Temporary migration and wages of Ph.D.s. stay longer or come back sooner?

Di Cin io, Ma co; G assi, Emanuele
A icle
Tempo a y mig a ion and wages o Ph.D.s. s ay longe o
come back soone ?
Re iew o Economic Analysis (REA)
P o ided in Coope a ion wi h:
In e na ional Cen e o Economic Analysis (ICEA), Wa e loo, On a io
Sugges ed Ci a ion: Di Cin io, Ma co; G assi, Emanuele (2025) : Tempo a y mig a ion and wages
o Ph.D.s. s ay longe o come back soone ?, Re iew o Economic Analysis (REA), ISSN 1973-3909,
In e na ional Cen e o Economic Analysis (ICEA), Wa e loo (On a io), Vol. 17, Iss. 1, pp. 1-35,
h ps://doi.o g/10.15353/ ea. 17i1.6084
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/328176
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by-nc/4.0/
Re iew o Economic Analysis 17 (2025) 1-35 1973-3909/2025001
1
www.Ro EA.o g
Tempo a y Mig a ion and Wages o Ph.D.s.
S ay Longe o Come Back Soone ?
MARCO DI CINTIO
Uni e si y o Salen o
Emp y 15
EMANUELE GRASSI
Uni e si y o Salen o
Emp y 15
This pape examines he wage implica ions o empo a y mig a ion o wo coho s o
I alian Ph.D.s. Special a en ion is gi en o he du a ion o he in e na ional expe ience, i s
con ibu ion o ea ned wages and he selec i i y o e u nees. Re u nees a e ound o be a
sel -selec ed g oup whose unobse able cha ac e is ics a e simul aneously associa ed wi h
bo h highe wages and a highe p opensi y o mig a e. Mo eo e , we ind posi i e e u ns
o hose who s ay longe and nega i e e u ns o hose who come back soone . The esul s
a e con i med in se e al obus ness and sensi i i y checks.
Emp y 10
Keywo ds: back mo e s; educa ion; high-skilled mig an s; in e na ional e u nees;
selec i i y; sel -selec ion; human capi al; wage-p emia.
JEL Classi ica ions: I23, J3, J61, F22
1 In oduc ion
The ising end in skilled mig a ion lows has signi ican ly enewed he in e es o schola s
and policy make s on opics ela ed o he b ain d ain, b ain gain and b ain ci cula ion (Beine
e al. 2011; Docquie & Rapopo 2012). As pa o an op imal li e-cycle planning, indi iduals
mo e wi hin and ac oss na ional bo de s o each he loca ions in which skills can be acqui ed
e ec i ely and/o a e be e ewa ded. Such mobili y, in u n, plays a c ucial ole in de e mining
he expansion and con ac ion o he supply o high-skilled labo o a coun y. In his con ex ,
while college and uni e si y g adua es ha e become a e e ence g oup in he esea ch on high-
skilled mig a ion (Faggian & McCann 2009; Haapanen & Te o 2011; Co co an e al. 2010;
Venho s e al. 2011; G oen 2004), we know ela i ely li le abou loca ion choices and labo
 We acknowledge he I alian Na ional Ins i u e o S a is ics o supplying us wi h he da a.
Di Cin io: Dipa imen o di Scienze dell’Economia, ma co.dicin io@unisalen o.i ; G assi: Dipa imen o
di Scienze dell’Economia, emanuele.g[email p o ec ed]
© 2025 Ma co Di Cin io and Emanuele G assi. Licensed unde he C ea i e Commons A ibu ion -
Noncomme cial 4.0 Licence (h p://c ea i ecommons.o g/licenses/by-nc/4.0/.
A ailable a h p:// o ea.o g.
Re iew o Economic Analysis 17 (2025) 1-35
2
www.Ro EA.o g
ma ke ou comes o indi iduals holding a Ph.D. Mo eo e , he li e a u e on g adua e mig a ion
has so a mainly ocused on pe manen mig a ion expe iences and has la gely o e looked
equally impo an episodes o empo a y mig a ion. This pape is in ended o s udy hese wo
la gely unexplo ed dimensions by in es iga ing he ole o empo a y in e na ional mig a ion
in pos -educa ion pe iods o indi idual wage o wo coho s o I alian Ph.D.s.
Ou ocus on doc o a e ecipien s is mo i a ed on se e al g ounds. Ph.D.s a e among he
mos quali ied wo ke s in he economy, as hey a e endowed wi h high le els o human capi al.
They con ibu e o he dissemina ion o knowledge, a e o en employed in inno a i e sec o s
and play a c ucial ole o economic de elopmen (G ogge & Hanson 2015). Fu he mo e,
Ph.D.s a e ained in academic en i onmen s opened o in e na ional exchanges, and his makes
doc o a e holde s a peculia popula ion o sc u inize in e ms o pos -educa ion mobili y
choices. Mo eo e , he labo ma ke o doc o a e holde s has aced bo h quan i a i e and
quali a i e changes. While in he pas decades he wo ldwide de elopmen o highe educa ion
sys ems has yielded a g owing numbe o s uden s ecei ing a doc o a e deg ee,
1
oppo uni ies
o academic employmen ha e g own a slowe a es, wi h an inc easing numbe o Ph.D.s
employed in he p i a e sec o (Ga cia-Que edo e al. 2011; Gae a e al. 2022). Then, o he
ex en ha lea ning and employmen oppo uni ies a e geog aphically dis ibu ed a a na ional
and in e na ional le el, ea ly mobili y pa e ns may acili a e be e job ma ches in he labo
ma ke o Ph.D.s. Occupa ional misma ch has been p o ed o ha e de imen al consequences
in e ms o bo h ea nings (Gae a e al. 2023; Cul e a e al. 2023) and job sa is ac ion le els
(Al ano e al. 2021) o Ph.D. holde s, especially in ields o s udies (such as humani ies and
social science) o which he oppo uni ies o be employed ou side o academia a e less e iden
(Hancock 2023). On he o he hand, bo h in e nal (A onica e al. 2023) and in e na ional (Ghosh
& G assi 2020) mig a ions could educe job misma ch, imp o e he job sa is ac ion and boos
he wage p emia o PhD g adua es. A he indi idual le el, mobili y allows esea che s o access
be e -equipped ins i u ions, collabo a e wi h leading schola s, and engage in in e na ional
ne wo ks (Passe e a e al., 2018), he eby enhancing hei academic epu a ion, esea ch
p oduc i i y and hei employabili y bo h in he public and p i a e sec o s. Geog aphic mobili y
is a pi o al ma e also o cha ac e ize he in e play be ween human capi al de elopmen and
labo ma ke dynamics o Ph.D. holde s. Fo ins ance, F anzoni e al. (2012) epo ha ou
ou o en emig a ed scien is s om I aly Spain, F ance, Ge many and Swi ze land a e willing
o e u n o hei home coun y condi ional on alid job oppo uni ies. These kinds o mobili y
a e ce ainly ele an o he I alian labou ma ke whe e he e is a lack o demand o PhDs in
he p i a e sec o and employmen dispa i ies be ween No h and Sou h a e pa icula ly ma ked
1
Fo ins ance, Au iol e al. (2013) documen a 38% g ow h in he numbe o Ph.D.s g adua ed om
uni e si ies in OECD coun ies o e he pe iod 2000-2009. Fo he I alian case, he Na ional Ins i u e
o S a is ics (ISTAT) has documen ed ha in 2000 a ound 4000 s uden s we e g an ed a Ph.D. deg ee,
while he numbe had isen o o e 12000 a e only 8 yea s.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
3
www.Ro EA.o g
(Pa en i e al. 2022). Howe e , his pape is speci ically conce ned wi h episodes o e u n
mig a ion. Resea ch highligh s how Ph.D. holde s e u ning o hei home coun ies o en b ing
wi h hem ad anced skills and in e na ional ne wo ks, con ibu ing o local academic and
indus ial ecosys ems (Bø ing e al., 2015). Howe e , his bene i is condi ional on he leng h
and quali y o hei o eign expe iences, making i impe a i e o analyze mobili y's impac on
bo h wages and employmen ou comes in di e en con ex s.
Despi e he well known measu emen di icul ies o assess he eal magni ude o e u nee
lows among mig an s (Dus mann & Weiss 2007; Amb osini e al. 2015), he li e a u e has
ecognized also he impo ance o e u n mig a ion as a possible channel h ough which sou ce
coun ies may bene i om in e na ional mig a ion.
2
While ab oad, indi iduals may enhance
hei human capi al endowmen (Dus mann e al. 2011), accumula e inancial asse s and
assimila e aluable cul u al and social no ms (Be oli & Ma che a 2015). Then, upon e u n,
hey migh enjoy signi ican wage p emium (Reinhold & Thom 2013; de Coulon & Pi acha
2005; Co e al. 2000) o success ully unde ake en ep eneu ial ac i i ies (McCo mick &
Wahba 2001; Dus mann & Ki chkamp 2002; Ma che a 2012). In he case o doc o a e holde s,
expe ience ab oad can also be pa o an in es men plan aimed a inc easing he scien i ic
capi al s ock, epu a ion o in ol emen in scien i ic ne wo ks, all o which may acili a e hei
ca ee in bo h p i a e and public ins i u ions in he domes ic labo ma ke .
All o his sugges s ha a c ucial ac o , la gely neglec ed in p e ious esea ch, is he amoun
o ime indi iduals choose o spend ab oad. Empi ical models implemen ed in he mig a ion
li e a u e o en dicho omize he choice o s ay o come back, bu do no conside he leng h o
ime spen in o he coun ies. Ne e heless, he heo y sugges s ha obse ed ou comes in he
labo ma ke a e likely o e lec human capi al in es men s, which a e ob iously a ime-
consuming ac i i y. I ollows ha ewa ds o o eign expe iences may depend also on how
long indi iduals decide o emain in a o eign coun y. To he bes o ou knowledge, his s udy
is he i s a emp ha sheds ligh on he ela ionship be ween wages o Ph.D. holde s and he
leng h o empo a y mig a ion.
To sum up, his pape con ibu es o he exis ing li e a u e in se e al ways. Fi s , we
documen he size and composi ion o he e u n mig a ion low o he popula ion o I alian
Ph.D.s. Second, we show ha ailing o accoun o he spell o expe ience ab oad yields
misleading es ima es o he impac o mobili y on indi idual wages, as he ue e ec s ongly
depends on he ime spen ab oad. Thi d, his s udy add esses he endogenei y o bo h he
e u nee s a us and he du a ion o he s ay ab oad and concludes ha each choice needs o be
conside ed as endogenous by p ac i ione s o cap u e unobse ed selec i i y among e u nees.
2
O he possible bene i s, which a e no discussed in his wo k, a e ela ed o emi ances. See Docquie
and Rapopo (2006), Ha on (2014) and sugges ed e e ences o u he de ails on he opic.
Re iew o Economic Analysis 17 (2025) 1-35
4
www.Ro EA.o g
In de ail, his pape is he i s documen ing posi i e selec ion ela ed o he choice o mig a ing
and nega i e selec ion ela ed o he leng h o s ay ab oad.
The implica ion o ou indings should be e alua ed in ligh o he ongoing academic and
policy deba e ocused on he in e na ional compe i ion o alen s (Docquie & Machado 2015).
As skilled labo mig a ion p oduces conside able knowledge lows among coun ies, wi h
ob ious epe cussions on agg ega e p oduc i i y, inno a ion and g ow h (Pe i e al. 2015),
many coun ies ha e adop ed quali y-selec i e immig a ion policies, such as ax bene i s and
simpli ied immig a ion measu es,
3
aimed a a ac ing and e aining alen s on a global scale
(Beine e al. 2008). I he ne g ow h o he domes ic high-skilled popula ion is slow, nega i e
epe cussions can be expec ed o i ms and coun ies alike, as i ms’ abili y o inno a e and
succeed s ongly depends on he quali y o he a ailable wo k o ce. Ou es ima es sugges ha
policies aimed a inc easing he incen i es o e u n o he home coun y may ac ually educe
he bene i s o hose who had o he wise planned longe pe iods in a hos coun y. Such policies
migh lead indi iduals o spend less ime han hey would op imally do, educing hei
in es men in human capi al and, in u n, hei gains om mig a ion.
We also wish o acknowledge up on some ca ea s ha should be bo ne in mind when
in e p e ing he esul s. High-achie e s in he educa ion sys em a e only pa o he high-skilled
mig an popula ion. Thus, he esul s canno be sla ishly ex ended o he whole ca ego y o
skilled e u nees. Ye , we belie e Ph.D. holde s dese e a speci ic esea ch ocus due o hei
impo ance in he c ea ion and di usion o scien i ic knowledge on a global scale. This s udy
has also some da a limi a ions. Fi s , da a a e e ospec i e and do no p o ide all he de ails
ela ed o he pe iod o ime spen ab oad, such as ea nings and employmen cha ac e is ics.
Second, e en i he da a epo s he exac leng h o s ay in mon hs, indi iduals we e no asked
abou he s a ing da e o he pe iod. Thi d, we a e awa e ha ou analysis ocuses exclusi ely
on e u n mig a ions ha may occu wi hin he sho ime pe iod be ween he comple ion o he
PhD and he momen o he in e iew. Thus, i is impo an o ake in o accoun ha he analyzed
popula ion o e u nees excludes hose indi iduals ha may e en ually e u n o I aly in he
mo e dis an u u e. Finally, i is wo h no icing he coho s analyzed in his s udy c ossed
signi ican mac oeconomic and ins i u ional challenges. The a e ma h o he 2008 inancial
c isis and subsequen so e eign deb c isis subs an ially educed academic employmen
oppo uni ies and eshaped labo ma ke condi ions in I aly. Simul aneously, academic e o ms,
such as he Gelmini e o m, inc eased p eca iousness in en y-le el academic oles (Passa e a
e al., 2018). These con ex ual elemen s likely in luenced mig a ion decisions and employmen
ou comes, u he unde sco ing he alue o examining sho - e m e u n mig a ion wi hin his
u bulen pe iod.
3
The EU Council Di ec i e 2009/50/EC in oduced he Blue Ca d, a simpli ied wo k-pe mi allowing
high-skilled non-EU ci izens o wo k in EU coun ies.

DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
5
www.Ro EA.o g
The es o he pape is o ganized as ollows. Sec ion 2 is conce ned wi h he posi ioning o
he pape in he ele an li e a u e. Sec ion 3 desc ibes he empi ical model. Sec ion 4 con ains
a b ie desc ip ion o he da a along wi h summa y s a is ics. Resul s and obus ness checks a e
p esen ed and discussed, espec i ely, in sec ion 5 and 6. Sec ion 7 concludes.
2 Backg ound
The scien i ic deba e o mig a ion is adi ionally amed in o he human capi al heo y, as i s ly
discussed in Sjaas ad (1962) and Becke (1962), and mos heo e ical models o mig a ion (and
hei empi ical implemen a ions) ha e been p edominan ly de eloped o explain pe manen
loca ion changes. Howe e , mig a ion can also be a e e sible decision ha b ings indi iduals
back o hei coun y o o igin.
4
In his espec , se e al au ho s ha e hus a con ibu ed o
p o ide he heo e ical unde pinnings o he decision-making o e u nees. In ea ly
con ibu ions o Hill (1987) and Djajić (1988), e u n mig a ion esul s om he balancing o
he ade-o be ween highe wages enjoyed while ab oad and o gone u ili y ela ed o highe
p e e ences o home consump ion. Thus, e en i li e ime income would be highe by emaining
in a gues coun y, indi iduals maximize hei li e ime u ili y by spending some ime in he hos
coun y and hen e u ning home. Ne e heless, e u n mig a ion is obse ed also in he absence
o a e e sal o he ela i e wages o he sending and ecei ing coun ies (S a k e al. 1997),
and i s a ional may well depend on o he ac o s, such as a ge -sa ings mo i es (Be ninghaus
& Sei e -Vog 1989), highe pu chasing powe in he home coun y (S a k e al. 1997), c edi
cons ain s (Mesna d 2004) and un ul illed expec a ions abou oppo uni ies in he hos coun y
(Bo jas & B a sbe g 1996).
As he p esen esea ch is ocused on doc o a e holde s, human capi al conside a ions a e
shown o be o pa icula in e es . I e u ns o human capi al in es men s made in he hos
coun y a e highe in he home coun y, some indi iduals may decide o eloca e o hei coun y
o o igin (Dus mann 1997). Mo eo e , knowledge and skills could be acqui ed ab oad mo e
e icien ly o as e han in he home coun y (Dus mann e al. 2011). In e es ingly, since human
capi al g ows o e ime, he leng h o ime spen ab oad is po en ially a key ac o in explaining
he impac o e u n mig a ion on labo ma ke ou comes a home. Thus, in con as o Bo jas
and B a sbe g (1996) who assume ha lea ning ab oad aises local ea ning by a ixed p opo ion
i espec i ely o he du a ion o he s ay ab oad, o he schola s ha e ex ended he e e ence
amewo k o he case o human capi al g ow h o e ime (Dus mann & Weiss 2007; San os &
Pos el-Vinay 2003; de Coulon & Pi acha 2005). This iew has been explo ed bo h heo e ically
and, o a lesse ex en , empi ically. In Dus mann (2002), mig a ion du a ion and a e -mig a ion
ac i i ies a e op imally chosen o maximize li e ime u ili y. The model p edic s ha he planned
4
Resea ch has also in es iga ed o he o ms o mig a ion. A o mal axonomy can be ound in Dus mann
and Weiss (2007).
Re iew o Economic Analysis 17 (2025) 1-35
6
www.Ro EA.o g
du a ion esponds o changes in hos and home coun y wages and ha he beha io al esponse
a ies depending on he pos -mig a ion ac i i y. May and Pe i (2009) de elop a model o
op imal human capi al in es men s, mig a ion and e u n mig a ion as unc ions o pe sonal
abili ies o analyze he e ec s o mig a ion policies on human capi al and wages in sending
coun ies. No ably, he model s esses he impo ance o human capi al accumula ion and i s
economic e u ns in shaping dis inc mig a ion pa e ns. Ye , longe spells o empo a y
mig a ion may also be associa ed wi h wo se ou comes in he domes ic labo ma ke . To he
ex en ha indi iduals lose aluable social capi al while ab oad (Ma che a 2012), e u n
mig a ion would de e io a e he pe o mance o e u nees. Also, i skills acqui ed ab oad ha e
limi ed ans e abili y ac oss bo de s (Chiswick & Mille 2009), indi iduals may accep unde -
quali ied jobs in he home coun y upon e u n. In a ecen s udy on g adua e (in e nal)
mig a ion, Di Cin io and G assi (2013) ind ha indi iduals choosing o mo e back home o
wo k (ins ead o emaining in he a ea o s udy) expe ience a wage loss. The au ho s poin ou
ha while mo e s ob ain a en because o mobili y, back mo e s choose o gi e up his en by
e u ning o he a ea o o igin. Gibson and McKenzie (2012) collec da a on i e coun ies
expe iencing high a es o high-skilled emig a ion and ind ha , hough mig a ion spu s human
capi al accumula ion, he e a e nega i e poin es ima es in wo ou o h ee cases and no
signi ican income gains o he emaining coun ies.
Empi ical e idence on he e ec s o e u n mig a ion o in e na ionally mobile indi iduals
is g owing apidly, bu i is la gely ocused on he expe ience o less de eloped coun ies.
Reinhold and Thom (2013) use da a on mig an s who e u n o Mexico a e spending some
ime in he Uni ed S a es and es ima e an ea nings inc ease o app oxima ely 2-3% o e e y
yea spen ab oad. Co e al. (2000) ocus on Hunga ian da a and ind a 40% ea ning p emium
o emale e u nees while no e ec is ound o men. Di e en ly, Ba e and O’Connell (2010)
epo a 7% wage p emium o bo h gende s inc easing wi h educa ion a ainmen s.
5
Amb osini
e al. (2015) analyzes Romanian da a and ind ha no only e u nees enjoy highe wages bu
also ha he p emium is inc easing wi h hei le el o skills.
The p esen s udy also s ands wi hin he on-going deba e conce ning he mobili y o college
and uni e si y g adua es. Bo h doc o a e holde s and g adua es lea e he educa ion sys em wi h
a highe han a e age human capi al endowmen and eel he u gency o making his in es men
paying o also h ough addi ional in es men s in mobili y pa e ns (Venho s e al. 2011). In
his espec , inc easing e idence sugges ing posi i e economic e u ns o mig a ion has been
documen ed mos ly o in e nal mig a ion pa e ns o college and uni e si y g adua es,
6
while
ew s udies ha e ied o shed ligh on he in e na ional mobili y o Ph.D. ecipien s and hei
5
A summa y o u he empi ical e idence o de eloping coun ies can be ound in May and Pe i
(2009).
6
See, among o he s, Ab eu e al. (2015) and Di Cin io and G assi (2013).
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
7
www.Ro EA.o g
pe o mance in he labo ma ke . Indeed, while he esea ch pe o mance o Ph.D.s ha e
a ac ed he a en ion o many schola s (A hey e al. 2007; G o e & Wu 2007), only ecen ly
he li e a u e has s a ed examining he mobili y o doc o a e holde s. In pa icula , i has been
shown how he p opensi y o mig a e ab oad esponds o age and gende di e ences (Di Cin io
& G assi 2016), he ype o jobs Ph.D.s a e willing o accep (Da is & Pa e son 2000), he
p esence o bo h ameni y ac o s (Go lieb & Joseph 2006) and wo ld-leading esea ch
o ganiza ions (G ogge & Hanson 2015). We ex end his discussion by deepening he
unde s anding o he labo ma ke ou comes associa ed wi h he in e na ional empo a y
mig a ion o doc o al holde s.
3 Resea ch me hods
To examine he implica ions o in e na ional empo a y mig a ion on wages, we se up an
empi ical amewo k in which (log) mon hly wages o empo a y mig an s 𝑦𝑖1 and non-
mig an s 𝑦𝑖0 a e ela ed o a se o explana o y a iables associa ed o bo h pe sonal and job
cha ac e is ics 𝑥𝑖𝑗, so ha :
𝑦𝑖1 =𝜇1+𝛽1
′𝑥𝑖1+𝛾𝑡𝑒𝑚𝑝𝑖+𝑢𝑖1
𝑦𝑖0 =𝜇0+𝛽0
′𝑥𝑖0+𝑢𝑖0 (1)
whe e 𝜇 is a scala , 𝛽 and 𝛾 a e pa ame e s o be es ima ed, 𝑡𝑒𝑚𝑝 is he numbe o mon hs he
𝑖−𝑡ℎ indi idual has spen ab oad a e he Ph.D. and 𝑢 is he e o e m. Cohe en ly wi h he
li e a u e on p og am e alua ion o iginally de eloped by Rubin (1974) and Holland (1986), he
p e ious se up can be in e p e ed in e ms o wo mu ually exclusi e ou comes associa ed o
p og am pa icipa ion. He e, an e alua ion p oblem a ises, o he esea che can only obse e
he wage o he ealized ou come, so ha he obse ed wage can be ead as ollows:
𝑦𝑖=𝑦𝑖0+𝑡(𝑦𝑖1 −𝑦𝑖0), (2)
whe e is he bina y indica o o empo a y mig a ion expe ience. By plugging equa ions (1)
in equa ion (2), he ollowing wage equa ion is immedia ely de i ed:
𝑦𝑖=𝜇0+𝛽0
′𝑥𝑖0+𝑢𝑖0+𝑡(𝜇1−𝜇0)+𝑡𝛾𝑡𝑒𝑚𝑝𝑖+𝑡(𝛽1
′𝑥𝑖1−𝛽0
′𝑥𝑖0)+𝑡(𝑢𝑖1−𝑢𝑖0). (3)
As a s a ing poin o discussion, we assume a homogenous ea men esponse (𝛽1=𝛽0=𝛽)
and a homogenous e a ic componen (𝑢1=𝑢0=𝑢). Fu he mo e, we ini ially assume ha all
ha ma e s o wages is whe he any expe ience ab oad has been done. Thus, when we neglec
he e ec o he leng h o s ay ab oad, equa ion (3) collapses o:
𝑦𝑖=𝜇0+𝛽′𝑥𝑖+𝑡(𝜇1−𝜇0)+𝑢𝑖. (4)
Re iew o Economic Analysis 17 (2025) 1-35
8
www.Ro EA.o g
E en in his simpli ied se ing, in e nal alidi y is h ea ened by he ac ha empo a y mig an s
a e sel -selec ed a he han a andom g oup o indi iduals. Thus, in he absence o an
exogenous sou ce o a ia ion o he incen i e o mig a e, he bina y indica o o mobili y is
suspec ed o endogenei y, i.e. indi iduals chose o mig a e only o he ex en ha he expec ed
bene i s associa ed wi h his choice is no less han he cos s o mo ing (Sjaas ad 1962; Bo jas
1987).
7
To ackle his p oblem, we implemen an endogenous bina y ea men e sion o
Heckman’s (1976; 1979) wo-s ep model,
8
whe e he obse ed mig a ion choice depends om
an unobse able la en a iable 𝑡∗ ha is assumed o be linea ly ela ed o a se o co a ia es,
𝑧𝑖, so ha : 𝑡𝑖∗=𝛼′𝑧𝑖+𝑒𝑖 (5)
and
𝑡𝑖={1  i 𝑖∗>0
0  o he wise , (6)
whe e equa ion (6) is he selec ion ule. We es ima e model (4), (5) and (6) wi hin a wo-s ep
amewo k de i ed in Maddala (1986) in which indi iduals di e in unobse able ways ha
con ibu e o de e mine bo h selec ion in o mig a ion and wages. Thus, we con ol o selec ion
in a mince ian- ype eg ession by es ima ing a selec ion ule ha p edic s whe he a Ph.D.
g adua e mig a es ab oad.
The wo-s ep es ima o elies on he assump ion ha he unobse ed he e ogenei y is
cap u ed by he co ela ion s uc u e be ween 𝑢𝑖 and 𝑒𝑖, i.e. he co ela ion be ween
unobse ables a ec ing and unobse ables a ec ing y. In pa icula , 𝑢𝑖 and 𝑒𝑖 a e bi a ia e
no mal dis ibu ed wi h co ela ion coe icien 𝜌. Es ima ion p oceeds as ollows. Fi s , we
ob ain p obi es ima es om:
𝑃𝑟(𝑡𝑖=1|𝑧𝑖)=𝛷(𝛼′𝑧𝑖) (7)
and, hen, we eco e he haza d h o each obse a ion acco ding o he o mula:
7
A di e en possible sho coming is ela ed o he ac ha indi iduals sel -selec in o employmen . As
commonly poin ed ou in many empi ical s udies, i wages a e only obse ed o indi iduals ha
ac ually ha e a job, hen sample selec ion bias a ises. Howe e , in ou s udy, we belie e ha his sou ce
o bias should play a li le ole due o he small numbe o unemployed indi iduals. P ecise s a is ics
a e gi en in sec ion 4.
8
An al e na i e app oach would be o model he mig a ion decision in a mul inomial con ex and co ec
o selec i i y as in Lee (1983) and Dahl (2002). Un o una ely, he numbe o obse a ions o each
possible des ina ion is oo low gi en he small ac ion o mo e s in ou da a. Thus, we decided o apply
a model ha is s ill able o accoun o selec i i y.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
15
www.Ro EA.o g
As a as he wage equa ion is conce ned, he se o con ol a iables is iche , as we add,
compa ed o he a o emen ioned equa ions, also many job- ela ed a iables. In pa icula , we
con ol o yea s o job expe ience (bo h wi h a linea and quad a ic e m) and we include
a iables ha indica e i indi iduals ca y ou R&D and/o eaching ac i i ies in hei job, i
hey hold a pos -doc posi ion, i hey had access he job h ough a public compe i ion, in which
economic sec o (indus y, se ices and ag icul u e) hey a e employed. In addi ion, since he
su ey deli e s in o ma ion ela ed o he scien i ic p oduc i i y in e ms o published a icles
and pa en s, we a e able o pa ially accoun o abili y in conduc ing esea ch. Mo eo e , we
p oxy de e mina ion and mo i a ion using a dummy equal o one whene e s uden s g adua ed
on ime and we also conside whe he hey we e al eady wo king a he ime o hei deg ee.
Las ly, we add ixed e ec s o 53 p o inces in which he deg ee was awa ded, 110 wo k
p o inces and 22 indus y in e cep s o con ol o di e ences in economic ac i i ies.
Gi en he la ge numbe o explana o y a iables, we check he deg ee o mul i- collinea i y
wi h he a iance in la ion ac o (VIF). In de ail, he squa e oo o he VIF indica es how much
la ge he s anda d e o is, compa ed o wha i would ha e been i ha a iable was
unco ela ed wi h he o he independen a iables. In ou analysis, all o he VIFs a e lowe
han 10 (many a e lowe han 2) and he mean VIF is 1.64. Thus, since all o he VIFs a e
ela i ely low, we can be con iden ha mul icollinea i y is no an issue in ou analysis.
5 Es ima ion esul s
OLS es ima es o he ela ionship be ween log mon hly wages and empo a y mig a ion a e
p esen ed in he i s column o able 3. The main eg esso s o in e es a e he dummy o he
mig a ion expe ience and he numbe o mon hs spen ab oad. Bo h a e ea ed as exogenous
a iables. Being a e u n mig an is associa ed wi h lowe wages, bu he e ec is mi iga ed by
he posi i e coe icien on he du a ion. In pa icula , only hose who s ay ab oad o mo e han
6.8 mon hs end o enjoy a wage gain. While we only epo he es ima es wi h he ull se o
con ol a iables, in mo e pa simonious speci ica ions o he empi ical model
17
wo in e es ing
pa e ns eme ge. Fi s , as we p og essi ely add mo e con ols, he magni ude o he coe icien
on he indica o o empo a y mig a ion ends o all. Since some o he con ols ha we
inc emen ally add a e po en ially co ela ed o unobse able indi idual ai s, he endogenei y
bias in OLS es ima es is likely o be se e e and he es ima ed impac o empo a y mig a ion is
con ounded wi h he e ec s o unobse ed cha ac e is ics. Second, also he coe icien o he
mig a ion du a ion is declining ac oss di e en model speci ica ions, sugges ing ha he e is a
speci ic sou ce o selec ion bias ha migh no be ully cap u ed by sol ing he endogenei y
17
The able is a ailable upon eques .

Re iew o Economic Analysis 17 (2025) 1-35
16
www.Ro EA.o g
p oblem o he dicho omous a iable alone.
18
We hus expec di e ences be ween he es ima es
o he endogenous dummy a iable model and he es ima es o he model in which we
endogenize also he mig a ion du a ion.
Table 3: Re u ns o empo a y mig a ion
OLS
Endogenous dummy
a iable model
Endogenous dummy
a iable model
Endogenous du a ion
(1)
(2)
(3)
(4)
Re u n-mig an s
-0.029**
(0.013)
-0.221***
(0.061)
-0.312***
(0.062)
-0.397***
(0.085)
Du a ion (# o mon hs)
0.004***
(0.001)
0.005***
(0.001)
0.016***
(0.004)
Female
-0.098***
(0.007)
-0.108***
(0.007)
-0.109***
(0.007)
-0.105***
(0.007)
Age a PhD (base: < 30)
Age a PhD (30-34 yea s)
-0.026***
(0.008)
-0.042***
(0.009)
-0.043***
(0.009)
-0.041***
(0.010)
Age a PhD (>34 yea s)
-0.059***
(0.014)
-0.090***
(0.016)
-0.093***
(0.016)
-0.086***
(0.016)
Teaching du ing PhD
0.005
(0.006)
0.007
(0.006)
0.007
(0.006)
0.008
(0.007)
No he n Uni e si y
0.133***
(0.038)
0.117***
(0.038)
0.119***
(0.038)
0.131***
(0.040)
PhD in ime
0.027***
(0.010)
0.027**
(0.011)
0.027**
(0.010)
0.026**
(0.011)
Scien i ic p oduc i i y
Jou nal a icles
0.002***
(0.000)
0.002***
(0.000)
0.002***
(0.000)
0.002***
(0.000)
Pa en s
0.020***
(0.005)
0.020***
(0.005)
0.020***
(0.005)
0.020***
(0.005)
Job cha ac e is ics
PhD equi ed o job
0.032***
(0.009)
0.034***
(0.009)
0.032***
(0.009)
0.034***
(0.009)
Job access (public
compe i ion)
0.136***
(0.010)
0.133***
(0.010)
0.135***
(0.010)
0.134***
(0.011)
Pos -doc con ac
0.025***
(0.009)
0.026***
(0.009)
0.024***
(0.009)
0.025***
(0.010)
Job wi h eaching
-0.01
(0.008)
-0.012
(0.008)
-0.01
(0.008)
-0.011
(0.008)
Job wi h RD
0.093***
(0.009)
0.092***
(0.010)
0.092***
(0.009)
0.092***
(0.010)
Al eady wo king be o e PhD
0.046***
(0.012)
0.050***
(0.012)
0.047***
(0.012)
0.051***
(0.013)
Wo k expe ience
0.030***
(0.009)
0.029***
(0.009)
0.031***
(0.009)
0.029***
(0.009)
Wo k expe ience (squa ed)
-0.002**
(0.001)
-0.002**
(0.001)
-0.002**
(0.001)
-0.002**
(0.001)
Job sec o (base: Ag icul u e)
Indus y
0.258***
(0.037)
0.253***
(0.036)
0.256***
(0.037)
0.251***
(0.038)
Se ices
-0.138***
(0.036)
-0.140***
(0.036)
-0.138***
(0.036)
-0.138***
(0.037)
cons an
7.115***
(0.103)
7.186***
(0.105)
7.177***
(0.108)
7.172***
(0.113)
haza d
0.134***
(0.0329)
0.153***
(0.033)
0.228***
(0.046)
18
Fo ins ance, di e ences in p e-mig a ion skill le els can be mi o ed in di e en mig a ion du a ion.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
17
www.Ro EA.o g
Table 3 con inued
OLS
Endogenous dummy
a iable model
Endogenous dummy
a iable model
Endogenous du a ion
FE: PhD NUTS-3 code
Yes
Yes
Yes
Yes
FE: wo k NUTS-3 code
Yes
Yes
Yes
Yes
bachelo g adua ion yea
Yes
Yes
Yes
Yes
Indus y FE
Yes
Yes
Yes
Yes
PhD coho
Yes
Yes
Yes
Yes
bachelo g ade
Yes
Yes
Yes
Yes
mo he educa ion
Yes
Yes
Yes
Yes
high school
Yes
Yes
Yes
Yes
R_sq
0.324
0.323
0.326
0.324
Obse a ions
8981
8981
8981
8981
No es: The dependen a iable is he log mon hly wage. ***, **, * deno e, espec i ely,
signi icance le els a 1, 5 and 10 %. Co ec ed s anda d e o s a e in pa en hesis.
Reg ession esul s also e eal ha o he obse ed cha ac e is ics a e impo an p edic o s o
wages in acco dance wi h p e ious li e a u e. In de ail, we ind ha emales ea n a ound 10%
less han males
19
. Wages a e dec easing in age a Ph.D. and a e highe o hose who ecei ed
hei doc o a e deg ee om no he n uni e si ies. As expec ed, we ind diminishing e u ns o
wo k expe ience, wi h a posi i e coe icien on he linea e m and a nega i e coe icien on he
quad a ic e m. P oxies o he job-ma ch quali y indica e highe wages o hose ca ying ou
R&D ac i i ies in hei job and o hose employed in jobs o which he Ph.D. was equi ed.
Ha ing comple ed he Ph.D. in ime has also a posi i e impac on wages, as employe s migh
in e p e i as a signal o e iciency and commi men .
To explo e in de ail issues ela ed o selec i i y o mig an s, we s a wi h he es ima ion o
he endogenous dummy a iable model desc ibed by equa ion (9) whe e, ini ially, we neglec
he e ec o he leng h o s ay ab oad. In pa icula , in he i s s age, we es ima e a selec ion
ule (i.e. he mig a ion equa ion) ha p edic s he p obabili y o be a e u n mig an , and hen
we p oceed o he analysis o he ea ning equa ion augmen ed by he haza d a e. Resul s o he
i s s age a e p esen ed in able A1, column 1, in he Appendix. In line wi h p e ious esea ch,
he p opensi y o mig a e is signi ican ly lowe o emales. This is usually unde s ood as
e idence o mo e binding amily ies o emales in con as o men being mo e commi ed o
19
I he dummy swi ches om 0 o 1, he pe cen age impac on Y is equal o 100[exp(𝛽󰆹) - 1].
Re iew o Economic Analysis 17 (2025) 1-35
18
www.Ro EA.o g
ca ee conce ns.
20
As expec ed, age a Ph.D. is a good p edic o o he mobili y choice.
Compa ed o he baseline ca ego y (being younge han 30), all he coe icien s a e s a is ically
signi ican , ha e a nega i e sign and hei magni ude is inc easing in age. In de ail, holding all
o he con ol a iables a hei means, he p obabili y o empo a y mig a ion is 4% lowe o
indi iduals aged 30-34 and 9% lowe o indi iduals aged mo e han 34. Ha ing changed ci y
o a end he Ph.D. inc eases he p obabili y o subsequen mobili y, while he p esence o
child en in he amily lowe he odds o mo e ab oad. Al hough no explici ly epo ed, he se
o dummies cap u ing he impac o he ields o s udy u n ou o be a ela i ely poo p edic o
in he mig a ion equa ion, wi h he excep ion o physics and indus ial enginee ing. Finally,
no e ha o accoun o di e ences ela ed o Ph.D. g an ing ins i u ions, p o ince ixed-e ec s
we e included in he es ima ion.
We now examine he esul s o he ea ning equa ion, in which he log wage is eg essed
agains he bina y indica o o in e na ional empo a y mobili y, indi idual cha ac e is ics, job
cha ac e is ics, amily backg ound, academic backg ound and a ull se o o igin and des ina ion
ixed e ec s. As discussed in Sec ion 3, we deal wi h he endogenei y o empo a y mig a ion
h ough an endogenous dummy a iable model. Thus, he ea ning equa ion has been augmen ed
wi h he haza d a e desc ibed in equa ion (8) and compu ed a e he i s s age p obi
eg ession desc ibed in equa ion (7).
By neglec ing he leng h o expe ience ab oad, esul s in able 3, column 2, con i m he
nega i e ela ionship be ween log wages and empo a y mig a ion. Howe e , be o e
p oceeding, we s ess he ac ha , despi e he es ima ion p o ocol ies o accoun o selec ion
in o mig a ion, iden i ica ion could s ill be h ea ened by he possible esidual co ela ion
be ween he leng h o s ay ab oad and he e o e m in he ea ning equa ion. The e o e, since
he esul s migh s ill be biased and indi iduals a e likely o di e in he leng h o s ay ab oad,
we y o cap u e also he e ec o he mig a ion du a ion on wages.
In able 3, column 3, he coe icien o he main a iable o in e es has s ill a nega i e sign
and is s a is ically signi ican . Quan i a i ely, empo a y mig a ion is associa ed wi h a
educ ion in log wages o 0.221, which inc eases o 0.312 when we also include he (exogenous)
numbe o mon hs spen ab oad. By compa ing bo h eg essions (columns 2 and 3 in able 3),
we no iced small e ec s on he es ima ed coe icien s o he o he eg esso s and a posi i e
and highly signi ican e ec on he selec ion-co ec ion e m. Thus, we canno ejec he null
hypo hesis ha he e o e ms o he mig a ion and employmen equa ion a e co ela ed.
Mo eo e , he esul poin s o posi i e selec ion o mig an s, sugges ing ha empo a y
mig an s a e a sel -selec ed g oup whose unobse ables cha ac e is ics a e simul aneously
20
In a s udy on UK uni e si y g adua es, Faggian e al. (2007) ha e documen ed a case in which
emales a e mo e mig a o y han hei male coun e pa s.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
19
www.Ro EA.o g
associa ed wi h bo h highe wages and a highe p opensi y o mig a e. Al hough he deba e on
mig an s selec i i y is s ill con o e sial, se e al empi ical s udies ha e ound e idence
consis en wi h he posi i e selec ion o mo e gene al g oups o mig an s (Belo & Ha on 2012;
G ogge & Hanson 2011; Chiquia & Hanson 2005). Ou esul co obo a es such li e a u e by
i s documen ing posi i e selec ion o Ph.D. mig an s. Mo eo e , his esul is aligned wi h
he hypo hesis es ed in Gibson and McKenzie (2012) ha he bes o he b igh es conside he
op ion o u he in es men s in human capi al soon a e hey inish s udying.
The las column o able 3 epo s he es ima es o he model in which we also accoun o
he endogenei y o he leng h o ime spen ab oad. As p e iously explained, we ackle his
p oblem by i s unning a p obi model (as in he i s s age epo ed in able A1, column 1, in
he Appendix) om which we eco e he in e se Mill’s a io (which co esponds o he i s
b anch in equa ion 8). Then, we use a nega i e binomial model o eg ess he leng h o ime
spen ab oad on he in e se Mill’s a io and o he co a ia es o ob ain he selec ion-co ec ed
es ima es o he leng h o s ay ab oad. Finally, we es ima e equa ion (10) whe e he expec ed
impac o empo a y mig a ion on wages depends also on he du a ion o s ay and he es ima es
con ol o bo h sou ces o endogenei y.
Table A1 (column 2) in he appendix shows he esul s o he second s age. The in e se
Mill’s a io u ns ou o be highly signi ican , indica ing ha e u nees a e a sel -selec ed g oup
among he pool o Ph.D.s. Mo eo e , we ind a nega i e coe icien , meaning ha unobse able
ai s ha lowe indi iduals’ ea ning capaci y a e also ela ed o p olonged leng hs o s ay. I
we in e p e he selec ion mechanism as de e mined by abili y ai s, nega i e selec ion implies
ha less able Ph.D.s end o s ay longe ab oad o acqui e mo e skills in he hos coun y and,
hen, enjoy highe wages when back o hei coun y o o igin. A he same ime, he nega i e
selec ion wi h espec o he leng h o ime spen ab oad may signal some di icul ies in adap ing
o o eign con ex s. Thus, indi iduals wi h less abili y o adap o un amilia con ex s would
end o spend sho e pe iods ab oad.
Es ima ed coe icien s on o he con ols sugges ha emales end o emain ab oad longe
han males, p obably o coun e -balance gende disc imina ion in he home coun y. This is in
line wi h indings in Al ano e al. (2021) whe e he au ho s sugges ha a gende gap in hou ly
wages exis s among Ph.D. holde s, wi h sizeable di e ences by sec o o employmen and ield
o specializa ion.
Gi en he selec ion mechanism, du a ion inc eases wi h age and i he e a e child en in he
amily. Di e en ly, pas mobili y o a end he Ph.D. is nega i ely associa ed wi h du a ion, as
well as he eg esso e e ed o eaching ac i i ies du ing he Ph.D. E en i no explici ly
epo ed, while ields o s udy we e poo p edic o s in he bina y mig a ion choice, hey a e
e y ele an in p edic ing he amoun o ime spen ab oad. In pa icula , we ind ha hose
wi h a Ph.D. in ei he physics, chemical sciences, psychology, law, economics and s a is ics o
social sciences end o emain ab oad sho e pe iods han g adua es om di e en disciplines.
Re iew o Economic Analysis 17 (2025) 1-35
20
www.Ro EA.o g
Tu ning he a en ion o he wage equa ion, ou es ima es ( able 3, column 4) do no p oduce
sensible changes in he es ima ed coe icien s o he con ol a iables compa ed wi h ou
p e ious esul s.
21
Howe e , as i can be eadily seen, esul s on he main eg esso s indica e
ha he e is now a speci ic ewa d associa ed o empo a y mig a ion, bu only o hose
indi iduals who spend no less han 24 mon hs in a o eign coun y, as each mon h spen ab oad
yields a ma ginal inc ease in a e age mon hly ea nings o 1.61%. This esul is pa ly in line
wi h p e ious empi ical e idence o a wage p emium o e u n mig an s, as in Ba e and
O’Connell (2010) and Amb osini e al. (2015). A he same ime, we e ine p e ious e idence
by documen ing he wage penal y su e ed by ea ly e u nees. In ou da a, neglec ing he
impo ance o he leng h o ime spen ab oad would deli e he misleading conclusion ha
empo a y mig a ion is o e all a bad in es men o high-skilled indi iduals. Ins ead, he e u ns
a e nega i e only o hose indi iduals who choose o e u n home ea ly, while he emaining
e u nees may achie e hei goals in e ms o skills acquisi ion in he hos coun y and enjoy
highe wages when back o hei coun y o o igin.
To some ex en , ou empi ical esul s can be econciled wi h heo e ical models o
empo a y mig a ion. F om one side, he nega i e wage e ec associa ed wi h sho e pe iods
ab oad may e lec he ou come o he ea ly e u nees who mis akenly chose an in e na ional
mig a ion pa h, as in Bo jas and B a sbe g (1996). On he o he side, he associa ion be ween
p olonged pe iods o expe ience ab oad and highe wages in he o igin coun y is in suppo o
models ha explain e u n mig a ion as a decision ha esponds o whe e human capi al can be
acqui ed mo e e ec i ely, and whe e he e u n o human capi al is highes (Dus mann e al.
2011).
Al e na i ely, p o ac ed leng hs o s ay may be associa ed wi h a p ocess o skill
di e si ica ion and, hus, o an imp o emen o indi iduals’ abili y o adap o di e en ypes
o occupa ions equi ing di e en knowledge. Fo ins ance, in Cha lo e al. (2005), he numbe
o indi idual skills inc eases wi h he du a ion o schooling, yielding be e expec ed
pe o mances in he labo ma ke . Ins ead, hose who go back home soone do no accumula e
a su icien ly di e en ia ed amoun o skills and abili ies, esul ing in poo e pe o mances in
he domes ic labou ma ke .
Ou es ima es also deli e no el policy p esc ip ions. Policies aimed a inc easing he
incen i es o e u n o he home coun y may educe he bene i s o hose who had o he wise
planned longe pe iods in a hos coun y. By inc easing he incen i e o e u n home, such
policies migh lead indi iduals o spend less ime han hey would op imally do, educing hei
in es men s in skills and dec easing indi idual gains om mig a ion. Fo ins ance, in Dus mann
(2011), e u n mig a ion can mi iga e he wel a e loss o he b ain d ain o he ex en ha
21
As he F s a is ic on he join signi icance o he indus y dummies is equal o 37.66, di e ences in
economic ac i i ies a e also pa icula ly ele an . Comple e ables a e a ailable upon eques .

DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
21
www.Ro EA.o g
e u nees ac ually augmen local skills. Thus, decision make s should calib a e public
in e en ions by aking in o accoun he leng h o ime mig an s ha e al eady spen in a hos
coun y.
5 Robus ness and sensi i i y
This sec ion p esen s se e al obus ness and sensi i i y checks o alida e he esul s p esen ed
so a .
5.1 Checking o he e ogeneous e u ns
As ecen ly no ed in he p og am e alua ion li e a u e, i he gains om p og am pa icipa ion
a y acco ding o indi iduals’ cha ac e is ics, es ima es may su e om he so called
he e ogeneous ea men bias (Heckman e al. 2006). To ackle his issue, we allow o a mo e
lexible model by aking in o accoun he possible he e ogenei y in ea men esponse (𝛽1≠
𝛽0) and by elaxing he hypo hesis o limi ed unobse ed he e ogenei y (𝜖1≠𝜖0). In
pa icula , his la e hypo hesis le us sepa a ely es ima e he co ela ions be ween each
ea men s a us and wages. Unobse ed cha ac e is ics may include he se o skills and
abili ies ha con ibu es o an indi idual’s wage and i s p opensi y o empo a ily mo e ab oad,
which may be di e en in he subpopula ions o mig an s and non-mig an s
22
.
The a iabili y in ea men esponse can be cap u ed in a eg ession amewo k wi h he
inclusion o he e m (𝑥−𝑥)𝛾𝑡, which is i sel endogenous. Fo mally, we s ill ely on a wo-
s ep selec ion model whe e equa ion (10) can be e o mula ed as ollows
23
:
𝑦𝑖=𝜇0+𝛽′𝑥𝑖+𝑡𝑖(𝜇1−𝜇0)+𝑡𝑖𝛾1𝑡𝑒𝑚𝑝
𝑖+𝑡𝑖𝛾0
′(𝑥𝑖−𝑥)+
+ρ1𝜎𝑖𝑡ℎ𝑖𝑡+ρ0𝜎𝑖(1−𝑡)ℎ𝑖𝑡+𝑣𝑖, (11)
whe e ρ1 and ρ10 a e he co ela ions be ween each ea men s a us and wages, is he e o
e m and ℎ𝑡 is he haza d a e in equa ion (8). We pos ula e ha indi iduals wi h highe
scien i ic p oduc i i y a e mo e in o med abou esea ch unding and wo k expe iences ab oad
and so hey a e also likely o ha e g ea e knowledge o (po en ial) cos s and bene i s associa ed
wi h he choice o empo a y mig a ion. Hence, hey could ul ima ely be able o ob ain highe
wages once hey e u n home. Mo eo e , younge Ph.D.s may bene i mo e om mobili y
because ha ing ob ained he Ph.D. while younge is o en pe cei ed as a measu e o
e ec i eness and commi men , which in u n can be ewa ded wi h highe wages in he labo
22
Fo ins ance, Al ano e al. (2022) documen he high he e ogenei y e en among highly skilled I alian
mig an s.
23
See Woold idge (2010).
Re iew o Economic Analysis 17 (2025) 1-35
22
www.Ro EA.o g
ma ke . A he same ime, being younge is also associa ed wi h a highe p opensi y o mig a e,
hus we le he ea men indica o in e ac wi h age a Ph.D. Fo comple eness, we also use
gende o cap u e o he dimensions along which he he e ogeneous ea men bias could deploy
i s e ec .
Table 4 p esen s wo se s o esul s which e e , espec i ely, o he endogenous dummy
a iable model (column 1) and he model wi h endogenous du a ion (column 2). Bo h models
ackle ea men he e ogenei y. As i can be eadily seen, bo h models do no p oduce sensible
changes in he es ima ed coe icien s o he main eg esso s and he con ol a iables alike
compa ed wi h ou p e ious esul s. Mo eo e , he coe icien s accoun ing o he e ogeneous
e u ns a e always s a is ically insigni ican , sugges ing ha he he e ogeneous ea men bias
is no ele an in ou case. The es ima es also con i m p e ious indings on unobse ed
he e ogenei y.
Table 4: Re u ns o empo a y mig a ion - He e ogeneous e ec s
Endogenous dummy
a iable model
Endogenous
du a ion
(1)
(2)
Re u n-mig an s
-0.272***
(0.0721)
-0.516***
(0.110)
Du a ion (# o mon hs)
0.00435***
(0.000667)
0.0328***
(0.00640)
Female
-0.116***
(0.00782)
-0.118***
(0.00920)
Age a PhD (base: < 30)
Age a PhD (30-34 yea s)
-0.0482***
(0.0103)
-0.0550***
(0.0123)
Age a PhD (>34 yea s)
-0.108***
(0.0171)
-0.121***
(0.0204)
Teaching du ing PhD
0.00788
(0.00691)
0.0123
(0.00821)
No he n Uni e si y
0.112***
(0.0385)
0.128***
(0.0453)
PhD in ime
0.0265**
(0.0105)
0.0253**
(0.0124)
Scien i ic p oduc i i y
Jou nal a icles
0.00220***
(0.000389)
0.00191***
(0.000461)
Pa en s
0.0209***
(0.00519)
0.0205***
(0.00611)
Job cha ac e is ics
PhD equi ed o job
0.0323***
(0.00888)
0.0342***
(0.0105)
Job access (public compe i ion)
0.135***
(0.0102)
0.134***
(0.0120)
Pos -doc con ac
0.0236**
(0.00922)
0.0233**
(0.0109)
Job wi h eaching
-0.00963
(0.00800)
-0.00952
(0.00941)
Job wi h RD
0.0918***
(0.00948)
0.0916***
(0.0112)
Al eady wo king be o e PhD
0.0459***
(0.0122)
0.0480***
(0.0143)
Wo k expe ience
0.0303***
(0.00901)
0.0271**
(0.0106)
Wo k expe ience (squa ed)
-0.00234**
(0.00101)
-0.00212*
(0.00119)
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
23
www.Ro EA.o g
Table 4 con inued
Endogenous dummy
a iable model
Endogenous
du a ion
(1)
(2)
Indus y
0.256***
(0.0366)
0.250***
(0.0431)
Se ices
-0.138***
(0.0362)
-0.138***
(0.0425)
cons an
7.209***
(0.112)
7.236***
(0.143)
emale
0.0226
(0.0182)
0.0246
(0.0216)
age a PhD (base: < 30)
age a PhD (30-34 yea s)
-0.0107
(0.0203)
-0.00511
(0.0241)
age a PhD (>34 yea s)
0.00298
(0.0407)
0.00248
(0.0481)
eaching du ing PhD
0.00465
(0.0176)
0.00366
(0.0210)
Scien i ic p oduc i i y
jou nal a icles
-0.000794
(0.000819)
-0.000865
(0.000983)
Pa en s
-0.00805
(0.00944)
-0.00709
(0.0115)
unobs. he e ogenei y: non-
mig an s
-0.258***
(0.0498)
-0.589***
(0.0943)
unobs. he e ogenei y: mig an s
0.113***
(0.0391)
0.243***
(0.0567)
FE: PhD NUTS-3 code
Yes
Yes
FE: wo k NUTS-3 code
Yes
Yes
bachelo g adua ion yea
Yes
Yes
Indus y FE
Yes
Yes
PhD coho
Yes
Yes
bachelo g ade
Yes
Yes
mo he educa ion
Yes
Yes
high school
Yes
Yes
R_sq
0.327
0.326
F s a is ic (unob. he e .)
14.64
19.77
Obse a ions
8981
8981
No es: The dependen a iable is he log mon hly wage. ***, **, * deno e, espec i ely,
signi icance le els a 1, 5 and 10 %. Co ec ed s anda d e o s a e in pa en hesis
Re iew o Economic Analysis 17 (2025) 1-35
24
www.Ro EA.o g
6 Ins umen al a iables
To u he co obo a e he alidi y o ou esul s, we include wo ins umen s o p edic he
mig a ion du a ion. We sea ched o alid ins umen s wi hin he da a and in ex e nal da a
sou ces as well. Fi s , su ey esponden s we e asked o epo i du ing he Ph.D. hey had been
in ol ed in some o m o aining ab oad o a leas one mon h. Second, we use he numbe o
pa en s g an ed by he Eu opean Pa en O ice disagg ega ed a he NUTS-3 le el.
P e ious mobili y pa e ns ha e been ho oughly used in simila s udies as p edic o s o
u u e mobili y.
24
In ou case, we use he dummy on o eign aining du ing he Ph.D. as an
exclusion es ic ion. Ha ing spen ime ab oad du ing he Ph.D. can in p inciple be associa ed
wi h u u e mobili y. Du ing pe iods ab oad, indi iduals may lowe he psychological cos o
being mobile, acqui e p o iciency in o eign languages
25
, and inc ease hei knowledge o
possible u u e des ina ions. On he o he hand, pas mobili y could help indi iduals o e alua e
and judge hei ap i ude o adap o he cul u e and con ex o a o eign coun y. Those who
ha e al eady expe ienced mobili y ab oad may ha e de eloped p ac ical adap a ion skills,
unde s anding di e en social no ms, and managing new o unexpec ed si ua ions. Mo eo e ,
exposu e o o eign cul u es allows indi iduals o de elop compe encies such as cul u al
empa hy, men al lexibili y, and he abili y o unde s and di e en pe spec i es. These skills
a e c ucial o e ec i e adap a ion o a new con ex . Also, pas expe iences can e eal pe sonal
s eng hs and a eas o imp o emen , helping indi iduals ecognize hei lea ning and
adap a ion s yles, as well as po en ial challenges hey migh ace in he u u e. Success ully
acing and o e coming challenges ela ed o pas mobili y s eng hens con idence in one’s
abili y o li e in a new cul u al en i onmen . Thus, his ins umen may con ain su icien
in o ma ion o p edic u u e mobili y. Ne e heless, i will be a alid ins umen only i pe iods
o aining ab oad in a s uden ’s cu iculum can be e ec i ely excluded om he wage
eg ession. F om his poin o iew, we can a gue, i s , ha in mos cases, aining p og ams
o Ph.D.s a e di ec ed o acqui e speci ic skills ha migh ha e a ce ain dep ecia ion a e.
Since we obse e Ph.D.s a e h ee and i e yea s a e g adua ion, he alue o hose skills
p obably educes and, hus, is less ela ed o cu en wages. Second, o he ex en ha wages
a e ad e ised as in models o wage pos ing, ou ins umen al a iable should no play a c i ical
ole in wage de e mina ion. Fo ins ance, models o di ec ed sea ch as in Moen (1997) o
Shime (2005) ypically assume ha he e is wage pos ing, and empi ical e idence o his
mechanism can be ound in Hall and K uege (2010) o he USA and B enzel e al. (2013) o
Ge many. In bo h s udies, he au ho s epo ha wo- hi ds o hi ings a e cha ac e ized by wage
24
Fo ins ance, Ab eu e al. (2015) use mig a ion o a end uni e si y and mig a ion a e g adua ion as
ins umen s o pos -g adua ion mo es and in e -indus y mobili y o a sample o UK g adua es.
25
Gibson and McKenzie (2011), o ins ance, ind ha s uden s who s udy a o eign language a e mo e
likely o mo e ab oad.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
31
www.Ro EA.o g
Faggian, A. & McCann, P., 2009. Human capi al, g adua e mig a ion and inno a ion in B i ish
egions. Camb idge Jou nal o Economics, 33(2), pp.317–333.
Faggian, A., McCann, P. & Sheppa d, S., 2007. Some e idence ha women a e mo e mobile
han men: Gende di e ences in U.K. g adua e mig a ion beha io . Jou nal o Regional
Science, 47(3), pp.517–539.
F anzoni, C., Scella o, G., & S ephan, P., 2012. Fo eign-bo n scien is s: mobili y pa e ns o
16 coun ies. Na u e Bio echnology, 30, pp.1250–1253.
Gae a, G.L., Lub ano La ade a, G. & Pas o e, F., 2022. The E ec o Job–Educa ion Ve ical
Misma ch on Wages Among Recen PhD G adua es: E idence F om an Ins umen al
Va iable Analysis. I alian Economic Jou nal 8, 197–225.
Gae a, G.L., Lub ano La ade a, G. & Pas o e, F., 2023. O e educa ion wage penal y among
Ph.D. holde s: an uncondi ional quan ile eg ession analysis on I alian da a, In e na ional
Jou nal o Manpowe , 44(6), pp.1096-1117.
Ga cia-Que edo, J., Mas-Ve dú, F. & Polo-O e o, J., 2011. Which i ms wan PhDs? An
analysis o he de e minan s o he demand. Highe Educa ion, 63(5), pp.607–620.
Ghosh, S., & G assi, E. 2020. O e educa ion and o e skilling in he ea ly ca ee s o Ph.D.
g adua es: Does in e na ional mig a ion educe labo ma ke misma ch? Pape s in Regional
Science, 99, 915–944.
Gibson, J. & McKenzie, D., 2012. The Economic Consequences o “B ain D ain” o he Bes
and B igh es : Mic oeconomic E idence om Fi e Coun ies. The Economic Jou nal,
122(560), pp.339–375.
Gibson, J. & McKenzie, D., 2011. The mic oeconomic de e minan s o emig a ion and e u n
mig a ion o he bes and b igh es : E idence om he Paci ic. Jou nal o De elopmen
Economics, 95(1), pp.18–29.
Go lieb, P.D. & Joseph, G., 2006. College- o-wo k mig a ion o echnology g adua es and
holde s o doc o a es wi hin he Uni ed S a es. Jou nal o Regional Science, 46(4), pp.627–
659.
G eene, W., 1995. Sample selec ion in he poisson eg ession model. Wo king Pape EC-95-
06, Depa men o Economics, S e n School o Business, New Yo k Uni e si y.
G oen, J.A., 2004. The e ec o college loca ion on mig a ion o college-educa ed labo .
Jou nal o Econome ics, 121(1-2), pp.125–142.
G ogge , J. & Hanson, G.H., 2015. A ac ing Talen : Loca ion Choices o Fo eign-Bo n PhDs
in he Uni ed S a es. Jou nal o Labo Economics, 33(S1), pp.S5–S38.
G ogge , J. & Hanson, G.H., 2011. Income maximiza ion and he selec ion and so ing o
in e na ional mig an s. Jou nal o De elopmen Economics, 95(1), pp.42–57.
G o e, W.A. & Wu, S., 2007. The Sea ch o Economics Talen : Doc o al Comple ion and
Resea ch P oduc i i y. The Ame ican Economic Re iew, 97(2), pp.506–511.

Re iew o Economic Analysis 17 (2025) 1-35
32
www.Ro EA.o g
Haapanen, M. & Te o, H., 2011. Mig a ion o he highly educa ed: e idence om esidence
spells o uni e si y g adua es. Jou nal o Regional Science, 52(4), pp.587–605.
Hall, R.E. & K uege , A.B., 2010. E idence on he De e minan s o he Choice be ween Wage
Pos ing and Wage Ba gaining. NBER Wo king Pape 16033.
Hancock, S., 2023. Knowledge o science-based economy? The employmen o UK PhD
g adua es in esea ch oles beyond academia. S udies in Highe Educa ion, 48(10), 1523–
1537.
Ha on, T.J., 2014. The economics o in e na ional mig a ion: A sho his o y o he deba e.
Labou Economics, 30, pp.43–50.
Heckman, J.J., 1979. Sample Selec ion Bias as a Speci ica ion E o . Econome ica, 47(1),
pp.153–161.
Heckman, J.J., 1976. The common s uc u e o s a is ical models o unca ion, sample selec ion
and limi ed dependen a iables and a simple es ima o o such models. Annals o Economic
and Social Measu emen , 5(4), pp.475–492.
Heckman, J.J., U zua, S. & Vy lacil, E., 2006. Unde s anding Ins umen al Va iables in Models
wi h Essen ial He e ogenei y. The Re iew o Economics and S a is ics, 88(3), pp.389–432.
Hill, J.K., 1987. Immig an decisions conce ning du a ion o s ay and mig a o y equency.
Jou nal o De elopmen Economics, 25(1), pp.221–234.
Holland, P., 1986. S a is ics and causal in e ence. Jou nal o he Ame ican S a is ical
Associa ion, 81(396), pp.945–960.
Maddala, G.S., 1986. Limi ed-Dependen and Quali a i e Va iables in Econome ics,
Camb idge Uni e si y P ess.
Ma che a, F., 2012. Re u n Mig a ion and he Su i al o En ep eneu ial Ac i i ies in Egyp .
Wo ld De elopmen , 40(10), pp.1999–2013.
May , K. & Pe i, G., 2009. B ain d ain and b ain e u n: heo y and applica ion o Eas e n-
Wes e n Eu ope. The BE Jou nal o Economic Analysis & Policy, 9(1), pp.1–50.
McCo mick, B. & Wahba, J., 2001. O e seas Wo k Expe ience, Sa ings and En ep eneu ship
Amongs Re u n Mig an s o LDCs. Sco ish Jou nal o Poli ical Economy, 48(2), pp.164–
178.
Mesna d, A., 2004. Tempo a y mig a ion and capi al ma ke impe ec ions. Ox o d Economic
Pape s, 56(2), pp.242–262.
Moen, E.R., 1997. Compe i i e sea ch equilib ium. Jou nal o Poli ical Economy, 105(2),
pp.385–411.
Pa en i, B., Pin o, M. & Sa no, D., 2022. Job Sa is ac ion Among Ph.D. Holde s: How much
do Regional Di ides and Employmen Sec o s ma e ?. Highe Educa ion Policy , 35,
pp.318–374.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
33
www.Ro EA.o g
Passa e a, G., T i ella o, P., & T i en i, M., 2018. Be ween academia and labo ma ke — he
occupa ional ou comes o PhD g adua es in a pe iod o academic e o ms and economic
c isis. Highe Educa ion, pp.1-19.
Pe i, G., Shih, K. & Spa be , C., 2015. STEM wo ke s, H-1B isas, and p oduc i i y in US
ci ies. Jou nal o Labo Economics, 33(S1), pp.S225–S255.
Reinhold, S. & Thom, K., 2013. Mig a ion Expe ience and Ea nings in he Mexican Labo
Ma ke . Jou nal o Human Resou ces, 48(3), pp.768–820.
Rubin, D.B., 1974. Es ima ing causal e ec s o ea men s in andomized and non andomized
s udies. Jou nal o Educa ional Psychology, 66(5), pp.688–701.
San os, Dos, M.D. & Pos el-Vinay, F., 2003. Mig a ion as a sou ce o g ow h: he pe spec i e
o a de eloping coun y. Jou nal o Popula ion Economics, 16, pp.161–175.
Shime , R., 2005. The assignmen o wo ke s o jobs in an economy wi h coo dina ion ic ions.
Jou nal o Poli ical Economy, 113(5), pp.996–1025.
Sjaas ad, L.A., 1962. The Cos s and Re u ns o Human Mig a ion. Jou nal o Poli ical
Economy, 70(5), pp.80–93.
S a k, O., Helmens ein, C. & Yego o , Y., 1997. Mig an s' Sa ings, Pu chasing Powe Pa i y,
and he Op imal Du a ion o Mig a ion. In e na ional Tax and Public Finance, 4(3),
pp.307–324.
S ock, J.H. & Yogo, M., 2005. Tes ing o weak ins umen s in Linea I eg ession. In D. W.
K. And ews & J. H. S ock, eds. Iden i ica ion and In e ence o Econome ic Models:
Essays in Hono o Thomas Ro henbe g. Essays in Hono o Thomas Ro henbe g.
Camb idge: Camb idge Uni e si y P ess, pp. 80–108.
Te za, J., 1998. Es ima ing coun da a models wi h endogenous swi ching: Sample selec ion
and endogenous ea men e ec s. Jou nal o Econome ics, 84(1), pp.129–154.
Venho s , V., Van Dijk, J. & Van Wissen, L., 2011. An Analysis o T ends in Spa ial Mobili y
o Du ch G adua es. Spa ial Economic Analysis, 6(1), pp.57–82.
Re iew o Economic Analysis 17 (2025) 1-35
34
www.Ro EA.o g
Appendix
Table A1: Mig a ion decision and mig a ion du a ion
I s age
Mig a ion decision
II s age
Mig a ion du a ion
(1)
(2)
emale
-0.201***
(0.037)
1.569**
(0.612)
age a PhD (base: < 30)
age a PhD (30-34 yea s)
-0.234***
(0.041)
1.916***
(0.698)
age a PhD (>34 yea s)
-0.616***
(0.059)
4.951***
(1.899)
eaching du ing PhD
0.053
(0.036)
-0.486***
(0.185)
no he n Uni e si y
-0.232
(0.221)
1.161
(0.872)
change ci y o PhD
0.103***
(0.037)
-0.754**
(0.319)
child en
-0.315***
(0.039)
2.622***
(0.957)
cons an
-0.618**
(0.262)
16.701***
(4.606)
in mills (II s age)
-12.146***
(3.744)
FE: PhD NUTS-3 code
Yes
Yes
PhD ields o s udy
Yes
Yes
PhD coho
Yes
Yes
bachelo g ade
Yes
Yes
mo he educa ion
Yes
Yes
high school
Yes
Yes
R_sq
0.071
Obse a ions
8981
8981
No es: A he i s s age, he dependen a iable is he bina y indica o o empo a y
mobili y. A he second s age, he dependen a iable is he numbe o mon hs spen
ab oad. ***, **, * deno e, espec i ely, signi icance le els a 1, 5 and 10 %. S anda d
e o s in he II s age (in pa en hesis) a e co ec ed o he wo-s age p ocedu e.
DI CINTO, GRASSI Tempo a y mig a ion and Wages o PhDs
35
www.Ro EA.o g
Table A2: IV mig a ion du a ion
IV1: aining ab oad du ing PhD
1.025***
(0.082)
IV2: numbe o pa en s (NUTS-3)
0.010***
(0.003)
emale
1.420**
(0.622)
age a PhD (base: < 30)
age a PhD (30-34 yea s)
1.694**
(0.705)
age a PhD (>35 yea s)
4.516**
(1.917)
eaching du ing PhD
-0.437**
(0.191)
no he n Uni e si y
0.822
(0.838)
change ci y o PhD
-0.685**
(0.322)
child en
2.335**
(0.967)
cons an
10.955**
(4.992)
in mills (II s age)
-11.005***
(3.783)
FE: PhD NUTS-3 code
Yes
PhD ields o s udy
Yes
PhD coho
Yes
bachelo g ade
Yes
mo he educa ion
Yes
high school
Yes
Chi-squa e (ins umen s)
166.86***
Obse a ions
8981
No es: The dependen a iable is he numbe o mon hs spen ab oad. ***, **, * deno e,
espec i ely, signi icance le els a 1, 5 and 10 %. S anda d e o s (in pa en hesis) a e
co ec ed o he wo-s age p ocedu e.