j.1467-9663.2011.00669.x Excess Travel in Non Professional Trips_why Look for It Miles Away

Embed Size (px)

Citation preview

  • EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS:WHY LOOK FOR IT MILES AWAY?tesg_669 20..38

    KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    Geography Department, Ghent University, Krijgslaan 281/S8, 9000 Gent, Belgium.E-mails: [email protected]; [email protected]; [email protected]

    Received: July 2010; accepted March 2011

    ABSTRACTBased on the spatial distribution of some quasi-daily destination classes and survey-reported tripdistances, regional variation in excess travel in non-professional trips in Flanders (Belgium) isassessed. To this end, proximity to various quasi-daily destinations is compared with the reporteddistance that is actually travelled to reach similar, but alternative, facilities. We note that in ruralareas (compared with urban areas) larger distances are travelled, although the closest facility ischosen more often. In the most urbanised areas, however, we note that spatial proximity is also animportant aspect in destination choice. Quantification of these phenomena can support thepractice of sustainable spatial planning by distinguishing areas that are toomono-functional or tooremote, and therefore need more functional diversity, and by identifying areas where densifica-tion is useful because the location is close to most quasi-daily destinations, reducing the need totravel over large distances.

    Key words: travel behaviour, Flanders, excess commuting, sustainable spatial development,accessibility

    INTRODUCTION

    Nss (2003) reduces the relationship betweenspatial proximity and mobility to its geometricessence: in an area with a high density of peopleand services, distances that are to be coveredbetween potential origins and destinationsare small. As the trip distance is related to theamount of energy required, empirical researchnot surprisingly shows that fuel consumptionfor transport per capita is actually lower in areaswith a high density than in regions with a lowdensity (Newman & Kenworthy 1989, 1999;Nss et al. 1996). This is a logical consequence,as Nss (2003, p. 59) states: The absence of anysuch influence would also have been quite sen-sational. But reality is obviously more complexthan a geometric problem. All kinds of factors,such as infrastructure configuration, routes ofpublic transport, or the lack of parking space,

    distort this obvious logic. Also, an unbalancedspatial mix, often caused by functional cityplanning, may cancel out the positive potentialof high density. Moreover, mode choice plays arole: cities with a high proportion of pedestri-ans, cyclists and public transport users will haveless traffic problems. Besides, on the regionallevel a clear linear relationship exists betweenfuel consumption and the number of kilome-tres travelled per person (Boussauw & Witlox2009).

    The main deviation between travel behav-iour and geometry is due to the fact that a highdegree of spatial proximity, and thus betteraccessibility, gives rise to new needs (Nss2003). Gains, in terms of both time and money,yielded by a higher level of accessibility arepartly offset by the individual who will makeuse of the increased choice range. When thenearest supermarket is located just 100 m from

    Tijdschrift voor Economische en Sociale Geografie 2012, DOI:10.1111/j.1467-9663.2011.00669.x, Vol. 103, No. 1, pp. 2038. 2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAGPublished by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

  • ones front door, then the threshold for visit-ing the second nearest supermarket, at forexample, 500 m, is particularly low, certainlywhen the latter offers more products or is alittle cheaper. But if the nearest store is locatedat 10 km, and the second nearest is at 20 km,the same person will probably go shopping atthe closest store. Although the use of a widerrange of accessible destinations, as well as anincrease in the number of trips may offset thepotential efficiency gains of the compact city,the aforementioned empirical studies suggestthat this is only partially the case.

    Handy et al. (2005) argue that the reason forthe emergence of a trip can be situated along achoice-necessity continuum. Driving aroundjust for fun (Ory & Mokhtarian 2005), is locatedat the choice end of this continuum, whilebuying a loaf of bread at the bakery around thecorner is at the need end of the same spec-trum. In these terms, excess travel is defined astravel beyond what is required for householdmaintenance, given choices about residentiallocation, job location, and activity participa-tion. The required trip length is then definedas the shortest route to the closest destinationpossible. The ratio between choice and neces-sity determines the excessive nature of any par-ticular trip, meaning the extent to which thetrip distance exceeds the distance to the closestfacility that could possibly satisfy the needs ofthe traveller. The degree of spatial proximitybetween the sites that are potential origins ordestinations of trips defines the actual travelleddistance. But spatial structure itself is also oneof the factors that influence the decision. In anarea with many options nearby, choice willoutweigh need. Furthermore, in this first area with a wide choice range the total trip lengthmight still be less than in the second area;where there is little choice.

    Vilhelmson (1999) develops a frameworkthat classifies activities based on the degree offlexibility in terms of physical location andpoint in time. Education is an example of anactivity where both location and time are speci-fied exactly. For an activity such as jogging thereverse is true. The last kind of activities arelargely determined by the choice of an indi-vidual, and may thus lead to excess travel. Vil-helmson (1999) finds that access to a car, nothaving children and having a part-time job is

    associated with an increase in the number ofkilometres travelled during this kind of freeactivity. It is evident that also the type of tripwill play a role in the degree of excess travel.Horner and OKelly (2007) quote the exampleof the difference between shopping trips forcomparison goods (which are only occasion-ally purchased), and convenience goods. Inthe latter category, attempts to minimise thedistance travelled will play a greater role than inthe first category. Based on interviews, Nss(2006) shows a willingness to travel longer dis-tances for work, education and visiting familyor friends in comparison with for example,schools, kindergartens and grocery shops.

    The excess commuting research frameworkoffers opportunities for studying this phenom-enon at a regional scale. In recent decades,excess commuting has become a major studytopic within the discipline of transport research(Ma & Banister 2006). The excess commute isthat share of the commute flow (in terms ofphysical distance or time distance), that cannotbe attributed to the spatial separation betweenjob locations and residential locations ofemployees, and is thus rooted in the travellersfreedom of choice.

    The main goal of our research is examiningregional variations in the relationship betweenthe length of quasi-daily trips and spatial prox-imity by applying an excess commutingapproach. The paper is structured as follows.First, we provide a summary of the excess com-muting literature and extend the concept tonon-professional travel. Second, we develop amethodology to define theoretical minimumnon-professional trip lengths and to distinguishspatial categories within reported trip lengths.Subsequently, results are obtained by compar-ing the reported distance travelled with thetheoretical minimum distance travelled, withineach spatial category. Finally, we draw conclu-sions from our findings and derive recom-mendations for sustainable spatial planningpractice.

    EXCESS COMMUTING ANDEXCESS TRAVEL

    The concept of wasteful commuting or excesscommuting was first introduced by Hamilton(1982). Hamilton defined excess commuting as

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 21

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • the difference between the actual commutingdistance and the theoretical minimum com-muting distance, suggested by the spatial struc-ture of the considered city. The attentionpaid by Hamilton to minimised commutingdistances stems from the successive oil crisesof 1973 and 19791980, when the availability,and in particular, the affordability of fossil oilproducts was at stake. Daily trips over largedistances were suddenly considered problem-atic, because of their particularly high energyconsumption and costs.

    As transport research progressed, theconcept of excess commuting was extendedand applied in different ways. The line ofinquiry that was started by White (1988) com-pares the spatial structure of different cities onthe basis of the minimum required commute; amethod that was later expanded with the ideaof the maximum possible commute (Horner2002). Both concepts are measurable pro-perties of spatial structure, which can beapplied not only to compare the morphologyof cities, but also to examine time series andthus measure suburbanisation and evolu-tions in commuting behaviour (Horner 2007;Boussauw et al. 2011b). Further, we can distin-guish between the more economically-inspired research direction that uses travel timeas a variable, and the more environmentalapproach that focuses on travel distance (Ma &Banister 2006). In both cases the minimumcommuting distance may be considered as ameasure of proximity, in terms of accessibility(when travel time is studied), or in the senseof spatial proximity (when physical distance isstudied).

    An interesting extension of this is thespatially disaggregated approach where theminimum commuting distance is mapped, con-sidering this variable as a measure of spatialproximity. Again, we can distinguish betweenanalyses that are based on time distance (anapproach introduced by Giuliano & Small1993), and more environmentally-orientedstudies where physical distance is used as ameasure (Niedzielski 2006; Yang & Ferreira2008; Boussauw et al. 2011a). When this type ofresearch is conducted at a regional scale, it maycontribute significantly to the sustainability ofproposed land developments, and to the detec-tion of regions that are vulnerable because of

    their extreme remoteness. This approach is rel-evant in light of peak oil theory. In the course ofhistory, the cost of transport has shown a nearlycontinuous downward trend, with only a rippleat the time of oil crises. Since today transportrelies almost entirely on finite fossil fuels, wesuspect that one day the cost of transport willevolve in the opposite direction, increasing sys-tematically as oil supplies decline. The sudden,albeit temporary, surge in oil prices in 2008seemed to forecast this hard reality. But eventhough there is little point in thinking indoomsday scenarios, it remains a fact that overtime highly car-dependent spatial structuresmay be particularly vulnerable to oil shortages(Dodson & Sipe 2008).

    To date, the excess commuting researchframework (hereafter extended to excesstravel), has to our knowledge only been appliedon the study of the home-to-work commute.There is no doubt about the primordial econ-omic importance of the commute, which rep-resents a significant proportion of the numberof car kilometres travelled. In Flanders, thehome-to-work commute represents 18.6 percent of trips. Yet, the average commuting triplength amounts to 19.0 kilometres, which ismuch higher than the average trip length of12.5 kilometres (for all purposes combined;Zwerts & Nuyts 2004). Moreover, we know thatcommuting trips are much less price elastic andthus more inert than other trips. All these argu-ments emphasise the importance of studyingcommuting behaviour.

    In contrast, in the Western world the share ofcommuter traffic in overall mobility is decreas-ing. Leisure travel, and by extension: touristtrips are on the rise. In essence, this evolutionoriginates from the ever growing prosperityand the improved accessibility of high speedtravel modes for an increasingly larger share ofthe population. Even if the penetration of theprivate car and the coverage of the motorwaynetwork had reached its structural limits, wewould still continue to use more and moreaircraft and high speed trains for recreationaland tourist trips. In the continuum of Handyet al. (2005), these trips are situated near thechoice end, and are much less emerging bynecessity. The changeable nature of the desti-nations and the consequent difficulties in dataacquisition represent perhaps the real reason

    22 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • why researchers have not yet ventured intothe study of excess leisure travel, and havethus confined themselves to the study of thehome-to-work commute. Similar reasons canbe 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 notentirely escape attention in the excess com-muting discourse. Recently, Horner andOKelly (2007) suggested that the study ofexcess travel for non-professional, but more orless daily trips could become an interestingextension, possibly shedding more light on therelationship between non-professional travelbehaviour and spatial structure. Examples areinnumerable: bringing children to day-care orschool, doing the groceries, or going to sportsor hobby clubs. The study of non-commutingtrips, however, entails considerable method-ological problems. The following problems canbe identified immediately:

    the capacity of many of the mentioned faci-lities is deemed elastic, compared withemployment centres that are characterisedby a relatively constant number of jobs;

    many non-professional trips are made fre-quently, but not daily;

    there are often multiple destinations for onepurpose, as is for example the case for mul-tipurpose shopping trips (Handy & Clifton2001);

    many leisure trips are part of a trip chainthat partly includes commuting, making thedistinction between professional and non-professional travel vague;

    commuting trip lengths are the result of adouble selection procedure, being a com-bination of the preference of an employerfor an employee, and the preference ofthe worker for the job. In contrast, non-commuting trips are based on a singleselection procedure, comprising only thecustomers preference for the visited facility.Consequently, excess commuting rates willby definition be higher, and thus not compa-rable with values found for non-commutingtrips;

    for some destination classes the trade-offbetween the accessibility of the widest pos-sible range of customers and the cost of an

    additional establishment is inherent in thelocation of the facility. This is the case withbranches of large chain stores, which areoften sited based on a facility location model;and

    there is often no area covering sample dataavailable on travel behaviour of individualsand families, so it is not possible to aggregatedata within relatively small traffic analysiszones. Moreover, available data often con-tains only information on reported triplengths, without mentioning the address ofthe visited facilities. In the latter case it is onlypossible to estimate excess travel that is gen-erated by residences (an important origin ofmany trips), in a particular area, and not forexample, by stores or schools (examples ofdestinations of non-professional trips).

    The conventional calculation of theminimum commuting distance involves origin-destination matrix optimisation through tech-niques of linear programming (White 1988).This method assumes that destinations (i.e. joblocations), have a fixed capacity, and do notadapt in case demand changes. In contrast withjobs, however, daily service and facility destina-tions do not fulfil these conditions, an issue thatisalsorecognisedbyFan et al. (2010).Therefore,we need to adapt the definition of excess travelto the context of non-work-related destinations.

    To our knowledge, Fan et al. (2010) are theonly authors who have analysed a case ofexcess travel in non-commuting trips. Theyapproach the concept of excess travel as thedifference between the distance travelled bya household to get through its activity pro-gramme (given the current destinations andtravel frequencies), and the distance thatwould be travelled in order to achieve thesame activity programme if case home loca-tion were optimised in a geometrical sense(meaning that the family would move houseto a location that is more centrally relative totheir activities).

    In contrast, our own approach, which weexplain in detail in the next section, comparesproximity of destinations, considered as aspatial characteristic of the studied location,with the travel behaviour of its residents andusers. We assess the ratio between the actualdistance travelled and the distance that would

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 23

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • be travelled for any trip if an alternative, nearbydestination were chosen. Summarising, thedefinition of excess travel in non-commutingtrips by Fan et al. (2010) is household-oriented,while our definition is location-oriented. Bothapproaches are therefore complementary. Afinal aspect worth mentioning is the compara-bility of results between various analyses. Itis known that the magnitude of the obtainedvalues in excess commuting research are largelysubject to the modifiable areal unit problem(MAUP), making attempts to compare resultsof research in areas with surveys based on adifferent zoning system virtually meaningless(Horner & Murray 2002). This problem is evenexacerbated in the study of non-commutingexcess travel: here, the extent to which a desti-nation category is representative of a particulardestination determines the obtained resultstoo, and will additionally bias comparisons.

    METHODOLOGY

    Our research takes the Flemish region and to some extent the Brussels capital region,together constituting the northern half ofBelgium, as a study area. The study consists ofseveral stages. Given the relatively scarce data itis impossible to use traffic analysis zones asspatial units, which would be in line with thespatially disaggregated study of excess commut-ing. Therefore we first have to define the spatialclasses that should be distinguished. We do thisby mapping the transition between more andless urbanised areas as accurate as possible.

    Then we develop a non-professional equiva-lent to the minimum commuting distance inthe form of a proximity map. This is done bydefining, for each statistical ward (corres-ponding with a neighbourhood), and for eachdefined spatial class, the minimum distancethat should be covered in order to reach allfacilities that are visited by an average Flemishhousehold during a week. Obviously, thismethod implies some simplifications. Thechoice of the number of destination categoriesor travel purposes is limited by the amount ofavailable data. So we need to consider thosefacilities for which data is available as represen-tative for a certain category of destinations.This choice is to a certain extent arbitrary.Furthermore, the activity pattern of a house-

    hold is also determined by the spatial context inwhich this family lives, reducing the assumptionof an average activity pattern that is applicableto every household to a major simplification.However, assuming an average householdtravel pattern, regardless of the location of resi-dence, is a deliberate choice we make in orderto map variations in spatial proximity in anobjective way. So, the average activity patternof a resident of the Flanders region is used asa reference in order to quantify proximity as aspatial characteristic.

    Third, we examine the effective distance trav-elled by households during non-professionaltrips, based on available travel survey data,while distinguishing our predefined spatialclasses and destination categories. Eventuallywe compare reported travel distances with cal-culated minimum distances, resulting in a ratiothat represents a measure for excess travel.We consider variation of spatial proximity andexcess travel as a spatial characteristic indicat-ing how sustainable a certain physical structureis in relation to non-professional trips.

    DETERMINATION OF SPATIAL CLASSES

    Unlike data on commuting, that is based ona census, data for non-professional trips inFlanders are only available in the form of asample Travel Behaviour Survey for Flanders20002001 (Onderzoek VerplaatsingsgedragVlaanderen [OVG]; Zwerts & Nuyts 2004). Thismeans that we cannot make a spatially continu-ous analysis on the basis of a map for the wholestudied region, as opposed to the study ofexcess commuting (Boussauw et al. 2011a).

    However, we want to relate the observedtravel behaviour to different types of spatialstructures. To retain a survey sample that islarge enough for every spatial class, we lookfor a meaningful spatial classification for theFlanders region. We base our argument on theexisting literature. Depending on the point ofview, two major formats exist. Based on empi-rical data (a combination of morphologicalcharacteristics and data on commuting andmigration flows), Luyten and Van Hecke(2007) assign each municipality to one of thefollowing four categories: urban agglomera-tion (agg); suburban (sub); commuter area(comm); and rural (rur; which is a residual

    24 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • category with very limited urban characteris-tics). We will call this the urban region classi-fication. The urban agglomeration consists ofthose municipalities where more than half ofthe population lives in an urban core or in theurban fringe, characterised by a continuouslybuilt-up environment. The suburban area isthe outer zone of the city, characterised by anextensive, rural morphology, combined with anurban functionality. Agglomeration and subur-ban area together compose the urban region.The commuter area is attached to the urbanregion and relies on this urban region for animportant part of its employment. The demar-cation of the different classes follows themunicipal borders, causing some loss of accu-racy. An overview is represented in Figure 1.

    The Spatial Structure Plan for Flanders(Ruimtelijk Structuurplan Vlaanderen [RSV];Ministry of the Flemish Community 1997/2004) offers a second classification, which ismuch more policy-oriented. This means thatthis format not only takes into account thecurrent situation, but incorporates also a visionfor future development which is adopted by theFlemish government. The main direction ofdevelopment, favoured by RSV, is indicated bythe dichotomy between urban areas and outly-

    ing areas. Urban areas are those areas thatshould receive most of the additional housingand businesses, and are very accurately delin-eated (on the ward level). The selection makesa distinction between metropolitan areas (ma;agglomerations with more than 300,000 in-habitants: Antwerp and Ghent, and the part ofthe Brussels agglomeration that is located inFlanders), regional urban areas (rua; between50,000 and 150,000 inhabitants) and smallurban areas. Within the latter category a dis-tinction is made between structure supportingsmall urban areas (ssua; being relatively impor-tant attraction and development poles) andsmall urban areas at provincial level (psua;being a development pole of minor impor-tance). The demarcation of urban areas isbased on a consultation process and is consoli-dated on the basis of cadastral boundaries. Fora number of urban areas this demarcationprocess is still ongoing. We used a provisionaldefinition, translated to the ward level. For sim-plicity, we consider everything that is not withinthe definition of any urban area as outlyingarea (oa). In the outlying area we can stilldistinguish selected residential nuclei (noa),generally corresponding to villages. In totalwe thus, distinguish six categories within the

    Source : Luyten and Van Hecke (2007)

    Figure 1. Spatial classification according to the urban region approach.

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 25

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • framework offered by RSV. A cartographicoverview is shown in Figure 2.

    DEVELOPING A PROXIMITY MAP

    Method and selection of destinations In a firstphase of our research we develop a methodto quantify the proximity of non-professionalquasi-daily destinations. We use the ward as aspatial unit, considering the centre of gravity(centroid) of the ward as a starting point tocalculate the network distance to the closestappropriate facility. The 9,708 wards whichcover Flanders and Brussels are thereforeregarded as residential locations. We usenetwork data consisting of the seven highestcategories of the Streetnet skeleton file, whichcontains almost all passable connectingroads.

    We select 18 types of facilities for which alocation dataset is available. Each facility type isjudiciously assigned to one of the travel pur-poses that are applied in the OVG. We onlyused purposes that are non-professional (workand business visits are thus excluded), and havea destination for which alternative locationscan be found (so the purposes walking/driving

    around and visiting someone were not consid-ered). The remaining purposes are: shopping(SHP); education (EDU); picking up/takingsomething/someone (PCK); leisure/sports/culture (LSC); and services, including visiting adoctor or a bank (SRV).

    The various selected facility types, datasources and links with OVG purposes can befound in Table 1. Shopping is represented bythree classes of supermarkets and some morespecialised types of shops. For the inter-pretation of education, the higher grades ofsecondary education and higher education (ingeneral: education for students over 14 yearsold), were not taken into account, since weconsider these facilities as too specialisedand thus, rather being part of the commute.Leisure/sports/culture is represented by cafs,restaurants, sports centres and cinemas, whilethe category services is represented by doctorsand banks. We construct picking up/takingsomething/someone by a combination ofeducation and leisure/sports/culture, supple-mented with nursery. This is of course, only anapproximation, where we assume that mainlychildren are taken to and collected from theiractivities.

    Source : (Ministry of the Flemish Community, 1997/2004)

    Figure 2. Spatial classification according to the RSV approach.

    26 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • With regard to the quality of the data wemention that the data retrieved from GoogleMaps (2009) are based on commercial informa-tion which is less complete than the other datasets that were used (Federal Public Service ofEconomy 2009; Ministry of Education 2009;Child & Family 2009; Cinebel 2009), whichclaim to be exhaustive. The location data fromGoogle Maps include geographic co-ordinates.The other data sets used consist of address liststhat were geocoded with the help of Yahoo!(2009). Finally we calculated the network dis-tance between each wards centroid and thenearest location within each selected type offacility using Dijkstras shortest path algorithm(implemented in the Closest Facility Tool ofArcGIS Network Analyst, ESRI, Redlands, CA)as follows:

    T i f T TwfO wf iO wfO,min , ,min: (1)

    in which: Tmin = minimum trip length; w = ward;f = type of facility; i = any possible destinationbelonging to type of facility f; and O = indicatesthat the spatial unit is always considered as thebase (origin) of the trip.

    Weighting The closest facility calculationprovides a proximity map for each type of

    facility, showing the accessibility, in terms ofphysical distance, from every considered ward.However, we want to limit the number of mapsto the five mentioned OVG purposes, and ulti-mately we want one summary map. We calcu-late spatial proximity per spatial class and pertravel purpose as follows:

    TT

    nsfO

    wfO

    w

    n

    ,min

    ,min

    ==

    1

    (2)

    in which: s = spatial class; and n = number ofwards in spatial class.

    To join the proximity of different purposesinto one map it is necessary to assign weights tothe various facility types:

    T a TwHO ff

    m

    wfO

    ,min ,min=

    =

    1

    (3)

    in which: H = average weekly haul; af =weight bytype of facility; and m = total number of facilitytypes.

    The weighting is determined by the weeklyvisit frequency to the respective facilities by anaverage Flemish household. As a starting point

    Table 1. Selection of facilities and purposes.

    Type of facility N Purpose OVG

    Bakera 3747

    ShoppingSupermarket class 1b (hypermarket) 54Supermarket class 2b (supermarket) 1484Supermarket class 3b (superette) 869Clothes shopb 660Do-it-yourself shopb 199Household appliancesb (electrical) 180Kindergartenc 2913

    Education and picking up/takingsomething/someone

    Primary schoolc 2861Middle schoolc (1st grade high school) 681Adult educationc 111Nurseryd 2844 Picking up/taking something/someoneCaf/bara 4746

    Leisure/sports/cultureRestauranta 6907

    Sports centrea 1581Cinema e 49Doctora 9713 ServicesBanka 3391

    Sources : a Google Maps (2009); b Federal Public Service of Economy (2009); c Ministry of Education (2009);d Child & Family (2009); e Cinebel (2009).

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 27

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • we take the number of trips per household perpurpose, as reported in the OVG. Overall anaverage household generates 42.95 trips perweek, of which 23.26 meet our criteria. Basedon a number of other data (demographic sta-tistics and market research), we estimate theaverage visit frequency. Visit frequencies ofsome destinations are extrapolated to fit inwith the OVG data (visit frequency per OVGpurpose). This means that some facility typesare considered as representative for similardestinations: for example, clothes shops,do-it-yourself and household appliance shopstogether are considered being representativefor non-food specialist shops. Estimated visitfrequencies are shown in Table 2. A trip is seenas a single move, which on average should be

    partially attributed to a trip chain. Based onOVG, we expect that visiting a facility generateson average 1.68 trips, since often more thanone facility is visited within one trip chain.Picking up/taking something/someone in thetable was reduced to the facility type nursery.For comparison with reported trip lengthsthis purpose is extended pro rata with the pur-poses education and leisure/sports/culture(see Table 2).

    Mapped proximity per spatial class Bymapping the result of Equation (3) for everyward, we get a weighted proximity map(Figure 3). This map provides an overview ofthe spatial variation in the minimum distance,expressed in kilometres that should at least becovered by an average Flemish household tocomplete its weekly programme.

    To compare these calculated minimum dis-tances with the distances reported in the OVG,we calculate the average values for each spatialclass: both the values per purpose (Figures 4and 5), and the weighted values (Figure 6).Note that the Brussels capital region wasomitted in these tables in order to match thestudy area of OVG.

    In general, shorter minimum trip lengthsseem to be associated with higher degreesof urbanisation, even if not all distinguishedclasses are useful in describing this phenom-enon (e.g. the distinction between the classessub and comm is not expressed in our find-ings). An ANOVA test, applied to both spatialclassifications, indicates that minimum dis-tances significantly differ among spatial classes(significance level of 0.01), and thus confirmsthe observed general trend.

    As expected agglomerations and metro-politan areas score points in terms of spatialproximity. The differences between suburbanand commuter areas are minimal, as well as thedifferences between the structure support-ing small urban areas and small urban areasat provincial level. The rural area and outly-ing area (including the residential nuclei inthe outlying area), score poorly, as was alsoexpected. In particular the spatial classificationaccording to RSV shows a systematic increaseof the minimum distances to be covered whenwe move from a more urbanised to a lessurbanised area.

    Table 2. Estimated weekly visit frequency by type of facility,per household.

    Purpose OVG Number oftrips/

    household-week

    Shopping 8.99Bakery 2.11Supermarket class 1 (hypermarket) 0.69Supermarket class 2 (supermarket) 3.26Supermarket class 3 (superette) 0.42Clothes shop 0.84Do-it-yourself shop 0.84Household appliances (electrical) 0.83

    Education (without highersecondary and highereducation)

    3.09

    Kindergarten 0.68Primary school 1.70Middle school (1st grade high

    school)0.52

    Adult education 0.19Picking up/taking something/

    someone (limited to nursery)0.34

    Nursery 0.34Leisure/sports/culture 9.00

    Caf/bar 1.84Restaurant 5.95Sports centre 1.00Cinema 0.21

    Services (e.g. doctor, bank) 1.85Doctor 0.62Bank 1.23

    Total 23.26

    28 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • REPORTED TRIP LENGTHS

    Data Data on the effective length of tripsmade by the inhabitants of the respectivespatial classes are obtained from the OVG. Thissurvey reports on travel behaviour of a sampleof 3,028 households over two consecutive days.The sample does not contain households fromBrussels (which is administratively not part ofthe Flemish region) and Ghent (for which aseparate survey was conducted). For our studywe added a random selection of data fromGhent to the Flemish data.

    We selected only those trips originating fromor ending at the residence of the respondent,with a destination or origin corresponding toone of the five selected purposes. Thus, we donot only consider journeys from home to thefacility and back, but also parts of trip chainsbetween the house and the facility. For eachtrip we know the reported distance travelled,the ward where the respondent resides, andthus the spatial class in which the residence islocated. However, we do not know the locationof the visited facility. This trip selection methodis a deliberate choice, aiming to retain a

    Figure 3. Weighted proximity map for Flanders and Brussels (shortest average weekly haul per household, for selectedfacilities).

    Figure 4. Estimated minimum distance by purpose (km)(urban region classification).

    Figure 5. Estimated minimum distance by purpose (km)(RSV classification).

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 29

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • maximum amount of useful information fromthe available data. Nevertheless, this selectionentails some specific biases. Trips of respon-dents making complex trip chains may beunder-represented, while the distance travelledin retained parts of trip chains may be lessrepresentative. Furthermore, this methodassumes that the most efficient trips have theirdestination close to home. However, in the casewhere the respondent is a commuter, a destina-tion that is chosen nearby the job location maylead to a very efficient trip. In the latter caseit should not be excluded that some of thereported excess travel, for example, in a shop-ping trip, is in fact due to the fact that therespondent works far from home. It is impor-tant to keep in mind these possible biases wheninterpreting the results of the analyses.

    Since we are looking for quasi-daily travelbehaviour, trips that cover extremely large dis-tances are considered as outliers. We eliminatedoutliers per purpose, while setting the thresholdat three standard deviations above the meantrip length. Table 3 shows the remainingnumber of observations after selection.

    MethodTolinkuptheminimumdistancetobecovered and the reported travelled distances(Witlox 2007), we follow two different approa-ches, in parallel. First we calculate the averagereported distance per purpose, per spatial class:

    TT

    qsp obsO

    sp rO

    r

    q

    ,

    ,

    ==

    1 (4)

    in which: Tobs = average reported distance perpurpose; Tr = reported length of trip r; p =purpose; and q = total number of reported tripsbased in spatial class s and with purpose p.

    In a next phase we compare this with theminimum distance to be covered, according toEquation (2).

    This procedure gives an indication of theinfluence of the degree of proximity of acertain type of facility on the actual distancetravelled to reach a similar facility. However,the information obtained in this way is notsufficient if we want to understand the rela-tionship between sustainability of travel pat-terns and spatial structure. In this second case,information about trip frequencies is alsoimportant, since it is conceivable that peoplewho make relatively short trips will compen-sate their benefit in terms of time and costs

    Figure 6. Estimated shortest average weekly haul per household (km).

    Table 3. Summary of the retained trips and correspondingOVG purposes.

    SHP EDU PCK LSC SRV

    agg 1,977 777 936 1,377 413sub 816 427 530 483 166comm 1,273 623 669 794 308

    rur 1,632 835 904 1,141 352ma 941 365 451 636 203rua 850 334 390 556 188ssua 411 171 158 222 96psua 281 119 143 202 61noa 1,990 1,022 1,104 1,291 429oa 1,225 651 793 888 262

    30 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • by making more trips, cancelling out gainsin fuel consumption (as an indicator ofsustainability).

    Therefore we calculate the average distancetravelled by a household during one week byadding up the distances covered by all selectedreported trips per spatial class, and extrapolatethis sum to a time frame of one week:

    TT

    tsH obsO

    sh rO

    r

    q

    h

    t

    ,

    ,

    .= ==

    3 5 11(5)

    in which: h = individual household; t = totalnumber of households in spatial class s andincluded in the survey (factor 3.5 extrapolatingthe two-day survey to a time frame of 7 days).

    In the next step we will compare this distancetravelled with the weighted average minimumdistance to be covered by an average householdto satisfy its needs (Equation 3).

    Results For the observed (reported) valueswe also calculate averages and display these byspatial class. Again, the ANOVA test indicatesthat significant differences between spatialclasses exist (p = 0.00, except for leisure trips byRSV class, where p = 0.03). The averages foreach purpose are shown in Figures 7 and 8,while the total observed distances (extra-polated to a full week) are shown in Figure 9.

    Figures 7 and 8 give a rather surprisingpicture. The differences between the spatialclasses are much smaller than what we mightexpect based on the major differences inspatial proximity. For most purposes, the least

    urbanised classes do not necessarily provide thegreatest distances travelled: the largest triplengths are rather recorded in the suburbanand commuter areas. Also the minor differ-ences between the metropolitan and regionalurban areas stand out. Despite the result of theANOVA test, the often wide confidence inter-vals also suggest that the link between observedtrip length and spatial class is rather weak. Pair-wise t-tests indeed show non-significant distinc-tions between some pairs of classes, confirmingthe picture provided by Figures 7 and 8.

    The results that are shown in Figure 9 takeinto account the spatial differences in trip fre-quency. Although, while assessing the results,we must not forget that our selection methodtakes no account of trips that do not haveeither their origin or their destination at therespondents home. In environments thatwould be associated with trip chains that aremore complex than average, this could leadto an underestimation of the number of tripsper household. Yet, the literature is not un-animous on this issue. According to Krizek(2003), a high degree of urbanisation is asso-ciated with more, but less complex, trip chains,while Maat and Timmermans (2006) found ahigher journey complexity in more urbanisedareas. Moreover, complex trip chains areusually very efficient journeys, conducting aseries of activities within a minimum journeydistance.

    When we consider the urban regions classi-fication, it appears that agglomerations, asexpected, yield the shortest average weekly haul.The commuter area thus not the rural area yields the longest weekly haul. When examining

    Figure 7. Reported trip length by purpose (km) (urbanregion classification).

    Figure 8. Reported trip length by purpose (km) (RSVclassification).

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 31

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • the RSV classification, metropolitan areas con-stitute theshortestweeklyhaul.Theweeklyhaulsare much longer in the regional urban areas, butstill shorter in comparison with the small urbanareas (the structure supporting small urbanareas in particular). In the outlying area, werecord the longest weekly haul.

    These relations are quite consistent withBritish research, recording the shortest traveldistances in the major British cities (more than250,000 inhabitants), except London. Smalltowns and rural areas score poorly (Banister1999). Our study adds a new element: com-muter areas which fit morphologically with therural area but are still within the sphere ofinfluence of the agglomeration are scoringworse than the more remote real rural area.

    Parts of the results are probably explainedby the utilisation of the available choice poten-tial due to spatial proximity. This is more oftenthe case in metropolitan and regional urbanareas as they are both functionally and mor-phologically more urbanised than the rest ofFlanders. At a lower geographical scale thestructure supporting small urban areas gener-ate more mileage than the small urban areas atprovincial level. However the nature of theformer class is more urban than the second.Perhaps the structure supporting small urbanareas are functionally more focused on thelarger cities.

    EXCESS TRAVEL

    We find that the influence of spatial proximitydoes not work out in the same way for each

    spatial class. We examine this phenomenonin analogy with excess commuting researchmethods. We define excess travel as the differ-ence between the minimum distance thatmust be covered to visit the desired type offacility (e.g. the nearest supermarket), and theobserved distance covered. The observed triplength is always larger than the minimum triplength because of non-spatial factors that aredetermined by for example, personal prefer-ences, transport cost, price differences betweensimilar facilities, or the organisation of tripchains. We choose to express this difference asthe ratio between the minimum distance to becovered and the observed distance travelled.This ratio is called the excess rate:

    ETT

    spO sp obs

    O

    spO

    =,

    ,min(6)

    ETT

    sHO sH obs

    O

    sHO

    =,

    ,min(7)

    in which: Ep = excess rate by purpose; and EH =excess rate for an average weekly haul.

    As shown by Equations (6) and (7), we calcu-late excess rates in two ways. First we determineper purpose and for each spatial class the rela-tionship between the reported average distanceof a trip with the considered purpose (as shownin Figures 7 and 8), and the minimum distanceto be covered to reach a similar destination(as shown in Figures 4 and 5). This excess rate bypurpose is shown in Figures 10 and 11. Second,

    Figure 9. Reported weekly haul length per household (km).

    32 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • we determine the ratio between the totalreported distance travelled, extrapolated to afull week (shown in Figure 9), and the weightedminimum weekly haul (shown in Figure 6). Thisweekly haul excess rate is shown in Figure 12.

    In Figures 10 and 11 we note for most pur-poses a systematic downward trend of the excessrate when we watch the various spatial classes inorder of decreasing urban nature. In short, thismeans that a high degree of spatial proximity isonly partially reflected in short trip lengths,because the increased choice of possible desti-nations generates comparatively long journeys.In metropolitan areas the average householdgoes shopping almost three times further fromhome than strictly necessary, while in the outly-ing area this rate amounts to one and a halfonly. Thus, a higher degree of spatial proximitycreates greater choice, compensating for asignificant proportion of the potential gains

    (in terms of external costs caused by traffic).Noteworthy is that the differences in excesstravel between the spatial classes are rathersmall. This is in contrast to what was foundpreviously in the case of excess commuting(Boussauw et al. 2011a), with very high values inurban areas, compared to very low values in ruralareas. Regarding differences between purposes,we notice a very high degree of excess travel inleisure trips, where strong personal preferencesplay a part. This applies to some extent also fortrips to services, although the low visit frequencyand the overall high degree of spatial proximity(there is a doctor in every street, say), play theirrole in the obtained excess rate.

    When assessing the excess rate of a combinedweekly haul, then the downward trend is lessevident. Following the spatial classificationaccording to the RSV, regional and small urbanareas report the highest values of excess travel.

    Figure 10. Excess rate by purpose (urban region classifica-tion ratio).

    Figure 11. Excess rate by purpose (RSV classificationratio).

    Figure 12. Weekly haul excess rate ratio.

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 33

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • The rates are much lower in the metropolitanareas which are mainly explained by a largershare of chained trips. Although this puts thelow excess rate of metropolitan residentssomewhat into perspective, it also means thatthe activity pattern of this group is alreadyvery efficient. A household in a regional urbanarea covers a weekly distance that is more than12 times longer than the minimum distancerequired by our model. By contrast, a house-hold living in the outlying area covers only sixtimes the minimum required distance.

    We can interpret the excess rate as a measurethat indicates to what extent a travel patterncan be made more efficient, given the spatialcontext. In this case gaining efficiency meansshortening travel distances by choosing similardestinations closer to home, within the existingspatial configuration of housing and facilities.Such an adjustment of households travel pat-terns may happen as transport would becomemore expensive, for example, by a severe con-gestion policy or a stringent environmentalpolicy, or by energy scarcity. In the outlyingareas, distances are relatively great, and theexcess rate is low. This means that those areasare most vulnerable to a price increase in trans-port. In regional urban and small urban areas,the distances are not only smaller, but thereis also more margin, leaving the possibilityto choose destinations closer to home. In themetropolitan areas, non-professional tripsseem comparatively efficient, apart from beingshort anyway.

    POSSIBLE BIASES IN THE RESULTS

    The discussion above assesses the number ofkilometres travelled per household. However,there are substantial differences betweenaverage households in the various studiedspatial classes. It is generally assumed thathouseholds in urban areas are relatively small,while household income peaks in the suburbanareas near large cities. These factors may play arole in the sustainability of travel behaviour,even though the precise effect is often unclear.

    A larger number of family members may leadto more kilometres travelled per household.Yet, within these families carpooling occursmore often, while children travel only a few

    kilometres independently. Thus, calculatedper person, larger households are expected toproduce less kilometres. When examining theinfluence of spatial structure, calculation of thenumber of kilometres travelled can be justifiedboth per household and per person, albeitfrom different viewpoints.

    Household income plays a role too. At themacroeconomic level, there is a linear relation-ship between income and the number ofkilometres travelled (Schafer & Victor 1997).If there were significant income differencesbetween the various spatial classes, it wouldmake sense to control for this variable. Deter-mining income, however, raises additionalmethodological problems. It is for example,possible that the effect of a higher income in anurban environment is primarily reflected inincreased tourist travel, and not in longer dailyjourneys (Holden & Norland 2005). Moreover,we are also downplaying car ownership, anintermediary variable that is influenced bothby income, by the surroundings (supply ofalternative transport means and parking),and by household size (Van Acker & Witlox2010).

    These assumptions add much more com-plexity to the study of the role of spatial struc-ture in the sustainability of travel behaviour.Should we measure the distance travelled perhousehold or per person? Is it useful to takeincome into account, and if so, how do wetackle this issue? To avoid oversimplification,we did not incorporate these variables in a sta-tistical analysis. However, below we shed morelight on the possible role of spatial structure initself in relation to the mentioned issues byproviding a few basic figures.

    Household size Some Flemish policy plansargue that the average family size is smaller inurban areas than in suburban and rural areas,and consider this phenomenon as a socialproblem (Boudry et al. 2003). It is also assumedthat the phenomenon of shrinking householdsize occurs more rapidly in the urban areas,increasing the identified problem. Champion(2001) paints a more balanced picture. Smallhouseholds, particularly one-person house-holds, are only rarely based on the stereotypicalyoung career maker with a very urban lifestyle.Young singles stay continuously longer living

    34 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • in the parental home, while more one-personhouseholds than before are the result of adivorce, and are in many cases located in asuburban area. In contrast, in Western citycores, especially the immigrant population, iskeeping average family size at a relatively highlevel (see Table 4).

    In urban agglomerations and metropolitanareas households are actually smaller thanaverage. Yet, an exploratory linear regression,trying to explain weekly distance travelled byvariation in family size does not yield any sig-nificant results. When we calculate the averagenumber of kilometres travelled per person,rather than per household, then differencesbetween spatial classes (as shown in Figure 9)are somewhat smaller. When using the spatialclassification according to RSV, structure sup-porting small urban areas stand out more,making travel patterns of the inhabitants of thisclass now appear the least sustainable. The out-lying areas and small urban areas at provinciallevel follow shortly. The agglomerations andmetropolitan areas still score best.

    Income By way of illustration we examine theaverage household income, based on the assess-ment forms of direct taxation for the year 2006.The available data are aggregated for eachmunicipality. This level of aggregation allows usto regroup the data according to the urbanregion classification, but not to the RSV classi-fication (Table 5). It is important to keep inmind, that here too the Brussels Capital Regionis not included in the analysis.

    There seems to exist a slightly downwardtrend when we move from more urbanised intoless urbanised areas. This is particularly thecase when we would take into account house-

    hold size (income per person). Yet, also in thiscase an exploratory linear regression doesnot yield any significant effect of the level ofincome per household or per person on theweekly number of kilometres travelled. We con-clude therefore that macro-economic theory,arguing that higher income results in morekilometres, does not apply at the regional scaleof our study area. This finding provides an addi-tional argument for the proposition that spatialstructure indeed plays an important role in thegenesis of travel behaviour.

    CONCLUSIONS

    The excess commuting research frameworkproves to be very useful in examining the rela-tionship between non-professional trips andspatial structure. By mapping the minimum dis-tance that an average household needs to coverin order to complete its weekly programme, weget an idea of the variation in spatial proximitybetween housing and facilities. This combinedminimum weekly haul varies from 5 km to392 km, depending on the residence location.This wide range of spatial proximity classesindicates that the distinction between moreand less urbanisation is still a reality andremains important in terms of mobility. Theproximity map we obtained in this way(Figure 3) can be used as a guidance for newdevelopments. Additional densification ofareas where the degree of spatial proximityis already high, or areas that are immediatelyadjacent to these, make an excessive increase ofnewly generated traffic the least likely. In areaswith a relatively low degree of spatial proximity,the situation could be improved by planninga better functional mix for the future. Yet,additional housing in areas characterised by alow degree of spatial proximity will generate

    Table 4. Average household size per spatial class inFlanders (2006).

    Class (urban regions) N Class (RSV) N

    agg 2.19 ma 2.13sub 2.55 rua 2.23comm 2.44 ssua 2.24rur 2.48 psua 2.33

    noa 2.50oa 2.61

    Table 5. Average household income by spatial class inFlanders (2006).

    Class (urban regions)

    agg 28,448sub 28,950comm 26,778rur 24,862

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 35

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • more traffic. These findings are in line withwhat Banister (1999, p. 318) suggests: Newdevelopment should be of a substantial size andlocated near to (or within) existing urban areasso that critical size thresholds can be achieved.

    As stated in the introduction, spatial proxim-ity is only one aspect of the overall picture. Thedegree to which choice behaviour is driven byspatial structure is equally important. Figure 9shows that the relationship between spatialproximity and the number of kilometrestravelled is not linear. Residents of agglome-rations, metropolitan and regional urban areastravel over relatively short distances, but theinhabitants of the suburban and commuterareas and small urban areas appear to have aless sustainable travel pattern than is suggestedby the rather urbanised spatial structures inwhich these people live. What is also surprisingis that these variations cannot be explained bydifferences in family size or income.

    From research on spatially disaggregatedexcess commuting we know that less urbanisedareas are characterised by a higher minimumcommuting distance along with a lower excessrate. In other words, residents of rural areas goto work further from home than urban resi-dents, but they opt more often for the closestjob they can find than city-dwellers do (Bous-sauw et al. 2011a). By analogy, we expected tofind a similar phenomenon in the study ofnon-professional trips. Based on Figure 12we see that this expectation is only partiallyconfirmed. In particular metropolitan areas,agglomerations and suburban areas are charac-terised by a relatively low excess rate, indicatingthat residents of these areas are still heavilyinfluenced by spatial proximity when choosingtheir non-professional destinations. Possibly,modal choice has to do something with this: inan urban environment more non-professionaltrips are made on foot or by bike, slow transportmodes for which trip length is crucial.

    We conclude that spatial structure anddegree of urbanisation is of great impor-tance to spatial proximity and length of non-professional trips. Particularly in metropolitanareas, but also in regional urban areas or thesuburban areas that are adjacent to both urbanclasses, households look for non-professionalactivities relatively close to home, especiallywhen it is possible to walk or bicycle there in

    a pleasant way. By attributing a role to thisaspect in spatial planning practice, generationof additional traffic can be avoided and thevulnerability of spatial developments to moreexpensive transport (by rising fuel prices, con-gestion problems and congestion policy) canbe reduced.

    Acknowledgements

    This research was funded by the Policy ResearchCentre on Regional Planning and Housing Flanders. We would also like to thank the two anony-mous referees for their useful comments. All remain-ing errors are ours.

    REFERENCES

    Banister, D. (1999), Planning More to Travel Less.Town Planning Review 70, pp. 313338.

    Boudry, L., P. Cabus, E. Corijn, F. De Rynck, C.Kesteloot & A. Loeckx (2003), De Eeuw van deStad Over Stadsrepublieken en Rastersteden. Brussels:Ministry of the Flemish Community.

    Boussauw, K., T. Neutens & F. Witlox (2011a),Minimum Commuting Distance as a Spatial Char-acteristic in a Non-monocentric Urban System:The Case of Flanders. Papers in Regional Science 90,pp. 4765.

    Boussauw, K., T. Neutens & F. Witlox (2011b),Measuring Spatial Separation Processes throughthe Minimum Commute: The Case of Flanders.European Journal of Transport and InfrastructureResearch 11, pp. 4260.

    Boussauw, K. & F. Witlox (2009), Introducinga Commute-energy Performance Index forFlanders. Transportation Research Part A 43, pp.580591.

    Champion, A.G. (2001), A Changing DemographicRegime and Evolving Polycentric Urban Regions:Consequences for the Size, Composition andDistribution of City Populations. Urban Studies 38,pp. 657677.

    Child & Family (2009), Opvangadressen.Available from . Accessed on 24 August 2009.

    Cinebel (2009) Available from . Accessed on 24 August 2009.

    Dodson, J. & N. Sipe (2008), Shocking the Suburbs:Urban Location, Homeownership and Oil

    36 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • Vulnerability in the Australian City. Housing Studies23, pp. 377401.

    Fan, Y., A. Khattak & D. Rodrguez (2010),Household Excess Travel and NeighbourhoodCharacteristics: Associations and Trade-offs. UrbanStudies doi: 10.1177/0042098010369712, pp. 119.

    Federal Public Service of Economy (2009),Planologie van de Handel. Available from. Accessed on 24 August2009.

    Google (2009) Google Maps. Available at . Accessed on 24 August 2009.

    Giuliano, G. & K.A. Small (1993), Is the Journey toWork Explained by Urban Structure? Urban Studies30, pp. 14851500.

    Hamilton, B.W. (1982), Wasteful Commuting. TheJournal of Political Economy 90, pp. 10351053.

    Handy, S.L. & K.J. Clifton (2001), Local Shoppingas a Strategy for Reducing Automobile Travel.Transportation 28, pp. 317346.

    Handy, S., L. Weston & P.L. Mokhtarian (2005),Driving by Choice or Necessity? TransportationResearch Part A 39, pp. 183203.

    Holden,E.&I.T.Norland(2005),ThreeChallengesfor the Compact City as a Sustainable Urban Form:Household Consumption of Energy and Transportin Eight Residential Areas in the Greater OsloRegion. Urban Studies 42, pp. 21452166.

    Horner, M.W. (2002), Extensions to the Concept ofExcess Commuting. Environment and Planning A34, pp. 543566.

    Horner, M.W. (2007), A Multi-scale Analysis ofUrban Form and Commuting Change in a SmallMetropolitan Area (19902000). Annals of RegionalScience 41, pp. 315332.

    Horner, M.W & M.E. OKelly (2007), Is Non-workTravel Excessive? Journal of Transport Geography 15,pp. 411416.

    Horner, M.W. & A.Y. Murray (2002), Excess Com-muting and the Modifiable Areal Unit Problem.Urban Studies 39, pp. 131139.

    Krizek, K.J. (2003), Neighborhood Services, TripPurpose, and Tour-based Travel. Transportation.30, pp. 387410.

    Luyten, S. & Van Hecke, E. (2007), De BelgischeStadsgewesten 2001. Brussels: Statistics Belgium.

    Ma, K.-R. & Banister, D. (2006), Excess Commut-ing: A Critical Review. Transport Reviews 26, pp.749767.

    Maat, K. & H. Timmermans (2006), Influence ofLand Use on Tour Complexity. In: Transportation

    Research Record 1977, pp. 234241. Washington,DC: National Research Council.

    Ministry of Education (2009), Onderwijsaanbodin Vlaanderen. Available at . Accessed on24 August 2009.

    Ministry of the Flemish Community (1997/2004), Ruimtelijk Structuurplan Vlaanderen Gecordineerde Versie. Brussels: Ministry of theFlemish Community.

    Nss, P. (2003), Urban Structures and Travel Behav-iour. Experiences from Empirical Research inNorway and Denmark. European Journal of Transportand Infrastructure Research 3, pp. 155178.

    Nss, P. (2006), Accessibility, Activity Participationand Location of Activities: Exploring the Linksbetween Residential Location and Travel Behav-iour. Urban Studies 43, pp. 627652.

    Nss, P., S.L. Sandberg & P.G. Re (1996), EnergyUse for Transportation in 22 Nordic Towns.Scandinavian Housing & Planning Research 13,pp. 7997.

    Newman, P. & J. Kenworthy (1989), Cities and Auto-mobile Dependence. A Sourcebook. Aldershot: Gower.

    Newman, P. & J. Kenworthy (1999), Sustainabilityand Cities: Overcoming Automobile Dependence. Wash-ington DC: Island Press.

    Niedzielski, M.A. (2006), A Spatially DisaggregatedApproach to Commuting Efficiency. Urban Studies43, pp. 24852502.

    Ory, D.T. & P.L. Mokhtarian (2005), When isGetting There Half the Fun? Modeling the Likingfor Travel. Transportation Research Part A 39, pp.97123.

    Schafer, A. & D. Victor (1997), The Past andFuture of Global Mobility. Scientific American 277,pp. 5861.

    Van Acker, V. & F. Witlox (2010), Car Ownershipas a Mediating Variable in Car Travel BehaviourResearch using a Structural Equation ModellingApproach to Identify its Dual Relationship. Journalof Transport Geography 18, pp. 6574.

    Vilhelmson, B. (1999), Daily Mobility and theUse of Time for Different Activities: The Case ofSweden. GeoJournal 48, pp. 177185.

    White, M.J. (1988), Urban Commuting Journeysare Not Wasteful. Journal of Political Economy 96,pp. 10971110.

    Witlox, F. (2007), Evaluating the Reliability ofReported Distance Data in Urban Travel Behav-iour Analysis. Journal of Transport Geography 15, pp.172183.

    EXCESS TRAVEL IN NON-PROFESSIONAL TRIPS 37

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG

  • Yahoo! (2009) Maps Web Services. Available from. Accessedon 24 August 2009.

    Yang, J. & J. Ferreira (2008), Choices versusChoice Sets: A Commuting Spectrum Method for

    Representing Job-housing Possibilities. Environ-ment and Planning B 35, pp. 364378.

    Zwerts, E. & E. Nuyts (2004), Onderzoek Verp-laatsingsgedrag Vlaanderen 20002001. Brussels-Diepenbeek: Ministry of the Flemish Community.

    38 KOBE BOUSSAUW, VERONIQUE VAN ACKER & FRANK WITLOX

    2011 The AuthorsTijdschrift voor Economische en Sociale Geografie 2011 Royal Dutch Geographical Society KNAG