1
Global c oss-coun y in es men ela ionships –
using ne wo k analysis and in e ac i e
isualiza ion echniques
Bojan S a ikj
Disse a ion p esen ed as pa ial equi emen o ob aining he
Mas e ’s deg ee in Da a Science and Ad anced Analy ics,
Majo in Business Analy ics
2
NOVA In o ma ion Managemen School
Ins i u o Supe io de Es a ís ica e Ges ão de In o mação
Uni e sidade No a de Lisboa
Global c oss-coun y in es men ela ionships – using ne wo k
analysis and in e ac i e isualiza ion echniques
by
Bojan S a ikj
Disse a ion p esen ed as pa ial equi emen o ob aining he Mas e ’s deg ee in Da a Science
and Ad anced Analy ics, Majo in Business Analy ics
Co-supe iso s: Dou o Flá io Luís Po as Pinhei o & Dou o João Miguel Falcão Pin o da Sil a
No embe 2021
3
Abs ac . One impo an ea u e o he globaliza ion p ocess is he inc ease in he economic and
inancial in e dependencies ac oss coun ies. In his con ex , i is c ucial o measu e he connec ions
be ween coun ies in o de o iden i y whe e he inancial cen e s a e loca ed, o cha ac e ize he
o eign in es men s, and de ine he se o wo ld coun ies ha ha e s onge linkages, among
o he s. In his con ex , o eign di ec and po olio in es men s play a c ucial ole in measu ing
in e na ional in es men s and unde s anding hese dynamics. Such an in e connec ed web o
o eign in es men ela ionships is di icul o measu e due o hei complexi y, bu as well as he
lack o uni ied da a sou ces. This hesis aims o use ne wo k analysis o map he po olio and
o eign di ec in es men global ela ionships in o de o iden i y pa e ns, p e e en ial pa hs o
in es men , es ablish ends and desc ibe he ela ions be ween coun ies o e ime. Secondly, i
ga he s he esul s o he ne wo k analysis and p esen s hem in an in ui i e web applica ion, whe e
he mos impo an indings a e highligh ed allowing he use s o in e ac wi h he da a and ex ac
insigh s o e all he a ailable yea s.
Keywo ds. Fo eign di ec in es men , Po olio in es men , Ne wo k analysis, In e ac i e
web applica ion
4
Table o Con en s
1 INTRODUCTION .................................................................................................................................. 5
2 LITERATURE REVIEW ON THE ECONOMIC PERSPECTIVE & NETWORK ANALYSIS ................................... 10
3 DATA ................................................................................................................................................ 16
4 NETWORK ANALYSIS ......................................................................................................................... 19
5 HYPOTHESIS FORMULATION.............................................................................................................. 25
6 RESULTS ........................................................................................................................................... 26
CDIS INWARD NETWORK ......................................................................................................................................... 28
PORTFOLIO LIABILITIES INVESTMENT NETWORKS .......................................................................................................... 32
COMBINED LIABILITIES NETWORK .............................................................................................................................. 35
SUMMARY ............................................................................................................................................................. 38
7 THE PLATFORM ................................................................................................................................. 39
THE STACK ............................................................................................................................................................ 39
NARRATIVE ............................................................................................................................................................ 41
7.1.1 Landing Page ................................................................................................................................... 42
7.1.2 Ne wo k Page .................................................................................................................................. 43
7.1.3 Me hodology ................................................................................................................................... 48
7.1.4 Abou and Downloads ..................................................................................................................... 48
8 CONCLUSIONS .................................................................................................................................. 49
9 BIBLIOGRAPHY.................................................................................................................................. 52
5
1 In oduc ion
Wi h he e e -g owing globaliza ion, an inc ease in ade, and he possibili y o in es ing in
di e en coun ies a he ip o ou inge s, i is becoming inc easingly di icul o ack he low
o money be ween coun ies. In such an in e connec ed web o ela ionships cha ac e ized by many
playe s, ma ke s, and in es men oppo uni ies, i is complex o map all he linkages be ween he
o igin and des ina ion o each in es men .
The ex e nal s a is ics compiled by cen al banks is an impo an ool o moni o he economic
ope a ions be ween one coun y and he es o he wo ld. This is c ucial o policymake s when
e alua ing he exis ence o in es men oppo uni ies, inancial/cu ency c ises, he
o igins/des ina ions o ex e nal in es men s among o he in e na ional ela ionships. In his con ex ,
he in e na ional accoun s aim o ack all in e na ional mone a y and non-mone a y ansac ions
du ing a gi en pe iod. The Balance o Paymen s Manual 6 h edi ion (BPM6) om he In e na ional
Mone a y Fund (IMF) aims o es ablish a s anda d amewo k o s a is ics o eco ding
ansac ions and posi ions be ween a coun y and he es o he wo ld.
Acco ding o he in e na ional s anda ds, he Balance o Paymen s (BoP) is composed by h ee
main componen s - he cu en accoun , he capi al accoun , and he inancial accoun . The cu en
accoun deals wi h he ade o goods and se ices, p ima y income, and seconda y income; he
capi al accoun deals wi h he low o capi al ac oss bo de s. While he inancial accoun is
conce ned wi h in e -coun y mone a y lows ela ed o inancial in es men s om ( o) he coun y
o ( om) he es o he wo ld, in he o m o deb secu i ies, equi y, loans, amongs o he s. The
inancial accoun o he BoP is also an impo an ac o in he In e na ional In es men Posi ions
amewo k, which will be explained nex .

6
The In e na ional In es men Posi ions (IIP) e e o he end-o -pe iod s ocks. I co esponds o
he alue, o all inancial asse s o esiden s o an economy ha ha e claims on non- esiden s, as
well as he liabili ies o esiden s o non- esiden s a a speci ic ime (IMF, 2009). The di e ence
be ween he asse and liabili y sides gi es ise o he ne posi ion o he IIP. The changes in he IIP
a e explained by he c oss-bo de lows, including ansac ions, p ice changes, exchange a e
changes, and o he changes in olume. The amoun o ansac ions co esponds o he inancial
lows eco ded unde he inancial accoun in he BoP. While, he es o he lows co espond o
he change in he ma ke alua ion o inancial asse s/liabili ies, app ecia ion and dep ecia ion o
cu encies, and o he changes no explained by he o he lows. The e o e, he combined a ia ion
o he low elemen s explain he di e ence in IIP be ween wo pe iods (IMF, 2009).
No wi hs anding, he s anda d measu es used o eco d he ex e nal s a is ics a e usually ela ed
be ween one coun y and he es o he wo ld, on an agg ega ed basis.
This hesis ocuses on he bila e al analysis o one coun y is-à- is i s main pa ne s aiming o
analyze he c oss-bo de inancial in es men s in he o m o deb secu i ies and equi y. These
inancial ins umen s (deb secu i ies and equi y) a e eco ded in he inancial accoun unde he
unc ional ca ego ies o he Fo eign Di ec In es men (FDI) and he Po olio In es men (PI).
FDI includes he ini ial in es men and all he o he inancial linkages om esiden s o one
coun y in an en e p ise loca ed in a o eign coun y, when he in es o owns a minimum o 10
pe cen o he o ing powe (OECD, 2008). Such in es men s aim o es ablish a long-las ing
in e es in a o eign business. On he con a y, he Po olio In es men co esponds o c oss-bo de
in es men s in he o m o deb secu i ies and equi y ha alls below he FDI h eshold. This ype
o in es men is o ien ed owa d small in es o s ha aim o ge sho - e m e u ns om hei
in es men s (OECD, 2008). These wo indica o s a e ele an as hey co espond o o eign
7
in es men s. Figu e 1 ep esen s he global o eign di ec in es men (asse s) and he po olio
in es men (asse s) measu ed as % o he wo ld GDP.
Figu e 1 – To al FDI asse s (le cha ) and po olio in es men s asse s ( igh cha ) as % o Wo ld GDP
As shown in he g aphs o Figu e 1, he FDI and PI ep esen a signi ican p opo ion o he
wo ld GDP. The global FDI asse s and PI asse s co espond o 45.36% and 16.71% o wo ld GDP
in 2019 espec i ely.
I is no iceable ha as ime passes and economies and ma ke s educe he ba ie s o en y
(Sobel, J.R., & Dwig h, 2007), he signi icance o di ec and po olio in es men s as a p opo ion
o e GDP is g owing. Some o hese obse a ions can be explained by he apid expansion o
mul ina ional companies, in e na ional in es men s, as well as bila e al ea ies be ween coun ies
(Egge & P a e may , 2004). The g owing impo ance o FDI and Po olio in es men is
showcased in he g aphs o Figu e 1, making hese complex ela ionships ele an and in e es ing
o s udy.
8
Wi h he e e -g owing olume o in es men s and he endency o co po a ions o unnel money
h ough o sho e cen e s, i is di icul o ully g asp in e na ional in es men connec ions
(Damgaa d & Elkjae , 2017). In he case o po olio in es men , many ansac ions occu in
seconda y ma ke s, while in he case o di ec in es men acco ding o he BPM6, he ope a ions
a e eco ded is-á- is he immedia e coun e pa and no he ul ima e in es o (IMF, 2009). In ha
sense, i an en e p ise in economy A has a subsidia y
1
in economy B and B has a subsidia y in
economy C, he ul ima e in es o o C is A, al hough only he di ec in es men ela ionship is
eco ded (ie. he FDI s a is ic would show a di ec ela ionship be ween A and B and B and C
sepa a ely). The e o e, making i di icul o ace he ul ima e in es o and, consequen ly, he “ eal”
sou ces and des ina ions o he in e na ional in es men s.
Acco ding o he exis ing documen a ion, IMF p o ides addi ional in o ma ion by coun e pa
coun y bo h o FDI and Po olio In es men , consis en wi h he BPM6: he Coo dina ed Di ec
In es men Su ey (CDIS) and Coo dina ed Po olio In es men Su ey (CPIS). The CDIS and
he CPIS co espond o annual (CDIS) and semi-annual (CPIS) su eys whe e immedia e
coun e pa ela ionships a e obse able. Since ul ima e ela ionships a e mo e complex,
unde lying s uc u es a e di icul o ex ac wi hou applying ad anced models ha can imply
hese ela ionships. Hence his hesis aims o iden i y pa e ns, p e e en ial pa hs o in es men ,
es ablish ends and desc ibe he ela ions be ween coun ies o e ime using he a ailable
in o ma ion om CPIS and CDIS.
1
A subsidia y is a company ha is owned o con olled by ano he company, usually e e ed o as he pa en
company. (OECD, 2001)
9
We use ne wo k analysis o model he linkages be ween he sou ces and inal des ina ions o
he o eign di ec and po olio in es men s on a global le el. This ool is commonly used in
analyzing complex social in e ac ions, whe e many ac o s ela e h ough bila e al ies ha c ea e a
ela ionship (Sco , Social Ne wo k Analysis, 1988). Ne wo ks ha e been used in analyzing
co po a e powe (Sco , 2003), in e na ional ies (Ha ne -Bu on , Kahle , & Mon gome y, 2009),
class s uc u e (P o a & Be s o d, 2011), e c. Thei complex sys ems oo s, help unde s and how
indi idual in e ac ions lead o he c ea ion o a mac oscopic s uc u e ha exposes he
undamen als o he sys em beha io as a whole (Lima, Pinhei o, Sil a, & Ma os, 2020).
The impo ance o ne wo k analysis o his hesis lies in i s po en ial o map he o eign
in es men linkages based on CDIS and CPIS da a, bo h on an indi idual le el agains he es o
he wo ld as well as globally. Applying ne wo k analysis o hese da ase s exposes he unde lying
s uc u e o he ne wo ks. The e o e, his hesis will le e age on he p ope ies o ne wo k analysis,
allowing us o ans o m he bila e al connec ions o a close o eali y co e s uc u e o global
in es men in e connec ions. Ul ima ely, iden i ying he mos impo an playe s in a global
ne wo k o o eign di ec in es men s and po olio in es men s.
This hesis is s uc u ed in 8 main sec ions: a e he in oduc ion sec ion, we p esen he
li e a u e e iew, hen he da a sec ion, a e which he ne wo k analysis sec ion ollows. The ea e ,
he hypo hesis o mula ion sec ion will gene a e he main ques ions his hesis is aiming o answe .
The esul s sec ion will discuss he indings o he analysis and highligh he esul s and di e ences
be ween ne wo ks. The pla o m sec ion will jus i y he choices behind he web applica ion
isualiza ions, hei unc ionali y, pu pose, and s o yline. And las ly, he conclusion will
summa ize he indings, and poin owa ds limi a ions bu also imp o emen s ha can be done in
he ield.
16
in e ing he pa h o o eign in es men bo h o o eign di ec and po olio in es men s by
applying ne wo k analysis wi hou bu dening coun ies o epo supplemen a y da a on ul ima e
in es men a ge s. As well as communica ing he esul s h ough an in e ac i e web applica ion.
3 Da a
We use da a sou ced om IMF’s da abase on Coo dina ed Di ec In es men Su ey (CDIS)
and Coo dina ed Po olio In es men Su ey (CPIS). They a e su eys ha collec da a wi h
ega ds o FDI and po olio in es men espec i ely, a e a ailable on an annual basis (in he case
o FDI) and semi-annual basis in he case o Po olio In es men a e 2013.
Each da ase ollows a simila s uc u e. The CDIS and CPIS da ase s con ain 6 main a iables
and a e summa ized in Table 1.
Table 1 — Summa y o he main a iables om he CDIS and CPIS da ase s ha we e used in his hesis
Va iable Name
Desc ip ion
Da a Type
Coun y Name
The coun y o o igin, o he in es o
Ca ego ical
Coun y Code
The coun y o o igin’s ISO code
Ca ego ical
Coun e pa Coun y Name
The coun e pa , o he ecipien o he
in es men
Ca ego ical
Coun e pa Coun y Code
The coun e pa coun y’s ISO code
Ca ego ical
Time Pe iod
The pe iod as a yea
Con inuous
Value
The in es men amoun
Con inuous
The indica o names, codes, hei naming con en ion in he web applica ion we ha e de eloped
a e summa ized in Table 2. Onwa ds, we use he e ms CDIS Inwa d, CDIS Ou wa d, CPIS Asse s,
and CPIS Liabili ies o e e o each o hese da ase s.

17
Table 2 — Summa y o indica o s used in his hesis. The i s ou indica o s a e sou ced di ec ly om he IMF
da abase, while he las wo a e a pai ing o he o he ou indica o s.
Since bo h CDIS and CPIS co espond o o al amoun s in US dolla s and no p oximi ies, which
is he desi able me ic o ou analysis, i is he e o e necessa y o ans o m hei alues o
p oximi ies.
Le us s a by de ining 𝑓𝑖𝑗as he o al in es men o coun y 𝑖 in coun y 𝑗, which can in gene al
e ms conce n any o he indica o s shown in Table 2. The in es men s be ween wo coun ies can
be asymme ical, ha is, 𝑓𝑖𝑗 ≠ 𝑓𝑗𝑖 implying he di ec ionali y o he in es men s. While, in mos
cases, he alue o 𝑓𝑖𝑗 is posi i e, in ce ain si ua ions i can also be nega i e. Fo ins ance, suppose
coun y 𝑖 ( he pa en ) in es ed and is holding a posi ion in coun y 𝑗 ( he a ilia e); he pa en can
use he a ilia e o unding ope a ions back a home. When he o al amoun o unding ha lows
back o he pa en , exceeds he o al amoun o in es men done in he a ilia e i is s anda d o
epo i as a nega i e low. As such, we shall conside he absolu e alue o 𝑓𝑖𝑗, ha is, |𝑓𝑖𝑗|. The
Web App
Indica o desc ip ion
Indica o Code
Sou ce
CDIS Inwa d
Inwa d Di ec In es men Posi ions,
US Dolla s
IIW_BP6_USD
IMF
CDIS Ou wa d
Ou wa d Di ec In es men Posi ions,
US Dolla s
IOW_BP6_USD
IMF
CPIS Asse s
Asse s, To al In es men , BPM6, US
Dolla s
I_A_T_T_T_BP6_USD
IMF
CPIS Liabili ies
Liabili ies, To al In es men , BPM6,
US Dolla s
I_L_T_T_T_BP6_USD
IMF
Asse s Combined
CDIS Ou wa d + CPIS Asse s
IOW_BP6_USD +
I_A_T_T_T_BP6_USD
Au ho s’
calcula ions
Liabili ies
Combined
CDIS Inwa d + CPIS Liabili ies
IIW_BP6_USD +
I_L_T_T_T_BP6_USD
Au ho s’
calcula ions
18
main eason o aking he absolu e alue ins ead o emo ing such obse a ions is o p e en he
loss o in o ma ion abou impo an in es men pa ne s.
In o de , o ob ain a p oxy o p oximi y we conside he ecip ocal o he absolu e alue o he
di ec ional in es men . The e o e, he p oximi y calcula ion o each in es men is as ollows:
𝜙𝑖𝑗 =1
|𝑓𝑖𝑗|(1)
whe e 𝜙𝑖𝑗 ≠ 𝜙𝑗𝑖. In ha sense, we say ha he la ge he in es men amoun be ween wo coun ies,
he close hey a e o each o he . Ul ima ely a p oximi y ma ix is ob ained, which o ms he basis
o building a di ec ed weigh ed ne wo k ha ep esen s global ade lows in o eign in es men .
Addi ionally, we combine bo h Inwa d and Ou wa d lows om CDIS and CPIS da ase s. I is
in e es ing o unde s and how he ne wo ks change once he main c oss-coun y in es men
da ase s a e combined, o map possible in e na ional in es men linkages ac oss coun ies. These
da ase s do no e e o he same s a is ical concep s because FDI is eco ded in he di ec ional
p inciple, whe eas he CPIS is eco ded on he asse s/liabili ies pe spec i e. Al hough he e is no
a ailable in o ma ion on he FDI asse s/liabili ies by coun e pa coun y, he e o e, he bes p oxy
ha can be used o unde s and he in e na ional in es men in he o m o deb /equi y is o
agg ega e CDIS Ou wa d wi h CPIS Asse s as Combined Asse s, and CDIS Inwa d wi h CPIS
Liabili ies as Combined Liabili ies.
19
4 Ne wo k Analysis
Ne wo k analysis has ound applica ions in many ields o s udy (Wasse man & Faus , 1994).
In pa icula , i has been use ul o he mapping and analysis o he web o in e connec ed
ela ionships be ween ac o s ha ela e h ough bila e al ies (Sco , Social Ne wo k Analysis,
1988). Ne wo ks ha e been used in analyzing co po a e powe (Sco , 2003), in e na ional ies
(Ha ne -Bu on , Kahle , & Mon gome y, 2009), class s uc u e (P o a & Be s o d, 2011), e c.
Ne wo k analysis’s oo s in he s udy o complex sys ems ha e helped us o un eil he link be ween
indi idual in e ac ions and he eme gence o seemingly disconnec ed mac oscopic pa e ns and
beha io s (Lima, Pinhei o, Sil a, & Ma os, 2020).
A ne wo k, G, is composed o wo di e en bu complemen a y elemen s: a se o N
e ices/nodes and a se o links/edges. Edges connec a pai o nodes and iden i y he exis ence o
a ela ionship be ween hem. Ve ices ep esen he uni o analysis, while he edges ep esen he
ela ionship be ween hem. In a social ne wo k amewo k, a node would be an indi idual and an
edge could ep esen he exis ence o iendship, co-wo k, o us ie be ween pai s o indi iduals.
O a node could ep esen a i m and he edge, a inancial ela ionship be ween hem. In his hesis,
we shall use nodes o a ne wo k as ep esen a ions o coun ies, while edges ep esen he exis ence
o an in es men be ween a pai o coun ies.
20
Figu e 2 – G aphical ( op) and ma ix (bo om) ep esen a ion o ne wo ks ep esen ing s uc u es ha ha e edges o
di e en na u e. Sou ce: (Lima, Pinhei o, Sil a, & Ma os, 2020).
Depending on he na u e o he ela ionships being modeled, he e can be h ee main ypes o
ne wo ks, namely: undi ec ed; di ec ed; and weigh ed. One way o ep esen such ne wo ks is
h ough he adjacency ma ix, see bo om panel o Figu e 2. The adjacency ma ix, A, o a ne wo k
in o ms on he exis ing ela ionships be ween nodes/ e ices. In ha sense, he en y aij o A is
ze o i he e is no ela ionship be ween nodes i and j, being non-ze o i a ela ionship exis s be ween
such a pai o nodes. In a weigh ed g aph, aij ep esen s he weigh (s eng h) o he ela ionship
be ween he nodes
4
. In he case o an undi ec ed ne wo k aij = aji. Fo he di ec ed ne wo k case
he ma ix is no symme ic along he diagonal and indica es a ela ionship and i s di ec ion,
he e o e aij  aji. The diagonal en ies o each ma ix A indica e sel - ela ionships, and as gene al
p ac ice a e se o ze o. In he case o all ne wo k ypes, he deg ee o a node is de ined as he
numbe o edges i has in common wi h all o he nodes. While, he a e age deg ee e e s o he
ne wo k as a whole and aims o measu e he ela ionship be ween he numbe o nodes and he
4
In an unweigh ed g aph his alue is one. In o he wo ds, ei he he ela ionship/link exis s o no .
21
numbe o edges. This measu e is ob ained by di iding he o al numbe o edges by he numbe
o nodes.
When building a g aphical ep esen a ion o a ne wo k, each edge ep esen s a ela ionship
be ween wo nodes (i.e., pe son, coun y, ins i u ion). In a di ec ed ne wo k, edges a e ep esen ed
wi h a ows o indica e he di ec ion o he ela ionship ( om sou ce o des ina ion). In he case o
di ec ed ne wo ks, wo nodes can ha e wo links (a ows) be ween each o he . Addi ionally, in a
weigh ed g aph, he hickness o he edge can be used o ep esen he s eng h o he ela ionship.
Addi ional a ibu es can be associa ed wi h each ela ionship and each ac o , which can be
associa ed wi h he edges and nodes o he ne wo k (e.g., we migh wan o conside he gende o
age o indi iduals). Howe e , hese addi ional a ibu es do no a ec he co e s uc u e o he
ne wo ks, bu hey add a dimension ha allows o classi y ela ionships and p o ile explana o y
ac o s o he c ea ion o ela ionships.
Weigh s can be in e p e ed ei he as p oximi ies/simila i ies o dis ances. I is impo an o
es ablish which measu e is being used in a ne wo k, as hey ha e opposi e in e p e a ions. Howe e ,
he choice hinges on a balance be ween he a ailable da a and he analy ical pu pose o he ne wo k
s uc u e unde s udy.
Las ly, i is also common o s udy a simpli ied p ojec ion o he ne wo k, o ins ance by using
an unweigh ed p ojec ion o a weigh ed ne wo k by applying a h eshold o edges. In he case o
his hesis, a di ec ed-weigh ed ne wo k will be cons uc ed o he analy ical pa . This implies
ha each edge/link ca ies in o ma ion on he s eng h o he ela ionship, as well as i s
di ec ionali y. Fo he case o analyzing global FDI and po olio in es men pa e ns, bo h
di ec ionali y and s eng h ma e . The e o e, di ec ed-weigh ed ne wo ks a e cons uc ed o de ine
and explain he unde lying s uc u e and iden i y he mos cen al coun ies and hei cha ac e is ics.

22
O en, i is he case ha we wan o iden i y he ole o each ac o in he o e all sys em h ough
i s posi ion in he ne wo k. In ne wo k analysis, his is done by es ima ing he cen ali y o nodes.
Se e al measu es exis o ha pu pose. Fo ins ance, one can a gue ha he mos cen al/impo an
node is he one wi h he highes deg ee/connec i i y, which is he numbe o links ha a e
connec ed/connec o a node. In he case o a weigh ed ne wo k, we o en measu e he s eng h,
ins ead o he deg ee, as he o al weigh o incoming/ou going edges. Mo eo e , in a di ec ed
ne wo k he measu e can be analyzed sepa a ely in o incoming and ou going connec ions. Hence,
allowing us o de ine h ee measu es: deg ee cen ali y, in-deg ee cen ali y, and ou -deg ee
cen ali y. The s anda dized o mula o deg ee cen ali y is:
𝐶𝐷(𝑎)=𝑣𝑎
𝑛−1
(2)
Whe e 𝑣𝑎 is he numbe o nodes a is connec ed o, and n is he o al numbe o nodes in he
ne wo k. In e ms o in/ou -deg ee cen ali y, he o mula ollows he same logic, excep aking
in o conside a ion only incoming o ou going edges in he nume a o .
Howe e , he numbe o connec ions an edge holds can ell li le abou he ole o a node in
media ing in o ma ion be ween di e en pa s o he ne wo k. To ha end, be weenness cen ali y
assumes ha a node is mo e impo an he mo e sho es pa hs (pa hs connec ing pai s o nodes in
he ne wo k) i media es. The highe he be weenness cen ali y o a node he mo e cen al/ ele an
i is. The o mula o be weenness cen ali y is:
𝐶𝐵(𝑎)= ∑ 𝜎(𝑖,𝑗|𝑎)
𝜎(𝑖,𝑗)
𝑖,𝑗∈𝐴
(3)
23
Whe e a is he node (coun y), 𝜎(𝑖,𝑗) ep esen s he numbe o sho es (I, j)-pa hs, and
𝜎(𝑖,𝑗|𝑎) is he numbe o sho es pa hs passing h ough node a, o he han i, j. I i = j hen
𝜎(𝑖,𝑗)=1, and i 𝑎 ∈𝑖,𝑗 hen 𝜎(𝑖,𝑗|𝑎)= 0.
Conside ing he dis ance be ween nodes in a ne wo k, closeness cen ali y measu es he
impo ance o a node depending on how close i is o he o he nodes in he ne wo k. The closes
i is, on a e age, he mo e cen al i is. I s o mula is:
𝐶𝐶(𝑎)= ∑ 𝑛−1
𝑑(𝑣,𝑎)
𝑛−1
𝑣=1
(4)
Whe e 𝑑(𝑣,𝑎) is he sho es pa h dis ance be ween and a, and n is he numbe o nodes ha
can each a.
Las ly, eigen ec o cen ali y measu es he cen ali y o a node by he p obabili y o a andom-
walke spending ime in ha node i he walke is allowed inde ini ely o mo e h ough he ne wo k
by a e sing i s edges. I o e s a dynamical measu e o cen ali y, in ha we a e saying ha a
node is mo e ele an he mo e ime a andom low o in o ma ion ans e ses i .
𝐴𝑥 =𝜆𝑥
(5)
Whe e A is an adjacency ne wo k and 𝜆 he eigen alue.
Figu e 3, isually shows how hese di e en measu es o cen ali y can classi y di e en nodes
as he mos cen al, hus highligh ing ha each one plays a di e en ole on di e en s uc u al
dimensions.
24
Figu e 3 – Simple ne wo k wi h he mos impo an cen ali y measu es Sou ce: (Lima, Pinhei o, Sil a, & Ma os, 2020)
In his hesis, we will use nodes o abs ac coun ies and edges o iden i y inancial ela ionships
be ween pai s o coun ies. Mo eo e , edges will ep esen e ealed p oximi ies be ween coun ies.
We will pe o m all compu a ions in he en i e y o he ne wo k (e.g., node cen ali y and sho es
pa hs), wi h all i s links, howe e , o isualiza ion pu poses (because he ne wo ks a e e y dense)
we will ep esen only he mos ele an edges.
To ha end we shall ollow he ollowing s eps: 1) we iden i y he Minimum Spanning T ee,
which is a se o edges ha ensu es all nodes a e in e connec ed while minimizing he sum o
p oximi ies be ween he selec ed edges; 2) hen we en ich he Spanning T ee wi h he edges ha
iden i y he closes ela ionships un il we each a minimum a e age deg ee o 3.5 links, which we
ake as a humb ule o a ne wo k densi y ha would allow o in e p e able ne wo k isualiza ion.
25
5 Hypo hesis Fo mula ion
The discussion so a gi es ise o he o mula ion o he main ques ions his hesis aims o
answe . Can we use cen ali y measu es gene a ed by applying ne wo k analysis o ack he
posi ions o he wo ld coun ies o e ime? Mo e speci ically, can he esul s o ne wo k analysis
applied on each indica o o he CDIS and CPIS da ase s be used o answe ques ions such as:
- Which coun ies ha e in e media ed he mos inwa d di ec in es men and po olio
in es men liabili ies pa hs o 2009 and 2019?
- Which coun ies expe ienced he highes ank a iance o e he las 10 yea s as op
in e media o s?
- Wha is he sho es in es men pa h be ween, o ins ance, China and he U.S. o any o he
coun y pai ?
- Wha is he b eakdown o inwa d FDI in es men pe coun y, g ouped by con inen ?
- Wha coun ies a e he op di ec in es men and po olio in es men in e media o s
globally?
The ollowing sec ions aim o ex ac he esul s o he analysis and answe he o mula ed
ques ion and in he discussion on how he web app can p o ide a ool o egula o s and cen al
bank o icials o quickly ob ain answe s o hose ques ions.
32
in he wo ld. Thei poli ical and economic ies can be s udied h ough he sho es pa hs o
in es men gene a ed by he CDIS Inwa d Ne wo k. Figu e 7 ep esen s he e olu ion o his
ela ionship be ween 2009 and 2019. F om 2009 up un il 2017 Japan was he only in e media o
in his ela ionship. While om 2018 o 2019 he U.K. and Hong Kong became in e media o s
be ween China and he U.S.
Figu e 7 – CDIS Inwa d Sho es Pa h Visualiza ion, whe e each node ep esen s a coun y as pa o he sho es pa hs
o in es men be ween he U.S. and China.
Po olio Liabili ies In es men Ne wo ks
Le us now ocus on he po olio liabili ies in es men ne wo ks. In 2009 he mos impo an
in e media o s we e: Japan, Aus alia, Is ael, Li huania and Romania. The isualiza ions in Figu e
8 ep esen he CPIS Liabili ies ne wo k in he yea s o 2009 and 2019 whe e he op in e media o s
a e shown acco ding o hei node size, based on be weenness cen ali y. Fo 2019 hese a e Japan,
Aus alia, Finland, Es onia and Bulga ia. The s uc u e o he CPIS Liabili ies ne wo k is
signi ican ly di e en o he CDIS Inwa d ne wo k. He e Japan and Aus alia a e he mos
dominan in e ms o in e media y powe based on be weenness cen ali y. Japan is consis en ly
anked 1s h oughou he en i e yea ange, while Aus alia has ne e d opped below 5 h posi ion,
as shown in he ank e olu ion cha in he web app. Aus alia’s ank alls o 5 h place only in 2013
and is ou anked by Is ael, Romania and Li huania. Fu he mo e, hey a e, espec i ely, pa o 74%
and 46% o all sho es in es men pa hs in 2019. This showcases hei abili y o a ac sho - e m

33
in es men s aimed a inancial gain, by e ail o ins i u ional in es o s. Excluding, Finland he op
ou in e media o s o 2019 ha e ne e anked below 9 h be ween 2009 and 2019.
Figu e 8 – CPIS Liabili ies Ne wo ks wi h node size ep esen ing ela i e amoun s o Be weenness Cen ali y o 2009
( op) and 2019 (bo om). Sou ce: web app.
34
I is wo h men ioning ha he U.K., he Ne he lands, and he U.S. a e op anked 19 h, 21s and
23 d in e ms o be weenness cen ali y ank. Thei ank in e ms o be weenness cen ali y has no
been below 31 in he pas 10 yea s. While Finland who anks 3 d in 2019 became an impo an
po olio liabili ies in es men in e media o only in 2015.
The s eng h o he op 3 coun ies in e ms o po olio liabili ies in es men s can u he be
ein o ced by looking a he p opo ion o hei po olio liabili ies in es men ela i e o he o al
o 2019. Figu e 9 shows ha he Japan is he mos dominan , wi h a 63.5% o he o al po olio
liabili ies in es men s o 2019. I is ollowed by Aus alia and Finland. These coun ies make up
87.76% o o al po olio liabili ies in es men o 2019.
Figu e 9 – T eemap ep esen a ion o he o al CPIS Liabili ies b oken down by coun y and con inen o 2019. Sou ce:
web app.
In e ms o closeness cen ali y, he e a e some mo e di e ences as compa ed o he CDIS
Inwa d ne wo k. The op coun ies o he CPIS Liabili ies in 2019 include he U.S., Luxembou g,
35
I eland, he Ne he lands, and Singapo e. Al hough, bo h he CDIS Inwa d ne wo k anks
p edominan ly o sho e cen es he highes based on closeness cen ali y measu es. The CPIS
Liabili ies ne wo k anks he U.S. he highes , bu also includes ax hea ens like Luxembou g,
I eland, he Ne he lands and Singapo e. The only common coun ies in he op 10 a e he B i ish
Vi gin Islands and Saudi A abia. In bo h cases, some o he main wo ld economies including China,
Ge many, F ance, and he U.K. ank ou side o he op 90.
Compa ing he sho es pa hs o in es men be ween he U.S. and China o CDIS Inwa d and
CPIS Liabili ies, he e is a sligh di e ence. As Figu e 10 indica es, Japan ac s as he in e media o
be ween he wo majo economies h oughou he en i e yea ange.
Figu e 10 – The sho es pa h be ween he U.S. and China o CPIS Liabili ies o e be ween 2009 and 2019. Sou ce:
web app.
Conce ning he in and ou -deg ee cen ali ies, he e a e some di e ences. The ou -deg ee
cen ali y iden i ies Bulga ia, Malaysia, Japan, Mongolia and La ia as he 5 coun ies wi h he
leas amoun o a iance in hei ank be ween 2009 and 2019. They ha e ne e been below he
15 h ank. In e ms o in-deg ee cen ali y he coun ies wi h he leas a iance include Ge many,
he Ne he lands, he U.S., Luxembou g and Swi ze land.
Combined Liabili ies Ne wo k
In he combined liabili ies ne wo k we ob ain a simila ne wo k s uc u e o he examples
discussed in Sec ion 5.1. Bo h in 2009 and 2019, he Uni ed S a es, he Ne he lands, Luxembou g
and China a e anked he highes as in e media o s based on be weenness cen ali y. In ac , he
36
op 10 coun ies a e he same o bo h he CDIS Inwa d and Combined Liabili ies ne wo ks, excep
o Hong Kong and Japan. In compa ison, he CDIS Inwa d ne wo k anks Hong Kong as a mo e
impo an in e media o ; while he Combined Liabili ies ne wo k anks Japan highe . Ne e heless,
i s s uc u e is e y simila wi h he CDIS Inwa d ne wo k, which is o be expec ed since i is one
o he main building blocks o Combined Liabili ies. The op coun ies in e ms o be weenness
cen ali y, a e s able in he highes anks, wi h Luxembou g being he only one ha anks lowes
a he 15 h posi ion be ween 2009 and 2019.
Figu e 13 – Combined Asse s Ne wo k wi h node size ep esen ing ela i e amoun s o Be weenness Cen ali y o
2019, whe e he Japan and i s connec ions a e highligh ed. Sou ce: web app
37
Japan’s posi ion is no su p ising since i is known o be one o he mos indeb ed coun ies in
he wo ld, hence i s high ank in a ne wo k ela ed o liabili ies. Figu e 13 highligh s Japan’s
posi ion in he Combined Liabili ies ne wo k and shows i s ole as a connec o be ween Asia,
Eu ope and No h Ame ica. Fu he mo e, he sho es pa h g aph in he web app, showcases ha
Japan is he main in e media o be ween he U.S. and China be ween 2009-2019. I s dominan
posi ion in he CPIS Liabili ies ne wo k u he ein o ces Japan’s in luence he e, since CPIS
Liabili ies is he o he building block o he Combined Liabili ies ne wo k. This is also ein o ced
by Japan’s p opo ion o o al Combined Liabili ies, adding up o 9.18% as shown in Figu e 14.
Figu e 14 – T eemap ep esen a ion o he o al Combined Liabili ies b oken down by coun y and con inen o 2019.
Sou ce: web app.
In e ms o in e media ion, he Ne he lands 66% o he sho es pa hs o in es men , while he
U.S. 56% o he ime and Japan 43% o 2009. Compa ed o 2019 he U.S. and he Ne he lands
swi ch posi ions, whe e he U.S. is pa o 67%, he Ne he lands is pa o 47% o he sho es pa hs
o in es men ; while Japan is hi ds wi h 31%.

38
Wi h ega ds o he closes coun ies wi hin he ne wo k, he coun ies wi h lenien ax laws a e
dominan in 2019. These include Samoa, Cayman Islands, Be muda, Je sey, and he B i ish Vi gin
Islands. In 2009 simila ly, Samoa, Be muda, Ne he lands An illes, Cayman Islands and B i ish
Vi gin Islands a e anked highes . These esul s a e close o he ou pu s o he CDIS Inwa d
ne wo k, al hough all h ee ne wo ks o sho e cen e s a e he closes o he es o he coun ies.
This is expec ed as hey a e able o e icien ly sp ead in es men s h ough he ne wo k.
The mos impo an coun ies in e ms o ou -deg ee cen ali ies o 2019 a e I aly, China,
Thailand, he Ne he lands, and Bulga ia. Tha is he same op 5 as he CDIS Inwa d ne wo k.
Simila ly o in-deg ee cen ali y, he U.S., U.K., F ance, Swi ze land, and he Ne he lands ank
he highes in bo h ne wo ks and ha e a s able ank ne e d opping below 6 h posi ion o e he
2009 o 2019 pe iod.
Summa y
Th oughou he esul s sec ion we ha e shown he indings and compa isons o h ee di e en
ne wo ks, namely CDIS Inwa d, CPIS Liabili ies and Combined Liabili ies ne wo ks. This
discussion has p o en he possibili y o applying ne wo k analysis and using he cen ali y
measu es, o ack he posi ion o coun ies and hei impo ance in a global in es men ne wo k.
Addi ionally, we ha e shown ha i is possible o unde s and he mos impo an coun ies in e ms
o acili a ion o in es men s, hei impo ance a a speci ic poin in ime bu as well as hei
e olu ion o e ime.
39
7 The Pla o m
On op o he analy ical pe spec i e, he main p oduc o his hesis was he de elopmen o a
web applica ion ha allows o 1) communica e hese indings in an in e ac i e da a isualiza ion
pla o m; and 2) allow o he quick explo a ion and disco e y o ele an pa ne ships and
in es men pa hs.
In his sec ion, we de ail he s eps conduc ed and hei a ionale o building in e ac i e and
isualiza ion-d i en web-applica ions o explo a ion o he ne wo ks o o eign in es men . In
ha sense, we s a by explaining he echnological s ack used, he na a i e employed, he choice
o pa icula isualiza ion o suppo he na a i e, he inal p oduc , and u u e s eps.
The S ack
T adi ionally da a has been communica ed and sha ed in a abula o ma . Indeed, many open
da a pla o ms hese days s ill ely on ha o ma and hus make complex da a explo a ion di icul
o he use . Mo eo e , abula da a ep esen a ions a e no only di icul o in e p e bu also
di icul o compa e ac oss yea s. In con as , in o ma ion abou size, magni ude, p oximi y,
compa abili y, and empo al e olu ion a e easie o g asp when da a is ep esen ed in he
app op ia ed isualiza ion o ma . In ha sense, a common solu ion comes in he o ma o a
Business In elligence dashboa d, which o e s a layou o quickly p esen mul iple ele an
isualiza ions and key indica o s o he use and communica e insigh s o a specialized audience.
Se e al choices we e conside ed o he isual communica ion o he esul s. One op ion would
ha e been o use an exis ing amewo k, such as Powe BI o Plo ly Dash. The ad an ages o using
such amewo ks a e he ease o use and he sho p oduc ion imes. Howe e , hese op ions usually
40
conside a igid can as o de elopmen and o e a limi ed ange o isualiza ion possibili ies and
cus omizabili y. Ano he signi ican d awback in using hese op ions is he cos o licensing he
so wa e (e.g. Powe BI) and hos ing equi emen s (Dash).
The o he choice in ol es he ull-s ack de elopmen o a web app. The ad an age o his line
o wo k is ha i o e s ull lexibili y in he de elopmen o he applica ion o mee he
de elopmen goals. Addi ionally, since i is ully de eloped by he eam i c ea es an asse ha can
hen be u he de eloped and hus b ing g ea e added alue. The ools used in his line o
de elopmen a e ee and open-sou ce, allowing he inal p oduc o be eplica ed and sha ed wi h
a wide audience. One o he disad an ages is he ime i akes o de elop such an applica ion, as
well as i s complexi y, and a s eepe lea ning cu e o beginne s.
In he con ex o he mas e hesis, he choice ul ima ely ell in o he de elopmen o a web
applica ion om sc a ch using ee and open-sou ce ools.
The explo a ion analysis o he da a was ini ially conduc ed in py hon. All he calcula ions we e
pe o med in Jupy e no ebook using common da a science lib a ies, such as Pandas and
Ne wo kX. In hese ea ly s eps wo k consis ed o da a cleaning and expo in json o cs o ma s,
which would eed he isualiza ions o be de eloped ahead. The speci ic aspec s o each
isualiza ion will be discussed in hei sub-sec ion.
The web applica ion was buil in a s ack ha comp ised aspec s o HTML 5, CSS, and
ja asc ip . In pa icula , we used se e al handy ja asc ip lib a ies, such as jQue y, D3.js, and
d3Plus. Fo he s yling o he HTML elemen s, we eso ed o he boo s ap lib a y, a well-known
amewo k de eloped by Twi e (see h ps://ge boo s ap.com/). Addi ionally, we used Fon
Awesome icons h oughou he websi e o s yle dynamic ac ions (e.g., mouse ho e ing) elemen s
o bu ons.
41
Conce ning ja asc ip lib a ies — D3.js and d3Plus — hey ope a e as DOM manipula o s
while aking le e age on he use o Scalable Vec o G aphics (SVG’s) ha allows d awing shapes
in a b owse window. The main ad an age o using SVG’s is ha hey don’ lose quali y when
hey a e scaled, and hey ha e low memo y equi emen s. Mo eo e , SVG elemen s can be
manipula ed simila ly o o he DOM elemen s.
Wi h ega ds o he da a loading and s o age, wo op ions we e conside ed. One was o load he
da a om a olde s uc u e h ough json and cs iles. The o he op ion was o build a da abase
and e ch he da a h ough a da abase connec ion, which would add complexi y o he de elopmen .
The i s op ion was chosen because he da a does no e ol e and does no need o be cons an ly
upda ed. Ini ially, he e we e some conce ns o e he loading speed due o la ge iles which would
ake a long ime o ende e e y ime he e is new use inpu . The wo ka ound o his issue was o
o ganize he da a iles in sepa a e yea s o coun ies and load a di e en ile o speci ic selec ions,
which signi ican ly imp o ed he loading speed. Fo academic pu poses, he cu en choice is
su icien , al hough i he web app ge s de eloped in a ully- ledged p oduc , an impo an
imp o emen would be o build a da abase.
All he echnologies combined, allowed us o build a web app and communica e he esul s
ac oss i e pages, including a landing, ne wo k, me hodology, abou , and downloads page.
Na a i e
This sub-sec ion will ou line he mos impo an isualiza ions pe page and jus i y he choices
behind he ypes o isualiza ions and hei pu pose.
48
The U.S. in he selec ion o CDIS Inwa d in 2019 is in he second posi ion as op in e media o ,
being pa o 53.73% o sho es in es men pa hs (blue ba ). While being he i s in e media y
18.17% o he ime (yellow ba ).
The las sec ion o he ne wo k page p esen s all he ne wo ks esul s in abula o m. I p esen s
each cen ali y measu e, hei ank, and alues, o all a ailable ne wo ks and yea s. He e, he use s
ha e he op ion o download he da a o u he explo a ion.
7.1.3 Me hodology
The me hodology sec ion ou lines all he in o ma ion a use needs in o de o unde s and
e e y hing ha is p esen ed h oughou he web applica ion. I is b oken down in o se e al sec ions,
mainly one pe isualiza ion. In o ma ion such as de ini ions o e ms and da a, as well as an
explana ion o wha ne wo k analysis is and he easoning behind da a agg ega ions and
ans o ma ions, is also p esen ed on his page.
7.1.4 Abou and Downloads
The abou sec ion gi es he use s a summa y o he web applica ion, i s pu pose, and mo i a ion.
I also gi es an in oduc ion o he au ho s and de elope s.
The downloads page has se e al links ha lead o he sou ce code, da ase s, and ela ed
ins i u ions.

49
8 Conclusions
The esul s ob ained in his hesis allow us o answe he main esea ch ques ion. The simple
answe is, i is possible o use he cen ali y measu es ob ained by applying ne wo k analysis o
ack he posi ion o coun ies o e ime.
Mo e speci ically, global economies and s a egic pa ne s such as he Uni ed S a es, China,
Hong Kong, Ne he lands, and Luxembou g ank he highes in he di e en ne wo ks in e ms o
in e media y powe based on be weenness cen ali y o 2019. The esul s also show ha he
a ia ion in hese anks is no he same o all he lis ed coun ies. Hong Kong and China, o
ins ance, ha e become some o he mos in luen ial in e media o s only in he las se e al yea s.
While he U.S., Ne he lands, and Luxembou g a e in he op 20 anks o mos o he yea s be ween
2009 and 2019. In e ms o closeness cen ali y, coun ies cha ac e ized by lenien co po a e ax
laws such as he Cayman Islands, Be muda, Je sey, he and B i ish Vi gin Islands a e among he
op anks in he las ew yea s. Las ly, he in/ou -deg ee cen ali ies ank he bigges global
economies he highes wi hou signi ican ola ili y be ween 2009 and 2019.
In addi ion o he analy ical pe spec i e, he web applica ion has p o en o ease he acking o
indi idual coun ies’ posi ions in global in es men ne wo ks o e ime. I gi es cen al banke s
and policymake s an easy way o iden i y unde lying in es men pa hs om a global in es men
ne wo k ha is buil based on CDIS and CPIS. One o he ad an ages is ha he ne wo ks and
measu es a e p e-compu ed, and hei esul s can be quickly ende ed on he page. Addi ionally,
he isualiza ion o he esul s makes i easie o unde s and and ex ac aluable in o ma ion om
i .
50
The e o e, his hesis showcases he possibili y o applying ne wo k analysis o analyze la ge
da ase s and p oduce signi ican esul s. Mo e speci ically, he powe o ne wo k analysis o use
immedia e coun e pa CDIS and CPIS da a o ace he ul ima e in es o s, as well as analyze any
coun y’s ank, in luence, and connec edness in a global ne wo k. Las ly, he combina ion o he
ne wo k analysis esul s wi h complex isualiza ion echniques p o es hei abili y o p o ide
use ul insigh s on economic analysis o in e na ional in es men ela ionships such as hose based
on Fo eign Di ec In es men and Po olio In es men .
E en hough, his hesis has been a success i also has some limi a ions. F om an analy ical
poin o iew, some new insigh s could be gained by analyzing a la ge ime ame. In e es ing
esul s could be ob ained when looking a he di e ences in he ne wo ks be ween ce ain e en s
such as he inancial c isis o 2008/2009 o e y ecen ly he s uc u e o he ne wo ks be o e and
a e he Co id-19 pandemic. As a u u e imp o emen , i could also be good o collec he da a on
an ul ima e coun e pa basis and compa e he models gene a ed om his hesis o see how well
i iden i ies he unde lying in es men pa hs. Cu en ly, he analysis is ocused on FDI and
po olio in es men ela ionships be ween coun ies as a whole, al hough zooming in and
collec ing da a based on ypes o in es men s di e en ia ed by he sec o in which he in es men
is pa o would add ano he dimension. Ano he d awback o he hesis is in he way Combined
Liabili ies and Combined Asse s alues a e c ea ed. They a e a combina ion o CDIS and CPIS
indica o s, while he indica o s used o hei c ea ion a e eco ded on di e en p inciples – namely
di ec ional and asse /liabili y p inciples espec i ely. The e o e, o u u e expansions as an
imp o emen , i would be be e o collec he CDIS da a on he asse /liabili y p inciple o he
combina ion o he indica o s o be mo e signi ican . The da a exhibi s also some limi a ions as he
o al amoun o he asse s/liabili ies is no equal due o he lack o in o ma ion and no ull co e age.
51
In e ms o he pla o m, i is cu en ly de eloped as an academic p ojec al hough i has he
po en ial o become a ully- ledged p oduc . The da a o his hesis is s a ic and is no e ol ing,
hence he e was no necessi y o s o e and e ch he da a om a da abase. Cu en ly, he da a is
s o ed in a olde s uc u e and is e ched as such in he web applica ion. Fo u he imp o emen
and as a bes p ac ice i would be ad isable o c ea e a da abase. I also has he po en ial o include
e en mo e isualiza ions ha would add alue o he use s.
Las ly, use expe ience su eys and in e iews ha e no been pe o med due o ime limi a ions.
Such exe cises would be e y use ul in be e unde s anding he posi i e and nega i e aspec s o
he design and communica ion in he web app. I would be in e es ing o ga he his in o ma ion
om expe s in he ields, such as cen al banke s and policymake s.
52
9 Bibliog aphy
Ba abási, A.-L. (2013). Ne wo k science. Philosophical T asac ions o he Royal Socie y A.
Bolí a , L. M., Casanue a, C., & Cas o, I. (2019). Global Fo eign Di ec In es men : A ne wo k
pe spec i e. In e na ional Business Re iew.
B anco, J. (2017). Shedding ligh on a libe alized ma ke . Lisbon: Uni e sidade Ca ólica
Po uguesa.
Chuluun, T. (2016). Global po olio in es men ne wo k and s ock ma ke como emen . Else ie .
Damgaa d, J., & Elkjae , T. (2017). The Global FDI Ne wo k: Sea ching o Ul ima e In es o s.
In e na ional Mone a y Fund.
E., B., De G ego io, J., & Lee, J.-W. (1997). How does o eign di ec in es men a ec economic
g ow h? Jou nal o In e na ional Economics.
Egge , P., & P a e may , M. (2004). The impac o bila e al in es men ea ies on o eign di ec
in es men .
Habe ly, D., & Wójcik, D. (2015). Regional Blocks and Impe ial Legacies: Mapping he Global
O sho e FDI Ne wo k.
Ha ne -Bu on , E. M., Kahle , M., & Mon gome y, A. H. (2009). Ne wo k Analysis o
In e na ional Rela ions. Camb idge Uni e si y P ess, 559 - 592.
Hakeem, M. M., & Suzuki, K.-i. (2016). Fo eign Po olio In es men and Economy: The Ne wo k
Pe spec i e. In e na ional Con e ence on Business, Economics, Managemen and
Ma ke ing.
IMF. (2009). BPM6 - Balance o Paymen s and In e na ional In es men Posi ion Manual 6 h
Edi ion. Washing on, D.C: In e na ional Mone a y Fund.
Jackson, M. O. (2008). Social and Economic Ne wo ks. P ince on Uni e si y P ess.
53
Lima, F., Pinhei o, F., Sil a, J. F., & Ma os, P. (2020). Fo eign di ec in es men – using ne wo k
analysis o unde s and he posi ion o Po ugal in a global FDI ne wo k.
No g en, A., & Olsson, M. (2021). Ins i u ional Dynamics in he Global FDI Ne wo k. Linköping.
OECD. (2001, Sep embe 25). Glossa y o S a is ical Te ms. Re ie ed om OECD:
h ps://s a s.oecd.o g/glossa y/de ail.asp?ID=2092
OECD. (2008). OECD Benchma k De ini ion o Fo eign Di ec In es men . OECD.
P o a, L., & Be s o d, M. (2011). Eme ging Class Rela ions in he Mekong Ri e Del a o Vie nam:
A Ne wo k Analysis. Jou nal o Ag a ian Change, 60-80.
Sco , J. (1988). Social Ne wo k Analysis.
Sco , J. (2003). Ne wo ks o Co po a e Powe : A Compa a i e Assessmen . Annual Re iew o
Sociology, 181-203.
Sobel, R. S., J.R., C., & Dwig h, L. R. (2007). F eedom, ba ie s o en y, en ep eneu ship, and
economic p og ess.
Te Wal, A. L., & Boschma, R. A. (2008). Applying social ne wo k analysis in economic
geog aphy: aming some key analy ic issues.
Wasse man, S., & Faus , K. (1994). Social Ne wo k Analysis: Me hods and Applica ions.
Camb idge: Camb idge Uni e si y P ess.

Page | i