Global Natural Gas Market Integration: The Role of LNG Trade and Infrastructure Constraints

Fa ag, Ma kos; Jeddi, Sami ; Kopp, Jan Hend ik
A icle — Published Ve sion
Global Na u al Gas Ma ke In eg a ion: The Role o LNG
T ade and In as uc u e Cons ain s
The Wo ld Economy
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Fa ag, Ma kos; Jeddi, Sami ; Kopp, Jan Hend ik (2025) : Global Na u al Gas
Ma ke In eg a ion: The Role o LNG T ade and In as uc u e Cons ain s, The Wo ld Economy,
ISSN 1467-9701, Wiley, Hoboken, NJ, Vol. 48, Iss. 6, pp. 1405-1417,
h ps://doi.o g/10.1111/ wec.13699
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/323789
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
The Wo ld Economy, 2025; 48:1405–1417
h ps://doi.o g/10.1111/ wec.13699
1405
The Wo ld Economy
ORIGINAL ARTICLE OPEN ACCESS
Global Na u al Gas Ma ke In eg a ion: The Role o LNG
T ade and In as uc u e Cons ain s
Ma kosFa ag1 | Sami Jeddi2 | JanHend ikKopp2
1Facul y o Managemen , Economics and Social Sciences, Uni e si y o Cologne, Cologne, Ge many | 2Ins i u e o Ene gy Economics, Uni e si y o
Cologne, Cologne,Ge many
Co espondence: Ma kos Fa ag (ma kos. a [email protected])
Recei ed: 17 Sep embe 2024 | Accep ed: 9 Janua y 2025
Funding: The au ho s ecei ed no speci ic unding o his wo k.
Keywo ds: a bi age| LNG ade| ma ke in eg a ion| na u al gas p ice| supply dis up ions
ABSTRACT
This pape analyses he in eg a ion o global na u al gas ma ke s ac oss No h Ame ica, Eu ope, and Asia om 2016 o 2022.
The analysis ocuses on he impac o he Uni ed S a es eme ging as a majo lique ied na u al gas (LNG) expo e and signi ican
supply dis up ions, including he sha p educ ion in Russian pipeline supplies o Eu ope. We iden i y a s uc u al b eak on 1
Oc obe 2021, coinciding wi h hese supply dis up ions and a igh ening global LNG ma ke . Using bo h linea and nonlinea
coin eg a ion echniques, we assess p ice con e gence ac oss he h ee egions in wo subsamples: be o e and a e he b eak. In
he i s subsample, we ind s ong in eg a ion be ween all h ee egional gas ma ke s, d i en by g owing LNG ade and sha ed
exposu e o global spo ma ke dynamics. Howe e , in he second subsample, he deg ee o in eg a ion be ween he Asian and
Eu opean ma ke s weakens, wi h US p ices decoupling om bo h. G ange causali y analysis e eals ha LNG in as uc u e
conges ion, pa icula ly in he US and No hwes Eu ope, signi ican ly d i es he widening p ice sp eads be ween he US and
Eu opean ma ke s. These indings sugges ha physical in as uc u e plays a cen al ole in ene gy ma ke in eg a ion, espe-
cially du ing pe iods o igh ma ke condi ions, whe e in as uc u e bo lenecks limi a bi age oppo uni ies.
JEL Classi ica ion: Q37, Q41, F14, C32, L95
1 | In oduc ion
In e na ional ade in na u al gas has adi ionally been di ided
in o h ee main egional ma ke s: Asia, Eu ope, and No h
Ame ica (Melamid 1994; Economides and Wood 2009; Geng
e al.2014). His o ically, his segmen a ion has been d i en by
limi ed lique ied na u al gas (LNG) anspo capaci y. Howe e ,
he li e a u e sugges s ha hese ma ke s a e g adually becom-
ing mo e in eg a ed (Neumann2009; Li e al.2014). Ma ke in-
eg a ion e e s o he ex en o which egional ma ke s sha e
in o ma ion and align p ices (McNew and Fackle 1997; Fackle
and Goodwin2001). In es iga ing his phenomenon has signi -
ican implica ions o supply secu i y, as ma ke pa icipan s in
one egion mus inc easingly conside condi ions in o he e-
gions o ensu e hei own supply.
The in eg a ion p ocess among he h ee egional gas ma -
ke s has been d i en by se e al key ac o s. Fi s , some e-
gions ha e expe ienced su plus na u al gas p oduc ion, while
o he s ha e seen inc easing consump ion.1 This imbalance
has necessi a ed expanding he in e na ional gas ade, wi h
LNG eme ging as a c i ical solu ion. Inc easing expo capac-
i ies and he g ow h o he LNG lee ha e signi ican ly im-
p o ed he echnical and economic easibili y o in e - egional
ade (Ba nes and Boswo h 2015; Li e  al. 2014). Second,
many comme cial ag eemen s ha e shi ed om adi ional
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium, p o ided he o iginal wo k is
p ope ly ci ed.
© 2025 The Au ho (s). The Wo ld Economy published by John Wiley & Sons L d.
1406 The Wo ld Economy, 2025
oil- indexed p icing in long- e m con ac s o g ea e eliance
on hub- based p icing. Fo example, he sha e o Gas- on- Gas
(G- o- G) compe i ion2 in global gas consump ion ose om
31% in 2005 o 49% in 2021, while oil indexa ion declined om
24% o 19% o e he same pe iod (IGU2021). The li e a u e
also sugges s ha he ela ionship be ween oil and na u al gas
p ices has become mo e ola ile, indica ing a decoupling o
he wo commodi ies (Chiappini e al.2019; Neumann2009).
A he same ime, he G- o- G compe i ion has seen a ise in
spo and sho - e m ansac ions, whe e shi s in egional sup-
ply and demand p omp LNG expo e s o edi ec spo ol-
umes (IGU2021). These de elopmen s ha e inc eased ma ke
liquidi y, enhanced oppo uni ies o spa ial a bi age, and
boos ed he p esence o physical ade s. Thi d, ad ancemen s
in shale gas explo a ion echnology ha e uelled a apid in-
c ease in p oduc ion in No h Ame ica, commonly e e ed
o as he shale gas e olu ion. As a esul , he Uni ed S a es
began expo ing LNG in 2016 and has quickly become a majo
playe in he global ma ke , wi h expo capaci ies expanding
yea o e yea (Melikoglu2014; Wiggins and E ienne2017).3
Finally, Eu opean and Asian coun ies ha e adop ed supply
di e si ica ion s a egies ha combine pipelines and LNG im-
po s o mi iga e supply isks (Fa ag and Zaki2024; Ri z2019;
Hinchey2018).
Se e al s udies ha e ocused on global gas ma ke in eg a ion,
p ima ily elying on p ice da a o measu e he deg ee o in e-
g a ion. The hypo hesis is ha g ea e con e gence be ween
gas p ices signi ies s onge spa ial a bi age and highe le els
o ma ke in eg a ion. The mos commonly used me hodolog-
ical app oach o es his hypo hesis is he coin eg a ion ech-
nique,4 which examines he exis ence o a long- un ela ionship
be ween p ices.5
Sili e s o s e  al. (2005) in es iga ed he in eg a ion o he
No h Ame ican, Eu opean, and Asian gas ma ke s using
mon hly p ices om No embe 1993 o Ma ch 2004. Thei
coin eg a ion analysis p o ided e idence o in eg a ion be-
ween he Asian and Eu opean ma ke s, while he No h
Ame ican ma ke emained decoupled. The au ho s explain
ha he Eu opean and Asian na u al gas ma ke s a e in e-
g a ed due o simila long- e m con ac s and oil- indexed
p icing mechanisms, which align p ice mo emen s in hese e-
gions. In con as , he No h Ame ican ma ke ope a es unde
a di e en , mo e compe i i e p icing sys em ha decouples i
om he oil- linked Eu opean and Asian ma ke s, esul ing in
a lack o in eg a ion ac oss he A lan ic. A simila conclusion
was eached by Li e al.(2014), who examined he in eg a ion
o in e na ional na u al gas ma ke s ac oss No h Ame ica,
Eu ope, (a) and Asia om 1997 o 2011, using a con e gence
es and Kalman il e analysis.6 In con as , Neumann(2009)
ound e idence o inc easing in eg a ion be ween No h
Ame ican and Eu opean gas ma ke s. Using he Kalman il e
o analyse da a om 1999 o 2008, Neumann obse ed ising
p ice con e gence, pa icula ly a e 2003. This end was a -
ibu ed o he ole o LNG in linking p e iously segmen ed
ma ke s ac oss he A lan ic du ing his pe iod.
Howe e , Nick and Tischle  (2014) poin ed ou ha linea
coin eg a ion models, which assume symme ic adjus men s,
may be misspeci ied o na u al gas ma ke s whe e adjus men s
o p ice de ia ions can be asymme ic. Fac o s such as ans-
ac ion cos s and di e en esponses o widening o na owing
sp eads con ibu e o his asymme y, making nonlinea coin e-
g a ion a mo e app op ia e app oach. To add ess his, hey ex-
amined he deg ee o in eg a ion be ween No h Ame ican and
Eu opean gas p ices using a nonlinea coin eg a ion app oach
ha accoun s o ansac ion cos s. Thei esul s p o ided s ong
e idence o nonlinea i y in he sub- samples analysed (2000–
2008 and 2009–2012). Mo e ecen ly, Chiappini e al.(2019) ap-
plied he momen um- h eshold au o eg ession (M- TAR) model
o Ende s and Siklos(2001) wi h daily p ice da a om 2004 o
2018, con i ming he p esence o nonlinea i ies and asymme-
ies in p ice adjus men s in he global gas ma ke . Thei anal-
ysis also shows ha he deg ee o in e dependence be ween he
No h Ame ican and Eu opean ma ke s has inc eased, whe eas
his has no occu ed be ween he No h Ame ican and Asian
ma ke s.
The e iewed li e a u e shows ha conclusions on egional
gas ma ke in eg a ion depend on he me hods used and he
key ma ke mechanisms a play du ing he analysed pe iod.
Rega ding ma ke mechanisms, LNG ade o e s mo e oppo -
uni ies o spa ial a bi age, con ibu ing o inc eased p ice
con e gence among he No h Ame ican, Eu opean, and Asian
ma ke s. Howe e , i emains unclea how ecen de elopmen s
in he global gas ma ke —especially he eme gence o he
Uni ed S a es as a majo LNG expo e since 2016 and he sup-
ply dis up ions caused by geopoli ical ensions be ween Eu ope
and Russia, amid a igh LNG ma ke —ha e impac ed ma ke
in eg a ion.
This pape con ibu es o he li e a u e by analysing he in-
eg a ion o he global gas ma ke om 2016 o 2022, using
daily u u es p ices ac oss he h ee main egional gas ma -
ke s. The No h Ame ican ma ke is ep esen ed by he Hen y
Hub (HH) benchma k, he No hwes Eu opean ma ke by he
Ti le T ans e Facili y (TTF) benchma k, and he Eas Asian
ma ke by he Eas Asian Index (EAX). This analysis is pa ic-
ula ly ele an o wo main easons. Fi s , his pe iod coin-
cides wi h he en y o he Uni ed S a es in o he global LNG
ade, a de elopmen ha may ha e eshaped ela ionships
wi hin he global gas ma ke . P e ious esea ch on ma ke in-
eg a ion la gely ocused on pe iods when he Uni ed S a es
was a ne impo e o gas. The e o e, his s udy p o ides new
insigh s in o in e dependencies and coin eg a ion unde di -
e en ma ke condi ions. Second, his pe iod has seen se e al
ac o s ha suppo a bi age in he global gas ma ke , pa ic-
ula ly be ween he Uni ed S a es and he o he wo egions,
d i en by he expansion o US LNG expo in as uc u e and
he ise in spo LNG ade7. Howe e , i has also wi nessed
ac o s ha hinde a bi age, such as US LNG expo in a-
s uc u e and Eu opean impo in as uc u e ope a ing
a maximum capaci y. In his con ex , his s udy p o ides a
o mal s a is ical analysis o he p ice di e en ials be ween
egional gas ma ke s and iden i ies a s uc u al b eak in
hese di e en ials on 1 Oc obe 2021. This iming aligns wi h
signi ican ma ke dis up ions, such as Russia educing gas
lows o Eu ope and a igh ening global LNG ma ke due o
supply ou ages and capaci y cons ain s (Fulwood e al.2022;
McWilliams e  al. 2023). To cap u e he po en ial e ec s o
hese dynamics, we conduc he coin eg a ion analysis o e
1407
wo subsamples, spli ing he da a on 1 Oc obe 2021. This
iming aligns wi h signi ican ma ke dis up ions, including
Russia's educ ion o gas lows o Eu ope and he igh ening
global LNG ma ke due o supply ou ages and capaci y con-
s ain s (Fulwood e al.2022; McWilliams e al.2023).
Ou esul s show ha du ing he i s subsample (Janua y
2016 o Sep embe 2021) he Asian and Eu opean gas p ices
a e coin eg a ed. This inding is consis en wi h p e ious
s udies, such as Chiappini e al.(2019), which also iden i ied
coin eg a ion be ween Eu opean and Asian gas ma ke s in
ea lie pe iods. The pe sis ence o his in eg a ion du ing ou
sample pe iod can be a ibu ed o he g ow h o LNG ade,
which has acili a ed a bi age oppo uni ies be ween Eu ope
and Asia. Bo h egions a e subjec o global supply–demand
dynamics and spo ma ke p icing mechanisms, ein o c-
ing hei in eg a ion. While Chiappini e al.(2019) ound no
coin eg a ion be ween Ame ican and Asian p ices bu did ind
coin eg a ion be ween Ame ican and Eu opean p ices, ou
analysis e eals ha Ame ican p ices we e coin eg a ed wi h
bo h Eu opean and Asian p ices du ing his pe iod. These di -
e ences likely e lec changes in he global gas ma ke , pa -
icula ly he ansi ion o he Uni ed S a es o a ne expo e ,
which has eshaped i s ela ionship wi h he Eu opean and
Asian ma ke s. Howe e , ou indings a e consis en wi h
hose o Nick and Tischle (2014) and Chiappini e al.(2019),
suppo ing he conclusion ha egional gas p ices a e non-
linea ly coin eg a ed. This implies ha adjus men s owa ds
equilib ium happen a di e en speeds based on he di ec ion
o de ia ion om he equilib ium.
In he second subsample (Oc obe 2021—No embe 2022),
we ind no e idence o linea coin eg a ion o any p ice pai s
based on he Engle–G ange app oach. Howe e , when we
apply he Ende s–Siklos h eshold coin eg a ion me hod, we
ind e idence o h eshold coin eg a ion only o he EAX–
TTF p ice pai . This sugges s he p esence o a nonlinea ,
asymme ic ela ionship be ween hese wo ma ke s du ing
he second subsample pe iod, while he o he pai s (HH–
TTF and HH–EAX) do no exhibi such a ela ionship. The
lack o coin eg a ion be ween he Ame ican and Eu opean
ma ke s may be a ibu ed o LNG in as uc u e conges ion
du ing his pe iod, which ac s as a physical ba ie o a bi-
age.8 To u he in es iga e he decoupling o he Ame ican
and Eu opean gas ma ke s obse ed in he second subsam-
ple, we examine he ela ionship be ween he HH–TTF p ice
sp ead and LNG in as uc u e conges ion. Using he Toda
and Yamamo o (1995) app oach, we analyse he p edic i e
ela ionship be ween LNG in as uc u e conges ion and he
HH–TTF p ice sp ead. In he second subsample, we ind sig-
ni ican G ange causali y om conges ion o he HH–TTF
sp ead, sugges ing ha in as uc u e cons ain s a e in lu-
encing p ice di e en ials. These indings unde sco e he c i -
ical ole o in as uc u e capaci y in acili a ing o impeding
ma ke in eg a ion be ween egional gas ma ke s.
The emainde o he pape is o ganised as ollows: Sec ion2
ou lines he concep ual backg ound. Following his, Sec ion3
discusses s uc u al changes in he egional gas ma ke , ocus-
ing on egional p ice pa e ns, LNG in as uc u e u ilisa ion,
and LNG ade dynamics in No hwes Eu ope (NWE), Eas
Asia, and he Uni ed S a es. Sec ion3.4 de ails ou me hodol-
ogy. Sec ion5 p esen s he baseline esul s o ou analysis, while
Sec ion6 p o ides u he esul s. Finally, Sec ion7 p o ides a
discussion and conclusion.
2 | Concep ual Backg ound
The concep o ma ke in eg a ion can be aced back o
Cou no , who s a ed ha i is ‘an en i e e i o y, o which he
pa s a e so uni ed by he ela ions o un es ic ed comme ce,
ha p ices ake he same le el h oughou wi h ease and apid-
i y’ (Cou no  1838). Empi ical s udies ha e examined ma ke
in eg a ion along e ical (p ices a di e en s ages o he supply
chain), ho izon al (p ices ac oss loca ions) and in e - empo al
(spo and u u e ma ke p ices) dimensions, o en employing
coin eg a ion me hods (Ihle and on C amon- Taubadel 2008;
Roman and Žáko á K oupo á2022). This s udy ocuses on he
ho izon al dimension o ma ke in eg a ion, which is heo e i-
cally mo i a ed by he Enke–Samuelson–Takayama–Judge spa-
ial equilib ium model (Enke1951; Samuelson1952; Takayama
and Judge1971).
In he p esence o ansac ion cos s, he condi ion o a bi age
can be ep esen ed as ollows:
whe e
pA
and
pB
deno e p ices in ma ke s A and B, espec i ely,
𝜏B,A
ep esen s he ansac ion cos o expo ing na u al gas om
ma ke B o ma ke A. The e o e, a bi age ac i i y may only be
igge ed i he implied g oss p o i o he ade co e s ansac-
ion cos s.
Howe e , his spa ial equilib ium model does no accoun o
he in as uc u e cons ain s on a bi age be ween ma ke s.
I he impo in as uc u e in ma ke
A
and he expo in a-
s uc u e in ma ke
B
a e ully u ilised, he p ice di e ence
canno be mi iga ed h ough a bi age. The impac o in a-
s uc u e cons ain s on p ice ela ionships undamen ally
di e s om ha o ansac ion cos s. While ansac ion cos s
ep esen angible expenses incu ed du ing ade—such as
anspo a ion and handling ees—in as uc u e cons ain s
ac as physical ba ie s ha limi a bi age, ega dless o he
p ice di e en ial o he associa ed ansac ion cos s (Kupe
and Mulde 2016). This dis inc ion is c ucial, as i highligh s
a bounda y o ma ke in eg a ion: e en i he p ice di e ence
(
p
A
−p
B)
exceeds ansac ion cos s
(
𝜏
B,A)
, no a bi age mecha-
nism can equilib a e he ma ke s i he in as uc u e is ully
u ilised.
The abo e equa ion is app op ia e o unde s anding p ice a -
bi age be ween he Uni ed S a es and Eu ope o he Uni ed
S a es and Asia, whe e di ec LNG ade occu s. The Uni ed
S a es, being a ne expo e , di ec ly supplies LNG o bo h e-
gions. Howe e , be ween Asia and Eu ope, no signi ican di ec
LNG ade can be obse ed du ing he s udy pe iod. Ins ead,
a bi age occu s h ough indi ec ade ia hi d- pa y LNG
ade s who e ou e shipmen s based on ma ke condi ions. To
(1)
pA>pB+𝜏B,A
1408 The Wo ld Economy, 2025
ep esen his, le
pA
and
pE
be he p ices in Asia and Eu ope,
espec i ely,
𝜏S,A
and
𝜏S,E
be he ansac ion cos s (including
anspo a ion) om swing supplie
S
o Asia and Eu ope, e-
spec i ely, and
qS,A
and
qS,E
be he olumes expo ed om swing
supplie
S
o Asia and Eu ope, espec i ely. Assuming ha he e
a e no binding in as uc u e cons ain s a bo h he swing sup-
plie and he impo ing egions, we ha e he ollowing condi-
ion
pA−𝜏S,A>pE−𝜏S,E
. This means ha swing supplie
S
will
expo LNG o Asia i he ne back p ice in Asia
(pA
−𝜏
S,A)
is
g ea e han he ne back p ice in Eu ope (
pE−𝜏S,E
). I he ne -
back is highe in Eu ope, he supplie will p e e o expo he e.
3 | S uc u al Changes in he In e na ional
Na u al Gas Ma ke
This sec ion p o ides a desc ip i e o e iew o key ends in
he global gas ma ke , ocusing on egional p ice luc ua ions,
LNG e minal u ilisa ion, and shi ing ade dynamics. These
elemen s a e c ucial o unde s anding he ac o s in luencing
ma ke in eg a ion. By examining p ice pa e ns ac oss No h
Ame ica, Eu ope, and Eas Asia, along wi h he impac o in a-
s uc u e cons ain s, his sec ion lays he g oundwo k o he
subsequen empi ical analysis.
3.1 | T ends and Fluc ua ions in Regional Na u al
Gas P ices
Figu e1 shows he loga i hmic p ices o he Hen y Hub (HH)
in he No h Ame ican ma ke , he Ti le T ans e Facili y (TTF)
in he Eu opean ma ke , and he Eas Asian Index (EAX) in he
Eas Asian egion.9 The igu e demons a es ha he e was a
subs an ial decline in p ices in Ma ch 2020, likely due o he ou -
b eak o COVID- 19 and i s impac on na u al gas demand. This
was u he exace ba ed by his o ically mild empe a u es.10
Figu e1 also shows ha Eu opean and Asian gas p ices began
o ise in he second hal o 2021. This can be a ibu ed o he
esu gence o demand om he indus ial and hea ing sec o s
as economic ac i i y ebounded and ex eme wea he e en s oc-
cu ed. F om his pe iod onwa ds, i is also e iden ha he HH
se ies was no signi ican ly a ec ed by hese inc eases.11
3.2 | LNG Te minal U ilisa ion and Shi ing Gas
T ade Dynamics
Figu e2 highligh s he majo shi s in he na u al gas ma ke
s a ing in Oc obe 2021 (indica ed by he g ey- shaded a ea),
coinciding wi h educed gas lows om Russia o Eu ope
FIGURE 1 | Na u al gas p ices in log le el. (a) Log p ice o TTF, (b) Log p ice o EAX, and (c) Log p ice o HH. [Colou igu e can be iewed a
wileyonlinelib a y.com]
FIGURE 2 | (a) Δ YoY (%) in Russian gas expo s, (b) Ga e e minal u ilisa ion a es, and (c) U.S. expo e minals u ilisa ion a es. Own cons uc-
ion based on da a ob ained om ENTSOG(2023), GIE(2024), and EIA(2023). [Colou igu e can be iewed a wileyonlinelib a y.com]

1409
(Hende son and Chyong 2023; Fa ag and Ruhnau 2024).
Figu e2a shows a sha p yea - on- yea (YoY) decline in Russian
gas expo s o Eu ope, e lec ing a delibe a e educ ion in daily
lows o he le el o nomina ions om long- e m con ac s, wi h
no addi ional olumes supplied o he Eu opean spo ma ke
(Fulwood e al.2022). This educ ion o ced Eu ope o inc ease
i s eliance on LNG impo s, as seen in Figu e2b, which de-
pic s a no able inc ease in he u ilisa ion a e o he Ga e e mi-
nal in he Ne he lands, he la ges impo e minal o LNG in
No hwes Eu ope. The heigh ened demand o LNG also caused
conges ion a o he Eu opean impo e minals (GIE 2024).
Simul aneously, as shown in Figu e2c, he u ilisa ion a e o
US gas expo e minals inc eased, nea ing ull capaci y and e-
lec ing a high le el o expo s. Howe e , capaci y cons ain s a
bo h Eu opean impo and US expo e minals limi ed he abil-
i y o signi ican ly inc ease LNG ade be ween he wo egions.
3.3 | Di e ging LNG Impo Dynamics in
No hwes Eu ope and Eas Asia
This subsec ion p esen s he yea - on- yea changes in LNG im-
po s om 2016 o 2022, wi h sepa a e g aphs o No h Wes
Eu ope, Japan and Ko ea (combined), and China. Figu e 3a
shows ha LNG impo s o Eu ope inc eased in he las qua -
e o 2021, likely d i en by educed Russian gas supplies, as
discussed in he p e ious sec ion. This educ ion led o ene gy
secu i y conce ns and e o s o di e si y away om adi ional
pipeline sou ces (Ai ken and E soy2023). In con as , Figu e3b
indica es ha he g ow h a e o LNG impo s in China began o
slow, while pipeline impo s om Russia inc eased, indica ing
a po en ial s abilisa ion o shi in he ene gy consump ion mix
(Rys ad Ene gy2023). Meanwhile, Figu e3c shows ha LNG
impo s in Japan and Ko ea emained ela i ely s able, e lec ing
s eady demand in hese ma u e ma ke s (Rys ad Ene gy2023).
These a ying impo pa e ns unde sco e di e ing egional de-
mand dynamics and sugges ha Eu ope, Japan & Ko ea, and
China a e expe iencing unique d i e s and p essu es in hei
LNG ma ke s, likely in luenced by geopoli ical, economic and
policy- ela ed ac o s.
3.4 | US LNG Expo Dynamics o No hwes
Eu ope and Eas Asia
This subsec ion illus a es he yea - on- yea changes in LNG
expo s om he Uni ed S a es o No h Wes Eu ope and Eas
Asia. Figu e4a shows a signi ican inc ease in US LNG expo s
o No hwes Eu ope in he las qua e o 2021, coinciding wi h
e o s o eplace Russian gas pipeline supplies. Howe e , despi e
his inc ease, Eu opean gas p ices con inued o ise sha ply,
sugges ing ha exis ing in as uc u e and ma ke capaci ies
we e ully u ilised, limi ing he po en ial o u he impo s o
s abilise o educe p ices. In con as , Figu e4b highligh s a de-
c ease in US LNG expo s o Eas Asia du ing he same pe iod.
This decline e lec s a shi in LNG ade dynamics, possibly
due o changes in compe i i e p essu es wi hin he global LNG
ma ke . These ends unde sco e he e ol ing ole o he Uni ed
S a es as a key LNG supplie and he di e ing impac s on e-
gional ma ke s.
FIGURE 3 | Δ YoY (%) in LNG impo s o majo egions: (a) No hwes Eu ope (NWE), (b) China, and (c) Japan & Ko ea. Own cons uc ion based
on da a ob ained om JODI(2024). [Colou igu e can be iewed a wileyonlinelib a y.com]
FIGURE 4 | Δ YoY (%) in US LNG expo s o: (a) No hwes Eu ope and (b) Eas Asian ma ke s. Own cons uc ion based on da a ob ained om
he EIA(2023). [Colou igu e can be iewed a wileyonlinelib a y.com]
1410 The Wo ld Economy, 2025
4 | Me hodology
Ou analysis examines he in eg a ion o he h ee egional gas
ma ke s using he coin eg a ion app oach. Be o e conduc ing
his analysis, we apply he augmen ed Dickey–Fulle (ADF) es ,
he Phillips–Pe on (PP) es , and he Kwia kowski–Phillips–
Schmid –Shin (KPSS) es o he h ee gas p ice se ies o e alu-
a e hei s a iona i y p ope ies. The ADF and PP es s examine
he null hypo hesis o non- s a iona i y (i.e., he p esence o a
uni oo ), while he KPSS es examines he null hypo hesis o
s a iona i y. I he esul s indica e ha he p ice se ies exhibi s
a uni oo , we p oceed wi h coin eg a ion analysis o examine
equilib ium ela ionships.
P e ious s udies ha e o en u ilised he adi ional symme -
ic coin eg a ion amewo k o analyse he in eg a ion o gas
p ices a bo h egional and in a egional le els (e.g., Sili e s o s
e al.2005; Asche e al.2002). Howe e , con en ional coin e-
g a ion es s may be misspeci ied when he adjus men p o-
cess is asymme ic. The me hodology p oposed by Ende s and
Siklos(2001) ex ends he widely used Engle and G ange (1987)
wo- s ep coin eg a ion p ocedu e by inco po a ing asymme -
ic adjus men s in he long- un ela ionships be ween gas
p ices. This ex ension, known as he Momen um Th eshold
Au o eg essi e (M- TAR) model, has been shown o pe o m be -
e in he p esence o asymme y, p o iding mo e eliable esul s
han me hods ha assume symme ic p ice adjus men s. This
app oach has been ex ensi ely applied o analyse asymme ic
adjus men in coin eg a ion ela ionships be ween a ious en-
e gy p ices (e.g., Hammoudeh e al.2008; Chiappini e al.2019;
Chang e al.2012).
In bo h he symme ic Engle and G ange  (1987) amewo k
and he asymme ic Ende s and Siklos(2001) ex ension, he i s
s ep is o es ima e he ollowing model, which ep esen s he
equilib ium ela ionship be ween wo egional gas p ice se ies,
using o dina y leas squa es (OLS):
whe e
P1
and
P2
ep esen he loga i hmic o ms o wo gas
p ice se ies. We es ima e h ee se s o gas p ice pai s: (TTF,
EAX); (HH, TTF); and (HH, EAX). The esiduals,
𝜀
, ob-
ained om Equa ion2, a e subsequen ly used in he second
s ep o he Engle and G ange  (1987) linea coin eg a ion
analysis (Equa ion3) and in he second s ep o he M- TAR
model o nonlinea coin eg a ion as p oposed by Ende s and
Siklos(2001) (Equa ion4):
The adjus men speed coe icien s, ρ0, ρ1, and ρ2, co espond
o he symme ic
(
𝜌
0)
and asymme ic (ρ1 and ρ2) coin eg a-
ion models. Addi ionally, he inclusion o lagged alues o
Δ𝜀
helps o ensu e ha he esiduals a e se ially unco ela ed. The
Hea iside indica o unc ion,
I
, is de ined as 1 i
Δ𝜀 −1
≥
𝜏
and 0
i
Δ�𝜀 −1<𝜏
, whe e
𝜏
is he h eshold alue, es ima ed using he
consis en sea ch me hod o Chan(1993).
We es o e idence o asymme ic adjus men s using wo
hypo heses. Fi s , we es he join null hypo hesis o no-
coin eg a ion (
H0:𝜌1=𝜌2=0
), wi h he c i ical alues ob ained
om Ende s and Siklos (2001). I he null hypo hesis o no-
coin eg a ion is ejec ed, we es o he null hypo hesis o sym-
me y (
H0:𝜌1=𝜌2
) using a s anda d F- es .
5 | Empi ical Resul s
This sec ion p esen s he empi ical analysis o he in eg a ion o
he h ee egional gas ma ke s. The analysis is s uc u ed as ol-
lows: Fi s , we es o a po en ial s uc u al b eak on 1 Oc obe
2021, using he Chow es on he log p ice di e en ials be ween
he p ice pai s, mo i a ed by signi ican de elopmen s in he
global gas ma ke . Nex , we examine he linea coin eg a ion e-
la ionships be ween he gas p ice pai s using he Engle–G ange
wo- s ep app oach. We hen conduc a nonlinea coin eg a ion
analysis, employing he MTAR model o in es iga e po en ial
asymme ies in p ice adjus men s. Finally, we es ima e sym-
me ic and asymme ic e o co ec ion models o assess he
sho - e m dynamics and adjus men s owa ds he long- un
equilib ium o each p ice pai .
5.1 | S uc u al B eak Analysis
We hypo hesise ha he ela ionships among ou a iables o
in e es may be a ec ed by a po en ial s uc u al b eak on 1
Oc obe 2021. This b eak da e is d i en by majo de elopmen s in
he na u al gas ma ke s, as discussed in he p e ious sec ion. Fo
ins ance, in he la e hal o 2021, geopoli ical ensions—pa ic-
ula ly Russia's delibe a e educ ion o gas expo s o Eu ope—
signi ican ly dis up ed supply. This caused a majo shock in he
global gas ma ke , leading o igh e ma ke condi ions.
To o mally es o his b eak, we ollow Büyük şahin
e al.(2013) and Luong e al.(2019), conduc ing a Chow(1960)
es on he log p ice di e en ials be ween he p ice pai s. We
pe o m he es using he ollowing speci ica ion:
He e,
TR
ep esen s a linea end,
𝜃
is a cons an e m, and
𝜙S −1
,
𝜙S −2
and
𝜙S −3
ep esen he lagged alues o he depen-
den a iable, whe e
𝜙1
,
𝜙2
and
𝜙3
a e he coe icien s on he
lagged e ms.
The esul ing F- s a is ics a e 17.413 (signi ican a he 1%
le el), 4.328 (signi ican a he 1% le el), and 2.867 (signi ican
a he 5% le el) o he sp eads EAX–TTF, HH–TTF and HH–
EAX, espec i ely, wi h 5 and 1764 deg ees o eedom. Gi en
hese signi ican s a is ics, we conclude ha a s uc u al b eak
occu ed a ound 1 Oc obe 2021. Consequen ly, he analysis is
conduc ed o e he pe iod om 1 Janua y 2016 o 1 No embe
2022, di ided in o wo subsamples, wi h 1 Oc obe 2021, as
he spli da e.
(2)
P1
=𝛽
0
+𝛽
1P2
+𝜀
(3)
Δ
𝜀 =𝜌0𝜀 −1+
p
∑
j
=
1
𝛿jΔ𝜀 −j+u
(4)
Δ
𝜀 =𝜌1I 𝜀 −1+𝜌2(1−I )𝜀 −1+
K
∑
i=1
𝜗iΔ𝜀 −i+u
(5)
S =𝜃+𝜆TR +𝜙1𝜙S −1+𝜙2𝜙S −2+𝜙3𝜙S −3+ϵ
,
1411
Summa y s a is ics o he h ee p ice se ies o e he wo sub-
samples a e p o ided in TableA1 in AppendixA, which shows
a shi owa ds highe p ices and g ea e a iabili y in he gas
ma ke s a e Sep embe 2021. The esul s o he uni oo es s
o he log le els and hei di e ences a e also p esen ed in
TableS2. The esul s show ha all he ime se ies in log le -
els a e I (1) a iables, meaning hey a e non- s a iona y in le els
bu become s a iona y a e i s di e encing. The e o e, coin e-
g a ion analysis is an app op ia e ool o in es iga e hei join
p ope ies.
5.2 | Examining he Linea Coin eg a ion
In he con ex o es ing o linea coin eg a ion, we apply he
wo- s ep app oach p oposed by Engle and G ange (1987). This
app oach in ol es i s es ima ing he equilib ium ela ionship
o each p ice pai acco ding o he speci ica ion in Equa ion2.
In he second s ep, we ob ain he esiduals om his eg ession
and apply he Engle–G ange esidual- based coin eg a ion es
o de e mine whe he he esiduals a e s a iona y.12
Table1 p esen s he esul s o he wo- s ep analysis, wi h he las
column p o iding he es s a is ics o he s a iona i y o he e-
siduals, which indica e whe he he a iables a e coin eg a ed.
Fo he EAX–TTF pai , he es ima ed
𝛽1
is 0.973 in he i s sub-
sample, indica ing ha a 1% inc ease in he TTF p ice is associ-
a ed wi h a 0.973% inc ease in he EAX p ice. Howe e ,
𝛽1
d ops
o 0.663 in he second subsample. The es ima ed
𝛽1
coe icien s
om he coin eg a ion eg essions o HH agains EAX and TTF
a e ela i ely lowe . In he i s subsample, he es ima ed
𝛽1
is
0.397 o HH–TTF and 0.376 o HH–EAX. In he second sub-
sample, he es ima ed
𝛽1
o HH–TTF emains s able a 0.385,
while HH–EAX declines sha ply om 0.376 o 0.214, indica ing
a weakening p ice linkage.
The esul s also show ha he es ima ed coe icien
𝛽0
a ies
ac oss he h ee pai s (EAX–TTF, HH–TTF, and HH–EAX)
and he wo subsamples, wi h a no able inc ease in he second
subsample. Fo example,
𝛽0
ises om 0.244 o 1.122 o EAX–
TTF, om 0.323 o 0.404 o HH–TTF, and om 0.283 o 1.036
o HH–EAX. This inc ease sugges s ha baseline p ice le -
els ac oss he h ee ma ke s ha e isen o e ime, indica ing a
g owing di e gence in ma ke condi ions. This di e gence may
be due o di e ences in egional supply and demand balances,
anspo a ion cos s, o ma ke - speci ic ac o s such as egula-
o y changes a ec ing na u al gas p icing.
The Engle–G ange es s a is ics in he las column o Table1
indica e ha he h ee p ice pai s a e coin eg a ed in he i s
subsample. Howe e , in he second subsample, he es s a is ics
a e no s a is ically signi ican , p o iding no e idence o linea
coin eg a ion.
5.3 | Examining he Nonlinea Coin eg a ion
In he p eceding subsec ion, he Engle–G ange es , which as-
sumes a linea and symme ic adjus men p ocess, indica es ha
he e is no e idence o coin eg a ion o any o he p ice pai s
du ing he second pe iod. This subsec ion examines es ima es
om he MTAR model p oposed by Ende s and Siklos(2001),
which explici ly accoun s o po en ial asymme ies in he ad-
jus men p ocess owa ds equilib ium. This analysis aimed
o de e mine whe he he e is e idence o asymme ies in he
i s subsample, which would sugges ha he adjus men p o-
cess occu s a di e en speeds depending on he di ec ion o
de ia ions om equilib ium (posi i e s. nega i e), a he han
symme ically. Addi ionally, we seek o es ablish whe he he
MTAR model p o ides e idence o coin eg a ion o any o he
p ice pai s in he second subsample pe iod.
Table 2 p esen s he esul s o he MTAR coin eg a ion es .
Column (1) shows he es ima ed h eshold alues, which indi-
ca e he poin a which adjus men s swi ch be ween egimes o
posi i e and nega i e de ia ions om equilib ium. Al hough he
es ima ed h esholds a e close o ze o, ou analysis shows ha
models wi h an es ima ed h eshold alue pe o m be e —ac-
co ding o he in o ma ion c i e ia— han models assuming a
ixed h eshold o ze o. Columns (2) and (3) show he es ima ed
pa ame e s o
𝜌1
and ρ2, as speci ied in Equa ion 4. He e,
𝜌1
ep esen s he speed o adjus men in esponse o posi i e de-
ia ions om equilib ium, whe eas
𝜌2
ep esen s he speed o
adjus men o nega i e de ia ions. I he absolu e alue o
𝜌1
is
g ea e han ha o ρ2, his indica es ha he adjus men p o-
cess is as e in esponse o posi i e de ia ions om equilib ium.
Con e sely, i
∣𝜌2∣
is g ea e , he adjus men is as e in esponse
o nega i e de ia ions om equilib ium. Fo example, in he
TABLE 1 | Linea coin eg a ion analysis.
P ice pai Subsample β0β1R2EG (1987)
EAX- TTF Fi s 0.244a[0.015] 0.973a[0.009] 0.895 −5.087a
Second 1.122a[0.098] 0.663a[0.027] 0.687 −2.841
HH- TTF Fi s 0.323a[0.012] 0.397a[0.007] 0.682 −4.478a
Second 0.404b[0.160] 0.382a[0.044] 0.215 −1.908
HH- EAX Fi s 0.284a[0.014] 0.376a[0.007] 0.645 −4.718a
Second 1.036a[0.214] 0.212a[0.060] 0.043 −1.628
No e: The i s subsample includes da a om 1 Janua y 2016 o 30 Sep embe 2021, while he second subsample includes da a om 1 Oc obe 2021 o 1 No embe
2022. S anda d e o s o he es ima ed coe icien s a e gi en in he squa e b acke s. The column i led ‘R2’ gi es he goodness o i o he eg essions. The las column
displays he Engle and G ange (1987) es s a is ic (EG (1987)) o coin eg a ion, wi h a signi ican es s a is ic sugges ing ha he esiduals a e s a iona y, hus
con i ming coin eg a ion be ween he a iables. The numbe o lags o he Engle–G ange coin eg a ion es was selec ed using he AIC. The c i ical alues o his es
a e ob ained om MacKinnon(2010). The symbols a and b deno e signi icance a he 1% and 5% le els, espec i ely.
1412 The Wo ld Economy, 2025
ela ionship be ween EAX and TTF in he i s subsample, he
es ima ed h eshold is −0.023, wi h adjus men coe icien s o
−0.026 o posi i e de ia ions and −0.120 o nega i e de ia-
ions. This esul indica es ha posi i e de ia ions om equi-
lib ium (whe e
Δ𝜀 −1≥−0.023
) a e elimina ed a a ela i ely
slowe a e o 2.6% pe day. In con as , nega i e de ia ions om
equilib ium a e adjus ed a a much as e a e o 12% pe day.
Consequen ly, he e is subs an ially slowe con e gence owa ds
equilib ium o posi i e de ia ions (abo e he h eshold) han
o nega i e de ia ions (below he h eshold). These indings
sugges ha a bi ageu s a e mo e ac i e in exploi ing la ge
p o i able oppo uni ies depending on he di ec ion he sp ead
is mo ing om i s equilib ium posi ion. This also implies ha
du ing he i s subsample pe iod, he ma ke adjus s mo e ap-
idly when EAX p ices a e dec easing ela i e o TTF p ices. This
conclusion is consis en wi h Chiappini e al.(2019), al hough
he es ima ed speeds o adjus men in bo h egimes du ing ou
sample pe iod a e highe han hei es ima es. In he second
subsample, he es ima ed h eshold is 0.015, wi h adjus men
coe icien s o −0.208 o posi i e de ia ions and −0.070 o neg-
a i e de ia ions. This ou come indica es ha posi i e de ia ions
om equilib ium a e elimina ed apidly, a a a e o 20.8% pe
day. The esul s o nega i e de ia ions do no show signi ican
adjus men , as he coe icien o nega i e shocks is s a is ically
insigni ican .
Column (4) in he able p esen s he es o he join null hypo he-
sis o no coin eg a ion wi h MTAR adjus men
(
H
0
:𝜌
1
=𝜌
2
=0
)
.
The esul s indica e ha his null hypo hesis is ejec ed o each
p ice pai in he i s subsample, as he es s a is ic exceeds he
c i ical alues p o ided by (Ende s and Siklos2001). Gi en his
esul , we p oceed o es he null hypo hesis o
H0:𝜌1=𝜌2
. The
esul s, shown in Column (5), indica e ha his null hypo hesis
is ejec ed, suppo ing he p esence o asymme ic adjus men .
Howe e , in he second subsample, Column (4) shows ha
he null hypo hesis o no coin eg a ion is ejec ed only o he
EAX–TTF p ice pai . The lack o nonlinea coin eg a ion o he
HH–TTF and HH–EAX pai s sugges s a decoupling o he US
gas ma ke om he Eu opean and Asian ma ke s du ing his
pe iod.
5.4 | Resul s o he (a)symme ic
E o Co ec ion Model
In his s ep, we es ima e bo h symme ic and asymme ic e o
co ec ion models (ECMs) o examine he adjus men p ocesses
o indi idual p ices owa ds equilib ium. We es ima e he sym-
me ic o asymme ic ECM o each p ice pai based on he
coin eg a ion esul s om he p e ious subsec ion.
Table3 p esen s he es ima ion esul s o he wo subsamples.13
The magni ude o he e o co ec ion e m (ECT) indica es he
speed a which de ia ions om equilib ium a e co ec ed. Fo
ins ance, i he ECT is −0.250, i sugges s ha app oxima ely
25% o he de ia ion is co ec ed each day, implying ha ull
co ec ion o equilib ium would ake abou 4 days. The esul s
indica e ha , o he EAX–TTF pai in he i s subsample, he
ECT o EAX in he high egime is −0.003 and s a is ically insig-
ni ican , sugges ing no adjus men o posi i e de ia ions. In he
low egime, he ECT o EAX is −0.090 and s a is ically signi -
ican , indica ing a co ec ion owa ds equilib ium o nega i e
de ia ions. Fo TTF, he ECT is −0.020 and s a is ically signi i-
can in he high egime and −0.020 and s a is ically signi ican
in he low egime, indica ing adjus men s in bo h cases. In he
second subsample, he ECT o EAX is −0.116 and s a is ically
TABLE 2 | Nonlinea coin eg a ion analysis.
P ice pai Subsample
(1) (2) (3) (4) (5)
Th eshold ρ1ρ2Φ (H0: ρ1 = ρ2 = 0) F (H0: ρ1 = ρ2)
EAX- TTF Fi s −0.023 −0.026a−0.120b27.555b25.492b
(−2.390) (−7.276) [0.000]
Second 0.050 −0.208b−0.070 5.909c3.665c
(−3.230) (−1.622) [0.057]
HH- TTF Fi s 0.012 −0.008 −0.044b12.849b5.586b
(−0.671) (−5.035) [0.018]
Second −0.068 −0.020 −0.090a2.915 2.218
(−1.308) (−2.029) [0.138]
HH- EAX Fi s 0.033 −0.083b−0.025b15.521b8.665b
(−4.480) (−3.427) [0.003]
Second 0.010 −0.002 −0.031a2.029 1.393
(−0.119) (−2.011) [0.239]
No e: The i s subsample includes da a om 1 Janua y 2016 o 30 Sep embe 2021, while he second subsample includes da a om 1 Oc obe 2021 o 1 No embe 2022.
Column (1) p o ides he es ima ed h eshold alues. Columns (2) and (3) p o ide he es ima ed coe icien s in Equa ion4. - s a is ics o he es ima ed coe icien s a e
gi en in b acke s. Column (4) shows he null hypo hesis es s o he h eshold coin eg a ion wi h he c i ical alues om Ende s and Siklos(2001) as ollows: C.V (1%)
is 8.310; C.V (5%) is 6.050; C.V (10%) is 5.060. Column (5) gi es he second null hypo hesis. The symbols a, b, and c deno e signi icance a he 1%, 5%, and 10% le els,
espec i ely.