You are on page 1of 14

Journal of Transport Geography 9 (2001) 173186

www.elsevier.com/locate/jtrangeo

Travel behaviour in Dutch monocentric and


policentric urban systems
Tim Schwanen *, Frans M. Dieleman, Martin Dijst
Urban Research centre Utrecht (URU), Faculty of Geographical Sciences, Utrecht University, Members of the Netherlands
Graduate School for Housing and Urban Research (NETHUR), P.O. Box 80.115, 3508 TC Utrecht, Netherlands

Abstract
In this paper, we analyse how monocentric and policentric urban structures aect modal choice and travel distances for dierent
travel purposes in the Netherlands. The analysis is based on data from the Dutch National Travel Survey 1998. Here we distinguish
four kinds of urban systems: one monocentric and three types of policentric systems. The evidence on how the structure of urban
regions aects travel behaviour is mixed. Regarding modal choice, deconcentration of urban land uses encourages driving and
discourages the use of public transport as well as cycling and walking. However, in terms of distance travelled per person, the results
of the relocation of jobs and residences to suburban locations are less commuting in some urban regions, and longer commuting
distances in others. The longer commute may also be an eect of the strong spatial planning policies in the Netherlands. 2001
Elsevier Science Ltd. All rights reserved.
Keywords: Modal split; Distance travelled; Urban form; Policentrism; Multivariate analysis

1. Introduction
Metropolitan areas both in old industrial countries
and in developing countries have evolved from mononodal or monocentric structures into multi-nodal or
policentric urban forms. Nonetheless, models of urban
land use still generally refer to monocentric urban
forms. These models are now being reformulated to
reect the present structure of metropolitan regions (cf.
Bailey, 1999; Clark, 2000). At the same time, the restructuring of metropolitan areas has prompted a new
debate mainly in an Anglo-American setting on the
eciency of diverse urban forms.
Much of the interest in new urban forms and policentricity relates to issues of travel behaviour. Specically, the eciency of urban form in terms of
commuting distances and times is at the heart of the
debate. There are two camps. On one side are the liberals, like Gordon and Richardson (1997), who believe
in the eciency of market mechanisms. On the other
side are the regulators, like Ewing (1997), who propagate
the interference of planners to solve the environmental

Corresponding author. Tel.: +31-30-253-22-24; fax: +31-30-25406-04.


E-mail address: t.schwanen@geog.uu.nl (T. Schwanen).

and congestion problems caused by travel behaviour.


The extensive use of the private car is seen as the main
cause of these problems.
The discussion on urban form and travel behaviour,
especially the monocentricpolicentric issue, is mainly
based on the analysis of American data sources. It is
relevant to bring in more evidence from Europe, as the
pattern of urbanisation there diers from that in the
USA. In Europe, relatively more people live in cities of
between 200,000 and one million inhabitants. In the
European Union as a whole, 170 cities lie in this range.
Only 32 cities have a population of over one million, and
very few of these have more than two million (Dieleman
and Faludi, 1998; Commission of the European Communities, 1994). In addition, most European cities were
founded in the Middle Ages; the dierent histories create
dierent contexts for travel behaviour. Moreover, over
the past decades, spatial planning measures in various
European countries have limited urban sprawl (see
Dieleman et al., 1999).
In this paper, we analyse the way in which monocentric and policentric urban structures aect modal
choice and travel distances for dierent purposes in the
Netherlands. The analyses will throw some light on the
usefulness of the `co-location' hypothesis which states
that rms and households periodically change their location in order to reduce commuting costs. We use the

0966-6923/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 6 - 6 9 2 3 ( 0 1 ) 0 0 0 0 9 - 6

174

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

Fig. 1. Daily urban systems: structure and urbanisation levels (based


on Van der Laan, 1998).

classication of monocentric and policentric urban


forms, made by Van der Laan (1998), who divides 26
daily urban systems into four functional types (Fig. 1).
Within these four types of metropolitan area, we look at
both modal choice and distances travelled to and from
work, for shopping and for leisure. The Dutch National
Travel Survey (OVG), which is used for the analyses,
allows us to simultaneously examine the inuence of
personal attributes and of urban structure. This combination is quite unusual in this type of research. First,
we review the relevant literature in Section 2. An overview of the data and methods used is presented in Section 3. Then we turn to a discussion of the results,
spread over Sections 4 and 5. On the basis of this
evidence, we go on to formulate some conclusions in
Section 6.

2. Travel behaviour and monocentricpolicentric urban


structures in the literature
Over the past decades, several forces particularly
the use of information and communication technologies,
globalisation and deregulation (Anas et al., 1998) have
led to a deconcentration of population and employment
from the core of the city to the periphery. In the process,
many urban regions became policentric (Ingram, 1998;

Batten, 1995; Clark and Kuijpers-Linde, 1994; Dieleman


and Faludi, 1998; Gordon et al., 1986). Instead of a
strong concentration of jobs in the city centre, urban
regions now have several employment and shopping
centres and have developed into policentric structures.
As a result, patterns of daily travel became tangential
instead of radial in many metropolitan areas. For example, in the late 1980s, 57% of all the inter-municipal
trips taken by the inhabitants of Randstad Holland were
between suburban communities. Travel between the
central cities and the suburban communities of the
Randstad was less frequent: 41% of the inhabitants'
trips (Cortie et al., 1992). Comparable evidence is
available for Paris (Bolotte, 1991; Jansen, 1993), Helsinki (Jansen, 1993), German cities (Schmitz, 1993),
Swedish cities (Asmervik and Naess, 1995) and American cities (Gordon and Richardson, 1996).
The literature on how monocentric and policentric
metropolitan structures aect travel behaviour is polarised. Some authors (e.g., Gordon and Richardson,
1997; Levinson and Kumar, 1994) suggest that a deconcentrated structure tends to reduce commuting distance and commuting times. The scholars in this group
adhere to the `co-location' hypothesis, which states that
rms and households periodically readjust spatially to
achieve balanced average commuting distances and duration (Gordon et al., 1991). Other authors (e.g., Cervero, 1996; Newman and Kenworthy, 1989; Ewing,
1997) refute this positive view of the eect of policentricity on travel behaviour. Ewing (1997), for example,
stated that average commuting times increased in US
metropolitan areas in the 1980s and were signicantly
longer in the suburbs than in the central cities. According to this author, decentralisation of rms and
households has been a disaster for travel behaviour.
Drawing on data from the Nationwide Personal
Transportation Study (NPTS), Gordon and Wong
(1985) have analysed the commuting distances for private
vehicles in American metropolitan areas with more than
100,000 inhabitants. They found that the average journey to work lengthens in accordance with city size in the
Northeast of the USA. For cities in the West, the average distance decreases with increasing city size for
metropolitan areas with at least three million inhabitants. Compared with the Northeast, the distance travelled in Western cities is smaller in most size classes
during the morning peak. Gordon and Wong explain
this dierence by referring to the fact that the western
cities are more policentric than cities in the Northeast.
Schmitz (1993) endorses this explanation. Examining
commuting regions in the former West Germany, he
found that commuters who cross the municipal borders
in monocentric regions travel farther on average than
those who live in policentric regions. In addition, travel
distance increases with the population size of German
cities.

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

In the analysis of the monocentricpolicentric relation, the spatial scale is critical. For instance, as Naess
(1995) concludes for Swedish commuting regions (encompassing all towns and larger villages situated within
35 km from the regional centre), when controlling for a
number of socio-economic variables, a policentric settlement structure makes the least demand on energy for
transport. This applies for all travel purposes. The
condition is that each individual town and village must
have a suciently high population density (see also
Naess et al., 1996). Apparently, people living in remote
sub-regional centres take advantage of local job, service
and leisure opportunities. Naess (1995) assumes that
Swedes in the countryside are largely adapted to the fact
that they cannot have the same number of options
concerning service facilities and specialised jobs as in the
cities.
However, Naess and Sandberg (1996) found that the
degree of decentralisation had the opposite eect on
energy use in transport on a lower spatial scale (Greater
Oslo). Elsewhere, Gordon et al. (1989a) saw that suburban residents within metropolitan areas have longer
commuting distances than central-city residents (see also
Frost et al., 1997). However, for employment centres in
the San Francisco Bay Area, Cervero and Wu (1998)
came to a dierent conclusion. They found that the
mean commuting distance of workers in suburban employment centres is shorter than for people working in
downtown San Francisco. Yet, during the 1980s this
dierence has been narrowing. When analysing the effects of job decentralisation in the same area, Cervero
and Landis (1991) stratied the respondents by their
place of residence at the time of the job relocation and
whether workers moved house after this relocation.
Commuting distances decreased among workers who
were already living in the suburbs. Those who remained
in San Francisco and became reverse commuters saw
their travel distances rise enormously. Former centralcity inhabitants who moved to the suburbs were better
o, but their commuting distances tended to be longer
than before the relocation of both job and home.
Commuting distances are important, but so are
commuting times. With respect to policentric metropolitan areas in the American West, Gordon and Wong
(1985) conclude that these travel times are on average
lower than in the monocentric metropolitan areas in the
Northeast. Within metropolitan areas, Gordon et al.
(1989a) have found shorter commuting times for suburban residents than for central-city residents in metropolitan areas with at least three million inhabitants.
In another publication (Gordon et al., 1989b), these
authors applied the regression analysis and came to a
comparable conclusion. Because of the stronger congestion eects in high-density monocentric cities, travel
times are higher there than in low-density suburban
communities in policentric cities. Giuliano and Small

175

(1993) point out that a policentric pattern of employment centres has a potential for shorter commuting
times. Yet for Los Angeles, they concede that commuters in suburban centres actually travel only a few
minutes less than downtown workers. According to
Cervero and Wu (1998), commuting times in the San
Francisco Bay Area have increased far more rapidly in
suburban employment centres than in downtown San
Francisco. Cervero and Landis (1991) draw the same
conclusion for commuting times as home-to-work distances; compared with commuting distances, however,
the dierences among population categories tend to be
smaller. Their results suggest that, although overall
commuting times and distances may be stable or fall
over time, the process of job decentralisation may have
increased the variation between individual workers.
Cervero and Landis (1991) and Cervero and Wu
(1998) have also analysed the eect of deconcentration
on modal split for the San Francisco Bay Area. Shifts in
commuting behaviour from using public transport to
solo driving were seen as the main eect of job decentralisation. The same eects were found for Oslo
(Naess and Sandberg, 1996; Usterud Hanssen, 1995)
and Melbourne (Bell, 1991). Yet for the Paris region,
Bolotte (1991) arrives at a dierent conclusion. For the
years 19711989, he found a growth in motorised suburban travel for all purposes, comprising 63% of the
total. However, he found hardly any change in the
modal split. The explanation lies in the investment in
new public transport infrastructure in Paris. This kept
the market share of public transport stable at a relatively
high level of 31%. However, access to public transport
after suburbanisation of the workplace is not the only
factor. Access to parking tends to be a particularly important factor for the increase in car use. Free access to
parking induces people to drive more, even when the
new suburban location is well served by public transport
(Aarhus, 2000).
The research eld has been fairly silent on the question of how deconcentration aects travel for purposes
other than commuting. Naess et al. (1996) do take other
travel purposes into account, but do not analyse them
separately. To our knowledge, the only exception is the
work of Gordon et al. (1988, 1989a). In their 1988 study,
they state that a policentric structure facilitates an increase of travel for family, personal, social and recreative purposes. Furthermore, within metropolitan areas,
they report that travel distances for non-work motives
are generally lower in central cities than in suburban
communities.
In the light of our review of the literature on the
impact that mono- and policentricity of urban regions
have on travel behaviour, we conclude that empirical
studies of this relationship are relatively scarce and
sometimes contradictory. There is a substantial dierence between the theory and travel behaviour in the

176

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

actual cities. The literature also shows that the eect of


policentricity on travel behaviour is scale dependent. In
our analyses, we will conne ourselves to the structure
of metropolitan regions. Within these regions, we take
into account the dierence between more central and
more suburban locations. We will also include analyses
of non-work travel purposes, which are almost entirely
absent in the literature. Finally, we will examine the
inuence of personal characteristics on travel behaviour.
As we note elsewhere with regard to studies on other
aspects of urban form and travel behaviour (Dieleman
et al., 2001), the individual level is rarely taken into
account in studies of travel in policentric urban regions.
3. Research design
In this paper, we use the Dutch National Travel
Survey (in Dutch: het Onderzoek VerplaatsingsGedrag
(OVG)) to address the research questions introduced
above. The OVG has been carried out since 1978. Since
1995, 7000 households in the Netherlands have been
interviewed each month. Each member of a household is
asked to ll in a travel diary for one day. Every year,
more than 150,000 people (including children) participate in this survey (CBS, 1998). For the analyses presented here, we use the OVG 1998 survey. We selected
heads of households and their partners, if present, for
inclusion in our analysis. This is 82.3% of all complete
individual records in the OVG, the other cases mainly
being children above the age of 5. This will enable us to
make more sound comparisons across household types
and spatial contexts. Furthermore, heads of households
and their partners are most likely to hold regular employment and carry out most of the household maintenance tasks, such as shopping.
Fig. 2 shows which factors are related to travel behaviour in our analyses. Two travel behaviour characteristics were selected as dependent variables: modal
choice and distance travelled. In these analyses, we distinguish ve transport modes: walking, cycling, public
transport (by bus, subway, tram and train), private car
(as driver), and other (mainly car passengers, but also
motorcycle, moped and taxi). The distinction between
drivers and passengers allows us to obtain a better understanding of the number of vehicle kilometres travelled by car.
The personal/household attributes are operationalised
in terms of the household type to which a person belongs. This typology of households is based on a combination of three characteristics: the size of the
household, the presence of young children (<12 years
old), and the number of adult members who participate
in the labour market. Based on these characteristics,
nine household types are distinguished (see Table 2 for
an example). Besides this typology, we also use other

Fig. 2. Variables in analyses.

characteristics of the household or its members: `annual


household income' (six classes), `car ownership of the
household' (yes or no), and a person's `highest educational level' (four classes), `gender' and `age' (10 classes).
Our classication of residential environments of the
respondents is based on two criteria: the structure of the
urban system; and the location and urbanisation level of
the municipality. Van der Laan (1996, 1998) constructed
a typology of daily urban systems based on two dimensions: the degree to which the suburban commuters
are oriented toward the central city; and the degree to
which central-city commuters are oriented toward suburban municipalities. The scores on both dimensions
were used to distinguish four types of functional daily
urban systems (see Fig. 1):
Central: Just as in a monocentric system, commuter
streams are predominantly oriented toward the central city.
Decentral: The central city is not a magnet for commuters. The suburbs attract commuters from other
suburbs and from the central city.
Cross-commuting: Many suburban commuters work
in other suburbs, and many inhabitants of the central
city are locally employed. In this case, the labour
market has a dual spatial character. The suburban
part and the central-city part do not have many relations between them.
Exchange-commuting: Many suburban commuters
travel to the central city, while many central-city residents work in the suburbs. According to Van der
Laan (1996, p. 10), this is indicative of a qualitative
spatial job mismatch between demand and supply.

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

177

4. Modal choice

This typology of Van der Laan is well suited for our


analysis of travel behaviour in monocentric and policentric urban systems, because it takes into account not
only the monocentric or policentric structure of the urban system, but also looks at its eects on commuting
patterns. Of course, this classication is especially suitable for the analysis of travel behaviour for work trips
and is probably less relevant for travel for shopping
purposes.
Three levels of urbanisation are distinguished: core
cities, suburbs, and growth centres. Growth centres are
the Dutch equivalent to the new town concept and were
developed in the 1970s to accommodate the overspill of
the population from the big cities. The OVG data only
refer to the municipal level. Therefore, we could not
include the characteristics of neighbourhoods within
those municipalities in our analysis.
To analyse trip purposes, we created three categories:
commuting (excluding business trips), shopping trips,
and leisure trips (social calls, culture, recreation, sports,
and walking and driving around).
On the basis of the OVG data, we rst review some
descriptive statistics on the use of transport modes and
the distances travelled in four types of urban systems in
the Netherlands. Then we apply multivariate statistical
analyses to study the relative importance of each factor
(personal attributes, type of residential environment,
trip purpose) for travel behaviour.

In this section, we analyse modal choice and its relationship to household characteristics, residential environment and trip purpose (see Fig. 2). First, we
present some cross-tabulations of modal choice with
each of the determinants considered. Then we present
three logistic regressions, one for each of the trip purposes.
Adults in the Netherlands make on average 3.5 trips
per day, most frequently by car (Table 1). For all trip
purposes, the private car is the favourite means of
transportation. But people also make extensive use of
the bicycle, especially for shopping and leisure activities.
In terms of the total number of trips, walking accounts
for a signicant share, far exceeding the use of public
transport.
Modal split and number of trips vary substantially
across the types of households distinguished here (Table
2). The more workers there are in a household, the
larger the number of trips made. Of course, this is
mainly due to the higher number of commuting trips
they make. For leisure trips, the reverse is true: households with one worker make more trips than twoworker households. Adults belonging to households
with children make the most trips, probably because
they bring the kids to school and to their leisure activities. Families in general, and two-worker families in

Table 1
Modal split by trip purpose (average number of trips per person) in the Netherlands, 1998
Travel purpose

Car driver

Public transport

Bicycle

Walking

Other

Total

Work
Shopping
Leisure
Other

0.35
0.33
0.38
0.41

0.06
0.03
0.04
0.03

0.17
0.28
0.21
0.18

0.03
0.19
0.24
0.12

0.06
0.12
0.22
0.09

0.67
0.95
1.09
0.83

Total

1.47

0.16

0.84

0.58

0.49

3.54

Source: OVG (1998).


Table 2
Modal split for all purposes by type of household (average number of trips per person) in the Netherlands, 1998
Household type

Car driver

Public transport

Bicycle

Walking

Other

Total

Family, one worker


Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other households

1.87
2.15
1.49
1.44
1.77
0.85
1.64
0.66
1.52

0.06
0.10
0.05
0.16
0.19
0.09
0.34
0.28
0.12

1.14
0.96
0.84
0.77
0.70
0.66
1.00
0.78
0.82

0.73
0.65
0.67
0.56
0.45
0.53
0.70
0.73
0.45

0.52
0.50
0.51
0.53
0.65
0.58
0.34
0.32
0.46

4.32
4.37
3.56
3.45
3.77
2.70
4.02
2.77
3.36

Total

1.47

0.16

0.84

0.58

0.49

3.54

Source: OVG (1998).

178

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

particular, are also the most frequent users of the private


car and the bicycle. Rarely do they use public transport,
probably for the same reasons mentioned above. In
contrast, single persons and couples use public transport
relatively frequently.
The number of trips made each day varies slightly by
the type of urban system people live in or the level of
urbanisation of the municipality (Table 3). Living in an
exchange-commuting or central region has the opposite
eect on the number of trips. This number is high in
exchange regions and low in central urban systems.
Modal choice varies substantially by urbanisation
level, as we have demonstrated in an earlier paper
(Dieleman et al., 2001). Most obvious are the dierences
between those living in the core cities of urban systems
and those in the suburbs and growth centres. Adults
living in the core cities use the car less frequently and
travel more by public transport. This reects the higher
availability of public transport in the core cities of the
Netherlands. Growth centres also have a relatively high
share of public transport, because these planned communities were designed to have railway stations. Residents of the core cities walk and use the bike more often
than those living in the suburbs and growth centres. In
addition, modal choice seems to be inuenced by the
type of daily urban system. In urban systems with a high
level of cross-commuting, where many commuters from
the suburbs also live in the suburbs, the use of the private car is high and the use of public transport low. In

the Netherlands, the public transport system is still


mainly oriented toward the core cities, while suburban
communities are less connected by public transport.
Therefore, the relatively high level of car use in crosscommuting urban regions makes intuitive sense.
Of course the results shown in Tables 13 are only
preliminary. They do not reect any interrelationships
that may also exist between household attributes and the
residential environment. Up to this point in the analysis,
we only considered the eect of household type. As
mentioned earlier, other attributes also play a role in
modal choice (Fig. 2), and modal choice will vary by trip
purpose. In that light, we now present three multinomial
logistic regression models, one for each trip purpose
(Tables 46). The dependent variable `main travel mode'
consists of three categories: car driver, public transport,
and bicycle and walking. When the same reference category is used, only two submodels need to be generated;
all coecients are relative to the reference category
(Wrigley, 1985). Here driving a private car is taken as
the point of reference and the parameters shown in
Tables 46 express the use of public transport and cycling and walking relative to driving. For example, in
the rst model for public transport (Table 4), the parameter B of )3.707 indicates a sharp decline in the
likelihood of using that mode relative to driving, if a car
is available in the household. The odds ratio expresses
the eect of the independent variable on the likelihood
of using that transport mode in comparison to the

Table 3
Modal split for all purposes by residential environment (average number of trips per person) in the Netherlands, 1998
Residential environment

Car driver

Public transport

Bicycle

Walking

Other

Total

Central

Core city
Suburb
Growth centre
Group total

1.30
1.62
1.42
1.44

0.24
0.10
0.29
0.18

0.78
0.75
0.48
0.76

0.69
0.51
0.62
0.61

0.49
0.52
0.54
0.50

3.50
3.50
3.34
3.49

Decentral

Core city
Suburb
Growth centre
Group total

1.05
1.51
1.63
1.29

0.38
0.16
0.23
0.28

0.92
0.84
0.63
0.86

0.72
0.63
0.58
0.67

0.45
0.48
0.56
0.48

3.52
3.62
3.63
3.57

Cross-commuting

Core city
Suburb
Growth centre
Group total

1.54
1.66
1.64
1.61

0.11
0.08
0.14
0.10

0.84
0.83
1.00
0.84

0.58
0.53
0.53
0.55

0.48
0.52
0.53
0.50

3.54
3.62
3.86
3.60

Exchange-commuting

Core city
Suburb
Growth centre
Group total

1.05
1.70
1.56
1.26

0.24
0.10
0.18
0.20

1.20
0.88
0.77
1.08

0.74
0.55
0.60
0.68

0.44
0.45
0.54
0.45

3.68
3.68
3.65
3.67

All urban areas

Core city
Suburb
Growth centre

1.18
1.58
1.58

0.29
0.12
0.23

0.90
0.80
0.65

0.70
0.57
0.59

0.46
0.50
0.55

3.54
3.57
3.60

Outside urban systems

1.62

0.07

0.84

0.49

0.50

3.51

Total

1.47

0.16

0.84

0.58

0.49

3.54

Source: OVG (1998).

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

179

Table 4
Logistic regression model for work trips
Public transport
B
Car ownership
Car in household

Bicycle/walking
Odds ratio



)3.707

0.02

Chi-square
Odds ratio



)3.099

0.05

298.6

Household type
Family, one worker (ref.)
Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other households

0.145
)0.157
0.239
0.402
)0.423
)0.160
0.189
0.217

1.16
0.86
1.27
1.50
0.66
0.85
1.21
1.24

)0.026
0.046
)0.095
)0.083
)0.292
)0.633
)0.322
0.004

0.98
1.05
0.91
0.92
0.75
0.53
0.72
1.00

Household income

)1.87E ) 04

1.00

)1.09E ) 02

0.99

Education
Higher (ref.)
Medium, higher
Medium, lower
Lower
Gender
Female
Age
Age squared
Residential environment
Central, core city (ref.)
Central, suburb
Decentral, core city
Decentral, suburb
Cross-commuting, core city
Cross-commuting. suburb
Exchange-commuting, core city
Exchange-commuting, suburb
Outside urban system
Constant


a 0:10
a 0:05

a 0:01


Reference category car driver.

5261.3

173.7
366.1

)0.484
)0.530
)1.001

0.62
0.59
0.37

)0.117
0.033
0.049

0.89
1.03
1.05

0.527

1.69

0.677

1.97

)1.17E ) 02
1.45E ) 04

0.99
1.00

4.24E ) 02
)3.78E ) 04

1.04
1.00

1109.7
22.1
36.5
828.3

)0.268
0.582
0.318
)0.309
)0.029
0.165
)0.623
)0.065

0.77
1.79
1.37
0.73
0.75
1.18
0.54
0.52

1.603
)2 log likelihood null model
)2 log likelihood nal model
v2
Nagelkerke q2

likelihood of using the reference transport mode. For


instance, the ratio of 1.69 for gender in the rst model
indicates that women are 1.69 times more likely to use
public transport than men. The chi-squared statistics
provides an indication of the relative weight of that
variable in the model. Each model uses the same set of
independent variables, permitting comparison across the
models. Because there are so few growth centres, they
are combined with the suburbs.
In all the three models, the independent variables have
a clear and (fairly) strong relationship to modal choice
(Tables 46). As indicated by the pseudo R-square statistics Nagelkerke's q2 the models explain a fair share

)0.308
0.164
)0.194
0.065
)0.281
0.287
)0.161
)0.199

0.74
1.18
0.82
1.07
0.76
1.33
0.85
0.82

1.836
38432.5
28940.8
9491.8
0.186

of the variation in mode choice, although the largest part


remains unexplained. The model for shopping performs
somewhat better than those for leisure and work trips.
The models are consistent in showing that car ownership
reduces using public transport as well as cycling and
walking. They also consistently show that women use
public transport and walk and cycle more than men.
There are two reasons for this dierence. One is that
working women live closer to their place of employment.
The other is that women take care of two-thirds of all
maintenance and caring tasks in the household (Dijst,
1995, 1999). Use of public transport increases along with
the level of education, but the relationship is not linear. A

180

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

Table 5
Logistic regression model for shopping trips
Public transport
B
Car ownership
Car in household

Bicycle/walking
Odds ratio



)4.824

0.01

Chi-square
Odds ratio



)3.711

0.03

466

Household type
Family, one worker (ref.)
Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other households

0.069
)0.281
1.189
1.017
1.238
0.573
0.755
0.965

2.00
0.76
3.29
2.77
3.45
1.77
2.13
2.62

)0.096
)0.458
0.085
)0.154
0.315
)0.256
)0.370
0.004

0.96
0.77
1.17
0.91
1.48
0.85
0.77
1.06

Household income

)3.29E ) 03

1.00

)5.49E ) 03

1.00

Education
Higher (ref.)
Medium, higher
Medium, lower
Lower

)0.091
)0.172
0.095

0.91
0.84
1.10

)0.236
)0.092
0.076

0.82
0.95
1.17

1.340

3.82

1.104

3.12

Age

2.17E ) 02

1.02

1.01E ) 02

1.01

)0.304
0.546
)0.146
)0.640
)1.035
0.016
)0.163
)1.143

Constant

)0.785

)2 log likelihood null model


)2 log likelihood nal model
v2
Nagelkerke q2

Reference category car driver.

4333.9
155.7
951.8

Residential environment
Central, core city (ref.)
Central, suburb
Decentral, core city
Decentral, suburb
Cross-commuting, core city
Cross-commuting, suburb
Exchange-commuting, core city
Exchange-commuting, suburb
Outside urban system

a 0:10
a 0:05

a 0:01

76.4
202.1

Gender
Female



8240.7

0.74
1.73
0.86
0.53
0.36
1.02
0.85
0.32

higher income lowers the probability of walking and cycling for all trip purposes considered. When income rises,
the use of public transport becomes less likely for shopping and, to a lesser extent, for work. Yet, for leisure activities, public transport tends to be used more frequently
by higher-income households. The relationship between
modal choice and the type and age of the household varies
according to the purpose of the trip. Working couples use
public transport relatively often but bike and walk infrequently for their journey to work. For shopping and leisure trips singles and couples use public transport more
frequently than families, while the frequency of biking
and walking varies more across the household types.

)0.245
0.353
)0.066
)0.110
)0.188
0.476
)0.078
)0.236

0.84
1.54
1.00
1.01
0.92
1.77
1.03
0.84

3.098
52761.7
34887.4
17874.3
0.262

Even after taking the strong inuence of personal


characteristics on modal choice into account, the inuence of residential environment on modal choice
remains strong for all trip purposes. The chi-square of
this variable in the various models is high and indicates that this factor is as important for modal choice
as the personal characteristics (Tables 46). For all
three types (working, shopping, leisure), driving is
most important in suburban locations, while public
transport and slow modes (walking and biking) are
more common in the core cities. This dierence is
clearly related to the density of land use in the core
cities and the wider availability of public transport

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

181

Table 6
Logistic regression model for leisure trips
Public transport
B
Car ownership
Car in household
Household type
Family, one worker (ref.)
Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other household
Household income
Education
Higher (ref.)
Medium, higher
Medium, lower
Lower
Gender
Female
Age
Age squared
Residential environment
Central, core city (ref.)
Central, suburb
Decentral, core city
Decentral, suburb
Cross-commuting, core city
Cross-commuting, suburb
Exchange-commuting, core city
Exchange-commuting, suburb
Outside urban system
Constant


a 0:10
a 0:05

a 0:01


Reference category car driver.

Bicycle/walking
Odds ratio



)4.487

0.01

Chi-square
Odds ratio



)3.152

0.04

8373.3
583.7

0.097
)1.497
0.828
0.501
0.793
0.288
0.485
0.477
7.03E ) 03

1.10
0.22
2.29
1.65
2.21
1.33
1.62
1.61

)0.076
)0.285
0.057
)0.150
0.111
)0.554
)0.503
)0.061

0.93
0.75
1.06
0.86
1.12
0.57
0.60
0.94

1.01

)2.49E ) 03

1.00

53.7
83.9

)0.200
)0.244
)0.341

0.82
0.78
0.71

)0.116
)0.040
0.088

0.89
0.96
1.09

0.973

2.65

0.754

2.13

)1.56E ) 02
2.88E ) 04

0.98
1.00

5.08E ) 02
)4.64E ) 04

1.05
1.00

2472.9
203.1
191.4
471.8

)0.297
0.293
)0.289
)0.620
)0.720
)0.006
)0.377
)0.827

0.74
1.34
0.75
0.54
0.49
0.99
0.69
0.44

0.420
)2 log likelihood null model
)2 log likelihood nal model
v2
Nagelkerke q2

there, as we argue elsewhere (Dieleman et al., 2001).


But the type of urban system also seems to inuence
modal choice. In cross-commuting urban regions, the
use of public transport is low both in central and
suburban locations. That is because the public transport system is not geared to this pattern of travel
behaviour. Conversely, for commuting and shopping,
people use public transport relatively often in urban
systems with a decentralised form of interaction,
though biking and walking are also common for such
trips. These slower modes of transport are also popular in urban systems with exchange commuting. The
reason may be that most urban systems in these cat-

0.048
0.089
)0.026
0.187
0.074
0.284
)0.065
0.099

1.05
1.09
0.97
1.21
1.08
1.33
0.94
1.10

1.611
53323.7
38536.8
14786.8
0.207

egories are located in the Randstad Holland. There,


densities are higher and more activity places can be
reached by cycling and walking.

5. Distance travelled
As in the discussion of modal choice, we rst present
some cross-tabulations of distance travelled against
three important dimensions: trip purpose, household
type, and type of urban system and level of urbanisation
within this system. We then present regression models

182

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

One reason is their relatively long commute; another is


that they have to take their children to school or other
activity places. Persons in households where no one is
employed for at least 12 h a week (mainly senior citizens) have much lower travel distances. This nding is in
line with the outcomes of earlier research (Schwanen,
1999). As Table 8 illustrates, we must take personal
circumstances into account when studying the relationship between urban structure and travel behaviour, because these attributes are strongly correlated with travel
behaviour.
Table 9 shows travel distances for the journey to
work. The gures are broken down by type of urban
system, level of urbanisation within these systems and
mode of travel. We focus here on work-related trips for
two reasons. First, we may assume that the inuence of
urban structure on travel distance is most operative in
the journey to work. Secondly, the classication of urban structure is based on commuting patterns. One
would expect to nd relatively long commuting distances in urban systems with exchange commuting, because few persons work locally within such areas. In
cross-commuting urban regions, the reverse is true.
Thus, according to the `co-location hypothesis', commuting distances should be fairly low. Table 9 (last
column) gives some evidence for this. In exchangecommuting urban systems, people travel relatively far,
no matter where they live within the region. In cross-

for work trips. These models include all explanatory


variables shown in Fig. 2.
In 1998, each household head and partner interviewed in the OVG survey travelled on average some 39
km per day for various purposes (Table 7). Not surprisingly, in terms of distance travelled per day, the
private car is by far the most important means of
transportation. Table 7 further shows that commuting
and leisure travel account for the largest shares of the
mean distance travelled per day. The high number of
kilometres for the transport mode `other' for leisure
purposes (4.8, mainly car passengers) reects that adults
frequently participate in out-of-home leisure activities in
company with others. In the light of the gures in Table
7, one would expect that if urban structure has any eect
on the distance travelled, it would be for commuting and
leisure trips. Trips for shopping tend to be so short that
the main activity is evidently at the local level.
Household type is clearly a major factor in the distance individuals travel (Table 8). Two-earner couples
and single working persons are very mobile and travel
long distances per day, mainly by car. Singles frequently
travel long distances by public transport. If we were to
break Table 8 down by trip purposes (not done here), we
would see that members of these two types of household
generally travel long distances, because they tend to live
far away from their work. Families with one or two
workers also travel fairly long distances, often by car.

Table 7
Average number of kilometres travelled per person by trip purpose and travel mode in the Netherlands, 1998
Purpose
Work
Shopping
Leisure
Other
Total

Car driver

Public transport

Bicycle

Walking

Other

Total

7.1
2.3
6.8
5.8

1.8
0.4
1.8
1.1

0.6
0.6
1.0
0.4

0.0
0.2
0.4
0.1

1.2
1.1
4.8
1.6

10.8
4.6
14.8
8.9

22.0

5.2

2.6

0.7

8.7

39.1

Source: OVG (1998).

Table 8
Average number of kilometres travelled by person for all purposes by household type and travel mode in the Netherlands, 1998
Household type

Car driver

Public transport

Bicycle

Walking

Other

Total

Family, one worker


Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other households

24.4
29.5
16.2
23.6
30.3
11.9
28.6
8.7
22.7

2.4
3.6
1.2
5.4
6.9
2.9
11.6
7.7
4.0

2.6
2.5
1.9
2.7
2.3
2.5
3.2
2.4
2.6

0.8
0.7
0.6
0.8
0.6
0.8
0.7
0.9
0.6

8.6
9.2
6.8
10.2
12.0
10.1
6.8
4.8
7.9

38.7
45.6
26.8
42.7
52.1
28.3
51.0
24.5
37.7

Total

22.0

5.2

2.6

0.7

8.7

39.1

Source: OVG (1998).

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

183

Table 9
Average number of kilometres travelled per person for work trips by residential environment and travel mode in the Netherlands, 1998
Residential environment

Car driver

Public
transport

Bicycle

Walking

Other

Total

Central

Core city
Suburb
Growth centre
Group total

5.7
6.9
8.1
6.3

2.6
1.1
2.0
1.9

0.7
0.6
0.6
0.6

0.0
0.0
0.0
0.0

1.1
1.2
1.9
1.2

10.1
9.8
12.6
10.1

Decentral

Core city
Suburb
Growth centre
Group total

5.5
6.5
8.7
6.3

3.5
1.6
3.4
2.8

0.8
0.7
0.5
0.7

0.0
0.0
0.0
0.0

0.9
0.9
1.2
0.9

10.7
9.8
13.8
10.7

Cross-commuting

Core city
Suburb
Growth centre
Group total

4.9
7.2
8.2
6.3

2.1
1.5
1.5
1.7

0.7
0.7
0.7
0.7

0.0
0.0
0.0
0.0

0.7
1.0
1.0
0.9

8.4
10.4
11.4
9.6

Exchange-commuting

Core city
Suburb
Growth centre
Group total

5.7
9.0
7.9
6.7

3.7
0.8
1.3
2.7

0.9
0.7
1.1
0.9

0.0
0.0
0.0
0.0

0.9
0.6
1.7
0.9

11.1
11.2
12.1
11.3

All urban areas

Core city
Suburb
Growth centre

5.6
6.9
8.4

3.1
1.3
2.7

0.8
0.6
0.6

0.0
0.0
0.0

1.0
1.0
1.4

10.4
10.0
13.1

Outside urban systems

8.2

1.1

0.5

0.0

1.5

11.3

Total

7.1

1.8

0.6

0.0

1.2

10.8

Source: OVG (1998).

commuting urban systems, people living in the core


cities travel relatively short distances. The two other
types of urban systems are somewhere in between. A
similar picture may be sketched for the other two trip
purposes. Comparable conclusions can be drawn for
leisure trips. For shopping trips, the eect of the urban
system on the distance travelled is negligible.
Generally speaking, the level of urbanisation of the
place of residence within the daily urban systems seems
to have a stronger inuence on distance travelled by
mode than the type of urban system has. When transport modes are not taken into account, the average
distance travelled in core cities resembles the average for
the suburbs. However, when transport modes are considered, the distance travelled by pubic transport turns
out to be larger in core cities than in suburbs. People
living in growth centres travel long distances for work
and leisure, both by car and, to a lesser extent, by public
transport. For shopping trips, the distance travelled is
highest for growth centres in central and decentralised
systems. The distance for shopping activities is also high
for suburban communities in cross- and exchangecommuting systems.
To measure the eects of attributes of persons and the
type of urban system they live in on travel distances, we
estimated three regression models (distance travelled by
car, public transport and bicycling/walking) for each of
the three trip purposes. The natural logarithm of the

distance travelled was taken as the dependent variable.


In this paper, we limit the discussion to the models for
work purposes (Table 10). This narrower focus reects
our nding that only for commuting trips do the type of
urban regions and the level of urbanisation of the municipality have a clear impact on distances travelled by
the three modes. Among the other purposes, no clear
eects of the structure of the urban system on the distance travelled could be noticed. However, the models
showed that people in core cities travel relatively long
distances by public transport, because such transport is
easily available there.
Although distance travelled for work trips is impacted by spatial structure, the models in Table 10
perform relatively poor, especially for the distance
travelled by bicycle or on foot and as a car driver. Only
part of the variation in commuting distances can be
attributed to dierences in the personal and household
situation or the residential environment.
The results of the regression models demonstrate the
importance of including personal attributes in the
analysis of travel behaviour. Gender is a strong determinant of distance travelled by each of the three travel
modes. In the Netherlands, women travel much shorter
distances to work than men, regardless of which mode
of transport they use. This is consistent with many
studies on the household responsibility hypothesis
(Johnston-Anumonwo, 1992; Turner and Niemeier,

184

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

Table 10
Regression models for kilometres travelled by mode, for work trips
Car driver
B
Car ownership
Car in household
Household type
Family, one worker
(ref.)
Family, two workers
Family, no worker
Couple, one worker
Couple, two workers
Couple, no worker
Single, worker
Single, no worker
Other households
Household income

0.253

Public transport
Beta

0.022

)0.089
0.161
)0.015
0.055
)0.359
0.032
0.105
)0.002
3.92E ) 03

)0.033
0.008
)0.004
0.023
)0.033
0.009
0.006
)0.001
0.053

)0.156
)0.244
)0.309

)0.073
)0.105
)0.051

Gender
Female

)0.486

)0.220

Age
Age squared

)4.14E ) 05 )0.034

Constant

0.001

Beta

0.055

0.026

0.006

Education
Higher (ref.)
Medium, higher
Medium, lower
Lower

Residential
environment
Central, core city
(ref.)
Central, suburb
Decentral, core city
Decentral, suburb
Cross-commuting,
core city
Cross-commuting,
suburb
Exchange, core city
Exchange, suburb
Outside urban
systems

R2
change

0.001

2.59E ) 03

0.045

0.013
0.046
0.035
)0.021

)0.321
)0.474
)0.504

)0.157
)0.195
)0.063

)0.403

)0.208

)1.57E ) 03 )0.017

Beta

)0.140

)0.050

)0.026
)0.106
0.014
0.085
)0.185
)0.121
)0.218
)0.040
0.008

0.042
0.000

)0.122
)0.239
)0.126
0.243

0.013

0.173

0.070
0.227
0.186

0.012
0.039
0.090

0.259
)0.176
0.237

7.57E ) 04

0.012

a 0:10
R2 0.070
a 0:05

a 0:01
Dependent variable is natural log kilometres travelled.


1997). This proposition states that women tend to have


shorter commutes and engage in trip chaining more
frequently when commuting. Education is also strongly
related to distance travelled by car or public transport.
The higher the level of education, the longer the commuting distances will be (e.g., Rouwendal and Rietveld,
1994). Similarly, a household's income is positively related to the distance travelled to work by car and public

0.001
0.000

0.056
0.063
)0.074

0.025
0.027
)0.014

)0.405

)0.192

3.42E ) 02
)3.49E ) 04

0.329
)0.286

0.036
0.002
0.006

)0.024
)0.053
)0.060
0.005

)0.007
)0.016
)0.019
0.001

0.029

)0.048

)0.008

)0.042
0.030
)0.195

)0.010
0.005
)0.091

3.959

0.000

)0.009
)0.005
0.004
0.033
)0.018
)0.038
)0.022
)0.016

)0.035
)0.101
)0.053
0.032

0.067
)0.024
0.108

R2
change

0.005

0.040

0.072

2.970

0.045

0.006

0.044
0.182
0.106
)0.158

0.006

)0.036
0.026
0.000
)0.015
0.005
)0.040
)0.017
)0.027

0.009

0.045

R2
change

0.008
)0.098
0.796
0.000
)0.031
0.069
)0.099
)0.127
)0.67

0.004

Bicycling/walking

0.971

R2 0.148

R2 0.050

transport. While household type remains important, it is


less so than in Table 8, now that other variables have
been brought into the analysis. Two-worker families
travel less than one-worker families. Of all household
types considered, dual-earner couples travel the longest
distance by car and by bicycle and on foot. In addition,
single workers walk or bike the shortest distance for
work. Car ownership and age have a weaker relation-

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

ship with distance travelled, probably because this relationship is captured by the other personal attributes
already.
The type of daily urban system and the level of urbanisation of the residential environment within these
areas remain important determinants of commuting
distance, especially for trips by public transport. In exchange-commuting regions, people living in the core
cities travel relatively long distances by public transport,
whereas those living in the suburbs drive long distances.
Also in decentralised urban regions, distance by car is
relatively long, while the distance travelled by public
transport is relatively short. In cross-commuting urban
systems, the distance travelled by car is relatively short
and by public transport fairly long. Although not shown
here, comparable conclusions can be drawn for decentralised and exchange-commuting urban systems with
respect to leisure trips.
6. Conclusions
In this paper we focused on the question of how
monocentric and policentric urban structures aect
modal choice and travel distance for dierent travel
purposes. Based on characteristics of commuting ows,
we divided the urban systems of the Netherlands into
four types: one monocentric type and three types of
policentric urban systems. The main conclusions may be
summarised as follows.
First of all, it is useful to include trips for purposes
other than commuting in studies of travel behaviour in
cities. The journey to work accounts for a mere fth of
all trips. In terms of the distance travelled, the daily
commute is roughly one-fourth of the total distance
travelled. In the Netherlands, the number of trips and
the distance travelled for leisure (social calls, culture,
recreation, sports, walking and driving around) comprise most of the travel among adults. Personal attributes prove to be strongly related to both modal choice
and distances covered. It is therefore useful to include
these characteristics in studies of travel behaviour, even
when the focus is on the relationship between urban
form and travel behaviour, as in this paper.
Secondly, our analysis yields mixed results on the
structure of urban regions with respect to modal
choice and distance travelled. Consider modal choice.
Deconcentration of urban land use to suburban locations and new towns almost certainly promotes the
use of the private car for all purposes. It also leads to
less use of public transport as well as of cycling and
walking, which are still very important travel modes in
Dutch cities. The urban structure of cross-commuting,
whereby many suburban residents work in suburban
locations, has the same eect. Decentralised and exchange-commuting urban systems seem to promote

185

public transport and biking. Evidently, these policentric urban structures are still of limited size in the
Netherlands. They thus lend themselves well for bicycle use, which is still widespread.
Let us now consider distance travelled instead of
modal choice. The outcomes seem to point in a different direction. In cross-commuting urban systems,
the distance travelled to work is relatively small. This
situation comes close to the `co-location' hypothesis of
scholars like Gordon and Richardson (1997). They
expect that rms and households periodically readjust
their location to achieve more balanced average
commuting distances. Longer commutes were observed
in the exchange-commuting urban systems in the
Netherlands, where the location of homes and jobs is
unbalanced (spatial mismatch). These imbalances
could well be related to the country's strong spatial
planning policies. By inuencing both the location of
residences and jobs in some urban regions, these
policies may have obstructed the spatial readjustment
processes of rms and households.
In sum, the deconcentration of urban land use and
the development of policentric urban structures seem
to lead to more use of the private car for all purposes.
At the same time, there is some evidence that the
distance travelled to work will not necessarily increase
by the development of policentric urban structures
and can be relatively small in cross-commuting urban
systems.

References
Aarhus, K., 2000. Oce location decisions, modal split and the
environment: the ineectiveness of Norwegian land use policy.
Journal of Transport Geography 8, 287294.
Anas, A., Arnott, R., Small, K.A., 1998. Urban spatial structure.
Journal of Economic Literature 26, 14261464.
Asmervik, S., Naess, P., 1995. The sustainable city and the use of
energy. In: The European City: Sustainable Urban Quality, part 2:
Selected Conference Papers. Ministry of Environment and Energy,
Spatial Planning Department, Copenhagen, pp. 719.
Bailey, T.J., 1999. Modelling the residential sub-market: breaking the
monocentric mould. Urban Studies 36 (7), 11191135.
Batten, D.F., 1995. Network cities: creative agglomerations for the
21st century. Urban Studies 32 (2), 313327.
Bell, D.A., 1991. Oce location city or suburbs? Travel impacts
arising from oce relocation from city to suburbs. Transportation
18, 239259.
Bolotte, L., 1991. Transport in Paris and the Ile de France. Built
Environment 17 (2), 160171.
Centraal Bureau voor de Statistiek, 1998. Onderzoek verplaatsingsgedrag 1998 documentatie ten behoeve van tape-gebruikers. CBS,
Voorburg/Heerlen.
Cervero, R., 1996. Jobs-housing balance revisited: trends and impacts
in the San Francisco Bay Area. Journal of the American Planning
Association 62 (4), 492511.
Cervero, R., Landis, J., 1991. Suburbanization of jobs and the journey
to work. Working Paper, No. 83. The University of California
Transportation Center, Berkeley.

186

T. Schwanen et al. / Journal of Transport Geography 9 (2001) 173186

Cervero, R., Wu, K.-L., 1998. Sub-centring and commuting: evidence


from the San Francisco bay area. Urban Studies 35 (7), 10591076.
Clark, W.A.V., 2000. Monocentric to policentric: new urban forms
and old paradigms. In: Bridge, G., Watson, S. (Eds.), A Companion to the City. Blackwell companions to geography, pp. 141154.
Clark, W.A.V., Kuijpers-Linde, M., 1994. Commuting in restructuring
urban regions. Urban Studies 31 (3), 465483.
Commission of the European Communities, 1994. Europe 2000+:
Cooperation for European Territorial Development. Oce for the
Ocial Publications of the European Communities, Luxembourg.
Cortie, C., Dijst, M., Ostendorf, W., 1992. The Randstad a metropolis?
Tijdschrift voor Economische en Sociale Geograe 83, 278288.
Dieleman, F.M., Dijst, M., Burghouwt, G., 2001. Urban form and
travel behaviour: micro-level household attributes and residential
context. Urban Research centre Utrecht, Utrecht University.
Dieleman, F.M., Dijst, M., Spit, T., 1999. Planning the compact city:
the Randstad Holland experience. European Planning Studies 7
(5), 605621.
Dieleman, F.M., Faludi, A., 1998. Polynucleated metropolitan regions
in Northwest Europe: theme of the special issue. European
Planning Studies 6 (4), 365378.
Dijst, M., 1995. An elliptical life: action space as integral measure for
accessibility and mobility. Koninklijk Nederlands Aardrijkskundig
Genootschap/Faculteit Bouwkunde Technische Universiteit Delft,
(in Dutch, with an executive summary in English).
Dijst, M., 1999. Two-earner families and their action spaces: a case
study of two Dutch communities. GeoJournal 48, 195206.
Ewing, R., 1997. Is Los Angeles-style sprawl desirable? Journal of the
American Planning Association 63 (1), 107126.
Frost, M., Linneker, B., Spence, N., 1997. The energy consumption
implications of changing worktravel in London, Birmingham and
Manchester: 1981 and 1991. Transportation Research A 31 (1),
119.
Giuliano, G., Small, K.A., 1993. Is the journey to work explained by
urban structure? Urban Studies 30 (9), 14851500.
Gordon, P., Kumar, A., Richardson, H.W., 1988. Beyond the journey
to work. Transportation Research A 6, 419426.
Gordon, P., Kumar, A., Richardson, H.W., 1989a. Congestion,
changing metropolitan structure and city size in the United States.
International Regional Science Review 12 (1), 4556.
Gordon, P., Kumar, A., Richardson, H.W., 1989b. The inuence of
metropolitan spatial structure on commuting time. Journal of
Urban Economics 26, 138151.
Gordon, P., Richardson, H.W., 1996. Beyond polycentricity. The
dispersed metropolis, Los Angeles. Journal of the American
Planning Association 62 (3), 289295.
Gordon, P., Richardson, H.W., 1997. Are compact cities a desirable
planning goal? Journal of the American Planning Association 63
(1), 95106.

Gordon, P., Richardson, H.W., Jun, M.J., 1991. The commuting


paradox: evidence from the top twenty. Journal of the American
Planning Association 57 (4), 416420.
Gordon, P., Richardson, H.W., Wong, H.L., 1986. The distribution of
population and employment in a polycentric city: the case of Los
Angeles. Environment and Planning A 18, 167173.
Gordon, P., Wong, H.L., 1985. The cost of urban sprawl: some new
evidence. Environment and Planning A 17, 661666.
Ingram, G.K., 1998. Patterns of metropolitan development: what have
we learned? Urban Studies 35 (7), 10191035.
Jansen, G.R.M., 1993. Commuting: home sprawl, job sprawl, trac
jams. In: Salomon, I., Bovy, P., Orfeuil, J.-P. (Eds.), A billion trips
a day: tradition and transition in European travel patterns. Kluwer
Academic Publishers, Dordrecht, pp. 101127.
Johnston-Anumonwo, I., 1992. The inuence of household type on
gender dierences in work trip distance. Professional Geographer
44 (2), 161169.
Levinson, D.M., Kumar, A., 1994. The rational locator: why travel
times have remained stable. Journal of the American Planning
Association 60 (3), 319332.
Naess, P., 1995. Urban form and energy use for transportation. A
nordic experience. Norwegian Institute of Technology, Trondheim.
Naess, P., Sandberg, S.L., 1996. Workplace location, modal split and
energy use for commuting trips. Urban Studies 33 (3), 357380.
Naess, P., Sandberg, S.L., Re, P.G., 1996. Energy use for
transportation in 22 Nordic towns. Scandinavian Housing &
Planning Research 13, 7997.
Newman, P.W.G., Kenworthy, J.R., 1989. Gasoline consumption and
cities. A comparison of US cities in a global survey. Journal of the
American Planning Association 55 (1), 2436.
Rouwendal, J., Rietveld, P., 1994. Changes in commuting distances of
Dutch households. Urban Studies 31 (9), 15451557.
Schmitz, S., 1993. Tangential statt radial: langfristige Veranderungen
der Berufspendlerstr
ome in Deutschland. BfLR-Mitteilungen. 1/93
(Februar).
Schwanen, T., 1999. Met de auto op weg naar de toekomst? Het
verplaatsingsgedrag van ouderen in hun vrije tijd. Faculty of
Geographical Sciences: Utrecht University.
Turner, T., Niemeier, D., 1997. Travel to work and household
responsibility: new evidence. Transportation 24, 397419.
Usterud Hanssen, J., 1995. Transportation impacts of oce relocation.
Journal of Transport Geography 3 (4), 247256.
Van der Laan, L., 1996. Commuting in multinodal systems? Erasmus
University Rotterdam, Centre for Labour Market Analysis
(ECLA), Rotterdam.
Van der Laan, L., 1998. Changing urban systems: an empirical analysis
at two spatial levels. Regional Studies 32 (3), 235247.
Wrigley, N., 1985. Categorical Data Analysis for Geographers and
Environmental Scientists. Longman, London/New York.

You might also like