Geo ge, Sa ah; Salomo, Ka ja; Helbig, Ma cel
A icle — Published Ve sion
Spa ial ad an ages o highly educa ed indi iduals in
Ge many: Is sus ainable mobili y an exp ession o
p i ilege?
Ci ies
P o ided in Coope a ion wi h:
WZB Be lin Social Science Cen e
Sugges ed Ci a ion: Geo ge, Sa ah; Salomo, Ka ja; Helbig, Ma cel (2025) : Spa ial ad an ages o
highly educa ed indi iduals in Ge many: Is sus ainable mobili y an exp ession o p i ilege?, Ci ies,
ISSN 1873-6084, Else ie , Ams e dam, Vol. 156, pp. 1-11,
h ps://doi.o g/10.1016/j.ci ies.2024.105507
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/308505
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h p://c ea i ecommons.o g/licenses/by/4.0/
Spa ial ad an ages o highly educa ed indi iduals in Ge many: Is
sus ainable mobili y an exp ession o p i ilege?
Sa ah Geo ge
a,*
, Ka ja Salomo
a
, Ma cel Helbig
b
a
Be lin Social Science Cen e, Ge many
b
Leibniz Ins i u e o Educa ional T ajec o ies, Ge many
ARTICLE INFO
Keywo ds:
Sus ainabili y
Daily mobili y
Socio-spa ial inequali y
Educa ion
ABSTRACT
To e ec i ely comba clima e change i is c ucial o encou age daily en i onmen ally iendly beha iou ac oss
la ge pa s o he popula ion. This includes daily mobili y beha iou , since p i a e anspo is one o he la ges
con ibu o s o g eenhouse emissions. P e ious s udies sugges ha highly educa ed indi iduals exhibi mo e
en i onmen ally iendly mobili y beha iou , a ac ha is usually explained by hei highe en i onmen al
awa eness. We ins ead explo e he ex en o which his beha iou is d i en by hei socio-spa ial ad an ages. We
use comp ehensi e da a on daily mobili y: ou analy ical sample includes 16,419 jou neys om 4168 indi iduals
in 2002 and 102,774 jou neys om 26,036 indi iduals in 2017. The da a is ep esen a i e o Ge man esiden s in
la ge ci ies aged 18 o 59. We employ mul ile el OLS eg ession, logis ic eg ession, and ac ional mul inomial
logi models o analyse changes in a el pa e ns among highly educa ed indi iduals o e ime. Ou indings
e eal ha uni e si y g adua es end o eside no only mo e o en in la ge ci ies bu in he mos cen al
neighbou hoods wi hin hese a eas, leading o sho e daily a el dis ances. Consequen ly, hei daily jou neys
ake less ime and hey a e able o use slowe , mo e sus ainable mobili y op ions when commu ing, unning
e ands, o engaging in leisu e ac i i ies wi hou incu ing highe a el ime cos s han o he g oups. Ou esul s
highligh he impo ance o add essing esiden ial inequali ies as a key s ep in enabling a b oade popula ion o
adop sus ainable li es yles.
1. In oduc ion
Clima e change is one o he mos p essing challenges o ou ime
(IPCC, 2022). The anspo sec o , along wi h he ene gy and indus ial
sec o s, is among he majo con ibu o s o clima e change (Geo ga zi
e al., 2019). Pa icula ly in Ge many, whe e he numbe o egis e ed
ca s is high and g eenhouse gas emissions om ca s and mo o cycles
cons i u e a subs an ial po ion o he coun y's o al emissions, e ec i e
CO
2
educ ion e o s equi e a shi om p i a e ca s o sus ainable
modes o anspo (I ano a e al., 2020). A he same ime, daily a el
dis ances pe pe son a e inc easing wo ldwide (Odyssee-Mu e, 2015).
The e o e, o c ea e a mo e sus ainable anspo sys em i is necessa y
o educe bo h o e all a el dis ances and p i a e ca use (Banis e ,
2011).
One social g oup appea s o be a he o e on o hese ansi ions:
highly educa ed indi iduals. They use sus ainable modes o anspo
mo e o en han o he socio-economic g oups in Wes e n socie ies and
a e mo e suppo i e o policies aiming o educe ca use (Hudde, 2022;
Kim e al., 2016; Roos e al., 2020). A highe le el o educa ional
a ainmen is posi i ely associa ed wi h mo e en i onmen ally iendly
beha iou in gene al (Al-Nuaimi & Al-Ghamdi, 2022; Meye , 2015). The
link be ween educa ional a ainmen and en i onmen ally iendly
beha iou has been he subjec o conside able deba e. One p ominen
hypo hesis is ha indi iduals wi h highe le els o educa ion a e mo e
awa e o he dange s posed by clima e change and, as a esul , engage in
mo e en i onmen ally iendly beha iou s (Kim e al., 2016).
We a gue ha educing highly educa ed people's en i onmen ally
iendly beha iou o hei g ea e en i onmen al awa eness is an
o e simpli ica ion. Ins ead, en i onmen ally iendly beha iou , espe-
cially ega ding daily mobili y choices, equi es ce ain esou ces and
imposes ce ain cos s. Fo ins ance, o people o choose bikes o e ca s,
hey need sui able oads and easonably sho dis ances, o he wise
cycling becomes dange ous and o e ly ime-consuming. In his pape we
aim o explo e he ex en o which he en i onmen ally iendly daily
mobili y beha iou o hose wi h highe educa ion can be explained by
hei socio-spa ial ad an ages. Pe sonal a el needs, p e e ences, and
* Co esponding au ho a : Reichpi schu e 50, 10785 Be lin, Ge many.
E-mail add ess: [email p o ec ed] (S. Geo ge).
Con en s lis s a ailable a ScienceDi ec
Ci ies
jou nal homepage: www.else ie .com/loca e/ci ies
h ps://doi.o g/10.1016/j.ci ies.2024.105507
Recei ed 8 Sep embe 2023; Recei ed in e ised o m 24 July 2024; Accep ed 5 Oc obe 2024
Ci ies 156 (2025) 105507
A ailable online 23 Oc obe 2024
0264-2751/© 2024 The Au ho s. Published by Else ie L d. This is an open access a icle unde he CC BY license (
h p://c ea i ecommons.o g/licenses/by/4.0/ ).
choices a e s ongly connec ed wi h esiden ial loca ion (B uns &
Ma hes, 2019). Impo an ly, esea ch on esiden ial seg ega ion in-
dica es ha indi iduals wi h highe educa ional a ainmen a e mo e
p one o eside in la ge ci ies, pa icula ly in well-connec ed inne -ci y
loca ions (Becke s & Boschman, 2017; L´
opez-Gay e al., 2020). Highly
educa ed people's socio-spa ial ad an ages migh explain why hey can
a o d o op o mo e sus ainable modes o anspo and why i e en
migh be in hei bes pe sonal in e es .
To answe his ques ion, we use uniquely comp ehensi e mobili y
da a om he Mobili y in Ge many (MiG) su eys in 2002 and 2017. The
MiG p o ides de ailed in o ma ion on daily mobili y beha iou –
including daily a el dis ances, a el ime, modal choice, a el pu -
poses, e c. – in addi ion o socio-economic and socio-spa ial cha ac e -
is ics, and i is ep esen a i e o Ge man esiden s o all ages. We look a
la ge ci ies due o he signi ican di e ences in he a ailabili y o
anspo a ion op ions ac oss egions; we aim o examine places whe e
indi iduals ha e he eedom o choose be ween p i a e, public and
sha ed anspo a ion, ecognising ha smalle ci ies ace nume ous
challenges and lack he ange o sus ainable mobili y op ions a ailable
in la ge ci ies.
P e ious esea ch on mobili y beha iou has o en been limi ed o
he choice o anspo mode and does no include a el dis ance o
a el ime (Hudde, 2022, 2023). I also gene ally ails o accoun o he
educa ion o esponden s (Ngah e al., 2021). In he ollowing analysis,
we link exis ing li e a u e on socio-spa ial inequali y o daily mobili y
beha iou and show how socio-economic and spa ial ac o s a ec daily
mobili y pa e ns.
2. Resea ch backg ound
Indi iduals wi h highe le els o educa ion (A-le el o abo e)
demons a e enhanced en i onmen al awa eness, conce n o socie al
wel a e (Meye , 2015), and an unde s anding o he en i onmen al
impac o indi idual beha iou (Al a ez-Su´
a ez, Vega-Ma co e, & Mi a,
2014). This os e s en i onmen ally iendly beha iou by people wi h
highe le els o educa ion, especially in he a eas o household, ood and
anspo choices (Al-Nuaimi & Al-Ghamdi, 2022; Asiksoy e al., 2020),
as well as g ea e suppo o sus ainabili y policies (El-Menoua &
Unzicke , 2021). These ends ex end o mobili y beha iou and a i-
udes owa ds anspo policies: G ea e en i onmen al awa eness and
educa ion posi i ely in luence suppo o policies aimed a educing
mo o ised indi idual anspo (Kim e al., 2016). Fu he mo e, highly
educa ed indi iduals a e mo e likely o use bicycles and public anspo
(Roos e al., 2020). In Ge man ci ies, o ins ance, mo e highly educa ed
people use bicycles mo e o en han people o lowe educa ional
a ainmen doubled hei use o bicycles be ween 1990 and 2018
(Hudde, 2022).
O e all, en i onmen al awa eness signi ican ly in luences indi idual
daily mobili y beha iou (Hamidi & Zhao, 2020). I explains why highly
educa ed indi iduals di e om o he s wi h a high socio-economic
s a us. While highe -income households ely on p i a e ca s (Follme
& G uschwi z, 2019), he en i onmen ally conscious beha iou among
highly educa ed indi iduals p omp s g ea e bicycle and public ans-
po usage (Roos e al., 2020). Howe e , sus ainable mobili y comes a a
cos . Mobili y- ela ed ime cos s a e highe o use s o public ans-
po a ion o bikes; in ac , e idence om Sweden, B azil, Aus alia, and
he Ne he lands sugges s ha commu ing by public anspo o en akes
wice as long as by p i a e ca (Liao e al., 2020). T a el imes o public
ansi in majo Ge man ci ies mos ecen ly ha e been es ima ed o be
app oxima ely h ee imes longe han hose o mo o ised indi idual
anspo commu ing (Mocanu e al., 2021). The ime cos s o sus ain-
able mobili y a e an obs acle o a b oade sus ainabili y ans o ma ion.
High a el ime expendi u es, u he mo e, ha e a de imen al impac
on bo h men al and physical heal h (Giu ge e al., 2020; Lo enz, 2018;
Milne e al., 2017; Xiao e al., 2020), hey inc ease s ess (S azdins
e al., 2016), lowe p oduc i i y (Giu ge e al., 2020; Lo enz, 2018;
Milne e al., 2017; Xiao e al., 2020), and limi he ime a ailable o
heal hy ea ing habi s (Senia e al., 2017), physical ac i i y (B ownson
e al., 2005) as well as o es and leisu e (Xiao e al., 2020).
Mo e so han socio-economic ac o s, he esiden ial con ex shapes
mobili y beha iou (Gao e al., 2022; Lucas e al., 2018). U ban a eas
usually o e a dense public anspo ne wo k and highe accessibili y
o local se ices daily such as supe ma ke s o schools wi hin close
dis ances, han u al a eas (Ca mo e al., 2017; Holian & Kahn, 2015).
Wi hin ci ies, gi en hei concen ic s uc u e, a el dis ances o poin s
o in e es (POI) dec ease when people li e close o he ci y cen e (Pa k
e al., 2021). Neighbou hoods wi h a g ea e p opo ion o high-income
households o e a mo e ex ensi e ange o anspo a ion al e na i es
and a close p oximi y o essen ial se ices. Con e sely, disad an aged
neighbou hoods a e less well connec ed o public anspo a ion,
employmen oppo uni ies, and ins i u ions o highe educa ion
(Nicole i e al., 2023).
Con enien access o essen ial se ices o en guides esidency de-
cisions (Hesse & Scheine , 2010), likely d i ing up en s in a eas wi h
high accessibili y (Gup a e al., 2022; Renne , 2022). E idence om
Munich (Ge many) and o he coun ies sugges s ha low-income esi-
den s a e inc easingly p iced ou o mo e cen al u ban a eas (Dohnke
e al., 2012; S ini asan e al., 2020), which has a nega i e impac on
hei access o public anspo a ion and o he se ices (Bü ne e al.,
2013; Hochs enbach & Mus e d, 2018). In gene al, po e y seg ega ion
in Ge man ci ies has inc eased du ing he las wo decades (Helbig &
J¨
ahnen, 2018). C ucially, seg ega ion and gen i ica ion a e no d i en
by income alone. Indi iduals wi h highe le els o educa ion p e e o
li e in cen al neighbou hoods, as e idences by s udies on a ious Eu-
opean ci ies (Booi & Bo e man, 2020; Dohnke e al., 2012; Hoch-
s enbach & Mus e d, 2018; L´
opez-Gay e al., 2020; S e ze , 2017). Thei
decision is in luenced by a desi e o gain imp o ed access o public
anspo a ion, POIs, cul u al o e ings and, mos no ably, knowledge
indus y employmen oppo uni ies in u ban a eas Flo ida (2019);
Konie zka and Ma yno ych (2023); Reckwi z (2021); Vos e al. (2016).
Agains his backg ound, we a e in e es ed in how highly educa ed
indi iduals' daily mobili y in Ge man la ge ci ies has changed du ing he
las wo decades. We assume ha highly educa ed indi iduals mo e
o en li e in la ge ci ies and ha his end has inc eased o e he yea s
(1a). In addi ion, we hypo hesise ha wi hin la ge ci ies, highly
educa ed indi iduals li e in mo e cen al a eas han o he social g oups
(1b). We u he assume ha a el dis ances o POIs o commu es,
e ands, leisu e, and shopping a e sho e o highly educa ed in-
di iduals han o o he social g oups (2a) due o hei mo e cen al
esiden ial loca ion (2b). These sho e dis ances should allow highly
educa ed indi iduals o choose sus ainable modes o daily mobili y mo e
o en (3a) wi hou incu ing highe a el ime expendi u es han o he
social g oups (3b).
3. Me hod and da a
The ollowing sec ion in oduces he Mobili y in Ge many (MiG)
su ey ha p o ides in o ma ion on daily mobili y beha iou , socio-
economic and socio-demog aphic cha ac e is ics o u ban esiden s in
2002 and 2017. We explain he sampling p ocedu e, da a collec ion and
weigh ing, and in oduce a iables such as a el dis ance, ime and
speed. The s a is ical app oach employs mul i-le el eg ession models o
sc u inise changes in mobili y pa e ns, whe eas ac ional mul inomial
logi models a e employed o examine he change in anspo mode
p e e ence.
3.1. Sampling, da a, and weigh ing
We use indi idual-le el da a on daily mobili y beha iou , he so-
cioeconomic s a us and socio-demog aphic backg ound o Ge man es-
iden s in la ge ci ies om he MiG su ey. Conduc ed by he social
esea ch ins i u e in as in 2002 and 2017, he MiG su ey, commissioned
S. Geo ge e al.
Ci ies 156 (2025) 105507
2
by he Fede al Minis y o T anspo and Digi al In as uc u e, is he la ges
and mos comp ehensi e o i s kind in Ge many. I is ep esen a i e o
he en i e Ge man popula ion ac oss all ages, and we ob ained access
h ough he Ge man Ae ospace Cen e.
The sampling o he MiG employs a wo-s age app oach, s a ing
wi h household in e iews ollowed by indi idual in e iews. House-
holds we e andomly selec ed ac oss Ge many based on popula ion
egis e s as well as, in 2017, andom-digi -dialling o bo h landline and
mobile phone numbe s ( iple- ame design). A ques ionnai e was
adminis e ed o each household, eques ing in o ma ion abou he
household and o e ing he op ion o a mo e de ailed ollow-up in e -
iew wi h each membe . These in e iews could be conduc ed ia pos al
mail, compu e -assis ed elephone in e iews, o online, and we e
a ailable in mul iple languages. All esponden s consen ed o he ano-
nymised use o hei da a o scien i ic s udies on mobili y beha iou
(in as ;, 2019b). In o ma ion abou daily jou neys we e collec ed ia
jou ney logs ha esponden s we e asked o ill ou a a andomly
selec ed day wi hin he span o 1.5 yea s du ing he ield phase o each
su ey o ensu e he da a is nei he biased by day-o -week no seasonal
biases (in as ;, 2019a). The ne esponse a e was 39 % in 2002 (Follme
& Kune , 2003) and 6 % in 2017 a he indi idual le el (RP3 esponse
a e acco ding o AAPOR 2016 which assumes he p opo ion o cases o
unknown eligibili y ha a e eligible o be equal o he p opo ion o
eligible uni s among all uni s in he sample). The decline in esponse
a es in such su eys has been no ed by se e al s udies e.g., (Czajka &
Beyle , 2016). The iple- ame design employed in 2017 should ensu e
ep esen a i eness despi e a dec ease in esponse a es.
Ou sample only includes esponden s aged 18 o 59, as child en,
adolescen s, and pensione s' daily mobili y beha iou di e s signi i-
can ly om hose wi hin wo king-age. (The ypical e i emen age in
Ge many, which was 62 yea s in 2002 and 64 yea s in 2017, could no be
accu a ely implemen ed due o he ac ha he in o ma ion on he age
o esponden s was made a ailable by he Ge man Ae ospace Cen e as a
ca ego ical a iable only). We u he es ic ou sample o la ge ci ies
in Ge many, which a e, in acco dance wi h he de ini ion p o ided by
he Ge man Fede al Minis y o Digi al A ai s and T anspo , ci ies wi h
mo e han 500,000 inhabi an s: Be lin, Hambu g, B emen, Do mund,
Essen, Duesseldo , Hanno e , Cologne, Bonn, F ank u am Main,
Mannheim, Nu embe g, S u ga , Munich, Leipzig, and D esden
(BMVI;, 2011). Ou analy ical sample consis s o 4168 indi iduals om
2437 households ha eco ded 16,419 jou neys in 2002, and 26,036
indi iduals om 15,846 households ha wen on 102,774 di e en
jou neys in 2017. In o de o es hypo hesis 1a, we ha e included e-
sponden s om ou side he 16 ci ies as well. Ga he ing in o ma ion
abou daily mobili y poses signi ican challenges, pa icula ly in e ms o
de e mining exac jou ney de ails. To add ess hese challenges, e-
sponden s we e p o ided wi h no ebooks o online op ions o eco ding
hei jou neys. When esponden s eco ded hei jou neys e ospec-
i ely, an in e iewe was ypically p esen o assis and suppo was
a ailable du ing he whole da a collec ion pe iod. Jou ney endpoin s
we e de e mined by in e ac i e lis s o by w i ing down add esses.
T a el dis ance, du a ion and speed we e la e on calcula ed based on
he in o ma ion p o ided by esponden s.
The da a a e weigh ed acco ding o he season and day o he week o
he su ey, he esponden 's place o esidence, household size,
employmen s a us, educa ion, age, and gende . This ed essmen ad-
jus s o mino imbalances in he ep esen a ion o speci ic social g oups
o esponden s who we e su eyed abou hei daily mobili y on a
pa icula weekday o du ing a speci ic season. Bu he da a weigh s also
add ess he di e en selec ion p obabili ies associa ed wi h he su ey's
iple- ame design and adjus o non- esponse a es among speci ic
sub-popula ions (Follme & G uschwi z, 2019). (Fo example, young
single males a e less likely o consen o pa icipa e in he su ey and a e
less accessible ia elephone.) A non- esponse su ey was u ilised o
collec p elimina y in o ma ion on ini ial non- esponden s. The e we e
no signi ican di e ences in a el pa e ns be ween esponden s and
non- esponden s. Howe e , he main su ey exhibi ed a sligh unde -
ep esen a ion o esponden s wi h a high olume o daily jou neys
(Follme & G uschwi z, 2019). Ano he po en ial bias could a ise om
non-mobile su ey pa icipan s. Table A1 in he Online Appendix com-
pa es he educa ion le el o mobile and non-mobile u ban esiden s aged
18 o 59. Non-mobile esponden s end o ha e lowe educa ion le els.
3.2. Va iables
Ou analyses examines h ee main ou come a iables: a el dis-
ance, a el ime, and a el speed o each jou ney. To exclude
implausible alues, we immed he op one pe cen ile o each o hese
a iables. As shown in Fig. A1 and A2 in he Appendix, all h ee ou come
a iables ha e a skewed dis ibu ion, which is why we chose o log-
ans o m hem o analysis (Wes , 2022). Responden s' jou neys we e
di ided in o ou di e en pu poses: commu ing, leisu e, e ands, and
shopping. The main mode o anspo indica es he p edominan mode
o anspo used by esponden s o an indi idual jou ney (by ca ,
cycling/walking, o by public anspo a ion). The dis ance o he ci y
cen e gi es he ai line dis ance o he geome ic cen e o esponden s
1 km-by-1 km neighbou hood g id o he geome ic cen e o he ci y.
Ou s a is ical es ima es o he e ec s o esponden s le el o
educa ional a ainmen a e con olled o e ec s o household income,
employmen s a us (employed o no ), gende , age, whe he o no he e
a e child en in he household, and whe he o no he household owns
ca s. Household income was ca ego ised in o quan iles based on he
equi alised income acco ding o household size. To es hypo hesis 1a,
we use a a iable o dis inguish be ween esponden s who li e wi hin
one o he 16 ci ies included in ou main sample and hose who li e
elsewhe e in Ge many; all o he hypo heses a e es ed only on e-
sponden s who li e wi hing hese 16 ci ies. Table A2 in he Appendix
p o ides mo e in o ma ion on he exac de ini ion o each a iable while
Table 1 p o ides summa y s a is ics.
3.3. S a is ical app oach
The da a a e clus e ed a di e en le els (mul iple ips pe each
indi idual, indi iduals clus e ed wi hin households), he e o e we
applied mul i-le el app oaches ha accoun o his da a s uc u e.
Beyond ha , ou dependen a iables equi e di e en app oaches as
hey a e measu ed a di e en s a is ical scales. Two o ou main
dependen a iables - a e age a el dis ance and a el ime pe
jou ney - a e me ic bu highly skewed and whe e log- ans o med o be
used in a linea eg ession analysis which mos closely es s ou as-
sump ions exp essed in hypo heses 2a, 2b, and 3b, wi hou iola ing
s a is ical model equi emen s. We es ima ed linea andom in e cep
ixed slope mul ile el models wi h educa ion as he main explana o y
a iable and a el dis ance and a el ime as dependen a iables, plus
a ious con ol a iables (see below). We speci ied ixed slopes because
we do no es whe he he s a is ical associa ion be ween, o example,
educa ion and a el dis ance a ies be ween di e en households.
Howe e , we assume ha di e en households di e in e ms o he
a e age a el dis ance o hei membe s, which we accoun o by
speci ying andom in e cep s (and by con olling o some household
cha ac e is ics, see below). We es ima ed hese models sepa a ely o
2002 and 2017 as well as sepa a ely o each mobili y pu pose
(commu ing, e ands, shopping, leisu e). The linea mul ile el eg es-
sion models a e de ined as ollows:
Yijk, =b000, +b0p0, Xpk, +b00q, Zq, +eijk, +u0j0, + 00k, (1)
Whe e Yijk a e a el dis ance/ ime/speed o jou ney i o indi idual j
who is a membe o household k and b000 is he g and ac oss-households
in e cep . 1ˆ
a
€
¦P a e p edic o s X a he indi idual le el (e.g., educa ional
a ainmen ) and b0p0 each o hei slopes ixed ac oss households. 1ˆ
a
€
¦Q
a e p edic o s Z a he household le el (e.g., household income, dis ance
S. Geo ge e al.
Ci ies 156 (2025) 105507
3
o ci y cen e o place o esidency, see below), wi h each o hei slopes
b00q. eijk a e esidual e o s a he jou ney le el, u0j0 esidual e o s a
indi idual le el and 00k esidual e o s a he household le el. Las ly,
s ands o he yea 2002 o 2017. We es ima e mul i-le el OLS eg ession
models using he MIXED ou ine o S a a 17. We calcula ed linea p e-
dic i e ma gins o bo h he a e age a el dis ance and a e age a el
ime pe jou ney based on he eg ession models o isually highligh he
di e ences be ween he ou educa ional g oups. As hese dependen
a iables we e log- ans o med o be included in he eg ession models,
we back- ans o med hem o p o ide he a el dis ance in kilome es
and a el ime in minu es acco ding o he ollowing o mula:
Y=
Yloge(2)
Whe e
Y is he back- ans o med p edic i e alue o he independen
a iable,
Ylog he log- ans o med p edic i e alue o he independen
a iable, and e is he ma hema ical cons an e (Eule 's numbe ).
To es hypo heses 3a ha examines he main mode o ans-
po a ion, we employed ac ional mul inomial logi models o in es i-
ga e he p ima y mode o anspo chosen by esponden s o di e en
a el pu poses. Fo his, we examined he equency o using speci ic
modes o anspo - ca , public anspo , bike/walking - o all jou neys
on he day o ques ioning and calcula ed a pe cen age be ween 0 and 1
o each mode o anspo and pe son. F ac ional mul inomial logi
models a e sui able in cases whe e he dependen a iable ep esen s he
p obabili y o choosing a speci ic al e na i e ou o mu ual exclusi e
al e na i es. This is ue in ou case, whe e choosing a ca as he main
mode o anspo o a ip means deciding agains public anspo o
cycling/walking as he main mode o anspo . By es ima ing ac ional
mul inomial logi models, we a e able o examine he ela i e p oba-
bili ies o choosing di e en modes o anspo o each a el pu pose,
while accoun ing o he in e dependencies be ween anspo choices.
The ull model equa ion is as ollows:
q(b) = ∑
N
n=1∑
M
m=1
yimlog(gm(xn,b) ) (3)
gm(xn,b) =
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
exnbm
1+∑M−1
k=1exnbk
,i m<M
1
1+∑M−1
k=1exnbk
,i m=M
⎫
⎪
⎪
⎪
⎪
⎬
⎪
⎪
⎪
⎪
⎭
(4)
Whe e bj. =(b1m, …, bpm.) ep esen s a ec o o coe icien s o indi-
idual, household and neighbou hood p edic o s P and he m,m <M
mode o anspo a ion. S anda d e o s a e clus e ed a he household
le el. We ha e calcula ed hese models sepa a ely o each mobili y
pu pose (commu ing, e ands, shopping, leisu e) and sepa a e o 2002
and 2017.
Rega ding hypo heses 1a and 1b uni s o analysis a e indi iduals, no
ips, and he dependen a iables - place o esidency in la ge ci ies and
dis ance o cen e o place o esidency - a e household le el cha ac-
e is ics, which ende s mul ile el models obsole e. Wi h ega d o hy-
po hesis 1b, he dis ance o he ci y cen e o he esponden s' place o
esidence is highly skewed and was log- ans o med. We es ed hy-
po heses 1b by applying linea eg ession models wi h dis ance o he
ci y cen e as he dependen a iable. Hypo heses 1a equi ed a logis-
ical eg ession model as he dependen a iable is bina y (li ing in la ge
ci ies o no ).
To add ess po en ial con ounding ac o s, we in oduced se e al
con ol a iables which p e ious esea ch has iden i ied as in luencing
daily mobili y beha iou and a el ime expendi u es. This includes
household income, employmen s a us (employed o no ), gende , age,
he p esence o child en in he household, and whe he o no he
household owns ca s. Highe household income is gene ally associa ed
wi h a highe le el o educa ional a ainmen and also in luences a el
pa e ns (Ca mo e al., 2017; Delcl`
os-Ali´
o & Mi alles-Guasch, 2018).
Women end o engage in mo e sus ainable and ac i e a el beha iou s,
leading o inc eased a el ime, while men end o a el longe dis-
ances and use ca s mo e o en (Goel e al., 2023; Roos e al., 2020). Age
a ec s a el beha iou h ough physical mobili y and agili y (D˙
edel˙
e
e al., 2020; Ga cía Rom´
an & G acia, 2022). The employmen s a us o
indi iduals, oo, is associa ed wi h dis inc commu ing pa e ns (Roos
e al., 2020). Las ly, indi iduals wi h child en o en need o a el longe
dis ances o child- ela ed e ands and a e mo e likely o use ca s
(D˙
edel˙
e e al., 2020; Delcl`
os-Ali´
o & Mi alles-Guasch, 2018; Lee e al.,
2018).
To es hypo heses 1b and 2b, we equi e da a on he dis ance o he
ci y cen e o esponden s' place o esidency which he MiG su ey only
p o ides o 91 % o ou main sample and only o 2017. The e o e, o
examine hypo hesis 2b, we i s eplica e he eg ession model
explaining a el dis ances wi hou he a iable ‘dis ance o he ci y
cen e’ wi h he educed sample size and compa e he esul s wi h he
eg ession model o he ull sample. This app oach enables us o elimi-
na e he likelihood ha a ia ions in he ou comes a e due o he al e ed
Table 1
Summa y S a is ics.
2002 2017
N Mean/
Pe cen
N Mean/
Pe cen
Da a Uni : Jou neys
T a el dis ance in kilome es 4033 11.17 23,377 13.53
T a el ime in minu es 4131 29.80 25,396 35.54
T a el speed in km/h 4016 19.57 23,181 19.02
Dis ance o he ci y cen e
A
– – 24,110 5.85
Main mode o anspo
1. Ca 1808 43.57 8440 35.38
2. By bike o by oo 1298 31.29 8480 35.65
3. Public anspo 1043 25.14 6867 28.87
T a el pu pose
1. Leisu e 945 22.82 5900 23.05
2. Commu es 1717 41.44 11,146 43.54
3. E ands 635 15.33 4636 18.11
4. Shopping 790 19.06 3208 12.53
Da a Uni : Indi iduals
Educa ion
1. Elemen a y school/no
deg ee
598 16.89 1934 9.19
2. Seconda y school 1005 28.41 4818 22.88
3. A-le el 405 11.43 4412 20.95
4. Uni e si y 1531 43.27 9891 46.97
Gende : Female 2129 51.07 12,758 49.75
Age g oups
1. 18–30 yea s 1072 25.73 7601 29.64
2. 31–40 yea s 1337 32.08 6871 26.80
3. 41–50 yea s 967 23.20 6045 23.57
4. 51–67 yea s 791 18.99 5125 19.99
Employed 2980 71.64 19,850 77.45
Da a Uni : Households
Household income
1. Ve y low 195 5.35 1996 7.78
2. Low 567 15.56 2650 10.34
3. Middle 1418 38.93 9779 38.13
4. High 1118 30.71 8708 33.96
5. Ve y high 344 9.45 2510 9.79
Households wi h kids 1426 34.35 9031 35.38
Households wi h ca s 1041 25.32 5264 21.04
No es. The a iables a el dis ance, a el ime and a el speed a e p esen ed
he e in hei o iginal o m be o e unde going log- ans o ma ion o ou
analysis.
A
The a iable o he dis ance o he ci y cen e is exclusi ely accessible
wi hin he subse o da a, he MiD-local da a se om 2017. This da a se con-
ains indi iduals who ha e chosen o documen hei exac add ess.
S. Geo ge e al.
Ci ies 156 (2025) 105507
4
sample a he han he explana o y a iable o he dis ance o he ci y
cen e o esponden s neighbou hoods. Subsequen ly, we inse he
a iable ‘dis ance o he ci y cen e’ o de e mine i s impac on he
a ia ion in hese dis ances be ween educa ional g oups.
4. Resul s
This chap e p esen s he esul s o ou h ee hypo heses. We begin
by examining he place o esidence o highly educa ed indi iduals.
Secondly, we analyse he a e age a el dis ance and, hi dly, a el
ime pe jou ney.
4.1. Educa ional a ainmen and esiden ial choice
(1a) Highly educa ed indi iduals mo e o en li e in la ge ci ies and
his end has inc eased o e he yea s. (1b) Wi hin la ge ci ies, highly
educa ed indi iduals li e mo e cen al han o he social g oups.
We compu ed he linea p edic i e ma gins o he bina y ou come
a iable esiding in la ge ci ies o 2002 and 2017 based on he logis ical
eg ession model o hypo hesis (1a). Resul s in Fig. 1 show ha in-
di iduals wi h a uni e si y deg ee a e mo e likely o eside in la ge ci ies
han o he g oups, and his end inc eased be ween 2002 and 2017. In
2017, indi iduals wi h a uni e si y deg ee we e h ee imes mo e likely
o li e in la ge ci ies in Ge many han hose wi h elemen a y school
deg ee. These indings a e consis en wi h hose o simila s udies con-
duc ed in Spain (Gonz´
alez-Leona do e al., 2019) and England (B idge,
2006).
Rega ding hypo heses (1b), we es ed i people wi h highe educa-
ional a ainmen gene ally li e mo e cen al wi hin he 16 la ge ci ies o
ou main sample, meaning i hei 1 km-by-1 km neighbou hoods a e
close o he geog aphical ci y cen e. Resul s shown in Fig. 2 suppo
ha assump ion. In la ge ci ies, uni e si y g adua es li e closes o he
geog aphical ci y cen e ou o all educa ional g oups, be ween 0.53 and
1.52 km close han o he educa ional g oups. Indi iduals wi h highe
educa ion no only mo e o en li e in la ge ci ies bu also mos cen al
wi hin hese a eas.
4.2. T a el dis ances and esiden ial choice
(2a) a el dis ances o POIs o commu es, e ands, leisu e, and
shopping, a e sho e o highly educa ed indi iduals compa ed o o he
social g oups (2b) due o hei mo e cen al esiden ial loca ion.
We es ima ed a se ies o mul ile el OLS eg essions wi h indi iduals
in la ge ci ies o analyse he daily a el dis ances o di e en educa-
ional g oups, p esen ed in Table 2. Fig. 3 displays he linea p edic i e
ma gins o hese models o a el dis ances pe jou ney, sepa a ely o
each a el pu pose o indi iduals esiding in la ge ci ies, and inde-
penden o he e ec s o all con ol a iables (employmen s a us,
gende , age, child en in he household, ca s in he household). In 2002,
a el dis ances did no a y signi ican ly among educa ional g oups. In
2017, howe e , people wi h uni e si y deg ees a elled signi ican ly
sho e dis ances pe jou ney o leisu e, shopping and e ands han
indi iduals wi hou highe educa ion. Fo example, uni e si y g adua es
a el 20 % ewe kilome es o shopping ips and 15 % ewe kilo-
me es o e ands compa ed o people wi h a seconda y school deg ee.
Rega ding commu es o and om wo k in 2017, we did no ind any
signi ican di e ences be ween educa ional g oups.
To es hypo hesis (2b), whe he he p oximi y o he ci y cen e
accoun s o he educed a el dis ances among he highe educa ed, we
eplica ed he eg ession model o hypo heses (2a) bu now including
he dis ance o he ci y cen e o indi iduals' place o esidency as an
addi ional independen a iable. Since his a iable is only a ailable o
a subse o ou sample om 2017, we i s eplica ed he eg ession
model o hypo heses (2a) o ensu e ha he e a e no meaning ul di -
e ences o his sub-sample compa ed o ou main sample. These Resul s
can be ound in Table A5 in he Online Appendix and show no ele an
di e ences compa ed o he main sample. Nex , we epea ed he
eg ession model o es hypo hesis (2a), including esiden ial dis ance
o he ci y cen e as an independen a iable. As expec ed, cen ali y has
a signi ican impac on a el dis ance o di e en a el pu poses (see
Fig. 4): Including esiden ial dis ance o he ci y cen e in he models
educes he di e ences ega ding he a e age dis ance pe jou ney be-
ween he di e en educa ional g oups, suppo ing Hypo hesis 2b.
.09
.12
.17
.23
.11
.14
.21
.33
.00
.05
.10
.15
.20
.25
.30
.35
Elemen a y School Seconda y School A-Le el Uni e si y
a
e
ana
ilop
o
e
ma
n
i
gni
il
o
y i
l
ibabo
P
2002 2017
Fig. 1. P obabili ies o educa ional g oups li ing in la ge ci ies.
No es.Linea p edic i e ma gins de i ed om he eg ession models in Table A3 in he Online Appendix. In e p e a ion: In 2017, indi iduals wi h highe educa ion
had a 33 % chance o li e in a la ge ci y (and a 67 % chance o li e in o he a eas o Ge many, espec i ely) whe eas indi iduals wi h seconda y educa ion had a 14 %
change o li e in la ge ci ies, each ne o all o he e ec s.
S. Geo ge e al.
Ci ies 156 (2025) 105507
5
Fig. 2. Dis ance o ci y cen es in la ge ci ies in 2017.
No es. Linea p edic i e ma gins de i ed om eg ession models in Table A4 in he Online Appendix. In e p e a ion: The mean dis ance be ween he esidence o
uni e si y g adua es in la ge ci ies and he geome ic ci y cen e (ai line dis ance) is 5.03 km, compa ed o 6.46 km o indi iduals wi h seconda y educa ion, ne o
all o he e ec s.
Table 2
ML-OLS eg essions: Kilome es pe jou ney.
(1) (2) (3) (4) (5) (6) (7) (8)
Yea 2002 2017 2002 2017 2002 2017 2002 2017
Va iables Commu es Leisu e E ands Shopping
Elemen a y school o no deg ee −0.064 0.025 0.090 0.048 0.140 0.049 0.062 0.161
(0.092) (0.057) (0.101) (0.080) (0.120) (0.095) (0.111) (0.090)
A-le els −0.026 −0.026 0.277** −0.055 0.104 −0.080 −0.101 −0.118*
(0.093) (0.038) (0.106) (0.045) (0.143) (0.054) (0.130) (0.052)
Uni e si y −0.137 0.023 0.104 −0.111** −0.018 −0.157** 0.050 −0.201***
(0.073) (0.034) (0.085) (0.042) (0.103) (0.049) (0.093) (0.046)
Household income: Ve y low −0.091 0.125 −0.205 0.079 −0.027 0.185 0.166 −0.038
(0.187) (0.078) (0.178) (0.087) (0.256) (0.100) (0.189) (0.103)
Low −0.066 −0.087 0.181 −0.056 −0.157 0.111 0.073 −0.038
(0.113) (0.051) (0.125) (0.066) (0.145) (0.069) (0.121) (0.067)
High 0.027 −0.013 0.102 0.015 0.130 0.002 0.187*0.031
(0.071) (0.027) (0.093) (0.037) (0.106) (0.044) (0.095) (0.040)
Ve y high 0.354** −0.008 −0.117 0.074 0.524** 0.026 −0.096 −0.044
(0.112) (0.038) (0.157) (0.051) (0.183) (0.056) (0.164) (0.055)
Female −0.216*** −0.220*** −0.044 −0.058*−0.187*−0.084*−0.093 0.030
(0.055) (0.022) (0.059) (0.024) (0.079) (0.033) (0.074) (0.031)
Age G oups: 18–30 0.338*** 0.065 0.137 0.025 0.277*0.079 −0.006 −0.026
(0.101) (0.038) (0.119) (0.043) (0.141) (0.058) (0.133) (0.051)
31–40 0.153 0.063 −0.035 −0.030 0.310*−0.125*0.080 −0.065
(0.091) (0.033) (0.113) (0.042) (0.134) (0.051) (0.113) (0.047)
41–50 0.058 0.076*0.072 −0.029 0.192 −0.068 0.041 −0.039
(0.095) (0.033) (0.106) (0.040) (0.126) (0.050) (0.108) (0.043)
Employed 0.240*0.184*** 0.059 0.047 −0.037 0.088 −0.039 0.072
(0.106) (0.050) (0.080) (0.037) (0.089) (0.046) (0.079) (0.043)
Household wi h kids −0.026 −0.127*** −0.124 −0.182*** −0.215*−0.281*** −0.058 0.010
(0.074) (0.028) (0.093) (0.037) (0.108) (0.040) (0.091) (0.039)
Household wi h ca s 0.352*** 0.374*** 0.498*** 0.289*** 0.315** 0.317*** 0.602*** 0.426***
(0.088) (0.029) (0.104) (0.039) (0.124) (0.046) (0.109) (0.042)
Cons an 1.343*** 1.486*** 0.840*** 1.342*** 0.646*** 1.024*** −0.090 0.146*
(0.160) (0.068) (0.146) (0.066) (0.181) (0.082) (0.151) (0.072)
N jou neys 2913 21,152 3325 20,863 2203 14,275 2625 12,032
N indi iduals 1137 8282 1221 8086 770 5297 1068 5505
No es. Robus s anda d e o s in pa en heses.
The a iable kilome es pe jou ney was log- ans o med; In e p e a ion: On a e age, indi iduals wi h uni e si y educa ion a el 20.1 % sho e dis ances on jou neys
o shopping des ina ions compa ed o indi iduals wi h seconda y school deg ees ( e e ence ca ego y) in 2017, ne o all o he e ec s.
***
p<0.001.
**
p<0.01.
*
p<0.05 ( wo- ailed es s).
S. Geo ge e al.
Ci ies 156 (2025) 105507
6
Fig. 3. T a el dis ance pe a el pu pose by educa ional g oups.
No es. Linea p edic i e ma gins o a el dis ances o each a el pu pose by educa ional le el, ne o all o he e ec s. The ma gins a e based on eg ession models
om Table 2 in he Online Appendix. As he a el dis ance pe jou ney was log- ans o med o inco po a ion in o he eg ession model, i was back- ans o med o
kilome es pe jou ney o his igu e (see S a is ical App oach). In e p e a ion: Uni e si y g adua es a el, on a e age, 1.28 km pe jou ney o and om shopping
des ina ions, ne o all o he e ec s.
1.78
1.54
1.16
.58
1.83
1.54
1.16
.46
1.73
1.46
1.10
.33
1.84
1.39
.96
.26
.00
.20
.40
.60
.80
1.00
1.20
1.40
1.60
1.80
2.00
Wo k Leisu e E ands Shopping
go
l(
s
e
e
mo
liK -)dem o sna
Elemen a y School Seconda y School A-Le el Uni e si y
Fig. 4. T a el dis ance pe a el pu pose by educa ional g oups, con olled o esiden ial dis ance o he ci y cen e.
No es. Linea p edic i e ma gins o a e age a el dis ances pe jou ney sepa a ed o each a el pu pose by educa ional g oups in la ge ci ies, including esiden ial
dis ance o he ci y cen e as an independen a iable, ne o all o he e ec s. The ma gins a e based on eg ession models om Table A6. As he a el dis ance pe
jou ney was log- ans o med o inco po a ion in o he eg ession model, i was back- ans o med o kilome es pe jou ney o his igu e (see S a is ical App oach).
In e p e a ion: Uni e si y g adua es a el, on a e age, 1.30 km pe jou ney o and om shopping des ina ions, ne o all o he e ec s including esiden ial dis ance
o he ci y cen e.
S. Geo ge e al.
Ci ies 156 (2025) 105507
7
4.3. Sus ainable a el and ime cos s
(3a) Sho e a el dis ances allow highly educa ed indi iduals o
choose sus ainable modes o anspo (3b) wi hou incu ing highe
ime cos s.
To es hypo hesis (3a), we examined he likelihood o indi iduals
wi h di e en le els o educa ional a ainmen in la ge ci ies o choose
be ween h ee main modes o anspo o any gi en jou ney: d i ing by
ca , public anspo , and walking/cycling. We es ima ed ac ional
mul inomial logi models o each ip pu pose and yea . We included all
con ol a iables excep whe he he household owns ca s, as his is
al eady a pe quisi e o he ca ego y d i ing by ca . The esul s indica e
ha uni e si y g adua es p e e walking/cycling as well as using public
anspo o e d i ing by ca o commu es om and o wo k. Fo
shopping, hey p e e walking/cycling o e d i ing and using public
anspo . Indi iduals wi h highe educa ion do no show a p e e ence
o d i ing by ca o e walking/cycling o using public anspo o any
o he a el pu poses. Sus ainable modes o anspo such as public
anspo , cycling, and walking incu highe a el ime expendi u es
due o hei ela i ely slowe a el speed. To ensu e ha his gene al
assump ion applies o ou cu en sample, we calcula ed he a e age
a el speed pe jou ney in km/h (log- ans o med) and applied a
mul ile el eg ession model o show he associa ion be ween mo e
sus ainable modes o anspo and a el speed (see Table A9 in he
Online Appendix). We ind ha using sus ainable modes o anspo as
main o anspo is associa ed wi h highe a el ime expendi u es and
ha indi iduals wi h highe le el o educa ional a ainmen in la ge
ci ies indeed a el a slowe speed.
We applied a mul ile el eg ession model o es hypo heses (3b) and
ound, ha hose wi h highe le els o educa ion do no incu highe
a el ime expendi u es o e ands, shopping, o leisu e ac i i ies han
o he educa ional g oups (see Fig. 5). In ac , hey spen on a e age
abou 7 % less ime pe jou ney o e ands and shopping han hose o
seconda y educa ion o lowe , abou 11 % less ime han hose wi h
elemen a y school educa ion on leisu e ac i i ies in 2017 (see Table A8
in he Online Appendix). Howe e , indi iduals wi h highe le els o
educa ion spen mo e ime on commu ing in 2017, which is o be ex-
pec ed since hei commu ing dis ance does no signi ican ly di e om
o he educa ional g oups. Despi e hei p e e ence o slowe , sus ain-
able modes o anspo , people wi h highe le els o educa ional
a ainmen ha li e in la ge ci ies do no incu highe a el ime ex-
pendi u es o daily jou neys han o he educa ional g oups due o hei
sho e dis ances o POIs in 2017. We do no ind hese e ec s in ou
sample o 2002 when he le el o educa ion was no ye as meaning ul in
p edic ing indi iduals' place o esidency (Booi & Bo e man, 2020).
5. Discussion
Ou esul s ex end cu en esea ch on sus ainable daily mobili y in
h ee di e en a eas: Fi s , while we ind ha wi hin la ge ci ies in-
di iduals wi h a highe le el o educa ion a el a slowe speeds and
mo e o en op o sus ainable modes o anspo han o he educa ional
g oups (Hudde, 2022), ou esul s show ha hey ne e heless do no
ha e o in es mo e ime in hei daily mobili y. Second, we show ha
his can be explained by hei sho e dis ances o POIs, which, hi d, can
pa ly be a ibu ed o hem li ing close o he geog aphical cen e o
la ge ci ies. In his ega d, people wi h highe le els o educa ional
a ainmen di e om people wi h high household incomes. While he
Fig. 5. T a el du a ion pe a el pu pose by educa ional g oups.
No es. Linea p edic i e ma gins o minu es pe jou ney o each a el pu pose by le el o educa ional a ainmen o esiden s o la ge ci ies, ne o all o he e ec s.
Ma gins a e based on eg ession models om Table A8 in he Online Appendix. As he a el ime pe jou ney was log- ans o med o inco po a ion in o he
eg ession model, i was back- ans o med o minu es pe jou ney o his igu e (see S a is ical App oach). In e p e a ion: Uni e si y g adua es spen , on a e age,
11.36 min pe jou ney o and om shopping des ina ions, ne o all o he e ec s.
S. Geo ge e al.
Ci ies 156 (2025) 105507
8
