Fou akis, Dimi is
A icle
Iden i ica ion and isualiza ion o clus e s using ne wo k
heo y me hods: The case o he G eek p oduc ion sys em
Economies
P o ided in Coope a ion wi h:
MDPI – Mul idisciplina y Digi al Publishing Ins i u e, Basel
Sugges ed Ci a ion: Fou akis, Dimi is (2025) : Iden i ica ion and isualiza ion o clus e s using
ne wo k heo y me hods: The case o he G eek p oduc ion sys em, Economies, ISSN 2227-7099,
MDPI, Basel, Vol. 13, Iss. 1, pp. 1-31,
h ps://doi.o g/10.3390/economies13010015
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/329295
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/
Academic Edi o s: Pe iklis Gogas and
Theophilos Papadimi iou
Recei ed: 2 Decembe 2024
Re ised: 24 Decembe 2024
Accep ed: 27 Decembe 2024
Published: 11 Janua y 2025
Ci a ion: Fou akis, D. (2025).
Iden i ica ion and Visualiza ion o
Clus e s Using Ne wo k Theo y
Me hods: The Case o he G eek
P oduc ion Sys em. Economies,13(1),
15. h ps://doi.o g/10.3390/
economies13010015
Copy igh : © 2025 by he au ho .
Licensee MDPI, Basel, Swi ze land.
This a icle is an open access a icle
dis ibu ed unde he e ms and
condi ions o he C ea i e Commons
A ibu ion (CC BY) license
(h ps://c ea i ecommons.o g/
licenses/by/4.0/).
A icle
Iden i ica ion and Visualiza ion o Clus e s Using Ne wo k
Theo y Me hods: The Case o he G eek P oduc ion Sys em
Dimi is Fou akis
Depa men o Su eying and Geoin o ma ics Enginee ing (Se es), In e na ional Hellenic Uni e si y,
62124 Se es, G eece; [email p o ec ed] o [email p o ec ed]
Abs ac : The in e es in clus e s in he economy and egional space, which has pe sis ed
o nea ly h ee decades, has eigni ed he unde s anding o he economy as a sys em o
in e dependencies be ween indus ies. Al hough he concep o clus e s can be aced back
o con ibu ions da ing om he ea ly 20 h cen u y, hey ha e become a cen al ocus o
egional de elopmen policies in ecen decades, as hey ha e been linked o enhance-
men s o inno a ion, he knowledge economy, and ul ima ely, e i o ial compe i i eness.
A guably, he mos e ec i e and comp ehensi e way o p esen he sys emic na u e o
he economy is h ough inpu –ou pu ables. The main ea u e o hese ables, on which
his wo k is based, is ha hey desc ibe he ela ionships and lows be ween indus ies
(o p oduc s) du ing he p oduc ion p ocess. These undamen al ela ionships among he
indus ies in he p oduc ion sys em a e depic ed in he in e -indus y (and in a-indus y)
ansac ion ma ix o an economy’s inpu –ou pu ables. To analyze hese ela ionships,
we use ne wo k heo y, in he con ex o which he ansac ion ma ix can be seen as he
adjacency ma ix o a di ec ed, weigh ed g aph (o ne wo k) wi h loops. In his s udy,
clus e s a e iden i ied o he case o G eece, using wo di e en app oaches based on he
modula i y o he ne wo k, u ilizing he 2010 inpu –ou pu ables o his coun y. As
a esul , i e clus e s o indus ies ha s uc u e he coun y’s p oduc ion sys em ac oss
62 indus ies a e iden i ied, which a e also p esen ed h ough g aphical isualiza ions.
Keywo ds: clus e s; inpu –ou pu ables; g aph/ne wo k heo y; p oduc ion sys-
em; G eece
1. In oduc ion
The aim o his a icle is o p esen a me hod o iden i ying and isualizing clus e s,
using he p oduc ion sys em o G eece as a case s udy. Al hough he concep o clus e s
appea ed in economic heo y as “indus ial dis ic s” (Ma shall,1964, Chap e 10) as ea ly
as he la e 19 h cen u y (1890), i e-eme ged in wo k by Po e (1998) and o he s a he
end o he 20 h cen u y wi h a a he ambiguous con en .
1
Howe e , he geog aphical le el
o e e ence emains unclea : i could e e o a neighbo hood in a me opoli an a ea, a
egion (howe e de ined), a coun y, o e en g oups o coun ies (Ma in & Sunley,2003).
Fu he mo e, ambigui y exis s ega ding he ypes o indus ies in ol ed, whe he hey
a e simila o complemen a y, as well as he na u e o hei ela ionships. Despi e hese
unce ain ies, clus e s a e conside ed c ucial elemen s o egional policy wi hin a ious
e sions o new egionalism (Amin & Th i ,1995;Amin,1999;Cook,2002;Cook & Mo gan,
1998;Flo ida,1995;Mo gan & Nauwelae s,1999;Mo gan,1997;S o pe ,1997), as hey a e
seen as key d i e s o p omo ing inno a ion, he “knowledge economy”, and ul ima ely,
he compe i i eness o egions and coun ies.
Economies 2025,13, 15 h ps://doi.o g/10.3390/economies13010015
Economies 2025,13, 15 2 o 31
Rega dless o any concep ual ambigui ies o weaknesses, he discussion ega ding
clus e s b ings o he o e on he issue o he di ision o labo and he in e dependen
na u e o economic ac i i y. In his ega d, he o e all economic ac i i y unc ions as
a sys em o in e dependen , sepa a e economic ac i i ies ca ied ou by independen
economic uni s linked h ough exchange ela ionships, ei he as inpu s o as ou pu s
o in e media e o inal goods, means o p oduc ion, o se ices. The in ensi y o hese
ela ionships is no uni o mly dis ibu ed among indus ies bu ends o concen a e a ound
speci ic indus ies o g oups o indus ies and has a dynamic and changing cha ac e . The
geog aphical scale o indus ies’ concen a ion, which is add essed by con empo a y clus e
heo y, emains an open esea ch ques ion.
In his pape , a me hodological app oach o iden i ying and isualizing clus e s is
i s p esen ed. Then, in Sec ion 3, his app oach is applied, using G eece as a case s udy,
based on he analysis o he ansac ion ma ix om he 2010 G eek inpu –ou pu ables
(Eu os a ,2013), using ools om g aph/ne wo k heo y. In his sec ion, he esul s o
applying his me hodology a e also p esen ed, which leads o he iden i ica ion o i e
clus e s in he G eek p oduc ion sys em. Sec ion 4 ea u es a de ailed discussion abou
he composi ion and in e nal unc ion o hese clus e s, and he pape hen p oceeds o
Sec ion 5.2
2. Me hod o he Iden i ica ion and Visualiza ion o Clus e s, and Da a
Handling
2.1. Iden i ica ion o Clus e s
Iden i ying subdi isions o clus e s wi hin a ne wo k (g aph) is bo h concep ually and
compu a ionally complex, wi h nume ous me hods and algo i hms p oposed o sol ing
his p oblem (Fo una o,2010). In he con ex o inpu –ou pu ables, he exis ence o
unc ional subdi isions has been known abou since he 1950s and 1960s (see Leon ie ,
1986), aking o ms such as s a egic indus ies, indus y hie a chies, o cohesi e g oupings
(clus e s).
3
The me hodological app oach o his s udy iews he inpu –ou pu able as a
g aph (o ne wo k), a pe spec i e ha aligns wi h a scien i ic end om he ea ly 2000s,
which es ablished ne wo k science as a sepa a e ield ocused on eal-wo ld ne wo ks
(social, echnological, biological, cogni i e) (Newman,2010;Ba abási,2014).4
The ansac ion ma ix o he inpu –ou pu able is ea ed as he adjacency ma ix o
a di ec ed, weigh ed ne wo k (wi h o wi hou loops), allowing he use o he ne wo k’s
opological cha ac e is ics and associa ed quan i a i e da a (weigh s o links). This ame-
wo k enables he iden i ica ion o subg oups o clus e s based on he ansac ions be ween
indus ies. The sea ch o cohesi e s uc u es wi hin a complex ne wo k, desc ibed by a
ma ix o g aph, is a p oblem in ma ix and g aph heo y, which is known as pa i ioning.
The goal is o di ide he adjacency ma ix in o subma ices (subg aphs) in a way ha
minimizes he links be ween hem, o ming g oupings wi h mo e in e nal links (edges)
han ex e nal ones.
This p oblem is e y challenging o sol e p ac ically, e en in i s simples o m, which
is he bisec ion (spli ing in o wo equal pa s) o a g aph. In his seemingly s aigh o -
wa d case, he g aph, i.e., i s co esponding adjacency ma ix, mus be di ided in o wo
subg aphs (subma ices) wi h an equal numbe o nodes ( e ices). Howe e , he numbe
o possible ways o bisec he g aph is app oxima ely 2
(n+1)
/
√
n, whe e nis he numbe o
nodes/ e ices (Newman,2010, p. 359; Ba abási,2014, Chap e 9, pp. 8–9). This means ha
he numbe o solu ions inc eases exponen ially wi h he inc ease in he numbe o nodes
o he g aph, wi h he esul being ha when he numbe o nodes o he g aph exceeds a
ew dozen, inding and e alua ing solu ions is compu a ionally in easible in p ac ice. I he
numbe o di isions is no p ede e mined, he numbe o possible solu ions inc eases e en
Economies 2025,13, 15 3 o 31
as e han exponen ially, as desc ibed by he Bell numbe , which app oxima es he numbe
o pa i ions (Ba abási,2014, Chap e 9, p. 9; Fo una o,2010, p. 87).
The e a e nume ous app oaches o sol ing he p oblem o ne wo k di ision and nu-
me ous co esponding algo i hms, as de ailed in an ex ensi e e iew by Fo una o (2010).
The la ge numbe o me hods and algo i hms can be a ibu ed no only o he compu a-
ional complexi y o he p oblem—being NP-ha d, which undoub edly con ibu es o he
de elopmen o many algo i hms—bu also o he inhe en ambigui y in de ining wha
cons i u es a cohesi e subg oup. This ambigui y leads o mul iple, some imes sligh ly
a ied de ini ions and e ms being used o desc ibe he same phenomenon (o on ology).
In he li e a u e, e ms such as g aph pa i ioning,modules,clus e s, and communi ies a e used
in e changeably o al e na i ely (Newman,2010;Fo una o,2010;Fo una o & Cas ellano,
2012;Ba abási,2014). The e m “communi ies” is he mos ecen and is o en p e e ed in ap-
p oaches ha iden i y subs uc u es wi hou p ede ining he numbe o o he cha ac e is ics
o hese subg oups.
The algo i hms de eloped om such an app oach eme ged p ima ily du ing he ea ly
2000s, d i en signi ican ly by he in ol emen o physicis s in he ield. These “physicis s
en e ed he game, b inging in hei ools and echniques: spin models, op imiza ion,
pe cola ion, andom walks, synch oniza ion, e c., became ing edien s o new o iginal
algo i hms” (Fo una o & Cas ellano,2012, p. 492). In con as , ea lie algo i hms, which
elied on he p io speci ica ion o ce ain pa ame e s o he subdi isions o a g aph o
ne wo k unde s udy, we e p ima ily he esul o esea ch wi hin he domain o social
ne wo k analysis (Wasse man & Faus ,1994;Sco ,2000). K-means clus e ing and hie a chical
clus e ing a e he mos commonly used algo i hms in social ne wo k analysis. The i s
has he se ious limi a ion ha he numbe o clus e s o be o med mus be speci ied in
ad ance. The second does no ha e his limi a ion bu “i does no p o ide any a way o
disc imina e be ween he many pa i ions ob ained by he p ocedu e, and o choose ha o
hose ha be e ep esen he communi y s uc u e o he g aph” (Fo una o & Cas ellano,
2012, p. 498).
This di e en ia ion among algo i hms enables an ini ial classi ica ion o he a ious
app oaches based on a seemingly “ echnical” c i e ion. Speci ically, hese me hods a e
di ided in o (a) hose ha p ede e mine he numbe o pa i ions o he size o he g oupings
(numbe o nodes) o be o med, o he deg ee cen ali y (numbe o edges o links) o each
node in he subg aphs c ea ed, and (b) hose ha pa i ion he g aph wi hou any p io
speci ica ion o he numbe , size, o o he cha ac e is ics o he subdi isions (Newman,
2010, pp. 354–358).
The i s class o me hods ely on a amewo k o assump ions o a model ega ding
he g aph’s s uc u e, exp essed h ough pa ame e s ha de ine he numbe and ea u es
o he pa i ions. In con as , he second app oach does no make any assump ions abou
he g aph’s s uc u e o he numbe o size o i s cohesi e s uc u es and elies solely on
he g aph’s inhe en da a, p oducing a solu ion (o solu ions) by op imizing an e alua-
ion index de i ed om compa isons wi h a “null model”. This null model is a andom
g aph ha main ains ce ain s uc u al cha ac e is ics o he g aph unde s udy (Fo una o,
2010, p. 86).
5
I is c ucial o no e ha in bo h app oaches, pa i ioning is based solely on
opological ea u es.
This second me hod uses an index called he ne wo k modula i y, which compa es he
links (edges) wi hin each subg oup (clus e /communi y) in a p oposed pa i ion o he
numbe expec ed i he ne wo k was andom (Newman & Gi an,2004). Ma hema ically,
Economies 2025,13, 15 4 o 31
modula i y quan i ies his compa ison (Fo una o & Cas ellano,2012, p. 493; Ba abási,2014,
Chap e 9, p. 20) as ollows:
Q=∑nc
c=1"lc
m−dc
2m2#
whe e n
c
is he o al numbe o subg oups, l
c
is he numbe o edges wi hin subg oup c,m
is he o al numbe o edges in he ne wo k, and d
c
is he o al deg ee o he nodes wi hin
he same subg oup c. This ma hema ical o mula ion cla i ies ha modula i y depends on
he di e ence be ween wo quan i ies: he i s is he a io o he edges wi hin a subg oup
o he o al numbe o edges in he ne wo k, and he second is he a io o he expec ed
edges wi hin he same subg oup o he o al numbe o edges in he ne wo k, assuming
ha he ne wo k’s nodes a e andomly connec ed bu e ain hei o iginal deg ee.
In his s udy, he me hod o maximizing modula i y is used o iden i y clus e s wi hin
G eece’s p oduc ion sys em. Despi e ha ing some disad an ages, his me hod has a
signi ican ad an age: i does no equi e any p econcei ed judgmen o assump ion abou
he ne wo k’s s uc u e. This is unlike many es ablished me hods om social ne wo k
analysis and hie a chical clus e ing, which depend on such assump ions.
Howe e , his me hod is no wi hou i s p oblems: (a) i misclassi ies subg oups
ha a e smalle han a ce ain esolu ion limi , as i canno dis inguish hem om he
null model (Fo una o & Cas ellano,2012, p. 501), and (b) a he han ha ing a single
op imal maximum, he e is o en a ange o nea -op imal solu ions o ming a pla eau
ins ead o a unique peak (Ba abási,2014, Chap e 9, p. 24). Ne e heless, algo i hms based
on modula i y maximiza ion a e gene ally us ed by he scien i ic communi y and i is
conside ed ha “modula i y o e s a i s p inciple unde s anding o a ne wo k’s communi y
s uc u e” (Ba abási,2014, Chap e 9, p. 24).
Modula i y always akes alues less han one. Posi i e alues (Q> 0) indica e a good
pa i ion, wi h highe alues ep esen ing be e esul s. A alue o ze o (Q= 0) means
he e is no signi ican clus e ing, while nega i e alues (Q< 0) imply ha he andom
g aph’s in e nal connec ions a e be e han hose o he gi en subdi ision, indica ing an
unaccep able pa i ion. This subdi ision is conside ed unaccep able because a andom
g aph yields be e esul s. The index akes s ongly nega i e alues when each node in he
ne wo k is ea ed as a sepa a e subg oup. Gene ally, nega i e modula i y alues indica e
he absence o meaning ul subg oups o communi ies wi hin he ne wo k s uc u e. This
is a clea signal ha he di ision does no exhibi he expec ed communi y cha ac e is ics,
implying ha he ne wo k lacks cohesi e in e nal g oupings. In p ac ice, accep able
modula i y alues ange om 0.3 o 0.7, al hough highe alues may occasionally appea
(Newman & Gi an,2004, p. 7).
Maximizing modula i y is known o be an NP-ha d p oblem, meaning ha i can only
be app oxima ed using heu is ic and app oxima ion algo i hms (Fo una o & Cas ellano,
2012, p. 500). Fo una o (2010), in his comp ehensi e e iew o communi y de ec ion in
g aphs, discusses a ious classes o heu is ic algo i hms, such as g eedy, gene ic, simula ed
annealing, spec al, and abu sea ch, each o e ing di e en le els o success depending on
he ne wo k’s size and complexi y.
To di ide G eece’s p oduc ion sys em in o clus e s, wo di e en modula i y-based
app oaches we e used: (a) he Lou ain me hod epo ed by Blondel e al. (2008) and (b) he
Gi an–Newman me hod modi ied by A enas e al. (Gi an & Newman,2002;Newman,
2004;A enas e al.,2007,2008). Bo h me hods a e ailo ed o di ec ed, weigh ed ne wo ks
and use he modula i y index o e alua e he ne wo k pa i ions hey p oduce, ul ima ely
p oposing an op imal di ision in o clus e s.6
Economies 2025,13, 15 5 o 31
The i s me hod, Lou ain, agg ega es each node wi h o he s in he ne wo k, ecal-
cula ing he modula i y a each s ep un il he di ision ha maximizes he modula i y is
ound (Blondel e al.,2008, pp. 3–4). The second me hod successi ely spli s he ne wo k
by emo ing edges wi h he highes be weenness and ecalcula es he modula i y a e -
e y s age. These wo me hods di e in he algo i hms hey use o calcula e modula i y.
Gi en ha he compu a ion o modula i y alls in o he ca ego y o NP-ha d p oblems
(Fo una o & Cas ellano,2012, p. 500), i can only be add essed using s ochas ic, heu is ic
op imiza ion algo i hms. The Lou ain me hod employs a g eedy algo i hm (Fo una o,
2010,
pp. 101–102
), while he modi ied Gi an–Newman me hod, as adap ed by A enas
e al., uses a di e en op imiza ion app oach called abu sea ch. E en hough bo h me hods
use he modula i y as he c i e ion o e alua ing ne wo k pa i ions, hey undamen ally
di e in hei s a egies o achie e pa i ioning and in how hey op imize he modula i y.
The Lou ain me hod was implemen ed using Pajek ne wo k analysis so wa e 4.10 (M a
& Ba agelj,2016). A enas e al.’s me hod—a modi ied e sion o he Gi an–Newman
me hod—was implemen ed using Rada ools so wa e sui e 4.0 (Gómez & Fe nández,2016).
Thus, e en hough bo h app oaches seek o maximize he modula i y, hei me hodologies
and algo i hmic implemen a ions a e dis inc , a ec ing hei pe o mance and esul s in
e ms o ne wo k pa i ioning.
2.2. Visualiza ion o Resul s
The isualiza ion o g aphs/ne wo ks, and consequen ly, o clus e s, is an inhe en
ad an age and key ea u e o ne wo k analysis. Fo g aphs wi h a ela i ely limi ed numbe
o nodes ( e ices), such as a ew dozen, ep esen a ion can be ela i ely simple, o en
aking he o m o a g id. Ano he op ion is a ci cula diag am, which o e s an aes he ically
appealing esul (K zywinski e al.,2009;C no sanin e al.,2014). Howe e , o g aphs o
ne wo ks wi h se e al dozen o e en hund eds o nodes, hese me hods become imp ac ical.
In such cases, isualiza ion in ol es linking nodes/ e ices wi h lines whe e links exis .
The challenge in hese mo e complex ne wo ks is ha he e is no single way o
ep esen he ne wo k g aphically (Di Ba is a e al.,1994, p. 236). Ins ead, he e a e
heo e ically in ini e ways o a ange he nodes and edges, as shown in he ollowing igu e.
Fo ins ance, e en a simple g aph wi h ou e ices and i e edges can be ep esen ed in a
leas h ee equally alid ways (Figu e 1). These ep esen a ions main ain he ela ionships
be ween nodes ( e ices), as indica ed by he connec ing lines (edges/links), bu esul in
isually di e en diag ams.
Economies 2025, 13, x 5 o 32
s age. These wo me hods diffe in he algo i hms hey use o calcula e modula i y. Gi en
ha he compu a ion o modula i y alls in o he ca ego y o NP-ha d p oblems (Fo u-
na o & Cas ellano, 2012, p. 500), i can only be add essed using s ochas ic, heu is ic op i-
miza ion algo i hms. The Lou ain me hod employs a g eedy algo i hm (Fo una o, 2010,
pp. 101–102), while he modi ied Gi an–Newman me hod, as adap ed by A enas e al.,
uses a diffe en op imiza ion app oach called abu sea ch. E en hough bo h me hods use
he modula i y as he c i e ion o e alua ing ne wo k pa i ions, hey undamen ally di -
e in hei s a egies o achie e pa i ioning and in how hey op imize he modula i y.
The Lou ain me hod was implemen ed using Pajek ne wo k analysis so wa e 4.10
(M a & Ba agelj, 2016). A enas e al.’s me hod—a modi ied e sion o he Gi an–New-
man me hod—was implemen ed using Rada ools so wa e sui e 4.0 (Gómez & Fe nández,
2016). Thus, e en hough bo h app oaches seek o maximize he modula i y, hei me h-
odologies and algo i hmic implemen a ions a e dis inc , affec ing hei pe o mance and
esul s in e ms o ne wo k pa i ioning.
2.2. Visualiza ion o Resul s
The isualiza ion o g aphs/ne wo ks, and consequen ly, o clus e s, is an inhe en
ad an age and key ea u e o ne wo k analysis. Fo g aphs wi h a ela i ely limi ed num-
be o nodes ( e ices), such as a ew dozen, ep esen a ion can be ela i ely simple, o en
aking he o m o a g id. Ano he op ion is a ci cula diag am, which offe s an aes he i-
cally appealing esul (K zywinski e al., 2009; C no sanin e al., 2014). Howe e , o
g aphs o ne wo ks wi h se e al dozen o e en hund eds o nodes, hese me hods become
imp ac ical. In such cases, isualiza ion in ol es linking nodes/ e ices wi h lines whe e
links exis .
The challenge in hese mo e complex ne wo ks is ha he e is no single way o ep-
esen he ne wo k g aphically (Di Ba is a e al., 1994, p. 236). Ins ead, he e a e heo e i-
cally in ini e ways o a ange he nodes and edges, as shown in he ollowing igu e. Fo
ins ance, e en a simple g aph wi h ou e ices and i e edges can be ep esen ed in a
leas h ee equally alid ways (Figu e 1). These ep esen a ions main ain he ela ionships
be ween nodes ( e ices), as indica ed by he connec ing lines (edges/links), bu esul in
isually diffe en diag ams.
Figu e 1. Diffe en ep esen a ions o he same g aph.
The inhe en lexibili y in isualizing g aphs/ne wo ks, and consequen ly, clus e s,
in oduces subjec i i y in o how g aphs/ne wo ks a e ep esen ed. This cha ac e is ic o
g aphs ce ainly quali ies he alue o immediacy and ease in unde s anding he ela ion-
ships depic ed h ough a g aphical ep esen a ion. The inal isualiza ion o a g aph is no
a single, “objec i e” ma hema ical ou come bu a he a esul o a c ea i e design and,
ul ima ely, he esea che ’s in en . This lexibili y can po en ially lead o he “misguiding”
o he eade (o e en he esea che hemsel es) since i allows o he p esen a ion o
da a in a way ha migh emphasize ce ain aspec s o he ela ionships—which may no
be he mos impo an ones— ha a e desc ibed by he g aph.
This issue has led o he de elopmen o a sub ield wi hin g aph heo y ocused spe-
ci ically on isualiza ion. This a ea add esses wo in e connec ed co e conce ns: (a)
Figu e 1. Di e en ep esen a ions o he same g aph.
The inhe en lexibili y in isualizing g aphs/ne wo ks, and consequen ly, clus e s,
in oduces subjec i i y in o how g aphs/ne wo ks a e ep esen ed. This cha ac e is ic o
g aphs ce ainly quali ies he alue o immediacy and ease in unde s anding he ela ion-
ships depic ed h ough a g aphical ep esen a ion. The inal isualiza ion o a g aph is
no a single, “objec i e” ma hema ical ou come bu a he a esul o a c ea i e design and,
ul ima ely, he esea che ’s in en . This lexibili y can po en ially lead o he “misguiding”
o he eade (o e en he esea che hemsel es) since i allows o he p esen a ion o da a
Economies 2025,13, 15 6 o 31
in a way ha migh emphasize ce ain aspec s o he ela ionships—which may no be he
mos impo an ones— ha a e desc ibed by he g aph.
This issue has led o he de elopmen o a sub ield wi hin g aph heo y ocused speci -
ically on isualiza ion. This a ea add esses wo in e connec ed co e conce ns: (a) aes he ic
ules ha ensu e he eadabili y and unde s anding o he g aph and (b) he c ea ion o
algo i hms o e icien compu e -based g aph ende ing. The i s aspec is di ec ly ela ed
o he cla i y and in e p e abili y o g aphs, while he second ocuses on he compu a ional
e iciency o algo i hms used o isualiza ion. As Di Ba is a e al. (1999, p. 14) no e,
“aes he ics speci y g aphic p ope ies o he d awing ha we would like o apply, as much
as possible, o achie e eadabili y”. Examples o aes he ic ules include minimizing he
c ossing o edges, educing he o al diag am a ea, and ensu ing uni o mi y in he edge
leng h (Di Ba is a e al.,1999, pp. 15–16).
When i comes o algo i hms o implemen ing g aph design ules, i should be no ed
ha hese a e essen ially op imiza ion p oblems ha may no be simul aneously sol able
and a e also compu a ionally di icul (NP-ha d). Conside ing he issue o compu a ional
e iciency, i becomes necessa y o p io i ize he aes he ic c i e ia o g aph layou . As a
esul , he inal ou pu o hese algo i hms o en in ol es a combina ion o “app oxima ion
s a egies and heu is ics” (Di Ba is a e al.,1999, pp. 16–17). Despi e he a ie y o
algo i hms a ailable o g aph d awing, hey gene ally sha e a common ounda ion. Many
algo i hms used o achie e he inal layou ely on a o ce-di ec ed app oach, a model
inspi ed by he basic p inciples o mechanics (Eades,1984). The co e idea in ol es ea ing
he edges (links) as sp ings, which helps o dis ibu e he nodes ( e ices) in a balanced
way ac oss a wo-dimensional space, such as a compu e sc een, a p in e , o a plo e .
Va ia ions o his app oach unde pin he algo i hms used in mode n so wa e o social
ne wo k analysis and specialized ma hema ical ools.
Fo his a icle, he F uch e man and Reingold (1991) algo i hm was used, inco po-
a ing Lomba di-s yle cu es o aes he ically enhance he diag ams by gi ing he edges a
cu ed shape (Duncan e al.,2012). The ne wo k diag am was c ea ed using Gephi e . 0.901
so wa e (Bas ian e al.,2009;Jacomy e al.,2014). The g id-based layou was c ea ed using
Pajek e . 4.10 so wa e (M a & Ba agelj,2016), and he ci cula diag am was c ea ed wi h
Ci cos Table Viewe . 0.63-10 so wa e (K zywinski e al.,2009).
2.3. Handling o Da a
The da a used in his s udy we e om he domes ic inpu –ou pu able o he yea 2010
(p oduc by p oduc ) in cu en basic p ices (in millions o EUR). This able o igina es om he
o icial inpu –ou pu ables o G eece in 2010 (Eu os a ,2013), p o ided by he Eu opean
S a is ical O ice (Eu os a ).
7
The classi ica ion o p oduc s and indus ies ollows he
Classi ica ion o P oduc s by Ac i i y-CPA 2008 (Eu opean Union,2008) o goods and se ices,
and he NACE Re . 2 S a is ical classi ica ion o economic ac i i ies in he Eu opean Communi y
(Eu os a ,2008) o economic ac i i ies. The Eu os a da a ca ego ize 65 p oduc s/indus ies:
64 acco ding o CPA/NACE classi ica ions plus 1 addi ional indus y, “L68A: Impu ed
en s o owne -occupied dwellings”.
Fo simpli ica ion, h ee indus ies wi h ze o o excep ionally low alues in he ans-
ac ions ma ix we e emo ed, as well as i e nega i e alues ha we e economically
meaningless, because hey e e sed he di ec ion o lows and impac s. These alues
appea ed in he ou pu o wo indus ies: N78 (“Employmen se ices”) and I (“Accommo-
da ion and ood se ices”). Speci ically, he nega i e alues amoun ed o EUR 0.48 million
o N78 (ac oss h ee en ies) and EUR 1.06 million o I (ac oss wo en ies). These we e
adjus ed o ze o, wi h a negligible o e all impac , as he ma ix’s o al ansac ions amoun
o app oxima ely EUR 111 billion, esul ing in an e ec o a ound 0.01‰.
Economies 2025,13, 15 7 o 31
Addi ionally, h ee indus ies wi h negligible o ze o p esence we e comple ely e-
mo ed: U (“Se ices p o ided by ex a e i o ial o ganiza ions and bodies”), L68A (“Im-
pu ed en s o owne -occupied dwellings”), and T (“Se ices o households as employe s;
undi e en ia ed goods and se ices p oduced by households o own use”). These indus-
ies made no con ibu ion o he added alue and did no show any signi ican ansac ions.
Following he adjus men s men ioned p e iously, he dimensions o he ma ix became
62
×
62 ( educed om he 65
×
65 e sion o icially published by Eu os a ), con aining
a o al o 3844 en ies. O hese, 416 en ies (10.8%) we e ze os, compa ed o 780 ze os
(18.5%) in he o iginal ma ix. Fu he simpli ica ion elimina ed e y small ansac ion
alues, which we e eplaced wi h ze o. As a esul , he numbe o ze o en ies inc eased
o 2328 (60.6% o he o al). I should be no ed ha his adjus men p ese ed 98% o he
o iginal sum o alues om he 62
×
62 ma ix be o e he in e en ions, amoun ing o EUR
109.077 billion ou o EUR 111.614 billion. This modi ied ma ix was subsequen ly used o
clus e iden i ica ion and isualiza ion using he me hod desc ibed ea lie .
In he p esen a ion and commen a y on he esul s, indus ies appea ing in clus e s
we e ca ego ized by echnology le el and knowledge in ensi y. This ca ego iza ion ollowed
Eu os a ’s guidelines (see Eu os a ,2016). The in ensi y o esea ch and echnological
de elopmen (R&D) was used o classi y wo-digi manu ac u ing indus ies, using R&D
expendi u es as a pe cen age o g oss alue added. Fou ca ego ies eme ged: High Tech-
nology (HT), Medium–High Technology (MHT), Medium–Low Technology (MLT), and
Low Technology (LT).
Fo se ice indus ies, he c i e ion was he pe cen age o employees in he co espond-
ing wo-digi economic ac i i y indus ies who ha e a e ia y educa ion deg ee. These
se ice indus ies we e classi ied in line wi h he wo main ca ego ies in manu ac u ing
(i.e., high and low echnology) in o wo p ima y g oups: Knowledge-In ensi e Se ices
(KIS) and Less-Knowledge-In ensi e Se ices (LKIS). Fu he mo e, he i s ca ego y (KIS)
was subdi ided in o Knowledge-In ensi e Ma ke Se ices (KI_m_S), High Technology
Knowledge-In ensi e Se ices (H _KIS), Knowledge-In ensi e Financial Se ices (KI_ _S),
and O he Knowledge-In ensi e Se ices (O_KIS). Less-Knowledge-In ensi e Se ices
we e subdi ided in o Less-Knowledge-In ensi e Ma ke Se ices (LKI_m_S) and O he
Less-Knowledge-In ensi e Se ices (O_LKIS).
I should be emphasized ha all he p e ious echnology and knowledge in ensi y
ca ego ies exclude he h ee indus ies o he p ima y sec o as well as ce ain seconda y
sec o indus ies, namely Mining (B), Elec ici y–Gas (D35), Wa e Supply (E36), Was e
Managemen (E37–E39), and Cons uc ion (F). Due o he absence o an “o icial” classi ica-
ion om Eu os a , we designa ed hese indus ies as “ adi ional” and ma ked hem wi h
he label “TR” in he ela ed ables.
In he ables and commen a y ha ollow, indus ies a e iden i ied using he NACE
Re . 2 classi ica ion codes. The desc ip ion o he indus ies used o p esen a ion pu poses
is a sho ened e sion o hei o icial desc ip ion, which is p o ided in Appendix A.
3. Resul s
3.1. The Clus e s and he Reliabili y o he Di ision
The ou come o di iding he ansac ion ma ix in o i e clus e s is shown in Table 1. I
is impo an o no e ha , using bo h me hods o maximizing he modula i y ( he Lou ain
me hod and he Gi an–Newman me hod modi ied by A enas e al.), he ini ial di ision
p oduced a six h “clus e ”, which consis ed o he “isola ed” indus y H53 Pos al Se ices.
A e a sepa a e analysis o i s connec ions and o achie e a mo e cohe en p esen a ion o
he esul s, his indus y was in eg a ed in o he “Mega-clus e ”. Due o he la ge size o
Economies 2025,13, 15 8 o 31
he Mega-clus e (24 indus ies), inco po a ing Pos al Se ices as i s 25 h membe did no
a ec i s o e all cha ac e .
Table 1. Clus e s in he p oduc ion sys em o G eece, 2010.
Code
Desc ip ion o Clus e and Indus y
Code
Desc ip ion o Clus e and Indus y
A. AGRICULTURE–TOURISM D. KNOWLEDGE–EDUCATION (con .)
1. A01 Ag icul u e 6. M72 Resea ch and de elopmen
2. A02 Fo es y 7. M73 Ad e ising
3. A03 Fishing 8. M74_M75 * O he scien i ic ac i i ies
4. C10–C12 Food–Be e ages 9. N78 Employmen ac i i ies
5. C33 Repai /ins alla ion o machine y
10.
P85 Educa ion
6. I Accommoda ion–Res au an s
11.
R90-R92 C ea i e ac i i ies—Gambling
7. S94 Membe ship o ganiza ions
12.
R93 Spo s–Rec ea ion
8. S96 Pe sonal se ices E. MEGA-CLUSTER
B. ENERGY–TRANSPORT 1. C13–C15 Tex iles–Appa el
1. B Mining 2. C17 Pape
2. C19 Pe oleum p oduc s 3. C20 Chemicals
3. D35 Elec ici y–Gas 4. C21 Pha maceu icals
4. E36 Wa e supply 5. C22 Plas ic p oduc s
5. H49 Land anspo 6. C26 Compu e s–Elec onics
6. H50 Wa e anspo 7. C29 Mo o ehicles
7. H51 Ai anspo 8. C30 O he anspo equipmen
8. H52 Wa ehousing 9. C31_C32 Fu ni u e–O he manu ac u ing
9. N77 Ren al/leasing ac i i ies
10.
E37–E39 Was e managemen
C. CONSTRUCTION
11.
G45 T ade and epai o mo o ehicles
1. C16 Wood
12.
G46 Wholesale ade
2. C23 Non-me allic mine al p oduc s
13.
G47 Re ail ade
3. C24 Basic me als
14.
H53 ** Pos al ac i i ies
4. C25 Me al p oduc s
15.
K64 Financial se ices
5. C27 Elec ical equipmen
16.
K65 Insu ance
6. C28 Machine y
17.
K66 O he inancial se ices
7. F Cons uc ion
18.
L68 Real es a e
8. M71 A chi ec s–Enginee s
19.
M69_M70 Legal, accoun ing, managemen
ac i i ies
D. KNOWLEDGE–EDUCATION
20.
N79 T a el agencies
1. C18 P in ing
21.
N80-N82 Secu i y, se ices o buildings
2. J58 Publishing
22.
O84 Public adminis a ion, de ense
3. J59_J60 Cinema–Tele ision
23.
Q86 Heal h
4. J61 Telecommunica ions
24.
Q87_Q88 Social ca e
5. J62_J63 Compu e –In o ma ion se ices
25.
S95 *
Repai o compu e s and household goods
Sou ce: Modi ied ansac ion ma ix. Da a p ocessed wi h he Lou ain and A enas e al. me hods. (*) These
indus ies we e appoin ed by he Lou ain me hod: (a) M74_M75 o clus e E, (b) S95 o clus e C. (**) The indus y
H53 in he o iginal di isions o bo h me hods was a sepa a e “clus e ”.
Di e ences be ween he wo me hods we e obse ed in only wo cases:
1.
Indus y M74_M75 O he Scien i ic Se ices: The Lou ain me hod placed i in he
“Mega-clus e ”, while he A enas e al. me hod (i.e., he modi ied G-N me hod) placed
i in he “Knowledge–Educa ion” clus e .
2.
Indus y S95 Repai o Compu e s and Household Appliances: The Lou ain me hod placed
i in he “Cons uc ion” clus e , while he A enas e al. me hod placed i in he
“Mega-clus e ”.
Economies 2025,13, 15 15 o 31
p ominence o A01 Ag icul u e, C10–C12 Food–Be e ages, I Accommoda ion–Res au an s,
and S94 Membe ship O ganiza ions, which p ima ily ecei e ou pu s om indus ies I
and C10–C12.
Economies 2025, 13, x 15 o 32
di ec ed owa d in e na ional ma ke s, pa icula ly hose o indus ies A03, A01, and C10-
C12, and I, al hough o he la e ( ou ism om ab oad), his is no cap u ed in he inpu –
ou pu ables.
F om he ci cula diag am (Figu e 5), i is appa en ha his clus e ’s s uc u e e ol es
a ound h ee key indus ies, namely Ag icul u e, Food–Be e ages, and Accommoda ion–
Res au an s, wi h he o he indus ies playing a complemen a y ole. This s uc u e is il-
lus a ed diffe en ly in he ne wo k diag am (Figu e 6), whe e he impo ance o each in-
dus y is depic ed by he size o he ne wo k nodes ( e ices) and labels. This impo ance
is measu ed using he eigen ec o cen ali y index (“PageRank”), and he hickness o he
lines ep esen s he olume o ansac ions. The ne wo k diag am highligh s he p omi-
nence o A01 Ag icul u e, C10–C12 Food–Be e ages, I Accommoda ion–Res au an s, and
S94 Membe ship O ganiza ions, which p ima ily ecei e ou pu s om indus ies I and
C10–C12.
Figu e 5. Ci cula diag am o he Ag icul u e–Tou ism clus e .
Figu e 6. Ne wo k diag am o he Ag icul u e–Tou ism clus e .
Figu e 5. Ci cula diag am o he Ag icul u e–Tou ism clus e .
Economies 2025, 13, x 15 o 32
di ec ed owa d in e na ional ma ke s, pa icula ly hose o indus ies A03, A01, and C10-
C12, and I, al hough o he la e ( ou ism om ab oad), his is no cap u ed in he inpu –
ou pu ables.
F om he ci cula diag am (Figu e 5), i is appa en ha his clus e ’s s uc u e e ol es
a ound h ee key indus ies, namely Ag icul u e, Food–Be e ages, and Accommoda ion–
Res au an s, wi h he o he indus ies playing a complemen a y ole. This s uc u e is il-
lus a ed diffe en ly in he ne wo k diag am (Figu e 6), whe e he impo ance o each in-
dus y is depic ed by he size o he ne wo k nodes ( e ices) and labels. This impo ance
is measu ed using he eigen ec o cen ali y index (“PageRank”), and he hickness o he
lines ep esen s he olume o ansac ions. The ne wo k diag am highligh s he p omi-
nence o A01 Ag icul u e, C10–C12 Food–Be e ages, I Accommoda ion–Res au an s, and
S94 Membe ship O ganiza ions, which p ima ily ecei e ou pu s om indus ies I and
C10–C12.
Figu e 5. Ci cula diag am o he Ag icul u e–Tou ism clus e .
Figu e 6. Ne wo k diag am o he Ag icul u e–Tou ism clus e .
Figu e 6. Ne wo k diag am o he Ag icul u e–Tou ism clus e .
Finally, om a echnological pe spec i e, as shown in Table 3, his is a low- echnology
and less-knowledge-in ensi e clus e . Only C33 Repai /Ins alla ion o Machine y is a
medium–low echnology indus y, while he o he s a e ei he adi ional ( he h ee p ima y
sec o indus ies) o low- echnology and less-knowledge-in ensi e indus ies.
4.3. Ene gy–T anspo Clus e
The Ene gy–T anspo clus e (Figu es 7and 8) consis s o nine indus ies: B Mining,
C19 Pe oleum P oduc s, D35 Elec ici y–Gas, E36 Wa e Supply, H49 Land T anspo , H50
Economies 2025,13, 15 16 o 31
Wa e T anspo , H51 Ai T anspo , H52 Wa ehousing, and N77 Ren al and Leasing Ac i i-
ies. As shown in Table 4, his clus e could al e na i ely be desc ibed as he Expo clus e ,
as i accoun s o mo e han hal o he coun y’s expo s (52.5%). Howe e , conside ing
ha all he anspo and logis ics se ices along wi h he ene gy p oduc ion indus ies a e
ound in his clus e , i is app op ia e o name i he Ene gy–T anspo clus e .
Economies 2025, 13, x 17 o 32
D35 Elec ici y–Gas, while H50 Wa e T anspo ecei es inpu s mainly om H52 Wa ehous-
ing and C19 Pe oleum P oduc s, wi h minimal ou pu s.
Figu e 7. Ci cula diag am o he Ene gy–T anspo clus e .
Figu e 8. Ne wo k diag am o he Ene gy–T anspo clus e .
Conside ing he o e all cha ac e is ics o he Ene gy–T anspo clus e and he in-
dus ies ha comp ise i , his clus e eme ges as a c i ical componen o he p oduc ion
sys em, p ima ily due o i s s ong o ien a ion owa d he in e na ional ma ke . Se e al o
i s indus ies a e in e na ionally compe i i e, no ably H50 Wa e T anspo , C19 Pe oleum
P oduc s, and H52 Wa ehousing.
Figu e 7. Ci cula diag am o he Ene gy–T anspo clus e .
Table 4. Ene gy–T anspo clus e : main economic cha ac e is ics.
NACE
Code Desc ip ion
Technology
T ansac ions
Inside Clus e Empl. (%) Demand
(%)
Added Value Expo s
Ou pu
Inpu (%) Technical
Coe icien (%)
Ex o e sion
B Mining TR 0.98 0.30 0.3 0.1 0.3 47.5 0.4 12.4
C19 Pe oleum p oduc s MLT 0.56 0.48 0.2 3.8 1.0 14.5 10.7 31.3
D35 Elec ici y–Gas TR 0.42 0.72 0.6 1.8 2.3 53.8 0.5 2.2
E36 Wa e supply TR 0.10 0.50 0.2 0.2 0.2 48.1 0 0
H49 Land anspo LKI_m_S 0.19 0.38 2.4 2.1 1.5 43.3 0.5 3.0
H50 Wa e anspo KI_m_S 0.41 0.47 0.7 6.2 3.7 47.0 37.4 96.0
H51 Ai anspo KI_m_S 0.28 0.54 0.2 0.6 0.3 28.1 1 18.2
H52 Wa ehousing LKI_m_S 0.80 0.45 0.9 0.3 0.6 42.0 1.9 27.6
N77 Ren al/leasing ac i i ies LKI_m_S 0.38 0.16 0.1 0.1 0.3 51.1 0.1 3.4
To al Ene gy–T anspo Clus e 0.50 5.6 15.3 10.2 38.3 52.5 39.5
This clus e has mode a e cohesion, wi h he alue o he in e -indus y ansac ions
di ided equally be ween in e nal and ex e nal ansac ions (50–50). A he indus y le el,
Mining (B),Elec ici y–Gas (D35),Pe oleum P oduc s (C19), and Wa ehousing (H52) display a
clea “in o e si e” o ien a ion, meaning ha mos o hei ansac ion alue occu s wi h
o he indus ies wi hin he clus e a he han wi h he es o he p oduc ion sys em, as
indica ed in he ele an columns (Ou pu s, Inpu s) o he able. The emaining indus ies
in he clus e exhibi a ying deg ees o olume o ex e nal ansac ions, in e ac ing mo e
wi h indus ies ou side he clus e .
Economies 2025,13, 15 17 o 31
Economies 2025, 13, x 17 o 32
D35 Elec ici y–Gas, while H50 Wa e T anspo ecei es inpu s mainly om H52 Wa ehous-
ing and C19 Pe oleum P oduc s, wi h minimal ou pu s.
Figu e 7. Ci cula diag am o he Ene gy–T anspo clus e .
Figu e 8. Ne wo k diag am o he Ene gy–T anspo clus e .
Conside ing he o e all cha ac e is ics o he Ene gy–T anspo clus e and he in-
dus ies ha comp ise i , his clus e eme ges as a c i ical componen o he p oduc ion
sys em, p ima ily due o i s s ong o ien a ion owa d he in e na ional ma ke . Se e al o
i s indus ies a e in e na ionally compe i i e, no ably H50 Wa e T anspo , C19 Pe oleum
P oduc s, and H52 Wa ehousing.
Figu e 8. Ne wo k diag am o he Ene gy–T anspo clus e .
The Ene gy–T anspo clus e makes a ela i ely low con ibu ion o employmen
(5.6%) and has a low echnical coe icien o added alue (38.3%). This indica es ha his
clus e is cha ac e ized by low added alue o , con e sely, a high in ensi y o in e media e
inpu s. Mos o he indus ies in his clus e a e classi ied as low- echnology and less-
knowledge-in ensi e. Only indus ies H50 Wa e T anspo and H51 Ai T anspo a e
classi ied as knowledge-in ensi e (KI_m_S), while he C19 Pe oleum P oduc s indus y is
medium–low echnology (MLT). In Figu es 7and 8, he clus e ’s s uc u e is illus a ed,
which is cen e ed a ound Mining (B),Pe oleum P oduc s (C19),Elec ici y–Gas (D35), and
T anspo a ion (p ima ily ma i ime H50 bu also land H49). Complemen ing his co e
s uc u e a e indus ies such as H52 Wa ehousing,H51 Ai T anspo ,N77 Ren al/Leasing
Ac i i ies, and E36 Wa e Supply. The D35 Elec ici y–Gas indus y se es as a supplie o
all he o he indus ies wi hin he clus e , as do C19 Pe oleum P oduc s and N77 Ren al and
Leasing Ac i i ies. The B Mining indus y p ima ily supplies C19 Pe oleum P oduc s and D35
Elec ici y–Gas, while H50 Wa e T anspo ecei es inpu s mainly om H52 Wa ehousing and
C19 Pe oleum P oduc s, wi h minimal ou pu s.
Conside ing he o e all cha ac e is ics o he Ene gy–T anspo clus e and he in-
dus ies ha comp ise i , his clus e eme ges as a c i ical componen o he p oduc ion
sys em, p ima ily due o i s s ong o ien a ion owa d he in e na ional ma ke . Se e al o
i s indus ies a e in e na ionally compe i i e, no ably H50 Wa e T anspo ,C19 Pe oleum
P oduc s, and H52 Wa ehousing.
4.4. Cons uc ion Clus e
The Cons uc ion clus e (Figu es 9and 10) is composed o eigh indus ies, including
six manu ac u ing indus ies: C16 Wood,C23 Non-Me allic P oduc s,C24 Basic Me als,C25
Me al P oduc s,C27 Elec ical Equipmen , and C28 Machine y. I also includes F Cons uc ion
and M71 A chi ec s–Enginee s om he se ice sec o . This clus e is pa icula ly cohesi e,
wi h o e wo- hi ds o i s ansac ions (0.68) occu ing in e nally (Table 5). Mos indus ies,
excep o C28 Machine y and M71 A chi ec s–Enginee s, engage in in e nal ansac ions in
e ms o ei he ou pu s o inpu s. I s name unde lines he cen al ole o he Cons uc ion
indus y in he clus e .
Economies 2025,13, 15 18 o 31
Economies 2025, 13, x 18 o 32
4.4. Cons uc ion Clus e
The Cons uc ion clus e (Figu es 9 and 10) is composed o eigh indus ies, including
six manu ac u ing indus ies: C16 Wood, C23 Non-Me allic P oduc s, C24 Basic Me als, C25
Me al P oduc s, C27 Elec ical Equipmen , and C28 Machine y. I also includes F Cons uc ion
and M71 A chi ec s–Enginee s om he se ice sec o . This clus e is pa icula ly cohesi e,
wi h o e wo- hi ds o i s ansac ions (0.68) occu ing in e nally (Table 5). Mos indus-
ies, excep o C28 Machine y and M71 A chi ec s–Enginee s, engage in in e nal ansac-
ions in e ms o ei he ou pu s o inpu s. I s name unde lines he cen al ole o he Con-
s uc ion indus y in he clus e .
Figu e 9. Ci cula diag am o he Cons uc ion clus e .
The Cons uc ion clus e is he smalles o he i e clus e s in e ms o p oduc ion
olume, con ibu ing only 7.3% o he o al g oss alue added (GVA). Nea ly hal o his
ou pu (3.6%) comes solely om he Cons uc ion (F) indus y. The o he hal comes om
he o he i e indus ies, each con ibu ing less han 1% o he o al GVA o he p oduc ion
sys em. The added alue echnical coefficien is he lowes among all i e clus e s (33.5%)
due o he excep ionally low alues ac oss i s indus ies, some o he lowes in he p o-
duc ion sys em.
Gene ally, apa om he Cons uc ion indus y, his clus e consis s o ela i ely
small indus ies. This is e lec ed in he modes sha e o o al employmen (12.7%) and
demand (10.7%) o he clus e , wi h he Cons uc ion indus y accoun ing o he majo i y
(7.5% in employmen and 8.1% in demand). Despi e he small size o i s indus ies, he
Cons uc ion clus e has a no able expo con ibu ion (11.9% o o al expo s o he p o-
duc ion sys em), mainly om i s six manu ac u ing indus ies. The C24 Basic Me als
Figu e 9. Ci cula diag am o he Cons uc ion clus e .
Economies 2025, 13, x 19 o 32
indus y leads expo s (5.2%), ollowed by he C27 Elec ical Equipmen indus y (2.1%),
wi h he emaining ou Manu ac u ing indus ies con ibu ing 2.9%. This expo s eng h
is e iden in he high expo index (Ex o e sion) o mos Manu ac u ing indus ies in he
clus e .
Table 5. Cons uc ion clus e : main economic cha ac e is ics.
NACE
Code Desc ip ion Technology
T ansac ions
Inside Clus e Empl.
(%)
Demand
(%)
Added Value Expo s
Ou pu Inpu (%) Technical
Coe icien (%) Ex o e sion
C16 Wood LT 0.67 0.54 0.5 0.0 0.2 29.7 0.1 3.1
C23 Non-me allic mine al
p oduc s MLT 0.96 0.34 0.6 0.2 0.7 50.5 0.8 11.1
C24 Basic me als MLT 0.83 0.52 0.5 0.8 0.6 24.4 5.2 43.5
C25 Me al p oduc s MLT 0.65 0.48 1.2 0.4 0.7 33.3 0.7 6.4
C27 Elec ical equipmen MHT 0.62 0.47 0.3 0.4 0.3 42.8 2.1 67.9
C28 Machine y MHT 0.31 0.42 0.2 0.5 0.3 50.1 1.3 42.9
F Cons uc ion TR 0.50 0.64 7.5 8.1 3.6 31.8 1.4 2.5
M71 A chi ec s–Enginee s ΚΙ_m_S 0.58 0.09 1.5 0.3 0.9 35.4 0.3 2.3
To al Cons uc ion Clus e 0.68 12.2 10.7 7.3 33.5 11.9 10.9
Figu e 10. Ne wo k diag am o he Cons uc ion clus e .
Technologically, mos indus ies in he Cons uc ion clus e a e o low o medium–
low echnology (LT, MLT), apa om M71 A chi ec s–Enginee s, which is knowledge-
in ensi e (ΚΙ_m_S), while he Cons uc ion (F) indus y is adi ional (T). Howe e , he
cen al ole o he Cons uc ion indus y enables i o ac i a e o he indus ies in he clus-
e h ough i s backwa d linkages, 65% o which a e in e nal. Mos indus ies, wi h he
in e nal o wa d linkages anging om 58% o 96%, a e in e connec ed wi hin he clus e .
This means ha changes in he demand o he Cons uc ion indus y can p opaga e
h ough in e -indus y ela ionships and s imula e demand o highe - echnology and
knowledge-in ensi e indus ies, such as C27 Elec ical Equipmen (medium–high echnol-
ogy), C28 Machine y (medium–high echnology), and M71 A chi ec s–Enginee s
(knowledge-in ensi e).
Figu e 10. Ne wo k diag am o he Cons uc ion clus e .
Economies 2025,13, 15 19 o 31
Table 5. Cons uc ion clus e : main economic cha ac e is ics.
NACE
Code Desc ip ion
Technology
T ansac ions
Inside Clus e
Empl. (%) Demand
(%)
Added Value Expo s
Ou pu
Inpu (%) Technical
Coe icien (%)
Ex o e sion
C16 Wood LT 0.67 0.54 0.5 0.0 0.2 29.7 0.1 3.1
C23
Non-me allic mine al p oduc s
MLT 0.96 0.34 0.6 0.2 0.7 50.5 0.8 11.1
C24 Basic me als MLT 0.83 0.52 0.5 0.8 0.6 24.4 5.2 43.5
C25 Me al p oduc s MLT 0.65 0.48 1.2 0.4 0.7 33.3 0.7 6.4
C27 Elec ical equipmen MHT 0.62 0.47 0.3 0.4 0.3 42.8 2.1 67.9
C28 Machine y MHT 0.31 0.42 0.2 0.5 0.3 50.1 1.3 42.9
F Cons uc ion TR 0.50 0.64 7.5 8.1 3.6 31.8 1.4 2.5
M71 A chi ec s–Enginee s KI_m_S 0.58 0.09 1.5 0.3 0.9 35.4 0.3 2.3
To al Cons uc ion Clus e 0.68 12.2 10.7 7.3 33.5 11.9 10.9
The Cons uc ion clus e is he smalles o he i e clus e s in e ms o p oduc ion
olume, con ibu ing only 7.3% o he o al g oss alue added (GVA). Nea ly hal o his
ou pu (3.6%) comes solely om he Cons uc ion (F) indus y. The o he hal comes
om he o he i e indus ies, each con ibu ing less han 1% o he o al GVA o he
p oduc ion sys em. The added alue echnical coe icien is he lowes among all i e
clus e s (33.5%) due o he excep ionally low alues ac oss i s indus ies, some o he lowes
in he p oduc ion sys em.
Gene ally, apa om he Cons uc ion indus y, his clus e consis s o ela i ely small
indus ies. This is e lec ed in he modes sha e o o al employmen (12.7%) and demand
(10.7%) o he clus e , wi h he Cons uc ion indus y accoun ing o he majo i y (7.5% in
employmen and 8.1% in demand). Despi e he small size o i s indus ies, he Cons uc ion
clus e has a no able expo con ibu ion (11.9% o o al expo s o he p oduc ion sys em),
mainly om i s six manu ac u ing indus ies. The C24 Basic Me als indus y leads expo s
(5.2%), ollowed by he C27 Elec ical Equipmen indus y (2.1%), wi h he emaining ou
Manu ac u ing indus ies con ibu ing 2.9%. This expo s eng h is e iden in he high
expo index (Ex o e sion) o mos Manu ac u ing indus ies in he clus e .
Technologically, mos indus ies in he Cons uc ion clus e a e o low o medium–low
echnology (LT, MLT), apa om M71 A chi ec s–Enginee s, which is knowledge-in ensi e
(KI_m_S), while he Cons uc ion (F) indus y is adi ional (T). Howe e , he cen al ole
o he Cons uc ion indus y enables i o ac i a e o he indus ies in he clus e h ough
i s backwa d linkages, 65% o which a e in e nal. Mos indus ies, wi h he in e nal
o wa d linkages anging om 58% o 96%, a e in e connec ed wi hin he clus e . This
means ha changes in he demand o he Cons uc ion indus y can p opaga e h ough
in e -indus y ela ionships and s imula e demand o highe - echnology and knowledge-
in ensi e indus ies, such as C27 Elec ical Equipmen (medium–high echnology), C28
Machine y (medium–high echnology), and M71 A chi ec s–Enginee s (knowledge-in ensi e).
4.5. Knowledge–Educa ion Clus e
The Knowledge–Educa ion clus e is composed o 12 indus ies; one is a Manu ac u -
ing indus y and 11 a e om he Se ice sec o (Figu es 11 and 12). I s name highligh s
wo p ima y cha ac e is ics: he economic signi icance o he Educa ion (P85) indus y
and he high concen a ion o se ice indus ies in his clus e , wi h 11 ou o a o al o
21 knowledge-in ensi e se ice indus ies in he p oduc ion sys em being in his clus e .
Al e na i ely, i could also be called he Knowledge–Communica ion clus e , as a leas six
o i s indus ies a e in ol ed in p oducing, suppo ing, and managing adi ional mass
and newe In e ne -based in e ac i e communica ion media. These include C18 P in ing,
J58 Publishing,J59_J60 Cinema–TV,J61 Telecommunica ions,J62_J63 Compu e –In o ma ion
Se ices, and M73 Ad e ising. Addi ionally, his clus e inco po a es ac i i ies om he
R90-R92 C ea i e Ac i i ies and Gambling indus y, which include he R90 C ea i e A s and
En e ainmen indus y.
Economies 2025,13, 15 20 o 31
Economies 2025, 13, x 21 o 32
coefficien a ises om he ex emely high alues in ce ain indus ies. Mos indus ies in
his clus e use minimal in e media e inpu s, and in some cases, such as J61 Telecommuni-
ca ions, J58 Publishing, and R90–R92 C ea i e Ac i i ies–Gambling, a signi ican po ion o
hei ansac ions a e in a-indus y.
Figu e 11. Ci cula diag am o he Knowledge–Educa ion clus e .
Figu e 12. Ne wo k diag am o he Knowledge–Educa ion clus e .
On he o he hand, M73 Ad e ising (9.8%) and M72 Resea ch and De elopmen (27.1%)
ha e ela i ely high olumes o inpu s. This clus e pe o ms pa icula ly poo ly in
Figu e 11. Ci cula diag am o he Knowledge–Educa ion clus e .
Economies 2025, 13, x 21 o 32
coefficien a ises om he ex emely high alues in ce ain indus ies. Mos indus ies in
his clus e use minimal in e media e inpu s, and in some cases, such as J61 Telecommuni-
ca ions, J58 Publishing, and R90–R92 C ea i e Ac i i ies–Gambling, a signi ican po ion o
hei ansac ions a e in a-indus y.
Figu e 11. Ci cula diag am o he Knowledge–Educa ion clus e .
Figu e 12. Ne wo k diag am o he Knowledge–Educa ion clus e .
On he o he hand, M73 Ad e ising (9.8%) and M72 Resea ch and De elopmen (27.1%)
ha e ela i ely high olumes o inpu s. This clus e pe o ms pa icula ly poo ly in
Figu e 12. Ne wo k diag am o he Knowledge–Educa ion clus e .
Many o hese indus ies ely on elecommunica ions se ices p o ided by J61 Telecom-
munica ions as well as on he p oduc ion and se ices o C18 P in ing and J58 Publishing o
c ea e and manage mass communica ion and en e ainmen . These indus ies ep esen he
majo i y o he so-called “c ea i e indus ies”, which employ he membe s o he “c ea i e
class” (Flo ida,2012, pp. 35–62). Ul ima ely, he name Knowledge–Educa ion clus e was
Economies 2025,13, 15 21 o 31
p e e ed because o he economic s uc u e o he clus e and he signi icance o knowledge
p oduc ion and dissemina ion, which is p ima ily a esul o educa ional p ocesses.
One dis inguishing ea u e o he Knowledge–Educa ion clus e , as shown in Table 6,
is i s ela i ely low cohesion (0.39) compa ed o he o he ou clus e s in he p oduc ion
sys em. I also makes he smalles con ibu ion o demand (10.5%) and is he second smalles
in e ms o employmen (12.3%) and g oss alue added (13.4%). The signi icance o P85
Educa ion is e iden in i s impac on hese indica o s, as i accoun s o app oxima ely hal
o mo e o he clus e ’s o al alue in each case. Aside om Educa ion, no o he indus y
makes a no able con ibu ion o employmen , wi h Educa ion i sel ep esen ing 7.5%.
Table 6. Knowledge–Educa ion clus e : main economic cha ac e is ics.
NACE
Code Desc ip ion
Technology
T ansac ions
Inside Clus e Empl.
(%)
Demand
(%)
Added Value Expo s
Ou pu
Inpu (%) Technical
Coe icien (%)
Ex o e sion
C18 P in ing LT 0.26 0.04 0.6 0.0 0.2 44.6 0.0 0.2
J58 Publishing O_KIS 0.43 0.43 0.4 1.2 1.3 62.0 0.4 4.1
J59_J60 Cinema–Tele ision H _KIS 0.77 0.59 0.4 0.5 0.3 30.5 0.3 5.1
J61 Telecommunica ions H _KIS 0.37 0.58 0.7 1.9 2.7 63.6 0.7 3.5
J62_J63 Compu e –In o ma ion se ices H _KIS 0.32 0.49 0.5 0.4 0.6 64.0 0.8 17.3
M72 Resea ch and de elopmen H _KIS 0.83 0.50 0.2 0.2 0.1 27.1 0.2 12.5
M73 Ad e ising KI_m_S 0.34 0.49 0.4 0.1 0.2 9.8 0.3 4.5
M74_M75
O he scien i ic ac i i ies KI_m_S 0.11 0.52 0.5 0.2 0.5 47.5 0.3 7.0
N78 Employmen ac i i ies KI_m_S 0.65 0.30 0.1 0.0 0.1 92.8 0.0 0.0
P85 Educa ion O_KIS 0.27 0.48 7.5 4.8 5.7 94.5 0.1 0.2
R90-R92 C ea i e ac i i ies—Gambling O_KIS 0.64 0.94 0.7 1.2 1.6 79.8 0.1 0.7
R93 Spo s–Rec ea ion O_KIS 0.62 0.49 0.4 0.1 0.1 34.9 0.0 0.2
To al Knowledge–Educa ion Clus e 0.39 12.3 10.5 13.4 66.0 3.2 3.3
This is a clus e ha ope a es wi h a e y high echnical coe icien o added alue
(66%), he highes among he i e clus e s and abo e he o e all p oduc ion sys em a e age
(55.6%). Acco ding o he de ailed indus y-speci ic da a in Table 6, his high coe icien
a ises om he ex emely high alues in ce ain indus ies. Mos indus ies in his clus e
use minimal in e media e inpu s, and in some cases, such as J61 Telecommunica ions,J58 Pub-
lishing,and R90–R92 C ea i e Ac i i ies–Gambling, a signi ican po ion o hei ansac ions
a e in a-indus y.
On he o he hand, M73 Ad e ising (9.8%) and M72 Resea ch and De elopmen (27.1%)
ha e ela i ely high olumes o inpu s. This clus e pe o ms pa icula ly poo ly in expo s
(3.2%), wi h he highes expo -con ibu ing indus y, J62_J63 Compu e –In o ma ion Se ices,
con ibu ing only 0.8% o he o al expo s o he p oduc ion sys em. Simila ly, his clus e
has a low ex e nal o ien a ion (Ex o e sion) index (3.3%), wi h nea ly all i s indus ies,
excep o M72 Resea ch and De elopmen (12.5%) and J62_J63 Compu e –In o ma ion Se ices
(17.3%), ha ing signi ican ly lowe alues han he p oduc ion sys em a e age (11.1%). This
indica es ha while he clus e is composed o knowledge-in ensi e se ices, i lacks an
expo o ien a ion. The excep ions a e M72 Resea ch and De elopmen and J62_J63 Compu e –
In o ma ion Se ices, which show po en ial o expo s.
4.6. Mega-Clus e
The Mega-clus e , as sugges ed by i s name, is he la ges clus e in he p oduc ion
sys em in e ms o almos all aspec s excep o expo s (Table 7). I consis s o 25 indus ies
in o al: 10 om he seconda y sec o (8 o which a e manu ac u ing) and 15 om he
se ice sec o . I includes some o he la ges indus ies by employmen , such as G47 Re ail
T ade (12.3%),O84 Public Adminis a ion (8.5%), and Q86 Heal h (4.7%), as well as some o
he s onges in e ms o demand and added alue, like L68 Real Es a e (10.3% and 15.1%,
espec i ely), O84 Public Adminis a ion–De ense (9.9% and 8.7%), and G46 Wholesale T ade
(6.3% and 6.0%).
Economies 2025,13, 15 22 o 31
Table 7. Mega-clus e : main economic cha ac e is ics.
NACE
Code Desc ip ion
Technology
T ansac ions Inside
Clus e Empl.
(%)
Demand
(%)
Added Value Expo s
Ou pu
Inpu (%) Technical
Coe icien (%)
Ex o e sion
C13–C15 Tex iles–Appa el LT 0.83 0.85 1.1 0.8 0.5 42.9 2.9 45.5
C17 Pape LT 0.57 0.85 0.2 0.3 0.2 29.5 0.3 11.1
C20 Chemicals MHT 0.52 0.70 0.3 0.5 0.3 27.6 2.2 45.5
C21 Pha maceu icals HT 0.97 0.78 0.4 0.5 0.5 58.6 1.9 45.7
C22 Plas ic p oduc s MLT 0.50 0.79 0.3 0.2 0.1 18.1 1.0 24.9
C26 Compu e s–Elec onics HT 0.37 0.76 0.1 0.1 0.1 58.4 0 0
C29 Mo o ehicles MHT 0.87 0.69 0.1 0.1 0.1 48.0 0.1 15.7
C30 O he anspo equipmen MHT 0.18 0.64 0.2 0.1 0.1 65.0 0.1 14.0
C31_C32 Fu ni u e—o he manu ac u ing LT 0.92 0.52 1.0 0.5 0.3 33.6 0.5 11.7
E37–E39 Was e managemen TR 0.58 0.40 0.5 0.4 0.7 58.1 0.6 9.4
G45 T ade and epai o mo o ehicles LKI_m_S 0.32 0.77 2.0 1.9 2.3 66.4 1.2 6.6
G46 Wholesale ade LKI_m_S 0.38 0.69 3.7 6.3 6.0 46.3 6.2 9.5
G47 Re ail ade LKI_m_S 0.40 0.68 12.3 3.3 3.7 55.2 3.2 9.4
H53 Pos al ac i i ies O_LKIS 0.65 0.83 0.5 0.0 0.3 37.4 0.0 1.1
K64 Financial se ices KI_ _S 0.61 0.64 1.8 1.2 3.6 69.3 0.8 3.1
K65 Insu ance KI_ _S 0.39 0.90 0.4 0.5 0.4 44.6 0.8 18.2
K66 O he inancial se ices KI_ _S 0.62 0.74 0.4 0.0 0.6 76.4 0 0
L68 Real es a e LKI_m_S 0.69 0.61 0.1 10.3 15.1 93.2 0 0
M69_M70 Legal, accoun ing, managemen
ac i i ies KI_m_S 0.57 0.62 2.4 0.4 2.1 62.2 0.8 4.5
N79 T a el agencies LKI_m_S 0.58 0.59 0.3 0.4 0.3 31.6 0 0
N80–N82 Secu i y, se ices o buildings LKI_m_S 0.55 0.53 1.2 0.2 1.7 53.0 0.2 1.2
O84 Public adminis a ion, de ense, social
secu i y O_KIS 1.00 0.59 8.5 9.9 8.7 71.8 0 0
Q86 Heal h O_KIS 0.86 0.87 4.7 5.5 4.6 67.6 0.1 0.3
Q87_Q88 Social ca e O_KIS 0.81 0.48 0.9 0.4 0.3 55.8 0 0
S95 Repai o compu e s and household
goods LKI_m_S 0.35 0.77 0.3 0.3 0.4 81.6 0.1 4.8
To al Mega-Clus e 0.53 43.6 44.2 53.2 65.1 23.0 5.6
In e ms o expo s, he Mega-clus e anks second (23.0%), bu his is la gely due o
he shee numbe o i s indus ies a he han he s eng h o any indi idual expo indus y.
I is a cohesi e clus e , he hi d highes in e ms o cohesion among he i e clus e s, wi h
53% o i s ansac ion olume occu ing in e nally. This cohesion pe sis s despi e he clus e
ha ing wice o e en mo e han iple he numbe o indus ies (25) compa ed o he o he
clus e s (8 o 12). Almos all o i s indus ies ha e a ansac ion index o bo h inpu s and
ou pu s ha exceeds 0.5.
The ela i ely high deg ee o cohesion makes u he di ision a emp s un eliable.
When he clus e was subjec ed o he same di ision p ocess as he o e all ne wo k (a
p oduc ion sys em composed o 62 indus ies), using he same me hod (modula i y op-
imiza ion algo i hms including he Lou ain and A enas e al. based on he modi ied
G-N me hod), he esul s we e inconsis en . The Lou ain me hod p oduced h ee new
sub-clus e s: one wi h a single isola ed indus y (C13-C15 Tex iles–Appa el), a second wi h
six indus ies ha could be e e ed o a “heal hca e-pha maceu ical sub-clus e ” (including
C20 Chemicals,C21 Pha maceu icals,C22 Plas ics,C31_C32 Fu ni u e, and heal hca e and
social wel a e se ices Q86 Heal h,Q87_Q88 Social Ca e), and a hi d sub-clus e con aining
he emaining 18 indus ies. The A enas e al. me hod esul ed in i e sub-clus e s, wi h
only he isola ed C13-C15 Tex iles–Clo hing sub-clus e ma ching he Lou ain esul s. The
emaining ou sub-clus e s a ied, wi h wo con aining ou indus ies each and wo
con aining nine indus ies each.
The signi ican a ia ion in he esul s be ween he wo me hods indica es ha u he
di ision o he Mega-clus e is pa icula ly un eliable. The ou come om he Lou- ain
me hod seems o make mo e economic sense, as i iden i ies a g ouping cen e ed a ound
heal hca e and social wel a e se ices, along wi h he pha maceu ical indus y, d aw-
ing in ela ed indus ies like chemicals and plas ics. Addi ionally, his di ision has a
highe modula i y index (Q= 0.2681) compa ed o he second me hod (Q= 0.1575). I
should be no ed, howe e , ha he modula i y alues in bo h cases a e signi ican ly lowe
han hose o he main di ision, which we e Q= 0.3005 o he Lou ain me hod and
Q= 0.2923 o he A enas e al. me hod. O e all, hese esul s a e insu icien o con iden ly
Economies 2025,13, 15 23 o 31
claim ha he p oduc ion sys em consis s o se en o , e en mo e unlikely, nine clus e s.
Ne e heless, hese indings can guide a be e unde s anding o how he Mega-clus e
unc ions, especially when combined wi h diag amma ic ep esen a ions.
The diag ams e eal he densi y o he in e - and in a-indus y ela ionships wi hin
he Mega-clus e , con ibu ing o i s ela i ely high cohesion index. In he ci cula diag am
(Figu e 13), he quan i a i e p ominence o he T ade indus ies (G45,G46,G47) is e iden ,
ma ked wi h b own-shaded a cs and connec ions, bo h o wa d and backwa d. The second
mos signi ican g oup is he Financial sec o (K64 Financial Se ices,K65 Insu ance, and K66
O he Financial Se ices), depic ed in g ay shades. Business se ice p o ide s (M69_M70 and
N80–N82) and N79 T a el Agencies, ma ked in shades o blue, o m ano he majo g oup
wi h s ong ies o bo h he T ade and Financial sec o s.
Economies 2025, 13, x 24 o 32
indus y ones, explaining hei isola ion as indi idual clus e s in he ini ial p oduc ion
sys em di ision and he Mega-clus e sub-di ision a emp s.
The ne wo k diag am (Figu e 14) o he Mega-clus e emphasizes he opological sig-
ni icance o indus ies a he han he olume o in e -indus y ela ions highligh ed in
he p e ious ci cula cho d diag am. In his isualiza ion, as wi h he co esponding dia-
g ams o he o he clus e s, he size and placemen o he indus ies (nodes/ e ices) in-
dica e hei impo ance o “in luence” wi hin he ne wo k (clus e ). This in luence is sum-
ma ized by he eigen ec o cen ali y (PageRank), which conside s bo h he numbe and
s eng h o he linkages as well as he signi icance o he connec ed indus ies.
Figu e 13. Ci cula diag am o he Mega-clus e .
Figu e 13. Ci cula diag am o he Mega-clus e .
The L68 Real Es a e indus y, shown in da k g een, also has no able links wi h T ade
and Finance. Public sec o indus ies (O84 Public Adminis a ion,Q85 Heal h, and Q86_Q87
Social Ca e), shaded in pu ple, mainly se e as ecipien s o o wa d linkages om o he
indus ies wi hin he clus e . The linkage be ween Q86 Heal h and C21 Pha maceu icals
Economies 2025,13, 15 24 o 31
and C31_C32 Fu ni u e is no able, which jus i ies hei appea ance as a sepa a e “heal hca e-
pha maceu ical sub-clus e ” in he Lou ain esul s. The Manu ac u ing indus ies, ma ked in
ed, ha e a ela i ely small size bu main ain a signi ican numbe o in e -indus y linkages
wi hin he clus e . Las ly, he diag ams show ha H53 Pos al Ac i i ies and C13-C15 Tex iles–
Appa el ha e a highe olume o in a-indus y linkages compa ed o in e -indus y ones,
explaining hei isola ion as indi idual clus e s in he ini ial p oduc ion sys em di ision
and he Mega-clus e sub-di ision a emp s.
The ne wo k diag am (Figu e 14) o he Mega-clus e emphasizes he opological
signi icance o indus ies a he han he olume o in e -indus y ela ions highligh ed
in he p e ious ci cula cho d diag am. In his isualiza ion, as wi h he co esponding
diag ams o he o he clus e s, he size and placemen o he indus ies (nodes/ e ices)
indica e hei impo ance o “in luence” wi hin he ne wo k (clus e ). This in luence is
summa ized by he eigen ec o cen ali y (PageRank), which conside s bo h he numbe
and s eng h o he linkages as well as he signi icance o he connec ed indus ies.
Economies 2025, 13, x 25 o 32
Figu e 14. Ne wo k diag am o he Mega-clus e .
The ne wo k diag am o he Mega-clus e e eals h ee g oups o indus ies:
1. Public Sec o Indus ies: Including O84 Public Adminis a ion and De ense and Q86
Heal h.
2. T ade Indus ies: Comp ising G46 Wholesale T ade, G47 Re ail T ade, and G45 Au omo i e
T ade and Repai s.
3. Financial Sec o Indus ies: Consis ing o K64 Financial Se ices and K65 Insu ance.
The L68 Real Es a e indus y is less p ominen in his ne wo k iew compa ed o he
p e ious ci cula diag am, while he M69_M70 Legal–Accoun ing Se ices and N80–N82
P o ec ion and O he Se ices indus ies, despi e being posi ioned pe iphe ally, ha e some
in luence on he clus e ’s dynamics. No ably, C21 Pha maceu icals and C31_C32 Fu ni u e,
loca ed nea Q86 Heal h a he op o he diag am, and he nea by Q86_Q87 Social Ca e,
C20 Chemicals, and C22 Plas ics o m he “heal hca e-pha maceu ical sub-clus e ” p e iously
iden i ied.
The Mega-clus e exhibi s a balance be ween high- echnology and knowledge-in en-
si e indus ies on he one hand and low- echnology and less-knowledge-in ensi e indus-
ies on he o he . Speci ically, his clus e includes i e high- and medium–high- echnol-
ogy indus ies and i e low- and low–medium- echnology indus ies (including E37–E39
Was e Managemen ). In he se ice sec o , he e a e eigh knowledge-in ensi e indus ies
and se en less-knowledge-in ensi e ones, o aling 13 high- echnology and knowledge-
in ensi e indus ies e sus 12 low- echnology and less-knowledge-in ensi e indus ies.
Despi e his nume ical balance, he ou comes o he clus e a o low- echnology and
less-knowledge-in ensi e p oduc ion. High- echnology (HT and MHT) and knowledge-
in ensi e indus ies (KI_ _S and O_KIS) con ibu e 21.5% o he o al g oss alue added
(GVA) o he p oduc ion sys em, compa ed o 31.7% o low- echnology (LT, MLT and
TR) and less-knowledge-in ensi e indus ies (LKI_m_S and O_LKIS). Simila ly, in e ms
o employmen , high- echnology (HT and MHT) and knowledge-in ensi e indus ies
(KI_ _S and O_KIS) accoun o 20%, while less-in ensi e ones accoun o 23.6%. Fo he
o e all demand, he igu es a e 19.2% e sus 25%, and o expo s, hey a e 6.8% e sus
Figu e 14. Ne wo k diag am o he Mega-clus e .
The ne wo k diag am o he Mega-clus e e eals h ee g oups o indus ies:
1.
Public Sec o Indus ies: Including O84 Public Adminis a ion and De ense and Q86 Heal h.
2.
T ade Indus ies: Comp ising G46 Wholesale T ade,G47 Re ail T ade,and G45 Au omo i e
T ade and Repai s.
3. Financial Sec o Indus ies: Consis ing o K64 Financial Se ices and K65 Insu ance.
The L68 Real Es a e indus y is less p ominen in his ne wo k iew compa ed o
he p e ious ci cula diag am, while he M69_M70 Legal–Accoun ing Se ices and N80–
N82 P o ec ion and O he Se ices indus ies, despi e being posi ioned pe iphe ally, ha e
some in luence on he clus e ’s dynamics. No ably, C21 Pha maceu icals and C31_C32
Fu ni u e, loca ed nea Q86 Heal h a he op o he diag am, and he nea by Q86_Q87
Social Ca e,C20 Chemicals, and C22 Plas ics o m he “heal hca e-pha maceu ical sub-clus e ”
p e iously iden i ied.
Economies 2025,13, 15 31 o 31
Newman, M. E. J., & Gi an, M. (2004). Finding and e alua ing communi y s uc u e in ne wo ks. Physical Re iew E—S a is ical,
Nonlinea , and So Ma e Physics,69(2 P 2), 026113. [C ossRe ]
Pe oux, F. (1950). Economic space: Theo y and applica ions. The Qua e ly Jou nal o Economics,64, 89–104. [C ossRe ]
Pe oux, F. (1970). No e on he concep o ‘g ow h poles’. In D. L. MacKee, R. D. Dean, & W. H. Leahy (Eds.), Regional economics
(pp. 93–103). The F ee P ess. (O iginal wo k published 1955).
Po e , M. E. (1998). The compe i i e ad an age o he na ions (New Edi ion wi h upda es in an In oduc ion by he Au ho ). Macmillan.
Rasmussen, P. N. (1957). S udies in in e -sec o al ela ions (E. Ha ck, Ed.; 2nd p in ing). Re ue Économique.
Sco , J. (2000). Social ne wo k analysis, a handbook (2nd ed.). Sage Publica ions.
Sonis, M., & Hewings, G. J. D. (2009). New De elopmen s in Inpu –ou pu analysis: Fields o in luence o changes, he empo al
leon ie in e se and he econside a ion o classical key sec o analysis. In M. Sonis, & G. J. D. Hewings (Eds.), Tool ki s in egional
science: Theo y, models, and es ima ion (pp. 69–117). Sp inge . Chap e 3.
S o pe , M. (1997). The egional wo ld: Te i o ial de elopmen in a global economy. The Guil o d P ess.
T aag, V. A., Wal man, L., & an Eck, N. J. (2019). F om Lou ain o Leiden: Gua an eeing well-connec ed communi ies. Scien i ic
Repo s,9, 5233. [C ossRe ]
Tseke is, T. (2017). Ne wo k analysis o in e -sec o al ela ionships and key sec o s in he G eek economy. Jou nal o Economic In e ac ion
and Coo dina ion,12, 413–435. [C ossRe ]
Wasse man, S., & Faus , K. (1994). Social ne wo k analysis. Camb idge Uni e si y P ess.
Disclaime /Publishe ’s No e: The s a emen s, opinions and da a con ained in all publica ions a e solely hose o he indi idual
au ho (s) and con ibu o (s) and no o MDPI and/o he edi o (s). MDPI and/o he edi o (s) disclaim esponsibili y o any inju y o
people o p ope y esul ing om any ideas, me hods, ins uc ions o p oduc s e e ed o in he con en .
