Is School Travel too Complex to Handle Without a Car? Assessing “Child-Friendliness” as a Pathway to Reducing Private Car use for Children’s Transport

ABSTRACT

Recognising the multifaceted challenges of addressing the dominance of cars for children’s transport, this paper introduces and tests a “child-friendliness index” (CFI) to assess the performance of neighbourhoods and school catchments to enable or constrain how children travel to school. The findings demonstrate that car trips both to and from school are negatively associated with child-friendliness. However, counter-intuitively there was no significant association between distance and car use, which remained high for shorter distances (< 4 km). These findings imply school travel decisions are susceptible to many factors. We discuss the policy pathway provided by CFI for school accessibility planning.

摘要

基于对解决儿童交通中汽车占主导地位一事之多方面挑战的认识，本文引入并测试了一个“儿童友好指数”（CFI），以评估社区和学校招生地区的表现，从而使儿童能够采用或限制儿童上学的方式。研究结果表明，往返学校的汽车行程与儿童友好性呈负相关。然而，距离和汽车使用之间反直观地没有显著的关联，在较短的距离（<4公里）内，这一关联仍然很高。这些发现表明，上学交通方式的选择容易受许多因素影响。我们探讨了CFI为学校无障碍规划提供的政策途径。

KEYWORDS:

• Child-friendliness
• sustainable mobility
• active school travel
• local accessibility
• car dependence
• mobility environment

1. Introduction

The increased number of car trips taken by children and young people as passengers globally has been well documented (Behrens et al. 2017, Malone and Rudner 2017, SRTS 2023) The associated decline in walking, cycling and public transport trips to school is most pronounced for the younger age groups living in developed countries (McMillan 2013, Garrard 2016, Jacob et al. 2021), with car use noticeably greater in households with young children compared to other household types (Fyhri et al. 2011, Oakil et al. 2016, Bayart et al. 2020).

In the Australian context, the heavy reliance on cars to transport children to their daily destinations, including school has been normalised as part of contemporary lifestyles. In terms of honouring children’s needs and rights to access non-car transport modes as part of their daily lives for their physical, psychological, emotional and social development (Freeman 2010, Schoeppe 2013, Ramanathan 2014, Waygood 2017, Gilliland et al. 2019, Keall et al. 2020, Aliyas et al. 2022, Gilbert and te Brömmelstroet 2023), planning and transport policies continue to fail substantively.

Although there are no nationwide datasets regarding the scale of private car usage to transport children to their daily destinations, individual studies and household surveys consistently report that the majority of children’s trips to and from school occur via car. Figures suggest that car usage is most prevalent when accessing primary school (Garrard 2016, The Royal Children’s Hospital National Child Health Poll 2019). In contrast, active transport rates for primary school trips only constitute around 25% of total trips, of which 9 out of 10 are walking (or 22% of the total) (Garrard 2016). In the context of the overall population of children attending a primary school in Australia (2,075,224 children at primary schools at the 2021 Census, ABS 2021c), this translates to over 1.5 million children nationwide who are being driven in cars to access a primary school on school days. Tellingly, these trips are typically within walking (up to 1 km) and cycling (up to 5 km) distances from school (Garrard 2009, Ridgewell et al. 2009, Sharpe and Tranter 2010).

The importance of multifaceted approaches to more effectively reduce the reliance on cars, has long been recognised (Wind 2012, Jensen 2013, Cortes-Morales and Christensen 2014), with various successful policies and programmes such as walking school buses, bike trains and active transport school programmes being facilitated by local governments (for instance Open Streets by Merri-bek City Council or Octopus Schools programme by City of Darebin in Victoria, Australia). However, supportive policy and programme development has been fragmented, with governance arrangements characterised by policy silos that hamper the capacity of such initiatives to create substantial change at scale (Gilbert et al. 2018a).

The concept of the local school has a long history in planning, going back at least as far as Clarence Perry’s Neighbourhood Unit from the 1920s, and despite a century of dramatic changes to the size, composition and structure of the cities in which an increasing majority live, schools continue to be “expected to be local” by communities (Brookfield 2017, p. 51). This is despite the reality that they are not always accessed and utilised at a local level due to the many factors influencing families’ decision-making processes when choosing schools (e.g. school type, curriculum offerings, socio-cultural or religious reasons) (Cucchiara and Horvat 2014, Jackson 2019).

The concept of the local school has also been a focal element within the planning and transport policy realm, with local living policies listing them as a key community facility. However, the simplistic assumption is often that proximity to home on its own will induce access by active transport (Gilbert and Woodcock 2022), while the actual decisions regarding school choice are complex with many factors independent of closeness to home such as traffic safety, school type and so on being relevant. These can then override any affordances for active modes of transport related to point-to-point distance. Proximity considerations become more meaningful when supported by a range of other features that improve the likelihood that any given place will foster active transport use in general since school access is often just one destination among many across children’s increasingly complex activity spaces. In the case of a school, this could be spatial elements, as in the provision of convenient, legible and safe active transport infrastructure; or temporal, such as starting and finishing times for after-school activities that enable access by non-car modes (Gilbert et al. 2018b).

A more sophisticated approach would be premised on a much wider suite of factors that support active travel modes rather than simplistic “buffers” represented by as-the-crow-flies 400 and 800 m circles around key activities. With this, consideration of network-based accessibility for active and public transport for key activity nodes in the pattern of daily life such as shops, sports and recreation, public open space as well as schools would become central to policy and planning, and especially the operation of school zones. To demonstrate the benefits of moving away from a simplistic home-to-school proximity-based approach, this paper combines the social and physical environmental features associated with “child-friendliness”, primarily comprising: positive indicators of child-oriented environmental quality such as “safety and free movement”, “peer gathering places” and “varied activity settings” informed by definitions developed by UNICEF (2013); the work of scholars such as Chawla (2002) and Chatterjee (2005) on children’s positive engagement with places with a diverse range of physical and social settings; the multi-scale importance of child-friendliness from the front door to streets and city-wide public spaces (Woolley and Johns 2001, Oktay 2004, Christensen et al. 2017). Thus, child-friendliness is conceptualised multidimensionally across scales in the present article rather than being framed by the limited scope and scale of child-specific places segregated from everyday urban places. In doing so, the paper advances the framework proposed by Gilbert et al. (2018a) from the theoretical to the practical by quantifying a range of built environment variables of child-friendliness, using a range of spatial databases. This framework utilises the concept of “child-friendliness” to incorporate broader understandings of children’s rights to the city manifested through a range of “agglomeration” effects that arise from the spatial co-presence of activities and sustainable modes of transport that enable children and their caregivers to decide not to drive when making transport decisions. With this, we argue that children have a right to sustainable mobility that is not dependent on access to being transported by car to go about their daily lives; honouring these rights requires more than short distances between a limited suite of child-specific spaces and instead, independent access to a much wider range of urban spaces and experiences.

The paper reports on children’s journeys to school across 10 schools in Melbourne and Adelaide, the capital cities of the Australian states of Victoria and South Australia. To understand how travel behaviour is impacted by the interaction of land use and transport contexts, or “mobility environments” (Bertolini and Le Clercq 2003, Bertolini 2017) related to home and school, a “child-friendliness index” was devised to assess how distance, mode, land use distribution and accessibility could be related.

2. Methods

As part of a broader research project (PhD thesis conducted at the University of South Australia titled “Children’s everyday mobilities and low carbon living”), 296 surveys were completed (response rate of 33%) by children aged 9–18 years, across 6 schools (3 primary, 2 secondary, 1 combined) in Adelaide, South Australia and 4 schools (2 primary, 1 secondary, 1 combined) in Melbourne, Victoria during 2017 and 2018. 84 surveys were also completed by parents of children aged 9–18 years in the same schools (response rate of 28%). Data collection took place between April 2017 and June 2018. In both years, the weather patterns were as usual for each city, though Melbourne experienced below-average rainfall, while Adelaide’s rainfall was close to the average. The temperatures in both cities were slightly above average for both cities (Bureau of Meteorology 2024).

In terms of metropolitan population contexts, Melbourne’s was 4,917,750 and Adelaide’s was 1,387,290 in the 2021 Census (ABS 2021a), reflecting the larger and smaller ends of Australian mainland state capitals. In terms of the travel mode to work, the state of Victoria had a lower car use (53.8%) than both South Australia (68.7%) and Australia as a whole (57%). The share of public transport for work trips was 4.4% for Victoria and 5.3% for South Australia (compared to 4.6% for Australia), while the share of walking and cycling in work journeys was 3% for Victoria and 3.3% for South Australia (compared to 3.2% for Australia) (ABS 2021b).

Schools were selected to represent local catchment areas (preference given to schools with school zoning rules as school zones are not a consistent policy across different types of schools), diverse socio-economic characteristics, based on the Socio-economic Index for Areas Index of Relative Socio-economic Disadvantage (SEIFA) (ABS 2016) and mixed residential density. SEIFA summarises the indicators on a range of economic and social conditions of people and households within an area which measures relative disadvantage. This study was approved by the University of South Australia Ethics Committee and gained the required approval from educational authorities as well as approval from school principals and teachers.

2.1. Distance Between Home and School

In this study, children and parents were asked to provide their home addresses in the form of the street name or closest intersection as well as the suburb. These addresses were geocoded using Google Earth Pro and the resultant keyhole markup language (kml) files were converted to shape files in QGIS software (version 3.8.0). In terms of the geographical information on where children live, the direction function on Google Earth Pro was used to calculate the shortest modal network distance between home and the school they attended for each respondent. This distance was calculated for driving as well as walking which included the pedestrian network. In most instances, driving distances were found to be equal to or longer than the walking distances: out of the 296 participants, there were only four cases (1.4%) where the driving distance was shorter than the walking distance (by up to 0.5 km). Children’s driving distance to school was recoded into two categories for further analyses as follows: 0 = within 4 km, in line with the overall active transport scale (i.e. walkable when less than 2.25 km, cyclable when less than 4 km) (Mandic et al. 2020). The sample included a higher proportion of children living within 4 km of their schools (64.2%) than those who lived further away (32.4%) Travel mode of participant children concerning the distance was given in Figures 1 and 2.

Figure 1. Travel mode to school by distance.

Figure 2. Travel mode from school by distance.

2.2. Child-Friendly Built Environment

The measurements used in this index correspond with the following built environment variables in the operational framework developed by Gilbert et al. 2018a: Land use and built environment, provision of nature, local accessibility and mobility provision of local services. These spatial measures are those previously found to be associated with children’s mobilities as explained in the following section and in Table 1 with the details of sources for each variable listed in Table A-1.

Table 1. Spatial measures of children’s car-based mobilities.

Themes and Variables	Units of Measurement	Method
Active Transport (AT) and Public Transport (PT):
Walkability index	Walkability Index (connectivity, Land Use Mix, population density) on AURIN Portal	Sum scores for all buffers collapsed into deciles and the groups recoded from 1 to 10, with 1 = least walkable and 10 = most walkable
Connectivity (by bike lanes)	Count of three (or more) way path intersections via the connectivity tool on AURIN Portal	Sum scores of the connectivity by bike paths collapsed into deciles and groups recoded from 1 to 10, with 1 = least connected and 10 = most connected
PT stops	Count of PT stops (bus, tram, train)	Results collapsed into deciles and groups recoded from 1 to 10, with 1 = lowest 10 = highest number of PT stops.
TRAFFIC SAFETY:
Crash data	Count of traffic crashes for each participant student’s home-based buffer	Results collapsed into deciles and groups reverse recoded from 1 to 10, with 1 = highest and 10 = lowest number of crashes.
Traffic lights	Count of traffic lights	Results collapsed into deciles and groups recoded from 1 to 10, with 1 = least and 10 = highest number of traffic lights.
Traffic exposure	Ratio of high order roads (301, 302, 303, 304) to low order roads (305, 309, 400)	Scores collapsed into deciles and groups reverse recoded from 1 to 10, with 1 = highest and 10 = lowest traffic exposure.
PROXIMITY TO LOCAL SERVICES AND GREEN SPACES:
Schools		Results for each collapsed into deciles and groups recoded from 1 to 10, with 1 = least and 10 = highest number of the service/space.
Libraries	Count of schools, childcare centres, libraries, health services and green spaces.
Health services
Childcare centres
Green spaces (playgrounds, parks, community gardens, reserves)
Retail at street level (eyes on the street)	Ratio of the total commercial area (e.g. commercial, mixed use, community facilities and activity centres) to the total area of each buffer area (home and school based)	Groups recoded from 1 to 10, with 1 = smallest and 10 = largest amount of local destinations.
DEMOGRAPHIC DIVERSITY:
Housing density	Ratio of the total number of dwellings to the total buffer area (hectare)	Results collapsed into deciles and the groups recoded from 1 to 10, with 1 = lowest and 10 = highest density.
Housing mix	The diversity index in the AURIN portal	Results collapsed into deciles and groups reverse recoded from 1 to 10, with 1 = highest diversity and 10 = lowest diversity.
Child population density	Ratio of the total number of children aged 0–19 to the total population	Results collapsed into deciles and groups recoded from 1 to 10, with 1 = lowest and 10 = highest child population density.

2.2.1. Active Transport Infrastructure

2.2.1.1. Walkability

The findings from previous studies both for children and adults indicate that connectivity of streets, population density and diversity of land uses (land use mix) in a neighbourhood promote increased rates of walking and improved accessibility to local destinations (Sallis 2006, van Loon 2014, Buck et al. 2015, Moudon 2006, Dovey and Pafka 2020). The walkability index tool within the Australian Urban Research Infrastructure Network (AURIN) portal (AURIN 2019) which combined all of these three variables was used to calculate a walkability index for each participant’s home-based and school-based buffer. The databases and the categories used for these composite elements within this tool were:

• Land Use Mix: Mesh Block 2016, land use categories of residential, commercial, education, hospital/medical and parkland.

• Population: Mesh Block 2016¹, total resident population

• Network: OpenStreetMap (OSM) 2018 streets

2.2.1.2. Bike and Public Transport Accessibility

In order to estimate the provision of cycling infrastructure, the bike network data was used to calculate the number of intersections to show the potential mobility of cyclists within the neighbourhood using the connectivity tool within the AURIN portal. For public transport accessibility, the number of stops (and stations) for bus, tram and train network were calculated.

2.2.2. Traffic Safety

Concern about traffic safety is one of the key factors contributing to reduced walking and cycling rates amongst school-aged children (Carver et al. 2013, Kottyan et al. 2014). In order to assess the actual likelihood of traffic danger, crash statistics and traffic exposure data were used. Crash statistics are commonly used to analyse road safety (BITRE 2015, Beck et al. 2016, State Government of Victoria 2019, New South Wales Government 2020). In addition, the number of traffic lights was counted to estimate the provision of traffic calming measures. The provision of traffic lights to assist crossing was previously found to be in positive association with the sense of safety amongst parents in relation to their children’s non-car-based mobilities (Veitch 2017) and similarly have been found in general to contribute to stronger associations of safety with urban street environments (White 2021).

2.2.2.1. Traffic Exposure

In order to estimate the level of traffic exposure, the ratio of the total length of main roads to the length of local roads was used. This method was adapted from previous studies that used the road hierarchy to calculate the traffic exposure in Western Australian context (Giles-Corti et al. 2011). The transport hierarchy within the Public Sector Mapping Agencies (PSMA) (2019) street network is used to identify higher-order (high traffic) and lower-order (low traffic) roads²:

• Higher order = 301, 302, 303 and 304

• Lower order = 305, 309 and 400

Using the QGIS software (sum of lines in a network buffer), the total length of these roads within each home and school based buffer was calculated.

2.2.2.2. Proximity to Local Services and Green Spaces

Selected basic services that facilitate the everyday life of families with children were counted for each participant’s home-based buffer. These services were schools, childcare centres, libraries, health services and public open spaces. Public open spaces are important local destinations for families with children as well as contributing to an attractive public realm (Broberg and Sarjala 2015, Van Kann 2015).

2.2.3. Demographic Diversity

This segment of the index includes data on housing density, housing diversity and child population density.

2.2.3.1. Housing Density

Housing density is one of the most commonly used variables in relation to environments promoting the use of non-car travel modes and increased levels of physical activity (Buck et al. 2015, Oliver 2015, van Leslie et al. 2007, Loon and Frank 2011). In this study, the housing density for each buffer area was calculated through the ratio of total number of dwellings to the total buffer area (hectare).

2.2.3.2. Housing mix

In recognition of the link between housing density and diversity, for instance, low-density housing areas generally not conducive to active and public transport use and growing preference for non-detached houses by families with children (Easthope and Tice 2011), “housing mix”(mix of different house types) has been added to the CFI. The diversity index in the AURIN portal was used to calculate the degree of housing diversity for each buffer area. This index computes a measure of the degree of diversity across different housing types such as separate, semi-detached, row terrace house, townhouses, flats and apartments. Smaller scores indicated an increasing degree of diversity.

2.2.3.3. Child Population Density

Given the population density was already included in the walkability index tool within the AURIN portal, only the child population density has been added to the demographic diversity component of the CFI. As per the previous studies (Cortes-Morales and Christensen 2014, Rothman 2014) on the importance of other children’s presence as part of children’s everyday mobilities, in this study, it has been hypothesised that a higher number of children in the neighbourhood creates better opportunities to interact with other children and to make friends as part of children’s everyday mobilities.

2.3. Calculation of Spatial Unit for Analysis: Network Buffers

In recognition of the importance of children’s membership of multiple groupings beyond their home as highlighted by ecological systems theory (Bronfenbrenner 1979) and the importance of local environments surrounding both home and school when assessing the environmental correlates of children’s active transport (Easton and Ferrari 2015, Larouche 2018), the buffers used for the CFI included both home-based and school-based areas. This method was proven useful for detecting associations in relation to after-school trips (from school) that would not have been captured through the inclusion of home-based local environments only.

Conceptualising individuals’ local environments in the context of measuring certain activities (e.g. physical activity levels) within their overall activity spaces is challenging as the full geographic extent often reaches well beyond local environments associated with school and home (Kwan 2012). However, our travel patterns and their spatial scales, invariably do include the local environments around our daily spaces and so, network-based buffers offer a more consistent and accurate scale for examining travel patterns. In the context of this study, the buffers related to a short walk from home and school represent the environments to which children are regularly exposed and are therefore useful in measuring the provision of child-friendly infrastructure and services.

The use of network-based buffers is useful in overcoming another issue with the conceptualisation of local environments, that is the misalignment between the areas on which data releases are based and the actual activity spaces. For example, the usual methods for such conceptualisation are often influenced by the availability of data for individual administrative units such as suburbs or other predetermined statistical areas. However, these administrative units rarely align with the individuals’ actual lived local environments and so illustrate the need for alternative scales and zones.

The neighbourhood generator tool within the AURIN portal was used to create the 1000 m network buffer for each participant’s home distance representing a generic 10 to 15-minute walk or 5-minute bicycle ride. This tool, rather than radial buffers, produces buffers for a street network using the maximum distance to travel along the network and a buffer value that determines the width of the buffer around the traversed network segments.

The buffer size of 1000 m is commonly used in studies related to children’s mobilities in Australia and overseas (De Meester et al. 2014, Oliver 2015, Lin et al. 2017, Williams et al. 2018). An example of school and home based buffers along with school zones is given in Figure 3.

Figure 3. Home and school based buffers based on participant children’s residential addresses (Note: School 1’s 1 km network buffer and school zone as shown in blue)

The OpenStreetMap (OSM) network dataset (2018) was used to measure the walkable catchment area within 1000 m of each participant’s home location. The information on the home addresses was sought both from the children’s and parents’ surveys. In total, 286 buffers were created with missing information from 10 participants (total n = 296). The same process was repeated for the schools (1000 m network buffer around schools) where participant children were attending to develop school-based indices.

2.4. Development of a Child Friendliness Index

The Child Friendliness Index (CFI) was calculated by summing the results for each measurement listed above both for the home-based and school-based buffers. The total score (1–10) for each measurement was summed for each buffer area ranging between 15 (lowest child-friendliness) and 150 (highest child-friendliness). The resulting child-friendliness indices (home-based and school-based) were further classified into three equal groups to rescale the results from low, medium and high child-friendliness. These results were used when analysing the relationship between the travel mode reported by participant children and parents.

In the absence of an evidence base indicating the differential impacts of individual indicators used, for this study, we used an equal weighting approach. Equal weighting is a commonly used approach for numerical indices due to the wide-ranging complexities of the development of weighted indices, for example in the absence of a single robust method that eliminates bias from the process (Greco et al. 2019, Gómez-Limón et al. 2020).

To date, no single approach has been developed to address such challenges in transport indices. One strategy which involves drawing on the experiences of practitioners and researchers assesses the relevance of each factor based on its impact on walkability, essentially determining its ability to predict the level of walkability. Some studies employing this kind of approaches in creating weighted walkability indices have assigned higher importance to factors like land use diversity and street connectivity compared to other elements within the built environment (Frank et al. 2010, Al Shammas and Escobar 2019). However, complexities arise due to contextual differences. For instance, in a study on the determinants of walkability in the Netherlands, Liao and colleagues (2020) found that elements related to the environment's attractiveness, such as the presence of natural features which are often overlooked in such indices, were more salient than factors like land use mix.

Another layer of complexity with such indices is the presence of collinearity where some independent variables are correlated, potentially leading to double counting. In the case of the CFI, this issue may arise concerning the factors of traffic exposure measured by the ratio of high-order roads to low-order roads and the crash statistics which are likely to increase with the higher presence of high-order roads.

The challenge of ranking the importance of contributing factors to active and public transport usage is further complicated by the presence of subjective elements, such as perceptions of safety in specific streets or places. This is particularly the case with decision-making processes around children’s daily movements, as perceptions of safety both by parents and children significantly influences travel behaviour (Aliyas et al. 2022, Gilbert et al. 2022, Swain 2023).

2.5. Dependent Variable: Children’s Mobilities

Children were asked to report the travel mode to and from school on the day of data collection. The response categories included, walk, ride a bike, bus, train, tram and car. The responses were recoded at 1 = active transport, 2 = public transport and 3 = car.

3. Data Analysis

The chi-square test for independence was used to identify statistically significant differences between the variables. Since the categories in each variable are more than those in a 2 × 2 table, the post-hoc tests were used to identify which combinations of results were statistically significant. The first step of the post-hoc test for chi-square analysis is the calculation of the adjusted p value (Bonferroni adjusted alpha value) which sets an adjusted significance level in order to make multiple comparisons. Calculated by dividing the standard p value (0.05) by the number of comparisons or tests, this value produces a more stringent p value for each comparison which is critical to prevent the possibility of a Type 1 error due to the large number of comparisons being made (Pallant 2011, Chen et al. 2017).

The second step in conducting the post-hoc tests was the calculation of the adjusted residual values (z scores) for each comparison or analysis being made and the individually adjusted p values for each analysis.³

4. Results

4.1. Distance and Private Car Usage

In order to identify to what degree the travel modes to and from school reported by children on the day of data collection were dependent on the distance between home and school, chi-square tests for independence were carried out for the travel modes (AT, PT and Car), both to and from school on the day of data collection. The initial chi-square test for independence results indicated that there was a significant difference in children’s travel mode to and from school on the day of data collection χ² (2, n = 285) = 54.04, p < 0.05 to school and from school χ² (2, n = 286) = 42.48, p < 0.05.

After the adjustment of residual values and individual p values (Bonferroni correction), the significance was observed between distance and active transport as well as public transport while no significant difference between distance and car usage (p < 0.0083). Using the adjusted residual values and adjusted individual p values (Bonferroni correction), for the distance within 4 km, there was no significantly lower car usage than expected by chance while there was no significantly higher car usage than expected by chance for the distance longer than 4 km (p < 0.0083). Comparing car and non-car modes to each other at the shortest distance (<4 km), there was no significantly higher AT or PT usage than car usage (Table A-2).

4.2. Child Friendliness and Private Car Usage

In order to identify if the travel modes to and from school reported by children on the day of data collection were dependent on the child-friendliness of their immediate neighbourhood and school, chi-square tests for independence were carried out for travel modes (AT, PT and Car), both to and from school on the day of data collection. All variables were assessed against home-based and school-based child-friendliness.

The chi-square test for independence results indicated that travel modes to and from school on the day of data collection were not independent of the child-friendliness (low, medium and high) of the area around home and school (p < 0.05). The results for each test, travel mode and home-based CFI and travel mode and school-based CFI, are explained in the next two sections.

4.2.1. Travel Mode and Home-Based CFI

A chi-square test for independence indicated a significant difference in children’s travel mode to school on the day of data collection (1 = AT, 2 = PT and 3 = Car), in relation to home-based child-friendliness (1 = low, 2 = medium, 3 = high), χ² (4, n = 285) = 34.18, p < 0.05 and from school χ² (4, n = 286) = 33.11, p < 0.05.

Using the adjusted residual values and adjusted individual p values (Bonferroni correction), at the low level of home-based child-friendliness (CFI_home = 1), there was significantly higher car usage than expected by chance (both to and from school), conversely significantly lower AT (to school) and PT usage (to and from school) than expected by chance (p < 0.0056). At the high level of child-friendliness (CFI_home = 3), there was significantly lower car usage than expected by chance and significantly higher AT usage than expected by chance both to and from school (Table A-3).

4.2.2. Travel Mode and School Based CFI

A chi-square test for independence indicated a significant difference in children’s travel mode to school on the day of data collection (1 = AT, 2 = PT and 3 = Car), in relation to school-based child-friendliness (1 = low, 2 = medium, 3 = high), χ² (4, n = 293) = 53.81, p < 0.05 and from school χ² (4, n = 294) = 52.95, p < 0.05.

Using the adjusted residual values and adjusted individual p values (Bonferroni correction), at the low level of school-based child-friendliness (CFI_school = 1), there was significantly higher car usage than expected by chance, conversely significantly lower PT usage than expected by chance both to and from school, with no significant difference in AT usage (p < 0.0056). By contrast, at the high level of child-friendliness (CFI_school = 3), both to and from school, there was significantly lower car usage than expected by chance and significantly higher PT usage than expected by chance; however, no significant difference was found for AT usage (p < 0.0056). There were no differences in the travel modes at the medium (CFI_home = 2) level of child-friendliness (Table A-4).

5. Discussion

The results from the spatial analysis demonstrated that children’s travel modes to and from school vary depending on the degree of child-friendliness in their home-based and school-based local environments with higher car usage and lower non-car usage for children living or attending a school in an area with a relatively lower level of child-friendliness.

The CFI in this study combines measures of the social and built environment which are associated with child-friendly, sustainable mobility. The combination of such correlates in a single index and its effectiveness in promoting sustainable modes of transport in accessing schools aligns with the recognition of the multifaceted changes that are required to counteract the dominance of cars in children’s mobilities. Against the background of the high proportion of short car trips in accessing schools, it is evident that proximity alone does not translate to a sustainable “school journey”, as the private car usage remains high regardless of the distance. The use of composite indices such as the CFI described here can prove useful in assessing the performance of children’s local environments in enabling the conversion of these short car trips to non-car modes and in actualising the proximity-related affordances. The CFI also expands our understanding of the complexities of travel patterns and the importance of addressing children’s specific needs and rights to use sustainable modes of transport as it represents a shift from the stand-alone objective-based guidelines (e.g. proximity to school) to a set of minimum standard-based guidelines. It offers an alternative technique to assess the performance of neighbourhoods and school catchments from the point of view of accessibility for children, a critical factor in achieving the widely but loosely defined concept of “local school”, the centrepiece of all local living policies (Gilbert and Woodcock 2022). The reliance on simplistic proximity-based approaches in local living policies is unlikely in the majority of instances to achieve meaningful outcomes because the complexities of achieving a “local school” involve so many more factors. Examples of these problematic approaches can be seen in way the “school travel plans” required in many Australian states for new schools have highly ambitious active travel mode shares but are produced independently from the land use, transport planning and urban design schemes for the new housing estate catchments the schools are to serve. For instance, the traffic report for Sydney Olympic Park new high school indicates a target of 73% for active transport mode (New South Wales Government 2024), noticeably higher than the Australian and New South Wales averages (e.g. around 25%).

By contrast the CFI proposed here specifically considers the factors at stake in affording sustainable accessibility for children in neighbourhoods and school catchment areas, enhances our understanding of the multidimensional factors that make a “local school” and provides concrete pointers towards the specific actions and interventions required to increase the likelihood of achieving such schools.

The index can also be applied to the catchments of other daily destinations for children, such as venues for extra-curricular activities, in supporting the idea that it is critical to develop clear and precise policies aiming to create a wider range of environments within the sustainable transport reach of children. This will constitute an important step toward the recognition of children’s rights to sustainable transport (Whitzman et al. 2010, Gilbert 2019) when accessing their daily destinations and much-needed attention to non-instrumental enablers and constraints (Gilbert et al. 2022). Foregrounding child-friendliness requires considering what might be referred to as the “agglomeration effects” of certain combinations of land use and sustainable transport that impact the decision-making contexts for children’s school journeys, and all the other activities in children’s lives where they ought not to be dependent on cars to engage with them.

Complementing other indices such as the school walkability index developed by Giles-Corti et al. 2011 and by Christiansen et al. 2014, which are associated with increased rates of active transport, the CFI captures additional factors, including various demographic variables, cycling and public transport accessibility, which all influence children’s everyday mobilities. The CFI can also be linked to other conceptually driven frameworks that assess neighbourhoods from the health and wellbeing point of view, such as the Neighbourhood Early Childhood Development Framework developed by Badland and colleagues (2023).

Despite the rigorous selection and analysis of the measures in developing the CFI, the index can be further improved and expanded to include additional measures: the more elements an index has, the greater potential there is for different aspects of performance to be analysed for the effects on children’s mobility patterns, and the more useful it is likely to be for policymakers, planners and urban designers. For example, the analysis of fine grain fabric of the built environment, including factors such as intersection density, ratio of T-intersections to crossroads, street widths, prevalence of pedestrian crossings (and by type), safe cycling infrastructure, the cleanliness of public places (e.g. sweeping schedules for specific areas), more fine-grained data on active frontages relative to total street length, actual non-residential land use diversity and data to indicate the provision of soft edges, footpath widths and conditions, street trees, shading and weather protection that promote further foot traffic and increased passive surveillance from well-used public spaces.

To enhance the accuracy of the CFI, future considerations could include assigning weights to individual factors within it. For example, given the evidence around the significance that parents attribute to “personal and traffic safety”, future iterations of the CFI could benefit from not only assigning greater weight to safety measures such as the volume and speed of vehicle traffic, the condition and distribution of footpaths and bike lanes and presence of safe crossings but also take into account community perceptions of safety as evidenced through media reporting and opinion surveys about the places in question (te Brömmelstroet 2020).

In relation to the use of crash statistics in developing the CFI, the underlying assumption was that fewer crashes meant relatively safer environments for pedestrians and cyclists. However, this assumption does not consider the possibility of the subject location being extremely dangerous and therefore being avoided by pedestrians and cyclists. Because crash statistics databases rely on police officer generated reports, it is important to acknowledge the inherent biases within these crash datasets, particularly concerning the underreporting of the number and types of injuries sustained by non-car users (Poulos 2015 Shinar 2018, Oxley 2020,). Future studies could use an approach where the crash dataset is checked against the datasets on a range of built environment variables such as traffic volume, vehicle speed, and provision of footpaths and bike lanes. Furthermore, future studies assessing the social and spatial determinants of private car usage in transporting children to their everyday destinations could benefit from the use of the CFI in investigating differences in socio-economically advantaged and disadvantaged areas.

A note of caution must be sounded, however. While the rise of urban analytics based on digital data has greatly enabled certain types of policy-relevant knowledge about cities to be gained, there are other trends and countervailing forces when it comes to the potential to create “local schools”. For example, the neoliberal ideology of “choice” that increasingly dominates education, and in particular primary and secondary schools tends to change the spatial dynamics of mobility and land use distribution in ways that may greatly complicate the ability to plan the creation of more integrated sustainable mobility environments for children. For example, the expansion of the independent school sector to cater to demands for greater “choice” has been shown to result in increased travel to school distances and accordingly, journeys by car (Yan et al. 2019). In the state school sector, selective entry programmes could have similar effects. Likewise, the proliferation of ethno-religious independent schools that are an increasing element of contemporary multicultural Australia form part of this cohort. As the CFI specifically looks at neighbourhood characteristics at both ends of the home to school journey, no matter how spatially proximate or distant they are, it could also be deployed to better understand the extent to which such social dynamics are impacting the creation of child-friendly contexts in Australia and elsewhere.

6. Conclusions

As Kullman (2010) puts it, a journey to school (or any other destination for a child) is more than simply transport as it offers many ways for children to relate to their everyday environments, including feelings of “self-possession” and as “an opportunity to make their own decisions” (p. 834). Similarly, children’s emotional and embodied engagements are not just with their destinations but with the various social, natural and built elements of their journeys (Porter et al. 2021). Integrated transport and planning policies and programmes that aim to create sustainable journeys to children’s everyday destinations such as schools have much to gain from a consideration of more all-encompassing metrics to assess the performance of these places in promoting sustainable mobility. The CFI could be a good starting point to de-emphasise instrumental aspects such as the use of proximity as a stand-alone objective and recognise the importance of other factors to unlock the full potential of “proximity”. Such an approach would also promote a move towards more elaborate solutions to complex problems such as the heavy reliance on cars when transporting children, require. The recognition of children’s right to access their daily destinations without being dependent on cars needs to be at the core of such approaches. By combining a myriad of factors that are required to create a child-friendly environment, the CFI serves as a crucial step towards displaying this recognition.

Acknowledgement

The first author also extends her gratitude to Andrew Allan, Carolyn Whitzman, Stephen Berry and Johannes Pieters for their supervision throughout this PhD project.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Research Node for Low Carbon Living, University of South Australia as part of a PhD scholarship provided by the CRC Low Carbon Living Limited.

Notes

1 Mesh blocks which contain between 30 and 60 dwellings, are the smallest geographical area defined by the Australian Bureau of Statistics (ABS) (ABS 2016)

2 301: National or State Highway, 302: Arterial road, 303: Sub-arterial road, 304: Collector road, 305: Local road, 309: Access road, 400: Pedestrian thoroughfare access

3 The function for “one tailed probability of the chi-squared distribution” in Excel (2010) [CHISQ.DIST.RT (z², deg_freedom)] is used for the calculation of these individually adjusted p values for each comparison (Carlberg 2011) and used to assess the statistical significance.

APPENDIX

Table A-1. Databases for each variable used.

Attributes	Databases
AT and PT	South Australia	Victoria
Walkability index (connectivity, LUM, pop'n density)	OpenStreetsMap – Lines (Australia) 2018; ABS, Mesh Block 2016 Census for Australia
Connectivity (by bike lanes)	Government of South Australia, SA Bike Direct 2013	The State of Victoria, Vicmac Transport – Bicycle Paths 2018
PT stops (tram, train and bus)	Government of South Australia, Adelaide Public Transport Stop Data 2019	Public Transport Victoria (PTV) Metro Bus Stops, PTV Metro Tram Stops, PTV Metro Train Stations 2019
TRAFFIC SAFETY
Crash data	Government of South Australia, Road Crashes 2013–2017	VicRoads – Crashes (Point) 2012–2016
Traffic lights	Government of South Australia, Traffic Signals 2015	The State of Victoria, Traffic Signals 2019
Traffic exposure: ratio of high order roads (301, 302, 303, 304) to low order roads (305, 309, 400)	PSMA Australia Limited, PSMA Street Network, 2018
PROXIMITY TO LOCAL SERVICES & GREEN SPACES
Schools (excluding special schools)	Australian Curriculum, Assessment, Reporting Authority, School Profile 2008–2016
Libraries	Government of South Australia, LGA SA Libraries 2014	Department of Premier and Cabinet, Public Libraries of Victoria, 2019
Health services (GP clinic, dental clinic and hospital)	Healthdirect Australia Ltd, National Health Services Directory 2018
Childcare centres	Australian Children's Education and Care Quality Authority (ACECQA), 2017: National Register Approved Providers in Australia
Local shops	Government of South Australia – Department of Planning, Transport and Infrastructure, 2014: Land Development Zones 2013 for South Australia	The State of Victoria, Department of Environment, Land, Water & Planning 2018: Vicmap Planning – Planning Scheme Zones
Green spaces (playgrounds, parks, community gardens, reserves)	PSMA Australia Limited, PSMA Greenspace (Point), 2018
DEMOGRAPHIC DIVERSITY
Housing density	ABS, SA1 – G32 Dwelling Structure – Census 2016
Housing mix (mix of different house types e.g. separate, semi-detached, flat/apartment, etc.)	ABS, SA1 – P31 Dwelling Structure – Census 2016
Child population density	ABS, SA1 – G01 Selected Person characteristics by Sex – Census 2016

Table A-2. Pearson’s chi-square test results for travel mode to/from school and distance.

		to school		from school
Travel mode		within 4km	more than 4km	within 4km	more than 4km
AT	Count	66a	2b	66a	3b
	Expected Count	45.3	22.7	45.8	23.2
	% within Distance	34.7%	2.1%	34.7%	3.1%
	Adjusted Residual	6.1	−6.1	5.9	−5.9
	p value	0.0000	0.0000	0.0000	0.0000
PT	Count	20a	37b	31a	39b
	Expected Count	38.0	19.0	46.5	23.5
	% within Distance	10.5%	38.9%	16.3%	40.6%
	Adjusted Residual	−5.7	5.7	−4.5	4.5
	p value	0.0000	0.0000	0.0000	0.0000
Car	Count	104a	56a	93a	54a
	Expected Count	106.7	53.3	97.7	49.3
	% within Distance	54.7%	58.9%	48.9%	56.3%
	Adjusted Residual	−0.7	0.7	−1.2	1.2
	p value	0.4995	0.4995	0.2433	0.2433

Each subscript letter denotes a subset of Distance categories whose column proportions do not differ significantly from each other at the .05 level, *significant at the Bonferroni adjusted alpha value of 0.0083

Table A-3. Pearson’s chi-square test results for travel mode to and from school (AT, PT and car) on day of data collection (reported by children) and child-friendliness (home-based)

		Travel mode to school this morning			Travel mode from school to home today
CFI_home		1 AT	2 PT	3 Car	1 AT	2 PT	3 Car
1	Count	11a	8a	74b	16a	9a	68b
	Expected Count	22.2	18.6	52.2	22.4	22.8	47.8
	Adjusted Residual	−3.3	−3.3	5.5	−1.9	−4	5.1
	p value	0.0009*	0.0008*	0.0000*	0.0575	0.0001*	0.0000*
2	Count	24a	23a	49a	20a	31a	46a
	Expected Count	22.9	19.2	53.9	23.4	23.7	49.9
	Adjusted Residual	0.3	1.2	−1.2	−1	2.1	−1
	p value	0.7475	0.2338	0.2164	0.3206	0.035	0.3352
3	Count	33a	26a	37b	33a	30a	33b
	Expected Count	22.9	19.2	53.9	23.2	23.5	49.3
	Adjusted Residual	3	2.1	−4.3	2.9	1.9	−4.1
	p value	0.0030*	0.0331	0.0000*	0.0040*	0.0582	0.0000*

Each subscript letter denotes a subset of travel mode to school this morning AT, PT, Car categories whose column proportions do not differ significantly from each other at the 0.05 level, *association is significant at the 0.0056 level.

Table A-4. Pearson’s chi-square test results for travel mode to school (AT, PT and car) on day of data collection reported by children and child-friendliness (school-based).

		Travel mode to school this morning			Travel mode from school to home today
CFI_school		1 AT	2 PT	3 Car	1 AT	2 PT	3 Car
1	Count	18a	2b	85c	23a	4b	78c
	Expected Count	24.4	22.2	58.4	24.6	26.8	53.6
	Adjusted Residual	−1.8	−6	6.5	−0.5	−6.4	5.9
	p value	0.0661	0.0000*	0.0000*	0.6371	0.0000*	0.0000*
2	Count	17a	25a	40a	17a	26a	39a
	Expected Count	19	17.4	45.6	19.2	20.9	41.8
	Adjusted Residual	−0.6	2.4	−1.5	−0.7	1.5	−0.7
	p value	0.5313	0.0148	0.1412	0.4909	0.1295	0.4605
3	Count	33a	35a	38b	29a	45a	33b
	Expected Count	24.6	22.4	59	25.1	27.3	54.6
	Adjusted Residual	2.4	3.7	−5.1	1.1	4.9	−5.2
	p value	0.0156	0.0002*	0.0000*	0.2662	0.0000*	0.0000*

Each subscript letter denotes a subset of travel mode from school to home today categories whose column proportions do not differ significantly from each other at the 0.05 level. *association is significant at the 0.0056 level.