con ol g oup using mul i a ia e ma ched sam-
pling me hods ha inco po a e he p opensi y sco e. The Ame ican S a is ician, 39, 33–38.
Rö hel, A. (2009). BGH, 18. 3. 2009 — XII ZR 74/08. Daue des nachehelichen Be euungsun e hal s
nach neuem Un e hal s ech .
Sianesi, B. (2004). An e alua ion o he swedish sys em o ac i e labo ma ke p og ams in he 1990s.
The Re iew o Economics and S a is ics, 86(1), 133–155. h ps:// doi. o g/ 10. 1162/ 00346 53043 23023
723
S a is isches Bundesam . (2020). Rech sp lege. Familienge ich e. Fachse ie 10 Reihe 2.2. Wiesbaden:
S a is isches Bundesam .
Tambo ini, C. R., Couch, K. A., & Reznik, G. L. (2015). Long- e m impac o di o ce on women’s ea n-
ings ac oss mul iple di o ce windows: A li e cou se pe spec i e. Ad ances in Li e Cou se Resea ch,
26, 44–59.
Thielemans, G., & Mo elmans, D. (2019). Female labou o ce pa icipa ion a e di o ce: How employ-
men his o ies ma e . Jou nal o Family Economic Issues, 40, 180–193.
Thomas, M. J., Mulde , C. H., & Cooke, T. J. (2017). Linked li es and cons ained spa ial mobili y: The
case o mo es ela ed o sepa a ion among amilies wi h child en. T ansac ions o he Ins i u e o
B i ish Geog aphe s, 42(2), 597–611.
Van Damme, M., Kalmijn, M., & Uunk, W. (2009). The employmen o sepa a ed women in Eu ope:
Indi idual and ins i u ional de e minan s. Eu opean Sociological Re iew, 25(2), 183–197.
Walpe , S., En lei ne -Phleps, C., & Langmeye , A. (2021). Sha ed physical cus ody a e pa en al sepa-
a ion: E idence om Ge many: In L. Be na di, D. Mo elmans (Eds.), Sha ed physical cus ody,
Eu opean s udies o popula ion (Vol.25, pp.285–330).
Weiss, R. S. (1984). The impac o ma i al dissolu ion on income and consump ion in single-pa en
households. Jou nal o Ma iage and he Family, 46(1), 115–127.
Wes , S. G., Biesanz, J. C., & Pi s, S. C. (2000). Causal in e ence and gene aliza ion in ield se ings:
Expe imen al and quasi-expe imen al designs. In H. T. Reis & C. M. Judd (Eds.), Handbook o
esea ch me hods in social and pe sonali y psychology (pp. 40–84). Camb idge Uni e si y P ess.
Wes eich, D., Lessle , J., & Funk, M. J. (2010). P opensi y sco e es ima ion: Neu al ne wo ks, suppo
ec o machines, decision ees (CART), and me a-classi ie s as al e na i es o logis ic eg ession.
Jou nal o Clinical Epidemiology, 63(8), 826–833. h ps:// doi. o g/ 10. 1016/j. jclin epi. 2009. 11. 020
Willenbache , B. (2010). Die Umges al ung des Geschlech e kon ak es du ch das nacheheliche Un e h-
al s ech . In M. Co ie , J. Es e mann& M. W ase (Eds.). Wie wi k Rech ? Baden-Baden: Nomos:
369–389.
Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps
and ins i u ional a ilia ions.