Getting to school on your own: the correlates of independence and modes of active school travel

ABSTRACT

Young people’s decreased active and independent travel to schools has prompted many countries to devise various policies, initiatives, and programs to counter the associated health detriments. Meanwhile, scholarly studies have identified how children were walking and biking to school benefits physical and mental health, social and cognitive development, and local government finances. Contributing to the broader spectrum of academic research concerning active travel to school, this study explored independent and active school-travel correlates and analyzed the difference between walking and biking. Survey responses from 367 children in North Carolina indicated that walking was sensitive to mixed land use and positive utility while biking was more connected to physical settings. Perceived environmental safety influenced independent active travel, indicating the need for future programs and initiatives to take different actions when targeting modes and independence of active travel to school.

KEYWORDS:

• Independent active travel to school
• neighborhood design
• mixed land use
• sidewalks
• positive utility
• subjective safety settings

1. Introduction

In many countries, there has been a decline in children’s active and independent travel to school—i.e., walking and biking. According to Aranda-Balboa et al. (2020), children’s active school travel between 2001 and 2011 reduced from 44% to 21% of total school travel in the United States, from 39% to 31% in Canada, and from 71% to 62% in England. However, the benefits of active school travel—including physical (e.g., reduced risk of cardiovascular disease, high cholesterol, hypertension, and obesity), social (e.g., increased interaction, urban vitality, and community connections), and mental (e.g., increased positive emotions, enjoyment, and self-esteem) – have prompted the implementation of many interventions to promote active travel (Pang, Kubacki, and Rundle-Thiele 2017; Stark et al. 2018). For instance, the Safe Routes to School (SRTS) program, funded by the U.S. Department of Transportation, received $612 million between 2005 and 2009 for investment in infrastructure and non-infrastructure components such as tools, safety education, and incentives (Safe Routes Partnership 2021). This program was replaced by the Transportation Alternatives Program in 2012. McDonald et al. (2016) estimated that approximately $30 billion is spent annually on motorized transportation to and from schools, which includes school district expenditure, private sector vehicle operating costs and travel time, and other externalities (i.e., traffic congestion, air pollution, and green gas) in the United States. Consequently, they argued that promoting non-motorized active travel to school would help the social economy by reducing the need for bus or car transport, and the cost savings would outweigh the expense of walking and bicycling infrastructure.

Given the recent growing interest in children’s active school travel (AST), we attempted in this study to explore the factors that relate to (1) overall AST and independent (without parents or guardians) AST (IAST); and (2) walking AST and biking AST. Using survey responses from 367 4th and 5th-grade students in North Carolina, we considered the roles played by different correlates, including subjectively and objectively measured aspects of the built environment (B.E.), demographic characteristics, automobile possession, BMI, physical activity (P.A.), destination, equipment, parental permission, and attitude toward AST and IAST. Zero-inflated Poisson, zero-inflated negative binomial, and negative binomial analyses were conducted. The outcomes of the study are expected to guide decision-makers in situating new schools, implementing neighborhood designs, promoting social support systems, and encouraging educational institutions to promote AST, especially IAST, and walking as well as biking.

2. Children’s active travel to school

In middle childhood (aged 6–12 years), children begin to venture away from home independently to explore their environment. This discovery and learning process enables children to overcome challenges and handle appropriate risks (NLI 2010). The level of a child’s independent mobility – especially independent active travel using non-motorized transportation – is conceptualized by three major factors: 1) territorial range (Hart 1979; Moore 1986; van Vliet 1983); 2) freedom, ability, or license to move around independently (Hillman, Adams, and Whitelegg 1990; O’Brien et al. 2000; Page et al. 2009); and 3) the frequency (Prezza et al. 2009), route, and destinations pertaining to their independent outdoor movement (Mavoa et al. 2011; Prezza et al. 2009). For example, independent active travel increases when the geographical distance from a child’s home to places for playing and socializing is closer, and when parents perceive their neighborhood to be safe (Zougheibe et al. 2021) and permit the child to move independently (Trapp et al. 2011). In addition, independent active travel increases when a child has more experience with non-motorized travel to school and child-friendly environments such as green spaces, low traffic areas, or low-density buildings (Kytta 2004).

Previous studies have empirically determined that independent active travel supports children’s physical, social, cognitive, and emotional development (Kytta 2004), with spontaneous outdoor play influencing motor development and physical health (Armstrong 1993; Page et al. 2010). Children exposed to more independent active travel acquire superior spatial skills and play with peers more often, both indoors and outdoors (Rissotto and Giuliani 2006). Moreover, independent active travel can impact emotional bonds with both peers and the natural environment (Bixler, Floyd, and Hammitt 2002). Meanwhile, several systematic reviews have found a positive association between AST and various P.A. outcomes (Faulkner et al. 2009; Larouche et al. 2014; Lee, Orenstein, and Richardson 2008; Lubans et al. 2011). While independence is considered beneficial for increasing AST, research has also demonstrated that being escorted while walking might promote AST, and that children escorted by an adult possess greater environmental knowledge than children who travel to school unescorted (Joshi, MacLean, and Carter 1999). By acknowledging the value of escorted AST, interventions such as “walking school buses” (Pang, Kubacki, and Rundle-Thiele 2017; Smith et al. 2020) (where parents or other volunteers take turns escorting children by walking or riding bicycles as a group) have attempted to shift school travel behavior from driving to escorted and independent walking.

Acknowledged possible correlates of independent active travel include the child’s demographic characteristics (e.g., age, gender, and ethnicity), parental characteristics, and B.E. Notably, the choice of travel mode is strongly influenced by age, with older children more likely to use non-auto modes and less likely to be escorted to or from school by their parents (McDonald 2008b; McMillan 2005; Vovsha and Petersen 2005). Moreover, boys have been found to be granted more freedom at an earlier age than girls (Moore 1986; van Vliet 1983), with parents more willing to allow boys to be outdoors on their own or with peers and more likely to permit circulation at greater distances from home (Timperio et al. 2006). However, regardless of parental influence, girls have been found to be less likely to walk or bike (Evenson et al. 2006, 2003; Marzi, Demetriou, and Reimers 2018; McDonald 2008b). A systematic literature review by Marzi, Demetriou, and Reimers (2018) looked at characteristics connected to independent mobility for boys and girls individually, taking into account sex/gender variations. According to this review, sex/gender inequalities in mobility are regularly observed, with girls being less active or mobile than boys. According to the 2001 Nationwide Household Transportation Survey, White respondents made more automobile trips and fewer walking trips than other ethnic groups, and biking was highest among White (Larouche et al. 2016) and Hispanic respondents (Pucher and Renne 2003). This is not only a matter of ethnicity but also of socioeconomic status, which incorporates automobile access, bicycle ownership, and household income (McDonald 2008a). Thus, neighborhood-level socioeconomic status was associated with AST, with students attending schools with the higher socioeconomic status being significantly less likely to travel actively than those attending schools with lower levels (Ikeda et al. 2018b). Car availability has been negatively associated with the probability of walking or biking (Heinen, Van Wee, and Maat 2010; Mitra 2013), and bicycle availability has been positively associated with biking (Fraser and Lock 2011; Handy, Van Wee, and Kroesen 2014; Heinen, Van Wee, and Maat 2010). A study by Larouche et al. (2016) explored the relationship between children’s demographic characteristics and their preference for either walking or biking separately, suggesting that factors such as higher P.A. and greater independent mobility were associated with children’s preference for cycling.

Several studies have only highlighted either the presence or absence of adult AST (Christensen et al. 2011; Faulkner et al. 2010), with other studies emphasizing how parental decision-making processes are impacted by parental characteristics. Parental employment status (He 2013; Yarlagadda and Srinivasan 2008), work flexibility (Yarlagadda and Srinivasan 2008), work arrangement and workplace location (He and Giuliano 2017), working hours (Motte-Baumvol, Bonin, and Belton-Chevallier 2017), and intrahousehold interactions between adults and children (Gupta et al. 2014; Vovsha and Petersen 2005; Yoon, Doudnikoff, and Goulias 2011) have also been acknowledged to be significant. For example, children and teenagers are less likely to walk to school when parents travel to work in the morning (McDonald 2008b), and mothers who are employed full-time are less likely to walk their children to school compared to part-time workers and non-workers (Yarlagadda and Srinivasan 2008). A systematic meta-reviews done by Riazi et al. (2022) introduced the “perceived competence” category to incorporate children’s perceptions of their safety experience, maturity, and confidence related to independent mobility. The freedom to be independently mobile was continuously positively connected with parents’ perceptions of their child’s confidence and attitude toward independent mobility, according to the study. Marzi, Demetriou, and Reimers (2018) also indicated that the interpersonal-level factor mainly affected by parental perception and social environment seemed to be more influential in ensuring independent mobility than the physical environment.

Researchers have increasingly explored the influence of B.E. on AST, reflecting the growth of new urbanism and walkability considerations in planning, landscape architecture, public health, and other related fields. Elsewhere, systematic meta-reviews have found that proximity to schools is consistently positively associated with AST, while residential density, land use mix, and connectivity have demonstrated inconsistent results (Bocarro et al. 2009; Ikeda et al. 2018a; Lin and Chang 2010; Panter, Jones, and van Sluijs 2008; Smith et al. 2020; Ton et al. 2019; Yarlagadda and Srinivasan 2008). On the other hand, a recent systematic review found strong evidence for the impact of multiple streetscape components (including two or more: crosswalk/sidewalk improvements, improved/covered bike parking, traffic calming features and safe places to walk) on children’s active travel (to school and other destinations) (Smith et al. 2017). Wong, Faulkner, and Buliung (2011) noted that differences in methodologies—in terms of the study area, measurements, and controls—might have been responsible for the inconsistency of the results, while other researchers have suggested that AST is moderated by other factors, such as demographics or parental perception of school distance and traffic safety (CDC 2005). Although the parents who participated in studies tended to claim distance as the greatest barrier or facilitator, a study (Nelson et al. 2008) noted that many children actually live within a reasonable walking distance from school.

Notably, different AST modes require different B.E. settings. While short distances have been found to be important for utilitarian travel, including both walking and biking (including AST), different conditions are generally required for each mode of transport. Pedestrians require good sidewalks and crosswalks, while bikers need either specialized bike paths or dedicated on-street bike lanes. While pedestrians desire interesting spaces along their route (e.g., storefronts and building facades), this may be less likely for bikers (Frank, Engelke, and Schmid 2003). Although some studies have positively associated aesthetic factors—such as the presence or absence of parks, street plantations, playgrounds, benches, and garbage bins—with both walking and biking (Fraser and Lock 2011; Heinen, Van Wee, and Maat 2010; Wang et al. 2016), one study presented mixed results (Ton et al. 2019).

Inconsistent findings were found in studies using children’s biking as an outcome measure (Moran, Plaut, and Baron-Epel 2016). For example, one study identified a favorable link between children’s biking and street connection (Trapp et al. 2011), while others found negative links (Van Dyck et al. 2009). In addition, walking and biking involve different types and levels of safety concerns; for young people, cycling is perceived as less safe for school travel than walking (Mandic et al. 2017). Working with the community to support active school travel initiatives (e.g., walking school buses and cycle trains) may require different interventions for cyclists versus walkers due to the differences in considerations. Bikes in Schools (a bike track created on school premises) is an example of a program that not only improves biking skills, safety knowledge, and school involvement but also provides an area for communities to socialize after school and raises awareness of cycling (Smith et al. 2020). Previous empirical studies, however, differentiating correlates of biking are mainly limited to the adult population. Moreover, most empirical studies that differentiated walking and biking AST dealt mainly with physical environmental variables (Kaplan, Nielsen, and Prato 2016; Schlossberg et al. 2006) rather than variables related to social environment or perception.

Despite a rich literature on AST, with most studies considering demographics, parental perceptions, B.E., and the social environment (Ikeda et al. 2018b; Marzi, Demetriou, and Reimers 2018; Rinne et al. 2022), a limited number of studies have focused on independence and modes of AST in one study. While some research looked at the impact of environmental factors on walking and biking, others looked at the impact of environmental attributes when combined with composite AST measures (Moran, Plaut, and Baron-Epel 2016; Schlossberg et al. 2006). Furthermore, few studies compared children’s IAST behavior and modes of AST in new urbanist settings with those in conventional ones. However, it is apparent that the factors determining AST are multidimensional. Accordingly, this study examines a broader spectrum of children’s AST according to subjectively and objectively measured B.E., demographics, automobile possession, neighborhood type (i.e., conventional and new urbanist), P.A. destinations and equipment, child’s attitude, and parental permission. Furthermore, the study compares the impact of different correlates on AST and IAST and walking and biking, AST’s two most prevalent modes in both new urbanist and conventional settings, confirming the heterogeneity of the included samples.

3. Materials and methods

3.1. Data and variables

This study was approved by the University of North Carolina Institutional Review Board (IRB# 3225). A sample of 367 4th – and 5th-grade students (aged 9–11 years) were recruited from four schools in Chapel Hill, North Carolina, United States. This age group was selected because it represents a stage of growing independence and development of spatial-cognitive skills that facilitate independent and active travel (Piaget 2013), with parents having previously indicated that children between 9 and 11 years old could be independently active without adult supervision (Jago et al. 2009). In Chapel Hill, the local elementary schools, local government, and a national program, Go! Chapel Hill Active Living by Design, has collaborated to encourage the spread of the SRTS program. The four schools in our study (Ephesus Road Elementary, Estes Hills Elementary, Rashkis Elementary, and Mary Scroggs Elementary) were the only schools participating in the SRTS program at the time of data collection. Students were asked about their AST pattern during the seven days previous to participating in the study in a group-administered survey at school. During the information session, each respondent was asked to complete the survey while in the room. If the respondents were unclear about the meaning of a question, they were allowed to ask the researcher or their physical education teachers for clarification. The data were collected between April and May in 2013; thus, it was presumed there were no seasonal or climatic effects on participant AST. Further, considering the small sample size and nonrandom nature of the sampling method, we acknowledge that the findings of this study are not readily generalizable to other samples and populations.

Questionnaires on AST were developed and modified from the existing parent-child survey, Active Where? (Joe, Carlson, and Sallis 2008). Two sections of the questionnaire asked about school travel mode and independent AST. The first question asks the following: In a typical school week, how many days do you go to/from school in the following ways? Response options were 0 to 5 days for walk, bike, car, and bus (although we focused on walking and biking), and participants were asked to respond separately for travel to school and from school. Second, the questionnaire asked about independent, active school travel with the question: If you walk or bike to/from school, how many days do you walk or bike without an adult? Again, options were 0 to 5 days for each mode of transportation, indicating both to school and from school.

Table 1 presents all the study variables and their measurements, with B.E. variables including distance to school, child population density (i.e., child population per unit area), mixed land use, street length, intersections, sidewalk density, and P.A. locations (measured within a quarter-mile radius of each respondent’s home). Matching with school zones, two large new urbanist neighborhoods were selected with two groups of conventional suburban neighborhoods located in the Chapel Hill–Carrboro area. These two neighborhoods were declared as “new urbanist neighborhoods” in the previous studies of Brown, Khattak, and Rodriguez (2008), Demir (2006), and Rodriguez, Khattak, and Evenson (2006). The selected new urbanist neighborhoods were greenfield developments built in the late 1990s and early 2000s according to new urbanism design principles (Demir 2006; Rodriguez, Khattak, and Evenson 2006).

Table 1. Descriptive statistics.

Variable	Measurment	n	Mean	S.D.	Min	Max
[Responses]
AST	Days of active travel (i.e., walking or biking) to or from school	367	3.51	4.26	0	10
IAST	Days of independent, active travel (i.e., walking or biking) to or from school	367	2.99	4.05	0	10
AST_W	Days of walking to or from school	367	2.35	3.63	0	10
AST_B	Days of biking to or from school	367	1.16	2.85	0	10
[Predictors]
BE_NEI	Neighborhood type (0 = suburban and 1 = new urbanism)*	367	0.31	0.46	0	1
BE_SCH	Distance to school along the circulation system (network distance, ratio)	367	1.71	1.76	0.00	15.05
BE_POP	Population of those aged 6–11 years in the census block	367	4.34	2.87	0.01	15.25
BE_MIX	Land use mix [0 = otherwise and 1 = if the neighborhood has at least one parcel (entirely or partially) with a retail, institutional, or office land use]*	367	0.27	0.44	0	1
BE_CON	Street connectivity (3-to-4 or 3-to-5-way intersections)	367	33.91	25.93	0	164
BE_SID	Sidewalk proportion (ratio of the sidewalk length per street length)	367	0.92	0.94	0	4
BE_DES	Physical activity locations (parks, greenway, trails, stream, pond, sports field, YMCA, school, shop, indoor sports facilities (martial arts, dance), cemetery, and bike lane	367	1.26	0.81	0	3
SD_AGE	Age (1 = 9 years, 2 = 10 years, and 3 = 11 years)	367	2.16	0.71	1	3
SD_FEM	Female (0 = male and 1 = female)*	367	0.49	0.50	0	1
SD_WHI	White (0 = others and 1 = White)*	367	0.55	0.50	0	1
SD_AUT	Automobiles in the household	367	2.23	1.17	0	9
SD_BMI85	BMI above the 85th percentile (0 = no and 1 = yes)*†	367	0.34	0.47	0	1
SD_PER	Parent’s permit to walk (let you walk anywhere without an adult?)*‡	367	0.32	0.47	0	1
SD_SIC	Sick last week (0 = no and 1 = yes)*‡	367	0.76	0.43	0	1
AT_1	Safe to walk (it is safe to walk or jog)	367	3.56	0.73	1	4
AT_2	Safe to bike (it is safe to ride a bike)	367	3.57	0.71	1	4
AT_3	Traffic (there is so much traffic that it makes it hard to walk)	367	1.35	0.71	1	4
AT_4	Crime (there is a lot of crime)	367	1.39	0.74	1	4
AT_5	See others (I often see other girls or boys playing outdoors)	367	3.23	1.00	1	4
AT_6	Street light (my neighborhood streets are well lit at night)	366	2.92	0.98	1	4
AT_7	Trees (there are trees along the streets)	367	3.29	0.96	1	4
AT_8	Interesting things (there are many interesting things to look at while walking)	367	3.01	0.96	1	4
AT_9	Sports equipment (at home there is enough sports equipment to use for physical activity)	367	3.26	0.96	1	4
AT_10	Sidewalk (there are sidewalks on most of the streets)	367	3.07	1.20	1	4
AT_11	Trails (there are bicycle or walking trails)	367	2.90	1.19	1	4
AT_12	Destinations (there are many places I like to go within easy walking distance of my home)	367	3.38	0.91	1	4
AT_13	Parents worry [my parents (or guardians) worry about something happening to me if I go somewhere on my own]	367	2.56	1.06	1	4
AT_14	Easy to walk to transit [it is easy to walk or bike to a transit stop (bus) from my home]	366	2.85	1.22	1	4
AT_15	People get along [people in this neighborhood generally get along with each other]	367	3.25	0.89	1	4
AT_16	Enjoy active travel to school [I enjoy (or would enjoy) walking or biking to school]	367	3.28	1.04	1	4
AT_17	Enjoy active travel to places [I enjoy (or would enjoy) walking or biking to places (park, playground, or shop)]	367	3.31	0.96	1	4
AT_18	Enjoy active travel without adults [I enjoy (or would enjoy) walking or biking to places without an adult]	367	2.93	1.06	1	4
AT_19	Active travel without adults is cool [other kids my age think walking or biking without an adult is “cool”]	367	2.86	0.98	1	4

* Binary variables

† BMI above 18.60 (19.08) for 9-year-old boys (girls), 19.36 (19.95) for 10-year-old boys (girls), and 20.16 (20.83) for 11-year-old boys (girls), which makes f (non-overweight) = 243 and f (overweight) = 124 (other typical percentiles such as 90th and 95th and BMI as a continuous variable were entirely insignificant)

‡ Inflation variables

Although both categories featured similar populations of families with children aged 9–11 years of similar socioeconomic status, new urbanist neighborhoods were designed according to “new urbanism” standards as residential planned unit developments (PUDs). While all students of Ephesus and Estes Hills schools reside in conventional neighborhoods, most of those from Rashkis and Mary Scroggs schools live in two new urbanist PUDs. The school zones for both Rashkis and Mary Scroggs schools also include residences outside the new urban PUDs; those participants were classified as conventional suburban residents. The new urbanist neighborhoods featured shorter distances to school, more intersections, longer sidewalks, and more mixed-use spaces, such as centralized commercial zones. The results of multiple t-tests identified a statistical difference between the two types of neighborhoods (see Appendix A).

The home neighborhood was defined as the area within a quarter-mile crow-fly buffer around each participant’s home location. Except for the distance to school, all variables were measured within the home neighborhood B.E. Most B.E. measures were derived from 2013 Chapel Hill open data using ArcGIS 10.1 (Environmental Systems Research Institute Inc., Redlands CA). Distance to school was estimated as the distance based on the shortest route possible along the circulation system (including roads, trails, and pathways) measured by Network Analyst 10.1 in ArcGIS. Child population density was derived from 2010 census data based on the number of children aged 6–11 years per block. When a buffer around a participant’s home was not fully contained within a census block polygon, the data were assigned in direct proportion to the area of the polygon contained within the buffer. The land-use mix included the presence of mixed land uses within the neighborhood and was coded using a binary variable: a value of 1 if the neighborhood had at least one parcel (entirely or partially) with retail, institutional, or office land-use code designation, and 0 otherwise. Variables regarding connectivity comprised street connectivity (number of intersections) and sidewalk density within the buffer. Last, destination access targets measured the availability of P.A. facilities. The number of P.A. destinations included parks, greenways, trails, streams, ponds, sports fields, YMCAs, schools, shops, indoor sports facilities (martial arts, dance), cemeteries, and bike lanes within a quarter-mile buffer.

The scale used to assess the perceived physical environment and attitude toward AST was derived from the work of Evenson et al. (2006) and the parent-child survey, Active Where? (Joe, Carlson, and Sallis 2008). For this study, 15 of the 21 items in the Evenson scale (which includes four subscales of perceived safety, aesthetics, facilities, and transportation) were used. Items related to garbage, bad smell, and scary dogs were dropped because those were considered inappropriate for the study site context. Neighborhoods utilized in this study were found to be relatively well-maintained through site examinations for the pilot study. The expert group survey also confirmed that these variables were less likely to be relevant to the site context.

Three items regarding parental permission were combined into one question. Items regarding attitude focused on how much respondents enjoy walking or biking to school and other places. Possible response options were provided on a 4-point scale: strongly agree, somewhat agree, somewhat disagree, or strongly disagree.

To examine AST by independence and by mode, this study included a total of 33 predictors: seven objectively measured B.E. variables (B.E._#), seven sociodemographic variables (S.D._#), and 19 attitudinal or perception variables (AT_#), including the perceived B.E. and perceived social environment of the neighborhood. Given that response variables were measured as weekly trip frequencies, the study used a Poisson-family regression for the count data (n = 367). This involved Poisson and negative binomial (N.B.) models and their zero-inflated forms, zero-inflated Poisson (ZIP), and zero-inflated N.B. (ZINB) models. For the zero-inflated models, SD_PER (parent’s permission to walk) and SD_SIC (sick last week) were set as inflation variables, with 67.8% of SD_PER and 24.0% of SD_SIC coded as zero.

3.2. Analysis procedure

First, as part of the psychometrics process, a common factor analysis was conducted to reduce the 19 variables (AT_1~ AT_19) for perceptions or attitudes concerning the environment (by applying the varimax axis rotation). The eigenvalues for the three factors (AT_F1, AT_F2, and AT_F3) – indicators for the same grouped factors – were 2.066, 1.940, and 1.754, respectively. The factor loadings of all indicator variables were above the minimum level of 0.3. Considering the structuring of factors and guided by Gim (2018), these were named (1) physical settings (AT_F1), (2) subjective and safety settings (AT_F2), and (3) positive utility (AT_F3). These three composite variables were subsequently included in the empirical models instead of the 19 initial variables for perceptions or attitudes (i.e., variable values were replaced by factor scores).

As is generally recommended, this study selected the final form of the count data model by comparing the log-likelihood, BIC, and AIC (see Appendix B). Initially, it was also tested using the Vuong test, a conventional way to determine whether the standard model can be replaced by its zero-inflated counterpart (if the test result is significant, the zero-inflated model can be employed). Although Wilson (2015) recently recommended not reviewing the Vuong test, for this study, the test result was consistent with that of the other fit indices. Regarding overall AST, the two Vuong tests supported the zero-inflated models rather than the standard forms. Accordingly, this study evaluated ZIP and ZINP using the likelihood-ratio test. It was found that ZINB did not significantly improve the fit, prompting the ultimate selection of ZIP. All fit indices delivered the same result, with ZIP being superior to the other three models, featuring a higher log-likelihood and lower AIC and BIC values. With regard to the IAST model, the two Vuong tests recognized that ZIP was superior to the Poisson regression, but that ZINB was not necessarily better than N.B. Similarly, using the likelihood-ratio test to compare the two zero-inflated models determined that ZIP was superior to ZINB, implying that ZIP was the best-fit model. The fit indices (log-likelihood, AIC, and BIC) also indicated that ZIP was the most desirable of the four alternatives. By estimating walking AST, the two Vuong tests demonstrated that zero-inflated count data models were more appropriate than their standard models. This study subsequently compared ZIP and ZINB through the likelihood-ratio test, which revealed that ZINB significantly improved the model fit. Similarly, the three indices all produced the best results for ZINB.

Finally, the likelihood-ratio test comparing Poisson and N.B. regressions found that walking AST was best estimated using the latter. In addition, comparing the actual distribution of the data and the distribution proposed by the model, deviance and Pearson chi-squares were both higher for N.B. than for the Poisson model. After determining whether the zero-inflated form could replace the standard N.B. through the Vuong test, the z-value (= 0.25) was found to be insignificant, supporting the standard model (see Appendix B), which also shows the results of the Vuong test comparing standard Poisson and ZIP; however, given the N.B. form was substantially superior, this comparison was unnecessary. Consistently for all cases, the fit indices produced a better fit for N.B. than for Poisson regression.

4. Results

For significance testing, this study employed the 90% significance level for each of the four active travel models (Table 2), with AST days including both AST and IAST. Overall, AST days were found to increase with a more diverse land-use mix, a greater proportion of sidewalks, and a higher positive utility rating (AT_F3). Overweight children tended to report fewer AST days. Meanwhile, IAST days were increased in census blocks with a larger child population and were particularly linked with higher sidewalk proportion, older, and male respondents. Notably, subjective and safety settings (AT_F2) had a negative impact on IAST. That is, IAST can be explained by negative loading for crime, traffic, and parental concern for the factor. When separately observing AST with two modes, walking (AST_W) and biking (AST_B), the significant factors and signs were dissimilar. Walking days to school were likely to increase for children in suburban neighborhoods with a shorter distance to school, fewer children, more mixed-use, and a higher positive utility rating. Compared to other ethnicities, White children conducted less AST by walking and more by biking. Moreover, males tended to do more biking AST. Biking tended to increase in neighborhoods with new urbanist features and a higher positive rating of physical settings.

Table 2. Count data models (n = 365).

	AST (model: ZIP)		IAST (model: ZIP)		AST_W (model: ZINB)		AST_B (model: NB)
	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.
Neighborhood type (BE_NEI)†	−0.004	0.120	0.135	0.124	−0.368*	0.197	1.784**	0.682
Distance to school (BE_SCH)	0.018	0.032	0.033	0.033	−0.122*	0.064	0.055	0.116
Size of population aged 6–11 (BE_POP)	0.006	0.013	0.029**	0.013	−0.038*	0.021	−0.089	0.079
Land use mix (BE_MIX)†	0.152*	0.069	0.070	0.079	0.256**	0.106	−0.334	0.553
Street connectivity (BE_CON)	0.000	0.001	−0.001	0.001	0.003	0.003	0.005	0.011
Sidewalk proportion (BE_SID)	0.074**	0.044	0.081*	0.046	0.042	0.076	0.032	0.318
Physical activity locations (BE_DES)	−0.038	0.054	−0.013	0.059	−0.056	0.086	−0.155	0.331
Age (SD_AGE)	0.004	0.044	0.089*	0.051	0.001	0.068	0.062	0.302
Female (SD_FEM)†	−0.051	0.062	−0.148**	0.072	0.073	0.097	−1.287**	0.440
White (SD_WHI)†	0.025	0.071	0.081	0.081	−0.197*	0.108	1.533***	0.421
Automobiles (SD_AUT)	−0.017	0.028	0.019	0.030	0.008	0.038	−0.166	0.235
BMI above the 85th percentile (SD_BMI85)†	−0.124*	0.071	0.033	0.075	−0.007	0.113	−0.297	0.455
Parent’s permit to walk (SD_PER)†	−0.049	0.072	0.047	0.081	−0.081	0.113	−0.658	0.414
Sick last week (SD_SIC)†	0.030	0.074	0.020	0.082	0.135	0.115	0.289	0.502
Attitudes to physical settings (AT_F1)‡	0.023	0.048	−0.042	0.054	0.065	0.071	0.690**	0.330
Attitudes to subjective and safety settings (AT_F2)‡	−0.054	0.051	−0.130**	0.054	−0.080	0.076	0.290	0.324
Attitudes to positive utility (AT_F3)‡	0.095*	0.052	0.047	0.060	0.272**	0.091	0.279	0.287
(Constant)	1.922	0.170	1.377	0.201	1.968	0.258	−1.098	1.192
Inflate
Parent’s permit to walk (SD_PER)*	0.763***	0.236	1.026***	0.250	0.625**	0.250
Sick last week (SD_SIC)*	−0.408	0.255	−0.350	0.261	−0.116	0.266
(Constant)	0.231	0.226	0.331	0.231	0.397	0.238
Ln(α)					−2.5327	0.488	1.8881	0.171
α					0.0794	0.039	6.6070	1.132
LR χ^2 (17) Pseudo R^2 ML R^2 Cragg-Uhler R^2 Note:	26.19 (p = 0.071) 0.028 0.100 0.103 converged in 3 iterations		35.12 (p = 0.006) 0.043 0.137 0.142 converged in 3 iterations		34.67 (p = 0.007) 0.035 0.108 0.112 converged in 6 iterations		71.96 (p = 0.000) 0.096 0.179 0.205 converged in 4 iterations

† Binary

‡ Psychometric factors identified through common factor analysis in Table 3

* p < 0.1; ** p < 0.05; *** p < 0.001

Note: For variable descriptions, see Table 1.

Table 3. Common factor analysis.

	AT_F1 (physical settings)	AT_F2 (subjective and safety settings)	AT_F3 (positive utility)
Sidewalk (AT_10)	0.546	−0.033	0.248
Trails (AT_11)	0.509	0.089	0.224
Trees (AT_7)	0.473	0.142	−0.007
Street light (AT_6)	0.467	0.128	−0.049
Interesting things (AT_8)	0.442	0.147	0.052
Sports equipment (AT_9)	0.384	0.196	0.156
See others (AT_5)	0.382	0.066	0.276
Destinations (AT_12)	0.341	0.140	0.322
Safe to walk (AT_1)	0.365	0.714	0.079
Safe to bike (AT_2)	0.299	0.670	0.172
People get along (AT_15)	0.307	0.493	0.019
Crime (AT_4)	0.016	−0.449	−0.037
Traffic (AT_3)	−0.070	−0.384	−0.116
Parents worry (AT_13)	−0.047	−0.363	−0.139
Enjoy active travel to places (AT_17)	0.154	0.195	0.664
Enjoy active travel to school (AT_16)	0.216	0.150	0.613
Enjoy active travel without adults (AT_18)	0.040	0.255	0.566
Active travel without adults is cool (AT_19)	−0.001	−0.004	0.363
Easy to walk to transit (AT_14)	0.254	0.050	0.305
Eigenvalues	2.066	1.940	1.754

Note: Converged in 6 iterations

Meanwhile, without parental permission, children did not conduct AST throughout the week, whether AST or IAST. That is, parental permission was a significant inflation variable separating zeros from the other possibilities (i.e., binary choice), with permission also functioning as an inflation variable for the walking AST model (presented at the bottom of Table 2). In contrast, a child respondent having been sick in the previous week did not inflate zero counts.

5. Discussion

The results of this study align with several previous studies and also point to unique results that could support discussions on differences between AST and IAST and different AST travel modes.

First, residing in a new urbanist or conventional suburban neighborhood did not impact overall AST. However, classifying AST according to mode, neighborhood type was seen to exert an opposite effect on walking compared to biking, with children living in new urbanist neighborhoods engaging in less walking and more biking AST. Accordingly, in the overall AST model (in which travel modes were not considered), neighborhood type was demonstrated to be insignificant as the opposing effects canceled each other out.

While some new urbanist features—a high proportion of sidewalks and mixed-use land—positively impacted overall AST, mixed-use land only impacted increased walking AST. This could be explained by assuming that all new urbanist design features (e.g., longer street length and more intersections in this study) are associated with AST or that this study’s two “new urbanist-ish” neighborhoods were not fully representative of the whole concept of the new urbanist town. This requires closer consideration of the specific context of the study area. Chapel Hill and its residents tend to be health-conscious and proactive in terms of health and active travel, with many programs for active living (e.g., Active Living by Design) having been initiated in the area (Transtria 2012), and different active travel interventions in school, neighborhood, and business level had been implemented according to community action (5P) model (Pang, Kubacki, and Rundle-Thiele 2017). In addition, all four schools were participating in the SRTS program (although each school determined if and how much they would participate each year). Thus, the effect of neighborhood type was unclear in this study.

Some studies (whilst not necessarily about AST) have compared new urbanist neighborhoods with conventional suburbs in terms of adult health outcomes, including BMI or walking behaviors (Brown, Khattak, and Rodriguez 2008; Rodriguez, Khattak, and Evenson 2006), with mixed results. For example, Brown, Khattak, and Rodriguez (2008) found no direct relationship between new urbanist neighborhoods and lower BMIs, with only household heads having lower BMI, while Rodriguez, Khattak, and Evenson (2006) found no difference in P.A. measures across neighborhood types. Meanwhile, Zhu et al. (2014) observed a 40% increase in the neighborhood walking and biking activities of neighborhoods in Austin, Texas, that moved to a planned development embracing several new urbanist design principles that focused on activity-friendly characteristics (e.g., hiking and bike paths, sidewalks, grid street patterns, and interconnected streets). They also documented a dramatic increase in walking and biking among people who lived in areas unfavorable to these activities.

These results indicate that increasing AST requires the implementation of an aggressive walking-and-biking favorable design implementation for the target population. Notably, residents who choose to move into new urbanist neighborhoods are more willing to embrace new urbanism principles such as walking and biking (Podobnik 2011). Future studies should consider the self-selection bias (Mokhtarian and Cao 2008) arising among residents of new urbanist neighborhoods; given that residents who prefer walking choose to live in neighborhoods with more walkable settings, they are more likely to walk and bike, with children of active parents tending to be more active (Petersen et al. 2020).

Second, the land-use mix increased walking AST but not biking AST. This might be in part because walking is more favorable to trip chaining than biking, with pedestrians more likely to desire interesting sites along their route than bikers (Wood, Frank, and Giles-Corti 2010). Another factor, in the context of escorted AST, is parents running errands, shopping, or stopping at a café or other social gathering place between home and school when AST is escorted. This explanation is supported by the above finding that land use mix is only significant for the overall AST model and not the IAST model. Notably, non-residential use might also mediate the fear of crime because there tend to be more onlookers (Park 2017). Considering the average number of mixed-land use within a quarter-mile (= 0.27) and distance to school (= 1.71 miles), children could find less than two retail, institutional, or office buildings (= 1.71 miles/0.25 miles * 0.27 = 1.85); in other words, the intensity of mixed-use was not very frequent. From this, we can assume that although shops or offices encountered along the route may promote a certain level of safety, the scarce “natural surveillance” from one or two shops would be insufficient to encourage children to go to school independently.

Another notable result is that the effect of the distance to school was negative for walking AST and insignificant for biking AST. This can be attributed to walkable distances being generally shorter than bikeable distances (Maat, Van Wee, and Stead 2005; Schlossberg et al. 2006), which are generally calculated as 0.25 miles to 0.5 miles (Mejia et al. 2015) for walking distance and 5 miles for bikeable distance (Sallis et al. 2004). Thus, the modification of zoning and subdivision ordinances for a certain level of mixed-land use and sidewalk proportion would be recommended. In addition, stakeholders should consider the impact of a school siting guide on creating favorable walking environments that accord with a neighborhood’s B.E. characteristics to promote AST.

Third, among the three attitudinal factors, positive utility (AT_F3) increased walking AST, what was in accordance with the findings of previous studies (Gim 2018; Mokhtarian and Salomon 2001). This suggests that travel may be motivated by the desire for travel-based multitasking, positive emotions, or fulfillment. In this research, respondents who enjoyed active travel to places including school and felt “cool” about engaging in active travel without adults were more likely to walk to school independently. Physical settings (AT_F1), such as trees, trails, street lights, or destinations, encouraged biking AST. Both results are easy to explain: (1) travel utility is better improved by walking, which requires more time than biking to travel the same distance to school; and (2) compared to walking, biking can be fully served by infrastructures such as bike-only lanes and racks. The negative association between subjective safety settings (AT_F2) and IAST suggests that children may not be involved in IAST if the area has features that cause worry for parents, such as a high crime rate or a high traffic volume, even if a neighborhood is equipped with favorable settings for pedestrian/biker safety and good neighbors. Notably, crime, traffic, and parental concern were negatively loaded, while both perceived safety of walking and biking and social trust from neighbors were positively loaded in AT_F2.

Overall, the three factors assumed different roles, being significant only when certain conditions were met. Favorable physical settings (AT_F1) facilitated biking AST, children reporting high positive utility of travel (AT_F3) increased walking AST, and negative safety settings; that is, concern about crime or traffic safety from children or parents (AT_F2) discouraged IAST. These findings could be implemented by policymakers. For example, considering biking received less social and infrastructure support compared to walking (Mandic et al. 2017), improving bike-safe infrastructure and introducing or expanding trails are critical for bikers. Promoting walking AST calls for more delicate efforts, including improving crime and traffic safety, promoting a positive attitude toward active travel, and creating exciting environments along walking routes. Enhancing sidewalk connectivity, improving street lighting, improving the safety of street crossings, traffic calming, and enhancing streetscape aesthetics could also be prioritized. Notably, fear of crime, rather than actual crime or risk, is more likely to result in people avoiding active travel (Park and Garcia 2019); thus, informational outreach and educational programs explaining the risks and benefits of AST should focus on building children’s skills and confidence, helping parents gain confidence in their children’s abilities (Riazi et al. 2022).

The statistical models used in this study also associate AST with non-BE variables. Male students tended to do more biking AST, which, combined with the stated findings concerning IAST, indicates that boys are more likely to engage in walking IAST than girls. Previous studies have also observed greater freedom regarding IAST among boys (Evenson et al. 2006; Marzi, Demetriou, and Reimers 2018; NLI 2010), thus indicating gender-specific interventions, especially for girls, are critical for promoting overall AST (Marzi, Demetriou, and Reimers 2018; Zougheibe et al. 2021). Compared to other ethnic groups, White children are less likely to engage in walking AST (negative coefficient) and more likely to engage in biking AST (positive coefficient), which confirms previous findings (Larouche et al. 2016; Pucher and Renne 2003). Elsewhere, McDonald (2008a) found that ethnic distinctions might be best explained by transportation accessibility and household income. Future studies may explore whether ethnic differences in AST resulted from inequitable socioeconomic status or inequitable distribution of the transportation system.

Finally, confounding our initial expectations, the walking AST model demonstrated a negative impact on the child population in the neighborhood. As discussed, the variable increased IAST but not overall AST, so this particular result suggests that a child-dense population reduces walk commuting accompanied by parents/guardians. Ultimately, the two models (the walking AST and IAST models) may together imply that with schoolmates living nearby, children tend to walk to school with other students rather than with their parents, especially considering that the four study schools were participating in the SRTS program. This supports the need for policies that establish “buddy” systems or similar strategies to provide companionship, friendship, and support in the context of IAST. In fact, school travel policies and practices (e.g., drop-off/pick-up rules, school patrols) were created solely to guarantee the safety of students (Smith et al. 2020). Thus, programs addressing AST include KidsWalk (CDC), Walk to School, Walking School Bus, and Safe Routes to School. These interventions directly mediate parental worries by having participating group members host workshops to promote parental awareness of the positive impact of their children’s active travel (Kingham and Ussher 2007).

6. Conclusions

In this paper, we confirm previous findings such as the positive impact of mixed land use, sidewalks, and utility rating on days of active travel to school. Meanwhile, the days of independent, active travel to school are heavily influenced by the child population nearby, gender, and safety settings. Another interesting aspect of this study was the differences identified in relation to AST modes. Walking was sensitive to mixed land use and positive utility, while biking was more connected to physical settings. This finding underscores the need for future programs and initiatives to take different actions to encourage walking and biking. Moreover, prospective studies should investigate the driving factors of IAST by biking and walking.

Although the outcomes of this study support the future complete street design for AST and IAST, there are some limitations that future research could consider. First, the (social) ecological model considers diverse variables such as individual, relationship, environment, and policy. Therefore, the relationship between children and parents (currently only including parental permission) or policies of educational authorities should be included in the future model. In that model, the “interaction term” between individual, interpersonal, environmental, and policy variables could be added; in turn, sufficient sample size must be secured. Increased sample size would enable more meaningful results due to the increased variation of variables. Spatial variables require a specific unit of aggregation by their own characteristics. Since walking distance and bicycle traffic distance are somewhat different, examining how variation in the observation unit of the built environment influences outcomes can suggest implications for biking and walking policies. The influence of residential self-selection, the current key topic of the built environment-travel interaction studies, could not be controlled in this study. In particular, future research is needed to investigate more closely the possibility of choosing a residence taking into consideration the location of the children’s school and the level of misestimation. Lastly, because the survey data were collected in 2013, they were somewhat outdated; in this sense, planners should be cautious in applying the empirical findings of this study, as this limitation may critically affect external validity. However, being cross-sectional, it was not the intention to generalize the findings to other times and areas (or examine the longitudinal relationship between predictor and response variables), but rather to secure their variations across different neighborhoods for the statistical inference of their inherent relationship. This aim was arguably well served.

Acknowledgments

The paper is based on the first author’s unpublished dissertation entitled “The Association of Urban Form and Design with Children’s Physical Activity and Active Travel” and was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2021S1A3A2A01087370).

Disclosure statement

The authors declare that they have no significant competing financial, professional, or personal interests that might have influenced the performance or presentation of the work described in this manuscript.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Notes on contributors

Jong Seon Lee

Jong Seon Lee is a design examiner at Korean Intellectual Property Office. She received her Ph.D. in Design from North Carolina State University. She has work experience at Graduate School of Public Administration, Seoul National University investigating roles of public administration creating healthy communities. She also led several urban regeneration projects through tourism in Korean Tourism Organization.

Yunmi Park

Yunmi Park is an assistant professor at Ewha Womans University. Dr. Park earned her Ph.D. in Urban and Regional Science from Texas A&M University and also has professional planning experience in South Korea, which enabled her to become a certified planner in both South Korea and the United States (AICP). Her areas of expertise and research interests include urban planning and design, urban shrinkage and revitalization, and spatial analytics.

Tae-Hyoung Tommy Gim

Tae-Hyoung Tommy Gim is an associate professor and the chair of the Department of Environmental Planning at Seoul National University. He received a Ph.D. in City and Regional Planning degree from the Georgia Institute of Technology. His fields of expertise include land use–transportation–environment interactions, social innovation, and quantitative analysis.

Appendix

Appendix A. T-test of B.E. variables by neighborhood type.

	Conventional		New Urbanist		t	df	p
	M	SD	M	SD
Distance to school (BE_SCH)	2.16	1.88	0.72	0.86	10.1	365	.000
Size of population aged 6–11 (BE_POP)	4.52	3.27	3.92	1.55	2.4	364	.017
Land use mix (BE_MIX)	0.15	0.36	0.52	0.50	−7.0	166	.000
Street connectivity (BE_CON)	20.72	11.67	63.54	24.71	−17.6	135	.000
Sidewalk proportion (BE_SID)	0.43	0.44	2.01	0.85	−18.6	139	.000
Physical activity locations (BE_DES)	0.94	0.64	1.97	0.67	−13.7	206	.000

Appendix B. Fit of alternative models.

Models	Tests
Overall active school travel (AST_A)	Vuong test (Poisson vs. ZIP)	z = 7.02 (p = 0.0000)
	Vuong test (NB vs. ZINB)	z = 1.80 (p = 0.0356)
	Likelihood-ratio test (ZIP vs. ZINB)	χ-bar2 (1) = 5.4e-05 (p = 0.4971)
		Poisson	NB	ZIP	ZINB
	Log-likelihood	−1,018.400	−746.726	−662.307	−662.307
	AIC	5.679	4.196	3.744	3.750
	BIC	−10.464	−547.912	−704.950	−699.050
Independent active school travel (IAST_A)	Vuong test (Poisson vs. ZIP)	z = 6.95 (p = 0.0000)
	Vuong test (NB vs. ZINB)	z = 0.25 (p = 0.4007)
	Likelihood-ratio test (ZIP vs. ZINB)	χ-bar2 (1) = 1.2e-05 (p = 0.4986)
		Poisson	NB	ZIP	ZINB
	Log-likelihood	−950.448	−670.138	−601.870	−601.870
	AIC	5.307	3.776	3.413	3.418
	BIC	−146.369	−701.088	−825.825	−819.925
Walking active school travel (AST_W)	Vuong test (Poisson vs. ZIP)	z = 7.43 (p = 0.0000)
	Vuong test (NB vs. ZINB)	z = 3.46 (p = 0.0003)
	Likelihood-ratio test (ZIP vs. ZINB)	χ-bar2 (1) = 7.31 (p = 0.0034)
		Poisson	NB	ZIP	ZINB
	Log-likelihood	−928.207	−604.838	−569.686	−566.033
	AIC	5.185	3.418	3.237	3.222
	BIC	−190.851	−831.688	−890.192	−891.600
Biking active school travel (AST_B)	Vuong test (Poisson vs. ZIP)	z = 5.44 (p = 0.0000)
	Vuong test (NB vs. ZINB)	z = 0.25 (p = 0.3995)
	Likelihood-ratio test (Poisson vs. NB)	χ-bar2 (1) = 546.50 (p = 0.000)*
		Poisson	NB	ZIP	ZINB
	Log-likelihood	−613.297	−340.046	−338.926	−337.540
	AIC	3.459	1.967	1.972	1.970
	BIC	−820.670	−1,361.272	−1,351.713	−1,348.584

* Deviance goodness-of-fit = 991.1902 [p > χ2(347) = 0.0000] and Pearson goodness-of-fit = 1,952.528 [p > χ2(347) = 0.0000] also support NB.