Cc

Ka

b

a

ARRA

KTPPCI

1

iTtaaeeoZtbdblv1Ken

0d

Accident Analysis and Prevention 45 (2012) 39– 49

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

j ourna l h o mepage: www.elsev ier .com/ locate /aap

ontrolling factors of the parental safety perception on children’s travel modehoice

ristof Nevelsteena,∗, Thérèse Steenberghena, Anton Van Rompaeyb, Liesbeth Uyttersprota

Spatial Application Division Leuven, K.U. Leuven, Celestijnenlaan 200E, 3001 Heverlee, BelgiumGeography Research Group, K.U. Leuven, Department Earth and Environmental Science, Celestijnenlaan 200E, 3001 Heverlee, Belgium

r t i c l e i n f o

rticle history:eceived 5 April 2011eceived in revised form 10 October 2011ccepted 10 November 2011

eywords:raffic safetyerception

a b s t r a c t

The travel mode of children changed significantly over the last 20 years, with a decrease of children trav-elling as pedestrians or cyclists. This study focuses on six to twelve year old children. Parents determineto a large extent the mode choice of children in this age category. Based on the analysis of an extensivesurvey, the research shows that traffic infrastructure has a significant impact on parental decision makingconcerning children’s travel mode choice, by affecting both the real and the perceived traffic safety. Realtraffic safety is quantified in terms of numbers of accidents and road infrastructure. For the perceivedtraffic safety a parental allowance probability is calculated per road type to show that infrastructure

arentshildren

nfrastructure

characteristics influence parental decision making on the children’s mode choice. A binary logistic modelshows that this allowance is determined by age, gender and traffic infrastructure near the child’s home ornear destinations frequently visited by children. Since both real and perceived traffic safety are influencedby infrastructure characteristics, a spatial analysis of parental perception and accident statistics can beused to indicate the locations where infrastructure improvements will be most effective to increase the

lling

number of children trave

. Introduction

The last 20 years, a decline in the number of traffic accidentsnvolving children can be observed in several European countries.he annual statistical report of the European Road Safety Observa-ory shows a reduction of 25% in the number of fatalities in generalnd a 50% reduction in the age group 0–15 years between 1997nd 2006 (ERSO, 2008). This decline seems to indicate safer trafficnvironments. However, according to Roberts, 1993 and Diguiseppit al., 1997 this decline can be attributed to a decline of child auton-my in the transport system. In several European countries e.g.,eiher, 2001 the movements of children are restricted to ‘safe’ loca-ions such as schools, playgrounds and in-house, and the transportetween those safe islands is often supervised by an adult. Chil-ren between 6 and 12 years old are more often transported in theackseat of the car than 20 years ago. Increases up to 40% over the

ast 10 years, even for trips less than a mile, have been reported inarious case studies in Northern America and Europe (Carlin et al.,997; Ewing et al., 2004; Hillmann et al., 1990; Kaesemans, 2002;

arsten et al., 2001; Mcmillan, 2006; O’Brien et al., 2000; Pooleyt al., 2005; Zwerts and Nuyts, 2001). In general the decrease of theumber of child pedestrian or cycling trips is perceived as negative

∗ Corresponding author. Tel.: +32 16 32 97 27; fax: +32 16 32 97 24.E-mail address: Kristof.nevelsteen@sadl.kuleuven.be (K. Nevelsteen).

001-4575/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.oi:10.1016/j.aap.2011.11.007

– safely – as pedestrians or cyclists.© 2011 Elsevier Ltd. All rights reserved.

because (1) it implies more car traffic and (2) autonomous travellingon foot or by bicycle enhances children’s motor system devel-opment, stimulates the development of their social identity andimproves their physical condition (Cooper et al., 2005; Timperioet al., 2004). On top of this, independent travelling raises the chanceof having superior traffic skills at a later age. In the light of thesefindings it is important to increase (autonomous) travel of chil-dren.

Prezza et al. (2006) mention the following reasons for anincreased car dependency of young children (1) increasing carownership, (2) greater complexity in lifestyle, (3) increasing timepressure, and (4) parental concern about children’s safety, bothbecause of traffic and possible abuse by strangers. The first threereasons can be related to ‘travel time to destination’ (Mcmillan,2006; McDonald, 2007; Schlossberg et al., 2006; Black et al., 2001).The parental concern about traffic safety is particularly importantfor children travelling as pedestrians or cyclists (Johansson, 2006;Ewing et al., 2004; Fotel and Thomsen, 2004; Martin and Carlson,2005). These reasons complement each other and overlap to someextent in explaining travel mode choice. The study of parental con-cern about traffic safety can help to better understand children’stravel mode choice in different environments.

Proponents of smart growth and liveable community con-cepts developed a persuasive hypothesis attributing the changein children’s travel mode choice to the urban form of our com-munities and believe that the decrease of trips as a pedestrian

4 alysis

oeAbeobt(tmdfsPfimccvrtcotcibdrdc2

mbmab

ecdatc

‘otaotrmtDpet

sRflf

0 K. Nevelsteen et al. / Accident An

r as a cyclist will be stopped if traffic environments are remod-lled in order to increase the safety for children (Mcmillan, 2005).ccording to Mcmillan (2005) such reasoning is rather naïveecause it is not clear to what extent modifications in the trafficnvironment impact the perceived safety of the parents. She devel-ped a framework which outlines the pathways and interactionsetween infrastructure and children’s travel mode choice. Infras-ructure influences indirectly parental decision-making throughreal and perceived) neighborhood safety, (real and perceived)raffic safety and household transportation options. These are all

oderated by social/cultural norms, parental attitudes and socio-emographics. Parents or caregivers are assumed in Mcmillan’sramework to make the final trip decision. This is supported bytudies of e.g., Fotel and Thomsen (2004), Kerr et al. (2006) andrezza et al. (2006). It is clear that both real and perceived traf-c safety together should be tackled to better understand whateasures can be taken to increase the number and safety of

hildren’s pedestrian and cycle trips. Fyhri and Hjorthol (2009)oncluded that measures aiming to increase walking and cyclingia improved traffic safety will be effective only if parents ‘expe-ience of traffic safety’ is improved. Noland (1995) even foundhat there is evidence that perceived safety improvements in bicy-le transportation have an aggregate elasticity value greater thanne (i.e. a 10% increase in perceived safety results in greaterhan 10% increase in the share of people commuting by bicy-le). So to reverse the trend of increasing car dependency, a shiftn parental perception is desirable. Parental concern is typicallyased on the perception of the potential risk for traffic acci-ents, but this perception does not necessarily correspond to theeal risks. This observation has been subject of many academiciscourses on whether our behavior is guided by real or per-eived risks (Beck, 1992; Giddens, 1999; Luhmann, 1991; Rasborg,002).

In order to understand parent’s perception of the traffic environ-ent and related children’s travel mode decision, two issues must

e addressed: (1) what is the risk perception of the traffic environ-ent by parents? And (2) how accurate is the perceived risk? Such

n accuracy assessment requires both real and perceived risks toe measured in a comparable manner.

Different definitions of risk are used in various disciplines. Inconomics, opportunities whose returns are not guaranteed areommonly described as ‘risks’ (Yates, 1994). In medicine and epi-emiology, risk is commonly defined as ‘the propensity of somedverse outcome’ (e.g. the contraction of a particular disease). Inraffic safety, the definition of risk is similar, and the adverse out-ome is a traffic accident.

Traffic safety statistics typically list occurrences of accidents. Asafe road’ could for example be defined as a road with a low numberf accidents or as a road with a low number of fatal accidents. Addi-ional uncertainty is introduced by the inaccuracy, completenessnd representativeness of traffic safety statistics. Traffic data areften subject to over- or under-representation of certain accidentypes. Especially child related accident statistics are often under-epresented (Adams, 1988; Leonard et al., 1999) mainly because inost cases police reports are the base of national accident statis-

ics. Minor accidents with children are often not officially reported.espite the drawbacks several studies succeeded in pinpointingroblematic locations based on accident statistics (e.g. Steenbergent al., 2010) or found variables that could explain spatial distribu-ion of accidents (e.g. Geurts et al., 2005).

Road accident risk is determined by traffic conditions, and expo-ure (Carlin et al., 1997; Macpherson et al., 1998; Posner et al., 2002;

oberts et al., 1997). Traffic exposure of children cannot be derived

rom accident statistics or traffic data since it depends largely onocal conditions and circumstances. Exposure data can be collectedrom extensive interview campaigns in which the travel paths of

and Prevention 45 (2012) 39– 49

children are reconstructed in detail and confronted with evenlydetailed traffic data.

Perceived risk can be defined as ‘the expectation of an undesir-able outcome’ (i.e. a traffic accident) (Hamed and Al Rousan, 1998).In other words, perceived risk is the estimated chance of beinginvolved in a traffic accident. The assessment of traffic risk per-ception is often subjective because the perception of risk is highlyindividual and depends on past experiences with accidents andpotential rewards of risk-taking (Adams, 1988). Therefore it can beuseful to use ‘balancing behavior’ to measure subjective risk per-ception. In the application to parental risk perception, these arebehavioral measures which address both the concern for children’ssafety and the concern that they recover the individual freedomthey need (Adams, 1988).

Existing research on the relation between children’s travel modeand the traffic environment is relatively limited because of (1) a lackof reliable data (2) the relatively narrow focus on variables typicallystudied within a given discipline (3) a lack of focus on children,walking and bicycling. Therefore the main objectives of this studyare: (1) to collect a data on the travel mode of young children intraffic and (2) to analyse the factors that have an impact on theparental decision making related to the children’s travel mode. Thisis elaborated through a case study.

2. Case study Flanders

The region Flanders in the northern part of Belgium whichis characterized by a high population density and a highly frag-mented urbanized or semi-urbanized landscape was used as a casestudy for this research. Out of 3963 elementary schools (for chil-dren between 6 and 12 years old) in Flanders 150 were selectedrandomly and asked to co-operate by distributing a questionnaireamong parents of their pupils. Teachers distributed the question-naire (one per household) to be filled in by the parents and collectedthem afterwards. Surveys were sent to the researchers in sealedenvelopes. The surveys were then manually digitized. The ques-tionnaire asked parents about their children’s travel mode and theirown perception of the traffic infrastructure in their environment.The questionnaire consisted of two parts.

The first part aimed at the detection of controlling factors ofparental perception of traffic safety with a focus on traffic infras-tructure, such as the role of road types (Table 1 and Table 2).

- Subsection 1 asked general questions about the child (age, gen-der) and household characteristics (brothers or sisters >12 yearsold in the household, number of cars, frequency of car use in thehousehold, child possession of a bike)

- Subsection 2 asked if the parents would allow their child to travelautonomously on eight different road types. They answered with1 = allowed or 0 = not allowed.

- Subsection 3 presented a set of theorems on different issuesconcerning traffic safety and traffic infrastructure where parentsrecorded if they fully agreed = 1, partially agreed = 2, partially dis-agreed = 3 or disagreed = 4 (Table 2).

The second part aimed at mapping children’s travel mode choiceby asking the parents to keep a diary of every trip on Monday,Wednesday and Saturday their child made. For every trip: start-ing point, destination, travel mode, travel distance and level ofguidance had to be recorded.

Descriptive details about the survey are presented in Table 1.

In total 25,000 households were addressed via the 150 selectedschools. 5637 sent a correct, completed first part back, whichresulted in data on 7968 children. In every child’s age group therewere more than 1000 responses with a frequency distribution over

K. Nevelsteen et al. / Accident Analysis and Prevention 45 (2012) 39– 49 41

Table 1Descriptive statistics: survey part 1 and 2.

Survey part 1

Variable Units Value Mean Std. Dev. Min. Max.

Children 79686 year 12877 year 11358 year 12929 year 123010 year 126711 year 135212 year 405

Boys 3904Girls 4064Age Years 8.71 1.87 6 12Households 5637Cars Number/Household 1.59 0.57 0 5Frequency of car use (Days/Week)/Household 5.91 2.04 0 7Older (>12) brothers or sisters in the household No = 0; Yes = 1 0.26 0.44 0 1Children own a bike No = 0; Yes = 1 0.98 0.12 0 1

Survey part 2

Variable Units Value Mean Std. Dev. Min. Max.

ad(srFr

engto

wat

TD

Fully filled in diaries 3040Trips Number/3 days 36833

ge groups equal to the demographic structure of children in Flan-ers. Only the age group of 12 year-olds was underrepresentedN = 405) because half of the children leave the undergraduatechool at the age of 11. The category “girls” was slightly over-epresented in comparison with the demographic structure oflanders. Responses were equally distributed over the study areaesulting in circa 1500–1800 children per Flemish province.

Table 2 presents the distribution of the Likert scale answers par-nts gave on the fifteen theorems (part 1 – subsection 3). When theumbers of the first and third quartile lie close together, a largeroup of parents answered similar. When the numbers are diverse,he answers are more dispersed. The median indicates the directionf the answer (fully agree disagree).

Unfortunately many diaries (part 2) were not correct filled in,

hich results in a dataset of 3040 diaries. The distribution over age

nd gender was similar to that of the first part of the survey, onlyhe amounts were lower.

able 2escriptive statistics: survey part 1 – subsection 3.

First part – subsection 3: Likert scale (0 = No opinion, 1 = Fully agree, 2 = Partially agree,

Theorems Median

Neighborhood is traffic safe 2

Neighborhood is socially safe 1

Our street is safer than most 2

Enough signed crossings 3

In the neighborhoodEnough play and sport facilities 2

In the neighborhoodChildren can move safely on 3Current public roadsIn general safe but some unsafe areas 2

Determine no autonomous trips for childrenLots of traffic at school area 2

Zone 30 km/h is useful 1

Zone 30 km/h should be extended 1

Only in the presence of other children 3

Can my child walk or cycle autonomousWe live in a safe street 2

In current traffic I feel safe 2

Enough is done to create safe traffic 2EnvironmentsTraffic infrastructure is adapted to the needs of my children 3

12.12 3.2 0 23

The classification of road types is based on maximum speed andinfrastructural measures (presence of cycling and pedestrian lanes).The eight different road types used in this study are presented inFig. 1. Pictures are added to present an idea of the appearance ofthe Flemish road network. A cycling lane is defined as a separatestrip where no motor vehicles are allowed (except certain types ofmotorbikes). The separation can be an elevated lane, a barrier or awhite painted dashed line.

Data on objective traffic safety were derived from Belgiannational accident statistics, georeferenced and corrected on theroad network by the regional public works department of Flanders.This road network is attributed with speed limits, vehicle restric-tions and the distance of every road segment. The available accidentdataset was compiled from official police reports, which implies an

underestimation of the total number of accidents (Adams, 1988;Leonard et al., 1999). In this study statistics from 2002 till 2005 wereused because data from before 2002 were not fully georeferenced

3 = Partially disagree, 4 = Disagree)

First quartile (25%) Third quartile (75%)

2 41 21 42 4

1 3

2 4

1 3

1 31 11 22 4

1 42 32 3

2 4

42 K. Nevelsteen et al. / Accident Analysis and Prevention 45 (2012) 39– 49

astruc

asrTw

3

3

wtm

ois(iwt

A

Wabtt

Fig. 1. Road infr

nd data from 2006 onwards were not yet available. Since the studyearched for general perception of the traffic environment it seemseasonable that this perception is generated over several years.herefore it is possible to compare 2002–2006 accident statisticsith 2008 perception data.

. Methods

.1. Children’s characteristics/perceived traffic safety by parents

Based on the trip diaries, general children’s travel mode choiceas studied using descriptive statistics for general quantitative

rends and Chi-square tests for qualitative associations betweenode choice and social variables.According to (Adams, 1988) ‘the proportion of children of vari-

us ages who are allowed autonomous travel on a certain road type’s a ‘balancing behavior measure’ which can be used for the expres-ion of risk perception. The answers of the allowance questionsurvey: part 1 – subsection 2) were reformed into an allowancendex per road type (i.e. the proportion of children) and could in that

ay be used as a balancing behavior measure to express perceivedraffic risk.

i = (˙Ap)Np

(1)

ith Ai the allowance index, Ap the answers parents gave (0 or 1)

nd Np the total number of questionnaires. Eq. (1) results in a num-er between 0 and 1, Ai = 1 in the case of full allowance and Ai = 0 inhe case of no allowance. If 50% of the parents allow autonomousravelling Ai = 0.5. This allowance index was also calculated per age.

ture in Flanders.

3.2. Controlling factors of perceived traffic safety by parents

Several binary logistic models were used to find controlling fac-tors for the parental risk perception of the traffic infrastructureexpressed as allowance per road type. For every road type a logisticmodel was constructed that shows which factors play an importantrole in the decision to allow children on that specific road type. Thepredictors consisted of variables associated to the children and theparents. The children’s variables are both interval and categoricalvariables (age, gender, activity level = total amount of trips per 3days, average distance to destination, older brother or sisters) andwere all coded with natural numbers (see text beneath Table 5).The parental variables are derived from the theorems concerningtraffic infrastructure (survey: part 1 – subsection 3) (Table 2).

In essence the models predict the probability that a parent witha specific answering pattern on the theorems, who is in a certainhousehold situation and has a child that meets certain characteris-tics will allow (answered with 1 in the survey: part 1 – subsection 2)his child on a specific road type. This probability can be calculatedusing Eq. (2).

Pi = exp(ˇ0 + ˇ1xi)

(1 + exp(ˇ0 + ˇ1xi))(2)

With Pi the predicted probability of the event which is coded with1 (allowing the child on that road type) and x1, x2, . . . the predict-ing variables of our model (see Table 5). ˇ0,ˇ1,. . . are the modelscoefficients, listed in Table 5.

K. Nevelsteen et al. / Accident Analysis and Prevention 45 (2012) 39– 49 43

Table 3Travel modes of Flemish children by age and gender; guidance level per mode by age.

Age Boys Girls

6 7 8 9 10 11 12

Pedestrian (%) 8 8 9 9 9 9 8 8 10Cyclist (%) 12 9 12 13 17 23 26 21 16Car passenger (%) 79 82 77 77 72 66 62 70 73Public transport (%) 0 1 1 1 1 1 3 1 1

Guided (all modes) (%) 94 93 91 88 79 73 68Autonomous (all modes) (%) 6 7 9 12 21 27 32

Guided cyclist (%) 86 85 71 59 27 19 15Autonomous cyclist (%) 14 15 29 41 73 81 85

Guided pedestrian (%) 86 77 57 49 31 32 385

S

3

tTb

4

4

iatiii

aiyat(egbatd

btTfowtt(iwtTc

b

Autonomous pedestrian (%) 14 23 43

um of columns percentages equal 100% at every line break.

.3. Real traffic safety

The frequency distributions of children’s traffic injuries (objec-ive traffic safety) over age, gender and road type were analysed.he data express relative numbers of accidents, no traffic risksecause exposure data are not available.

. Results

.1. Children’s characteristics and travel mode choice

General travel behavior data of Flemish children can be foundn the ‘travel survey Flanders’ (Hajnal and Miermans, 1996; Zwertsnd Nuyts, 2001). Our 2008 data from the diaries extend and updatehe travel mode choice data and serve as a contribution to sparsenternationally literature on travel mode of children. The data aren line with trends observed and described in both national andnternational literature.

The data show that the travel mode distribution changes withge (Table 3). The decrease of car use is compensated by an increasen bicycle use at later age. There are as many 6 year olds as 12ear olds who walk, yet younger children walk more under guid-nce than older ones (Table 3 (guided pedestrian)). In general morehan 80% of the young children are escorted when being road userscyclists or pedestrians), while only 15–30% of the older ones arescorted. The guidance level for all modes includes the obviouslyuided trips made as car passenger which explains the low num-ers for autonomous trips. From both the ‘travel survey Flanders’nd our data, it appears that despite the fact that children starto cycle and walk more autonomously at later age, there is still aecrease of children who travel by bicycle or as pedestrian.

There is a significant difference between the travel mode distri-ution for school trips and other trips. Car use is 13% less for schoolrips and especially bicycle use increases when travelling to school.his observation can be explained by the average travel distanceor school trips compared to the average travel distance for leisurer family trips. Only 14% of the school trips are more than 5 kmhile 42% of the leisure trips and 48% of the family trips are more

han 5 km. A Chi-square test between travel mode and travel dis-ance shows a significant dependency between the two variablesP < 0.001). It should be noticed that the questionnaire was heldn November, which means that a part of leisure and family trips

ere carried out after dark. This may be an additional reason forhe difference in travel mode choice between school and other trips.

hese numbers confirm the importance of distance on travel modehoice.

No significant differences in travel mode could be detectedetween girls and boys apart from the age group of older children.

1 69 68 62

Within the age group 6–9 years no differences in travel mode forall trips can be observed (t = 1.88; p > 0.05), within the 10–12 groupthe difference in travel mode for all trips is significant (t = −2.95;p < 0.05) with boys travelling more by bike than girls (+5%), butgirls walking more than boys (+1.8%). For school trips this trend iseven more pronounced with 9% more boys travelling by bike and3% more girls who walk. Similar trends were observed in Zwertset al. (2009) for Flanders, in Fyhri and Hjorthol (2009) for Norwayand in O’Brien et al. (2000) for Great Britain.

Large international differences exist in children’s travel modechoices and consequently the exposure to risk (Carlin et al., 1997;Posner et al., 2002; Roberts et al., 1997). Compared to other coun-tries, Flanders has a cycling culture; especially when consideringschool trips (23% in total and 40% for children in the age group11–12 year cycle to school). Although 20% of all trips are less than1 km, Flemish children do not walk much, 9% for all trips and only13% of the children walk to school which is for 30% of the childrenless than 1 km.

4.2. Parental perception of traffic safety of children

4.2.1. General parental perceptionLarge differences exist between the allowance indices for dif-

ferent ages (Fig. 2) and this for every road type. The differencebetween 6 year olds and 12 year olds is on average 0.6 or 60%for every road type, meaning that 60% more parents allow 12year old children autonomous on a specific road than they allow6 year olds. Roads with a cycling or pedestrian lane are perceivedsafer in comparison to roads where the speed limit is lower butwithout cycling or pedestrian lanes (Fig. 3). This difference inperception between speed and infrastructure measures is morepronounced for pedestrian lanes than for cycling lanes. As expectedthe allowance index both for pedestrians and cyclist decreaseswhen speed limit increases, meaning that children are less allowedon roads with higher speed limits. Table 4 shows that there is adifference in allowance index between cyclists and pedestrians,suggesting that walking is perceived safer than cycling. This dif-ference is larger on roads without a cycling or pedestrian lane andit is particularly applicable for younger children. So the perceptionof walking on a pedestrian lane is perceived safer than cycling is ona cycling lane for all speed limits and especially for younger chil-dren. When there is no pedestrian or cycling lane, the differencebecomes negative for older children and so the safest travel mode

becomes cycling. Boys are significantly more allowed as a cyclist totravel on any road. The difference ranges from 3 to 10% more boysallowed. As pedestrian the allowance is similar between boys andgirls.

44 K. Nevelsteen et al. / Accident Analysis and Prevention 45 (2012) 39– 49

Table 4Difference in allowance index between cycling and walking.

Road Type Age

6 7 8 9 10 11 12

Vehicle restricted with cycling/pedestrian lane (%) 10.4 11.4 11.6 6.2 3.8 1.4 1.8Vehicle restricted (%) 6.6 7.8 8.2 3.8 4.4 −0.2 0Zone 30 km/h with cycling/pedestrian lane (%) 12.2 20.6 19.8 14.4 6.8 3.4 1.8Zone 30 km/h (%) 1.4 3.2 2.8 0 −3.6 −6 −10.4Build up area (50 km/h) with cycling/pedestrian lane (%) 8.6 18 21 18.8 11.8 7.6 4.4build up area (50 km/h) (%) 0.4 0.2 0.4 −2.4 −7 −11.2 −15.8Outside the build-up area (≥70 km/h) with cycling/pedestrian lane (%) 2.4 4.4 7.4 11.6 7.4 6 1Outside the build-up area (≥70 km/h) (%) 0.6 0 0.4 1.4 1.2 −1.2 −3.4

Pedestrian allowance index − Cycling allowance index × 100.Positive numbers mean that walking is perceived safer than cycling.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

6 7 8 9 10 11 12

Allo

wan

ce in

dex

vehicle restricted with cycling/pedestrian lanevehicle restricted

zone 30 km/h with cycling/pedestrian lane

zone 30 km/h

build up area (50 km/h) with cycling/pedestrian lanebuild up area (50 km/h)

outside the build up area (≥70 km/h) with cycling/pedestrian laneoutside the builde up area (≥ 70 km/h)

h cycl

4(

W

Age

Fig. 2. Average allowance index of bot

.3. Controlling factors of perceived traffic safety by parents

binary logistic regression models)

Coefficients from the logistic model are presented in Table 5.hen used in Eq. (2), these coefficients predict the probability

Fig. 3. Allowance inde

ing and walking by age and road type.

that a specific parent will allow a specific child on that specific

road. For every road type a model with the same variables wasconstructed. Not every variable is however significant for everyroad type. From the fifteen theorems (survey part 1 – subsection3) used for the derivation of the variables related to the parents,

x by road type.

K.

Nevelsteen

et al.

/ A

ccident A

nalysis and

Prevention 45 (2012) 39– 49

45

Table 5Binary logistic model coefficients.

Cycling Walking

vr vr cl 30 30 cl bua bua cl obua obua cl vr vrpl 30 30 pl bua buapl obua obuapl

Number of cars 0.133 0.143 0.178 0.188 0.221 0.158 0.016 0.148 0.207 0.142 0.094 0.117 0.087 0.252 -0.029 0.242Car use −0.044 −0.025 −0.061 0.001 −0.043 0.029 −0.035 0.015 −0.065 −0.088 −0.027 −0.128 0.018 −0.088 −0.031 0.026Neighborhood is traffic

safe−0.070 −0.112 −0.190 −0.134 −0.229 −0.173 −0.110 −0.152 −0.058 −0.122 −0.163 −0.166 −0.165 −0.180 −0.078 −0.103

Our street is safer thanmost

−0.011 0.048 −0.016 −0.025 0.083 −0.017 −0.067 0.041 0.003 −0.015 0.030 −0.068 0.030 −0.108 0.009 0.053

Enough signedcrossings in theneighborhood

0.008 0.004 0.007 −0.029 0.008 0.002 0.028 0.013 0.006 −0.003 −0.011 −0.065 0.002 −0.048 0.000 −0.043

Enough play and sportfacilities in theneighborhood

−0.108 −0.039 −0.050 0.016 −0.018 0.015 −0.059 0.009 −0.083 −0.018 −0.032 −0.003 0.038 0.047 −0.062 0.018

Safe but some unsafeareas determine noautonomous trips

−0.005 −0.085 −0.061 −0.051 −0.023 −0.110 0.129 −0.039 −0.052 −0.085 −0.020 −0.096 0.023 −0.097 0.069 −0.037

Lot of traffic at schoolarea

0.210 0.220 0.336 0.293 0.199 0.242 −0.015 0.114 0.217 0.219 0.275 0.297 0.194 0.293 −0.026 0.049

Zone 30 km/h is useful 0.084 −0.018 0.083 −0.022 0.179 0.057 0.288 0.240 0.064 −0.042 0.125 0.018 0.168 0.053 0.263 0.205Zone 30 km/h should

be extended0.014 −0.015 −0.074 −0.103 −0.168 −0.131 0.030 0.031 −0.021 −0.028 −0.171 −0.085 −0.091 −0.108 −0.028 0.004

We live in a safe street −0.021 −0.017 0.006 −0.009 0.015 0.008 0.118 0.099 −0.031 −0.026 −0.075 0.004 −0.034 0.066 0.077 −0.009Traffic infrastructure is

adapted to the needsof my children

0.021 0.044 −0.128 −0.101 −0.184 −0.133 −0.222 −0.156 0.033 0.056 −0.111 0.042 −0.177 −0.073 −0.200 −0.138

Activity level 0.064 0.025 0.039 0.008 0.032 0.009 −0.015 0.010 0.062 0.009 0.023 0.014 0.008 0.018 −0.024 −0.005Average distance to

destination−0.034 −0.074 −0.127 −0.068 −0.129 −0.082 0.087 0.064 −0.018 −0.077 −0.069 −0.046 −0.041 −0.070 0.016 0.034

Gender 0.328 0.412 0.276 0.257 0.199 0.259 0.389 0.115 0.162 0.299 0.157 0.168 0.200 0.026 0.172 −0.028Age 0.563 0.668 0.819 0.813 0.843 0.846 0.785 0.774 0.491 0.682 0.656 0.787 0.679 0.776 0.739 0.633Older brother or sister −0.064 −0.221 −0.157 −0.241 −0.064 −0.186 0.286 −0.085 −0.142 −0.049 −0.242 −0.299 −0.162 −0.190 −0.119 −0.106Constant −5.366 −4.757 −7.583 −6.255 −8.728 −7.222 −10.673 −9.281 −4.373 −4.073 −5.941 −4.611 −7.529 −5.483 −8.972 −6.925

N 3035 3042 3039 3040 3034 3036 3036 3037 3034 3037 3036 3040 3035 3038 3035 3037LL 3369.52 2791.23 2850.99 2856.77 2582.64 2888.96 1416.26 2520.86 3425.96 2509.35 3070.26 2589.65 2581.46 2842.63 1435.36 2968.58�2 763.58 835.69 1240.71 1264.68 1001.10 1314.76 334.74 737.03 620.89 728.63 941.28 1088.23 686.79 1189.02 302.19 689.67Pr(�2) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001Pseudo R2 0.299 0.345 0.435 0.459 0.406 0.469 0.238 0.238 0.251 0.325 0.364 0.429 0.310 0.441 0.216 0.292

Age: 6–12; gender: 1 = boy 0 = girl; average distance: 1 = <1 km, 2 = 1–2 km, 3 = 2–5 km, 4 = >5 km; older brothers or sisters: 1 = yes, 0 = no; number of cars: 0–3 or more cars; car use: 0 = never, 1 = once a week, 2 = 1–2 times a week,3 = 3–4 times a week, 4 = 5–6 times a week, 5 = every day.Bold = significant at 99%CI (p < 0.01).vr = vehicle restricted; 30 = 30 km/h; bua = build up area; obua = outside the build area; pl = pedestrian lane; cl = cycling lane.

4 alysis

olTamtoataaooitdrt

arnBiaflosaodtdsbToem

l(vltlr

4

aictgapetg(oi

6 K. Nevelsteen et al. / Accident An

nly ten were significant for at least one road type. Only these areisted in Table 5 together with the variables related to the children.he variables that contribute significantly for a road type, whichre variables significantly increasing the predicting power of theodel, are indicated in bold (99%CI (p < 0.01)). Variables related to

he children with a significant contribution to the regression modelf almost every road type and both for cycling and walking are ‘age’nd ‘gender’. ‘Gender’ and ‘age’ were also found as significant con-ributors by McDonald (2008). Older boys are more likely to travelutonomous on every road type both as a cyclist or a pedestrian. Asn example consider the case of a 6-year-old boy versus a 6-year-ld girl who travel by bike. The boy has 3.4% chance of being allowedn a 50 km/h road without a cycling lane and the girl 2.7%. Keep-ng the other factors equal, an 11-year-old boy has 71% chance onhat same road while an 11-year-old girl has only 64% chance. Theifference between boys and girls enlarges with age, enlarges onoads with no cycling or pedestrian lane and is smaller for walkinghan for cycling.

‘Lot of traffic at school area’ and ‘neighborhood is traffic safe’re significant predictors in the developed models for almost everyoad type. Although the coefficients are positive for the first andegative for the second, both variables express a similar perception.oth the area of departure and destination has to be traffic safe to

ncrease the allowance to travel autonomous on a certain road types a cyclist or a pedestrian. For example: a 10-year-old boy of aamily who lives in a safe neighborhood and who’s school area hasittle traffic has a chance of being allowed autonomous as a cyclistn a 50 km/h road without a cycling of 55% while the 10-year-oldon of family who live in a neighborhood which is not traffic safend where the school area has a lot of traffic will only have a chancef being allowed of 25%. Allowing the child on a certain road typeepends for a great part on the perception of ‘near-by areas’. Whenhose areas are perceived dangerous the probability of allowancerops significantly. It is remarkable that theorems likes ‘enoughigned crossings in the neighborhood’ and ‘zone 30 km/h shoulde extended’ are not significant in the prediction of the allowance.he theorem saying ‘the traffic infrastructure is adapted to the needf my children’ becomes significant for high speed limited roads,xpressing that infrastructure is important for parental decisionaking especially on roads with higher speed limits.The models are stronger for middle speed limited roads than for

ow or high speed limited road. This is expressed by the pseudo R2

Nagelkerke R square) which ranges from 0.3 to 0.55. The higher thisalue the better the model predicts the parental allowance. This isogical since the variance of the parental allowance is smaller onhose roads because the largest part of the children is allowed onow speed limited roads and not allowed on high speed limitedoads.

.4. Accident statistics

In Table 6 the number of accidents of a group is expressed rel-tive to the total number of accidents (only bicycle injuries arencluded, pedestrian injuries are not accurate enough) involving ahild, which happened between 2002 and 2005, on a road belongingo one of the 8 road types. The numbers are always differentiated byender, age and road type. This was possible because every reportedccident in Flanders has data on the location, circumstances andersons involved. Although no risks can be compared, some inter-sting trends become clear from this table: (1) on every road type,he number of accidents is significantly higher for boys than for

irls, with the largest differences on roads without cycling lanes.2) The accidents of the different age groups are similar distributedver the different road types, which means that a dangerous roads dangerous for all ages. (3) When comparing roads with the same

and Prevention 45 (2012) 39– 49

speed limits it is clear that more accidents happen on roads withouta cycling or a pedestrian lane.

4.5. Spatial distribution of the allowance index

Because the allowance index is related to specific road types, itcan be used to map roads according to the proportion of childrenof various ages who are allowed autonomous trips. This map ofthe allowance index can be compared to road accident statisticsto make a comparison between real and perceived risks. Heist-op-den-Berg, a municipality of 40,000 inhabitants in the centerof Flanders is used as illustration. Heist-op-den-Berg consists ofone urbanized center surrounded by several small village centers.Ribbon development is common on roads connecting the centers.Every road segment was classified according to the categories usedin the survey. Next all segments were attributed with the aver-age allowance index (only for cycling) for children between 6 and12 years old (allowance index per age category is also possible),and overlaid with traffic accident locations (Fig. 4). Roads with anallowance index greater than 0.75 are indicated as ‘perceived safe’on the map. The map shows discrepancies, such as roads whichare perceived as being safe, yet where traffic accidents occurredbetween 2002 and 2005. These are located in the main center, nearelementary schools. The resulting road map can be used for net-work analyses related to mobility of children. Examples are shortestperceived safe route calculations, finding closest facilities whichare perceived as safely accessible and finding service areas of play-grounds and other facilities intended for children which have a lownumber of accidents in their vicinity. A Question such as ‘how manypercent of the road network is perceived safe within the directneighborhood of elementary school x?’ can now be answered.

5. Discussion

5.1. Parents determine the travel mode choice

Based on the observed allowance index and the general travelmode distribution of Flemish children, the parental perception ofthe traffic safety (expressed as allowance) seems to play an impor-tant role in the travel mode used by children. Indeed a similardifference in the allowance index and the travel mode by age isobserved, with an increase of bicycle use by age. No increase ofwalking is observed although the allowance index for walking alsodecreases with age. The reason for this status quo in walking lies inthe fact that the percentages of children walking to school bothinclude independent and guided trips while parental allowanceindex is based on independent travelling alone. Children, as theyget older, start to walk more unattended because parents allowmore independent walking at later age. In that sense also walk-ing correlates well with the allowance index. When the allowanceindex is differentiated by gender and compared to the travel modedistribution, it is clear that parents differentiate in travel mode forboys and girls. Boys are more allowed to travel alone as a cyclist andthey do so in reality as well. These observations confirm that chil-dren’s travel mode choice is partly determined by parental decisionmaking, as was already indicated by Fotel and Thomsen (2004), Kerret al. (2006) and Prezza et al. (2006). They also are a call for moreresearch in the age group 6–12 years old. The differences within thegroup are so large that it would be wrong to see this group as one.Requirements in terms of traffic infrastructure, traffic education ortravel mode choice, differ by age and gender.

5.2. Controlling factors of parental traffic safety perception

Descriptive numbers of the subjective allowance index by roadtype show that the presence of a bicycle or pedestrian lane is

K. Nevelsteen et al. / Accident Analysis and Prevention 45 (2012) 39– 49 47

Table 6Relative number of bicycle accidents by road type, age and gender (2002–2005).

Road type Age Gender

6 7 8 9 10 11 12 Boys Girls

Vehicle restricted with cycling lane (%) 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.1 0.0Vehicle restricted (%) 0.0 0.0 0.0 0.1 0.1 0.3 0.3 0.2 0.7Zone 30 km/h with cycling lane (%) 0.2 0.2 0.5 0.3 0.8 1.0 1.7 2.8 1.8Zone 30 km/h (%) 1.1 1.9 1.7 2.0 2.8 3.1 5.2 11.8 6.1Build up area (50 km/h) with cycling lane (%) 0.6 0.6 0.7 1.1 2.0 3.1 6.6 8.8 5.7Build up area (50 km/h) (%) 2.1 2.8 2.6 4.0 5.3 8.2 13.8 25.3 13.8

S

pasdipasF

Outside the build-up area (≥70 km/h) with cycling lane (%) 0.4 0.4Outside the build-up area (≥70 km/h) (%) 0.4 0.8

um of values under ‘Age’ equals 100 and sum of values under ‘Gender’ equals 100.

erceived safer than speed limits. Physical separation of childnd motor vehicles is perceived more important than loweringpeed limits. Through binary logistic models significant factorsetermining the safety perception of different road types were

dentified. Variables related to children and variables related to

arents are significant contributors to predict the allowance prob-bilities on different road types. Variables related to the childrenhow that age and gender are significant predictors of allowance.ollowing the models the differences in allowing probabilities

Fig. 4. classified parental average allowance index (6–12 years old)

0.5 0.6 2.0 2.2 5.6 7.8 3.80.5 0.9 1.3 2.6 4.4 7.3 3.8

enlarge between the groups, as roads have no cycling or pedes-trian lane. Differences are also large when the speed limit is50 km/h. This is an illustration that children between 6 and 12years old cannot be seen as one group and that the type ofroad plays a role in the decision of allowance. ‘Neighborhood

is traffic safe and’ and ‘lots of traffic at school area’ seem mosteffective in predicting allowance and this for every road typeand both for walking and cycling. This suggests that the deci-sion to allow a child on a certain road type mainly depends

and bicycle accidents for the municipality Heist-op-den-Berg.

4 alysis

oa

iieaiarAcd

5e

tacsdtdsdlTa(wrSaira

5

aliiiliwTct

5

spdokatt(

8 K. Nevelsteen et al. / Accident An

n the traffic situation in well-known (i.e. near-by and school)reas.

This observation is important although it has a double mean-ng. On the one hand this is good because it suggests that specifiednfrastructural measure in ‘near-by areas’ will have a positiveffect on the parental safety perception and can increase children’sutonomous travelling. On the other hand the danger exist of focus-ng infrastructural measures only on residential areas and schoolreas, leaving the roads in between intact, even if these are part ofoutes taken by children and/or have a higher number of accidents.s is illustrated in the example of Heist-op-den-Berg (Fig. 4) it islear that several accidents involving children do not happen in theirect neighborhood of schools.

.3. Increasing travelling as a cyclist or pedestrian requires safenvironments

When studying measures for increasing the number of childrenhat travel as a cyclist or pedestrian, it is necessary to take intoccount that such an increase is only desirable if children’s safetyan be guaranteed to an acceptable level. Therefore real trafficafety was included in the study, expressed as numbers of acci-ents. The relative distribution of accidents shows some similarrends with the ones observed in the allowance index. More acci-ents happen on roads without a cycling lane and this for everypeed limit category, both for boys and girls. It is so that more acci-ents happen on roads without a cycling lane with a lower speed

imit than on roads with a cycling lane but with a higher speed limit.hese numbers do not express risk because no exposure data wasvailable. But considering the fact that parental decision makingthe allowance) has a great impact on children’s travel mode choice,e can assume that if parental allowance decreases for a certain

oad type, fewer children will be exposed on that type of road.o with parental allowance significantly lower for roads without

cycling lane and the number of accidents higher on those roads,t is reasonable to say that the distribution of risk over the differentoad types would confirm the trends the accident numbers providend probably even strengthen them.

.4. Infrastructure and traffic safety

Both subjective and objective traffic safety measures express call for targeted measures instead of the more general speedowering measures, or in other words infrastructural measures aremportant both for real and perceived risk. Programs like educat-ng the driver are in this opinion less effective and will have a lowermpact on changing children’s travel mode, because parents beliefess in measures where people’s behavior is important (speed lim-ts). These programs will no doubt affect the number of accidents

ith children but have less impact on the parental perception.he strength of infrastructure measures on children’s travel modehoice lies in the combination of its impact on real and perceivedraffic risk.

.5. Practical realization is a spatial problem

As children’s travel mode choice strongly follows parental deci-ion making, aspects that influence parental thinking can deliverositive effects for increasing the number of trips made by chil-ren as a cyclist or as a pedestrian. Taken into account the necessityf providing a safe traffic environment, traffic infrastructure is aey element for success because both the number of accidents

nd the parental allowance is influenced by it. Keeping in mindhe above discussed topics, our starting point is in essence a spa-ial problem. Infrastructure measures taken in ‘well known areas’e.g. schools, residential neighborhoods,. . .) are effective on the

and Prevention 45 (2012) 39– 49

perceptual side, but lots of accidents still happen in areas betweenthose ‘known places’. Probably this is partially due to the fact thatmore infrastructural measures were already taken in e.g., school-areas in the past. Of course this is positive but it does not seemto be enough observing the decrease in children’s travelling as acyclist or as a pedestrian over the last two decades. So the combi-nation is important and infrastructural measures are necessary onlocations known to be dangerous (e.g. based on accident statistics)and known to have a positive effect on the parental perception. Bymapping parental allowance index in combination with accidentlocation such locations can become clear (e.g. Fig. 4). The work pre-sented here specified only the presence of a walking or cycling laneas infrastructural variable and showed that this already has a sig-nificant impact. In future work more infrastructural variables canbe added to enhance detail on parental perception and so narrowsdown the number of locations where measures can or should betaken.

6. Conclusions

In general the results of this study confirm that parental per-ception of the traffic environment is one of the most importantaspects to achieve an increase in children’s trips as a cyclist or asa pedestrian. The perception (here studied as the allowance) dif-fers by age of the child, road type, travel mode (cycling or walking),and the perceived safety in ‘near-by areas’. The study supports theconceptual framework of Mcmillan (2005) and demonstrates thatthe practical realization of changing children’s travel mode choiceis also a spatial problem, where perceived and real safety shouldbe combined in order to change the travel mode choice of chil-dren between 6 and 12 years old. In this study only the presenceof cycling and pedestrian lanes and speed regimes are studied. Theresearch showed that these infrastructural features and measure-ments have a significant impact on both parental traffic perceptionand on the number of accidents. Future research can focus on moreinfrastructural features which will extend the insight in which typeof environments the parental perceived risks are high or low. Bygathering information on exposure, real traffic safety can be stud-ied in more detail. So with more data available, a similar study,which uses the same methods, could lead to an extended, morecomplete study on the travel mode choice of children. The out-come of this study shows that combining both parental perceptionand accident statistics in a spatial relation can demonstrate loca-tions where infrastructural improvements will be most effective toincrease trips made by children as a cyclist or as a pedestrian in asafe way. Children will receive safe traffic environments adapted totheir needs and they will be allowed to cycle and walk more, whichhas positive effects on their development. In the end it is in every-body’s interest that traffic environments are suitable for childrenso they can develop into appropriate road users.

Role of the funding source

This research is funded by ‘FondsvoorWetenschappelijkonder-zoek (FWO)’.

Acknowledgement

This research could not be performed without the help of the>5000 household who kindly filled out the questionnaire. The studyis part of the ‘Levenslijn’ project, which focuses on children’s mobil-ity in Flanders.

alysis

R

A

BB

C

C

D

E

E

F

F

G

G

H

H

H

J

K

K

K

L

K. Nevelsteen et al. / Accident An

eferences

dams, J., 1988. Evaluating the effectiveness of road safety measures. Traffic Engi-neering and Control 21 (6), 344–352.

eck, U., 1992. Risk Society. Towards a New Modernity. Sage, London.lack, C., Collins, A., Snell, M., 2001. Encouraging walking: the case of journey-to-

school trips in compact urban areas. Urban Studies 38 (7), 1121–1141.arlin, J.B., Stevenson, M.R., Robert, I., Bennett, C.M., Gelman, A., Nolan, T., 1997.

Walking to school and traffic exposure in Australian children. Australian andNew Zealand Journal of Public Health 21 (3).

ooper, A.R., Andersen, L.B., Wedderkopp, N., Page, A.S., Froberg, K., 2005. PhysicalActivity levels of children who walk, cycle, or are driven to school. AmericanJournal of Preventive Medicine 29 (3).

iguiseppi, C., Roberts, I., Leah, Li., 1997. Influence of changing travel patternson child death rates from injury: trend analysis. Bio-Medical Journal 314,710.

RSO European Road Safety Observatory, 2008. Annual Statistical Report.www.erso.eu.

wing, R., Schroeer, w., Greene, W., 2004. School location and students travel: anal-yses of factors affecting mode choice. Transportation Research Record: Journalof the Transportation Research Board 1895, TRB, National Research Council,Washington, D.C., 55–63.

otel, T., Thomsen, T.U., 2004. The surveillance of children’s Mobility. Surveillanceand Society 1 (4), 535–554.

yhri, A., Hjorthol, R., 2009. Children’s independent mobility to school, friends andleisure activities. Journal of Transport Geography 17, 377–384.

eurts, K., Thomas, I., Wets, G., 2005. Understanding spatial concentrations ofroad accidents using frequent item sets. Accident Analysis and Prevention 37,787–799.

iddens, A., 1999. Runaway World: How Globalization is Reshaping our Lives. Pro-file, London.

ajnal, I., Miermans, W., 1996, onderzoek verplaatsingsgedrag Vlaanderen, Diepen-beek, opdracht ministerie van de Vlaamse gemeenschap, mobiliteitscel.

amed, M.M., Al Rousan, T.M., 1998. The impact of perceived risk on urban com-muters’ route choice. Road and Transport Research 7 (4), 46–63.

illmann, M., Adams, J., Whitelegg, J., 1990. One False Move: A Study of Children’sIndependent Mobility. Policy Studies Institute, London, England.

ohansson, M., 2006. Environment and parental factors as determinants of mode forchildren’s leisure travel. Journal of Environmental Psychology 26, 156–169.

aesemans, G., 2002. Tienerkliks. In: Het verhaal van het Vlaamse Kliksonsonder-zoek. Garant, Leuven/Apeldoorn.

arsten, L., Kuiper, E., Reubsaet, H., 2001. Van de straat? In: De relatie jeugd enopenbare ruimte verkend. Van Gorcum, Assen.

err, J., Rosenberg, D., Sallis, J.F., Saelens, B.E., Frank, L.D., 2006. Active commuting toschool: associations with environmental and parental concerns. The American

College of Sports Medicine.

eonard, P.A., Beattie, T.F., Gorman, D.R., 1999. Under representation of morbidityfrom paediatric bicycle accidents by official statistics-a need for data collectionin the accident and emergency department. Biomedical Journal, Injury Preven-tions 5, 303–304.

and Prevention 45 (2012) 39– 49 49

Luhmann, N., 1991. Soziologie des risikos [The sociology of risk]. De Gruyter, Berlin.Macpherson, A., Roberts, I., Pless, B., 1998. Children’s exposure to traffic and pedes-

trian injuries. American Journal of Public Health 88, 1840–1845.Martin, S., Carlson, S., 2005. Barriers to children walking to or form school – Unites

states. MMWR, Morbidity and Mortality Weekly Report 54 (38), 949–952.McDonald, N.C., 2007. Children’s mode choice for the school trip: the role of distance

and school location in walking to school. Transportation 35, 23–35.McDonald, N.C., 2008. Household interactions and children’s school travel: the effect

of parental work patterns on walking and biking to school. Journal of TransportGeography 16, 324–331.

Mcmillan, T.E., 2005. Urban form and a child’s trip to school: the current literatureand a framework for future research. Journal of Planning Literature 19, 4.

Mcmillan, T.E., 2006. The relative influence of urban form on a child’s travel modeto school. Transportation Research Part A 41, 69–79.

Noland, R., 1995. Perceived risk and modal choice: risk compensation in transporta-tion systems. Accident Analysis & Prevention 27, 503–521.

O’Brien, M., Jones, D., Sloan, D., Rustin, M., 2000. Children’s independent social mobil-ity in the urban public realm. Childhood 7 (3), 257–277.

Pooley, C.G., Turnbull, J., Adams, M., 2005. The journey to school in Britain since the1940s: continuity and change. Area 37 (1), 43–53.

Posner, J.C., Liao, E., Winston, F.K., Cnaan, A., Shaw, K.N., Durbin, D.R., 2002. Expo-sures to traffic among urban children injured as pedestrians. Biomedical Journal,Injury Preventions 8, 231–235.

Prezza, M., Romana, A., Cristallo, C., Luigi, S., 2006. Parental perception of social riskand of positive potentiality of outdoor autonomy for children: the developmentof two instruments. Journal of Environmental Psychology 25, 437–453.

Rasborg, K., 2002. Beyond the risk society. In: Paper presented at Nordisk Soci-ologkongrs, Reykjavik, August, pp. 15–17.

Roberts, I., 1993. Why have child pedestrian death rates fallen? BMJ 306, 1737–1739.Roberts, I., carlin, J., Bennett, C., Bergstrom, E., Guyer, B., Nolan, T., Norton, R., Pless, B.,

Rao, R., Stevenson, M., 1997. An international study of the exposure of childrento traffic. Biomedical Journal, Injury Preventions 3, 89–93.

Schlossberg, M., Greene, J., Paulsen, P., Johnson, B., Parker, B., 2006. School trips:effects of urban form and distance on travel mode. Journal of American PlanningAssociation 72 (3), 337–346.

Steenbergen, T., Aerts, K., Thomas, I., 2010. Spatial clustering of events on a network.Journal of Transport Geography 18, 411–418.

Timperio, A., Salmon, J., Telford, A., Crawford, D., 2004. Perceptions of local neighbor-hood environments and their relationship to childhood overweight and obesity.International Journal of Obesity 29, 170–175.

Yates, J.F., 1994. Risk-taking Behavior, second ed.,. John Wiley & Sons Ltd., England.Zeiher, H., 2001. Chidren’s island in space and time: the impact of spatial differenti-

ations on children’s ways of shaping social life. In: Kruger, H.H., Bois-Reymond,M.D., Sunker, H. (Eds.), Childhood in Europe: Approaches, Trends, Findings. PeterLang, New York.

Zwerts, E., Nuyts, E., 2001. Onderzoek verplaatsingsgedrag Vlaanderen Diepenbeek,opdracht ministerie van de Vlaamse gemeenschap, mobiliteitscel.

Zwerts, E., et al., 2009. How children view their travel behaviour: acase study from Flanders (Belgium). Journal of Transport Geography,doi:10.1016/j.jtransgeo.2009.10.002.