The gender gap in lifetime earnings: A microsimulation approach

Glaubi z, Rick; Ha nack‐Ebe , As id; We e , Mi iam
A icle — Published Ve sion
The gende gap in li e ime ea nings: A mic osimula ion
app oach
LABOUR
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Glaubi z, Rick; Ha nack‐Ebe , As id; We e , Mi iam (2024) : The gende gap in
li e ime ea nings: A mic osimula ion app oach, LABOUR, ISSN 1467-9914, Wiley Pe iodicals, Inc.,
Hoboken, NJ, Vol. 38, Iss. 4, pp. 425-474,
h ps://doi.o g/10.1111/lab .12274
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/313683
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/
ORIGINAL ARTICLE
The gende gap in li e ime ea nings:
A mic osimula ion app oach
Rick Glaubi z
1,2
| As id Ha nack-Ebe
1
| Mi iam We e
1
1
School o Business & Economics, F eie
Uni e si ä Be lin, Be lin, Ge many
2
Cen e o E idence-Based Heal hca e
(ZEGV), Uni e si y Hospi al and Facul y
o Medicine Ca l Gus a Ca us a TU
D esden, D esden, Ge many
Co espondence
Rick Glaubi z, Cen e o E idence-Based
Heal hca e (ZEGV), Uni e si y Hospi al
and Facul y o Medicine Ca l Gus a
Ca us a TU D esden, Fe sche s aße
74, 01307, D esden, Ge many.
Email: [email p o ec ed]
Abs ac
To ob ain a mo e comple e unde s anding o he
pe sis ing gende ea nings gap in Ge many, his pape
in es iga es bo h he c oss-sec ional and li e ime dimen-
sion o gende inequali ies. Based on a dynamic mic o-
simula ion model, we analyse how gende di e ences
accumula e o e wo k li es o examine he li e ime
dimension o he gende gap. We es ima e an a e age
gende gap in li e ime ea nings o 51.5 pe cen o bi h
coho s 1964–72. We show ha his unadjus ed gende
li e ime ea nings gap inc eases s ongly wi h he numbe
o child en, anging om 17.3 pe cen o childless
women o 68.0 pe cen o women wi h h ee o mo e
child en. Resul s om a coun e ac ual analysis app oach
show an adjus ed gende gap in li e ime ea nings o
a ound 10 pe cen , sugges ing ha he gende gap in li e-
ime ea nings is a he d i en by gende di e ences in
obse able cha ac e is ics han by di e ences in ewa ds.
JEL CLASSIFICATION
D15, D31, J16, J22, J31
1|INTRODUCTION
As mos esea ch on he gende pay gap has ocused on di e ences in mon hly o annual ea n-
ings da a, e idence on how gende inequali ies add up o e he li e cou se is s ill limi ed. In
Accep ed: 26 Ap il 2024
DOI: 10.1111/lab .12274
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s License, which pe mi s
use and dis ibu ion in any medium, p o ided he o iginal wo k is p ope ly ci ed, he use is non-comme cial and no modi ica ions o
adap a ions a e made.
© 2024 The Au ho s. LABOUR published by Fondazione Giacomo B odolini and John Wiley & Sons L d.
Labou . 2024;38:425–474. wileyonlinelib a y.com/jou nal/lab 425
con as o c oss-sec ional analysis, which gene ally only ocuses on a snapsho o an indi id-
ual's employmen ca ee , a li e cycle pe spec i e acknowledges ha ea nings a e ansien
ac oss indi iduals' ca ee s. Fo example, ea nings can be empo a ily low du ing educa ional
aining o e en ze o du ing imes o unemploymen o labou ma ke inac i i y. The analysis
o accumula ed ea nings o e he en i e ca ee (‘li e ime ea nings’) o e s mo e comp ehensi e
insigh s in o indi iduals' long- e m posi ion in he ea nings dis ibu ion and is mo e closely
linked o indi iduals' li e chances (see, e.g., Co neo, 2015; Tambo ini e al., 2015). The e o e,
he concep o li e ime ea nings is sui able o examine he ex en o which gende inequali ies
accumula e o e he li e cycle. Howe e , due o high da a equi emen s, he e is only sca ce
empi ical e idence on gende li e ime ea nings gaps (e.g., Boll e al., 2017; Gu enen e al., 2021,
2022). In addi ion, hese s udies a e o en limi ed by hei use o adminis a i e da a and subse-
quen lack o amily- ela ed in o ma ion (e.g., numbe o child en, ma i al s a us). Because he
a e age labou ma ke pa icipa ion o women is lowe han ha o men a bo h he in ensi e
and ex ensi e ma gin due o, e.g., di e en e ec s o pa en hood (see, e.g., Goldin, 2014; Kle en
e al., 2019), an analysis o he household con ex is necessa y o a mo e comp ehensi e unde -
s anding o he unde lying d i e s o gende di e en ials in li e ime ea nings.
This s udy uses he Socio-Economic Panel (SOEP) o shed ligh on he ole o women's am-
ily backg ounds in gende di e ences, om bo h a c oss-sec ional and a li e ime pe spec i e.
Using an Oaxaca–Blinde decomposi ion, we ind ha he gaps can la gely be explained by bo h
he ex ensi e and in ensi e ma gins o labou . On a e age, women ha e less wo k expe ience
and wo k ewe hou s han men, explaining a la ge pa o women's lowe ea nings.
To u he ake ad an age o he de ailed socioeconomic and amily backg ound in o ma-
ion in he SOEP su ey compa ed wi h adminis a i e da a sou ces, we use a dynamic mic o-
simula ion model o ob ain ull employmen biog aphies, and subsequen ly li e ime ea nings
da a. In con as o he exis ing s udies using adminis a i e da a, his allows us o analyse he
ex en o which gende gaps in li e ime ea nings a y by amily backg ound (numbe o chil-
d en). Fu he mo e, his app oach also leads o a mo e comp ehensi e sample han he ones o
ea lie s udies o Ge many (Boll e al., 2017; Bönke e al., 2015) as we a e, o he i s ime,
able o include sel -employed indi iduals, ci il se an s and women wi h longe unemploy-
men /inac i i y spells. We ind ha women ea n on a e age 51.5 pe cen less han men o e
hei wo k li e. This unadjus ed gende gap in li e ime ea nings co ela es la gely wi h he num-
be o child en and anges om 17.3 pe cen o childless women o 68.0 pe cen o women
wi h h ee child en o mo e.
To in es iga e which pa o he obse ed gende gap in li e ime ea nings can be explained
by he obse able di e ences in he dis ibu ion o cha ac e is ics (e.g., wo k expe ience, le el
o educa ion) ac oss gende and which pa is due o di e ences in labou ma ke e u ns o
cha ac e is ics, we es ima e women's coun e ac ual li e ime ea nings. We ind ha a ound
80 pe cen o he obse ed li e ime ea nings gap can be explained by di e en cha ac e is ics
ac oss men and women, leading o an adjus ed gende li e ime ea nings gap o 10 pe cen .
Ou pape is ela ed o h ee di e en s ands o li e a u e. Fi s and mos gene ally, i con-
ibu es o he ex ensi e li e a u e on he gende gap in pay and i s d i e s. Exis ing s udies
show ha a la ge ex en o he pay gap can be a ibu ed o ewe hou s wo ked and highe dis-
con inui y o emale employmen biog aphies (e.g., Be and e al., 2010; Blau & Kahn, 2017).
1
The pe sis ence o his gende ea nings inequali y is mainly due o di e en e ec s o pa en -
hood on men's and women's labou ma ke beha iou , and consequen ly hei ea nings (see, e.
g., Angelo e al., 2016; Bü iko e e al., 2018; Kle en e al., 2021; Kle en & Landais, 2017;
Wald ogel, 1998). In line wi h p e ious s udies (e.g., Gallen e al., 2019; Goldin, 2014; Juhn &
426 GLAUBITZ ET AL.
McCue, 2017), we con i m ha gende di e ences in annual ea nings inc ease du ing he
pe iod o amily o ma ion, peak a ound age 40 and slowly dec ease un il e i emen , leading o
an in e se U shape o he gende annual ea nings gap o e he wo k li e.
S udies o Ge many show ha he c oss-sec ional ea nings gap be ween mo he s and non-
mo he s a e la gely d i en by domes ic wo k and childca e du ies (e.g., Beblo & Wol , 2002;
Ej næs & Kunze, 2013). S ikingly, child penal ies on women's pay a e high in Ge many com-
pa ed wi h o he coun ies (see, e.g., Kle en e al., 2019). This is o en a ibu ed o longe
ma e nal lea e en i lemen and a highe a e o pa - ime wo k o women in Ge many (see, e.
g., Gangl & Zie le, 2009; Ha kness & Wald ogel, 2003). Howe e , mo e ecen s udies also s ess
he in luence o ela i e conse a i e gende no ms in Ge many in his con ex (e.g., Kle en
e al., 2019,2020).
Second, ou s udy adds o he sca ce li e a u e on li e ime ea nings and speci ically o wha
ex en hese di e by gende . Due o he high da a equi emen s, he li e a u e on he gende
pay gap and i s e olu ion has p ima ily ocused on c oss-sec ional hou ly wages, annual ea n-
ings o ea nings o e a sho ime pe iod. Using adminis a i e da a o he Uni ed S a es,
Gu enen e al. (2021) show ha he ac ion o women among li e ime op ea ne s is signi i-
can ly lowe han ha o men o bi h coho s 1956–58. On a e age, li e ime op ea ne s in he
Uni ed S a es end o be indi iduals who expe ience high ea nings g ow h o e he i s hal o
hei li e cycle— he pe iod when he gende gap inc eases he mos , likely due o amily- ela ed
easons. In a la e s udy, Gu enen e al. (2022) p o ide e idence ha he la ge gende li e ime
ea nings gap is na owing o e ime, wi h women's median li e ime ea nings inc easing while
men's median li e ime ea nings dec eases o younge bi h coho s.
Using adminis a i e da a om he Ge man Pension Regis e (VSKT), Bönke e al. (2015)
ind e idence ha in agene a ional li e ime ea nings inequali y o Wes Ge man men bo n
be ween 1935 and 1969 has inc eased, la gely due o losses in he bo om o he li e ime ea n-
ings dis ibu ion. They also supplemen hei wo k wi h addi ional esul s on Wes Ge man
women. Howe e , due o da a es ic ions, hei da a only includes women wi h s able employ-
men biog aphies. The e o e, he VSKT da a is no ep esen a i e o mos women mainly due
o he high a e o inac i i y among women o olde coho s and should no be used o es ima -
ing he gende li e ime ea nings gap in Ge many. Closes o ou pape is he s udy by Boll e al.
(2017) analysing he gende li e ime ea nings gap in Ge many. Using he adminis a i e Sample
o In eg a ed Labou Ma ke Biog aphies (SIAB), hey es ima e an unadjus ed gende li e ime
ea nings gap o 46 pe cen o Wes Ge man bi h coho s 1950–64. They show ha he gende
gap widens signi ican ly du ing he age o amily o ma ion and ha gende di e ences in wo k
expe ience and hou s wo ked explains a ound wo- hi ds o his o e all gende li e ime ea n-
ings gap. Howe e , SIAB da a does no o e any in o ma ion abou indi iduals' amily back-
g ound. Hence, o he bes o ou knowledge, ou s udy is he i s o ex ensi ely examine he
in luence o pa en hood in he con ex o gende di e en ials in li e ime ea nings.
Thi d, ou s udy con ibu es om a me hodological poin o iew o he li e a u e on he
implemen a ion o dynamic mic osimula ion models o he simula ion o missing in o ma ion
(e.g., Le ell & Shaw, 2016; Li & O'Donoghue, 2013; Zucchelli e al., 2012). A dynamic mic o-
simula ion app oach e e s o a eg ession-based simula ion which p edic s he ansi ion p ob-
abili ies o di e en uni s (e.g. indi iduals o households) o mo ing om one s a e o ano he
be ween wo di e en poin s in ime. The e o e, in con as o s udies using a splicing app oach
(e.g., G abka & Goebel, 2017; Wes e meie e al., 2012) whe e sequences o exis ing biog aphies
a e s i ched oge he o cons uc ull li e-cycle da a, he mic osimula ion app oach ypically
‘ages’ he da a yea by yea (Li & O'Donoghue, 2013). We apply a dynamic mic osimula ion
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 427
model o SOEP su ey da a o ob ain comple e ea nings biog aphies, which acili a es li e ime
ea nings analyses. Combining simula ion models wi h su ey da a is a well-es ablished me hod
o deal wi h missing obse a ions and panel a i ion, which o en impede using su ey da a o
conduc long- e m analyses (see, e.g., B own e al., 2009; Co onado e al., 2011). Fo Ge many,
e.g., he e a e exis ing s udies simula ing employmen biog aphies using SOEP da a (e.g., Bonin
e al., 2015; Geye & S eine , 2014; Hänisch & Klos, 2016).
The nex sec ion in oduces ou da ase and s a s by analysing c oss-sec ional gende di e -
ences in hou ly wages and annual ea nings o e he wo k li e by using an Oaxaca–Blinde
decomposi ion. Sec ion 3desc ibes ou mic osimula ion app oach o ob ain ull wo k biog a-
phies and p esen s ou es ima es o he unadjus ed and adjus ed gende li e ime ea nings gap.
Sec ion 4concludes.
2|CROSS-SECTIONAL ANALYSIS
The c oss-sec ional analysis explo es how gende gaps in hou ly wages and annual ea nings
de elop wi h inc easing age and o in es iga e i sho - e m di e ences al eady ollow ce ain
pa e ns ac oss gende . This i s s ep is c ucial o subsequen ly be e unde s and how gende
inequali ies in labou ma ke cha ac e is ics and ea nings add up o equalize o e he en i e
wo k li e.
2.1 |Da a and me hodology
Ou s udy is based on he Ge man SOEP. The SOEP is a ep esen a i e annual panel su ey
ques ioning abou 30,000 indi iduals ac oss 15,000 households since 1984. In con as o admin-
is a i e da a, he SOEP includes a ich se o socioeconomic a iables, de ailed labou ma ke
in o ma ion and household backg ound including in o ma ion on he pa ne and child en.
2
Speci ically, we use he 35 h wa e o he SOEP comp ising da a o he yea s 1984–2018.
We es ic ou c oss-sec ional analysis o bi h coho s 1940–79. These a e he same bi h
coho s used o he unde lying eg essions o ou mic osimula ion model in Sec ion 3.We
obse e hese coho s a leas once be ween he ages o 38 and 44 in he SOEP. This age es ic-
ion is c ucial as i is he age ame when indi iduals' c oss-sec ional ea nings show he highes
co ela ion wi h li e ime ea nings (Bjö klund, 1993; Bönke e al., 2015) and is he e o e needed
o success ully simula e li e-cycle p o iles in Sec ion 3. Fu he mo e, we ocus on Wes Ge man
indi iduals because hose bo n in Eas Ge many we e only included in he SOEP a e
Ge man euni ica ion in 1990. The poo compa abili y o he Fede al Republic o Ge many and
he Ge man Democ a ic Republic wi h espec o labou ma ke ins i u ions and economic sys-
ems does no allow us o simula e missing in o ma ion o Eas Ge mans be o e 1990. I is
impo an o no e ha , by ocusing only on employed indi iduals, we a e solely analysing he
obse ed dis ibu ion o ea nings and no he coun e ac ual dis ibu ion ha would be
obse ed i e e yone we e in employmen and had posi i e ea nings.
Sec ion 2 ocuses on he e olu ion o c oss-sec ional hou ly wages and annual ea nings wi h
inc easing age o e he wo k li e. This app oach sheds ligh on wo main componen s o he
gende gap in li e ime ea nings; he gende gap in hou ly wages shows he di e ences in
he compensa ion be ween women and men o 1 h o hei wo k, while he gap in annual ea n-
ings e eals dissimila i ies d i en by he a ia ion in wo king hou s.
428 GLAUBITZ ET AL.

We use an Oaxaca–Blinde decomposi ion (see Blinde , 1973; Oaxaca, 1973) o in es iga e
how much o he di e ence in he obse ed gende gap is d i en by di e en obse able cha ac-
e is ics be ween men and women and how much can be a ibu ed o di e en e u ns o cha -
ac e is ics wi hin he labou ma ke .
3
Using his decomposi ion app oach he gende gap Gin
he labou ma ke ou come a iable L(he e, loga i hmic hou ly wage and loga i hmic annual
ea nings) is de ined as:
Gx¼EL
mx
ðÞEL
x

:ð1Þ
The e o e, Gis he gende di e en ial be ween he means o ou come L o men m
ðÞ
and
women ðÞa age x. We can hen di ide he gende gap in o wo pa s. Fi s , he endowmen
pa , which is he componen o he gende gap which is due o di e ences in he dis ibu ion
o obse able cha ac e is ics be ween men and women. And second, he coe icien pa , which
accoun s o di e ences in e u ns o cha ac e is ics. Hence, he coe icien pa shows he gen-
de d i en di e ence o he labou ma ke 's willingness o pay o he same cha ac e is ics
ob ained by ei he men o women. Howe e , no e ha he coe icien pa may also include
gende di e ences ha emain unexplained in ou model due o da a and model es ic ions.
We un he ollowing eg ession model sepa a ely by sex sðÞand age xðÞ o he labou ou -
come L
4
:
Ls,i,x¼αs,i,xþβs,i,xZs,i,xþϵs,i,x,Eεs,x
ðÞ¼0, sF,M
g
,x20,60½,ð2Þ
whe e Zis a ec o o con ol a iables including wo k expe ience measu ed as numbe o
wo king yea s, ull- ime o pa - ime wo k, wo k sec o , highes educa ion le el, ma i al s a us
and numbe o child en. In addi ion, we con ol o coho and ime e ec s.
5
2.2 |Hou ly wage
O e all, employed men ha e signi ican ly highe hou ly wages han employed women (see
Table A1). A he beginning o hei wo k li e a age 20, men ea n on a e age €9.37 pe hou
while women's a e age wage is only €7.97 pe hou . In line wi h esul s ound by he Fede al
S a is ical O ice (S a is isches Bundesam , 2017), he a e age hou ly wages o men in ou sam-
ple hen almos iples o e he wo k li e o €26.13 pe hou a age 60. In con as , women's
hou ly wages only inc ease o €17.48, al eady showing signi ican gende di e ences in wage
g ow h o e he wo k li e.
The solid line in Figu e 1shows he e olu ion o he gende gap in hou ly wages in log
poin s om age 20 o 60. No ably, he gende gap emains s able o e he ea ly yea s o wo k
li e. A age 25, men's hou ly wages a e only 0.059 log poin s highe han women's and he di e -
ence is s ill insigni ican (see also Table 1). Howe e , du ing he ime o amily o ma ion and
childca e, his gap d as ically widens up o a highly signi ican di e ence o 0.378 log poin s a
age 45.
6
A e wa ds, he g ow h o he gende gap in hou ly wages slows down and emains el-
a i ely s able wi h a peak a age 55. This inding is consis en o all coho s (see Figu e A2). In
line wi h ou indings, p e ious s udies also documen ed a widening o he gende wage gap
o e he li e cycle (e.g., Ande son e al., 2002; Angelo e al., 2016; Ty owicz e al., 2018).
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 429
The esul s o he Oaxaca–Blinde decomposi ion a e displayed by he g ey lines in Figu e 1
and also in Table 1. Visibly, he widening o he gende gap in hou ly wages o e he wo k li e
can be explained by he inc ease in he endowmen pa , while he coe icien pa o he gende
gap shapes i s o e all end. A younge ages, he di e en dis ibu ion o cha ac e is ics does
no play a ole ye . The e o e, a he beginning o wo k li e all wage di e ences be ween men
and women a e due o di e en e u ns o labou ma ke cha ac e is ics. Main di e ences in
cha ac e is ics such as wo k expe ience o amily backg ound widen only la e in li e; a e age
25, he high and signi ican coe icien s o wo k expe ience in Table 1show ha he inc ease
o he endowmen pa can la gely be explained by women's lowe gain o wo k expe ience wi h
inc easing age. By he age o 60, men ha e accumula ed on a e age 37.32 yea s o ull- ime and
1.09 yea s o pa - ime wo k expe ience, whe eas women ha e accumula ed on a e age only
19.65 yea s o ull- ime and 13.32 yea s o pa - ime wo k expe ience (see Table A1). Ou esul s
show ha hese la ge di e ences in wo k expe ience a e c ucial in explaining he gende gap in
hou ly wages. By he end o wo k li e, di e ences in wo k expe ience accoun o 0.309 log
poin s o he o e all gende wage gap o 0.340 log poin s. Hence, a ound 90 pe cen o he o e -
all gende gap o 40.5 pe cen in hou ly wages can be explained by di e ences in wo k
expe ience.
In con as o he s able g ow h o he endowmen pa , he e olu ion o he coe icien pa
ollows a sligh in e se U-shape. A age 20, he gende gap canno be explained h ough di e -
ences o cha ac e is ics ac oss gende s, bu he coe icien pa amoun s o 0.126 log poin s. This
means ha e en i we obse ed he same labou ma ke cha ac e is ics in women and men,
men's wages would be 0.126 log poin s (13.4 pe cen ) highe han women's wages a his age.
The coe icien pa o he gende gap peaks a 0.247 log poin s (28.0 pe cen ) a age 45 and
hen declines again o a di e ence o 0.042 log poin s (4.3 pe cen ) jus be o e e i emen .
7
In
con as o he endowmen pa , none o he a iable g oups ha e a cons an signi ican in lu-
ence on he o e all gende gap, including he cons an i sel .
8
The e o e, no one indi idual
FIGURE 1 Gende gap in hou ly wages. Only employed indi iduals a e conside ed. Coho s 1940–79,
weigh ed sample. Sou ce: Own calcula ions based on SOEP 35.
430 GLAUBITZ ET AL.
TABLE 1 Oaxaca–Blinde decomposi ion o hou ly wage gende gap.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 Age 25 Age 30 Age 35 Age 40 Age 45 Age 50 Age 55 Age 60
O e all
Men 1.963***
(0.040)
2.563***
(0.021)
2.771***
(0.015)
2.912***
(0.012)
2.980***
(0.012)
3.008***
(0.013)
3.019***
(0.016)
3.054***
(0.022)
3.003***
(0.026)
Women 1.945***
(0.033)
2.503***
(0.024)
2.586***
(0.019)
2.628***
(0.017)
2.637***
(0.016)
2.630***
(0.017)
2.634***
(0.017)
2.622***
(0.020)
2.663***
(0.031)
Di e ence 0.018 (0.052) 0.059 (0.031) 0.186***
(0.024)
0.284***
(0.021)
0.343***
(0.020)
0.378***
(0.021)
0.385***
(0.024)
0.432***
(0.030)
0.340***
(0.041)
Endowmen 0.108**
(0.040)
0.031
(0.018)
0.033*
(0.015)
0.107***
(0.018)
0.107***
(0.020)
0.131***
(0.022)
0.196***
(0.026)
0.200***
(0.028)
0.297***
(0.043)
Coe icien 0.126*
(0.050)
0.091**
(0.033)
0.152***
(0.028)
0.177***
(0.026)
0.235***
(0.029)
0.247***
(0.028)
0.189***
(0.036)
0.231***
(0.040)
0.042
(0.060)
Endowmen
Child en 0.003 (0.005) 0.000
(0.003)
0.001 (0.001) 0.002
(0.002)
0.004
(0.003)
0.009*
(0.004)
0.015*
(0.007)
0.014
(0.011)
0.013
(0.013)
Ma ied 0.001 (0.004) 0.001
(0.004)
0.004 (0.002) 0.002 (0.002) 0.003 (0.002) 0.005* (0.002) 0.000 (0.002) 0.002
(0.003)
0.010
(0.008)
Expe ience 0.068*
(0.030)
0.024*
(0.011)
0.084***
(0.013)
0.167***
(0.019)
0.207***
(0.022)
0.224***
(0.028)
0.228***
(0.026)
0.264***
(0.032)
0.309***
(0.049)
Pa ime 0.002 (0.008) 0.041**
(0.013)
0.027
(0.015)
0.067***
(0.019)
0.096***
(0.019)
0.066***
(0.019)
0.026
(0.019)
0.056*
(0.024)
0.030
(0.024)
Educa ion 0.008
(0.007)
0.019**
(0.006)
0.009
(0.006)
0.021**
(0.007)
0.019**
(0.006)
0.022***
(0.006)
0.031***
(0.007)
0.030***
(0.008)
0.045***
(0.011)
Coho 0.001
(0.005)
0.002
(0.003)
0.000
(0.002)
0.001 (0.002) 0.003 (0.002) 0.001 (0.001) 0.002 (0.001) 0.003
(0.002)
0.001
(0.002)
Sec o 0.037
(0.029)
0.009 (0.011) 0.019*
(0.008)
0.017**
(0.006)
0.025***
(0.007)
0.045***
(0.007)
0.024***
(0.007)
0.024*
(0.010)
0.022*
(0.011)
Coe icien
(Con inues)
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 431
TABLE 1 (Con inued)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 Age 25 Age 30 Age 35 Age 40 Age 45 Age 50 Age 55 Age 60
Child en 0.003 (0.005) 0.030 (0.020) 0.095***
(0.025)
0.063 (0.034) 0.002 (0.038) 0.007 (0.046) 0.075
(0.044)
0.048
(0.053)
0.080
(0.088)
Ma ied 0.006
(0.010)
0.021 (0.027) 0.041 (0.031) 0.052 (0.034) 0.090* (0.038) 0.008
(0.041)
0.100* (0.045) 0.047
(0.056)
0.020
(0.074)
Expe ience 0.243***
(0.063)
0.207 (0.129) 0.067 (0.128) 0.094
(0.127)
0.058
(0.112)
0.005 (0.170) 0.150
(0.226)
0.196
(0.280)
0.973
(0.995)
Pa ime 0.013 (0.027) 0.031 (0.037) 0.008 (0.023) 0.021
(0.019)
0.032 (0.021) 0.032 (0.021) 0.030 (0.025) 0.145***
(0.036)
0.036
(0.054)
Educa ion 0.329 (0.195) 0.144
(0.174)
0.210
(0.141)
0.105
(0.230)
0.564*
(0.256)
0.183 (0.237) 0.568 (0.295) 0.131
(0.840)
0.722
(0.648)
Coho 0.083 (0.044) 0.023 (0.085) 0.002 (0.036) 0.007
(0.125)
0.091 (0.049) 0.019
(0.034)
0.030 (0.041) 0.030
(0.038)
0.002
(0.040)
Sec o 0.222 (0.193) 0.445**
(0.139)
0.436**
(0.160)
0.053
(0.116)
0.066
(0.147)
0.139
(0.142)
0.037
(0.124)
0.085
(0.188)
0.153
(0.192)
Cons an 0.762**
(0.276)
0.368 (0.283) 0.587*
(0.255)
0.343 (0.308) 0.708* (0.322) 0.186 (0.317) 0.276
(0.383)
0.125
(0.887)
0.308
(1.090)
N765 1782 3053 4323 5356 5592 4304 2866 1758
No e: S anda d e o s in pa en heses; The s a s e e o he ollowing signi icance le el: *p< 0.05; **p< 0.01; ***p< 0.001. The di e en d i e s a e summa ized as ollowed: ‘Child en’:
Numbe o child en; ‘Ma ied’: Dummy a iable on ma i al s a us, ‘Expe ience’: To al yea s o wo king ull ime, pa ime o being inac i e (also squa ed); ‘Pa ime’: Dummy a iable
indica ing ull ime o pa ime wo k; ‘Educa ion’: Dummy a iables indica ing highes le el o educa ional a ainmen , ‘Sec o ’: Occupa ional sec o ; ‘Coho ’: Coho dummies. Coho s
1940–79, weigh ed sample. Sou ce: Own calcula ions based on SOEP 35.
432 GLAUBITZ ET AL.
The Family Module hen consis s o wo s eps: p edic ing ma i al s a us, including a pa -
ne ing module when necessa y, and p edic ing bi hs o child en o indi iduals wi h missing
in o ma ion. Fi s , we un logis ic eg essions sepa a ely by gende s( emale o male) and ma i-
al s a us m(single o pa ne ed) in yea o p edic he indi idual ansi ion p obabili y pma ied
o change he ma i al s a us om yea o he missing yea þ1:
pma ied
m,s, þ1¼β0þβ1Xm,s, þϵm,s, ,Eϵm,s,
ðÞ¼0,mS,P
g
,sF,M
g
, 1984,2017½:ð3Þ
The eg ession consis s o a se o explana o y a iables X including socioeconomic cha ac-
e is ics (e.g., educa ion, age, mig a ion backg ound) and labou ma ke beha iou (e.
g., employmen s a us). In addi ion, we con ol o he numbe o yea s ha an indi idual's
ma i al s a us has emained unchanged un il yea . Table A7 gi es a de ailed o e iew abou
all co a ia es included in each eg ession-based simula ion s ep.
Recall ha i Pi ≤Ni , he ma i al s a us s ays he same and i Pi >Ni , he ma i al s a us
changes. The e o e, his simula ion s ep has ou possible ou comes: Fi s , a pe son who is sin-
gle in yea can emain single in þ1. Second, ma ied indi iduals can s ay ma ied. He e we
assume ha hei pa ne s emain he same. Thi d, ma ied indi iduals in pe iod can ge
di o ced and become single in þ1.
11
And ou h, singles in yea can ge ma ied in þ1. In
his las case, we un a Pa ne Module o assign a pa ne .
12
This allows us o accoun o pa -
ne s' cha ac e is ics when simula ing amily and labou ma ke decisions. Using Mahalanobis
dis ance ma ching (Mahalanobis, 1936) we iden i y i e ‘bes ’pa ne s based on age, educa ion
and egion o each obse a ion. We hen andomly assign one o he i e po en ial pa ne s o
he indi idual. Ou ma ching p ocedu e is no unique, i.e., one indi idual can se e mul iple
imes as a ‘dono ’ o pa ne cha ac e is ics. In his way, we ensu e a su icien pool o po en-
ial pa ne s.
Nex , we simula e whe he a woman will gi e bi h o a child in he nex non-obse ed
pe iod þ1 by ma i al s a us m:
pbi h
m, þ1¼β0þβ1Xm, þϵm, ,Eϵm,
ðÞ¼0,mS,P
g
, 1984,2017½:ð4Þ
Again, X ep esen s a se o explana o y a iables including socioeconomic cha ac e is ics
like in o ma ion on exis ing child en and labou ma ke in o ma ion. The simula ion is simila
o he app oach desc ibed in he simula ion o he ma i al s a us. A e wa ds, he in o ma ion
on an indi idual's numbe o child en is upda ed acco dingly. In con as o ou ma iage simu-
la ion, bi hs a e only simula ed o women. Child en a e hen a ached o men depending on
women's amily backg ound.
Because we es ima e ansi ion likelihoods o þ1 by using in o ma ion a ailable in pe iod
, he likelihood o a change o he ma i al s a us o a childbi h in þ1 do no in luence he
ansi ion p obabili y o one ano he . The e o e, he o de in which we implemen e ili y and
ma i al ansi ions is i ele an and does no al e ou esul s.
Comple ing he Family Module o yea s 1984–2032 esul s in a sample wi h ull in o ma-
ion on amily cha ac e is ics. Figu e 4shows ha ou simula ed da a (dashed line) eplica es
he ini ial dis ibu ions be o e he simula ion (solid line) e y accu a ely. In Panel A, he
women's a e age numbe o child en inc eases s ongly un il age 35. Then, he g ow h a e
slows down and comes o a na u al s op be ween ages 40 and 45 due o biological easons.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 439

Panel B displays he pe cen age change in ma i al s a us by age. Ob iously, bo h men and
women ollow he same end o e he li e cycle. Mos changes in ma i al s a us happen in he
beginning o li e.
3.1.3 | Module 2: Labou Ma ke Module
The Labou Ma ke Module gene a es comple e in o ma ion on an indi idual's employmen
biog aphy h ough i e s ages: labou ma ke pa icipa ion, employmen s a us, ype o wo k
a angemen ( ull- ime o pa - ime), annual wo king hou s and annual ea nings. In his mod-
ule, we use bo h o wa d and backwa d simula ion as he in oduc o y su ey ques ionnai es
do no allow us o cons uc su icien wo k his o ies. Ou model desc ip ion will ocus on he
o wa d-looking simula ion componen . Howe e , he backwa d-looking pa o he simula ion
ollows he same me hodology.
In gene al, he logic and s uc u e o his module is e y simila o ou app oach in he Fam-
ily Module. We s a wi h he es ima ion o plmp
m, þ1ðÞ
, he p obabili y o an indi idual o ma i al
s a us m o change he labou ma ke pa icipa ion lmp om yea o yea þ1. The labou
ma ke pa icipa ion dummy a iable is equal o 1 i indi iduals a e unemployed o employed
and equal o 0 i hey a e no a ached o he labou ma ke (e.g., due o pa en al o sick lea e).
We un he es ima ion sepa a ely by gende sand ma i al s a us m:
0 .5 1 1.5 2
A e age numbe o child en
20 25 30 35 40 45 50 55 60
Age
(a) Numbe o child en
0 .02 .04 .06 .08 .1
Change in ma i al s a us (in %)
20 25 30 35 40 45 50 55 60
Age
Women
0 .02 .04 .06 .08 .1
Change in ma i al s a us (in %)
20 25 30 35 40 45 50 55 60
Age
Men
(b) Change in ma i al s a us
Be o e simula ion A e simula ion
FIGURE 4 Family in o ma ion be o e and a e simula ion. Panel A shows he a e age numbe o child en
o women by age be o e and a e he simula ion. Panel B demons a es he sha e o indi iduals in ou sample
changing hei ma i al s a us be o e and a e he simula ion. Sou ce: Own calcula ions based on SOEP 35.
440 GLAUBITZ ET AL.
plmp
s,m, þ1¼β0þβ1plmp
s,m, þβ2plmp
s,m, 1þβ3Xs,m, þϵs,m, ,
Eϵs,m,
ðÞ¼0, sF,M
g
,mS,P
g
, 1984,2017½:ð5Þ
Xs,m, ðÞ
is again a ec o o con ol a iables wi h socioeconomic cha ac e is ics like ma i al
s a us, pa ne 's ea nings and hei own labou ma ke in o ma ion. Fu he mo e, we include
lagged dependen a iables o accoun o pa h dependencies o e he wo k li e while s ill
modelling a dynamic da a gene a ing p ocess.
13
I indi iduals a e eco ded as no pa icipa ing
in yea þ1, we di ec ly eco d hei ea nings as ze o o þ1 and do no include hem in he
subsequen s eps. Fo indi iduals who a e ac i e in he labou ma ke , we nex un a eg ession
o es ima e he p obabili y o change hei employmen s a us pemp
s,m,e, þ1ðÞ
(employed/unem-
ployed) om yea o yea þ1. The ollowing model is un sepa a ely by gende s, ma i al s a-
us mand employmen s a us e:
pemp
s,m,e, þ1¼β0þβ1pemp
s,m,e, þβ2pemp
s,m,e, 1þβ3Xs,m,e, þϵs,m,e, ,
Eϵs,m,e,
ðÞ¼0, sF,M
g
,mS,P
g
,e0,1
g
, 1984,2017½:ð6Þ
Once mo e, he eg ession con ains a se o explana o y a iables Xs,m,e, ðÞ
including in o ma-
ion on amily and he socioeconomic backg ound. Also included in he con ol ec o is he
wo k his o y o indi iduals. To his end, we measu e wo k expe ience by yea s o ull- ime
wo k, pa - ime wo k and yea s wi hou any wo k un il yea o accoun o he di e en le els
o labou ma ke expe ience.
Indi iduals eco ded as unemployed in yea þ1 a e his i s eg ession s ep ecei e ze o
ea nings in þ1 and a e excluded om u he es ima ions. Fo all employed indi iduals, he
dynamic mic osimula ion mo es o wa d wi h a logis ic eg ession simula ing i indi iduals
wo ked ull- o pa - ime in yea þ1. In he nex s ep, we es ima e he p obabili y o changing
ull- ime o pa - ime a angemen s om yea o yea þ1:
pw
s,m, þ1ðÞ
¼β0þβ1pw
s,m, þβ2pw
s,m, 1þβ3Xs,m, þϵs,m, ,
Eϵs,m,
ðÞ¼0, sF,M
g
,mS,P
g
, 1984,2017½:ð7Þ
Again, Xs,m, ðÞ
includes he usual con ol a iables in addi ion o he labou ma ke his o y.
We can now mo e on o es ima e he p ecise numbe o annual wo king hou s in þ1 sepa-
a ely o pa - ime and ull- ime wo ke s. We use an OLS eg ession model ollowing he same
logic as he ea nings eg ession model as in oduced in Equa ion (2.8).
14
Finally, we use an ea nings eg ession o es ima e he annual ea nings ys,m, þ1ðÞ
by gende s
and ma i al s a us m:
ys,m, þ1ðÞ
¼β0þβ1ys,m, þβ2ys,m, 1þβ3Xs,m, þϵs,m, ,
Eϵs,m,
ðÞ¼0, sF,M
g
,mS,P
g
, 1984,2017½:ð8Þ
Xs,m, ðÞ
now includes in o ma ion abou he wo k his o y in yea s o ull- ime wo k, pa -
ime wo k o unemploymen , wo king hou s in and, i applicable, pa ne and child in o ma-
ion. All ea nings a e p ice-adjus ed and p esen ed in 2015 Eu o. The simula ion hen mo es o
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 441
he nex yea , e.g., þ2o 2. A e comple ing all i e s eps o he Labou Ma ke Module
be ween 1984 and 2017, all indi iduals ha e comple e employmen and ea nings in o ma ion
o p e iously unobse ed yea s. A e wa ds, we con inue he simula ion un il 2032 o ob ain
comple e biog aphical da a up o age 60.
Figu e 5shows ha ou simula ed da a (dashed line) eplica es he o iginal SOEP da a (solid
line) well, pa icula ly o Panel D (Full- ime wo k), Panel E (Wo king hou s) and Panel F
(Ea nings). Panel A (Labou Ma ke Pa icipa ion), Panel B (Employmen ) and Panel C
.5 .6 .7 .8 .9 1
LMP a e (in %)
20 30 40 50 60
Age
Women
.5 .6 .7 .8 .9 1
LMP a e (in %)
20 30 40 50 60
Age
Men
(a) Labo ma ke pa icipa ion (LMP)
.8 .85 .9 .95 1
Employmen a e (in %)
20 30 40 50 60
Age
Women
.8 .85 .9 .95 1
Employmen a e (in %)
20 30 40 50 60
Age
Men
(b) Employmen
0.05 .1 .15 .2
Unemploymen a e (in %)
20 30 40 50 60
Age
Women
0.05 .1 .15 .2
Unemploymen a e (in %)
20 30 40 50 60
Age
Men
(c) Unemploymen
.2 .4 .6 .8 1
Full ime a e (in %)
20 30 40 50 60
Age
Women
.2 .4 .6 .8 1
Full ime a e (in %)
20 30 40 50 60
Age
Men
(d) Full- ime wo k
1200 1600 2000 2400
Annual wo king hou s
20 30 40 50 60
Age
Women
1200 1600 2000 2400
Annual wo king hou s
20 30 40 50 60
Age
Men
(e) Annual wo king hou s
0 20000 40000 60000
Annual ea nings (in Eu o)
20 30 40 50 60
Age
Women
0 20000 40000 60000
Annual ea nings (in Eu o)
20 30 40 50 60
Age
Men
( ) Annual ea nings
Be o e simula ion A e simula ion
FIGURE 5 Labou ma ke in o ma ion be o e and a e simula ion. Only employed indi iduals a e
conside ed. Does no include alues o ze o annual ea nings. Coho s 1940–79, weigh ed sample. Sou ce: Own
calcula ions based on SOEP 35.
442 GLAUBITZ ET AL.
(Unemploymen ) show small de ia ions. Mos o hese di e ences occu in he beginning o he
wo k li e. These di e ences do no necessa ily diminish he quali y o ou mic osimula ion o
he ollowing wo easons: Fi s , ou sample es ic ion o indi iduals obse ed a leas once a
age 30 o olde leads o ewe obse a ions in indi iduals' ea ly wen ies. As a esul , ou SOEP
sample be o e he simula ion is no e y eliable o his age ange due o a small sample size,
and he e o e compa isons may be misleading. Second, as depic ed in Figu e 5, ea nings a e on
a e age ela i ely low a he beginning o an indi idual's wo k li e and hey inc ease o e hei
ca ee s. Consequen ly, ea nings a young age only accoun o a small sha e o li e ime ea n-
ings. Gene ally, li e ime ea nings es ima es migh be mo e eliable o indi iduals whom we
obse e in hei mid-30s o mid-40s as exis ing s udies p o ide e idence ha du ing his pe iod
he ( ank) co ela ion be ween c oss-sec ional and li e ime ea nings is pa icula ly high (see,
e.g., Bjö klund, 1993; Bönke e al., 2015; Haide & Solon, 2006).
15
A e he comple ion o bo h modules o ou dynamic mic osimula ion model, we ob ain all
ele an labou ma ke and household in o ma ion o bi h coho s 1964–72 om age 20 o
60 o p oceed wi h ou li e ime analysis.
16
O e all, he simula ed da a mi o s he da a pa e ns
be o e simula ion and ou simula ion esul s a e obus . Addi ional obus ness checks based on
a Mon e Ca lo simula ion app oach (Figu e A5 and Figu e A6) and he simula ion o pseudo-
missings (Figu e A4) can be ound in he Appendix.
3.2 |Li e ime analysis
Al hough we ha e al eady shown ha women ace lowe hou ly wages and annual ea nings,
and a e less ac i e on he labou ma ke , he c oss-sec ional analysis only shows a snapsho o
an indi idual's employmen biog aphy. A c oss-sec ional analysis does no e eal how hese di -
e en ac o s add up o e he li e cycle. Fo a be e unde s anding o when and how in li e he
gende gap de elops, we in es iga e di e ences in accumula ed ea nings o e he li e cycle o
bi h coho s 1964–72 using hei comple e biog aphy da a om age 20 o 60 ob ained om ou
mic osimula ion. To analyse he accumula ion o ea nings o e he wo k li e we ollow Bönke
e al. (2015) and use he ‘up- o-age-X’(UAX) concep . UAX ea nings e e o accumula ed
p ice-adjus ed (in 2015 Eu o) g oss annual ea nings up o a ce ain age X. In line wi h he s udy
by Bönke e al. (2015), we de ine li e ime ea nings as UA60 ea nings.
3.2.1 | Gende gap in li e ime ea nings
To analyse he gende gap in li e ime ea nings, we now ocus on non-loga i hmic incomes
a he han loga i hmic incomes as used in he Oaxaca–Blinde decomposi ion in Sec ion 2.
17
Using loga i hmic incomes would lead o he exclusion o ze o ea nings and, hus, pe iods o
inac i i y.
18
Because especially women accumula e pe iods o inac i i y o e li e h ough mo h-
e hood and child ea ing, hose pa s o hei employmen biog aphies wi hou any ea nings
play a c ucial ole o he gende li e ime ea nings gap and need o be included in his analysis.
The gende gap Gin he labou ma ke ou come a iable L(he e, hou ly wages, annual
ea nings, UAX ea nings) in pe cen o men mand women a age xis now de ined as:
Gx¼Lm,xL ,x

=Lm,x

100:ð9Þ
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 443
Based on ou new sample ob ained om he mic osimula ion, Figu e 6shows he gende
gaps in hou ly wages, annual ea nings and UAX ea nings o ages 20–60 o bi h coho s 1964–
72. As expec ed, despi e he same end, we see se e al di e ences when we compa e he gende
gaps in hou ly wages and annual ea nings using his mic osimula ion sample o ou esul s
based on he c oss-sec ional sample discussed in Sec ion 2.
A ea ly ages, he gende gap in hou ly wages a he low bu hen inc eases s eadily un il
e i emen . Howe e , we can obse e di e ences in le els, which a e d i en by he mo e con-
ined coho es ic ion in ou mic osimula ion sample and he a ying de ini ion o he gende
gap (loga i hmic s. non-loga i hmic income). Compa ing he gende gaps in annual ea nings
e eals mo e p onounced di e ences be ween he c oss-sec ional and li e ime app oach. Fi s ,
he in e sely U-shaped gende gap in annual ea nings in Figu e 6is signi ican ly la ge han
he gende gap shown in Figu e 2. This di e ence is la gely d i en by he inclusion o inac i e
labou pe iods wi h ze o ea nings in his li e ime analysis, while we excluded hose in ou
c oss-sec ional analysis in Sec ion 2.
19
Including pe iods wi h ze o ea nings leads o a decline in
women's a e age ea nings, and hus o an inc ease in he gende gap. Na u ally, his di e ence
is especially p onounced in he yea s o amily o ma ion when women, on a e age, ha e longe
spells o labou ma ke inac i i y due o child ea ing. Second, in con as o he gende gap we
obse e in ou c oss-sec ional da a, Figu e 6shows a p onounced decline o he gende gap in
annual ea nings be ween ages 40 and 60. Again, hese di e en esul s a e d i en by he di e -
en composi ion o ou wo samples. While he c oss-sec ional sample includes all bi h coho s
1940–79, he li e ime sample is es ic ed o younge coho s. Due o he highe labou ma ke
pa icipa ion a es o women o younge coho s, he gende gap in annual ea nings declines
again be o e e i emen once we es ic ou sample o younge coho s, because mo e women
een e ed he labou ma ke a e imes o inac i i y du ing amily o ma ion.
FIGURE 6 Gende gaps in wages, annual ea nings and UAX ea nings o e he li e cycle. Indi iduals wi h
ze o UAX ea nings a e included in he calcula ion. Fo annual ea nings, employed and unemployed indi iduals
a e conside ed. Fo hou ly wages, only employed indi iduals a e conside ed. Coho s 1964–72. Sou ce: Own
calcula ions based on SOEP 35.
444 GLAUBITZ ET AL.

Finally, he solid line in Figu e 6shows he gende gap in UAX ea nings as he sum o he
annual ea nings up o age X. Ul ima ely, he UA60 ea nings coincide wi h ou de ini ion o li e-
ime ea nings. Hence, he highe he age X, he close UAX ea nings a e o li e ime ea nings. A
he beginning o he wo k li e, women ea n on a e age 20 pe cen less han men do. The di e -
ence in ea nings accumula es o e he li e cou se and inc eases o a gende gap in UA40 ea n-
ings o 52.7 pe cen . A e ha , he gap emains s able, which esul s in a gende gap in
li e ime ea nings o 51.5 pe cen (UA60). A his poin in li e, women ha e ea ned on a e age
a ound €732,000—sligh ly less han hal o he a e age income ha men we e able o accumu-
la e (€1,510,000).
20
The e olu ion o he gende gap in UAX ea nings is by cons uc ion ollowing he shape o
he annual ea nings gende gap cu e. UAX ea nings a e less ola ile because he ma ginal
e ec o adding an addi ional yea o annual ea nings o he UAX ea nings dec eases wi h
inc easing age. Hence, he gende gaps in annual and UAX ea nings bo h expe ience la ge
g ow h un il age 40, bu when he gende gap in annual ea nings declines again, he UAX gen-
de gap emains a i s high le el.
The p o ound di e ence in li e ime ea nings is la gely he esul o di e ences in he ex en-
si e and in ensi e ma gin o labou supply o women o e hei li es. One can discuss how
labou supply is in luenced by own decisions o o ced by pe sonal and social ci cums ances.
P e ious s udies ha e shown a s ong ela ionship be ween gende gaps in income and child en
(e.g., Adda e al., 2017; Angelo e al., 2016; Kle en & Landais, 2017). This can be pa ially
explained by he close connec ion be ween women's labou ma ke decisions and he numbe
o child en hey ha e (Ej næs & Kunze, 2013; Kühhi & Ludwig, 2012). In line wi h hese s ud-
ies, we also ind ha mo he s ace highe ea ning losses wi h e e y addi ional child, while
a he hood does no seem o a ec men's ea nings. Hence, obse ed ea nings di e ences
be ween childless women and men a e smalles and g ow wide wi h e e y addi ional child (see
Figu e A8). This obse a ion also holds ue when we analyse he e olu ion o UAX ea nings by
numbe o child en (Figu e A9).
Figu e 7shows he gende gap in hou ly wages (Panel A), he gende gap in hou s wo ked
(Panel B), he gende gap in annual ea nings (Panel C) and he gende gap in UAX ea nings
(Panel D) o e he li e cycle by numbe o child en. In he beginning, he gende gap in hou ly
wages shows only small gende di e ences o men and women wi h and wi hou child en bu
widens o e he li e cycle. In Sec ion 2, we ha e shown ha his can be la gely explained by he
lesse wo k expe ience women wi h child en gain o e hei li e cou ses. The gende gap in
annual ea nings clea ly di e s by he numbe o child en h oughou he en i e li e cycle (see
Figu e 7, Panel C), exace ba ing he gap in hou ly wages associa ed wi h mo he 's lowe in en-
si e ma gin o wo k (see Figu e 7, Panel B).
The gende gap in li e ime ea nings also inc eases wi h he numbe o child en. While child-
less men and women expe ience a gende gap o 17.3 pe cen , he gap is signi ican ly highe
o men and women wi h h ee o mo e child en (68.0 pe cen a age 60). The signi ican wid-
ening o he gende gap be ween UA20 and UA35 ea nings he eby coincides wi h he inc ease
in he c oss-sec ional gende gaps in annual hou s wo ked, and consequen ly annual ea nings.
These esul s a e in line wi h exis ing s udies inding e idence o mo he hood penal ies and
a he hood p emiums (e.g., Budig & England, 2001; Killewald & Ga ca-Manglano, 2016;
Killewald & Gough, 2013). The e o e, desc ip i e e idence clea ly hin s ha mo he hood migh
be a key d i e o gende ea nings inequali y o e he li e cycle.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 445
3.2.2 | Coun e ac ual analysis
In he las s ep, we wan o es ima e, which pa o he obse ed gende gap in li e ime ea nings
can be associa ed wi h di e ences in he dis ibu ion o obse able cha ac e is ics ac oss gende
and which pa is associa ed wi h di e ences in e u ns o cha ac e is ics. To in es iga e his
issue u he , we will p edic coun e ac ual li e ime ea nings o women in he ollowing wo
s eps.
21
Fi s , we use a sligh ly modi ied e sion o he ea nings eg ession esul s om ou mic o-
simula ion model, es ima ed o male Mand emale Findi iduals sepa a ely
22
:
^
ys, ¼^
β0,sþþ^
β1,sXs, ,sF,M
g
and 1984,2017½:ð10Þ
Second, we hen se all annual ea nings o men and women o missing and e-es ima e
women's coun e ac ual annual ea nings ^
yC
by using he coe icien s ob ained om he male
eg ession model in he women's Mince ea nings eg ession:
^
y , ¼^
β0,mþ^
β1,mX , , 1984,2017½:ð11Þ
Fo he es ima ion o male ea nings, we use coe icien s om he male eg ession model.
Women's coun e ac ual annual ea nings in yea hen ep esen he sala y women would ha e
0 20 40 60 80
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Age
(a) Hou ly wages
0 20 40 60 80
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Age
(b) Annual wo king hou s
0 20 40 60 80
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Age
(c) Annual ea nings
0 20 40 60 80
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
(d) UAX ea nings
Childless One child
Two child en Th ee o mo e child en
O e all a e age
FIGURE 7 Gende gaps o e he li e cycle by child en. Numbe o child en e e s o he o al numbe a age
60. Gende gaps in accumula ed ea nings a e ea nings up o a gi en age. Indi iduals wi h ze o annual and UAX
ea nings a e included in he calcula ion. Sou ce: Own calcula ions based on SOEP 35.
446 GLAUBITZ ET AL.
ea ned i hei cha ac e is ics we e ewa ded he same as men's. Adding up he coun e ac ual
annual ea nings o each woman o e he li e cou se hen yields women's coun e ac ual UAX
ea nings. Fu he mo e, we also calcula e women's UAX ea nings o a baseline scena io whe e
we use he coe icien s om he emale ea nings eg ession model (Equa ion 10). We hen cal-
cula e he baseline UAX ea nings gap o which men's and women's espec i e coe icien s a e
used and he coun e ac ual UAX ea nings gap whe e women's ea nings a e es ima ed using
male coe icien s.
23
The coun e ac ual gende UAX gap is he e o e solely based on di e en
cha ac e is ics o men and women and no by di e en e u ns o cha ac e is ics.
Figu e 8depic s bo h he baseline (solid line) and coun e ac ual (dashed line) gende gap
in UAX ea nings. The di e ence be ween hose wo concep s can be in e p e ed as he
unexplained pa o he gende gap in UAX ea nings (adjus ed gende gap). In he beginning o
he wo k li e, he di e ence be ween bo h gaps shown in Figu e 8is 14.9 pp. The e o e, in ea ly
yea s, app oxima ely one- hi d o he gende gap in UAX ea nings is due o a di e en alloca-
ion o cha ac e is ics and wo- hi ds is due o a di e en ewa d o paymen o cha ac e is ics.
The adjus ed gende gap hen inc eases o abou 20.8 pe cen o UA30 ea nings and declines
a e wa ds o 10.0 pe cen o li e ime ea nings (UA60). Thus, un il he yea s o amily o ma-
ion, he unexplained di e ence be ween women's and men's pay g ows, whe eas i declines
owa ds e i emen . O e all, 80 pe cen o he ( e-es ima ed) baseline gende li e ime ea nings
gap a age 60 can be explained by a di e en dis ibu ion o labou ma ke cha ac e is ics o
men and women. Consequen ly, one- i h o he ( e-es ima ed) baseline gende li e ime ea n-
ings gap a age 60 can be explained by less a ou able ewa ds o women's labou ma ke cha -
ac e is ics, leading o an o e all adjus ed gende li e ime ea nings gap o a ound 10 pe cen .
FIGURE 8 Coun e ac ual es ima ion o he li e ime ea nings gap. Baseline and coun e ac ual gende gap
in UAX ea nings. Gende gaps in accumula ed ea nings a e ea nings up o a gi en age. Indi iduals wi h ze o
UAX ea nings a e included in he calcula ion. Sou ce: Own calcula ions based on SOEP 35.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 447
The e olu ion o he adjus ed gende gap indica es ha ewa ds a e leas a ou able o women
in he i s hal o hei wo k li e.
Nex , we wan o in es iga e how pa en hood in luences he adjus ed gende gap in li e ime
ea nings. Hence, Figu e 9compa es he baseline and coun e ac ual gende gaps by he numbe
o child en. As al eady shown in Figu e 7(Panel D), he baseline gende gap in li e ime ea n-
ings is lowes o childless women and inc eases s ongly wi h he numbe o child en women
ha e. Bu how much o he gende gap in li e ime ea nings o women wi h and wi hou chil-
d en can be explained by a di e en dis ibu ion o cha ac e is ics, and wha is he in luence o
he ole o mo he hood on he adjus ed gende gap in li e ime ea nings?
In s a k con as o he baseline UA60 gende gap, he adjus ed UA60 gende gap only
sligh ly di e s be ween men and women wi h di e en numbe s o child en. The adjus ed gen-
de gap es ima es a e 7.7 pe cen o one child and sligh ly highe o men and women wi h
wo child en (7.3 pe cen ) and h ee o mo e child en and women wi h h ee o mo e child en
(9.9 pe cen ). Hence, he la ge di e ences in he obse ed gende gaps o women wi h
child en a e mainly d i en by he di e en accumula ion o cha ac e is ics a he han an addi-
ional unexplained penal y o mo he hood. Ou esul s in Sec ion 2and Figu e 7(Panel B) indi-
ca ed ha hese di e ences migh be mainly due o ewe wo king hou s and less wo k
expe ience, which women wi h child en accumula e o e hei wo k li e. Howe e , ou esul s
sugges he opposi e o childless indi iduals, o which he en i e ( e-es ima ed) gende gap in
li e ime ea nings (a ound 20 pe cen ) appea s o be d i en by di e ences in ewa ds.
0 15 30 45 60
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
(a) Childless
0 15 30 45 60
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
(b) One child
0 15 30 45 60
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
(c) Two child en
0 15 30 45 60 75
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
(d) Th ee o mo e child en
Baseline
FIGURE 9 Coun e ac ual es ima ion o he li e ime ea nings gap by numbe o child en. Es ima ed and
coun e ac ual gende gaps in UAX ea nings. Gende gaps in accumula ed ea nings a e ea nings up o a gi en
age. Indi iduals wi h ze o UAX ea nings a e included in he calcula ion. Sou ce: Own calcula ions based on
SOEP 35.
448 GLAUBITZ ET AL.
APPENDIX A
A.1 |Oaxaca Blinde decomposi ion
The Oaxaca Blinde decomposi ion was simul aneously in oduced by Oaxaca (1973) and
Blinde (1973) and di ides he gende di e en ial in labo ma ke ou comes (he e: hou ly wage
o annual ea nings) in o an endowmen pa and a coe icien pa . The endowmen pa o he
gende di e en ial accoun s o he pa o he gap which can be a ibu ed o di e ences in he
alloca ion o cha ac e is ics (e.g., wo king hou s, highes le el o educa ion) be ween men and
women. In con as , he coe icien pa cap u es he gende di e ences in labo ma ke e u ns
o cha ac e is ics, and he e o e in hei coe icien s. In o he wo ds, i s a es he gende di e -
ences o wha he labo ma ke is willing o pay o he same cha ac e is ics. This pa is also
called he aw o adjus ed gende wage/ea nings di e en ial. This adjus ed gap, howe e , also
con ains he e ec s o gende di e ences in unobse ed p edic o s (Jann, 2008). The Oaxaca-
Blinde decomposi ion app oach enables us o analyze whe he he gende gap in wages/ea n-
ings is mainly d i en by he di e en dis ibu ions o p oduc i i y cha ac e is ics o by di e en
ewa ds o hese cha ac e is ics by gende .
The gende gap Gxis de ined as he di e ence be ween he means o he labo ma ke ou -
comes La age xo men m and women :
Gx¼EL
mx
ðÞEL
x
 ð12Þ
Ls o ei he sex sðÞis based on he linea model
Lsx ¼Z0
sxβsx þϵsx,Eϵsx
ðÞ¼0, S ,m g,ð13Þ
whe e he ec o Zincludes all ele an cha ac e is ics, βis he es ima ion ec o and ϵis he
e o e m. Inse ing Equa ion (2.13) in o Equa ion (2.12), he ea nings di e en ial can also be
w i en as:
Gx¼EZ
mx
ðÞ
0βmx EZ
x

0β x:ð14Þ
Fo he decomposi ion o he esul s, a non-disc imina o y coe icien ec o is needed,
called β. Following Neuma k (1988), he ec o is de e mined as a pooled eg ession o e bo h
sexes. The gende gap can hen be ew i en as:
Gx¼EZ
mx
ðÞEZ
x

0β
x
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Endowmen pa
þEZ
mx
ðÞ
0βmx β
x

þEZ
x

0β
xβ x
hi
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Coe icien pa
ð15Þ
whe e he i s pa o Equa ion (2.15) is he endowmen pa and he second pa is he coe i-
cien componen o he gende gap in he labo ma ke ou come.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 455

A.2 |Robus ness: Mic osimula ion
Pseudo missings To es he obus ness o ou simula ion model u he , we use he concep o
pseudo missings. To ha end, we se uly obse ed in o ma ion o some pa o he sample
missing (pseudo missings) and p edic hei now missing obse a ions again by using ou
dynamic mic osimula ion and he eg ession coe icien s p e iously ob ained. As we need a
s a ing poin o a leas wo obse a ions o ou models due o he lagged e ms, we use he
i s wo uly obse ed yea s o e e yone be o e s a ing o c ea e pseudo missings. Figu e A4
shows he di e ences be ween he simula ed pseudo missings (dashed line) and he uly
obse ed in o ma ion (solid line) o labo o ce s a us, employmen s a us, annual wo king
hou s and annual ea nings. In mos g aphs, he le el o accu acy o he model is so high ha i
is ha d o e en ell he solid and dashed line apa . Fo labo ma ke s a us, he model p edic s
99.9% o all pseudo missings co ec ly. And e en o employmen s a us, whe e he e appea o
be bigge di e ences be ween pseudo missing and obse a ions a a i s glance, o e all 97.7% o
all cases a e simula ed co ec ly. These esul s u he suppo he obus ness o ou simula ion
model.
Mon e Ca lo simula ion Ano he way o alida e he obus ness o ou dynamic mic o-
simula ion model is o make use o he unde lying andom p ocess desc ibed in Subsec ion
3.1.1. We implemen a Mon e Ca lo simula ion app oach by simula ing each indi idual's
employmen biog aphies 100 imes. By doing so, due o he unde lying andom p ocess de e -
mining ansi ions in labo ma ke ou come a iables be ween 1 and , we simula e up o
100 di e en employmen biog aphies o each indi idual. Howe e , due o limi ed compu a-
ional capaci ies we only simula e he employmen a iables (labo ma ke s a us, employmen
s a us, ull- ime/pa - ime wo k, annual wo king hou s and annual ea nings) and keep he am-
ily in o ma ion (numbe o child en and ma i al s a us) cons an o each o he 100 i e a ions.
In he nex s ep, we calcula e li e ime ea nings o each o he 100 simula ed ca ee pa hs pe
indi idual and compu e he a e age li e ime ea nings and he esul ing UAX ea nings gende
gap in he popula ion o each o he 100 uns. By de i ing he 95% con idence in e als we can
analyze whe he a e age li e ime ea nings a y signi ican ly o di e en unde lying andom
p ocesses o whe he hey a e obus . The esul s a e p esen ed in Figu es A5 and A6. Figu e
A5 shows ha li e ime ea nings by coho s a e e y obus . Howe e , li e ime ea nings o
women a y mo e s ongly han men's. Figu e A6 p o ides e idence o a e y na ow 95% con-
idence in e al o he gende gap in UAX ea nings. Consequen ly, he esul s o he Mon e
Ca lo simula ion con i m he high obus ness o ou simula ion ou comes.
456 GLAUBITZ ET AL.
TABLE A1 Desc ip i e s a is ics—means by age
Age
Men 20 25 30 35 40 45 50 55 60
Annual ea nings 15748.13 27727.89 37925.13 45217.80 51615.70 54204.14 54747.55 53969.63 51535.02
(10972.17) (13306.99) (18571.57) (24095.07) (31182.61) (38951.27) (35380.01) (33505.90) (50496.35)
Hou ly wage 9.37 15.13 18.12 20.72 23.06 23.95 24.24 25.83 26.13
(7.72) (18.11) (20.76) (16.32) (16.25) (17.91) (14.47) (31.96) (28.33)
Hou s wo ked pe week 34.55 38.29 42.81 43.49 44.39 44.10 43.50 42.65 39.34
(13.42) (14.48) (12.47) (11.38) (11.01) (10.61) (11.17) (12.13) (14.38)
Yea s in ull- ime wo k 1.20 4.75 8.54 12.97 17.71 22.58 27.43 32.69 37.32
(1.28) (2.60) (3.77) (4.37) (4.85) (5.31) (5.71) (5.81) (5.74)
Yea s in pa - ime wo k 0.14 0.33 0.55 0.56 0.61 0.65 0.75 0.68 1.09
(0.47) (0.98) (1.56) (1.64) (1.93) (2.10) (2.44) (2.46) (2.90)
Yea s in unemploymen 0.13 0.31 0.39 0.43 0.45 0.47 0.51 0.49 0.45
(0.38) (0.70) (0.98) (1.21) (1.37) (1.64) (1.85) (1.78) (1.65)
Yea s o educa ion 8.97 10.61 11.84 12.44 12.62 12.65 12.67 12.57 12.73
(3.86) (3.03) (3.25) (3.17) (3.03) (2.96) (2.92) (2.84) (2.92)
Age
Women 20 25 30 35 40 45 50 55 60
Annual ea nings 12773.34 21115.69 22975.43 21925.18 22944.75 24975.61 26705.30 26475.69 24659.61
(8683.31) (12332.56) (16720.75) (19512.82) (18626.65) (20497.00) (21713.56) (25559.13) (21236.77)
Hou ly wage 7.97 12.87 15.19 15.63 16.23 16.23 16.82 16.54 17.48
(6.58) (9.59) (12.19) (13.18) (12.02) (10.88) (12.49) (13.18) (14.99)
Hou s wo ked pe week 31.75 32.28 29.91 26.68 27.38 28.87 30.09 29.42 26.97
(13.02) (14.34) (15.60) (15.26) (14.29) (13.99) (13.98) (13.64) (14.49)
(Con inues)
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 457
TABLE A1 (Con inued)
Age
Women 20 25 30 35 40 45 50 55 60
Yea s in ull- ime wo k 1.20 4.36 6.73 8.04 9.63 11.61 14.00 16.70 19.65
(1.23) (2.71) (4.12) (5.23) (6.46) (7.99) (9.69) (11.75) (13.99)
Yea s in pa - ime wo k 0.21 0.82 1.91 3.81 5.69 7.60 9.45 11.66 13.32
(0.55) (1.61) (2.60) (3.76) (4.87) (6.16) (7.75) (9.77) (11.79)
Yea s in unemploymen 0.17 0.28 0.40 0.50 0.56 0.58 0.64 0.71 0.57
(0.40) (0.73) (0.95) (1.19) (1.52) (1.60) (1.77) (1.99) (1.86)
Yea s o educa ion 9.17 11.17 12.07 12.42 12.48 12.39 12.34 12.11 12.02
(3.91) (2.98) (3.31) (3.00) (2.89) (2.91) (2.78) (2.59) (2.71)
No es: Only employed indi iduals wi h hou ly wages and annual ea nings g ea e han ze o we e included. Coho s 1940–1979, weigh ed sample. Annual ea nings and hou ly wages a e p ice-
adjus ed and p esen ed in 2015 Eu o. S anda d e o s in pa en heses.
Sou ce: Own calcula ions based on SOEP 35.
458 GLAUBITZ ET AL.
TABLE A2 Reg ession esul s o hou ly wages - women
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
One child 0.019 0.061 0.159*** 0.023 0.003 0.075** 0.068*0.104** 0.058
(0.158) (0.055) (0.046) (0.036) (0.038) (0.034) (0.040) (0.048) (0.070)
Two child en 0.461 0.234** 0.154*** 0.058 0.000 0.073** 0.138*** 0.088*0.014
(0.501) (0.107) (0.058) (0.041) (0.039) (0.036) (0.040) (0.048) (0.071)
3 o mo e child en 0.171 0.111 0.167*** 0.030 0.036 0.129*** 0.103*0.051
(0.195) (0.093) (0.055) (0.050) (0.043) (0.048) (0.058) (0.083)
Ma ied 0.033 0.038 0.054 0.008 0.004 0.068** 0.052*0.031 0.055
(0.100) (0.036) (0.034) (0.030) (0.029) (0.027) (0.029) (0.035) (0.048)
Yea s FT 0.445*** 0.055** 0.059*** 0.028*** 0.035*** 0.026*** 0.027*** 0.014** 0.030***
(0.061) (0.022) (0.014) (0.009) (0.007) (0.005) (0.005) (0.006) (0.007)
Yea s FT (sq) 0.054*** 0.003 0.002*** 0.000 0.001*0.000 0.000 0.000 0.000**
(0.016) (0.002) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Yea s PT 0.023 0.039 0.028 0.019*0.013 0.020*** 0.002 0.001 0.011
(0.103) (0.029) (0.017) (0.011) (0.008) (0.006) (0.006) (0.006) (0.008)
Yea s PT (sq) 0.012 0.001 0.003*0.002** 0.001** 0.001*** 0.000*0.000 0.000**
(0.028) (0.004) (0.002) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000)
Yea s UE 0.689*** 0.101** 0.174*** 0.013 0.062*** 0.096*** 0.076*** 0.058*** 0.044*
(0.209) (0.043) (0.039) (0.021) (0.018) (0.018) (0.018) (0.017) (0.024)
Yea s UE (sq) 0.236*0.000 0.034*** 0.005** 0.004** 0.007*** 0.004** 0.001 0.001
(0.140) (0.006) (0.009) (0.003) (0.002) (0.002) (0.002) (0.001) (0.002)
Pa - ime 0.244*** 0.272*** 0.110*** 0.182*** 0.161*** 0.102*** 0.023 0.068*0.069
(0.081) (0.049) (0.042) (0.032) (0.031) (0.027) (0.031) (0.039) (0.054)
(Con inues)
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 459
TABLE A2 (Con inued)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
Educa ion 0.068*** 0.007 0.033** 0.063*** 0.018 0.008 0.023 0.012 0.052
(0.024) (0.017) (0.015) (0.015) (0.023) (0.015) (0.021) (0.029) (0.036)
Educa ion (sq) 0.004*0.001 0.003*** 0.006*** 0.002*** 0.003*** 0.004*** 0.002** 0.005***
(0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Cons an 2.241*** 1.714*** 1.233*** 1.700*** 1.825*** 1.508*** 1.033*** 1.614*** 1.898***
(0.770) (0.266) (0.211) (0.222) (0.199) (0.161) (0.313) (0.373) (0.463)
Obs. 382 882 1307 1859 2493 2653 2043 1320 778
R-squa ed 0.323 0.127 0.187 0.240 0.192 0.219 0.205 0.213 0.248
Coho -FE YES YES YES YES YES YES YES YES YES
Sec o -FE YES YES YES YES YES YES YES YES YES
No es: S anda d e o s in pa en heses; The s a s e e o he ollowing signi icance le el:
*p< 0.1, **p< 0.05, ***p< 0.01.
Sou ce: Own calcula ions based on SOEP 35.
460 GLAUBITZ ET AL.

TABLE A3 Reg ession esul s o hou ly wages—men
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
One child 0.160 0.089 0.017 0.017 0.007 0.029 0.043 0.026 0.053
(0.349) (0.055) (0.031) (0.024) (0.022) (0.024) (0.029) (0.040) (0.053)
Two child en 0.952 0.134*0.065*0.069*** 0.033 0.103*** 0.000 0.034 0.010
(0.751) (0.079) (0.035) (0.024) (0.022) (0.023) (0.027) (0.037) (0.049)
3 o mo e child en 0.006 0.139*0.024 0.015 0.050*0.049*0.013 0.075 0.203***
(0.173) (0.075) (0.045) (0.033) (0.028) (0.029) (0.035) (0.049) (0.067)
Ma ied 0.026 0.015 0.164*** 0.085*** 0.101*** 0.069*** 0.084*** 0.020 0.105**
(0.166) (0.041) (0.027) (0.023) (0.021) (0.023) (0.026) (0.035) (0.048)
Yea s FT 0.737*** 0.164*** 0.105*** 0.058*** 0.053*** 0.045*** 0.032*** 0.040*** 0.030
(0.061) (0.023) (0.013) (0.008) (0.007) (0.008) (0.009) (0.012) (0.036)
Yea s FT (sq) 0.108*** 0.012*** 0.006*** 0.002*** 0.001*** 0.001*** 0.001*** 0.001** 0.001
(0.015) (0.002) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.001)
Yea s PT 0.225 0.197*** 0.074*** 0.020*0.056*** 0.038*** 0.069*** 0.070*** 0.057***
(0.181) (0.036) (0.018) (0.012) (0.010) (0.011) (0.010) (0.014) (0.017)
Yea s PT (sq) 0.079 0.021*** 0.006*** 0.001 0.003*** 0.003*** 0.003*** 0.003*** 0.001*
(0.071) (0.005) (0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Yea s UE 0.279 0.108** 0.178*** 0.113*** 0.117*** 0.105*** 0.094*** 0.069*** 0.094***
(0.195) (0.050) (0.029) (0.015) (0.013) (0.012) (0.014) (0.019) (0.036)
Yea s UE (sq) 0.099 0.005 0.023*** 0.004*** 0.006*** 0.006*** 0.004*** 0.002** 0.006
(0.101) (0.014) (0.006) (0.001) (0.001) (0.001) (0.001) (0.001) (0.004)
Pa  ime 0.271*** 0.422*** 0.257*** 0.173*** 0.336*** 0.251*** 0.189*** 0.389*** 0.201***
(0.075) (0.054) (0.043) (0.034) (0.031) (0.031) (0.039) (0.045) (0.054)
(Con inues)
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 461
TABLE A3 (Con inued)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
Educa ion 0.031 0.038*** 0.071*** 0.039*** 0.051*** 0.040*0.060*** 0.002 0.020
(0.030) (0.013) (0.010) (0.010) (0.013) (0.022) (0.023) (0.055) (0.067)
Educa ion (sq) 0.003 0.003*** 0.005*** 0.003*** 0.004*** 0.001*0.000 0.002 0.002
(0.003) (0.001) (0.000) (0.000) (0.000) (0.001) (0.001) (0.002) (0.002)
Cons an 1.996*** 1.591*** 2.263*** 2.331*** 2.519*** 1.703*** 1.673*** 1.875*** 1.318
(0.504) (0.163) (0.183) (0.172) (0.173) (0.187) (0.231) (0.437) (0.926)
Obs. 383 900 1746 2464 2863 2939 2261 1546 980
R-squa ed 0.449 0.231 0.185 0.229 0.277 0.283 0.252 0.184 0.208
Coho -FE YES YES YES YES YES YES YES YES YES
Sec o -FE YES YES YES YES YES YES YES YES YES
No es: S anda d e o s in pa en heses; The s a s e e o he ollowing signi icance le el:
*p< 0.1, **p< 0.05, ***p< 0.01.
Sou ce: Own calcula ions based on SOEP 35.
462 GLAUBITZ ET AL.
TABLE A4 Reg ession esul s o annual ea nings—women
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
One child 0.113 0.033 0.130*** 0.008 0.005 0.101*** 0.059 0.086 0.044
(0.150) (0.051) (0.045) (0.036) (0.037) (0.034) (0.039) (0.047) (0.071)
Two child en 0.485 0.234** 0.205***** 0.044 0.007 0.092*** 0.131*** 0.066 0.024
(0.486) (0.100) (0.059) (0.041) (0.039) (0.035) (0.039) (0.048) (0.071)
3 o mo e child en 0.056 0.174*0.160*** 0.011 0.081*0.132*** 0.079 0.034
(0.183) (0.093) (0.056) (0.049) (0.043) (0.047) (0.058) (0.083)
Ma ied 0.040 0.062*0.038 0.044 0.004 0.102*** 0.036*** 0.013 0.048**
(0.096) (0.034) (0.034) (0.031) (0.029) (0.026) (0.028) (0.035) (0.048)
Yea s FT 0.466*** 0.098*** 0.065*** 0.031*** 0.036*** 0.025*** 0.023*** 0.009*0.031***
(0.060) (0.021) (0.014) (0.009) (0.007) (0.005) (0.005) (0.006) (0.007)
Yea s FT (sq) 0.059*** 0.006*** 0.003*** 0.000 0.001** 0.000 0.000 0.000 0.000**
(0.015) (0.002) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Yea s PT 0.099 0.033 0.005 0.001 0.002 0.010*0.002 0.001 0.011
(0.103) (0.027) (0.016) (0.011) (0.008) (0.006) (0.006) (0.006) (0.008)
Yea s PT (sq) 0.034 0.002 0.001 0.001 0.001 0.001*** 0.000 0.000 0.000**
(0.028) (0.004) (0.002) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000)
Yea s UE 0.524*** 0.076*0.174*** 0.010 0.075*** 0.097*** 0.079*** 0.050*** 0.047*
(0.197) (0.040) (0.039) (0.021) (0.018) (0.018) (0.018) (0.017) (0.024)
Yea s UE (sq) 0.141 0.002 0.033*** 0.006** 0.005*** 0.007*** 0.004** 0.001 0.001
(0.134) (0.006) (0.009) (0.003) (0.002) (0.002) (0.002) (0.001) (0.002)
Weekly hou s 0.066*** 0.054*** 0.081*** 0.093*** 0.091*** 0.087*** 0.105*** 0.113*** 0.093***
(0.009) (0.004) (0.004) (0.003) (0.003) (0.002) (0.003) (0.004) (0.006)
(Con inues)
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 463
TABLE A4 (Con inued)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Age 20 25 30 35 40 45 50 55 60
Weekly hou s (sq) 0.001*** 0.000*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Educa ion 0.046** 0.002 0.034** 0.071*** 0.013 0.007 0.024 0.012 0.051
(0.023) (0.016) (0.015) (0.015) (0.023) (0.015) (0.020) (0.028) (0.036)
Educa ion (sq) 0.002 0.001 0.003*** 0.006*** 0.003*** 0.003*** 0.004*** 0.002** 0.005***
(0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Cons an 7.583*** 8.273*** 8.012*** 7.637*** 7.141*** 7.141*** 7.140*** 6.986*** 7.550***
(0.197) (0.137) (0.130) (0.126) (0.169) (0.119) (0.153) (0.213) (0.277)
Obs. 382 882 1307 1859 2493 2653 2043 1320 778
R-squa ed 0.573 0.540 0.627 0.663 0.578 0.599 0.674 0.681 0.660
Coho -FE YES YES YES YES YES YES YES YES YES
Sec o -FE YES YES YES YES YES YES YES YES YES
No es: S anda d e o s in pa en heses; The s a s e e o he ollowing signi icance le el:
*p< 0.1, **p< 0.05, ***p< 0.01.
Sou ce: Own calcula ions based on SOEP 35.
464 GLAUBITZ ET AL.
0 2 4 6 8 10 12 14 16
Pe cen
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Yea s o pa icipa ion
FIGURE A3 Dis ibu ion o pa icipa ion yea s in he SOEP. No es: Re e s o pa icipa ion yea s o he SOEP
sample used in 2 o his pape . Res ic ions o he mic osimula ion in Sec ion 3 a e no applied he e. Sou ce:
Own calcula ions based on SOEP 35.
.6 .7 .8 .9 1
LMP a e (in %)
20 25 30 35 40 45 50
Age
Women
.6 .7 .8 .9 1
LMP a e (in %)
20 25 30 35 40 45 50
Age
Men
Labo ma ke pa icipa ion (LMP)
.5 .6 .7 .8 .9 1
Employmen a e (in %)
20 25 30 35 40 45 50
Age
Women
.5 .6 .7 .8 .9 1
Employmen a e (in %)
20 25 30 35 40 45 50
Age
Men
Employmen
0 600 1200 1800 2400
Annual wo king hou s
20 25 30 35 40 45 50
Age
Women
0 600 1200 1800 2400
Annual wo king hou s
20 25 30 35 40 45 50
Age
Men
Annual wo king hou s
0 20000 40000 60000
Annual ea nings (in Eu o)
20 25 30 35 40 45 50
Age
Women
0 20000 40000 60000
Annual ea nings (in Eu o)
20 25 30 35 40 45 50
Age
Men
Annual ea nings
Obse ed Pseudo
FIGURE A4 Pseudo missings o labo ma ke ou comes. The g aphs compa ing uly obse ed and
simula ed pseudo in o ma ion o annual wo king hou s and annual ea nings only ocus on employed
indi iduals. Sou ce: Own calcula ions based on SOEP 35.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 471

700000 720000 740000 760000
Li e ime ea nings (in Eu o)
1964 1965 1966 1967 1968 1969 1970 1971 1972
Coho
Women
1500000 1600000 1700000
Li e ime ea nings (in Eu o)
1964 1965 1966 1967 1968 1969 1970 1971 1972
Coho
Men
95% CI A e age li e ime ea nings
FIGURE A5 Mon e Ca lo simula ion o ea nings Sou ce: Own calcula ions based on SOEP 35.
20 30 40 50 60
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Up o age
95% CI Gende gap in UAX ea nings
FIGURE A6 Mon e Ca lo simula ion o he gende gap in UAX ea nings Sou ce: Own calcula ions based
on SOEP 35.
472 GLAUBITZ ET AL.
0 10 20 30 40 50 60
Gende gap (in %)
20 25 30 35 40 45 50 55 60
Age
Hou ly wage gap Ea nings gap - all
Ea nings gap - employed only UAX ea nings gap
FIGURE A7 Gende gaps in ea nings by di e en concep s. Indi iduals wi h ze o UAX ea nings a e
included in he calcula ion. Fo annual ea nings gap, all employed and unemployed indi iduals a e conside ed.
Coho s 1964–1972. Sou ce: Own calcula ions based on SOEP 35.
0 20000 40000 60000
Annual ea nings (in Eu o)
20 25 30 35 40 45 50 55 60
Age
Women
0 20000 40000 60000
Annual ea nings (in Eu o)
20 25 30 35 40 45 50 55 60
Age
Men
Childless One child
Two child en Th ee o mo e child en
O e all a e age
FIGURE A8 Annual ea nings by gende and numbe o child en. Employed and unemployed indi iduals
a e conside ed. Numbe o child en e e s o he o al numbe a age 50. Coho s 1964–1972. Sou ce: Own
calcula ions based on SOEP 35.
EARNINGS GENDER GAP: MICROSIMULATION APPROACH 473
0 500000 1000000 1500000
UAX ea nings (in Eu o)
20 25 30 35 40 45 50 55 60
Up o Age
Women
0 500000 1000000 1500000
UAX ea nings (in Eu o)
20 25 30 35 40 45 50 55 60
Up o Age
Men
Childless One child
Two child en Th ee o mo e child en
O e all a e age
FIGURE A9 UAX ea nings by gende and numbe o child en. Employed and unemployed indi iduals a e
conside ed. Numbe o child en e e s o he o al numbe a age 50. Coho s 1964–1972. Sou ce: Own
calcula ions based on SOEP 35.
474 GLAUBITZ ET AL.