Observed patterns of free-floating car-sharing use

Fab a, Na alia; Pin assilgo, Ca a ina; Souza, Ma eus
A icle
Obse ed pa e ns o ee- loa ing ca -sha ing use
SERIEs - Jou nal o he Spanish Economic Associa ion
P o ided in Coope a ion wi h:
Spanish Economic Associa ion
Sugges ed Ci a ion: Fab a, Na alia; Pin assilgo, Ca a ina; Souza, Ma eus (2024) : Obse ed pa e ns o
ee- loa ing ca -sha ing use, SERIEs - Jou nal o he Spanish Economic Associa ion, ISSN 1869-4195,
Sp inge , Heidelbe g, Vol. 15, Iss. 3, pp. 259-297,
h ps://doi.o g/10.1007/s13209-024-00298-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/326957
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/
SERIEs (2024) 15:259–297
h ps://doi.o g/10.1007/s13209-024-00298-2
ORIGINAL ARTICLE
Obse ed pa e ns o ee- loa ing ca -sha ing use
Na alia Fab a1·Ca a ina Pin assilgo2·Ma eus Souza3
Recei ed: 12 May 2023 / Accep ed: 15 Ap il 2024 / Published online: 9 June 2024
© The Au ho (s) 2024
Abs ac
F ee- loa ing ca -sha ing (FFCS) se ices allow use s o en elec ic ehicles by he
minu e wi hou es ic ions on pick-up o d op-o loca ions wi hin he se ice a ea
o he en al company. Beyond enla ging he choice se o mobili y op ions, FFCS
may educe conges ion and emissions in ci ies, depending on he se ice’s usage and
subs i u ion pa e ns. In his pape , we shed ligh on his by analyzing he uni e se
o FFCS ips conduc ed h ough a leading company in Mad id du ing 2019. We
co ela e FFCS usage pa e ns wi h da a on a ic condi ions, demog aphics, and
public ansi a ailabili y ac oss he ci y. We ind complemen a i ies be ween FFCS
and public anspo in middle-income a eas wi h sca ce public anspo op ions.
Mo eo e , we ind ha he use o FFCS peaks ea lie han o e all a ic and is b oadly
used du ing he summe mon hs. This sugges s ha FFCS may ha e smoo hed oad
a ic in Mad id, con ibu ing o a educ ion in o e all conges ion.
Keywo ds Ca -sha ing ·Sha ed mobili y ·Road conges ion ·Elec ic ehicles
JEL Classi ica ion R41 ·Q52
1 In oduc ion
Road anspo is associa ed wi h se e al ypes o nega i e ex e nali ies, including a -
ic acciden s (Edlin and Ka aca-Mandic 2006), conges ion (ECA 2019), and pollu ion,
such as ine pa icula e ma e (PM2.5) (ECA 2019; Böhm e al. 2022; Li and Managi
2021). These pollu an s, in u n, ha e been linked o inc eased mo ali y among adul s
The eplica ion ma e ial o he s udy is a ailable a h ps://doi.o g/10.5281/zenodo.7707902.
BMa eus Souza
[email p o ec ed]
1Depa men o Economics, Uni e sidad Ca los III de Mad id, Mad id, Spain
2Depa men o Economics, Uni e si y o Wa wick, Wa wick, UK
3Depa men o Economics, Uni e si y o Mannheim, Mannheim, Ge many
123
260 SERIEs (2024) 15:259–297
and in an s (Cu ie and Neidell 2005; De yugina e al. 2019; Geh si z 2017; Lelie eld
e al. 2015) h ough a a ie y o mechanisms, anging om espi a o y (Wu e al. 2018;
Wei and Tang 2018) o ca dio ascula diseases (Al-Kindi e al. 2020; Sanidas e al.
2017). The anspo sec o also accoun s o app oxima ely 23% o ene gy- ela ed
CO2 emissions globally (IEA 2022). Addi ionally, ca owne ship in ci ies may lead o
ine icien land use. On a e age, p i a e ca s a e pa ked 96% o he ime, o 23 h pe
day, hus ha ing ela i ely low usage a es (Nagle 2021). This ine iciency is exac-
e ba ed by he inc easing single-occupancy ehicle commu ing a es (Machado e al.
2018).
Wo ldwide, a b ead h o policies a e being implemen ed o mi iga e hese ex e nal-
i ies, including conges ion p icing (Anas and Lindsey 2011; Bö jesson e al. 2012),
subsidies o elec ic ehicle pu chases (Rapson and Muehlegge 2023), o low-
emission zones in ci ies (Galdon-Sanchez e al. 2022; Bö jesson e al. 2021), among
o he s. Reliance on hese policies is expec ed o g ow as coun ies seek o comply
wi h hei commi men s o achie e ne -ze o emissions by 2050 (Eu opean Commis-
sion 2020; U.S. DoS and EOP 2021). To imp o e hei e ec i eness, we s ill need
a deepe unde s anding o he bene i s and cos s o he a ious policy op ions. Fo
equi y pu poses, we mus also unde s and who bene i s and who is ha med by hose
policies. Signi ican headway has been made in some on s (Pa y 2002), ye li le is
known abou he po en ial impac s o some apidly p oli e a ing echnologies. This
s udy con ibu es in his di ec ion by analyzing he cha ac e is ics o use s and usage
pa e ns o one such echnology, ee- loa ing ca -sha ing (FFCS).
In pa hanks o ad ances in communica ions and acking echnologies, as well
as he widesp ead adop ion o sma phones, companies a e now able o en ca s by
he minu e, a a es anging om app oxima ely 0.19 e/min o 0.31 e/min.1In ea lie
s ages o he deploymen o hese se ices, use s we e equi ed o d op o he ehicles
a speci ic loca ions o a cha ging s a ions. Howe e , in ecen yea s, he e has been
a apid g ow h o demand o FFCS se ices allowing use s o pick up he ehicle
and d op i o a any place wi hin he se ice a ea o he en al company (Ampudia-
Renuncio e al. 2020a). Impo an ly, he companies p o iding hese se ices o en do
so wi h elec ic ehicles.
The e o e, FFCS can po en ially a ec conges ion and en i onmen al ex e nali ies
in opposing di ec ions, depending on which mode o anspo i is subs i u ing. On he
one hand, FFCS may alle ia e u ban conges ion and pollu ion h ough he ex ensi e
and in ensi e ma gins, i.e., i hey educe ca owne ship (ca shedding) o he numbe
o ips wi h p i a e pollu ing ca s (Bucsky and Juhász 2022). This possibili y is
suppo ed by he ac ha sha ed ehicles a e mos ly elec ic and ha e much highe
u iliza ion a es han p i a e ehicles (Habibi e al. 2017). On he o he hand, ca -
sha ing may inc ease conges ion i used as a subs i u e o public anspo , walking,
o cycling (Machado e al. 2018). Whe he one e ec o he o he domina es is an
empi ical ques ion ela ed o he usage pa e ns o FFCS.
As a s a ing poin o shedding ligh on he wel a e e ec s o FFCS, his s udy
p o ides a comp ehensi e desc ip ion o ca -sha ing usage pa e ns, seasonali y o
1Ra es published on he websi es o he main FFCS companies ope a ing in Mad id (Emo , Sha eNow,
WiBLE, Zi y), as o Decembe 2023.
123
SERIEs (2024) 15:259–297 261
use, use s’ incomes, and p ima y usage pu poses. In pa icula , we in es iga e he
co ela ions be ween he a iables o in e es wi hou claiming causa ion. While mos
o he ea ly li e a u e on ca -sha ing has ocused on Ge man ci ies,2o elied mos ly
on su ey da a (Ami nazmia sha and Diana 2022; Mülle e al. 2017; Schmölle e al.
2015; Kopp e al. 2015), in his pape , we p o ide new e idence based on housands
o obse ed ips in he ci y o Mad id. Acco ding o su eys ac oss se e al Eu opean
and No h-Ame ican ci ies (Sp ei e al. 2019; Habibi e al. 2017), Mad id had he
highes FFCS u iliza ion a e (be ween 17 and 21.6% o he ime when ca s can be
po en ially d i en pe day). Addi ionally, oge he wi h Ams e dam, Mad id is one o
he only ci ies in Eu ope whe e he FFCS lee is ully elec ic (Ampudia-Renuncio
e al. 2020a).
One o he s eng hs o his s udy is ha i combines a ich se o da a sou ces.
Impo an ly, we ha e access o a unique p op ie a y da abase o he uni e se o ca -
sha ing ips (mo e han 1,500,000) ca ied ou in 2019 by one o he leading FFCS
companies ope a ing in Mad id.3To co ela e ca -sha ing usage pa e ns wi h socio-
demog aphic cha ac e is ics, we use neighbo hood-le el da a on popula ion, income,
ca owne ship, public anspo ne wo k, and pa king a ailabili y da a. We also analyze
daily oad a ic da a, compa ing i agains he seasonali y o ca -sha ing ips.
Ou mainda ase con ainsp ecisegeo- e e encedda aon heca -sha ing ip iming,
o igin, and des ina ion bu lacks in o ma ion on he cus ome s’ demog aphics. In
pa icula ,whilewecan ackuse s h oughuniqueiden i ie s,wedono obse ewhe e
hey li e. Hence, we de eloped a s a egy o impu e use s’ neighbo hood o esidence
based on epea ed ips om and o he same des ina ion du ing commu ing imes.
Ou app oach cons i u es an imp o emen ela i e o he p e ious li e a u e, which
was es ic ed o analyzing he cha ac e is ics o he place o o igin/des ina ion o
ips (Schmölle e al. 2015; Ampudia-Renuncio e al. 2020a) and he e o e abs ac ed
away om use s’ place o esidence. Besides impu ing membe s’ esidences, we also
exploi he hou ly equency o epea ed o igins and des ina ions o in e he ips’
pu poses, as ei he commu e o leisu e. These s a egies allow us o ma ch he ip da a
and he socio-demog aphics o s udy he de e minan s o ca -sha ing usage and he
di e ences ac oss income g oups.
We ob ain meaning ul insigh s in o he usage pa e ns and cha ac e is ics o he
FFCS use s. Fi s , we ind ha ca -sha ing use s mos ly li e in middle- and high-
income neighbo hoods. This inding is pa ly d i en by he se ice a ea es ic ion
om he FFCS company (Fig. 1), e ealing ha FFCS companies expec o ha e
g ea e demand in highe -income a eas. Hence, lowe -income neighbo hoods do no
ha e access o FFCS, which is a s ong o ce owa d a posi i e co ela ion be ween
neighbo hood income and ca -sha ing use. Usage in ensi y in he co e ed neighbo -
hoodsp o idesa di e en pic u e, as he mos loyal use s end o li einmiddle-income
2Some ea lie s udies wi hin Mad id exis (e.g., Ampudia-Renuncio e al. 2020b), howe e , wi h a sho e
pe iod o analysis and smalle co e age a ea. The da a used in hose s udies we e ob ained om he
companies’ websi es.
3Du ing he pe iod ha we analyze, he company cha ged ixed p ices pe minu e, such ha o al cos s
depended exclusi ely on he ip du a ion. The e was no p ice disc imina ion ac oss use s and no eba es
o equen use s.
123
262 SERIEs (2024) 15:259–297
Fig. 1 Co e age a ea o he
ca -sha ing se ice, in 2019.
No es The igu e p esen s he
s udy a ea o his pape and he
co e age a ea o he ca -sha ing
se ice, i.e., he a ea in which
cus ome s can s a o end a ip.
Use s a e allowed o d i e
ou side he co e age a ea and
lea e he ca in s andby mode i
hey need o make a s op ou side
he co e age a ea. All e i o ial
uni s ep esen ed a e
neighbo hoods inside Mad id
Municipali y, excep o he
da ke shaded uni s:
Alcobendas, Coslada, Pozuelo
de Ala cón, and San Sebas ian
de los Reyes municipali ies
Ca sha ing co e age a ea
Neighbo hoods inside Mad id Municipali y
Ou side Mad id Municipali y
neighbo hoods, i.e., condi ionally on ha ing access o FFCS, he e is a nega i e co -
ela ion be ween neighbo hood income and he equency o ca -sha ing use.
Based on he iming o use, we also p o ide e idence on he pu poses o ca -sha ing
ips. We a e able o classi y almos 536 housand commu ing ips and 386 housand
leisu e ips. This allows us o explo e he e ogenei y along hese dimensions. Fo
example, we ind ha income isposi i ely co ela ed wi h commu ing wi h ca -sha ing
bu nega i ely co ela ed wi h leisu e ips.
Since mos loyal cus ome s li e in neighbo hoods wi h high ca owne ship a es
and ewe public anspo op ions, ou analysis sugges s ha hese cus ome s use he
se ice o complemen public anspo . I ca -sha ing membe s a e ca owne s who
p e e o lea e hei ca a home, o e en e en ually sell i , FFCS would imply an
o e all educ ion o ehicles in he ci y and, consequen ly, educed use o pa king
space, dec eased oad a ic, and imp o ed ai quali y.
Addi ionally, we ind ha he seasonali y o ca -sha ing ips closely ma ches he
ends o o he mo o ized ehicles, al hough ca -sha ing usage peaks sligh ly ea -
lie . Mo eo e , while oad a ic dec eases du ing he summe mon hs, he numbe
o ca -sha ing ips inc eases in he mon hs o July and Sep embe . These pa e ns
sugges ha FFCS con ibu es o smoo hing o e all a ic and, hus, educing con-
ges ion in he ci y. Impo an ly, du ing summe mon hs, usage by he highes -income
cus ome s dec eases subs an ially. We see a highe usage o he ca -sha ing se ice
du ing wo kdays, pa icula ly on F idays. Du ing weekends, ips a e less equen bu
longe . O e all, ca -sha ing ips a e gene ally sho , bo h ega ding he du a ion and
dis ance.
Finally, we implemen g a i y- ype eg ession speci ica ions o unco e po en ial
ic ions o acili a o s o adop ing ca -sha ing ac oss he ci y. This analysis is pe -
o med a he le el o Mad id’s “ anspo zones” (“zonas de anspo e,” in Spanish),
which a e smalle han neighbo hoods bu la ge han census ac s. We obse e nea ly
500 anspo zoneswi hin heca -sha ingcompany’s a ea, co esponding o abou 250
123

SERIEs (2024) 15:259–297 263
housand o igin/des ina ion pai s (o dyads). Ou esul s sugges ha he likelihood o
aking a ca -sha ing ip dec eases wi h he dis ance be ween o igin and des ina ion.
This could be due o he high cos s associa ed wi h long ca -sha ing ips. Con e sely,
he likelihood o ca -sha ing is highe o o igin/des ina ion dyads ha a e poo ly con-
nec ed by public ansi (i.e., dyads wi h long expec ed a el imes by public ansi ).
2 Da a sou ces
Table 1p o ides a summa y o he main a iables used in his s udy, including sample
sizes and da a sou ces. We desc ibe each da a sou ce in de ail below.
2.1 F ee- loa ing ca -sha ing ips
We use he uni e se o ca -sha ing ips made h ough one o he leading companies
ope a ing in Mad id du ing 2019. This da abase, which he company p o ided di ec ly
o he au ho s, con ains unique ehicle and cus ome IDs, he s a ing and ending ime
and loca ion o each ip, and he du a ion and dis ance a eled. Du ing he analyzed
sample pe iod, he company had a lee o 656 ope a ing elec ic ehicles, which
i ually emained cons an h oughou he yea 2019 (Fig. 11). We excluded ips
by company employees and ocused only on ips made by cus ome s. We exclude
ips ha las less han 2 minu es o mo e han 180 minu es. We u he excluded
obse a ions o which he same membe used he se ice mo e han 6 imes on he
same day.4
2.2 Demog aphics
In e ms o demog aphics, we p ima ily use da a on popula ion and income, ob ained
om he Mad id Ci y Council (AM 2018c) and he Na ional S a is ics Ins i u e (INE
2018a;b). This en ails combining da a a di e en spa ial agg ega ion le els. Fo he
ci yo Mad id,weob ainneighbo hood-le eldemog aphics o i s131neighbo hoods.
We also collec municipali y-le el da a o he o he municipali ies ha a e pa o he
Communi y o Mad id (me opoli an a ea). We also collec da a on o he demog aphic
a iables, such as gende and age. Howe e , we ind no signi ican he e ogenei y along
hose dimensions, such ha hey a e omi ed om he emainde o his pape . No e
ha we do no di ec ly obse e demog aphics o he ca -sha ing use s, which limi s
he scope o ou he e ogenei y analysis. Rega dless, in Sec . 3we p esen a p oposal
o impu ing demog aphics.5
4These a e ou lie cases ha may ep esen bugs in he eco ding o ip du a ions.
5P e ious s udies, mos ly elying on su ey da a, ound ha ca -sha ing use s a e p edomina ely male,
high-income, and young (Ami nazmia sha and Diana 2022; Vine and Polak 2020). Simila ly, Schmölle
e al. (2015) a gue ha ca -sha ing se ices can be especially appealing o young adul s, since owning a
p i a e ca is becoming less aluable now, compa ed o he pas .
123
264 SERIEs (2024) 15:259–297
Table 1 Desc ip i e s a is ics
Va iable Obs Mean S d.
De
Min Max Sou ce Yea
Indi idual-le el da a
Full sample o FFCS
ips
Numbe o ips 1,474,287 T ip
da a
2019
Numbe o cus ome s 143,315 T ip
da a
2019
Cus ome loyal y
No. o days used in
2019
143,315 8.224 14.239 1 344 T ip
da a
2019
No. o imes used in
2019
143,315 10.287 20.306 1 726 T ip
da a
2019
Pe cen age o days
used in 20191(%)
143,315 4.027 7.212 .274 100 T ip
da a
2019
Weigh ed no. o days
used in 20192143,315 14.699 26.323 1 365 T ip
da a
2019
Sub-sample o ips wi h iden i ied esidence
Numbe o iden i ied
neighbo hoods
95 T ip
da a
2019
Numbe o ips 511,070 T ip
da a
2019
Numbe o cus ome s 12,143 T ip
da a
2019
Annual ne income
(e)312,143 20,994 5,247 8,758 31,239 AM,
INE
2018
Cus ome loyal y
No. o days used in
2019
12,143 31.925 28.209 1 344 T ip
da a
2019
No. o imes used in
2019
12,143 42.088 43.256 1 724 T ip
da a
2019
Pe cen age o days
used in 20191(%)
12,143 10.74 9.45 0.28 94.25 T ip
da a
2019
Weigh ed no. o days
used in 2019212,143 39.21 34.48 1.01 344 T ip
da a
2019
Su ey sample o EDM 4
Numbe o ca s pe
household
28,922 0.96 0.80 0 5 EDM 2018
Inadequa e public
anspo (%)536,030 21.34 37.85 0 100 EDM 2018
% indi iduals wi h
d i ing license
119,423 0.64 0.48 0 1 EDM 2018
123
SERIEs (2024) 15:259–297 265
Table 1 (con inued)
Va iable Obs Mean S d.
De
Min Max Sou ce Yea
Neighbo hood/municipali y a e ages
Neighbo hood
popula ion size
135 26,589 17,971 1,176 116,037 AM,
INE
2018
A e age annual ne
income pe capi a
(e)
135 17,425 6,318 6,712 32,242 AM,
INE
2018
A e age age 135 43.60 3.58 32 48.91 AM,
INE
2018
Pe cen age o emales
(%)
135 53.31 1.86 48.30 58.22 AM,
INE
2018
A e age numbe o
ca s pe household
135 0.95 0.31 0.42 1.91 EDM 2018
% indi iduals wi h
d i ing license
135 59.58 8.83 32.13 78.71 EDM 2018
Inadequa e public
anspo (%)5135 18.74 8.45 3.44 52.17 EDM 2018
No. o me o s a ions
pe km2135 1.51 2.26 0 15.73 CRTM 2018
No. o ain s a ions
pe km2135 0.15 0.40 0 2.58 CRTM 2018
No. o bus s a ions pe
km2135 24.6 12.05 0.12 53.16 CRTM 2018
No. o egula ed
pa king slo s pe ca 648 0.36 0.35 0.12 2.55 AM 2018
Neighbo hoods in he s udy a ea
Ca sha ing membe
densi y (%)
95 0.56 0.58 0.01 3.66 T ip
da a
2019
Neighbo hood
popula ion size
95 27,471 18,994 1,176 116,037 AM,
INE
2018
A e age annual ne
income pe capi a
(e)
95 19,220 5,783 8,758 31,239 AM,
INE
2018
A e age age 95 44.04 3.40 32.5 48.91 AM,
INE
2018
Pe cen age o emales
(%)
95 53.66 1.92 48.30 58.22 AM,
INE
2018
A e age numbe o
ca s pe household
95 0.96 0.29 0.42 1.71 EDM 2018
% indi iduals wi h
d i ing license
95 61.72 8.04 32.13 78.71 EDM 2018
Inadequa e public
anspo (%)595 17.65 8.33 5.69 52.17 EDM 2018
123
266 SERIEs (2024) 15:259–297
Table 1 (con inued)
Va iable Obs Mean S d.
De
Min Max Sou ce Yea
No. o me o s a ions
pe km295 1.74 2.08 0 13.07 CRTM 2018
No. o ain s a ions
pe km295 0.11 0.37 0 2.58 CRTM 2018
No. o bus s a ions pe
km295 26.2 11.2 1.7 53.16 CRTM 2018
No. o egula ed
pa king slo s pe ca 648 0.36 0.35 0.12 2.55 AM 2018
1Compu ed as he numbe o days he cus ome used he ca -sha ing se ice di ided by he o al numbe o
days le in 2019 since he i s day hey used he se ice. 2Compu ed as he pe cen age o days a cus ome
used he ca -sha ing se ice, mul iplied by o al numbe o days in a yea . 3Assuming all esiden s in
a ce ain neighbo hood ha e an annual ne income equal o he annual income pe capi a o ha same
neighbo hood. 4Su ey da a om EDM 2018, es ic ed o ou ou municipali ies o analysis (Mad id,
Alcobendas, Coslada, Pozuelo de Ala cón, and San Sebas ian de los Reyes). The obse a ion uni s o
he a iable “% indi iduals wi h d i ing license” co espond o an indi idual, o “numbe o ca s pe
household” o a household, and o “inadequa e public anspo ” o a ip made by ca by one o he su ey
esponden s. 5Using su ey da a om EDM 2018, we c ea ed a measu e o adequacy o he public anspo
ne wo k. I measu es he pe cen age o ips, ou o o al ips made by ca , in which he esponden s o
EDM a i m using his a el mode because hey ha e no public anspo al e na i e. 6Numbe o slo s
om he ci y’s egula ed pa king se ice (Se icio de Es acionamien o Regulado, SER), no malized by he
numbe o egis e ed ehicles. In 2018, only 48 neighbo hoods in he Municipali y o Mad id we e co e ed
by he SER
2.3 Mobili y- ela ed da a
Ano he key da a sou ce is Mad id’s Mobili y Su ey (Encues a Domicilia ia de
Mo ilidad,CRTM2018b). This is a ecall su ey o o e 58 housand households,
aiming o assess mobili y pa e ns in he Communi y o Mad id du ing wo king days.6
The mobili y su ey includes in o ma ion on he numbe o ca s pe household, which
we use o p oxy ca owne ship in each neighbo hood. Mo eo e , using ip da a om
hemobili y su ey, webuilda measu eo he adequacyo hepublic anspo ne wo k
o assess he deg ee o complemen a y and subs i u abili y be ween he ca -sha ing
se ice and public anspo (see Sec . 3 o u he de ails).
We also downloaded da a om Mad id’s Regional T anspo Conso ium (Conso -
cio Regional de T anspo es de Mad id,CRTM2018a) con aining he geog aphical
coo dina e o each me o, ain, and bus s a ion in he Mad id Au onomous Com-
muni y, o measu e neighbo hood-le el densi y o public anspo s a ions in Mad id
and neighbo ing municipali ies. The Conso ium also p o ides de ailed geo e e enced
6Responden s we e asked o desc ibe all o hei ips om he day p io o he in e iew. Da a a e
ep esen a i e o ips aken om Monday o Thu sday.
123
SERIEs (2024) 15:259–297 273
ha s a o end in A ocha a e mul i-modal. Fo example, use s migh ake a ain o
A ocha and, om he e, ake he ca -sha ing se ice o hei inal des ina ion. Thus,
as explained in Sec . 3, we exclude A ocha om ou analyses. Likewise, we exclude
he Cas illa and Ae opue o neighbo hoods, which a e almos en i ely occupied by he
Chama ´ın ain s a ion and Mad id-Ba ajas Ai po . Acco ding o ou me hodology,
he Cas illa and Ae opue o neighbo hoods would be he 7 h and 5 h a eas wi h he
highes membe densi y (1.52 and 1.66%, espec i ely).
Figu e 2panel (b) shows ha mo e loyal ca -sha ing membe s li e in he pe iphe y
o ou s udy a ea, pa icula ly in he ou municipali ies ou side Mad id and he sou h-
eas e n neighbo hoods. A compa ison o panels (a) and (b) e eals ha neighbo hoods
wi h a highe pe cen age o ca -sha ing cus ome s ha e a lowe loyal y o he se ice,
which migh sugges ha hose cus ome s use he se ice o dis inc pu poses. The
sec ions below explo e his u he .
4.2 Income dis ibu ion
P io li e a u e sugges s ha FFCS in Mad id is abou h ee imes mo e expensi e
han public anspo when conside ing single- ip a es (Ampudia-Renuncio e al.
2020b). The a ailabili y o mon hly passes o public anspo a ion inc eases he cos
di e ence. Howe e , ca -sha ing is cheape han ide-hailing and, depending on he
usage in ensi y, ca -sha ing can be a mo e a o dable op ion han owning a ca . The
implica ion o hese cos di e ences is ha ca -sha ing migh no be accessible o
a ac i e o some income g oups while appealing o o he s.
In his sec ion, we p o ide insigh s in o how income ela es o ca -sha ing usage
in he ci y o Mad id. Ha ing iden i ied he use s’ esidence loca ion h ough he ca -
sha ing usage pa e ns, we hen impu e hei income based on neighbo hood-le el
da a (see Table 1). This shows ha he a e age annual ne income o ca -sha ing
use s in ou sample is close o e20 housand, which is subs an ially highe han
he a e age income pe capi a o he en i e municipali y o Mad id (e16,700) (INE
2018b).13This is in line wi h he ac ha he ca -sha ing se ice a ea (Fig. 1) excludes
some o he lowes -income neighbo hoods in he ci y. Ou indings a e consis en wi h
p io li e a u e showing ha ca -sha ing use s end o ha e an abo e-a e age income
(Ami nazmia sha and Diana 2022; Caul ield and Kehoe 2021).
Al hough ou da a sugges he exis ence o a posi i e co ela ion be ween ca -
sha ing membe densi y and income, we would expec o ind a nega i e co ela ion
be ween membe loyal y and income, gi en ha , as discussed in Sec . 4.1, neighbo -
hoods wi h a highe pe cen age o ca -sha ing cus ome s seem o ha e a lowe loyal y
o he se ice. We o mally es hese ela ionships wi h eg essions as ollows:
Yi=α+4
g=2βg1i∈income qua ile g+εi(2)
whe e Yiis he ou come o in e es (cus ome densi y o loyal y) o he uni o analysis
i; he uni o analysis is he neighbo hood o he densi y a iable, and he cus ome
13 The a e age annual income pe capi a o he ou municipali ies o he s udy a ea (Mad id, Alcobendas,
Coslada, Pozuelo de Ala con, and San Sebas ian de los Reyes) was e17,425.
123

274 SERIEs (2024) 15:259–297
o he loyal y a iable; αis a eg ession cons an ; he indica o s 1[i∈income qua ile
g] a e equal o one when ibelongs o a gi en income qua ile, ze o o he wise; no e
ha he 1s qua ile (income om e8000–e14,500) is he omi ed compa ison g oup;
and εiis an idiosync a ic e o e m. The coe icien s βg hus cap u e how densi y o
loyal y changes depending on he income qua iles o neighbo hoods o cus ome s.
Reg ession es ima es om Eq. (2) a e p esen ed in Table 4. Column (1) shows
ha he highes -qua ile neighbo hoods ha e a membe densi y o abou 0.72 p.p.
highe han he lowes -qua ile neighbo hoods. Con e sely, column (2) sugges s ha
he highe -income cus ome s a e less loyal, i.e., hey use he se ice almos 6.6 days
less han he lowe -income cus ome s. In Figs. 3and 4, panel (a), hese opposi e
ela ionships a e e iden . While neighbo hoods wi h highe income end o ha e a
highe p opo ion o ca -sha ing use s, hose wi h he lowes income in ou s udy a ea
concen a e he mos loyal membe s. Rega dless, in esul s p esen ed in appendix (see
Table 11), we show ha bo h loyal y and income a e posi i ely co ela ed wi h he
p obabili y o aking a ca -sha ing ip.
Table 4 Reg ession esul s o pu pose o ip: ip le el (sub-sample wi h iden i ied esidence)
Membe loyal y
Membe densi y Days used Times used Days used (weigh ed)
(1) (2) (3) (4)
Annual ne income (e)
14,500–19,500e0.104
(0.071)
−2.443**
(0.960)
−4.129***
(1.498)
−4.439***
(1.190)
19,500–24,000e0.379***
(0.106)
−4.920***
(0.878)
−8.565***
(1.366)
−8.597***
(1.072)
+24,000e0.722***
(0.118)
−6.601***
(0.848)
−11.286***
(1.305)
−10.385***
(1.045)
Obse a ions 92 11,818 11,818 11,818
R20.358 0.007 0.008 0.011
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
The abo e linea eg essions a e es ic ed o he sub-sample o ca -sha ing use s wi h iden i ied esidence,
excluding hose li ing in A ocha, Ae opue o, and Cas illa neighbo hoods. The eg essions p o ide co e-
la ions be ween ca sha ing membe densi y and loyal y wi h he neighbo hood income qua iles, using as
base ca ego y he lowes qua ile: e8,000–e14,500. Column (1) uses neighbo hood obse a ions, while in
columns (2–4) each obse a ion ep esen s a cus ome
The ou come a iable in column (1) measu es he pe cen age o ca -sha ing use s in a neighbo hood,
i.e., o al numbe o use s di ided by o al esiden popula ion
Columns (2–4) p o ide di e en measu es o membe loyal y. The dependen a iable in column
(2) Measu es he o al numbe o days a cus ome used he ca -sha ing se ice in 2019, while column
(3) Measu es he numbe o imes she/he used he se ice. In column (4), he ou come a iable measu es
he weigh ed numbe o days used, gi en by he numbe o days he cus ome used he ca -sha ing se ice,
weigh ed by he o al numbe o days le in 2019 since he i s day hey used he se ice, mul iplied by
o al numbe o days in a yea
123
SERIEs (2024) 15:259–297 275
2
1.5
1
.5
0
10,000 15,000 20,000 25,000 30,000
A e age annual income pe capi a (€)
(a) A e age annual income
2
1.5
1
.5
0
.5 1 1.5 2
A e age numbe o ca s pe household
(b) Ca owne ship
P opo ion o ca -sha ing use s (%)
P opo ion o ca -sha ing use s (%)
Fig. 3 Co ela ion plo s be ween membe densi y and neighbo hood-le el a ibu es. No es The sca e plo s
co ela e ca -sha ing membe densi y wi h wo neighbo hood a ibu es. Membe densi y is compu ed as
he numbe o ca -sha ing use s li ing in each neighbo hood, weigh ed by he o al popula ion. Panel
(a) co ela es membe densi y wi h he a e age annual ne income pe capi a (INE 2018b;AM2018c),
while panel (b) p o ides he co ela ion wi h he a e age numbe o ca s pe household (CRTM 2018b)
in each neighbo hood o he s udy a ea. We exclude he neighbo hoods o A ocha and Cas illa (whe e he
main ain s a ions a e) and Ae opue o.
4.3 E ec on mobili y pa e ns
One c ucial ac o o conside when assessing he implica ions o FFCS in ci ies is
he impac o he ca -sha ing se ice on mobili y pa e ns. Fo example, ully elec-
ic FFCS ca s could accele a e he deca boniza ion o ci ies h ough ca shedding. In
his scena io, he majo i y o ca -sha ing membe s would be ca owne s conside ing
123
276 SERIEs (2024) 15:259–297
A e age numbe o days a cus ome used he se ice
70
70
60
60
50
50
40
40
30
30
20
10,000 15,000 20,000 25,000 30,000
A e age annual income pe capi a (€)
(a) A e age annual income
20
.5 1 1.5 2
A e age numbe o ca s pe household
(b) Ca owne ship
A e age numbe o days a cus ome used he se ice
Fig. 4 Co ela ion plo s be ween membe loyal y and neighbo hood-le el a ibu es. No es The sca e plo s
co ela e ca -sha ing membe loyal y wi h wo neighbo hood a ibu es. Membe loyal y is measu ed as
he weigh ed a e age numbe o days he membe s li ing in he neighbo hood used he se ice in 2019.
This loyal y measu e is compu ed by aking he o al numbe o days each cus ome used he se ice,
weigh ed by he o al numbe o days le in 2019 since hey i s used he se ice, and mul iplied by he
o al numbe o days in a yea . Panel (a) co ela es membe loyal y wi h he a e age annual ne income
pe capi a (INE, 2018b;AM,2018c), while panel (b) p o ides he co ela ion wi h he a e age numbe o
ca s pe household (CRTM, 2018b) in each neighbo hood o he s udy a ea. We exclude he neighbo hoods
o A ocha and Cas illa (whe e he main ain s a ions a e) and Ae opue o. We also exclude he Pa ones
neighbo hood, which exhibi s ou lie loyal y abo e 100
123
SERIEs (2024) 15:259–297 277
he possibili y o selling hei ehicle and using he ca -sha ing se ice as an al e -
na i e (Jochem e al. 2020; Becke e al. 2018).14 This may be pa icula ly likely o
households owning mo e han one ca . Ano he po en ial mechanism is ha d i e s
may swi ch hei pollu ing ehicle o an elec ic one ollowing a posi i e ca -sha ing
expe ience. In ha case, e en when shedding is limi ed, ca -sha ing may con ibu e
o educing pollu ion by accele a ing he elec i ica ion o he lee . Mo eo e , FFCS
migh p e en u u e ehiclepu chasesi d i e sconside ha heFFCSse iceal eady
ul ills hei needs.
Besides he e ec on ca owne ship, howe e , i is essen ial o analyze he ela ion-
ship be ween he use o ca -sha ing se ices and public anspo . On he one hand, i
use s conside hem o be subs i u es, FFCS could educe he use o collec i e public
anspo , leading o an o e all inc ease in he numbe o ca s in he ci y. Fo ins ance,
use s may see FFCS and public anspo a ion as subs i u es o sho ips be ween
nea by a eas, o which he a el ime by public anspo is gene ally highe han by
ca (Ampudia-Renuncio e al. 2020b). Howe e , when conside ing he ime equi ed
o each he ehicle and o ind a pa king slo , he a el ime di e ence be ween FFCS
and public anspo may no be signi ican (Sp ei e al. 2019). Indeed, his ou come is
unlikely, acco ding o p io li e a u e (Habibi e al. 2017; Becke e al. 2018; Tyndall
2019).
On he o he hand, FFCS could be a complemen o public anspo , pa icula ly
in a eas whe e he public anspo ne wo k is sca ce (Vine and Polak 2020). In his
case, FFCS would p e en he use o p i a e ehicles, con ibu ing o educed land
occupancy and lowe ing emissions. Howe e , e en whe e ca -sha ing complemen s
incomple e public anspo ne wo ks, FFCS se ices mus be implemen ed whe e
al e na i e anspo modes exis , gi en he unce ain y ega ding he a ailabili y o
ca s o he e u n ip (Ampudia-Renuncio e al. 2020a). This is ano he impo an
sou ce o complemen a i y.
Building on da a om se e al sou ces, we p o ide insigh s in o which o he abo e
mechanisms is mo e likely o domina e in Mad id. We ela e FFCS usage wi h ca
owne ship and he public anspo ne wo k in he ci y. To measu e he a ailabili y
o public anspo se ices, we i s ocus on he densi y o me o, ain, and bus
s a ions/s ops, i.e., he numbe o s a ions/s ops pe Km2in each neighbo hood. Since
he p esence o se e al s a ions migh no imply a p ope connec ion be ween di e en
des ina ions, we u he c ea ed a measu e o public anspo adequacy based on a
ques ion om Mad id’s Mobili y Su ey (see Sec . 3).
We assess he co ela ion be ween ca -sha ing usage and mobili y pa e ns h ough
he ollowing eg ession:
Yi=γ+Xθi+εi,(3)
14 Acco ding o a su ey conduc ed among ca -sha ing use s in 11 Eu opean ci ies (Jochem e al. 2020),
he sha e o su ey pa icipan s selling a ca a e ying he se ice anged om 3.6% o 16.0%, wi h he
lowes alue being ound in Mad id. Howe e , compa ed o s a ion-based ca -sha ing, FFCS membe s a e
less likely o dec ease ehicle owne ship, as hese use s see he FFCS se ice as a complemen a he han
a eplacemen o hei p i a e ca (Ami nazmia sha and Diana 2022).
123
278 SERIEs (2024) 15:259–297
whe e Yiis he ou come o in e es (cus ome densi y) o he neighbo hood i;γis
a eg ession cons an ; Xis a ec o o explana o y a iables; and εiis an idiosyn-
c a ic e o e m. The coe icien s θicap u e how ca -sha ing membe densi y changes
acco ding o neighbo hood cha ac e is ics and ameni ies, no ably, he annual income
pe capi a, a e o ca owne ship, and adequacy o he public anspo ne wo k.
Table 5p o ides he esul s o he abo e eg ession. Fi s ly, we no e ha neigh-
bo hoods wi h highe a es o ca owne ship end o ha e a highe pe cen age o
ca -sha ing membe s (see also panel (b) o Fig. 3). In pa icula , columns (1) and (5)
sugges ha an addi ional ca pe household is associa ed wi h a 0.56 p.p. inc eased
membe densi y. Since hese eg essions al eady con ol o he income le el o he
neighbo hood, hese esul s sugges ha ca -sha ing use s a e ca owne s, subs i u ing
hei p i a e ehicle o a sha ed one, when using he se ice. In column (2) we es he
in luence o he public anspo ne wo k on ca -sha ing usage. The es ima es indica e
ha neighbo hoods wi h ewe me o and bus s ops pe Km2ha e a highe densi y o
ca -sha ing membe s.
In Table 9in appendix, we es ima e Eq. 3 o a di e en ou come a iable: membe
loyal y. In his speci ica ion, θicap u es changes in cus ome loyal y, o cus ome i,
acco ding o he same neighbo hood cha ac e is ics and ameni ies. The esul s suppo
ou p e iousconclusions,neighbo hoods wi h ewe public anspo s a ions/s opspe
Km2and wi h mo e complain s o inadequa e public anspo ne wo k ha e he mos
loyal cus ome s. Howe e , hese e ec s a e mu ed once we con ol o ca owne ship,
which is nega i ely co ela ed wi h he a ailabili y o public anspo . These esul s
sugges ha ca -sha ing migh be used in Mad id as a subs i u e o p i a e ehicles,
and o complemen he exis ing public anspo ne wo k. In ac , he nega i e co ela-
ion be ween ca owne ship and he a ailabili y o public anspo could indica e ha
esiden s in a eas wi h poo e public anspo ne wo ks own a ca o coun e insu i-
cien public anspo op ions. Ou indings could imply ha FFCS se ices dec ease
commu ing/ a el imes, imp o e connec ions, and p omo e ca shedding.
In Table 10 in appendix, we es ima e he o iginal Eq. 3, ex ending he p e ious
analysis o all neighbo hoods in Mad id municipali y. In neighbo hoods whe e we did
no iden i y ca -sha ing membe s, we impu e a membe densi y o ze o. Ou esul s
a e obus o his sample modi ica ion.
Fo he comple e deca boniza ion o he ci y, p e e ence should be gi en o mo e
sus ainable a el modes, such as walking, cycling, and public anspo . The i s wo
modes a e no easible o long dis ances, and expanding he public anspo ne wo k
is gene ally cos ly and equi es a long implemen a ion pe iod. In his sense, elec ic
ee- loa ing ca -sha ing se ices migh be an op imal solu ion o complemen he
exis ing public anspo ne wo k. I is impo an o s ess ha hese sha ed ehicles
a e conside ably mo e e icien han p i a e ca s. Fi s ly, he FFCS lee in Mad id is
ully elec ic. Secondly, FFCS p o ides mo e e icien use o he sca ce space in ci ies,
pa icula ly pa king. On a e age, a p i a e ca is pa ked 23 hou s a day, ansla ing
in o a usage a e o 4% (Nagle 2021). By being sha ed among se e al indi iduals,
FFCS ehicles ha e a highe u iliza ion a e. In ou da a, ehicles’ a e age usage a e
was 23%, measu ed by he o al numbe o hou s each ca was used, di ided by he
maximum numbe o hou s i could ha e been used. Habibi e al. (2017) and Sp ei
123

SERIEs (2024) 15:259–297 279
Table 5 Reg ession esul s o mobili y pa e ns: neighbo hood le el (sub-sample wi h iden i ied esidence)
Membe densi y No. ips a des ina ion
(1) (2) (3) (4) (5) (6)
Log annual ne
income (e)
0.739*** 0.933*** 0.877*** 0.906*** 0.737*** 32.742***
(0.198) (0.123) (0.213) (0.126) (0.200) (6.908)
A g. no. o ca s
pe household
0.561*** 0.298 0.568**
(0.212) (0.244) (0.218)
% o esiden s
wi h d i ing
license
−0.003 −0.004 −0.003
(0.009) (0.010) (0.009)
No. me o s a ions
pe km2−0.033** −0.021
(0.013) (0.015)
No. ain s a ions
pe km20.030 0.048
(0.082) (0.092)
No. bus s a ions
pe km2−0.009** −0.006
(0.004) (0.004)
Inadequa e public
anspo
ne wo k (%)
0.008 −0.001
(0.006) (0.006)
No. o pa king
slo s pe ca
0.573***
(0.051)
Obse a ions 92 92 92 92 92 47
R2 0.459 0.467 0.480 0.376 0.459 0.715
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
No es The linea eg essions use neighbo hood obse a ions. Columns (1) o (5) a e es ic ed o he sub-
sample o ca -sha ing use s wi h iden i ied esidence, excluding hose li ing in A ocha, Ae opue o, and
Cas illa neighbo hoods. The ou come a iable in columns (1) o (5) measu es he neighbo hood membe
densi y, i.e., he pe cen age o ca -sha ing use s in each neighbo hood, ou o he o al esiden popula-
ion. These eg essions p o ide co ela ions be ween ca -sha ing membe ship and he cha ac e is ics and
ameni ies o he neighbo hood. The dependen a iable in column (6) measu es he o al numbe o ips
ending in each neighbo hood, weigh ed by he esiden popula ion, while he independen a iable is gi en
by he numbe o pa king slo s pe ca . These pa king slo s a e in eg a ed wi hin he Regula ed Pa king
Se ice (SER), which, in 2018, only co e ed 48 neighbo hoods in he Municipali y o Mad id. Reg ession
(6) p o ides he co ela ion be ween he numbe o ips and pa king a ailabili y a he des ina ion
123
280 SERIEs (2024) 15:259–297
e al. (2019), which s udied FFCS in se e al Eu opean and No h-Ame ican ci ies,
concluded ha Mad id had he highes usage a e o all ci ies analyzed.
4.4 Seasonali y and pu pose o ca -sha ing usage
Figu e 5shows he empo al dis ibu ion o ca -sha ing ips. Fo g aphs on he le -
hand side, we use he ull sample o ca -sha ing ips in 2019. Righ -hand side g aphs
a e o a es ic ed sample o ips o which we iden i ied use s’ esidences, such ha
we can co ela e seasonali y pa e ns wi h demog aphic cha ac e is ics (income).
Fo g aphs on he le , we con as he ca -sha ing ip seasonali y wi h oad a ic
on he main highway (M-30) and in u ban a eas inside he municipali y. We obse e
ha he seasonali y o ca -sha ing ips closely ollows he pa e n o o he mo o ized
ehicles in Mad id. We see a highe usage du ing peak hou s (mo ning and a e noon)
on weekdays (panel a). In Mad id, he e is a hi d peak hou a lunch ime, be ween 2
pm and 5 pm, when many s o es close (Ampudia-Renuncio e al. 2020b; Habibi e al.
2017). None heless, we no ice ha he mo ning peak o ca -sha ing use s happens
ea lie , be ween 6 am and 7 am, wi h he numbe o ca -sha ing ips d opping signi i-
can ly du ing he usual mo ning peak, be ween 8 am and 9 am. This migh sugges ha
ca -sha ing use s a oid high conges ion hou s o a oid paying highe bills, since he
ca -sha ing se ice is cha ged by he minu e. Addi ionally, ca -sha ing ips exhibi
a highe peak a lunch ime han oad a ic. Acco ding o he g aphs on he igh ,
du ing bo h hese peaks, mos ca -sha ing membe s a e abo e he medium income o
all ca -sha ing use s. The e a e only wo peaks a lunch ime and in he a e noon on
weekends (panel b). Simila ly, he peaks o ca -sha ing use s happen sligh ly ea lie
han usual oad a ic.
Looking a he pa e ns by day o he week (panel c), we see highe usage o
he ca -sha ing se ice on weekdays, pa icula ly on F idays. We also no e ha ips
du ing he weekend a e domina ed by below-a e age income cus ome s, while du ing
weekdays, especially a peak hou s, highe -income use s a e mo e p e alen . These
igu es sugges ha highe -income cus ome s use he se ice o commu ing a he
han leisu e. This is in line wi h indings o Ge many, whe e mos ee- loa ing ca -
sha ing ips happen on Sa u day, and he peak usage hou s happen conside ably la e
han hose o p i a eca s, pa icula lydu ing he a e noon(Schmölle e al. 2015).The
seasonali y o ca -sha ing ips seems o be ci y-speci ic, especially when compa ing
wo kdays e sus weekends (Sp ei e al. 2019).
Rega ding hedis ibu iono ips h oughou heyea (panel d),Julyand Sep embe
show mo e ca -sha ing usage. Howe e , oad a ic, which is ela i ely consis en
h oughou he yea , sees a d op in hese mon hs, in ensi ying in Augus . Al hough
we see he same d op in Augus o ca -sha ing ips, he educ ion is much less
p onounced,sugges ing ha ou is s alsouse hisse ice. As udyinBe linandMunich
oundad opinca -sha ingu iliza ionbe weenJuneandSep embe ,a guing ha du ing
hese mon hs, use s op o o he a el modes, such as walking and cycling (Schmölle
e al. 2015). The igh -hand side g aph o panel (d) is consis en wi h highe -income
cus ome s lea ing he ci y du ing he summe o holidays.
123
SERIEs (2024) 15:259–297 281
Annual ne income (€) Annual ne income (€) Annual ne income (€)
822000
621000
420000
219000
0
01234567891011121314151617181920212223
Hou o he Day
Ca sha ing ips M-30 U ban a eas
18000
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hou o he Day
A e age 95% CI
(a)
Pa e ns by Hou o Day – weekdays
822000
621000
420000
219000
0
01234567891011121314151617181920212223
Hou o he Day
18000
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hou o he Day
(b)
Pa e ns by Hou o Day – weekends
20 22000
15 21000
10 20000
519000
0
Sunday Monday Tuesday Wednesday Thu sday F iday Sa u day
Day o he Week
18000
Sunday Monday Tuesday Wednesday Thu sday F iday Sa u day
Day o he Week
(c)
Pa e ns by Day o Week
10 22000
8
21000
6
20000
4
19000
2
0
1 2 3 4 5 6 7 8 9 10 11 12
Mon h o he Yea
18000
123456789101112
Mon h o he Yea
(d)
Pa e ns by Mon h o Yea
Pe cen o T ips Pe cen o T ips Pe cen o T ips Pe cen o T ips
Annual ne income (€)
Fig. 5 Seasonali y o ca -sha ing ips. No es Le -hand side igu es use he ull sample, while igh -hand
side igu es a e es ic ed o he sub-sample o membe s wi h iden i ied esidence. The blue and ed lines in
he le -hand side igu es ep esen oad a ic seasonali y in he main highway (M-30) and in u ban a eas o
he municipali y o Mad id, espec i ely (AM 2018a). The ed lines in he igh -hand side igu es ep esen
he impu ed a e age annual ne income o he ca -sha ing membe s wi h iden i ied esidence, excluding
hose li ing in A ocha, Ae opue o, and Cas illa neighbo hoods (20,435e) (INE 2018b;AM2018c)
123
282 SERIEs (2024) 15:259–297
To es he hypo hesis ha highe -income cus ome s p edominan ly use he se ice
o commu e, we ca ego ized ips acco ding o hei pu pose, based on he equency o
usage a di e en imes o he day and he week (see Sec . 3). Wi h his speci ica ion,
we could ca ego ize 56% o he ips om membe s wi h an iden i ied esidence. O
hose,68.8%we e commu ing ips, while 31.2% hadaleisu e pu pose. The eg ession
es ima es om Table 6sugges ha highe -income cus ome s a e mo e likely o ake
ca -sha ing ips o commu e and less likely o use his se ice o leisu e pu poses.
Acco ding o Becke e al. (2017), who ocus on he case o Swi ze land, “ ee-
loa ing ca -sha ing is mainly used o disc e iona y ips, o which only subs an ially
in e io public anspo a ion al e na i es a e a ailable.” Schmölle e al. (2015) ind
ha ca -sha ing se ices in Be lin and Munich a e used p edomina ely o shopping
and social- ec ea ional ac i i ies. In Vine and Polak (2020), he au ho s a gue ha he
pu pose o ca -sha ing usage di e s acco ding o ca owne ship s a us. Ca owne s a e
mo e likely o use he se ice o business pu poses ( o ins ance, mee ings), while
non-ca owne s use he se ice o shopping pu poses.
Finally, Fig. 6 ep esen s he seasonali y o ips acco ding o hei dis ance and
du a ion. Ca -sha ing ips we e gene ally sho as a unc ion o dis ance and a el
ime. The a e age igu es we e 23 minu es and 7 Km pe ip (in line wi h p e ious
s udies,e.g., Ampudia-Renuncioe al.2020b;Sp ei e al. 2019). Asexpec ed, ips ake
longe du ing peak hou s due o hea ie oad a ic. On weekends and F idays, ips
a e ela i ely longe in dis ance and ime a eled. Addi ionally, o e ime, cus ome s
s a ed o ake longe ips (in e ms o dis ance).
4.5 G a i y speci ica ions
Tables 7and 8p esen esul s om es ima ion o g a i y- ype models such as Eq.
(1). Table 7is o baseline speci ica ions. In he i s column, we show esul s when
only log dis ance is included as a ic ion a iable, igno ing he a ailabili y o pub-
lic ansi . We ind ha ca -sha ing ips become less likely wi h inc eased dis ance
be ween o igin/des ina ion pai s. Acco ding o ou p e e ed speci ica ion (column 3),
he poin es ima e is (−0.47). Gi en ou Poisson speci ica ion, his coe icien needs
o be ans o med o ease o in e p e a ion. Fo example, i he dis ance be ween o i-
gins/des ina ions we e o double, hen he numbe o ca -sha ing ips would d op by
almos 28% [=exp(−0.47 ×log2)−1]. This may be pa ly explained by he ac ha
he cos s o ca -sha ing ips can also inc ease signi ican ly wi h dis ance (as he ip
du a ion inc eases, and en als a e paid by he minu e).
In e ms o he impac o public ansi a el imes, no e ha column (2) o Table 7
sugges s a nega i e elas ici y, while column (3) sugges s a posi i e elas ici y. Howe e ,
he nega i e elas ici y om column (2) may be con ounded by he e ec o dis ances.
Column (3) is he p e e ed speci ica ion in he sense ha i es ima es he elas ici y o
ca -sha ing wi h espec o public ansi a el imes, while simul aneously con olling
o he dis ance be ween he o igins/des ina ions o he ips. In ha case, he elas ici y
becomes posi i e, such ha ca -sha ing is mo e likely as o igins/des ina ions a e less
well-connec ed h ough public ansi . The poin es ima e is (0.0037), which can be
in e p e ed as ollows: i a el ime by public ansi inc eases by 10 minu es, hen
123
SERIEs (2024) 15:259–297 289
Table 9 Reg ession esul s o mobili y pa e ns: cus ome le el (sub-sample wi h iden i ied esidence)
Membe loyal y
(1) (2) (3) (4) (5)
Log annual ne
income (e)
−14.351*** −13.774*** −13.307*** −13.873*** −14.840***
(1.577) (1.255) (1.714) (1.257) (1.596)
A g. no. o ca s
pe household
13.212*** 10.403*** 14.189***
(1.279) (1.928) (1.387)
% o esiden s
wi h d i ing
license
−0.291*** −0.283*** −0.289***
(0.074) (0.075) (0.074)
No. me o s a ions
pe km2−0.567** −0.169
(0.221) (0.241)
No. ain s a ions
pe km2−3.182** −1.562
(1.251) (1.291)
No. bus s ops pe
km2−0.152*** −0.052
(0.035) (0.041)
Inadequa e public
anspo
ne wo k (%)
0.126*** −0.065
(0.039) (0.043)
Obse a ions 11,818 11,818 11,818 11,818 11,818
R20.020 0.018 0.020 0.012 0.020
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
The linea eg essions use cus ome obse a ions and a e es ic ed o he sub-sample o ca -sha ing use s
wi h iden i ied esidence, excluding hose li ing in A ocha, Ae opue o, and Cas illa neighbo hoods. The
loyal y measu e, used as ou come a iable, measu es he weigh ed numbe o days a cus ome used he
ca -sha ing se ice, gi en by he o al numbe o days used, weigh ed by he o al numbe o days le in 2019
since he i s day hey used he se ice, mul iplied by he o al numbe o days in a yea . The eg essions
p o ide co ela ions be ween he cha ac e is ics and ameni ies o he neighbo hood and he loyal y/in ensi y
o usage o he esiden cus ome s
123

290 SERIEs (2024) 15:259–297
Table 10 Reg ession esul s o mobili y pa e ns: neighbo hood le el ( ull sample)
Membe densi y
(1) (2) (3) (4) (5) (6)
Log annual ne
income (e)
0.745***
(0.110)
0.935***
(0.329)
0.841***
(0.128)
1.001***
(0.259)
0.761***
(0.114)
0.948***
(0.326)
A e age no. o ca s
pe household
0.474
(0.307)
0.218
(0.294)
0.358
(0.264)
% o esiden s wi h
d i ing license
−0.020
(0.021)
−0.014
(0.014)
−0.018
(0.019)
No. me o s a ions
pe km2−0.047*
(0.027)
−0.035*
(0.020)
No. ain s a ions
pe km20.443
(0.315)
0.379
(0.230)
No. bus s a ions
pe km2−0.002
(0.004)
−0.002
(0.005)
Inadequa e public
anspo
Ne wo k (%)
0.013
(0.011)
0.007
(0.010)
Obse a ions 135 135 135 135 135 135
R20.241 0.307 0.348 0.368 0.279 0.316
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
These eg essions p o ide co ela ions be ween densi y o ca -sha ing membe ship and he cha ac e is ics
and ameni ies o neighbo hoods. The linea eg essions use neighbo hood-le el obse a ions and include
all neighbo hoods o he municipali y o Mad id in addi ion o he ou municipali ies co e ed by ou s udy
a ea. The ou come a iable measu es he neighbo hood membe densi y, i.e., he pe cen age o ca -sha ing
use s in each neighbo hood, ou o he o al esiden popula ion. This densi y is se o ze o o neighbo hoods
whe e we did no iden i y ca -sha ing membe s
123
SERIEs (2024) 15:259–297 291
Table 11 Reg ession esul s o p obabili y o ip happening, linea p obabili y model: ip le el (ex ended
sub-sample wi h iden i ied esidence)
P obabili y o aking a ca sha ing ip
(1) (2) (3)
Log annual ne income (e)−0.03777***
(0.00063)
−0.00210***
(0.00060)
Membe loyal y10.00275***
(0.00001)
0.00275***
(0.00001)
Con ol o weekly seasonali y Yes Yes Yes
Con ol o mon hly seasonali y Yes Yes Yes
Obse a ions 3,682,648 3,682,648 3,682,648
R2 0.015 0.101 0.101
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
The able p o ides eg essions esul s using a linea p obabili y model, whe e he dependen a iable is an
indica o equal o 1 when a cus ome akes a ca -sha ing ip and 0 o he wise. We use cus ome -by-da e
obse a ions. The dependen a iable is se as missing o all da es p io o he i s usage da e o a gi en
cus ome . This a iable he e o e measu es he p obabili y o aking a ca -sha ing ip a e ying he se ice
o he i s ime in 2019
1Weigh ed numbe o days a cus ome used he ca -sha ing se ice in 2019, compu ed as he numbe o
days he cus ome used he ca -sha ing se ice weigh ed by he o al numbe o days le in 2019 since he
i s day hey used he se ice, mul iplied by o al numbe o days in a yea
123
292 SERIEs (2024) 15:259–297
Table 12 Reg ession esul s o ip du a ion and dis ance: ip le el (sub-sample wi h iden i ied esidence)
T ip dis ance T ip du a ion
(1) (2) (3) (4) (5) (6)
Log annual ne income (e)−2.10846*** −2.01185*** −1.93357*** −3.77156*** −3.85011*** −3.81258***
Membe loyal y1(0.02620) (0.02639)
0.00379***
(0.02640)
0.00249***
(0.10407) (0.10492)
−0.00308***
(0.10528)
−0.00467***
(0.00014) (0.00014) (0.00054) (0.00055)
Con ol o weekly seasonali y No No Yes No No Yes
Con ol o mon hly seasonali y No No Yes No No Yes
Obse a ions 497,009 497,009 497,009 497,011 497,011 497,011
R2 0.013 0.014 0.023 0.003 0.003 0.005
S anda d e o s in pa en heses: ∗p<0.1, ∗∗p<0.05, ∗∗∗p<0.01
No es The linea eg essions use ip obse a ions and a e es ic ed o he sub-sample o ca -sha ing use s wi h iden i ied esidence, excluding hose li ing in A ocha,
Ae opue o, and Cas illa neighbo hoods. The ou come a iables a e ip dis ance [in kilome e s, o columns (1) o (3)] and ip du a ion [in minu es, o columns (4) o (6)]
1Weigh ed numbe o days a cus ome used he ca -sha ing se ice in 2019
123
SERIEs (2024) 15:259–297 293
Numbe o cus ome s using he se ice
48
44
40
36
32
28
24
20
16
12
8
4
0
123456789 1011 12
Mon h o Yea 2019
(a) All use s
Numbe o new cus ome s joining he
se ice
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
123456789 1011 12
Mon h o Yea 2019
(b) New use s
Numbe o cus ome s ( housands)
Numbe o new cus ome s ( housands)
Fig. 7 Numbe o ca -sha ing cus ome s pe mon h, in 2019. No e The igu e po ays he numbe o ca -
sha ing cus ome s in 2019. Panel (a) shows he o al numbe o cus ome s ha used he se ice in each
mon h o he yea , while panel (b) p esen s he numbe o new cus ome s joining he ca -sha ing se ice
o he i s ime, in each mon h
123
294 SERIEs (2024) 15:259–297
50
40
30
20
10
0
0 50 100 150 200 250 300 350
Weigh ed numbe o days a cus ome used he se ice, since i s use
Pe cen o cus ome s
Fig. 8 F equency dis ibu ion o membe loyal y. No e The igu e shows he equency dis ibu ion o he
weigh ed membe loyal y measu e, which is gi en by he numbe o days a cus ome used he ca -sha ing
se ice, weigh ed by he o al numbe o days le in 2019 since he i s day he/she used he se ice,
mul iplied by he numbe o days in a yea . The e ical lines gi e he a e age loyal y o he ull sample o
cus ome s (14.7 days). The median loyal y is 6.8 days
Loca ion o me o and ain s a ions
T ain (Ce canias)
Me o
Neighbou hoods inside Mad id Municipali y
Ou side Mad id Municipali y
(a) T ain and me o
Loca ion o bus s ops
Bus s ops
Neighbou hoods inside Mad id Municipali y
Ou side Mad id Municipali y
(b) Bus
Fig. 9 Public anspo ne wo k (me o, ain, and buses). No e The igu es show he geog aphical loca ion o
each me o, ain, and bus s a ion/s op in he municipali ies o ou s udy a ea (CRTM 2018a). All e i o ial
uni s co espond o neighbo hoods inside Mad id Municipali y, excep o he dashed uni s: Alcobendas,
Coslada, Pozuelo de Ala cón, and San Sebas ian de los Reyes municipali ies
123

SERIEs (2024) 15:259–297 295
(0.42,2.55]
(0.34,0.42]
(0.28,0.34]
(0.26,0.28]
(0.23,0.26]
[0.14,0.23]
No da a
Fig. 10 Regula ed pa king slo s, in 2018. No e: The igu e po ays he numbe o egula ed pa king slo s,
weigh ed by he numbe o egis e ed ehicles in each neighbo hood. These pa king slo s a e in eg a ed
wi hin he Regula ed Pa king Se ice (Se icio de Es acionamien o Regulado, SER). In 2018, only 48
neighbo hoods in he Municipali y o Mad id we e co e ed by he SER. All e i o ial uni s ep esen ed a e
neighbo hoods inside Mad id Municipali y, excep o he municipali ies o Alcobendas, Coslada, Pozuelo
de Ala cón, and San Sebas ian de los Reyes. The yellow line delimi s ou ca -sha ing s udy a ea
Size o Flee
400
0 200 600 800
01jan2018
02ma 2018
01may2018
30jun2018
29aug2018
28oc 2018
27dec2018
25 eb2019
26ap 2019
25jun2019
24aug2019
23oc 2019
22dec2019
Fig. 11 Flee o ca -sha ing company. No e The igu e shows he FFCS en al company’s cumula i e numbe
o ca s ( lee ) ope a ing in Mad id du ing 2018/2019. No e ha ou main analyses a e pe o med o he
yea o 2019, when he lee was al eady s abilized
Re e ences
Al-Kindi SG, B ook RD, Biswal S, Rajagopalan S (2020) En i on- men al de e minan s o ca dio ascula
disease: lessons lea ned om ai pollu ion. Na Re Ca diol 17(10):656–672
123
296 SERIEs (2024) 15:259–297
Ami nazmia sha E, Diana M (2022) A e iew o he socio-demog aphic cha ac e is ics a ec ing he
demand o di e en ca -sha ing ope a ional schemes. T ansp Res In e disc Pe spec . h ps://doi.o g/
10.1016/j. ip.2022.100616
Ampudia-Renuncio M, Gui ao B, Molina-Sánchez R, B angança L (2020a) Elec ic ee- loa ing ca sha ing
o sus ainableci ies:Cha ac e iza iono equen ipp o ilesusingacqui ed en alda a.Sus ainabili y
12(3):1248. h ps://doi.o g/10.3390/su12031248
Ampudia-Renuncio M, Gui ao B, Molina-Sánchez R, Engel de Ál a ez C (2020b) Unde s anding he
spa ial dis ibu ion o ee- loa ing ca sha ing in ci ies: analysis o he new mad id expe ience h ough
a web-based pla o m. Ci ies 98:102593
Anas A, Lindsey R (2011) Reducing u ban oad anspo a ion ex e nali ies: oad p icing in heo y and in
p ac ice. Re En i on Econ Policy 5(1):66–88
Ayun amen o de Mad id (AM) (2018a) Po al de da os abie os del Ayun amien o de Mad id. His ó ico de
da os del á ico
Ayun amen o de Mad id (AM) (2018b) Po al de da os abie os del Ayun amien o de Mad id. Se icio de
es acionamien o egulado (SER)
Ayun amen o de Mad id (AM) (2018c) Po al web del Ayun amien o de Mad id. Es adís ica
Becke H, Cia i F, Axhausen KW (2017) Modeling ee- loa ing ca - sha ing use in Swi ze land: a spa ial
eg ession and condi ional logi app oach. T ansp Res Pa C 81:286–299
Becke H, Cia i F, Axhausen KW (2018) Measu ing he ca owne ship impac o ee- loa ing ca -sha ing
– a case s udy in basel, Swi ze land. T ansp Res Pa D 65:51–62
Böhm M, Nanni M and Pappala do L (2022) G oss pollu e s and ehicle emissions educ ion. Na Sus ain
5(8):699–707
Bö jesson M, Eliasson J, Hugosson MB, and B undell-F eij K (2012) The s ockholm conges ion cha ges–5
yea s on. e ec s, accep abili y and lessons lea n . T ansp Policy 20:1–12
Bö jesson M, Bas ian A and Eliasson J (2021) The economics o low emission zones. T ansp Res Pa A:
Policy P ac 153:99–114
Bucsky P and Juhász M (2022) Is ca owne ship educ ion impac o ca sha ing lowe han expec ed? A
Eu ope wide empi ical e idence. Case S ud T ansp Policy 10:2208–2217
Buggle J, Maye T, Sakalli SO, Thoenig M (2023) The Re ugee’s Dilemma: e idence om Jewish Mig a ion
ou o Nazi Ge many. The Qua J Econ 138(2):1273–1345
Caul ield B, Kehoe J (2021) Usage pa e ns and p e e ence o ca sha ing: a case s udy o Dublin. Case
S ud T ansp Policy 9:253–259
Conso cio Regional de T anspo es de Mad id (CRTM) (2018a) Da os abie os CRTM
Conso cio Regional de T anspo es de Mad id (CRTM) (2018b) Encues a domicilia ia de mo ilidad de la
Comunidad de Mad id (EDM2018)
Co eia S, Guima ães P, and Zylkin T (2020) Fas poisson es ima ion wi h high-dimensional ixed e ec s.
The S a a J 20(1):95–115
Cu ie J, Neidell M (2005) Ai pollu ion and in an heal h: Wha can we lea n om Cali o nia’s ecen
expe ience? Q J Econ 120(3):1003–1030
De yugina T, Heu el G, Mille NH, Moli o D, Rei J (2019) The mo ali y and medical cos s o ai pollu ion:
e idence om changes in wind di ec ion. Am Econ Re 109(12):4178–4219
Edlin A, Ka aca-Mandic P (2006) The acciden ex e nali y om d i ing. J Poli Econ 114(5):931–955
Eu opean Commission (2020) 2030 clima e a ge plan. EC Documen 52020DC0562 . Eu opean Cou o
Audi o s (ECA) (2019). Audi p e iew: U ban mobili y in he EU.
Galdon-Sanchez JE, Gil R, Holub F, and U iz-Uha e G (2022) Social bene i s and p i a e cos s o d i ing
es ic ion policies: he impac o Mad id Cen al on conges ion, pollu ion, and consume spending. J
Eu Econ Assoc j ac064
Geh si z M (2017) The e ec o low emission zones on ai pollu ion and in an heal h. J En i on Econ
Manag 83:121–144
Habibi S, Englund C, Vo ono A, Engdahl H, Sp ei F, Pe e sson S, and Wedlin J (2017) Compa ison o
ee- loa ing ca sha ing se ices in ci ies. Eu opean Council o Ene gy E icien Economy (ECEEE)
Summe S udy, P esqu’île de Giens, F ance, 29 May–3 June, 2017
Ins i u o Nacional de Es adís ica (INE) (2018a) Ins i u o Nacional de Es adís ica. Ci as de población
Ins i u oNacional de Es adís ica (INE) (2018b) Ins i u o Nacional de Es adís ica. Indicado es de en a media
y mediana
In e na ional Ene gy Agency (IEA) (2022) Global ene gy- ela ed CO2 emissions by sec o
123
SERIEs (2024) 15:259–297 297
Jochem P, F ankenhause D, Ewald L, Ensslen A, F omm H (2020) Does ee- loa ing ca sha ing educe
p i a e ehicle owne ship? The case o sha e now in Eu opean ci ies. T ansp Res 141(Pa A):373–395
Kopp J, Ge ike R, Axhausen KW (2015) Do sha ing people beha e di e en ly? An empi ical e alua ion o
he dis inc i e mobili y pa e ns o ee- loa ing ca -sha ing membe s. T anspo a ion 42:449–469
Lelie eld J, E ans JS, Fnais M, Giannadaki D, Pozze A (2015) The con ibu ion o ou doo ai pollu ion
sou ces o p ema u e mo ali y on a global scale. Na u e 525(7569):367–371
Li C, Managi S (2021) Con ibu ion o on- oad anspo a ion o PM2.5. Sci Rep 11(1):21320
Machado C, Hue N, Be ssane i F (2018) An o e iew o sha ed mobili y. Sus ainabili y 10:4342
Mülle J, de Almeida Co eia GH, and Bogenbe ge K (2017) An explana o y model app oach o he spa ial
dis ibu ion o ee- loa ing ca sha ing bookings: a case-s udy o ge man ci ies. Sus ainabili y 9:1–14
Nagle E (2021) S anding s ill. The Royal Au omobile Club Founda ion o Mo o ing L d
Pa y I (2002) Compa ing he e iciency o al e na i e policies o educing a ic conges ion. J Public Econ
85(3):333–362
Pe ei a RH, Sa ai a M, He szenhu D, B aga CKV, and Conway MW (2021) 5 : apid ealis ic ou ing on
mul imodal anspo ne wo ks wi h 5 in . Findings.
Rapson DS, Muehlegge E (2021) The economics o elec ic ehicles. Re En i on Econ Policy 17(2):274-
294
Sanidas E, Papadopoulos DP, G assos H, Velliou M, Tsiou is K, Ba be seas J, and Papademe iou V
(2017) Ai pollu ion and a e ial hype ension. a new isk ac o is in he ai . J Am Soc Hype ens
11(11):709–715
Schmölle S, Weikl S, Mülle J, and Bogenbe ge K (2015) Empi ical analysis o ee- loa ing ca sha ing
usage: he Munich and Be lin case. T ansp Res Pa C 56:34–51.
Sil a JMCS, Ten ey o S (2006) The log o g a i y. The Re Econ S a 88(4):641–658
Sp ei F, Habibi S, Englund C, Pe e sson S, Vo ono A, Wedlin J (2019) F ee- loa ing ca -sha ing elec i i-
ca ion and mode displacemen : a el ime and usage pa e ns om 12 ci ies in Eu ope and he Uni ed
S a es. T ansp Res Pa D 71:127–140
Tyndall J (2019) F ee- loa ing ca sha ing and ex empo aneous public ansi subs i u ion. Res T ansp Econ
74:21–27
Uni ed S a es Depa men o S a e and Execu i e O ice o he P esiden (U.S. DoS and EOP) (2021) The
long- e m s a egy o he Uni ed S a es: pa hways o ne -ze o g eenhouse gas emissions by 2050.
Whi e House Publica ions
Vine SL, Polak J (2020) The impac o ee- loa ing ca sha ing on ca owne ship: ea ly-s age indings om
London. T ansp Policy 75:119–127
Wei T, Tang M (2018) Biological e ec s o ai bo ne ine pa icula e ma e (PM2.5) exposu e on pulmona y
immune sys em. En i on Toxicol Pha macol 60:195–201
Wu J-Z, Ge D-D, Zhou L-F, Hou L-Y, Zhou Y, Li Q-Y (2018) E ec s o pa icula e ma e on alle gic
espi a o y diseases. Ch on Dis T ansl Med 4(2):95–102
Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps
and ins i u ional a ilia ions.
123