The ci y-ai po connec ion in he low-cos ca ie e a: Implica ions o u ban
anspo planning
José I. Cas illo-Manzano
*
Facul ad de Ciencias Económicas y Emp esa iales, Uni e si y o Se ille, A da. Ramón y Cajal, 1, 41018 Se ille, Spain
Keywo ds:
Ai po g ound access
U ban anspo planning
Passenge beha io
Low-cos ai lines
abs ac
This a icle, examines di e ences be ween he beha io o passenge s o low-cos and ne wo k ai lines
when choosing hei anspo mode o a el o ai po s. I is ound ha a passenge flying wi h a low-
cos ca ie is 6% less likely o ake a axi o he ai po , bu mo e han 4% mo e likely o d i e a en ed ca
and 2% mo e likely o use public anspo han a use o a ne wo k ca ie .
Ó2010 Else ie L d. All igh s ese ed.
1. In oduc ion
One o he ea u es o ai anspo de egula ion has been he
de elopmen o low-cos ca ie s (LCCs) and hei new manage-
men model, he main cha ac e is ics o which ha e unde mined
belie in he indus y’s p e ious s uc u e, p ocedu es, and business
models and b ough abou majo changes in he s a egies and
beha io o he di e en economic agen s on bo h he supply and
he demand sides o he ai anspo indus y.
On he supply side, o example, he e has been an upsu ge in
he use and impo ance o many seconda y, o en unde used
(F ancis e al., 2004), egional ai po s compa ed o he hubs
(Reynolds-Feighan, 2001). These ai po s ha e seen hei
ba gaining powe diminished by he agg essi e ba gaining
me hods o LCCs (Ba e , 2004) seeking o achie ing minimum
ai po cha ges. B oadly speaking, compe i ion be ween ai po s
o a ac LCCs can be said o be on he inc ease (Pels e al.,
2009).
LCCs’use o seconda y ai po s also means ai line a ficis
u he sca e ed ac oss mul iple ai po s se ing he same
me opoli an a ea. This is a d awback o ai po access planning,
he main pu pose o which is o educe he ma ke sha e o p i a e
ehicle use by bo h passenge s and ai po s a (Humph eys and
Ison, 2005 on policies o changing ai po employee a el
beha io ), as p i a e ehicles a e he modes o anspo a ion ha
mos con ibu e o noise, conges ion le els, and ai pollu ion
(G aham, 2008) in ai po hin e lands.
On he demand side, one o he mos impo an changes
b ough abou by he LCCs o ou analysis is he inc eased demand
o ai anspo se ices by younge and p ice-sensi i e a ele s
(O’Connell and Williams, 2005). A p io i his migh be su mised o
impac on he ma ke sha es o he a ious modes o anspo a ion
ha passenge s use o ge o ai po s and benefi he less expensi e.
This would explain he in e es ha some LCCs in Eu ope ha e in
ei he de eloping o wo king in coope a ion wi h bus and coach
companies (such as Te a ision, igh ly linked o he Ryanai g oup,
o he Easybus connec ions o he ai po s a ound London whe e
Easyje ope a es).
This pape looks a di e ences in he beha io o LCC and
ne wo k ai line passenge s when choosing o a mode o
anspo a ion o an ai po ? And i hese di e ences do exis ,
wha a e hey, and wha e ec s do hey ha e on ai po g ound
access planning? Apa om some heo e ical conside a ions,
he e ha e been ewe empi ical s udies looking a hese
ques ions.
1
2. Da a
The da abase comp ised 20,383 passenge s, 6247 o whom we e
LCC passenge s. All o hese we e in e iewed in depa u e lounges
*Tel.: þ34 678542833.
E-mail add ess: [email p o ec ed]
1
Eche a ne (2008) supposed ha logic would dic a e ha low-cos passenge s
would gene a e a g ea e demand o ca pa king acili ies a egional and
seconda y ai po s whils adi ional ai line passenge s would p esen a highe axi
usage. De Neu ille (2006) simila ly su mised ha p ice-conscious a ele s on low-
cos ca ie s a e he passenge s mos likely o use public anspo , and ha he
ma ke o ai po access is now mo e closely aligned wi h he adi ional ma ke
o u ban public anspo .
Con en s lis s a ailable a ScienceDi ec
Jou nal o Ai T anspo Managemen
jou nal homepage: www.else ie .com/loca e/jai aman
0969-6997/$ esee on ma e Ó2010 Else ie L d. All igh s ese ed.
doi:10.1016/j.jai aman.2010.02.005
Jou nal o Ai T anspo Managemen 16 (2010) 295e298
a se en di e en Spanish seconda y ai po s, none o which had
e ficien ail-based public anspo a ion o he ci y a he ime he
su ey campaign was conduc ed (Table 1).
2
The size o he sample
and he la ge numbe o passenge a ibu es and cha ac e is ics
p o ide an e ec i e da abase o an analysis o he ai anspo
indus y in Medi e anean coun ies (see Cas illo-Manzano and
López-Valpues a (in p ess) o ano he applica ion o his da abase).
Themain ea u eso AENA’s( heSpanish PublicAi po Au ho i y)
2005e2007 su ey ha is used o cons uc he da abase a e lis ed in
Table 1. AENA has o a ed annual su eys ound some o i s ai po s
since 1996. The me hodology used has, he e o e, finely honed and
he e has been su ficien unding o ob ain samples ha closely
app oxima e o a andom p ocess o a la ge popula ion and wi h
a low sampling e o . As wi h simila da abases, each obse a ion was
weigh ed acco ding o he numbe o passenge s on he fligh so ha
he sample could be expanded o ep esen he popula ion.
3. Me hodology
Ai po g ound access sys ems play an impo an ole in ai po
planning and managemen (Tsamboulas and Nikole is, 2008) and
a e inc easingly seen as a majo p oblem wo ldwide. Howe e , ew
s udies ha e been done o da e on he choice o means o ai po
access. T adi ionally, a ele s’choices o modes o anspo a ion
ha e been he objec o mic oeconome ic analyses using disc e e
choice models.
Since he ocus o ea lie s udies has been on ai po g ound
access and possible di e ences in he decisions and beha io o
passenge s acco ding o whe he hey a e business o ou ism/
leisu e a ele s i makes sense o complemen his wo k wi h
a s udy o di e ences in passenge s’ai po g ound access mode
choice beha io depending on whe he hey fly on low cos o
ne wo k ca ie s. LCC g ow h is al e ing he classic dis inc ion
be ween business and ou is passenge s. Fi s ly, he e a e gene -
ally no specific sea s assigned o business passenge s in hese
ai planes making i di ficul o dis inguish business a ele s om
o he passenge s on low-cos ai line fligh s. Despi e his, as he
ma ke sha e o LCCs ises, business passenge s inc easingly choose
hei se ices.
Wi hou ejec ing disc e e demand models, he p oposed
me hodology is amed by s a is ical causal in e ence and based on
an es ima ion o he e ec ha a specific measu e o ac can ha e
on one o mo e ele an a iables. In con as wi h adi ional
analyses, his me hodology allows consis en es ima o s o he
e ec s o he e alua ed measu e o be ob ained (Ro ni zky and
Robins, 1995) by de e mining and isola ing he possible impac o
addi ional con amina ing a iables.
S a ing wi h an N-size andom sample (Table 2)wedefined he
bina y a iable D ha indica es whe he he obse a ion co esponds
oapassenge flyingwi hanLCC(D
i
¼1)o a adi ionalai line(D
i
¼0).
Thus, ou Nobse a ions we e di ided in o N
1
and N
0
obse a ions
(LCC s. adi ional ai line). In ou case, N
1
s ands o he 6247
passenge s who used an LCC, while N
0
ep esen s he emaining
14,136passenge s. Thus, hecondi ion ha s a es ha “N
0
isa leas he
same o de o magni ude o N
1
”is sa isfied (Abadie and Imbens, 2006).
We defined he ou come a iable Y
j
as he decision o use
a specific mode o anspo a ion o a gi en ci y-ai po ans e .
In ou case, we conside ed fi e possible modes o anspo a ion:
axi, public bus, ho el bus, p i a e ca and en ed ca . Using he
po en ial-ou come no a ion o he RCM (Rubin, 1974), he esponse
a iable was gi en as Y
ij
(1) when ideno ed a passenge using an
LCC and as Y
ij
(0) when ico esponded o a passenge flying on
a adi ional ai line. Hence, Y
ij
was equal o
Yij ¼DijYijð1Þþ1DijYijð0Þ(1)
The A e age T ea men E ec (ATE) o he LCC on he selec ed
sample o each o he analyzed modes o anspo a ion was
(Imbens, 2004)
a
j¼EYijð1ÞYijð0Þ¼1
NX
N
i¼1Yijð1ÞYijð0Þ(2)
We also defined a K-dimensional ec o o obse ed co a ia es
as X. A iad was he e o e obse ed o each indi idual (D
ij
,Y
ij
,X
ij
).
The aim was o gua an ee wo condi ions in he e alua ion
p ocess. Fi s ly, he uncon oundedness condi ion (Rosenbaum and
Rubin, 1983), also known as he condi ional independence
assump ion,i.e.D ðYð1Þ;Yð0ÞÞjX. Secondly,we equi edcompliance
wi h he o e lap condi ion, acco ding o which e e y alue o ec o
Xis associa ed wi h a posi i e p obabili y ha allows (Ho z e al.,
2005). The condi ion can be ead as ollows: 0 <PðD¼1jXÞ<1.
To gua an ee bo h condi ions, he e alua ion p ocess con o med o
he ollowing phases (Heckman and Vy lacil, 2005).
3.1. Es ima ion o p opensi y sco e
Fi s ly, we no ed whe he he obse a ions co esponded o
a passenge using an LCC (D
i
¼1) o a legacy ai line (D
i
¼0). We
hen es ima ed he p opensi y sco e, defined by Rosenbaum and
Table 1
Technical da a su ey.
Ai po Alican e Bilbao Se ille Valencia San iago Valladolid Za agoza
Ai po a fic in 2008 9,578,308 4172,901 4,391,794 5779,336 1,917,434 479,716 594,952
In o ma ion
ga he ing
Ques ionnai e A ailable in 12 languages A ailable in
6 languages
A ailable in
5 languages
A ailable in
4 languages
Gene al Depa ing passenge s >15 yea s o age.
Sampling
(be o e weigh ing)
Sample size 2420 3182 4140 4965 3497 1042 1137
Sampling me hod S a ified by a fic segmen s wi h selec ions o fligh s o each ou e and g oups o passenge pa icipan s done by
sys ema ic sampling.
Sampling e o
a
2% 1.7% 1.5% 1.4% 1.5% 2.7% 2.5%
Numbe o wa es 1
Field wo k Time pe iod Sep . 22e28 May 4e10 June 6e12 July 12e18 June 28eJuly 4 June 15e21 June 15e21
Loca ion Depa u e lounges.
Time able MoneSun. 6ame10pm shi s ex ended du ing peak a fic pe iods.
Yea 2006 2007 2006 2006 2006 2005 2006
a
E o ¼Kffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðNnÞ=ðNnÞ
p ffiffiffiffiffiffiffiffiffiffiffi
pq=n
p;whe e: N¼popula ion size; n¼sample size; p¼q¼0.5 complemen a y p obabili ies o he answe o an e en a he poin o
g ea es inde e minacy; k¼pa ame e o he le el o answe o an e en , whe e k¼2 o a 95.45% confidence le el.
2
A subway line connec ing Valencia o i s ai po was opened a ew mon hs a e
he su ey was conduc ed.
J.I. Cas illo-Manzano / Jou nal o Ai T anspo Managemen 16 (2010) 295e298296
Rubin (1983) as he condi ional p obabili y o “pa icipa ing in he
e alua ed measu e,”gi en a ec o Xo obse ed co a ia es.
Di e en bina y esponse models can be used o es ima e he
p opensi y sco e depending on he choice o hypo hesis ega ding
he configu a ion o he Fdis ibu ion unc ion. In his case, we
used he bina y esponse model (e.g. logi o p obi ) ha maxi-
mized he log pseudo-likelihood:
3
(X)¼P(D¼1jX)¼F(
b
X) (3)
whe e
b
is he ec o o pa ame e s associa ed wi h X. He e, X
comp ised he 20 co a ia es p esen ed in Table 2 along wi h hei
desc ip i e s a is ics.
3.2. Es ima ion o a e age ea men e ec
In a second phase, we calcula ed he a e age ea men e ec o
he measu e being e alua ed on he esponse a iable, in ou case,
he p obabili y o a passenge choosing a specific mode o ans-
po a ion o ge o he ai po . The a e age e ec on he selec ed
sample was es ima ed using:
a
¼E½
a
ðXÞ (4)
In ou case, wi h Ybeing a disc e e choice a iable (using one o
he fi e possible modes o anspo a ion), we used a mul inomial
logi model o es ima e he a e age ea men e ec . The e o e,
acco ding o Hi ano and Imbens (2001), he mul inomial logi
p obabili y o mula o a passenge iwhen a pe son chooses a mode
o anspo a ion j o fi e ca ego y ou comes and equency
weigh s is
pij ¼P ðyi¼jÞ¼(1=1þS5
m¼2eðx
i
s
m
Þ;i j ¼1
eðx
0
i
s
m
Þ=1þS5
m¼2eðx
i
s
m
Þ;i js1(5)
whe e:
x’
i
s
m¼
s
m0þ
a
Diþ
s
m1b
3
ðxiÞþ
s
m2b
3
ðxiÞEhb
3
ðxÞiDiþuij
(6)
Table 2
Co a ia es and hei desc ip i e s a is ics.
Va iable Desc ip ion Mean S d. De .
a) Socio-demog aphic ac o s and employmen s a us. Base ca ego y includes he unemployed.
Sex 1 ¼male; 0 ¼ emale. 0.537 0.498
Age 1 ¼30 and unde ; 2 ¼31-49; 3 ¼50e64;
4¼65 and abo e. 1.985 0.825
Non-Spanish 1 ¼non-Spanish passenge ; 0 o he wise. 0.300 0.458
F equen flye Numbe o fligh s aken by passenge in p e ious wel e mon hs:
1¼0fligh s; 2 ¼1-3; 3 ¼4-12; and 4 ¼o e 12 fligh s.
2.433 1.003
Homemake 1 ¼passenge is homemake ; 0 o he wise. 0.032 0.175
Sel -employed 1 ¼passenge is non-sala ied, gene ally sel -employed; 0 o he wise. 0.168 0.374
Sala ied wo ke 1 ¼passenge is sala ied wo ke ; 0 o he wise. 0.584 0.493
S uden 1 ¼passenge is s uden ; 0 o he wise. 0.108 0.311
Re i ed 1 ¼passenge is e i ed; 0 o he wise. 0.082 0.274
b) T ip ca ego y. Base ca ego y includes passenge s isi ing iends and ela i es (VFR) on a ne wo k ca ie .
Low-cos ca ie 1 ¼passenge is flying wi h a low-cos ca ie (LCC); 0 o he wise. 0.306 0.461
Vaca ion 1 ¼ aca ion ip; 0 o he wise. 0.441 0.497
Business 1 ¼business ip; 0 o he wise. 0.310 0.462
Leng h o s ay (LOS) 1 ¼passenge e u ns a e one nigh o ea lie ; 2 ¼ wo nigh s
o a week; 3 ¼se en o 14 days; 4 ¼mo e han wo weeks bu
less han a mon h; 5 ¼mo e han a mon h.
2.408 0.955
c) Social in e ac ion. Base ca ego y includes passenge s a eling wi hou child en and no accompanied o/ om he ai po .
G oup size 1 ¼ a eling alone; 2 ¼ wo people; 3 ¼ h ee o mo e people. 1.684 0.741
Child en 1 ¼ a eling wi h child en; 0 o he wise. 0.078 0.267
Seen o 1 ¼someone sees passenge o a ai po ; 0 o he wise. 0.279 0.448
d) En i onmen . Base ca ego y includes passenge s a eling on wo kdays.
Weekend 1¼su ey aken on Sa u day o Sunday, when axi a e
(fla o egula ) is highe ; 0 o he wise.
0.265 0.441
Ho el 1 ¼passenge depa s om a ho el, boa ding house, o o he paid
accommoda ion; 0 o he wise.
0.214 0.410
Home 1 ¼passenge depa s om own p ima y o seconda y home; 0 o he wise. 0.523 0.499
F iends o amily 1 ¼passenge depa s om home o iends o ela i es; 0 o he wise. 0.142 0.349
Table 3
P obi es ima ion o he p opensi y sco e.
Co a ia e Coe ficien Co a ia e Coe ficien
Sex 0.031 Business 0.502***
Age 0.039 Leng h o s ay 0.077***
Non-Spanish 0.885*** G oup Size 0.063***
F equen flye 0.029*** Child en 0.110
Homemake 0.089 Seen o 0.004
Sel -employed 0.028 Weekend 0.092**
Sala ied wo ke 0.027 Ho el 0.183***
S uden 0.205** Home 0.035
Re i ed 0.018 F iends o amily 0.155***
Vaca ion 0.038 Cons an 0.907***
No. o obse a ions
(be o e weigh ing)
19930
Log pseudo-likelihood 14640197
Pseudo R
2
0.137
Wald Chi
2
wi hou clus e s
(p- alue)
4638753.72 (0.000)
No e: *, **, o *** indica e coe ficien significance a he 10%, 5%, and 1% le els
calcula ed om s anda d e o s obus o he e oskedas ici y and clus e ed by
ai po o o igin.
Table 4
Mul inomial es ima ion o ele an e ec s.
LCC (D
i
) Cons an b
3
ðx
i
Þð
b
3
ðx
i
ÞE½b
3
ðxÞÞD
i
Public Bus 0.227* 1.825*** 0.794 1.154***
Ho el Bus 0.017 3.290*** 4.584*** 2.912***
Ren -a- ca 0.169* 1.251 1.966* 0.178
Taxi 0.218*** 0.080 0.795* 0.428
No e: *,**,o *** indica ecoe ficien significancea he10%,5%, and1%le els,calcula ed
om s anda d e o s obus o he e oskedas ici y and clus e ed by ai po o o igin.
J.I. Cas illo-Manzano / Jou nal o Ai T anspo Managemen 16 (2010) 295e298 297
3.3. The ma ginal e ec o b
a
As in bina y ou come models, only he sign o he coe ficien
is di ec ly in e p e ed in mul inomial models. Thus, a posi i e
coe ficien in he mul inomial logi means ha , as he eg esso
inc eases, al e na i e jis mo e likely o be chosen han al e -
na i e k. In o de o make he eading o he esul s easie , he
odds a ios o ela i e- isk a ios o he bina y a iable D
i
(using
an LCC) a e also conside ed. Following Came on and T i edi
(2009), he ela i e p obabili y o odds a io o choosing al e -
na i e j a he han al e na i e 1, also called he base ou come, is
gi en by
P ðyi¼jÞ
P ðyi¼1Þ¼ex
0
i
s
j
(7)
Howe e , mul inomial logi coe ficien s and odds a ios only
allow he subs i u abili y ela ionship be ween pai s o op ions o
be s udied, ha is, he ela ionship be ween each op ion and he
base ca ego y, which in ou case is he p i a e ca . In o de o
o e come his ocus on pai wise opposi ions, we calcula ed he
ma ginal e ec s o D
i
ac oss all conside ed op ions. This enabled
he di ec subs i u abili y ela ionship, should one exis , o be
ob ained be ween he fi e modes o anspo a ion o LCC
passenge s (D
i
¼1), as opposed o he con ol g oup including all
passenge s flying wi h legacy ai lines. Acco ding o Came on and
T i edi (2009), he ma ginal e ec s a he mean (MEMs) o he
mul inomial logi model a e:
d
pij
d
Di
¼pij
a
j
a
i(8)
a
i¼Xpil
a
l;
whe e is a p obabili y-weigh ed a e age o
a
l
.
4. Resul s
Table 3 summa izes he esul s o p opensi y sco e es ima ion
(Model 3), in he con ex o he 20 co a ia es in Table 2. A p obi
specifica ion was op ed o as i maximized he log pseudo-
likelihood.
A mul inomial logi specifica ion (Models 5 and 6) was hen
used o es ima e he a e age ea men e ec s. The esul s a e
p esen ed in Table 4.
Finally, he ma ginal e ec a he mean o an LCC passenge is
es ima ed by applying Model 8 (Table 5).
5. Conclusions
The esul s show ha being a LLC passenge educes he p ob-
abili y o choosing a axi o go o he ai po by 5.85%, bu inc eases
he likelihood o a LLC passenge and ha o choosing a en ed ca
o a public mode o anspo a ion by abou 4% and 2%, espec i ely.
The ou come canno be explained by ac o s such as LCC passenge s
ha ing a lowe income le el han he ne wo k ca ie passenge s
because he sco es a e co ec ed using income le el p oxy
a iables. I would he e o e be mo e app op ia e o speak o
passenge s who a e inc easingly p ice-conscious. The e could,
howe e , be a h eshold ype p ice e ec on mode choice i he e is
a s ong psychological shock in paying mo e o a 10 km axi ide
han o a 1000 km ai plane fligh . The a e age ci y-ai po dis ance
o he se en ai po s in ou sample is 9.86 km, wi h a s anda d
de ia ion o only 1.21. Gene ally-speaking, 10 km is a easonable
ci y-ai po dis ance measu e o Spanish seconda y ai po s.
Acknowledgemen s
The au ho would like o exp ess his g a i ude o AENA o i s
suppo . The au ho is also g a e ul o Kenne h Bu on, Guido
Imbens and he anonymous e iewe s o e y help ul commen s.
Re e ences
Abadie, A., Imbens, G.W., 2006. La ge sample p ope ies o ma ching es ima o s o
a e age ea men e ec s. Econome ica 74, 235e267.
Ba e , S.D., 2004. How do he demands o ai po se ices di e be ween ull-
se ice ca ie s and low-cos ca ie s? Jou nal o Ai T anspo Managemen 10,
33e39.
Came on, A.C., T i edi, P.K., 2009. Mic oeconome ics Using S a a. S a a Publica ion,
College S a ion.
Cas illo-Manzano, J.I., López Valpues a, L. The decline o he adi ional a el agen
mode. T anspo a ion Resea ch E in p ess, doi:10.1016/j. e.2009.12.009.
De Neu ille, R., 2006. Planning ai po access in an e a o low-cos ai lines. Jou nal
o he Ame ican Planning Associa ion 72, 347e356.
Eche a ne, R., 2008. The impac o a ac ing low cos ca ie s o ai po s. In:
G aham, A., Papa heodou u, A., Fo sy h, P. (Eds.), A ia ion and Tou ism. Ashga e
Publishing, Hampshi e.
F ancis, G., Humph eys, I., Ison, S., 2004. Ai po s’pe spec i es on he g ow h o
low-cos ai lines and he emodelling o he ai po eai line ela ionship.
Tou ism Managemen 25, 507e514.
G aham, A., 2008. Managing Ai po s: An In e na ional Pe spec i e, hi d ed.
Bu e wo h Heinemann-Else ie , Ox o d.
Heckman, J.J., Vy lacil, E., 2005. S uc u al equa ions, ea men e ec s, and
econome ic policy e alua ion. Econome ica 73, 669e738.
Hi ano, K., Imbens, G.W., 2001. Es ima ion o causal e ec s using p opensi y sco e
weigh ing: an applica ion o da a on igh hea ca he e iza ion. Heal h Se ices
& Ou comes Resea ch Me hodology 2, 259e278.
Ho z, V.J., Imbens, G.W., Mo ime , J.H., 2005. P edic ing he e ficacy o u u e
aining p og ams using pas expe iences a o he loca ions. Jou nal o Econo-
me ics and S a is ics 125, 241e270.
Humph eys, I., Ison, S., 2005. Changing ai po employee a el beha iou : he ole
o ai po su ace access s a egies. T anspo Policy 12, 1e9.
Imbens, G.W., 2004. Nonpa ame ic es ima ion o a e age ea men e ec s unde
exogenei y: a e iew. Re iew o Economics and S a is ics 86, 4e29.
O’Connell, J.F., Williams, G., 2005. Passenge s’pe cep ions oflow cos ai lines and
ull se ice ca ie s: a case s udy in ol ing Ryanai , Ae Lingus, Ai Asia and
Malaysia Ai lines. Jou nal o Ai T anspo Managemen 11, 259e272.
Pels, E., Njego an, N., Beh ens, C., 2009. Low-cos ai lines and ai po compe i ion.
T anspo a ion Resea ch E 45, 335e344.
Reynolds-Feighan, A., 2001. T a fic dis ibu ion in low-cos and ull-se ice ca ie
ne wo ks in he US ai anspo a ion ma ke . Jou nal o Ai T anspo
Managemen 7, 265e275.
Rosenbaum, P.R., Rubin, D.B., 1983. The cen al ole o he p opensi y sco e in
obse a ional s udies o causal e ec s. Biome ika 70, 41e55.
Ro ni zky, A., Robins, J., 1995. Semipa ame ic eg ession es ima ion in he p esence
o dependen censo ing. Biome ika 82, 805e820.
Rubin, D.B., 1974. Es ima ing causal e ec s o ea men s in andomized and
non- andomized s udies. Jou nal o Educa ional Psychology 66, 688e701.
Tsamboulas, D.A., Nikole is, A., 2008. Passenge s’willingness o pay o ai po
g ound access ime sa ings. T anspo a ion Resea ch A 42, 1274e1282.
Table 5
Ma ginal e ec o b
a
:
P i a e Ca 0.0003 /¼0.03%
Public Bus 0.0185 /
D
1.85%
Ho el Bus 0.0010 /¼0.10%
Ren -a-Ca 0.0385 /
D
3.85%
Taxi 0.0585 /V5.85%
J.I. Cas illo-Manzano / Jou nal o Ai T anspo Managemen 16 (2010) 295e298298