Busch, Fabian; Ochsen, Ca s en
A icle — Published Ve sion
Local E ec s o Educa ion and Age G oups on
Unemploymen in Ge many
G ow h and Change
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Busch, Fabian; Ochsen, Ca s en (2025) : Local E ec s o Educa ion and Age
G oups on Unemploymen in Ge many, G ow h and Change, ISSN 1468-2257, Wiley Pe iodicals,
Inc., Hoboken, NJ, Vol. 56, Iss. 1,
h ps://doi.o g/10.1111/g ow.70011
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/313752
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
G ow h and Change
-
ORIGINAL ARTICLE
OPEN ACCESS
Local E ec s o Educa ion and Age G oups on
Unemploymen in Ge many
Fabian Busch
1
| Ca s en Ochsen
2
1
Depa men o Economics, Uni e si y o Ros ock, Ros ock, Ge many |
2
Depa men o Labou Economics, Uni e si y o Applied Labou S udies, Schwe in,
Ge many
Co espondence: Ca s en Ochsen ([email p o ec ed])
Recei ed: 9 No embe 2023 | Re ised: 10 Sep embe 2024 | Accep ed: 11 No embe 2024
Keywo ds: ageing | human capi al | labou mobili y | egional unemploymen | spa ial in e ac ions
ABSTRACT
This a icle p o ides a comp ehensi e analysis o how egional changes in he age and educa ion dis ibu ion o he labou o ce
a ec local and neighbou hood unemploymen a es. Based on heo e ical conside a ions, we a gue ha di e ences in job
sea ch, sepa a ion, and commu ing a e key ac o s in g oup di e ences, and he e o e, changes in ela i e g oup size a ec he
le el o unemploymen . The empi ical analysis ocuses on local labou ma ke s in Ge many, using a dynamic spa ial panel da a
model. Acco ding o he es ima es, an inc easing p opo ion o young and/o low‐educa ed wo ke s aises local unemploymen ,
while la ge p opo ions o olde p ime‐age and/o highly educa ed wo ke s aise unemploymen in neighbou ing labou
ma ke s. As a esul , he ecen ageing and educa ion de elopmen s in he Ge man labou o ce ha e led o a 25% educ ion in
he unemploymen a e.
JEL Classi ica ion: R23, R12, J11, J24, J61
1
|
In oduc ion
The unemploymen a e is o en used as an indica o o he
o e all economic si ua ion. Howe e , i can also be seen as an
a e age unemploymen isk ha akes in o accoun di e en
g oups o people wi h a ying le els o educa ion and age. Fo
ins ance, hose wi h lowe le els o educa ion a e mo e likely o
be unemployed han hose wi h highe le els o educa ion.
1
A
change in hei g oup sha es changes he o e all unemploymen
a e, e en i he unemploymen a e o bo h g oups does no
change. Simila ly, younge wo ke s a e mo e likely o be un-
employed han olde wo ke s.
2
O e he pas 2 decades, he
OECD a e age has shown ha unemploymen a es dec ease
wi h inc easing age and educa ion le els, ega dless o he size
o he g oups and economic cycles. This means ha changes in
he dis ibu ion o age and educa ion g oups a e c ucial ac o s
ha a ec he unemploymen a e in he long e m.
The aim o his a icle is o analyse how egional changes in he
age and educa ion dis ibu ion o he labou o ce a ec local and
neighbou hood unemploymen a es.
3
Based on heo e ical
conside a ions, we a gue ha di e ences in job sea ch, sepa a-
ion, and mobili y in e ms o commu ing a e key ac o s in g oup
di e ences, and he e o e, changes in ela i e g oup size a ec
he le el o o e all unemploymen . Bo h age and educa ion ha e
a composi ional impac on unemploymen in he local and
neighbou ing egions. Ou hypo hesis is ha he ecen changes
in he dis ibu ion o educa ion and age in he Ge man labou
o ce accoun o subs an ially educing he unemploymen a e.
Ou esea ch conside ably con ibu es o he li e a u e by
empi ically assessing composi ional e ec s on unemploymen a
he egional le el. We conside a spa ial econome ic app oach
and da a on he local dis ibu ion o age and educa ion. An
ageing p ocess and an inc eased educa ion le el o he labou
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly
ci ed.
© 2025 The Au ho (s). G ow h and Change published by Wiley Pe iodicals LLC.
G ow h and Change, 2025; 56:e70011 1 o 20
h ps://doi.o g/10.1111/g ow.70011
o ce cha ac e ise he pe iod unde s udy. Ou dynamic space‐
ime panel da a model (spa ial Du bin model) e eals ha a
ising sha e o you h and/o low‐educa ed wo ke s inc eases
local unemploymen in Ge many. Howe e , hese g oups do no
a ec unemploymen in he neighbou ing labou ma ke .
Con e sely, a la ge sha e o olde p ime‐age and/o highly
educa ed wo ke s in he su ounding a ea aises he unem-
ploymen a e in he local egion. These g oups ha e s a is ically
weak nega i e e ec s on unemploymen in hei home a ea. In
conclusion, he ageing o he labou o ce and he ising edu-
ca ion le el educe o e all unemploymen . Fu he mo e, he
cu en changes in age and educa ion g oups a e almos equally
impo an and accoun o an unemploymen a e educ ion o
abou 25%. These indings ha e p ac ical implica ions o poli-
cymake s and labou ma ke analys s in unde s anding and
add essing egional unemploymen .
The a icle is o ganised as ollows: Sec ion 2o e s a li e a u e
o e iew, and Sec ion 3p o ides some heo e ical conside -
a ions. Sec ion 4desc ibes he da a and he econome ic
app oach and epo s and discusses he es ima ed esul s. Sec-
ion 5concludes.
2
|
Li e a u e Re iew
Du ing he ea ly 1970s, he young popula ion inc eased in many
de eloped coun ies. Today, we a e wi nessing a subs an ial ise
in he p opo ion o olde wo ke s, which has implica ions o
he labou ma ke when younge and olde wo ke s a e
impe ec subs i u es (Eas e lin, Wach e , and Wach e 1978).
4
While Eas e lins' coho c owding hypo hesis p ima ily ocuses
on ma iage, e ili y, wages, and labou ma ke pa icipa ion,
Pe y (1970) ocussed on he ela ionship be ween popula ion
g oup size and employmen . Based on mac oeconomic da a, he
subsequen schola ly discussion concludes ha a la ge sha e o
he wo king‐age popula ion you h aises he o e all unem-
ploymen a e. This is because he young o en ha e he highes
unemploymen le el among age g oups.
5
In con as , Shime (2001) concludes ha he labou supply o
many young people can educe he o e all unemploymen a e
using US s a e‐le el panel da a. He a gues ha a highe sha e o
young indi iduals in he wo king‐age popula ion can lead o an
inc ease in job c ea ion due o hei highe sea ch in ensi y. A
simila esul is ound in No ds öm Skans (2005), who uses
Swedish da a and concludes ha younge wo ke s can bene i
om belonging o a la ge coho in e ms o educed
unemploymen .
Ga lo , Pohl, and Schanne (2013) conduc ed a s udy o analyse
he impac o smalle labou ma ke en y coho s on unem-
ploymen and he di ec e ec o he age s uc u e on unem-
ploymen in Wes Ge many. They ound ha i he age
dis ibu ion o he labou o ce emained unchanged, he un-
employmen a e would ha e been highe . As he size o he
younge gene a ion en e ing he labou ma ke in Ge many has
been dec easing, he demog aphic change could enhance job
oppo uni ies and educe he unemploymen a e. Ochsen (2021)
analyses he local e ec o he age dis ibu ion o he wo king‐age
popula ion on unemploymen . Using US coun y‐le el da a, he
applies a dynamic space‐ ime panel da a model and conside s
di e en age g oups in he local and neighbou ing egions. The
esul s p o ide s ong e idence ha (spa ial) age g oup changes
a e an impo an long‐ e m d i e o o e all unemploymen
change. Ageing o he wo king‐age popula ion educes o e all
unemploymen , and he p esen changing age s uc u e leads o a
long‐ e m educ ion o he US unemploymen a e.
A majo bene i o educa ion is ha he unemploymen isk de-
c eases wi h an inc easing educa ion le el (Mince 1991). Fo
example, Acemoglu (1998,2002a,2002b,2003) discusses he ole
o echnological change and wages o shi s o mo e skilled and
less unskilled wo ke s. Consequen ly, he dis ibu ion in an
economy shi s con inuously owa ds mo e educa ed wo ke s,
simila o he ageing p ocess. When employe s p e e college
g adua es o jobs ha equi e a high school deg ee, he ela i e
demand o college g adua es ises. Ea ly li e a u e ha inds
empi ical e idence o his a e Teulings and Koopman-
schap (1989), Howe (1993), and an, Ou s, and Ridde (1995).
The e idence ha la ge g oups o be e ‐educa ed indi iduals
c owd ou less‐educa ed indi iduals also has consequences o
he low‐educa ed unemployed, as, o example, Wolbe s (2000),
A be g (2003), Ges huizen and Wolbe s (2010), and Ab as-
sa (2015) poin ou . The e o e, c ea ing low‐skilled jobs may no
necessa ily imp o e he employabili y o low‐skilled wo ke s.
The li e a u e discussed analyses ei he he e ec s o ageing o
educa ion on unemploymen . An excep ion is a s udy by Biagi
and Luci o a (2008), who examine he e ec s o demog aphic
and educa ional changes on he e olu ion o unemploymen
a es o a panel o Eu opean coun ies. Thei esea ch indings
sugges ha demog aphic and educa ion changes a ec young
and adul wo ke s and mo e o less educa ed indi iduals
di e en ly. The s udy e eals ha changes in he popula ion age
s uc u e (baby bus ) posi i ely ela e o you h unemploymen
a e, whe eas changes in he educa ional s uc u e (educa ion
boom) educe unemploymen among he mo e educa ed.
Finally, na ional‐le el da a a e no app op ia e o co e ing
wi hin‐coun y mobili y. In small local egions, spa ial mobili y
(in e ms o commu ing) is ela ed o local labou ma ke
igh ness. In addi ion, spa ial mobili y is di e en o age g oups
(younge wo ke s is mo e mobile han olde wo ke s) and ed-
uca ion g oups ( he highly educa ed a e mo e mobile han low‐
educa ed wo ke s).
6
The e ec s o a changing age s uc u e in
he local labou o ce on unemploymen in a spa ial in e ac ion
model a e conside ed only in he s udy by Ochsen (2021).
Conce ning educa ional dis ibu ion and spa ial mobili y, Kulu,
Lundholm, and Malmbe g (2018) epo ha changes in popu-
la ion composi ion, mainly inc eased en olmen in highe edu-
ca ion, accoun o much o he ising spa ial mobili y using da a
o Sweden. Using F ench da a, Lemis e and Mo eau (2009)
ind ha e u ns o spa ial mobili y o men inc ease wi h
educa ion.
3
|
Theo e ical Conside a ions
Shi s in he labou o ce o la ge sha es o educa ed and olde
wo ke s a ec he unemploymen a e when g oup‐speci ic
2 o 20 G ow h and Change, 2025
unemploymen a es di e . To analyse his, we di ide he labou
o ce in o wo g oups, G oup 1 and G oup 2. Age G oup 1
ep esen s he younge wo ke s (y), and age G oup 2 ep esen s
he olde wo ke s (o). Fo educa ion, we dis inguish be ween
low‐educa ed (μ) as G oup 1 and high‐educa ed (h) as G oup 2.
Fo simplici y, he labou o ce consis s o hese wo g oups only
(ei he age o educa ion), wi h a labou o ce sha e o p o he
i s g oup and 1 −p o he second g oup. Wo ke s a e ei he
employed o unemployed; i hey a e unemployed, we assume
hey a e seeking a new job. The agg ega ed unemploymen a e
uconsis s o he g oup‐speci ic a es weigh ed a he espec i e
labou o ce sha e: u=pu
1
þ(1 −p)u
2
.
In he s anda d sea ch and ma ching amewo k, equilib ium
unemploymen is explained by wo low a es: he sepa a ion
a e sand he job inding a e (θ). While sis he isk o job loss
and he co esponding low o unemploymen , (θ) is he
p obabili y o an unemployed pe son inding new employmen
(wi h ma ke igh ness θ). Gi en ha u,u
1
, and u
2
a e in
equilib ium, we ha e:
u∗=pu∗
1+(1−p)u∗
2=ps1
s1+ 1(θ)+(1−p)s2
s2+ 2(θ)(1)
I is easy o see ha a ising pinc eases (dec eases) u* when
u∗
1>u∗
2(u∗
1<u∗
2).
In addi ion, o a gi en dis ibu ion o he g oups, he unem-
ploymen a e ises i a leas one g oup's unemploymen a e
inc eases. Finally, he low a es in o and ou o unemploymen
can explain he g oup‐speci ic unemploymen a e change.
To poin ou ha g oup‐speci ic lows and unemploymen a es
ma e , we use agg ega ed s ock and low da a p o ided by he
Ge man Fede al Employmen Agency. Table 1p o ides a e age
alues o you h (15–24 yea s) and olde wo ke s (50–64 yea s)
as well as o low (no app en iceship) and high‐educa ed
wo ke s (academic educa ion) o he pe iod 2008–2018 using
mon hly da a. u∗
iis calcula ed acco ding o Equa ion (1) and u
i
is he usual unemploymen a e, calcula ed using he g oup‐
speci ic s ock o he unemployed di ided by he g oup‐speci ic
s ock o he labou o ce.
The sepa a ion isk and job‐ inding a e a e abo e a e age o
you h. Sepa a ion is abo e he a e age o he low‐educa ed
wo ke s, bu job inding is below. This di e ence is he eason
o he as di e ences be ween hese wo g oups in e ms o he
unemploymen a e. Olde wo ke s and hose who a e highly
educa ed ha e sepa a ion a es below he a e age. Conce ning
job inding, hese wo g oups di e ; he highly educa ed a e
abo e he a e age, and olde wo ke s a e below. Again, his
di e ence explains he di e ence in unemploymen a es.
Ano he impo an inding is ha he age g oups ha e a clea ,
dynamic pa e n. Compa ed o olde wo ke s, you h is mo e
o en unemployed bu also as e eemployed. In con as ,
highly educa ed wo ke s bene i om a ou able low a es,
while he low‐educa ed su e om ad e se low a es con-
ce ning he unemploymen a e. Wi hin all g oups, bo h a e age
unemploymen a es in Table 1a e e y simila , no wi h-
s anding ha he equilib ium unemploymen a e is a simpli ied
concep . Hence, he low a es a e use ul in explaining wha
happens i he sha e o age o educa ion g oups changes. Fo
example, om an agg ega ed pe spec i e, we can a gue ha less
low‐educa ed and mo e high‐educa ed wo ke s mean less sep-
a a ion and as e job inding, which dec eases he unemploy-
men a e. Howe e , mo e olde and less younge wo ke s esul
in less sepa a ion bu also slowe job inding. Labou ma ke
dynamics ob iously decline, bu he e ec on he unemploy-
men a e is ambiguous.
7
The e ec s a e mo e complex when conside ing egions (e.g.,
coun ies) and allowing wo ke s o commu e be ween neigh-
bou ing egions.
8
Younge wo ke s a e egional mo e mobile
han olde wo ke s, and highly educa ed wo ke s a e mo e
mobile han low‐educa ed wo ke s. To accommoda e his
empi ical obse a ion, we ocus on egional labou ma ke
in e ac ions.
The sea ch a e σ=uþeis he sum o unemployed and employed
job seeke s di ided by he labou o ce, wi h e≤1−u. F om a
egional pe spec i e, i is ob ious ha people apply no only o
jobs in hei home egion bu also in su ounding egions. In his
case, wo ke s commu e be ween hei home and wo kplace e-
gion. We e e o commu ing and in e ‐ egional sea ches as
mobili y; hence, his de ini ion excludes mo es om one egion
o ano he . To main ain he model's simplici y, we conside job
seeke s and acancies only om he local egion land egions
adjacen o l, which we ea as one homogenous egion, n.
The igh ness (θ) o he local labou ma ke is gi en by
TABLE 1 |A e age low a es and unemploymen a es in Ge many.
Sepa a ion a e Job inding a e u∗
iu
i
You h 0.038 0.416 8.4 8.1
Olde wo ke s 0.018 0.172 9.8 10.2
Low educa ed 0.073 0.216 25.7 25.7
High educa ed 0.011 0.25 4.3 4.3
All 0.023 0.228 9.4 9.5
No e: Mon hly da a a e aken om s a is ics o he Fede al Employmen Agency. Job‐ inding a es a e calcula ed as he a io o lows om unemploymen o employmen
in he p e ious mon h, and sepa a ion a es a e calcula ed as lows om employmen o unemploymen in he p e ious mon h. Equilib ium unemploymen a es a e
calcula ed acco ding o Equa ion (1), and he (no mal) unemploymen a e is calcula ed as he numbe o unemployed di ided by he labou o ce. Pe iod: Janua y 2008
o Decembe 2018.
3 o 20
θl= l/(ul+el+u
∼n+e
∼n)= l/(σl+σ
∼n),
and he igh ness o he adjacen dis ic s' labou ma ke is
gi en by
θn= n/(un+en+u
∼l+e
∼l)= n/(σn+σ
∼l),
whe e
l
(
n
) deno es he local (neighbou hood) acancy a e
and ~ ep esen s spa ial sea ch ac i i ies. All job seeke s apply
o jobs in hei home egion. Because younge wo ke s and
mo e educa ed wo ke s a e mo e mobile, he numbe o
egional mobile job applican s depends on he age and educa-
ion s uc u e o he job seeke s. Only some o he olde and
low‐educa ed job seeke s om neighbou ing egions apply o
jobs in he local egion. We e e o σl=plσl
1+(1−pl)σl
2and
σn=pnσn
1+(1−pn)σn
2as local sea ch a es and
σ
∼n=[pnσn
1+(1−pn)σn
2α]Ln/Lland σ
∼l=[plσ1
l+(1−pl)
σ2
lα]Ll/Lnas spa ial sea ch a es.
All wo ke s esiden in he local egion, L
l
, a e no malised o 1.
The a e σ
∼n, ela ed o he labou o ce in he local labou
ma ke , has he same denomina o as σ
l
.σ
∼nand σ
n
di e
because hey a e ela ed o di e en labou o ce sizes, σ
∼n o L
l
and σ
n
o L
n
. The sha e o G oup 2 (low‐educa ed o olde
wo ke s) job seeke s is la ge in hei esiden egion. The
mobili y weigh ing ac o α, wi h 0 ≤α<1, accommoda es he
limi ed spa ial mobili y o olde and low‐educa ed wo ke s;
hence, σn
2>σn
2α. The di e ences be ween σ
∼land σ
l
a e analo-
gous. This a ec s he dis ibu ion o he job seeke s a ailable o
local i ms: plσl
1
σl+σ
∼n+pnσn
1
σl+σ
∼n≡pl. Hence, he job seeke
s uc u e depends on he g oup dis ibu ion (educa ion o age)
o he labou o ce in bo h egions.
Job seeke s om he local egion ind, on a e age, new
employmen a he a e l
i(θ
l
,pl)+ n
i(θ
n
,pn) because o he
spa ially mobile sea ch ac i i ies. F om his, i ollows ha he
spa ial co ela ion o unemploymen a es is posi i e, and he
(spa ial) co ela ion o acancy and unemploymen a es is
nega i e. Sepa a ions can di e ac oss g oups and egions.
Finally, he local labou o ce, L
l
, can be subdi ided in o h ee
g oups: local unemployed u
l
, esiden s employed in he local
egion ω
l,l
, and esiden s employed in he neighbou egion ω
l,n
.
Since L
l
=1, we ha e u
l
þω
l,l
þω
l,n
=1.
The local unemploymen a es o a g oup e ol e acco ding o
sepa a ion and job inding, wi h i=[y,o] o i=[μ,h]:
˙
ul
i=sl
i(1−ωl,n
i−ul
i)+sn
iωl,n
i− l
i(θl,pl)ul
i− n
i(θn,pn)ul
i.
sl
i(1−ωl,n
i−ul
i)is he g oup‐ ela ed low in o unemploymen
om local employmen . sn
iωl,n
iis he g oup‐ ela ed low in o local
unemploymen om jobs in he neighbou ing egion. On he
igh ‐hand side, he las wo e ms a e he p obabili ies o an-
si ion in o a new job in he local and neighbou ing labou ma ke .
Wi h ˙
ui=0 and he summa ion o he wo unemploymen a es
weigh ed a he espec i e local popula ion p opo ions, p
l
and
(1 −p
l
), we ob ain he local equilib ium unemploymen a e:
ul=ul
2+pl(ul
1−ul
2)
=sl
2+(sn
2−sl
2)ωl,n
2
sl
2+ l
2(θl,pl)+ n
2(θn,pn)
+pl(ul
1−ul
2),
includes spa ial and (spa ial) g oup e ec s. The second e m on
he ac ion line indica es ha local unemploymen inc eases as
he numbe o spa ially mobile wo ke s inc eases and sn
2>sl
2,
wi h ωl,n
2as he sha e o esiden s employed in he neighbou ing
egion (n). The e a e wo channels o he g oup‐ ela ed e ec s:
he i s e ec is ‘hidden’ in he (spa ial) job‐ inding a es, and
he second is ela ed o he di e ences in g oup‐ ela ed unem-
ploymen a es. This second e m disappea s i ul
1=ul
2. Fo
ul
1>ul
2(ul
1<ul
2), an inc easing p opo ion o G oup 1 wo ke s
inc eases (dec eases) o e all sepa a ion and unemploymen .
The i s e ec con ains g oup‐ ela ed ma ching e iciency and
mobili y e ec s on he neighbou ing labou ma ke . This e ec
means ha he mo e G oup 1 wo ke s a e in he neighbou ing
egion, he lowe he local ma ke igh ness and, hence, he
lowe he p obabili ies o ansi ion in o a new job o local
wo ke s (plis he sha e o job seeke s a ailable o local i ms).
Thus, he p opo ion o g oup‐speci ic wo ke s in he local and
su ounding labou ma ke s is impo an o he local unem-
ploymen a e. In he empi ical pa , we es ima e egional panel
da a wi h a spa ial panel model o analyse his empi ically.
4
|
Empi ical Analysis
This sec ion analyses he ela ionship be ween changes in age
and educa ion composi ion o he labou o ce and he unem-
ploymen a e using da a o Ge many. The p e ious sec ion
shows he s ock and low da a p o ided by he Ge man Fede al
Employmen Agency (Table 1). In his sec ion, we i s discuss
he de elopmen o he g oups ha a e conside ing using OECD
da a o Ge many. The educa ion g oups a e now di e en ia ed
by he ISCED classi ica ion, which is sligh ly di e en om he
Ge man Fede al Employmen Agency classi ica ion. Figu e 1
p o ides he e olu ion o wo age g oup sha es and wo edu-
ca ion g oup sha es o 1999–2018. The sha es o p ime‐age
wo ke s (25–49 yea s) and medium‐educa ed (ISCED 3–4) a e
no displayed.
The end in he da a shows ha he sha es o he younge and
low‐educa ed decline o e ime, while he sha es o olde and
highly‐educa ed wo ke s inc ease. Gi en ha he unemploy-
men a es o olde and highly educa ed wo ke s a e lowe ,
hese shi s mus dec ease he o e all unemploymen a e.
Focussing only on younge o low‐educa ed wo ke s, as ypi-
cally done in he li e a u e, is misleading because ends in
o he g oups a e no conside ed. The e olu ions o Ge man
unemploymen a es o age and educa ion g oups a e p o ided
in Appendix A(Figu es A1 and A2).
Based on his compa ison, analysing he ela ionship be ween
di e en age and educa ion g oups and unemploymen appea s
4 o 20 G ow h and Change, 2025
meaning ul. Since he analysis o mac oeconomic da a would
p o ide no subs an ial new indings, egional da a will be
applied because hey allow o conside ing a mo e di e en ia ed
pa e n. The e o e, he econome ic analysis will u ilise coun y‐
le el da a (NUTS‐3 le el).
4.1
|
Da a and Econome ic F amewo k
We use he Ge man Sample o In eg a ed Labou Ma ke Bi-
og aphies (SIAB‐7514), a andom sample d awn om he In e-
g a ed Employmen Biog aphies (IEB). This da a sou ce en ails
indi idual da a on labou ma ke biog aphies.
9
Co e ing
16 yea s om 1999 o 2014, we con e ed he aw da a in o
mon hly segmen s. Hence, we analyse he pe iod om Janua y
1999 un il Decembe 2014. We compu ed ou a iables a he
indi idual le el and agg ega ed hem a he adminis a i e dis-
ic le el. This gi es us a s ongly balanced panel consis ing o
402 c oss‐sec ion uni s (coun ies) and 77,184 obse a ions o
di e en sha es o he local labou o ce. We e e o
Da a Desc ip ion sec ion in Appendix A o a de ailed desc ip-
ion o he edi ing p ocess.
Table 2desc ibes how he a iables a e gene a ed and p o ides a
summa y s a is ic o hese a iables. We will use he age g oup
sha es in di e en combina ions o deal wi h di e en e e ence
g oups. The wo di e en ypes educa ion and schooling will be
used sepa a ely. Educa ion is ela ed o o mal educa ion a e
schooling and consis s o no app en iceship, app en iceship, and
academic deg ee. Schooling is ela ed o he las school lea ing
ce i ica e and is sepa a ed in o no ce i ica ion, ce i ica ion
wi hou a uni e si y en ance quali ica ion, and high school
deg ee. In bo h cases, he e e ence in he eg essions is he
medium g oup (in e ms o schooling, i is ce i ica ion wi hou
a uni e si y en ance quali ica ion, and in e ms o educa ion, i
is app en iceship).
We conside only employed and unemployed indi iduals o he
wo king‐age popula ion be ween 15 and 64 yea s. This is ela ed
o he e i emen age in Ge many. In addi ion, we do no
conside o he indi idual cha ac e is ics like gende o wo k
expe ience. Also, we abs ain om in e ac ing age wi h educa-
ion g oups, such as subdi iding you h in o h ee g oups: you h
wi hou app en iceship, you h wi h app en iceship, and you h
wi h academic deg ees.
10
This is caused by he limi ed numbe
o indi iduals a he egional le el. In some cases, we would no
ha e enough indi iduals wi hin a speci ic subg oup o ou
econome ic analysis. The e o e, we decided o use only age
g oups and educa ion g oups in he labou o ce.
We a e p ima ily in e es ed in he e ec s o a change in g oup
composi ions in he local and neighbou ing egions on local
unemploymen . Fo example, he local you h sha e cap u es
g oup‐speci ic job inding and sepa a ion in he local egion. In
addi ion, we conside he e ec o changes in you h sha e in he
neighbou ing egion on local unemploymen . Since he young
in bo h egions hold, on a e age, he same job‐ ele an cha -
ac e is ics, we do no a gue ha , o example, younge wo ke s
in neighbou ing egions a e mo e p oduc i e han hose in he
local egion. This is no possible because, in he es ima es, e e y
sha e is conside ed as a local egion and as a neighbou egion (I
am he neighbou o my neighbou ). Howe e , he neighbou ing
egion's you h needs o be spa ially mobile (in e ms o
commu ing) o wo k in he local egion.
11
This is why we wan
o conside you h sha e e ec s o bo h he local and he
neighbou ing egions on local unemploymen in he es ima es.
In addi ion, we also wan o con ol o o he age g oups o
di e en ia e be ween hese g oups and he e e ence g oup o
olde wo ke s. Fo educa ion, we a gue in he same way. He e,
we expec ha he low‐educa ed g oup is less mobile. Since his
g oup is o en small, we use he medium‐educa ed g oup as a
e e ence. To conside all hese aspec s, we use a lexible spa ial
econome ics app oach: The spa ial Du bin model. In he
ollowing, we desc ibe he s uc u e o he model.
To accoun o addi ional unobse ed ime and spa ial a ying
e ec s a he local le el, ime lagged and spa ial lagged e ec s o
he dependen a iable ln u
i
(unemploymen a e in loga i hm)
FIGURE 1 |Labou o ce sha e o age and educa ion in Ge many.
5 o 20
a e conside ed (Equa ion 2). To gene a e spa ially lagged
coun e pa s, we cons uc ed a spa ial weigh ma ix, W, ha
indica es he con igui y o egions and de ined con igui y be-
ween wo egions as hose ha sha e a common bo de . The
ma ix has he en y 1 i wo egions sha e he same bo de and
0 o he wise. Then, we ow no malise W, which ensu ed ha all
weigh s we e be ween 0 and 1 and ha weigh ing ope a ions
can be in e p e ed as an a e age o he neighbou ing alues.
ln u
i, −1
is he ime lagged dependen a iable and γ he au o -
eg essi e ime dependence pa ame e . Wln u
i
gene a es he
a e age alues o he egions adjacen o egion i, and λis he
spa ial dependence pa ame e — he spa ial lagged e ec o he
dependen a iable. Wln u
i, −1
is he combined spa ial and ime
lagged dependen a iable, and πis he spa io‐ empo al di u-
sion pa ame e . The inclusion o he spa ial and ime lagged
dependen a iable could se e as a con ol o omi ed a iables
o a leas educe omi ed a iable bias (LeSage and Pace 2009).
We will discuss he issue o endogenei y in Sec ion 4.4.
To sum up, we conside a spa ial and ime dynamic model ha
is also known as he dynamic spa ial Du bin model (wi h ime
and ixed e ec s):
ln ui =γln ui, −1+λW ln ui +πW ln ui, −1
+α1lnageg oup1i +β1Wlnageg oup1i
+α2lnageg oup2i +β2Wlnageg oup2i
+α3lnedug oup1i +β3Wlnedug oup1i
+α4lnedug oup2i +β4Wlnedug oup2i
+ci+θ +ei
(2)
whe e ln u
i
, ln ageg oup
i
, ln edug oup
i
and e
i
a e s acked
Tn �1 column ec o s, Wis a ow no malised n�nspa ial
weigh s ma ix ha is nons ochas ic and gene a es he spa ial
dependence be ween c oss‐sec ional uni s, c
i
a e egional, and θ
a e ime e ec s. ln ageg oup1 is he sha e o he i s age g oup
(e.g., you h) in he local egion, and Wln ageg oup1 is he
a e age sha e o he same g oup in he neighbou ing egions.
The same applies o he second age g oup and he wo educa ion
g oups (edug oup).
12
The bias‐co ec ed quasi maximum like-
lihood app oach p o ided by Yu, de Jong, and Lee (2008) is
conside ed o he dynamic models.
13
The e ec s o he ime and
spa ial lagged dependen a iable will no be discussed below.
14
Howe e , hese lags help a e wa ds o calcula e he dynamic
long‐ un e ec s. In all eg essions, coun y‐clus e obus s an-
da d e o s a e conside ed.
The pa ame e s αand βin Equa ion (2) canno be in e p e ed as
elas ici ies o pa ial de i a i es due o spillo e e ec s.
15
The e o e, we i s p o ide he es ima ed coe icien s and, sub-
sequen ly, he esul ing elas ici ies. Because o hei limi ed
mobili y, no all olde wo ke s o less educa ed wo ke s in he
neighbou ing egion apply o jobs in he local egion, and
he e o e, he spa ial g oup sha es se e mos ly as a p oxy a -
iable o mobili y in e ms o commu ing.
Fo example, le us assume ha p ime‐age wo ke s a e mo e
a ac i e o i ms han o he age g oups. An inc ease in he
neighbou ing p ime‐age sha e induces mo e job applica ions a
i ms in he local egion. This, in u n, dec eases sea ch cos s
and inc eases he acancy a e. Howe e , his also dec eases he
local ma ke igh ness and he p obabili y o ansi ioning in o a
new job o local job seeke s. This e ec is likely la ge han he
e ec on acancies (mo e jobs). Wi h espec o he pa ame e β,
we expec a posi i e e ec . In con as , he pa ame e αis
TABLE 2 |Va iable desc ip ion and basic s a is ics.
Va iable Desc ip ion Obs Mean SE Min Max
Unemploymen
a e
Numbe o unemployed indi iduals in he dis ic labou o ce o all
indi iduals in he dis ic labou o ce
77,184 0.0996 0.055 0.008 0.450
Age g oups
You h Numbe o indi iduals in he dis ic labou o ce aged 15–24 o all
indi iduals in he dis ic labou o ce aged 15–64
77,184 0.069 0.018 0.012 0.152
25–39 yea s Numbe o indi iduals in he dis ic labou o ce aged 25–39 o all
indi iduals in he dis ic labou o ce aged 15–64
77,184 0.356 0.052 0.217 0.512
25–49 yea s Numbe o indi iduals in he dis ic labou o ce aged 25–49 o all
indi iduals in he dis ic labou o ce aged 15–64
77,184 0.666 0.045 0.476 0.785
40–49 yea s Numbe o indi iduals in he dis ic labou o ce aged 40–49 o all
indi iduals in he dis ic labou o ce aged 15–64
77,184 0.310 0.032 0.184 0.411
Educa ion/schooling g oups
No
App en iceship
Numbe o indi iduals in he dis ic labou o ce wi hou oca ional
educa ion o all indi iduals in he dis ic labou o ce
77,184 0.134 0.057 0 0.337
Academics Numbe o indi iduals in he dis ic labou o ce wi h an academic
educa ion o all indi iduals in he dis ic labou o ce
77,184 0.115 0.063 0.016 0.518
No G adua ion Numbe o indi iduals in he dis ic labou o ce wi h no g adua ion o
all indi iduals in he dis ic labou o ce
77,184 0.013 0.010 0 0.090
High School Numbe o indi iduals in he dis ic labou o ce wi h uni e si y
en ance quali ica ion o all indi iduals in he dis ic labou o ce
77,184 0.183 0.092 0.030 0.691
No e: Da a a e aken om he SIAB‐7514 and agg ega ed o a balanced mon hly coun y le el om Janua y 1999 o Decembe 2014.
6 o 20 G ow h and Change, 2025
nega i e i he local sha e o p ime‐age wo ke s inc eases, and
his age g oup is mo e a ac i e o i ms o e all.
I he spa ial e ec s o he conside ed g oup s uc u es a e
essen ial, we ha e o conside he bias on αi we neglec β. Le ω
be he pa ame e o he local e ec when he spa ial e ec is
neglec ed. The s anda d esul is hen ω=αþβδ, whe e δ
measu es he co a iance o he local and he spa ial age o ed-
uca ion s uc u e. The la e is posi i e in he da a, and we
expec β o be posi i e, which yields a posi i e bias on ω.
Conce ning he s a is ical ele ance o ou es ima es, we p o ide
coun y‐clus e obus s anda d e o s o con ol o he e o-
skedas ici y, se ial co ela ion and c oss‐sec ional dependence
in he esiduals. We conside he False Posi i e Risk (FPR) o
p o ide in o ma ion o s a is ical e idence. In con as o p‐
alue, he FPR measu es he p obabili y o he null hypo hesis
being ue (Colquhoun 2017,2019). Fo a discussion o he
misin e p e a ion o p‐ alue, see, o example, Wasse s ein and
Laza (2016). We conside FPR =0.05 (equals p‐ alue o 0.0034)
and FPR =0.01 (equals p‐ alue o 0.0005). Fo he compu a ion
o he FPR, we e e o Compu a ion o he False Posi i e Risk
sec ion in Appendix A.
4.2
|
Resul s
Table 3p o ides di e en basic speci ica ions o Equa ion (2).
Fo easons o compa ison, eg essions (1), (5), and (9) a e
simple ixed and ime e ec s models. In eg essions (2), (6), and
(10), only he spa ial lagged dependen is conside ed (γ=π=0),
while in (3), (7), and (11), he ime lagged e ec is also included
(π=0). In (4), (8), and (12), all lagged e ec s o he dependen
a e conside ed. Since all spa ial and ime lagged e ec s p o ide
s ong empi ical e idence, we p e e (4), (8), and (12) as he bes
speci ica ion.
16
We conside only one g oup in hese es ima es
because he ocus he e is on model speci ica ion in gene al.
Howe e , a sho discussion o he esul s will help us ela e ou
indings o he exis ing li e a u e. The es ima ed elas ici y o he
s anda d ixed and ime e ec s es ima es can be compa ed wi h
he long‐ un elas ici ies p o ided below. While (1) and (5) a e in
line wi h he li e a u e, eg ession (9) p o ides no eliable es i-
ma es. This is because he sha e o hose who g adua e wi hou
a school lea ing ce i ica e is e y small (abou 1%). The e o e,
he esul s o eg ession (9)–(12) will no be in e p e ed u he .
Conce ning eg essions (2)–(4), we ind empi ical e idence o
he local e ec o he you h sha e on unemploymen . Due o he
s ong empi ical e idence o lagged dependen e ec s, we ely
mos on eg ession (4). In his case, he spa ial e ec o he
you h sha e p o ides no empi ical e idence. The eg essions
(6)–(8) conside o mal educa ion.
Independen o he speci ica ion, we ind ha he local sha e o
hose in he labou o ce wi h no app en iceship is posi i ely
ela ed o he unemploymen a e, while he spa ial e ec is
nega i e. Hence, compa ed o he e e ence g oup, his g oup is
less a ac i e o i ms.
TABLE 3 |Basic esul s o age, educa ion, and schooling.
Dependen a iable: log unemploymen a e
Re e ence g oup Age 25–64 A leas app en iceship A leas seconda y educa ion
Conside ed g oups (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Log you h 0.193
b
0.123
b
0.037
b
0.039
b
(0.016) (0.013) (0.003) (0.003)
W(log you h) 0.078
a
−0.016 −0.005
(0.024) (0.006) (0.005)
Log no app en iceship 0.316
b
0.253
b
0.079
b
0.078
b
(0.021) (0.018) (0.005) (0.005)
W(log no
app en iceship)
0.020 −0.042
b
−0.030
b
(0.031) (0.008) (0.007)
Log no g adua ion 0.002 0.001 0.001 0.001
(0.003) (0.002) (0.001) (0.001)
W(log no g adua ion) 0.009
a
0.001 0.001
(0.003) (0.001) (0.001)
Spa ial lag (λ) 0.477
b
0.162
b
0.241
b
0.478
b
0.166
b
0.244
b
0.499
b
0.172
b
0.243
b
Time lag (γ) 0.798
b
0.809
b
0.789
b
0.800
b
0.804
b
0.813
b
Spa ial‐ ime lag (π)−0.116
b
−0.113
b
−0.112
b
Wi hin R
2
0.808 0.158 0.919 0.919 0.814 0.601 0.920 0.920 0.803 0.133 0.919 0.918
Obse a ions 77,184 77,184 76,782 76,782 77,184 77,184 76,782 76,782 77,184 77,184 76,782 76,782
No e: Spa ial lag, ime lag, and spa ial‐ ime lag e e o he dependen ; all eg essions include ixed and ime e ec s; coun y‐clus e obus s anda d e o s a e in
pa en heses; pe iod: mon hly da a o Janua y 1999 o Decembe 2014; balanced coun y‐le el panel.
a
FPR ≤0.05.
b
FPR ≤0.01.
7 o 20
Table 4conside s g oups o di e en ages and o mal educa ion
(wi hou app en iceship, wi h app en iceship, academic deg ee).
In addi ion o he you h sha e, a second and hi d age g oup is
added wi h di e en age g oup anges and e e ence g oups. We
do his o addi ionally con ol o possible in e ac ions be ween
he educa ion and age g oups.
17
The educa ional e e ence is
always he g oup wi h app en iceship.
18
While we ind empi -
ical e idence o he local you h sha e e ec in all speci ica ions,
he spa ial e ec is no s a is ically ele an . Fo he age g oup
25–39 yea s, we ind nei he empi ical e idence o he local no
spa ial e ec .
19
When he e e ence age g oup is 50–64, he age
g oup 25–49 has a posi i e spa ial e ec ( eg ession (3)). In
eg ession (4), we subdi ide his age g oup in o 25–39 and 40–
49 yea s. He e, we ind posi i e empi ical e idence only o he
spa ial e ec o he age g oup 40–49 yea s. Olde p ime‐age
wo ke s (40–49 yea s) seem o be mo e mobile and compe i-
i e han he e e ence (50–64 yea s) and a ec he ma ke
igh ness.
20
In addi ion, when we conside he di ec ion o all
age coho e ec s be ween 25 and 49 yea s, he local e ec s a e
nega i e, while he spa ial e ec s a e posi i e. This inding in-
dica es ha his age g oup is mo e a ac i e han he e e ence
coho . We ind he opposi e di ec ion o he e ec s o he
young, which means ha he e e ence is mo e a ac i e o
i ms han young wo ke s.
TABLE 4 |Resul s o age and educa ion.
Dependen a iable: log unemploymen a e
Re e ence app en iceship and age g oup 25–64 40–64 50–64
Conside ed g oups (1) (2) (3) (4)
Log you h 0.036
b
0.036
b
0.034
b
0.034
b
(0.003) (0.003) (0.003) (0.003)
W(log you h) −0.013 −0.011 −0.007 −0.010
(0.006) (0.007) (0.006) (0.007)
Log 25–39 0.002 −0.003
(0.010) (0.012)
W(log 25–39) 0.010 0.040
(0.020) (0.021)
Log 25–49 −0.014
(0.020)
W(log 25–49) 0.119
b
(0.038)
Log 40–49 −0.005
(0.011)
W(log 40–49) 0.072
b
(0.021)
Log no app en iceship 0.077
b
0.077
b
0.078
b
0.078
b
(0.005) (0.005) (0.005) (0.005)
W(log no app en iceship) −0.030
b
−0.030
b
−0.030
b
−0.029
b
(0.008) (0.008) (0.008) (0.008)
Log academics −0.009 −0.009 −0.009 −0.011
(0.006) (0.006) (0.006) (0.006)
W(log academics) 0.049
b
0.050
b
0.044
b
0.041
b
(0.009) (0.009) (0.009) (0.009)
Spa ial lag (λ) 0.240
b
0.240
b
0.240
b
0.239
b
Time lag (γ) 0.796
b
0.796
b
0.795
b
0.795
b
Spa ial‐ ime lag (π)−0.120
b
−0.120
b
−0.122
b
−0.122
b
Wi hin R
2
0.919 0.920 0.920 0.920
Obse a ions 76,782 76,782 76,782 76,782
No e: Spa ial lag, ime lag, and spa ial‐ ime lag e e o he dependen ; all eg essions include ixed and ime e ec s; coun y‐clus e obus s anda d e o s a e in
pa en heses; pe iod: mon hly da a o Janua y 1999 o Decembe 2014; balanced coun y‐le el panel.
a
FPR ≤0.05.
b
FPR ≤0.01.
8 o 20 G ow h and Change, 2025
pa icipa ion in p og ams o ac i e labou ma ke policies (in he da a
since 2000). These da a come om di e en sou ces and a e me ged in
he IEB. The su ix 7514 s ands o he panel e sion co e ing 1975 un il
2014, wi h bo h yea s included in he panel.
Ou main goal was o eshape he aw da a in a way ha enabled us o
dis inguish be ween employmen and unemploymen a he adminis-
a i e dis ic le el (c oss‐sec ion) mon hly ( ime uni ) h oughou he
en i e pe iod o he analysis. We use he e i o ial alloca ion om
Decembe 31, 2014 o ou analysis, which means we ha e 402 c oss‐
sec ion uni s (K eise). We analyse he pe iod om Janua y 1999 o
Decembe 2014, 192 mon hs.
De ining employmen and unemploymen was he i s s ep owa ds
achie ing his pa icula panel s uc u e. Since he SIAB‐7514 consis s o
indi idual da a in i s aw e sion, we de ine an indi idual as employed i
and only i he indi idual epo s he cha ac e is ic a ibu e 101 in he
employmen s a us a iable. In u n, an indi idual is deemed unem-
ployed i and only i i epo s one o he cha ac e is ic a ibu es
1,2,31,32,41,51 in he employmen s a us a iable.
26
To agg ega e he indi idual da a o a mon hly adminis a i e dis ic
le el, we had o edi he o iginal panel subs an ially. In he i s s ep, we
con e he aw da a o sequen ial da a. The SIAB is o ganised in spells.
Each spell has a commencing da e and an end da e, which a e p ecise o
he day. The da es ma k he beginning and he end o an indi idual
episode. The e m ‘p ecise o he day’ implies ha a spell may s a and
end on any gi en day in any gi en mon h in any gi en yea be ween
Janua y 1999 and Decembe 2014. In he aw e sion, spells can o e lap
wi h each o he . Hence, spells o he same indi idual may co e he
same pe iod o o e lap in pa . The e o e, c ea ing sequen ial da a
means o ganising he spells so ha he e will (a) be no cong uency o
pa ial o e lapping and (b) ha each spell s a s exac ly one day a e he
end da e o he p e ious spell i he same indi idual is conside ed. We
ook he STATA code o gene a e sequen ial da a om he da a epo
o he SIAB‐7514.
27
Once he sequen ial da a s uc u e was es ablished, we c ea ed a a i-
able epo ing he single yea o he indi idual spell using he yea
speci ied in he a iable ha ma ked he beginning o an episode and
e ased all spells ha s a ed and ended be o e 1999. Nex , we pay
a en ion o hose spells ha o e lapped yea s. Tha is, spells ha s a ed
in 1 yea and ended in he ollowing yea o he yea a e ha . We ocus
solely on spells ha co e a maximum 2‐yea span based on he yea
epo ed in he column o he s a da e o he episode. All spells
exceeding his limi we e d opped om he panel. This is because he
ac ual numbe o hese spells is e y small. Less han 3% o he en i e
panel is a ec ed. Since hese spells almos exclusi ely epo a ibu es
in he a iable depic ing he employmen s a us we do no use o ou
analysis, almos all hose spells a e d opped la e . The emaining spells,
which o e lapped 1 yea o mo e, a e spli in o se e al episodes o a ach
a unique alue o he yea a iable o each episode. We hen d op hose
spell episodes ha s a ed be o e 1999. Fo example, i a spell s a ed in
1998 and ended in 2000, we duplica ed he o iginal spell wo imes,
ending wi h h ee iden ical spells. Each spell is assigned o a yea , ha
is, 1998, 1999, and 2000. We kep he wo la e spells and d opped he
i s . Once he spells a e in o de , we spli he panel in o indi idual
yea s, e ec i ely gi ing up he panel s uc u e.
In he nex s ep, we de elop a code ha enables us o b and an indi-
idual as employed o unemployed in any gi en mon h om Janua y
1999 o Decembe 2014. We employ a syn ax ha ope a es in a way ha
an indi idual is assigned o he s a us o being employed o unemployed
o a gi en mon h i he indi idual s a us exceeds hal o a gi en mon h.
Fo example, a pe son epo s an episode o employmen in a gi en yea
ha s a s on Janua y 1s and ends no la e han Ma ch 15 h. This
pe son is assigned o be employed h oughou Janua y and Feb ua y o
ha yea . Fo Ma ch, howe e , he indi idual could be assigned ei he
s a us. In Ma ch, an indi idual will be employed i he consecu i e
episode s a s he day a e he p e ious episode and expe iences a
change in any a iable, bu he a iable indica es he employmen s a-
us. On he o he hand, an indi idual is unemployed in Ma ch i he
a iable indica es a change in employmen s a us om 101 o any
numbe in 1,2,31,32,41,51.
We c ea e ou a iables once we ha e ans o med he indi idual epi-
sodes in o mon hly episodes. We ecode he o iginal indus y b anches
in he panel by ca ego ising hem in o 13 b anches o economic ac i i y
using he s anda ds p o ided in he FDZ da a epo o he Sample‐o ‐
In eg a ed‐Labou ‐Ma ke ‐Biog aphies Regional‐File 1975–2010 (SIAB‐
R 7510).
28
We compu ed he espec i e sha es o employmen and unemploymen
o each combina ion o c oss‐sec ion and ime uni o all a iables.
Finally, we agg ega e he indi idual da a a he adminis a i e dis ic
le el and ecei e a s ongly balanced panel, which is he basis o ou
empi ical analysis.
Compu a ion o he False Posi i e Risk
The alse posi i e isk (FPR) was in oduced by Colquhoun (2019,2017)
and measu es he p obabili y ha he esul occu ed by chance P(H
0
|
da a). The app oach is based on he Bayes heo em ha we exp ess in
odds:
pos e io odds on H1=Bayes ac o ×p io odds
This is equal o
P(H1|da a)
P(H0|da a)=P(da a |H1)
P(da a |H0)×P(H1)
P(H0)
Following Colquhoun, he Bayes ac o becomes a likelihood a io (LR),
and he p io odds can be exp essed using he p obabili y ha he e is a
eal e ec , P(H
1
): P(H
1
)/(1 −P(H
1
)). Among o he s, Sellke, Baya i,
and Be ge (2001) p o ide an app oach o calcula e he LR based on he
p‐ alue: LR =1/(−ep log (p)). Howe e , his measu e can be conside ed
only as long as p<1/e, wi h eas Eule 's numbe .
Taking hings oge he and conside ing P(H
0
| da a) =1−P(H
1
| da a)
gi es us he FPR:
FPR =1
1+1
−ep log(p)
P(H1)
1−P(H1)
Applying he FPR app oach equi es o speci y P(H
1
) i s . Howe e ,
speci ying he p io p obabili y in eg ession analysis is di icul , and we
should always be ca e ul when de ining his unknown numbe . We use
P(H
1
)/(1 −P(H
1
)) =0.5/(1 −0.5) =1, which means ha bo h p oba-
bili ies ha e he same weigh . This is equal o a 50:50 chance o a eal
e ec speci ied be o e he da a a e analysed. This seems easonable
when we do no know wha o choose o a e open o he esul s. When
he p io p obabili y o a eal e ec is 0.5, he FPR is much la ge han
he co esponding p‐ alue, and, o example, p=0.05 is equal o a FPR
o 0.2893.
15 o 20
FIGURE A1 |Unemploymen a es by age g oups in Ge many.
FIGURE A2 |Unemploymen a es by educa ion g oups in Ge many.
16 o 20 G ow h and Change, 2025
FIGURE A3 |Fi s di e ence o sha es and lagged i s di e ence o unemploymen a e.
17 o 20
TABLE A1 jFu he esul s using age, educa ion o schooling.
Age
Re e ence g oup 40–64 50–64 50–64 App en iceship Seconda y educa ion
Conside ed g oups (1) (2) (3) (4) (5)
Log you h 0.041
b
0.039
b
0.038
b
(0.003) (0.003) (0.003)
W(log you h) 0.001 0.001 −0.006
(0.006) (0.005) (0.006)
Log 25–39 0.016 0.013
(0.010) (0.010)
W(log 25–39) 0.034 0.053
(0.020) (0.021)
Log 25–49 0.023
(0.018)
W(log 25–49) 0.192
b
(0.034)
Log 40–49 0.004
(0.011)
W(log 40–49) 0.083
b
(0.019)
Log no app en iceship 0.082
b
(0.005)
W(log no app en iceship) −0.022
(0.008)
Log academics −0.006
(0.006)
W(log academics) 0.059
b
(0.009)
Log no g adua ion 0.001
(0.001)
W(log no g adua ion) 0.001
(0.001)
Log high school −0.012
(0.006)
W(log high school) 0.069
b
(0.009)
Spa ial lag (λ) 0.241
b
0.239
b
0.239
b
0.240
b
0.238
b
Time lag (γ) 0.809
b
0.807
b
0.807
b
0.799
b
0.811
b
Spa ial‐ ime lag (π)−0.116
b
−0.120
b
−0.121
b
−0.116
b
−0.119
b
Wi hin R
2
0.919 0.919 0.919 0.905 0.894
No e: Dependen a iable: log o unemploymen a e; spa ial lag, ime lag, and spa ial‐ ime lag e e o he dependen ; all eg essions include ixed and ime e ec s;
coun y‐clus e obus s anda d e o s a e in pa en heses; pe iod: mon hly da a o Janua y 1999 o Decembe 2014; balanced coun y le el panel.
a
FPR ≤0.05.
b
FPR ≤0.01.
18 o 20 G ow h and Change, 2025
TABLE A2 jSho ‐ un and long‐ un elas ici ies: age and schooling.
Sho ‐ un elas ici ies Long‐ un elas ici ies
Conside ed g oups Di ec Indi ec To al Di ec Indi ec To al
Dependen a iable: log unemploymen a e
Table 5: Reg ession (2): e e ence: seconda y educa ion and 40–64 yea s
You h 0.043 −0.008 0.035 0.234 0.157 0.391
(0.003) (0.008) (0.009) (0.020) (0.149) (0.161)
25–39 0.016 0.006 0.022 0.091 0.168 0.259
(0.010) (0.025) (0.027) (0.057) (0.363) (0.393)
No G adua ion 0.001 0.001 0.001 0.003 0.007 0.010
(0.001) (0.001) (0.001) (0.004) (0.013) (0.015)
High School −0.011 0.076 0.065 −0.020 0.765 0.745
(0.006) (0.011) (0.011) (0.037) (0.309) (0.327)
Table 5: Reg ession (4): e e ence: seconda y educa ion and 50–64 yea s
You h 0.040 −0.014 0.026 0.214 0.051 0.265
(0.003) (0.008) (0.009) (0.019) (0.105) (0.117)
25–39 0.018 0.035 0.053 0.115 0.429 0.544
(0.010) (0.027) (0.028) (0.059) (0.323) (0.351)
40–49 0.004 0.072 0.076 0.055 0.713 0.768
(0.011) (0.023) (0.022) (0.057) (0.287) (0.304)
No G adua ion 0.001 0.001 0.001 0.003 0.006 0.008
(0.001) (0.001) (0.001) (0.004) (0.012) (0.015)
High School −0.016 0.060 0.044 −0.061 0.504 0.443
(0.006) (0.011) (0.012) (0.036) (0.172) (0.186)
No e: Di ec e ec s come om he local egion, and he indi ec e ec s come om he neighbou ing egions. Long‐ un e ec s cumula e eedback o e he pe iod
conside ed. Robus s anda d e o s a e in pa en heses; pe iod: mon hly da a o Janua y 1999 o Decembe 2014; balanced coun y‐le el panel; obse a ions: 76,782.
19 o 20
TABLE A3 jResul s o age and educa ion wi h second o de
neighbou ma ix.
Dependen a iable: log
unemploymen a e
Re e ence age g oup 25–64 40–64 50–64
Conside ed g oups (1) (2) (3) (4)
Log you h 0.026
b
0.025
b
0.024
b
0.024
b
(0.003) (0.003) (0.003) (0.003)
W(log you h) −0.012 −0.011 −0.008 −0.012
(0.009) (0.011) (0.010) (0.011)
Log 25–39 −0.007 −0.013
(0.010) (0.012)
W(log 25–39) 0.011 0.041
(0.033) (0.036)
Log 25–49 −0.035
(0.020)
W(log 25–49) 0.137
(0.053)
Log 40–49 −0.012
(0.011)
W(log 40–49) 0.070
(0.025)
Log no app en iceship 0.079
b
0.080
b
0.080
b
0.080
b
(0.006) (0.006) (0.006) (0.006)
W(log no
app en iceship)
−0.059
b
−0.060
b
−0.058
b
−0.058
b
(0.010) (0.010) (0.010) (0.010)
Log academics −0.013 −0.012 −0.011 −0.011
(0.006) (0.006) (0.006) (0.006)
W(log academics) 0.054
b
0.054
b
0.045
a
0.040
(0.014) (0.014) (0.014) (0.015)
Spa ial lag (λ) 0.474
b
0.474
b
0.473
b
0.473
b
Time lag (γ) 0.788
b
0.788
b
0.788
b
0.788
b
Spa ial‐ ime lag (π)−0.304
b
−0.304
b
−0.306
b
−0.307
b
Wi hin R
2
0.919 0.919 0.920 0.920
Obse a ions 76,782 76,782 76,782 76,782
No e: Spa ial lag, ime lag, and spa ial‐ ime lag e e o he dependen ; all
eg essions include ixed and ime e ec s; coun y‐clus e obus s anda d e o s
a e in pa en heses; pe iod: mon hly da a o Janua y 1999 o Decembe 2014;
balanced coun y‐le el panel.
a
FPR ≤0.05.
b
FPR ≤0.01.
20 o 20 G ow h and Change, 2025
