Higher economic growth in poor countries, lower migration flows to the OECD – Revisiting the migration hump with panel data

Benček, Da id; Schneide heinze, Claas
A icle — Published Ve sion
Highe economic g ow h in poo coun ies, lowe
mig a ion lows o he OECD – Re isi ing he mig a ion
hump wi h panel da a
Wo ld De elopmen
P o ided in Coope a ion wi h:
Kiel Ins i u e o he Wo ld Economy – Leibniz Cen e o Resea ch on Global Economic Challenges
Sugges ed Ci a ion: Benček, Da id; Schneide heinze, Claas (2024) : Highe economic g ow h in poo
coun ies, lowe mig a ion lows o he OECD – Re isi ing he mig a ion hump wi h panel da a,
Wo ld De elopmen , ISSN 1873-5991, Else ie , Ams e dam, Vol. 182, pp. 1-20,
h ps://doi.o g/10.1016/j.wo ldde .2024.106655
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/301403
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Wo ld De elopmen 182 (2024) 106655
A ailable online 8 June 2024
0305-750X/© 2024 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
Regula Resea ch A icle
Highe economic g ow h in poo coun ies, lowe mig a ion lows o he
OECD – Re isi ing he mig a ion hump wi h panel da a
Da id Benˇ
cek
a
, Claas Schneide heinze
b
,
*
a
Kiel Ins i u e o he Wo ld Economy, Ge many
b
Kiel Ins i u e o he Wo ld Economy, Ge many
ARTICLE INFO
JEL Classi ica ion:
F22
F63
O15
Keywo ds:
In e na ional mig a ion
Economic de elopmen
De elopmen assis ance
ABSTRACT
Compa ing emig a ion a es o coun ies a di e en s ages o economic de elopmen , an in e se u-shape
eme ges. Since he “mig a ion hump” peaks a an a e age income o 6000 o 10 000 USD, economic p og ess in
de eloping coun ies is o en assumed o inc ease mig a ion consis en ly. Howe e , i is poo ly unde s ood o
wha ex end coun y-le el cha ac e is ics, indi idual incomes and o he dimensions o de elopmen e oke his
pa e n, which limi s i s alue o causal in e ence and conc e e policy ad ice. In his pape we ocus on he ole
o economic g ow h and in es iga e whe he in de eloping coun ies emig a ion indeed inc eases wi h economic
p og ess a sho e mo e policy- ele an ime pe iods o up o 10 yea s. Using 35 yea s o da a on mig a ion lows
o OECD des ina ions, we success ully ep oduce he hump-shape in he c oss-sec ion. Howe e , ou mo e
igo ous ixed e ec s panel es ima ions ha exploi he a ia ion o e ime obus ly ea u e con as ing esul s:
emig a ion a es all as incomes inc ease. This inding holds independen o he le el o income a coun y s a s
ou a . In con as o p e ailing de elopmen emig a ion na a i es, ou esul s imply ha ising indi idual in-
comes discou age emig a ion and hence conduci e economic policies can educe emig a ion. Ou indings do no
ule ou ha o he slow-mo ing de elopmen dimensions such as educa ional ad ancemen , demog aphic
change, and s uc u al economic ans o ma ion could s ill inc ease mig a ion in he long e m.
1. In oduc ion
In e na ional mig a ion is as old as na ion s a es. In ecen decades,
howe e , mig a ion has inc easingly ocused on a small numbe o
des ina ion coun ies. While he global sha e o in e na ional mig an s
inc eased only mode a ely om 2.9 pe cen in 1990 o 3.5 pe cen in
2019 (UN, 2019), mig a ion owa ds OECD des ina ions has inc eased a
a much highe pace (OECD, 2019). As a esul , abou hal o he 272
million in e na ional mig an s oday eside in jus 10 coun ies (UN,
2019). Acco ding o da a om he Gallup Wo ld Poll his end is un-
likely o shi : Globally 750 million indi iduals in end o mo e ab oad,
and wo hi ds o hem aim a one o jus 18 des ina ions (Esipo a e al.,
2018). In he yea s o come, clima e change and popula ion g ow h a e
o ecas o u he inc ease he pool o aspi ing mig an s (Ca aneo &
Pe i, 2016; Hanson & McIn osh, 2016). In many des ina ion coun ies,
immig a ion has become highly poli icized and a cul u al backlash uels
populis mo emen s (Hooghe & Ma ks, 2018; Ingleha & No is, 2016).
In sea ch o common g ound, policy make s emphasize he impo -
ance o ackling oo causes o mig a ion and ha e iden i ied po e y
and low economic de elopmen as majo d i e s.
1
In luencing mig a ion
indi ec ly h ough de elopmen coope a ion a he han di ec ly by
es ic i e immig a ion policy comes wi h poli ical and p ac ical ad-
an ages. De elopmen policies a e mo e likely o gain public suppo
om o e s h oughou he poli ical spec um. Mo eo e , upholding
es ic i e immig a ion policy egimes is ex emely expensi e and has
been shown o shi egula o i egula mig a ion (Czaika & Hobol h,
2016). Howe e , he idea o educe mig a ion by suppo ing economic
de elopmen has been hea ily c i icized by academics based on ecen
s udies showing middle-income coun ies o ha e he highes emig a ion
a es (Clemens & Pos el, 2018; de Haas, 2019). These au ho s a gue ha
coun y-le el income and emig a ion a e ela ed in a hump-shaped
pa e n (de Haas, 2010; Clemens, 2014; Djajic e al., 2016; Commis-
sion, 2018; Dao e al., 2018; Clemens & Pos el, 2018), combining c oss-
sec ional e idence wi h a plausible heo y: A low income le els, c edi
* Co esponding au ho .
E-mail add esses: [email p o ec ed] (D. Benˇ
cek), [email p o ec ed] (C. Schneide heinze).
1
The Mig a ion Pa ne ship F amewo k ini ia ed by he Eu opean Commission, he Global Compac s o Mig a ion, and Emmanuel Mac on in his speech a he
So bonne each exp ess he need o imp o e li ing condi ions in o igin coun ies o educe in e na ional mig a ion (Ec, 2016; 2018; Mac on, 2018).
Con en s lis s a ailable a ScienceDi ec
Wo ld De elopmen
jou nal homepage: www.else ie .com/loca e/wo ldde
h ps://doi.o g/10.1016/j.wo ldde .2024.106655
Accep ed 10 May 2024
Wo ld De elopmen 182 (2024) 106655
2
cons ain s p e en aspi ing mig an s om emig a ing, while a highe
income le els dec easing economic incen i es o emig a ion domina e
e e less binding c edi cons ain s (Dao e al., 2018). In consequence,
emig a ion a es a e assumed o ollow an in e se u-shape along he
economic de elopmen pa h o a coun y.
Such a ela ionship would ha e a - eaching implica ions: The peak
implied by he exis ing es ima es is loca ed oughly a he cu en pe
capi a income le el o Bulga ia, China o Colombia. Abou wo hi ds o
he wo ld’s popula ion li es in coun ies below his h eshold (Dao e al.,
2018). Hence, in e p e ing he mig a ion hump as a causal ela ionship
means ha economic g ow h in de eloping coun ies should be expec ed
o boos emig a ion in he u u e. E ec i e de elopmen policy could
hus aise immig a ion p essu es in mos p ima y des ina ion coun ies.
Clemens and Pos el (2018, p. 686) explici ly emphasize his ade-o :
“de elopmen assis ance o o igin coun ies, o he ex en ha i is success-
ul in os e ing sus ained de elopmen , is likely o c ea e addi ional p essu e
on hi d-coun y hos ing a angemen s by encou aging g ea e o e all
emig a ion.”
Ye , while his ela ionship is inhe en ly in e - empo al, many o he
abo e s udies ely almos exclusi ely on c oss-sec ional e idence. The
ac ha middle income coun ies expe ience highe emig a ion han
hei poo e coun e pa s migh be a di ec consequence o hei income
le el o i migh be due o undamen al di e ences be ween low and
middle-income coun ies ha simul aneously a ec bo h de elopmen
and emig a ion (Lucas, 2019). In ha espec , he mig a ion hump hy-
po hesis esembles one o he mos hea edly deba ed concep s in
de elopmen economics: he Kuzne s-cu e. Based on he obse a ion
ha middle-income coun ies expe ience highe economic inequali y
han hei poo e and iche coun e pa s, Kuzne s deduced ha eco-
nomic de elopmen in poo coun ies inc eases inequali y (Kuzne s,
1955). Only much la e i was shown ha he hump-shaped c oss-
coun y pa e n was la gely d i en by sys ema ic di e ences be ween
coun ies and does no ep esen a na u al ime pa h (Deininge &
Squi e, 1998; Field, 2002).
In his pape , we a gue along he same lines o he ela ionship
be ween economic g ow h and emig a ion o OECD coun ies. We
demons a e ha coun ies a he upwa ds-sloping pa o he mig a ion
hump, on a e age, di e ma kedly om iche coun ies wi h espec o
c ucial exogenous ac o s such as dis ance o OECD coun ies, size and
pas colonial ies. These exogenous cha ac e is ics a e well-known o
in luence bo h de elopmen and mig a ion in he same di ec ion and
hus likely con ound he c oss-sec ional ela ionship.
Mo eo e , he subs an i e a gumen abou he mig a ion hump aims
a he e y long un,
2
and migh hus no be e y in o ma i e abou sho
o medium e m policy aims. E en i he obse ed c oss-sec ional pa e n
was oo ed in a causal ela ionship a he coun y le el, an impo an
open empi ical ques ion is whe he and how sho o medium e m
dynamics de ia e om he long- e m ajec o y. Unde s anding such
sho o medium e m dynamics is c ucial o policymake s in ying o
an icipa e he ma ginal e ec s o de elopmen policy a easonable ime
ho izons.
In his pape , we employ a coun y-le el da a se ecen ly compiled
by Wesselbaum and Abu n (2019) ha co e s bila e al mig a ion lows
be ween 198 coun ies o o igin and 16 OECD des ina ions om 1980 o
2014 and es he exis ence o he mig a ion hump in panel da a. In
con as ing c oss-sec ional wi h panel es ima es, we a e, o ou knowl-
edge, he i s o sys ema ically analyze he dynamics unde lying he
mig a ion hump. Ou analysis ocuses speci ically on de eloping coun-
ies o which he c oss-sec ional e idence sugges s a posi i e ela-
ionship be ween economic p og ess and emig a ion. While we
success ully ep oduce he hump-shape in he c oss-sec ion, ou mo e
igo ous ixed e ec s panel es ima ions employing he wi hin a iance
o e ime obus ly yield con as ing esul s: emig a ion a es all as in-
comes inc ease. Ou esul s a e obus o using di e en income anges,
ime ends, and con ols. Mos impo an ly hey also hold o di e en
mig a ion da a and di e en ime pe iods (i.e. i e and en yea agg e-
ga es). Ou esul s do no imply ha inancial cons ain s would no be
binding o many indi iduals. Ye , when economic oppo uni ies
imp o e, ew o hem seem o u ilize hei inc easing capabili ies o
mig a e.
Ou inding cas s signi ican doub on he alidi y o he mig a ion
hump hypo hesis as a uni e sal in e - empo al ela ionship and conse-
quen ly ques ions i s ele ance o policy making. In con as o p e-
ailing de elopmen emig a ion na a i es, ou esul s imply ha
conduc i e economic policies in de eloping coun ies can educe
emig a ion.
The emainde o he pape p oceeds as ollows: In Sec ion 2 we
e iew he mig a ion and de elopmen li e a u e and c i ically discuss
bo h he heo e ical a gumen and he empi ical e idence ha unde pin
he mig a ion hump. A e in oducing he da a in Sec ion 3, Sec ion 4
p esen s he empi ical analysis as well as se e al obus ness checks.
Sec ion 5 sums up and concludes.
2. The de elopmen –emig a ion nexus: Theo y and exis ing
empi ical e idence
S udying he ela ionship be ween economic de elopmen and
mig a ion has a long adi ion in de elopmen economics (e.g. Ha is &
Michael, 1970). The as majo i y o he academic li e a u e used o
ocus on he in luence o mig a ion on de elopmen (Beine e al., 2001;
Giuliano & Ruiz-A anz, 2006). How de elopmen a ec s mig a ion has
ecei ed much less a en ion. As in e na ional mig a ion gained poli ical
ele ance in des ina ion coun ies due o la ge numbe s o i egula
a i als o mig an s om poo coun ies, he ocus s a ed o shi .
Se e al empi ical and heo e ical s udies ha e begun o analyze he ole
o economic de elopmen in emig a ion pa e ns mo e sys ema ically
(Docquie e al., 2014; Dus mann & Oka enko, 2014; Clemens, 2014;
Dao e al., 2018; de Haas e al., 2018; Clemens, 2020a; b). While hese
au ho s’ empi ical indings some imes di e ge, hey b oadly ag ee on
he main heo e ical a gumen : An indi idual’s decision o mig a e
gene ally depends on (i) aspi a ions and (ii) capabili ies o mo e
(Ca ling & Schewel, 2018).
A he mac o le el, nume ous ac o s sys ema ically in luence aspi-
a ions and capabili ies. These include economic, poli ical, cul u al,
en i onmen al, and demog aphic condi ions. Due o he complex ela-
ionship be ween economic p og ess and hese o he dimensions o
de elopmen , hei indi idual e ec s a e di icul o disen angle. La ge
pa s o he li e a u e ely on GDP as a uni e sal measu e o de elop-
men . Economic g ow h, o example, imp o es local incomes as well as
he s a e’s abili y o p o ide public goods.
A p io i, he o e all in luence o de elopmen on emig a ion is
ambiguous. I local li elihoods imp o e, mig a ion aspi a ions dec ease
(Dus mann & Oka enko, 2014). Howe e , highe disposable income
simul aneously elaxes budge cons ain s ha may p e iously ha e
p ohibi ed mig a ion. Hence, economic de elopmen dec eases mig a-
ion aspi a ions bu inc eases mig a ion capabili ies. Which o hese
e ec s domina es likely di e s ac oss coun ies and be ween di e en
g oups o indi iduals wi hin coun ies.
2.1. The mig a ion hump: Concep , e idence and in e p e a ion
The mig a ion hump hypo hesis (o mobili y ansi ion heo y) da es
back o Zelinsky (1971) and is among he bes known s ylized ac s
ega ding he de elopmen –mig a ion nexus. The hypo hesis posi s an
in e ed u-shaped ela ionship be ween de elopmen and emig a ion.
This undamen ally di e s om a adi ional neoclassical iew o
mig a ion, as o example employed in he g a i y li e a u e, which
2
Clemens and Pos el (2018) demons a e ha a ealis ic a es o economic
g ow h he poo es quin ile o coun ies migh no each he peak o he
mig a ion hump un il he yea 2198.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
3
omi s c edi cons ain s a he indi idual le el and hus assumes
emig a ion o dec ease along he de elopmen ajec o y as ising li ing
s anda ds a home ende mig a ion less a ac i e. Many schola s ha e
a gued in a o o a hump-shaped ela ionship be ween de elopmen
and emig a ion using di e en e ms, e.g. ‘mig a ion cu e’ (Ake man,
1976), ‘mig a ion ansi ion’ (Gould, 1979), ‘mig a ion hump’ (Ma in,
1993) and ‘emig a ion li e cycle’ (Ha on & Williamson, 1994).
3
While
hese schola s b oadly ag ee on he in e se-U shaped pa e n, hey hold
di e en ac o s esponsible o i (see Clemens (2014) o an excellen
e iew).
Among hese a e demog aphic change (Eas e lin, 1961; Ha on &
Williamson, 1994), inancial cons ain s (Faini & Ven u ini, 1994;
Ha on & Williamson, 1994), in o ma ion asymme ies (G eenwood,
1969; Massey e al., 1993; Eps ein, 2008), s uc u al economic ans-
o ma ion (Zelinsky, 1971), economic inequali y (S a k, 2006) and
immig a ion ba ie s ab oad (Ha on & Williamson, 2005). All hese
p oposed de e minan s a e s ongly ela ed o de elopmen and a gu-
ably also o emig a ion and he e a e di e en mechanisms h ough
which hey may gi e ise o a hump-shaped long- e m ela ionship be-
ween de elopmen and emig a ion. Ye , such a mig a ion hump is no a
unique ou come ha will always occu . E en i all hese ac o s ope a e
as sugges ed, he nega i e ela ionship be ween de elopmen and
emig a ion, ha is induced by imp o ing li ing s anda ds and inc easing
oppo uni y cos s o mig a ion migh s ill p e ail.
de Haas (2010) was he i s o se e al esea che s who p o ided
empi ical e idence in suppo o he mig a ion hump hypo hesis a a
global le el. Desc ip i ely and by means o bi a ia e and mul i a ia e
eg ession analysis, he de ec ed a non-linea , hump-shaped ela ionship
be ween pe capi a GDP and emig an s ocks wi h a peak a an income
le el o 12 000 USD pe capi a. Using c oss-sec ional da a om he
Wo ld Bank and he Uni ed Na ions, Clemens (2014) showed ha he
mig a ion hump also exis s in mig a ion low da a. The highes
emig a ion a es a e obse ed in coun ies in he middle o he global
income dis ibu ion, while he iches and he poo es coun ies expe-
ience sys ema ically less emig a ion. Acco ding o his non-pa ame ic
eg essions, he a e o emig a ion s eadily inc eases up o a peak
a ound a pe capi a income o 6000–8000 USD. This pa e n holds o
each o he decades om 1960 o 2010. In a mo e ecen s udy, Clemens
and Pos el (2018) loca e he peak o be a a somewha highe le el o
8000–10 000 USD.
Dao e al. (2018), Djajic e al. (2016), and he Eu opean Commission
(2018) p o ide simila desc ip i e e idence.
4
Ye , he loca ion o he
peak in hei s udies a ies be ween 4000 USD (Djajic e al., 2016) and
7000–13000 USD (Commission, 2018). Since hese s udies di e in
e ms o hei mig a ion and GDP da a, ime pe iods, and coun y se-
lec ion, a ying peak le els do no ques ion he gene al ela ionship.
Despi e di e ences in he loca ion o he peak, hese s udies con inc-
ingly demons a es: Emig a ion is, on a e age, highe in middle-income
coun ies han i is in ei he high o low-income coun ies.
Howe e , he mig a ion hump’s policy ele ance is based on i ’s
causal in e p e a ion. Suppo ed by he di e en heo e ical a gumen s
ha link de elopmen o ising emig a ion, he c oss-sec ional e idence
o he mig a ion hump is widely in e p e ed as a na u al ime pa h a
he coun y le el. Fo example, Clemens and Pos el (2018) sugges a
causal ela ionship when s a ing ha : “economic g ow h has his o ically
aised emig a ion in almos all de eloping coun ies”. This in e p e a-
ion ypically builds speci ically on he ole o indi idual incomes and
he easibili y o inance mig a ion. To explain he e ec o ising
incomes on emig a ion in he con ex o he mig a ion hump, he
mig a ion decision is depic ed as an in es men decision: Any inc ease in
indi idual income a ec s bo h he easibili y o mig a ion by easing he
inancial cons ain and he incen i e o s ay by inc easing he oppo -
uni y cos s. A low income le els, he o me e ec domina es, c ea ing
a posi i e income–mig a ion ela ionship un il income is su icien ly
high o discou age emig a ion. In consequence, o e he long- e m
de elopmen pa h o a coun y, emig a ion a es a e assumed o in-
c ease uni e sally un il pe capi a incomes o 6000–10 000 USD a e
eached. This e y in ui i e explana ion is backed up by mic oeconomic
e idence o Indonesia. Using census da a, Bazzi (2017) p o ides some
empi ical suppo o he exis ence o a capi al cons ain o in e na-
ional mig a ion in a causal se up. While in poo u al a eas o Indonesia
Bazzi (2017) inds posi i e income shocks o inc ease emig a ion, he
opposi e e ec occu s o he mos de eloped egions wi hin he coun-
y. I has o be no ed, howe e , ha his con incing e idence comes
om a single coun y whe e simila i y be ween di e en o igins is much
highe han in he global c oss-coun y samples ha unde lie he
mig a ion hump.
Mic oeconomic suppo , a ich and in ui i e heo e ical ounda ion
and he empi ical ep oducibili y ac oss da a se s and ime ha e c ea ed
a powe ul na a i e o in e p e he mig a ion hump as a uni e sal
ela ionship a he coun y le el. Howe e , a causal in e p e a ion based
on c oss-sec ional e idence, migh s ill be misleading, especially since
a ious omi ed a iables could go e n his ela ionship.
2.2. Risks o causal in e ence: Poo and middle income coun ies di e
sys ema ically
The causal in e p e a ion o he mig a ion hump hypo hesis is based
on he assump ion ha oday’s poo and middle income coun ies a e
undamen ally simila wi h espec o impo an ac o s such as mig a-
ion cos . Howe e , i oday’s poo coun ies di e om hei iche
coun e pa s in impo an omi ed ac o s such as o example
geog aphical loca ion, language, o cul u e, such an assessmen could be
misguided.
Economically speaking, sys ema ic he e ogenei y ac oss coun ies
may endange alid causal in e ence. As Lucas (2019, p. 18) pu s i : “In
he end, c oss-coun y e idence may ell us li le abou he ime-pa h o
emig a ion as de elopmen p oceeds; hose coun ies cu en ly in he
middle-income ange may simply di e in undamen al ways om wha
hei poo e coun e pa s a e e ol ing in o.”
We b ie ly examine di e ences in basic coun y cha ac e is ics ha
a e known o in luence bo h de elopmen and mig a ion. In doing so, we
ocus on he g oup o poo coun ies on he upwa d-sloping pa o he
hump and es i hese a e simila o hei iche coun e pa s (summa-
ized in Table 1). Speci ically, he i s g oup consis s o all coun ies
Table 1
Selec ed coun y cha ac e is ics by income g oup.
low income: <5000
GDP pc
N =69
emaining non-
OECD
N =84
p- alue
a . GDP pc as o 2010 (PPP
$2011)
2367 (1058) 16,489 (17564) <0.001
dis ance o OECD coun y
(km)
4744 (1754) 3872 (2359) 0.012
common bo de wi h
OECD
0.00 (0.00) 0.04 (0.19) 0.083
colonial ies wi h OECD 0.46 (0.50) 0.70 (0.46) 0.004
landlocked 0.31 (0.47) 0.10 (0.30) 0.002
a . popula ion (millions,
2010)
44.6 (156) 10.1 (23.1) 0.073
No e: Coun ies a e clus e ed by a e age income be ween 1960 and 2010; da a
sou ces: Penn Wo ld Tables 2015 and CEPII’s GeoDis Da abase.
3
In line wi h Clemens (2014) we use he e m ‘mig a ion hump’, which is he
mos illus- a i e in ou iew.
4
We label a eg ession ha simply c ea es a bes i in a wo-dimensional
model as “de- sc ip i e” because i is a way o desc ibing he ela ionship be-
ween he wo a iables and no an app oach ha aims a isola ing unde lying
componen s.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
4
wi h an a e age income pe capi a o less han 5000 USD be ween 1960
and 2010, while he second g oup includes all he emaining non-OECD
coun ies.
5
We exclude OECD o igin coun ies o his desc ip i e able
because geog aphical p oximi y o hese p ima y des ina ion coun ies
is among he ac o s we wan o in es iga e. The geog aphical measu es
a e aken om CEPII’s GeoDis Da abase (Maye , 2011). The dissimi-
la i ies a e s iking. Poo e coun ies le o he hump’s peak a e on
a e age loca ed signi ican ly u he away om OECD-coun ies, less
likely o ha e colonial ies wi h hem and a e mo e equen ly land-
locked. In addi ion, hese coun ies hos much la ge popula ions. E en
a e excluding China and India he a e age popula ion in he poo
coun y g oup is almos wice as high. Small coun ies o en exhibi
highe emig a ion a es han la ge ones no leas because o a lack o
oppo uni ies o specializa ion (de Haas, 2010). Sho dis ance mo es,
o example o he nex la ge ci y, a e a mo e likely o in ol e c ossing
in e na ional bo de s i he coun y’s land a ea is small. Fu he mo e,
lea ing a small coun y is much easie in e ms o mone a y and physical
e o as he nea es bo de is much close . I is impo an o no e ha all
hese ac o s a e well known o impac de elopmen and mig a ion and
a he same ime hey a e plausibly exogenous. Mo e speci ically, hey
a e nega i ely ela ed o bo h de elopmen and emig a ion, and hence,
p o ide a compe ing explana ion o low emig a ion a es in poo
coun ies. Such ac o s a e he e o e likely o con ound any empi ical
analysis o he ela ionship be ween de elopmen and emig a ion ha
does no accoun o hem.
These insigh s cas some doub on he hump’s alidi y as a uni e sal
ela ionship and ques ion in e ences based on c oss-sec ional da a. Fo a
obus iden i ica ion o he link be ween economic de elopmen and
emig a ion we need o con ol o di e ences ac oss coun ies. Tha is
he na u al domain o panel s udies.
2.3. The impac o de elopmen on emig a ion o e ime: Insigh s om
panel s udies
In con as o c oss-sec ional s udies, ime-se ies app oaches allow o
accoun o di e ences be ween coun ies by employing he a ia ion
wi hin coun ies o e ime. While economic de elopmen is included in
mos s udies on mig a ion as an impo an d i e , e y ew exis ing
s udies explici ly ocus on he impac o economic de elopmen on
emig a ion, and ha dly any s udy accoun s o non– linea ela ionships
o explici ly es s he mig a ion hump. P io o ou s udy we a e only
awa e o wo pape s ha speci ically es he mig a ion hump in ime
se ies da a (Vogle & Ro e, 2000; Telli, 2014). Howe e , hese s udies
only ocus on mig a ion o one speci ic des ina ion coun y (Ge many
and he UK, espec i ely), and bo h ely exclusi ely on annual da a.
Fu he mo e, and likely o be mos p oblema ic, bo h s udies use me ely
a squa ed e m in hei panel eg essions o accoun o a hump-shaped
ela ionship and do no es mo e lexible amewo ks, hus o cing he
da a o ei he ake a hump shape, a linea shape o no shape a all.
Recen econome ic s udies show ha using only a squa ed e m o
de ec (in e se) u-shaped ela ionships o en leads o alse conclusions
(Lind & Mehlum, 2010; Haans e al., 2016; Simonsohn, 2018). Mos o
oday’s g a i y-s yle mig a ion models ocus on he de e minan s o
bila e al mig a ion lows and hence on he des ina ion choice a he
han on oo causes o emig a ion in o igin coun ies. In consequence,
exis ing s udies yield inconclusi e esul s (Clemens, 2014). While o
example Bazzi (2017) and Dao e al. (2018) de ec a posi i e ela ionship
be ween GDP and mig a ion a low income le els, O ega and Pe i
(2013) and B¨
ohme e al. (2019) ind a uni e sal nega i e ela ionship.
O he s udies do no e u n a s a is ically signi ican ela ionship a all
(Mayda, 2010; Naud´
e, 2010; Ruyssen e al., 2012). Acco ding o
Clemens (2014), exis ing panel and ime se ies s udies ha seek o
explain he ela ionship be ween income a o igin and emig a ion ail o
de ec he mig a ion hump, because hey su e om h ee majo
sho comings. Fi s , he ime ho izon hey employ (15–20 yea s) is oo
sho o de ec long- e m pa e ns. Second, by using annual da a, sho -
e m economic luc ua ions mask he in luence o income le els and
long- e m ends. Thi d, as ime-se ies s udies ypically do no allow o
a non-linea e ec , he di e en di ec ion o impac (nega i e o iche
coun ies, posi i e o poo e coun ies) leads o inconsis en esul s and
coe icien s ha a e close o ze o. We ag ee wi h his e alua ion and
speci ically design ou empi ical me hodology below o add ess hese
limi a ions (see sec ion 4).
3. Da a
Da a a ailabili y is among he main cons ain s o quan i a i e
mig a ion esea ch in gene al. This is pa icula ly ele an o s udies
ha in es iga e long e m ends. Any conclusi e analysis mus be based
on a la ge ime dimension in o de o be able o iden i y subs an ial
changes and a oid elying on sho e m luc ua ions in mig a o y pa -
e ns due o exogenous shocks. Fu he mo e, a la ge sample o obse -
a ional uni s is desi able o p e en biased es ima es esul ing om
idiosync a ic cha ac e is ics o indi idual uni s.
The mig a ion panel da ase compiled by Abu n and Wesselbaum
(2019), mee s bo h o hese equi emen s. By me ging in o ma ion om
he 2015 Re ision o he Uni ed Na ions’ Popula ion Di ision wi h he
OECD’s mig a ion da abase and da a om O ega and Pe i (2013), he
au ho s compile one o he longes and mos exhaus i e panel da a se s
o bila e al ne mig a ion lows, co e ing 198 coun ies o o igin and 16
OECD des ina ions om 1980 o 2014. S ill, he panel is unbalanced
because o missing da a, especially in he ea ly 1980 s when da a is
a ailable o only abou hal o he coun y dyads. Bu since ou esea ch
ques ion ocuses on he ela ionship be ween incomes and emig a ion,
we a e no in e es ed in di ec ions o mig a ion lows bu a he hei
a ia ions in o al olumes o e ime (and income). The e o e, we
agg ega e all bila e al lows by hei o igin o calcula e he numbe o
emig an s pe coun y and yea . To a ce ain ex en , his agg ega ion
also mi iga es a po en ial selec ion bias om missing obse a ions ea ly
on in he obse a ion pe iod.
Ou main a iable o in e es is economic de elopmen o which we
ely on GDP da a p o ided he Penn Wo ld Table (Feens a e al., 2015).
In addi ion, ou empi ical analysis uses se e al common con ol a i-
ables: coun y sizes a e measu ed by hei o al popula ion (also
included in Penn Wo ld Table); o accoun o exis ing mig an ne -
wo ks, a signi ican de e minan o bila e al mig a ion lows, we con ol
o he size o a coun y’s diaspo a popula ion wi hin he 16 OECD
des ina ions in ou sample (based on decennial mig an s ocks published
by he Wo ld Bank and ¨
Ozden e al. (2011)). To inco po a e po en ial
shocks om con lic we use UCDP’s a med con lic da abase o cons uc
a ca ego ical a iable ha dis inguishes peace, mino con lic , and wa
(Pe e sson and Eck, 2018; Gledi sch e al., 2002). In addi ion, we also
con ol o a ying poli ical igh s and ci il libe ies using da a om
F eedom House (2018) (FH) ia Teo ell e al. (2019). In o de o accoun
o changing poli ical ends wi h espec o mig a ion, we include an
index om he In e na ional Mig a ion Policy in Compa ison P ojec
(IMPIC), measu ing he es ic i eness o mig a ion policies among he
OECD des ina ions conside ed in ou sample (Helbling e al., 2017);
las ly, we con ol o he changing cos o mig a ion du ing he s udy
pe iod by including he numbe o ai a el passenge s as a pe cen age
o wo ld popula ion in he model (Wo ld Bank, 2019). Fo a b ie
desc ip ion o he da a, Fig. 1 isualizes he mig a ion panel and Table 2
p o ides summa y s a is ics o all a iables we employ ac oss di e en
speci ica ions.
5
The di e ences be ween he g oups a e no sensi i e o he h eshold alue.
Table 10 in Appendix B ea u es di e en h eshold alues and he signi ican
g oup di e ence pe sis .
D. Benˇ
cek and C. Schneide heinze

Wo ld De elopmen 182 (2024) 106655
5
4. Empi ical analysis
4.1. Me hodology
The main objec i e o his pape is o es whe he he c oss-sec ional
inding o an in e se u-shaped ela ionship o mig a ion and de elop-
men holds o he sho e - e m ela ionship be ween economic g ow h
and emig a ion a he coun y le el. To in es iga e his ques ion we
employ a panel se up which exploi s only he a ia ion wi hin coun ies
o e ime. As a i s s ep, we eplica e he c oss-sec ional mig a ion
hump using ou da a. Fo one, his ensu es ha we can compa e ou
panel es ima es wi h p io c oss-sec ional analyses and po en ial dis-
c epancies do no simply esul om di e ences in da a sou ces. Fo
ano he , eplica ing he mig a ion hump enables us o iden i y he
c i ical income h eshold up o which emig a ion is hypo hesized o
inc ease and unca e ou sample acco dingly. As exis ing empi ical
s udies iden i y his u ning poin a di e en le els be ween 4000 and
13 000 USD, i is impo an o iden i y he upwa d-sloping ange o he
mig a ion hump o ou speci ic da a se . Hence, ou empi ical analysis
o he in luence o economic p og ess on emig a ion in poo coun ies
p oceeds in h ee s eps:
Employing he same me hodology as Clemens (2014), we ep oduce
he c oss-sec ional mig a ion hump wi h he OECD mig a ion da a se
compiled by Abu n and Wesselbaum (2019).
10 We unca e ou sample o coun ies and only include obse a-
ions whe e incomes ha e emained le o he c oss-sec ional peak
du ing he en i e obse a ion pe iod. This way we speci ically ocus on
he upwa d-sloping pa o he c oss-sec ional mig a ion hump whe e we
would expec a obus posi i e ela ionship be ween GDP and he
numbe o emig an s.
We es ima e a ange o ixed e ec s panel emig a ion models, which
a e based on he ecen li e a u e and aim a explaining changes in
emig a ion wi hin coun ies o e ime by changes in GDP and o he
con ol a iables.
Fo he co e o ou analysis (s ep 3), we employ a s aigh o wa d
panel emig a ion model o es i c oss-sec ional and panel es ima es o
he de elopmen – emig a ion nexus concu . The se up o ou model is
in luenced by Mayda (2010) and O ega and Pe i (2013). Depa ing
om hei se up, we only model emig a ion a he le el o o igin
coun ies ins ead o bila e al lows because we a e no in e es ed in he
des ina ion choice. Hence, we do no include des ina ion coun y ac-
o s. In ha sense ou empi ical model is e y simila o he “unila e al”
(o igin-coun y le el) model by B¨
ohme e al. (2019). While he decision
o model agg ega e emig a ion ins ead o bila e al lows is based on ou
2.00%
1.50%
1.00%
0.50%
A ica (N=50) Ame icas (N=34) Asia (N=47)
0.00%
2.00%
Eu ope (N=38) Oceania (N=3) 1980 1990 2000 2010
1.50%
1.00%
0.50%
0.00%
1980 1990 2000 2010 1980 1990 2000 2010
e a noi a gimE
Fig. 1. Emig a ion a es owa ds 16 OECD coun ies ac oss ime and con inen s. No e: Figu e 1 depic s he a ia ion in emig a ion a es owa ds 16 OECD des ina ion
coun ies ac oss con inen s and ime. The da a is based on he 2015 Re ision o he Uni ed Na ions’ Popula ion Di ision wi h he OECD’s mig a ion da abase and da a
om O ega and Pe i (2013) and was complied by Abu n & Wesselbaum (2019).
Table 2
Summa y s a is ics o he explana o y a iables.
S a is ic N Mean S . De . Min Pc l(25) Pc l(75) Max
Emig an s ( housands) 5, 768 17.34 37.05 0.00 0.81 18.05 949.10
Emig a ion a e (%) 5, 768 0.19 0.33 0.00 0.02 0.20 6.52
GDP (PPP billions $2011) 5, 768 324.89 1, 170.37 0.08 9.09 187.55 16, 395.20
GDP pe capi a (PPP $2011) 5, 768 12, 400.51 16, 194.58 223.09 2, 429.61 16, 822.63 215, 721.00
GDP pe capi a g ow h(%) 5, 745 2.53 9.12 −69.63 −1.11 6.17 142.68
Popula ion (millions) 5, 768 34.07 125.78 0.01 2.14 21.90 1, 382.79
Diaspo a (millions) 5, 768 0.33 0.77 0.00 0.01 0.30 13.12
Con lic 5, 768 1.20 0.50 1 1 1 3
FH index 5, 433 1.87 0.81 1.00 1.00 3.00 3.00
Ai passenge s 5, 768 25.52 7.71 15.41 19.50 30.60 42.99
Immig. pol. es ic i eness 5, 103 0.40 0.03 0.37 0.38 0.41 0.46
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
6
esea ch ques ion, i comes wi h he addi ional ad an age o ha ing
ha dly any ze os in he dependen a iable.
6
Ou main speci ica ion is
yi +1=
α
+βGDPi +γXi +δi+
τ
+
ε
i ,(1)
whe e y
i +1
deno es he numbe o emig an s om coun y i in yea +1.
GDP
i
is he main a iable o in e es and ep esen s o al GDP o a gi en
coun y and yea . X
i
ep esen s a se o con ol a iables ha a y o e
coun ies and ime. δ
i
and
τ
a e ec o s o coun y and yea ixed e ec s,
espec i ely, and
ε
i
ep esen s he e o e m.
In compa ison o c oss-sec ional eg essions, his panel se up is much
less likely o su e om omi ed a iable bias, since coun y and ime
ixed e ec s con ol o unobse ed he e ogenei y om ime-in a ian
ac o s such as geog aphical cha ac e is ics, language, na ional mig a-
ion na a i es o cul u al ela ions, which clea ly a ec mig a ion
lows. Hence, ou es ima es mo e likely ep esen a causal ela ionship.
Mo eo e , as he subsequen analysis e eals, adding di e en se s o
con ol a iables has e y li le impac on ou co e esul s.
In line wi h B¨
ohme e al. (2019), we ha e decided o model
emig a ion in absolu e e ms. Hence, we eg ess he absolu e numbe o
emig an s on absolu e GDP and con ol o popula ion size.
7
Using he
emig a ion a e and GDP pe capi a could impede he iden i ica ion o
he ue e ec o economic p og ess on emig a ion as, a leas in he
sho un, a ia ions in hese a ios may la gely be d i en by popula ion
g ow h. Mo eo e , popula ion g ow h exe s an in luence on emig a ion
beyond inc easing he pool o po en ial mig an s. I shapes he age
dis ibu ion wi hin coun ies, which a ec s a e age emig a ion p o-
pensi ies, and mo e populous coun ies yield highe oppo uni ies o
in e nal mig a ion and hus expe ience less emig a ion (de Haas e al.,
2018). Beyond ha , ou emig a ion da a ea u es he absolu e numbe o
emig an s o each coun y and yea as obse ed in des ina ion coun-
ies. Compu ing emig a ion a es om emig an numbe s and popula-
ion size isks in oducing a measu emen e o s emming om poo
quali y popula ion da a in de eloping coun ies.
We use explana o y a iables lagged by one yea on he igh hand
side o equa ion (1) in o de o accoun o ime-consuming p epa a ions
ha usually go along wi h mig a ion as well as o mi iga e issues o
e e se causali y.
Fo all high-magni ude a iables (i. e. emig an s, GDP, popula ion
and diaspo a) we use he in e se hype bolic sine (IHS) ans o ma ion
ins ead o a loga i hmic one in ou es ima ions. This has he ad an age
ha obse a ions wi h ze o alues do no need o be disca ded o al e ed
(by adding a cons an ) as IHS is de ined o any eal numbe (see Bu -
bidge & Magee, 1988; MacKinnon & Magee, 1990). A he same ime,
IHS e ains he p ope ies o a log ans o ma ion and we can in e p e
es ima ed coe icien s as pe cen age changes o elas ici ies (Pence,
2006; Bellema e & Wichman, 2019).
Ano he impo an ea u e o any long- e m mig a ion model is he
way o inco po a e ime ends in he easibili y o in e na ional
mig a ion. Global echnological p og ess in communica ion and ans-
po a ion likely dec eases mig a ion cos s o e ime; immig a ion pol-
icies change (de Haas e al., 2018). We use wo di e en app oaches.
Mos conse a i ely we include yea ixed e ec s. As an al e na i e o
he s ic yea ixed e ec s speci ica ion, we employ wo a iables ha
e lec mig a ion- ele an echnological and poli ical changes:
Dec easing anspo cos s and ease o a el a e app oxima ed by he
numbe o ai a el passenge s pe yea (as a pe cen age o wo ld
popula ion); changes in mig a ion policies a e e lec ed in he IMPIC-
index o es ic i eness o mig a ion policies among he des ina ion
coun ies.
Besides using yea ixed e ec s o abso b agg ega e changes o e
ime (which con ols o sudden global shi s in emig a ion), we ackle
he issue o yea ly luc ua ions on he igh hand side o equa ion (1) by
also speci ying a model based on 5-yea and 7-yea a e ages ha smoo h
he ime-se ies da a.
8
In Appendix D we pe o m a Mon e Ca lo Simula ion o demons a e
ha ou econome ic se up is well sui ed o iden i y he ela ionship
be ween economic g ow h and emig a ion in his speci ic da a s uc u e.
Speci ically we show ha ou se up ou pe o ms he mo e common pe
capi a speci ica ion, i emig a ion is di ec ly a ec ed by popula ion
g ow h o subjec o a ime end.
Based on he exis ing li e a u e, we would expec he panel es ima es
o esemble hei c oss-sec ional coun e pa s (Vogle & Ro e, 2000;
Telli, 2014; Clemens & Pos el, 2018; de Haas e al., 2018). Fo he
poo es coun ies we would expec a posi i e ela ionship be ween be-
ween GDP and emig a ion o OECD coun ies.
4.2. Resul s
E en hough ou mig a ion da a only ea u e OECD des ina ion
coun ies, we a e able o eplica e Clemens and Pos el (2018) c oss-
sec ional esul o a hump-shaped income–emig a ion ela ionship
e y closely in Fig. 2. Th oughou he decades, emig a ion peaks
somewhe e be ween 7000 and 14 000 USD pe capi a. Based on hese
es ima es, we es ic he da a in he es o ou analysis o hose coun-
ies wi h pe capi a incomes below 7000 USD h oughou he en i e
ime pe iod.
9
Tha lea es us wi h a balanced panel o 54 low-income coun ies. The
a e age pe capi a GDP o e he en i e ime pe iod is oughly 2000 USD,
mean annual economic g ow h and emig a ion a es equal 1.52 pe cen
and 0.06 pe cen , espec i ely. A comple e o e iew o summa y
3%
2%
1%
500 7,000 14,000 50,000
Ini ial GDP pe capi a (2011 PPP US$, log scale)
noi alupoplai ini/wol noi a gimeladaceD
Fig. 2. Non-pa ame ic eg ession o decadal emig a ion a es on ini ial eal
income pe capi a, 1980–2014. No e: Bold lines a e Nada aya-Wa son ke nel-
weigh ed local means (Epanechniko ke nel, bandwid h o 0.5 na u al log
poin s); anspa en lines depic a ying bandwid hs be ween 0.4 and 0.6 na -
u al log poin s; ini ial GDP pe capi a means a he beginning o he espec i e
decade; o co ec o he sho e 2010 decade in he da a we ha e scaled up he
es ima ed mig a ion low, allowing o a di ec compa ison.
6
Tha would bias es ima es unless accoun ed o , e.g., by using a poisson
pseudo-maximum likelihood es ima o . Gi en he absence o ze os in ou se up,
we can s ick o a linea panel model.
7
Ou esul s a e obus o modeling his ela ionship in pe capi a e ms (see
Table 11 in appendix C).
8
We e ain he lagged s uc u e o ou es ima ion by ma ching a e aged ime
pe iods ha a e shi ed by one yea , e.g. he a e age numbe o emig an s om
1981 o 85 eg essed on a e ages o ou RHS a iables om 1980 o 1984.
9
This sample es ic ion is based on da a exceeding he obse a ion pe iod o
ou analysis since Feens a e al. (2015) p o ide longe ime se ies o GDP and
popula ion da a. While his dis inc ion ba ely changes he se o coun ies unde
conside a ion and has no e ec on ou esul s, we choose o use all o he in-
o ma ion a ailable o us o es ic he sample.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
7
s a is ics o all ele an a iables is shown in Table 9 in Appendix B.
Ou main es ima ion esul s a e epo ed in Table 3, which i s in-
cludes wo pooled speci ica ions o equa ion (1) wi hou coun y ixed
e ec s as models 1 and 2. The es ima es a e in line wi h he c oss-
sec ional e idence in he mig a ion li e a u e and show a obus posi-
i e e ec o income on emig a ion ha co esponds o he upwa d-
sloping pa o he mig a ion hump. Hence, con olling o ime ends
and excluding all coun ies beyond 7000 USD o GDP pe capi a does no
change he posi i e c oss-sec ional ela ionship be ween GDP and
emig a ion o poo e de eloping coun ies. Explici ly including a ime
end based on global ai passenge s and OECD mig a ion policies e lec
he inc eased o al numbe o emig an s and a mild dampening e ec o
mig a ion es ic i eness, sugges ing a 1 pe cen dec ease in emig an s
be ween he mos open and he s ic es mig a ion policies obse ed in
he da a. The es ima ed e ec o income emains unchanged and co -
esponds o an inc ease o 0.8 pe cen in emig a ion wi h 1 pe cen GDP
g ow h. O e all, his pooled es ima ion is e y much in line wi h he
exis ing mig a ion hump e idence. Ye , he B eusch-Pagan es ad ises
agains he use o a pooled model due o he e oskedas ici y. The Haus-
man es a o s he ixed-e ec s es ima o .
Tu ning o he panel es ima es in models 3–6 o Table 3, we obse e
ha he c oss-sec ional ela ionship does no hold up a he coun y
le el. He e, ising incomes ac ually educe he o al numbe o emig an s
om a gi en coun y. This e ec is obus o he addi ion o ime- a ying
coun y-le el con ol a iables (models 5 and 6) as well as o using he
ime end a iables ins ead o ime ixed e ec s (models 4 and 6). I also
holds i we do no con ol o ime e ec s a all (see Table 12 in Appendix
C). In all cases, GDP g ow h a es o 1 pe cen educe emig a ion by
abou 0.5 pe cen . Ou es ima es o he e ec s o ins i u ional en i-
onmen s and he occu ence o iolen con lic show he expec ed signs:
Emig a ion inc eases as a med con lic s in ensi y, au oc a ic egimes
exhibi lowe emig a ion a es.
10
Mos impo an ly, hese esul s do no suppo a hump-shaped
income– emig a ion ela ionship and a he sugges ha economic
p og ess, on a e age, educes emig a ion owa ds OECD des ina ions.
Nex we in es iga e he sensi i i y o ou esul s o di e en sample
selec ions, i.e. we employ di e en GDP pe capi a h esholds a which
we unca e he sample. The ini ial h eshold (7000 USD pe capi a GDP)
is based on he c oss-sec ional peak and hus he co esponding sample is
well-sui ed o compa e c oss-sec ional and panel esul s. Howe e , by
design ou wo king sample is somewha unbalanced. I mainly consis s
o he poo es coun ies, obse a ions in he 4000 o 7000 USD income
ange a e unde ep esen ed. Inc easing he h eshold le el p o ides us
wi h a la ge sample size and addi ional coun y– yea obse a ions
along he inc easing segmen o he mig a ion hump. Fo example,
shi ing he cu -o le el om 7000 o 10 000 USD pe capi a gi es us 15
ex a coun ies and he a e age pe capi a GDP is s ill a below 7000.
Ye , i comes a he cos o including coun ies which ha e su passed he
peak in ecen yea s. Fo his sensi i i y es , we es ima e model 4 om
Table 3 wi h yea and coun y ixed e ec s on a ange o sub-samples
ha co espond o maximum GDP pe capi a h esholds be ween 4000
and 20 000 USD. The lowe bound is based on he lowes peak le el om
he espec i e li e a u e (Djajic e al., 2016). Ye , since ou ini ial sample
al eady comes wi h a low a e age income o abou 2000 USD, i is mo e
easonable o inc ease ou h eshold han o dec ease i .
We depic he es ima ed coe icien s o GDP and hei co esponding
95 pe cen con idence in e als in Fig. 3. This exe cise shows a signi i-
can ly nega i e associa ion s a ing om a h eshold le el o 4500 USD
pe capi a ha inc eases in size up o ou o iginal cu o poin o 7000
USD. A highe In Appendix C Figu e 4 we p o ide he esul s based on
Model 5 including he addi ional con ol a iables. Th esholds he
es ima e luc ua es sligh ly a ound he a e age alue o abou −0.5. The
changing size o he es ima ed coe icien hin s o somewha he e oge-
neous impac s ac oss coun ies. Tha is no su p ising as economic
p og ess may a ec he economic oppo uni ies o he espec i e pop-
ula ions di e en ly, and hus we should be ca e ul no o o e in e p e
he exac size o he coe icien s. Ye , and mo e impo an ly, he nega i e
ela ionship holds ac oss he whole cu -o ange and he size o he
es ima ed coe icien s does no change sys ema ically wi h income
le els.
Such agg ega e analysis migh s ill mask he e ogeneous ou comes
ac oss di e en coun ies since economic ajec o ies di e signi ican ly.
Mo e speci ically, he agg ega e analysis does no e eal whe he ou
es ima es a e pa icula ly d i en by high-g ow h o low-g ow h coun-
ies. Fo ins ance, he obse ed nega i e ela ionship be ween eco-
nomic g ow h and emig a ion migh be d i en by economic c ises
spu ing ou –mig a ion. To in es iga e he e ogeneous impac s ac oss
di e en le els o economic g ow h, we spli ou sample u he in o
high pe o me s and low pe o me s. The dis inc ion is made based on
he a e age GDP pe capi a g ow h (PPP) o e he en i e obse a ion
pe iod.
The speci ica ion is again iden ical o Model 4 in Table 3 including
coun y and ime ixed e ec s. We dis inguish ou di e en subse s o
coun ies o his analysis (p esen ed in Table 4): he low-pe o ming
coun ies wi h less han 1 pe cen a e age g ow h (column 1), o wi h
less han 2 pe cen (column 2); and he high-pe o ming coun ies wi h
mo e han 1 pe cen a e age eal economic g ow h (column 3), o wi h
mo e han 2 pe cen g ow h (column 4). An in e es ing pa e n eme ges:
Fo all bu he leas -pe o ming coun ies he associa ion be ween GDP
and he numbe o emig an s is again signi ican ly nega i e. The highe
he a e age economic g ow h, he highe is he es ima ed coe icien .
This makes in ui i e sense since low g ow h a es lea e mos ci izens
una ec ed in he sho - e m. In consequence, highe g ow hs a es a e
easie o pe cei e and hus may be mo e ele an o he mig a ion de-
cision. Mo eo e , he small and insigni ican coe icien o he wo s
pe o ming coun ies wi h e y li le economic p og ess (Column 1)
sugges s ha i is in ac economic g ow h discou aging emig a ion and
no ecessions spu ing emig a ion. To u he look in o his we es
ou lie dummies
11
o posi i e and nega i e g ow h yea s. The esul s
a e p o ided in Table 13 in Appendix C. No ably, hese dummies do no
e u n signi ican coe icien s and ha dly change he size o he gene al
ela ionship.
In s a k con as o mos o he p e ious li e a u e, ou indings hus
indica e a nega i e impac o ising incomes on emig a ion in poo
coun ies. We do no ind suppo o a non-linea ela ionship be ween
economic g ow h and emig a ion, i.e. a posi i e ela ionship a low
a e age incomes and a nega i e one a highe le els. Ins ead he nega-
i e ela ionship is independen o income le els. This sugges s ha he
mig a ion hump in he c oss-sec ion is due o omi ed a iables a he
coun y le el.
4.3. Robus ness checks
A emaining conce n wi h he obus ness o ou esul s may s em
om he use o annual da a. Clemens (2014) a gues ha in such kind o
panel analysis sho e m luc ua ions may o e shadow he ue ela-
ionship be ween economic de elopmen and emig a ion.
In o de o add ess hese conce ns we agg ega e ou da a in o i e
and se en yea ime in e als and un he same eg essions again (see
Table 5 Columns 3– 6). This way, ou es ima es a e much less ulne able
o sho e m luc ua ions in economic condi ions and mig a ion op-
po uni ies. Especially business cycle luc ua ions should ha e e y li le
10
Resul s emain unchanged when using Poli y IV da a ins ead o he F eedom
House index o measu e he ins i u ional amewo k in o igin coun ies (see
Table 15 in appendix C).
11
Speci ically, o each coun y we code yea s as posi i e g ow hs ou lie s i
he eal annual g ow h a e exceeds he a e age alue by a leas wo s anda d
de i a ions. Nega i e ou lie yea s a e compu ed in a simila ashion.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
8
impac on his speci ica ion. This es ima e can also be in e p e ed as he
mo e long- e m ela ionship be ween economic g ow h and emig a ion.
Na u ally, ha comes a he cos o educing he numbe o obse a ions
subs an ially, which isks insigni ican esul s. To u he in es iga e he
obus ness o ou ini ial es ima e we conside p i a e consump ion as an
al e na i e economic measu e. Especially o small coun ies, household
consump ion is o en conside ed o be a less ola ile wel a e measu e,
and i is less in luenced by exchange a e luc ua ions. The eg essions
a e p esen ed in Table 5 and suppo ou ini ial indings. Using house-
hold consump ion ins ead o GDP e u ns a sligh ly la ge coe icien ,
while he eg essions wi h agg ega ed da a yield somewha smalle co-
e icien s. Ye , he nega i e ela ionship be ween economic p og ess and
emig a ions is obus o hese al e a ions.
As a nex obus ness check, we epea ou es ima ion wi h di e en
se s o mig a ion da a (Table 6). In column 1 and 3, we u ilize he
bila e al mig a ion s ock da a p o ided by he Wo ld Bank (Wo ld Bank,
2018). This da a se comes wi h he addi ional ad an age o co e ing a
longe ime span, anging om 1960 o 2018. Mo eo e i co e s he ull
se o des ina ion coun ies.
Ye , in con as o ou main mig a ion da a, we only ge eigh poin s
in ime and inconsis en ime in e als. In o de o allow o a
Table 3
Main esul s: Pooled e sus panel eg essions.
Pooled Panel
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
GDP (PPP $2011) 0.775* 0.788* −0.532
***
−0.537
***
−0.481
***
−0.471
***
(0.339) (0.353) (0.067) (0.076) (0.059) (0.068)
Popula ion 0.222 0.219 1.591
***
2.673
***
0.921
***
1.768
***
(0.361) (0.372) (0.280) (0.290) (0.251) (0.268)
Ai passenge s 0.062
***
0.052
***
0.045
***
(0.015) (0.010) (0.009)
Immig. pol. es ic i eness −10.009
***
−4.157
***
−4.911
***
(1.930) (1.023) (0.949)
UCDP: Mino con lic 0.356
***
0.335
***
(0.055) (0.062)
UCDP: Wa 0.508
***
0.514
***
(0.076) (0.087)
FH: pa ly ee 0.055 −0.039
(0.072) (0.079)
FH: no ee −0.107 −0.250
**
(0.078) (0.086)
Diaspo a size 0.168
***
0.254
***
(0.038) (0.040)
Coun y FE no no yes yes yes yes
Yea FE yes no yes no yes no
Num. obs. 1858 1645 1858 1645 1769 1560
R
2
(o e all) 0.535 0.513 0.891 0.883 0.916 0.907
R
2
(wi hin) 0.457 0.513 0.047 0.522 0.096 0.565
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a o e he
en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he in e se hype bolic
sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical
a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .–.
0.00
-0.25
-0.50
5000 10000 15000 20000
GDP pe capi a h eshold
e ami se)1102$PPP(PDG
Fig. 3. Es ima ed coe icien o IHS- ans o med GDP (wi h 95 % con idence
in e al) condi ional on a ying GDP pe capi a h esholds o he unde lying
sample. No e: This igu e depic s coun y- ixed e ec s coe icien es ima es o
he ela ionship be ween GDP changes and changes in emig a ion o OECD
coun ies condi ional on he se o coun ies included. An x-axis alue o 10 000
indica es ha he sample consis o all coun ies ha do no exceed an income
le el o 10 000 USD a any momen in he obse a ion pe iod (1980–2014).
Mo ing o he igh inc eases he numbe o coun ies in he sample, mo ing o
he le depic s es ima es unde a s ic e exclusion c i e ion. he g ey a ea i-
sualizes he 955 con idence in e als. The es ima es o he in luence o GDP on
Emig a ion a e based on model 4 in Table 3.
Table 4
Panel es ima ion o a ying g ow h sub-samples.
max 1 %
g ow h
max 2 %
g ow h
min 1 %
g ow h
min 2 %
g ow h
GDP (PPP
$2011)
−0.206 −0.468
***
−0.599
***
−0.873
***
(0.132) (0.081) (0.091) (0.177)
Popula ion −0.850 −0.708* 2.726
***
5.810
***
(0.492) (0.353) (0.366) (0.533)
Coun y FE yes yes yes yes
Yea FE yes yes yes yes
Coun ies 18 36 36 18
Num. obs. 608 1238 1250 620
R
2
(o e all) 0.885 0.904 0.894 0.880
R
2
(wi hin) 0.017 0.046 0.070 0.197
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is Emig a ion ( o al numbe o emig an s). The
high magni ude a iables Emig a ion, GDP, Popula ions, and Diaspo a size a e
ans o med using he in e se hype bolic sine unc ion. The in e p e a ion o he
espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and
poli ical eedom (FH) a e cap u ed by ca ego ical a iables wi h h ee le els
each. Explana o y a iables a e lagged by one yea . The sample di e s ac oss
eg essions in his able, he selec ion is based on a e age GDP pe capi a g ow h
be ween 1980 and 2014.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
15
***
p <0.001,
**
p <0.01, *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a
o e he en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he
in e se hype bolic sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH) a e
cap u ed by ca ego ical a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .
Table 15
Panel eg essions wi h di e en ins i u ional a iables.
Model 1 Model 2 Model 3
GDP (PPP $2011) −0.481
***
−0.553
***
−0.423
***
(0.059) (0.061) (0.058)
Popula ion 0.921
***
1.257
***
1.136
***
(0.251) (0.247) (0.241)
UCDP: Mino con lic 0.356
***
0.231
***
0.292
***
(0.055) (0.060) (0.054)
UCDP: Wa 0.508
***
0.356
***
0.481
***
(0.076) (0.084) (0.074)
FH: pa ly ee 0.055
(0.072)
FH: no ee −0.107
(0.078)
Diaspo a size 0.168
***
0.191
***
0.212
***
(0.038) (0.038) (0.039)
Poli y IV 0.006
(0.005)
Poli ical Te o Sco e 0.177
***
(0.028)
Coun y FE yes yes yes
Yea FE yes yes yes
Num. obs. 1769 1721 1774
R
2
(o e all) 0.916 0.918 0.912
R
2
(wi hin) 0.096 0.141 0.090
***p <0.001; **p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela-
i ely poo coun ies wi h less han 7000 USD pe capi a o e he en i e obse a ion pe iod (1980 – 2014).
The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he
in e se hype bolic sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic
alues. Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical a iables wi h h ee le els
each. All explana o y a iables a e lagged by one yea .
Table 16
Panel eg essions: 5-yea a e ages wi h emig a ion a e and GDP pc.
Emig a ion a e Emig a ion a e (log)
Model 1 Model 2 Model 3 Model 4
GDP pe capi a −0.037
***
−0.034
**
−0.031
***
−0.029
***
(0.010) (0.011) (0.008) (0.009)
Ai passenge s 0.002
***
0.001
***
(0.000) (0.000)
Immig. pol. es ic i eness 0.109 0.068
(0.133) (0.106)
Coun y FE yes yes yes yes
Yea FE yes no yes no
Num. obs. 424 373 424 373
R
2
(o e all) 0.774 0.781 0.790 0.797
R
2
(wi hin) 0.039 0.050 0.042 0.063
***
p <0.001;
**
p <0.01; *p <0.05
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a
o e he en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he
in e se hype bolic sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH)
a e cap u ed by ca ego ical a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .
Table 17
Fixed e ec s s. Fi s di e ence eg essions.
Fixed E ec s Fi s Di e ence
Model 1 Model 2 Model 3 Model 4
GDP (PPP $2011) −0.359
**
−0.456
***
−0.381* −0.418
**
(0.133) (0.120) (0.150) (0.144)
Popula ion 2.177
***
1.373
**
2.978
***
2.768
***
(0.512) (0.468) (0.325) (0.345)
(con inued on nex page)
D. Benˇ
cek and C. Schneide heinze

Wo ld De elopmen 182 (2024) 106655
16
Table 17 (con inued)
Fixed E ec s Fi s Di e ence
Model 1 Model 2 Model 3 Model 4
UCDP: Mino con lic 0.388
***
−0.050
(0.097) (0.097)
UCDP: Wa 0.573
***
(0.134)
FH: pa ly ee −0.107 −0.318*
(0.153) (0.146)
FH: no ee −0.195 −0.503
**
(0.165) (0.156)
Diaspo a size 0.155* 0.011
(0.074) (0.046)
Coun y FE yes yes yes yes
Yea FE yes yes yes yes
Num. obs. 424 416 370 363
R
2
( ull model) 0.913 0.928
R
2
(p oj model) 0.063 0.134
R
2
0.010 0.034
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h
less han 7000 USD pe capi a o e he en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP,
popula ion, and diaspo a size a e ans o med using he in e se hype bolic sine unc ion. The in e p e a ion o he espec i e
coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical a iables wi h
h ee le els each. All explana o y a iables a e lagged by one yea .
Table 18
Pooled s. Panel eg ession: Using log ans o ma ion ins ead o IHS.
Pooled Panel
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
GDP (PPP $2011) 0.786* 0.799* −0.522
***
−0.527
***
−0.468
***
−0.459
***
(0.335) (0.349) (0.064) (0.073) (0.058) (0.067)
Popula ion 0.198 0.194 1.522
***
2.601
***
0.924
***
1.772
***
(0.355) (0.366) (0.269) (0.279) (0.247) (0.264)
Ai passenge s 0.061
***
0.052
***
0.044
***
(0.015) (0.010) (0.009)
Immig. pol. es ic i eness −9.677
***
−3.894
***
−4.686
***
(1.908) (0.983) (0.933)
UCDP: Mino con lic 0.351
***
0.331
***
(0.054) (0.061)
UCDP: Wa 0.499
***
0.505
***
(0.075) (0.085)
FH: pa ly ee 0.068 −0.023
(0.071) (0.078)
FH: no ee −0.079 −0.216*
(0.077) (0.084)
Diaspo a size 0.173
***
0.253
***
(0.038) (0.040)
Coun y FE no no yes yes yes yes
Yea FE yes no yes no yes no
Num. obs. 1858 1645 1858 1645 1769 1560
R
2
(o e all) 0.538 0.516 0.895 0.888 0.917 0.908
R
2
(wi hin) 0.460 0.516 0.048 0.531 0.096 0.567
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a o e he
en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables Emig a ion, GDP, Popula ions, and Diaspo a size a e ans o med using log ans o ma ion
Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .
Table 19
Pooled s. Panel eg ession: Al e na i e GDP measu e.
Pooled Panel
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
GDP UN (PPP $2011) 0.614* 0.624* −0.573
***
−0.572
***
−0.470
***
−0.464
***
(0.309) (0.312) (0.082) (0.094) (0.074) (0.086)
Popula ion 0.404 0.402 1.855
***
2.899
***
1.110
***
1.897
***
(0.319) (0.321) (0.289) (0.305) (0.259) (0.281)
Ai passenge s 0.068
***
0.048
***
0.041
***
(0.014) (0.011) (0.010)
Immig. pol. es ic i eness −9.546
***
−4.599
***
−5.406
***
(1.725) (1.053) (0.978)
(con inued on nex page)
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
17
Table 19 (con inued)
Pooled Panel
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
UCDP: Mino con lic 0.355
***
0.338
***
(0.057) (0.063)
UCDP: Wa 0.431
***
0.462
***
(0.082) (0.093)
FH: pa ly ee 0.080 −0.013
(0.073) (0.081)
FH: no ee −0.144 −0.296
***
(0.080) (0.088)
Diaspo a size 0.163
***
0.249
***
(0.040) (0.042)
Coun y FE no no yes yes yes yes
Yea FE yes no yes no yes no
Num. obs. 1783 1576 1783 1576 1697 1494
R
2
( ull model) 0.534 0.513 0.891 0.883 0.916 0.908
R
2
(p oj model) 0.452 0.513 0.044 0.520 0.089 0.563
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a o e he
en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he in e se hype bolic
sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical
a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .
Table 20
Pooled s. Panel eg ession: Al e na i e popula ion measu e.
Pooled Panel
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
GDP (PPP $2011) 0.823* 0.834* −0.622
***
−0.610
***
−0.478
***
−0.464
***
(0.324) (0.336) (0.064) (0.073) (0.060) (0.069)
Ai passenge s 0.058
***
0.083
***
0.056
***
(0.014) (0.010) (0.009)
Immig. pol. es ic i eness −9.088
***
−4.852
***
−5.604
***
(2.180) (0.971) (0.945)
UCDP: Mino con lic 0.353
***
0.334
***
(0.056) (0.062)
UCDP: Wa 0.513
***
0.516
***
(0.077) (0.087)
FH: pa ly ee 0.046 −0.037
(0.073) (0.080)
FH: no ee −0.132 −0.249
**
(0.080) (0.086)
Diaspo a size 0.179
***
0.265
***
(0.039) (0.041)
Coun y FE no no yes yes yes yes
Yea FE yes no yes no yes no
Num. obs. 1797 1635 1797 1635 1719 1560
R
2
( ull model) 0.537 0.518 0.895 0.886 0.916 0.906
R
2
(p oj model) 0.468 0.518 0.053 0.520 0.094 0.561
***
p <0.001;
**
p <0.01; *p <0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies wi h less han 7000 USD pe capi a o e he
en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he in e se hype bolic
sine unc ion. The in e p e a ion o he espec i e coe icien s is simila o loga i hmic alues. Con lic (UCDP) and poli ical eedom (FH) a e cap u ed by ca ego ical
a iables wi h h ee le els each. All explana o y a iables a e lagged by one yea .
Table 21
GMM es ima ion: Va ying popula ion lags.
Model 1 Model 2 Model 3
GDP −1.29*
(0.55)
−1.17*(0.57) −1.07*(0.52)
lag(Popula ion, 1) 10.44
(14.63)
lag(Popula ion, 2) 11.08
(12.02)
lag(Popula ion, 3) 16.54(14.37)
n 54 54 54
T 35 35 35
Num. obs. 1858 1858 1858
Num. obs. used 1750 1696 1642
Sa gan Tes : chisq 32.08 30.80 30.83
Sa gan Tes : d 159.00 157.00 154.00
(con inued on nex page)
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
18
Table 21 (con inued)
Model 1 Model 2 Model 3
Sa gan Tes : p- alue 1.00 1.00 1.00
Wald Tes Coe icien s: chisq 5.70 4.30 4.58
Wald Tes Coe icien s: d 2 2 2
Wald Tes Coe icien s: p- alue 0.06 0.12 0.10
Wald Tes Time Dummies: chisq 539.81 506.46 473.47
Wald Tes Time Dummies: d 33 32 31
Wald Tes Time Dummies: p- alue 0.00 0.00 0.00
***p < 0.001; **p < 0.01; *p < 0.05.
No e: The dependen a iable is emig a ion ( o al numbe o emig an s). The sample consis s o 54 ela i ely poo coun ies
wi h less han 7000 USD pe capi a o e he en i e obse a ion pe iod (1980 – 2014). The high magni ude a iables
emig a ion, GDP, popula ion, and diaspo a size a e ans o med using he in e se hype bolic sine unc ion. Con lic (UCDP)
and poli ical eedom (FH) a e cap u ed by ca ego ical a iables wi h h ee le els each. All explana o y a iables a e lagged
by one yea .
Compa ing di e en empi ical speci ica ions: A Mon e Ca lo simula ion
This pape ea u es a dis inc empi ical se up. In sligh con as o mos o he pape s in his ield we decided o use absolu e emig a ion as he
dependen a iable and con ol o absolu e GDP and Popula ion sepa a ely – ins ead o eg essing emig a ion a es on GDP pe capi a. W he eby
wan o be e accoun o he in luence o demog aphic ac o s on emig a ion and educe he likelihood o measu emen e o (see sec ion 4.1 o a
mo e de ailed discussion on hese issues). In his sec ion we use a basic, ye powe ul, Mon e Ca lo simula ion o compa e di e en empi ical se ups
ac oss di e en scena ios. This is o some ex en mo i a ed by ecen conce ns ha ou empi ical speci ica ion migh p oduce spu ious esul s
(Clemens, 2020b). Speci ically, Clemens (2020b) a gues ha including ime ixed e ec s could mask he ue ela ionship be ween pe capi a incomes
and emig a ion, and ha wo non-s a iona y eg esso s (GDP and Popula ion) would gi e ise o spu ious co ela ion. This simula ion exe cise is based
on he same eal wo ld GDP and Popula ion da a o 54 coun ies o e 35 yea s ha was used h oughou he pape (sou ce: Wo ld Penn Tables).
12
In
consequence, any issues ela ed o unde lying ime ends and a non-s a iona y ela ionship be ween hese wo a iables a e well-cap u ed. We han
simula e emig a ion lows as a unc ion o GDP pe capi a. No e ha GDP pe capi a is non-s a iona y and consequen ly he simula ed emig a ion
a iable is non-s a iona y as well. Fo his ask, we employ annual da a because a his highe equency ou es ima es a e mos ulne able o biases
associa ed wi h non-s a iona i y.
Speci ically we compu e emig a ion lows acco ding o ou di e en scena -ios:
•In he i s scena io emig a ion only depends on GDP pe capi a and an e o e m.
Scena io 1:
Emîg1i =β∗GDPpci +
μ
i ,(2)
•Fo he second scena io we assume popula ion o posi i ely in luence emig a ion. High popula ion g ow h migh pu p essu e on local labo
ma ke s and is ela ed o young popula ions, which plausibly inc eases a e age mig a ion p opensi ies.
Scena io 2:
Emîg2i =β∗GDPpci +γ∗Popi +
μ
i (3)
In scena io h ee, we assume a linea posi i e ime end o in e na ional mig a ion. Real wo ld mig a ion ends, anspo in as uc u e
de elopmen and an a e age global libe aliza ion o immig a ion policies make his a ealis ic assump ion.
Scena io 3:
Emîg3i =β∗GDPpci +δ∗
τ
+
μ
i ,(4)
•In he o h scena io, we induce au o-co ela ion. Tha is in line wi h he impo ance o ne wo ks o mig a ion.
Scena io 4:
Emîg4i =β∗GDPpci +E∗Emîg1i −1+
μ
i ,(5)
95 pe cen con idence in e als o he GDPpc es ima es om he espec i e eg essions. The i s wo boxplo s ep esen es ima ion esul s om
ou absolu e eg ession se up wi h and wi hou ime ixed-e ec s. The hi d and ou h boxplo in each panel a e based on es ima es om he
eg ession in pe capi a e ms (wi h and wi hou ime ixed-e ec s).
In he i s scena io (wi hou any addi ional in luence) he esul s a e e y simila ac oss he di e en me hods (uppe le panel). Despi e he non-
s a iona y na u e o all ou a iables o in e es all ou speci ica ions p oduce on a e age unbiased es ima es. Mo eo e , we do no de ec any dis-
o ions esul ing om he inclusion o ime ixed e ec s. In scena io wo i s sys ema ic di e ences become appa en (uppe igh panel). While ou
wo king model s ill pe o ms well, he pe capi a speci ica ion p oduces signi ican biases. Wi h ime ixed e ec s we obse e a downwa ds bias,
wi hou ime ixed e ec s an e en la ge upwa ds bias appea s. The hi d scena io shows he c ucial need o ime- ixed e ec s (bo om le panel). I
12
Simila o he es o he pape we use he in e se hype bolic sine unc ion o ans o m hese high magni ude a iables.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
19
we do no con ol o a ime end in in e na ional mig a ion ia ime ixed-e ec s, bo h models p oduce upwa ds biased es ima es. Wi h ime- ixed
e ec s bo h speci ica ions e ain unbiased es ima es again. A simila lesson can be lea ned om scena io ou (bo om igh panel). Time- ixed e ec s
e en help agains au oco ela ion in he dependen a iable. Al oge he , his simula ion ask p o ides aluable insigh s in o s eng hs and weaknesses
o di e en empi ical se ups in di e en scena ios. Omi ing ime ixed-e ec s c ea es subs an ial biases i emig a ion lows a e non-s a iona y. Using
he same GDP and Popula ion da a as in ou co e analysis, we do no ind any e idence o a highe isk o spu ious esul s wi h ou empi ical
app oach. To he con a y, ou speci ica ion clea ly ou pe o ms he mo e s anda d pe -capi a se up in wo o he ou scena ios.
Re e ences
Abu n, A., & Wesselbaum, D. (2019). Gone wi h he wind: In e na ional mig a ion gone
wi h he wind: in e na ional mig a ion. Global and Plane a y Change, 178, 96–109.
Ake man, Sune (1976). “Theo ies and me hods o mig a ion esea ch”. In: F om Sweden
o Ame ica, pp. 19–75.
Bazzi, S. (2017). Weal h he e ogenei y and he income elas ici y o mig a ion. Ame ican
Economic Jou nal: Applied Economics, 9(2), 219–255. h ps://doi.o g/10.1257/
app.20150548
Beine, M., Docquie , F., & Rapopo , H. (2001). B ain d ain and economic g ow h:
Theo y and e idence. Jou nal o De elopmen Economics, 64(1), 275–289. h ps://doi.
o g/10.1016/S0304-3878(00)00133-4
Bellema e, M. F., & Wichman, C. J. (2019). Elas ici ies and he in e se hype bolic sine
ans o ma ion. Ox o d Bulle in o Economics and S a is ics. h ps://doi.o g/10.1111/
obes.12325
Be hiaume, Nicolas, Naomi Lee mans, Nienke Oomes, Hugo Rojas-Romagosa, and
Tobias Ve lie (2021). “A Reapp aisal o he Mig a ion-De elopmen Nexus Tes ing
he Robus ness o he Mig a ion T ansi ion Hypo hesis”. In: Janua y.
B¨
ohme, Ma cus H., And ´
e G ¨
oge , and Tobias S ¨
oh (2019). “Sea ching o a be e li e:
P edic ing in e na ional mig a ion wi h online sea ch keywo ds”. In: Jou nal o
De elopmen Economics Ap il. doi: 10.1016/j.jde eco.2019. 04.002.
B ücke , He be , S ella Capuano, and Abdeslam Ma ouk (2013). “Educa ion, gende
and in e na ional mig a ion: Insigh s om a panel da ase 19802010”. In: Mimeo
Augus , p. 13. u l: h p://www.iab.de/en/da en/iabb ain-d ain-da a.aspx.
Bu bidge, John B., Lonnie Magee, and A. Leslie Robb (1988). “Al e na i e
T ans o ma ions o Handle Ex eme Values o he Dependen Va iable”. In: Jou nal o
he Ame ican S a is ical Associa ion 83.401, pp. 123–127. issn: 0162-1459. doi:
10.1080/01621459.1988.10478575.
Ca ling, J., & Schewel, K. (2018). Re isi ing aspi a ion and abili y in in e na ional
mig a ion. Jou nal o E hnic and Mig a ion S udies, 44(6), 945–963. h ps://doi.o g/
10.1080/1369183X.2017.1384146
Ca aneo, C., & Pe i, G. (2016). The mig a ion esponse o inc easing empe a u es.
Jou nal o De elopmen Economics, 122, 127–146. h ps://doi.o g/10.1016/j.
jde eco.2016.05.004
Clemens, Michael A. (2014). “Does De elopmen Reduce Mig a ion?” In: In e na ional
Handbook on Mig a ion and Economic De elopmen 8592, pp. 152–185. doi: 10.4337/
9781782548072.00010.
Clemens, Michael A (2020a). “Mig a ion om De eloping Coun ies: Selec ion, Income
Elas ici y and Simpson’s Pa adox”. In: IZA Discussion Pape s No. 13612. doi: 10.
2139/ss n.3688065.
Clemens, Michael A (2020b). “The Emig a ion Li e Cycle: How De elopmen Shapes
Emig a ion om Poo Coun ies”. In: IZA Discussion Pape s No. 13614. doi: 10.2139/
ss n.3679020.
Clemens, M. A., & Pos el, H. M. (2018). De e ing emig a ion wi h o eign aid: An
o e iew o e idence om low-income coun ies. Popula ion and De elopmen Re iew,
4, 667–693. h ps://doi.o g/10.1111/pad .12184
Czaika, Ma hias and Mogens Hobol h (2016). “Do es ic i e asylum and isa policies
inc ease i egula mig a ion in o Eu ope?” In: Eu opean Union Poli ics 17.3, pp.
345–365.
Dao, Thu Hien, F ´
ed´
e ic Docquie , Ch is Pa sons, and Gio anni Pe i (2018). “Mig a ion
and de elopmen : Dissec ing he ana omy o he mobili y ansi ion”. In: Jou nal o
De elopmen Economics 132.2018, pp. 88–101. doi: 10.1016/j.jde eco.2017.12.003.
de Haas, Hein (2010). “Mig a ion ansi ions: a heo e ical and empi ical inqui y in o he
de elopmen al d i e s o in e na ional mig a ion”. In: DEMIG p ojec pape 1.
de Haas, Hein (2019). Why De elopmen Will No S op Mig a ion. u l: h ps://www.macm
illanihe.com/blog/pos /why-de elopmen -will-no -s op-mig a ion-hein-de-haas/.
de Haas, Hein, Ma hias Czaika, Ma ie-Lau ence Flahaux, Edo Mahend a, Ka ha ina
Na e , Simona Vezzoli, and Ma ia Villa es-Va ela (2018). “In e na ional Mig a ion.
T ends, De e minan s and Policy.” In: IMI Wo king Pape Se ies 142, pp. 1–59.
Deininge , K., & Squi e, L. (1998). New ways o looking a old issues: Inequali y and
g ow h. Jou nal o De elopmen Economics, 57(2), 259–287. h ps://doi.o g/10.1016/
S0304-3878(98)00099-6
Djajic, S., Ki da , M. G., & Vinog ado a, A. (2016). Sou ce coun y ea nings and
emig a ion. Jou nal o In e na ional Economics, 99, 46–67. h ps://doi.o g/10.1016/j.
jin eco.2015.12.001
Docquie , F., Pe i, G., & Ruyssen, I. (2014). The c oss-coun y de e minan s o po en ial
and ac ual mig a ion. In e na ional Mig a ion Re iew, 48(s1), S37–S99. h ps://doi.
o g/10.1111/im e.12137
Dus mann, C., & Oka enko, A. (2014). Ou -mig a ion, weal h cons ain s, and he quali y
o local ameni ies. Jou nal o De elopmen Economics, 110, 52–63. h ps://doi.o g/
10.1016/j.jde eco.2014.05.008
Eas e lin, R. A. (1961). In luences in Eu opean o e seas emig a ion be o e wo ld wa I.
Economic De elopmen and Cul u al Change, 9(3), 331–351. h ps://doi.o g/10.1086/
449911
EC (2016). Mig a ion Pa ne ship F amewo k: A New App oach o Be e Manage Mig a ion.
Tech. ep. S asbou g: Eu opean Commission. u l: h ps://eeas.eu opa.eu/si es/eeas
/ iles/ ac shee %7B%5C_%7Dec%7B%5C_%7D o ma %7B%5C_%7Dmig a ion%7B
%5C_%7Dpa ne ship%7B%5C_%7D a%20mewo k%7B%5C_%7Dupda e%7B%5C_
%7D2.pd .
Eps ein, G. S. (2008). He d and ne wo k e ec s in mig a ion decision making. Jou nal o
E hnic and Mig a ion S udies, 34(4), 567–583. h ps://doi.o g/10.1080/
13691830801961597
Esipo a, Neli, Ani a Pugliese, and Julie Ray (2018). Mo e Than 750 Million Wo ldwide
Would Mig a e I They Could. Washing on D.C. u l: h ps:// news.gallup.com/poll/24
5255/750-million-wo ldwide-mig a e.aspx.
Eu opean Commission (2018). Many mo e o come? Mig a ion om and wi hin A ica.
Tech. ep. Luxembou g, Join Resea ch Cen e , Publica ions O ice o he Eu opean
Union, p. 36. doi: 10.2760/1702.
Faini, Ricca do and Alessand a Ven u ini (1994). “Mig a ion and g ow h: he expe ience
o Sou he n Eu ope”. In: CEPR Discussion Pape s 964.
Falco, Chia a, Ma zio Galeo i, and Alessand o Olpe (2019). “Clima e change and
mig a ion: Is ag icul u e he main channel?” In: Global En i onmen al Change 59.59.
issn: 09593780. doi: 10.1016/j.gloen cha.2019.101995.
Feens a, Robe C., Robe Inklaa , and Ma cel P. Timme (Oc . 2015). “The Nex
Gene a ion o he Penn Wo ld Table”. In: Ame ican Economic Re iew 105.10, pp.
3150–3182. issn: 0002-8282. doi: 10.1257/ae .20130954.
Field, G. (2002). Dis ibu ion and de elopmen : A new look a he de eloping wo ld. MIT
P ess.
F eedom House (2018). F eedom in he Wo ld 2018. u l: h ps:// eedomhouse.o g/%20 e
po - ypes/ eedom-wo ld.
Giuliano, P., & Ruiz-A anz, M. (2006). Remi ances, inancial de elopmen , and g ow h.
Jou nal o De elopmen Economics, 90(1), 144–152. h p://hdl.handle.ne /10
419/34035.
Gledi sch, Nils Pe e , Pe e Wallens een, Mikael E iksson, Ma ga e a Sollenbe g, and
Hå a d S and (2002). “A med Con lic 1946-2001: A New Da ase ”. In: Jou nal o
Peace Resea ch 39.5, pp. 615–637. issn: 0022-3433. u l: h ps://www.js o .o g/s abl
e/1555346 ( isi ed on 02/13/2019).
Gould, J. D. (1979). “Eu opean in e -con inen al emig a ion 1815-1914: Pa e ns and
causes”. In: Jou nal o Eu opean Economic His o y 8.3, p. 593. doi: h ps://sea ch.p o
ques .com/doc iew/1292863303?accoun id=200654.
G eenwood, M. J. (1969). An analysis o he de e minan s o geog aphic labo mobili y
in he Uni ed S a es. The Re iew o Economics and S a is ics. h ps://www.js o .o g/
s able/1926728.
Haans, R. F. J., Pie e s, C., & He, Z.-L. (2016). Thinking abou U: heo izing and es ing
U- and in e ed U-shaped ela ionships in s a egy esea ch. S a egic Managemen
Jou nal, 37(7), 1177–1195.
Hanson, G., & McIn osh, C. (2016). Is he medi e anean he new Rio G ande? US and EU
immig a ion p essu es in he Long Run. Jou nal o Economic Pe spec i es, 30(4),
57–82. h ps://doi.o g/10.1257/jep.30.4.57
Ha is, John R. and Michael P. Toda o (1970). “Mig a ion, Unemploymen and
De elopmnen : A Two-Sec o Analysis”. In: The Ame ican Economic Re iew 60.1, pp.
126–142. u l: h ps://www.js o .o g/s able/1807860.
Ha on, T. J., & Williamson, J. G. (2005). Wha undamen als d i e wo ld mig a ion ?
Jou nal o Popula ion Economics, 18, 719–740.
Ha on, T. J., & Williamson, J. G. (1994). Wha d o e he mass mig a ions om Eu ope in
he la e nine een h cen u y? Popula ion and De elopmen Re iew, 20(3), 533–559.
Helbling, M., Bje e, L., R¨
ome , F., & Zobel, M. (2017). Measu ing immig a ion policies:
The IMPIC Da abase. Eu opean Poli ical Science, 16(1), 79–98. h ps://doi.o g/
10.1057/eps.2016.4. issn: 1680–4333, 1682–0983.
Hooghe, Liesbe and Ga y Ma ks (2018). “Clea age heo y mee s Eu ope’s c ises: Lipse ,
Rokkan, and he ansna ional clea age”. In: Jou nal o Eu opean Public Policy 25.1,
pp. 109–135. doi: 10.1080/13501763.2017.1310279.
Ingleha , Ronald and Pippa No is (2016). “T ump, B exi , and he Rise o Populism:
Economic Ha e-No s and Cul u al Backlash”. In: SSRN Elec onic Jou nal, pp. 1–53.
doi: 10.2139/ss n.2818659.
Kuzne s, S. (1955). Economic g ow h and income inequali y. The Ame ican Economic
Re iew, 45(1), 1–28.
Lana i, M., & Thiele, R. (2018). The impac o o eign aid on mig a ion e isi ed. Wo ld
De elopmen , 111, 59–74. h ps://doi.o g/10.1016/j. wo ldde .2018.06.021
Langella, Monica and Alan Manning (2021). “Income and he desi e o mig a e”. In:
Discussion Pape No.1794, pp. 1–52. u l: h ps://news.gallup.com/poll/ 24
5255/750-million-wo ldwide-mig a e.aspx.
Lind, J. T., & Mehlum, H. (2010). Wi h o wi hou u? The app op ia e es o a U-shaped
ela ionship. Ox o d Bulle in o Economics and S a is ics, 72(1), 109–118. h ps://doi.
o g/10.1111/j.1468-0084.2009.00569.x
Lucas, Robe E. B. (2019). Mig a ion and de elopmen The ole o de elopmen aid, p. 112.
isbn: 9789188021403. u l: h ps://eba.se/wp-con en /uploads/2019/08/20195.pd
%7B%5C%%7D0Ah ps://www.delmi.se/en.
D. Benˇ
cek and C. Schneide heinze
Wo ld De elopmen 182 (2024) 106655
20
MacKinnon, James G. and Lonnie Magee (1990). “T ans o ming he Dependen Va iable
in Reg ession Models”. In: In e na ional Economic Re iew 31.2, pp. 315–339. issn:
0020-6598. doi: 10.2307/2526842.
Mac on, Emmanuel (2018). Ini ia i e o Eu ope: So bonne speech o Emmanuel Mac on.
Tech. ep.
Ma in, P. L. (1993). T ade and mig a ion: NAFTA and ag icul u e. USA: Policy ana.
Massey, D. S., A ango, J., & Hugo, G. (1993). Theo ies o in e na ional mig a ion: A
e iew and app aisal. Popula ion and De elopmen Re iew, 19(3), 431–466. h ps://
www.js o .o g/s able/2938462.
Mayda, A. M. (2010). In e na ional mig a ion: A panel da a analysis o he de e minan s
o bila e al lows. Jou nal o Popula ion Economics, 23(4), 1249–1274. h ps://doi.
o g/10.1007/s00148-009-0251-x
Maye , T. and S. Zignago (2011). “No es on CEPII’s dis ances measu es: he GeoDis
Da abase”. In: CEPII Wo king Pape 2011-25.
Migali, S., & Scipioni, M. (2019). Who’s abou o lea e? A global su ey o aspi a ions
and in en ions o mig a e. In e na ional Mig a ion, 57(5), 181–200. h ps://doi.o g/
10.1111/imig.12617
Naud´
e, W. (2010). The de e minan s o mig a ion om sub-saha an a ican coun ies.
Jou nal o A ican Economies, 19(3), 330–356. h ps://doi.o g/10.1093/jae/ejq004
OECD. (2019). In e na ional mig a ion ou look 2019 (43 d ed.). Pa is: OECD Publishing.
O ega, F., & Pe i, G. (2013). The e ec o isa policies on in e na ional mig a ion lows.
Mig a ion S udies, 1(1), 47–74. h ps://doi.o g/10.1093/mig a ion/mns004
¨
Ozden, Ça˘
gla , Ch is ophe R. Pa sons, Mau ice Schi , and Te ie L. Walmsley (Jan. 1,
2011). “Whe e on Ea h Is E e ybody? The E olu ion o Global Bila e al Mig a ion
1960–2000”. In: The Wo ld Bank Economic Re iew 25.1, pp. 12–56. issn: 0258-6770.
doi: 10.1093/wbe /lh 024.
Pence, Ka en M. (2006). “The Role o Weal h T ans o ma ions: An Applica ion o
Es ima ing he E ec o Tax Incen i es on Sa ing”. In: The B.E. Jou nal o Economic
Analysis & Policy 5.1. issn: 1935-1682. doi: 10.1515/15380645.1430.
Pe e sson, The ´
ese and K is ine Eck (July 1, 2018). “O ganized Violence, 1989–2017”.
In: Jou nal o Peace Resea ch 55.4, pp. 535–547. issn: 0022-3433. doi: 10. 1177/
0022343318784101.
Rikani, Albano, Ka ja F iele , and Jacob Schewe (2022). “Clima e change and
in e na ional mig a ion: Explo ing he mac oeconomic channel”. In: PLoS ONE 17.11
No embe , pp. 1–25. issn: 19326203. doi: 10.1371/jou nal.pone. 0276764.
Ruyssen, Ilse, Ge die E e ae , and Glenn Rayp (2012). “De e minan s and dynamics o
mig a ion o OECD coun ies in a h ee-dimensional panel amewo k”. In: Empi ical
Economics 46.1, pp. 175–197. doi: 10.1007/s00181-012-0674-1.
Simonsohn, U i (2018). “Two Lines: A Valid Al e na i e o he In alid Tes ing o U-
Shaped Rela ionships Wi h Quad a ic Reg essions”. In: Ad ances in Me hods and
P ac ices in Psychological Science 1.4, pp. 538–555. doi: 10. 1177/
2515245918805755.
S a k, O. (2006). Inequali y and mig a ion: A beha io al link. In: Economics le e s, 91(1),
146–152.
Telli, H. (2014). Less po e y, mo e emig a ion: Unde s anding mig an lows om
de eloping coun ies. Mig a ion and De elopmen , 3(1), 54–72. h ps://doi.o g/
10.1080/21632324.2013.831584
Teo ell, Jan, S e an Dahlbe g, S¨
o en Holmbe g, Bo Ro hs ein, Na alia Al a ado Pachon,
and Richa d S ensson (2019). The Quali y o Go e nmen S anda d Da ase , Ve sion
Jan19. Uni e si y o Go henbu g. The Quali y o Go e nmen Ins i u e. doi:
10.18157/qogs djan19.
UN (2018). Global Compac o Sa e, O de ly, and Regula Mig a ion D a Re 2. Tech. ep.
UN (2019). “Uni ed Na ions, Depa men o Economic and Social A ai s, Popula ion
Di ision (2019). In e na ional Mig an S ock 2019”. In: Uni ed Na ions da abase,
POP/DB/MIG/S ock/Re .2019.
Vogle , M., & Ro e, R. (2000). The e ec s o de elopmen on mig a ion: Theo e ical
issues and new empi ical e idence. Jou nal o Popula ion Economics, 13(3), 485–508.
Wesselbaum, Dennis and Amelia Abu n (2019). “Gone wi h he Wind: In e na ional
Mig a ion”. In: Global and Plane a y Change 178, pp. 96–109. issn: 0921-8181. doi:
10.1016/j.gloplacha.2019.04.008.
Wo ld Bank (2018). “Wo ld Bank Bila e al Mig a ion Ma ix 2017”. In: Mig a ion and
Remi ances Da a. u l: h ps://www.wo ldbank.o g/en/ opic/mig a ion emi anc
esdiaspo aissues/b ie /mig a ion- emi ances-da a.
Wo ld Bank (2019). “Ai anspo , egis e ed ca ie depa u es wo ldwide”. In: Wo ld
De elopmen Indica o s. u l: h ps://da a.wo ldbank .o g/indica o /IS.AIR.DPRT.
Zelinsky, W. (1971). The hypo hesis o he mobili y ansi ion. Geog aphical Re iew, 61
(2), 219. h ps://doi.o g/10.2307/213996
D. Benˇ
cek and C. Schneide heinze