Measu ing he subs i u ion e ec s be ween High Speed Rail and ai
anspo in Spain
José I. Cas illo-Manzano
a,
⇑
, Ra ael Pozo-Ba ajas
a
, Juan R. T ape o
b
a
Applied Economics & Managemen Resea ch G oup, Uni e si y o Se ille, Spain
b
Facul ad de Ciencias y Tecnologías Químicas, Uni e sidad de Cas illa-La Mancha, Spain
a icle in o
Keywo ds:
High Speed Rail
Ai lines
Subs i u ion e ec s
Dynamic Linea Reg ession
Time Va ying Pa ame e
Spain
abs ac
The main objec i e o his pape is o es ima e he impac ha he expansion o he HSR ne wo k has had
on ai anspo in Spain by es ima ing he subs i u ion e ec be ween he wo ypes o anspo a ion.
This pape conside s he way ha he HSR ne wo k has g own and how his g ow h could ha e a ec ed
ai anspo dynamically. The findings show ha a dynamic ision o his subs i u ion a e should be
adop ed, as opposed o assuming ha he a e is cons an , as has been he case in p e ious e e ences.
Al hough he a e a ies significan ly o e he s udy pe iod, only 13.9% o HSR passenge demand was
ound o ha e come om ai a el du ing he 1999–2012 pe iod, meaning ha HSR and ai lines would
seem o o e mo e independen se ices han a fi s i migh appea . This confi ms he hypo hesis as o
he HSR’s g ea abili y o gene a e i s own demand. The subs i u ion a e be ween he wo anspo
modes seems o be closely linked o he way ha any new s a ions a e inco po a ed in o he HSR ne -
wo k. Con e gence be ween he seasonali y o HSR and ai anspo has also been examined. The esul s
show ha i is di ficul o alk o a eal HSR anspo ne wo k in Spain.
Ó2015 Else ie L d. All igh s ese ed.
1. In oduc ion
In hese imes o economic ad e si ies (see T achanas and
Ka akilidis (2013) and Ali (2012) on his aspec ) while eme ging
coun ies a e p epa ing o s a building hei High Speed Rail
(he eina e HSR) ne wo ks, o he s, such as F ance, which we e
pionee s in his ype o anspo (Ma i-Hennebe g, 2013), ha e
decided o pos pone some schemes while con inuing wi h o he s
(Leheis, 2012; Bayon, 2013), o o upg ade hei con en ional ail
inc easing speeds on exis ing acks up o 200 km pe hou (De
Rus, 2012). In his con ex , he e a e g owing numbe s o pape s
ha place inc eased impo ance on he cos and economic de el-
opmen (Gi oni, 2006), on he need o a p ope p io cos assess-
men o he in es men (De Rus and Nombela, 2007; De Rus and
Roman, 2006), and he need o p e en he HSR ne wo k being
de eloped on he basis o poli ical, a he han economic c i e ia
(Albala e and Bel, 2012; Bel, 2011).
A majo pa o his discussion abou he sui abili y o
de eloping HSR ne wo ks o lines ocuses on he analysis o he
compe i ion o collabo a ion be ween he HSR and o he means
o anspo a ion. Compe i ion be ween he HSR and he ca has
been add essed by Gonzalez-Sa igna (2004b) and Roman e al.
(2007), and collabo a ion be ween he HSR and o he land ans-
po by Tapiado e al. (2009). Wi h espec o compe i ion
be ween he HSR and ai a fic, Kappes and Me ke (2013) indi-
ca e ha he wo bigges obs acles o b eaking in o he Eu opean
ai ma ke a e access o slo s and compe i ion om HSR lines.
The e a e also s udies ha add ess collabo a ion be ween ai ans-
po and he HSR, such as hose by Gi oni and Banis e (2006),
Ly hgoe and Wa dman (2002), Redondi e al. (2013) and Soco o
and Viecens (2013). O he pa hs o collabo a ion could eme ge i
seconda y ai po s we e linked o he main hub by HSR, as is he
case in Spain, as his would allow in e na ional fligh s o be dis-
pe sed a ound he coun y (Sismanidou e al., 2013).
Specifically, he implemen a ion o he HSR in Spain and he
way ha he coun y has become an expo e o a mode o ans-
po is a case s udy o in e na ional significance (see Campos and
De Rus (2009), Ma i-Hennebe g (2013), Redondi e al. (2013),
Soco o and Viecens (2013), and Ma in and Nombela (2007) on
his opic). When he mos ecen Mad id–Ba celona–Figue es
and Mad id–Valencia–Alican e lines came in o se ice Spain
quickly ad anced o he posi ion o ha ing he second la ges
HSR ne wo k in he wo ld behind China, a coun y ha is no easily
compa able o Spain due o i s size and popula ion (see Albala e
h p://dx.doi.o g/10.1016/j.j angeo.2015.01.008
0966-6923/Ó2015 Else ie L d. All igh s ese ed.
⇑
Co esponding au ho a : Applied Economics & Managemen Resea ch G oup,
Uni e si y o Se ille, A da. Ramón y Cajal, 1, 41018 Se ille, Spain. Tel.: +34 954
556727; ax: +34 954 557629.
E-mail add esses: [email p o ec ed] (J.I. Cas illo-Manzano), [email p o ec ed]
(R. Pozo-Ba ajas), [email p o ec ed] (J.R. T ape o).
Jou nal o T anspo Geog aphy 43 (2015) 59–65
Con en s lis s a ailable a ScienceDi ec
Jou nal o T anspo Geog aphy
jou nal homepage: www.else ie .com/loca e/j angeo
and Bel (2011) on his opic), wi h, specifically, o e 3100 km in
se ice in 2013 (Adi , 2014). Howe e , his expansion could impac
nega i ely on egional cohesion, al hough he ex en o any
impac s a ies depending on he a ea (see O ega e al., 2014)o
he planning le el. O ega e al. (2012) conclude ha he e ec s
a e posi i e on he na ionwide and co ido le els bu ha on
he egional le el, hey migh be nega i e.
In his con ex , he main objec i e o his pape is o es i-
ma e he e ec ha he expansion o he HSR ne wo k has
had on ai anspo in Spain by es ima ing he subs i u ion
e ec be ween he wo ypes o anspo . This will be done
o e he b oad ime ame o Janua y 1999–Decembe 2012
ha was ma ked by he expansion o he Low Cos Ca ie s
(he eina e LCCs) ha included bo h in e na ional – o
example, Ryanai and Easyje – and domes ic – especially
Vueling – companies (Cas illo-Manzano e al., 2012b; Bel and
Fageda, 2010). This wen hand-in-hand wi h he decline o he
leading na ional ai line, i.e., Ibe ia, which closed he yea
(2012) wi h losses o 351 million Eu os. Ob iously, his expan-
sion o he LCCs, especially on domes ic ou es, is an impo an
ac o ha should be aken in o accoun as, a p io i, he all in
he cos o ai a el should inc ease compe i ion wi h he HSR
(Yang and Zhang, 2012) by educing he appeal o he HSR com-
pa ed o ai anspo .
The a icle is o ganized as ollows: Sec ion 2p esen s a li e a-
u e e iew o he subs i u ion e ec s be ween he HSR and ai
anspo . Sec ion 3lays ou he da a and p esen s he me hodolog-
ical app oach. Sec ion 4p esen s he empi ical esul s and he dis-
cussion o hese findings and, finally, Sec ion 5p esen s he
conclusions o he s udy.
2. Li e a u e e iew
The analysis o he subs i u ion a e be ween he HSR and ai
anspo is especially ele an o he economy. Fi s ly, an es i-
ma e is equi ed o enable he demand o ecas s and/o he Social
Cos /Benefi Analysis (De Rus and Roman, 2006) o be d awn up
ha a e used o jus i y/no jus i y he iabili y o non iabili y o
a new HSR line. In his line, Ca e a-Gomez e al. (2006), o exam-
ple, indica e ha om he economic poin o iew he HSR
be ween Mad id and Se ille would no be socially p ofi able when
analyzing conges ion, main enance cos s, acciden s and en i on-
men al cos s o he ca , he ain, and he long dis ance bus along
he Se ille–Mad id co ido , and he elas ici y o p ices be ween
hem, using a unc ion o maximize he sum o consume and p o-
duce su pluses. F om ano he poin o iew, Gi oni and Banis e
(2006) conclude ha ai lines can use ailway se ices as spokes
o hei own ne wo k o se ices. In such cases he ailway in a-
s uc u e complemen s he ai anspo in as uc u e, and should
also be seen as pa o he la e . Soco o and Viecens (2013) also
use a heo e ical model o conclude ha plane– ain in eg a ion
is beneficial, o a leas no de imen al, by which hey a e a guing
ha he complemen a y na u e o he wo has p imacy o e he
subs i u ion e ec .
Ma in and Nombela (2007) use an agg ega e mul inomial
logi model o s udy compe i ion acco ding o dis ance. Thei
s udy concludes ha he HSR will mainly a ac passenge s
om ai lines and he long dis ance bus on jou neys o o e
500 km, while o sho e jou neys, he main compe i ion will
come om ca s. A simila conclusion can be ound in
A ms ong and P es on (2011). Using logi disc e e choice mod-
els González-Sa igna (2004a) also concludes ha he HSR may
be conside ed as a uly compe i i e p oduc o ai anspo
o e dis ances ha can be co e ed in a maximum o 3 h. How-
e e , none o hese s udies goes so a as o o e es ima ions
ha quan i y he subs i u ion ela ionships be ween ai anspo
and HSR.
In he case o he Mad id–Se ille HSR line, DeRus and Inglada
(1997) use passenge da a om Ibe ia and RENFE ( he Spanish
na ional ailway company) and demand elas ici y wi h espec
o he GDP o indica e ha ai passenge s be ween he wo ci ies
ell om 694,400 (25.1%) in 1992 o 352,200 (10.1%) in 1996, he
yea ha he HSR was b ough in o se ice, and ha passenge s
on his new mode o anspo s ood a 1,438,200 (41.3%). Using
an analysis based on disagg ega ed mode choice models wi h
in o ma ion om a mixed e ealed p e e ences/s a ed p e e -
ences da abase, Roman e al. (2007) es ima ed ha in he bes
Mad id–Ba celona scena io ha hey analyzed, HSR’s ma ke
sha e would no exceed 35% o all ai and ain passenge s. How-
e e , in eali y he HSR ma ke sha e o he Mad id–Ba celona
ou e s ood a a ound 50% in 2012, he las yea o which i
was analyzed.
Using a Two-S age Leas Squa e es ima o (2SLS-IV) o es ima e
he equa ions, Jiménez and Be anco (2012) indica ed ha he
in oduc ion o HSR in Spain led o ai ope a ions educing by
17% a he same ime ha o e all demand o anspo inc eased,
which mean an e en g ea e all in hei sha e.
Basing his ocus on compe i ion be ween he HSR and he
plane, Dob uszkes (2011) finds ha ai companies could educe
he numbe o passenge s pe fligh and inc ease equency o
espond o he in oduc ion o HSR, and ha his would p e en
a all in o e all passenge numbe s. In he same line, Fageda e al.
(2011) s a e ha ai companies would also lowe hei p ices o
inc ease hei compe i i eness. Fo his hey use a p icing equa-
ion wi h he wo-s age leas squa es es ima o . Yang and
Zhang (2012) use a a ia ion o he classic Ho elling model o
indica e ha he a iables ha ha e he g ea es e ec on he
decision o use one mode o anspo o he o he in China a e
p ice and equency. One las an eceden o his s udy ha can
be ci ed is F öidh (2008), which compa es gene alized cos s aced
by plane and HSR passenge s and a i es a he b oad conclusion
ha a eling ime is he mos impo an ac o o gaining ma -
ke sha e in Sweden.
The ocus o his pape o e s he ollowing ad an ages com-
pa ed o hese ea lie pape s. Fi s ly, i conside s he way ha
he HSR ne wo k has g own and how his g ow h could ha e
a ec ed ai anspo dynamically. This enables he e ec s o
new high-speed ou es and o new ai po e minals o be quan-
ified. In addi ion, p e ious e e ences ha e employed eg ession
ype models in o de o explain he influence o new HSR s a ions
on fligh a fic by means o dummy a iables and cons an coe -
ficien s. He e, we epo a mo e flexible and obus me hodology
ha is capable o allowing pa ame e s ha a y o e ime. The
idea behind his is o explo e he subs i u ion ac o be ween
HSR and planes on a dynamic basis. Essen ially, he subs i u ion
e ec be ween he wo anspo a ion me hods can be a non-s a-
iona y s ochas ic a iable ins ead o a cons an coe ficien . In
pa icula , he p oposed model is a Dynamic Linea Reg ession
ha belongs o he amily o Time Va ying Pa ame e (TVP) mod-
els (Wes and Ha ison, 1989; Ha ey, 1989). The ad an ages o
ou app oach o e he ocuses o he p io li e a u e will be ana-
lyzed in Sec ion 3.
Finally, as he analysis ex ends h ough Decembe , 2012, bo h
he new lines ha ha e been b ough in o se ice ( he Cuenca–
Albace e–Valencia line, o example) and he e ec o he economic
c isis can be included. The basis om which we s a is ha HSR is
a high cos means o anspo o he a ele , al hough in gene al
e ms a es will no co e he o e all cos o i s cons uc ion. The
explosion o he LCCs on o he Spanish ma ke as a whole also
led o LCCs aking 52.5% o comme cial passenge ai anspo in
he las mon h o he sample.
60 J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65
3. Da a and me hods
The da ase can be di ided in o wo g oups:
(A) Ai passenge s on domes ic fligh s flying o/ om Mad id-
Ba ajas, which is conside ed as he endogenous a iable
and is compu ed as planepas
. This in o ma ion is a ailable
om he Spanish Public Ai po Au ho i y (AENA, 2014).
We only used da a om Mad id-Ba ajas ai po as he as
majo i y o Spanish a fic had i s o igin o des ina ion a
he Mad id HSR s a ions (Mad id Pue a de A ocha and, o
a much lesse ex en , Mad id-Chama ín) du ing he pe iod
analyzed. To be specific, his was 84% in 2011 (Fe opedia,
2014) and, i can be assumed, a highe figu e du ing he p e-
ious yea s, as he e we e ewe hubs in he HSR ne wo k. As
p e ious s udies ha e s a ed (see o example Gui ao, 2013),
he Spanish HSR ne wo k oday has a clea ly adial a chi ec-
u e in which Mad id clea ly s ands ou as he cen al hub.
This is e en ue gi en ha no only is Mad id’s HSR s a ion
(Mad id Pue a de A ocha) i sel loca ed wi hin he hin e -
land o he ai po , bu also o he HSR s a ions, such as
Toledo, Valladolid and Ciudad Real.
(B) The exogenous a iables, namely:
a The numbe o ai ope a ions on domes ic fligh s wi h a sin-
gle pe iod delay (op
1
). We assume ha ai passenge s a
ime a e hea ily influenced by he numbe o ope a ions
du ing he p e ious pe iod. This a iable s ongly co ela es
wi h he e olu ion o LCCs in Spain,
1
and he e o e co ec s
o hei possible e ec . This in o ma ion is a ailable om
he Spanish Public Ai po Au ho i y (AENA, 2014).
b High Speed Rail passenge s (hs pas
). To be specific, his is
he o al numbe o passenge s who a eled on he Spanish
High Speed Rail ne wo k du ing pe iod . These da a we e
ob ained om he Minis y o Public Wo ks (2014).
c The popula ion o he p o inces connec ed by HSR (popu
).
Sou ce: Na ional S a is ics Ins i u e (INE, 2014).
d The economic cycle measu ed by he unemploymen a e
(unemp
) in he p o ince o Mad id, acco ding o da a om
he Spanish Public Employmen Se ice (SEPE, 2014).
e Indica o o dummy exogenous a iables:
i On 9 h Janua y, 2009 Mad id expe ienced a hea y snow-
all ha pa alyzed he ai po o se e al hou s. This
dummy a iable (snow
) is modeled as a pulse, whe e all
he alues a e ze o excep o Janua y, 2009, which is
equal o one (ABC na ional newspape , 2009).
ii Business. The numbe o ading days in a mon h can a y
conside ably and his may ha e a subs an ial e ec on ai
passenge s. When conside ing jus wo ypes o day, o
ins ance, weekdays and Sa u days and Sundays, his a -
iable is calcula ed as he numbe o business o ading
days in a gi en mon h minus he numbe o Sa u days
and Sundays in said mon h, mul iplied by 5/2. Mo e
de ails o his way o dealing wi h business day a ia ion
can be ound in Ha ey, 1989, p. 334. Any u he holidays
in he mon h a e sub ac ed om he business days.
iii Seasonal dummy a iables. 11 dummy a iables a e
employed o model he annual seasonali y expec ed in
he da a.
The da a we e collec ed mon hly om Janua y, 1996, o Decem-
be , 2012, esul ing in a o al sample o 204 obse a ions pe
a iable.
Fig. 1 shows passenge s on domes ic fligh s and HSR passenge s.
The s ong annual seasonal pa e n should be no ed in bo h ime
se ies. I is in e es ing o no e ha e en when bo h ime se ies
e eal yea ly seasonali y, hei e olu ion, especially du ing he fi s
yea s, di e s no ably. In ac , up o 2005, HSR seasonali y was mo e
ou ism-o ien ed, wi h ob ious peaks in Augus and Decembe ,
compa ed o he mo e business-o ien ed seasonali y o he plane,
wi h oughs in Augus and Decembe . Subsequen ly, om 2006
onwa ds, HSR seasonali y inc easingly app oaches ha o he plane.
Fu he mo e, by he beginning o 2008 he e is a significan
d op in domes ic fligh s ha ma ches he s iking inc ease in
HSR passenge s.
Addi ional in o ma ion o explain he subs i u ion e ec
be ween plane and ain is gi en in Tables 1 and 2. In pa icula ,
Table 1 shows he da es ha new HSR ou es came in o se ice.
Table 2 also shows he da es when new ai po e minals we e
opened in Mad id (T4) and Ba celona.
The ime se ies model employed in he analysis is in he class o
ime- a ying pa ame e s desc ibed in a S a e Space amewo k
(Wes and Ha ison, 1989). The p oposed Dynamic Linea Reg es-
sion model can be desc ibed as ollows:
planepas
¼b
0
þb
1
op
1
þb
2
hs pas
þb
3
unemp
þb
4
snow
þb
5
bus
þb
6
popu
þX
17
i¼7
b
i
Seas
i
þ
e
ð1Þ
whe e
1planepas
is he numbe o domes ic ai passenge s a ime a
Mad id-Ba ajas ai po .
2op
1
is he numbe o ai ope a ions on domes ic fligh s a
Mad id-Ba ajas ai po du ing pe iod 1.
3hs pas
is he numbe o HSR passenge s in he Spanish ailway
sys em a ime .
4unemp
is he unemploymen a e.
5snow
is a dummy a iable ha deno es ha he ai po was
pa alyzed by a hea y snow all in Janua y, 2009.
6popu
is he popula ion o he p o inces connec ed by HSR a
ime .
7bus
is a dummy a iable ha conside s he di e ences be ween
business o ading days and weekends.
8Seas
i
is a se o dummy a iables o i=6,...,16 and ime .
The las e m in exp ession (1) e e s o he model e o , which
is assumed o be no mally dis ibu ed wi h ze o mean and a iance
2
.
99 00 01 02 03 04 05 06 07 08 09 10 11 12 13
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4 x 106
Time (yea s)
plane
HSR
Fig. 1. Ai passenge s on domes ic fligh s and HSR passenge s.
1
A sca e plo showing e idence o he significan co ela ion be ween ope a ions
and low-cos plane passenge s is a ailable om he au ho s upon eques .
J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65 61
In o de o associa e he es ima ed coe ficien s wi h hei
espec i e elas ici ies, he a iables, excluding he dummies, a e
log ans o med. The main di e ence be ween he Dynamic Linea
Reg ession model and he s anda d eg ession model lies in he
model coe ficien s. No e ha he Dynamic Linea Reg ession may
be seen as an ex ension o he s anda d Linea Reg ession when
he coe ficien s a e no limi ed o emaining cons an . In pa icula ,
unlike he es o coe ficien s he coe ficien b
2
in model (1) has
he sub index . This means ha he pa ame e is no assumed o
be cons an and, hus, i may a y wi h ime. I should be no ed
ha mos o he s udies assume ha he coe ficien is cons an
(De Rus and Roman, 2006; Ma in and Nombela, 2007; Roman
e al., 2007). b
2
in (1) s ands o he subs i u ion e ec be ween
plane and ail passenge s. In gene al e ms, gi en he numbe o
changes ha he anspo indus y is expe iencing, such as new
ou es, he g ow h o LCCs, and agg essi e p ice policies, i seems
mo e sensible o assume a ime- a ying coe ficien a he han
one ha is cons an .
Typical leas squa es o maximum likelihood es ima ion p oce-
du es canno be applied o es ima e his pa ame e . Essen ially,
ecu si e es ima ion algo i hms, such as he Kalman Fil e , should
be employed ins ead. In o he applica ions, such as enginee ing,
he way ha he pa ame e e ol es o e ime can some imes be
known a p io i, bu in economics such knowledge is no usually
a ailable. So, in his case, a Gene alized Random Walk model
exp essed in a S a e Space amewo k (Jakeman and Young,
1981) is used o allow pa ame e b
2
o a y o e ime:
x
x


¼
a
1
a
2
0
a
3

x
1
x

1

þ
g
g

 ð2Þ
b
2
¼10ðÞ
x
x

 ð3Þ
He e (2) and (3) a e he s a e and obse a ion equa ions, espec-
i ely, whe e
a
1
,
a
2
and
a
3
a e cons an pa ame e s; b
2
is a
smoo hed signal componen consis ing o he fi s s a e x
; and x

is a second s a e a iable (gene ally known as he ‘‘slope’’); while
g
and g

a e ze o mean, se ially unco ela ed whi e noise a iables
wi h a cons an block diagonal co a iance ma ix.
Depending on he alues o
a
i
,i= 1, 2, 3, he model may be pa -
icula ized o special cases, such as he Random Walk; he
Smoo hed Random Walk; he In eg a ed Random Walk; he Local
Linea T end; and he Damped T end. Fo ins ance, i
a
1
=
a
2
=
a
3
=1;
g
= 0, he model is he In eg a ed Random Walk (IRW).
These models a e implemen ed in he MATLAB ‘CAPTAIN’ oolbox,
whe e he s a e es ima ion is achie ed by means o he Kalman Fil-
e and Fixed In e al Smoo hing algo i hms. A mo e de ailed
desc ip ion o he algo i hms implemen ed in CAPTAIN can be
ound in Taylo e al. (2007).
In ou pa icula case, he a o emen ioned IRW model has been
used o es ima e b
2
.
Finally, one posi i e aspec o he p oposed Dynamic Linea
Reg ession model ocus o e he s anda d eg ession model is ha
i is mo e pa simonious, easie o in e p e and, ela i ely-speak-
ing, p e en s issues wi h omi ed a iables. Fo ins ance, he s an-
da d eg ession model would equi e as many dummy a iables o
be defined p e iously as po en ial changes a e expec ed in he sub-
s i u ion e ec , such as he impac o a new HSR ain s a ion o /
and a new ai po e minal, o example (see Tables 1 and 2). Thus,
he numbe o dummy a iables would g ow subs an ially. Fu -
he mo e, a e emo ing any dummy a iables ha a e no s a is-
ically significan , he in e p e a ion o he esul s would be mo e
di ficul . In o he wo ds, whe eas he ime- a ying pa ame e
es ima ion o he DLR model p o ides a smoo h es ima e o he
ime- a ying subs i u ion e ec , he use o a cons an coe ficien s
s anda d eg ession wi h nume ous dummies would yield an
app oxima ion o he subs i u ion e ec based on ab up changes
and po en ial mul icollinea i y issues. Rega ding he obus ness
o he me hod o omi ed a iables, i should be no ed ha i he e
a e any omi ed a iables in he s anda d eg ession model ha
a ec he subs i u ion coe ficien , such as a dummy indica ing a
new HSR s a ion, o example, he esul s migh be misleading.
Howe e , he DLR cap u es he dynamics o he coe ficien wi hou
he need o such dummy indica o s, and hus, he esul s ob ained
a e mo e obus .
4. Resul s and discussion
Fig. 2 shows he es ima e o b
2 2
e sus ime. I can be concluded
ha , as we had assumed, he pa ame e is no cons an . In o de o
explo e he influence o new HSR s a ions as well as new ai po e -
minals, e ical lines ha e been added o he figu e. Dashed lines
ep esen he imes when new HSR s a ions we e opened. This in o -
ma ion is also e e ed o Table 1. Do ed lines s and o he new
e minal ai po s men ioned in Table 2.
As he a iables o plane and ain passenge s in Eq. (1) a e
changed o loga i hms, he b
2
pa ame e is a c oss pseudo elas ic-
i y, bu whe e he denomina o is a quan i y, and no a p ice. Basi-
cally, i p o ides in o ma ion abou he pe cen age all in ai
passenge s i HSR passenge s we e o inc ease by 1%. Fo example,
i mon hly HSR passenge s ha e inc eased on a e age by a figu e o
312,674 (20.17%) be ween 2007 and 2012, acco ding o ou models
his co ela es wi h a all o 67,074 (5.78%) ai passenge s. The e-
o e only 21.4% o he inc ease in HSR passenge s would come om
ai anspo . This pe cen age would all o 13.9% o he whole
o he 1999–2012 pe iod analyzed, as an a e age annual inc ease
o 385,960 HSR passenge s co ela es wi h an a e age annual
all o 53,558 ai passenge s.
I is easy o see how he subs i u ion a e be ween he wo
means o anspo has allen in absolu e alues o e ime, going
om a figu e o a ound 0.191 a he beginning o 1999 o a figu e
o 0.164 by he end o 2007. This all p ocess exis ed despi e he
ne wo k expanding du ing his pe iod wi h new connec ions o
Za agoza and Malaga and only slowed down empo a ily owa d
he end o 2001. This is a clea 9/11 e ec and was also el by
ai anspo du ing he mon hs a e he e o is a ack. In any
case, he elas ici y con inued o end owa d ze o wi h he opening
Table 1
S a da es o new HSR ou es.
HSR ou es S a da e
Mad id–Guadalaja a–Za agoza–Lleida 10/2003
Za agoza–Huesca 04/2005
Mad id–Toledo 11/2005
Mad id–Ta agona 12/2006
Mad id–Malaga 12/2007
Mad id–Sego ia–Valladolid 12/2007
Mad id–Ba celona 02/2008
Mad id–Cuenca–Albace e–Valencia 12/2010
Table 2
Opening da es o new ai po e minals in Mad id and Ba celona.
Ai po e minals Opening da e
Mad id T4 02/2006
Ba celona 07/2009
2
No e ha he es ima ion o b
2
is s a is ically significan a he 5% le el.
62 J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65
o he new HSR line o Za agoza (10/2003) and he opening o
Te minal T4 in Mad id (02/2006).
Ne e heless, when he Mad id–Malaga (Decembe , 2007) and
Mad id–Ba celona (Feb ua y, 2008) lines came in o ope a ion,
he e was a o al hal in he all in subs i u abili y in absolu e
e ms. This was also influenced by he all in o e all ain + plane
passenge numbe s due o he economic c isis ha a ec ed Spain
(Paglia a e al., 2012). The end u ned ound and began o ise
again in 2010. This e ec inc eased when he Mad id–Valencia line
came in o se ice (Decembe , 2010). In all hese h ee cases we a e
close o he condi ions ha Vicke man (1997) s a ed we e neces-
sa y o an HSR line o be socially p ofi able, as he h ee ci ies
in ol ed a e all la ge and he dis ances ange be ween 354.4 km
in he case o Valencia (wi h 2.5 million inhabi an s in he p o -
ince, compa ed o Za agoza’s 975 housand), 530.5 km in he case
o Malaga (wi h 1.6 million) and Ba celona’s 619.9 km (wi h
5.5 million). This e ec is u he heigh ened as a high- alue ne -
wo k hub is included ha has adi ionally been se ed by an ai
shu le wi h some o he highes passenge numbe s in Eu ope,
Mad id–Ba celona.
Howe e , i canno be ejec ed ha he subs i u ion e ec
be ween he wo ypes o anspo had p e iously been g ea e ,
du ing he yea s ollowing he opening o he fi s HSR line in
1992, which was a pe iod ma ked by he high p ice o domes ic
ai a es. This would help o explain why p e ious pape s o e
no iceably highe subs i u ion a es. Un o una ely, he lack o
mon hly ime se ies o he da ase used does no allow us o es
his hypo hesis.
The emaining pa ame e es ima es a e shown in Table 3. Fo
he sake o cla i y, he es ima ion o he dummy a iables ha
model he seasonali y has no been included in he able. I should
be no ed ha , e en when some seasonali y dummies a e no s a is-
ically significan , i is impo an o include hem in he model in
o de o cap u e he whole seasonali y pa e n.
I should be no ed ha he model in (1) assumes ha he ol-
ume o High Speed Rail passenge s de e mines he numbe o ai
passenge s. Howe e , his assump ion may gi e ise o an endoge-
nei y issue. A G ange non-causali y es was ca ied ou o es
his assump ion (see Lü kepohl, 2005 o de ails). Table 4 shows
ha he causali y di ec ion uns om high-speed ail passenge s
o plane passenge s, as he null hypo hesis based on a epas no
causing planepas is ejec ed wi h a P- alue o 0.01.
5. Conclusions
Compa ed o ea lie s udies his a icle o e s an absolu ely o i-
ginal me hodological app oach ha uses Dynamic Linea Reg es-
sion models based on ime- a ying pa ame e s de eloped in a
S a e Space amewo k. These enable he subs i u ion a e be ween
he anspo o HSR and ai anspo passenge s o be analyzed in
dep h. Using a ele an case s udy, he apid g ow h o he Spanish
high-speed ain ne wo k, p o ides he mos p ecise ision o da e
o he deg ee o subs i u abili y be ween hese wo modes o ans-
po . Empi ical e idence is also o e ed as o he eal o igin o he
passenge s who c ea e he demand o HSR.
Fi s ly, he findings confi m ha he dynamic ision o his sub-
s i u ion a e is he ision ha should be adop ed; con inual
changes in Spanish ail anspo a ion wi h he opening o new
lines mean ha ecu si e es ima ion echniques ha e o be used
o show he way ha hese changes impac on he subs i u ion ac-
o be ween ai anspo and HSR, as opposed o supposing ha
his a e is cons an , as was assumed by p e ious e e ences. The
a e a ies significan ly o e he s udy pe iod. Specifically, he sub-
s i u ion a e can be seen o all om i s maximum absolu e alue
du ing he pe iod o s udy, 0.191 in Janua y, 1999, o a minimum
alue o 0.164. This implies a mean a e age alue o 0.1767 o
said subs i u ion coe ficien o he pe iod o he analysis, 1999–
2012. This in u n equa es o an inc ease o only 13.9% in HSR pas-
senge demand coming om ai a el du ing his pe iod.
When his is examined alongside p e ious s udies, De Rus and
Roman (2006) can be seen o ha e p edic ed ha 91% o HSR pas-
senge s ha would come om o he modes o anspo on he
Mad id–Ba celona sec ion would come om he plane, which
would mean ha app oxima ely 60% o HSR passenge s would
swi ch om ai a el. Ma in and Nombela (2007) p edic ed ha
ai a el would lose 20% o ma ke sha e na ionwide by 2010,
and ha hese passenge s would ans e o he ain, which logi-
cally equa es o a clea ly highe pe cen age han 20% o HSR pas-
senge s. Meanwhile, Jiménez and Be anco (2012) es ima ed a
17% educ ion in he numbe o ai ope a ions due o HSR. In sho ,
he findings o his s udy a e mo e es ained han he esul s o
he abo e s udies, which we e some imes based on p io
p edic ions. In o he wo ds, HSR and he ai lines would seem o
o e mo e independen se ices han i a fi s migh appea .
99 00 01 02 03 04 05 06 07 08 09 10 11 12 13
-0.185
-0.18
-0.175
-0.17
-0.165
Time (yea s)
β2
Mad id-Lleida ou e
Mad id Te minal
Mad id –Malaga, Mad id-Valladolid ou es
Mad id -Ba celona ou e
Ba celona Te minal
Mad id-Valencia ou e
Fig. 2. Es ima ion o b
2
e sus ime.
Table 3
Es ima ion esul s o he Dynamic Linea Reg ession model
p oposed in (1).
Explana o y a iable Es ima e
b
0
14.4
⁄⁄⁄
(1.7)
op
1
0.43
⁄⁄⁄
(0.11)
unemp
0.04 (0.17)
snow
0.12
⁄⁄⁄
(0.04)
bus
0.002
⁄
(0.0012)
popu
0.05 (0.1)
R
2
0.97
2
1.8 10
3
Q(12) 13.04
Q(24) 19.48
KSL 0.07 (0.21)
2
s ands o he inno a ions a iance; Q(12) a e he Ljung-
Box Qs a is ics o 12; KSL is a Kolmogo o –Smi no –Lillie o s
Gaussiani y es (P- alues in b acke s).
One, wo, o h ee as e isks indica e coe ficien significance a
he 10%, 5% and 1% le els, espec i ely.
Table 4
G ange non-causali y es s.
Null hypo hesis F-s a is ic P- alue
G ange : hs pas does no cause planepas 3.7 0.01
G ange : planepas does no cause hs pas 0.35 0.788
No e: The e e ence model o he G ange es s is VAR(3) iden ified ia he
Schwa z Bayesian C i e ion.
J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65 63

Fu he mo e, he e a e changes in he a e be ween he wo as
ime goes on; fi s ly, i alls in absolu e alue un il he end o 2007,
despi e he ac ha he supposed Spanish high-speed ail ne wo k
has g own, specifically wi h he addi ion o he Mad id–Guadalaj-
a a–Za agoza–Le ida sec ion in Oc obe , 2003, he Za agoza–Hues-
ca sec ion in Ap il, 2005, he Mad id–Toledo sec ion in No embe ,
2005, and he Le ida–Ta agona sec ion in Decembe , 2006. The
limi ed popula ions o he new ci ies inco po a ed in o he ne -
wo k mean ha hese sec ions do no ulfill he equi emen s s a-
ed by Vicke man (1997) o jus i y an HSR connec ion.
The e was a change in he end wi h he opening o he HSR
lines o Malaga and Ba celona (end o 2007 and beginning o
2008). Despi e he ac ha he subs i u ion a e seems o ha e fla
lined wi h he coming in o ope a ion o he new ai po e minal a
Ba celona, he ac o he ma e is ha he HSR alue con inued o
ise, especially wi h he coming in o se ice o he Mad id–Valen-
cia line (12/2010).
These findings can also be conside ed indi ec empi ical p oo
ha he academic deba e o e whe e HSR passenge demand comes
om has been misguided, as was p e iously he case ega ding
whe e demand o LCCs was coming om. The la e ini ially
ocused on subs i u ion be ween LCCs and Ne wo k Ai lines, bu
an inc easing numbe o mo e ecen s udies (see Cas illo-
Manzano e al., 2012a; Cas illo-Manzano and Lopez-Valpues a,
2014; Gillen and Lall, 2004; Tapiado e al., 2008) a e showing ha
his subs i u ion is, in many cases, negligible compa ed o he new
demand ha he LCCs gene a e. Subsequen analyses should he e-
o e pu mo e emphasis on dis inguishing be ween he sha e o HSR
demand ha is new demand, and he sha e ha comes om o he
means o anspo , wi h he analysis o subs i u ion a es being
b oadened o ake in o he modes o anspo apa om he plane.
In addi ion, should he majo i y o passenge s no come om
he plane, as he findings o he p esen s udy would seem o sug-
ges , his would ha e a g ea impac on he scena ios and esul s o
Cos Benefi Analyses ha a e used o suppo decisions on cos ly
new HSR lines, especially as he willingness o pay o use s who
swi ch om he con en ional ain and, abo e all, he long dis ance
bus and he ca , will p esumably no co e he eal cos s o he HSR
(see De Rus and Roman (2006) o an analysis o he willingness o
pay o po en ial HSR use s depending on he mode o anspo
om which hey come).
To summa ize, he conclusions d awn om his s udy allow i
o be s a ed ha e en in he mos geog aphically ex ensi e HSR
‘ne wo k’ in he wo ld compa ed o he su ace a ea in ol ed,
he Ibe ian peninsula, he e is no empi ical e idence a all ha
he ne wo k has gene a ed any clea ne wo k e ec s ha will
a ac mo e passenge s om ai anspo , since he subs i u ion
coe ficien alue a he end o he pe iod o analysis is e en sligh ly
lowe han he alue a he beginning, when he Mad id–Se ille
line was he only line in se ice. Howe e , he sea ch o his pos-
sible absence o ne wo k e ec s would equi e his analysis being
comple ed wi h simila analyses o HSR ne wo ks in o he coun-
ies and, in he case o he Spanish ne wo k, igo ous analyses
by indi idual line. Nei he o hese asks would be easy gi en he
e y small numbe o HSR ne wo ks wo ldwide, and he lack o
anspa ency o Spanish da a.
In o he wo ds, he expansion o he ne wo k, wi h some lines
o e ing less han doub ul social p ofi abili y and clea ly ollowing
poli ical c i e ia (Bel, 2011), has seen he subs i u ion a e wi h ai
anspo all o e many yea s.
Acknowledgemen s
The au ho s a e g a e ul o P o . Tim Schwanen and he anony-
mous e iewe s o hei e y help ul commen s. The au ho s
would also like o exp ess hei g a i ude o he Eu opean Regional
De elopmen Fund and he Spanish Minis y o Economy and Com-
pe i i eness (G an Numbe ECO2012-36973) o hei suppo .
Re e ences
ABC, 2009. Imp e isión o al. ABC Mad id 10/1/2009, 1.
Adi , Adminis ado de In aes uc u as Fe o ia ias, 2014.Adi – Líneas de Al a
Velocidad. <h p://www.adi .es/es_ES/in aes uc u as/lineas_de_al a_
elocidad/lineas_de_al a_ elocidad.sh ml> (11.01.14).
AENA, 2014. Es adís icas áfico: Pasaje os, ope aciones y ca ga. <h p://www.aena-
ae opue os.es/csee/Sa elli e?Language=ES_ES&pagename=es adis icas>
(15.02.14).
Albala e, D., Bel, G., 2011. When economics does no ma e : ise and splendou o
he High Speedy Rail in Spain. Re . Econ. Apl. 19 (55), 171–190.
Albala e, D., Bel, G., 2012. The Economics and Poli ics o High-Speed T ain.
Lexing on Books.
Ali, T.M., 2012. The impac o he so e eign deb c isis on he eu ozone coun ies.
P ocedia – Soc. Beha . Sci. 62, 424–430.
A ms ong, J., P es on, J., 2011. Al e na i e ailway u u es: g ow h and/o
specialisa ion? J. T ansp. Geog . 19 (6), 1570–1579.
Bayon, M., 2013. Pa is-Nice TGV plans sc apped. The Ri ie a Times 11.07.2013.
<h p://www. i ie a imes.com/index.php/p o ence-co e-dazu -a icle/i ems/
pa is-nice- g -plans-sc apped.h mlcom/index.php/p o ence-co e-dazu -
a icle/i ems/pa is-nice- g -plans-sc apped.h ml> (27.09.14).
Bel, G., 2011. In as uc u e and na ion building: he egula ion and financing o
ne wo k anspo a ion in as uc u es in Spain (1720–2010). Bus. His . 53 (5),
688–705.
Bel, G., Fageda, X., 2010. P i a iza ion, egula ion and ai po p icing: an empi ical
analysis o Eu ope. J. Regul. Econ. 37 (2), 142–161.
Campos, J., De Rus, G., 2009. Some s ylized ac s abou high-speed ail: a e iew o
HSR expe iences a ound he wo ld. T ansp. Policy 16 (1), 19–28.
Ca e a-Gomez, G., Cas anedo-Galan, J., Co o-Millan, P., Inglada, V., Pesque a, M.A.,
2006. Cos –benefi analysis o in es men in high-speed ain sys em in Spain.
T anspo . Res. Rec. 1960, 135–141.
Cas illo-Manzano, J.I., Lopez-Valpues a, L., 2014. Li ing ‘‘up in he ai ’’: Mee ing he
equen flye passenge . J. Ai T ansp. Manage. 40, 48–56.
Cas illo-Manzano, J.I., Lopez-Valpues a, L., Ped egal, D.J., 2012a. How can he e ec s
o he in oduc ion o a new ai line on a na ional ai line ne wo k be measu ed?
J. T ansp. Econ. Policy 46 (2), 263–279.
Cas illo-Manzano, J.I., Lopez-Valpues a, L., Ped egal, D.J., 2012b. Wha ole will hubs
play in he LCC poin - o-poin connec ions e a? The Spanish expe ience. J.
T ansp. Geog . 24, 262–270.
De Rus, G., 2012. Economic E alua ion o he High Speed Rail. Repo o he Expe
G oup o En i onmen al S udies. Minis y o Finance, Sweden. 2012:1.
De Rus, G., Nombela, G., 2007. Is in es men in high speed ail socially p ofi able? J.
T ansp. Econ. Policy 41, 3–23.
De Rus, G., Roman, C., 2006. Economic e alua ion o he high speed ail Mad id-
Ba celona. Re . Econ. Apl. 14 (42), 35–79.
DeRus, G., Inglada, V., 1997. Cos –benefi analysis o he high-speed ain in Spain.
Ann. Reg. Sci. 31 (2), 175–188.
Dob uszkes, F., 2011. High-speed ail and ai anspo compe i ion in Wes e n
Eu ope: a supply-o ien ed pe spec i e. T ansp. Policy 18 (6), 870–879.
Fageda, X., Jiménez, J.L., Pe digue o, J., 2011. P ice i al y in ai line ma ke s: a s udy
o a success ul s a egy o a ne wo k ca ie agains a low-cos ca ie . J. T ansp.
Geog . 19 (4), 658–669.
Fe opedia, 2014. Núme o de iaje os AVE – La ga Dis ancia de g andes elaciones
de Ren e e ing esos. <h p://www. e opedia.es/wiki/N%C3%BAme o_de_
iaje os_AVE_-_La ga_Dis ancia_de_g andes_ elaciones_de_Ren e_e_ing esos>
(24.10.14).
F öidh, O., 2008. Pe spec i es o a Fu u e High-Speed T ain in he Swedish
Domes ic T a el Ma ke . J. T ansp. Geog .
Gillen, D., Lall, A., 2004. Compe i i e ad an age o low-cos ca ie s: some
implica ions o ai po s. J. Ai T ansp. Manage. 10, 41–50.
Gi oni, M., 2006. De elopmen and impac o he mode n high-speed ain: a
e iew. T ansp. Re . 26 (5), 593–611.
Gi oni, M., Banis e , D., 2006. Ai line and ailway in eg a ion. T ansp. Policy 13 (5),
386–397.
González-Sa igna , M., 2004a. Compe i ion in ai anspo : he case o he high
speed ain. J. T ansp. Econ. Policy 38 (1), 77–108.
Gonzalez-Sa igna , M., 2004b. Will he high-speed ain compe e agains he
p i a e ehicle? T ansp. Re . 24 (3), 293–316.
Gui ao, B., 2013. Spain: highs and lows o 20 yea s o HSR ope a ion. J. T ansp.
Geog . 31, 201–206.
Ha ey, A., 1989. Fo ecas ing S uc u al Time Se ies Models and he Kalman Fil e .
Camb idge Uni e si y P ess.
INE, Ins i u o Nacional de Es adís ica, 2014. Ci as de población y censos
demog áficos. <h p://www.ine.es/inebmenu/mnu_ci aspob.h m#1>
(23.10.14).
Jakeman, A.J., Young, P.C., 1981. Recu si e fil e ing and he in e sion o ill-posed
causal p oblems. U ili as Ma h. 25, 351–376.
Jiménez, J.L., Be anco , O., 2012. When ains go as e han planes: he s a egic
eac ion o ai lines in Spain. T ansp. Policy 23, 34–41.
Kappes, J.W., Me ke , R., 2013. Ba ie s o en y in o Eu opean a ia ion ma ke s
e isi ed: a e iew and analysis o manage ial pe cep ions. T anspo . Res. E:
Logis . T anspo . Re . 57, 58–69.
64 J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65
Leheis, S., 2012. High-speed ain planning in F ance: lessons om he
Medi e anean TGV-line. T ansp. Policy 21, 37–44.
Lü kepohl, H., 2005. New in oduc ion o mul iple ime se ies analysis. Sp inge ,
Be lin.
Ly hgoe, W.F., Wa dman, M., 2002. Demand o ail a el o and om ai po s.
T anspo a ion 29 (2), 125–143.
Ma i-Hennebe g, J., 2013. Eu opean in eg a ion and na ional models o ailway
ne wo ks (1840–2010). J. T ansp. Geog . 26, 126–138.
Ma in, J.C., Nombela, G., 2007. Mic oeconomic impac s o in es men s in high
speed ains in Spain. Ann. Reg. Sci. 41 (3), 715–733.
Minis e io de Fomen o, 2014. Ren e-Ope ado a. Viaje os anspo ados. Al a
elocidad y la ga dis ancia. <www. omen o.gob.es/BE/sedal/07103000.XLS>
(15.02.14).
O ega, E., Lopez, E., Monzon, A., 2012. Te i o ial cohesion impac s o high-speed
ail a di e en planning le els. J. T ansp. Geog . 24, 130–141.
O ega, E., Lopez, E., Monzon, A., 2014. Te i o ial cohesion impac s o high-speed
ail unde di e en zoning sys ems. J. T ansp. Geog . 34, 16–24.
Paglia a, F., Vassallo, J.M., Román, C., 2012. High-speed ail e sus ai anspo a ion.
Case s udy o Mad id-Ba celona, Spain. T anspo . Res. Rec.: J. T anspo . Res.
Boa d 2289, 10–17.
Redondi, R., Malighe i, P., Palea i, S., 2013. Eu opean connec i i y: he ole played
by small ai po s. J. T ansp. Geog . 29, 86–94.
Roman, C., Espino, R., Ma in, J.C., 2007. Compe i ion o high-speed ain wi h ai
anspo : he case o Mad id-Ba celona. J. Ai T ansp. Manage. 13 (5), 277–284.
SEPE, Se icio Público de Empleo Es a al, 2014. Es adís icas de empleo. <h p://
www.sepe.es/con enido/es adis icas/da os_es adis icos/empleo/index.h ml>
(15.02.14).
Sismanidou, A., Ta adellas, J., Bel, G., Fageda, X., 2013. Es ima ing po en ial long-
haul ai passenge a fic in na ional ne wo ks con aining wo o mo e
dominan ci ies. J. T ansp. Geog . 26, 108–116.
Soco o, M.P., Viecens, M.F., 2013. The e ec s o ai line and high speed ain
in eg a ion. T anspo . Res. A – Policy P ac . 49, 160–177.
Tapiado , F.J., Ma eos, A., Ma í-Hennebe g, J., 2008. The geog aphical e ficiency o
Spain’s egional ai po s: a quan i a i e analysis. J. Ai T ansp. Manage. 14, 205–
212.
Tapiado , F.J., Bu ckha , K., Ma í-Hennebe g, J., 2009. Cha ac e izing Eu opean
high speed ain s a ions using in e modal ime and en opy me ics. T anspo .
Res. A – Policy P ac . 43, 197–208.
Taylo , C.J., Ped egal, D.J., Young, P.C., Tych, W., 2007. En i onmen al ime se ies
analysis and o ecas ing wi h he Cap ain oolbox. En i on. Modell. So w. 22
(6), 797–814.
T achanas, E., Ka akilidis, C., 2013. Fiscal defici s unde financial p essu e and
insol ency: e idence o I aly, G eece and Spain. J. Policy Model. 35 (5), 730–
749.
Vicke man, R., 1997. High-speed ail in Eu ope: expe ience and issues o u u e
de elopmen . Ann. Reg. Sci. 31 (1), 21–38.
Wes , M., Ha ison, J., 1989. Bayesian Fo ecas ing and Dynamic Models. Sp inge -
Ve lag, New Yo k.
Yang, H., Zhang, A., 2012. E ec s o high-speed ail and ai anspo compe i ion on
p ices, p ofi s and wel a e. T anspo . Res. B: Me hodol. 46 (10), 1322–1333.
J.I. Cas illo-Manzano e al. / Jou nal o T anspo Geog aphy 43 (2015) 59–65 65