aspects of spatial proximity and - ugentkboussau/dissertation_boussauw_a... · 2011-02-16 · en...

1

Aspects of spatial proximity and

sustainable travel behaviour in Flanders:

A quantitative approach

2

Cover: The picture on the cover (© Rocky Zutterman) is a remake of the

painting “Il ciclista attraversa la città”, by the futurist Fortunato Depero

(1945). The Futurism movement revered speed, technology and industry,

and was thus not particularly advocating sustainable mobility, despite

the iconic use of a bicycle in this work of art.

Copyright © Kobe Boussauw, Department of Geography, Faculty of

Sciences, Ghent University, 2011. All rights reserved. No part of this

publication may be reproduced, stored in a retrieval system, or transmit-

ted, in any form or by any means, electronic, mechanical, photocopying,

recording, or otherwise, without permission in writing from the copyright

holder(s).

ISBN: 978-94-906-9553-8

Legal deposit: D/2011/12.134/6

NUR: 755/901/904/976

The research reported in this dissertation was conducted at the Social

and Economic Geography research unit, Department of Geography,

Faculty of Sciences, Ghent University, and funded by the Policy Research

Centre on Regional Planning and Housing - Flanders (Steunpunt Ruimte

en Wonen 2007-2011).

3

Aspects of spatial proximity and

sustainable travel behaviour in Flanders:

A quantitative approach

Aspecten van ruimtelijke nabijheid en

duurzaam verplaatsingsgedrag in Vlaanderen:

Een kwantitatieve benadering

Proefschrift

Proefschrift aangeboden tot het behalen van de graad van

doctor in de wetenschappen: geografie

vrijdag 4 maart 2011

door

ir. Kobe Boussauw

4

Promotor:

prof. dr. Frank Witlox, Universiteit Gent

Samenstelling examencommissie:

prof. dr. Georges Allaert, Universiteit Gent

prof. dr. David Banister, University of Oxford

prof. dr. Peter Cabus, Katholieke Universiteit Leuven

prof. dr. Ben Derudder (voorzitter), Universiteit Gent

prof. dr. Martin Dijst, Universiteit Utrecht

prof. ir. Dirk Lauwers, Universiteit Gent

prof. dr. ir. Jacques Teller, Université de Liège

dr. Veronique Van Acker, Universiteit Gent

5

Contents Preface 9

Chapter 1: Introduction 15

1.1 Summary 15

1.2 Sustainable mobility, climate change and peak oil 16

1.3 The time-distance-space relationship 25

1.4 The rebirth of distance 33

1.5 Flanders and Brussels: policy context 45

1.6 Research questions, conceptual framework and implementation 47

1.7 Overview of the research 51

References 55

Chapter 2: Introducing a commute-energy performance index 63

Abstract 63

2.1 Introduction 64

2.2 Energy use and urban spatial structure 64

2.3 Limitations of studying the home-to-work commute 66

2.4 Commute-energy performance (CEP) index 67

2.5 Geographical setting and data analysis 68

2.6 Results 72

2.7 Relation to spatial-morphological characteristics 81

2.8 Conclusions 84

References 85

Chapter 3: Minimum commuting distance as a spatial

characteristic in a non-monocentric urban system 89

Abstract 89

3.1 Introduction 90

3.2 Spatial variations in excess travel 92

3.3 Possible policy implications 94

3.4 Methodology 95

3.5 Case study area: Flanders and Brussels (Belgium) 102

3.6 Application and results of the case study 106

3.7 Possible biases 113

3.8 Conclusions 113

References 115

Contents

6

Chapter 4: Measuring spatial separation processes through

the minimum commute 119

Abstract 119

4.1 Introduction 120

4.2 Defining spatial separation processes 122

4.3 Measuring spatial separation by excess commuting characteristics 124

4.4 Spatial development and commuting in Flanders and Brussels 128

4.5 Data 131

4.6 Method 132

4.7 Results 134

4.8 Conclusions 143

References 145

Chapter 5: Excess travel in non-professional trips:

Why looking for it miles away? 149

Abstract 149


5.2 Excess commuting and excess travel 152

5.3 Methodology 156

5.4 Determination of spatial classes 157

5.5 Developing a proximity map 159

5.6 Reported trip lengths 166

5.7 Excess travel 171

5.8 Possible biases in the results 174

5.9 Conclusions 176

References 178

Data sources 181

Chapter 6: Relationship between spatial proximity and

travel-to-work distance: The effect of the compact city 183

Abstract 183


6.2 The relevant literature 187

6.3 Methods 190

6.4 Results 200

6.5 Discussion 212

6.6 Conclusions and pathways for further research 214

References 215

Contents

7

Chapter 7: Linking expected mobility production

to sustainable residential location planning 221

Abstract 221


7.2 Study area 225

7.3 Methodology and data 225

7.4 Analysis 230

7.5 Forecasting model for Flanders 233

7.6 Discussion 235

7.7 Conclusion and directions for further research 237

References 238

Chapter 8: Conclusions and policy recommendations 243

8.1 General conclusions 243

8.2 Options for spatial planning policy 251

8.3 Some directions for further research 257

References 258

Addendum: The spatial component of air travel behaviour:

An exploration 261

A.1 The aeroplane: the forgotten transport mode 261

A.2 Urban versus rural lifestyle 264

A.3 Rebound effect and policy implications 265

References 266

Samenvatting 269

S.1 Overzicht 269

S.2 Onderzoeksopzet 270

S.3 Bevindingen 272

S.4 Aanbevelingen voor het ruimtelijk beleid 277

S.5 Verder onderzoek 279

Referenties 279

Curriculum vitae 281

Contents

8

9

Preface (English version below)

Na ruim drie jaar werken aan dit proefschrift is het tijd om er de laatste

hand aan te leggen. Hoewel het niet de bedoeling lijkt dat een “woord

vooraf” áchteraf geschreven wordt, is deze omgekeerde volgorde nood-

zakelijk om een terugblik te kunnen werpen. Want wat bracht mij er

eigenlijk toe om me te verdiepen in mobiliteit en ruimtelijke ordening?

Volgens de overlevering was mijn eerste woordje “mama” en mijn

tweede “lamborghini”, verwijzend naar een paars matchbox-autootje dat

deel uitmaakte van mijn eerste verjaardagsuitzet. Een volgende stap in

deze evolutie was dat ik rond mijn dertiende op zaterdag op de fiets

sprong om aan mijn nieuwe hobby te werken, die bestond uit het

verzamelen van prospectussen voor auto’s. Hoewel dat doorgaans vlot

ging bij garages die gespecialiseerd waren in Renault of Volkswagen,

werden er al eens wenkbrauwen gefronst toen ik op zekere dag de moed

had om de toonzaal van Rolls-Royce te betreden.

Deze vroege interesse voor auto’s heeft echter weinig opgeleverd: tot

op de dag van vandaag heb ik mij nooit een wagen aangeschaft. Met het

ouder worden groeide mijn belangstelling voor het leefmilieu en begon

mijn interesse zich te verschuiven naar fietstechniek. Handig, want die

nieuwe sportfiets was niet alleen een perfect vervoermiddel voor de stad,

maar ik kon er ook de weekends en de vakanties mee vullen.

Tijdens mijn opleiding in architectuur en planning maakte ik een

ontwerp voor een fietsvriendelijker heraanleg van het kruispunt “De

Sterre” in Gent. Dat werd uiteraard nooit uitgevoerd, en ironisch genoeg

is dit de plek waar ik de laatste drie jaar weer elke dag tweemaal op

lichtjes suïcidale wijze langs fiets.

Een volgende stap was mijn eerste job, als “consultant

verkeerskunde”, waarbij ik ingeschakeld werd in de opmaak van mobili-

teitsplannen, en waarin ik de aanleiding vond om mijn studie in de

ruimtelijke planning aan te vullen met nog een opleiding verkeerskunde.

Jaren later, in het UN-Habitat-team in Kosovo, zou deze achtergrond

weer zeer goed van pas komen bij het ontwikkelen van ruimtelijke en

mobiliteitsplannen in een heel andere context.

En uiteindelijk bleken mobiliteit en planning ook een domein te

vormen waarin nog heel wat theoretische onderzoeksmogelijkheden braak

Preface

10

lagen. Mijn beslissing om daarin te stappen heb ik me niet beklaagd,

getuige daarvan dit proefschrift.

Dan rest mij nog de eer en het genoegen om een reeks mensen te

bedanken, zonder wie dit werk er vandaag niet zou liggen. In de eerste

plaats komt natuurlijk professor Frank Witlox, mijn promotor, die mij

eind 2007 een goede reden gaf om Kosovo weer voor België in te ruilen.

Hij zorgde voor het onderzoekskader en gaf mij de tijd en ruimte om

cursussen te volgen en aan congressen deel te nemen. Van hem kreeg ik

het vandaag in academische kringen zo gewaardeerde peer review- en

publicatievirus te pakken. En met resultaat: de zes basishoofdstukken van

het voorliggende proefschrift hebben een “peer review”-proces doorstaan

en zijn (of worden) gepubliceerd in een reeks hoogstaande weten-

schappelijke tijdschriften. Een snel optelsommetje leert dat er in de loop

van het schrijven van deze verhandeling maar liefst achtentwintig

internationale reviewers hebben bijgedragen tot de kwaliteit van dit werk.

Hoewel deze experts doorgaans anoniem optraden, heb ik hun inspanning

zeer gewaardeerd.

Maar dichterbij huis, op de werkvloer van de sociaal-economische

geografie (SEG), zijn er wel meer mensen die een plaats verdienen in een

dankwoord. Ik denk vooral aan Tijs Neutens, Veronique Van Acker,

Thomas Vanoutrive, Enid Zwerts en Nathalie Van Nuffel die op tijd en

stond methodologische ondersteuning boden, in het bijzonder in het eerste

jaar, toen de wondere wereld van statistiek, dataverwerking en GIS zich

nog aan mij aan het openbaren was. Daarnaast heb ik ook de inbreng van

andere SEG-onderzoekers van het eerste uur enorm gewaardeerd: David

Bassens (die nog steeds op zoek is naar een manier om de kloof tussen het

wereldstedenonderzoek en het mobiliteitsonderzoek te dichten), Heidi

Hanssens (die altijd klaar stond om de sociale omkadering te verzorgen),

Sven Vlassenroot (die de koffiepauzes steevast met nuttige weetjes over

de academische wereld kwam opvrolijken), Lomme Devriendt (die de

SEG vorm gaf) en professor Ben Derudder (die zijn nuttige tips van de

zijlijn gaf). Daarmee wil ik de rest van de Vakgroep Geografie uiteraard

niet vergeten; één speciale vermelding nog voor Helga Vermeulen, zonder

wiens organisatorische en administratieve nauwgezetheid ik wellicht niet

eens aan de universiteit was kunnen beginnen. Maar ook buiten de

Vakgroep Geografie heb ik heel wat ondersteuning gekregen. De tientallen

mensen (onderzoekers, promotoren en coördinatoren binnen de

verschillende universiteiten, en begeleiders van de Afdeling Ruimtelijke

Planning) achter het Steunpunt Ruimte en Wonen (dat mijn onderzoek

Preface

11

financierde) verdienen hier dan ook een extra vermelding, net zoals Jos

Zuallaert en Dirk Lauwers die mij de mogelijkheid boden om van tijd tot

tijd nog wat in de planningspraktijk te gaan werken in Kosovo.

Tot slot nog een bedankje voor zij die mijn activiteiten van buiten de

werkvloer volgden, maar er niet voor terugschrokken om over mijn

onderzoek in discussie te gaan: eerst en vooral mijn ouders, Johan en

Kristien, maar ook mijn drie zussen (Anna, Marieke en Mathilde), oma

Simonne, het wekelijkse badmintongezelschap (Filip, Wouter, Bram en

Piet) waar de evolutie van de olieprijs een steeds weer opduikend thema

was, Rocky (die de lichtjes controversiële kaft van dit boek leverde),

Hermes (die bijna wekelijks naar de stand van zaken informeerde) en dan

nog een hele reeks van vrienden, familie, reisgenoten, huisgenoten en

facebook-friends die in de loop van de laatste jaren hun interesse lieten

blijken. Merci, allemaal!

___________________________________________

After more than three years working on this thesis, the time has come to

add the finishing touch. Although the word “preface” does not sound as if

it is intended to be written afterwards, this reverse order is necessary to

allow looking back. Because why did I bury myself in the study of

mobility and spatial planning?

According to tradition, my first word was “mama” and my second

“lamborghini”, referring to a purple matchbox-car that was one of my

first-birthday presents. As a next step in this evolution, when I was

thirteen, on Saturdays I jumped on my bike to work on my new hobby,

which consisted of collecting commercial prospectuses for cars. Although

this went generally smoothly in garages that specialized in Renault or

Volkswagen, some eyebrows raised when on a certain day I had the

courage to enter the showroom of Rolls-Royce.

However, this early interest in vehicles did not yield that much: up to

now, I never bought a car. With age, my interest in the environment

increased and I started concentrating on bicycle technology. That was

quite convenient, because my new touring bike was not only a perfect

transport means in the city, but I could also use it to fill my weekends

and holidays.

During my training in architecture and planning, I devised a bicycle-

friendly redesign of the intersection “De Sterre” in Ghent. Of course, my

design was never implemented, and ironically, during the last three years,

Preface

12

this is the place where I cycled through every day twice in a slightly

suicidal manner.

A further step was my first job, as a “traffic and mobility

consultant”, where I was called in to develop municipal mobility plans. In

this job, I also found a good reason to supplement my studies in spatial

planning with an advanced course on traffic and mobility. Some years

later, in the UN-Habitat team in Kosovo, this background proved again

very useful in developing urban plans and mobility plans in a quite

different context.

Ultimately, mobility and planning turned out to be a domain where

many theoretical research opportunities were still present. And I can say

that I have no regret on my decision to get into this, as is demonstrated

today by this dissertation.

Of course, I would like to use this opportunity to thank a number of

people, without whom this work would not be accomplished today. In the

first place there is of course Prof. Frank Witlox, my supervisor, who gave

me at the end of 2007 a good reason to exchange Kosovo for Belgium

again. Frank provided the research framework and gave me the time and

space to follow courses and participate in conferences. He transferred the

“peer review and publication virus” to me, which is highly valued in

academic circles. This was not without success: all of the six basic

chapters of the present dissertation have gone through a peer review

process and are (or will be) published in a series of high ranking academic

journals. A quick summation shows that twenty-eight international

reviewers have contributed to the quality of this work. Although these

experts usually performed anonymously, I greatly appreciate their efforts.

Closer to home, at the Social and Economic Geography (SEG)

research cluster, there are more people who deserve a place in this

expression of gratitude. Tijs Neutens, Veronique Van Acker, Thomas

Vanoutrive, Enid Zwerts and Nathalie Van Nuffel have all offered

methodological support to me, especially in the first year, when the

wonderful world of statistics, data processing and GIS was still in the

process of revealing itself to me. In addition, I also have greatly

appreciated the input from other SEG-researchers who were there in the

very early stage of my study: David Bassens (still looking for a way to

bridge the gap between world city research and mobility research), Heidi

Hanssens (who was always ready to care about the social environment),

Sven Vlassenroot (who was invariably cheering up coffee breaks with

interesting academic rumours), Lomme Devriendt (who was responsible

Preface

13

for the SEG house style) and Prof. Ben Derudder (who contributed with

useful tips from the sideline). This is of course not to forget the rest of

the Geography Department, while I would like to add a special note to

Helga Vermeulen, without whose organizational and administrative

accuracy I may not even have started working at Ghent University. But

even outside the Geography Department, I got a lot of support. The

dozens of people (researchers, supervisors and coordinators within the

different universities, and the coaches of the Ministry’s Spatial Planning

Department) working for the Policy Research Centre on Regional

Planning and Housing (which funded my research) deserve a special

mention here too, just as Jos Zuallaert and Dirk Lauwers who offered me

the possibility to do still some practical planning work in Kosovo.

Finally, a word of thanks to all those who observed my professional

activities from outside, but were not afraid of getting involved in some

debate on my research topic. First and foremost I should mention my

parents, Johan and Kristien, but also my three sisters (Anna, Marieke

and Mathilde), my grandmother Simonne, the weekly badminton

company (Filip, Wouter, Bram and Piet) (where the oil price trend was

an ever resurfacing theme), Rocky (who made the slightly controversial

cover of this book), Hermes (who informed almost weekly on the state of

affairs), and of course the whole series of friends, family, travel compan-

ions, housemates and facebook-friends who demonstrated their interest

during the last couple of years. Merci, everyone!

Preface

14

15

Chapter 1:

Introduction

1.1 Summary

An often-heard statement says that the interaction between mobility

policy and spatial planning practice needs more coordination. Any

concerned politician or citizen feels that the perceived increase in car

traffic, and the growth of problems that are associated with car use, have

“something” to do with unorganized urban expansion, sprawling new

housing and industrial allotments, and ribbon development. Nevertheless,

it is less clear how this relationship exactly looks like, leaving alone the

question how planning should be used as a tool to improve accessibility

and steer mobility to a more sustainable course.

This dissertation wants to gain insight in the reciprocal relationship

between mobility and spatial development, taking into account the

societal context of climate targets and imminent peak oil. This will be

done through the development of a number of quantitative research

methods, which are embedded in a literature review and will be applied

to the case study of Flanders (Belgium).1

This broadly defined research objective is narrowed to the mobility of

people, and will focus on exploring the sustainability of spatial structure

with respect to travel behaviour, with particular attention to the daily

distances travelled. Sustainability is defined in terms of resilience, not

only for growing mobility but also for a possible declining future mobility.

A generally declining mobility is a scenario that may develop due to

rising energy costs (e.g. peak oil scenario) or stringent climate policies,

while a selective shrinkage of mobility (only affecting parts of the popula-

tion) may occur through increased saturation of the traffic system.

Moreover, spatial structure plays a role in the potential steering of travel

behaviour in a more sustainable direction.

1 The reason for this research is found in the ‘mission statement’ of the funding

Policy Research Centre on Regional Planning and Housing - Flanders (2007-

2011), a policy-oriented research consortium that is administered by the De-

partment of Planning, Housing and Heritage of the Flemish government.

Chapter 1

16

We refer to the research line on the connection between spatial struc-

ture and travel behaviour, which has developed mainly in the English-

speaking, German, Dutch and Scandinavian world, and whose prelimi-

nary conclusions can be formulated as follows: “A sustainable travel

pattern can only be realized within an appropriate spatial framework, but

other measures (financial and regulating) are needed to effectively change

travel behaviour. In other words, planning is necessary but not sufficient”

(after Zhang, 2002, p. 3). The spatial quality that facilitates a travel

pattern based on short distances is called “spatial proximity”, even

though this concept has not been clearly defined in the exploratory phase

of this research.

This introductory chapter is structured as follows. Section 2 provides

an overview of climate change and peak oil, two global phenomena that

are directly related to the external effects of mobility. Section 3 deals

with some captivating aspects of time and space perception and increased

prosperity that underpin the growth of mobility. Section 4 gives an

overview of the possible role of spatial structure in a future-oriented

approach to mobility. Section 5 presents a snapshot of the spatial policy

context in Flanders and Brussels. Section 6 focuses on the research

questions, establishes a conceptual framework and states how the research

will be conducted. Section 7 gives an outline of Chapters 2 to 7, each of

which studies a separate aspect of the problem and can therefore be read

as an individual article.

1.2 Sustainable mobility, climate change

and peak oil

The objective of coming to a less car-dependent and, by extension, a less

oil-dependent transport system can be argued from different perspectives.

Local environmental and safety problems caused by transport have

already been in the spotlight for several decades. The principal issues in

this debate are air pollution, noise, deterioration of the livability of

residential areas, accidents and ecological and landscape fragmentation.

This environmental approach is part of what is called “sustainable

mobility” in the recent transport literature. Banister (2008) identifies four

components that may contribute to the transition to a more sustainable

transport system: (1) reducing the need to travel through substitution,

(2) achieving a modal shift through transport policy measures, (3)

Introduction

17

distance reduction through land-use policy measures, and (4) efficiency

increase through technological innovation.

Especially in the field of air pollution, accidents and livability, we can

say that in the western world a lot of progress has been made by a

combination of the mentioned policies and technological developments.

We do not elaborate on this: for an overview we refer to Gilbert and Perl

(2008, pp. 189-264).

However, in terms of climate change (an environmental problem) and

peak oil vulnerability (an economic problem), there seems to be much less

progress. Below, we examine these two global phenomena, considering

that these could constitute a major incentive to reduce the oil dependence

of the transport system.

1.2.1 Climate change

It seems that around 1995, in the scientific community, a consensus was

reached on the acknowledgement of climate change as an important

human-caused problem. With the Kyoto Protocol (1997), which was

ratified by almost all concerned countries except the US, it became clear

that these countries recognized anthropogenic climate change as a

problem and, at least in a rhetorical sense, wanted to commit themselves

to cut greenhouse gas emissions. In the Kyoto Protocol, Belgium, for

instance, is committed to limit its emissions levels by 2012 to 92% of the

1990 level. Since 2001, we see that climate change both in the peer-

reviewed literature and outside it has become an established phenomenon

(Weart, 2010).

Climate change is caused by greenhouse gases, of which carbon diox-

ide (CO2) is the most important. In 2006, transport was worldwide

responsible for 23% of the energy-related greenhouse gas emissions. One

fifth of the projected increase in emissions comes on account of transport,

mainly in the form of increasing ownership and use of passenger cars in

non-OECD countries and a general increase in international air travel

and overseas cargo shipping (IEA, 2008).

In 2005, in the European Union (EU) 80% of all greenhouse emissions

was energy-related, including 24% (equalling 19% of total emissions) that

was transport related. The international bunkers (fuel used by interna-

tional aviation and shipping, both intra-European and intercontinental)

are not yet included in these figures. In the period 1990-2005 greenhouse

Chapter 1

18

gas emissions in the EU decreased in all sectors, except in transport

(+26%) and international bunkers (+64%) (EEA, 2008).

In Flanders, greenhouse gas intensity in transport (this is the amount

of emissions per person-kilometre, or per ton-kilometre for freight)

decreased slightly in the period 1999-2005, but the absolute growth of

traffic offset this efficiency gain well and ensured an absolute increase in

emissions (by 1.50 megatons of CO2 equivalent per year, an increase of

12% over the period 1990-2007) that was the greatest in the transport

sector (again, international bunkers are not yet included). This sector was

therefore largely responsible for not achieving the intermediate emission

reduction targets of the Flanders Region for 2005, even though in 2007

the Kyoto target was still met. This last evolution was due to the manu-

facturing industry which recorded over the last period a reduction of 5.33

megatons of CO2 equivalent per year (VMM, 2008, pp. 95-96).

Regarding the transport sector, the Flanders Climate Policy Plan

(LNE, 2006) focuses on achieving a modal shift, increasing overall

efficiency, and improving the economy of the fleet. However, no measures

that would curb the autonomous growth of traffic are proposed and

international aviation and shipping are even completely out of focus.

Although in Flanders the growth rate of personal car travel has almost

reached zero in recent years (SVR, 2010), the complete traffic volume

seems to be mainly associated with economic dynamics, rather than with

climate policy. Moreover, both freight (SVR, 2010) and international air

traffic with origins and destinations in and around Flanders was growing

steadily up to 2008 (Brussels Airport, 2009).

Transport is clearly bottom of the class when it comes to greenhouse

gas emissions. Although efficiency improvement in vehicles is indeed

enforceable in terms of regulation, it seems that traffic growth itself is

directly linked with increasing prosperity. When mobility growth stag-

nates, this is usually due to a (temporary) economic recession. Also

reaching a high level of congestion can lead to stagnation within a

particular segment of the transport sector, e.g. in road traffic. Structural

congestion may suppress autonomous growth, but particularly increases

the pressure to guide the growth in another direction, for example by

building more road infrastructure, through measures that improve the

spread of traffic across the day, by increasing the capacity and the

attractiveness of public transport, or by shifting the growth from com-

muter traffic towards the segment of recreational and tourist travel

(including airplane use).

Introduction

19

Overall, proposed policy measures aimed to reduce greenhouse emis-

sions from traffic can be divided into four packages:

1. Steering mobility to a less car-dependent course by encouraging a

modal shift towards alternative transport modes (other than passen-

ger car or truck), a more efficient utilization of vehicles, but also

substitution of transport by telecommunication. Known examples in

passenger transport are: car sharing, carpooling, the provision of pub-

lic transport, encouraging cycling and walking, and teleworking and

videoconferencing (Robèrt and Jonsson, 2006).

2. Increasing, and possibly varying, financial charges on those types of

mobility which emit most greenhouse gases, while alternatives may be

supported in parallel. Examples in passenger travel are raising fuel

taxes, introducing registration charges corresponding to the emission

level of the vehicle, but also road tolls, “smart” charging, free public

transport and bicycle allowances (Chapman, 2007).

3. Increasing vehicle efficiency by accelerating fleet replacement by more

fuel efficient vehicles or vehicles that run on alternative fuels (includ-

ing switching to renewable organic fuels, known as biofuels) (Anable

and Bristow, 2007).

4. Intervening in the spatial structure through land use planning with

the aim of bringing potential destinations closer together and making

long distance transport needless. With regard to person mobility it is

generally assumed that increasing density and developing a high de-

gree of land use mix leads to less use of private cars and shorter daily

distances travelled (Newman and Kenworthy, 2006). The latter form

of proposed policy is the background of the research which is reported

in this dissertation.

The objectives of these four packages are clear, and throughout the

western world many examples can be found where this kind of measures

are implemented and measurable successes were reported. However, the

finding that overall traffic and transport emissions continue to grow

(Cervero and Murakami, 2010), leaves room for some healthy scepticism.

Below, we give a number of concerns, again grouped by package.

1. In terms of modal shift many local success stories are known. An

example is the historical centre of Bruges (Belgium), where in the

period 1999-2004 car traffic in the city core declined under the influ-

ence of a rigorous mobility policy, while the number of cyclists and

users of public transport increased considerably in the same period.

The number of registered cars in the entire municipality (which con-

Chapter 1

20

tains six times more inhabitants than the old town), however, con-

tinued to increase. Thus, the success story remains very local, and led

to a geographical shift in the growth of car traffic (City of Bruges,

2007). It is difficult, if not impossible, to find a clear example where a

purposive mobility policy on a regional scale has led to a significant

substitution of individual car use. Nevertheless, the finding that the

modal split largely varies throughout the western world, depending

on the city, the country and the general context, is promising for po-

tential sustainability gains in mobility.

2. The main success story in terms of taxation is perhaps the London

Congestion Charge, which has led two years after the introduction to

a reduction in CO2 emissions in the charged area by 19.5% (Beevers

and Carslaw, 2005). Again, it may be assumed that a part of the

suppressed traffic finds its way outside the demarcated charged area.

Moreover, public support for additional charges on car use (also out-

side the city centres) is hardly present anywhere in the western

world.

3. Anable and Bristow (2007) show that UK greenhouse gas emissions

from cars remained almost constant in the period 1990-2005, while

both the kilometrage and the average weight per vehicle increased

significantly. So we are talking about a major rebound effect, where

efficiency gains are associated with an increase in activity, and finally

with a status quo of energy consumption. Indications exist that the

rebound effect occurs at the macroeconomic level too, meaning that

efficiency gains may lead to accelerated economic growth with an

overall increase (instead of a decrease) in energy consumption

(whether or not outside the transport sector) as a consequence

(Saunders, 1992). Also, the effectiveness of the use of so-called biofu-

els has in recent years led to a major controversy (Righelato and

Spracklen, 2007).

4. Although many studies have found that spatial structures with an

urban character are associated with a lower per capita energy con-

sumption for transport (Newman and Kenworthy, 1999), the

correlation between density or spatial diversity and sustainability of

travel patterns is in fact quite weak. One possible reason is a geo-

graphical form of the rebound effect. An increase in the choice range,

manifested as a better internal accessibility which is typical of urban

structures, will partly offset the relatively small mutual distances,

making people less inclined to choose the nearest possible destination

Introduction

21

(Handy et al., 2005). A second reason is that a high density city is

often connected with a serviced area that is spread out very widely,

feeding people who travel long distances to the city (Mees, 2010, pp.

24-26). A third possible reason is the occurrence of the classic re-

bound effect: cost savings through more efficient daily travel patterns

can lead to more or longer recreational trips (Holden and Norland,

2005).

These outlined reservations contain a series of possible explanations for

the failure to meet emission reduction targets for transport. Decoupling

transport-related emissions and prosperity is not a technical but a social

problem, for which an univocal response has not been found yet. More-

over, in the preceding argument we confined ourselves to the western

world. The largest relative growth of transport-related emissions, how-

ever, is realized in developing countries.

1.2.2 Peak oil

Although it was in the 1950s introduced by M. K. Hubbert, peak oil has

only recently become a common term, describing the evolution of the

production of an oil well over time, from discovery to exhaustion.

Hubbert (1956) pictured this evolution as a bell-shaped curve, a theory

according to which the peak production of oil in the US was predicted to

occur between 1966 and 1972. This forecast became true in 1970 (Gilbert

and Perl, 2008, p. 126), after which the share of imported oil into the US

increased dramatically.

Today, the term peak oil is used to refer to phenomena caused by

global scarcity of fossil oil. Although the report of the Club of Rome in

1972 predicted for 1992 a depletion of all oil reserves that were known at

that time (Meadows et al., 1972), in practice peak oil is not to be seen as

depletion, but as a mismatch between supply and demand (Campbell and

Laherrère, 1998). When demand for oil continues to rise by a steadily

growing global economy, but production does not follow this trend, then,

in principle, the result is a drastic increase of the relatively inelastic oil

price.

Although the phenomenon of peak oil is much less known in common

than climate change (Bardi, 2009), it may be much more restrictive and

may therefore have more far-reaching consequences, both for the economy

in general and for the transport sector in particular. The reason is that

Chapter 1

22

peak oil is about physical limitation, which can virtually not be controlled

by any policy.

A survey by ASPO (2010) shows that most scenarios locate the peak

in world oil production somewhere between 2006 and 2012, although

information is highly dependent on the source. IEA (2008) estimates the

world oil production in 2030 to 101.5 million barrels per day (some 20%

more than the production level of 2008) while an estimate by Aleklett et

al. (2010) gives 75.8 million barrels per day (which is 10% less compared

to the production level of 2008). The debate seems characterized by an

optimistic view of the IEA versus the pessimistic view that is dominant in

the peer-reviewed literature.

But it is not just the point in time when peak oil will occur that is

uncertain: the potential consequences are not clear too. Before 2008 it

was generally assumed that reaching the production peak would be

followed by strong price increases. Indeed, during 2007 oil prices rocketed

up, to reach a top of about 147 USD per barrel in June 2008 (wich is

about 125 USD expressed in June 2010 US dollars). Then the price

dropped again very quickly to stabilize at a price of around 75 USD in

the course of 2010 (Fig. 1.1). Although the rate of the price evolution

may be due to speculation, the intrinsic explanation for the increase may

be found in a mismatch between supply and demand. But this does not

yet explain the last price drop: contrarily, peak oil theory predicts a

continuous price increase.

Hamilton (2008) presents a part of the solution, by showing that nine

out of ten recessions since World War II in the US immediately followed

a sudden surge in the oil price. This finding suggests that a surge in oil

prices is followed by recession, shrinking the demand for oil until a price

level is reached that can be worn by the current economic system. This

means that the oil price remains high, when expressed in terms of

prosperity but not necessarily in monetary terms. Kopits (2009) argues

that the oil price is linked to the share of oil consumption in the GDP of

the US. According to this thesis the oil consumption in the US is limited

to 4% of the GDP. If the consumption level exceeds this threshold, a

recession should be expected. In 2009 this would have meant that the US

economy was not ready to absorb oil prices of more than 80 USD per

barrel. Consequently, to avoid future recessions, oil dependence should

diminish. Kopits (2009) therefore proposes to increase the price artificially

by introducing a carbon tax in those periods where the market oil price is

Introduction

23

below 80 USD per barrel, in order to both ensure price stability and

encourage energy saving developments.

Within the euro area, based on a simulation by the macro-

econometric model of the Eurosystem, the economy of Belgium appears to

be the most sensitive to fluctuations in oil prices: a price increase of 10%

for Belgium would after three years result in a cumulative decrease of

0.4% in the GDP (ECB, 2010).

0

20

40

60

80

100

120

1946 1956 1966 1976 1986 1996 2006

US

D/b

arr

el

nominal

adjusted for inflation

Fig. 1.1. Evolution of the oil price, based on annual averages.

Source: Inflationdata (2010)

But what does this story mean for the transport system? According to

IEA (2008), in 2006, 95% of all energy used worldwide for transport

comes from petroleum. IEA (2008) believes this share to drop to 93% in

2015 and to 92% in 2030, provided that biofuel substitution is widely

implemented. This means that the transport sector remains extremely

dependent on the availability, and thus the price, of oil.

Rodrigue et al. (2009) present the following possible effects of high oil

prices on the transport system:

• reduction of both the speed and the total amount of mobility

• shifts to alternative modes that rely less on oil

• changes in the organization of transport and distribution networks in

favour of more fuel-efficient vehicles and shorter total distances

• in the long-term: changes in location choices as a function of facilitat-

ing the mentioned adjustments

Chapter 1

24

Although there is little evidence for long-term effects of a high fuel price

on the transport system, a number of indications can be deduced from

the responses to the surge in oil price in the second half of 2007 and the

first half of 2008.

In Europe, the fuel for cars and trucks is relatively heavily taxed. In

Belgium 50% to 60% of the customer price are taxes. In most US states

taxes make up only around 17% of the final fuel price. These differences

in taxation have developed historically, and are linked with oil depend-

ence. Oil price increases on the international market are quickly perceived

by consumers in countries where the tax level is low. Therefore, especially

in the US, the oil price surge in the period 2007-2008 had appreciable

impact on the traffic. INRIX (2008) found in the first half of 2008, when

the retail fuel price in the US increased by 28% on average, a significant

reduction of congestion. The correlation between the increase in price and

the decrease in automobile traffic was strongest in cities with a lot of

recreational traffic (tourist destinations such as Las Vegas and Miami)

and in cities where the car is dominant but public transport offers a

decent alternative (e.g., Atlanta, Los Angeles). In cities where public

transport was already heavily used (e.g. New York, Washington DC), the

effect was less pronounced. Apparently the strongest impact was reported

in the segment of tourism and recreational traffic, and in places where

room for behavioural change is present.

Hamilton (2009) points out that those houses that were most isolated

were hit hardest by the real estate credit crunch in 2007-2008. Easily

accessible homes that are located close to a wide range of potential

destinations remained relatively attractive.

The French study of Gonzalez-Feliu et al. (2010) shows that a distri-

bution system based on large-scale retail chains (hypermarkets) is much

more dependent on car use and fossil fuels, in comparison with a network

of local shops or pickup points. However, the latter category requires

more labour force. This means that small-scale distribution would gain

importance at the expense of hypermarkets, in case the relative cost of oil

would rise faster than the relative cost of labour.

Influenced by the recent volatility of the oil price and peak oil theory,

apparently some literature on the theme of urban resilience has emerged.

The research question is which factors determine the vulnerability of an

urban economy in the context of unstable oil prices and how the damage

could be minimized. Transport obviously plays a major role in this theme

Introduction

25

(Dodson and Sipe, 2008; Gleeson, 2008), to which this dissertation wants

to contribute.

1.3 The time-distance-space relationship

Based on the above-mentioned findings, it seems worth to take a closer

look at the mechanisms behind the continuous increase in mobility and

the associated oil dependence. The annual distance covered by an average

individual is increasing every year, and this is done with ever-faster

means of transport. Nevertheless, most of the possible destinations where

everyone is hurrying to remain relatively static: most town centres have

been in the same place for decades, if not for centuries, and although

cities are sprawling, they seem much less expansive than the mobility

itself.

1.3.1 Time distance and travel time budget

In general, mobility is considered as an achievement of modern man. In

the western world the ability to quickly move in any desired direction has

become a prerequisite for leading a full life. This principle has become

popular on every conceivable scale level. In Belgium, for example, one

who travels abroad for holiday less than once a year, is regarded as

underprivileged. The physical location of a job has become a minor

decision factor when choosing a home. Other activities such as education,

recreation and shopping have become virtually footloose in the course of

the twentieth century. Nevertheless, note that the importance of regional

and local identity is not disappearing, and is in some cases perhaps even

more appreciated than before. Increased mobility has made it possible to

operate as a cosmopolitan from your own village, with the rural family

history still fresh in mind. Specifically in Belgium decades of focused

mobility policies have led to an economic shift towards manufacturing

and service industry without causing a spatial division between vivid

metropolitan and empty rural areas. Thus, daily travel over rather large

distances has, to a certain extent deliberately, become an essential part of

the Belgian society.

Parallel to the increase of the footlooseness of activities, the meaning

of the term “distance” blurred. It seems that the increase of mobility has

resulted in a replacement of the notion of distance by the measure of

time. Already in 1791 the meter (m) has been defined very accurately as

Chapter 1

26

the ten-millionth of the distance between the North Pole and the equator.

But the need for setting a standard time was only felt with the arrival of

the train. In the Belgian cities, people only switched late 19th century to

the railway time, which was the same across the country (Reynebeau,

2003, p. 55). In municipalities that were not connected to a railway line,

and were thus deprived of rapid transport, it took some more decades

before distances were expressed precisely in units of time.

This anecdote illustrates the strange evolution in which the percep-

tion of a spatial measure - distance - was transformed into a temporal

concept: time. From a physics point of view, this obviously makes no

sense. If we express a distance in the SI unit2 s (second, or a secondary

unit such as minute or hour) (as in: “I live half an hour from my work.”),

and we stick to the definition of speed as the quotient of distance and

time, the inherent meaning of speed changes. Indeed, the unit of meas-

urement for speed is no longer m/s, but: s/s, which means that the unit is

just omitted. This is less enigmatic than it seems. The advent of rapid

transport and the declining importance of physical distance has resulted

in a perception of speed no longer behaving as a variable but as a con-

stant. This explains immediately why unexpected traffic jams and train

delays are so annoying: the supposedly constant average speed is sud-

denly a variable again, so that the stability of the whole system crashes.

The unexpected traffic jam makes a physical impossibility happen,

namely the sudden stretching of the distance between two fixed points.

To handle the perception of time as a function of distance, in trans-

port studies the concept of “time distance” was introduced (Mérenne-

Schoumaker et al., 1999, p. 90). This is actually the combination of the

variable distance and the unit s. In this context, the assertion above in

which the location of a job plays only a limited role in the search for a

home, should be reviewed. It is not the physical distance, but especially

the mutual time distance that determines the geographic location of the

various activities in which an individual or a household participates.

Common sense says that one wants to minimize the time distance

between origins and destinations, assuming that travel is a derived

demand (Mokhtarian and Salomon, 2001). This reasoning has been

largely prompted by the fact that people, within certain financial limits,

always choose the fastest available transport mode. Workers with a

decent income only take public transport to commute if this is faster than

2 Système International d’Unités (International System of Units)

Introduction

27

the car. The market share of international rail traffic has dropped steeply

since the advent of low cost airlines. Rail projects that have gained

market share are invariably those whose speed is competitive with the

airplane (Rodrigue et al., 2009). And biking is especially popular in dense

urban centres, where it takes hours to manoeuvre a car through or to find

some parking space (Verhetsel et al., 2007, p. 5).

However, the above, intuitive, argument is only a fallacy by which

many urban planners and policy makers have been caught in the past.

The Greek urban planner Doxiadis wrote in 1976 that the daily travel

time taken by an individual had evolved from several hours in prehistoric

times to a constant of half an hour in the first urban civilizations. In the

nineteenth century, however, this optimum constant would have been

abandoned because of the rapid expansion of cities and the inadequate

transport systems. But Doxiadis’ argument was not supported by any

data (Hupkes, 1977, p. 257). Time budget surveys, primarily those of

Szalai et al. (1972), shed new light on the matter. At the aggregate level

(e.g. all inhabitants of a region) the average amount of time spent on

travel appeared to be regarded as a constant, irrespective of the geo-

graphical location of the studied region. Since both the US, Western

Europe and Eastern Europe were involved in the investigation, we may

conclude that this time budget is independent of the economic develop-

ment stage or income. Szalai et al. (1972) found an average personal

travel time budget of 1h13min per day, or 444h per year.

In his thesis Hupkes (1977) formulates the so-called BREVER-law,

based on the time budget survey of Szalai et al. (1972): the law of

conservation of travel time and trips (in Dutch: Behoud van REistijd en

VERplaatsingen). The conclusions of Szalai et al. (1972) and Hupkes

(1977) were confirmed by those of Zahavi et al. (1980) and especially by

Schafer (2000), who also involved Asian and African rural regions in his

research and demonstrated in this way the general validity of the

BREVER law. However, more disaggregated research, like that of Joly

(2004), shows that the travel time budget strongly varies according to the

social group to which one belongs. Van Wee et al. (2002, p. 5) believe

that the travel time budget is less constant than claimed by Hupkes: in

the Netherlands over the years 1980-1990 they observed a slight annual

increase. In Flanders, Glorieux et al. (2005, p. 7) found 1h00min spent on

travel on an average day in 1999, and 1h04min in 2004. If only the

respondents who actually moved on the survey day are considered, travel

times rise to 1h24min and 1h27 min respectively. These results are in line

Chapter 1

28

with the previously mentioned studies, and are also not in contradiction

with the statement of Van Wee et al. (2002, p. 5).

1.3.2 Travel speed

The tendency to choose the fastest manner to make a trip has not led to

a reduction of the time spent on travel, but to an increase in kilometrage.

Within the constant travel time budget (and the financial travel budget),

speed is usually maximized. So, in the long run, our earlier argument that

the speed factor should be considered as a constant, does not hold. The

average travel speed is increasing over the years, perhaps since the

invention of the wheel (Ma and Kang, 2011). In the economic centres of

the western world congestion and traffic regulations have somewhat

constrained the increase of the average speed today, but air travel and

fast public transport systems are still growing unabated. Moreover, it

should not be forgotten that motorists who cause congestion just take the

car because it is still faster than the alternatives. On average, the indi-

vidual speed is thus higher in a congested situation than in a similar

travel pattern system without congestion.

From different perspectives explanations can be found for the human

inclination to travel ever faster. First, there is a simple economic explana-

tion. Maintaining a higher speed automatically means a wider radius of

action, or more precisely, a larger space-time prism (Hägerstrand, 1970).

The wider the radius of action, the more likely that within the associated

space-time prism a suitable home, one or two nice and well-paid jobs, a

decent school, a number of relatives and friends and the desired recrea-

tional and shopping facilities are located. High speed allows for spatial

optimization of the utility of one’s travel pattern within the constant

travel time budget. But this economic approach may not explain every-

thing. The desire for speed is as much a psychological phenomenon that is

related with sensation and status. The doctrine of “dromology” according

to Virilio (1977) is one of the possible approaches. Marinetti (1909, p. 1)

uses the beauty of speed as one of the statements of his futurist mani-

festo: “Nous déclarons que la splendeur du monde s’est enrichie d’une

beauté nouvelle: la beauté de la vitesse. Une automobile de course avec

son coffre orné de gros tuyaux tels des serpents à l’haleine explosive... une

automobile rugissante, qui a l’air de courir sur de la mitraille, est plus

belle que la Victoire de Samothrace.”

Introduction

29

The foregoing considerations cast a different light on the utility of

travel, compared to what is common in the public debate. From time to

time, in the popular press the cost of congestion is calculated. The

number of hours spent by motorists in traffic jams is considered as idle

time and is multiplied by the value of the activities that one could have

performed otherwise (Blauwens et al., 2002, p. 360). When we take into

account the fact that the travel budget is constant within certain limits,

then we must conclude that the time one could have saved by escaping

from the traffic jam would have been spent on travel anyway. In particu-

lar for structural congestion, the BREVER law indicates that the lost

time is already reckoned in by the traveller beforehand, and is thus part

of the maximized utility of the trip. The capacity of the road network

should be seen as a limiting factor, which is in nature not very different

from other limiting factors such as the fuel price, the purchase cost of the

vehicle or the socially desirable level of road safety (which determines the

speed limits). What distinguishes congestion from these other constraints

is primarily the idea that the government has a hold over the capacity of

the roads, and in addition, the relatively large and very annoying unpre-

dictability of the phenomenon.

The discussion on the value of time and the relationship between time

and space has often led to philosophical reflections. Hupkes (1977) writes:

“The desire to raise the marginal utility of time spent on things other

than paid work, has in the western world first led to the virtual disap-

pearance of doing nothing, and second, in terms of personal labour, to

disposable consumer items, ready-to-eat food, and motorized and prefera-

bly automatical housekeeping equipment that requires little maintenance.

The motorization of the housekeeping, however, has ... not lead to

significant time savings.” Rising affluence causes a pursuit for maximizing

the utility of one’s available (leisure) time by using it increasingly more

intensive. This story is as true for travel and explains the urge to keep

accelerating.

Illich (1974, p. 42) understands the increasing energy-intensity of

transport as a mechanism that feeds social inequality: “Beyond a critical

speed, no one can save time without forcing another to lose it.”

Coolsaet (1990) argues that influencing speed is the solution for many

negative aspects of the car. Since there is already a social basis for the

imposition of speed limits by the government, he suggests to adjust the

standards downwards. This would not only yield serious benefits for road

Chapter 1

30

safety, emission levels and energy consumption, but would also reduce

congestion.

Marte (2003) elaborates on this, and states that based on the

BREVER law a reduction of the (maximum allowed or effective) speed

would in the long run lead to a reduction in the amount of traffic. Indeed,

the pressure to search for a job closer to home, or to move house near the

workplace, or to choose schools, shops and leisure activities less far from

home, would increase. Moreover, the travel time ratio between private

and public transport would be adjusted in favour of the latter mode so

that a modal shift towards public transport is to be expected. Not only

the consumption of fuel would go down, probably also the congestion

problem would be influenced significantly in a positive sense.

1.3.3 Travel cost

The average travel time budget is not the only factor that is relatively

constant. A less well studied, but similar phenomenon, exists with respect

to the financial cost of travel. For Belgian households, the proportion of

household income that was spent on fuel remained constant at around 3%

in the period 1995-2004. For transport spending in general a slightly

upward trend is noticeable, but this is almost entirely due to the pur-

chase of vehicles (fixed costs) (Fig. 1.2) (Statistics Belgium, 2008).

Detailed historical data on household spending on cars is not available,

but based on figures for car ownership, it can be concluded that the retail

price of an average car must have declined over the past decades in

relation to purchasing power. As soon as a faster - and thus more expen-

sive in relative terms - transport mode became cheaper in absolute terms,

there was a growing market for it. The modal shift from the bicycle and

public transport towards the car in the 1970s can be explained in this

way. In the 1990s, a similar modal shift occurred in the specific field of

tourism travel: from the car and the train to the plane.

Introduction

31

Fig. 1.2. Evolution of the transport share of household expenditures in

Belgium. Source: Statistics Belgium (2008)

1.3.4 The combined travel time and cost budget

The substitution of the metre by the second as a measure for expressing

perceived distance is not absolute. In fact, we may talk about a combined

travel time-cost budget, in which both the kilometre-dependent cost and

the required time are reckoned. The ratio between the temporal and the

financial component in the time-cost budget is determinant for the

perception of physical distance by an individual. In Belgium, the cost per

kilometre has systematically declined over the years, with the introduc-

tion of the railways and tramways in the nineteenth century as a main

milestone. This trend was then accelerated by the democratization of the

car, which became cheaper in comparison with the general prosperity

level and provided - at least until the early 1980s - definitely the fastest

way to make any domestic trip.

Regarding the cost component, the base price of fuel is an important

element. The price elasticity for fuel is highly dependent on the kind of

trip. It is known that business travel and commuting are hardly affected

by price increases. However, in freight transport and recreational and

tourist travel, the connection is much clearer. The relationship between

the amount of touristic air traffic and the price of flights is even almost

linear (Litman, 2007). When fuel prices rise through the peak oil phe-

Chapter 1

32

nomenon, the importance of the cost component in the combined time-

cost budget will increase. Also any climate policy that is rationing the

consumption of fossil fuels (e.g. through emissions trading) would signifi-

cantly rise the importance of the cost component (Keppens, 2006).

Since variable transport costs are almost directly linked to distance

travelled,3 the importance of physical distance in the perception of the

traveller will increase when the share of the financial component in the

time-cost budget is rising. The oil crises of 1973 and 1979 caused a ripple

in the readily availability of fuels, so that over the period 1973-1985 the

share of cost in the time-cost budget was lifted to a higher level (Infla-

tiondata, 2010). Between 1985 and 2003 the price level of oil, adjusted for

inflation, was again quite low. In May 2008 oil prices had approximately

quadrupled since 2003. Oil prices reached a record of 147 USD per barrel

on July 11, 2008, after which the price quickly fell. In the period 2009-

2010, after some fluctuations it remained relatively constant at a level of

about 70-80 USD per barrel.

The ratio between the cost component and the time component in the

time-cost budget depends also on external factors such as the develop-

ment stage of the considered region, and the general state of the world

economy. In countries where the cost of motorized transport swallows, or

would swallow, a large share of household income, physical distance

continues to be determinant. The effort it takes to make a trip on foot or

by bicycle, or even the marginal cost of fuel in the case of a motorized

trip, remains largely linked to the kilometrage. In contrast, in the western

world the cost of having a car, and the fuel cost, has become a basic

expense that is almost always outweighing the perceived benefit of

intensifying the use of the available travel time budget.

The regional stage of development determines the generalization of

car ownership and the performance of public transport, but also the

congestion level and the social sensitivity to external impacts. During the

1970s, for example, in Belgium the average time spent on commuting

decreased, while the length of the commute simultaneously increased. The

massive shift from slow transport modes to the car allowed for a signifi-

cant intensification of the use of the travel time budget. So it became

3 Although the correlation varies with the speed: in steady-flow traffic at

medium speed, an average car operates in the most efficient mode. At high

speeds, but also in stop-and-go traffic, fuel consumption and emissions per

kilometre go up quickly (Deakin, 2007).

Introduction

33

possible to go living further away from one’s workplace, or to look for a

job far from home. Since 1981, in Belgium the time spent on a particular

commute is again in line with the distance. The generalization of car

ownership and the saturation of the road infrastructure does not any

longer allow to increase the average commuting speed (Fig. 1.3)

(Mérenne-Schoumaker et al., 1999; Verhetsel et al., 2007).

0

5

10

15

20

25

30

35

1970 1981 1991 2001

km

/day a

nd

min

/day

average home-to-work

distance (km)

average home-to-work time

(min)

Fig. 1.3. Evolution of the average commuting time and distance in

Belgium. Sources: Mérenne-Schoumaker et al. (1999); Verhetsel et al.

(2007)

The external effects of transport, as an umbrella for road safety and

environmental effects, play their role too. As there is more traffic on the

roads, public support for government action is growing. Speed limits are

in this very significant because they directly affect the average speed.

Financial policies, including fuel taxes and fixed or variable tolls, empha-

size again the cost component.

1.4 The rebirth of distance

1.4.1 Absorption of rising energy prices

In the future, potentially rising energy prices could lead to a reappraisal

of physical distance. In other words, the unit s will slowly but surely be

pushed aside by the unit m. The spatial structure and the coherence and

distance between different functions and services will be more pronounced

Chapter 1

34

in the individual perception, since one will be inclined to minimize not

only travel time but also travel distance. “Proximity”, a concept that is

associated with physical distance (as opposed to the concept of “accessi-

bility” that is rather associated with time distance), will gain in value,

just as slower and thus more energy-efficient transport modes.

To date, no recent evolutionary data sets that support this claim are

known. Leaving aside the oil crises of the 1970s, significant increases in

fuel prices have only very recently manifested, and were also short-lived.

The data available for Belgium appears to show that covered distances

continue to increase every year. However, for the number of car kilome-

tres travelled, there is a certain slowdown in growth since 2000

(Goossens, 2008). This is probably due to saturation of the road network,

rather than to an increased level of costs. In contrast, the number of train

passengers, and users of public transport in general, continues to grow

steadily. The distance between home and workplace certainly continued

to rise until 2001 (Verhetsel et al., 2007). Also, the number of registered

vehicles in Belgium continues to increase: in 2009, 5.5% more vehicles

were driving around compared to 2005 (Statistics Belgium, 2010).

Moreover, there is a technological buffer that may absorb for some

time a considerable part of the potential price increase, since it is possible

to make the fleet a lot more efficient. From 2002 to 2005 the average fuel

consumption of new cars sold in Belgium decreased already from 6.3 to

5.9 l/100 km. Second, the average weight of a newly sold car increased

from 1993 to 2004 by 30%, while the consumption level of a standard

passenger car with hybrid engine technology is only 4.3 l/100 km (De

Vlieger et al., 2006), and the most economical of all small cars even get

by on 3.4 l/100 km. So there is still a margin that allows reducing fossil

fuel consumption and vehicle emissions significantly by using more

efficient engines in smaller and lighter cars. Nevertheless, history teaches

us that any improvement in energy efficiency of motor vehicles is more

than offset by an increase in overall kilometrage. This pattern is consis-

tent with the Jevons paradox, and the macro-economic extension of it -

known as the Khazzoom-Brookes postulate - which states that, at

constant energy prices, any technological improvement in energy effi-

ciency at the micro level, may result in an overall increase (and not a

decrease) in energy consumption (Saunders, 1992). Since around 2000 in

Belgium we observe a tendency towards stabilization, at least with

respect to energy consumption by passenger transport (De Vlieger et al.,

Introduction

35

2006). Also, the use of alternative fuels should be taken into account,

even though these will have little impact on price making.

Research on price elasticity indicates a clear link between cost and

consumption of fuel, although there are noticeable differences depending

on the type of trip (Bogaert et al., 2006). Short-term price increases are

mainly captured by travelling less kilometres by car. In the longer term,

effects are more important in terms of vehicle efficiency and car owner-

ship. Long-term price elasticities are usually larger than short-term

values. Based on a meta-analysis approach, Brons et al. (2008) find -0.12

as the short-term price elasticity for mileage per car, while the value for

fuel efficiency is 0.14, and -0.08 for car ownership. In the long term, these

values amount to -0.29, 0.31 and -0.08, while the aggregate long-term

price elasticity is -0.84. The absolute value of these figures is less than 1,

meaning that demand for fuel is relatively inelastic.

An important explanation for the inelasticity is the virtual absence of

possible substitutes. There are only few, if any, fully-fledged alternative

fuels available at an equal or lower price. Furthermore, regional varia-

tions exist: North American, Canadian and Australian data show lower

elasticities, which is almost entirely due to the smaller perceived effect on

car ownership. On the other hand, based on comparative research

between similar cities in the US and Australia, Wegener and Fürst (1999)

found that Australians consumed per capita only a little more than half

the volume of fuel that was consumed in the US. Differences in taxation

make the cost of petrol and diesel in Australia about twice as much as in

the US, explaining most of the difference in consumption.

Regarding the distant future, we can say that existing analyses of

price elasticities do not take into account very long term effects, such as

shifts in terms of regional economic structure, real estate prices and

urban development. Brons et al. (2008) recognize that current analyses do

not constitute grounds for very long term predictions. In addition, they

show that mainly travel distance is sensitive to price rises as the studied

time span increases. Another important element is the ceteris paribus

assumption, which is common in price elasticity studies. How mobility

would evolve if a sharp rise in oil prices leads to an economic recession

cannot be inferred from these studies.

Chapter 1

36

1.4.2 Mobility, accessibility and spatial structure

It is clear that reducing the oil dependence of the (person) transport

system has an important technological component: cars need to become

more fuel efficient and alternative fuels should get adopted wherever

possible. The second component, however, has to do with adapting travel

behaviour: a reduction of oil dependence means less car kilometrage (both

by travelling less kilometres, and by a modal shift), and to some extent

less mobility.

In the economic approach, however, mobility is considered as a sec-

ondary activity, and not as a purpose in itself. The assumed objective is

to optimize the accessibility, where accessibility should be seen as a

measure of the interaction potential of an individual, a family or an

organization. According to Geurs and Ritsema van Eck (2001, p. 36)

accessibility reflects the extent to which the “land-use transport system

enables (groups of) individuals or goods to reach activities or destinations

by means of a (combination of) transport mode(s)”.

It is possible to realize a wide range of interaction possibilities, and

thus potential destinations, within short distance. Some of the densest

urban centres in the world, such as Manhattan or Hong Kong, illustrate

this. This mechanism underlies the formation of most cities, increasing

accessibility based on a minimum of transport. Engwicht (1992, p. 12)

puts it as follows: “Cities are an invention to maximise exchange and

minimise travel.”

The work of Newman and Kenworthy (1989, 1999) was the first that

managed to identify a link between the spatial structure of a variety of

world cities and energy consumption for transport. Their graph, of which

Fig. 1.4 is a reproduction, has over the years become popular among

urban and traffic planners. Newman and Kenworthy (1989, 1999) note

that population density shows an inverse relationship with energy

consumption for transport. Residents of densely populated cities cover

shorter distances, since their daily destinations are closer to home. In

addition, they use more often public transport, compared to cities with a

low population density. This phenomenon is strengthened by the higher

level of congestion and parking limitations that are associated with a high

density.

Introduction

37

Fig. 1.4. Fuel consumption per capita versus urban density (1980).

Source: Newman and Kenworthy (1989)

The discourse of Newman and Kenworthy (1989, 1999) has given rise to a

line of research that investigates the relationship between spatial struc-

ture and sustainability of travel patterns in more detail. Some studies

seem to confirm broadly the argument of Newman and Kenworthy (1989,

1999), while other researchers sharply criticise the methodology originally

used, particularly on the demarcation of the urban areas that were

examined in the original study (Mindali et al., 2004; Mees, 2010). Al-

though there is little debate about the existence of a relationship between

spatial structure and travel patterns (Ewing and Cervero, 2010), there is

less unanimity about the sense or nonsense of conducting spatial policies

Chapter 1

38

aimed at sustainable mobility. Gordon and Richardson (1997) for exam-

ple believe that the connection is too weak to justify interventions.

According to them, as cities are sprawling spontaneously, this would be

the most efficient form of urban growth within the current market

economy.

A more nuanced approach is the assumption that planning is neces-

sary, but will not suffice alone to make mobility more sustainable (Zhang,

2002, p. 3). However, the spatial structure is a hard constraint that

determines whether it is possible to shorten trip lengths or to take public

transport or the bike at all.

1.4.3 Urban sprawl

The acknowledgement that uncontrolled and unplanned urban growth

entails problems is far from new. In the peer-reviewed literature, several

types of quasi-unorganized development have been lumped together as

“(urban) sprawl”. Inspired by Harvey and Clark (1965), Ewing (1997)

distinguished three types of sprawl: (i) leapfrog or scattered development,

(ii) commercial strip development, and (iii) large expanses of low-density

or single-use development. In addition, Ewing (1997) states that indica-

tors based on accessibility are perhaps more suitable to detect sprawl

than morphological measures, which is an important contribution com-

pared to Harvey and Clark (1965), even without putting this statement

into practice. Nevertheless, in most of the literature, the definition of

sprawl has a strong emphasis on the morphological character. Torrens

and Alberti (2000, p. 4) call sprawl “a relatively wasteful method of

urbanization, characterized by uniform low densities”. They put forward

seven different variables as indicators for sprawl, including six based on

morphological characteristics (ranging from density gradient to fractal

dimension) and only one based on accessibility. Galster et al. (2001, p.

685) developed a more comprehensive definition: “Sprawl is a pattern of

land use in an urban area that exhibits low levels of some combination of

eight distinct dimensions: density, continuity, concentration, clustering,

centrality, nuclearity, mixed uses, and proximity.” Moreover, they focus

on the evolutionary nature, viewing sprawl as a process and not as a

state.

In the peer-reviewed literature, and in many policy documents, a

strong link between sprawl and increased traffic is assumed (Ewing et al.,

Introduction

39

2002). This is logical bearing in mind that separation processes between

origins and destinations inevitably entail larger distances to be covered.

If we restrict ourselves to commuter traffic, we see that trip lengths

have increased systematically over the past decades. This trend was

observed in Belgium (see before), the US and the UK, and we may

assume that this evolution is manifest throughout the Western world

(Aguilera, 2005). Even though there seems to be a link between the

presence of sprawl and the average trip length, this reasoning does not

necessarily imply a one-way causality. Sprawl is caused by increasing

individual mobility, i.e. a wealth-related phenomenon that is the basis of

an autonomous growth of traffic. So, we might explain the wave of

suburbanization as a materialization of this increased mobility, which has

itself a mutual reinforcing effect on the growth of traffic. Gilbert and Perl

(2008, p. 235) formulate this phenomenon as follows: “Sprawl is believed

to be facilitated by car ownership and use and also to contribute to it, in

a positive feedback loop that reinforces both low-density development and

motorization.”

Primarily in the US the debate on the advantages and disadvantages

of sprawl has been developed into a highly polarized, dichotomous

discussion. From a sociological perspective we can consider James How-

ard Kunstler as advocate of the anti-sprawl school. Kunstler (1993) uses

cultural, aesthetic and ecological reasons for rejecting suburbia as a

human habitat. On the other side of the spectrum, we find Robert

Bruegmann, who describes the American suburb as a natural materializa-

tion of the American Dream, yielding both many benefits for residents

and a number of social and ecological problems (Bruegmann, 2006).

Surprisingly, this ideological duality also appears to exist when we

view sprawl and suburbanization from a mobility perspective. Newman

and Kenworthy (1989) advocate compact cities with a high density, since

these would require less energy per capita than suburbanized regions. The

main reason for this is that distances are reduced at high densities, and

that low residential densities may never provide a sufficient basis for the

organization of efficient public transport. At the other side, Gordon and

Richardson (1989) argue that suburban areas eventually evolve into full-

fledged urban areas, while the intra-suburban traffic is much more

efficient and less sensitive to traffic congestion than inner-city traffic.

They find the more difficult organization of public transport in the

suburbs of minor importance, since car traffic is relatively smooth thanks

to the abundance of space, reducing the need for alternatives. They find

Chapter 1

40

the argument of Newman and Kenworthy (1989, 1999), which is built on

fuel dependence, too weak, since there is no scarcity (yet) and energy

consumption levels are far more influenced by economic factors than by

spatial structure.

Although both Newman and Kenworthy (1989, 1999) and Gordon

and Richardson (1989, 1997) rely on data and scientific methods to

underpin their argument, both of their analytical frameworks are strongly

ideologically biased. In their studies, sprawl is associated with a specific

morphology, particularly monotonous suburban districts with a strict

separation of functions, characterized by store strips, commercial archi-

tecture and large internal distances. However, the extent to which spatial

separation leads to an effective increase of distances that need to be

covered is less clear, since this cannot be derived from local morphological

characteristics. A monotonous residential lot embedded in a major

employment centre could possibly lead to a more sustainable travel

pattern than a compact town that is immersed in a rural area. Particu-

larly in a context where average trip lengths, and in particular average

commuting distances, have become very large in practice, it is hard to tell

which kind of spatial developments are problematic in relation to mobil-

ity, and which are rather beneficial. Banister (1999) observes the issue in

a non-morphological way, and suggests that it is utmost important that

new developments are sufficiently large and are located in or immediately

subsequent to existing urban areas. Local morphology, density and spatial

diversity come only in second place.

Over the last decade, research on the link between travel and land

use has evolved and improved in a number of areas. Datasets have

expanded and have become more reliable over the years, statistical

methods have improved, and additional hypotheses (e.g. on causality, but

also on behavioural aspects such as attitudes and (residential) self-

selection) were incorporated. The dichotomy that was introduced by

Newman and Kenworthy (1989) and Gordon and Richardson (1989),

evolved to a more nuanced understanding which provided the basis for a

series of recent textbooks such as Boarnet and Crane (2001), Glaeser and

Kahn (2003) and Levinson and Krizek (2008).

1.4.4 The paradigm of the compact city

In Europe, since World War II, in a number of agglomerations tackling

urban sprawl is considered an inherent part of the basic principles of

Introduction

41

what constitutes proper urban planning. The main examples that fit into

this tradition are the New Towns Act (UK, 1946), the Finger Plan of

Copenhagen (1947), the various Notes on Spatial Planning (The Nether-

lands, from 1960) the Villes Nouvelles around Paris (since 1965), et

cetera. In other, relatively densely populated European regions that did

not develop around a single city, such as the Ruhrgebiet (Germany) and

the Flemish Diamond (Belgium), the focus on steering spatial develop-

ment came much later. In these suburbanized regions, which have

developed between and around an historically polycentric structure,

traditionally less important economic problems were involved with

additional constructing. The more dispersed nature of the historical

spatial structure and the presence of a dense railway network resulted in

a relatively late manifestation of congestion problems. In the US, the

focus on thorough urban planning is also a relatively recent phenomenon,

which has resulted in, e.g., Florida’s anti-sprawl rule (Ewing, 1997) or

Portland’s urban growth boundary (Song and Knaap, 2004).

In 1990, the European Union took a position against the further de-

velopment of sprawl by endorsing the paradigm of the compact city in

the Green Paper on the Urban Environment (CEC, 1990). The compact

city is presented as the ideal to be pursued, ensuring a sustainable

development. Mobility is one of the key guiding principles that form the

basis of the compact city policy, since slow traffic and public transport

would get more opportunities through shorter distances and higher

densities. The compact city remained an important issue in the European

Spatial Development Perspective (ESDP, 1999, p. 22).

The origin of the European compact city policy can be found in the

existing post-war spatial strategies of a number of EU member states,

whose objective was to guide population growth while safeguarding a

sufficiently high quality of life. In the period when the foundation for

these plans was laid, still mainly benefits were expected from the auto-

mobile, so we can say that the sustainable mobility story was only

subsequently added to the concept of the compact city. A second track

that formed the basis of the compact city policy, is the nostalgic image of

the medieval city-state, which is located in the surrounding countryside

and where all possible destinations are within walking distance of each

other.

In one form or another, the compact-city policy turns up in most pro-

gressive spatial policy plans for cities and regions in Europe. The Spatial

Structure Plan for Flanders (RSV, 1997/2004) is no exception.

Chapter 1

42

Although by most urban planners the compact city policy is consid-

ered valid in a European context, also a number of critiques have been

developed. Perhaps the main criticism is the reasoning that the creation

of spatial proximity by increasing density and mixing functions does not

necessarily lead to a reduction in the number and the length of car trips,

but merely provides the appropriate spatial framework. Moreover, the

compact city policy does not take into account the size of the city. In a

city like London, for example, which is by its size much more compact in

a geometrical sense than a medium sized city, per capita energy consump-

tion for transport is higher than in medium sized British cities (Breheny,

1992). According to Banister (1992), in remote small towns with a limited

degree of self-sufficiency, transport energy consumption per capita is

easily one third higher than average, while in rural areas this may rise up

to three times the level of a regional city.

In Belgium, there has been only very limited research into the possi-

ble effects of a compact city policy on travel behaviour. Verhetsel (2001)

compared a policy scenario according to the Spatial Structure Plan for

Flanders (RSV, 1997/2004) with a trend scenario for the spatial devel-

opment of the Antwerp urban region, and concluded that the effects of

planning are very limited. This study considered newly expected devel-

opments over the period 1991-2010. Obviously, these are engrafted onto

the existing spatial structure of the Antwerp urban region, so that any

structural change would be necessarily small in relative terms. It is

therefore not surprising that the direct impact of an infill scenario is

limited within the considered planning horizon.

Breheny (1996) discusses the centralist and decentralist movements,

which in the 1970s strongly determined the debate in the field of spatial

planning. He concludes that eventually the compromise is the best

solution. Compactness and diversity is important but should not lead to a

loss of environmental quality. Moreover, it is not feasible to exclude

greenfield developments completely, and there should also be a role for

urban networks.

With regard to movement, it has become clear that the spatial struc-

ture should provide the framework that allows for a sustainable travel

pattern, with a substantial role for short-distance trips, public transport

and non-motorized modes. However, it cannot be expected that travel

behaviour will be steered only by the morphology of space.

Introduction

43

1.4.5 Public transport and spatial structure

Although this dissertation focuses on measuring spatial proximity in

relation to the amount of mobility, expressed in person kilometres, still

some comments should be made on the possible role of spatial structure

in relation to the potential for public transport.

The main spatial condition for efficient public urban transport con-

sists of the presence of a critical mass of potential travellers and

destinations, situated within walking distance of the stops in the network.

To ensure a competitive position compared to the car, it is also important

that car traffic is relatively slow and unattractive, for example due to

congestion or a limited number of available parking spaces. These

conditions are associated with thresholds, both in terms of density and in

relation to the size of the city. According to Newman and Kenworthy

(1989) a density of about 30-40 inhabitants per hectare (as an average for

the whole metropolitan area) is the threshold below which it is impossible

to develop an efficient public transport system. Levinson and Kumar

(1997) mention a minimum of 10,000 people per square mile (equivalent

to 39 persons per hectare), while Mees (2010, pp. 51-54) cites several

sources that discuss a variety of figures ranging from 48 to 275 people per

hectare. According to the 4th RTD Framework Programme of the

European Commission, the share of car use in a city could only decrease

in favour of public transport starting from a size of 750,000 inhabitants

(Wegener and Fürst, 1999).

Also at a higher scale level efficient public transport is possible, albeit

within a spatial network structure. The nodes of the network consist of

towns or villages with a high density and a limited size, concentrated

around the stops (stations). With a sufficiently large mass of potential

travellers, located within the radius of action of available pre- and post-

transport, public transport may also play an important role within this

network. Within the towns or villages which are the nodes, however,

public transport will hardly play any role, given the limited size of the

node. Ideally, the size of the settlement, therefore, will be limited to the

range of a pedestrian. To be competitive with automobile traffic, high

frequency and speed is required, in combination with relatively slow car

traffic (e.g., due to congestion or other speed restrictions).

In all other cases (local traffic in small and regional cities, or traffic in

and from vast and sparsely populated suburban areas and the outlying

area) public transport will never perform a significant share of all trips,

Chapter 1

44

and won’t be able to play a role in the sustainability and efficiency of the

transport system. In all these cases, public transport will continue to take

only a niche of the transport market and will have primarily a social

function.

However, a spatial structure which is completely responsive towards

public transport opportunities, is not necessarily the most suitable for

other forms of sustainable transport such as walking, cycling or even

short car trips. Because of the crowded traffic large cities are scoring

relatively poorly in terms of bicycle use. Due to their large size these

cities also often exhibit a more thorough separation of functions than is

the case in small towns. Therefore, the distances covered in big cities are

not necessarily shorter than in smaller cities. Also, big cities generate

significant amounts of commute from the surrounding area. The theoreti-

cal potential for walking is therefore suppressed, and an important

volume of car traffic remains present. The finding by Breheny (1992) that

the fuel consumption per capita in London is slightly higher than in the

British regional cities, despite the efficient London public transport

system, is significant in this regard. The assumption that an urban

structure that is suitable to be served by a high quality public transport

network is, by definition, a sustainable system, is one of the weaknesses

in the compact city theory.

In Flanders the railway-grafted historically polycentric structure ful-

fils the spatial conditions for the operation of a regional public transport

system. However, most spatial developments that occurred after World

War II no longer meet these, just as the many settlements where the

existing rail infrastructure was removed or is no longer operated. Regard-

ing the role of local public transport, Brussels is the only agglomeration

in our study area that meets both the density requirement and the

criterion of minimum size. The metropolitan areas of Antwerp and to a

lesser extent, Ghent, are in compliance with the density requirement of

30-40 persons per hectare, but lack the critical mass needed to justify the

implementation of a high end public transport system (such as a metro

network).

Of course, even this situation may change if energy prices would start

soaring extremely one day, or in case climate change would become top

political priority. Public transport is generally much more energy efficient

than private transport, so especially an extensive rail network could

ensure also in these conditions a high degree of accessibility for large

parts of the population, apart from the car. Mees (2010, p. 53) states that

Introduction

45

public transport should play a role anyway in a future-oriented vision of

mobility, even if most neighbourhoods do not meet the specified density

requirements: “Arguments that densities many times current levels are

needed before transport trends can change are really arguments for

continuing with automobile dependence.”

1.5 Flanders and Brussels: policy context

The research which is reported in this dissertation, is funded by the

Policy Research Centre on Regional Planning and Housing - Flanders,

and is limited to the territory of the Flanders Region. Geographically

speaking, this region, with over 6 million inhabitants and an area of

about 13,500 km2, covers the northern half of Belgium, but excludes the

Brussels Capital Region which is an enclave of only 161 km2 and is home

to over 1 million inhabitants. Given the geographical context, Brussels

was included in most analyses in this dissertation, unless data limitations

did not allow incorporation. It is however important to note that policies,

both in terms of planning and in terms of mobility, are in Flanders and

Brussels based on different policy plans that were drafted and adopted by

two individual governments.

In 1997, the Spatial Structure Plan for Flanders (RSV, 1997/2004),

which may be considered as the first full-fledged spatial policy plan for

the Flanders Region, chose resolutely for strengthening the dichotomy

between urban areas and the countryside. This plan introduces “decon-

centrated clustering” as the first main principle for steering spatial

development (Albrechts et al., 2003; Allaert, 2005, p. 10). “Clustering”

means selectively concentrating the growth of living, working and other

social functions in cities and centres, while “deconcentrated” means

accounting for the existing (deconcentrated) development pattern and the

spread distribution of dynamic functions throughout Flanders. The

protection of open space and the revitalization of the urban fabric are

clearly paramount. By pursuing a spatial concentration of development in

precisely those areas that already possess a significant density, fragmenta-

tion of the (open) space is supposed to be combated, while existing

facilities and infrastructure will be used in a more efficient and more

sustainable way.

A concrete instrument proposed to implement these objectives is the

demarcation of urban areas, meaning that a line is drawn around those

Chapter 1

46

areas that should be reserved for the development of new highly dynamic

activities. Additional supply of industrial and residential land is provided

in these urban areas, and in new residential developments a minimum

density of on average 25 dwellings per hectare is aimed for. Also in the

nuclei of the countryside a minimum density is required, although this

requirement is only 15 dwellings per hectare. From a spatial perspective,

the alleged advantages of demarcating the urban areas are as follows:

scarce space is dealt with in a more economical way, facilities end up

closer to homes and are therefore more accessible and cheaper, contiguous

agricultural areas do not continue to fragment and ecological relation-

ships are preserved.

The RSV suggests that the demarcation provides also a number of

advantages in terms of mobility: “The principle of deconcentrated

clustering leads to a spatial clustering of needs for movement. This leads

to broadening of individual travel options (e.g. different destinations

remain accessible on foot, by bicycle). Consequently, the accessibility of

various facilities is basically higher. Spatial clustering is a prerequisite for

common transport. Through spatial concentration of origins and destina-

tions, an overall reduction of the amount of traffic is also possible.”

(RSV, 1997/2004, p. 472) Regarding the spatial component of mobility

policy in Flanders, generally the RSV is referred to (MOW, 2001).

The clear option to demarcate the urban areas demonstrates that the

RSV is undeniably following the compact city model, although this is

supplemented by explicit attention to the functioning of the compact

urban areas as nodes in urban networks (e.g. the Flemish Diamond,

connecting the metropolitan areas of Antwerp, Ghent and Brussels

Capital Region with their surroundings). Such an urban network can be

considered as a spatial concept for an urban structure at a different scale

level and as an extension to the compact city model.

In Brussels, the RSV is not in force. The second version of the over-

arching spatial policy plan for Brussels (Gewestelijk Ontwikkelingsplan /

Plan Régional de Développement (GewOP/PRD, 2002)) is in fact not a

regional plan but rather an urban development plan. Compared with

RSV, the GewOP/PRD is much clearer in relation to the regulation of

density and land use mix. Moreover, the development of public transport

and a bicycle network are essential parts of the plan, and one of the

strategic objectives is to reduce car traffic by 20%.

Introduction

47

1.6 Research questions, conceptual

framework and implementation

1.6.1 Research questions

Based on the foregoing considerations, we formulate the general research

question that is underlying this thesis, as follows:

• To what extent is mutual proximity of potential destinations deter-

minant for daily travelled distances in Flanders, and what does this

mean in a context of peak oil and climate change? (A)

In order to make the basic question operational, we subdivide it into

three questions:

• How can the influence of spatial structure on daily travel distances be

quantified? (B1)

• How can spatial proximity be defined, measured and applied in

sustainable spatial planning practice? (B2)

• What locations should be selected for additional housing or jobs, if

we want to keep additional traffic generation as small as possible?

(B3)

1.6.2 Conceptual framework

The research questions summarize the research framework which stems

from the broader context outlined. In this section we propose a concep-

tual framework that clarifies the links between different aspects of the

research. Fig. 1.5 is a schematic representation of this conceptual frame-

work.

The conceptual framework is policy oriented and uses the method

that Mayer and Greenwood (1980) developed specifically for policy

research. The conceptual framework examines the opportunities that are

offered by the policy area of spatial planning to meet the stated objec-

tives, within Flanders (where possible extended to Brussels). Possible

influences of other policy areas (mobility, environment, climate change

and taxation) are regarded as exogenous. Two policy objectives to be

pursued are formulated: (1) improving the sustainability of daily travel

patterns, and (2) mitigating the spatial and economic vulnerability to

potentially rising transport costs, and thus strengthening resilience.

Chapter 1

48

Fig. 1.5. Conceptual framework

As an instrumental objective, which is assumed to be strongly impres-

sionable by spatial planning policy, we propose an increase in spatial

proximity. Spatial proximity is represented by a non-exhaustive set of

different measurable spatial variables such as the jobs-housing balance

(the ratio between the number of jobs and the number of working

population), density (e.g. population density, or job density), diversity

(the degree of spatial mix of different functions), and so forth. Part of the

research will be to select variables based on their representativeness of

the characteristic called spatial proximity, and to find ways to measure

these accurately.

The relationship between spatial proximity and the two proposed ba-

sic policy objectives is based on two assumptions. First, we know from

the literature that a higher population density and a better mix of

functions usually accompanies shorter trip distances and a lower share of

car drivers, factors indicating a more sustainable travel pattern. Although

the found relationships are often weaker than would be expected intui-

tively, they are usually statistically significant. This means that

strengthening the spatial proximity is expected to result in a more

sustainable travel pattern. How strong this effect is, depends inter alia on

personal socioeconomic characteristics, such as income, family composi-

Introduction

49

tion or personal preferences. In short, in this context spatial planning is

used to pursue environmental and climate objectives.

Second, we argue that increasing characteristics of spatial proximity

reduces the vulnerability of the spatial-economic system to increasing

transport costs (Dodson and Sipe, 2008). Although there is no extensive

literature available on this proposition, it is logical that increasing

transport cost is critical for interaction possibilities in a spatial system

where everything is far away from everything else. In urban structures,

characterized by a high degree of spatial proximity, however, interaction

possibilities remain ample even when transport is limited. The reason why

this view has hardly been addressed in the scholarly literature is the

almost continuous decline in transport costs throughout history. It is only

after acknowledging the sense of reality of the peak oil theory that an

increase in transport cost in the long term can be seen as a likely sce-

nario.

This brings us to the right part of the scheme depicted in Fig. 1.5,

where the influences that are considered as exogenous in this conceptual

model are located. The oil price is represented as the most important

factor, since it is highly autonomous and can only to a very limited

extent be controlled. Technological developments are important too in

the sense that a combination of more efficient engines and alternative

fuels could eventually substitute a part of the growing demand for oil.

Technology will have both an indirect (through transport costs), and a

direct effect on the two key policy objectives. Finally, also some other

policy areas play their role in determining the cost of transport. Never-

theless, in the overall picture, the impact of these policy factors may

perhaps be less important, since these are to some extent the result of a

social consensus, unlike the oil price which is mainly determined by

physical supply limits. The combination of all these exogenous influences

determines the relative cost of transport. This relative cost must be seen

in an objectified form: for example, the average cost to move one person

over one kilometre, expressed as a percentage of GDP. Ultimately, the

relative cost partly determines the sustainability of daily travel (distance

travelled and transport mode) and the extent to which the economy may

potentially suffer from spatial vulnerability.

A final element that occurs in the conceptual framework consists of

the so-called unintended effects of strengthening spatial proximity, which

are summarized under the heading of “land use restrictions”. The deliber-

ate strengthening of certain characteristics of spatial proximity, such as

Chapter 1

50

population density and functional diversity, inevitably entails implica-

tions that are not universally perceived as positive, such as higher land

prices, smaller houses or more potential nuisance. Unintended effects are

included in the conceptual framework for completeness, but are no

subject of further research in this dissertation.

1.6.3 Implementation

The implementation of the conceptual framework follows two tracks. The

first track is the placement of the research in the scholarly literature.

Although the thesis does not contain a separate chapter on existing

literature, each chapter is embedded in a review of the relevant literature.

This was done by consulting mainly international peer-reviewed journals,

in most cases included in the ISI / Thomson Reuters Web of Science

index, and complemented with international standard works and Belgian

or Flemish research reports and policy documents.

As may be presumed from the title of the dissertation, the second

track consists of the actual quantitative analysis. The sources for the

various quantitative analyses are several data sets that were made

available from the Policy Research Centre on Regional Planning and

Housing - Flanders, or by the Flemish regional government itself, sup-

plemented with data that is freely available on the internet. The origin

and content of the data sets used is described in the various chapters.

The applied methods are inspired from the literature, but were ad-

justed according to the proposed research questions and the specific

characteristics of the study area. Applied quantitative methods range

from simple calculations and cartographic representation to data manipu-

lation with own algorithms and simple (t-test, ANOVA and correlation

analysis) or more advanced statistical techniques (multivariate regression

analysis, spatial econometrics and the construction of a forecasting

model). Repeatedly attention is paid to the cartographic representation of

variables used, while the maps themselves are considered as a major form

of outcome. Analyses were performed to the whole study area (the

Flanders Region), if possible supplemented with the Brussels Capital

Region (depending on the availability of data).

Introduction

51

1.7 Overview of the research

Each of the following six chapters contains a clearly defined sub-study

that is providing an answer to a part of the problem. Since each chapter

can be read separately, below an overview is presented in order to

establish the linkages between the chapters (Fig. 1.6). All off these six

chapters have been published, or will be published in international peer-

reviewed journals. A basic version of each of these chapters was presented

at least at one scholarly international conference.

In the second chapter, which is published as Boussauw and Witlox

(2009), energy consumption levels for commuting are calculated and

mapped on the basis of residential location. Regional differentiations in

commuting distances, modal shares of non-car travel modes and aspects

of infrastructure and population densities are used to explain some

relationships between energy consumption, commuting behaviour and

spatial-economic structure in Flanders and Brussels. It is found that

mode choice appears to be of little impact for the energy performance of

home-to-work travel on a regional scale. At the other hand, proximity

between home and work locations is paramount. Residential density plays

a part in this, although much depends on the specific situation. This is

also the case for the accessibility of the main road and rail network. In

some regions these infrastructures induce long-distance commuting,

whereas in the economic core areas this effect is much less pronounced.

All these are factors that are very much determined by infrastructural

and spatial policies of the past.

Chapter 3, which will be published as Boussauw et al. (2011a), as-

sesses the possibility to use a spatially disaggregated form of the so-called

minimum commuting distance as a spatial proximity characteristic with

regard to the commute. This paper focuses on regional variations in

commuting trip lengths by calculating minimum (required) commuting

distances, along with excess commuting rates. So, this paper contributes

to the excess commuting research framework from a regional perspective,

both by stressing the specific characteristics of urban networks with

overlapping commute areas, and by putting forward an alternative

method for calculating spatially disaggregated values. In the study area,

large variations in minimum commuting distances occur. This in turn

identifies to a large extent opportunities for shrinking commuting dis-

tances by influences such as rising fuel prices, compact urban planning,

extreme congestion or dissuasive traffic policies.

Chapter 1

52

Fig. 1.6. Dissertation outline

The fourth chapter (Boussauw et al., 2011b) extends the excess commut-

ing framework to a longitudinal assessment, aiming to measuring spatial

separation processes between jobs and housing in Flanders. It is known

that the average distance covered by individual commuting trips increases

year after year, regardless of the travel mode. Although increasing

prosperity is often invoked as the main reason, the discipline of spatial

planning also points to the relevance of land-use policies that enable

processes of suburbanization and sprawl. By calculating time series of

spatially disaggregated theoretical minimum commuting distances, this

paper offers a method to identify and quantify the process of spatial

separation between the housing market and the job market. We consider

the detected spatial separation as an indicator for the contribution of

spatial processes to the growth of traffic.

In the study area, it is found that over time the minimum commuting

distance increased in many municipalities, especially where population is

growing faster than job supply, or where traditionally high concentrations

Introduction

53

of employment still increase. Decreases are noticed in suburban areas that

are getting a more urban character by acquiring a considerable functional

mix. For the study area in its entirety, we do indeed register an increas-

ing spatial separation between home and work locations. However, this

separation evolves less rapidly than the increase in commuting distances

itself.

Regarding the methodology, we find that the use of municipalities as

a spatial entity is suitable for grasping regional transformations of the

economy and intermunicipal forms of suburbanization and peri-

urbanization. However, a similar methodology, applied at a more detailed

geographical scale, could be used to detect processes of sprawl in the

morphological sense.

As a third contribution from the excess travel perspective, Chapter 5

(Boussauw et al., 2011c) is applying the excess commute framework on

non-commuting daily travel. Based on the spatial distribution of some

quasi-daily destination classes and survey-reported trip distances regional

variation in excess travel in non-professional trips in Flanders is assessed.

To this end, proximity to various quasi-daily destinations is compared

with the reported distance that is actually travelled to reach similar, but

alternative, facilities.

We note that in rural areas (compared with urban areas) larger dis-

tances are travelled, although the closest facility is chosen more often. In

the most urbanized areas, however, we note that spatial proximity is also

an important aspect in destination choice.

Quantification of these phenomena can support the practice of

sustainable spatial planning by distinguishing areas that are too mono-

functional or too remote, and therefore need more functional diversity,

and by identifying areas where densification is useful because the location

is closely to most quasi-daily destinations, reducing the need to travel

over large distances.

In Chapter 6 (Boussauw et al., 2011d), an assessment is made of the

relationship between selected aspects of spatial proximity (density,

diversity, minimum commuting distance, jobs-housing balance and job

accessibility) and reported commuting distances. Part of this paper is

built on results from Chapter 2 and Chapter 3. Results show that correla-

tions may depend on the considered trip end. For example, a high

residential density, a high degree of spatial diversity and a high level of

job accessibility are all associated with a short commute by residents,

while a high job density is associated with a long commute by employees.

Chapter 1

54

A jobs-housing balance close to one is associated with a short commute,

both by residents and by employees. In general, it appears that the

alleged sustainability benefits of the compact city model are still valid in

a context of continuously expanding commuting trip lengths.

The seventh chapter (Boussauw and Witlox, 2011) extends the

methodology of Chapter 6 to all daily travel, both commuting and non-

commuting, and includes results from Chapter 5. Based on a set of spatial

proximity characteristics this paper develops a model that estimates for

every neighbourhood in Flanders the amount of traffic that would be

generated by an additional residential unit when socioeconomic variables

are held constant. The results show that residential density, land use

diversity and proximity of facilities influence daily travelled distances

when these variables are measured in the immediate vicinity of the

residential location of the respondent (within a radius of 1 km). When

aggregating these variables at a larger geographical scale, in most cases

the impact proves no longer significant. Variables based on the spatial

distribution of jobs, or on the global accessibility of the entire population

in the study area, do not show significant effects on the travel distance.

Despite the statistical significance only a fraction of the observed

variance in reported distances is explained by characteristics of spatial

proximity. However, we can assume that the importance of spatial

structure in the genesis of mobility patterns will increase in case the cost

of transport would rise (cf. peak oil). For this reason, the application of

the mapped results of the proposed model could contribute to the prac-

tice of sustainable spatial planning.

As a conclusion, the eighth chapter summarizes the findings, makes

the necessary critical nuance, and formulates policy recommendations.

Finally, an addendum explores the possible relationship between air

travel behaviour and residential location, an issue that was not included

in the main theme of the dissertation but may be considered as an

important issue when putting travel demand into a global perspective.

Introduction

55

References

Aguilera, A. (2005) “Growth in commuting distances in French polycen-

tric metropolitan areas: Paris, Lyon and Marseille.” Urban Studies.

42(9), pp. 1537–1547.

Albrechts, L., P. Healey and K. R. Kunzmann (2003) “Strategic spatial

planning and regional governance in Europe.” Journal of the Ameri-

can Planning Association. 69(2), pp. 113-129.

Aleklett, K., M. Höök, K. Jakobsson, M. Lardelli, S. Snowden and B.

Söderbergh (2010) “The peak of the oil age - Analyzing the world oil

production reference scenario in World Energy Outlook 2008.” Energy

Policy. 38(3), pp. 1398-1414.

Allaert, G. (2005) Wegwijs in Ruimtelijke Economie: Doorkijk naar

Planning en Management van Ruimte. Gent: Academia Press.

Anable, J. and A. L. Bristow (2007) Transport and Climate Change:

Supporting document to the CfIT report. Aberdeen - Loughborough:

The Robert Gordon University - Loughborough University.

ASPO (2010) Association for the Study of Peak Oil & Gas from

http://www.peakoil.net.

Banister, D. (1992) “Energy use, transport and settlement patterns”. In:

M. Breheny (Eds.) Sustainable Development and Urban Form. Lon-

don: Pion, pp. 160-181.

Banister, D. (1999) “Planning more to travel less.” Town Planning

Review. 70(3), pp. 313-338.

Banister, D. (2008) “The sustainable mobility paradigm.” Transport

Policy. 15(2), pp. 73-80.

Bardi, U. (2009) “Peak oil: The four stages of a new idea.” Energy. 34(3),

pp. 323-326.

Beevers, S. D. and D. C. Carslaw (2005) “The impact of congestion

charging on vehicle emissions in London.” Atmospheric Environment.

39(1), pp. 1-5.

Blauwens, G., P. De Baere and E. Van de Voorde (2002) Transport

Economics. Antwerpen: De Boeck.

Boarnet, M. G. and R. Crane (2001) Travel by Design: The Influence of

Urban Form on Travel. Oxford: Oxford University Press.

Bogaert, S., L. De Smet, F. Verdonck, K. Van Biervliet, S. Logghe, L.

Franckx, J. Eyckmans and S. De Jaeger (2006) Verkennende Studie

Chapter 1

56

naar Prijs- en Inkomens-Elasticiteiten van Milieugerelateerde Goe-

deren en Diensten in Vlaanderen. Aalst: Flemish Environment

Agency.

Boussauw, K. and F. Witlox (2009) “Introducing a commute-energy

performance index for Flanders.” Transportation Research Part A.

43(5), pp. 580-591.

Boussauw, K. and F. Witlox (2011) “Linking expected mobility produc-

tion to sustainable residential location planning: Some evidence from

Flanders.” Journal of Transport Geography. in press.

Boussauw, K., T. Neutens and F. Witlox (2011a) “Minimum commuting

distance as a spatial characteristic in a non-monocentric urban sys-

tem: The case of Flanders.” Papers in Regional Science. 90(1), in

press.

Boussauw, K., B. Derudder and F. Witlox (2011b) “Measuring spatial

separation processes through the minimum commute: The case of

Flanders.” European Journal of Transport and Infrastructure Re-

search. 11(1), pp. 42-60.

Boussauw, K., V. Van Acker and F. Witlox (2011c) “Excess travel in

non-professional trips: Why looking for it miles away?” Tijdschrift

voor Economische en Sociale Geografie. in press.

Boussauw, K., T. Neutens and F. Witlox (2011d) “Relationship between

spatial proximity and travel-to-work distance: The effect of the com-

pact city.” Regional Studies. in press.

Breheny, M. (1992) “The contradictions of the compact city”. In: M.

Breheny (Eds.) Sustainable Development and Urban Form. London:

Pion, pp. 138-159.

Breheny, M. (1996) “Centrists, decentrists and compromisers: Views on

the future of urban form”. In: M. Jenks, E. Burton and K. Williams

(Eds.) The Compact City: A Sustainable Urban Form? London:

E&FN SPON, pp. 13-35.

Brons, M., P. Nijkamp, E. Pels and P. Rietveld (2008) “A meta-analysis

of the price elasticity of gasoline demand. A SUR approach.” Energy

Economics. 30(5), pp. 2105-2122.

Bruegmann, R. (2006) Sprawl: A Compact History. Chicago: University

of Chicago Press.

Brussels Airport (2009) Brutrends 2008. Brussel: Brussels Airport.

Campbell, C. J. and J. H. Laherrère (1998) “The end of cheap oil.”

Scientific American. March 1998, pp. 78-83.

Introduction

57

CEC (1990) Green Paper on the Urban Environment - COM(90)218.

Brussels: Commission of the European Communities.

Cervero, R. and J. Murakami (2010) “Effects of built environments on

vehicle miles traveled: Evidence from 370 US urbanized areas.” Envi-

ronment and Planning A. 42(2), pp. 400-418.

Chapman, L. (2007) “Transport and climate change: A review.” Journal

of Transport Geography. 15(5), pp. 354-367.

City of Bruges (2007) Evaluatie Mobiliteitsplan. Brugge: Stad Brugge -

Mobiliteitscel.

Coolsaet, W. (1990) Naar een Oplossing voor het Verkeersprobleem.

Gent: Kritiek.

De Vlieger, I., E. Cornelis, L. Int Panis, L. Schrooten, L. Govaerts, L.

Pelkmans, S. Logghe, K. Vanherle, F. Vanhove, G. De Ceuster, C.

Macharis, T. Festraets, E. Pekin, L. Turcksin, K. Van Bladel, M. de

Jong, J. Van Mierlo, J. Matheys, J.-M. Timmermans, C. De Geest

and E. Van Walsum (2007) MIRA (2007) Milieurapport Vlaanderen,

Achtergronddocument 2007, Transport. Aalst: Flemish Environment

Agency.

Deakin, E. (2007) “The environmental consequences of sprawl”. In:

Transport, Urban Form and Economic Growth - Round Table 137.

Paris: European Conference of Ministers of Transport - OECD Pub-

lishing.

Dodson, J. and N. Sipe (2008) “Shocking the suburbs: urban location,

homeownership and oil vulnerability in the Australian city.” Housing

Studies. 23(3), pp. 377-401.

ECB (2010) “Oil prices - Their determinants and impact on euro area

inflation and the macroeconomy.” ECB Monthly Bulletin. 12(8), pp.

75-92.

EEA (2008) Data and Maps: EN01 Energy Related Greenhouse Gas

Emissions. European Environment Agency from http://www.eea.eu

ropa.eu/data-and-maps/indicators/en01-energy-related-greenhouse-

gas-emissions.

Engwicht, D. (1992) Towards an Eco-City: Calming the Traffic. Sydney:

Envirobook.

ESDP (1999) European Spatial Development Perspective. Luxembourg:

European Commission.

Ewing, R. (1997) “Is Los Angeles-style sprawl desirable?” Journal of the

American Planning Association. 63(1), pp. 107-126.

Chapter 1

58

Ewing, R. and R. Cervero (2010) “Travel and the built environment: A

meta-analysis.” Journal of the American Planning Association. 76(3),

pp. 265-294.

Ewing, R., R. Pendall and D. Chen (2002) Measuring Sprawl and Its

Impact. Washington, DC: Smart Growth America.

Galster, G., R. Hanson, M. R. Ratcliffe, H. Wolman, S. Coleman and J.

Freihage (2001) “Wrestling sprawl to the ground: Defining and meas-

uring an elusive concept.” Housing Policy Debate. 12(4), pp. 681-717.

Geurs, K. T. and J. R. Ritsema van Eck (2001) Accessibility Measures:

Review and Applications. Bilthoven: National Institute for Public

Health and the Environment.

GewOP/PRD (2002) Gewestelijk Ontwikkelingsplan / Plan Régional de

Développement. Brussels: Brussels Capital Region.

Gilbert, R. and A. Perl (2008) Transport Revolutions: Moving People and

Freight Without Oil. London: Earthscan / James & James.

Glaeser, E. L. and M. E. Kahn (2003) “Sprawl and urban growth”. In: J.

V. Henderson and J.-F. Thisse (Eds.) Handbook of Urban and Re-

gional Economics. Amsterdam: Elsevier, pp. 2481-2527.

Gleeson, B. (2008) “Critical commentary. Waking from the dream: An

Australian perspective on urban resilience.” Urban Studies. 45(13),

pp. 2653-2668.

Glorieux, I., J. Minnen and J. Vandeweyer (2005) Tijdsbesteding in

Vlaanderen: Veranderingen in Tijdsbesteding tussen 1999 en 2004.

Brussels: Vrije Universiteit Brussel.

Gonzalez-Feliu, J., F. Toilier and J.-L. Routhier (2010) “End consumer

goods movement generation in French medium urban areas.” Proce-

dia - Social and Behavioral Sciences. 2(3), pp. 6189-6204.

Goossens, S. (2008) Verkeerstellingen 1e Jaarhelft 2008. Brussels: Minis-

try of the Flemish Community.

Gordon, P. and H. W. Richardson (1989) “Gasoline consumption and

cities: A reply.” Journal of the American Planning Association.

55(3), pp. 342-346.

Gordon, P. and H. W. Richardson (1997) “Are compact cities a desirable

planning goal?” Journal of the American Planning Association.

63(1), pp. 95-106.

Hägerstrand, T. (1970) “What about people in regional science?” Papers

in Regional Science. 24(1), pp. 6-21.

Introduction

59

Hamilton, J. D. (2008) “Oil and the macroeconomy”. In: S. N. Durlauf

and L. E. Blume (Eds.) The New Palgrave Dictionary of Economics.

Basingstoke: Palgrave Macmillan.

Hamilton, J. D. (2009) “Causes and consequences of the oil shock of 2007-

08”. In: D. H. Romer and J. Wolfers (Eds.) Brookings Papers on

Economic Activity: Spring 2009. Washington, DC: Brookings Institu-

tion Press, pp. 215-284.

Handy, S., L. Weston and P. L. Mokhtarian (2005) “Driving by choice or

necessity?” Transportation Research Part A. 39(2-3), pp. 183-203.

Harvey, R. O. and W. A. V. Clark (1965) “The nature and economics of

urban sprawl.” Land Economics. 41(1), pp. 1-9.

Holden, E. and I. Norland (2005) “Three challenges for the compact city

as a sustainable urban form: Household consumption of energy and

transport in eight residential areas in the greater Oslo region.” Urban

Studies. 42(12), pp. 2145–2166.

Hubbert, M. K. (1956) Nuclear Energy and the Fossil Fuels. Houston,

TX - Washington, DC: Shell Development Company - American Pe-

troleum Institute.

Hupkes, G. (1977) Toekomstscenario’s voor ons Vervoerssysteem. PhD

Thesis. Amsterdam: Universiteit van Amsterdam.

IEA (2008) World Energy Outlook. Paris: International Energy Agency.

Illich, I. (1974) Energy and Equity. New York, Evanston, San Francisco,

London: Harper & Row.

Inflationdata (2010) Historical Crude Oil Prices from http://inflation

data.com/Inflation/Inflation_Rate/Historical_Oil_Prices_Table.asp.

INRIX (2008) National Traffic Scorecard - Special Report. Kirkland,

WA: INRIX.

Jenks, M., E. Burton and K. Williams (Eds.) (1996) The Compact City:

A Sustainable Urban Form? London and New York: E&FN Spon.

Joly, I. (2004) Travel time budget - Decomposition of the worldwide

mean. International Association of Time-Use Research Annual Con-

ference 2004. Rome: Italian National Statistical Institute.

Keppens, M. (2006) Modelling Private Road Transport of Flemish

Households. The Mobility and Welfare Effects of Fuel Taxes versus

Tradable Fuel Rights. PhD Thesis. Hasselt: Universiteit Hasselt.

Kopits, S. (2009) “Oil: What price can America afford?” Douglas-

Westwood Research Note. 22nd June.

Kunstler, J. H. (1993) The Geography of Nowhere: The Rise and Decline

of America’s Man-Made Landscape. New York: Simon and Schuster.

Chapter 1

60

Levinson, D. M. and K. J. Krizek (2008) Planning for Place and Plexus:

Metropolitan Land Use and Transport. New York: Routledge.

Levinson, D. M. and A. Kumar (1997) “Density and the journey to

work.” Growth and Change. 28(2), pp. 147-172.

Litman, T. (2007) Transportation Elasticities: How Prices and Other

Factors Affect Travel Behavior. Victoria: Victoria Transport Policy

Institute.

LNE (2006) Vlaams Klimaatbeleidsplan 2006-2012. Brussels: Ministry of

the Flemish Community.

Ma, K.-R. and E.-T. Kang (2011) “Time-space convergence and urban

decentralisation.” Journal of Transport Geography. doi:10.1016/

j.jtrangeo.2010.06.016.

Marinetti, F. T. (1909) “Manifeste du futurisme.” Le Figaro. 20 Febru-

ary.

Marte, G. (2003) “Slow vehicle traffic is a more attractive alternative to

fast vehicle traffic than public transport.” Journal of World Trans-

port Policy & Practice. 9(2), pp. 18-23.

Mayer, R. and E. Greenwood (1980) The Design of Social Policy Re-

search. Englewood Cliffs, NJ: Prentice-Hall.

Meadows, D. H., D. L. Meadows, J. Randers and W. W. Behrens III

(1972) The Limits to Growth. New York: Universe Books.

Mees, P. (2010) Transport for Suburbia: Beyond the Automobile Age.

London: Earthscan.

Mérenne-Schoumaker, B., H. Van der Haegen and E. Van Hecke (1999)

Algemene Volks- en Woningtelling op 1 maart 1991: Werk- en

Schoolpendel. Brussels: Statistics Belgium.

Mindali, O., A. Raveh and I. Salomon (2004) “Urban density and energy

consumption: A new look at old statistics.” Transportation Research

Part A. 38(2), pp. 143–162.

Mokhtarian, P. L. and I. Salomon (2001) “How derived is the demand for

travel? Some conceptual and measurement considerations.” Transpor-

tation Research Part A. 35(8), pp. 695-719.

MOW (2001) Mobiliteitsplan Vlaanderen - Ontwerp. Brussels: Ministry of

the Flemish Community.

Newman, P. and J. Kenworthy (1989) Cities and Automobile Depend-

ence. A Sourcebook. Aldershot: Gower.

Newman, P. and J. Kenworthy (1999) Sustainability and Cities: Over-

coming Automobile Dependence. Washington, DC: Island Press.

Introduction

61

Newman, P. and J. Kenworthy (2006) “Urban design to reduce automo-

bile dependence.” Opolis. 2(1), pp. 35-52.

Reynebeau, M. (2003) Een geschiedenis van België. Tielt: Lannoo.

Righelato, R. and D. V. Spracklen (2007) “Carbon mitigation by biofuels

or by saving and restoring forests?” Science. 317(5840), pp. 902.

Robèrt, M. and R. D. Jonsson (2006) “Assessment of transport policies

toward future emission targets: A backcasting approach for Stock-

holm 2030.” Journal of Environmental Assessment Policy and

Management. 8(4), pp. 451-478.

Rodrigue, J.-P., C. Comtois and B. Slack (2009) The Geography of

Transport Systems. New York: Routledge.

RSV (1997/2004) Ruimtelijk Structuurplan Vlaanderen - Gecoördineerde

Versie. Brussels: Ministry of the Flemish Community.

Saunders, H. D. (1992) “The Khazzoom-Brookes postulate and neoclassi-

cal growth.” Energy Journal. 13(4), pp. 131-148.

Schafer, A. (2000) “Regularities in travel demand: An international

perspective.” Journal of Transportation and Statistics. 3(3), pp. 1-31.

Song, Y. and G.-J. Knaap (2004) “Measuring urban form - Is Portland

winning the war on sprawl?” Journal of the American Planning As-

sociation. 70(2), pp. 210-225.

Statistics Belgium (2008) Levensstandaard - verbruik van goederen en

diensten from http://www.statbel.fgov.be/figures/d323_nl.asp.

Statistics Belgium (2010) Grootte van het voertuigenpark 2010 from

http://statbel.fgov.be/nl/modules/publications/statistiques/verkeer_

vervoer/evolution_du_parc_de_vehicules_2010.jsp.

SVR (2010) Cijfers. Studiedienst van de Vlaamse Regering. Ministry of

the Flemish Community from http://www4.vlaanderen.be/dar/svr/

Cijfers/Pages/Excel.aspx. Retrieved 26/08/2010.

Szalai, A. and P. Converse (1972) The Use of Time: Daily Activities of

Urban and Suburban Populations in Twelve Countries. The Hague:

Mouton.

Torrens, P. M. and M. Alberti (2000) Measuring Sprawl. London: Centre

for Advanced Spatial Analysis - University College London.

Van Wee, B., P. Rietveld and H. Meurs (2002) A Constant Travel Time

Budget? In Search for Explanations for an Increase in Average

Travel. Amsterdam: Vrije Universiteit Amsterdam.

Verhetsel, A. (2001) “The impact of planning and infrastructure measures

on rush hour congestion in Antwerp, Belgium.” Journal of Transport

Geography. 9(2), pp. 111-123.

Chapter 1

62

Verhetsel, A., I. Thomas, E. Van Hecke and M. Beelen (2007) Pendel in

België. Deel I: De Woon-Werkverplaatsingen. Brussels: Statistics Bel-

gium.

Virilio, P. (1977) Vitesse et Politique: Essai de Dromologie. Paris:

Editions Galilée.

VMM (2008) “MIRA-T 2008 Flanders Environment Report: Indicator

Report”. In: MIRA-T 2008 Indicatorrapport: Flemish Environment

Agency.

Weart, S. R. (2010) “The idea of anthropogenic global climate change in

the 20th century.” Wiley Interdisciplinary Reviews: Climate Change.

1(1), pp. 67-81.

Wegener, M. and F. Fürst (1999) “Land-use transport interaction: State

of the art.” Berichte aus dem Institut für Raumplanung(46), pp. 1-99.

Zahavi, Y. and A. Talvitie (1980) “Regularities in travel time and money

expenditures.” Transportation Research Record. (750), pp. 13-19.

Zhang, M. (2002) Conditions and Effectiveness of Land Use as a Mobility

Tool. PhD Thesis. Cambridge, MA: Massachusetts Institute of Tech-

nology.

63

Chapter 2:

Introducing a

commute-energy

performance index

This paper has been published as Boussauw, K. and F. Witlox (2009)

“Introducing a commute-energy performance index for Flanders.”

Transportation Research Part A. 43(5), pp. 580-591. Copyright ©

Elsevier. All rights reserved.

Abstract

Based on 2001-census data for Belgium, energy consumption levels for

commuting were calculated and mapped on the basis of residential

locations in the administrative regions of Flanders and Brussels. Com-

parison with regional differentiations in commuting distances, modal

shares of non-car travel modes and aspects of infrastructure and popula-

tion densities clarifies some relationships between energy consumption,

commuting behaviour and spatial-economic structure in the suburbanised

historic-polycentric spatial structure which characterises the northern

part of Belgium. It is found that mode choice appears to be of little

impact for the energy performance of home-to-work travel on the scale of

the Flanders region. At the other hand, proximity between home and

work locations is paramount.

Residential density plays a part in this, although much depends on

the specific situation. This is also the case for the accessibility of the main

road and rail network. In some regions these infrastructures induce long-

distance commuting, whereas in the economic core areas this effect is

much less pronounced. All these are factors that are very much deter-

mined by infrastructural and spatial policies of the past.

Chapter 2

64

Keywords: sustainable spatial development; commuting; energy per-

formance; Flanders

2.1 Introduction

Following Newman and Kenworthy (1989), many researchers have put

forward the energy efficiency of urban transport as a sustainability

indicator. Although Newman and Kenworthy (1989) were repeatedly

criticized because of methodological reasons, the rationale for the use of

energy performance as an indicator for measuring the sustainability of

transport in relation to spatial structure kept upright.

This paper investigates the link between spatial structure and energy

consumption for home-to-work travel. To this end the concept of a

commute-energy performance (CEP) index will be developed and tested

for Flanders (Belgium). This indicator is not only considered as a proxy

for the sustainability of the transport system in itself, but by extension

for those of the spatial-economic structure as a whole. The results can

constitute a basis for further research, which aims to determine the

robustness of spatial structures in a climate of incipient fuel scarcity. A

better understanding of this matter will uncover social and spatial

evolutions, and leads to a policy that facilitates a more sustainable

development.

The paper is structured as follows. In Section 2, we briefly discuss the

relationship between energy use and spatial structure in order to concep-

tualize our CEP index. Section 3 puts the home-to-work commute in the

context of all personal travel. The CEP index is then formally developed

in Section 4. In Section 5, we introduce our data and geographical setting

and discuss some spatial differentiations. For this purpose, we map the

average energy consumption for home-to-work travel in Flanders and

Brussels. Based on obvious regional differences, a number of hypothetical

relationships with the underlying spatial and economic structures can be

put forward. Finally, in Section 6 and 7, we summarize our main findings.

2.2 Energy use and urban spatial structure

The main thesis of Newman and Kenworthy (1989, 1999) is the existence

of an inverse relationship between urban density and energy consumption

for transport. Their research was based on data from 32 world cities. In a

Introducing a commute-energy performance index

65

critical reaction to Newman and Kenworthy’s (1999) conclusions, based

on a new analysis of the same data, Mindali et al. (2004) argue that the

assumed general correlation between density and energy consumption for

transport is in fact only valid for certain aspects of the urban structure,

i.e. in the central business district. Banister (1992) and Banister and

Banister (1995) applied a similar methodology as Newman and Kenwor-

thy (1989) on British cities, using data from the National Travel Survey

(1985-1986) and the 1981 census. For London, the city with the highest

overall density, the analysis does not support Newman and Kenworthy’s

thesis: energy consumption per capita is slightly higher in the capital

than the average in the other surveyed cities (> 25.000 inhabitants).

Dodson and Sipe (2008) on the other hand introduced the concept of an

“oil vulnerability index” as a quantification of the robustness of a spatial

entity to rising oil prices, and also take social factors (such as income)

into account. They found that those parts of the outer urban fringe where

no public rail transport is available, are the most vulnerable.

It is also our aim to develop a spatial sustainability indicator that

enables the mapping of regional and urban differences in energy consump-

tion for transport. There are two important arguments that can be put

forward to develop such an index. First, there is the growing importance

of the energy factor in the public debate. The sharp fluctuations in oil

prices, the debate about peak oil and the efforts made to reduce emissions

of greenhouse gases play a role in this discussion (Witze, 2007). Second,

the relationship between spatial structure and travel is a vexed issue.

Travel patterns are highly heterogeneous, and vary with the morphology

and the use of the space. Although concepts such as high density and

diversity are classically considered characteristics of a spatial structure

with a high potential for sustainable trips (Cervero and Kockelman,

1997), it is in fact very difficult to isolate spatial parameters and to

demonstrate causal relationships (Van de Coevering and Schwanen, 2006;

Van Acker et al., 2007; Hammadou et al., 2008). Existing literature

envisages almost always clearly demarcated urban areas. Newman and

Kenworthy (1989) for example, did not consider external flows, while in

some of the cities they studied a significant proportion of the jobs are

filled by commuters. Cervero and Kockelman (1997) and Schwanen and

Mokhtarian (2005) studied the San Francisco Bay Area, which is morpho-

logically much more homogeneous and extensive than, for example, an

average European urban area. Little is known however about the rela-

tionship between the suburbanised historic-polycentric spatial structure

Chapter 2

66

which characterises densely populated European regions such as Flanders,

and the travel patterns of its users. Moreover, travel behaviour is to a

large extent determinant for the energy consumption, distances and used

transport mode being the main parameter. In addition, the rate of car

ownership and - for car owners - the chosen type of car play their part

(Keppens, 2006). Hence, it is worthwhile to investigate the spatial

differentiation in energy consumption for transport in spatially highly

heterogeneous areas. The indicator to be developed should be able to

make explicit the relationship with spatial qualities, such as density,

characteristics of proximity or remoteness, or the presence or absence of

major transport infrastructure at different geographical scales.

2.3 Limitations of studying the

home-to-work commute

Accurate data on the home-to-work commute is more often available than

data on other trips. This is probably the main reason why many studies,

such as those of Dodson and Sipe (2008), focus on home-to-work travel to

quantify the sustainability of travel patterns. However, in studies focusing

on a small enclosed area, it is easier to incorporate different kinds of

trips, as was done by Saunders et al. (2008).

This paper too is based on home-to-work commuting data. This

commute is not representative of all trips, but does affect significantly

non-work related trips. From the theory of the constant travel time

budget (Schafer, 2000), we can say that commuters who spend a lot of

time travelling to work will spend less time on other trips. This means

that they will make more efficient chained trips and that they will look

for destinations closer to home, but also that they will choose more

frequently for fast means of transport (i.e. the car). Moreover, the home-

to-work commute is more inert than other trips are, which can be

illustrated on the basis of price elasticities (Mayeres, 2000). Given this

rigidity, changes in preconditions, such as fluctuations in fuel prices, will

be more problematic for commuting patterns than for non-work trips.

Furthermore, commuting trips cover more often large distances (Zwerts

and Nuyts, 2004), and thus contribute significantly to the negative effects

of traffic. The last two arguments indicate that the study of the home-to-

work commute remains very important.


67

It is important to understand that the rigidity of the commute, both

with respect to distance covered and with respect to modal choice, is not

only a spatial issue. The attitude of the commuter towards the destina-

tion and itinerary, and in particular its habits, determine this rigidity to

a large extent. Consequently, we should consider the travel pattern as a

result of the interaction between space, motivation and habit (Gardner,

2009).

2.4 Commute-energy performance

(CEP) index

In order to exemplify the relationship between the spatial configuration of

an urban region and energy use we develop a commute-energy perform-

ance (CEP) index. This index is obtained by dividing the total amount of

energy consumption for home-to-work travel per census block (i.e. the

smallest geographical research unit) by the working population (active

workforce) that lives in the census block. More formally

s

isiis

sN

cED

CEP∑ ⋅⋅

=

,

(2.1)

In which CEPs is the energy performance per member of the active

workforce for home-to-work travel from the considered (statistical) census

block s; Ds is the total distance travelled (one way) for home-to-work

travel from the considered census block s; iE is the mean energy con-

sumption per passenger for the considered mode i; ci,s is the correction

factor for the considered mode i, within the census block s; Ns is the

number of members of the active workforce in the considered census

block s.

In order to take into account the differences in energy efficiency be-

tween the different transport modes used, the home-to-work travel trips

are split up into motorized (fuel consuming) trips (car, public transport)

and non-motorized trips (on foot, bicycle). For public transport there are

significant differences in energy efficiency between bus, tram, metro, and

train. Hence, we formulate the mean energy consumption per passenger in

relation to the type of public transport used

∑ ⋅= iibtm EKE (2.2)

In which =iE mean energy consumption per passenger for the considered

mode i (bus, tram, metro); Ki is the share of the considered mode i (bus,

Chapter 2

68

tram, metro) in the total number of kilometres made by these three

modes.

For the train mode we can further distinguish between electric and

diesel trains (where α denotes the fraction of electric/diesel trains),

resulting in

trdtretr EEE ⋅−+⋅= )1( αα (2.3)

For car use a comparable subdivision in view of the type of combustion

fuel used (petrol, diesel, LPG) should be made as well. However, our

commuting data (see further) does not allow to distinguish between

different types of car trips, so although further subdividing iE is useful,

it will not add to the analysis. For all non-motorized trips iE is of course

equal to zero.

To keep the relationship between the mode and the distance trav-

elled, for each mode a correction factor is derived from the average trip

length

∑ ⋅

⋅

=

iisi

msm

smDS

DSc

,

,

, (2.4)

In which cm,s is the correction factor for the considered mode m, within

the census block s; Sm,s is the share of the considered mode m as main

transport mode in the total number of home-to-work trips from the

considered census block s; mD is the mean distance of a home-to-work

trip with the considered mode m; i is each of the considered modes.

Finally the resulting number of person kilometres per mode is multi-

plied with a standardized value for the energy consumption per mode.

2.5 Geographical setting and data analysis

Our developed CEP index is tested for Flanders and the Brussels Capital

Region. Fig. 2.1 is a schematic representation of the spatial structure of

Flanders, as defined in the Spatial Structure Plan for Flanders (Ruim-

telijk Structuurplan Vlaanderen (RSV), 1997/2004). The RSV is the

overarching spatial policy plan for the Flanders region. The RSV selects

three “metropolitan areas”, with more than 300,000 inhabitants, being

Ghent, Antwerp and the Flemish urban area around Brussels. Because of

the consistency of the research, we also integrate the Brussels Capital

Region.


69

Furthermore the RSV selects ten “regional urban areas” with a mag-

nitude which is situated between 50,000 and 150,000 inhabitants, as well

as five “urban networks”, an “economic network” along the Albert canal

and five “gates” (ports and airports). The main urban network is that of

the Flemish Diamond, which is bounded by the three metropolitan areas

and the regional urban area of Leuven, and is the economic core of

Flanders. The other areas are considered as countryside, still including a

number of small urban areas and economic nodes with rather limited

development perspectives.

Fig. 2.1. Schematic representation of the spatial vision on Flanders.

Source: RSV (1997/2004)

Chapter 2

70

Table 2.1. Average trip length for home-work travel by transport mode,

for the purpose of determining the correction factor for the trip length

(based on OVG 2001)

bicycle =bcD 4.07 km

moped, motorbike =mbD 10.85 km

transportation organized by the employer or school =emD 18.86 km

car driver =cdD 20.33 km

car passenger =cpD 17.43 km

train =trD 48.48 km

bus, tram, metro (underground) =btmD 18.86 km

on foot only =ofD 2.15 km

Table 2.2. Itemization of passenger kilometre share for urban and

regional public transport by mode, Flanders and Brussels

tram and trolley bus =tK 11.3%

metro (underground) =meK 2.0%

bus =bK 86.7%

Table 2.3. Default values for energy consumption per person kilometre

(source: De Vlieger et al. (2006))

tram =tE 0.06 kWh/pkm

metro (underground) =meE 0.08 kWh/pkm

city bus =cbE 0.25 kWh/pkm

coach =coE 0.32 kWh/pkm

aggregated: urban and regional public

transport

=btmE 0.26 kWh/pkm

electric train =treE 0.13 kWh/pkm

diesel train =trdE 0.18 kWh/pkm

aggregated: train =trE 0.14 kWh/pkm

city car =ccE 0.43 kWh/pkm

family car =cfE 0.53 kWh/pkm

aggregated: car =cE 0.48 kWh/pkm

bicycle, on foot =ofbc EE , 0.00 kWh/pkm


71

Also within the infrastructure network, especially the road network, a

selection and classification of roads is made. The metropolitan and

regional urban areas are all opened up by the main road network, and

served by at least one train station.

The data used to calculate the CEP index for Flanders and Brussels

are drawn from various sources. The so-called General Socio-Economic

Survey 2001 (SEE 2001, see: Verhetsel et al., 2007) is a comprehensive

survey of the Belgian population (excluding children younger than six

years old), which has its origin in the 10-yearly census. The response to

the survey is 95%. The questionnaire of SEE 2001 contains a number of

mobility related questions. The distance between home and work or

school is assessed. In addition, questions are raised about the means of

transport used for home-to-work or home-to-school trips, the number of

daily trips whether or not combined, car ownership and the perception of

the supply of public transport around.

Data on the average trip length per mode is based on the Travel Be-

haviour Research project in Flanders (OVG 2001) (Zwerts and Nuyts,

2004) (Table 2.1).

The standardized values for the energy consumption per mode are

taken from De Vlieger et al. (2006), and are based on the French research

by Enerdata (2004) (Table 2.3). All energy values are converted to

kilowatt hour per person kilometre (kWh/pkm). In each case the final

energy consumption by the vehicle is considered, meaning that for electric

vehicles the losses that occur in the production and transmission of the

electricity, which depend greatly on how it is generated, are not taken

into account. For the category “car as passenger” the same value is

applied as for the category “car driver”, since the default value is set per

person and already takes into account the average occupancy rate of the

vehicle. More specific variations in energy consumption, such as the

distinction between diesel and gasoline cars, or regional differences in the

composition of the fleet of personal cars or the ridership of buses and

trains, are not taken into account.

Data on the use of local public transport was split up on the basis of

the passenger statistics of the urban and regional public transport

companies in Flanders (De Lijn) and Brussels (STIB) for the year 2006

(De Lijn, 2006; STIB, 2006) (Table 2.2). The mode train is itemized into

78% electric and 22% diesel, based on De Vlieger et al. (2006).

Chapter 2

72

2.6 Results

2.6.1 Spatial distribution of the CEP index:

a first glance

We calculate the CEP index for home-to-work travel, based on the

departure zones. Because of the limitations of the available data, the

resulting map should only be interpreted as an approximation, which

aims to uncover the gradients with regard to energy consumption for

home-to-work travel in Flanders and Brussels. In a next step the relation-

ship with a number of spatial characteristics will be unveiled (see

further).

In order to interpret the CEP index, the relevant spatial characteris-

tics are mapped as well, based on data from SEE 2001. The main road

network and the railways are added to all maps. Fig. 2.3 shows the

average home-to-work distance per member of the active work force. For

this, data on home-to-work distances are aggregated for each neighbour-

hood (census block) and divided by the working population. It is possible

to draw up a similar map for the home-to-school distances. Figs 2.4-2.6

present the frequency with which the modes, that are known to be

energy-efficient, are used as main transport mode for home-to-work trips.

For the purpose of ease of interpretation, these maps are simplified in the

sense that, for example, any pre- and post-transportation is not consid-

ered while evaluating the use of the various modes. Moreover, not all

modes are included: pedestrians, car passengers and transport which is

organised by the employer are not incorporated in the set. The purpose of

the maps is thus to give a global overview of the variation of these

parameters on the territory of Flanders and Brussels.

According to the mapped CEP index, energy consumption for home-

to-work travel seems to be particularly high in those regions which in

spatial planning terminology are defined as the countryside (A1-8) (Fig.

2.2)1. These regions have in common that they possess a relatively rural

character, compared to the labour markets where they are focused on.

The regions A1 and A3, for example, are influenced by the labour

markets in the metropolitan and urban areas of Brussels, Ghent and

Leuven, even if those are relatively distant (Van Nuffel, 2007). In addi-

1 The applied codes refer to the respective zones on Figs 2.2-2.5.


73

tion, commuters in these rural regions have on average higher incomes

which allows them to live outside the city centres in the relatively quiet

and green environments, being less sensitive to the financial impact of the

large daily commuting distances.

Apart from that, some corridors along the motorways are strongly

reflected in the map. The locations B1-4 catch the eye. It is clear that in

these cases the increased accessibility by the presence of a motorway has

contributed to enlarge commuting distances and the increased importance

of the car as a transport mode. The area, in which the energy consump-

tion is pre-eminently low, is the Brussels Capital Region (C1). The

Flemish urban area around Brussels has a more or less comparable

pattern, but still scores worse than the Brussels’ municipalities. This

result concurs with what might be expected, as the Brussels region

represents the largest job market of the country, and also has the highest

population densities. It is therefore consistent with the idea that the

match in the labour market supply and demand is achieved within short

distances. Moreover, the metropolitan spatial structure is responsible for

the relatively large influence of other parameters on the energy consump-

tion, such as modal split and vehicle ownership. This will be discussed

below.

Similar patterns occur in the two other metropolitan areas of Ant-

werp (C2) and Ghent (C3), in which the effect of the metropolitan

structure of Antwerp clearly outreaches the case of Ghent. In all regional

urban areas, we also find lower energy consumption than the average.

But also outside the metropolitan and regional urban areas, there are a

number of regions that come out on the right side by their significantly

lower energy consumption. The most contiguous region we find at D1-2.

This region is characterized by a strong sprawl of less specialized labour,

and a strong spatial interweaving of the labour market with the residen-

tial structure. The importance of location-bound industries, in particular

in the agricultural sector, probably plays a part in this. So, the distance

between home and workplace remains relatively confined.

Furthermore, also the corridor Brussels-Mechelen-Antwerp (D3), an

important transport artery, scores remarkably well on the map, as well as

a part of the economic network of the Albert canal (D4). These economi-

cally strong areas have high concentrations of employment in a - on the

scale of Flanders - relatively good mix with the residential structure. We

see the same phenomenon, albeit on a smaller scale, arising in D5-D7.

Chapter 2

74

The rural areas D8-D11 show rather low figures. Apparently, the rela-

tively poor accessibility of these regions has caused only a few long-

distance commuters to settle here. In addition, the low population and

building density in these regions makes that a rather large share of the

population is still working in the local agribusiness.

Fig. 2.2. Daily energy consumption per capita for home-work travel

(kWh). Source: SEE 2001

2.6.2 Spatial patterns and relation to

home-to-work distance

The CEP map (Fig. 2.2) shows a remarkable resemblance with the map

that visualizes the home-to-work distances (Fig. 2.3). The Pearson’s

correlation coefficient between the two sets of indicators is close to 0.95.

We can therefore conclude that the energy consumption for home-to-work

travel is first and foremost determined by the distance between home and

workplace. Contrary to what is generally assumed, it appears that the

used transport mode plays only a very limited role. This can partly be

explained by the fact that the average distance covered by train commut-

ers (on average 48 km in 2000) is much larger than the average journey

that is made by car (on average 20 km). Secondly, the bicycle is only an

alternative for short trips, which makes this mode only marginally

represented in the total number of kilometres.


75

Fig. 2.3. Mean distance for home-work travel (km). Source: SEE 2001

2.6.3 Relation to the use of the bicycle

In Belgium, particularly in Flanders, the bicycle has always been an

important means of transport in short distance commuting. Belgium is

ranked third in Europe (after The Netherlands and Denmark (1997)) with

regard to the use of the bicycle (Witlox and Tindemans, 2004). This is

chiefly explained by the flat topography, as well as the strong railway-

bounded pre-war spatial development (in which the bicycle appeared to

be the perfect pre- and post-transportation means). The relief shows an

unmistakable determining factor with respect to the use of the bicycle in

home-to-work travel. The cycling map (Fig. 2.4) has therefore bright

spots in the hilly regions A1, A2, A3, D10 and E1. Also, in the Brussels

Capital Region and the Flemish urban area around Brussels, cycling to

work hardly occurs. The very dense traffic during peak-hours is an

important explanatory variable.

In all regional urban areas, the bike is still very important in the

home-to-work travel. That is also the case, albeit somewhat less explicit,

in the metropolitan areas of Antwerp and Ghent. Furthermore those

regions where a significant mix of living and working occurs, and the

distances are accordingly short, catch the eye (D1-7). In addition, there

are some more isolated areas that also score well in terms of bicycle use,

Chapter 2

76

such as the urban network of the coast (E2) and the peripheral regions

E3 and E4.

When looking at the cycling map (Fig. 2.4) and the distance map

(Fig. 2.3) at one glance, we see that the largest share of cyclists occurs at

first in those regions where the home-to-work distances are the shortest,

with the Brussels Capital Region and the Flemish urban area around

Brussels as a major exception to this rule. According to the OVG 2001,

the average length of a cycle trip in home-to-work travel amounts to 4

km. This figure is based on a survey, and its reliability as a reported

distance is relatively small given the small distance range (Witlox, 2007).

Fig. 2.4. Share of bicycle in home-work trips. Source: SEE 2001

Clearly, the bicycle plays a major role in those regions where the com-

muting distances are of this small magnitude. The influence of the bicycle

on the total energy consumption for home-to-work travel is very limited.

Even in regions with a relatively high share, the bike use share in com-

muting is less than 20% of the total number of trips. As trips with other

modes cover much larger distances per trip (see Table 2.1), the gain in

terms of energy consumption made by cyclists is of little significance. The

low energy consumption in areas with a high proportion of cyclists is

largely on the account of the small absolute distances, regardless of the

transport mode used. But the positive impact of the short distances in


77

these areas is reinforced by the larger share of cyclists that substitutes car

use on short distances, in comparison with other regions.

2.6.4 Relation to the use of the train

Figs 2.5 and 2.6 visualize the level of use of public transport. Fig. 2.5

deals with the share of train passengers, while Fig. 2.6 demonstrates the

share of tram, bus and metro travellers. High concentrations of train

commuters can be found along the railway axes which provide since a

long time a fast connection with the capital, particularly to the west and

south-west of Brussels (F1-F7). Also some railway axes that are focused

on Antwerp, catch the eye (F7-8). Apart from this, some local concentra-

tions of train commuters (F9-12) can be distinguished. The inhabitants of

the Brussels Capital Region and the Flemish urban area around Brussels

hardly use the train for home-to-work travel. This applies in general to

the regions which are relatively remote from Brussels too. In the outlying

regions of Flanders, and particularly the northeast, commuters hardly use

the train.

Interestingly, a significant portion of the concentrations of train

commuters is located in those areas where commuting distances are the

largest and corresponding energy consumption is the highest. This is

particularly the case in the area southwest of Brussels (G1), and the

infrastructure axes towards Ghent (G2) and Leuven (G3). This can be

explained by the finding from the OVG 2001 that the average length of a

home-to-work trip by train amounts to 48 km, which is very high (Table

2.1). The train is therefore mainly an alternative to long car trips.

Consequently, it stands to reason that the train is popular in regions with

large commuting distances.

However, this does not apply to all areas where the train is well posi-

tioned. In particular, a number of regional urban areas (G4-7) and Ghent

show relatively low energy consumption combined with a rather high

proportion of rail commuters. This applies to a certain extent also to the

railway axes that focus on Antwerp (G8-G11).

It is however not possible to make unequivocal statements about the

impact of the use of the train on energy consumption. First, it is true

that the average rail passenger covers larger distances than the average

car driver (48 km compared with 20 km). When we multiply the average

energy consumption per kilometre (0.14 kWh/pkm respectively 0.48

kWh/pkm, see Table 2.3) by the average number of kilometres per mode,

Chapter 2

78

it appears that in home-to-work travel the energy consumption of a train

trip per person is only 30% less than the energy consumption of a trip by

car. This difference is less impressive than the expectations raised by the

sustainable image of the train. Fast rail transport supply also induces

long-distance commuting, which means no gain in terms of energy

consumption. Finally, the majority of train journeys entail travel to and

from the station, which often means an additional trip by car.

On the other hand, the train substitutes long car trips, at least where

rail transport supply is present. Furthermore, the train is doing this -

calculated per kilometre - in a more energy-efficient way. Nevertheless,

the share of train commuters, even in the concentration areas, is limited

to a maximum of approximately 20%. In general it may be said that

there occurs only a positive impact on energy consumption by use of the

train in those regions where the average home-to-work distances are

already large. But the effect is still too small to be visible on the energy

performance map (Fig. 2.2).

Fig. 2.5. Share of train in home-work trips. Source: SEE 2001

2.6.5 Relation to the use of urban and regional

public transport

Fig. 2.6 shows that tram, metro and bus as a main transport mode in

home-to-work trips only play a significant part in the three metropolitan


79

areas. In Ghent, the smallest of these metropolitan areas, this mode

scores only in a few neighbourhoods higher than 10%. In Antwerp the

influence of the urban and regional transport reaches beyond that, with

values that are often well above 10%. In Brussels, where urban transport

is built on the backbone of the metro, the situation is completely differ-

ent. Many neighbourhoods show a share of around 30%.

Given the preponderance of Brussels in the use of this mode, we can-

not draw any conclusions about the average trip distance. This is because

the values in Table 2.1 are derived from a sample survey from Flanders.

The impact on energy consumption is difficult to predict. There is a

significant discrepancy between the energy consumption of electric rail

transport and that of (diesel) buses. Only in Brussels an extensive metro

and tram network exists.

On the other hand, those parts of Brussels and Antwerp that show

lower energy consumption per capita have also a high to very high

proportion of public transport users. It is clear that the high density of

inhabitants and jobs in those cities combines the positive effects of the

proximity of functions with an energy-efficient and well exploited urban

transport system. The relative good energy efficiency and the high

patronage of the metro, supplemented with the tram, account for this.

Fig. 2.6. Share of bus, tram, metro in home-work trips.

Source: SEE 2001

Chapter 2

80

In the rest of Flanders the urban and regional transport, based on diesel

buses, hardly plays a part in the energy performance of home-to-work

commuting. The supply, which is limited in comparison with the metro-

politan areas, and the non competitive speed of this kind of public

transport contributes to the limited success of the urban and regional

transport in the market segment of home-to-work travel. Although we

will not discuss this more profoundly, the relatively low ridership outside

the urban areas also works to the detriment of the energy performance of

public transport.

2.6.6 Relation to car ownership

Car ownership is expressed in number of available cars per household. In

the historic centres of the study area, in particular those that are part of

the metropolitan areas, car ownership is low. Among other things, spatial

factors are at the basis of this. The high density, the significant mix of

functions and good supply of public transportation makes it relatively

easy to live without a car in the city centres. Also, a number of social

elements play a part, since it is precisely in these areas that the family

sizes are small and household revenues are rather low. However, it is

difficult to isolate environmental and social factors. The environment and

the rent and real estate prices in the city are often not adapted to the

lifestyle of families with children. In addition, those families need more

often combined trips, for which the car is usually the appropriate means

of transport.

In spatial terms the zones with a remarkably high car ownership have

following characteristics: they are suburban areas of major employment

centres, which also have a quick access to the main road network. In

social terms those are areas with high incomes. All main concentrations

are located in the suburban belts around the three metropolitan and some

regional urban areas.

Despite those regions that catch the eye, the regional differentiation is

in fact very limited. The resulting map is fairly homogeneous. The

differences are usually too small to be able to determine unique relation-

ships between car ownership and energy consumption for home-to-work

travel. However, this is possible in the above mentioned areas that attract

the attention. Where the car ownership is significantly lower (the urban

centres), also energy consumption is lower, and vice versa in zones where

the car ownership is significantly higher (the mentioned suburban areas).


81

In the rest of Flanders the relationship is much less pronounced. In the

urban network of cities situated along the coast, with its atypically aged

demographic composition, there even seems to be an inverse relationship

between car ownership and energy consumption.

2.7 Relation to spatial-morphological

characteristics

The CEP index is also useful in interpreting the relation to a range of

spatial-morphological characteristics. As an exploration, we outline the

relationship between energy consumption and two notable spatial charac-

teristics: residential density and proximity to the main road and railway

network.

2.7.1 Relation to the population density

Density is perhaps the most widely used measure, which is also easily

quantifiable. In their research about the relationship between urban

density and fuel efficiency, Newman and Kenworthy (1989) included

Brussels. The study area was limited to what is today called the Brussels

Capital Region. External commuting flows were not considered, while we

know that most jobs in Brussels are taken up by people who commute.

Let us now repeat the Newman and Kenworthy (1989) exercise for

Flanders and Brussels based on our developed CEP. Fig. 2.7 plots the

energy consumption for home-to-work travel on the basis of residence in

relation to the population density of the block. The image is of course

only a facet. Further research will also provide relations with the density

of jobs on employment locations, and the focus could be expanded to all

types of travel (school, shopping, recreation, visiting, et cetera). Despite

the limited coverage of the research field, we can see a number of inter-

esting similarities and differences with what Newman and Kenworthy

(1989) have found.

Chapter 2

82

Fig. 2.7. Plot of residential density and energy consumption for home-

work commuting, per census block

In general, it appears that the basic argument is still valid: the energy

performance improves with increasing density. However, the large

variation at the root of the graph catches the eye. Even though the trend

line for energy consumption is rising with decreasing density, for a part of

Flanders, in fact, an inverse relationship applies. The part of the dotted

cloud that is situated near the root of the graph (A) represents the rural

areas (i.e. countryside). These areas are typified by sparsely populated

regions with a non-intensive home-to-work travel pattern. In these areas,

most people tend to work at or near their home: farmers, and local

economy workers.


83

The part of the dotted cloud that is situated above the left part of

the trend line (B) represents the peripheral suburban sprawl. These areas

have developed as a result of counter-urbanised delocalization, and are

mostly inhabited by people who exhibit a strong urban lifestyle. Their

professional lives are also mainly concentrated in the urban areas; hence,

they represent an important travel demand.

The areas which are situated around the trend line roughly represent

the urban areas. The suburban blocks that are adjacent to the traditional

city centres are represented around the left part of the trend line (C),

while the traditional city centres themselves can be found on the right

side of the graph (D). In contrast to the analysis of Newman and Ken-

worthy (1989), which considered only urban areas, we can thus deduce a

spatial typology from regional variations in the efficiency of the transport

system.

2.7.2 Relation to the access to the main

road network

In the analysis of the spatial variation in energy performance of home-to-

work travel we already referred to a possible link with the location of the

motorways. In order to gain further insight into the potential effects of

the present infrastructure, the main road network and the rail network

was already presented in Fig. 2.2. Interestingly, certain main roads stand

out on the map, while others do not. Increased energy consumption is

clearly visible in the regions I1-10. All of these cases are rural regions that

became very well connected by the construction of a main road, but do

not have any large or diversified employment. The new access to the

main road network has encouraged moving to the countryside, has

generated traffic, has increased commuting distances and has put the

focus more exclusive on car travel. Eventually, all this has led to in-

creased energy consumption. For the other main roads, which are

traversing a structurally different socio-economic landscape, this finding

does not apply. Overall, it seems that two types of roads can be identified

using our framework. The first type of main road crosses an area with a

fairly diverse economy, where the residential structure is to a large extent

mixed with the job market. This is particularly true between Brussels and

Antwerp (I11, I12), but also in a number of other places (I13-15). More-

over, all main road segments crossing the metropolitan areas, harbour

and airport areas are of this type. The second type of main road runs

Chapter 2

84

through a relatively remote area, without providing a smooth connection

with a major employment centre in the urban network of the Flemish

Diamond. Examples of these corridors can be found in I16-20.

2.7.3 Relation to the access to the rail network

The link with the rail network is also very visible. Stations are located

next to the historic centres, which have the highest residential density

and often also the largest functional mix. Of the lower energy consump-

tion for home-to-work travel that we often observe in the vicinity of the

railway stations, only a tiny share can be attributed to the train com-

muters. This is explained in that the average train trip in home-to-work

travel still represents an energy quantity of 70% of that of an average car

commuting trip.

But the link is not unambiguous. A number of important rail lines are

flanked by an - in terms of traffic volume - much more important motor-

way, which eventually leaves its heavy mark on local commuting

behaviour. This is illustrated in the counter-urbanized area around J1,

where good train and motorway connections are present. The energy

consumption is thus high.

In the region southwest of Brussels (A1), a different phenomenon

emerges. Despite good rail links with the Flemish Diamond and the

absence of a main road, we find very high energy consumption levels. The

census blocks in which the main stations are located, score a lot better in

terms of energy than the surrounding areas. The blocks in the immediate

vicinity of Ghent’s main railway station (J2) have a very large share of

train commuters as well as relatively high energy consumption. Note that

this station has a very good train connection to Brussels. Areas which

manage to combine a large share of train commuters with low energy

consumption in home-to-work travel are rare: J3-8 fall into this category.

2.8 Conclusions

We have argued that the energy performance of the transport system is

an important approximate indicator for the sustainability of a spatial

structure. This is certainly true when advocating a so-called low carbon

economy is put increasingly higher on the political agenda. Obviously the

link with the spatial or urban (re)development of cities should be made as

well. Having a better understanding of the mechanisms that cause the


85

major observed regional variations in energy consumption will lead to

better land use planning in practice.

The issue of proximity in planning remains very important. In home-

to-work travel, the distance between home and workplace is to a very

large extent determinant for the energy performance of the commuting

system. Contrary to the conventional belief, the mode used is of much

less importance. In this respect we notice a discrepancy with the current

mobility policy of the Flemish government, which is very much focused

on the reduction of the share of car drivers, but much less on a reduction

of the number of kilometres, despite an increase by 10% of the average

commuting distance between 1991 and 2001 (Verhetsel et al., 2007).

Hence, the opportunity for someone to find a suitable job nearby his

or her living environment, or the ease with which someone can move in

the vicinity of his or her work will increasingly determine the robustness

of a spatial economic system in a climate of rising oil prices. It appears

that travel behaviour remains largely determined by the rigidity of the

housing stock, which makes short term policy intervention not easy. This

applies too - mutatis mutandis - for non-work related trips, although

these show higher elasticities. Therefore, we argue that it is important to

develop a more profound insight into the regional variations in the energy

performance of the whole transport system. So we will get a better

understanding of the processes that led to the current situation, and we

will be able to assess the policies that played a part or can still play a

part in this. In this respect the development and implementation of a

commute-energy performance index seems a useful indicator to assess

both transport and spatial planning policies with respect to inducing

sustainable development.

References

Banister, D. (1992) “Energy use, transport and settlement patterns”. In:

M. Breheny (Eds.) Sustainable Development and Urban Form. Lon-

don: Pion, pp. 160-181.

Banister, D. and C. Banister (1995) “Energy consumption in transport in

Great Britain: Macro level estimates.” Transportation Research Part

A. 29(1), pp. 21-32.

Chapter 2

86

Cervero, R. and K. Kockelman (1997) “Travel demand and the 3Ds:

Density, diversity, and design.” Transportation Research Part D.

2(3), pp. 199-219.

De Lijn (2006) Jaarverslag 2006. Mechelen: De Lijn - Flemish Public

Transport Agency.

De Vlieger, I., E. Cornelis, L. Int Panis, L. Schrooten, L. Govaerts, L.

Pelkmans, S. Logghe, K. Vanherle, F. Vanhove, G. De Ceuster, C.

Macharis, T. Festraets, E. Pekin, L. Turcksin, K. Van Bladel, M. de

Jong, J. Van Mierlo, J. Matheys, J.-M. Timmermans, C. De Geest

and E. van Walsum (2006) MIRA (2006) Milieurapport Vlaanderen,

Achtergronddocument 2006, Transport. Aalst: Flemish Environment

Agency.

Dodson, J. and N. Sipe (2008) “Shocking the suburbs: Urban location,


Studies. 23(3), pp. 377-401.

Enerdata S.A. (2004) Efficacité Energétique des Modes de Transport:

Rapport Final. Grenoble-Gières: Enerdata S.A.

Gardner, B. (2009) “Modelling motivation and habit in stable travel

mode contexts.” Transportation Research Part F. 12(1), pp. 68-76.

Hammadou, H., I. Thomas, A. Verhetsel and F. Witlox (2008) “How to

incorporate the spatial dimension in destination choice models: The

case of Antwerp.” Transportation Planning and Technology. 31(2),

pp. 153-181.

Keppens, M. (2006) Modelling Private Road Transport of Flemish

Households. The Mobility and Welfare Effects of Fuel Taxes versus

Tradable Fuel Rights. PhD Thesis. Hasselt: Universiteit Hasselt.

Mayeres, I. (2000) “The efficiency effects of transport policies in the

presence of externalities and distortionary taxes.” Journal of Trans-

port Economics and Policy. 34(2), pp. 233-260.


consumption: a new look at old statistics.” Transportation Research

Part A. 38(2), pp. 143–162.








87

Saunders, M. J., T. Kuhnimhof, B. Chlond and A. N. R. da Silva (2008)

“Incorporating transport energy into urban planning.” Transportation

Research Part A. 42(6), pp. 874-882.



Schwanen, T. and P. L. Mokhtarian (2005) “What affects commute mode

choice: Neighborhood physical structure or preferences toward

neighborhoods?” Journal of Transport Geography. 13(1), pp. 83-99.

STIB (2006) Activiteitenverslag 2006. Brussels: Société des Transports

Intercommunaux de Bruxelles / Maatschappij voor het Intercommu-

naal Vervoer te Brussel.

Van Acker, V., F. Witlox and B. van Wee (2007) “The effects of the land

use system on travel behavior: A structural equation modeling ap-

proach.” Transportation Planning and Technology. 30(4), pp. 331-

353.

Van de Coevering, P. and T. Schwanen (2006) “Re-evaluating the impact

of urban form on travel patterns in Europe and North-America.”

Transport Policy. 13(3), pp. 229-239.

Van Nuffel, N. (2007) “Determination of the number of significant flows

in origin-destination specific analysis: The case of commuting in

Flanders.” Regional Studies. 41(4), pp. 509-524.



gium.

Witlox, F. and H. Tindemans (2004) “Evaluating bicycle-car transport

mode competitiveness in an urban environment. An activity-based

approach.” Journal of World Transport Policy & Practice. 10(4), pp.

32-42.

Witlox, F. (2007) “Evaluating the reliability of reported distance data in

urban travel behaviour analysis.” Journal of Transport Geography.

15(3), pp. 172-183.

Witze, A. (2007) “That’s oil, folks….” Nature. 445(4), pp. 14-17.

Zwerts, E. and E. Nuyts (2004) Onderzoek Verplaatsingsgedrag Vlaan-

deren 2000-2001. Brussels-Diepenbeek: Ministry of the Flemish

Community.

Chapter 2

88

89

Chapter 3:

Minimum commuting

distance as a spatial

characteristic in a

non-monocentric

urban system

This paper will be published as Boussauw, K., T. Neutens and F. Witlox

(2011) “Minimum commuting distance as a spatial characteristic in a

non-monocentric urban system: The case of Flanders.” Papers in Re-

gional Science 90(1). Copyright © Regional Science Association

International - Blackwell Publishing. All rights reserved.

Abstract

This paper focuses on regional variations in commuting trip lengths by

calculating minimum (required) commuting distances, along with excess

commuting rates. The study contributes to the excess commuting re-

search framework from a regional perspective, both by stressing the

specific characteristics of urban networks with overlapping commute

areas, and by putting forward an alternative method for calculating

spatially disaggregated values. A case study in the north of Belgium

shows that large variations in minimum commuting distances occur. This

in turn identifies to a large extent opportunities for shrinking commuting

distances by influences such as rising fuel prices, compact urban planning,

extreme congestion or dissuasive traffic policies.

Keywords: excess commuting; spatial proximity; sustainable spatial

development; Flanders

Chapter 3

90

3.1 Introduction

The concept of excess commuting or wasteful commuting was initially

introduced by Hamilton (1982), and has become a well-established line of

inquiry in transportation research in the last decades (Ma and Banister,

2006). Hamilton (1982, p. 1040) defined wasteful commuting as the

difference between the actual commuting distances and the theoretical

minimum (required) commuting distances, which are suggested by the

spatial structure of a considered city. Hamilton’s interest for minimized

commuting distances probably stemmed from the consecutive oil crises of

1973 and 1979-1980, when the availability, and in particular, the afforda-

bility of fossil oil products was at stake. Daily trips over large distances

were suddenly considered problematic, because of their particularly high

energy consumption.

There is a broad consensus that spatial structure as a combination of

morphological elements and activities (e.g. size, shape and functional

mix) is a key determinant in explaining travel pattern generation (Giuli-

ano and Small, 1993; Van Acker et al., 2007). In many policy documents

on mobility and transport much hope is set on achieving an “adequate

spatial planning” as an effective means to improve the efficiency and

sustainability of mobility. However, while the spatial structure is gener-

ally recognized as a prerequisite for trip generation, observed travel

behaviour and trip distances in particular are additionally induced by

other factors.

We know that the observed average commuting distances in Euro-

pean and North American metropolitan areas increase year after year

(Banister et al., 1997; Aguilera, 2005). There is no doubt that the in-

creased prosperity implicitly or explicitly plays a role in the growth of the

travelled distances. However, it appears that the possibility to travel over

increasingly larger distances is systematically materialized in the form of

the physical separation of functions. Given that the public debate ac-

knowledges that an important part of the traffic problems is related to

land use policies, it is important to estimate what share of the actual

travel is caused by the spatial structure itself and what share is in fact an

extrapolation that originated from other elements, such as the general

prosperity, but also the level of congestion, the quality of roads, or the

price of fuel. Hamilton (1982) explored only home-work commuting, but

the concept may, mutatis mutandis, be extended to all daily travel

categories (Horner and O’Kelly, 2007).

Minimum commuting distance as a spatial characteristic

91

Researchers in this field usually consider excess commuting as a char-

acteristic of a specific city, regarded as a monocentric, polycentric or

dispersed urban system. Building on the exploratory work by Niedzielski

(2006) and Yang and Ferreira (2008) we want to put forward an exten-

sion to this line of reasoning, in which minimum commuting distance and

excess commuting are to be seen as properties of specific spatial entities

(spatially homogeneous areas) within a non-monocentric spatial system.

To this end, the local minimum commuting distance is considered as a

measure for spatial proximity of each relevant area, as embedded in the

studied region. This measure can be used to quantify relations to other

spatial characteristics such as density, spatial diversity or accessibility.

Quantification of traffic volumes is essential to assess the extent of

excess travel. As a proxy for traffic volumes we will use the number of

kilometres travelled per person to or from a considered zone within a pre-

defined time frame. As shown in Boussauw and Witlox (2009), at least at

the regional level (macro scale), distance travelled per person can be

deemed a good approximation of energy consumption, sustainability of

travel behaviour and total traffic volume. Depending on the spatial

characteristics of the study area and the research scale level, other

elements such as modal split, the composition of the fleet and the level of

congestion might be incorporated to avoid running the risk of oversimpli-

fication.

In relation to Niedzielski (2006) and Yang and Ferreira (2008), the

novelty of our approach is twofold. First, we develop a method that

extends the use of linear programming techniques and is more adequate

in studying spatial variations. In respect to the latter we are able to

simulate the case where all commuters start to look simultaneously for a

job closer to their homes. Second, we apply the spatially disaggregated

approach at a regional scale, on an urban network with overlapping

commuter areas. Ultimately, the methodology is applied to the study area

(Flanders and Brussels) and the results are interpreted.

Chapter 3

92

3.2 Spatial variations in excess travel

As a starting point, Hamilton (1982) considered a monocentric urban

model, as developed by Mills (1967) and Muth (1969). Within this model,

there is a balance between accessibility to the central business district

(CBD) and the bid rent, which is materialized in a density gradient of

both jobs and houses. The monocentric model allows predicting the

minimum home-work distances on the basis of the geometric characteris-

tics of a circle. Hamilton (1982) calculated these minimum distances for a

number of American and Japanese cities. He found that these distances

differed substantially from the actual commuting distances drawn from

survey data of the considered cities, questioning the validity of the

monocentric city model. Hamilton (1982) found an average actual

commuting distance of 8.7 miles (13.92 km), corresponding to a minimum

commuting distance of only 1.1 miles (1.76 km). Thus, in this case the

excess rate amounts to 7.9 or, put differently, 87% (= (8.7 - 1.1)/8.7) of

the actual commute is excessive. Hamilton’s (1982) methodology proved

very controversial, and was criticized by White (1988) and Small and

Song (1992), whose results showed large deviations compared to the

values calculated by Hamilton. The common motivation behind these

studies is the investigation of the predictive power of the monocentric

urban model (later extended to polycentric and dispersed urban models;

Song, 1995), with respect to commuting cost minimizing behaviour.

A major dichotomy appearing among studies on excess commuting is

the choice of either travel distance or travel time as a proxy for the travel

cost. While travel time is an intuitively appealing proxy for travel cost,

there are, from environmental policy perspectives, well-grounded reasons

to use distance as a parameter. First, the adverse external effects of travel

are more closely related to travel distance than to travel time. Second,

there is the constant nature of the personal travel time budget, which

means that over time urban travellers spend a constant amount of time

on their daily travel (Schafer, 2000).

Among the studies in which distance is used as a parameter, extreme

variations are recorded across various study areas (Ma and Banister,

2006). Frost et al. (1998) found an excess commuting rate of only 18.9%

for London (including all inward commuting), while the ratio found by

Song (1995) amounts to 81.6% for Los Angeles. Given the lack of uni-


93

formity in the methodologies used, comparing results of different studies

can hardly be justified. It was found that calculated excess commuting

rates decrease as the number of constraints which are explicitly controlled

increases; examples include limitations originating from two-worker

households and the divide between tenants and home-owners (Kim,

1995), or accounting for the expectation of future job locations (Crane,

1996). Furthermore, the modifiable areal unit problem (MAUP), intrinsic

to the practice of spatial analysis, adds to the difficulty of comparing the

results of different studies in the sense that the size and configuration of

the zones used may bias the excess commuting rate (Horner and Murray,

2002).

Ma and Banister (2007) examined the theoretical relationship be-

tween variations in the spatial distribution of population and jobs, and

excess commuting. They showed that the excess commuting rate is a

good proxy for the potential reduction of commuting distances within a

specified study area, but that the minimum commuting distances (and by

extension also maximum commuting distances) are more suitable to use

in comparisons between different cities or points in time. Niedzielski

(2006) and Yang and Ferreira (2008), studied for the first time spatially

disaggregated minimum commuting distances within an urban area.

None of the above-mentioned contributions, however, analysed em-

pirically the spatial variations of the minimum commuting distance and

excess commuting rate on a regional scale. In the monocentric city model,

access to the CBD is the main determinant of the urban structure, while

in a polycentric or dispersed model accessibility to a diverse range of job

and service locations is determinant. For the detection of spatial varia-

tions of excess travel within the suburbanized historically polycentric

spatial structure that characterizes many urbanized regions in Europe,

the application of the monocentric city model to derive minimum com-

muting distances makes little to no sense.

We hypothesize that there exist important regional variations of

minimum commuting distances and excess travel, for which a link with

spatial characteristics (e.g., density, functional mix or proximity to major

transportation infrastructure), can be found. Earlier research shows that

there are important differences between cities (Charron, 2007), even if it

is still unclear how these values evolve through a region with multiple

centres and suburban areas.

Chapter 3

94

3.3 Possible policy implications

In a number of studies, such as Scott et al. (1997), Frost et al. (1998) and

O’Kelly and Niedzielski (2008), the focus shifts from the explanatory

nature of an urban economic model to the possible relevance for spatial

policies, aimed at assessing the potential reduction of undesired external

effects of the traffic, such as emissions or energy consumption. In particu-

lar, Frost et al. (1998) examined the evolution of the minimum

commuting distance and excess commuting rate for several British cities,

and linked this evolution with spatial developments.

For a given area, the excess commuting rate can be considered as an

indicator for the extent to which the particular spatial structure of this

area is able to absorb shrinkage of the total traffic volume, without

incurring severe economic damage. We assume that such a reduction can

be caused by external factors such as rising fuel prices, extreme conges-

tion or dissuasive policy measures (e.g., to limit emissions; Scott et al.,

1997, or congestion). Alternatively, we can also state that shrinkage of

excess travel volumes might become part of a scenario that, in the long

run, will ensue from an expected continued increase of oil prices (peak

oil). In addition, minimum commuting distances are an indicator of the

imbalance between the residential function (considered as the origin of

travel) and other functions, such as jobs, schools, shops and recreational

attraction poles (which may be regarded as destinations). The effect of

encouraging or just discouraging certain new developments can thus be

tested against the expected impact on minimum commuting distances. A

reduction of these theoretical minimum commuting distances will eventu-

ally make an effective reduction of travelled distances under the

aforementioned changing external influences much easier.

A final interesting feature of our approach is the ability to assess the

impact of various autonomous development scenarios on commuter

traffic. In this way, a significant growth or decline in both population and

employment under the influence of external factors can lead to a better

understanding of the need for investments in infrastructure or public

transport.


95

3.4 Methodology

3.4.1 Premises

The premise of our method implies that any observed departure will be

matched to the nearest observed arrival, within a pre-defined time frame

(e.g. morning of a working day). For each trip purpose (e.g., work), the

number of departures per zone, as well as the number of arrivals, is

retained, but the observed connection between origins and destinations

will be cut through with the aim of minimizing the distance between

those two. To apply the methodology on a real case, we assume that an

origin-destination (OD) matrix is available for the selected zones, as well

as a distance matrix, providing the shortest physical network distance

between each pair of zones.

This theoretical exercise does not take into account the match

between origin and destination that exists in the real world. For home-

work travel, this would imply that everyone who is part of the active

population can be considered suitable to perform any job. Although this

theoretical assumption does not necessarily correspond to a real world

situation, we have made this assumption deliberately because we want to

gain an insight into the theoretically maximum reduction of travelled

distances.

Apart from job qualification, there are some more possible biases that

should be kept in mind. We do not consider residential self-selection

(Mokhtarian and Cao, 2006; Van Acker et al., 2010), the trade-off

between real estate prices, accessibility and environmental quality, or the

income and the composition of households (Van Ommeren, 2000). Also,

the presence or absence of rapid transport infrastructure, such as com-

muter rail, was not included in the calculation. In Section 7 we will

discuss the possible impact of these simplifications for our case study.

Furthermore, chained trips or detours are not simulated. As a conse-

quence, results should be considered an underestimation. In the discussion

of the case study (infra), a comparison will be made between the calcu-

lated distances travelled and those reported by survey respondents.

Chapter 3

96

3.4.2 Linear programming and the Monge-

Kantorovich mass transportation problem

White (1988) first suggested to define the problem of the minimum

required commuting distance as a conventional Monge-Kantorovich mass

transportation problem, to be solved by standard techniques of linear

programming (e.g., a simplex algorithm). The definition of this type of

problem is often illustrated by mines that have to supply factories. All

mines together produce the ore needed by all factories. Because of

differences in location, the transport cost to deliver one shipment from a

mine to a factory varies across the possible pairs of mines and factories.

The problem is solved when an OD matrix is calculated, which yields the

smallest total transport cost. Today several commercial and non-

commercial solvers are available, offering good approximate solutions for

this global minimum cost (in our case: minimum distance) and yielding a

corresponding OD matrix. A popular non-commercial solver, based on the

simplex algorithm, is for example lp_solve (Berkelaar et al., 2003).

In most of the excess commuting literature, the focus is on the calcu-

lated minimum commuting distance, which is usually compared with the

observed total commuting distance. In his spatially disaggregated ap-

proach Niedzielski (2006) also uses the calculated OD matrix to obtain

local values for the minimum commuting distance.

However, this procedure has some disadvantages. First, there is no

unique solution for the mass transportation problem (Feldmann and

McCann, 2002), making the resulting OD matrix dependent on the

software used and, to a certain extent, on the sort order of rows and

columns in the cost matrix. Second, and perhaps more problematic, is

that the algorithm fills as many cells as possible with zeros, mostly

assigning traffic flows from one zone to only one corresponding zone. This

situation is incompatible with our approach which aims at simulating the

implications of the case where all commuters start to look simultaneously

for a job closer to their homes. Apart from minimizing the total travelled

distance (which is equivalent to the cost in the transportation problem),

it is in this case also important that the optimization process is done in a

geographically balanced way, obtaining the global optimum through

multiple parallel local optimization processes. Because of this additional

premise, we develop our own procedure instead of using a standard solver

software package. Our method results in a geographically much more


97

“smooth” solution, avoiding results where adjacent zones show extremely

differing values.

3.4.3 Algorithm

The algorithm developed here has been implemented to assess the extent

of excess commuting. The approach adopted is as follows. First, per zone

trips are made as much as possible intra-zonal. Thereafter in each run,

one trip from the remaining surplus or deficit is exchanged with the

nearest matching zone. The distance covered by the exchanged trip is

recorded. This cycle is repeated until all departures are matched with an

arrival. We give a more detailed description of the process in Fig. 3.1.

Initially, the physical distance dij over the network between any pair

of zones (i, j) is calculated. Each zone is represented by its centroid.

Using the Dijkstra algorithm, as implemented in the Network Analyst

extension of ESRI ArcGIS 9.2, all shortest paths between any pair of

centroids over the network were calculated. The resulting product is a

symmetrical distance matrix in which both rows and columns represent

every zone of the dataset. The distance for an intrazonal trip, which is

originally calculated as zero, is simulated by taking half the network

distance between the centroid of the considered zone and the centroid of

the nearest zone. In this way, intrazonal network distance is also taken

into account, which is not always done by alternative methods such as

developed in O’Kelly and Lee (2005).

The combination of the OD matrix and the distance matrix provides

the total distances travelled between each pair of zones, both viewed as

outbound (every zone considered as origin) as well as inbound (every zone

considered as destination).

The departures and arrivals within the OD matrix are then summed,

so that we obtain for each zone i the total number of departures Oi as

well as the total number of arrivals Di. The minimization process starts

by equalizing the number of internal trips Ii to the number of departures

Oi (in the case where there are fewer departures than arrivals) or to the

number of arrivals Di (in the opposite case). This number is then multi-

plied by the distance of the simulated internal trip HIi, and stored as a

basic travelled distance, both outbound (HOi) and inbound (HD

i). For the

next step in the process, the smaller of those two values is subtracted

from the number of departures from and arrivals at the considered zone.

Trips in one of the two directions are thus reduced to zero.

Chapter 3

98

For the remaining surplus (if originally there were more arrivals than

departures) or deficit (in the opposite case), we question whether a deficit

or otherwise a surplus exists in the nearest zone j. If a deficit (or surplus)

can be found in zone j, then one trip from the surplus (or deficit) of zone

i is used to fill up (or to receive) a part of the deficit (or surplus) of zone

j. The distance dij, which should be covered to eliminate this trip, will be

added to the outgoing (or incoming) distances (HOi or HD

i) of the consid-

ered zone as well as to the incoming (or outgoing) distances (HDj or HO

j)

of the nearby zone.

This process is pursued step by step until all zones i are given an ini-

tial chance to exchange trips ends. Then the cycle is repeated until all

surpluses and deficits are eliminated. We are particularly interested in the

total minimized outbound distance TOi and the total minimized incoming

distance TDi. In combination with the observed outbound distance

(TOi(obs)) and the observed inbound distance (TD

i(obs)) for the consid-

ered zone i, which we previously calculated, we can also map the

corresponding excess rates.

The main advantage of our algorithm is that the structure of the cal-

culated OD matrix is not arbitrary, but that the global optimum is

achieved through many cycles of local optimization. Tests show that the

impact of the sort order of the rows and columns on the resulting matrix

is negligible. However, there is also a disadvantage to taking this ap-

proach. In comparison with a solution generated by lp_solve, the total

minimum commuting distance yielded by our algorithm is a bit larger (11

to 15% in our tests), making it suboptimal in a mathematical sense,

although still preferable from a geographical point of view.


99

for every zone i characterized by iO

departures and

iD arrivals:

is ii OD < or ii DO < ?

if ii OD < : if ii DO < :

internal movements ii DI = ,

0=iD , iii IOO −=

internal movements ii OI = ,

0=iO , iii IDD −=

external covered distance for departures 0=OiH ,

external covered distance for arrivals 0=DiH ,

select zone i nearest by j

look up network distance ijd in the distance matrix

calculate the internally covered distance 2

ijiIi

dI .H =

store IiH , iD and iO

select the first zone i

select zone i that is nearest by j for which:

not jj OD == 0

if 0=jD then 0≠iD

if 0=jO then 0≠iO

look up network distance ijd in the distance matrix

is 0=jD or 0=jO ?

if 0≠jD (and 0=jO ): if 0≠jO (and 0=jD ):

calculate the externally covered distances:

ijOi

Oi dHH += ,

ijDj

Dj dHH += ,

1−= ii OO ,

1−= jj DD

calculate the externally covered distances:

ijDi

Di dHH += ,

ijOj

Oj dHH += ,

1−= ii DD ,

1−= jj OO

store the calculated values

is 0== ii OD ?

if not 0== ii OD :

select the next zone i in the list if 0== ii OD :

is for all zones i : 0== ii OD ?

if not for all zones i : 0== ii OD :

select the next zone i in the list if for all zones i : 0== ii OD :

for all zones i :

total minimized outgoing covered distance Ii

Oi

Oi HHT +=

total minimized incoming covered distance Ii

Di

Di HHT +=

Fig. 3.1. Minimization algorithm

Chapter 3

100

3.4.4 Excess rate

The excess rate E is defined as the ratio between the observed travelled

distance T(obs) (in the model) and the calculated minimum commuting

distance T, and is calculated both for outbound (EOi) and inbound (ED

i)

trips of every zone:

Oi

OiO

iT

obsTE

)(= (3.1)

Di

DiD

iT

obsTE

)(= (3.2)

We calculate the excess rate E for the average departure from (EO) or the

average arrival at (ED) a typical zone i as follows:

∑∑ ⋅=

ii

i

iOi

OiO

O

O

T

obsTE

)( (3.3)

∑∑ ⋅=

ii

i

iDi

DiD

D

D

T

obsTE

)( (3.4)

While the average excess rate over the whole study area is given by:

∑

∑

∑

∑===

i

Di

i

Di

i

Oi

i

Oi

DO

T

obsT

T

obsT

EE

)()(

(3.5)

In case the data set is acquired by different sub models, as described in

the “data” section infra, a specific application of the MAUP might occur

leading to a small error so that in practice DO

EE ≠ . The reason is that

every sub model, of which the results will be combined into one entity,

provides only detailed zoning in its own focus area (in the case of our

data: a province). We overcome the aggregation error by means of the

following approximation:

∑ ∑

∑∑

+

+

=

i i

Di

Oi

i

Di

i

Oi

TT

obsTobsT

E

)()(

(3.6)

We define the observed average distance covered per trip )(obsh as

follows:

∑

∑

∑

∑===

ii

i

Di

ii

i

Oi

DO

D

obsT

O

obsT

obshobsh

)()(

)()( (3.7)

while we define the minimized average distance per trip:

∑

∑

∑

∑===

ii

i

Di

ii

i

Oi

DO

D

T

O

T

hh (3.8)


101

Again, an error may occur by aggregating sub models, so that in practice

)()( obshobsh DO≠ and DO hh ≠ . In that case, we approximate )(obsh

and h as follows:

∑∑

∑ ∑

+

+

=

ii

ii

i i

Di

Oi

DO

obsTobsT

obsh

)()(

)( (3.9)

and

∑∑

∑ ∑

+

+

=

ii

ii

i i

Di

Oi

DO

TT

h (3.10)

3.4.5 Spatial distribution and density ratio of

arrivals and departures

The minimum commuting distance consists of a combination of both the

spatial separation of functions and a difference in density between typical

residential areas and typical employment centres. We will quantify these

properties by defining density ratios between departures and arrivals.

We calculate the density C of the departures O and arrivals D for the

typical zone i with area A from which an average trip departs or at which

an average trip arrives as:

∑∑

⋅=

i

ii

i

i

iO

O

O

A

OC (3.11) and ∑

∑⋅=

i

ii

i

i

iD

D

D

A

DC (3.12)

While the average density over the entire study area is given by:

∑

∑

∑

∑===

ii

ii

ii

ii

DO

A

D

A

O

CC (3.13)

Spatial separation of the considered functions can thus be measured by

calculating the ratio between CO and CD:

D

OOD

C

CR = (3.14)

Chapter 3

102

3.5 Case study area: Flanders

and Brussels (Belgium)

Within the context of this research we want to test our model for the

Flanders and Brussels region. It is important to take into account that in

this region commute areas of different cities overlap (Van Nuffel, 2007),

and that many jobs are located far outside the CBDs, such as in port

areas, small towns, historically developed businesses outside of urban

areas or peripheral industrial sites. Even though spatial dispersal is

usually larger in the residential function than in other functions, a

polycentric and partly dispersed spatial distribution of both jobs and

other destinations should be taken into account. The polycentric nature

of regional employment has a historical basis, while sprawl is mainly a

post-war phenomenon that is still developing (Vandenbulcke et al., 2009).

Riguelle et al. (2007) note that contemporary polycentric development, in

the form of sub-centres in the periphery of major cities (“edge cities”) has

hardly occurred in Belgium. Moreover, Brussels, and to a lesser extent

Antwerp and Ghent, are dominant employment centres, putting the

importance of other historical centres into perspective (Aujean et al.,

2005).

Brussels (with more than 1 million inhabitants) is the main centre of

service industries and government activities and is also the largest

employment centre in Belgium. The economy of Antwerp (about 500,000

inhabitants), the second largest city in Belgium and one of the largest

ports in Europe, is based on port activities and industries (e.g., petro-

chemicals). Ghent (about 240,000 inhabitants) is the next urban area in

the ranking, with significant activity in industry, port operations and

research and development. Major centres in the immediate sphere of

influence of Brussels and Antwerp are Mechelen, Leuven, Aalst and Sint-

Niklaas. In the east we find the double-centre of Hasselt-Genk, which

developed around the long gone mining industry, but managed to attract

new businesses. In the southwest we find the Roeselare-Kortrijk region

characterized by smaller-scale industrial activities, while Bruges (in the

northwest) is oriented towards tourism and limited port handling. The

coastal area is dominated by an elongated urban network which is mainly

based on tourism.

The framework for the description of spatial structures at the macro

scale in this region is the so-called Spatial Structure Plan for Flanders


103

(RSV, 1997/2004). The RSV is the overarching spatial policy plan for the

Flanders region. Among other issues, the RSV selects and demarcates

urban areas, for which specific urban planning policies are defined.

3.5.1 Data

The methodology can be applied at different scale levels. Within the

scope of this paper we focus on the level of census wards, corresponding

with neighbourhoods. This unit division allows the development of a

detailed analysis of the theoretical minimum commuting distances and

excess traffic generation. This implies having detailed data. We use the

OD matrices of the multimodal model for Flanders (MMM) (Verhetsel,

1998). MMM is a macroscopic traffic simulation model that was commis-

sioned by the Flemish government and has been developed since 1998.

The model is essentially made up of five sub models, one for each prov-

ince of Flanders, including the Brussels Capital Region for consistency.

Every sub model consists of a GIS map that divides the province into

small traffic analysis zones (TAZs). In most places, TAZs correspond to

standardized census wards. To obtain homogeneous densities, some

repartitioning was done. In sparsely populated areas different census

wards were regrouped into one TAZ. In other places, such as the port

areas, a more refined zoning was applied. The surrounding areas (the

other sub modelled provinces, the Walloon Region, France, the Nether-

lands, Germany and Luxembourg) are also part of the GIS map, but are

divided into TAZs of which the size increases with the distance from the

study area. Properties of the used TAZs can be found in Table 3.1.

Table 3.1. Properties of the used TAZs

total mean st. deviation

number of TAZs 6,652.00 - -

area (km2) 13,751.36 2.07 2.80

number of departures

TAZ (4-11 a.m.)

2,356,461.00 354.00 406.00

number of arrivals

TAZ (4-11 a.m.)

2,356,461.00 354.00 692.00

The GIS map is linked to an OD matrix which indicates for a certain

period of time how many trips occur from every zone to any other zone in

the model. Each provincial model contains roughly between 1,400 and

Chapter 3

104

3,300 zones, and the associated matrices have as many rows as columns.

The matrices that were available for this study simulate traffic on an

average weekday between 4 a.m. and 11 a.m. (i.e., morning traffic). By

aggregating the data for a largely extended morning rush period, we

avoid inaccuracies caused by the calibration of the MMM, which is

designed to calculate traffic flows on an hourly basis.

The matrices were first built on data on home-work commuting and

home-school commuting from the General Socio-Economic Survey 2001

(SEE 2001; Verhetsel et al., 2007), that are available at the level of

census wards. SEE 2001 is an exhaustive survey of the Belgian population

(excluding children younger than six years), which has its origin in the

decennial census. The questionnaire of SEE 2001 gathers each individual’s

residence address and the address of the workplace or school. While

83.2% of the respondents provided the name of the municipality of the

workplace or school, only 56.4% filled out the full address. This is

significantly lower than the overall SEE 2001 response rate of 95%

(Verhetsel et al., 2007). The processed data were aggregated by

neighbourhood and supply a picture of the daily travelled distances to

and from each neighbourhood. This information can also be aggregated

for analysis at the municipal level. This data was geocoded; errors and

deficiencies such as the lack of addresses were corrected wherever possi-

ble. To this end, alternative socio-economic databases were used, which

were supplied by the Crossroads Bank for Enterprises1 (for home-work

travel) and the Flemish Department of Education (for home-school

travel). For other kinds of travel, grouped as recreation, shopping and

other traffic, synthetic matrices were built using a gravity model of which

the parameters were derived from the Travel Behaviour Survey in

Flanders and other relevant surveys conducted in Belgium. In this way, a

complete OD matrix for the base year 2007 was obtained.

Given the significant share of home-work and home-school travel dur-

ing the morning peak hours, one can state that the OD matrices

represent an adequate simulation of personal mobility in the morning.

Nevertheless, caution is required with respect to the interpretation of the

results for trip purposes other than work or school because for these trips

the data is of lower accuracy. To illustrate, on the basis of the datasets

used, we cannot make sufficiently accurate simulations for the traffic

1 The Crossroads Bank for Enterprises is the public data management service

of the Belgian social security system.


105

outside the peak hours (which consists mainly of other than home-work

and home-school based trips), particularly during weekends or vacation

periods.

The zoning of the MMM, which is based on census wards, was done

in a more refined way for densely populated areas than for those areas

that are sparsely populated. While it is logical to define zones with a

homogeneous density, this has the disadvantage that the effect of the

specific travel behaviour of a small population in a large but sparsely

populated zone could easily result in a disproportionately prominent spot

on the map. This problem can be partly alleviated by using a predeter-

mined density threshold below which no data is represented.

Even though the MMM provides data for non-work travel too (Ham-

madou et al., 2008), within the framework of this paper, the calculation

and discussion is done for home-work travel only. This is plausible since

home-work trips are in an economic sense the most crucial of all personal

trips. This can be illustrated using price elasticities. Home-work travel in

particular is a lot less price elastic, and thus more inert, than other types

of trips, such as leisure or shopping travel (Mayeres, 2000). An additional

reason to focus on home-work travel is the fact that the availability of

data about this commuting class is generally better, and that those

datasets usually are more complete and reliable (Witlox, 2007).

3.5.2 Network

We use Streetnet as a network to connect the various zones. Streetnet is

a detailed topological representation of the Belgian road network built up

by links and nodes to which various attributes, for example, road classifi-

cation, are attached. To calculate the shortest network-based path

between each pair of centroids of the considered zones, we use the seven

highest functional road classifications from the Streetnet network data.

These categories include all regional and local roads that could be used as

a connection. The two lowest categories which cover alleys and rural

roads are not included. The public transport network is also not included,

since we assume that the search for the shortest road between two points

usually results in a shorter path than a search for the shortest link

through a line of public transport.

For travel outside Belgium, where a lower accuracy is acceptable, we

manually extend the Streetnet file to areas outside the country borders

Chapter 3

106

with a square grid with a mesh of 5 km in the immediate vicinity of the

borders, and a mesh of 10 km in the more remote areas.

3.6 Application and results of the case study

From the available MMM data we deduce a data set containing only

home-work travel. The calculation is done for each sub model separately,

for the time frame of a weekday morning. The resulting maps of the five

sub models are then combined into one map, covering Flanders and

Brussels.

The detailed zoning of the MMM allows the detection of variations

both in the observed commuting distances and the minimum commuting

distances on Flanders’ scale, but also allows us to make a more detailed

analysis and to discover relations to spatial characteristics of neighbour-

hoods.

Further, it is also possible to make a typological classification (e.g.,

industrial area, suburban allotment, business district, nineteenth-century

belt, historic town centre, ribbon development, peripheral built-up area,

et cetera), and then to seek explanations for variations of (minimum)

commuting distances within a selected typology (e.g., distance to eco-

nomic cores, distance to the main road network, supply of public

transport, et cetera). This form of analysis, however, falls outside the

scope of this paper and will be subject to further research.

3.6.1 Spatial variations of the minimum

home-work distance

Figs 3.2 and 3.3 show the calculated minimum commuting distances,

based on departures and arrivals in the morning traffic. The zones with a

density of departures or arrivals below the 10th percentile were omitted

to avoid disproportional dominance of large but sparsely populated areas

on the map.

For the entire study area the following values were found:

=)(obsh 16.2 km and =h 6.9 km

The value for )(obsh (16.2 km) differs from the reported value of 19.0

km which is given by SEE 2001 (Verhetsel et al., 2007). Hence, an

underestimation of the reported situation by our model is found, which is

caused by chained trips and detours that were not taken into account.


107

3.6.1.1 Origins

Taking Fig. 3.2 as a reference, the following areas, considered as depar-

ture areas or origins, display low values (which we situate in the lowest

distance class on the map, e.g., 0 - 3.20 km).

This can be interpreted in a positive sense, since low values indicate a

high degree of spatial proximity between home and work locations, and

thus marks opportunities for short travel distances.

The metropolitan region of Brussels (1), Antwerp (2) and a wide belt

between these two cities stand out. In Antwerp, the port plays an

important role. Parts of the metropolitan region of Ghent (3) show rather

low values. However, some rather mono-functional residential areas score

relatively poorly, including some densely populated parts of the nine-

teenth-century belt. Like in Antwerp, in the northern part of Ghent the

nearby port is decisive. The highly dispersed southwestern region of

Roeselare-Kortrijk (4), the urban network of the coast (5) and the

regional urban areas and small urban areas are in general in the same

case. There is a wide variety of ranges of influence between the different

regional urban areas. A number of specific areas show low scores as well.

These are usually characterized by a rather low population density in

combination with a local concentration of employment.

Following areas, among others, display high values (which we situate

in the higher distance classes on the map, i.e., > 8.20 km). The broad

north-south oriented belt located between Ghent (3) and Brussels (1)

stands out, with the highest values being recorded in the south (6). Also

a corresponding belt located east of Brussels shows high values (7).

Finally, the rather remote parts of the east-west oriented axis of the E40

highway (8, 9) are in the same case. The high values, obtained for these

areas, can be interpreted in a negative sense, since these indicate a poor

degree of spatial proximity and actual travel distances that are necessar-

ily large.

Chapter 3

108

Fig. 3.2. Theoretical minimum commuting distance in home-work travel,

origin zones during morning traffic

Fig. 3.3. Theoretical minimum commuting distance in home-work travel,

destination zones during morning traffic


109

3.6.1.2 Destinations

Also in the arrival zones (destinations), in this case to be considered as

employment centres, there are significant differences in the distance

within which employees can be found for the available jobs. From the

explained perspective, low values (situated in the lowest distance class on

the map) can be interpreted as positive. Typically residential areas, where

the supply of active labour force is high but only few jobs are available,

show low values. This is the case in most areas because jobs occur in

stronger spatial concentrations than houses do. Urban neighbourhoods

with a good balance between housing and jobs are in the same case.

Some specific areas display high values (situated in the higher dis-

tance classes on the map), which can be interpreted as negative. Port

areas and other industrial areas with a high concentration of jobs stand

out in this sense. Also parts of administrative city centres, and especially

the districts in the Brussels Capital Region, that comprise a large share of

offices, show high values.

3.6.2 Spatial variations of the excess rate

Figs 3.4 and 3.5 show the excess rate E for departure zones and arrival

zones. The same density thresholds were used as in Figs 3.2 and 3.3.

For the study area the following values were found:

=OE 16.9 , =

DE 16.0 and =E 2.33

On the one hand, the high values which we find for EO and ED indi-

cate that for a typical trip a lot of profit could be made. On the other

hand, optimization of commuting in areas with a relatively high density

would in the first instance have a negative impact on the areas with low

density, which explains the rather low figure we find for the global excess

rate E (= 2.33, or, put differently, 57% of the actual commute is exces-

sive). Despite the proportionally large profits which could be obtained in

high density areas, a minimization of the commuting distances would lead

to an overall reduction of only a factor of 0.43.

3.6.2.1 Origins

Areas with a high excess rate are typically located in urban areas or,

more specifically, near major concentrations of employment. That is

because in these regions the minimum commuting distances are very

small. At the same time the accessibility of jobs is usually higher than in

Chapter 3

110

the more remote areas, so that the physical distance criterion will be less

preponderant in job choice.

These findings are in line with what Hamilton (1982) has found,

namely, that city-dwellers go to work many times further from home than

is suggested by the spatial distribution of housing and jobs. The explana-

tion is found, on the one hand, in the theory of the constant travel time

budget and, on the other hand, in the financial travel budget being a

constant share of the household income (Schafer, 2000). Departing from

an urban area, there are a lot more jobs available within reach of the

available generalized personal travel budget than departing from a more

rural area. Viewed from the urban area, the job that yields the greatest

benefit will often not be the nearest job, compared to the viewpoint from

the countryside.

Nevertheless, net commuting distances departing from the urban ar-

eas are still shorter than average. In summary, the spatial structure of

urban areas ensures that the commuting distances are relatively low, but

that there still exists a wide margin that allows an additional reduction of

travelled distances.

Note that some more suburban regions can be found that have both a

high excess rate and long commuting distances. Those are generally areas

that are easily accessible and have high incomes so that the barrier to

travel over long distances is far lower there.

As opposed to these urban and suburban areas are the more remote

municipalities, mostly belonging to the rural areas. Most of these munici-

palities have an excess rate of around 1, or often even lower than 1.

Again, these low values could be explained by the high excess rate in the

core municipalities. Many workers who live in those core municipalities

still go to work far from home, and thus make the nearby jobs available

for residents of the surrounding municipalities. The very high excess rates

in the core municipalities are responsible for the counterintuitively low

excess rate in the surrounding municipalities.


111

Fig. 3.4. Excess rate in home-work travel, origin zones

during morning traffic

Fig. 3.5. Excess rate in home-work travel, destination zones

during morning traffic

Chapter 3

112

The municipalities with a very low excess rate are the most vulner-

able to changes in external factors that steer travel behaviour. When the

generalized cost of trips would increase (e.g., by rising fuel prices, conges-

tion, road pricing or deterioration of the supply of public transport),

employees will be inclined to look for a job closer to home. At the time

that residents of core municipalities are going to work closer to home, this

would mean that the residents of the surrounding municipalities with a

low excess rate would have to go to work even further from home.

3.6.2.2 Destinations

Zones with a low excess rate are typically found in areas with high

concentrations of employment, such as the port areas and other large-

scale industrial cores and city centres. In the case of large-scale industry,

this is somewhat remarkable because in most cases the observed commut-

ing distances to such locations are already considerably higher than

average. The physical separation of functions plays here: for the indus-

tries that are established at these, often remote, locations it appears

difficult to attract employees from a small recruitment area. A similar

phenomenon as in the large industrial cores occurs in the office centres in

Brussels.

Also in the region Roeselare-Kortrijk we notice low values, which are

in this case, however, linked to short observed commuting distances. In

this region the actual travelled distances are more often than average

approaching the optimum. In the north-south oriented belt between

Ghent and Brussels, we notice again low values, but given the limited

employment there, this should not be evaluated positively.

High excess rates are typically found in both sparsely and densely

populated residential areas, where the relatively limited number of

available jobs is often occupied by employees who do not necessarily live

in the vicinity.

3.6.3 Spatial distribution and density ratio

of arrivals and departures

The density of a zone from which an average trip departs CO is 1,187

departures/km2, while for the average arrival CD is 4,756 arrivals/km2.

The ratio between the two density measures ROD = 0.25. The average

density of arrivals and departures DO

CC = = 171 trips/km2.


113

The large average disparity in concentrations of jobs and houses sheds

light on the spatial background of the minimum commute distance. In the

specific case of home-work travel, this ratio is of course strongly related

to the more commonly used jobs-housing balance parameter (Peng, 1997).

3.7 Possible biases

In the context of Flanders and Brussels, the mentioned possible biases,

originating from our premises, will be more serious when we consider the

larger cities, particularly Brussels. In Belgium, highly specialized, well-

paid jobs are mostly centralized in the Brussels region, where the CBD

plays an important role. The long distance rail accessibility to this CBD

is excellent, while the geographically central location in the Brussels

agglomeration ensures the interaction with a large number of potential

employees who live in the surroundings. These factors result in specialist

workers in Brussels not living in the city. Instead, they prefer the green

suburban neighbourhoods in Brussels’ periphery, or in the less densely

built municipalities of the large commuting region. The spatial mismatch

between locations of work and residence of specialized employees is an

additional obstacle to a possible reduction of commuting distances,

occurring particularly in the large cities, and especially in Brussels. A

similar bias is found in the spatial variation of household sizes. One-

person households, which are encountered more often in cities, face more

freedom regarding the choice of job and residence location than families

with two breadwinners.

3.8 Conclusions

In this paper, we have studied the spatial variation of the minimum

commuting distance and the excess rate as indicators of spatial proximity

of functions, in particular housing and jobs. We have elaborated a

methodology to calculate these indicators, and indicated their relevance

for spatial planning policies. Methodological problems associated with the

use of those indicators still occur, especially when different regions or

cities are compared with each other. However, the spatial variation of the

minimum commuting distance seems suitable to measure the extent to

which a given area can operate on the basis of short distances. The main

Chapter 3

114

reason for these variations is the systematic differences in the spatial

distribution of the job market with respect to the housing market.

In Flanders and Brussels jobs, but also services, show a much

stronger pattern of concentration than dwellings do. Moreover, subur-

banization of both functional groups happens in a different way.

Employment is mainly situated in or in the immediate sphere of influence

of urban areas. Furthermore, extremely high local concentrations of

employment exist, such as in the Brussels office districts, the seaports or

the national airport. Employers who are located in these areas are usually

unable to recruit workers living on average close to the company.

The suburbanization of dwellings is for a major part located in mu-

nicipalities in the countryside, often far away from the economic core

areas. Although we find the highest concentrations of this residential

function in the urban areas, housing is much more homogeneously spread

over the entire study area than jobs. This means that for inhabitants of

the more remote regions with a low jobs-housing ratio it is very difficult

to find a job close to home.

The minimum commuting distance can be considered as a measure of

proximity to the labour market, viewed from the housing market, or vice

versa, as a measure of proximity to the housing market, viewed from the

labour market. The model that is discussed in this paper suggests that

the minimum average distance for a home-work trip within the current

job and housing market in Flanders and Brussels is fixed at 6.9 km, to be

compared to a calculated real world value of 16.2 km. However, in this

variable important regional gradients can be observed. Employees living

in the vicinity of the economic core areas can easily find a job close to

home, whereas the inhabitants of remote rural municipalities must

necessarily commute over long distances.

The regional variations of the excess rate show that people who live

in the vicinity of major employment centres could still significantly

reduce their daily commuting distance, relatively spoken. However, for

residents of outlying regions this would be difficult or even impossible:

they will instead be required to travel even longer distances in case a

general contraction of commuting distances would happen. Such a

reduction of commuting distances is a scenario that may occur when the

absolute cost of transportation would increase.

To date, in the Western world we have only observed a trend of in-

creasing commuting distances, an evolution originating in the

democratization of the car. Since the fastest way to make a typical


115

journey to work is by car, this has led to higher speeds and ultimately to

trips covering increasingly longer distances.

The spatial segregation of functions in Flanders and Brussels mainly

developed in the era of upcoming cheap, fast transportation. We can

therefore say that the longer distances travelled partly materialized in the

form of the physical separation of functions. To a certain extent, this

spatial development shuts the door to a potential shrinkage of commuting

distances. Regions with a large minimum commuting distance are there-

fore very sensitive to rising transport costs, leading to a reduction of

mobility. In the long run, however, such an increase in costs is likely to

occur in the light of peak oil.

Although the paper only deals with home-work commuting, a similar

logic may be valid for services such as schools and shops. Spatial prox-

imity of functions is a paramount prerequisite for a sustainable travel

pattern on the basis of relatively short distances. Towards policy making,

this can be translated as the importance of providing an adequate spatial

and functional mix. Given the relatively large average daily distance that

is covered per person, the role of this spatial mix is probably more

important at a regional level than at the level of, for example, a historical

urban structure (compact city). Concretely, this could mean that the

stimulation of additional jobs and services in areas with a low jobs-

housing balance should be given priority, but also that suburbanization of

the residential function in remote rural municipalities should be discour-

aged.

Mapping the development of minimum commuting distances is an

important research field. For a better understanding of the evolution of

this indicator, it is necessary to compare longitudinal data. Further

research should also include non-work-related trips, and should take into

account other biases, such as the influence of accessibility (in terms of

travel time), the role of chained trips, household composition and income,

and modal choice.

References



42(9), pp. 1537–1547.

Chapter 3

116

Aujean, L., E. Castiau, M. Roelandts and C. Vandermotten (2005)

Central Belgium: Commuting & the Definition of Functional Urban

Regions. London: Insitute of Community Studies/The Young Foun-

dation & Polynet Partners.

Banister, D., S. Watson and C. Wood (1997) “Sustainable cities, trans-

port, energy and urban form.” Environment and Planning B. 24(1),

pp. 125-143.

Berkelaar, M., K. Eikland and P. Notebaert (2003) lp_solve, version 4.0

from http://groups.yahoo.com/group/lp_solve/.



43(5), pp. 580-591.

Charron, M. (2007) “From excess commuting to commuting possibilities:

More extension to the concept of excess commuting.” Environment

and Planning A. 39(5), pp. 1238-1254.

Crane, R. (1996) “The influence of uncertain job location on urban form

and the journey to work.” Journal of Urban Economics. 39(3), pp.

342-356.

Feldman, M. and R. J. McCann (2002) “Uniqueness and transport

density in Monge’s mass transportation problem.” Calculus of Varia-

tions and Partial Differential Equations. 15(1), pp. 81-113.

Frost, M., B. Linneker and N. Spence (1998) “Excess or wasteful com-

muting in a selection of British cities.” Transportation Research Part

A. 32(7), pp. 529-538.

Giuliano, G. and K. A. Small (1993) “Is the journey to work explained by

urban structure?” Urban Studies. 30(9), pp. 1485-1500.

Hamilton, B. W. (1982) “Wasteful commuting.” Journal of Political

Economy. 90(5), pp. 1035-1053.

Hammadou, H., I. Thomas, A. Verhetsel and F. Witlox (2008) “How to

incorporate the spatial dimension in destination choice models: The

case of Antwerp.” Transportation Planning and Technology. 31(2),

pp. 153-181.

Horner, M. W. and A. T. Murray (2002) “Excess commuting and the

modifiable areal unit problem.” Urban Studies. 39(1), pp. 131-139.

Horner, M. W. and M. E. O’Kelly (2007) “Is non-work travel excessive?”

Journal of Transport Geography. 15(6), pp. 411-416.

Kim, S. (1995) “Excess commuting for two-worker households in the Los

Angeles metropolitan area.” Journal of Urban Economics. 38(2), pp.

166-182.


117

Ma, K. R. and D. Banister (2006) “Excess commuting: A critical review.”

Transport Reviews. 26(6), pp. 749-767.

Ma, K. R. and D. Banister (2007) “Urban spatial change and excess

commuting.” Environment and Planning A. 39(3), pp. 630-646.

Mayeres, I. (2000) “The efficiency effects of transport policies in the

presence of externalities and distortionary taxes.” Journal of Trans-

port Economics and Policy. 34(2), pp. 233-260.

Mills, E. S. (1967) “An aggregative model of resource allocation in a

metropolitan area.” American Economic Review. 57(2), pp. 197-210.

Mokhtarian, P. and X. Cao (2006) “Examining the impacts of residential

self-selection on travel behavior: A focus on methodologies.” Trans-

portation Research B. 42(3), pp. 204-228.

Muth, R. F. (1969) Cities and Housing: The Spatial Pattern of Urban

Residential Land Use. Chicago: University of Chicago Press.

Niedzielski, M. A. (2006) “A spatially disaggregated approach to com-

muting efficiency.” Urban Studies. 43(13), pp. 2485-2502.

O’Kelly, M. E. and M. A. Niedzielski (2008) “Efficient spatial interaction:

Attainable reductions in metropolitan average trip length.” Journal


O’Kelly, M. E. and W. Lee (2005) “Disaggregate journey-to-work data:

Implications for excess commuting and jobs-housing balance.” Envi-


Peng, Z. R. (1997) “The jobs-housing balance and urban commuting.”

Urban Studies. 34(8), pp. 1215-1235.

Riguelle, F., I. Thomas and A. Verhetsel (2007) “Measuring urban

polycentrism: A European case study and its implications.” Journal

of Economic Geography. 7(2), pp. 193-215.





Scott, D. M., P. S. Kanaroglou and W. P. Anderson (1997) “Impacts of

commuting efficiency on congestion and emissions: Case of the Hamil-

ton CMA, Canada.” Transportation Research Part D. 2(4), pp. 245-

257.

Small, K. A. and S. Song (1992) “Wasteful commuting: A resolution.”

Journal of Political Economy. 100(4), pp. 888-898.

Chapter 3

118

Song, S. (1995) “Does generalizing density functions better explain urban

commuting? Some evidence from the Los Angeles region.” Applied

Economics Letters. 2(5), pp. 148-150.

Van Acker, V., B. van Wee and F. Witlox (2010) “When transport

geography meets social psychology: Toward a conceptual model of

travel behaviour.” Transport Reviews. 30(2), pp. 219-240.




353.

Van Nuffel, N. (2007) “Determination of the number of significant flows

in origin-destination specific analysis: The case of commuting in

Flanders.” Regional Studies. 41(4), pp. 509-524.

Van Ommeren, J. (2000) “Job and residential search behaviour of two-

earner households.” Papers in Regional Science. 79(4), pp. 375-391.

Vandenbulcke, G., T. Steenberghen and I. Thomas (2009) “Mapping

accessibility in Belgium: A tool for land-use and transport planning?”


Verhetsel, A. (1998) “The impact of spatial versus economic measures in

an urban transportation plan.” Computers, Environment and Urban

Systems. 22(6), pp. 541-555.



gium.

White, M. J. (1988) “Urban commuting journeys are not ‘wasteful’.”




15(3), pp. 172-183.

Yang, J. and J. Ferreira (2008) “Choices versus choice sets: A commuting

spectrum method for representing job-housing possibilities.” Envi-

ronment and Planning B. 35(2), pp. 364-378.

119

Chapter 4:

Measuring spatial

separation processes

through the minimum

commute

This paper is published as Boussauw, K., B. Derudder and F. Witlox

(2011) “Measuring spatial separation processes through the minimum

commute: The case of Flanders.” European Journal of Transport and

Infrastructure Research 11(1), pp. 42-60. Copyright © The authors. All

rights reserved.

Abstract

The average distance covered by individual commuting trips increases

year after year, regardless of the travel mode. The causes of this phe-

nomenon are diverse. Although increasing prosperity is often invoked as

the main reason, the discipline of spatial planning also points to the

relevance of land-use policies that enable processes of suburbanization

and sprawl. By calculating time series of spatially disaggregated theoreti-

cal minimum commuting distances, this paper offers a method to identify

and quantify the process of spatial separation between the housing

market and the job market. We identify the detected spatial separation

as one of the possible indicators for the contribution of spatial processes

to the growth of traffic.

In the case study area of Flanders and Brussels (Belgium), it is found

that over time the minimum commuting distance increased in many

municipalities, especially where population is growing faster than job

supply, or where traditionally high concentrations of employment still

increase. Decreases are noticed in suburban areas that are getting a more

urban character by acquiring a considerable functional mix. For the study

Chapter 4

120

area in its entirety, we do indeed register an increasing spatial separation

between home and work locations. However, this separation evolves less

rapidly than the increase in commuting distances itself.

Regarding the methodology, we find that the use of municipalities as

a spatial entity is suitable for grasping regional transformations of the

economy and intermunicipal forms of suburbanization and peri-

urbanization. However, a similar methodology, applied at a more detailed

geographical scale, could be used to detect processes of sprawl in the

morphological sense.

Keywords: excess commuting; Flanders; sustainable spatial develop-

ment; urban sprawl

4.1 Introduction

Over the last century, there has been a mutually reinforcing relation

between rising prosperity and a general increase in individual mobility:

the growth of traffic volumes are both an effect and a cause of mounting

affluence. For instance, faster transport modes (especially but not exclu-

sively, the car) have consistently gained ground because the individual

budget spent on transport has in absolute terms continuously been

growing. The net result is that the average distance covered per person

has been systematically increasing (Bleijenberg, 2003).

With regards to the commute, this means that, in general, workers

have been looking for daily occupations increasingly further away from

their home, or - conversely - that they have been moving house further

away from their jobs. The logical result is a stronger separation between

house and job location: if travel costs decrease, then travel consumption

increases (Rietveld and Vickerman, 2003). This is especially true when

the spent amount of time can be kept constant by increasing average

travel speed, which seems generally to be the case in the Western world

(Schafer, 2000).

Interestingly, this evolution apparently has an important spatial

component. That is, it looks like changing travel behaviour is partly

materialized in suburban and peri-urban developments, implying that

possible origins and destinations lost relative proximity to each other. In

many regions in the developed world, this materialization may be respon-

sible for a certain degree of irreversibility of the expansion of travel

Measuring spatial separation processes through the minimum commute

121

patterns, as this limits the possibility of travel distances being reduced

again in the future, even if environmental or congestion policies would

aim for this. Since the adverse external effects of traffic are linked to the

distance travelled (especially when travel is by car), this evolution makes

it harder to address traffic problems at the source. In addition, spatial

separation of functions is probably a long term economic disadvantage,

since it makes the economic system vulnerable to possible future circum-

stances where transport costs would increase, for example by rising oil

prices (peak oil) (Boussauw and Witlox, 2009) or growing structural

congestion.

The perspective of this paper is the study of the mobility component

of spatial separation of the job and housing market. Our research ques-

tion is: how can spatial separation between residential locations and job

locations be measured by means of a spatial proximity indicator, and

what is the connection with the growth of observed commuting distances?

We hereby hypothesize that the well-established line of inquiry on excess

commuting (Ma and Banister, 2006) can provide a methodology to

quantify evolutions of spatial separation in a functional sense, and to

provide better insight in the contribution of spatial structure to the

growth of traffic volumes over the last decades.

In practice, this will be done through calculating spatially disaggre-

gated evolutions in minimum commuting distances, defined as the

theoretical minimum distance that each worker would have to cover in

order to find a job as close to home as possible under the assumption that

actual residential locations and job locations are maintained and the total

distance travelled (by all workers together) is minimized. After applying

the method to a case study area in northern Belgium, we evaluate the

observed evolution based on both occurring sprawl and regional-economic

shifts in the labour market, aiming to explain regional variations.

Although assessing time series of the minimum commuting distance in

order to describe spatial separation between the housing market and the

labour market has been done before (Frost et al., 1998; Horner, 2007;

Yang, 2008), the approach adopted in his paper is novel for two reasons.

First, we adopt a study area which could be qualified as a suburbanized

historically polycentric spatial structure - as there are many in Europe -

and shows therefore significant spatial variations (i.e. it includes many

historical urban areas, suburban and peri-urban developments, and major

industrial zones), although it can be demarcated in a fairly consistent

way based on economic and political criteria. Second, we refine the

Chapter 4

122

method in order to detect variations in the minimum commuting distance

in a spatially disaggregated way, which certainly contributes to a critical

understanding of the role of the present variations of spatial structure in

comparison with Frost et al. (1998), Horner (2007) or Yang (2008) who

define only one figure for the entire city or region.

The remainder of this paper is structured as follows. First, we clarify

the possible reasons behind evolutions in spatial proximity by putting

this in the perspective of the existing literature. Second, we present an

overview of the case study area (Flanders and Brussels), including the

main socio-economic and spatial processes and changes that may play a

role in our analysis. Then we develop our methodology, including calcula-

tion and evaluation methods, and discuss the available data. After

applying the method, we present the results, discuss possible biases and

draw some final conclusions.

4.2 Defining spatial separation processes

One can think of several possible reasons leading to changes in spatial

proximity between homes and jobs. For planning policies, the most

relevant possible causes are suburbanization and sprawl. Ewing (1994)

assumes a strong link between sprawl and increased traffic. But also

thorough zoning policies may lead to increasing mutual distances. These

phenomena are local in nature and manifest themselves at a small

geographical scale, meaning that in quantitative analysis, these will

particularly manifest when data are available at a detailed spatial

division (e.g. traffic analysis zones).

Newman and Kenworthy (1989) and Gordon and Richardson (1989)

associate sprawl with a specific morphology, particularly consisting of

monotonous suburban districts with a strict separation of functions,

characterized by store strips, commercial architecture and large internal

distances. However, the extent to which spatial separation leads to an

effective increase of distances that need to be covered is less clear, since

this cannot be derived from local morphological characteristics. For

instance, a monotonous residential lot embedded in a major employment

centre could possibly lead to a more sustainable travel pattern than a

compact town that is immersed in a rural area.

Therefore it is important to view spatial separation not only in a

morphological way but also in a functional sense. Particularly in a


123

context where average trip lengths, and in particular average commuting

distances, have become very large in practice, it is hard to tell which kind

of spatial developments are problematic in relation to mobility, and

which are rather beneficial. Banister (1999) therefore observes the issue at

hand in a non-morphological way, and suggests that it is of the utmost

importance that new developments are sufficiently large and are located

in or immediately subsequent to existing urban areas. As a consequence,

local morphology, density and spatial diversity come only in second place.

Another phenomenon that could cause this kind of functional separa-

tion processes, although particularly at a larger geographical scale level

and consequently only partly related to suburbanization processes,

consists of regional economic shifts within the labour market. An eco-

nomic transformation, e.g. towards a more service-based industry, could

lead to a different spatial distribution of jobs, e.g. through centralization.

But also an absolute increase in the number of jobs in one zone can lead

to a spatial distribution where work locations are on average closer to

residences of potential employees. Since these kinds of regional transfor-

mations take place at a higher geographical scale than suburbanization

processes, these will rather become clear when relatively large zones (e.g.

districts or municipalities) are studied.

If we restrict ourselves to commuter traffic, we see that trip lengths

have increased systematically over the past decades. This trend was

observed in Belgium (see below), the US and the UK, and we may

assume that this evolution is manifest throughout the Western world

(Aguilera, 2005). The basic mechanism that underlies this growth is an

increase in travel speed. Basically, people do not spend more time

commuting than they did before, nor do they spend a greater proportion

of their income on transport (Schafer, 2000). Rather, it is the general

prosperity growth that has led to an absolute increase in resources

devoted to transport, resulting in more car ownership, more car use (at

the expense of slower transport modes), an extension of the motorway

network and a pushed up speed of public transport. In the surroundings

of large agglomerations, congestion has slightly slowed down the overall

increase of travel speed, although it did certainly not stop it (Van Wee et

al., 2002).

Even though spatial separation processes and increasing trip lengths

appear to be associated, this does not necessarily imply a one-way

causality. The fact that employees live on average further from their work

location than before, a phenomenon which occurs partly in the form of

Chapter 4

124

sprawl, is caused by increasing individual mobility, i.e. a wealth-related

phenomenon that is the basis of an autonomous growth of traffic. So, we

might explain these spatial separation processes as a materialization of

this increased mobility, which has itself a mutual reinforcing effect on the

growth of traffic.

4.3 Measuring spatial separation by

excess commuting characteristics

The perceived increase in traffic volume due to spatial separation proc-

esses, whether or not perceived as sprawl, remains a main concern in

most spatial policy plans (Sultana and Weber, 2007). Given the impor-

tance attached to mobility, it may be surprising that quantification of

this phenomenon is usually confined to measuring morphological charac-

teristics. However, in this paper we want to measure the process of

spatial dispersion that results in further separating origins and destina-

tions over the years, or, as might be the case in some areas, decreasing

separation between origins and destinations. By measuring this develop-

ment over a certain time interval, we have a good idea what proportion

of the traffic growth in that period is due to spatial expansion. An

important methodological issue concerns the definition of these origins

and destinations. This obstacle disappears when restricting ourselves to

the commute, where origins (residential locations of workers) and destina-

tions (job locations) can be clearly defined.

Cross-sectional analyses that compare travel behaviour of commuting

residents or workers in suburban areas with commuting behaviour in

areas with urban characteristics are common in the literature. Sultana

and Weber (2007), for instance, recorded large differences in actual travel

times and trip lengths, depending on the direction of the commuter flow.

Flows within urban areas appear to be the shortest, both in distance and

in time. Commuting trips within the suburban areas, however, appear

shorter than commuting trips from the suburbs towards the city.

In order to map the spatial separation between the housing market

and the labour market, regardless of actual commuting patterns, other

methods are needed. An elementary indicator of the spatial distribution

of housing and jobs with an alleged impact on mobility is the jobs-

housing balance (the ratio between the number of jobs in an area and the

number of workers living in the same zone). In its simplest form, this


125

indicator is measured by zone, making it sensitive to variations in zoning

and insensitive to effects of embeddedness in the surrounding region or

the presence or absence of transport infrastructure. Peng (1997) tried to

overcome these problems partly by calculating the jobs-housing balance

based on the surrounding area of which every considered zone is the

centre.

A more advanced line of inquiry is the field of “excess commuting”,

which originally focuses on the study of spatial structure in relation to

commuting behaviour by comparing observed commuting distances with

theoretical calculated minimum commuting distances. An overview of

existing research on this topic is given by Ma and Banister (2006). One of

the spatial characteristics, provided by the excess commuting literature,

is the minimum commuting distance, which incorporates both the spatial

distribution of jobs and housing, and the infrastructure network. It is

related to the job-housing balance, but is a more sophisticated indicator

as it also takes into account the embeddedness of the study area in the

wider region, as well as the transport network (Horner and Murray,

2003). We consider the minimum commuting distance as a measure of the

proximity of the housing market to the labour market (Horner, 2004), or

as a variable that indicates to what extent the system could absorb

shrinkage of commuting distances. The latter is relevant when we want to

understand the potential impact of the commute getting more expensive,

e.g. under the influence of rising fuel prices (Boussauw and Witlox, 2009).

The calculation of the minimum commuting distance implies that the

existing residential and job locations are retained, but that workers are

assigned to jobs in a way that leads to a minimization of the total

distance travelled. More formally:

minimize ∑∑= =

=

n

i

n

jijijtdH

1 1

(4.1)

given: j

n

iij Dt =∑

=1

(4.2), i

n

jij Ot =∑

=1

(4.3), and 0≥ijt (4.4)

in which: H = total distance travelled within the system by

matching workers and jobs

n = number of zones

Oi = number of workers living in zone i

Dj = number of jobs in zone j

dij = network distance between centroids of zone i and zone j

tij = number of trips between zone i and zone j

Chapter 4

126

Ma and Banister (2007) published an in-depth study on the relationship

between excess commuting and urban form. Horner (2007) and Yang

(2008) used also the maximum commuting distance as a measure of

spatial dispersion, while Charron (2007) and Yang (2008) introduced the

terms “proportionally matched commute” and “random average com-

mute”, based on a more extensive theoretical exploration on possible

commuting ranges, and proposed as more realistic alternatives to the

maximum commuting distance.

The maximum commuting distance should be seen as the conceptual

inverse of the minimum commuting distance: it is the value obtained

when all employees within the study area simultaneously would exchange

jobs or houses aiming to live as far as possible from their jobs. In an

urban system consisting of a single city that is isolated from neighbouring

cities, this measure is a good complement to the minimum commuting

distance. In the context of our assessment, however, the notion of maxi-

mum commuting distance is rather abstract. Not only will employees

never be inclined to spontaneously maximize home to work trip length,

also outcomes seem to be largely dependent on the considered region

frontier and are highly correlated with the size of the study area

(Charron, 2007). Moreover, the amount of traffic that is associated with a

maximized commuting distance is not compatible with the existing

transport infrastructure or any form of pursuing economic efficiency. In

this paper, we will calculate the global value of the maximum commuting

distance as an additional support of the observed trends in minimum

commuting distance, but we will not elaborate on spatially disaggregated

values.

The contribution of a time series approach of the proportionally

matched commute is mainly in quantifying the extent to which a regional

system evolves from mono-centric to dispersed (Yang, 2008). Since the

possible interpretation of the meaning of spatially disaggregated values of

the proportionally matched commute in a polycentric region is subject to

further research that is beyond the scope of this paper, we will no further

elaborate on this.

In the literature, the average minimum commuting distance is usually

calculated as a single measure for an entire city or region (Ma and

Banister, 2006). However, it is also possible to calculate this in a spatially

disaggregated way in order to reveal regional variations. In the latter

case, for each zone (traffic analysis zone, municipality, district ...) the

distance outcome is calculated twice: once for the outgoing commute, and


127

once for the incoming commute. Niedzielski (2006) elaborated on a

spatially disaggregated analysis of Warsaw on the basis of minimum

commuting distances. When taken the locations of residence as a view-

point, the largest minimum commuting distances were recorded in areas

with suburban characteristics. Viewed from the perspective of job loca-

tions, the highest values are recorded downtown (CBD). Yang and

Ferreira (2009) included spatially disaggregated values in their informa-

tion system for planners in the Boston area. A similar analysis, albeit on

a regional scale, was made for Flanders and Brussels (Belgium) by

Boussauw et al. (2011). A spatially disaggregated calculation of the

minimum commuting distance on cross-sectional data allows obtaining

these values per zone for a given point in time. In a next step, a similar

analysis based on time series data would then reveal trends, indicating for

a certain zone if the housing market moved on average further away from

the job market over the chosen time span, or if, maybe, the opposite has

happened.

Horner and Murray (2002) devoted considerable attention to the im-

pact of the underlying zoning system on the calculated minimum

commuting distance, both with regard to size and to location. This issue

is known as the modifiable areal unit problem (MAUP). The use of larger

zones means that more trips become intrazonal, leading to an overestima-

tion of the minimum commuting distance (Horner and Murray, 2002).

Both interzonal and intrazonal trips originating in large zones can only be

simulated with limited precision, leading to significant inaccuracies.

However, when excess commuting is studied, this problem can partly be

accommodated by recalculating observed commuting distances using the

same origin-destination matrix that is used to calculate the minimum

commuting distance.

For our research, however, the main problem associated with the se-

lection of zone sizes is the geographical scale at which the studied

developments occur. This means that the spatial separation between jobs

and housing can both be caused by sprawl (at a small geographical scale)

and by regional economic shifts within the labour market (at a large

geographical scale). The first phenomenon will particularly manifest when

detailed zoning (e.g. traffic analysis zones) is used, while the second

phenomenon will become clear when relatively large zones (e.g. districts)

are studied.

While a considerable volume of research on excess commuting has

been published yet, the literature is still fast evolving. Recent inquiries

Chapter 4

128

include attainability of trip length reductions (O’Kelly and Niedzielski,

2008), differentiation in excess rate depending on mode choice (Murphy,

2009), the relationship with the jobs-housing balance as a suitable

indicator for measuring spatial expansion (Layman and Horner, 2010),

the influence of uncertainties in travel time measurements when minimiz-

ing time distance (Horner, 2010) and optimization of spatially

disaggregated calculation methods (Boussauw et al., 2011).

Based on the properties of the excess commuting framework, we find

a time series approach of the minimum commute useful to study proc-

esses of suburbanization. Existing research in this field is rare, however.

Frost et al. (1998) described the evolution of the minimum commuting

distance for a set of demarcated British cities. They compared the 1981

situation with 1991. Yang (2008) reported on similar research in the

metropolitan areas of Boston and Atlanta, based on data from 1980, 1990

and 2000. Both studies find an increase of both minimum commuting

distances and actual commuting distances, in all involved cities. Horner

(2007) compared commuting data from Tallahassee, Florida, from 1990

and 2000. The average reported commuting distance and maximum

commuting distance increased during this period, while the minimum

commuting distance showed only a statistically non-significant increase.

However, neither Frost et al. (1998) nor Horner (2007) or Yang (2008)

applied a spatially disaggregated method, so it is not clear which parts of

the urban areas are the most affected by spatial separation. Furthermore,

no typical peri-urban areas, which are relatively far from the CBD, were

incorporated in the analyses. As stated in the introduction, our approach

consists of an extension to the empirical diversity of the three aforemen-

tioned authors by including suburban and rural areas and an extension of

their methodology by applying a spatially disaggregated method.

4.4 Spatial development and commuting

in Flanders and Brussels

The case study area consists of the administrative regions of Flanders and

Brussels, which together form the northern part of Belgium (Fig. 4.1).

Historically, the area is built around three major cities (Brussels, Ant-

werp and Ghent), a dozen regional cities and several dozen smaller towns

and central municipalities. The economic core is borne by the wide

surroundings of the axis Brussels-Antwerp, a region known as the Flemish


129

Diamond. The rapid development of an extensive railway network in the

nineteenth century and the introduction of cheap season tickets formed

the backbone of a policy aimed at an industrialization of the country

based on limited urbanization (Verhetsel et al., 2010). The new working

class was able to continue living outside major cities, in a house with a

garden and under the watchful eye of the village priest, while commuting

every day to the factory or the office. This form of institutionalized

commute laid the foundation for the spatial separation between housing

and work locations that has only developed more distinctly since the

advent of the motorway and general car ownership. In the period before

World War II, this phenomenon has materialized in the form of spatial

developments that were strongly clustered around the railway stations. In

the post-war period, however, this structure fanned out into a network of

suburbanized and peri-urbanized areas around the ancient settlements.

Fig. 4.1. Situation of infrastructure and urban areas

In this context, suburbanization refers to sprawl that is occurring in or

adjacent to urban areas, while peri-urbanization consists of developments

in the countryside, free-standing or in line with existing villages. These

postwar developments gradually led to the emergence of a variety of

forms of urbanization, ranging from historical densely populated cities

and sparsely populated suburban residential belts to commuter areas and

Chapter 4

130

pure countryside (Vanneste et al., 2008). Depending on the definition,

57% (Vanneste et al., 2008) to 97% (Antrop, 2004) of the Belgian popula-

tion lives in urbanized areas.

The problem of spatial separation, supposedly mainly in the form of

sprawl, is not a new issue in Flanders. In the period between 1980 and

2000 planning became a major political topic, as open space became

scarce, and urban flight appeared to feed a wave of suburban and peri-

urban development. Newly emerging issues in the field of landscape

ecology, water pollution, increasing distribution costs, road safety,

congestion and social segregation were partly attributed to sprawl. The

political response, in the form of the Spatial Structure Plan for Flanders

(RSV, 1997/2004) had to wait until 1997, but did offer an answer in the

form of the demarcation of urban areas and the focus on encouraging

additional building in the cities and existing settlements.

The policy measures that are imposed by this plan are based on a

rather intuitive analysis of the problem, with often no thorough scientific

rationale behind it. Since then, several quantitative spatial analyses have

been carried out. The main research report on sprawl, which was issued

by the Flemish Environmental Agency (Gulinck et al., 2007), approaches

the phenomenon mainly from a landscape-ecological perspective. It

focuses on morphological changes that are related to spatial fragmenta-

tion. In the process, sprawl detection is based on maps and satellite

images from different periods. The main indicator is the overall total

built-up area, which for Flanders increased over the period 1990-2000

from 13.2% to 14.6%. Other indicators were the length of the road

network (increasing by 6.4% between 1991 and 2001), morphological

grain size and global proximity to buildings and infrastructure. A time

series analysis of these indicators points to a systematic increase in

sprawl.

The successive censuses show that commuting distances in Belgium

systematically increase. Based on an assessment of the respondents, the

perceived average distance between home and workplace evolved from

11.9 kilometres in 1970 (Mérenne-Schoumaker et al., 1999) to 19.0

kilometres in 2001 (Verhetsel et al., 2007), an increase of no less than

60%. Although non-commuting trips and freight transport grew even

faster over this period, it is clear that the increase in trip length is largely

responsible for the overall growth in traffic.


131

4.5 Data

The decennial census, organized by the Belgian federal government,

assesses the municipality of residence and the work municipality of all

working citizens. Furthermore, the (perceived) distance travelled every

day to work is also registered. Commuting data for 1981, 1991 and 2001

is available in the form of origin-destination matrices, with the municipal-

ity as a spatial unit. The small municipalities of the Brussels Capital

Region were combined into one zone, in order to avoid biases in relation

to the two other metropolitan municipalities of Antwerp and Ghent. As a

result, the study area is divided into 309 zones with a mean area of 44

km2 (standard deviation: 29 km2).

For the particular purposes of our research, the matrices were cleaned

by removing all trips that have their origin or destination outside the

study area (Wallonia and the neighbouring countries). Apart from these

inter-regional commuters, also home workers (including teleworkers) were

removed from the matrices, as well as respondents with an itinerant

occupation. The exclusion of these records is justified because we only

want to measure the extent to which residential and labour structure

separate within Flanders-Brussels.

Although the method of data collection in the three survey moments

was performed in a similar manner, there are rather large differences

between the three data sets. Important variations in the number of

commuting trips are found, as well as in the number of unknown or

incomplete registrations. Moreover, the number of home workers declined

dramatically over the three consecutive survey moments, and the struc-

ture of the inter-regional commute (between Flanders-Brussels and the

neighbouring region and countries) changed. Possibly more important is

the influence of the regional economic transformations that occurred. In

the period 1981-2001 the economy shifted towards a more service-based

system, and lost a number of important industrial employment centres

such as the Kempen coal mines in the east of Flanders and the Hainaut

steel industry south of the study area. Also, the port of Antwerp and the

airport of Brussels went through an era of major growth, and a high

standard service industry (especially in information technology and

international public institutions) developed in the urban areas, in particu-

lar in the Brussels agglomeration. This restructuring of the labour market

cannot be separated from the suburbanization of the residential structure:

Chapter 4

132

both transformations add to the spatial separation that we want to

assess.

Regarding the development of sprawl in the studied period, also the

growth of the infrastructure network is undoubtedly important. Although

most motorways that exist today where constructed in the period prior to

1981 (roughly between 1961 and 1981, see Fig. 4.1), it can be expected

that many developments along the motorways have been built in the

period 1981-2001.

4.6 Method

To answer the research question, we investigate the evolution of the

spatial distribution of residential locations of workers and job locations in

Flanders and Brussels (Belgium), mainly relying on the minimum com-

muting distance concept, and comparing the trend found with the

evolution of observed commuting distances. To this end we first take the

study area as a whole, then we reiterate the calculation in a spatially

disaggregated way (for each municipality separately). Regarding indica-

tors for the study area as a whole, we will also provide the global

maximum commuting distance.

As we consider the minimum commuting distance as a measure of

spatial proximity between job market and housing market, the possible

mismatch between qualifications and job preferences of workers and

requirements of employers is not taken into account, which is an impor-

tant deviation from reality. Although in the literature attempts were

made to disaggregate excess commuting characteristics based on a

classification of workers and jobs (O’Kelly and Lee, 2005), this approach

falls outside the scope of our paper. Nevertheless it is important to keep

this constraint in mind when interpreting the results. O’Kelly and Lee

(2005) found e.g. that service workers are subject to shorter minimum

commuting distances than industrial workers, but that actual commuting

behaviour of the former group gives rise to larger excess rates than the

latter. However, it is unclear whether these findings also apply outside

the surveyed American cities, e.g. in a Belgian context. Also, the unem-

ployed population and home workers do not affect the results since they

are excluded from the dataset, which is important to keep in mind when

interpreting the results.


133

An increase of the minimum commuting distance towards or from a

certain area is usually indicative of the spatial expansion of the residen-

tial market or a trend of concentration in the labour market, while a

decline usually will point to a greater functional mix. The underlying

causes may be both of spatial development nature (sprawl), and of

regional-economic nature (e.g. the economic shift towards service indus-

tries, increased tertiarisation).

According to our hypothesis, the rate at which the minimum com-

muting distance increases is an indicator for the impact of spatial

transformations on the actual average commuting distance. The difference

between the growth rate of the minimum commuting distance and the

growth rate of the actual commuting distance is a measure of the non-

spatial component in the overall growth of the commute. Consequently,

an increase in commuting distances that would develop faster than the

increase of the minimum commuting distances points to a mobility

growth that is relatively independent of changes in spatial structure. In

the opposite case, we would be dealing with a spatial evolution that is in

and by itself responsible for the entire growth in commuter traffic

volume.

To calculate the minimum commuting distance, most authors make

use of one of the minimization algorithms for the so-called “transporta-

tion problem”, as available in various software packages. However, we

applied an alternative method that has been used before for a cross-

sectional analysis of the same study area, and is explained in Boussauw et

al. (2011). This method is conceived as an iterative process that simulates

the behaviour of commuters who simultaneously look for a job closer to

home. The disadvantage of this method is that the achieved “optimum”

remains slightly higher than the mathematical minimum. The advantage

of the method is that the produced origin-destination matrix holds a

much greater realism, since it is not influenced by algebraic tricks that

are applied to find the mathematical minimum but do not make sense in

a real world approach. An example is the allocation of all resident

workers from one zone to jobs in only one corresponding zone. Moreover,

the structure of origin-destination matrices that are produced by the

iterative process on the basis of various data sets (representing different

points in time) is similar, meaning that comparing these matrices using

statistical methods (such as Student’s t-test) makes sense. This is not

necessarily true when applying a standard “transportation problem”

Chapter 4

134

algorithm, where the structure of the resulting origin-destination matrix

depends on the used software package.

Confronting the minimum commuting distance for an area with the

observed (calculated) commuting distance is known as the study of excess

commuting. In practice we implement this confrontation by calculating

the excess rate, defined as the quotient of the mean observed commuting

distance and the mean minimized commuting distance.

The mean observed distance used in the analysis is actually a calcu-

lated (thus not perceived by the respondent) distance. For each

municipality, the centre of gravity (centroid) is determined. Then a

shortest-path matrix is calculated by means of a skeleton file of the

Belgian road network, making the shortest network distance between

each possible pair of municipalities available. Intra-municipal distances

are simulated by taking for every municipality half of the network

distance to the centroid of the nearest municipality. Given that the road

network in 1981 was already very dense, we use the same skeleton file for

the three time periods.

4.7 Results

4.7.1 Evolution of the mean observed

commuting distance

We calculate for each municipality the mean trip length for both outgo-

ing (outbound mean calculated distance: omcd) and incoming (inbound

mean calculated distance: imcd) commuting trips. A weighted average

can be found in Table 4.1, along with the commuting distance as per-

ceived by the respondents (mean perceived distance: mpd).

Table 4.1. Observed commuting trips and commuting distances

1981 1991 2001

number of commuting trips

(matrix)

1,950,477 2,331,090 1,907,197

mcd (matrix) 12.2 kma 12.6 kmb 14.7 kmb

mpd (Belgium) 14.6 kma 17.2 kmb 19.0 kmb a source: Mérenne-Schoumaker et al. (1999), p. 80 b source: Verhetsel et al. (2007), p. 60


135

The important differences between the calculated (mcd) and the observed

mean commuting distance (mpd) can be explained as follows:

• mcd is calculated based on the shortest network distance while the

shortest route is usually not the fastest, and thus not the path that is

chosen by the commuter;

• mcd is calculated from centroids of municipalities, which is an

important simplification, particularly with regard to intra-municipal

trips;

• mcd ignores interregional, usually relatively long, commuting trips;

• mpd includes commuting trips in the south of Belgium, which are

longer on average.

Other possible biases in perceived trip length are discussed by Witlox

(2007).

Taken together, the mcd variable should therefore be regarded as a

theoretical measure, which only makes sense when compared to similar

calculations, such as the minimum commuting distance. For comparison

with survey data from other countries, mpd will naturally be better

suited. More important than the absolute figure is the trend that both

quantities exhibit.

Fig. 4.2. Evolution of the commuting distance (1981-2001) based on

municipality of residence

Chapter 4

136

Fig. 4.3. Evolution of the commuting distance (1981-2001) based on

municipality of workplace

A closer look reveals that over the period 1981-2001 mcd increases in 280

of the 309 municipalities when considering the outbound commute (Fig.

4.2), and increases in 291 of the 309 municipalities when looking at the

incoming commute (Fig. 4.3). Although municipalities with a relative

growth show a certain spatial clustering, the virtually general upward

trend suggests that at least some of the causes of the growth in trip

length are of non-spatial nature.

4.7.2 Evolution of the mean minimum

commuting distance

Given the significant variation in the number of recorded trips between

the three snapshots (1981, 1991 and 2001), we first examine whether this

sample variation affects the calculated mean minimum commuting

distance (mmid). We did this by weighing commuting trips for the year

1981 and 1991 so that the three tables on which the minimization

procedure was then performed all contain 1,907,197 trips (equalling the

number of trips in the origin-destination matrix for 2001, see Table 4.1).


137

The obtained results are thereupon compared with the results that are

based on the unweighted tables. No significant differences were found,

and Pearson’s correlation coefficient between the two sets of spatially

disaggregated results is 1.00. Next, the global mmid for the entire study

area is calculated for the three different points in time. The obtained

figures are presented in Table 4.2, along with the evolution of mcd, and

the maximum commuting distance (mmad).

Table 4.2. Evolution of the global minimum commuting

distance and excess rate

1981 1991 2001

mmid 9.0 km 8.9 km 9.4 km

evolution mmid (base 1981) -1.4% +4.3%

mcd 12.2 km 12.6 km 14.7 km

evolution mcd (base 1981) +3.4% +20.1%

excess rate (mcd/mmid) 1.36 1.43 1.57

evolution excess rate

(base 1981)

+4.8% +15.1%

mmad 19.5 km 18,6 km 21.6 km

evolution mmad (base 1981) -4.5% +10.8%

A salient element in Table 4.2 is the negative evolution of mmid over the

period 1981-1991 (-1.4%), which is associated with an increase of mcd

(+3.4%). So, the studied residential locations and job locations would

have come 1.4% closer together over this period, while commuting

distances continuously increased. Nevertheless, over the entire period

under investigation (1981-2001) we find, as expected, an increase in

mmid.

The second striking result is the growth rate of mmid: this is much

lower (+4.3%) than the growth rate of mcd (+20.1%). Logically, this

trend is in line with an increase of the excess rate. This result suggests

that over the studied period the average worker became less inclined to

seek a job close to home, or to look for a home close to work. This also

means that only a small part of the increase in commuter traffic can be

attributed to the expansion of the spatial structure.

The evolution of mmad confirms the observed trend of mmid, al-

though the differences are more pronounced. For a theoretical approach

to the possible interpretation of the differences between mmid and mmad,

Chapter 4

138

which is beyond the scope of this paper, we refer to Charron (2007) and

Yang (2008).

Note that in our literature overview only three papers (i.e. Frost et

al., 1998, Horner, 2007 and Yang, 2008) studied time series of minimum

commuting distance. The results of Horner (2007) and Yang (2008) are

analogous with our findings: in Tallahassee, in Boston and in Atlanta

both minimum commuting distances and observed commuting distances

grew, while the former increased more slowly than the latter. Frost et al.

(1998), however, found for all analysed British cities that the growth of

the minimum commuting distance was faster than the growth of the

observed commuting distance, suggesting that changes in the spatial

structure could be held fully responsible for the increase in commuter

traffic volume, along with an improved efficiency in the commute itself

(given the reduction of the excess rate). In contrast to our research,

however, Frost et al. (1998) limited their study area to demarcated cities,

to which only incoming commuting trips were added. So, the spatial

structure of the commuter area around the considered cities was not fully

grasped. Within these demarcated cities urban sprawl hardly occurs, and

also the growth in traffic itself is far more constrained by congestion than

is the case in the suburban areas. While this alternative approach may

explain the difference in results, apparently great caution in interpreting

the results is needed.

4.7.3 Evolution of the spatially disaggregated

minimum commuting distance

A second step in the analysis is the calculation of spatially disaggregated

values of mmid. For each municipality and for each of the three points in

time mmid was calculated twice: once for the outgoing commute (ommd)

(these are the outgoing and internal trips together) and once for the

ingoing commute (immd) (these are the incoming and the internal trips

together). So, for each municipality we obtain two time series.

Then, for each time series the existence of a clear trend for the period

1981-2001 was examined. First, a Student’s t-test was applied to compare

columns and rows of the produced origin-destination matrices. Non-

significant differences (significance level: p < 0.05) are considered as a

status quo. The decision rules that were used to determine the existence

of a trend are presented in Table 4.3. As a general principle it is assumed

that a trend is only acknowledged if the evolution over the period 1981-


139

1991 is not contradictory to the evolution over the period 1981-2001.

Moreover, the differences have to be statistically significant in at least

one of the two periods.

Table 4.3. Decision table trend

evolution

1981-2001

evolution

1981-1991

evolution

1991-2001

conclusion

+ + + / - / not sig. +

+ - + 0

+ not sig. + +

- - + / - / not sig. -

- + - 0

- not sig. - -

not sig. + / - / not sig. + / - / not sig. 0

For the municipalities where a significant trend was found, the absolute

differences in mmid over the period 1981-2001 were mapped (Fig. 4.4 and

Fig. 4.5). The following applies:

ommd81-01 = ommd01 - ommd81 (4.5)

and immd81-01 = immd01 - immd81 (4.6)

4.7.4 Outbound minimum commuting distance

by municipality

Fig. 4.4 shows the evolution of the minimum commuting distance over

the time frame 1981-2001 for those municipalities that contain more

working residents than jobs. Municipalities with a job surplus are omitted

since the applied method implies that the minimum commuting distance

for workers living in such a municipality remains constant as long as the

job surplus exists. Following issues on the map stand out:

• Most municipalities in the economic core around the triangle Ant-

werp-Brussels-Leuven (A) show a decrease of ommd. Exceptions are

some municipalities with a more rural character that are located on

the edge of the conurbation and have received a large share of hous-

ing suburbanization (Kapellen (B), Nijlen and Berlaar (C), Zemst

(D), Oud-Heverlee (E), Beersel and Sint-Pieters-Leeuw (F)).

Chapter 4

140

• The former mining region in the Kempen (G) shows a sharp increase

of ommd, perhaps due to economic transformation processes that

have led to a decline of job supply in this area.

• In some areas apparently the construction of the motorway has

indeed led to a suburban housing development that entailed only few

jobs. This is clearly the case along the E40 between Ostend and

Ghent (H), and in some municipalities in the Voorkempen (Brecht

and Zoersel (I)) and Limburg (such as Maasmechelen (J), Riemst

(K), Heers (L)).

• In other municipalities also a lot of employment developed in the

proximity of the motorway. This is the case in Leie valley south of

Ghent (M), or in Lummen (N).

Fig. 4.4. Evolution of the minimum commuting distance (1981-2001)

based on municipality of residence

4.7.5 Incoming minimum commuting distance

by municipality

Fig. 4.5 shows the evolution of the minimum commuting distance over

the period 1981-2001 for those municipalities that contain more jobs than

working residents. In this case, municipalities with a surplus of workers

are omitted since the applied method implies that the minimum commut-


141

ing distance for workers employed in such a municipality remains con-

stant as long as the worker surplus exists. Following issues on the map

stand out:

• The value of immd increases in almost all non-peripheral cities, and

the Brussels region has the strongest growth. This means that these

cities need to cover an expanding recruitment area to have their

available jobs occupied.

• A number of cities in the more remote areas, however, show a

reduction of the minimum commuting distance. Thus, in these places

the concentration of employment is proportionally shrinking. This is

the case in the cities Kortrijk (O), Ostend (P) and Bruges (Q) in the

west, and in the cities of Hasselt and Genk (R), Sint-Truiden (S) and

Tongeren (T) in the east.

Fig. 4.5. Evolution of the minimum commuting distance (1981-2001)

based on municipality of workplace

4.7.6 General interpretation and possible biases

The results of our analysis seem to indicate that at the macro level the

distance between job locations and residential locations of the working

population did not dramatically increase, and that commuter traffic

Chapter 4

142

shows an autonomous growth that can only partially be explained by

changes in the spatial structure.

There are two, possibly complementary, explanations for the rela-

tively limited detected increase of the average distance between home and

work locations. First, jobs have partly followed the suburbanization

process of housing. Sprawl does not simply consist of new residential

allotments, but equally of new business parks. Moreover, many new jobs

engrafted onto the additional residential areas, e.g. new employment that

developed in schools, childcare, local health care, supermarkets and public

services in growing municipalities. At the macro level, suburban areas are

relatively multifunctional: sprawl does not necessarily imply the absence

of a functional mix, even if it is at a lower density in comparison with the

city. Nevertheless, the regional job market remains more spatially

concentrated than the housing market. This is the case in many industrial

activities, but in particular the growth of specialized services (such as

financial services, technology and consultancy) led to an increase in the

number of jobs in the Brussels region. This growth explains a significant

portion of the found increase in the general minimum commuting dis-

tance.

A second, perhaps equally important reason is the large-mesh nature

of the used dataset. Our analysis cannot possibly grasp the sprawl that

occurs within a municipal boundary. Although, given the rather small

area of an average municipality, suburbanization certainly plays a role

too at an intermunicipal scale level, it appears that the intramunicipal

share of these transformation processes cannot be detected from a low

resolution data set. To identify sprawl on a lower scale level time series

data at a much denser spatial aggregation level is needed. Thus, the

discrepancy between our analysis and the morphological analyses men-

tioned above is mainly due to the difference in scale. It is likely that the

increase in the minimum commuting distance, and therefore the expan-

sion of the functional space, would be better reflected in the figures when

examining this micro level. This is due to the modifiable areal unit

problem (MAUP) of which the consequences for the minimum commuting

distance were discussed above. What is certain is that the use of a fine-

mesh data set leads to lower absolute figures for the minimum commuting

distance. On the basis of cross-sectional approximate data for 2007

available for the same study area (Flanders and Brussels) at the level of

(fine-meshed) traffic analysis zones, and using the same algorithm as we

did, Boussauw et al. (2011) found a mean minimum commuting distance


143

of 6.9 kilometres (to be compared with our calculated 9.4 kilometres

(2001)). Unfortunately, time series data are not available at this micro-

scopic scale level, so it is not possible to verify our assumption that

spatial separation may be mainly taking place at this smaller geographi-

cal scale level.

The inability to outline developments based on fine-mesh data might

lead to an overrepresentation of the impact of regional economic trans-

formations in the results, pushing the effect of the spatial shift in the

housing stock to the background. The decline of some industrial activi-

ties, the development of logistics in the port areas and the general shift

towards service industries play perhaps a more important role at the

studied scale level with respect to spatial proximity between home and

work locations than the suburbanization of the housing stock does.

When we consider the spatially disaggregated evolution of the mini-

mum commuting distance, we notice that the major conurbations in the

economic core of Flanders and Brussels gain importance in terms of

employment, while a reduction of jobs in more peripheral industrial

sectors has led to a local increase of the minimum commuting distance.

The phenomenon of residential suburbanization in the municipalities

along the motorways has led only here and there to spatial separation,

particularly where jobs did not follow the spatial shift in housing.

4.8 Conclusions

The applied method detects spatially disaggregated evolutions in mini-

mum commuting distance, identifying local increases or decreases in

spatial separation between home and work locations at the level of the

municipality. The results show that there is indeed a general loss in

spatial proximity between housing and jobs for the study area in its

entirety, although the pace of this separation process is on average much

lower than the growth in observed commuting distances. It is found that

the minimum commuting distance increased in many municipalities,

especially where population is growing faster than job supply, or where

traditionally high concentrations of employment still increase. Further-

more, there are also municipalities where a decrease is noticed, especially

in suburban areas that are getting a more urban character by acquiring a

considerable functional mix.

Chapter 4

144

The development of the motorway network has undoubtedly contrib-

uted significantly to the growth of actual commuting distances. In those

municipalities where mono-functional planning practice has facilitated

residential development with access to the motorway, the minimum

commuting distance increased, with negative consequences for the spatial

proximity between residential and work locations. In municipalities where

a better balance between different functions was achieved, the develop-

ment of multifunctional sprawl has not necessarily led to an increase of

the minimum commuting distance at the supra-municipal scale level.

Nevertheless, the influence of sprawl on commuting behaviour seems

to be only secondary to the effects of regional-economic transformations,

which for example led to the loss of employment in the Kempen region

and an increase in employment concentration in Brussels and (to a lesser

extent) in Antwerp. Still, in spatial and economic planning it is impor-

tant to ensure the local balance between jobs and inhabitants as good as

possible. Horner and Murray (2003) argue that the most effective way to

do this is raising residential density in areas with a decent job supply:

through the deliberate reallocation of workers’ residences a significant

decrease of the minimum commuting distance can be attained. However,

Yang (2008) shows that job decentralization may also be responsible for

the growth of the excess rate itself, and thus for a weakening land use-

transport connection. From this finding Yang (2008) argues that a policy

of ensuring a good job-housing balance is insufficient: concentrations of

employment both in cities and in suburbs should take the form of com-

pact, yet relatively large, centres and subcentres.

In order to draw valid conclusions, the degree of detail and the con-

sistency of the used data is crucial. We find that the use of municipalities

as a spatial entity is suitable to grasp regional transformations of the

economy, but is far from perfect to detect sprawl in the morphological

sense. If the same analysis could be repeated at a lower scale level,

probably more concrete guidelines for spatial planning could be given.

This is particularly true when mode choice would be taken into account

too. In that case, concentration of activities around the stops of public

transport would remain equally important. For cyclists, at the other

hand, proximity and the availability of infrastructure are also of interest

at the micro level.


145

References



42(9), pp. 1537–1547.

Antrop, M. (2004) “Landscape change and the urbanization process in

Europe.” Landscape and Urban Planning. 67(1-4), pp. 9-26.


Review. 70(3), pp. 313-338.

Bleijenberg, A. (2003) “The driving forces behind transport growth and

their implications for policy”. In: Managing the Fundamental Drivers

of Transport Demand: Proceedings of the International Seminar.

Paris: OECD Publishing.

Boussauw, K., T. Neutens and F. Witlox (2011) “Minimum commuting



press.



43(5), pp. 580-591.



and Planning A. 39(5), pp. 1238-1254.

Ewing, R. H. (1994) “Characteristics, causes, and effects of sprawl: A

literature review.” Environmental and Urban Studies. 21(2), pp. 1-15.

Frost, M., B. Linneker and N. Spence (1998) “Excess or wasteful com-

muting in a selection of British cities.” Transportation Research Part

A. 32(7), pp. 529-538.

Gordon, P. and H. W. Richardson (1989) “Gasoline consumption and

cities: A reply.” Journal of the American Planning Association.

55(3), pp. 342-346.

Gulinck, H., S. Meeus, J. Peymen, L. Stalpaert, K. Bomans, V. Dewael-

heyns and S. Heremans (2007) Milieurapport Vlaanderen,

Achtergronddocument Versnippering. Aalst: Flemish Environment

Agency.

Horner, M. W. (2004) “Spatial dimensions of urban commuting: A review

of major issues and their implications for future geographic research.”

The Professional Geographer. 56(2), pp. 160-173.

Chapter 4

146

Horner, M. W. (2007) “A multi-scale analysis of urban form and commut-

ing change in a small metropolitan area (1990-2000).” Annals of

Regional Science. 41(2), pp. 315-332.

Horner, M. W. (2010) “Exploring the sensitivity of jobs - housing statis-

tics to imperfect travel time information.” Environment and Planning

B. 37(2), pp. 367-375.



Horner, M. W. and A. T. Murray (2003) “A multi-objective approach to

improving regional jobs-housing balance.” Regional Studies. 37(2),

pp. 135-146.

Layman, C. and M. W. Horner (2010) Comparing methods for measuring

excess commuting and jobs-housing balance: an empirical analysis of

land use changes. Transportation Research Board Annual Meeting

2010. Washington, DC.





Mérenne-Schoumaker, B., H. Van der Haegen and E. Van Hecke (1999)

Algemene Volks- en Woningtelling op 1 maart 1991: Werk- en

Schoolpendel. Brussels: Statistics Belgium.

Murphy, E. (2009) “Excess commuting and modal choice.” Transporta-

tion Research Part A. 43(8), pp. 735-743.





O’Kelly, M. E. and W. Lee (2005) “Disaggregate journey-to-work data:

Implications for excess commuting and jobs-housing balance.” Envi-


O’Kelly, M. E. and M. A. Niedzielski (2008) “Efficient spatial interaction:

Attainable reductions in metropolitan average trip length.” Journal



Urban Studies. 34(8), pp. 1215-1235.

Rietveld, P. and R. Vickerman (2003) “Transport in regional science: The

‘death of distance’ is premature.” Papers in Regional Science. 83(1),

pp. 229-248.


147





Sultana, S. and J. Weber (2007) “Journey-to-work patterns in the age of

sprawl: Evidence from two midsize southern metropolitan areas.” The

Professional Geographer. 59(2), pp. 193-208.

Van Wee, B., P. Rietveld and H. Meurs (2002) A Constant Travel Time

Budget? In Search for Explanations for an Increase in Average

Travel. Amsterdam: Vrije Universiteit Amsterdam.

Vanneste, D., I. Thomas and L. Vanderstraeten (2008) “The spatial

structure(s) of the Belgian housing stock.” Journal of Housing and

the Built Environment. 23(3), pp. 173-198.

Verhetsel, A., I. Thomas and M. Beelen (2010) “Commuting in Belgian

metropolitan areas: The power of the Alonso-Muth model.” Journal

of Transport and Land Use. 2(3), pp. 109-131.



gium.



15(3), pp. 172-183.

Yang, J. (2008) “Policy implications of excess commuting: Examining the

impacts of changes in US metropolitan spatial structure.” Urban

Studies. 45(2), pp. 391-405.

Yang, J. and J. Ferreira (2009) “Informing the public of transportation-

land use connections.” Transportation Research Part C. 17(1), pp.

27-37.

Chapter 4

148

149

Chapter 5:

Excess travel in non-

professional trips: Why

looking for it miles away?

This paper will be published as Boussauw, K., V. Van Acker and F.

Witlox (2011) “Excess travel in non-professional trips: Why looking for it

miles away?” Tijdschrift voor Economische en Sociale Geografie. Copy-

right © The authors. All rights reserved.

Abstract

Based on the spatial distribution of some quasi-daily destination classes

and survey-reported trip distances, regional variation in excess travel in

non-professional trips in Flanders (Belgium) is assessed. To this end,

proximity to various quasi-daily destinations is compared with the

reported distance that is actually travelled to reach similar, but alterna-

tive, facilities.

We note that in rural areas (compared with urban areas) larger dis-

tances are travelled, although the closest facility is chosen more often. In

the most urbanized areas, however, we note that spatial proximity is also

an important aspect in destination choice.

Quantification of these phenomena can support the practice of sus-

tainable spatial planning by distinguishing areas that are too mono-

functional or too remote, and therefore need more functional diversity,

and by identifying areas where densification is useful because the location

is close to most quasi-daily destinations, reducing the need to travel over

large distances.

Keywords: travel behaviour; Flanders; excess commuting; sustainable

spatial development; accessibility

Chapter 5

150

5.1 Introduction

Næss (2003) reduces the relationship between spatial proximity and

mobility to its geometric essence: in an area with a high density of people

and services, distances that are to be covered between potential origins

and destinations are small. As the trip distance is related to the amount

of energy required, empirical research not surprisingly shows that fuel

consumption for transport per capita is actually lower in areas with a

high density than in regions with a low density (Newman and Kenwor-

thy, 1989 and 1999; Næss et al., 1996). This is a logical consequence, as

Næss (2003) states: “The absence of any such influence would also have

been quite sensational.” But reality is obviously more complex than a

geometric problem. All kinds of factors, such as infrastructure configura-

tion, routes of public transport, or the lack of parking space, are

distorting this obvious logic. But also an unbalanced spatial mix, often

caused by functional city planning, may cancel out the positive potential

of high density. Moreover, mode choice plays a role: cities with a high

proportion of pedestrians, cyclists and public transport users will have

less traffic problems. Besides, on the regional level a clear linear relation-

ship exists between fuel consumption and the number of kilometres

travelled per person (Boussauw and Witlox, 2009).

The main deviation between travel behaviour and geometry is due to

the fact that a high degree of spatial proximity, and thus better accessi-

bility, gives rise to new needs (Næss, 2003). Gains, in terms of both time

and money, yielded by a higher level of accessibility are partly offset by

the individual who will make use of the increased choice range. When the

nearest supermarket is located just 100 m from one’s front door, then the

threshold for visiting the second nearest supermarket, at e.g. 500 m, is

particularly low, certainly when the latter offers more products or is a

little cheaper. But if the nearest store is located at 10 km, and the second

nearest is only at 20 km, the same person will for sure go shopping in the

closest store. Although the use of a wider range of accessible destinations,

as well as an increase in the number of trips may offset the potential

efficiency gains of the compact city, the aforementioned empirical studies

suggest that this is only partially the case.

Handy et al. (2005) argue that the reason for the emergence of a trip

can be situated along a choice-necessity continuum. Driving around just

for fun (Ory and Mokhtarian, 2005) is located at the “choice” end of this

continuum, while buying a loaf of bread at the bakery around the corner

Excess travel in non-professional trips: Why looking for it miles away?

151

is at the “need” end of the same spectrum. In these terms, excess travel is

defined as travel beyond what is required for household maintenance,

given choices about residential location, job location, and activity partici-

pation. The required trip length is then defined as the shortest route to

the closest destination possible. The ratio between choice and necessity

determines the excessive nature of any particular trip, meaning the extent

to which the trip distance exceeds the distance to the closest facility that

could possibly satisfy the need of the traveller. The degree of spatial

proximity between the sites that are potential origins or destinations of

trips defines the actual travelled distance. But spatial structure itself is

also one of the factors that influence the decision. In an area with many

options nearby, “choice” will outweigh “need”. Furthermore, in this first

area - with a wide choice range - the total trip length might still be less

than in the second area - where there is little choice.

Vilhelmson (1999) develops a framework that classifies activities

based on the degree of flexibility in terms of physical location and point

in time. Education is an example of an activity where both location and

time are specified exactly. For an activity like jogging the reverse is true.

It are the activities in this last quadrant that are largely determined by

the choice of an individual, and thus may lead to excess travel. Vilhelm-

son (1999) finds that access to a car, not having children and having a

part-time job is associated with an increase in the number of kilometres

travelled during this kind of “free” activities. It is evident that also the

type of trip will play a role in the degree of excess travel. Horner and

O’Kelly (2007) quote the example of the difference between shopping

trips for “comparison goods” (which are only occasionally purchased) and

“convenience goods”. In the latter category, attempts to minimize the

distance travelled will play a greater role than in the first category. Based

on interviews, Næss (2006) shows a willingness to travel longer distances

for work, education and visiting family or friends in comparison with e.g.

schools, kindergartens and grocery shops.

The excess commuting research framework offers opportunities for

studying this phenomenon at a regional scale. In recent decades, excess

commuting has become a major study topic within the discipline of

transport research (Ma and Banister, 2006). The excess commute is that

share of the commute flow (in terms of physical distance or time dis-

tance) that cannot be attributed to the spatial separation between job

locations and residential locations of employees, and is thus rooted in the

travellers’ freedom of choice.

Chapter 5

152

The main goal of our research is examining regional variations in the

relationship between the length of quasi-daily trips and spatial proximity

by applying an excess commuting approach. The paper is structured as

follows. First, we provide a summary of the excess commuting literature

and extend the concept to non-professional travel. Second, we develop a

methodology to define theoretical minimum non-professional trip lengths

and to distinguish spatial categories within reported trip lengths. Subse-

quently, results are obtained by comparing the reported distance travelled

with the theoretical minimum distance travelled, within each spatial

category. Finally, we draw conclusions from our findings and derive

recommendations for sustainable spatial planning practice.

5.2 Excess commuting and excess travel

The concept of “wasteful commuting” or “excess commuting” was first

introduced by Hamilton (1982). Hamilton defined excess commuting as

the difference between the actual commuting distance and the theoretical

minimum commuting distance, suggested by the spatial structure of the

considered city. The attention paid by Hamilton (1982) to minimized

commuting distances stems from the successive oil crises of 1973 and

1979-1980, when the availability, and in particular, the affordability of

fossil oil products was at stake. Daily trips over large distances were

suddenly considered problematic, because of their particularly high energy

consumption and costs.

As transport research progressed, the concept of excess commuting

was extended and applied in different ways. The line of inquiry that was

started by White (1988), compares the spatial structure of different cities

on the basis of the minimum required commute; a method that was later

expanded with the idea of the maximum possible commute (Horner,

2002). Both concepts are measurable properties of spatial structure,

which can be applied not only to compare the morphology of cities, but

also to examine time series and thus measure suburbanization and

evolutions in commuting behaviour (Horner, 2007; Boussauw et al.,

2011b). Further, we can distinguish between the more economically-

inspired research direction that uses travel time as a variable, and the

more environmental approach that focuses on travel distance (Ma and

Banister, 2006). In both cases the minimum commuting distance may be

considered as a measure of proximity, in terms of accessibility (when


153

travel time is studied), or in the sense of spatial proximity (when physical

distance is studied).

An interesting extension of this is the spatially disaggregated ap-

proach where the minimum commuting distance is mapped, considering

this variable as a measure of spatial proximity. Again, we can distinguish

between analyses that are based on time distance (an approach intro-

duced by Giuliano and Small, 1993), and more environmentally-oriented

studies where physical distance is used as a measure (Niedzielski, 2006;

Yang and Ferreira, 2008; Boussauw et al., 2011a). When this type of

research is conducted at a regional scale, it may contribute significantly

to the sustainability of proposed land developments, and to the detection

of regions that are vulnerable because of their extreme remoteness. This

approach is, among other, relevant in the light of the peak oil theory. In

the course of history, the cost of transport has shown a nearly continuous

downward trend, with only a ripple at the time of oil crises. Since today

transport relies almost entirely on finite fossil fuels, we suspect that one

day the cost of transport will evolve in the opposite direction, increasing

systematically as oil supplies decline. The sudden, albeit temporary, surge

in oil prices in 2008 seemed to forecast this hard reality. But even though

there is little point in thinking in doomsday scenarios, it remains a fact

that over time highly car-dependent spatial structures may be particu-

larly vulnerable to oil shortage (Dodson and Sipe, 2008).

To date, the excess commuting research framework (hereafter ex-

tended to excess travel) has to our knowledge only been applied on the

study of the home-to-work commute. There is no doubt about the

primordial economic importance of the commute, which represents a

significant proportion of the number of car kilometres travelled. In

Flanders, the home-to-work commute represents 18.6% of trips. Yet, the

average commuting trip length amounts to 19.0 kilometres, which is

much higher than the average trip length of 12.5 kilometres (for all

purposes combined) (Zwerts and Nuyts, 2004). Moreover, we know that

commuting trips are much less price elastic and thus more inert than

other trips. All these arguments emphasize the importance of studying

commuting behaviour.

In contrast, in the western world the share of commuter traffic in the

overall mobility is decreasing. Leisure travel, and by extension: tourist

trips, are on the rise. In essence, this evolution originates from the ever

growing prosperity and the improved accessibility of high speed travel

modes for an increasingly larger share of the population. Even if the

Chapter 5

154

penetration of the private car and the coverage of the motorway network

would have reached its structural limits, we still continue to use more

and more often aircrafts and high speed trains for recreational and tourist

trips. In the continuum of Handy et al. (2005), these trips are situated

near the “choice” end, and are much less emerging by “necessity”. The

changeable nature of the destinations and the consequent difficulties in

data acquisition represent perhaps the real reason why researchers have

not yet ventured to the study of excess leisure travel, and have thus

confined themselves to the study of the home-to-work commute. Similar

reasons can be found concerning the study of other non-professional trips,

such as shopping, home-to-school travel or visiting public services.

However, non-professional trips did not entirely escape attention in

the excess commuting discourse. Recently, Horner and O’Kelly (2007)

suggested that the study of excess travel for non-professional, but more or

less daily trips could become an interesting extension, possibly shedding

more light on the relationship between non-professional travel behaviour

and spatial structure. Examples are innumerable: bringing children to

day-care or school, doing the groceries, or going to sports or hobby clubs.

The study of non-commuting trips, however, entails considerable meth-

odological problems. Following problems can be identified immediately:

• The capacity of many of the mentioned facilities is deemed elastic,

compared with employment centres that are characterized by a rela-

tively constant number of jobs.

• Many non-professional trips are made frequently, but not daily.

• There are often multiple destinations for one purpose, as is e.g. the

case for multipurpose shopping trips (Handy, 2001).

• Many leisure trips are part of a trip chain that partly includes

commuting, making the distinction between professional and non-

professional travel vague.

• Commuting trip lengths are the result of a “double” selection proce-

dure, being a combination of the preference of an employer for an

employee, and the preference of the worker for the job. In contrast,

non-commuting trips are based on a “single” selection procedure,

comprising only the customer’s preference for the visited facility.

Consequently, excess commuting rates will by definition be higher,

and thus not comparable with values found for non-commuting trips.

• For some destination classes the trade-off between the accessibility of

the widest possible range of customers and the cost of an additional

establishment is inherent in the location of the facility. This is the


155

case with branches of large chain stores, which are often sited based

on a facility location model.

• There is often no area covering sample data available on travel

behaviour of individuals and families, so it is not possible to aggre-

gate data within relatively small traffic analysis zones. Moreover,

available data often contains only information on reported trip

lengths, without mentioning the address of the visited facilities. In

the latter case it is only possible to estimate excess travel that is gen-

erated by residences (an important origin of many trips) in a

particular area, and not e.g. by stores or schools (examples of desti-

nations of non-professional trips).

The conventional calculation of the minimum commuting distance

involves origin-destination matrix optimization through techniques of

linear programming (White, 1988). This method assumes that destina-

tions (i.e. job locations) have a fixed capacity, and do not adapt in case

demand would change. In contrast with jobs, however, daily service and

facility destinations do not fulfil these conditions, an issue that is also

recognized by Fan et al. (2010). Therefore, we need to adapt the defini-

tion of excess travel to the context of non-work-related destinations.

To our knowledge, Fan et al. (2010) are the only authors who have

analyzed a case of excess travel in non-commuting trips. They approach

the concept of excess travel as the difference between the distance

travelled by a household to get through its activity programme (given the

current destinations and travel frequencies), and the distance that would

be travelled in order to achieve the same activity programme in case

home location would be optimized in a geometrical sense (meaning that

the considered family would move house to a location that is more

centrally relative to their activities).

In contrast, our own approach, which we explain into detail in the

next section, compares proximity of destinations, considered as a spatial

characteristic of the studied location, with the travel behaviour of its

residents and users. We assess the ratio between the actual distance

travelled and the distance that would be travelled in case for any trip

purpose an alternative, nearby destination would be chosen. Summariz-

ing, the definition of excess travel in non-commuting trips by Fan et al.

(2010) is household-oriented, while our definition is location-oriented.

Both approaches are therefore complementary.

A final aspect worth mentioning is the comparability of results be-

tween various analyses. It is known that the magnitude of the obtained

Chapter 5

156

values in excess commuting research are largely subject to the modifiable

areal unit problem (MAUP), making attempts to compare results of

research in areas with surveys based on a different zoning system virtu-

ally meaningless (Horner and Murray, 2002). This problem is even

exacerbated in the study of non-commuting excess travel: here, the extent

to which a destination category is representative of a particular destina-

tion is determining the obtained results too, and will additionally bias

comparisons.

5.3 Methodology

Our research takes the Flemish Region and - to some extent - the Brus-

sels Capital Region, together constituting the northern half of Belgium, as

a study area. The study consists of several stages. Given the relatively

scarce data it is impossible to use traffic analysis zones as spatial units,

which would be in line with the spatially disaggregated study of excess

commuting. Therefore we first have to define the spatial classes that

should be distinguished. We do this by mapping the transition between

more and less urbanized areas as accurate as possible.

Then we develop a non-professional equivalent to the minimum com-

muting distance in the form of a proximity map. This is done by defining,

for each statistical ward (corresponding with a neighbourhood) and for

each defined spatial class, the minimum distance that should be covered

in order to reach all facilities that are visited by an average Flemish

household during a week. Obviously, this method implies some simplifica-

tions. The choice of the number of destination categories or travel

purposes is limited by the amount of available data. So we need to

consider those facilities for which data is available as representative for a

certain category of destinations. This choice is to a certain extent arbi-

trary. Furthermore, the activity pattern of a household is also determined

by the spatial context in which this family lives, reducing the assumption

of an average activity pattern that is applicable to every household to a

major simplification. However, assuming an average household travel

pattern, regardless of the location of residence, is a deliberate choice we

make in order to map variations in spatial proximity in an objective way.

So, the average activity pattern of a resident of the Flanders region is

used as a reference in order to quantify proximity as a spatial characteris-

tic.


157

In a third phase, we examine the effective distance travelled by

households during non-professional trips, based on available travel survey

data, while distinguishing our predefined spatial classes and destination

categories. Eventually we compare reported travel distances with calcu-

lated minimum distances, resulting in a ratio that represents a measure

for excess travel. We consider variation of spatial proximity and excess

travel as a spatial characteristic, indicating how sustainable a certain

physical structure is in relation to non-professional trips.

5.4 Determination of spatial classes

Unlike data on commuting, which is based on a census (SEE 2001), data

for non-professional trips is in Flanders only available in the form of a

sample (Travel Behaviour Survey for Flanders 2000-2001 (Onderzoek

Verplaatsingsgedrag Vlaanderen (OVG)) (Zwerts and Nuyts, 2004). This

means that we cannot make a spatially continuous analysis on the basis

of a map for the whole studied region, as opposed to the study of excess

commuting (Boussauw et al., 2011a).

However, we want to relate the observed travel behaviour to different

types of spatial structures. To retain a survey sample that is large enough

for every spatial class, we look for a meaningful spatial classification for

the Flanders region. We base our argument on the existing literature.

Depending on the point of view, two major formats exist. Based on

empirical data (a combination of morphological characteristics and data

on commuting and migration flows) Luyten and Van Hecke (2001) assign

each municipality to one of the following four categories: urban agglom-

eration (“agg”), suburban (“sub”), commuter area (“comm”) and rural

(“rur”) (which is a residual category with very limited urban characteris-

tics). We will call this the “urban region” classification. The urban

agglomeration consists of those municipalities where more than half of the

population lives in an urban core or in the urban fringe, characterized by

a continuously built-up environment. The suburban area is the outer zone

of the city, characterized by an extensive, rural morphology, combined

with an urban functionality. Agglomeration and suburban area together

compose the urban region. The commuter area is attached to the urban

region and relies on this urban region for an important part of its em-

ployment. The demarcation of the different classes follows the municipal

Chapter 5

158

borders, causing some loss of accuracy. An overview is represented in Fig.

5.1.

Fig. 5.1. Spatial classification according to the “urban region” approach

(Luyten and Van Hecke, 2001)

The Spatial Structure Plan for Flanders (Ruimtelijk Structuurplan

Vlaanderen (RSV), 1997/2004) offers a second classification, which is

much more policy-oriented. This means that this format does not only

take into account the current situation, but incorporates also a vision for

future development, which is adopted by the Flemish government. The

main direction of development, favoured by RSV, is indicated by the

dichotomy between urban areas and outlying areas. Urban areas are those

areas that should receive most of the additional housing and businesses,

and are very accurately delineated (on the ward level). The selection

makes a distinction between metropolitan areas (“ma”) (agglomerations

with more than 300,000 inhabitants: Antwerp and Ghent, and the part of

the Brussels agglomeration that is located in Flanders), regional urban

areas (“rua”) (between 50,000 and 150,000 inhabitants) and small urban

areas. Within the latter category a distinction is made between “structure

supporting small urban areas” (“ssua”) (being relatively important

attraction and development poles) and “small urban areas at provincial

level” (“psua”) (being a development pole of minor importance).


159

Fig. 5.2. Spatial classification according to the RSV approach

(Ruimtelijk Structuurplan Vlaanderen, 1997/2004)

The demarcation of urban areas is based on a consultation process and is

consolidated on the basis of cadastral boundaries. For a number of urban

areas this demarcation process is still ongoing. We used a provisional

definition, translated to the ward level. For simplicity, we consider

everything that is not within the definition of any urban area as outlying

area (“oa”). In the outlying area we can still distinguish selected residen-

tial nuclei (“noa”), generally corresponding to villages. In total we

distinguish thus six categories within the framework offered by RSV. A

cartographic overview is shown in Fig. 5.2.

Since the two proposed classifications have their pros and cons, we

decided to include both systems in our analysis.

5.5 Developing a proximity map

5.5.1 Method and selection of destinations

In a first phase of our research we develop a method to quantify the

proximity of non-professional quasi-daily destinations. We use the ward

as a spatial unit, considering the centre of gravity (centroid) of the ward

Chapter 5

160

as a starting point to calculate the network distance to the closest

appropriate facility. The 9708 wards which cover Flanders and Brussels

are therefore regarded as residential locations. Used network data consist

of the seven highest categories of the Streetnet skeleton file, which

contains almost all passable connecting roads.

We select 18 types of facilities for which a location dataset is avail-

able. Each facility type is judiciously assigned to one of the travel

purposes that are applied in the OVG. We only used the purposes that

are non-professional (“work” and “business visit” are thus excluded) and

have a destination for which alternative locations can be found (so the

purposes “walking/driving around” and “visiting someone” were not

considered, just as the indefinite “other purpose”). The remaining

purposes are: shopping (SHP), education (EDU), picking up/taking

something/someone (PCK), leisure/sports/culture (LSC), services (e.g.

medical doctor, commercial bank) (SRV).

The various selected facility types, data sources and links with OVG

purposes can be found in Table 5.1. The purpose “shopping” is repre-

sented by three classes of supermarkets and some more specialized types

of shops. For the interpretation of the purpose “education” the higher

grades of secondary education and higher education (in general: education

for students over 14 years old) were not taken into account, since we

consider these facilities as too specialized and thus rather being part of

the commute. The purpose “leisure/sports/culture” is represented by

cafés, restaurants, sports centres and cinemas, while the category “ser-

vices” is represented by medical doctors and banks. We construct the

purpose “picking up/taking something/someone” by a combination of

education and leisure/sports/culture, supplemented with nursery. This is

of course only an approximation, where we assume that mainly children

are taken to and collected from their activities.

With regard to the quality of the data we mention that the data re-

trieved from Google Maps (2009) is based on commercial information

which is less complete than the other data sets that were used (Federal

Public Service of Economy (2009), Ministry of Education (2009), Child &

Family (2009), Cinebel.be (2009)), which claim to be exhaustive. The

location data from Google Maps include geographic coordinates. The

other data sets used consist of address lists, which were geocoded with

the help of Yahoo! Maps Web Services (2009). Finally, we calculated the

network distance between each ward’s centroid and the nearest location

within each selected type of facility using Dijkstra’s shortest path algo-


161

rithm (implemented in the Closest Facility Tool of ArcGIS Network

Analyst) as follows: Owf

Oiwf

Owf TTfiT min,,min, : ≥∈∀ (5.1)

in which:

Tmin = minimum trip length

w = ward

f = type of facility

i = any possible destination belonging to type of facility f

O = indicates that the spatial unit is always considered

as the base (origin) of the trip

Table 5.1. Selection of facilities and purposes

type of facility n source purpose OVG

baker’s 3747 Google Maps

supermarket class 1

(hypermarket)

54

supermarket class 2

(supermarket)

1484

supermarket class 3

(superette)

869

clothes shop 660

do-it-yourself shop 199

household appliances

(electrical)

180

Federal

Public

Service of

Economy

shopping

kindergarten 2913

primary school 2861

middle school (1st

grade high school)

681

adult education 111

Ministry of

Education

education +

picking up/taking

something/someone

nursery 2844 Child &

Family

picking up/taking

something/someone

café/bar 4746

restaurant 6907

sports centre 1581

Google Maps

cinema 49 Cinebel.be

leisure/sports/culture

medical doctor 9713

commercial bank 3391 Google Maps services

Chapter 5

162

5.5.2 Weighting

The closest facility calculation provides a proximity map for each type of

facility, showing the accessibility, in terms of physical distance, from

every considered ward. However, we want to limit the number of maps to

the five mentioned OVG purposes, and ultimately we want one summary

map. We calculate spatial proximity per spatial class and per travel

purpose as follows:

n

T

T

n

w

Owf

Osf

∑=

=1

min,

min, (5.2)

in which:

s = spatial class

n = number of wards in spatial class

To join the proximity of different purposes into one map it is necessary to

assign weights to the various facility types:

Owf

m

ff

OwH TaT min,

1min, ∑

=

⋅= (5.3)

in which:

H = average weekly haul

af = weight by type of facility

m = total number of facility types

The weighting is determined by the weekly visit frequency to the respec-

tive facilities by an average Flemish household. As a starting point we

take the number of trips per household per purpose, as reported in the

OVG. Overall an average household generates 42.95 trips per week, of

which 23.26 meet our criteria.

Based on a number of other data (demographic statistics and market

research), we estimate the average visit frequency. Visit frequencies of

some destinations are extrapolated to fit in with the OVG data (visit

frequency per OVG purpose). This means that some facility types are

considered as representative for similar destinations: e.g. clothes shops,

do-it-yourself and household appliance shops together are considered

being representative for non-food specialist shops. Estimated visit fre-

quencies are shown in Table 5.2. A trip is seen as a single move, which on

average should be partially attributed to a trip chain. Based on OVG we

expect that visiting a facility generates on average 1.68 trips, since often

more than one facility is visited within one trip chain. The purpose


163

“picking up/taking something/someone” in the table was reduced to the

facility type “nursery”. For comparison with reported trip lengths this

purpose is extended pro rata with the purposes “education” and “lei-

sure/sports/culture” (see below).

Table 5.2. Estimated weekly visit frequency by type of facility,

per household

purpose OVG

• representative facility

# trips/household-

week

shopping 8.99

• bakery 2.11

• supermarket class 1 (hypermarket) 0.69

• supermarket class 2 (supermarket) 3.26

• supermarket class 3 (superette) 0.42

• clothes shop 0.84

• do-it-yourself shop 0.84

• household appliances (electrical) 0.83

education (without higher secondary and

higher education)

3.09

• kindergarten 0.68

• primary school 1.70

• middle school (1st grade high school) 0.52

• adult education 0.19

picking up/taking some-

thing/someone (limited to nursery)

0.34

• nursery 0.34

leisure/sports/culture 9.00

• café/bar 1.84

• restaurant 5.95

• sports centre 1.00

• cinema 0.21

services (e.g. doctor, bank) 1.85

• medical doctor 0.62

• commercial bank 1.23

SUM 23.26

Chapter 5

164

5.5.3 Mapped proximity per spatial class

By mapping the result of equation (5.3) for every ward, we get a

weighted proximity map (Fig. 5.3). This map provides an overview of the

spatial variation in the minimum distance, expressed in kilometres, that

should at least be covered by an average Flemish household to complete

its weekly programme.

Fig. 5.3. Weighted proximity map for Flanders and Brussels (shortest

average weekly haul per household, for selected facilities)

To compare these calculated minimum distances with the distances

reported in the OVG, we calculate the average values for each spatial

class: both the values per purpose (Figs. 5.4 and 5.5) and the weighted

values (Fig. 5.6). Note that the Brussels Capital Region was omitted in

these tables in order to match the study area of OVG.

In general, shorter minimum trip lengths seem to be associated with

higher degrees of urbanization, even if not all distinguished classes are

useful in describing this phenomenon (e.g. the distinction between the

classes “sub” and “comm” is not expressed in our findings). An ANOVA

test, applied to both spatial classifications, indicates that minimum

distances significantly differ among spatial classes (significance level of

0.01) and thus confirms the observed general trend.


165

Fig. 5.4. Estimated minimum distance by purpose (km)

(urban region classification)

Fig. 5.5. Estimated minimum distance by purpose (km)

(RSV classification)

Fig. 5.6. Estimated shortest average weekly haul per household (km)

Chapter 5

166

As expected, agglomerations and metropolitan areas score points in terms

of spatial proximity. The differences between suburban and commuter

areas are minimal, as well as the differences between the structure

supporting small urban areas and small urban areas at provincial level.

The rural area and outlying area (including the residential nuclei in the

outlying area) score poorly, as was also expected. In particular, the

spatial classification according to RSV shows a systematic increase of the

minimum distances to be covered when we move from a more urbanised

to a less urbanised area.

5.6 Reported trip lengths

5.6.1 Data

Data on the effective length of trips made by the inhabitants of the

respective spatial classes are obtained from OVG. This survey reports on

travel behaviour of a sample of 3028 households over two consecutive

days. The sample does not contain households from Brussels (which is

administratively not part of the Flemish Region) and Ghent (for which a

separate survey was conducted). For our study we added a random

selection of data from Ghent to the Flemish data.

We selected only those trips originating from or ending at the resi-

dence of the respondent, with a destination or origin corresponding to one

of the five selected purposes. Thus, we do not only consider tours from

home to the facility and back, but also parts of trip chains between the

house and the facility. For each trip we know the reported distance

travelled, the ward where the respondent resides, and thus the spatial

class in which the residence is located. However, we do not know the

location of the visited facility. This trip selection method is a deliberate

choice, aiming to retain a maximum amount of useful information from

the available data. Nevertheless, this selection entails some specific biases.

Trips of respondents making complex trip chains may be under-

represented, while the distance travelled in retained parts of trip chains

may be less representative. Furthermore, this method assumes that the

most efficient trips have their destination close to home. However, in case

the respondent is a commuter, also a destination that is chosen nearby

the job location may lead to a very efficient trip. In the latter case it

should not be excluded that some of the reported excess travel, e.g. in a


167

shopping trip, is in fact due to the fact that the respondent works far

from home. It is important to keep in mind these possible biases when

interpreting the results of the analyses.

Since we are looking for quasi-daily travel behaviour, trips that cover

extremely large distances are considered as outliers. We eliminated

outliers per purpose, while setting the threshold at three standard

deviations above the mean trip length. Table 5.3 shows the remaining

number of observations after selection.

Table 5.3. Summary of the retained trips and matching OVG purpose

n SHP EDU PCK LSC SRV

agg 1977 777 936 1377 413

sub 816 427 530 483 166

comm 1273 623 669 794 308

rur 1632 835 904 1141 352

ma 941 365 451 636 203

rua 850 334 390 556 188

ssua 411 171 158 222 96

psua 281 119 143 202 61

noa 1990 1022 1104 1291 429

oa 1225 651 793 888 262

5.6.2 Method

To link up the minimum distance to be covered and the reported trav-

elled distances (Witlox, 2007), we follow two different approaches, in

parallel. First we calculate the average reported distance per purpose, per

spatial class:

q

T

T

q

r

Orsp

Oobssp

∑=

=1

,

, (5.4)

in which:

Tobs = average reported distance per purpose

Tr = reported length of trip r

p = purpose

q = total number of reported trips based in spatial class s and with

purpose p

Chapter 5

168

In a next phase we compare this with the minimum distance to be

covered, according to equation (5.2).

This procedure gives an indication of the influence of the degree of

proximity of a certain type of facility on the actual distance travelled to

reach a similar facility. However, the information obtained in this way is

not sufficient if we want to understand the relationship between sustain-

ability of travel patterns and spatial structure. In this second case also

information about trip frequencies is important, since it is conceivable

that people who make relatively short trips will compensate their benefit

- in terms of time and costs - by making more trips, cancelling out gains

in fuel consumption (as an indicator of sustainability).

Therefore we calculate the average distance travelled by a household

during one week by adding up the distances covered by all selected

reported trips per spatial class, and extrapolate this sum to a time frame

of one week:

t

T

T

t

h

q

r

Orsh

OobssH

∑∑= =

⋅=1 1

,

, 5.3 (5.5)

in which:

h = individual household

t = total number of households in spatial class s and included in the

survey

(factor 3.5 extrapolating the two-day survey to a time frame of 7

days)

In a next step we will compare this distance travelled with the weighted

average minimum distance to be covered by an average household to

satisfy its needs (equation (5.3)).

5.6.3 Results

For the observed (reported) values we also calculate averages and display

these by spatial class. Again, the ANOVA test indicates that significant

differences between spatial classes exist (p = 0.00, except for leisure trips

by RSV class, where p = 0.03). The averages for each purpose are shown

in Figs. 5.7 and 5.8, while the total observed distances (extrapolated to a

full week) are shown in Fig. 5.9.

Figs. 5.7 and 5.8 give a rather surprising picture. The differences be-

tween the spatial classes are much smaller than what we might expect

based on the major differences in spatial proximity. For most purposes,


169

the least urbanized classes do not necessarily provide the greatest dis-

tances travelled: the largest trip lengths are rather recorded in the

suburban and commuter areas. Also the minor differences between the

metropolitan and regional urban areas stand out. Despite the result of the

ANOVA test, the often wide confidence intervals also suggest that the

link between observed trip length and spatial class is rather weak.

Pairwise t-tests indeed show non-significant distinctions between some

pairs of classes, confirming the picture provided by Figs. 5.7 and 5.8.

Fig. 5.7. Reported trip length by purpose (km)

(urban region classification)

Fig 5.8. Reported trip length by purpose (km) (RSV classification)

Chapter 5

170

Fig. 5.9. Reported weekly haul length per household (km)

The results that are shown in Fig. 5.9 take into account the spatial

differences in trip frequency. Although, while assessing the results, we

must not forget that our selection method takes no account of trips that

do not have either their origin or their destination at the respondent’s

home. In environments that would be associated with trip chains that are

more complex than average this could lead to an underestimation of the

number of trips per household. Yet, the literature is not unanimous on

this issue. According to Krizek (2003), a high degree of urbanization is

associated with more, but less complex, trip chains, while Maat and

Timmermans (2006) found a higher tour complexity in more urbanized

areas. Moreover, complex trip chains are usually very efficient tours,

conducting a series of activities within a minimum tour distance.

When we consider the “urban regions” classification, it appears that

agglomerations, as expected, yield the shortest average weekly haul. The

commuter area - thus not the rural area - yields the longest weekly haul.

When examining the RSV classification, metropolitan areas constitute the

shortest weekly haul. The weekly hauls are much longer in the regional

urban areas, but still shorter in comparison with the small urban areas

(the structure supporting small urban areas in particular). In the outlying

area, we record the longest weekly haul.

These relations are quite consistent with British research, recording

the shortest travel distances in the major British cities (> 250,000

inhabitants), except London. Small towns and rural areas score poorly

(Banister, 1999). Our study adds a new element: commuter areas which

fit morphologically with the rural area but are still within the sphere of

influence of the agglomeration are scoring worse than the more remote

“real” rural area.


171

Parts of the results are probably explained by the utilization of the

available choice potential due to spatial proximity. This is more often the

case in metropolitan and regional urban areas, as they are both function-

ally and morphologically more urbanized than the rest of Flanders. At a

lower geographical scale the structure supporting small urban areas

generate more mileage than the small urban areas at provincial level.

However the nature of the former class is more urban than the second.

Perhaps the structure supporting small urban areas are functionally more

focused on the larger cities.

5.7 Excess travel

We find that the influence of spatial proximity does not work out in the

same way for each spatial class. We examine this phenomenon in analogy

with excess commuting research methods. We define excess travel as the

difference between the minimum distance that must be covered to visit

the desired type of facility (e.g., the nearest supermarket) and the

observed distance covered. The observed trip length is always larger than

the minimum trip length because of non-spatial factors that are deter-

mined by e.g. personal preferences, transport cost, price differences

between similar facilities, or the organization of trip chains. We choose to

express this difference as the ratio between the minimum distance to be

covered and the observed distance travelled. This ratio is called the

excess rate:

Osp

OobsspO

spE

EE

min,

,= (5.6)

OsH

OobssHO

shE

EE

min,

,= (5.7)

in which:

Ep = excess rate by purpose

EH = excess rate for an average weekly haul

As shown by equations (5.6) and (5.7), we calculate excess rates in two

ways. First we determine per purpose and for each spatial class the

relationship between the reported average distance of a trip with the

considered purpose (as shown in Figs. 5.7 and 5.8), and the minimum

distance to be covered to reach a similar destination (as shown in Figs.

5.4 and 5.5). This excess rate by purpose is shown in Figs. 5.10 and 5.11.

Secondly, we determine the ratio between the total reported distance

travelled, extrapolated to a full week (shown in Fig. 5.9), and the

Chapter 5

172

weighted minimum weekly haul (shown in Table 5.6). This weekly haul

excess rate is shown in Fig. 5.12.

Fig. 5.10. Excess rate by purpose (urban region classification) (ratio)

Fig. 5.11. Excess rate by purpose (RSV classification) (ratio)

Fig. 5.12. Weekly haul excess rate (ratio)


173

In Figs. 5.10 and 5.11 we note for most purposes a systematic downward

trend of the excess rate when we watch the various spatial classes in

order of decreasing urban nature. In short, this means that a high degree

of spatial proximity is only partially reflected in short trip lengths,

because the increased choice of possible destinations generates compara-

tively long journeys. In metropolitan areas the average household goes

shopping almost three times further from home than strictly necessary,

while in the outlying area this rate amounts to one and a half only. Thus,

a higher degree of spatial proximity creates greater choice, compensating

for a significant proportion of the potential gains (in terms of external

costs caused by traffic). Noteworthy is that the differences in excess

travel between the spatial classes are rather small. This is in contrast to

what was found previously in the case of excess commuting (Boussauw et

al., 2011a), with very high values in urban areas, compared to very low

values in rural areas. Regarding differences between purposes, we notice a

very high degree of excess travel in leisure trips, where strong personal

preferences play. This applies to some extent also for trips to services,

although the low visit frequency and the overall high degree of spatial

proximity (there is a doctor in every street, say) play their role in the

obtained excess rate.

When assessing the excess rate of a combined weekly haul, then the

downward trend is not evident anymore. Following the spatial classifica-

tion according to the RSV, regional and small urban areas report the

highest values of excess travel. The rates are much lower in the metro-

politan areas, which is mainly explained by a larger share of chained

trips. Although this puts the low excess rate of metropolitan residents

somewhat into perspective, it also means that the activity pattern of this

group is already very efficient. A household in a regional urban area

covers a weekly distance that is more than 12 times longer than the

minimum distance required by our model. By contrast, a household living

in the outlying area covers only 6 times the minimum required distance.

We can interpret the excess rate as a measure that indicates to what

extent a travel pattern can be made more efficient, given the spatial

context. In this case gaining efficiency means shortening travel distances

by choosing similar destinations closer to home, within the existing

spatial configuration of housing and facilities. Such an adjustment of

households’ travel pattern may happen in case transport would become

more expensive, e.g. by a severe congestion policy or a stringent environ-

mental policy, or by energy scarcity. In the outlying areas, distances are

Chapter 5

174

relatively great, and the excess rate is low. This means that those areas

are most vulnerable to a price increase in transport. In regional urban

and small urban areas, the distances are not only smaller, there is also

more margin leaving the possibility to choose destinations closer to home.

In the metropolitan areas, non-professional trips seem comparatively

efficient, apart from being short anyway.

5.8 Possible biases in the results

The discussion above assesses the number of kilometres travelled per

household. However, there are substantial differences between average

households in the various studied spatial classes. It is generally assumed

that households in urban areas are relatively small, while household

income peaks in the suburban areas near large cities. These factors may

play a role in the sustainability of travel behaviour, even though the

precise effect is often unclear.

A larger number of family members may lead to more kilometres

travelled per household. Yet, within these families carpooling occurs more

often, while children travel only few kilometres independently. Thus,

calculated per person, larger households are expected to produce less

kilometres. When examining the influence of spatial structure, calculation

of the number of kilometres travelled can be justified both per household

and per person, albeit from different viewpoints.

Household income plays a role too. At the macroeconomic level, there

is a linear relationship between income and the number of kilometres

travelled (Schafer and Victor, 1997). If there would exist significant

income differences between the various spatial classes, it would make

sense to control for this variable. Determining income, however, raises

additional methodological problems. It is for example possible that the

effect of a higher income in an urban environment is primarily reflected in

increased tourist travel, and not in longer daily journeys (Holden and

Norland, 2005). Moreover, we are also downplaying car ownership, an

intermediary variable that is influenced both by income, by the surround-

ings (supply of alternative transport means and parking) and by

household size (Van Acker and Witlox, 2010).

These assumptions add much more complexity to the study of the

role of spatial structure in the sustainability of travel behaviour. Should

we measure the distance travelled per household or per person? Is it


175

useful to take income into account, and if so, how do we tackle this issue?

To avoid oversimplification, we did not incorporate these variables in a

statistical analysis. However, below we shed more light on the possible

role of spatial structure in itself in relation to the mentioned issues by

providing a few basic figures.

5.8.1 Household size

Some Flemish policy plans argue that the average family size is smaller in

urban areas than in suburban and rural areas, and consider this phe-

nomenon as a social problem (Boudry et al., 2003, p. 114). It is also

assumed that the phenomenon of shrinking household size occurs more

rapidly in the urban areas, increasing the identified problem. Champion

(2001) paints a more balanced picture. Small households, particularly

one-person households, are only rarely based on the stereotypical young

career maker with a very urban lifestyle. Young singles stay continuously

longer living in the parental home, while more one-person households

than before are the result of a divorce, and are in many cases located in a

suburban area. In contrast, in western city cores especially the immigrant

population is keeping average family size at a relatively high level. For

Flanders (2006) we find the following values (Table 5.4):

Table 5.4. Average household size per spatial class (2006)

class (urban regions) n

agg 2.19

sub 2.55

comm 2.44

rur 2.48

class (RSV) n

ma 2.13

rua 2.23

ssua 2.24

psua 2.33

noa 2.50

oa 2.61

In urban agglomerations and metropolitan areas households are actually

smaller than average. Yet, an exploratory linear regression, trying to

explain weekly distance travelled by variation in family size does not

yield any significant results. When we calculate the average number of

kilometres travelled per person, rather than per household, then differ-

ences between spatial classes (as shown in Fig. 5.9) are somewhat smaller.

When using the spatial classification according to RSV, structure sup-

porting small urban areas stand out more, making travel patterns of the

Chapter 5

176

inhabitants of this class now appearing the least sustainable. The outlying

areas and small urban areas at provincial level follow shortly. The

agglomerations and metropolitan areas still score best.

5.8.2 Income

By way of illustration we examine the average household income, based

on the assessment forms of direct taxation for the year 2006. The avail-

able data are aggregated for each municipality. This level of aggregation

allows us to regroup the data according to the “urban region” classifica-

tion, but not to the RSV classification. It is important to keep in mind

that here too the Brussels Capital Region is not included in the analysis.

Table 5.5. Average household income by spatial class (2006)

class (urban regions) €

agg 28448

sub 28950

comm 26778

rur 24862

There seems to exist a slightly downward trend when we move from more

urbanized into less urbanized areas. This is particularly the case when we

would take into account household size (income per person). Yet, also in

this case an exploratory linear regression does not yield any significant

effect of the level of income per household or per person on the weekly

number of kilometres travelled. We conclude therefore that the macro-

economic theory, arguing that higher income results in more kilometres,

does not apply at the regional scale of our study area. This finding

provides an additional argument for the proposition that spatial structure

indeed plays an important role in the genesis of travel behaviour.

5.9 Conclusions

The excess commuting research framework proves to be very useful in

examining the relationship between non-professional trips and spatial

structure. By mapping the minimum distance that an average household

needs to cover in order to complete its weekly programme, we get an idea

of the variation in spatial proximity between housing and facilities. This


177

combined minimum weekly haul varies from 5 km to 392 km, depending

on the residence location. This wide range of spatial proximity classes

indicates that the distinction between more and less urbanisation is still

reality and remains important in terms of mobility. The proximity map

we obtained in this way (Fig. 5.3) can be used as a guidance for new

developments. Additional densification of areas where the degree of

spatial proximity is already high, or areas that are immediately adjacent

to these, make an excessive increase of newly generated traffic the least

likely. In areas with a relatively low degree of spatial proximity, the

situation could be improved by planning a better functional mix for the

future. Yet, additional housing in areas characterized by a low degree of

spatial proximity will generate more traffic. These findings are in line

with what Banister (1999, p. 318) suggests: “New development should be

of a substantial size and located near to (or within) existing urban areas

so that critical size thresholds can be achieved.”

As stated in the introduction, spatial proximity is only one aspect of

the overall picture. The degree to which choice behaviour is driven by

spatial structure is equally important. Fig. 5.9 shows that the relationship

between spatial proximity and the number of kilometres travelled is not

linear. Residents of agglomerations, metropolitan and regional urban

areas travel over relatively short distances, but the inhabitants of the

suburban and commuter areas and small urban areas appear to have a

less sustainable travel pattern than is suggested by the rather urbanized

spatial structures in which these people live. What is also surprising is

that these variations cannot be explained by differences in family size or

income.

From research on spatially disaggregated excess commuting we know

that less urbanized areas are characterized by a higher minimum com-

muting distance along with a lower excess rate. In other words, residents

of rural areas go to work further from home than urban residents, but

they opt more often for the closest job they can find (Boussauw et al.,

2011a) than city-dwellers do. By analogy, we expected to find a similar

phenomenon in the study of non-professional trips. Based on Fig. 5.12 we

see that this expectation is only partially confirmed. In particular metro-

politan areas, agglomerations and suburban areas are characterized by a

relatively low excess rate, indicating that residents of these areas are still

heavily influenced by spatial proximity when choosing their non-

professional destinations. Possibly, modal choice has to do something

Chapter 5

178

with this: in an urban environment more non-professional trips are made

on foot or by bike, slow transport modes for which trip length is crucial.

We conclude that spatial structure and degree of urbanization is of

great importance to spatial proximity and length of non-professional

trips. Particularly in metropolitan areas, but also in regional urban areas

or the suburban areas that are adjacent to both urban classes, households

look for non-professional activities relatively close to home, especially

when it is possible to walk or bike there in a pleasant way. By attributing

a role to this aspect in spatial planning practice, generation of additional

traffic can be avoided and the vulnerability of spatial developments to

more expensive transport (by rising fuel prices, congestion problems and

congestion policy) can be reduced.

References


Review. 70(3), pp. 313-338.

Boudry, L., P. Cabus, E. Corijn, F. De Rynck, C. Kesteloot and A.

Loeckx (2003) De Eeuw van de Stad - Over Stadsrepublieken en Ras-

tersteden. Brussels: Ministry of the Flemish Community.

Boussauw, K., T. Neutens and F. Witlox (2011a) “Minimum commuting



press.




search. 11(1), pp. 42-60.



43(5), pp. 580-591.

Champion, A. G. (2001) “A changing demographic regime and evolving

polycentric urban regions: Consequences for the size, composition and

distribution of city populations.” Urban Studies. 38(4), pp. 657-677.



Studies. 23(3), pp. 377-401.


179

Fan, Y., A. Khattak and D. Rodríguez (2010) “Household excess travel

and neighbourhood characteristics: Associations and trade-offs.” Ur-

ban Studies. doi: 10.1177/0042098010369712 pp. 1-19.

Giuliano, G. and K. A. Small (1993) “Is the journey to work explained by

urban structure?” Urban Studies. 30(9), pp. 1485-1500.

Hamilton, B. W. (1982) “Wasteful Commuting.” Journal of Political

Economy. 90(5), pp. 1035-1053.

Handy, S. L. and K. J. Clifton (2001) “Local shopping as a strategy for

reducing automobile travel.” Transportation. 28(4), pp. 317-346.

Handy, S. L., L. Weston and P. L. Mokhtarian (2005) “Driving by choice

or necessity?” Transportation Research Part A. 39(2-3), pp. 183-203.

Holden, E. and I. T. Norland (2005) “Three challenges for the compact

city as a sustainable urban form: Household consumption of energy

and transport in eight residential areas in the greater Oslo region.”

Urban Studies. 42(12), pp. 2145–2166.

Horner, M. W. (2002) “Extensions to the concept of excess commuting.”

Environment and Planning A. 34(3), pp. 543-566.



Horner, M. W. (2007) “A multi-scale analysis of urban form and commut-

ing change in a small metropolitan area (1990-2000).” Annals of

Regional Science. 41(2), pp. 315-332.

Horner, M. W. and M. E. O’Kelly (2007) “Is non-work travel excessive?”


Krizek, K. J. (2003) “Neighborhood services, trip purpose, and tour-based

travel.” Transportation. 30(4), pp. 387-410.

Luyten, S. and E. Van Hecke (2007) De Belgische Stadsgewesten 2001.

Brussels: Statistics Belgium.

Ma, K.-R. and D. Banister (2006) “Excess commuting: A critical review.”


Maat, K. and H. Timmermans (2006) “Influence of land use on tour

complexity.” Transportation Research Record. 1977, pp. 234-241.

Mokhtarian, P. L. and I. Salomon (2001) “How derived is the demand for

travel? Some conceptual and measurement considerations.” Transpor-

tation Research Part A. 35(8), pp. 695-719.

Næss, P. (2003) “Urban structures and travel behaviour. Experiences

from empirical research in Norway and Denmark.” European Journal

of Transport and Infrastructure Research. 3(2), pp. 155-178.

Chapter 5

180

Næss, P. (2006) Accessibility, activity participation and location of

activities: Exploring the links between residential location and travel

behaviour.” Urban Studies. 43(3), pp. 627-652.

Næss, P., S. L. Sandberg and P. G. Røe (1996) “Energy use for transpor-

tation in 22 nordic towns.” Scandinavian Housing & Planning

Research. 13(2), pp. 79-97.









Schafer, A. and D. Victor (1997) “The past and future of global mobil-

ity.” Scientific American. 277(4), pp. 58-61.

Van Acker, V. and F. Witlox (2010) “Car ownership as a mediating

variable in car travel behaviour research using a structural equation

modelling approach to identify its dual relationship.” Journal of

Transport Geography. 18(1), pp. 65-74.

Vilhelmson, B. (1999) “Daily mobility and the use of time for different

activities: The case of Sweden.” GeoJournal. 48(3), pp. 177-185.

White, M. J. (1988) “Urban commuting journeys are not ‘wasteful’.”




15(3), pp. 172-183.

Yang, J. and J. Ferreira (2008) “Choices versus choice sets: A commuting

spectrum method for representing job-housing possibilities.” Envi-

ronment and Planning B. 35(2), pp. 364-378.

Zwerts, E. and E. Nuyts (2004) Onderzoek Verplaatsingsgedrag

Vlaanderen 2000-2001. Brussels-Diepenbeek: Ministry of the Flemish

Community.


181

Data sources

Cinebel.be (2009), Cinebel.be from http://www.cinebel.be/nl/. Retrieved

24/08/2009.

Google Maps (2009), Google Maps from http://maps.google.be/. Re-

trieved 24/08/2009.

Ministry of Education (2009), Onderwijsaanbod in Vlaanderen from http:

//www.ond.vlaanderen.be/onderwijsaanbod/. Retrieved 24/08/2009.

Child & Family (2009), Opvangadressen from http://www.kindengezin

.be/Themas/Kinderopvang/Opvangadressen/opvangadressen.jsp. Re-

trieved 24/08/2009.

Federal Public Service of Economy (2009), Planologie van de handel from

http://mineco.fgov.be/ministry/publications/publ_planologie_nl.htm

. Retrieved 24/08/2009.

Yahoo! Maps Web Services (2009), Yahoo! Maps Web Services from

http://developer.yahoo.com/maps/. Retrieved 24/08/2009.

Chapter 5

182

183

Chapter 6:

Relationship between

spatial proximity and

travel-to-work distance:

The effect of the

compact city

This paper will be published as Boussauw, K., T. Neutens and F. Witlox

(2011) “Relationship between spatial proximity and travel-to-work

distance: The effect of the compact city.” Regional Studies. Copyright ©

Regional Studies Association - Routledge. All rights reserved.

Abstract

In this paper, an assessment is made of the relationship between selected

aspects of spatial proximity (density, diversity, minimum commuting

distance, jobs-housing balance and job accessibility) and reported com-

muting distances in Flanders (Belgium). Results show that correlations

may depend on the considered trip end. For example, a high residential

density, a high degree of spatial diversity and a high level of job accessi-

bility are all associated with a short commute by residents, while a high

job density is associated with a long commute by employees. A jobs-

housing balance close to 1 is associated with a short commute, by both

residents and by employees. In general, it appears that the alleged

sustainability benefits of the compact city model are still valid in a

context of continuously expanding commuting trip lengths.

Keywords: compact city; spatial proximity; commuting; sustainable

spatial development; Flanders

Chapter 6

184

6.1 Introduction

Although the spatial development model of the ‘compact city’ gained

momentum in the planning literature particularly during the 1980s, there

are still many prevailing spatial policy plans that draw from this concept.

One of those is the Spatial Structure Plan for Flanders, which was

adopted in 1997 (Ministry of the Flemish Community, 1997/2004) and

was inspired, among others, by the European Union’s “Green paper on

the urban environment” (Commission of the European Communities,

1990) and the Netherlands’ “Vierde Nota over de Ruimtelijke Ordening

(Fourth Report on Spatial Planning; Ministry of Housing, Spatial Plan-

ning and the Environment, 1988).

In Europe, the concept of the compact city emerged from a visionary

quest for a model of sustainable urban development, based on a city

tailor-made for pedestrians and cyclists, with a relatively high density, a

high degree of functional mix and efficient public transport (Jenks et al.,

1996, p. 5). In North America, the New Urbanism concept can be consid-

ered the counterpart of the compact city model, offering an express

alternative for the typically American, extensive form of suburbanisation

(Ellis, 2002), although there are differences in terms of scale (New

Urbanism occurs usually in small-scale developments that are strongly

oriented towards walking). Apart from the protection of open space and

economics of scale, the motivation for compact city development is to a

large extent grounded on the sustainability of mobility patterns. Encour-

aging trips over short distances and creating spatial conditions that

stimulate walking, cycling and using public transport are key ingredients

of the compact city model. The model has been criticized because of the

potential of larger social problems in residential neighbourhoods with high

densities, the concentration of pollution in living environments and the

increasing risk of congestion (Burton, 2000). However, the benefits of

enhanced spatial proximity and reduced car dependence are rarely

questioned.

In reality, the distinctly demarcated, quasi-walled, compact city has

gradually disappeared since the nineteenth century. The density gradient

from the Alonso-Muth-Mills model (describing the equilibrium between

the distance to the central business district and real estate demand in a

monocentric urban system) corresponds much better to reality, even

though in the post-war Western world it was overtaken by the develop-

ment of urban sprawl, in which historical centres are usually embedded.

Relationship between spatial proximity and travel-to-work distance

185

However, this does not mean that mobility-influencing characteristics

that are attributed to the compact city are completely absent in subur-

banized and even sprawled areas.

Sustainable urban development implies offering more opportunities

for travel over short distances, and encouraging the use of alternative

transport modes (instead of the private car). These two objectives are not

independent. If the demand for long-distance travel is reduced, slower

transport modes, which are usually less environmentally stressful, are

likely to be chosen more often. Boussauw and Witlox (2009) showed that

at a regional scale the distance travelled per person can be considered a

good approximation for the overall sustainability of the commute, since

positive consequences of modal shifts (especially towards train) are often

counterbalanced by increases in average trip length, while a reduction of

trip lengths may result in a modal shift towards low-impact modes such

as cycling and walking. These patterns result in a correlation between

energy consumption (as a sustainability indicator) and distance travelled

that is even stronger than intuitively expected. Furthermore, many recent

policy plans (as in the study area: the Mobility Plan for Flanders and the

Flemish Climate Policy Plan) overemphasize encouraging a modal shift,

while research into the number of kilometres travelled has faded some-

what into the background. This justifies an approach in which the

distance travelled is considered a sole sustainability indicator, although

the authors are well aware that reality is significantly simplified by not

taking into account modal split, congestion and other factors that

influence the environmental impact of travel.

Within the scope of this paper, it is examined whether and to what

extent there is a discernible link between the alleged qualities of the

compact city and the travel patterns of its users on the basis of commut-

ing data for Flanders and Brussels. To this end, the paper focuses not

only on cities, but also on the presence of characteristics attributed to the

compact city model throughout the suburbanized historically polycentric

spatial structure that characterizes this region. More specifically, an

attempt is made to gain a deeper understanding of the observed spatial

variation in trip lengths on the basis of commuting data available for

Flanders and Brussels. The restriction to home-to-work travel is moti-

vated by the need for accurate data: in Belgium, commuting is the only

category of travel that is surveyed area-wide. Also, the study of the

commute is particularly relevant from an environmental point of view,

Chapter 6

186

given that the longest average trip lengths are recorded in this travel

category (for example, Zwerts and Nuyts, 2004).

Although today a large body of literature describes aspects of the re-

lationship between spatial structure and trip lengths, this paper adds

considerably to the issue. One of the major difficulties regarding the

realistic interpretation of previous studies is the uncertainty about the

spatial scale at which the influence of certain spatial characteristics

manifest. Much research has been done on monocentric urban structures,

in a way neglecting the spatial characteristics that reflect the embedded-

ness in the larger region (for example, Cervero, 1996; Peng, 1997; Wang,

2000; Schwanen and Mokhtarian, 2005a). By studying a region that is

characterized by a large spatial diversity, and by comparing a series of

various spatial variables - which account for the characteristics of the

surrounding area in different ways - this paper adds to the general

research framework on proximity and trip length and thus partly ad-

dresses the mentioned difficulty. In this way, the validity of various

computational methods for spatial proximity is examined explicitly.

Moreover, the analysis is consistently applied to both origins of trips (in

this case, residential locations) and destinations (in this case, work

locations) in order to obtain more insight into the role played by the

deviation in spatial distribution between both location patterns through-

out the region. Although the approach adds to the complexity of the

issue, it contributes to the practice of spatial planning by complementing

the classical triad ’density, diversity and design’ that was introduced by

Cervero en Kockelman (1997).

A second important contribution is to be found in the specific atten-

tion that is paid to methodological problems such as the modifiable areal

unit problem (MAUP), spatial autocorrelation, non-linearity and multi-

collinearity. Although this paper does not aim to solve these issues

directly, it does provide additional insight into the influence of the

analytical methods used on obtained correlations and significances. For

example, it is not inconceivable that in previous studies the apparent

non-significance of assumed relationships between spatial phenomena was

related to a non-deliberate choice of the geographical scale.

While the approach disregards personal variations in travel behaviour

that are not inherently spatial - which is indeed a major simplification - a

purely geographical approach to the concept of proximity remains of

major importance. The present argument for this is grounded in peak oil

theory, which dictates that limited oil production will in the long run


187

lead to an important increase in the mileage-related costs of mobility and

give rise to the importance of spatial proximity (Dodson and Sipe, 2008).

The paper is structured as follows. A summary of the literature on

compact city characteristics and commuting trip length is presented in

the second section. A research methodology is developed, taking into

account statistical issues and data limitations, in the third section. The

results are analysed and confronted with the knowledge of the spatial

structure of the study area in the fourth section. The fifth section dis-

cusses theoretical issues of scale level and origin versus destination.

Finally, the sixth section provides concluding remarks and outlines

possible improvements by further research.

6.2 The relevant literature

6.2.1 Characteristics of the compact city

According to Neuman (2005), the most important mobility-influencing

characteristics of the compact city consist of a high density, a high degree

of functional mix and a fine-grained land-use pattern. The rationale

behind the pursuit of high densities stems from the work of Newman and

Kenworthy (1989, 1999), who pointed out that global cities with more

inhabitants per square kilometre (km2) consume less fuel per capita for

transportation needs. A higher density ensures a critical mass of public

transport patronage, but also enhances the potential for spatial interac-

tion between people based on short distances. While Newman and

Kenworthy’s work has been criticized on many occasions (for example,

Mindali et al., 2004), their thesis is still widely supported in spatial

planning practice.

A sound spatial mix of functions is another important, yet often

equivocal, feature of the compact city. In particular, it is unclear at which

geographical scale this mixture can play its full role. In the most extreme

case, a high degree of spatial mix boils down to self-sufficiency of jobs,

shops, schools and other services per neighbourhood. While several

authors have studied the influence of spatial characteristics on the travel

behaviour of residents at this district level (among them, Frank and Pivo,

1994; Cervero and Kockelman, 1997; Crane and Crepeau, 1998; and

Schwanen and Mokhtarian, 2005a, 2005b), it may also be apposite to

observe this kind of spatial mix at the level of a city or a region, at least

Chapter 6

188

with regard to the more specialized functions. That is because a small

grain size may neglect economic agglomeration benefits at larger scales.

Moreover, some activities (for example, heavy industry) obviously need a

lot of space, especially if desired or legally enforced environmental

buffering is to be taken into account.

Although by the end of the 1990s the debate on the compact city was

transformed into a discourse on sustainable urban development (Williams

et al., 2000), density and diversity were retained as important elements.

Cervero and Kockelman (1997) argued that density, diversity and design

are important spatial factors underlying travel behaviour, and they

focused on the scale of the neighbourhood, attributing an important role

to the design of public space. Stead et al. (2000) found that the size of

the city and the proximity to key infrastructure are additional factors

that determine the relationship between spatial structure and travel;

while Van Acker et al. (2007) focused on the explanatory power of a

range of social characteristics.

Based on these findings, it is argued here that high density, a sound

land-use mix and a small grain size are all spatial elements, amenable to

modification by policy, that are part of an overarching spatial quality

that is called here ‘spatial proximity’.

A less frequently discussed, but nevertheless important, aspect is the

spatial scale at which the influence of certain spatial characteristics is at

its maximum. In the literature on the compact city, often no distinction

is made between small towns and metropolises. In Flanders, the distance

covered by an average trip (all purposes combined) amounts to 12.5

kilometres (one way). The average commuter even covers a distance of

about 19 kilometres per ride between home and work (Zwerts and Nuyts,

2004). For the United Kingdom, Lyons and Chatterjee (2008) found an

average commuter trip length of 13.7 km in 2002-2003. According to

Banister et al. (1997), commuter trip lengths increase year after year in

European and North American metropolitan areas. It is patent that the

average trip length no longer corresponds with the scale of a compact

city, at least in a Belgian context.

The scale problem is acknowledged by many authors. Van Wee

(2002), for example, suggested that the findings of Newman and Kenwor-

thy (1989, 1999), which are based on demarcated cities, cannot be applied

to the Dutch Randstad conurbation, as this urban system is operating at

a higher, regional-scale level. Alberti (1999), on the other hand, referred

to several possible approaches to a sustainable spatial structure, stating


189

that at the regional level the settlement pattern is of interest, while at

the local level neighbourhood design is paramount. In short, the scale

problem is well recognized, but not yet explicitly addressed. As will be

shown in the present paper, this is partly due to the fact that the quanti-

tative analysis of scale effects is hampered by substantial methodological

issues.

Based on these arguments, we hypothesize that the main compact

city variables (density and land-use mix) are insufficient to represent the

differences in the spatial distribution of residential locations and job

locations at the scale of the region, or the embeddedness in the surround-

ing area. Therefore, three additional variables that should grasp the issue

in a more comprehensive way are introduced in the third section.

6.2.2 Commuting trip length

Although it may be intuitively understood that spatial proximity is a

predominant determinant for commuting trip lengths, many studies

indicate that personal, economic and behavioural factors play an impor-

tant role too, making aggregate non-spatial variables even prevalent in

explaining commuting trip lengths. Cervero (1996), Peng (1997), and Van

Acker and Witlox (2011) found a positive, albeit relatively limited, effect

of income level on commuting distance. Wang (2003) was more specific:

above a certain income threshold, employees again tend to live closer to

their work, making the discerned relationship non-linear. Other socio-

economic characteristics that were investigated in the literature are

gender, race, number of workers per household, education and property

status of the residence (Wang, 2000). In the same analyses, urban

characteristics (in contrast to rural characteristics) are considerably

associated with short commuting distances. Irrespective of the type of

model, the construction of the spatial proximity variables and the

available data used, generally low coefficients of determination are

obtained, indicating that a number of unknown or non-quantifiable

‘spurious’ variables are responsible for the largest share of the explained

variance.

However, this is not to say that the spatial aspect is unimportant.

Rather, spatial structure is a rigid constraint that is not able to undergo

rapid change. If transport were to become more expensive, a scenario that

is suggested by peak oil theory, then the role of spatial proximity will

undoubtedly gain importance (Dodson and Sipe, 2008). The present

Chapter 6

190

research focuses on the relationship between spatial proximity and

commuting distance by testing the applicability of several possible

indicators on a regional scale.

6.3 Methods

6.3.1 Research design

The aim is to examine relationships between commuting trip lengths and

a number of spatial characteristics of the zones where the surveyed trips

have their origin and destination. Since the analysis focuses on regional

variations, possible relationships are first explored visually by confronting

maps with knowledge of the study area. A quantitative analysis is then

conducted by calculating correlations between the average trip length and

each of the other selected variables. The variables are aggregated within

geographic zones, introducing effects of spatial autocorrelation and the

modifiable areal unit problem (MAUP). To address this problem, the

choice of an aggregation level that is adapted to the analysis is para-

mount.

The two most obvious variables that are included in this research are

directly obtained from the compact city literature. These are density and

spatial diversity. However, three additional spatial parameters must also

be added: the theoretical minimum commuting distance (as a more

sophisticated measure of proximity), the jobs-housing balance (as a

measure of self-sufficiency of a zone regarding job supply), and the

number of accessible jobs (as a measure of accessibility). In the next

section we will explain how these three additional variables have been

calculated. In our analysis, the dependent variable is trip length both for

trips that depart from and arrive in each considered zone.

Relationships are deduced in two steps. First, all variables are visual-

ized at an intermediate aggregation level to obtain an overview. This map

allows for an initial impression of possibly present links. Subsequently, an

exploratory spatial correlation analysis is applied at several aggregation

levels, with the intention of quantifying a number of relationships and

testing their significance. Based on the initial results, the most appropri-

ate aggregation level for the analysis is selected. Finally, expected

multicollinearity in the variables and the assumed linearity of relation-


191

shipsare asessed, the aim being to detect redundancies and thus improve

the interpretation of the results.

6.3.2 Used data and variable construction

In what follows, the origin and composition of the six used variables:

distance travelled per trip, density, diversity, minimum distance travelled

per trip, jobs-housing balance, and number of potentially accessible jobs,

are discussed.

6.3.2.1 Distance travelled per trip (dpt)

To determine the distance travelled per trip, the origin-destination (OD)

matrices of the Flanders Multimodal Model (MMM) are used. The MMM

is a macro-traffic model that is developed since 1998 and is commissioned

by the Flemish government. The matrices provided for this study simu-

late traffic on an average weekday between 04.00 and 11.00 hours

(morning traffic). The zones corresponding with the matrix are in most

places in line with census wards. The matrices are built on the basis of

the 2001 Census, which is an exhaustive survey of the Belgian population

(excluding children under six years of age), assessing the address of

residence and the address of the workplace (Verhetsel et al., 2007). The

processed data are aggregated by ward and they present a picture of the

daily travelled distances to and from each neighbourhood.

To obtain trip lengths, the shortest distance over the road network is

calculated between the centroids of the connecting zones. Note that the

distances calculated in this way are a slight underestimation of the real

distances travelled, since detour factors associated with faster routes or

the public transport network are not included (Witlox, 2007). Since the

fastest route depends on varying congestion levels, the aim is to avoid

added complexity and thus stick to calculating the shortest path. The

average distance travelled per trip )(obsh to and from each traffic

analysis zone i is calculated as follows:

∑∑

=

i

OiO

iO

Hobsh )( (6.1)

∑

∑=

i

DiD

iD

Hobsh )( (6.2)

where O and D are the number of departing and arriving trips; and H is

the distance covered by each trip.

Chapter 6

192

6.3.2.2 Density (dens)

As an approximation for density, the number of departing commuting

trips, as well as the number of arriving trips per square kilometre, is

calculated based on the MMM. Residential density is approximated by

counting the number of outbound trips per square kilometre in the

morning traffic. Job density is approximated by the density of inbound

trips. The densities C are calculated based on the number of outbound

trips O and the number inbound trips D belonging to traffic analysis zone

i with area A as follows:

COi = Oi/Ai (6.3) CD

i = Di/Ai (6.4)

6.3.2.3 Diversity (div)

As a source, the Strucnet file of the NGI (National Geographical Institute

of Belgium, 2009), which contains all buildings represented on the official

Belgian topographic maps with scale 1:10,000, was used. Different

categories of buildings are distinguished, but the accuracy of categoriza-

tion is limited. All buildings that are morphologically part of a group of

houses are listed as ‘ordinary building’. All other as such recognizable

buildings (those used for industry, schools, hospitals, public services et

cetera) have their own feature class. In practice this means that many

commercial functions, offices and services that are interwoven with

housing are not recognizable. Nevertheless, this inventory can be used to

approximate the diversity of functions in a given zone, and is without

doubt the best currently available area-wide data set in Belgium.

To calculate the spatial-functional diversity per zone, the Shannon

index was applied. This index is used in landscape ecology as a measure

of morphological diversity (Nagendra, 2002), and is in this case also called

spatial entropy (Batty, 1974). An extension to urban diversity is obvious.

The Shannon index is calculated as follows:

∑=

⋅−=

N

nnni ppS

1

ln

(6.5)

where N is the number of features included within the considered aggre-

gation zone i; and pn is the proportion of each function that occurs within

this zone.


193

6.3.2.4 Minimum distance travelled per trip (mdpt)

The theoretical minimum commuting distance is introduced as a proxy

for spatial proximity. The minimum commuting distance stems from the

research on excess commuting (Hamilton, 1982; Ma and Banister, 2006,

2007; Charron, 2007), and is in this case considered as a spatial charac-

teristic (Niedzielski, 2006; Boussauw et al., 2011).

The principle of the method implies linking any observed departure

(in this case, in the morning traffic) to the nearest observed arrival, also

in the morning traffic. Per traffic analysis zone the number of departures,

as well as the number of arrivals, is retained, but the existing relationship

between origins and destinations is cut to minimize the total travelled

distance within the system. This theoretical exercise deliberately disre-

gards non-spatial factors that determine the real-world match between

origin and destination. When commuting trips are considered, this means

that everyone who is part of the active population is considered suitable

to perform any job.

Boussauw et al. (2011) developed an algorithm to calculate the local

values of minimum commuting distance, based on the MMM, for the

commute occurring between 04.00 and 11.00 hours. The algorithm obtains

a general minimization of travel distances via local optimization, simulat-

ing individuals pursuing a job closer to home. For each traffic analysis

zone, the calculation was performed twice: once with the zone considered

as an origin (outbound travel), and again with the zone considered as a

destination (inbound travel). The first computed value is seen as an

approximation for the proximity of housing within the zone in relation to

the Flemish and Brussels labour market, while the second value is

representative of the proximity of the labour market of the zone in

relation to housing in Flanders and Brussels. The average minimum

distance to be covered (min)h per trip from and to a traffic analysis zone

i is calculated as follows:

( )

∑

∑=

i

OiO

iO

Hh

min(min) (6.6)

( )

∑∑

=

i

DiD

iD

Hh

min(min) (6.7)

where min(H) is the minimized commuting distance per zone.

Chapter 6

194

6.3.2.5 Jobs-housing balance (jhb)

The jobs-housing balance is the ratio between the number of jobs local-

ized in a zone and the number of working people who live in the same

zone. Because of its simplicity, this indicator is often used as a spatial

characteristic in commuting research (Cervero, 1989; Peng, 1997; Horner

and Murray, 2003; for an overview, see Horner, 2004).

The jobs-housing balance for every traffic analysis zone i, based on

the MMM, is calculated by dividing the number of arrivals in the morn-

ing traffic by the number of departures:

Bi = Di/Oi (6.8)

6.3.2.6 Number of potentially accessible jobs (potjob)

The accessibility of the job market is also included in the analysis, and is

expressed as the number of jobs located in Belgium that can be reached

from every zone. Accessibility, departing from a particular zone, is

determined by combining a probability curve with the travel time

between the considered zone and all other zones in Belgium, after which

the values for all these other zones are summed. The probability distribu-

tion was deducted from empirical data; while the travel time was

calculated on the basis of a network with attributed impedances. The

basic data on the location of jobs date from 2001 and were provided by

the Statistics Belgium (2009) and is aggregated by borough (that is, a

municipality in the former administrative system).

The number of jobs accessible from a borough i is defined as follows:

( )∑=

⋅=

n

jijji cFJA

1

(6.9)

where Ai is the number of jobs accessible from the considered borough i;

Jj the number of jobs located in each borough j; and F(cij) is the imped-

ance function, based on the modelled probability curve.

The calculated results from Vandenbulcke et al. (2007, 2009) are

adopted, and these two papers are used for the method of exact calcula-

tion. This accessibility index is related to the jobs-housing balance, but

also takes into account the travel time needed to reach the core of the

labour market.


195

6.3.3 Level of aggregation and MAUP

In order to discover correlations, the various parameters must be aggre-

gated within the same zonal classification. The traffic analysis zones of

the MMM are not suitable for this procedure. These zones are small in

high-density regions and vice versa, which apparently results - among

other objections - in a non-normal distribution of the variables linked to

these zones. The large dispersion in area size makes the traffic analysis

zones unsuitable for calculation of the Shannon index, which is area

dependent, and also the calculation of the number of accessible jobs was

initially based on another zoning area (boroughs). For the sake of uni-

formity, we choose to aggregate all data into a grid of square cells.

However, the choice of the size of these cells is not evident. It is a

well-known phenomenon that statistical correlations change whenever a

different level of spatial aggregation is chosen, and that these become

generally stronger when the aggregation level increases, even if accuracy

of the data is in fact lost by increasing the aggregation level (Amrhein,

1995). This mechanism is called the scale effect of the MAUP, and is

inherent in any quantitative analysis of spatially aggregated data.

Openshaw and Taylor (1979) argue that the choice of the aggregation

level should depend on the expected significance for the studied variables.

However, this suggestion does not solve the question. The main variable,

average trip length, shows important dispersion. Even if one chooses the

mesh size of the grid so that the average trip has its destination in a zone

adjacent to the zone of origin, this will often not be the case for trip

lengths deviant from this average. In this context, ‘mesh size’ points to

the length of a side of a cell in a uniform square grid. Regarding the

independent variables it is even less clear what an appropriate level of

aggregation would be.

To support the choice of a suitable aggregation level, four different

grids with mesh sizes of 1, 4, 8 and 16 km were applied in the exploratory

stage. These options are illustrated in Figs 6.1-6.4 for one example (that

is, trip length). The results of a spatial correlation analysis (see below)

indicate that the absolute values of the coefficients increase and that

significances weaken along with an increase of the aggregation level. This

finding confirms the expected influence of the MAUP.

The highest level of aggregation (16 x 16 km) seems inappropriate

because of the large number of non-significant relationships that are

revealed by means of spatial regression, due to the reduced dataset size.

Chapter 6

196

The lowest level of aggregation (1 x 1 km) also seems inapplicable: the

relatively high degree of discontinuity between adjacent cells makes it

sometimes difficult to clearly distinguish local patterns, resulting in often

very low coefficients. Therefore, for the current analysis, an intermediary

level of aggregation with a mesh size of 4 km was chosen (Fig. 6.2). If one

takes into account a general detour factor of 1.40 (Rietveld et al., 1999;

Witlox, 2007), then this choice implies that about 70% of all commuters

have their destination in another zone than the origin zone of their trip

(Zwerts and Nuyts, 2004).

Fig. 6.1. Average commuting distance (dpt), origin zones,

aggregation level = 1 km


197





Chapter 6

198



For the construction of the grid, the border zones of the study area are

joined so that any cell size approaches the area of a square cell. The

aggregation of those data that were originally available at the level of

traffic analysis zones (that is, travelled distance, density, minimum

commuting distance and jobs-housing balance) or borough (access to

jobs) is done proportionally to the geographical overlap between two

geographical divisions. The Shannon index (diversity) is calculated once

for the grid with a mesh size of 1 km, and then aggregated to the applied

scale level.

6.3.4 Spatial regression

Given the nature of the examined spatial characteristics, an important

degree of positive spatial autocorrelation occurs between the zones

themselves. Spatial autocorrelation is the correlation between values of a

variable that has its origin in the vicinity of the locations where these

values are measured. Positive spatial autocorrelation, meaning that

neighbouring areas are similar, is a phenomenon that is typically present

in most empirical spatial datasets, and is explained by Tobler’s (1970)

first law of geography, in that ‘Everything is related to everything else,


199

but near things are more related than distant things.’ (p. 236) The

presence of spatial autocorrelation is a violation of the assumption of the

independence of successive observations, which must be satisfied before

applying classical statistical techniques such as correlation and regression

analysis. Positive spatial autocorrelation leads too often to positive results

of significance tests, and thus to the unjustified rejection of null hypothe-

sises, and to an overestimation of regression and determination

coefficients (Anselin and Griffith, 1988).

Like the MAUP, spatial autocorrelation has often been ignored in

past spatial planning studies. One of the reasons for the neglect of these

issues in research applications is that the theoretical study of useful

spatial alternatives to traditional statistical methods has been debated for

a long time (Cliff and Ord, 2009). An alternative regression method that

was adopted by many authors is the so-called ‘spatial regression’ tech-

nique (Anselin and Bera, 1998). This method starts from a significance

test for spatial autocorrelation based on the calculation of Moran’s I, a

measure of spatial autocorrelation. If this test is not significant, the

application of an ordinary least-squares (OLS) regression is suggested. If

the Moran’s I test is significant (this is usually the case), the occurring

kind of spatial dependence (‘spatial lag’ or ‘spatial error’) should be

detected. The occurrence of spatial lag implies that the dependent

variable in a cell is affected not only by the independent variables in the

cell itself, but also by those of the neighbouring cells. In the case of

spatial error, only the residuals in the regression analysis of neighbouring

cells are correlated. Depending on the occurrence of spatial lag or spatial

error, a regression model is selected that reduces the influence of these

phenomena. In some cases, both the spatial lag and the spatial error

model can be applied. Each of these calculations provides, inter alia, a

regression coefficient, a significance test and a pseudo-coefficient of

determination (pseudo-R2). The relative popularity of this method in the

recent literature is probably due to the user-friendly implementation in

the freely distributed software tool GeoDa (Anselin et al., 2006). In this

study, models were estimated by means of GeoDa 0.9.5-i5, which is a

public domain software package making spatial regression techniques

available in a geographical information system (GIS) environment. The

software was developed by Dr Luc Anselin of the School of Geographical

Sciences and Urban Planning, Arizona State University, Phoenix, Ari-

zona, USA.

Chapter 6

200

6.4 Results

6.4.1 Cartographic analysis

In order to gain a better understanding of the spatial variations of the

variables, a cartographic representation is shown in Figs 6.5-6.13,

aggregated at an intermediate scale in a grid of 4 x 4 km. Fig. 6.14 shows

a schematic reference of the study area (Flanders and Brussels). The

variables distance travelled per trip, minimum commuting distance and

density are each shown twice: once the zones are regarded as origins; and

once the zones are regarded as destinations. The variables diversity, jobs-

housing balance and number of potentially accessible jobs are displayed

once.

Fig. 6.5. Average commuting distance (dpt), origin zones


201

Fig. 6.6. Average commuting distance (dpt), destination zones

Fig. 6.7. Average minimum commuting distance (mdpt), origin zones

Chapter 6

202

Fig. 6.8. Average minimum commuting distance (mdpt),

destination zones

Fig. 6.9. Residential density (dens)


203

Fig. 6.10. Job density (dens)

Fig. 6.11. Shannon index (div)

Chapter 6

204

Fig. 6.12. Jobs-housing balance (jhb)

Fig. 6.13. Number of accessible jobs (potjob)


205

Fig. 6.14. Schematic representation of the spatial vision on Flanders.

Source: RSV (1997/2004)

Before using these data in a statistical spatial analysis, a number of

interesting findings from the maps can already be noted.

The most obvious relationships can be seen in the maps where the

zones are considered origins, or residential zones (Figs 6.5, 6.7 and 6.9).

The spatial variations of the distance travelled per trip (Fig. 6.5) is in

most regions similar to the spatial pattern of the minimum commuting

distance (Fig. 6.7), although values of the second variable are of course

considerably lower. The map showing residential density (Fig. 6.9) is in

many regions nearly the opposite of Figs 6.5 and 6.7. Therefore, there

seems to be a link between high residential densities and short commut-

ing distances. The map of the spatial diversity distribution (Fig. 6.11)

seems to indicate the same kind of relationship, at least as far as the

origin zones are considered. The low measured spatial diversity in the

province of Limburg (in the east) is noticeable.

On the maps where the zones are regarded as a destination or labour

zone (Figs 6.6 and 6.8), the patterns are less consistent. Here the thor-

Chapter 6

206

ough spatial concentration of jobs, in comparison with the more dispersed

structure of housing, plays an important role. In areas with a low job

density, the minimum commuting distance is of little significance.

A comparison of Fig. 6.10 with Figs 6.6 and 6.8 indicates that job

density is correlated with commuting distance, as well as with the

minimum commuting distance to the destination areas. For a better

insight, the jobs-housing balance (Fig. 6.12) should be taken into account.

In theory, a jobs-housing balance fluctuating around a value of 1 would

be most likely to yield short commuting distances (Peng, 1997). Because

of the large differences in spatial concentration between housing and jobs,

Fig. 6.12 does not seem to support this theory at first sight. The jobs-

housing balance therefore requires more specific attention in the following

paragraphs. In the economic core of Flanders, there also seems to be a

correlation between the accessibility of the job market (Fig. 6.13) and the

distance travelled, while this relationship is much less clear in the rest of

Flanders.

6.4.2 Correlation analysis for the average distance

travelled per trip

All the independent variables are considered approximations for spatial

proximity, in this case between the housing market and the job market.

A Pearson’s correlation matrix shows that most coupled variables in the

origin zones exhibit correlations in the range of 20-50%, amounting even

to 72% (between job density and minimum commuting distance) in the

destination zones. The couple job accessibility and jobs-housing balance is

an exception, with a correlation coefficient of only 9.7%.

Because of this inherent multicollinearity, and because of the difficul-

ties regarding interpretation, multiple regression techniques will not be

applied, but the paper confines itself to a bivariate correlation analysis.

All variables are normalized to z-scores and subsequently included in

a bivariate spatial regression using GeoDa. The average distance travelled

is always regarded as the dependent variable. For the spatial lag and

spatial error models, spatial weight matrices are constructed on the basis

of the grids, according to the so-called ‘queen’ method. This means taking

into account the influence of all adjacent cells, including those cells that

touch the studied cell at only one point, such as the adjacent squares in

chess that are covered by the queen.


207

First, an OLS regression is performed, as well as a test on spatial

autocorrelation (through a significance test for Moran’s I), and the

presence of spatial lag and spatial error is detected. In case no significant

spatial autocorrelation is found, the results of the OLS regression are

selected. In the other case, the spatial lag or spatial error model is

applied, depending on which phenomenon is significant. If both phenom-

ena appear significant, both models are applied. Each time the regression

coefficient r, which can be considered as a spatial correlation coefficient,

and the value of the significance test are selected. The zones are first

regarded as origins and then as destinations. The results are presented in

Table 6.1.

Table 6.1. Results of the correlation analysis for the average distance

travelled per trip

Origin dens div mdpt jhb potjob

MI 0.628 0.609 0.643 0.680 0.631

pMI 0.000 0.000 0.000 0.000 0.000

rOLS (-0.318) (-0.448) (0.255) (0.040) (-0.409)

pOLS 0.000 0.000 0.000 0.237 0.000

rSL -0.098 -0.162 0.080 0.019 (-0.093)

pSL 0.000 0.000 0.000 0.322 0.000

rSE (-0.109) -0.229 0.070 0.029 -0.292

pSE 0.000 0.000 0.003 0.223 0.000

Destination dens div mdpt jhb potjob

MI 0.655 0.677 0.624 0.576 0.653

pMI 0.000 0.000 0.000 0.000 0.000

rOLS (0.209) (0.095) (0.426) (0.467) (0.292)

pOLS 0.000 0.005 0.000 0.000 0.000

rSL 0.056 0.020 0.185 0.191 (0.004)

pSL 0.004 0.302 0.000 0.000 0.829

rSE (0.045) 0.012 0.224 0.202 0.171

pSE 0.108 0.678 0.000 0.000 0.014

Note:

origin: each zone is considered as an origin

destination: each zone is considered as a destination

Chapter 6

208

between parentheses: the GeoDa model suggests that one should not

use this method

MI: Moran’s I

pMI: p-value associated with Moran’s I

rOLS: correlation coefficient (ordinary least squares

method)

pOLS: p-value associated with OLS method

rSL: correlation coefficient (‘spatial lag’-method)

(bold = significant at the 0.05-level)

pSL: p-value associated with SL-method

rSE: correlation coefficient (‘spatial error’-method)

(bold = significant at the 0.05-level)

pSE: p-value associated with SE-method

6.4.2.1 Interpretation regarding the origin zones

In this case, each zone is considered as a residential area, acting as a

source of home-to-work trips in the morning traffic. For all variables, a

significant amount of spatial autocorrelation was found, making the

application of a simple linear regression (OLS) inappropriate. The applied

spatial regression models were suggested by GeoDa. Except for the jobs-

housing balance, all variables show a significant relation with the distance

travelled per trip. An overview of the findings now follows.

• A higher population density is associated with shorter commuting

distances.

• A greater diversity is also associated with shorter commuting dis-

tances. The link with spatial diversity is much stronger than the

relationship with population density.

• The minimum commuting distance shows a positive correlation with

the observed commuting distance, the magnitude of r being similar as

for the relationship with population density.

• There is no significant correlation between the observed commuting

distance and the jobs-housing balance. Given the specific characteris-

tics of the job-housing balance, a separate section will be devoted to

this variable below.

• Better access to the job market is associated with shorter commuting

distances. In comparison with the other independent variables, this

relationship can be called strong.


209

6.4.2.2 Interpretation regarding the destination zones

This subsection considers each zone as a labour area, a destination of the

commute during the morning. Again, significant spatial autocorrelation

for all variables is found. The calculation of the spatial regression models

suggested by GeoDa results in following findings:

• In contrast to the population density in the origin zones, a higher

density of jobs in destination zones is associated with longer commut-

ing distances. The relationship with job density is, however, less

strong than the relationship with population density in the origin

zones.

• There is no demonstrable correlation with spatial diversity.

• There is a relatively strong positive correlation between minimum

commuting distance and the observed commuting distance.

• Even if the jobs-housing balance appears not to be relevant in the

origin zones, this variable provides a relatively strong correlation in

the destination zones. The higher is the jobs-housing balance, the lar-

ger the commuting distances. This finding should, however, be

qualified (see the next section).

• The interpretation of the influence of the number of potentially

accessible jobs is not obvious, since this variable represents the num-

ber of jobs accessible departing from the considered zone, and not

from the other zones. This parameter yields a positive r. Better ac-

cess to the job market seen from the considered zone is associated

with a concentration of jobs, and thus with longer commuting dis-

tances to this zone.

6.4.3 Non-linearity in the jobs-housing balance

The use of spatial correlation analysis assumes a linear relationship

between the corresponding variables. This was assessed by the estimation

of a Lowess curve in a scatter plot, showing indeed linearity for most of

the used datasets. Lowess or Loess (locally weighted/estimated scatter-

plot smoothing) is a regression technique that applies linear least squares

regression on segments of the data, yielding a smoothly fitted curve.

However, slopes tend to flatten in the border areas of the sample curve.

This is particularly the case when dealing with extremely high densities

or large minimum commuting distances, or extreme values (both high and

low) of accessibility. Only with regard to the jobs-housing balance is the

relationship clearly not linear, especially when considering origin zones.

Chapter 6

210

On the one hand, it seems logical that a higher jobs-housing balance

leads to a larger share of nearby jobs, reducing average commuting

distances. On the other hand, it is clear that a high jobs-housing balance

in the destination zones is accompanied by an (over)concentration of jobs,

leading to larger commuting distances. Thus, a high jobs-housing balance

would be associated with both long and short commuting distances. This

paradox stems from the assumption of linearity, which appears unjusti-

fied.

When one considers the total amount of travelled distances produced

per zone, adding the distances covered by inbound trips to these covered

by the outbound trips, and plot this summed variable against the jobs-

housing balance, a ‘U’-shaped curve should be obtained (Peng, 1997)

(Fig. 6.15). The minimum distance corresponds to an equilibrium point

where the jobs-housing balance is close to 1, representing a perfectly

equal spatial distribution of housing and jobs.

To verify the validity of this hypothesis, the link between the total

generated distance per zone (4 x 4 km) and the jobs-housing balance is

visualized by means of a scatter plot for which a Lowess curve and a

best-fit quadratic curve are estimated (Fig. 6.16).

The curves appear to correspond to the hypothesis, that is, the total

commuting distance associated with one zone reaches a minimum at an

optimal jobs-housing balance, which is situated around a value of 1.

However, the fit is rather weak. This is partly because there are many

more zones with a low jobs-housing balance than with a high jobs-housing

balance. However, a more important reason is that the natural relation-

ship between spatial segregation and trip length is disturbed and

smoothed by non-spatial influences, as discussed in the Introduction. The

same reason applies to the generally low values of r that were found in

the previous section.


211

Fig. 6.15. Expected distribution of the total amount of commuter travel

and the jobs-housing balance

Fig. 6.16. Observed distribution of the total amount of commuter travel

and the jobs-housing balance

Chapter 6

212

6.5 Discussion

6.5.1 Origin versus destination zones

For most parameters, a positive correlation with the outbound trip length

corresponds to a negative correlation with the inbound trip length. A

higher population density in an origin zone means shorter trips. A higher

job density, by contrast, just means longer trips. A similar pattern is

observed when considering the number of accessible jobs. With respect to

spatial diversity, there seems to be no discernible relationship in the

destination zones, while one should approach the jobs-housing balance in

a non-linear way.

These results derive from the unequal spatial distribution of housing

and jobs. Jobs occur more often than housing in high concentrations in

urban areas, where the density and the spatial diversity are generally

high. In most zones the jobs-housing balance is lower than a value of 1,

making job supply in the other zones, which are much less numerous,

overconcentrated. Over-concentration means that employees more often

come from very far away to take up the available jobs. On the origin side,

however, employees living in an area with urban characteristics (high

density and spatial diversity) more often find a job close to home.

Therefore, Newman and Kenworthy’s (1989, 1999) thesis that fuel

consumption is negatively correlated with density is only valid when the

residential zones are considered. In the case of the labour zones, an

inverse correlation is found. However, modal choice was not taken up in

this study. It can be assumed that zones with a very high job concentra-

tion are accessed more than average by public transport. Consequently,

energy consumption (and other undesired external effects of traffic) is

expected to grow slower than the travelled distance itself. But because

the highest job concentrations entail the most extreme forms of long-

distance commuting, this mitigating effect on fuel consumption will be

rather limited.

Spatial diversity does not explain commuting distance viewed from

the labour zones. This can be explained by the large differences in types

of labour areas: high job concentrations generating long trips occur both

in urban centres with a high spatial diversity and in isolated monofunc-

tional industrial and port areas. This is different with regard to the

residential areas, where diversity is an explaining factor: a high residen-


213

tial density is usually associated with a large supply of jobs, and thus a

sound spatial mix of both functions.

When considering the relationship between the observed commuting

distance and the minimum commuting distance, values of r are signifi-

cantly higher for labour areas compared to residential areas. However, the

relationship with job density (in the labour areas) is weaker if compared

with the relationship with housing density (in residential areas).

When the origin zones are examined, it can be concluded that job

accessibility and spatial diversity should be considered the most appro-

priate indicators influencing commuting trip length. In this case,

residential density and minimum commuting distance have a relatively

low predictive value, while the jobs-housing balance is not a linear

predictor at all. In the destination zones, however, the minimum commut-

ing distance is a relatively good indicator, while the jobs-housing balance

and job density have only conditional predictive value.

6.5.2 Relevant scale levels

The paper has taken into account the fact that the relationships that are

derived from an analysis at the level of zones of 4 x 4 km cannot neces-

sarily be extended to processes that play at a different aggregation level.

The importance of scale may mainly depend on the considered trip-length

classes.

Density, diversity and proximity are spatial characteristics that are

important for traffic generation, but mainly on a scale level that can be

associated with the considered distance class. Influencing the travel

behaviour of motorists should be done especially at higher scale levels. To

influence the trips of pedestrians, cyclists and users of local public

transport, the lowest scale levels are perhaps the most relevant.

The travel behaviour of the twenty-first-century commuter deviates

strongly from that of the historical inhabitants of a medieval or nine-

teenth-century compact city, so that only a limited overall environmental

impact is to be expected from a policy that only focuses on strengthening

these old structures. Potentially more impact could be expected from a

policy that focuses on strengthening density, diversity, the jobs-housing

balance and proximity within the higher levels of aggregation, also

comprising suburban areas. In addition, the quality of the urban structure

at the lowest scale levels may still have an impact on the share of pedes-

Chapter 6

214

trians, cyclists and users of public urban transport, and may in that sense

have a positive impact on the urban environment.

6.6 Conclusions and pathways

for further research

Daily trips on foot, by bicycle or public urban transport, but also short

car trips, fit into the spatial model of the compact city. By strengthening

the characteristics of the compact city, with the emphasis on residential

density and proximity of functions, a spatial framework may be created

that makes trips over long distances less necessary or even unnecessary.

However, the category of short trips has become a niche market, espe-

cially within the commute. In Flanders, the average commuter covers a

distance of about 19 km per trip, a trip length that is well beyond the

compact urban scale. However, a number of qualities that are attributed

to the compact city are still valuable, even if they also influence spatial

processes at the regional scale. Viewed from a sustainability perspective,

an equal jobs-housing balance and a high residential density are para-

mount. A high degree of spatial diversity and accessibility of the labour

market also play a positive role. By contrast, high concentrations of

employment are not desirable. High job densities seem to give rise to

overconcentration easily, so that workers should be recruited from a wide

hinterland. These features are reflected in the minimum commuting

distance, which is a good measure for the proximity of housing and the

job market and clearly evolves parallel with the observed commuting

distances. Since the minimum commuting distance is determined by the

spatial patterns of housing, jobs and infrastructure, it can be concluded

that commuting over short distances can be facilitated by changes in the

spatial structure.

The outlined research design, however, also calls for further elabora-

tion. The results for the commute cannot simply be extrapolated to other

types of travel. Trips to school, to shops or social activities usually are

much shorter than home-to-work trips (Zwerts and Nuyts, 2004). There-

fore, it might be possible that the correlations for this type of travel are

stronger at the lower scale levels in comparison to home-to-work trips.

Another extension concerns the feasibility of an equal jobs-housing

balance: it has to be taken into account that a neighbourhood’s jobs-

housing balance that fluctuates area-wide around 1 may not be compati-


215

ble with agglomeration trends that are important for many economic

sectors.

Another improvement could be incorporating the modal split into the

comparisons. Long commuting distances are often covered by train, which

means that the negative impact of these trips increases at a slower rate

than the number of kilometres travelled. Moreover, it may well be that

increasing spatial proximity at the lowest scale levels will result in a

larger share of pedestrians, cyclists and users of urban public transport in

urban areas, strengthening the positive effect of shorter trips by a modal

shift. The authors hope to address these and related issues in future

research.

References

Alberti, M. (1999) “Urban patterns and environmental performance:

What do we know?” Journal of Planning Education and Research.

19(2), pp. 151-163.

Amrhein, C. G. (1995) “Searching for the elusive aggregation effect:

Evidence from statistical simulations.” Environment and Planning A.

27(1), pp. 105-119.

Anselin, L. and A. Bera (1998) “Spatial dependence in linear regression

models with an introduction to spatial econometrics”. In: A. Ullah

and D. Giles (Eds.) Handbook of Applied Economic Statistics. New

York: Marcel Dekker, pp. 237-290.

Anselin, L. and D. A. Griffith (1988) “Do spatial effects really matter in

regression analysis?” Papers in Regional Science. 65(1), pp. 11-34.

Anselin, L., I. Syabri and Y. Kho (2006) “GeoDa: An introduction to

spatial data analysis.” Geographical Analysis. 38(1), pp. 5-22.

Banister, D., S. Watson and C. Wood (1997) “Sustainable cities, trans-

port, energy and urban form.” Environment and Planning B. 24(1),

pp. 125-143.

Batty, M. (1974) “Spatial entropy.” Geographical Analysis. 6(1), pp. 1-

31.




press.

Chapter 6

216



43(5), pp. 580-591.

Burton, E. (2000) “The potential of the compact city for promoting social

equity”. In: K. Williams, E. Burton and M. Jenks (Eds.) Achieving

Sustainable Urban Form. London and New York: E&FN Spon, pp.

19-29.

Cervero, R. (1989) “The jobs-housing balance and regional mobility.”

Journal of the American Planning Association. 55(2), pp. 136-150.

Cervero, R. (1996) “Mixed land-uses and commuting: Evidence from the

American housing survey.” Transportation Research Part A. 30(5),

pp. 361-377.



2(3), pp. 199-219.



and Planning A. 39(5), pp. 1238-1254.

Cliff, A. D. and J. K. Ord (2009) “What were we thinking?” Geographical

Analysis. 41(4), pp. 351-363.

Commission of the European Communities (1990) Green Paper on the

Urban Environment - COM(90)218. Brussels: Commission of the

European Communities.

Crane, R. and R. Crepeau (1998) “Does neighborhood design influence

travel? A behavioral analysis of travel diary and GIS data.” Trans-

portation Research Part D. 3(4), pp. 225-238.



Studies. 23(3), pp. 377-401.

Ellis, C. (2002) “The new urbanism: Critiques and rebuttals.” Journal of

Urban Design. 7(3), pp. 261-291.

Frank, L. D. and G. Pivo (1994) “Impacts of mixed use and density on

utilization of three modes of travel: Single-occupant vehicle, transit,

and walking.” Transportation Research Record. 1466, pp. 44-52.

Hamilton, B. W. (1982) “Wasteful commuting.” Journal of Political

Economy. 90(5), pp. 1035-1053.





217

Horner, M. W. and A. T. Murray (2003) “A multi-objective approach to

improving regional jobs-housing balance.” Regional Studies. 37(2),

pp. 135-146.

Jenks, M., E. Burton and K. Williams (1996) “Compact cities and

sustainability: An introduction”. In: M. Jenks, E. Burton and K. Wil-

liams (Eds.) The Compact City: A Sustainable Urban Form? London

and New York: E&FN Spon, pp. 3-8.

Lyons, G. and K. Chatterjee (2008) “A human perspective on the daily

commute: Costs, benefits and trade-offs.” Transport Reviews. 28(2),

pp. 181-198.







Part A. 38(2), pp. 143-162.

Ministry of Housing, Spatial Planning and the Environment (1988)

Vierde Nota over de Ruimtelijke Ordening. Den Haag: Ministry of

Housing, Spatial Planning and the Environment.

Nagendra, H. (2002) “Opposite trends in response for the Shannon and

Simpson indices of landscape diversity.” Applied Geography. 22(2),

pp. 175-186.

National Geographical Institute of Belgium (2009) Structuur en codering

van de gegevens top10v-gis en top50V-gis. National Geographical In-

stitute of Belgium from

http://www.ngi.be/Common/articles/CA_Top10V-GIS_TOP50V-

GIS/restruct_doc_nl.htm.

Neuman, M. (2005) “The compact city fallacy.” Journal of Planning

Education and Research. 25(1), pp. 11-26.







Openshaw, S. and P. Taylor (1979) “A million or so correlation coeffi-

cients: Three experiments on the modifiable areal unit problem”. In:

Chapter 6

218

N. Wrigley (Eds.) Statistical Methods in the Spatial Sciences. Lon-

don: Routledge and Kegan Paul, pp. 127-144.


Urban Studies. 34(8), pp. 1215-1235.

Rietveld, P., B. Zwart, B. van Wee and T. van den Hoorn (1999) “On

the relationship between travel time and travel distance of commut-

ers. Reported versus network travel data in the Netherlands.” Annals

of Regional Science. 33(3), pp. 269-287.



Schwanen, T. and P. L. Mokhtarian (2005a) “What if you live in the

wrong neighborhood? The impact of residential neighborhood type

dissonance on distance traveled.” Transportation Research Part D.

10(2), pp. 127-151.

Schwanen, T. and P. L. Mokhtarian (2005b) “What affects commute

mode choice: Neighborhood physical structure or preferences toward

neighborhoods?” Journal of Transport Geography. 13(1), pp. 83-99.

Stead, D., J. Williams and H. Titheridge (2000) “Land use, transport and

people: Identifying the connections”. In: K. Williams, E. Burton and

M. Jenks (Eds.) Achieving Sustainable Urban Form. London and New

York: E&FN Spon, pp. 174-186.

Tobler, W. R. (1970) “A computer movie simulating urban growth in the

Detroit region.” Economic Geography. 46 (supp.), pp. 234-240.

Van Acker, V. and F. Witlox (2011) “Commuting trips within tours: How

is commuting related to land use?” Transportation. doi:

10.1007/s11116-010-9309-6.




353.

Van Wee, B. (2002) “Land use and transport: Research and policy

challenges.” Journal of Transport Geography. 10(4), pp. 259-271.

Vandenbulcke, G., T. Steenberghen and I. Thomas (2007) Accessibility

Indicators to Places and Transports - Final Report. Leuven / Lou-

vain-la-Neuve: Federal Public Service of Mobility / Federal Science

Policy.

Vandenbulcke, G., T. Steenberghen and I. Thomas (2009) “Mapping

accessibility in Belgium: A tool for land-use and transport planning?”



219



gium.

Wang, F. (2000) “Explaining intraurban variations of commuting by job

proximity and workers’ characteristics.” Environment and Planning

B. 28(2), pp. 169-182.

Wang, F. (2003) “Job proximity and accessibility for workers of various

wage groups.” Urban Geography. 24(3), pp. 253-271.

Williams, K., E. Burton and M. Jenks (2000) “Achieving sustainable

urban form: An introduction”. In: K. Williams, E. Burton and M.

Jenks (Eds.) Achieving Sustainable Urban Form. London and New

York: E&FN Spon, pp. 1-5.



15(3), pp. 172-183.

Zwerts, E. and E. Nuyts (2004) Onderzoek Verplaatsingsgedrag Vlaande-

ren 2000-2001. Brussels-Diepenbeek: Ministry of the Flemish

Community.

Chapter 6

220

221

Chapter 7:

Linking expected

mobility production to

sustainable residential

location planning

This paper will be published as Boussauw, K. and F. Witlox (2011)

“Linking expected mobility production to sustainable residential location

planning: Some evidence from Flanders.” Journal of Transport Geogra-

phy. Copyright © Elsevier. All rights reserved.

Abstract

Based on a set of spatial proximity characteristics this paper develops a

model that estimates for every neighbourhood in Flanders (Belgium) the

amount of traffic that would be generated by an additional residential

unit when socio-economic variables are held constant. The results show

that residential density, land use diversity and proximity of facilities

influence daily travelled distances when these variables are measured in

the immediate vicinity of the residential location of the respondent

(within a radius of 1 km). When aggregating these variables at a larger

geographical scale, in most cases the impact proves no longer significant.

Variables based on the spatial distribution of jobs, or on the global

accessibility of the entire population in the study area, do not show any

significant effects on the travel distance.

Despite the statistical significance only a fraction of the observed

variance in reported distances is explained by characteristics of spatial

proximity. However, we can assume that the importance of spatial

structure in the genesis of mobility patterns will increase in case the cost

of transport would rise (cf. peak oil). For this reason, the application of

Chapter 7

222

the mapped results of the proposed model could contribute to the prac-

tice of sustainable spatial planning.

Keywords: spatial proximity; travel behaviour; sustainable spatial

development; Flanders

7.1 Introduction

Research into the relationship between spatial structure and travel

behaviour exists in many forms. Scholars typically focus on the search for

statistical associations between aspects of travel behaviour (such as choice

of mode or destination, travel time or trip length) and spatial characteris-

tics (such as density and degree of mix of homes, jobs and other facilities,

or mere morphological features such as street patterns and neighbourhood

layout). Ewing and Cervero (2010) present an extensive literature review

on this.

An important part of the existing research in this field focuses on the

potential application of the obtained results in the development of a more

sustainable urban and regional spatial structure that can operate on the

basis of minimal energy needs for transport (Ewing et al., 2008). Al-

though many policy plans still refer to Newman and Kenworthy (1989,

1999), who argue that there is a strong inverse relationship between

population density and transport energy consumption per capita, later

research shows that this statement is a serious simplification. Criticisms

of Newman and Kenworthy (1989, 1999) (Mindali et al., 2004; Mees,

2010, pp. 24-26) rely mostly on methodological issues, such as the chosen

demarcation of the studied cities, while quantitative research into the

relationship between spatial characteristics and energy consumption in a

regional network structure finds much more complex interactions (Bous-

sauw et al., 2011a). Energy consumption by transport is partly

determined by the modal split, and partly by the total distance travelled

within the studied system. Previous research shows that in regional

studies (which go beyond the urban scale) in a western context, the daily

distance travelled per person is a good approximation of sustainability of

travel patterns (Boussauw and Witlox, 2009), while the influence of

modal split is only secondary. In the following sections, we will therefore

focus on the relationship between distance travelled and land use charac-

teristics that measure mutual proximity between possible destinations.

Linking expected mobility production to residential location planning

223

In this over the years adequately documented line of research, we can

distinguish two important constants: i.e., (i) the assumed relationships

always appear to be statistically significant, but (ii) explain only a small

share of the observed variance. The first of these two findings is actually

trivial: it would be quite remarkable if the influence of the spatial distri-

bution of different types of destinations, which among others defines

mutual distances that need to be covered, would not pass significance

tests (Næss, 2003). The second finding, however, is a lot less comforting:

the explained variance (in many analyses represented by the coefficient of

determination (R2) of a regression equation) is usually very low (Handy

et al., 2005a; Cervero and Kockelman, 1997; Cervero, 1996; Næss and

Sandberg, 1996). Obviously, this means that spatial characteristics

explain travel behaviour to an only very limited extent.

In socio-geographically inspired research, spatial features are usually

only one of the considered clusters of explanatory variables in the model.

By combining many socio-demographic and economic variables (such as

income, car ownership, family composition, lifestyle or job preference)

with spatial characteristics, a relatively satisfactory fit may be obtained

(Van Acker and Witlox, 2011; Maat and Timmermans, 2006). An

advantage of this approach is the accurate estimation of the model

coefficients since the influence of any potential correlation between spatial

and socio-economic variables is filtered out. An example of such a

correlation is the inverse relationship between income class and residen-

tial density. A major drawback of upgrading a spatial model to a socio-

economic model to explain travel behaviour is that the influence of the

spatial structure, which is present anyway, seems to fade into the back-

ground.

A model built on mere spatial features is nevertheless useful for spa-

tial policy. Although spatial characteristics explain only a small part of

the assessed travel patterns, the built environment is still determining the

physical preconditions for sustainable mobility patterns. Moreover, we

argue that the importance of the spatial component in the genesis of

travel patterns is not constant throughout history, but is linked to the

cost and the speed of mobility. Over the centuries, the absolute cost to

move an individual over a distance of one kilometre has almost continu-

ally been decreasing, if we neglect the slight ripples in the cost curve

during the oil crises in the seventies. Moreover, the average speed of

travel has been continuously increasing world wide (Schafer, 2000), a

phenomenon that is largely explained by the growth in car ownership and

Chapter 7

224

the extension of the road network. Both developments have led to a

systematic decline of the importance of physical distance between poten-

tial destinations (Rietveld and Vickerman, 2004), which in turn resulted

in the weakening of the transport-land use connection (Giuliano, 1995).

In statistical analyses based on spatial characteristics this phenomenon is

reflected in a low coefficient of determination.

The continuing decline in transport costs is only possible through the

abundant availability of energy in the form of fossil fuels and is therefore

finite (Wegener, 2010). According to the peak oil theory (Witze, 2007)

the relative cost of oil products may significantly increase over time,

leading to a reduction in mobility and a growing importance of mutual

spatial proximity of destinations (Dodson and Sipe, 2008). The propor-

tionately small share of the variance in travel patterns that is explained

by spatial structure should not be considered unimportant. It is exactly

physical space that is the most rigid component, and thus the slowest to

adapt to changing economic conditions, in contrast to e.g. behavioural

elements that are more subject to an individual’s choice.

The aim of the current research is the development of a model for the

study area of Flanders (Belgium), based on mere spatial characteristics,

that indicates what level of mobility production (expressed as daily

distance travelled per individual) is associated with the location of an

additional housing unit in a certain area. The use of the residential

location as a reference for the study is taken by the abundance of avail-

able residence-based travel data. We use data from the 2007-2008 Travel

Behaviour Survey for Flanders (Janssens et al., 2009) and a number of

additional data sets containing spatial variables.

In a first phase, the relationship between characteristics of spatial

proximity, as measured in the area of residence, and individually reported

daily travel distances is assessed through a linear regression model. The

model accounts for variability related to the applied aggregation level by

incorporating various geographical scale levels. In a second phase, results

obtained from the regression analysis are used to construct a map which

represents for each statistical ward the expected mobility production by

an inhabitant in this location if only spatial variables are taken into

account. Obviously, the model will explain only a limited share of the

expected variance, an aspect which should be taken into account in the

interpretation. This means that the applicability of the model is largely

relying on the assumption that observed relationships may become

stronger in the future.


225

Unlike the usual approach taken to the assessment of land use-

transport connections, which aims to detect statistical relationships, our

study expands the findings immediately to an application that is useful in

location policy at a regional scale level.

7.2 Study area

The research focuses on the Flanders region, which is together with the

Brussels Capital Region composing the north of Belgium. The main

borders of the Flanders region are constituted by the North Sea, the

Netherlands and the Walloon region (south of Belgium). The Brussels

Capital Region, which has over one million inhabitants, is the largest

agglomeration in the region, and is in geographical terms centrally

located in Flanders.

In addition to Brussels, the metropolitan areas of Antwerp (400,000

inhabitants) and Ghent (250,000 inhabitants) are located in Flanders, as

are ten regional cities (with a population of around 100,000 inhabitants)

and a series of smaller urban centres and municipalities.

An interesting, typical Belgian, aspect is found in the history of the

institutionalized commute, through government support for the construc-

tion of an extended railway network and cheap commuter tickets, aiming

for the industrialization of the country based on a minimum of urbaniza-

tion (Verhetsel et al., 2010). In the 19th century and early 20th century,

this policy led to a clustering of housing and amenities in the towns and

villages that were connected to the railway network. After World War II,

these structures have fanned out into car-oriented suburban develop-

ments, an evolution that is associated with ever increasing mutual

distances between homes, jobs and daily facilities, and has created a

major source of dispersed traffic (Boussauw et al., 2011b).

7.3 Methodology and data

7.3.1 Analysis and model structure

The objective of the paper is to develop a model that forecasts regional

variations in mobility production based on characteristics of spatial

proximity at the residential location. We use regression analysis, with

daily kilometrage per person as the dependent variable. Explanatory

Chapter 7

226

variables consist of a number of measures of spatial proximity that are

observed at various aggregation levels around the individual residential

locations. In addition, a number of socio-economic variables are used as

control variables. The applied data sets are described below.

We start from a full model that includes all considered variables.

Then, we trim the model and ultimately only retain those variables and

scale levels that show statistically significant. If necessary, transforma-

tions are applied to address potential deviations from the normal

distribution or prominent non-linear relationships.

After building and trimming the model, the obtained equation is used

to estimate the mobility generating character of each neighbourhood (i.e.

census ward) in Flanders. For each ward the relevant spatial variables are

recalculated, from which the expected daily number of generated kilome-

tres per person is regressed. These values are then displayed in the form

of a map. When interpreting the map, it is important to realize that the

extent to which spatial structure explains the mobility of a (new) resident

of any area is indicated by the coefficient of determination (R2) of the

regression equation.

7.3.2 Dependent variable (PKM)

The daily kilometrage per person is used as the dependent variable. The

data source is the Travel Behaviour Survey for Flanders (OVG3)

(Janssens et al., 2009). OVG3 is a mobility survey conducted during

2007-2008 in 8,800 respondents over the age of 6 years and living in the

Flanders region (excluding the Brussels Capital Region). The selection is

based on a sample from the national register. The home address of the

respondents is recorded. Respondents are asked to keep track of all their

trips during a predetermined random day by means of a travel diary. Of

the 8,800 respondents, 7,273 have actually moved on that day, and have

reported the perceived distance covered by their trips. In our analysis we

use the sum of the lengths of all trips reported by the respondent.

Because of the nature of the data possible biases inherent in the use of

travel diaries should be taken into account (Witlox, 2007).

For the sake of calculating the values of the explanatory spatial vari-

ables, address data was geocoded (converted into XY coordinates) using

the Google Maps web service. An overview of the residential locations of

the respondents is given by Fig. 7.1.


227

Fig. 7.1. Situation of the residential location of the respondents

in the study area

7.3.3 Explanatory variables

A total of six explanatory variables have been selected (in addition to the

control variables, which are discussed subsequently), each of which can be

considered as a measure for the mutual spatial proximity with regard to

potential destinations. The variables are: (i) accessibility, (ii) residential

density, (iii) land use diversity, (iv) job density, (v) minimum commuting

distance, and (vi) proximity of facilities. The construction of these

variables is explained in the following paragraphs.

Since we are using spatially aggregated data, the modifiable areal unit

problem (MAUP) (Openshaw and Taylor, 1979) should be taken into

account. To reduce distortion of the results by the influence of the spatial

scale at which data is aggregated, each variable has been determined at

three different levels of aggregation. To this end, per respondent three

circular zones have been drawn of which the midpoint is the reported

residential location, with a respective radius equalling 1 km, 4 km and 8

km. Although the choice of these three levels may seem rather arbitrary,

the analysis will clearly show that the used variables do only have clear

impact at the lowest scale level. Consequently, the two higher aggrega-

tion levels will be removed from the regression, thus avoiding

multicollinearity problems. At the other hand, given the resolution of the

Chapter 7

228

basic data sets, using an even smaller zone than a circle with r = 1 km

would not be justified.

Within these circles, data is then averaged on the basis of the propor-

tional overlap with the original zones associated with the used data sets

(these are census wards, traffic analysis zones (TAZ’s) and a one kilome-

tre square grid respectively).

7.3.3.1 Accessibility (ACC)

To define regional geographical accessibility, we start from the 2007

population data in Flanders and Brussels, aggregated by census ward. A

distance matrix is calculated between each possible pair of census wards,

based on a shortest path calculation over the road network (Streetnet).

Finally, for each census ward, the total distance that should be covered to

visit each resident of any other census ward in the study area once and

return back home, is summed. This accessibility index thus gives a

measure of the interaction opportunities with all other inhabitants of

Flanders and Brussels, based on physical distance.

A disadvantage of this measure is that neighbouring countries and

regions are not accounted for. However, the nature of a cumulative

accessibility measure requires a clear delineation of the study area. The

applied boundary is justified by rather strong language and cultural

differences making daily travel crossing the outer borders of the Flanders

region relatively rare. For example, in 2007, only 2.0% of Flanders’

employed labour force worked in Wallonia and only 1.5% worked abroad.

(Policy Research Centre on Work and Social Economy, 2010) Also, since

this is a distance based accessibility measure, it does not take into

account variations in travel speed e.g. due to choice options in travel

mode and route, or the presence of congestion.

7.3.3.2 Residential density (POPD)

The residential density is based on government population data for 2007,

aggregated by census ward in Flanders.

7.3.3.3 Land use diversity (DIV)

To approximate the degree of land use mix, the Strucnet file of the

National Geographical Institute of Belgium (2009) was used, containing

all buildings that are represented by the official topographic maps with


229

scale 1:10,000. The buildings are divided into categories. Although the

accuracy of the categorization is limited, this inventory can be used to

calculate approximate land use diversity in a given area. Since this

dataset contains a functional classification and is available at a high

resolution, it prevails on satellite imagery and is thus the best area

covering dataset currently available.

To calculate spatial-functional diversity, we employ the Shannon in-

dex. This index is used in landscape ecology as a measure of

morphological diversity (Nagendra, 2002), and is sometimes called spatial

entropy (Batty, 1974). The calculation was done for a square grid based

on an area of 1 km2, after which results were proportionally aggregated

within the three described circular zones. In this way, the possible

additional bias caused by the property of the Shannon index to increase

with larger area coverage is avoided.

7.3.3.4 Job density (JOBD)

Job density is based on commuting data as provided by the Multimodal

Model for Flanders (MMM, version 2007). MMM is a simulation of all

personal trips in the Flanders region formatted as an origin-destination

(OD) matrix and is based on a combination of various sources of socio-

economic data. MMM aggregates arrivals of all commuting trips between

4 am and 11 am in the morning traffic within TAZ’s, which are compara-

ble to, but typically slightly larger than, census wards.

7.3.3.5 Minimum commuting distance (MCD)

This variable was constructed based on the OD-matrices for commuting

between 4 am and 11 am, as they were simulated in the MMM. The

principle of the method implies that any departure (in this case in the

morning traffic) is linked to the nearest possible arrival (also in the

morning traffic). Per TAZ, the number of departures as well as the

number of arrivals are retained, but the in reality existing tie between

origins and destinations is cut in order to minimize the total distance

travelled within the system. This theoretical exercise provides a good

measure of the spatial proximity between the housing market and the

labour market. The data are results provided by Boussauw et al. (2011c),

where details on the calculation can be found.

Chapter 7

230

7.3.3.6 Proximity to facilities (SPROX)

This variable was constructed based on the spatial distribution of non-

work related destinations that are often visited by an average Flemish

household, such as schools, shops, cafes, sports clubs, banks, and medical

services. Per census ward the minimum distance was calculated that

needs to be covered by an average Flemish family to get its weekly

programme done when always opting for the closest facility within each

destination class. This weekly programme for an average family was

determined based on data from the second phase of the Travel Behaviour

Survey for Flanders (OVG2) (Zwerts et al., 2004). The data are results

provided by Boussauw et al. (2011d), to which we refer for further

calculation details.

7.3.4 Control variables

The OVG3 (Janssens et al., 2009) contains a number of socio-economic

data that may explain part of the variance in the reported distance.

These variables are: education level (EDU), income level (INC), age

(AGE) and gender (GND). We include these in the model as control

variables. This means that our research does not focus on the explanatory

power of these socio-economic variables, although it is supposed that they

make the regression equation more fitting. The selected control variables

all exhibit a statistically significant relationship with the reported travel

distance and make an important contribution to the model fit.

Education and income levels are included as continuous variables.

Because of the assumed non-linear influence of the respondent’s age, the

age variable is recoded into four dummy variables. Following categories

are considered: 0-19 years, 20-39 years, 40-59 years and 60-79 years, while

80 years or older is used as the reference category. Gender is obviously a

dummy variable; male is considered as the reference group.

7.4 Analysis

Based on the described variables, a multivariate linear regression equa-

tion has been framed. A logarithmic transformation was applied on the

dependent variable PKM, resulting in an adequate approximation of the

normal distribution. After an exploratory test for the presence of non-

linear relationships, a linear regression appeared to provide the best


231

match with reality if the non-linear effect of the age of the respondent is

modelled by means of dummy variables. As mentioned, the six explana-

tory variables were repeatedly constructed at three separate levels of

aggregation (circles with r = 1 km, r = 4 km and r = 8 km). Ultimately,

the basic equation is composed of eighteen independent variables and four

control variables, and is expressed formally:

εγγγ

γγγγ

βββ

βββα

+⋅+⋅+⋅+

⋅+⋅+⋅+⋅+

⋅+⋅+⋅+

⋅+⋅+⋅+=

−−−−

===

===

∑∑∑

∑∑∑

INCEDUGND

AGEAGEAGEAGE

SPROXMCDJOBD

DIVPOPDACCPKM

rrr

rrr

rrr

rrr

rrr

rrre

765

7960459403392021901

8,4,16

8,4,15

8,4,14

8,4,13

8,4,12

8,4,11)(log

(7.1)

For most spatial variables, no significant effects are yielded. In particular,

those variables that are constructed on the same basis but at a different

level of aggregation (e.g., ACC1, ACC4 and ACC8) appear to be highly

correlated and thus causing effects of multicollinearity. The best results

are achieved by applying only the first level of aggregation (circles with r

= 1 km). Subsequently the variables related to the spatial distribution of

jobs (JOBD1 and MCD1) do not significantly affect the results. Although

this outcome is unexpected, it can be explained by the small proportion

of today’s commuter traffic in the total number of trips (20.6%) and total

distance travelled (34.5%) (Janssens et al., 2009). Also the accessibility

variable ACC1 was excluded from the equation, since no significant

correlation between ACC1 and loge(PKM) was found. The purified

regression equation is as follows:

εγγγ

γγγγ

βββα

+⋅+⋅+⋅+

⋅+⋅+⋅+⋅+

⋅+⋅+⋅+=

−−−−

INCEDUGND

AGEAGEAGEAGE

SPROXDIVPOPDPKMe

765

7960459403392021901

131211)(log

(7.2)

The results of the regression analysis are given in Table 7.1. The results

are consistent with the literature: significances are satisfactory (all results

are within the 0.01 confidence level) at a low coefficient of determination

(R2 = 14.3%). The relationships found meet the expectations. A higher

population density and a higher degree of spatial diversity are associated

with shorter travel distances. Also, a larger minimum distance to reach

Chapter 7

232

daily facilities is associated with shorter real travel distances. The age

group between 20 and 59 years exhibits the most intensive travel pattern,

while women are less mobile than men. Both a higher level of education

and a higher income are associated with increased mobility.

Table 7.1. Coefficients of the regression analysis

R2 = 0.143 coefficient p-value

(constant) 1.502 0.000

POPD1 -3.99 . 10-5 0.000

DIV1 -0.278 0.001

SPROX1 0.004 0.000

AGE0-19 0.847 0.000

AGE20-39 1.066 0.000

AGE40-59 0.969 0.000

AGE60-79 0.624 0.000

GND -0.245 0.000

EDU 0.173 0.000

INC 0.111 0.000

The relatively small share of the observed variance that is explained by

the model, is common for mobility research. Although this phenomenon is

in part due to data deficiencies (including underreporting and randomiza-

tion of reporting days), the truth lies perhaps in the importance of the

many random factors that form the underlying reason for a significant

share of individual trips, but are difficult or even impossible to model. An

example of this is the so-called random taste variation that is accounted

for in many discrete choice modelling techniques (Train, 2003, p. 46). In

Flanders, we find similar difficulties in travel behaviour modelling

attempts in Witlox and Tindemans (2004).

If we redo the regression analysis on the basis of only the control

variables, then we obtain an R2 equalling 11.9%. The same analysis based

on only the spatial variables yields an R2 of 2.1%. This means that within

this last, reduced, model only 2.1% of the observed variance in distance

travelled could be explained by characteristics of spatial proximity of the

residential environment of the respondent. As explained in the introduc-

tion, this is the restrictive but nevertheless relevant context within which

the results should be interpreted.


233

The sum of the two coefficients of determination of both reduced

models is less than 14.3%, which indicates the occurrence of suppression

and suggests that combining the two sets of explanatory variables (socio-

economic and spatial) indeed yields some added value. However, it is

clear that the explanatory value of spatial variables largely subordinates

to that of the socio-economic variables. This means, for example, that

lowering the average income would be more effective in combating

excessive mobility than increasing housing density.

7.5 Forecasting model for Flanders

In order to develop a forecasting, area covering, model based on the

results of the regression analysis, we isolate the spatial variables. To this

end, the control variables are made constant by equalling these to the

mean value of the considered variable in the dataset. Formally:

133.3

111.0173.0245.0624.0

969.0066.1847.0502.1

7960

59403920190

=

⋅+⋅+⋅−⋅+

⋅+⋅+⋅+=

−

−−−

INCEDUGNDAGE

AGEAGEAGEctrlα

(7.3)

Based on the regression coefficients for the spatial variables the expected

amount of generated kilometres per inhabitant PKMw for each census

ward in Flanders w is determined as follows:

)004.0278.00.0000399133.3exp( wwww SPROXDIVPOPDPKM ⋅+⋅−⋅−=

(7.4)

The mapped result is shown in Fig. 7.2. The expected amount of daily

generated kilometres per inhabitant based on characteristics of spatial

proximity and averaged by census ward is approximately normally

distributed and is characterized by the values that are shown in Table

7.2.

Chapter 7

234

Fig. 7.2. Spatial distribution of the estimated daily generated mobility

per capita based on characteristics of spatial proximity

Table 7.2. Features of the distribution of daily generated mobility per

capita as expected by the model, based on census wards in Flanders

N = 9205 km

km 05% percentile 15.3

mean 23.0 25% percentile 20.2

median 23.0 75% percentile 25.8

standard deviation 5.1 95% percentile 30.1

The 95-percentile value is almost twice as large as the 5-percentile value.

This means that based on characteristics of spatial proximity, the 5%

best-located census wards are estimated to generate only half of the

mobility of the 5% worst-located wards.

As expected, and as shown in Fig. 7.2, urban areas yield the lowest

values, particularly in the historical city centres and a number of 19th-

century neighbourhoods in Ghent and Antwerp. Among the regional

urban areas mainly Leuven, Mechelen, Aalst, Brugge and Oostende score

well. Also the edge of the Brussels conurbation scores quite well, although

the agglomeration effect decays rapidly while moving away from the

centre of the capital. When we examine regions instead of cities, we see

that typically rural areas as well as green and wooded areas with scat-


235

tered development score badly. Conversely, the immediate vicinity of

large agglomerations score well, just as the highly suburbanized areas

Kortrijk-Leie (in the south-west) and the so-called Flemish Diamond (the

area cornered by Ghent, Antwerp, Leuven and Brussels).

7.6 Discussion

The results can be summarized as follows. Residential density, land use

diversity and proximity of facilities affect the daily distance travelled if

these variables are measured in the immediate vicinity of the residential

location of the respondent (within a radius of 1 km). When these vari-

ables are aggregated at a higher geographical scale, the impact is no

longer significant in most cases. This is also the case for variables that are

based on the overall accessibility of the entire population in the study

area. From this we can deduce that the overall travel pattern of an

average resident of Flanders is to a larger extent determined by local

accessibility of possible destinations than by regional accessibility or the

embeddedness in the wider region.

In addition, also variables that are based on the spatial distribution

of jobs do not show a significant impact on travel distance. This latter

finding is somewhat surprising given the adequate volume of literature

that is specifically focusing on the relationship between commuting and

the spatial distribution of jobs and housing. A comprehensive overview is

given by Horner (2004). Although the commute continues to represent a

significant share of overall travel, it should not be forgotten that a large

part of the population does not commute. Moreover, the share of the

commute in the total traffic volume is systematically decreasing (Pisarski,

2006, p. 2). Also, non-business trips tend to be considerably influenced by

the local supply of potential destinations (Handy et al., 2005b), whereas

commuting is only limitedly influenced by the local supply of jobs. For

the average resident of Flanders, the specific spatial structure of the job

market plays a relatively minor role compared to more generalized

characteristics of overall spatial proximity.

The results of the regression analysis indicate that only a fraction of

the observed variance in distance travelled is explained by spatial fea-

tures. A combination of some very basic socio-economic characteristics

(education, income, age and gender) is already explaining a much larger

Chapter 7

236

share of the variance. Nevertheless, both sets of explanatory variables are

clearly complementary.

The low proportion of variance explained by a model that contains

only spatial variables (R2 = 2.1%) entails in practice the risk of neglect-

ing the importance of spatial structure as a framework for the genesis of

potentially sustainable travel patterns. However, we argue that the

importance of the spatial distribution and thus the mutual distance

between potential destinations has decreased in the course of history as

the cost of transport fell. The cheaper transport is, the larger individual

freedom is in choosing a particular destination from a range of potential

locations where the occurred need can be fulfilled. The more expensive

transport is, the more often the nearest potential destination will be

chosen (Handy et al., 2005b). In the case where transport is expensive,

the distance from the residential location to this nearest facility, which is

derived from the spatial distribution of the whole range of potential

destinations, will largely determine the distance travelled by the consid-

ered individual.

The spatial separation of destinations, a trend which is often desig-

nated as “sprawl” is based on a rapid decline in transport costs combined

with an increase in travel speed (Ewing, 1994). However, based on the

peak oil theory (Witze, 2007), in time, an increase in transport costs is

expected due to oil scarcity. Although there is little discussion on the fact

that future fossil energy supply will have difficulties in matching global

demand, many uncertainties are present. Both the point in time when

peak oil will occur and the severity of the economic implications are

unclear. In addition, technology that reduces the oil dependence of the

transport system, such as the ongoing development and implementation

of the electric car, may mitigate these effects (Van Ruijven and Van

Vuuren, 2009). However, IEA (2008) estimates that the oil dependence of

the global transport system will only drop from 95% (in 2006) to 92% in

2030, mainly through biofuel substitution. So, even after taking into

account the development of alternative fuels, it is very likely that the

energy component of the cost of transport will gradually increase, with

implications for accessibility. Anticipating this by recognizing the princi-

ple of spatial proximity in the practice of spatial planning is thus

important.

Although much literature has yet emphasized the relationship be-

tween spatial characteristics and travel distance, in this paper we have

extrapolated the results of the analysis to a model that calculates and


237

visualises in which areas an additional residential unit would contribute

the least to mobility growth. Not unexpectedly, the most urbanized areas

turn out to be the most resilient and sustainable locations. This means

that a further increase of residential density and land use mix in urban

areas is the best guarantee for curbing excessive mobility and preparing

for the end of cheap oil. However, this conclusion requires some qualifica-

tion: there are limits to increasing density and land use mix targeted to

sustainable mobility patterns, primarily by environmental standards and

social desirability (Gordon and Richardson, 1997). Moreover, our results

suggest that a policy of compact development - compared to less steered,

dispersed development - will yield an only marginal direct impact on the

distances travelled. This finding is confirmed by numerous policy studies

(e.g. Echenique et al., 2009, p. 81).

7.7 Conclusion and directions

for further research

Although the influence of locally measurable features of spatial structure

on the distances travelled by individuals is statistically significant, the

explanatory power is small. However, in a context of possibly increasing

transport cost, but also of climate policies and congestion problems, this

does not mean that the spatial component is unimportant. Moreover, the

explanation that is provided by spatial variables is complementary to the

variance explained by socio-economic variables.

Using the analysis results in a policy instrument that can be used for

steering additional residential development is therefore useful. In order to

keep undesired mobility growth within certain limits, it is appropriate to

strengthen the urban character of existing cities as much as possible.

However, it is still possible to improve the model based on spatial and

functional disaggregation. This would enable the development of more

accurate sub-models that are optimized for one city or a part of a region,

or for one category of travel (e.g. the commute), or for one section of the

population (e.g. school children). Moreover, a refinement would not only

allow modelling the spatial distribution of residential locations, but also

of destination categories such as schools, jobs, shops or public services.

Chapter 7

238

References

Batty, M. (1974) “Spatial entropy.” Geographical Analysis. 6(1), pp. 1-

31.

Boussauw, K., T. Neutens and F. Witlox (2011a) “Relationship between






search. 11(1), pp. 42-60.

Boussauw, K., T. Neutens and F. Witlox (2011c) “Minimum commuting



press.

Boussauw, K., V. Van Acker and F. Witlox (2011d) “Excess travel in

non-professional trips: Why looking for it miles away?” Tijdschrift

voor Economische en Sociale Geografie. in press.



43(5), pp. 580-591.

Cervero, R. (1996) “Mixed land-uses and commuting: Evidence from the

American housing survey.” Transportation Research Part A. 30(5),

pp. 361-377.



2(3), pp. 199-219.



Studies. 23(3), pp. 377-401.

Echenique, M., A. Hargreaves, Y. Jin, G. Mitchell and A. Namdeo (2009)

London and the Wider South East Regions Case Study. Cambridge-

London-Bristol: Solutions Project.

Ewing, R. (1994) “Characteristics, causes, and effects of sprawl: A

literature review.” Environmental and Urban Studies. 21(2), pp. 1-15.

Ewing, R., K. Bartholomew, S. Winkelman, J. Walters and D. Chen

(2008) Growing Cooler: The Evidence on Urban Development and

Climate Change. Washington, DC: Urban Land Institute.


239

Ewing, R. and R. Cervero (2010) “Travel and the built environment: A

meta-analysis.” Journal of the American Planning Association. 76(3),

pp. 265-294.

Giuliano, G. (1995) “The weakening transportation-land use connection.”

Access. 6(1), pp. 3-11.

Gordon, P. and H. W. Richardson (1997) “Are compact cities a desirable

planning goal?” Journal of the American Planning Association.

63(1), pp. 95-106.

Handy, S., X. Cao and P. Mokhtarian (2005a) “Correlation or causality

between the built environment and travel behavior? Evidence from

Northern California.” Transportation Research Part D. 10(6), pp.

427-444.

Handy, S., L. Weston and P. Mokhtarian (2005b) “Driving by choice or

necessity?” Transportation Research Part A. 39(2-3), pp. 183-203.




IEA (2008) World Energy Outlook. Paris: International Energy Agency.

Janssens, D., E. Moons, E. Nuyts and G. Wets (2009) Onderzoek Ver-

plaatsingsgedrag Vlaanderen 3 (2007-2008). Brussels-Diepenbeek:

Ministry of the Flemish Community.

Maat, K. and H. Timmermans (2006) “Influence of land use on tour

complexity.” Transportation Research Record. 1977, pp. 234-241.

Mees, P. (2010) Transport for Suburbia: Beyond the Automobile Age.

London: Earthscan.



Part A. 38(2), pp. 143-162.

Næss, P. (2003) “Urban structures and travel behaviour. Experiences

from empirical research in Norway and Denmark.” European Journal

of Transport and Infrastructure Research. 3(2), pp. 155-178.

Næss, P. and S. L. Sandberg (1996) “Workplace location, modal split and

energy use for commuting trips.” Urban Studies. 33(3), pp. 557-580.

Nagendra, H. (2002) “Opposite trends in response for the Shannon and

Simpson indices of landscape diversity.” Applied Geography. 22(2),

pp. 175-186.

National Geographical Institute of Belgium (2009) Structuur en codering

van de gegevens top10v-gis en top50V-gis. National Geographical In-

stitute of Belgium from

Chapter 7

240

http://www.ngi.be/Common/articles/CA_Top10V-GIS_TOP50V-

GIS/restruct_doc_nl.htm.





Openshaw, S. and P. Taylor (1979) “A million or so correlation coeffi-

cients: Three experiments on the modifiable areal unit problem”. In:

N. Wrigley (Eds.) Statistical Methods in the Spatial Sciences. Lon-

don: Routledge and Kegan Paul, pp. 127-144.

Pisarski, A. E. (2006) Commuting in America III. Washington, DC:

Transportation Research Board.

Policy Research Centre on Work and Social Economy (2010) Gewestelijke

uitgaande pendel. Policy Research Centre on Work and Social

Economy from http://www.steunpuntwse.be/view/nl/18767. Re-

trieved 06/08/2010.

Rietveld, P. and R. Vickerman (2004) “Transport in regional science: The

‘death of distance’ is premature.” Papers in Regional Science. 83(1),

pp. 229-248.



Train, K. (2003) Discrete Choice Methods with Simulation. Cambridge,

UK: University Press.



10.1007/s11116-010-9309-6.

Van Ruijven, B. and D. P. Van Vuuren (2009) “Oil and natural gas

prices and greenhouse gas emission mitigation.” Energy Policy.

37(11), pp. 4797-4808.

Verhetsel, A., I. Thomas and M. Beelen (2010) “Commuting in Belgian

metropolitan areas: The power of the Alonso-Muth model.” Journal

of Transport and Land Use. 2(3), pp. 109-131.

Wegener, M. (2010) The future of mobility in cities: Challenges for urban

modelling. Proceedings of the World Conference on Transport Re-

search. Lisbon.



15(3), pp. 172-183.


241

Witlox, F. and H. Tindemans (2004) “The application of rough sets

analysis in activity-based modelling. Opportunities and constraints.”

Expert Systems with Applications. 27(4), pp. 585-592.

Witze, A. (2007) “That’s oil, folks….” Nature. 445(4), pp. 14-17.

Zwerts, E. and E. Nuyts (2004) Onderzoek Verplaatsingsgedrag Vlaande-

ren 2000-2001. Brussels-Diepenbeek: Ministry of the Flemish

Community.

Chapter 7

242

243

Chapter 8:

Conclusions and

policy recommendations

8.1 General conclusions

8.1.1 Introduction

For several decades now, analyzing the relationship between travel

behaviour and spatial structure has been a rewarding research topic.

Although its history goes back to Von Thünen (early 19th century) and

Alonso (early sixties) (Arnott and McMillen, 2006), this research line has

only been applied in an environmental framework since the eighties. For

an extensive, recent literature review in this context, we refer to Hickman

et al. (2009), which includes almost 250 studies. In the mean time, the

definition of the environmental problem has also evolved. Tangible

environmental and social effects of traffic such as air pollution, noise and

accidents have been constant concerns since the seventies. In the eighties,

these were complemented with the potential scarcity of fossil fuels; while

issues of global warming completed the debate in the nineties. The peak

oil phenomenon is thus a repackaged form of expected shortage of fossil

fuels that has only recently gained interest.

From the approach that views spatial planning as a means to influ-

ence mobility in a more sustainable direction, both “optimistic” and

“pessimistic” research results have surfaced: some researchers state that

there is a significant influence of spatial structure on travel behaviour,

while others argue the opposite (Van Acker, 2010, p. 273). When we rely

on the conclusions formulated in Chapter 7 and Chapter 8, it seems that

our own research finds only a weak, albeit statistically significant connec-

tion: i.e. the proportion of the variance in the distances travelled that is

explained by spatial structure is small. So, the assumed links are present

indeed, but the processes that underpin these may be too complex to be

steered by simple planning principles. The main reason for the only small

share of observed travel that is explained by spatial characteristics is

Chapter 8

244

however known from the literature on travel behaviour. In Chapter 7 it

has been demonstrated that the role of the spatial distribution of housing

and potential destinations is only secondary to the importance of socio-

economic characteristics. This means that e.g. age, gender, household

composition, disposable income and sector of employment are more

decisive for an individual’s travel pattern than the physical structure of

the residential environment. Recent research additionally involves aspects

of social psychology, and demonstrates that the variance which is still

present within coherent social groups is largely determined by individual

preferences, attitudes and lifestyles (Van Acker, 2010, p. 10). Hence, it is

important to keep these unpredictable human characteristics in mind

when discussing the link between travel behaviour and spatial structure.

A second caveat relates to the reliance on cross-sectional data. Chapter 4,

which is the only chapter that analyses time series, suggests that the link

between mobility and spatial structure weakened between 1981 and 2001.

Although the analysis of Chapter 4 (on commuting) cannot simply be

extrapolated to all travel, observed correlations should not be expected to

remain constant over time.

For the US, Ewing et al. (2008, p. 9) estimate the average difference

between the number of kilometres produced by compact development in

comparison to standard development at about 30%. This estimated

difference would be caused by a combination of a reduction in distance

travelled and a relative increase in the use of public transport. The real

gains that can be achieved also depend on the turnover in the real estate

market and the historical spatial structure. For the US, Ewing et al.

(2008, p. 9) estimate that compact development could lead to a 7 to 10%

reduction of transport-related emissions by 2050. In Europe, though, it is

quite possible that the potential gains are lower than in the US since the

European historical spatial structure is more compact, making opportuni-

ties for infill development rare. In addition, certain forms of the rebound

effect, in particular shifts from daily travel to occasional tourist and

recreational traffic, are usually not included in the analyses on which

these figures are based.

When we compare the expected impact of spatial policy with pricing

policy, the latter appears a much more efficient and controllable way to

curb traffic in the short term (Anas and Rhee, 2006; Rodier, 2009).

Nevertheless, pricing policies seem to bear more fruit within a spatial

structure that allows switching to alternative modes or choosing closer

Conclusions and policy recommendations

245

destinations. In that sense, compact development may contribute to

enhancing the effectiveness of non-spatial measures (Zhang, 2002, p. 3).

A context of rising energy prices, and hence rising transport costs, is

much more compelling than a mere policy objective aimed at reducing

emission levels. A lack of spatial proximity may turn oil price shocks

really problematic and strengthen the difference in accessibility between

metropolitan and rural areas in a way that has serious consequences for

the local economy and quality of life in remote areas. The six sub-studies

that make up this doctoral research can be interpreted from two policy

perspectives: (1) the required reduction in emissions from transport, and

(2) mitigation of accessibility problems caused by rising energy prices.

Below, in this context, we give a systematic overview of the conclusions

that we draw from these sub-studies, both for the commute and for non-

work related quasi-daily mobility.

In the next section we link our findings to the spatial policy that was

devised in the Spatial Structure Plan for Flanders (RSV, 1997/2004). The

last section, finally, gives an overview of research gaps and subsequent

avenues for further research.

8.1.2 Commuting

Chapters 2, 3, 4 and 6 focus on commuter travel. In the literature on

person mobility, commuting is clearly over-represented in respect to other

types of travel. As it is the case in this dissertation, the choice to study

the commute is in most cases data-driven. In many western countries,

and Belgium is no exception, the commute is particularly well docu-

mented. In Belgium, the decennial censuses (which basically cover all

commuters), supplemented with social security data and some special

surveys (Vanoutrive et al., 2010) constitute a very extensive database.

Quasi-daily non-commuting trips are only documented through survey

samples. Besides, in Belgium, these surveys are split up into the Travel

Behaviour Survey for Flanders (Onderzoek Verplaatsingsgedrag

Vlaanderen (OVG)) and the Belgian Mobility Survey (MOBEL) that is

mainly focusing on Wallonia and Brussels.

Chapter 2 can be considered as an exploratory study based on rela-

tively simple data, assumptions and methods. This chapter shows that

there are significant regional variations regarding commuting distance

and commuting energy consumption, viewed from the residential location.

Moreover, on a regional scale, energy consumption for commuting is

Chapter 8

246

highly correlated with commuting distance, which means that the

importance of mode choice is only of second order. These findings are put

in the context of the argument of Newman and Kenworthy (1989, 1999),

who found a strong negative correlation between population density and

per capita energy consumption for transport. Also, in Flanders, this

correlation exists, albeit much weaker.

The relatively narrow focus of Chapter 2 raises the question whether

the spatial variation in commuting distances can be explained by varia-

tions in the spatial proximity between houses and jobs, and if so, how

this proximity can be measured. This question is the basis for Chapter 3,

where a method is developed to calculate the spatial proximity between

houses and jobs based on spatially disaggregated values for the minimum

commuting distance. The minimum commuting distance is a concept

derived from the excess commuting literature. Unlike Chapter 2, the data

set used in Chapter 3 does not only contain the location of residence

(origin), but also the job location (destination) of the commuting flow. It

is therefore possible to develop two spatial proximity maps for the

commute: one that represents the spatial proximity of residential loca-

tions in relation to the job market, and another map showing the spatial

proximity of work sites in relation to the housing market. These maps

demonstrate that inhabitants of relatively remote areas with a low jobs-

housing balance have difficulties to find a job close to home. Moreover,

the inhabitants of these regions would even be more affected when all

employees would collectively look for a job closer to home, or would move

house to live closer to their jobs. Residents of urban areas, in contrast,

have wide margins to adapt their commuting behaviour, for example

under the influence of rising transport costs. The main cause of this

phenomenon is the large difference in spatial distribution between the

housing market and the job market. Jobs occur usually in a more concen-

trated form than houses do, and many agglomerations of jobs (especially

in industrial estates) are located relatively far from the residential

concentrations. Furthermore, the results of Chapter 3 seem to confirm the

assumptions that were raised in Chapter 2: i.e. those regions that are

characterized by a relatively low degree of spatial proximity, viewed from

the residential location, are broadly consistent with the regions where

above average commuting distances are reported. Regarding the job

locations, however, most regions with a high density of jobs score rela-

tively poorly in terms of spatial proximity.


247

In Chapter 2 a possible link is suggested between the increase in av-

erage commuting distance over the years and the evolution of spatial

proximity between the housing stock and the job market. In Chapter 4,

this hypothesis is assessed through time series of the minimum commut-

ing distance and partly also through the maximum commuting distance.

The spatial resolution of the used commuting matrices is limited to the

municipal level, restricting the utility of the results to the supra-

municipal level. Nevertheless, in many municipalities, a clear trend may

be discerned, indicating an average increase in spatial separation between

housing stock and job market. The reason is that the increase in jobs is

mainly found in the agglomerations, while the housing market is charac-

terized by a trend of further dispersion. More important, however, is the

finding that the growth of the commute is much faster than the observed

processes of spatial separation. We can therefore conclude that the

growth of the commute is a quasi-autonomous, prosperity driven process

that exhibits interaction with spatial separation processes although there

is no unidirectional causal link between these two phenomena.

Chapter 2 suggests an inverse relationship between population density

and per capita energy consumption for transport. However, this relation-

ship is not quantified, nor is there a link to the spatial characteristics of

the job location. Chapter 6 elaborates on this problem by selecting a

number of potential variables for spatial proximity, and by calculating

correlations between commuting distance and these selected variables (on

both ends of the commuting flow). Viewed from the residential location,

residential density, functional mix and overall accessibility are negatively

correlated with commuting distance. The minimum commuting distance,

as expected from Chapter 3, is positively correlated with the reported

commuting distance: this is the case both in residential locations and in

job locations. Job density (in job locations) is positively correlated with

commuting distance, as opposed to residential density in the residential

locations. This means that high density does not lead to a more sustain-

able commuting pattern if it is not accompanied with a high degree of

functional mix, while a skewed jobs-housing balance contributes to an

increase in commuting distance. With respect to the jobs-housing balance

we can say that the average commute is the shortest when the origin or

the destination of the trip is in an area where the jobs-housing balance is

close to or equal to 1.

However, the results of these four chapters should be interpreted

within the limitations of the data available and the methods used. The

Chapter 8

248

main constraint is the lack of differentiation based on sector, income class

or level of education. It is for example quite possible that highly educated

employees, who often perform specialized jobs, live less frequently in the

vicinity of the companies that require their qualifications. This is for

example the case for office jobs in central Brussels that are taken by

workers living in the green suburban belt outside the city. On the other

hand, low-skilled jobs occur relatively common in industrial areas that are

not always near a residential centre or agglomeration. According to Van

Acker and Witlox (2011) higher incomes are associated with longer

commuting distances. Differentiation would clarify this issue and allow

mapping the role of the spatial mismatch between qualifications and

preferences of employees and job requirements.

Moreover, in the sub-studies, too little attention has been paid to the

role of public transport. A high concentration of jobs around a major

railway station will have a smaller impact in terms of energy consumption

than a concentration of jobs on a remote industrial area that is virtually

only accessible by car. Therefore, in many planning strategies for office

development it is argued that specialized jobs which necessarily occur in

strong spatial concentrations because of the required agglomeration

benefits, should preferably be located in the vicinity of a main railway

station (Schwanen et al., 2004). Examples are the North quarter and the

European quarter in Brussels, or the office district of Amsterdam-Zuidas.

However, an analysis of the effectiveness of this planning concept is

beyond the scope of this dissertation.

8.1.3 Daily non-commuting travel

Although the findings of Chapters 2, 3, 4 and 6 clearly demonstrate a link

between the commute and the spatial proximity of housing and jobs,

these conclusions should not simply be extrapolated to other forms of

quasi-daily mobility. Chapters 5 and 7 try to extend the methodology to

travel patterns that are not part of the commute.

In Chapter 5 the concepts of the minimum commuting distance and

excess commuting are applied to quasi-daily trips that are not part of the

commute. The results are displayed in the form of a spatial proximity

map, representing for every census ward the relative proximity to a

bundle of facilities. In addition, for each spatial class an aggregated excess

rate is calculated, which indicates how strong the relationship is between


249

the degree of spatial proximity to facilities and the actual distance

travelled.

The proximity map indicates which areas are in fact too remote to

possibly achieve sustainable travel patterns in terms of access to everyday

amenities. The comparison of the minimum required travel distance to

the observed travel distance indicates that the relationship between

spatial proximity and travel patterns is stronger in the most remote areas

compared to more urbanized areas, even though the distance travelled is

on average larger in the more remote areas. In other words, residents of

rural areas are more likely to choose the closest possible destination than

residents of more urbanized areas. Metropolitan areas and urban agglom-

erations, which together constitute the most urbanized of all assessed

areas, however, are an exception to this rule: here the actual distances

travelled are short, while the nearest possible destination is chosen

relatively often.

Chapter 7 combines the variables that were developed in Chapter 3, 5

and 6 with the method of analysis developed and explained in Chapter 6.

The spatial correlation analysis of Chapter 6 is extended to a multivariate

regression analysis based on a sample provided by the 2007 phase of OVG

(Janssens et al., 2009). In order to get an overall picture of the impact of

all quasi-daily trips together, commuting and non-commuting trips were

combined and so the total distance travelled per respondent was ana-

lyzed. By extrapolating the obtained regression equation to the whole

study area, a new, more sophisticated, proximity map is generated.

Unlike the analysis in Chapter 6, Chapter 7 only includes the spatial

characteristics of the residential location of the respondents, limiting the

validity of the conclusions to the spatial characteristics of the residential

environment. A major finding of Chapter 7 is that the importance of the

spatial distribution of jobs fades completely into the background when

studying the aggregated travel pattern. So, the impact of the spatial

distribution of jobs on total mobility is relatively unimportant in relation

to the spatial distribution of other quasi-daily destinations. Population

density, functional mix and spatial proximity of facilities, however,

remain intact as determining factors. However, the overall explanatory

value of a model based on spatial properties is rather limited. This means

that attempts to shrink mobility patterns based on spatial policies are

doomed to lead to poor results unless combined with other measures (e.g.,

pricing) or concomitant with (autonomously) rising energy prices.

Chapter 8

250

Chapters 5 and 7 also have their limitations. The main shortcoming

can be summarized as the absence of an analysis based on the spatial

characteristics of the destination. In Chapters 3, 4 and 6 job locations are

also involved in the assessment. In Chapters 5 and 7 we do not gain

insight on the influence of the location of the visited shops, schools,

recreation facilities, et cetera. The reason why such analysis is beyond the

scope of this dissertation is, on the one hand, the lack of a sufficiently

large data set that contains the necessary information, and on the other

hand, the high level of complexity of the issue. Nevertheless, this is an

important track for possible future research. To our knowledge, in

Flanders or Belgium there is no research available that is for example

quantifying the total difference in energy efficiency between suburban

hypermarkets and small urban convenience stores, or between large and

small schools, or between more and less specialized educational institu-

tions.

8.1.4 Non-daily travel

Occasional (i.e. non-daily) trips, including excursions, weekends out, city

breaks, holidays and business trips are not the subject of the six sub-

studies of this dissertation. Although the study of this type of trips is

usually counted as tourism or business travel research, and not as

mobility research in the narrow sense, we still want to give this issue

some thought. As suggested in the introduction, it is quite possible that

the slow-down of the growth of mobility in the segment of quasi-daily

trips is compensated by continued growth within the segment of occa-

sional mobility. This shift would also be speed-driven, mainly based on

the far-reaching democratization of air travel. Moreover, there may also

be a link between the amount of travel (and thus the associated level of

energy consumption) that results from occasional mobility, and the

spatial structure of the residential environment of the traveller (Holden

and Norland, 2005). In order not to leave this hypothesis without any

evidence, in an addendum we present a brief analysis, based on the

methodology used in Chapter 5.


251

8.2 Options for spatial planning policy

8.2.1 The sustainable mobility paradigm

Banister (2008) describes a new mobility paradigm, in which sustainable

development is paramount and the sense of a further increase in mobility

is questioned. Urry (2002) depicts proximity as a necessary condition for

the occurrence of social interaction, which he calls co-presence, without

automatically being linked with extensive mobility. In policy terms, this

means that improving accessibility should become the objective, as

opposed to intensifying mobility. The ultimate goal is to achieve a high

degree of accessibility based on a minimum of mobility and therefore a

minimum amount of traffic. The above shows that steering spatial

development should be viewed first as a way to enhance resilience to oil

scarcity and more expensive transport, and second as working on a

spatial framework suitable to facilitate climate policies.

Based on the scheme of Banister (1999, pp. 316-319) we make a brief

translation of our research results in recommendations towards spatial

planning policy:

• Wherever possible, developments should be based on a high residen-

tial density, also in areas where, apart from housing, other activities

occur. The desired level of residential density depends on the level of

ambition. Density thresholds mentioned in the literature are usually

linked to the desired role of public transport and were discussed in

Chapter 1. The eco-districts Vauban and Rieselfeld in the German

Freiburg-im-Breisgau, known as prototypes of green contemporary

residential urban extensions, have a gross residential density that

goes up to about 150 inhabitants per hectare (Ryan and Throgmor-

ton, 2003), which is comparable to the most densely populated

neighbourhoods in the 19th century belts of the larger Belgian cities.

However, in Flanders, the aspiration level of the Spatial Structure

Plan for Flanders (RSV, 1997/2004) is limited to 25 dwellings

(equivalent to approximately 55 people) per hectare in urban areas

and 15 dwellings (equivalent to approximately 38 people) per hectare

in the outlying area. Moreover, a progressive shrink of the average

household size threatens to bring down the effective residential den-

sity on the basis of these standards (which are expressed in dwellings

per hectare).

Chapter 8

252

• The size of the nuclei must be large enough to assure a relatively high

degree of self-sufficiency, based on agglomeration economies. Follow-

ing Banister (1999), we could argue that towns or villages with less

than 25,000 residents should in principle not be allowed to grow

anymore, while isolated cities that are not part of an agglomeration

should in time reach a size of at least about 50,000 inhabitants. In

addition, growth should be directed as much as possible in or imme-

diately subsequent to the agglomerations (> 250,000 inhabitants),

where public transport can play an important role.

• New developments should be located within or next to existing urban

areas. Facilities and jobs should be planned and developed simultane-

ously with housing, preferably concentrated in local centres.

Moreover, new developments should have good non-car accessibility.

8.2.2 Possible adjustments to the Spatial

Structure Plan for Flanders

As such, these principles, relying on compact development and economies

of agglomeration, are not new. In developing the RSV (1997/2004), a

number of these items were already taken into account:

• The RSV requires that 60% of additional housing units should be

built in urban areas, and only 40% in the nuclei of the outlying area.

This rule differs from the trend that was observed in the early nine-

ties, when most new homes were built outside the urban areas. To

meet this objective in practice, the urban areas are demarcated and

additional residential land is designated in those urban areas.

• The RSV requires a minimum residential density for new develop-

ments (25 dwellings per hectare in urban areas and 15 dwellings per

hectare in the outlying area).

• Interweaving of functions and activities is a guiding principle in the

development of urban areas.

• Offices are preferably concentrated at nodes of public transport.

• Amenity levels should be attuned to the importance of the respective

urban areas.

Based on our own research results and literature review we suggest some

possible adjustments to the perspective of the RSV regarding the steering

of new developments. We summarize these adjustments in Fig. 8.1. The

main research results we have been drawing from in order to produce Fig.

8.1, are represented in Figs 3.1 and 3.2, in Fig. 5.3, and in Fig. 7.2. The


253

schematic map that is presented in Fig. 8.1 and is explained below, can

be considered as a possible spatial development perspective that is

motivated from optimizing accessibility based on a minimum amount of

traffic, by maximizing spatial proximity and valorising economies of

agglomeration. This development perspective includes the following

elements:

• The distinction of areas suitable for additional residential develop-

ment and areas where less residential development, or even a status

quo, is desirable, should be based on the existing agglomerations. The

overall degree of spatial proximity of a municipality located near a

metropolitan area is much higher than in a remote small urban area.

It is therefore important to exploit and strengthen existing agglom-

eration effects. The demarcation of the areas that are designated as

“conurbation” in Fig. 8.1 is done on the basis of various indicators of

spatial proximity, and these should be considered as search areas for

additional housing. The areas classified as “potential conurbation”,

are today probably too small or too remote to be considered yet as a

conurbation, but could become a conurbation in the future through

directed growth. Perhaps other areas exist that rather belong to this

category, but were not indicated as such on the map. The term “ur-

ban area” should therefore be extended to a conurbation, with the

intention to grant the existing suburban areas in the conurbations a

more urban character (by increasing density and functional mix)

while stopping the growth of peri-urban developments.

• The proportion of all new dwellings that should be directed into these

conurbations will depend on arguments from sectors other than mo-

bility. From the accessibility perspective, preferably 100% of

additional homes would be located in the conurbations. A variant of

this could allow a minimal amount of new housing outside the conur-

bations to meet the locally occurring shrink of family size, or could

allow for natural growth of the local population. The greater the pro-

portion of additional dwellings that will be located in the

conurbations, the wider the demarcation of the external border of the

conurbation can be. In case there would still be built houses outside

the conurbations, these should be built first in small urban areas, and

only in second order in the nuclei of the outlying area. To ensure the

compactness of the conurbation, the desired housing density should

be adjusted upward: an average gross density of 80 to 120 inhabitants

per hectare appears feasible.

Chapter 8

254

• By concentrating residential development in the conurbations, a

reduction in the total number of vehicle kilometres travelled will be

facilitated. Since it is mainly the mobility which would otherwise

have occurred outside the conurbation that will be suppressed, such a

development will not result in relief of congestion in the conurbations.

The enhanced agglomeration effect, however, provides a stronger ba-

sis for an efficient public transport system. Each demarcated

conurbation should aim to increase the internal accessibility by public

transport (based on restricted tram and (trolley)bus lanes and (light)

rail), and also by bike. By discouraging car traffic that enters the

conurbation, additional space can be created to improve the internal

accessibility. Additional budget for public transport should mainly be

directed at developing these conurbation networks, complemented

with transferia to receive car traffic from the outlying area.

• Within the conurbations, a well-balanced spatial distribution of

facilities should be pursued, preferably concentrated in centres and

sub-centres. Upscaling, in the sense that several smaller establish-

ments such as schools, shops and workshops are replaced by one

campus, one hypermarket or one industrial estate should be discour-

aged because of the consequent reduction of the level of spatial

proximity.

• In small urban areas, as in the outlying area, the amount of addi-

tional houses to be built should be minimized (preferably to zero),

pursuing a status quo of the population. The small urban areas

should serve as subregional centres with the fullest possible range of

amenities and a variety of jobs that meet the needs of the service

area that is surrounding every small urban area. By concentrating

facilities in small urban areas, and also by expanding the supply of

services, the level of spatial proximity in the surrounding service area

will increase, reducing the need for residents to travel to a conurba-

tion. By enlarging the supply of jobs, the average distance between

home and work locations will be reduced. The emphasis should be on

non-specialized jobs (which have a weaker link with the conurbations

and also generate less long-distance commuting), itinerant and home-

based occupations (the latter jobs also fit in the actual countryside).

For commuters from the small urban areas and the surrounding ser-

vice area, rail links to the conurbations should be optimized. Based

on these principles, in Fig. 8.1 the selection of small urban areas is

reproduced from the RSV, unless included in a conurbation. Based on


255

arguments from other sectors this selection may be changed, or possi-

bly supplemented with certain economic nodes or main villages.

• Since highly specialized office jobs generate long commuting dis-

tances, these should be directed as much as possible to central

locations near major railway stations. In order not to encourage long-

distance commuting, these sites are less suitable for non-specialized

jobs where the aim should be to recruit staff locally wherever possi-

ble. Central locations that have good access by train are limited in

number: we are talking about the major stations in Brussels and

Antwerp, possibly supplemented with Mechelen, Gent and Leuven.

The surroundings of the other major railway stations in Flanders are

also suitable for the establishment of institutions of higher education.

• In an integrated policy plan, the selection and demarcation of urban

development areas should of course not only be determined by mobil-

ity. Economic aspects (such as the siting of major port and industrial

areas that may entail long distances between home and work loca-

tions), the presence of valuable natural areas, forests and landscapes,

agricultural areas and heritage are all factors that should be taken

into account when planning future development. Moreover, the neces-

sary supply of additional housing will be determined on the basis of

demographic trends.

• Spatial policies aimed at strengthening spatial proximity may be

supported by an appropriate fiscal policy. Enhancing the variability

of the transport cost as a function of the distance travelled may help

mitigating differences in the real estate prices between the conurba-

tions and the outlying areas.

Although the basic principles are similar, the spatial development per-

spective outlined in Fig. 8.1 contains a few noticeable differences

compared to existing policy plans and studies. Below, we touch upon the

most striking elements, comparing with both the RSV (1997/2004) and

the Spatial-Economic Main Structure (Cabus et al., 2001), which is a

spatial perspective on the current and future economic development in

Flanders.

Compared with the RSV perspective, which is shown schematically in

Fig. 6.14, we note the following:

• The demarcation of the regional urban areas and metropolitan areas

is expanded, some of which may even grow together.

Chapter 8

256

• The small urban areas are no longer regarded as development areas

for additional homes, but rather as employment and service centres

for the surrounding outlying areas.

• The requirements in terms of residential density and functional mix

in the (re)development of urban neighbourhoods are revised signifi-

cantly upwards.

Fig. 8.1. Spatial development perspective based on maximum access and

minimum mobility

The main difference compared with the vision of the Spatial-Economic

Main Structure (Cabus et al., 2001) is that the economic development

zones, which coincide with the conurbations as suggested in Fig. 8.1,

supplemented with the major ports and industrial areas (which are not

shown in Fig. 8.1) and to a lesser extent the small urban areas, are

viewed much more compact.

While this study has analyzed a number of potential problems and

has issued some possible guidelines based on this analysis, bringing these

suggestions into practice is not obvious. In Flanders and Belgium, the

role of the government is largely limited to the designation of locations

where housing and industry is allowed to develop, along with the plan-

ning of the infrastructure network and public transport. The density of

residential (re)development is hardly centrally controlled, and the legal

and fiscal framework is rather targeted at providing legal certainty than


257

at creating future-oriented dynamics. Moreover, no efficient economic

planning tools are available that can significantly influence the location of

jobs. Also, the location and size of schools and shops are hardly steered

by a spatial development perspective. So, the main challenge is perhaps

not to be found in research and planning theory, but in the implementa-

tion of the principle of compact urban development.

8.3 Some directions for further research

Results of research in terms of spatial planning and transport policy will

never be entirely satisfactory. Each new analysis raises additional ques-

tions. Moreover, history tells us that planning is not a science. Analyses

and models may underpin policy choices, but remain only one element in

decision making processes. Nevertheless, quantitative research is impor-

tant, and should not fade into the background of planning processes.

Throughout this dissertation, research gaps have been pointed out

continuously, along with the necessary caution that should be observed

when drawing conclusions from the research results. Below we give a list

of possible avenues for further research that build on this dissertation.

Within commuting research, the following directions for further re-

search can be put forward:

• The application of the concept of minimum commuting distance

(based on the method described in Chapter 3) and correlation analy-

sis (as in Chapter 6) based on commuting data that differentiate

between economic sector, level of education and income class.

• The analysis of time series of minimum commuting distance based on

detailed zoning (traffic analysis zones or census wards), with the aim

of measuring processes of spatial separation and sprawl (based on the

method described in Chapter 4).

• Calculating variations in the energy efficiency of the commute to

offices and industrial estates based on the type of location and taking

into account the mode choice, the economic sector, the level of educa-

tion and the income class of staff members.

Within the study of quasi-daily mobility outside the commute, the

following future research tracks may be distinguished:

• Refinement of the concept of the spatial proximity map based on

residential locations (see Chapter 5) and of the underlying methods of

measurement.

Chapter 8

258

• The development of a similar map based on the spatial proximity of

destinations (complementary to the current map which relies only on

residential locations, considered as origins).

• The analysis of time series of the spatial proximity map(s) with the

aim of measuring processes of spatial separation and sprawl (as a

combination of the methods from Chapters 4 and 5).

• Modelling the differences in overall energy efficiency for transport

between small scale and large scale retailing (taking into account

both energy consumption by goods supply and by customer’s trips).

• Modelling the differences in energy efficiency of the school commute

between networks of more numerous, but smaller, schools versus less

numerous, but larger, schools.

Further research based on these suggestions would contribute to a well-

informed practice of spatial planning, which aims to ensure a high level of

accessibility in combination with a sustainable form of mobility.

References

Anas, A. and H.-J. Rhee (2006) “Curbing excess sprawl with congestion

tolls and urban boundaries.” Regional Science and Urban Economics.

36(4), pp. 510-541.

Arnott, R. and D. P. McMillen (2006) A Companion to Urban Econom-

ics. Malden, MA - Oxford - Victoria: Blackwell Publishing.


Review. 70(3), pp. 313-338.


Policy. 15(2), pp. 73-80.

Cabus, P., G. Allaert and W. Vanhaverbeke (2001) De Ruimtelijk-

Economische Hoofdstructuur van Vlaanderen: Concept in Hoofdlij-

nen. Gent: Academia Press.

Ewing, R., K. Bartholomew, S. Winkelman, J. Walters and D. Chen

(2008) Growing Cooler: The Evidence on Urban Development and

Climate Change. Washington, DC: Urban Land Institute.

Hickman, R., C. Seaborn, P. Headicar, O. Ashiru, S. Saxena, D. Banister

and T. Pharoah (2009) Land Use and Transport: Settlement Patterns

and the Demand for Travel. Oxford: Commission for Integrated

Transport.


259

Holden, E. and I. Norland (2005) “Three challenges for the compact city

as a sustainable urban form: Household consumption of energy and


Studies. 42(12), pp. 2145–2166.

Janssens, D., E. Moons, E. Nuyts and G. Wets (2009) Onderzoek Ver-

plaatsingsgedrag Vlaanderen 3 (2007-2008). Brussels-Diepenbeek:

Ministry of the Flemish Community.





Rodier, C. (2009) “Review of international modeling literature: Transit,

land use, and auto pricing strategies to reduce vehicle miles traveled

and greenhouse gas emissions.” Transportation Research Record.

(2132), pp. 1-12.



Ryan, S. and J. A. Throgmorton (2003) “Sustainable transportation and

land development on the periphery: A case study of Freiburg, Ger-

many and Chula Vista, California.” Transportation Research Part D.

8(1), pp. 37-52.

Schwanen, T., M. Dijst and F. M. Dieleman (2004) “Policies for urban

form and their impact on travel: The Netherlands experience.” Urban

Studies. 41(3), pp. 579-603.

Urry, J. (2002) “Mobility and proximity.” Sociology. 36(2), pp. 255-274.

Van Acker, V. (2010) Spatial and Social Variations in Travel Behaviour.

PhD Thesis. Ghent: Ghent University.



10.1007/s11116-010-9309-6.

Vanoutrive, T., L. Van Malderen, B. Jourquin, I. Thomas, A. Verhetsel

and F. Witlox (2010) “Mobility management measures by employers:

Overview and exploratory analysis for Belgium.” European Journal of

Transport and Infrastructure Research. 10(2), pp. 121-141.



nology.

Chapter 8

260

261

Addendum:

The spatial component

of air travel behaviour:

An exploration An earlier version of this addendum has been published (in Dutch) as Boussauw (2009a) in the journal Agora and as Boussauw (2009b) on the website Low Tech Magazine.

A.1 The aeroplane: the forgotten

transport mode

While urban and interurban person mobility still seems to increase, fuel consumption and emissions are stagnating. Cleaner cars and improved public transport play a role in this evolution, but it is also true that mobility is slowly but surely clashing with structural capacity limits. Good news for the climate, so it seems. However, data on this, reported by the Flemish Environment Agency (Vlaamse Milieumaatschappij (VMM)) show an important gap (De Vlieger et al., 2007). International aviation is systematically not included in the statistics. However, air travel is the fastest growing segment of the transport sector, which is also responsible for bafflingly large volumes of CO2 emissions from the com-bustion of equally large volumes of fossil fuels. In 2008, the number of passengers in Belgium increased by 5.8% over the previous year, while road traffic, exceptionally, slightly decreased (-1%) (Statistics Belgium, 2009).

But what amount of emissions are we talking about? Since there are no published statistics on fuel consumption by flying inhabitants of the Flanders region, we made a rough estimate based on flight data for 2008 from the Belgian aviation authorities (Brussels Airport, 2009; Statistics Belgium, 2009). According to the demographic composition we assumed that out of the 45% Belgian passengers using Belgian airports, 58% are

Addendum

262

living in Flanders. The travel distances were estimated based on distance classes, provided by Brussels Airport (2009), and complemented with data on the destinations served by the regional airports. Some examples: for flights to Mediterranean resorts, the average length of a flight was estimated at 2,000 km, while for North American destinations this amounted to 7,500 km, and to 1,500 km for Eastern European destina-tions.

Based on a report by the Danish Environmental Protection Agency (2003) we assumed an average of 5 litres of kerosene fuel consumption per 100 km, for each occupied seat. For flights of less than 1,000 km, fuel consumption is generally higher. Moreover, consumption depends strongly on the type of aircraft, and the occupancy rate. No detailed information is available on the used equipment. This is also the case for the occu-pancy rate, which might be much lower since the beginning of the financial crisis, than before. All these factors make our estimate of fuel consumption perhaps rather cautious (underestimate).

Converted to PJ (petajoule = 109 MJ (megajoule)), we find a total consumption by Flemish air travellers of approximately 69.0 PJ. For overland person transport, VMM (for 2006) reports a total consumption of 121.6 PJ (De Vlieger et al., 2007). This means that aviation is respon-sible for more than one third of total energy consumption and hence CO2 emissions for transport of inhabitants of Flanders. And this proportion is increasing rapidly.

It should not be forgotten that one litre of kerosene burned at high altitudes contributes much more to global warming than the consumption of the same amount of energy on the ground level. At altitude also the emission of water vapour and nitrogen oxides has a significant contribu-tion, in addition to the effect of carbon dioxide. This has, among other things, to do with cloud formation, caused by jet engines. According to Åkerman (2005), the total impact of aviation on global warming is 2.7 times greater than the impact of CO2 emissions from aviation alone.

Despite the rapid increase in air traffic, the aircraft remains an excep-tional transport mode for the average inhabitant of Flanders. In the Travel Behaviour Survey for Flanders (OVG) (2000-2001) 52.6% of the respondents indicate not to fly at all (Zwerts and Nuyts, 2004). Less than half of the travellers who used Brussels Airport in 2008 were flying more than four times a year (Brussels Airport, 2009). The vast majority of the emissions from air travel is therefore caused by a small minority of the population.

The spatial component of air travel behaviour: An exploration

263

A natural upper limit to the demand for air travel is not yet in sight. On average, an inhabitant of Flanders flies annually once to, say, Egypt and back. Still, there seems to be a lot of growth opportunity before the saturation level of the average tourist will be reached. The price of air tickets continues to fall, thanks to declining fixed costs. Only the price of fuel, or the introduction of taxes could constitute an impediment to growth. For the time being, however, the oil price remains at an accept-able level, and important eco-taxes are almost nonexistent. Within the current framework a rapid growth in air travel, both in the western world and globally, is inevitable.

Attempts to provide technological solutions have little success. The environmental problem of aviation has in fact only a limited technological dimension: per person kilometre, a plane is about as efficient as a car. The essence of the problem is that the availability of cheap air transport induces long-distance travel, and thus leads to an explosion in the number of kilometres travelled. Given the huge volumes of fuel we are talking about, tentative experiments with aviation on biofuels are of little significance to the market.

Adding aviation figures to fuel consumption and emissions statistics of person transport leads to new insights. On the one hand, in the context of energy and climate issues it appears suddenly very easy to reduce energy consumption and emissions: imposing a high tax on flying would undoubtedly result in a significant reduction in the number of passengers. Besides, the economic downturn would be largely confined to the tourism sector (comprising about two thirds of air passengers), which is much more price sensitive than the business sector and also less critical to the domestic economy.

On the other hand, policy efforts in the field of alternative transpor-tation modes sink away when these are viewed in the light of the growth of aviation. Does it make sense to abandon the car for daily travel, as air traffic continues to grow as fast yet? Moreover, there is no significant societal support to curb air traffic growth, especially not in the non-western world.

Addendum

264

A.2 Urban versus rural lifestyle

Based on Norwegian research, Holden and Norland (2005) suggest that there is not only a link between spatial structure and daily travel pat-terns, but also with air travel. They argue that city dwellers make more unsustainable long distance trips compared to non-city dwellers. They quote several possible reasons for this. As city dwellers have less often a garden, they would need more vacation outdoors. Living in the city would more often be accompanied with an internationally oriented lifestyle. Moreover, city dwellers would have lower daily travel expenses, perhaps because they own less often a car, leaving them with more financial headroom to pay for, among other things, plane tickets.

While in Flanders the contrast between urban and rural areas may be less significant than in Norway, it is still worth to assess the validity of the assertion of Holden and Norland (2005) also in this context. However, the available data is quite limited. We rely on the OVG survey, which distinguishes between respondents who never fly (the reference category) and those who are occasional or frequent flyers. Of each respondent, we know the municipality of residence. Based on Luyten and Van Hecke (2007), we assigned all municipalities into four categories: urban agglom-eration, suburban area, commuter area and rural area (the residual category). This classification was made on an empirical basis, and therefore does not always reflect the policy oriented classification of the Spatial Structure Plan for Flanders (RSV, 1997/2004). In addition, the OVG survey asked a question on the respondent’s perception of his residential environment: we distinguish between people who think they live in a centre (dense built-up area) and those who think they live outside a centre.

We include the variables in two logistic regression models, considering air travel behaviour as the dependent variable. In the first model, the location of the municipality of residence is introduced as an explanatory variable, while in the second model the perception of the residential environment is used. We present the output of these two regressions in Table A.1 and Table A.2 respectively.

The column exp(B) can be interpreted as the odds ratio of the prob-ability that a respondent in this category is flying in comparison to the reference category. Table A.1 shows that the probability that a resident of the urban agglomeration flies is a factor 1.6 higher compared to a rural resident. For residents of suburban and commuter areas, this odds ratio is


265

1.2 to 1.3. From Table A.2 we learn that the probability that a resident of a centre or a dense built-up area is flying is 1.2-fold higher compared to someone living outside a centre. All results are statistically significant at the 0.05 level. Table A.1. Odds ratios of the probability of flying when not living in the rural area

category exp(B) p-value

intercept - 0.000

urban agglomeration 1.576 0.000

suburban area 1.206 0.014

commuter area 1.262 0.000

rural area (reference category) - -

Table A.2. Odds ratio of the probability of flying when living in a centre

category exp(B) p-value

intercept - 0.000

centre 1.157 0.002

outside centre (reference category) - -

Although the available data is too rough to allow quantifying variations in energy consumption and CO2 emissions, our analysis seems to confirm the argument of Holden and Norland (2005). If we take aviation into account, then the travel pattern of the city dweller may be a lot less sustainable than we thought.

A.3 Rebound effect and policy implications

The above findings fit the theory of the so-called rebound effect (Alcott, 2005). Energy savings in one field are compensated by more consumption in another field, as long as the disposable budget remains the same. A small local ecological footprint, which is made possible by an urban lifestyle, is compensated by a large global ecological footprint which in this case can be written on behalf of air travel.

A macro-economic extension of this thesis is known as the Khazzoom-Brookes postulate (Saunders, 1992). This states that at constant energy prices, increasing energy efficiency does not lead to a decrease, but rather

Addendum

266

to an increase in the volume of energy consumed globally. The reason for this is that greater efficiency leads to more prosperity with relatively low market prices for energy, compared to the baseline scenario. The conse-quent wealth surplus can be spent easily to energy-intensive consumption. Applied to our subject, this means that the money that is saved by an individual through better thermal insulation of his house or by driving a more fuel-efficient car, is easily invested in activities where efficiency seems less important, such as vacations. The rapid growth of air travel indicates that tourist flights play an important role in this mechanism. This observation adds a note of caution to a climate policy that is based solely on encouraging more energy efficiency.

Air traffic increase is conveniently overlooked in many mobility stud-ies and environmental policy plans. For example, the Climate Policy Plan for Flanders (LNE, 2006) includes no measures related to aviation. Nevertheless, in this plan efforts to promote cycling are considered relevant in terms of reducing CO2 emissions, even though the results expected from this are only marginal compared to the emissions from aviation.

Besides the climate issue, also the depletion of fossil energy is a good reason to deal judiciously with the available oil resources. In this light, inducing the demand for city breaks may not be the most sustainable strategy. However, this is exactly what happens today through partner-ships between low-cost airlines and diverse government agencies (Monbiot, 2009).

In addition, we have brought up a surprising aspect of the urban life-style. An urban spatial structure may encourage sustainable local travel behaviour, it has no control over global consumption patterns. Urban dynamics rather stimulate air travel demand. When local mobility clashes with structural capacity limitations, the desire for more interaction with the world results possibly in flying.

References

Åkerman, J. (2005) “Sustainable air transport - On track in 2050.” Transportation Research Part D. 10(2), pp. 111-126.

Alcott, B. (2005) “Jevons’ paradox.” Ecological Economics. 54(1), pp. 9-21.


267

Boussauw, K. (2009a) “Stadsmens onderweg: Een duurzaamheidspara-dox.” Agora - Magazine voor Sociaalruimtelijke Vraagstukken. 25(5), pp. 7-10.

Boussauw, K. (2009b) Hoe duurzaam leeft de stadsbewoner? Low Tech Magazine from http://www.lowtechmagazine.be/2009/12/hoe-duur zaam-is-de-stadsbewoner.html.

Brussels Airport (2009) Brutrends 2008. Brussels: Brussels Airport. Danish Environmental Protection Agency (2003) Greenhouse Gas

Emissions from International Aviation and Allocation Options. Co-penhagen: Danish Environmental Protection Agency.

De Vlieger, I., E. Cornelis, L. Int Panis, L. Schrooten, L. Govaerts, L. Pelkmans, S. Logghe, K. Vanherle, F. Vanhove, G. De Ceuster, C. Macharis, T. Festraets, E. Pekin, L. Turcksin, K. Van Bladel, M. de Jong, J. Van Mierlo, J. Matheys, J.-M. Timmermans, C. De Geest and E. Van Walsum (2007) MIRA (2007) Milieurapport Vlaanderen, Achtergronddocument 2007, Transport. Aalst: Flemish Environment Agency.

Holden, E. and I. Norland (2005) “Three challenges for the compact city as a sustainable urban form: Household consumption of energy and transport in eight residential areas in the greater Oslo region.” Urban Studies. 42(12), pp. 2145–2166.

LNE (2006) Vlaams Klimaatbeleidsplan 2006-2012. Brussels: Ministry of the Flemish Community.

Luyten, S. and E. Van Hecke (2007) De Belgische Stadsgewesten 2001. Brussels: Statistics Belgium.

Monbiot, G. (2009) “Subsidising the climate crash.” The Guardian. (7th July).

RSV (1997/2004) Ruimtelijk Structuurplan Vlaanderen - Gecoördineerde Versie. Brussels: Ministry of the Flemish Community.

Saunders, H. D. (1992) “The Khazzoom-Brookes postulate and neoclassi-cal growth.” Energy Journal. 13(4), pp. 131-148.

Statistics Belgium (2009) Mobiliteitsportaal from http://www.statbel. fgov.be/port/mob_nl.asp.

Zwerts, E. and E. Nuyts (2004) Onderzoek Verplaatsingsgedrag Vlaanderen 2000-2001. Brussels-Diepenbeek: Ministry of the Flemish Community.

Addendum

268

269

Samenvatting

S.1 Overzicht

Deze verhandeling wil het inzicht in de wederzijdse relatie tussen mobili-

teit en ruimtelijke ontwikkelingen vergroten, en het probleem situeren in

een context van klimaatbeleid en peak oil. Dit wordt gedaan op basis van

literatuurstudie en de toepassing van een aantal, voor een deel nieuw

ontwikkelde, onderzoeksmethoden op de case study Vlaanderen.1

Dit erg ruim geformuleerde onderzoeksveld werd vernauwd tot perso-

nenmobiliteit, waarbij we onze doelstelling richten op het onderzoeken

van de duurzaamheid van de ruimtelijke structuur ten aanzien van

verplaatsingsgedrag, met bijzondere aandacht voor de dagelijks afgelegde

afstanden. Duurzaamheid wordt gedefinieerd in de zin van het robuust

zijn, niet enkel voor een groeiende maar ook voor een in de toekomst

mogelijk krimpende mobiliteit. Een algemeen krimpende mobiliteit is een

scenario dat zich kan ontwikkelen tengevolge van stijgende energiekosten

(het peak-oil scenario) of een stringent klimaatbeleid, terwijl een selectie-

ve krimp van de mobiliteit (slechts van toepassing op een deel van de

bevolking) zich kan voordoen door een toenemende verzadiging van het

verkeerssysteem. Bovendien speelt de ruimtelijke structuur een rol in de

mogelijke sturing van het verplaatsingsgedrag in een meer duurzame

richting.

Deze verhandeling volgt de klassieke wetenschappelijke onderzoekslijn

over het verband tussen ruimtelijke structuur en verplaatsingsgedrag, die

al decennia in ontwikkeling is. Een voorlopige algemene conclusie van

deze onderzoekslijn in een context van duurzaamheid zou als volgt

kunnen worden geformuleerd: “Een duurzaam verplaatsingspatroon kan

slechts tot stand komen binnen een ruimtelijk kader dat daar geschikt

voor is, maar andere maatregelen (financieel en regulerend) zijn nodig om

het verplaatsingsgedrag effectief te wijzigen. Met andere woorden:

ruimtelijke ordening is nodig, maar het is niet voldoende” (naar Zhang,

1 De aanleiding voor dit onderzoek bevindt zich in de taakstelling van het

Steunpunt Ruimte en Wonen (2007-2011), beleidsondersteunend onder-

zoeksconsortium dat aangestuurd wordt door het Departement Ruimtelijke

Ordening, Woonbeleid en Onroerend Erfgoed van de Vlaamse overheid.

Samenvatting

270

2002, p. 3). De ruimtelijke kwaliteit die een verplaatsingspatroon op basis

van korte afstanden faciliteert noemen we “ruimtelijke nabijheid”, zelfs al

is dit begrip in de verkennende fase van dit onderzoek nog niet eenduidig

gedefinieerd.

De verhandeling bestaat uit acht hoofdstukken. Het inleidende hoofd-

stuk schetst de context van het onderzoek en formuleert de

onderzoeksvragen. De hoofdstukken 2 tot en met 7 bestuderen elk een

afzonderlijk aspect van de probleemstelling en kunnen dan ook als

afzonderlijke artikels gelezen worden. Deze zes hoofdstukken werden all

gepubliceerd, of zullen gepubliceerd worden, in internationale peer-

reviewed tijdschriften. Een basisversie van elk van deze hoofdstukken

werd op minstens één internationaal wetenschappelijk congres gepresen-

teerd. In het achtste hoofdstuk worden de bevindingen tenslotte

samengevat, wordt de noodzakelijke kritische nuancering aangebracht, en

worden aanbevelingen voor het beleid geformuleerd.

S.2 Onderzoeksopzet

Het inleidende hoofdstuk geeft een overzicht van de klimaatwijziging en

peak-oil, twee mondiale verschijnselen die rechtstreeks verband houden

met externe effecten van mobiliteit, en legt verbanden met enkele boeien-

de aspecten van tijd- en ruimtebeleving en welvaartstoename die aan de

basis liggen van de groei van de mobiliteit. Ook wordt de mogelijke rol

van de ruimtelijke structuur in een toekomstgerichte benadering van de

mobiliteit onderzocht binnen de beleidscontext in Vlaanderen en Brussel2.

Op basis van de geschetste context worden vervolgens de onderzoeks-

vragen geformuleerd, en wordt een conceptueel kader opgesteld dat de

verwachte verbanden tussen de verschillende concepten visualiseert.

S.2.1 Onderzoeksvragen

Op basis van de beschouwingen uit Hoofdstuk 1 formuleren we de

onderzoeksvraag, die aan de basis ligt van deze verhandeling, als volgt:

• Tot op welke hoogte is de wederzijdse ruimtelijke nabijheid tussen

potentiële bestemmingen bepalend voor de dagelijks afgelegde afstan-

2 Gezien de geografische context werd Brussel in de meeste analyses binnen

deze verhandeling mee opgenomen, tenzij beperkingen in de beschikbare data

dit niet toelieten.

Samenvatting

271

den in Vlaanderen, en wat betekent dit in de context van peak-oil en

klimaatwijziging? (A)

Om de basisvraag te kunnen operationaliseren, splitsen we deze onmiddel-

lijk op in de volgende drie sub-vragen:

• Hoe kan de invloed van de ruimtelijke structuur op de dagelijks

afgelegde afstanden gekwantificeerd worden? (B1)

• Hoe kan ruimtelijke nabijheid gedefinieerd en gemeten worden, en

toegepast worden in de praktijk van duurzame ruimtelijke planning?

(B2)

• Welke locaties zijn het meest geschikt om bijkomende woningen en

jobs te realiseren, als we het bijkomend gegenereerd verkeer tot het

minimum willen beperken? (B3)

S.2.2 Conceptueel kader

Fig. S.1. Conceptueel kader

De gestelde onderzoeksvragen vormen de gebalde verwoording van een

onderzoekskader dat voortspruit uit de ruimere geschetste context.

Hierboven geven we een schematische weergave van dit conceptueel

kader, dat de verbanden tussen de verschillende aspecten van het onder-

zoek verduidelijkt (Fig. S.1).

Samenvatting

272

S.3 Bevindingen

S.3.1 Pendel

Hoofdstukken 2, 3, 4 en 6 behandelen enkel het pendelverkeer. Hoofdstuk

2 kan beschouwd worden als een verkennend onderzoek op basis van

relatief eenvoudige basisgegevens, assumpties en methodes. Dit hoofdstuk

toont aan dat er belangrijke regionale variaties bestaan inzake pendelaf-

standen en energieverbruik, bekeken vanuit de woonlocatie. Bovendien

blijkt het energieverbruik voor de pendel op regionale schaal sterk

gecorreleerd te zijn met de pendelafstand, wat betekent dat het belang

van de moduskeuze slechts van de tweede orde is. Deze bevindingen

worden in de context geplaatst van de stelling van Newman en Ken-

worthy (1989, 1999), die een sterk negatief verband vinden tussen

bevolkingsdichtheid en energieverbruik per capita voor transport. Ook in

Vlaanderen lijkt er een dergelijk, zij het vrij zwak, verband te bestaan.

De relatief beperkte focus van Hoofdstuk 2 roept de vraag op of de

ruimtelijke variatie in pendelafstanden verklaard kan worden door

variaties in de ruimtelijke nabijheid tussen woningen en jobs, en zo ja,

hoe deze nabijheid gemeten kan worden. Deze vraag is de basis voor

Hoofdstuk 3, waarin een methode ontwikkeld wordt om de ruimtelijke

nabijheid tussen woningen en jobs te berekenen op basis van ruimtelijk

gedesaggregeerde waarden voor de minimale pendelafstand. De minimale

pendelafstand is een concept dat afkomstig is uit de literatuur rond

bovenmatige pendel (“excess commuting”). Anders dan in Hoofdstuk 2

bevat de dataset die in deze paper werd gebruikt niet enkel de woonloca-

tie (oorsprong) maar ook de werklocatie (bestemming) van de

pendelstromen. Het is dan ook mogelijk om twee ruimtelijke nabijheids-

kaarten voor het pendelverkeer te ontwikkelen: één die de ruimtelijke

nabijheid van een woonlocatie ten opzichte van de jobmarkt weergeeft, en

één die de ruimtelijke nabijheid van een werklocatie ten opzichte van de

woonmarkt toont. Deze kaarten geven aan dat inwoners van relatief

afgelegen gebieden met een lage arbeidsbalans3 erg moeilijk een job

dichtbij huis kunnen vinden. Bovendien zijn de inwoners van deze regio’s

extra benadeeld wanneer alle werknemers collectief een job dichter bij

huis zouden gaan zoeken of dichter bij hun werk zouden gaan wonen.

3 De arbeidsbalans is de verhouding tussen het aantal jobs en het aantal

werkende inwoners in een gebied.

Samenvatting

273

Inwoners van de agglomeraties daarentegen hebben ruime marges om hun

pendelgedrag aan te passen, bijvoorbeeld onder invloed van stijgende

vervoerskosten. De belangrijkste oorzaak van dit fenomeen is het grote

verschil in ruimtelijke spreiding tussen de woonmarkt en de jobmarkt.

Jobs zijn doorgaans sterker geconcentreerd dan woningen, en heel wat

concentraties van jobs (in het bijzonder in industriegebieden) zijn relatief

ver verwijderd van de woonconcentraties. Verder lijken de resultaten van

Hoofdstuk 3 de vermoedens die Hoofdstuk 2 opriep te bevestigen: de

regio’s die gekenmerkt worden door een relatief lage mate van ruimtelijke

nabijheid, bekeken vanuit de woonlocatie, komen in grote lijnen overeen

met de regio’s waar we bovengemiddelde pendelafstanden waarnemen.

Met betrekking tot de werklocaties zijn het echter voornamelijk de regio’s

met een hoge dichtheid aan jobs die relatief slecht scoren op het vlak van

ruimtelijke nabijheid.

In Hoofdstuk 2 wordt een mogelijk verband gesuggereerd tussen de

toename van de gemiddelde pendelafstand in de loop der jaren en de

evolutie van de ruimtelijke nabijheid tussen woonmarkt en jobmarkt. In

Hoofdstuk 4 wordt deze hypothese onderzocht op basis van tijdsreeksen

voor de minimale pendelafstand, en deels ook voor een variant hierop (de

maximale pendelafstand). De ruimtelijke resolutie van de gebruikte

pendelmatrices is beperkt tot het gemeenteniveau, wat de bruikbaarheid

van de resultaten beperkt tot het bovengemeentelijke schaalniveau.

Niettemin is er in heel wat gemeenten een duidelijk waarneembare trend,

die gemiddeld wijst op een toename van de ruimtelijke scheiding tussen

woonmarkt en jobmarkt. De reden daarvoor is dat de toename van jobs

vooral te vinden is in de agglomeraties, terwijl de woningmarkt geken-

merkt wordt door een vorm van verdere ruimtelijke uitspreiding.

Belangrijker is echter de vaststelling dat de groei van het pendelverkeer

een stuk sneller verloopt dan de processen van ruimtelijke scheiding. We

kunnen dan ook besluiten dat de groei van het pendelverkeer een quasi-

autonoom, welvaartsgestuurd proces is dat een wisselwerking vertoont

met ruimtelijke scheidingsprocessen zonder dat er een causaal éénrich-

tingsverband bestaat.

Hoofdstuk 2 suggereert een omgekeerd verband tussen bevolkings-

dichtheid en energieverbruik per capita voor transport. Dit verband

wordt echter niet gekwantificeerd, noch wordt er een verband gelegd met

de ruimtelijke kenmerken van de werklocatie. Hoofdstuk 6 gaat dieper in

op dit probleem door eerst een aantal potentiële grootheden voor ruimte-

lijke nabijheid te selecteren, en vervolgens correlaties te berekenen tussen

Samenvatting

274

de geobserveerde pendelafstand en deze geselecteerde variabelen (aan

beide uiteinden van de pendelstroom). Bekeken vanuit de woonlocatie is

de woondichtheid, de ruimtelijke mix en de algemene bereikbaarheid

negatief gecorreleerd met de pendelafstand. De minimale pendelafstand is,

zoals verwacht op basis van Hoofdstuk 3, steeds positief gecorreleerd met

de pendelafstand, zowel in de woonlocaties als in de joblocaties. Jobdicht-

heid (in de joblocaties) is echter positief gecorreleerd met de

pendelafstand, in tegenstelling tot woondichtheid in de woonlocaties. Dit

betekent dat een hoge dichtheid niet leidt tot een duurzamer pendelpa-

troon als deze niet gepaard gaat met een hoge graad van functionele mix,

en dat een scheve arbeidsbalans bijdraagt tot een toename van de

pendelafstand. Met betrekking tot de arbeidsbalans kunnen we stellen dat

de pendelafstanden gemiddeld het kortste zijn als de herkomst of de

bestemming van de verplaatsing zich in een gebied bevindt waar de

arbeidsbalans nagenoeg in evenwicht is (d.w.z. gelijk is aan 1).

De resultaten van deze vier hoofdstukken moeten echter geïnterpre-

teerd worden binnen de beperkingen van de gebruikte data en de

gehanteerde methode. De belangrijkste randvoorwaarde is het ontbreken

van een differentiatie op basis van sector, inkomen of opleidingsniveau.

Het is namelijk best mogelijk dat hoogopgeleiden, die vaker gespeciali-

seerde jobs uitvoeren, minder vaak in de buurt van de voor hen geschikte

jobs wonen. Dit is bijvoorbeeld het geval voor kantoorjobs in het centrum

van Brussel die door werknemers uit de villawijken in de groene rand

worden ingevuld. Anderzijds komen jobs voor laagopgeleiden relatief vaak

voor in industriegebieden die niet steeds in de buurt van een woonkern of

agglomeratie liggen. Volgens Van Acker en Witlox (2011) is een hoger

inkomen geassocieerd met grotere pendelafstanden. Een differentiatie zou

deze problematiek scherper kunnen stellen en de rol van de ruimtelijke

mismatch tussen de kwalificatie en voorkeuren van de werknemers en de

vereisten die een bepaalde job stelt beter in kaart kunnen brengen.

Bovendien is de rol van het openbaar vervoer in deze deelonderzoeken

onderbelicht. Een hoge concentratie van jobs rondom een belangrijk

station zal in termen van energieverbruik een minder grote impact

hebben dan een concentratie van jobs op een afgelegen industrieterrein

dat nagenoeg enkel per auto bereikbaar is.

Samenvatting

275

S.3.2 Dagelijkse niet-pendel

Hoewel de bevindingen van Hoofdstukken 2, 3, 4 en 6 duidelijk aantonen

dat er een verband bestaat tussen de ruimtelijke nabijheid van woningen

en jobs en het pendelverkeer, mogen deze conclusies niet zomaar geëxtra-

poleerd worden naar andere vormen van quasi-dagelijkse mobiliteit.

Hoofdstukken 5 en 7 proberen de gehanteerde methodiek uit te breiden

met verplaatsingspatronen die niet tot de pendel behoren.

In Hoofdstuk 5 wordt het concept van de minimale pendelafstand en

bovenmatige pendel toegepast op quasi-dagelijkse verplaatsingen die niet

tot de pendel behoren. De resultaten worden weergegeven in de vorm van

een nabijheidskaart, die voor elke statistische sector de relatieve nabijheid

tot een korf van voorzieningen weergeeft. Daarnaast wordt voor elke

ruimtelijke categorie een geaggregeerde excesfactor berekend, die aangeeft

hoe sterk het verband is tussen de ruimtelijke nabijheid van de voorzie-

ningen en de effectief afgelegde afstanden.

De nabijheidskaart geeft een indicatie van welke gebieden in feite te

afgelegen zijn om een duurzaam verplaatsingspatroon in functie van het

bereiken van dagelijkse voorzieningen mogelijk te maken. De vergelijking

van de minimaal af te leggen afstand met de geobserveerde afgelegde

afstand geeft aan dat het verband tussen ruimtelijke nabijheid en ver-

plaatsingspatronen sterker is in de meest afgelegen gebieden dan in de

meer verstedelijkte gebieden, ook al is de afgelegde afstand gemiddeld

groter in de meer afgelegen gebieden. Met andere woorden: inwoners van

het platteland zullen vaker de dichtstbijzijnde mogelijke bestemming

kiezen dan inwoners van meer verstedelijkte gebieden. De grootstedelijke

gebieden en de agglomeraties, die samen de meest verstedelijkte van alle

onderzochte gebieden zijn, vormen echter een uitzondering: hier zijn de

effectief afgelegde afstanden kort, terwijl er toch relatief vaak voor de

dichtstbijzijnde mogelijke bestemming geopteerd wordt.

Hoofdstuk 7, tenslotte, combineert de variabelen die ontwikkeld wer-

den in Hoofdstuk 3, 5 en 6 met de analysemethode van Hoofdstuk 6. De

ruimtelijke correlatieanalyse van Hoofdstuk 6 wordt uitgebreid tot een

multivariate regressieanalyse op basis van een steekproef die geleverd

wordt door het Onderzoek Verplaatsingsgedrag van 2007 (Janssens et al.,

2009). Om een globaal beeld te geven van de impact van alle quasi-dage-

lijkse verplaatsingen samen werden pendel- en niet-pendelverplaatsingen

samengevoegd en werd dus de totale afgelegde afstand per respondent

geanalyseerd. Door een extrapolatie op basis van de bekomen regressie-

Samenvatting

276

vergelijking wordt een nieuwe, verfijnde, nabijheidskaart gegenereerd. In

tegenstelling tot de analyse in Hoofdstuk 6, worden in de analyse van

Hoofdstuk 7 enkel de ruimtelijke kenmerken van de woonlocatie van de

respondenten opgenomen, zodat de conclusies enkel geldig zijn met

betrekking tot de ruimtelijke kenmerken van de woonomgeving. Een

belangrijke conclusie van Hoofdstuk 7 is dat het belang van de ruimtelijke

distributie van jobs helemaal naar de achtergrond verdwijnt wanneer we

het volledige verplaatsingspatroon gaan bestuderen. De impact van de

ruimtelijke distributie van jobs op de totale mobiliteit is dus relatief

onbelangrijk in verhouding tot de ruimtelijke distributie van andere

quasi-dagelijkse bestemmingen. Bevolkingsdichtheid, functionele mix en

ruimtelijke nabijheid van voorzieningen blijven echter wel overeind als

bepalende factoren. Globaal blijkt de verklarende waarde van een model

op basis van ruimtelijke eigenschappen echter zeer beperkt te zijn. Dat

betekent dat pogingen om mobiliteitspatronen te doen krimpen op basis

van een ruimtelijk beleid gedoemd zijn om tot povere resultaten te leiden,

tenzij in combinatie met andere maatregelen (zoals bv. prijsbeleid) of

samengaand met (autonoom) stijgende energieprijzen.

Ook Hoofdstukken 5 en 7 hebben echter hun beperkingen. De belang-

rijkste lacune kan samengevat worden als het ontbreken van een analyse

op basis van de ruimtelijke kenmerken van de bestemming. In Hoofdstuk-

ken 3, 4 en 6 wordt ook de joblocatie in het onderzoek betrokken. In

Hoofdstukken 5 en 7 hebben we echter geen zicht op de invloed van de

locatie van de bezochte winkels, scholen, ontspanningscentra, ... De reden

waarom een dergelijke analyse buiten het bestek van deze verhandeling

valt, is enerzijds het gebrek aan een voldoende grote dataset die de

benodigde informatie bevat, en anderzijds de erg grote complexiteit van

dit soort onderzoek. Niettemin ligt hier een belangrijke piste voor moge-

lijk vervolgonderzoek. Er is voor België of Vlaanderen geen onderzoek

beschikbaar dat bijvoorbeeld het totale verschil in energetische efficiëntie

tussen hypermarkten en buurtwinkels kwantificeert, of ook tussen grote

en kleine scholen, of tussen meer en minder gespecialiseerde onderwijsin-

stellingen.

S.3.3 Niet-dagelijkse verplaatsingen

Occasionele (dus niet-dagelijkse) verplaatsingen, zoals daguitstapjes,

weekendjes, citytrips, vakanties of zakenreizen vormen geen voorwerp van

de zes deelstudies van deze verhandeling. Hoewel dit soort verplaatsingen

Samenvatting

277

slechts zelden wordt bestudeerd binnen het mobiliteitsonderzoek, is het

best mogelijk dat de afvlakking van de groei van de mobiliteit binnen het

segment van de quasi-dagelijkse verplaatsingen gecompenseerd wordt

door een voortdurende groei binnen het segment van de occasionele

mobiliteit. Deze verschuiving zou dan opnieuw snelheidsgedreven zijn,

voornamelijk op basis van de verregaande democratisering van het

luchtverkeer. Bovendien is er mogelijk ook een verband tussen het

verkeer, en dus het energieverbruik, dat het gevolg is van occasionele

mobiliteit, en de ruimtelijke structuur van de woonlocatie van de reiziger

(Holden en Norland, 2005).

Deze hypothese wordt verkennend getoetst in een addendum. We stel-

len vast dat inwoners van meer verstedelijkte gebieden vaker het vliegtuig

nemen dan inwoners van minder verstedelijkte gebieden. Hoewel de

beschikbare gegevens niet van die aard zijn dat verschillen in energiever-

bruik en CO2-uitstoot kunnen worden gekwantificeerd, lijkt de analyse te

suggereren dat, als we vliegverkeer in de analyse zouden opnemen, het

verplaatsingspatroon van de stadsbewoner wellicht een stuk minder

duurzaam is dan algemeen gedacht.

S.4 Aanbevelingen voor het ruimtelijk beleid

Banister (2008) volgend, stellen we een ontwikkelingsperspectief voor dat

gemotiveerd wordt vanuit het optimaliseren van de bereikbaarheid op

basis van een minimaal verkeersvolume, door het maximaliseren van de

ruimtelijke nabijheid en het valoriseren van agglomeratie-effecten. Uit het

voorgaande blijkt dat het sturen van ruimtelijke ontwikkelingen ten

eerste moet gezien worden als een manier om het systeem robuuster te

maken voor olieschaarste en duurder wordend transport, en ten tweede

als het werken aan een ruimtelijk kader dat geschikt is om klimaatbeleid

te faciliteren. Aan de hand van het stramien van Banister (1999, p. 316-

319) kunnen we een beknopte vertaling maken van onze onderzoeksresul-

taten in aanbevelingen voor het ruimtelijk beleid:

• Ontwikkelingen moeten zoveel mogelijk gebeuren op basis van een

hoge woondichtheid, ook in gebieden waar andere activiteiten dan

wonen voorkomen. Hoe hoog de woondichtheid moet zijn hangt af

van het ambitieniveau. Dichtheidsdrempels die in de literatuur wor-

den genoemd zijn meestal gekoppeld aan de gewenste rol van het

openbaar vervoer en worden besproken in Hoofdstuk 1. Een aantal

Samenvatting

278

nieuwe eco-wijken in het buitenland hebben een bruto-woondichtheid

die gaat tot zo’n 150 inwoners per hectare, wat vergelijkbaar is met

de dichtstbevolkte wijken in de 19e-eeuwse gordels van de Belgische

steden. Het ambitieniveau van het RSV is echter beperkt tot 25 wo-

ningen (equivalent met ongeveer 55 inwoners) per hectare in

stedelijke gebieden en 15 woningen (equivalent met ongeveer 38 in-

woners) per hectare in het buitengebied. Een voortschrijdende

gezinsverdunning dreigt de effectieve woondichtheid op basis van deze

normen (die uitgedrukt zijn in woningen per hectare) zelfs nog verder

naar beneden te halen.

• De omvang van de kernen moet voldoende groot zijn om op basis van

agglomeratievoordelen een relatief hoge mate van zelfvoorziening te

kunnen verzekeren. Banister (1999) volgend zouden we kunnen stellen

dat kernen die minder dan 25000 inwoners huisvesten in principe niet

meer zouden mogen groeien, en zouden geïsoleerde steden, die geen

deel uitmaken van een agglomeratie, op termijn een omvang moeten

bereiken van tenminste zo’n 50000 inwoners. Daarnaast zou groei zo-

veel mogelijk in of onmiddellijk aansluitend bij de agglomeraties

(> 250000 inwoners) moeten plaatsvinden, waar het openbaar vervoer

een belangrijke rol kan spelen.

• De locatie van nieuwe ontwikkelingen zou moeten gebeuren aanslui-

tend bij of in bestaande stedelijke gebieden. Daarbij moeten

faciliteiten en jobs tegelijkertijd met woningen gepland en ontwikkeld

worden, bij voorkeur geconcentreerd in lokale centra. Bovendien moe-

ten nieuwe ontwikkelingen zoveel mogelijk bereikbaar zijn zonder

auto.

Op basis van deze principes wordt in Hoofdstuk 8 een vertaling gemaakt

naar mogelijke aanpassingen aan het Ruimtelijk Structuurplan Vlaande-

ren (RSV, 1997/2004), waarvan we de belangrijkste hieronder opsommen:

• Bijkomende woningen zouden zoveel mogelijk in of onmiddellijk

aansluitend bij de agglomeraties moeten worden gebouwd, met een

hoge woondichtheid en een goede mix van functies. De agglomeraties

bestaan in grote lijnen uit de grootstedelijke en regionaalstedelijke

gebieden (zoals geselecteerd in het RSV), inclusief de suburbane gor-

dels die onmiddellijk bij deze steden aansluit.

• Binnen deze agglomeraties moet een uitgebalanceerde ruimtelijke

distributie van voorzieningen worden nagestreefd, bij voorkeur gecon-

centreerd in centra en subcentra. Schaalvergroting waarbij

verschillende vestigingen van bijvoorbeeld scholen, winkels of ateliers

Samenvatting

279

vervangen worden door één campus, hypermarkt of industrieterrein

moet ontmoedigd worden wegens de reductie van de ruimtelijke na-

bijheid die hier het gevolg van kan zijn. Op het vlak van openbaar

vervoer zou de verbetering van het interne openbaar-vervoernetwerk

in de agglomeraties prioriteit moeten krijgen.

• In de kleinstedelijke gebieden moet een volwaardig aanbod aan jobs

en voorzieningen geboden worden voor de inwoners van het omlig-

gende buitengebied, zodat de noodzaak voor verplaatsingen naar de

agglomeraties beperkt wordt. Bijkomende woningen zijn echter niet

gewenst in de kleinstedelijke gebieden en het buitengebied, aangezien

deze bijkomend lange-afstandsverkeer genereren.

• Zeer gespecialiseerde kantoorjobs zouden zoveel mogelijk op centrale

locaties bij belangrijke spoorstations moeten gesitueerd worden (Brus-

sel, Antwerpen, eventueel ook Mechelen, Gent en Leuven). Om lange-

afstandspendel niet aan te moedigen, zijn deze locaties echter minder

geschikt voor niet-gespecialiseerde jobs waarbij het de bedoeling is om

het personeel zoveel mogelijk lokaal te recruteren.

S.5 Verder onderzoek

Tot slot wordt een overzicht gegeven van mogelijke richtingen voor

verder onderzoek dat een bijdrage zou kunnen leveren tot een goed

geïnformeerde praktijk van ruimtelijke ordening, gericht op de ontwikke-

ling van een goede bereikbaarheid in combinatie met een duurzame vorm

van mobiliteit.

Referenties


Review. 70(3), pp. 313-338.


Policy. 15(2), pp. 73-80.

Holden, E. en I. Norland (2005) “Three challenges for the compact city as

a sustainable urban form: Household consumption of energy and


Studies. 42(12), pp. 2145–2166.

Samenvatting

280

Janssens, D., E. Moons, E. Nuyts en G. Wets (2009) Onderzoek Ver-

plaatsingsgedrag Vlaanderen 3 (2007-2008). Brussel-Diepenbeek:

Ministerie van de Vlaamse Gemeenschap.

Newman, P. en J. Kenworthy (1989) Cities and Automobile Dependence.

A Sourcebook. Aldershot: Gower.

Newman, P. en J. Kenworthy (1999) Sustainability and Cities: Overcom-

ing Automobile Dependence. Washington, DC: Island Press.


Versie. Brussel: Ministerie van de Vlaamse Gemeenschap.

Van Acker, V. en F. Witlox (2011) “Commuting trips within tours: How

is commuting related to land use?” Transportation. doi: 10.1007

/s11116-010-9309-6.



nology.

281

Curriculum vitae Kobe Boussauw (°1978, Bruges) is a researcher

at the Geography Department of Ghent Univer-

sity (since November 2007). He graduated as a

civil engineer-architect (Ghent University, 2001)

and as a spatial planner (Ghent University,

2002), and obtained the certificate of post

academic traffic and mobility studies (University

of Antwerp, 2005). During his PhD research,

Kobe also finished the complete doctoral training

programme (Ghent University, 2011).

Kobe has worked as a consultant in a private company (iris consult-

ing, 2001-2003), as a civil servant for the Flemish Government (Planning

Department, 2003-2006), and as an advisor for the UN-Habitat pro-

gramme in Kosovo (2006-2007). Kobe’s PhD research was funded by the

Policy Research Centre on Regional Planning and Housing - Flanders

(Steunpunt Ruimte en Wonen 2007-2011).

Kobe is the author of several publications in international peer-

reviewed academic journals, and published also in various regional

journals. He acted as a referee for the journals Papers in Regional

Science, Regional Studies and Tijdschrift voor Economische en Sociale

Geografie, and for the Transportation Research Board’s annual meetings

(2010 and 2011). He presented his work at many national and interna-

tional conferences in Europe and the US. At the Transport Research

Arena conference in Brussels (2010), he was awarded a bronze medal for

his paper “Spatial variations in destination proximity: A regional case

study”. Kobe was also an occasional guest lecturer at Ghent University,

at the Artesis University College of Antwerp and at the University of

Prishtina.

Articles in international peer-reviewed journals included in the

ISI / Thomson Reuters Web of Science index



43(5), pp. 580-591.



Curriculum vitae

282


press.

Boussauw, K., T. Neutens and F. Witlox (2011) “Relationship between



Boussauw, K., B. Derudder and F. Witlox (2011) “Measuring spatial



search. 11(1), pp. 42-60.

Boussauw, K., V. Van Acker and F. Witlox (2011) “Excess travel in non-

professional trips: Why looking for it miles away?” Tijdschrift voor

Economische en Sociale Geografie. in press.

Boussauw, K. and F. Witlox (2011) “Linking expected mobility produc-

tion to sustainable residential location planning: Some evidence from

Flanders.” Journal of Transport Geography. in press.

Articles in regional journals

Boussauw, K. (2006) “Sprawl of duurzaamheid? Leren van utopische

concepten voor stedelijke mobiliteit.” Ruimte en Planning. 26(1), pp.

22-33.

Boussauw, K., D. Lauwers and F. Witlox (2008) “Ruimtelijke structuur

en energieverbruik voor vervoer: Een eerste verkenning voor Vlaande-

ren.” Ruimte en Planning. 28(3), pp. 35-48.

Boussauw, K. (2008) “Stedenbouw in Kosovo: Visievorming voor transi-

tie.” Agora - Magazine voor sociaalruimtelijke vraagstukken. 24(4),

pp. 12-14.

Boussauw, K. (2009) “Stadsmens onderweg: een duurzaamheidsparadox.”

Agora - Magazine voor sociaalruimtelijke vraagstukken. 25(5), pp. 7-

10.

Boussauw, K. and F. Witlox (2009) “Theoretische minimale pendelaf-

stand als ruimtelijke karakteristiek: Een denkoefening.” Tijdschrift

Vervoerswetenschap. 45(4), pp. 150-158.

Boussauw, K., T. Neutens and F. Witlox (2010) “Pendel in en om de

compacte stad: Een ruimtelijke analyse van de afstand tot het werk.”

Ruimte & Maatschappij. 2(2), pp. 5-22.

Curriculum vitae

283

International conference papers

Boussauw, K. and F. Witlox (2008) Introducing a commute-energy

performance index for Flanders. 48th Congress of the European Re-

gional Science Association. Liverpool.

Boussauw, K. and F. Witlox (2009) Variation and evolution of minimum

commuting distances in Flanders. 4th Kuhmo-Nectar Conference and

Summer School. Copenhagen.

Boussauw, K., T. Neutens and F. Witlox (2010) Spatial variations of the

minimum home-to-work distance in the north of Belgium. Transpor-

tation Research Board 2010 Annual Meeting. Washington, DC.

Boussauw, K., T. Neutens and F. Witlox (2010) Does spatial proximity

influence commuting trip length? An approach based on evidence

from Flanders and Brussels. Transportation Research Board 2010

Annual Meeting. Washington, DC.

Boussauw, K. and F. Witlox (2010) Spatial variations in destination

proximity: A regional case study. Transport Research Arena 2010.

Brussels.

Boussauw, K. (2010) Aspects of spatial proximity and sustainable travel

behavior in Flanders: A quantitative approach. Association of Colle-

giate Schools of Planning 2010 PhD Workshop. Atlanta, GA.

Boussauw, K. and F. Witlox (2010) Excess travel in non-commuting

trips: A regional case study. World Conference on Transport Re-

search. Lisbon.

Boussauw, K. and F. Witlox (2011) Regional variations in travel energy

consumption: Some evidence from Flanders. Mobil.TUM 2011 Con-

ference. Munich.

Boussauw, K. and F. Witlox (2011) The role of spatial proximity in daily

mobility production: A case study in the North of Belgium. Annual

Meeting of the Association of American Geographers. Seattle.

National and regional conference papers

Boussauw, K., D. Lauwers and F. Witlox (2008) Ruimtelijke structuur en

energieverbruik voor vervoer: Een eerste verkenning voor Vlaanderen.

Uitdagingen voor de ruimtelijke ordening in Vlaanderen. Brussels.

Boussauw, K. and F. Witlox (2008) L’introduction d’un indice de per-

formance énergétique pour la navette en Flandre et Bruxelles. Derde

Belgische dagen van de geografie - Troisièmes journées belges de la

géographie. Brussels.

Curriculum vitae

284

Boussauw, K., D. Lauwers and F. Witlox (2008) En wat als de olie op is?

De relatie tussen ruimte en energieverbruik voor vervoer. Colloquium

Vervoersplanologisch Speurwerk 2008. Santpoort (The Netherlands).

Boussauw, K. and F. Witlox (2008) Kilometers malen: Maatschappelijke

dwangneurose of ruimtelijk probleem? Duurzame mobiliteit Vlaande-

ren: De leefbare stad. Ghent, 205-228.

Boussauw, K. and F. Witlox (2009) De wereld om de hoek: Is ruimtelijke

nabijheid maakbaar? PlanDag 2009. G. Bouma, F. Filius, H. Lein-

felder and B. Waterhout. Brussels, pp. 425-434.

Boussauw, K., T. Neutens and F. Witlox (2009) “Excess commuting in

Flanders and Brussels”. In: C. Macharis and L. Turcksin (Eds.)

BIVEC-GIBET Transport Research Day 2009. Brussels: VUBPress,

pp. 157-171.

Boussauw, K., T. Neutens and F. Witlox (2009) Pendelgedrag en nabij-

heid: Speelt de compacte stad haar rol? Colloquium Vervoers-

planologisch Speurwerk 2009. Antwerp.

Boussauw, K. and F. Witlox (2010) Travel energy consumption and the

built environment: Evidence from Flanders. Colloque de la Con-

férence Permanente du Développement Territorial 2010. Liège.

Boussauw, K. and F. Witlox (2010) A residential location model based on

characteristics of spatial proximity: A case study in the North of Bel-

gium. 11th TRAIL Congress. Rotterdam.

Book chapters

Boussauw, K., Zwerts, E. and F. Witlox (2009) “Mobiel Vlaanderen”. In:

L. Vanderleyden, M. Callens and J. Noppe (Eds.) De Sociale Staat

van Vlaanderen 2009. Brussel: Studiedienst van de Vlaamse Reger-

ing, pp. 279-312.

Boussauw, K., D. Lauwers and F. Witlox (2009) “Ruimtelijke structuur

en energieverbruik voor vervoer: Een eerste verkenning voor Vlaande-

ren”. In: Re-Marc-able Landscapes - Marc-ante Landschappen: Liber

Amicorum Marc Antrop. Gent: Academia Press, pp. 235-246.

Witlox, F., K. Boussauw, W. Debauche, B. Derudder, C. Macharis and S.

Verlinde (2009) “Vandaag besteld, vannacht geleverd. Over de moge-

lijkheden van nacht- en daldistributie als oplossing voor het probleem

van het stedelijk goederenvervoer in België”. In: C. Kesteloot, M.

Goossens, H. Van der Haegen et al. (Eds.) Bas-Congo tot Dadizele.

Veelzijdigheid in de Geografie. Liber Amicorum Etienne Van Hecke.

Curriculum vitae

285

Leuven: KULeuven, Instituut voor Sociale en Economische Geografie,

pp. 227-242.

Others

Boussauw, K., D. Lauwers, E. Zwerts and F. Witlox (2009) Visie Ruim-

tegebruik en Ruimtebeslag 2020-2050: Sectornota Mobiliteit. Ghent:

Steunpunt Ruimte en Wonen.

Boussauw, K. (2009) Hoe duurzaam leeft de stadsbewoner? Low Tech

Magazine from http://www.lowtechmagazine.be/2009/12/hoe-duur

zaam-is-de-stadsbewoner.html.

Boussauw, K., R. Simoen and F. Witlox (2010) Focusnota Ruimte,

Logistiek en Multimodaliteit. Ghent: Steunpunt Ruimte en Wonen.

Zwerts, E., K. Boussauw, L. Bral, P. De Maeyer, B. Derudder, V. Van

Acker, L. Verdonck and F. Witlox (2010) Algemeen profiel Oost-

Vlaamse O&O-bedrijven. Ghent: POM Oost-Vlaanderen.

286

This dissertation aims to contribute to the insight in the reciprocal relationship

between person mobility and spatial development, taking into account the societal

context of climate targets and imminent peak oil. This is done through the

development of a number of quantitative research methods, which are embedded

in a literature review and is applied to the case study of Flanders (Belgium).

The research focuses on exploring the sustainability of spatial structure with

respect to travel behaviour, with particular attention to the daily distances

travelled. Sustainability is defined in terms of resilience, not only for growing

mobility but also for a possible declining future mobility, a scenario that is

suggested by peak oil theory or may be the consequence of a stringent climate

policy. Moreover, spatial structure plays a role in the potential steering of travel

behaviour in a more sustainable direction.

From this point of view, the dissertation assesses to what extent mutual spa-

tial proximity between potential destinations is determining the daily distances

covered in Flanders, and how spatial development can play a role in pursuing a

high degree of accessibility based on a minimum amount of traffic.

Kobe Boussauw (°1978, Bruges) is a researcher at the Geography Department of

Ghent University. He is a civil engineer-architect and a spatial planner. Before,

Kobe has worked as a consultant in a private company, as a civil servant for the

Flemish Government, and as an advisor for the UN-Habitat programme in

Kosovo. Kobe’s PhD research was funded by the Policy Research Centre on

Regional Planning and Housing - Flanders (Steunpunt Ruimte en Wonen 2007-

2011).

aspects of spatial proximity and - ugentkboussau/dissertation_boussauw_a... · 2011-02-16 · en...

Documents