1 Introduction

The World Health Organization [1] recommends an average of 60 min of moderate-to-vigorous physical activity for people between 5 and 17 years of age. Increased sedentary behaviour in children and teenagers—e.g., recreational screen time or being driven—can cause negative health effects. Worldwide, childhood obesity is estimated at 9% (60 million children) by 2020 [2]. In developing countries between 1980 and 2013, overweight and obese children and adolescents rose from 8 to 13% [3]. Conversely, the benefits of physical activity are many: cardiorespiratory fitness, muscular strength, bone health, lower body fatness and good mental health [4].

The built environment is widely considered a social determinant of health, including during childhood and adolescence [5]. It can facilitate an active lifestyle or contribute to the rise of non-communicable diseases, like obesity, when its design inhibits activity.

A good way to increase physical activity levels in children and adolescents is via active commuting to and from school. Those who engage in active transportation (non-motorized means that imply human power, such as walking and cycling) to/from school (ATS) present higher levels of overall physical activity [6]. Interest in active everyday utilitarian mobility has been increasing as the benefits for people, cities, and the environment, have been increasingly identified [7, 8].

Despite the numerous advantages of ATS, some research suggests it is decreasing, both in the United States [9] and Canada [10], and also in Europe—e.g., Spain [11] and Czechia [12].

Research provides evidence that urban characteristics influence children’s travel mode choice [13]. Two decades ago, Cervero and Kockelman [14] established the 3Ds—density, diversity and design—to identify the main built environment characteristics that influence adult transport mode choice. A decade later, Lee and Moudon [15] refined the attributes of the 3Ds and added a fourth element with their 3Ds + R (route characteristics) concept. These four measures of the built environment are widely accepted by researchers and have been further refined.

The D variables have been broadened to include destination accessibility and distance to transit [16]. Handy [17] suggests considering place accessibility, proposing that choices offered by the built environment greatly influence decisions about travel mode. The level of accessibility from a given place indicates the presence of destinations around it and those that can be reached by different means of transport (motorized or not). Thus, accessibility can be a predictor of active transportation.

A complementary approach postulates that attitudes towards active transportation determine the use of active travel modes [18, 19]. Accordingly, children’s willingness to go to school on foot or by bike might be influenced by parental preferences and habits. The built environment itself impacts people’s perception and attitudes about travel choice [20].

Finally, other academic research suggests socioeconomic characteristics, perceptions of the built environment, and personal attitudes—e.g., household transportation options, social and cultural norms, and caregiver attitudes—influence travel behaviour [13]. This includes individual factors [21]—e.g., age and gender—community, school and family factors [19], and other economic, socio-cultural and political characteristics [13, 18]. These factors mediate the influence of the built environment characteristics on physical activity, including ATS [8].

2 Research relevance, objective and methods

Despite significant efforts from many knowledge fields that have contributed to a better understanding of ATS, many questions remain about which factors impact it. Besides the inherent complexity of the urban environment, there is also the problem of individual characteristics, preferred lifestyles and travel options, whose correlations affect travel behaviour [22].

Children and youth represent a vulnerable group in urban planning, since cities are planned by adults and conceived mainly to meet their needs [5]. Urban correlates of active transportation have generally been studied in relation to adults, mostly considering work travel and seldom school commuting. Accordingly, most of the frameworks applied to school travel were conceived for adult work travel [23].

Aware of previous reviews [18, 21, 23, 24], the aim of this paper is to summarise the latest research methods, influencing factors and findings on the relationship between the built environment and ATS, focusing on both urban correlates and individual characteristics. An updated systematic review is required to inform about the status of the unanswered questions of previous research in this area. Prior conceptual frameworks on this topic reveal the need to observe not only environmental factors, but also personal characteristics and attitudes [21, 24]. For some factors—e.g., household travel patterns, socioeconomic status, social support, traffic safety…—there are still no consistent findings of its positive association with active transportation to school [24]. Multi-level design methodologies, which combine objective and perceived measures of the built environment [18], and which can be replicated internationally [23], are also demanded. Although geocoding systems have improved over the last decades, there is until now difficulty to estimate the routes to school [23] or to capture urban data of the microscale.

The following sections explain the research methods used to conduct this review and the influencing factors found.

2.1 Literature search

We conducted a systematic search for English- and Spanish-written peer-reviewed journal articles published since January 2010 until 2020. We chose those languages because they are predominant in the Western world. The articles were available in the following databases (number of articles in brackets): CINAHL Complete (2), Scopus (6), Gale Academic OneFile (1), Science Direct (14), Food Science Source (4), MEDLINE Complete (4), Open AIRE (2), SPORTDiscus (2), JSTOR Journals (2), APA PsycInfo (1), Sociology Source Ultimate (2) and Fuente Académica Premier (1). The period was chosen to make an update of the research over the last decade. Literature reviews and dissertations were excluded from the search. As we centre on the Western world, studies which take place in Asia were excluded. The search terms used were: (environment* or neighbo* or facilit* or walkab* or urban design or aesthetic* or infrastructur* or land use or connect* or level of service or urban form or cyclab* or street*) and (pedest* or foot or walk* or cycl* or bik* or transport* or commut* or active tra* or mobilit*) and (child* or boy* or girl* or student* or pupil* or young* or youth or adolescent* or teenage* or school).

The search provided 1340 results that were filtered based on the PRISMA Statement for Preferred Reporting Items for Systematic Reviews and Meta Analyses [25]. Once duplicates were removed, 398 articles were screened based on their title. The 76 most relevant articles were screened based on the abstract, resulting in 41 journal articles that were fully read and finally included in the study. The selection process is shown in Fig. 1.

Fig. 1
figure 1

Flow diagram of article selection

Full size image

The selected articles come from different disciplines, including architecture, urban planning, geography, transportation, medicine, sport science, nutrition, and public health, highlighting the transdisciplinary nature of the topic, previously noted by Pikora et al. [26], McMillan [24], and Timperio et al. [27]. Articles included investigate the built environment in relation to ATS in children or teenagers (6–18 years old) and the influence of individual and external factors. Studies dealing with a different destination or purpose, or focusing on students with disabilities, were discarded.

Parents play a key role in trip to school decision-making up to a certain age [24], acting as predictors of children’s active travel mode [28]. Among the additional records mentioned before, three of them [29,30,31] focus on the parental perception of the built environment and, for this reason, were included in this study.

2.2 Investigated factors

This review presents and discusses in the next sections the general research methodologies and characteristics of the selected studies and then analyses the specific research methods and factors that affects ATS, following the framework developed by Panter et al. [21]. First (from Sect.  4 to 5.4), we consider the group of the physical environmental attributes, dividing them in two classes, according to the Behavioural Model of the Environment by Lee and Moudon [15]: (a) the area characteristics around the origin and destination, and (b) the characteristics of the route connecting them. Second (from Sect.  6 to 6.4), we contemplate the individual and external factors. The former refers to demographic and socioeconomic characteristics, and individual attitudes and perceptions of the built environment. The latter, include variables such as the climate, the cost of travel or government policies.

According to the social-ecological model by Mcleroy [32], there are multiple levels of influence on travel behaviour: personal characteristics, psychosocial factors, environmental factors and policies [33]. Previous frameworks and reviews of active transportation [24, 26, 34] and many among the analysed studies [28, 35,36,37,38,39,40,41,42] adapt the social-ecological model to identify the characteristics affecting ATS.

2.3 Study characteristics

This section discusses three methodological aspects of the selected studies—(a) location, (b) data, participants, and sample size, and (c) travel mode focus—summarised in Table 1.

Table 1 Characteristics of the studies reviewed
Full size table

2.4 Location

Among the 41 studies, 20 (49%) were conducted in six EU countries, 12 in the USA and Canada, 7 in Australia and New Zealand, and the remaining 2 in Mexico and Brazil. This shows a growing interest in the matter in Europe, compared to a majority of North American studies reported by [43], who pointed out the need to conduct studies elsewhere.

Although the precise location is not always available—sometimes not even the city—it is clear that the studies involve cities of greatly varying size: some have no more than 100,000 inhabitants (e.g., Lünen, DE and Heerlen, NL) others have over 5 million (e.g., London and Mexico City). Kim and Heinrich [44] compare the environmental factors of cities as different in size as Manhattan, KS (pop. 52,000) and Austin, TX (pop. 790,000). A map of the locations is shown in Fig. 2.

Fig. 2
figure 2

Source: Map by authors, based on Google Maps base layer

Location maps of the reviewed studies. Blue circle = location unavailable, just country information, blue icon = study take place in one city, other colour icons = study takes places in multiple cities (each colour represents one paper).

Full size image

Some investigate full metropolitan areas [45, 46] and some focus only on particular boroughs, districts or census blocks, to compare the influence of different urban structures at the macro level [47,48,49].

Although not always clearly stated, the majority of the studies apparently take place in urban environments, with some exceptions. Race et al. [46], and Gallimore et al. [48] compare experiences and communities of urban and suburban children. Carver et al. [31], Macdonald et al. [50]; Nelson and Woods [51] and O’Loghlen et al. [52] compare urban and rural schools. Differences between built environment types reflect the variety of community settings and a variation in the factors which affect mode choice.

2.5 Data, participants and sample size

The origin and nature of morphological and sociologic data is not clearly defined in many studies. While some use external broader surveys from which selected information is extracted (54%), others develop their own methods to gather the required travel and socioeconomic information using questionnaires previously developed by the same or different researchers (5%), or original questionnaires specifically prepared for the given study (41%).

Different studies use different data sources: reported by the students (19%), or by adults, including parents (24%), caretakers or school representatives (2%), or mixed (55%).

Sample size varies considerably. The number of schools included varies from 3—Race et al. [46] in Vancouver—to 397—O’Loghlen et al. [52] in various Canadian cities. However, almost half the studies (45%) do not provide school numbers. The number of students ranges from 10—Zdrahalova and Boumova [53] in Prague—to greater than 6000—Dygrýn et al. [12] IPEN study, in the Czech Republic. The majority (52%), include over 1000 children and teenagers, and only three contain less than 100. The percentages of girls and boys are usually similar. Only two samples differ more than 6%, reaching almost 11% in Panter et al. [28].

Age is an important factor for ATS, since students’ autonomy increases with age [54]. Almost half the studies (45%) only include students with age differences of 4 years or less and three (7%) involve students attending the same grade—Gallimore et al. [48], Kim and Heinrich [44], Zdrahalova and Boumova [53]. However, about a quarter of the studies include students with a 6-year age span. The maximum age span (5–15 years) is analysed by Sener et al. [29]. When the range is big, age grouping is performed for statistical analysis, but with no homogeneous criteria. The mean age across samples is 10.2.

Since no longitudinal studies were identified for this review, only cross-sectional studies are included. Among those, only three included a follow-up after 1 to 10 years [12, 36, 55]

2.6 Transportation mode focus

More than half of the studies (54%) analyse both walking and cycling. In four cases, cycling rates are very low [28, 54, 56, 57]. Beside those, six studies (15%) include other low modal share active means of transport, like skateboards, roller-skates or scooters. Some studies (27%) consider only walking. Two focus exclusively on bikes [58, 59].

Most students travel to school using the same transport mode every day. Thus, most studies (72%) only take account of the most frequent travel mode used by each student. The others (28%) analyse recall periods (e.g., daily, various days a week, or weekly), and normally consider students active when they report more than 40% of the trips using active means.

3 Area definition and characteristics

Since the built environment influences travel choice and thus ATS, the physical characteristics of origin and destination areas are deemed important. Which physical characteristics and the exact role they play are questions that researchers still focus on. There is no consensus on factors, how to measure them, or how to define the geographic area of analysis.

In what follows, different approaches to defining the study area are analysed. We use ‘study area’ instead of ‘neighbourhood’ following the terminology of Lee and Moudon [15]. Subsequently, the physical characteristics being investigated and new evidence of how they impact ATS is summarised. The corresponding columns in the first part of Table 2 present the most relevant findings of this analysis.

Table 2 Methodological Characteristics and main findings of the studies reviewed
Full size table

3.1 Area definition

Researchers generally consider areas around the origin and destination—home and school—relevant to study, with 85% of the studies explaining which they study and how they are delimited.

While 10% of the studies [30, 39, 42, 46, 54] use the complete school attendance boundary to define the limits of the observed built environment, most researchers define the study area using a buffer around the student’s home or school, depending on which is considered as central. 50% consider both, 19% only the area around the school, and 16% only the area around the student’s home.

Two types of buffers are used: straight-line buffers (SLB)—also referred to as airline buffers and Euclidean buffers—and street-network buffers (SNB). The former is a circle around a given centre. The latter is an irregular polygon, obtained by joining the vertices at the end of a given distance measured along the street network in all directions. 16% of the studies use SLB, 57% use SNB.

Buffer size varies tremendously, ranging from 200 to 1600 m (m) (1 mile). According to Bejleri et al. [60] a 10-min walk (800 m or half mile) is a reasonable distance for elementary school children, but there is no consensus. As such, authors of other studies consider several different distances. Jauregui et al. [56] use 400 and 800 m SLB around schools, Oreskovic et al. [55] use 400, 800, and 1600 m SLB and SNB. Mitra and Buliung [38] use 250, 400, 800 and 1000 m SLB around the school and the student’s home.

A few researchers use other approaches to generate the buffers. Aarts et al. [61] use SNB buffers based on the parent’s perspective, as the reachable area in a 10-to-15-min walk or in a 5-to-8-min ride. Carver et al. [36] use 800 m SNB around the students’ homes but consider the whole pedestrian network rather than just streets, such as park paths.

Another approach defines the study area as a SLB along the student’s route between home and school. Among the five studies (12%) using this method, the buffer distance varies: 50 m [55], 80 m [40] and 100 m [28, 36, 62].

3.2 Area characteristics

Once the study area is defined, a wide range of methods are used to define and analyse its characteristics. Seven studies (17%) use walkability indices, generally based on dwelling density, land use mix and intersection density [35, 37, 45, 46, 49, 50, 56].

Christiansen et al. [37] categorized school areas in Denmark based on road connectivity, dwelling density and vehicular traffic exposure, but found no difference in active transportation between low, medium and high walkable areas for trips shorter than 0.5 km. Macdonald et al. [50] used the same factors to characterize the zones around the students’ homes in Scotland, and found that walkability had a positive association with ATS for all trips considered, even for those longer than 2 km. Mitra and Buliung [38] found a stronger association of environmental characteristics and mode choice for the 400 m SLB around school and child’s home in Toronto. Similarly, in Melbourne, Carver et al. [35] found a positive association with ATS for children living closer than 750 m (almost half a mile) to school in highly walkable neighbourhoods. School surroundings with high walkability are related in their study with more ATS.

Six studies (15%) compare urban versus suburban or rural sites, reporting different levels of impact on ATS. Race et al. [46] and Gallimore et al. [48] compare new urban developments versus suburban areas. The latter finds that children in new urban communities in Utah walk more than those in suburban areas. These places have connected streets and smaller lot sizes, versus the suburban environments where residential cul-de-sacs connect directly with main arterials, which could be hazardous for children. At the same time, children feel safe when they walk along typical suburban streets with single-family detached houses [46]. Among those investigating rural areas O’Loghlen et al. [52], find that schools in rural areas in Canada are usually situated on high-speed roads, making ATS difficult.

Area walkability can be described differently, such as infrastructure provision versus quality [63]. Most studies use urban form features to characterize study areas, but Carver et al. [36], van Kann et al. [41] and Rothman et al. [30] use on-foot street audits to observe the school environment during school hours. They gather information about universal accessibility characteristics, the presence and conditions of sidewalks, traffic lights, and aesthetics, and also evaluate schools’ policies promoting ATS. Three studies (7%) observe aesthetics. Van Kann et al. [41] focused on school area aesthetics in Dutch cities and found a correlation between ATS in higher grades and the presence of parks, park trails, and absence of litter. However, D’Haese et al. [64] measured similar factors in Belgian cities using questionnaires and found no significant association. In Irish cities, Nelson and Woods [51] found aesthetic perceptions are predictors of ATS only among male students. Following Gehl’s theories about the relation of the quality of the built environment and necessary activities, these dissimilar results suggest aesthetics do not play an important role.

4 Route characteristics

The characteristics of the route to school are key to understanding travel choice [21, 27]. Distance is the biggest determinant, with only two studies not including it, Dygrýn et al. [12] and Van Kann et al. [41]. Almost half the studies (42%) analyse other route characteristics, including route directness and traffic safety. The corresponding columns in the second part of Table 2 summarise these factors and the main findings.

4.1 Distance

Previous reviews [18, 21] declared distance the most significant predictor of ATS and this review confirms it. Waygood and Susilo [65] find it more important than car ownership, which is also a determinant factor for Ikeda et al. [34].

There are four ways to measure the distance between origin and destination: (a) the Euclidean distance—i.e., the length of the straight line between two points; (b) the shortest route along the street network excluding those elements not accessible for the considered active mode—e.g., expressways; (c) the shortest route along the active transportation network, includes pathing that are accessible for the considered active mode—e.g., pedestrian or cyclable paths through parks; and (d) the actual route taken by the students, either self-reported, through questionnaires or interviews, or using a GPS tracking device.

Only Mäki-Opas et al. [58] use Euclidean distance exclusively, and two [60, 66] use it in combination with the shortest route through the network. Most studies (55%) use shortest route, although it is not always clear which approach to defining shortest route is taken. Only two studies consider actual route. Smith et al. [40] asked children to map it, and Helbich [62] used GPS tracks.

Spatial distance and temporal distance are not always equivalent. Two researchers use time to characterize distance. Chillón et al. [11] use the duration of the trip, and Kelly and Fu [67] calculated the spatial distance by multiplying average journey time by average speed. Perceived distance may be important [68] but only three papers in this literature review consider it, Kim and Heinrich [44]; Yu and Zhu [42]; Zdrahalova and Boumova [53].

Distance is an inclusion/exclusion criteria or sample classification criteria in 24% of the studies. The threshold varies greatly, not only based on the active mode of interest. The use of both US customary units and SI units leads to use of multiples and divisors of the mile or the kilometre, respectively, making studies hard to compare. The classification of the sample based on the distance travelled varies considerably across researchers.

Oreskovic et al. [55] focus on children living within 1.6 km (1 mile) from school in Houston, considering both SNB and SLB. Other North American studies also consider a 1.6 km SNB [39] or SLB [69] around the school. In Belgium and Brazil, D’Haese et al. [64] and Silva et al. [70] established 1.5 km and 3 km as the limit distances for walking and cycling respectively. Nelson and Woods [51] set 4 km (2.5 miles) as a ‘realistic’ distance for adolescents ATS in Irish cities.

The reasonable ATS distance is still unclear and may be country- or cultural-specific. The threshold between active and non-active trips regarding travel on foot was found to be 0.8 in Scotland [65], 1 km in The Netherlands [61], 1.3 km in Spain [66], and 1.5 km in Belgium [64] and Brazil [70]. Cycling distances can be 2–4 times greater—e.g., 3 km in the same Brazilian study, pointing to the need to distinguish between walking and cycling. Among the six studies that account for other active means—e.g., skates, skateboards, and scooters—none consider specific distances in their analyses. In general, as distance increases, the negative association with active commuting to school is stronger. Longer trips are more likely to face traffic or personal safety problems [47, 54].

4.2 Directness

Route directness measures how directly it links origin and destination. Most measures of it result from comparison of network distance versus straight line distance between ends of a given route. Thus, it is related to the spatial configuration of the street network.

Bejleri et al. [60] and Campos-Sánchez et al. [66] use Pedestrian Route Directness (PRD) which is the ratio between the shortest distance through the pedestrian network and the straight line. The former observe that students take alternative paths or shortcuts outside the conventional street network. The latter identified greater levels of ATS when the PRD was low. Panter et al. [28] found that ten year-old children who lived in deprived neighbourhoods where routes to school were more direct (which means higher PRD) walked and cycled less to school, as did Jauregui et al. [56].

Different urban forms present different street directness. For example, suburban environments usually have higher PRD values due to having disconnected pathways and some cul-de-sacs that make it difficult to walk and cycle [48, 60]. Neighbourhoods with a dense connected grid have more direct origin–destination routes—i.e., lower PRD values. Thus, walking and cycling distances are shorter. Macdonald et al. [50] found greater ATS levels in areas with greater intersection density and fewer cul-de-sacs. A perceived diversity of routes was associated with higher cycling rates in The Netherlands [61].

4.3 Traffic safety

Traffic safety is another important determinant of active travel [34, 43] and especially relevant to ATS, since it is a common parental fear. Perceived traffic safety is considered as important as the level of traffic accidents and casualties, with the two often differing [30]. Researchers consider both approaches. 23% of the studies analyse traffic safety only through the perceptions of the students’ caretakers or other adults such as the school board members. Scheiner et al. [71], Kelly and Fu [67], Panter et al. [28] and Bosch et al. [47] include data on traffic safety from local statistics, to compare perceptions with accurate measures.

Among the urban features which can be related to road safety, the most frequently mentioned are crossing guards, stop (or traffic) lights, crosswalks, and speed limits. The presence of crossing aids or crossing guards is an important correlate of ATS for Jauregui et al. [56], because they increase the perception of safety [30]. Other researchers suggest that alternative non-vehicular based street design approaches—woonerven and shared space streets—are perceived as safe as conventional streets [72].

Some issues are unclear. Rothman et al. [30, 39] found a positive association between traffic lights and walking, while Panter et al. [28] identified higher density of traffic lights with less cycling to school. Frequent traffic lights oriented to car speeds may result in too many bicycle trip interruptions [73]. Two North American studies found that marked crosswalks facilitated pedestrian travel—Race et al. [46]; Gallimore et al. [48]—and one Irish study found this association with girls [51]. In Germany, by contrast, Scheiner et al. [71] identified necessary use of marked crosswalks by children with less walking.

There is more consensus about traffic calming solutions [40, 71] and speed limit signs [48] which are broadly associated with more walking and cycling, and a positive impact on traffic safety.

Van Kann et al. [41] highlights that traffic danger is a barrier to walking and cycling in the Netherlands, especially among older teenagers. Bosch et al. [47] reached similar results in London, where children were less likely to walk or cycle on routes with higher accident levels. However, Helbich [62]—in The Netherlands—and Race et al. [46]—in Canada—found moderate or insignificant relation between traffic safety and ATS, suggesting it can sometimes be a barrier but not always.

4.4 Other route characteristics

Four studies (10%) use street audits to analyse different route characteristics at a closer scale—e.g., number of crossings, sidewalk width, presence of entrances along the street, and street maintenance. They study either the school grounds [36], the street segments [53, 69], or the blocks [48] along the route to school. Other four researchers in this literature review include subjective route measures, especially in relation to perceived barriers and safety [37, 44, 57, 71].

The influence of presence of parks and green spaces on ATS was unclear in the review by Panter et al. [21], whereas D’Haese et al. [43] found positive association with leisure time walking but not cycling.

In Houston and in Prague, the presence of park space along the route presented a negative association with ATS [53, 55]. Not parks themselves, but people drinking or smoking in them are perceived as a danger children in the Czech Republic, and thus obstruct ATS [53]. Van Kann et al. [41] evaluated aesthetics factors through the perceptions of the children and found positive associations with ATS.

5 Individual and external factors

Previous ATS studies focused on the built environment, but they also considered broader socioenvironmental models, including particular factors of the students or the adults responsible for them [13, 21, 27]. As described in Pont et al. [18], that is because of the influence of these things on children’s modal choice. This section discusses the main literature findings on individual and external factors—summarised in the last column of Table 2. This data is usually collected through questionnaires.

5.1 Demographic characteristics

Age is a highly influential personal factor affecting ATS. Ten studies (24%) show greater levels of ATS in older students [29, 31, 41, 47, 54, 56, 57, 61, 62, 71] and two show lower probabilities of being driven with increasing age [31, 71]. Children begin to travel autonomously from age 10 in Ireland, New Zealand and the Czech Republic [53, 54, 67]. Oliver et al. [54] and Sener et al. [29] find ATS also peaks around this age, when children start middle school.

Most researchers report no significant differences in ATS based on gender [37, 47, 61, 62, 65]. However, two studies show more ATS among boys than girls [29, 59]. Chillón et al. [11] and Sener et al. [29] suggest parents accept less independent mobility of girls, perhaps to overprotect them [70] and thereby accompany them to school using their own non-active travel to work modes [36].

Ethnicity may be a mediating factor between built environmental factors and socioeconomic characteristics, according to Bosch et al. [47]. They found black children in London walked and cycled less and lived further away. The research of Aarts et al. [61], Mäki-Opas et al. [58], and Oliver et al. [54] in The Netherlands and New Zealand considered ethnicity but saw no impact. Chillón et al. [69] and Sener et al. [29] report some contradictory evidence in different US cities—the former showing positive association between the percentage of Hispanic students and ATS, the latter showing lower ATS in Texas for families with immigrant parents.

Other psychological characteristics related to child autonomy also play a role. Self-efficacy refers to one’s ability to succeed in a situation or in achieving a goal. In Valencia, Spain, Molina-García et al. [49] found less self-efficient children were less active. Especially among primary school children, parents drive their children because they do not feel confident that their children can use public transport or cycle alone [67]. However, Mitra and Buliung [38] suggest more research is needed on self-efficacy development.

5.2 Socioeconomic characteristics

According to Pang et al. [74] and McMillan [24], home socioeconomic characteristics impact modal choice. More than half the reviewed studies (64%) assessed variables such as household income, education and occupation of the parents, and car ownership. Typically, these characteristics are extracted from sources providing data for a whole area—e.g., census area, neighbourhood, district. Only two studies [28, 61] use family level data.

While a previous literature review by Pont et al. [18] reported higher household income had a negative association with ATS, two more recent studies [67, 71] show no significant influence. However, Sener et al. [29] found less ATS in low-income Texas families, and Panter et al. [28] found less probability of ATS in more deprived neighbourhoods in Norfolk, UK.

The role of parents’ education level remains unclear. Kim and Heinrich [44], so as Smith et al. [40] show a positive correlation between higher education and walking in different New Zealand and US cities. Sener et al. [29] report that less parental education corresponds with less active transportation in children.

Four studies find work-related factors impact ATS. Panter et al. [28], and Scheiner et al. [71] report a strong association between the mother’s employment and commute travel mode and their child’s transportation mode. However, Sener et al. [29] find the father’s job influences ATS more, noting that fathers who are self-employed walked and cycled more to school with the children. Single-parent households with a working adult are negatively associated with ATS [61].

Car ownership is mostly considered to have a negative impact on ATS [18, 71]. Among the publications, car ownership is high—more than 80% of the families have cars in Australia [31], Belgium [64] and the UK [28, 50]. This corresponds with low levels of ATS [38, 61, 67]. Not having a car shows the highest impact on ATS [64]. Yu and Zhu [42] found less car dependency and more use of alternative transportation modes in low socioeconomic status families who cannot easily afford a vehicle. A car dominant culture can lessen the influence of interventions promoting ATS [40].

5.3 Attitudes and perception of the built environment

Children’s own perceptions of the built environment and its barriers are considered only in ten studies (24%), with data sometimes gathered in recorded interviews [40, 46] and sometimes in discussion groups [53]. Children’s biggest fear is their personal safety [46] so they prefer to go to school with friends. If travelling alone, they prefer the perceived-as-shorter route [53], but travelling together they may choose longer routes [40].

More than half the studies (62%) consider parents’ attitudes and perceptions. Fear of personal danger is a concern [11], but the biggest concern is traffic safety, especially for primary school children [31]. Some parents drive their children because they find them unable to travel alone [71], like other literature review also confirm [75]. Trapp et al. [59] report that parental confidence in children’s skills mediates the association between perceived safety and cycling. Others find Children’s resistance to walking or cycling is negatively associated with ATS [59, 69]. Yu and Zhu [42] find parents worry about the heavy weight of the school backpacks. Time constraints and trip chaining are also a reason to drive [35, 42, 76].

Parental support positively affects ATS, especially when parents [46] or grandparents [57] accompany their children to school on foot or by bike. Children of parents who use active modes are more likely to use active means as well [77]. Some studies suggest parents encourage ATS more when siblings that can travel together [11, 61, 67].

5.4 External factors

School size, travel cost, mobility policies and weather can affect ATS [21] but are seldom studied. One study found weather, together with safety, the most common parental barrier to ATS [69]. However, Oliver et al. [54] found an inconsistent association between daily weather patterns and ATS in various New Zealand cities.

Policy may promote age-specific independent mobility [35], both in countries unused to a particular active mode and among parts of the population not familiar with walking or biking in countries with good active infrastructure and traditions [58, 70].

In some countries school administers decide school locations and characteristics [60]. And so, 10% of the studies include questionnaires for school boards and principals. These found that factors affecting ATS are: school administrator’s perceptions of traffic safety around the school [52]; adequate number and secure bicycle racks [61]; and access points or location of crossing guards [60].

However, school policies have not been investigated in combination with family and community support [19, 21, 78]. An active community culture—e.g. walk-to-school days and ‘walking school buses’—are important to support ATS [40]. The reduction of private car drop-offs also lessens traffic danger around schools, encouraging some families to use modes like walking school buses [67]. Follow-ups are necessary to test the effectiveness of these programs [36], as a peer literature review about school transport interventions also noted [79].

6 Discussion

This review focuses on the research methods and methodological characteristics of recent ATS studies, also including an overview of other external factors that might influence ATS, like socioeconomic characteristics, perceptions of the built environment, and personal attitudes. The first significant finding is the growing research interest outside North America, where most earlier studies took place. This review of recent studies reveals 49% investigated European cases and 29% North American. The second significant finding is the wide variation found in: (a) student population samples—by size and age; (b) the cases—by urban, geographic, and cultural environments; and (c) the analysed data—by sources and type.

As in previous reviews [18, 21], distance between the student’s home and school remains the most important environmental barrier to ATS. However, the threshold for children and adolescents remains unclear. The limit for pedestrians—more studied than cyclists—varies from 0.8 to 1.5 km (almost 1 mile), perhaps depending on regional or cultural factors.

Methodologically, the latest studies share a clear and new preference for measuring distance by the shortest route through the street network, an advantage over precedent direct line measures. This measuring approach impacts how study areas are delimited. While earlier research mostly used straight-line buffers, in the last decade street-network buffers are used in 57% of research. Since distance threshold is unclear, most studies considered buffers of 0.4 to 1.6 km (1 mile). Study areas are mostly defined around schools or the student’s homes, but 12% of the studies use 50–100 m buffers along home-to-school routes. As prior reviews also noted [23], future work should still look for a standard definition of buffer sizes, facilitating more international comparison.

Two related dimensions of the route are time and perceived distance. Their role is understudied because the necessary data is difficult to collect, but researching this may better explain modal choices.

Route directness—the relation between the shortest path through the network and the distance measured as the crow flies—also affects ATS. Research shows more direct routes present higher rates of active modes. However, this also goes for the same Euclidean distance. What is unclear and needs more research is if, given the same route distances, directness affects ATS.

Route directness results from the street network configuration, a key element of the urban form affecting ATS. The effect of urban form on ATS is mostly studied through factors such as dwelling density, land use mix and intersection density, which depends on the street network. Most studies show that higher levels of these three urban form variables positively affect ATS relating to distance, except for trips shorter than 0.5 km [37].

Walkability aspects of infrastructure provision and quality are considered in few studies (20%), probably because data is hard to obtain and analyse. Half of them consider objective-quantitative aspects and the others subjective-qualitative ones. While urban design aesthetics or quality differences might not greatly influence necessary activities like commuting to school, researchers do believe street design plays an important role in ATS, since it impacts both traffic safety and perceived risk, both important determinants of active travel. More research about the microscale of the built environment is still needed to address this gap, as mentioned in previous reviews [24].

Not many studies investigate the effect of other urban features or land uses (parks, retail, cafes) along the routes to school, and the results are contradictory. This is understandable since in different places similar features and activities can be perceived either more or less safe and attractive. A park can either be—or be perceived as—an enabler or disabler of ATS. It greatly depends on students’ and parents’—or adults in charge—perceptions. There is a need for more research on this topic.

Consistent findings of a positive or negative association with ATS were not found for individual factors like socioeconomic characteristics, perception of safety or social support. The combination of objective and perceived measures of the build environment should be considered in future studies.

Lastly, this review shows that school initiatives and policies—educational and physical—to facilitate and encourage ATS may help children perceive less barriers and increase traffic safety around schools. Effective programs to promote active transportation require a combination of school, family, and community factors [19]. Active parents and letting children contribute to design activities and actions to promote ATS seems important [46], but further research is needed, including long-term interventions, as other literature review of the last decade [79] also concluded.

7 Conclusion

The present systematic review builds on previous knowledge about environmental correlates of ATS, as well as personal and external factors, which can influence transport mode choice. It updates the latest research methods and findings, highlighting new insights into the complexity of transport behaviour in children and youth.

A limitation of this review is the exclusion of research articles undertake in Asia, which may result in an incomplete representation of the latest evidence about ATS. A second limitation relates to factors observed. Not every article focuses on both the built environment and personal characteristics, at the same time, making difficult their comparison. And third, all selected studies are cross-sectional and therefore they do not observe data over a long period of time.

The selection of the articles with the help of the systematic standards of PRISMA is one strength of this review. A second strength was the search for articles written, not only in English, but also in Spanish, widening the scope of the investigation. Finally, the selection of studies including built environmental factors, as well as individual characteristics and external aspects, meets the demand of prior research [21].

Understanding the built environment correlates of ATS is complex and requires different approaches. The physical transformation of the built environment itself does not change travel behaviour [80], but it can positively or negatively affect ATS. This review reveals a lack of longitudinal studies and a rise in interdisciplinary teams that we consider vital to further develop the required knowledge. The main findings are: (1) lack of generic definitions of buffer sizes and its characteristics, so as for threshold distances, (2) scarce information on infrastructure aspects and urban quality and (3) interventions which focus on aspects of the built environment and personal attitudes together.