Global linkages across sectors and frequency bands: a band spectral panel regression approach

Lyu, Jingjing; Süssmu h, Be nd
A icle — Published Ve sion
Global linkages ac oss sec o s and equency bands: a
band spec al panel eg ession app oach
Re iew o Wo ld Economics
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Lyu, Jingjing; Süssmu h, Be nd (2025) : Global linkages ac oss sec o s and
equency bands: a band spec al panel eg ession app oach, Re iew o Wo ld Economics, ISSN
1610-2886, Sp inge , Be lin, Heidelbe g, Vol. 161, Iss. 4, pp. 1421-1462,
h ps://doi.o g/10.1007/s10290-025-00590-8
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/330653
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/
ORIGINAL PAPER
Re iew o Wo ld Economics (2025) 161:1421–1462
h ps://doi.o g/10.1007/s10290-025-00590-8
Abs ac
We in oduce he echnique o band spec al panel eg ession (BSPR) o analyze
global linkages ac oss sec o s and equency bands. I elies on decomposing ime
se ies—allowably measu ed in mixed obse a ion equency—in o “de ia ion cycle”
dynamics by equency band. We use i o compu e measu es o eal co-mo emen ,
ade linkage, inancial ma ke in eg a ion, and policy coo dina ion band by band.
Conside ing in a-indus y as well as in e -indus y linkage indica o s, we apply i
o da a o con empo a y China and 22 o i s op-25 majo ading pa ne s in he
p e- ade wa and p e-pandemic e a. Band-speci ic ixed e ec s and band-indus y-
speci ic in e ac ion e ms a e included. Fo labo in ensi e indus ies co-mo emen
h ough in a-indus y ade linkages is ound o be band-speci ic. Mo eo e , ou
esul s cla i y he puzzle o inancial globaliza ion implying eal egionaliza ion o
con agious synch oniza ion o cyclical dynamics. We ind he la e o hold in he
4–6 yea s band and he o me in he 6–10 yea s ange.
Keywo ds Spec al eg ession · F equency domain · Cyclical co-mo emen ·
Sec o s
JEL Classi ica ion C32 · C49 · E32 · F40
Accep ed: 7 Ap il 2025 / Published online: 23 May 2025
© The Au ho (s) 2025
Global linkages ac oss sec o s and equency bands: a band
spec al panel eg ession app oach
JingjingLyu1· Be ndSüssmu h2,3
Be nd Süssmu h
[email p o ec ed]
Jingjing Lyu
[email p o ec ed]
1 No heas No mal Uni e si y, Changchun, China
2 Ins i u e o Empi ical Resea ch in Economics, Uni e si y o Leipzig, G immaische S . 12,
04109 Leipzig, Ge many
3 CESi o, Munich, Ge many
1 3
J. Lyu, B. Süssmu h
1 In oduc ion
A he la es since he collapse o he B e on Woods exchange a e sys em in he
ea ly 1970s, global linkages, co-mo emen , and spillo e e ec s ac oss na ions o he
wo ld economy go in he ocus o a b oad esea ch agenda bo h om a heo e ical
as well as empi ical pe spec i e (Cheung and Wes e mann, 2013). Since he las wo
decades he heo e ical and empi ical li e a u e s esses he dependency o hese phe-
nomena on he he e ogenei y o sec o s (Kalemli-Ozcan e al., 2001; Belke & Heine,
2006; Ko inek e al., 2010; Bie baume -Polly e al., 2016; Azcona, 2022; Sh awan &
Dubey, 2022) and on di e ences in pe iodici ies o analyzed dynamics, ha is, on he
speci ic equency bands conside ed (Ahmed e al., 2004; Aguia & Gopina h, 2007;
Kose e al., 2012; Blonigen e al., 2014; Nachane & Dubey, 2013, 2018, 2021). The
ecen Uni ed S a es (US)-China ade wa , he COVID-19 pandemic, and he Russia-
Uk aine wa subs an ially u he ed he discussion (Bohn e al., 2021; Bengu ia e al.,
2022; Li & Su, 2022). Due o he un o essen occu ence and mos ly exogenous na u e
o hese e en s, he ocus o he p esen s udy, howe e , is on he p e-US-China ade
wa e a.
Sec o s uc u es domina ed by b oadly de ined sec o s wi h la ge common shocks
end o cyclical coupling and ul ima ely o global in e dependence o dynamics a
sho - e m and long- e m ho izons. The opposi e applies o sec o s uc u es domi-
na ed by b oadly de ined sec o s wi h idiosync a ic small shocks. In he i s case
inancial cons ain s a e globally con agious ac oss sec o s and na ions, whe eas in
he second case hey a e no o only egionally con agious ac oss eme ging ma ke
economies (EME). The la e is due o some EME indus y po en ially bene i ing
om he decline o he same sec o o a ela ed indus y in an ad anced economy
ha compe es o he same inpu s. Asynch onous p oduc ion dynamics o decoupling
esul s; see, e.g., Ko inek e al. (2010).
Ahmed e al. (2004) can be in e p e ed as sugges ing in eg a ed in en o y manage-
men and o he business p ac ices–mos ly conce ning in a-indus y a he han in e -
indus y in e dependency– o imply synch onici y a ela i ely high equencies.
Howe e , i is unclea whe he his applies o bo h ad anced economies and EME.
Co-mo emen due o mone a y and iscal policy coo dina ion is expec ed p ima ily a
business-cycle equencies, whe eas co-mo emen due o echnological inno a ions
a all equencies alike. G ow h spillo e s, on he o he hand, can be o “cyclical
g ow h” o “secula g ow h” na u e. This dis inc ion is a well-known di icul y (Agu-
ia & Gopina h, 2007), in pa icula , in he EME con ex .1 These ype o spillo e s
a e likely o be a ibu able o ade in ensi y and specializa ion, echnology ans e s,
o in lows o o eign di ec in es men (FDI). Co-mo emen a dis inc , hough in
any case, a he low- equency bands migh be explained, o ins ance, by di e en
shades o liquidi y and di e en modes o en y, such as g een ield and me ge s and
acquisi ion (Aguia & Gopina h, 2007; Gawellek e al., 2021).
1 The ela ed s and o li e a u e on he P ebisch-Singe hypo hesis and secula low equency supe cycles
(Ha ey e al., 2010; E en & Ocampo, 2013) in eal commodi y p ices has made some me hodological
p og ess in his ega d. I consis s in subs an ially inc easing he powe o augmen ed Dickey-Fulle (ADF)
es s by including a equency domain componen , e e ed o as lexible Fou ie componen (FFC) in he
ADF es ing p ocedu e (Ende s & Lee, 2012; Winkel ied, 2018).
1 3
1422
Global linkages ac oss sec o s and equency bands: a band spec al…
As onishingly, hese wo s ands o ecen li e a u e, s udying he ole o sec o s
o global in e dependence on he one hand and o equency bands on he o he , ha e
no been sa is yingly in eg a ed so a . Ou s udy seeks o con ibu e o he li e a u e
in his ega d and ies o shed some ligh on global linkages ac oss sec o s and e-
quency bands. To his end, we conside i e b oadly de ined sec o s classi ied by ac-
o in ensi y and ou di e en equency bands in he empi ical sec ion o ou s udy.
Ou second cen al con ibu ion is o me hodological na u e as we in oduce he
echnique o band spec al panel eg ession (BSPR). The band spec al panel eg es-
sion (BSPR) model is, in some sense, a panel e sion o he mo e gene al band spec-
al eg ession model (Engle, 1972, 1974; Co bae e al., 2002; Assenmache -Wesche
& Ge lach, 2008a). I elies on dissec ing ime se ies ha can be measu ed in di e en
equency–e.g. in mon hly, qua e ly, and annual equency– in o “de ia ion cycle”
dynamics by equency band (A is e al., 2004). The dissec ed componen s allow us
o compu e measu es o co-mo emen , ade linkage, ma ke in eg a ion, and policy
coo dina ion by equency band i espec i e o he obse a ion- equency o he
unde lying ime se ies. The esul ing panel s uc u e consis s in en i ies, e e ing
o economies, and equency bands, e e ing o pe iodici ies o cyclical dynamics,
a he han en i ies and ime as in s anda d panel models. BSPR models a e lexible in
allowing o band-speci ic ixed e ec s and band-indus y-speci ic in e ac ion e ms.
Conside ing in a-indus y as well as in e -indus y linkage indica o s, we apply he
p oposed me hod o da a o con empo a y China and i s majo ading pa ne econo-
mies om 1997 o 2016, i.e. p io o he US-China ade wa . Fo mo i es and conse-
quences o he la e agains he backd op o p o ec ionsim and global linkages; see
Bengu ia e al. (2022), Guo e al. (2018), Noland (2018), Sheng e al. (2019).
Ou me hodological app oach imp o es exis ing app oaches examining global
linkages:
(i) I is able o deal wi h mixed obse a ion equencies o ime se ies ha migh be
weekly, mon hly, qua e ly o annual in mos mac oeconomic applica ions. S an-
da d (dynamic) panel app oaches would equi e an a p io i empo al agg ega ion
o he lowes a ailable obse a ion equency. This agg ega ion migh well be
con o e sial and possibly obscu ing he con ained dynamics. Handling mixed
obse a ion equency da a, howe e , is also possible o al e na i e app oaches
such as dynamic co ela ion (Fid muc e al., 2013) o Koopman mode analysis
(Hua e al., 2016).
(ii) In con as o exis ing app oaches, i goes beyond a pu ely desc ip i e ea men
o co-mo emen and linkages as i also decomposes a wide a ie y o dynamic
co a ia es in o di e en equency bands. The app oach, hus, opens he oppo -
uni y o go beyond a desc ip i e analysis. I is explo a i e in na u e and comes
wi h a s a is ical signi icance basis. By ying o explain economic beha io and
ela ions a di e en equencies, i s ands in a cen u y long adi ion o he p es-
en jou nal. See, o ins ance, he seminal con ibu ion on cyclical ship cons uc-
ion by Nobelis Jan Tinbe gen in 1931 published by he Wel wi scha liches
A chi , i.e. he p ecu so y jou nal o he Re iew o Wo ld Economics (Tinbe gen,
1931).
1 3
1423
J. Lyu, B. Süssmu h
(iii) Due o (ii) ou app oach can cla i y enigma ic and allegedly con adic o y ind-
ings om he heo e ical as well as applied li e a u e. I is, o ins ance, unclea
om he li e a u e whe he inancial globaliza ion implies eal egionaliza ion
h ough a home bias o in es o s and, hus, mo e decoupling (Hea hco e & Pe i,
2004). O whe he inancializa ion and inancial ma ke con agiousness a he
come wi h mo e coupling. See, e.g., Nachane and Dubey (2018). The BSPR
app oach is able o disen angle he esul ing o al e ec o di e en equency
bands.
The emainde o he pape is o ganized as ollows. Sec ion 2 s a s wi h some p in-
ciples o spec al analysis and band spec al eg ession analysis unde lying he BSPR
app oach. I con inues wi h an ou line o how o implemen he BSPR model elying
on de ia ion-cycle il e ing, In Sec . 3 an applica ion o he p oposed echnique s udy-
ing global linkages ac oss sec o s and equency bands is gi en. Sec ion 4 concludes.
2 Band spec al panel eg ession
2.1 P inciples o spec al analysis and band spec al eg ession
The me hodological s a ing poin o spec al analysis as canonical analogue o au o-
co ela ion analysis in he ime domain is he ep esen a ion o a ime se ies
X
as a
pe iodic (sinusoidal) componen wi h known pe iod leng h, i.e.
X =
R
cos (
ω
+
ϕ
)+
z
,
(1)
whe e
z
is assumed o ep esen a s a iona y mean-ze o andom a iable, R deno es
ampli ude, angle
(ω +ϕ)
is measu ed in adians wi h
π
adians (i.e. 180 deg ees), and
ω
is deno ing he angula equency (o equency exp essed in adians), i.e. he num-
be o adians pe uni o ime. Angula equency
ω
is ela ed o o dina y equency ,
i.e. he numbe o comple ed cycles pe uni o ime, by
=ω
2π
. Pe iod o pe iodici y
T o a cyclic pa e n is gi en by
T= −1
. Thus, he highes measu able equency–
e e ed o as “Nyquis equency”– co eponds o
ω=
π
⇔
=1
2⇔
T
=2
.
I desc ibes a wo-pe iod cycle, i.e., o a mean-ze o s a iona y se ies a dynamics
al e na ing om nega i e suppo o posi i e suppo wi h peak- o-peak o ough- o-
ough dis ance equaling wo pe iods (
T=2
). Gi en ha any se ies can be exp essed
as he supe posi ion o se e al such pe iodic componen s wi h di e en ampli ude,
equency, and phase shi , we may w i e
E
(X )=
k
∑
j=1
Rjcos (ωj +ϕ)
.
(2)
F om a cen al p ope y o igonome ic unc ions,
cos (ω +ϕ) = cos ω ·cos ϕ−sin ω ·sin ϕ
, i ollows ha
1 3
1424

Global linkages ac oss sec o s and equency bands: a band spec al…
E
(X )=
k
∑
j=1
(ajcos ωj +bjsin ωj )
wi h
a
j=
R
jcos
ϕ
j
;
b
j=−
R
jsin
ϕ
j
,
(3)
ha is, ampli udes a e hemsel es now gi en by sinusoidal laws o mo ion.
Pe iodog am analysis o , in gene al, he spec um decomposes he a iance o
s ochas ic p ocess
X
in o i s “p- h ha monics” wi h
ωp=2π·p
N
, i.e. p-mul iples o
2π
N
wi h p ep esen ing he sha e o he N- h slice o he uni ci cle “cake.” I can be
o malized as
I
(ωp)=
1
Nπ
[(
∑X cos
2
πp
N
)2
+
(
∑X sin
2
πp
N
)2]
=1
Nπ
{[∑
(x
−
x) cos ωp
]
2+
[∑
(x
−
x) sin ωp
]
2
},
(4)
whe e we d opped he sum ope a o indices o no a ional ease.
F om polynomial mul iplica ion and conside ing ha he c oss-
p oduc s o cosine and sine unc ions sum o ze o, i ollows ha
I(
ωp
)= 1
Nπ ∑s, −1(
x
−
x
)(
xs
−
x
) (cos
ωp
cos
ωps
+ sin
ωp
sin
ωps
).
Con-
side ing
γ
k
=1
NN−k
=1 (
x
−
x
)(
x
+
k
−
x
)
as de ining he sample au oco e-
la ion unc ion (SACF) and ano he cen al p ope y o igonome ic unc ions,
cos ωp cos ωp( +k) + sin ωp sin ωp( +k) = 2 cos ωp( +k− ) = 2 cos ωpk
,
as well as Eule ’s o mula allows us o e-w i e
I(ωp)
as he disc e e Fou ie ans-
o ma ion (DFT) o he SACF
I
(ωp)=
1
π

c0+2
N
−1

k=1 γkcos ωpk

=
N−1

−
(N
−
1)

γke−1
π·iωpk.
(5)
Hence, he un-smoo hed pe iodog am-es ima e o he spec um is gi en by

xx (ω)=

γ0+2
∞

k=1 
γk·cos ωk
.
(6)
As shown by Engle (1972, 1974), i
y =x′
β+ε
o
=1,...,N
is a alid eg es-
sion model in he ime domain, i can be ans o med in o he equency domain by
applying a DFT o bo h i s dependen a iable and i s independen a iables. Deno -
ing acco dingly ans o med a iables as
y
,
x
, he eg ession in he equency domain
is
y=x′β+ε
. The DFT no ably does no a ec he s anda d eg ession s uc u e.
The es ima o

β
can be w i en as
1 3
1425
J. Lyu, B. Süssmu h

β
=

N−
1

k=0 
xx (ωk)
−1
N−
1

k=0 
xy (ωk)
,
(7)
whe e

xy (ω)
is a ec o o c oss-pe iodog ams. No e, since

β
a e ages o e pe i-
odog ams, he e is no need o smoo h hese as is necessa y when es ima ing he
spec um.2 In con as o he FFC-en iched ADF es ing li e a u e (Ende s & Lee,
2012; Winkel ied, 2018), DFT componen s a e no included as addi ional eg esso s.
Ou s a egy a he consis s in a equency domain ans o ma ion o he eg ession
model in i s en i e y. The bene i o ansla ing he en i e eg ession model in o he
equency domain is he oppo uni y o check whe he a speci ic model applies o
some bu no o all equencies. To do so, he eg ession model is mul iplied by an
N×N
ma ix A wi h uni y on he main diagonal o each included equency and
ze o en ies elsewhe e
Ay=Ax′β+Aε, whe e E(Aε)(Aε)∗=σ2A
(8)
wi h as e isk ‘*’ deno ing complex conjuga e anspose. Thus, o compu e ec o

β
we sum o e a pa icula equency band a he han o e he ull ange o equen-
cies as in (7). I (7) is es ima ed only o a subse o equencies, bu holds ue o all
equencies, he es ima o is consis en bu ine icien as i does no use all a ailable
in o ma ion.
The logics o Engle’s a gumen can be analogously applied o a alid pe iod-
speci ic (o ime- ixed-e ec s) panel model
yi =α +x′
i β+νi+εi
, whe e
i=1,...,I
deno es c oss-sec ional en i ies and
νi
ixed e ec s wi h b obse a ions
pe g oup i, o jus as well o a equency-band-speci ic (o band- ixed-e ec s) panel
model in he equency domain
y
jb
=
αb
+
x
′
jb
β
+
νj
+
εjb
wi h
E
(
εjb
|
αb,ν
j,x
jb
H,...,x
jb
L
)=0,
(9)
and
bH
co esponding o he equency band comp ising he highes equencies
including he Nyquis (o nea -Nyquis ) equency,3 and
bL
con aining he lowe
bound alue o conside ed equency o uppe bound alue o pe iodici y, i.e. o
TH
,
whe e
TH
migh be chosen such ha a co esponding cyclicali y eplica es i sel , a
leas , once o e he conside ed pe iod o leng h N. He e,
j=1,...,J
and
νj
deno e
c oss-sec ions and co esponding ixed e ec s, espec i ely.
2 Fo his poin and he a gumen a ion in he emaining pa o he p esen pa ag aph see Assenmache -
Wesche and Ge lach (2008b, pp. 423–424).
3 Ac ually, Nyquis co eponds o he lowes pe iodici y,
TL
, a s ake. I he aw se ies
y
and
x
o he
same o di e en en i ies a s ake a e o di e en equency o obse a ion (“mixed equency”),
TL
migh
be chosen so as o cap u e a pe iodici y ha co esponds o he Nyquis equency o he se ies wi h he
lowes esolu ion o obse a ion- equency; e.g., a wo yea s pe iodici y in he case o annual se ies ep e-
sen ing he lowes esolu ion se ies in e ms o equency o obse a ion.
1 3
1426
Global linkages ac oss sec o s and equency bands: a band spec al…
2.2 Implemen ing band spec al panel eg essions
In he ollowing, we de elop a p ocedu e o implemen BSPR model (9). I makes use
o he no ion o band-pass de ia ion cycles. A is e al. (2004) de ine de ia ion cycle
dynamics in e ms o a cyclical dynamics de ia ing om end o po en ial. Thei
de ini ion, hus, implies ha he de ia ion cycle ep esen s an unobse ed componen
wi hin an addi i e o mul iplica i e unobse ed componen s model, ha is, a signal-
noise decomposi ion o , mo e speci ically, a end-cycle decomposi ion.
The smoo hed minimum mean squa e es ima o o he signal, i.e. he end compo-
nen
µ
, o he local linea end model o se ies
y
gi en by
y =µ +ϵ , ϵ
i.i.d.
∼N
(
0,σ2
ϵ
)
µ =µ −1+β −1+η ,η
i.i.d.
∼N(0,σ2
η)
β =β
−
1+ζ , ζ
i.i.d.
∼
N
(
0,σ2
ζ
)
(10)
o
=1,2,...,N
wi h es ic ions
σ2
η=0
and
σ2
ϵ
/σ
2
ζ=λ
, minimizes he penal-
ized leas squa e (PLS) c i e ion
PLS
=
N
∑
=1
(y −µ )2+λ
N
∑
=3 (
∆2µ
)
2
,
(11)
whe e Lag ange mul iplie
λ
cap u es he a iabili y o he noise, i.e. he cyclical,
componen ela i e o ha o
µ
. Fo
σ2
η
app oaching ze o,
λ
goes o in ini y, and he
limi ing ep esena ion o
µ
is a s aigh line (Hod ick & P esco , 1997). The noise,
i.e. i egula o cyclical, componen is
y −µ
.
Assuming he a ailabili y o a double-sided in ini e sample,
y +j,
j=−∞,...,+∞
,
he Wiene -Kolmogo o il e (Ha ey & P oie i, 2005) p o ides he minimum mean
squa e linea es ima o o
µ
, ha is
µ
|∞ =w(L)y wi h w(L)= σ
2
ζ
σ2
ζ
+
|
1
−
L
|
2σ2
ϵ
=
1
1+λ
|
1
−
L
|
4 (12)
and
|1−
L
|2= (1 −
L
)(1−
L
−1)
. Le
L=1
, i can be seen ha he weigh s o he
il e sum up o one. The equency esponse unc ion o his il e is
w(
e−iω
)
=
1
1+4
λ
(1 −cos
ω
)
2
.
(13)
I equals one a ze o equency and dec eases mono onically o
ω
app oaching
π
( i.e.
he Nyquis equency). Hence, (12) is o be in e p e ed as a low-pass il e wi h co -
esponding high-pass il e
1−w(L)
. The implici cu -o equency
ωc
co esponds
o a gain

w

e
−iω=1
2
. I sa is ies
1 3
1427
J. Lyu, B. Süssmu h
λ
=
[
4 (1 −cos ωc)2
]−1
=
0.25
(1 −cos ωc)
2
.
(14)
F om (14), i is s aigh o wa d o cons uc an app oxima e band-pass il e wi hou
su e ing om una ailabili y o end-o -sample es ima es, as is he case o wo-sided
(cen e ed) o one-sided MA- il e s such as he il e s p oposed by B y e al. (1971),
Bax e and King (1999), Ch is iano and Fi zge ald (2003), which seem inapp op i-
a e gi en he no o iously sho pe iod o obse a ion o economic ime se ies. The
la e applies, in pa icula , in he case o EME. The app oxima e band-pass il e is
achie ed by wha is widely known in he enginee ing sciences as a pa allel ci cui
applica ion o a low-pass il e . I is gi en in he p esen con ex by
y

λL,λ
H

=

µ

λL

−

µ

λH

wi h





λL=

4

1−cos

2π/TL

2
−1
λH=

4

1
−
cos

2π/TH

2

−1
.
(15)
Each o hese ans o med se ies a ies c oss-sec ionally wi h
j=1,...,J
and
ac oss equency bands indexed by
b=1,...,B
acco ding o
y
jb

λ
L
b,λ
H
b
=
µ j

λ
L
b

−µ j

λ
H
b
wi h

λL
b=

4

1
−
cos 2π
b
·
δ

2

−1λH
b=

4

1
−
cos 2π
(b+1)δ

2

−1
,
(16)
whe e
δ
depends on he obse a ion- equency o
y
and always ep esen s mul iples o
he lowes esolu ion equency o obse a ions o he J di e en
y
( and
x
) se ies. I
he la e is, o example, annual, we a e gi en wi h
δ∈{δm= 24; δq= 8; δa=2}
,
whe e supe sc ip m, q, and a deno es mon hly, qua e ly, and annual equency o
obse a ion, espec i ely. Fo ins ance, o
B=4
,
y jb
e ains cyclicali ies wi h pe i-
odici y o 2–4 yea s ( o
b=1
), 4–6 yea s ( o
b=2
), 6–8 yea s ( o
b=3
) and 8–10
yea s ( o
b=4=B
), espec i ely. The only emaining pa ame e ha needs o be
chosen in ad ance and app op ia ely, i.e. o he obse a ion- equency o he unde -
lying se ies, which is decomposed in o band-componen s, is
δ
. We p oceed analo-
gously wi h all exogenous se ies a s ake ende ing
x jb
band-speci ic ans o ms.
3 A BSPR applica ion: global linkages
In he ollowing h ee pa ag aphs, we b ie ly su ey he ecen li e a u e on he heo-
e ical a ionale o and co esponding empi ical e idence o he h ee co e ac o s
de e mining global linkages: ade linkages, inancial in eg a ion, and policy coo di-
na ion. The adi ional heo e ical iew on global linkages es s on linkages induced
by ade o inancial in eg a ion o he in e -indus y ype. I is in he spi i o he pop-
ula Hecksche -Ohlin ade models (Baldwin, 2013) ha a e oo ed in compa a i e
ad an age easoning. The linchpin mechanism o hese models and o a ionaliza-
ions o co esponding empi ical indings is he inc easing specializa ion in p oduc-
1 3
1428
Global linkages ac oss sec o s and equency bands: a band spec al…
Fig. 4 Bila e al IIP-co ela ions: de eloping coun ies
Fig. 3 Bila e al IIP-co ela ions: EME coun ies
1 3
1435

J. Lyu, B. Süssmu h
in so a as i s bila e al IIP-co ela ion inc eases wi h implied pe iodici ies o e he
whole ange.
Clea ly con ex pa e ns—wi h a dec ease a he in e media e bands and an inc ease
in he long- un equencies—a e gi en o B azil, Sou h A ica (Fig. 3), and Malaysia
(Fig. 4). Fo he emaining majo i y o analyzed ading pa ne economies, we ind a
“hockey s ick”-like shape wi h IIP-co ela ion peaking a he highes equency band,
dec easing up o he 6–8 yea s band, and han ei he s agna ing o sligh ly inc easing
in he 8–10 yea s cyclical g ow h equency band. The e is no a gene al in ui ion o
simila i ies and di e ences o hese shapes o band-speci ic pai wise IIP co ela ions.
Any in e p e a ions o such simple desc ip i e measu es would mainly be specula-
i e in na u e and ep esen na a i es. Howe e , as al eady men ioned abo e, Saudi
A abia s ands ou as ading pa ne o China. I is he s onges coupled wi h Chinese
cyclical g ow h in he 8–10 yea s band. See he bolded dashed-do ed g aph in Fig. 4.
This clea ly speaks in a o o i s ole as cen al Chinese oil supplie in he pe iod
o analysis. A ole ha is pa icula ly ele an in he cyclical g ow h (8–10 yea s),
neighbo ing he secula g ow h (
>10
yea s), equency band. The opposi e applies o
India. See he magen a colo ed dashed g aph in Fig. 3. I sugges India o couple wi h
he Chinese economy mos ly a high business cycle equencies coined by, among
o he s, dynamics in logis ics, sha ed business p ac ices, IT, and enginee ing.
In echnical e ms, he e seems o be o e all enough a ia ion in bila e al IIP-co -
ela ions ac oss economies and equency bands o jus i y a mo e sys ema ic in e en-
ial analysis applying he BSPR echniques p oposed in Sec . 2. Such an explo a i e
analysis would go beyond me e na a i es and specula ions. I unde sco es ou poin
(ii) se ou in he in oduc o y sec ion.
3.2 Cons uc ion o explana o y indica o a iables
Ou i s cen al block o explana o ies, con ained in
x′
jb
in ep esen a ion (9) o he
BSPR model ou lined in Sec . 2, a e ade linkages. As a gued abo e, hey can be o
wo gene al ypes: in e -indus y and in a-indus y ade linkages.
The in e -indus y ade linkage indica o s ha we conside a e qui e s anda d and
equen ly used in he empi ical li e a u e (Nachane & Dubey, 2013). We de ine hem
as a e age bila e al expo in ensi y (ExI), impo in ensi y (ImI ), and o al ade
in ensi y o ela i e openness (T I) wi h ou e e ence economy China
In e -indus y
ExIjb =1
N
N
∑
(
ExC,jb
ExCb +Exjb
)
(17)
In e -indus y Im
Ijb =1
N
N
∑
(
ImC,jb
ImCb +Imjb
)
(18)
1 3
1436
Global linkages ac oss sec o s and equency bands: a band spec al…
In e -indus y
T Ijb =1
N
N
∑
(
T C,jb
T Cb +T jb
)
=1
N
N
∑
ExC,jb +ImC,jb
(ExCb +ImCb )+(Exjb +Imjb )
,
(19)
whe e subsc ip b e e s o he b- h equency band de ia ion cycle componen , C
deno es China, and
j=1,...,J
i s majo ading pa ne s.
ExC,j
,
ImC,j
, and
T C,j
deno e o al nominal expo s om China o j, o al nominal impo s om j o China,
and o al ade (in nominal e ms) be ween China and coun y j, espec i ely. All
unde lying Chinese se ies a e a ailable in mon hly equency and can be agg ega ed
o lowe esolu ion obse a ion- equency o ob ain co esponding de ia ion cycle
componen s and allowing o band-speci ic a e ages also in he mixed equency
case.
Fo in a-indus y ade linkages (TL) ou baseline bila e al measu e is he obse -
a ion pe iod a e age o he also equen ly used indica o by G ubel and Lloyd
(1975):
In a-indus y
TLjb =1
N
N
∑
(
1−
∑
s|ExC,js −ImC,js |
∑
s
|
ExC,js +ImC,js
|),
(20)
whe e
s=1,...,S
deno es conside ed sec o s o commodi ies. I is bound o he
(0,1) in e al. Bo h expo and impo quan i ies a e agg ega ed o e indus ies (see
Sec . e BSPRspssubsec ionspslinkages o de ail). We also cons uc and use he
indica o as an indus y-speci ic measu e, which is gi en o each sec o s by
In a-indus y
TLjbs =1
N
N
∑
(
1−|ExC,js −ImC,js |
|
ExC,js +ImC,js
|).
(21)
How indus ies s a e exac ly de ined and which sec o al quan i ies we ac ually use in
he BSPR es ima es is de ailed in Sec . 3.3.
Simila o ou bila e al dependen a iable, IIP-co ela ion, he emaining explana-
o ies ep esen bila e al co ela ions. We conside wo measu es o inancial in eg a-
ion. Fi s , bila e al co ela ions o some measu e o inancial openness, ha is, o
M2/GDP a ios, ac oss espec i e band-speci ic de ia ion cycle componen s b o he
M2/GDP se ies o coun y j wi h i s analogue Chinese componen -se ies,5 We e e o
his measu e as Financial In eg a ion (FI) I. Secondly, we conside bila e al co ela-
ions o espec i e s ock p ice index se ies ha we dissec in o equency band com-
ponen se ies be o e band-wisely compu ing co ela ions. No ably, Chinese inancial
5 As M2 is a ela i ely liquid mone a y agg ega e including sho - e m asse s held by non-banks, i is a
equen ly used p oxy in his con ex ((Nachane & Dubey, 2018) p. 11). Howe e , i should be no ed ha
his choice comes wi h a ca ea . While o he main pa o ou pe iod o analysis China implemen ed an
M2 a ge , a good chunk o i s majo ading pa ne s did no . This is p oblema ic i he Chinese cen al
bank eac s o changes in a o eign expo posi ion.
1 3
1437
J. Lyu, B. Süssmu h
ma ke s a e ela i ely closed o . Howe e , as ou measu e is bila e al, his conce ns
e e y coun y pai wi h China o ou sample alike. A summa y o unde lying s ock
ma ke index se ies and co esponding sou ces is gi en in Table 10 in he Appendix.
Hence o h, we e e o his indica o as FI II.
Analogously, we cons uc bila e al measu es o mone a y policy coo dina ion
based on pai -wise and band-speci ic M2 g ow h a e co ela ions and o iscal pol-
icy coo dina ion based on pai -wise and band-speci ic public de ici , i.e. go e nmen
su plus o GDP a io, g ow h a e co ela ions.
Finally, we compu e a band-speci ic indica o o in la ion co-mo emen using
bila e al co ela ions o de ia ion cycle componen s o espec i e CPI se ies (le els).
3.3 The sec o al dimension o in a-indus y linkages
As a gued a he beginning o Sec . 3, ac o in ensi y as a de ining p ope y o sec o s
is assessed c ucial o in a-indus y ade linkages in p edic ing in e na ional busi-
ness cycle co-mo emen a di e en equencies in he li e a u e. We, hus, conside
sec o s classi ied by ac o in ensi y also in ou BSPR analysis. Ou p ima y da ase
o he espec i e se ies is he Gene al Adminis a ion o Cus oms o he P.R. China
(GACC, hence o h ‘China Cus oms’), om which we e ie e a o al o 968 mon hly
se ies. Hal o hese e e o impo quan i ies, he o he hal o expo quan i ies,
espec i ely. They a e compiled a he wo-digi China Ha monized Commodi y
Desc ip ion and Coding Sys em (HS) code le el comp ising 22 commodi ies o 22
ou o ou 25 conside ed majo ading pa ne economies o China.6 The ac o in en-
si y classi ica ion is done on he basis o he UNCTAD/WTO In e na ional T ade
Cen e classi ica ion using he s anda d indus ial ade classi ica ion (SITC) ‘ e . 2
codes’ and dis inguishing i e main g oups o sec o s a he h ee-digi le el ollow-
ing he scheme o he Empi ical T ade Analysis Cen e (ETA) o E asmus Uni e si y
Ro e dam. A summa y o he applied scheme is gi en in Table 2.
Besides conside ing (21) o he abo e lis ed i e sec o s by ac o in ensi y as
explana o y, we also compu e (20) o all
S= 22
in China Cus oms a ailable and in
he sub ac i e pa o (20) agg ega ed commodi ies and espec i e expo and impo
quan i ies. We e e o i as o e all in a-indus y linkages in ou BSPR es ima es.
6 Due o da a issues and p oblems, which can ha dly be aken ca e o o co ec ed o , his led us o abs ac
om se ies om Taiwan, Hong Kong, and he EU.
Fac o in ensi y classi ica ion (ETA p oduc
g oup)
Commodi y HS
indus y codes
P ima y p oduc s (A) I, II, II, IV
Na u al- esou e in ensi e p oduc s (B) V, VIII, IX, XV,
XIV
Unskilled-labo in ensi e p oduc s (C) XI, XII, XIII, XX
Technology in ensi e p oduc s (D) VI, XVI, XVIII,
XIX, XVII
Human-capi al in ensi e p oduc s (E) X, XXI, XXI, XXII
Table 2 Sec o s calssi ied by
ac o in ensi y o sec o al
in a-indus y TL compu a ion
Fo de ail on HS classi ica ion
indus y code and sou ces see
Table 10 in he Appendix
1 3
1438
Global linkages ac oss sec o s and equency bands: a band spec al…
3.4 BSPR es ima es and in e p e a ion
The columns o Table 3 labeled Model I, II, and III e lec how we p oceed in iden i-
ying he de e minan s o bila e al global linkages o ou e e ence economy China.
The h ee conside ed co e models migh be ep esen ed in an ex ension o baseline
BSPR model (9). I eads
y
jb
=
αb
+
x
′
jb
β
+
x
′
jbs
βs
+
z
′
jb
βb
+
z
′
jbs
βbs
+
νj
+
ε
jb
(22)
wi h
E(εjb|αb,ν
j,x,z)=0
, whe e
x={xjb;xjbs}∧z={zjb;zjbs}
wi h
s=1,...,S
deno ing sec o s classi ied by ac o in ensi y, we es ima e–besides
band- ixed and coun y- ixed e ec s,
αb
and
νj
– gene al
β
e ec s, e ec s e e ing
o quan i ies o pa icula sec o s
βs
, o band-speci ic e ec s
βb
o ce ain indica o s,
and o band-speci ic e ec s e e ing o a iables o pa icula sec o s
βbs
, espec-
i ely. As he se o
z
a iables is echnically gene a ed by band-(speci ic-)in e ac ion
e ms, lea ing ou as e e ence band he 2–4 yea s pe iodici y in e al,
z⊋x
, i.e.,
x
is a eal sub-se o
z
. A ully in e ac ed model allowing o
z⊇x
would boil down o
single-equa ion es ima ions.
In gene al, magni udes and signs o poin es ima es o ou BSPR speci ica ions
can be in e p e ed in analogy o and wi h simila sho comings as linea p obabili y
model (LPM) speci ica ions applied o panel da a. Simila o an LPM, whe e he
dependen is bound o he [0, 1] in e al, ou dependen a iable is a pai wise co -
ela ion coe icien bound o
[−1,+1]
. Fo ins ance, a band e ec in he 4–6 yea s
ange o −.280 as opposed o a co esponding one in he 6–8 yea s ange o −.691 can
be ead as—all else equal—a 41.1 pe cen age poin s less lowe ing associa ion o he
de aul co ela ion coe icien in he highe equency band (p o ided ha he de aul
coupling o cons an is posi i e).
Howe e , simila o LPM in e p e a ions, such eading has o be aken wi h a
pinch o sal . As a gene al ca ea he e is no model ea u e ha p e en s he linea
models o p edic alues o he IIP co ela ions ha do no all in o he
[−1,+1]
in e al. Mo eo e , i has o be kep in mind ha he e e ence band h oughou is
he 2–4 yea s band. This implici ly implies ha he > 10 yea s band is ega ded as a
il e ed low equency o de e minis ic end componen ha gene ally plays no ole
o business cycle equencies.
In Table 3, beginning wi h he second column down o line ‘T ade In ensi y’ gen-
e al
β
e ec s a e gi en. The ollowing ‘In a-Indus y’ block displays es ima ed e ec
sizes e e ing o sec o al quan i ies
βs
. The p oceeding ow labeled ‘Band E ec s’
oge he wi h he op ow (‘Cons an ’) depic band- ixed e ec s, ollowed by he
band-indus y e ec s
βbs
and inally some band-speci ic e ec s
βb
( ‘Band-In e -
Indus y,’ ‘Band-IC,’ and ‘Band-FI II’) ha ha e o be in e p e ed in conjunc ion wi h
i s e e ence
β
e ec s a he op o Table 3. The model ypes I o III a y wi h consid-
e ed band-speci ic e ec s
βb
: While Model I conside s in a-indus y ade linkages
as sole band-speci ic de e minan s, Model II addi ionally conside s in e -indus y
measu es o bila e al expo s, impo s, and ade band-wisely. Finally, Model III on
op o his speci ies band-speci ic in la ion co-mo emen and band-speci ic inancial
1 3
1439
J. Lyu, B. Süssmu h
in eg a ion measu ed by indica o FI II, which come ou ai ly sizable and clea ly
signi ican ly di e en om ze o in ou es ima es.
In con as , bo h mac o-policy acco dance indica o s o mone a y policy (MP)
and iscal policy (FP)–as well as bila e al M2/GDP a io co ela es cap u ed by FI I–
a e no es ima ed as s a is ically di e en om ze o i in e ac ed wi h equency band
iden i ie s in ex ensions o ou models (no shown in Table 3). Ne e heless, we keep
hem in he
xjb
-pa o he empi ical model as, a leas , non-in e ac ed FP is es ima ed
wi h a nega i e coe icien (−.152) s a is ically signi ican a a en pe cen le el in
model III.The nega i e ela ionship migh be a ionalized by he ac ha he syn-
ch onized c ea ion o public de ici s does no necessa ily embody he in o ma ion o
wha he espec i e economies used hese means. Using de ici s o consump i e o
in es i e go e nmen al spendings, deb se ice o combina ions o i can ha e qui e
idiosync a ic e ec s on ou pu dynamics. This led us o abs ac om equency-
band-in e ac ed e sions o hese eg esso s in he
z′
jb
βb-pa s ensu ing
z⊋x
.
In e ms o in o ma ion c i e ia and o he alues o i o da a, Model III seems an
adequa e choice.7 Howe e , i is wo h o assess i also agains he o he wo speci-
ica ions. This becomes e iden when looking a he es ima ed coe icien alues o
he gene al
β
e ec s block in he op ows o Table 3. As men ioned abo e, policy
coo dina ion indica o s o ei he mone a y o iscal na u e a e nea ly h oughou no
signi ican ly associa ed wi h bila e al global linkages ac oss speci ica ions.
In la ion co-mo emen (IC) seems nega i ely ela ed o global linkages; see, a
i s , he co esponding coe icien es ima es in he columns e e ing o Model I and
II, espec i ely. As speci ica ion III shows in he penul ima e ow o coe icien es i-
ma es, his seems o ha e i s o igin a low equencies as bo h he es ima ed BSPR
coe icien o IC wi hou band-in e ac ion ( e e ing o he 2–4 yea s band) amoun -
ing o 0.202 and he one e e ing o he 4–6 yea s pe iodici y (−.054) a e insigni i-
can , while he e a e indica ions o a signi ican nega i e associa ion in he 6–8 yea s
and 8–10 yea s band, espec i ely. Acco ding o Wang and Wen (2007) and Cicca-
elli and Mojon (2010), i is an empi ical ac ha global co-mo emen in in la ion
is highe han he one in cyclical ou pu dynamics. I is a ionalized in a a ie y o
New Keynesian wo-coun y models by Wang and Wen (2007). The la e s udy also
jus i ies IC as exogenous o , a leas , no as endogenous. I shows ha , a leas , in he
con ex o New Keynesian open-economy models, in e na ional spill-o e s a e no
he o igin o IC. In ou esul s, hough only o highe pe iodici y dynamics, he s yl-
ized ac ha global in la ion co-mo emen is highe han cyclical ou pu dynamics is
cap u ed by he signi ican nega i e coe icien s o IC in he 6–8 yea s (−.344) and
8–10 yea s (−.628) ange in he column o Table 3 e e ing o Model III.
Speci ica ion III is also highly ins uc i e in explaining he inancial globaliza-
ion– eal egionaliza ion s. eal con agion puzzle. As we ha e a gued abo e, wha
he deba e o he puzzle so a igno es is ha he RBC iewpoin (FI p omo ing
7 In gene al, he o e all R-squa es calcula ed as squa ed co ela ions be ween p edic ed and ac ual depen-
den alues a e easonably high lying be ween abou 30 and app oxima ely 40 pe cen . As all h ee speci-
ica ions ep esen ixed-e ec s eg essions maximizing wi hin-R-squa es, co esponding alues by a
ou numbe he espec i e R-squa ed be ween alues.
1 3
1440

Global linkages ac oss sec o s and equency bands: a band spec al…
Model I Model II Model III
Cons an 1.052*** 1.019*** −0.158
(0.175) (0.195) (0.242)
Mone a y Policy 0.008 −0.342 −0.037
(0.429) (0.416) (0.369)
In la ion co-mo emen (IC) −0.514** −0.556*** 0.202
(0.180) (0.194) (0.234)
Fiscal Policy 0.009 0.163 −0.152*
(0.130) (0.161) (0.074)
Financial In (FI) I −0.111 −0.376 −0.077
(0.171) (0.272) (0.212)
Financial In (FI) II −0.548*** −0.431*** 1.171***
(0.140) (0.103) (0.256)
In e -Indus y
Exp In ensi y (ExI) −0.138** −0.175 −0.924
(0.063) (0.111) (0.384)
Imp In ensi y (ImI) 0.234** 0.275* 0.051
(0.110) (0.137) (0.059)
T ade In ensi y (T I) −0.230 −0.674* −2.356***
(0.726) (0.834) (0.699)
In a-Indus y
O e all 1.677 1.663 −2.053
(1.759) (2.235) (1.387)
P ima y 0.309 0.501 0.239
(0.557) (0.400) (0.372)
Na u al Resou ces −0.895 −0.549 0.374
(0.705) (0.738) (0.476)
Unskilled −0.381* −0.463* −0.165
(0.199) (0.235) (0.186)
Table 3 Band spec al panel eg ession models: es ima es
1 3
1441
J. Lyu, B. Süssmu h
Model I Model II Model III
Technology 0.149 −0.211 1.139**
(0.846) (0.631) (0.522)
Human Capi al 0.449 0.351 −0.143
(0.336) (0.251) (0.140)
Band E ec s
Band 4–6 yea s 6–8 yea s 8–10 yea s 4–6 yea s 6–8 yea s 8–10 yea s 4–6 yea s 6–8 yea s 8–10 yea s
−0.254** −0.744*** −0.477*** −0.280** −0.691*** −0.679*** 0.427 0.347 0.761***
(0.096) (0.164) (0.120) (0.124) (0.148) (0.149) (0.254) (0.229) (0.198)
Band-In a-Indus y
O e all −1.130 −1.765 −5.778* −1.761 −3.061 −5.219* 1.262 0.464 −3.539*
(1.484) (2.747) (2.848) (1.690) (2.266) (2.556) (1.233) (1.687) (1.720)
P ima y −0.691 −0.332 0.022 −0.928*** −0.497 −0.347 −0.291 −0.007 −0.793
(0.474) (0.609) (0.684) (0.299) (0.350) (0.573) (0.359) (0.347) (0.463)
Na u al Resou ces 1.113 1.156 1.899* 0.273 0.710 0.784 −1.138* −0.787 0.413
(0.929) (0.898) (0.992) (0.667) (0.747) (1.007) (0.608) (0.558) (0.634)
Unskilled 0.711** −0.065 1.573* 0.664*** −0.251 0.407 −0.197 −0.287 2.515***
(0.297) (0.481) (0.852) (0.213) (0.476) (1.023) (0.253) (0.509) (0.776)
Technology 0.054 −0.256 1.533 0.909 0.730 1.248 −0.715 −0.995 1.272*
(0.851) (1.587) (1.289) (0.825) (1.231) (0.991) (0.570) (0.798) (0.749)
Human Capi al −0.425 −0.516 −0.062 −0.343 −0.407 −0.013 0.425* 0.256 0.586***
(0.341) (0.360) (0.291) (0.359) (0.284) (0.299) (0.238) (0.210) (0.177)
Band-In e -Indus y
Exp In ensi y (ExI) 0.087 −0.405 0.125 0.004 −0.947*** 0.319**
(0.122) (0.435) (0.179) (0.124) (0.296) (0.128)
Imp In ensi y (ImI) −0.193 −0.872 0.137 0.078 0.127 0.016
(0.182) (0.683) (0.682) (0.099) (0.4317) (0.348)
T ade In ensi y (T I) 1.556 4.107* 5.627** 1.673* 4.909*** 0.768
Table 3 (con inued)
1 3
1442
Global linkages ac oss sec o s and equency bands: a band spec al…
Model I Model II Model III
(1.004) (2.837) (2.120) (0.949) (1.422) (1.183)
Band-IC −0.054 −0.344*** −0.628***
(0.104) (0.113) (0.116)
Band-FI II −0.924** −1.275*** −1.585***
(0.384) (0.242) (0.219)
N obs 88 88 88
Log L 67.811 90.033 136.394
AIC −93.621 −138.067 −230.786
BIC −41.597 −86.042 −178.763
R-squa ed Wi hin Be ween O e all Wi hin Be ween O e all Wi hin Be ween O e all
0.8250 0.0002 0.2999 0.8940 0.0002 0.2997 0.9632 0.0358 0.3921
Numbe o coun ies 22 22 22
Fixed-e ec s (wi hin) eg ession; g oup a iable: coun y; obus s anda d e o s in pa en heses; ***
p<0.01
, **
p<0.05
, *
p<0.1
Table 3 (con inued)
1 3
1443
J. Lyu, B. Süssmu h
(1) (2) (3) (4)
Base
Model
Model I Model II Model III
Cons an 0.0788 0.156 0.233 0.316
(0.0785) (0.302) (0.337) (0.194)
Mone a y Policy 0.0507 −0.107 −0.0741 −0.125
(0.0704) (0.0849) (0.100) (0.111)
In la ion co-
mo emen (IC)
0.0480 −0.0172 0.0448 −0.0376
(0.0396) (0.0734) (0.0814) (0.256)
Fiscal Policy 0.0311 0.0240 0.0493 0.00736
(0.0330) (0.0376) (0.0417) (0.0587)
Financial In
(FI) I
−0.0159 −0.000326 0.000812 0.00144
(0.0284) (0.0266) (0.0250) (0.0253)
Financial In
(FI) II
0.0332 −0.0108 −0.00506 −1.095***
(0.0380) (0.0374) (0.0531) (0.392)
In e -Indus y
Exp In ensi y
(ExI)
−8.600 −12.08*** 68.89 36.85
(8.509) (4.239) (129.8) (93.39)
Imp In ensi y
(ImI)
−6.528 −10.21*** 50.50 24.88
(7.420) (3.839) (111.0) (79.61)
T ade In ens
(T I)
16.49 22.36*** −121.4 −61.93
(15.85) (7.603) (242.0) (172.9)
In a-Indus y
O e all −0.398 0.714 1.240 0.459
(0.534) (1.843) (1.916) (1.831)
P ima y −0.0164 −0.364 −0.685 −0.552
(0.126) (0.603) (0.861) (0.663)
Na u al
Resou ces
0.194 −0.149 −0.380 −0.232
(0.186) (0.684) (0.696) (0.634)
Unskilled 0.0378 0.443 0.701* 0.310
(0.134) (0.286) (0.401) (0.420)
Technology 0.253 −0.737 −1.168 −0.952
(0.346) (0.939) (0.956) (0.932)
Human Capi al 0.000113 0.0138* 0.0250* 0.0151
(0.0228) (0.00777) (0.0143) (0.0105)
Obse a ions 420 420 420 420
R-squa ed 0.0212 0.577 0.643 0.687
Numbe o
coun ies
21 21 21 21
Table 4 Time domain com-
pa ison I: s anda d s a ic panel
models
Random-e ec s eg ession;
coun y obus s anda d e o s;
***
p<0.01
, **
p<0.05
,
*
p<0.1
1 3
1444
Global linkages ac oss sec o s and equency bands: a band spec al…
Model I Model II Model III
Cons an 0.822*** 0.833*** −0.158
(0.141) (0.156) (0.196)
Mone a y Policy −0.352 −0.245 −0.809**
(0.355) (0.348) (0.364)
In la ion co-mo emen (IC) −0.247 −0.318 0.267
(0.208) (0.232) (0.223)
Fiscal Policy 0.022 0.199 −0.0998
(0.102) (0.154) (0.0991)
Financial In (FI) I 0.144 −0.199 0.0554
(0.134) (0.262) (0.177)
Financial In (FI) II −0.536*** −0.413*** 0.944***
(0.113) (0.0866) (0.204)
In e -Indus y
Exp In ensi y (ExI) −0.116*** −0.100 −0.00378
(0.0286) (0.0997) (0.0620)
Imp In ensi y (ImI) 0.0984 0.132 −0.172***
(0.102) (0.105) (0.0592)
T ade In ensi y (T I) 0.0131 −0.0142 −0.563
(0.646) (0.838) (0.464)
In a-Indus y
O e all 2.226* 0.820 −2.228
(1.094) (1.319) (1.335)
P ima y −0.319 −0.479 −0.500*
(0.509) (0.408) (0.284)
Na u al Resou ces −0.849* −0.438 0.396
(0.492) (0.536) (0.395)
Unskilled −0.281 −0.168 0.0626
(0.184) (0.252) (0.231)
Table 7 Robus ness check II: a e aged Jan and Feb obse a ions be o e X-12-SA o China
1 3
1451

J. Lyu, B. Süssmu h
Model I Model II Model III
Technology −0.628 −0.287 0.911
(0.608) (0.692) (0.576)
Human Capi al 0.194 0.184 −0.265*
(0.331) (0.235) (0.134)
Band E ec s
Band 4–6 yea s 6–8 yea s 8–10 yea s 4–6 yea s 6–8 yea s 8–10 yea s 4–6 yea s 6–8 yea s 8–10 yea s
−0.284*** −0.621*** −0.406*** −0.371*** −0.667*** −0.601*** −0.0743 0.216 0.676***
(0.0793) (0.103) (0.0981) (0.109) (0.136) (0.113) (0.228) (0.179) (0.200)
Band-In a-Indus y
O e all −0.758 −1.389 −3.285 0.186 −0.767 −1.942 3.325*** 2.236 −0.961
(1.284) (2.098) (2.103) (1.265) (1.714) (1.890) (0.906) (1.549) (1.323)
P ima y −0.281 0.359 0.674 −0.199 0.542 0.976* 0.007 0.785** 0.397
(0.504) (0.539) (0.563) (0.366) (0.359) (0.489) (0.343) (0.292) (0.308)
Na u al Resou ces 0.520 0.791 1.700* −0.114 0.373 0.894 −1.187*** −1.045** 0.614
(0.884) (0.771) (0.948) (0.800) (0.679) (1.045) (0.419) (0.460) (0.696)
Unskilled 0.240 −0.104 0.957 −0.007 −0.524 −0.220 −0.618** −0.382 1.958*
(0.294) (0.410) (1.055) (0.238) (0.565) (1.424) (0.220) (0.589) (0.973)
Technology −0.069 −0.094 0.174 −0.357 −0.318 −0.782 −2.085*** −1.875* −0.559
(0.779) (1.204) (1.020) (0.820) (1.068) (0.954) (0.637) (0.966) (0.683)
Human Capi al −0.287 −0.343 −0.0331 −0.337 −0.421 −0.0615 0.386** 0.178 0.523***
(0.327) (0.330) (0.309) (0.296) (0.287) (0.248) (0.144) (0.171) (0.140)
Band-In e -Indus y
Exp In ensi y (ExI) 0.066 −0.457 −0.085 0.134 −1.044** 0.077
(0.108) (0.510) (0.171) (0.0974) (0.382) (0.102)
Imp In ensi y (ImI) −0.067 −0.753 −0.197 0.240* 0.0288 −0.269
(0.139) (0.626) (0.494) (0.115) (0.413) (0.335)
T ade In ens (T I) −1.331 2.540 4.166** −1.083 3.391** −0.307
Table 7 (con inued)
1 3
1452
Global linkages ac oss sec o s and equency bands: a band spec al…
Model I Model II Model III
(1.005) (2.578) (1.821) (0.980) (1.481) (1.337)
Band-IC 0.007 −0.238** −0.531***
(0.113) (0.0990) (0.0980)
Band-FI II −0.201 −1.094*** −1.370***
(0.351) (0.223) (0.216)
N obs 88 88 88
Log L 77.715 96.297 140.213
AIC −113.429 −150.593 −238.426
BIC −61.405 −98.569 −186.402
R-squa ed Wi hin Be ween O e all Wi hin Be ween O e all Wi hin Be ween O e all
0.8373 0.0101 0.4668 0.8933 0.0000 0.4085 0.9607 0.0044 0.5080
Numbe o coun ies 22 22 22
Fixed-e ec s (wi hin) eg ession; g oup a iable: coun y; obus s anda d e o s in pa en heses; ***p < 0.01, **p < 0.05, *p < 0.1
Table 7 (con inued)
1 3
1453
J. Lyu, B. Süssmu h
lexible in allowing o band-speci ic ixed e ec s and band-indus y-speci ic in e ac-
ion e ms, which bo h a e o high ele ance in he s udy o global linkages. Techni-
cally, BSPR models can con ol o unobse ed he e ogenei y ac oss c oss-sec ional
en i ies and equency bands. They ha e he po en ial o emo e omi ed a iable
bias p oblems i omi ed eg esso s a e c oss-sec ional and equency band in a ian .
In ou BSPR applica ion on bila e al ou pu co-mo emen o he Chinese economy
wi h i s majo ading pa ne economies, we ind e idence o he associa ion o
bo h in e -indus y ade in ensi ies and in a-indus y ade linkages wi h in e na-
ional ou pu co-mo emen . In pa icula , we ind suppo o Hecksche -Ohlin in e -
indus y decoupling due o specializa ion– ha can be a ionalized wi h compa a i e
ad an age a gumen s– o op-high as well as o op-low equency bands. E idence
o New T ade Theo y in a-indus y “ex e nal economies o scale” is ound o (a)
ela i ely low equencies and (b) sec o s wi h high labo in ensi y. Spillo e s in
echnology in ensi e indus ies a e no equency-speci ic in he sho and medium
un. I a all, hey a e only in ensi ied in he cyclical g ow h ela ed equency band
o 8–10 yea s pe iodici ies.
Fu he mo e, we ind no con incing e idence o policy coo dina ion indica o s
o ei he mone a y o iscal na u e o be associa ed wi h bila e al global linkages. We
also con i m he well-documen ed and heo e ically a ionalized empi ical ac ha
global in la ion co-mo emen is highe han co-mo emen in cyclical ou pu dynam-
ics by es ima ing a nega i e associa ion be ween bila e al ou pu linkages and in la-
ion co-mo emen . Howe e , we ind i o be only signi ican in he ela i ely low
equency bands.
Finally, ou BSPR es ima es a e mos help ul in explaining he inancial globaliza-
ion– eal egionaliza ion s. eal con agion puzzle. Agains he backd op o ou ind-
ings, bo h he RBC iew, which sees inancial in eg a ion os e ing egionaliza ion o
ou pu dynamics, and he inancial con agion pe spec i e, which sugges s he oppo-
si e (i.e. os e ing synch oniza ion), can be jus i ied a di e en equency bands. Ou
BSPR es ima es ind indica ions o he i s in low equency bands and indica ions o
he la e in he highes equency band. We a ibu e his o he RBC a ionaliza ion
being mo e symme ic, less ansi o y, and es ing a he on p opaga ed han di ec
impac s o shocks. The opposi e applies o models o inancial con agion.
1 3
1454
Global linkages ac oss sec o s and equency bands: a band spec al…
The con empo a y poli ical economy discussion su ounds decoupling, geo-eco-
nomic agmen a ion, a e e sed o slowe globaliza ion (“deglobaliza ion” o “slow-
baliza ion”) implied by a pos -pandemic e-newed Eas –Wes di ide and coined by
decele a ing ade and in es men as well as smalle global alue chains. The slow-
baliza ion phenomeon has been accele a ed due o he US push o con ain Chiona
wi hin he con ex o s a egic compe i ion du ing he incumben pe iod o he 45 h
p esiden o he US. The ac ha he la e is also he 47 h p esiden o he US high-
ligh s he impo ance o p esidencies, poli ical and pa isan cycles as well as planning
and ime- o-build ho izons. All o hese dynamic no ions can be seen as ul ima ely
ep esen ed by he e ogenous equency bands making BSPR a cohesi e amewo k
o adequa ely and explo a i ely s udy globaliza ion-slowbaliza ion phenomena.
Fo u u e wo k, we see a wide a ay o applica ions o he p oposed BSPR
me hodology wi hin he ealms o possibili y. I comp ises de e minan s o poli ical
and/o pa isan business cycles and hei synch oniza ion, e.g. ac oss ede al s a es,
ela i ely low equency-con ingen inancial cycles and hei co-mo emen , e.g.
ac oss di e en comodi ies and ma ke s, as well as cycles in capi al o ma ion, labo
demand o mig a ion lows and hei synch onici y a he egional o in e na ional
le el. Howe e , i also emains a u u e ask o ho oughly s udy and analyze he
s a is ical p ope ies and assump ions o he BSPR model.
Appendix
See Figs. 5, 6 and Tables 8, 9, 10.
1 3
1455
J. Lyu, B. Süssmu h
Fig. 5 Sample band componen s I: qua e ly M2 se ies o Ge many
1 3
1456

Global linkages ac oss sec o s and equency bands: a band spec al…
Fig. 6 Sample band componen s II: Annual GDP se ies o Saudi A abia
1 3
1457
J. Lyu, B. Süssmu h
Table 8 Selec ion o conside ed 22 majo ading pa ne economies o China
1 Aus alia 6 Ge many 11 Malaysia 16 Singapo e 21 Uni ed Kingdom (UK)
2 Belgium 7 India 12 Ne he lands 17 Sou h A ica 22 Uni ed S a es (US)
3 B azil 8 Indonesia 13 Philippines 18 Sou h Ko ea
4 Canada 9 I aly 14 Russia 19 Spain
5 F ance 10 Japan 15 Saudi A abia 20 Thailand
In alphabe ical o de ; e e ence economy ( h oughou ) is China. The selec ion o iginally emana ed om
he op 25 majo ading pa ne economies o China du ing he pe iod o obse a ion. I excludes Hong
Kong, Taiwan, and he EU en i y
Fac o in ensi y classi ica-
ion (ETA g oup)
China cus oms: commodi y (HS
code)
P ima y p oduc s (A) Li e animals / animal p oduc s (I)
Vege able p oduc s (II)
Animal / ege able oils and a s (III)
Food, be e ages, and obacco (IV)
Na u al- esou e in ensi e
p oduc s (B)
Mine al p oduc s (V)
Lea he and ela ed p oduc s (VIII)
Wood, cha coal and ela ed p od-
uc s (IX)
Base me al and ela ed p oduc s
(XV)
Pea ls and (semi-)p ecious s ones
(XIV)
Unskilled-labo in ensi e
p oduc s (C)
Tex iles and ex ile a icles (XI)
Foo wea , head gea and ela ed
p oduc s (XII)
S one, ce amics, and glass (XIII)
Miscellanous manu ac u ed a icles
(XX)
Technology in ensi e
p oduc s (D)
Chemicals and allied indus ies (VI)
Rubbe s and plas ics (VII)
Machine y, elec ical/elec onic
equipmen (XVI)
P ecision/musical ins umen s and
clocks (XVIII)
A ms and ammuni ion (XIX)
Vehicles, ai c a , and anspo a ion
(XVII)
Human-capi al in ensi e
p oduc s (E)
Pulp, pape , and ela ed p oduc s
(X)
A wo k and an iques (XXI)
A icles o special ade (XXII)
Table 9 Sec o s calssi ied by
ac o in ensi y o sec o al
in a-indus y TL compu a ion
HS classi ica ion indus y code
as p o ided by Expo - o-
China (ETCN, China Cus oms)
Fac o in ensi y classi ica ion
(FIC) om Empi ical T ade
Analysis Cen e (ETA)
Sou ce: China Cus oms
(se ies); FIC-ETA: w w w 2 . e c o
n . u u . n l / u s e s / m a e w i j k / e a / i n
e n s i y . h m
1 3
1458
Global linkages ac oss sec o s and equency bands: a band spec al…
Acknowledgemen s We would like o hank he Edi o , Galina B. Hale, as well as an anonymous e e ee
o many help ul commen s and sugges ions ha ma kedly imp o ed ou pape . We a e also g a e ul o
Bas ian Gawellek and Ma co Sunde o hei excellen esea ch assis ance, o A gimi o A a ia, Jö g B ei-
ung, Maximilian Jage , Helmu Lü kepohl, and G ego on Schweini z o many aluable commen s and
ema ks as well as pa icipan s o he ITISE In e na ional Wo k-Con e ence on Time Se ies Analysis a he
Uni e si y o G anada, he IWH-CIREQ-GW Mac oeconome ic Wo kshop on Mixed F equency Da a in
Mac oeconomics and Finance, he Wo kshop on Mac oeconomic Resea ch a UEK C acow Uni e si y o
Table 10 S ock ma ke index se ies by coun y: codes and sou ces
Coun y Sou ce Code Index name
Aus alia Rese e Bank o
Aus alia
AUSHRPRCF S&P/ASX 200
Belgium Eu onex B ussels BGSHRPRCF BXS, B ussels S ock Exchange Cash
Ma ke Re u n Index
B azil Reu e s BRSHRPRCF The Bo espa Index(Indice Bo espa)
Canada Reu e s CNSHRPRCF S&P/TSX, To on o S ock Exchange
Composi e Sha e P ice Index
China Na ional Bu eau o
S a is ics o China
CHSHRPRCF Shanghai S ock Exchange Composi e
Index China
EU Thomson Reu e s EMSHRPRCF Da as eam EURO Sha e P ice Index
(Eu o Zone)
F ance Main Economic
Indica o s
FRSHRPRCF SBF250
Ge many Reu e s BDSHRPRCF Deu sche Boe se, DAX 30
Hongkong Census and S a is ics
Depa men , Hong
Kong
HKSHRPRCF Hong Kong Heng Seng Sha e P ice
Index
India Cen al S a is ical
O ganisa ion, India
INSHRPRCF Bombay S ock Exchange Na ional 100
Sha e P ice Index
Indonesia Reu e s IDSHRPRCF Jaka a S ock Exchange Index (JSX)
I aly Bo sa I aliana ITSHRPRCF Milan COMIT Gene al Sha e P ice Index
Japan Reu e s JPSHRPRCF Tokyo SE, TOPIX Index
Malaysia Reu e s MYSHRPRCF Financial Times S ock Exchange Bu sa
Malaysia KLCI
Ne he lands S a is ics Ne he lands NLSHRPRCF Ams e dam SE All Sha e S ock P ice
Index
Philippines Cen al Bank
Philippines
PSECOMP PSEI Index, de i ed om daily se ies
by me
Russia Reu e s RSSHRPRCF MICEX Sha e P ice Index
Saudi A abia Saudi A abian Mon-
e a y Agency
SISHRPRCF Tadawul All Sha e Index (TASI)
Singapo e Thomson Reu e s SPSTDSCAF Singapo e STRAITS T.DS, Da as eam
Sou h A ica Da as eam SASHRPRCF To al S ock Ma ke S ock P ice Index
Sou h Ko ea Reu e s KOSHRPRCF Ko ea Composi e S ock P ice Index
(KOSPI)
Spain Minis y o he Econo-
my and Finance,
Spain
ESSHRPRCF Mad id SE Gene al Index
Taiwan Reu e s TWSHRPRCF TSE Capi aliza ion Weighed S ock Index
(TAIEX)
Thailand Reu e s THSHRPRCF SET Index, Bangkok SE P ice Index
UK Reu e s UKSHRPRCF Financial Times All Sha e Index
USA Reu e s USSHRPRCF Dow Jones Indus ial Sha e P ice Index
1 3
1459
J. Lyu, B. Süssmu h
Economics, and he Leipzig Uni e si y Economics Lunch ime Semina o many help ul discussions. The
usual disclaime applies.
Funding Open Access unding enabled and o ganized by P ojek DEAL.
Jingjing Lyu epo s inancial suppo was p o ided by he Minis y o Cul u e and Tou ism o he P.R.
China (Social Science Resea ch P ojec ; G an N . 25DY13) and by No heas No mal Uni e si y (Social
Science You h Cul i a ing P ojec ; G an N . 24QN004).
Decla a ions
Con lic o in e es The e a e no addi ional ela ionships, pa en s o ac i i ies o disclose.
Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License,
which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long
as you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e
Commons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his
a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line
o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use
is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission
di ec ly om he copy igh holde . To iew a copy o his licence, isi h p : / / c e a i e c o m m o n s . o g / l i c e n
s e s / b y / 4 . 0 / .
Re e ences
Aguia , M., & Gopina h, G. (2007). Eme ging ma ke business cycles: The cycle is he end. Jou nal o
Poli ical Economy,115, 69–102.
Ahmed, S., Le in, A., & Wilson, B. (2004). Recen U.S. mac oeconomic s abili y: Good policies, good
p ac ices, o good luck? Re iew o Economics and S a is ics,86, 824–832.
A kolakis, C., & Ramana ayanan, A. (2009). Ve ical specializa ion and in e na ional business cycle syn-
ch oniza ion. Scandina ian Jou nal o Economics,111, 665–680.
A is, M., Ma cellino, M., & P oie i, T. (2004). Da ing business cycles: A me hodological con ibu ion
wi h an applica ion o he Eu o A ea. Ox o d Bulle in o Economics and S a is ics,66, 537–565.
A is, M., & Okubo, T. (2011). Does in e na ional ade eally lead o business cycle synch oniza ion? A
panel da a app oach. The Manches e School,79, 318–332.
Assenmache -Wesche, K., & Ge lach, S. (2008). In e p e ing Eu o A ea in la ion a high and low equen-
cies. Eu opean Economic Re iew,52, 964–986.
Assenmache -Wesche, K., & Ge lach, S. (2008). Money g ow h, ou pu gaps and in la ion a low and high
equency. Jou nal o Economic Dynamics & Con ol,32, 411–435.
Azcona, N. (2022). T ade and business cycle synch oniza ion: The ole o common ade pa ne s. In e -
na ional Economics,170, 190–201.
Backus, D., Kehoe, P., & Kydland, F. (1995). In e na ional business cycles: Theo y and E idence. In T. F.
Cooley (Ed.), F on ie s o business cycle esea ch (pp. 331–357). Be lin: De G uy e .
Baldwin, R. (2013). The de elopmen and es ing o Hecksche -Ohlin ade models. A e iew. Camb idge,
MA: MIT P ess.
Bax e , M., & King, R. (1999). Measu ing business cycles: App oxima e band-pass il e s o economic
ime se ies. Re iew o Economics and S a is ics,81, 575–593.
Belke, A., & Heine, J. (2006). Specialisa ion pa e ns and he synch onici y o egional employmen pa -
e ns. In e na ional Economics and Economic Policy,3, 91–104.
Bengu ia, F., Choi, J., Swenson, D. L., & Xu, M. J. (2022). Anxie y o pain? The impac o a i s and
unce ain y on Chinese i ms in he ade wa . Jou nal o In e na ional Economics,137, 103608.
Bie baume -Polly, J., Hube , P., & Rozmahel, P. (2016). Regional business-cycle synch oniza ion, sec o
specializa ion and EU accession. Jou nal o Common Ma ke S udies,54, 544–568.
Blonigen, B. A., Pige , J., & Sly, N. (2014). Como emen in GDP ends and cycles among ading pa -
ne s. Jou nal o In e na ional Economics, 94, 239–247. h p s : / / d o i . o g / 1 0 . 1 0 1 6 / j . j i n e c o . 2 0 1 4 . 0 6 . 0 0 8
1 3
1460