EN中文

Journal of Geographical Sciences >

2024 , Vol. 34 >Issue 4: 804 - 833

DOI: https://doi.org/10.1007/s11442-024-2228-9

Research Articles

Exploring zonal heterogeneities of primary school students’ commute-mode choices through a geographically weighted regression model

  • WU Dawei , 1, 2 ,
  • MA Lu , 1, 2, * ,
  • YAN Xuedong 1, 2
Expand
  • 1. School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044, China
  • 2. Ministry of Transport Key Laboratory of Transport Industry of Big Data Application Technologies for Comprehensive Transport, Beijing 100044, China
* Ma Lu, Professor, specialized in transportation planning and transportation big data. E-mail:

Wu Dawei (1999-), PhD Candidate, specialized in transport geography and travel behaviour. E-mail:

Received date: 2023-03-26

  Accepted date: 2023-11-03

  Online published: 2024-04-24

Supported by

National Natural Science Foundation of China(71971023)

Beijing Social Science Foundation(21DTR055)

Fold

Abstract

Commuting is an important part of primary school students’ travel behavior, which has been concerned for a long time. We found that the commute-mode choice behavior of primary school students in the context of regional segmentation shows strong characteristics in space, but has not yet been considered in traditional research. To fill this gap, this study summarizes the commute-mode choices of primary school students with different characteristics based on the Beijing School Commute Survey. And the geographically weighted regression (GWR) model is built to analyze the zonal heterogeneity of the impact of personal characteristics, family factors and school factors on the primary school students’ commute- mode choices from a low-carbon perspective. The results demonstrate that the possibility of primary school students choosing low-carbon commuting modes is positively correlated with the grade, commuting time, commuting escort type and housing category, but is inversely related with the commuting distance and the distance from the school to the city center. The coefficient estimates of explanatory variables vary across regions. Finally, we put forward policy suggestions regarding promoting the low-carbon commuting structure, such as developing the urban slow traffic system, which serve as a reference for policymakers.

Key words： primary school students; school commute; mode choice; zonal heterogeneity; geographically weighted regression

Cite this article

WU Dawei , MA Lu , YAN Xuedong . Exploring zonal heterogeneities of primary school students’ commute-mode choices through a geographically weighted regression model[J]. Journal of Geographical Sciences, 2024 , 34(4) : 804 -833 . DOI: 10.1007/s11442-024-2228-9

1 Introduction

The growth of students is related to the well-being of the family and the stability of society. To some extent, their development determines the future of society and the nation. Countries around the world have formulated laws and policies to protect students’ safety (Chui and Jordan, 2018; Rayens et al., 2022). Students, especially primary school students, are not yet fully mature. The international community is committed to providing help for their daily behavior in many aspects, such as travel activities and epidemic prevention (Mei et al., 2019; Nguyen, 2020; Mickells et al., 2021). Commuting is the main component of students’ travel activities. In China, students commute on the daily basis in enormous numbers, showing high frequency and time stability, and the scale is expanding (Gao et al., 2018).
Currently, the booming urban economy has offered city commuters ever more travel modes. Those living within short commuting distance can opt for walking, cycling, motorcycling and other travel modes, while others living within long commuting distance can turn to subways, buses, private cars and other travel modes (Fitch et al., 2019). The emergence of various travel modes provides many possibilities for citizens to travel, greatly increasing the convenience of travel. However, negative consequences ensue as a result of the massive commuter flow. Taking motor vehicles as an example, a large number of commuter vehicles in motion often bring heavy traffic, resulting in serious exhaust gas and noise pollution (Liu et al., 2015; Miao et al., 2015). Urban sustainability is confronted with severe challenges.
The commuting modes of minors, especially primary school students, differ from the regular commuting behavior, and tend to be diversified. In addition to conventional transportation modes such as bicycles and buses, the commuting mode of primary school students also includes shuttle buses and school buses provided by schools and families (Nirupama and Hafezi, 2014; Shafahi et al., 2019). The commuting behavior of students mainly occurs at a fixed time between school and home. It is worth noting that the characteristics of primary school students’ commuting behavior vary across regions given the regional difference in economy and the difference in school location. Primary school students in different regions will also have different choices in commuting modes and influence each other. It is necessary to analyze and master the characteristics of primary school students’ commute-mode choices in different regions and the zonal heterogeneity of the impact of factors. However, most previous studies tend to consider the characteristics of the commuting mode of the whole group of primary school students, a comparison between different regions is lacking in literature. The purpose of this paper is to take the urban area of Beijing as an example to compare the differences in the choice of commuting modes among primary school students with different characteristics from a low-carbon perspective. On this basis, this paper studies the influence of personal characteristics, family factors and school factors on the choice of commuting modes of primary school students, as well as the spatial variation of the influencing factors. This is achieved by building a scientific and advanced model — geographically weighted regression (GWR) model, based on the commuting survey data of primary school students in Beijing. The main achievements of this paper are:
(1) Summarizing the characteristics of the commute-mode choices of primary school students with different peculiarities (grade, commuting time, commuting distance, housing category, commuting escort type and the distance from the city center).
(2) Constructing a GWR model to analyze the zonal heterogeneity of the impact of the grade, commuting time, commuting distance, family housing category, commuting escort type and the distance from the city center on the commuting mode choice of primary school students from a low-carbon perspective.
(3) Proposing six policy recommendations to promote the low-carbon commuting modes among primary school students and alleviate road congestion based on the above two research results.
The remaining chapters of this paper are organized as follows. In Section 2, the article reviews the characteristics of primary school students’ commuting modes, policies and factors that influence primary school students’ choices of commuting modes. In Section 3, the article introduces the study area and the results of primary school students’ commuting survey, and makes the descriptive statistical analysis of the data. The basic principles of the GWR model and OLS (ordinary least squares model) model used in this study are described in Section 4. In Section 5, the paper discusses the results of the two models. Next, the regression results of the GWR model are analyzed in detail spatially. In Section 6, six policy recommendations are put forward. The last part of the article summarizes conclusions of the study and put forward future research recommendations.

2 Literature review

In addition to the regular commuting modes such as bicycles, buses and subways, primary school students’ commuting modes also include school buses and shuttle buses. This group’s commuting mode choices vary greatly from one another due to factors such as region, parents, the grade, commuting distance, etc. Here is a brief review that focuses on the commuting characteristics of primary school students, policies affecting the commuting of primary school students and the influencing factors of primary school students’ commute-mode choices.

2.1 Characteristics of commuting among primary school students

Compared with adult commuting, primary school students’ commuting characteristics are more diversified (Kobus et al., 2015; Soltani et al., 2018; Schonbach et al., 2019). In terms of frequency, primary school students return home for lunch and then go to school in addition to morning and evening commutes, and the frequency of commuting is, therefore, significantly higher than that of other groups (Chillon et al., 2017; Herrador-Colmenero et al., 2019). In terms of time, the commuting time of primary school students depends on the start time of the course (Chen et al., 2021), which may vary for seasonal climate and other factors, and thus primary school students’ commuting time is more flexible (Whalen et al., 2013; Fast, 2020). In terms of commuting modes, some primary school students do not have the ability to commute independently, so their commuting mode choices are vulnerable to the influence of commuting companions (e.g. parents and elders) (Ermagun and Samimi, 2016; Rodriguez-Rodriguez et al., 2019; Jurak et al., 2022).

2.2 Overview of policies affecting the commuting of primary school students

As an important way of governance, the policies formulated by governments in many cities around the world have had indirect or direct impacts on primary school students’ commuting. In China, the distance between school and home is the main factor determining the commuting mode (Yin et al., 2019). The government has introduced a policy of nearby enrollment. This policy can help to promote low-carbon commuting among primary school students and enhance the rational distribution of educational resources (McMillan, 2005; Liang et al., 2019). In Minnesota, USA, Sirard et al. (2015) evaluated a change in the government’s school selection policy and found that the stricter school selection policy reduced students’ commuting distance, but had no significant impacts on the enthusiasm for commuting. Marshall et al. (2010) discovered that policies such as school choice and school siting may conflict with increasing the proportion of non-motorized school commuting. Liao and Dai (2022) revealed that the policy of “attending nearby school” did not achieve the expected effects, and this policy cannot guarantee to reduce primary school students’ commuting distance. In Nanjing, Liu et al., (2020) found that despite the policy of “attending nearby school”, some students living in the study area still have long commuting distance, which increases students’ possibility of choosing cars to commute. In Israel, Elias et al. (2014) revealed that the government’s policy of encouraging primary school students to walk will increase the probability that children choose to commute on foot and by bike, but it also needs to consider the parents’ behavior patterns and their cognition of the way children go to school.

2.3 Influencing factors of primary school students’ commuting modes

There is a lot of literature about the influencing factors of primary school students’ commuting modes, and it can be concluded that the influencing factors of primary school students’ commuting mode choices mainly include personal factors and external factors (Mandic et al., 2015).
Personal factors involve age, gender, personal preferences, exercise and so on. In terms of age, the age of primary school students is positively related to their enthusiasm for commuting (Galvez-Fernandez et al., 2022). In terms of gender, girls are more willing to walk than boys, while boys are more willing to ride bicycles than girls (Leslie et al., 2010). At the exercise level, Tudor-Locke et al. (2001) concluded that moderate exercise can enhance the enthusiasm of primary school students to choose walking or cycling to commute. In addition, the most important thing is that primary school students would like to choose commuting modes according to their own preferences and habits (De Bruijn and Gardner, 2011).
Compared with the personal factors affecting the commute-mode choices of primary school students, the external factors are more complex and changeable (Pocock et al., 2019). At the level of environment, Ferri-Garcia et al. (2020) found that commuting distance and climate conditions will become factors that cannot be ignored to affect students’ choices of commuting modes. Larsen et al. (2009) concluded that higher land utilization and the existence of street trees increase the possibility of primary school students’ low-carbon commuting (walking or cycling). In addition to the environmental level, there is also the family level that is worth considering. Xiao et al. (2021) discovered that household vehicle ownership plays an important role in the choice of commuting modes for primary school students, and its influence exceeds personal characteristics and the building environment around the school. Corral-Abos et al. (2021) found that parents’ attitude and family social status are significant factors influencing primary school students’ commuting enthusiasm. In addition to attitude, parents’ education level, occupation and commuting modes are also closely related to the way their children commute (Chillon et al., 2009; Rodriguez-Lopez et al., 2013). Many studies have proved that the density of bus stops, the connectivity of streets, the status of schools and parents’ concerns have a significant impact on primary school students’ commute-mode choices (Kerr et al., 2006; Carlson et al., 2014; Ikeda et al., 2018; Peng et al., 2021).
Therefore, various factors, both internal and external, can affect primary school students’ commute-mode choices. Most of the above research has studied primary school students’ commuting behavior from the perspectives of characteristics, policies and other common influencing factors, and has obtained many clear conclusions. But there are also some problems, such as insufficient sample size, and ignoring the impact of commuting time (Broberg and Sarjala, 2015). The results of most studies only reveal the relationship between some perceptible factors and the commute-mode choices of primary school students (Beck and Nguyen, 2017).
In addition, the models used in some studies, such as the multinomial Logit model (MNL) and ordinary least squares (OLS) model, do not consider the zonal heterogeneity of primary school students’ commuting behavior (Khan et al., 2013; Ipek et al., 2019; Janacek and Rybacek, 2020). However, we found that the choice of primary school students’ commuting modes showed strong characteristics in space under the background of regional segmentation. This problem has not been taken into account in existing studies. Thus, it is difficult for us to master the zonal heterogeneity of primary school students’ commute-mode choices.
This study will explore the zonal heterogeneity of primary school students’ commute-mode choices under the background of regional segmentation based on the GWR model that can express zonal heterogeneity through the commuting survey data of primary school students in Beijing. In addition to considering the direct factors such as commuting time and commuting distance, this paper also takes the lead in studying the impact of some complex indirect factors (such as family housing category and commuting escort types) on the commuting mode of primary school students.

3 Study area and data

This section introduces the study area and data, and makes a descriptive statistical analysis of the data.

3.1 Study area

Beijing, the capital city of China and the economic hub of northern China as well. It governs 16 districts, including Dongcheng and Xicheng, with a population of 21.89 million in 2020 and a total area of 16,410.54 square kilometers. In 2021, Beijing’s GDP reached more than 4 trillion yuan, second only to Shanghai in China (Netease, 2022b). Fueled by its robust economy, the travel mode appears to be diversified in Beijing. Residents living in urban areas have shared bicycles, buses, subways and other travel modes for commuting (Zhang et al., 2014; Jin et al., 2019).
The central urban area (districts of Dongcheng, Haidian, Xicheng, Fengtai, Chaoyang, and Shijingshan), as the population gathering area of Beijing, covers an area of 1385 square kilometers, with a population of 10.988 million, that is, 8.4% of the area covers 50.2% of the city’s population (PGBM, 2022). In 2021, there were more than 270 primary schools in the central urban area, with more than 610,000 primary school students (BXN, 2022; BYN, 2022). In this area, various commuting modes are available to primary school students, including walking, bicycle, electric bicycle, subway, school bus, shuttle bus, motorcycle, taxi and private cars, which is of great research value (Ma et al., 2019). Therefore, the study area of this paper is concentrated in the central urban area of Beijing. In the study area, we explore the characteristics of primary school students’ commuting modes based on the data from Beijing School Commute Survey. The traffic analysis zones (TAZs) in this paper are divided in 2023 according to the spatial distribution of the data, which is based on the administrative divisions of the central urban area of Beijing. The administrative divisions have been in use since the adjustment in 2015. The study area contains a total of 29 TAZs. Figures 1a and 1b are maps of the study area and the TAZs respectively.
Figure 1 Map of central urban area of Beijing (a) and TAZs (b)

3.2 Data

This section first introduces the source of the research data (Section 3.2.1), reveals the survey data results, makes descriptive statistics on the results (Section 3.2.2), and sets explanatory variables and dependent variables (Section 3.2.3).

3.2.1 Data source

The data adopted in this study originates from Beijing School Commute Survey in the Fifth Comprehensive Transportation Survey of Beijing, which is the latest comprehensive transportation survey data available in Beijing. The survey involves 198 primary schools in Beijing’s central urban area with some of them being top primary schools, such as Zhongguancun No.3 Primary School and Wuyi Primary School. The survey was conducted based on online questionnaires, using an undifferentiated simple random sampling method to sample schools and students within schools. The sampled students are required to fill out an online questionnaire and they must provide detailed commuting information before completing the questionnaire. Figure 2 shows the spatial distribution of primary schools involved in the study. Each dot represents a school, and the color of the dot represents the sample size collected in this school.
Figure 2 Spatial distribution of schools and number of samples
The questionnaire covers personal characteristics, family factors and school factors related to primary school students’ commuting, mainly including the longitude, the latitude, the region of school, commuting time and distance of primary school students, family housing category and commuting escort types. The data were preprocessed before analysis, and the data that did not meet the research purpose were deleted, such as schools located beyond the study area, students who attended classes at home, and students whose commuting distance was less than 1 km (Dai et al., 2019; Jiang et al., 2022). A total of 21,232 valid data were obtained in the final survey. Table 1 shows some results of the primary school students’ commuting survey.
Table 1 Part of the results of the commuting survey among primary school students
Object ID Mode School Student Family
Name Longitude Latitude Region Grade Commuting time Commuting distance Housing category Commuting escort type
498** Walk B** 116.36** 39.9** Xicheng Five AM7:30 1.167 km Self-owned None
39** Bicycle T** 116.32** 40.0** Haidian Six AM7:15 2.709 km Tenancy None
304** Car T** 116.32** 40.0** Haidian Six AM7:30 3.454 km Self-owned Parents

Note: To protect the privacy of students, object ID, name, longitude and latitude of the school are not fully displayed in this table.

3.2.2 Data description

The survey results show that there are 11 commuting modes for primary school students in the central urban area of Beijing, including walking, cycling, two-wheel electric bicycles, three-wheel electric bicycles, buses, subways, school buses, shuttle buses, motorcycles, taxis and private cars. Among them, the shuttle bus is a commuting mode provided by the parents’ companies, and parents are permitted to pick up and drop off their children on the way of the shuttle buses. The survey covers six districts, namely Dongcheng, Xicheng, Haidian, Chaoyang, Fengtai and Shijingshan. The respondents were all primary school students from grade four to grade six. There are three commuting escort types: parents, elders and nobody. Commuting time is basically concentrated between 6:00 and 9:00. The commuting distance ranges from 1 to 20 km. The distance from the city center to school range from 1.6 to 53.9 km. The housing category is divided into three categories, including self-owned, tenancy and lodge.
Relevant research based on the questionnaire shows that personal characteristics, family characteristics and school characteristics are all important factors that affect the choice of commuting modes of primary school students (Molina-Garcia and Queralt, 2017; Masoumi et al., 2018; Buli et al., 2022). The 11 commuting modes, from walking to car, are labeled 1 to 11 based on the survey results from a low-carbon perspective. Environmentally friendly commuting modes were assigned low values and vice versa for high values. It is worth mentioning that low-carbon travel in this paper is defined as behavior that helps reduce carbon emissions from the transportation sector, which is consistent with the research by Song et al. (2022). Previous studies (Nakamura and Hayashi, 2013; Li et al., 2018; Rong et al., 2022) have confirmed that walking, cycling, electric bicycles, buses, school buses, and subways are low-carbon travel modes because these travel modes significantly reduce carbon emissions from the transportation sector. Therefore, the commuting behavior of primary school students choosing these travel modes is identified as low-carbon travel in this paper. And the paper makes a descriptive statistical analysis of six types of characteristics: grade, commuting time and distance of primary school students, the region of schools, family housing category and commuting escort types. Figures 3 to 6 show the statistical results.
Figure 3 Low carbon characteristics of commuting modes of primary school students in different regions
Figure 4 Spine chart of primary school students’ grade (a), commuting time (b) and the commuting mode choice (Note: 1 to 11 in the vertical coordinate represent walk, bicycle, two-wheel electric bicycle, three-wheel electric bicycle, bus, subway, school bus, shuttle bus, motorcycle, taxi and car, respectively.)
Figure 5 Violin chart of primary school students’ escort type (a), housing category (b) and the commuting mode choice (Note: 1 to 11 in the vertical coordinate represent walk, bicycle, two-wheel electric bicycle, three-wheel electric bicycle, bus, subway, school bus, shuttle bus, motorcycle, taxi and car, respectively.)
Figure 6 Distribution of commuting distance among primary school students
Previous studies have shown that students’ choices for commuting by bike will also be different due to the impact of the regional environment (Wati et al., 2015). Figure 3 compares the low-carbon characteristics of primary school students’ commuting modes in different regions, which is represented by the average value of the commuting modes after quantification (the darker the green, the stronger the low-carbon characteristics). As shown in Figure 3, the commuting mode coefficients of primary school students in the six regions are all lower than 6, indicating that primary school students’ commuting modes in the six regions tend to be low carbon. Among them, primary school students in Shijingshan show the most obvious low-carbon characteristics, which are significantly stronger than other regions. In contrast, there is decline in the low-carbon characteristics of the commuting mode of primary school students in Xicheng, Dongcheng and Fengtai. Haidian and Chaoyang exhibit the weakest low-carbon characteristics compared with other districts, which may be related to the economic level of these regions (Zhang et al., 2014; Wang et al., 2021; Zhang et al., 2021).
Figures 4a and 4b are two spine charts showing the commuting modes selected by primary school students of different grades and commuting time. As shown in Figure 4a, primary school students of three grades tend to be consistent in their choices of commuting modes. The majority of primary school students choose to commute on foot, followed by cars. The proportion of primary school students in Grade 5 who choose to commute on foot is the highest compared with primary school students in other grades (Grades 4 and 6), more than 30%. In addition to two commuting modes of walking and cars, the primary school students in Grades 4 and 5 prefer two-wheel electric bicycles, while the primary school students in Grade 6 prefer to commute by bus. The students in the three grades who choose to commute by bike account for a relatively similar proportion. In contrast, students of three grades rarely choose three-wheel electric bicycles, subways, school buses, shuttle buses, motorcycles and taxis to commute. It also shows that the proportion of primary school students choosing low-carbon commuting modes (walking, bicycle, two-wheel electric bicycle, three-wheel electric bicycle, bus, subway) increased with age.
Figure 4b shows primary school students’ commute-mode choices with different commuting times. It can be found that most primary school students commute between 6:00 AM and 8:00 AM. Among them, students who commute from 6:00 AM to 6:59 AM mostly choose cars, followed by buses and walking. Most students who commute from 7:00 AM to 7:59 AM are willing to walk to school, followed by cars and two-wheeled electric bicycles. As commuting modes for a minority of adults, shuttle buses, motorcycles and taxis also are rarely chosen by primary school students to commute. It is not difficult to find that, with the delay in commuting time, the proportion of low-carbon commuting modes is gradually increasing.
Figures 5a and 5b reveal the results of commute-mode choices of primary school students with different commuting escort types and different housing categories through two violin pictures. It can be seen from Figure 5a that primary school students who commute with their parents prefer to commute by bike, followed by walking. The proportion of this group choosing cars is significantly higher than that of other groups. The percentage of primary school students who go to school alone or commute with elders is inversely proportional to the degree of motorization of commuting modes. On the whole, the black bar in the first picture is more widely distributed than those in the other two pictures, indicating that primary school students who choose motorized commuting modes with their parents account for more than primary school students who commute alone or with their elders. The median point (white dot) of the three violin charts is concentrated at the bottom, indicating that most primary school students choose low-carbon commuting, whether accompanied by parents or elders or going to school alone.
Figure 5b shows primary school students’ commute-mode choices under different housing categories in the form of the violin chart. The proportion of primary school students with their own houses who choose to commute by bike is the largest, followed by walking and cars, which shows polarization. The number of primary school students who rent and lodge a house is proportional to the degree of low carbonization of the commuting mode. Students who lodge in houses have a smaller difference in the choice of different commuting modes. In addition, the position of the median point of the last two pictures in the three pictures is significantly lower than that of the first, indicating that the proportion of lodging and renting primary school students choosing low-carbon commuting modes is more than that of primary school students with their own houses.
As shown in Figure 6, most primary school students’ commuting distance is within 5 km, which belongs to short-distance commuting (De Angelis et al., 2021). Among them, more than 50% of primary school students have a commuting distance of no more than 2.5 km.
Short-distance commuting can alleviate students’ fatigue in commuting, and also help students put more energy into their studies. However, some students experience extreme commuting (Raza et al., 2021). Their commuting distance is more than 10 km, or even more than 15 km. This group deserves attention.

3.2.3 Variable setting

In this paper, the choice of commuting modes of primary school students is set as the dependent variable, which is a categorical variable. The lower the value, the more low-carbon it is. In order to comprehensively explore the impact mechanism of personal characteristics, family factors and school factors on the commuting mode of primary school students, grade, commuting time and distance of primary school students, family housing category, distance from the city center and commuting escort types are taken as explanatory variables. Except that commuting distance is a continuous variable, other variables are classified variables. Table 2 describes the definitions of the dependent variable and explanatory variables, and assigns values to different situations of each classified variable.
Table 2 Description of the dependent variable and explanatory variable
Type Attribution Name Definition Assignment Category
Dependent variable Classified
variable		 Commuting
mode		 Primary school students’
commute-mode choices		 1 Walk
2 Bicycle
3 Two-wheel electric bicycle
4 Three-wheel electric bicycle
5 Bus
6 Subway
7 School bus
8 Shuttle bus
9 Motorcycle
10 Taxi
11 Car
Explanatory
variable		 Classified
variable		 Grade Grades of primary school
students		 4 Grade four
5 Grade five
6 Grade six
Classified
variable		 Commuting
time		 Primary school students’
commuting time		 2 Commuting time from 6:00 to 6:59
3 Commuting time from 7:00 to 7:59
4 Commuting time from 8:00 to 8:59
5 Commuting time after 9:00
Classified
variable		 Commuting
distance		 Primary school students’
commuting distance		 [1, 20) Commuting distance range from 1 to 20 km
Classified
variable		 Distance from
the city center		 Distance from the city
center to school		 [1.6, 53.9] Distance from the city center to school range from 1.6 to 53.9 km
Classified
variable		 Commuting
escort type		 Primary school students’
commuting escort type		 1 Parents
2 None
3 Elderly
Classified
variable		 Housing category Housing categories of primary school students’ families 1 Self-owned
2 Tenancy
3 Lodge

Note: The center of Beijing is Tiananmen Square at coordinates (39.9087N, 116.3974E).

4 Research methodology

This section will introduce the research methods of the influence of explanatory variables on primary school students’ commute-mode choices, mainly including multicollinearity (Section 4.1), and spatial autocorrelation (Section 4.2). The basic principles of OLS (Section 4.3) and GWR (Section 4.4) are introduced and compared.

4.1 Multicollinearity

Multicollinearity refers to the high correlation between two or more variables in a linear regression model, which is a common problem in linear regression models. Multicollinearity will affect the estimation of ordinary least squares and the interpretation of the model, and this problem needs to be avoided. To ensure the reliability and effectiveness of the model, we use variance inflation factor (VIF) to verify whether there is multicollinearity or the degree of multicollinearity between variables. Previous studies have shown that variables with VIF values greater than 10 have multicollinearity and should be deleted (Lavery et al., 2019; Tamura et al., 2019; Ghorbani, 2020).

4.2 Spatial autocorrelation

Spatial autocorrelation is a method to test the spatial dependence of variables according to their values and spatial positions. This test usually evaluates the significance of variables by Moran’s I index, Z-score and P-value (Moran, 1950; Chen, 2021). The expression formula of spatial autocorrelation is as follows:
$I=\frac{n}{{{S}_{0}}}\frac{\sum\limits_{i=1}^{n}{\sum\limits_{j=1}^{n}{{{W}_{i,j}}{{Z}_{i}}{{Z}_{j}}}}}{\sum\limits_{i=1}^{n}{Z_{i}^{2}}}$
where Zi is the difference value between the attribute value of location i and its average value $\bar{x}$, Wi,j is the spatial weight between location i and j, n is the number of spatial units. S0 is the set of all spatial weights:
${{S}_{0}}=\sum\limits_{i=1}^{n}{\sum\limits_{j=1}^{n}{{{W}_{i,j}}}}$
In the spatial autocorrelation test, the value of Moran’s I index will be distributed between -1 and 1 after normalization. A value of Moran’s I greater than 0 suggests a spatial positive correlation. That is, the correlation is more significant with the aggregation of spatial distribution positions (distance). Moran’s I is a negative value, which means that the correlation increases with the increase of the dispersion of the position (Moran, 1950; Darand et al., 2017). In addition, the proposed null hypothesis is that variables are randomly distributed and cannot be predicted. The Z-score and P-value are the basis for judging whether the null hypothesis can be overturned. For example, if the Z-score is not between -2.58 and +2.58, it means that the confidence level for rejecting the null hypothesis is 99% (Cima et al., 2018). The calculation formula of the Z-score is as follows:
$Z(I)=\frac{I-E(I)}{\sqrt{Var(I)}}$
where the calculation formula of E(I) and Var(I) is:
$E(I)=-1/(n-1)$
$Var(I)=E({{I}^{2}})-E{{(I)}^{2}}$
The confidence level set in this study to overturn the null hypothesis is 99%, that is, the significance level set in this study is p<0.01.

4.3 Geographically weighted regression

There are correlations between many variables. For example, the degree of mobility of commuting will increase with the increase in commuting distance. The OLS model is commonly used to analyze the relationship between explanatory variables and dependent variables. Many studies (Ogura, 2010; Molina et al., 2020) have studied the impact of various factors on students’ commuting modes with the help of the OLS model. The regression coefficient value of this model is estimated by the minimum value of the square of the residual error (Brunsdon et al., 1996), and its expression is as follows:
${{Y}_{i}}={{\beta }_{0}}+\sum\limits_{k}{{{\beta }_{k}}{{X}_{ki}}+{{\varepsilon }_{i}}}$
where i(i=1, 2, …, n) represents the ith TAZ, Yi is the dependent variable, β0 is the intercept, βk represents the regression coefficient of the kth explanatory variable, Xki is the value of the kth explanatory variable in the ith sample and εi is the random error.
It should be realized that OLS regression cannot express spatial characteristics, and the estimated regression coefficient of the model remains constant in different traffic analysis areas throughout the study area (Li et al., 2020; He et al., 2021). The commuting mode of primary school students is affected by the regional economy and transportation infrastructure, and will change with the variation in spatial location. The GWR model considers spatial autocorrelation and zonal heterogeneity between variables, so it is usually used to process spatial data regression (Lu et al., 2017; Ma et al., 2020; Comber et al., 2022). Compared with the OLS model, the GWR model considers the longitude and latitude of the sample in the regression coefficient, effectively connects the explanatory variables with the geographical location, and quantifies the spatial effect by carrying out a separate OLS regression for each region. It can not only effectively estimate the data with spatial autocorrelation, but also reflect the spatial nonstationarity of parameters in different regions (Matthews and Yang, 2012; Du et al., 2022). The calculation formula of the GWR model is as follows (Brunsdon et al., 1996):
${{Y}_{i}}={{\beta }_{0}}({{u}_{i}},{{v}_{i}})+\sum\limits_{k}{{{\beta }_{k}}({{u}_{i}},{{v}_{i}}){{X}_{ki}}+{{\varepsilon }_{i}}}$
where i(i=1, 2, …, n) represents the ith TAZ, Yi is the dependent variable, β0(ui, vi) is the intercept, βk(ui, vi) denotes the regression coefficient of the kth explanatory variable in the ith sample, Xki is the value of the kth explanatory variable in the ith sample, ui represents the longitude, vi denotes the latitude, (ui, vi) represents the longitude and latitude coordinates of TAZ in space and εi is the random error.
According to the first law of geography (Tobler, 2004), everything is spatially related, and the closer things are, the greater the spatial correlation is. This theory is implemented in the GWR model by calculating the weights of different TAZs through local weighted least square estimation. In the model construction, take β(u0, v0) as an example, the local least squares estimate of β(u0, v0) at (u0, v0) is:
$\hat{\beta }({{u}_{0}},v{}_{0})={{[{{X}^{T}}W({{u}_{0}},v{}_{0})X]}^{-1}}{{X}^{T}}W({{u}_{0}},v{}_{0})Y$
where $W({{u}_{0}},{{v}_{0}})=\text{Diag}{{[{{w}_{1}}({{u}_{0}},{{v}_{0}}),\ {{w}_{2}}({{u}_{0}},{{v}_{0}}), ...,\ {{w}_{n}}({{u}_{0}},\ {{v}_{0}})]}^{T}}$; $X=({{X}_{0}},{{X}_{1}},...,\ {{X}_{k}})$; Xj = ${{({{x}_{0j}},{{x}_{1j}},...,{{x}_{nj}})}^{T}};$ $Y={{({{Y}_{1}},{{Y}_{2}},...,{{Y}_{n}})}^{T}},$ wi(uj, vj) (1≤j≤n) is the spatial distance decay function, which is described as follows (Yacim and Boshoff, 2019):
${{w}_{i}}({{u}_{j}},{{v}_{j}})=\left\{ \begin{align} {{\left[ 1-{{\left( \frac{{{d}_{ij}}}{h} \right)}^{2}} \right]}^{2}},\ \left| {{d}_{ij}} \right|<h & \\ 0,\left| {{d}_{ij}} \right|\ge h & \\ \end{align} \right.$
where dij is the spatial distance between TAZs i and j. It was obtained by using Euclidean distances, and the expression is as follows (Selby and Kockelman, 2013):
${{d}_{ij}}=\sqrt{{{({{u}_{i}}-{{u}_{j}})}^{2}}+{{({{v}_{i}}-{{v}_{j}})}^{2}}}$
In Eq.9, h is bandwidth, which is a non-negative attenuation parameter of the functional relationship between weight and distance. The larger the bandwidth is, the slower the spatial weight Wij decays with the increase of the two-point distance dij. In the GWR model, the commonly used criterion for selecting the optimal bandwidth is the cross-validation (CV) criterion (Hurvich et al., 1998; Da Silva and Mendes, 2018). Its expression formula is as follows:
$\text{CV}(h)=\sum\nolimits_{i=1}^{n}{{{({{Y}_{i}}-{{{\hat{Y}}}_{\ne i}}(h))}^{2}}}$
where n is the sample size, ${{\hat{Y}}_{\ne i}}(h)$ represents the predicted value of Yi (the observation value excluded from the regression model). The bandwidth corresponding to the lowest CV value is the optimal bandwidth. In addition, The Corrected Akaike Information Criterion can also be used to determine the optimal window width (Hurvich et al., 1998; Luo and Peng, 2016). The formula is as follows:
$\text{AICc}=2n\text{ln(}\hat{\sigma })+n\text{ln}(2\pi )+n\frac{n+tr(S)}{n-2-tr(S)}$
where $\hat{\sigma }$ is the estimated standard error of residuals, tr(S) is the trace of the hat matrix which is the function of the bandwidth (Tian et al., 2012). Similar to the CV value, the optimal bandwidth occurs when the AICc value is the lowest. In this paper, the AICc value is finally used as the method to determine the optimal bandwidth. R language is used to construct both the OLS model and the GWR model in this study.

5 Results

This section will focus on the fitting effect between the OLS model and the GWR model while visualizing the parameter values of the explanatory variables in the GWR model and analyzing the results. Section 5.1 compares the fitting effect of the OLS model and the GWR model on the commute-mode choices of primary school students by comparing R2 and AICc values. Section 5.2 visualizes the parameters of each explanatory variable obtained from the GWR model (Section 5.2.1), and analyzes the results of each variable (Section 5.2.2).

5.1 Model comparison

As the preliminary work of the OLS model and GWR model, it is indispensable to carry out multicollinearity and spatial autocorrelation tests around various explanatory variables. This paper calculates the VIF values of each explanatory variable to verify whether there is multicollinearity between variables (Kouda et al., 2019; Vorosmarty and Dobos, 2020). Table A1 lists the VIF values of variables. Then calculate the Moran’s I value, P-value and Z-score of each effective variable based on the results obtained in Table A1 to verify whether the variable has spatial autocorrelation (Dray, 2011; Yang et al., 2011; Hamylton and Barnes, 2018). The calculation results are given in Table A2.
In this paper, the GWR and OLS models are constructed to fit the impact of primary school students’ commute-mode choices, and the fitting results are shown in Table 3 with R2 and AICc values. The fitting results showed that the R2 value of the commuting mode model of primary school students increased from 0.2733 in the OLS to 0.4933 in the GWR, an increase of 180%. AICc value decreased from 112,456.9 in the OLS to 112,234.8 in the GWR. GWR apparently demonstrates more advantages in model fitting than the OLS model. This is due to the fact that the GWR has supplemented spatial information and fully considered the zonal heterogeneity of the influence of explanatory variables on the commute-mode choices of primary school students. Therefore, there is significant improvement in the persuasiveness of GWR interpretation results.
Table 3 Comparison of fitting effect between OLS and GWR
AICc R2
OLS 112456.9 0.2733
GWR 112234.8 0.4933

Note: R2 of the GWR is the optimal value.

On this basis, this paper examines the spatial distribution of the standard residuals of the GWR model for primary school students’ commute-mode choices. The inspection results are shown in Figure 7. All TAZs in the whole study area passed the residual test (less than 2.5 times the standard deviation) (Zhao et al., 2020; Du et al., 2022).
Figure 7 Spatial distribution of standard residuals for the GWR model
Some of the model results are echoed by findings from previous work. For example, the grade is inversely proportional to the degree of motorization of commuting modes (Booth et al., 2007), and the increase of commuting distance will promote the motorization of commuting modes (Kelly et al., 2014). In addition, this paper also puts forward some new insights: primary school students’ commute-mode choices in the core areas of Beijing (Dongcheng and Xicheng districts) are more low-carbon than that in the outer ring area. Primary school students who commute with their parents are more willing to choose motorized commuting modes. Compared with other housing categories, primary school students with self-owned housing are more motivated to choose motorized commuting modes.
The results of multicollinearity and spatial autocorrelation tests show that the variables involved in this paper are all effective variables with spatial autocorrelation and non-multicollinearity. Therefore, the next section will focus on the analysis of the parameter results of six variables: grade, the distance from the city center, commuting time, commuting escort type, commuting distance, and housing category. These variables cover personal characteristics, family factors and school factors that affect the commute-mode choices of primary school students. It has previously been proved that the GWR model is better than the OLS model, so to ensure the simplicity of the analysis, the subsequent analysis in this paper is based on the GWR parameter results. Six statistics are selected in this paper, namely maximum value (MAX), upper quartile (UQ), median (MED), mean value (MEAN), lower quartile (LQ) and minimum value (MIN). The specific parameter results are shown in Table 4.
Table 4 Estimated Value of the GWR Model for the commuting mode of primary school students
Variable MEAN MIN LQ MED UQ MAX
Grade -0.0767 -0.2053 -0.0781 -0.0732 -0.0697 -0.0026
Distance from the city center 0.0423 -0.0409 0.0355 0.0431 0.0504 0.0820
Commuting time -1.4454 -2.2218 -1.4709 -1.4489 -1.4062 -1.1439
Commuting escort type -1.5022 -1.7208 -1.5189 -1.4931 -1.4809 -0.8113
Commuting distance 0.3081 0.2225 0.3057 0.3061 0.3076 0.3336
Housing category -1.0908 -1.2200 -1.1242 -1.0991 -1.0803 -0.0617

5.2 Spatial feature of variable coefficients

The main difference between the GWR model and the OLS model is that it considers the spatial distribution of the coefficients, and the estimated coefficients can reflect the influence and its zonal heterogeneity of explanatory variables on the commuting mode of primary school students in different regions. Visualizing the parameter results is the most intuitive and effective way to analyze the fitting results of the GWR model. Thus, this study has drawn some spatial distribution maps of the impact on primary school students’ commute-mode choices, as shown in Figures 8-10. This section will analyze the parameter results and spatial distribution of each explanatory variable according to personal characteristics, family factors and school factors.
(1) Personal characteristics
The personal characteristics of primary school students mainly include grade, commuting distance and commuting time. Figures 8a, 8b and 9a reflect the spatial distribution of average estimation coefficients of the grade, commuting distance and commuting time.
As shown in Figure 8a, the average coefficient of the grade is negative in all TAZs, indicating that the grade is negatively related to the degree of motorization of primary school students’ commuting modes. This is mainly because with the increase of the grade, the safety awareness and independent commuting ability of primary school students will also be enhanced, and there will be a gradual rising tendency to choose low-carbon and convenient commuting modes such as walking or cycling in commuting hours (Van Goeverden et al., 2013). At the same time, it can be found that the influence of the grade change in the western region is stronger than that in the eastern and central regions. This is because the western region, especially Haidian district is home to hundreds of schools, and roads are susceptible to heavy congestion during commuting hours. Primary school students choose low-carbon commuting modes such as cycling or walking for efficiency. At the same time, their commuting habits will spread to other individuals, and the proportion of students choosing low-carbon commuting modes will further increase. In addition, Haidian district is aggressively promoting the development of the slow traffic system, with a focus on taking multiple measures to improve the convenience of the slow traffic system (Netease, 2022a), such as the construction of dedicated bicycles lanes, to further enhance the attractiveness of the bicycle commuting for primary school students. Therefore, the grade change in this area has a greater impact on the choice of low-carbon commuting modes for primary school students.
Figure 8 Spatial distribution of average coefficients of the grade (a) and commuting distance (b) for the GWR model
Figure 8b shows the spatial distribution of the average coefficient of commuting distance. Consistent with conventional cognition (Zhao et al., 2018), the coefficient value in the whole region is positive, that is, the longer the commuting distance is, the more likely the primary school students are to choose the motorized commuting modes. It can be seen from the whole study area that the positive correlation between the east and west sides is significantly stronger than that in the middle, and the significance in the northern region is greater than that in the southern region. Compared with the central region and the southern region, the eastern, western and northern regions are mainly economically developed districts such as Chaoyang and Haidian, with a lot of vehicles, more extensive bus and subway lines, and more high-quality schools with school bus operating conditions (Tencent, 2022). Thus, the longer the commuting distance of students in these areas, the greater the possibility of choosing motorized commuting modes.
Figure 9a represents the spatial distribution of the average coefficient of primary school students’ commuting time. The characteristics of the whole region show consistency, and the delay in primary school students’ commuting time will further enhance the possibility of using non-motorized commuting modes. The lag of commuting time indicates that students are not strict about commuting time, and low-carbon commuting modes will be prioritized under the same choice. It is worth noting that the absolute values of the coefficients in the central and southern regions are higher than those in other regions. It shows that with the delay in commuting time, the possibility of primary school students in this area choosing low-carbon commuting is also increasing. Primary school students in the southern region are more flexible in the choice of commuting modes since their commuting time is believed to be less restricted than that of students in Haidian district and other northern regions. Under this condition, the possibility of choosing low-carbon commuting modes increases. The central area, mainly Beijing’s old urban area, is featured by narrow roads, which make it more convenient to commute by bike at various times (YNET, 2022).
Figure 9 Spatial distribution of average coefficients of commuting time (a) and the distance from the city center (b) for the GWR model
(2) Family factors
The family factor is composed of the commuting escort type and the housing category. The spatial distribution of the average coefficients of the commuting escort type and the housing category is shown in Figures 10a and 10b.
Figure 10 Spatial distribution of average estimation coefficients of commuting escort types (a) and housing categories (b) for the GWR model
Figure 10a shows the spatial distribution of the average coefficient of commuting escort types. On the whole, there is a negative correlation between the commuting escort types and the motorized degree of the commuting modes of primary school students. Compared with commuting with parents, primary school students prefer walking, cycling and other low- carbon modes when commuting alone. This is in line with common sense. When primary school students use motorized commuting (such as private cars and motorcycles), they need escort from adults. However, they do not need their parents to follow them when they walk and ride bicycles to commute (Wen et al., 2008). It is worth mentioning that primary school students are more likely to choose low-carbon commuting modes with their elders than the first two. This is mainly related to the mobility difficulties and frugality of the elderly. On the whole, the influence of commuting escort types on the commuting mode of primary school students gradually decreases from west to east. On the one hand, this is because some primary school students have only one way to commute, whether alone or accompanied by parents or the elderly, such as walking, bus or subway. On the other hand, some primary school students’ commuting modes are always consistent, whether they are accompanied by their parents and the elderly, such as two-wheeled electric bicycles and private cars. The number of students in both types has also increased from west to east.
The spatial distribution of the average coefficient of the housing category is shown in Figure 10b. The coefficient values are negative in the whole region, indicating that the housing category is negatively related to the willingness of primary school students to choose motorized commuting modes. Self-owned housing means that their families have a good economic foundation, so primary school students of this type are more likely to choose motorized commuting modes such as cars. It can be found that the degree of negative correlation gradually increases from southeast to northwest, mainly because Haidian district is rich in educational resources. It has many famous schools, including Zhongguancun No.1 Primary School and Beijing Haidian Experimental Primary School (Wikipedia, 2022). Families in this area generally value their children’s education (Chen and Bai, 2015). To minimize children’s fatigue in commuting, parents will choose to rent or borrow houses near the school, greatly reducing commuting distance. Therefore, primary school students tend to choose low-carbon commuting modes, such as walking or cycling.
(3) School factors
The school factor mainly involves the distance from the city center. Figure 9b shows the spatial distribution of the average coefficient of the distance from the city center. It can be observed that all regions show positive correlation, except for a small portion of the central region showing weak negative correlation. In other words, the degree of low-carbon commuting modes among primary school students increases as the distance from the city center increases. The reason is that as the distance from the city center increases, the school becomes more remote. This leads to a decrease in the coverage of low-carbon commuting modes such as buses and shared bicycles, reducing the likelihood of primary school students choosing low-carbon commuting modes. It is worth noting that the positive correlation is gradually increasing from northwest to southeast. Compared to the northwest region, primary school students in the southeast region are more willing to choose non low-carbon commuting modes based on the same distance. This is mainly due to the fact that the developed economy in the southeastern region, and most families have a good economic foundation. Primary school students will give priority to convenient commuting modes such as private cars under the same conditions.

6 Policy implication

This paper discusses the impact of personal characteristics, family factors and school factors on primary school students’ commute-mode choices, and summarizes the research results on this basis. It has important policy significance, especially in areas with large numbers of students. According to the research results, this study puts forward detailed and targeted policy suggestions about primary school students, families and schools from the perspective of policymakers, aiming to enhance the care of primary school students, alleviate road congestion and promote low-carbon commuting of primary school students.

6.1 Design and develop the urban slow traffic system

As pointed out by Zhang and Wan (2010), the slow traffic system has the advantages of low travel cost, green and environmental protection, which is an indispensable part of the urban traffic system. The results of this study show that the degree of construction of the slow traffic system in a region determines primary school students’ interest in choosing low-carbon commuting modes. Thanks to the construction of the slow lane, the primary school students in Haidian have strong enthusiasm to choose low-carbon commuting modes. Based on this fact, we encourage all regions across the city to actively optimize the low-carbon commuter travel environment and improve the safety, continuity and convenience of the slow traffic system by taking measures such as creating slow traffic dedicated roads, widening bicycle lanes, setting up bicycle post station, reducing the crossing distance at intersections, placing the sign of “Pedestrian Priority” and emplacing safety island for pedestrians crossing.

6.2 Develop and promote students’ out-tracking system

As shown in the research results, primary school students tend to choose walking, cycling and other modes when commuting alone. However, due to weak cognitive ability and risk awareness of primary school students, and the large number of vehicles during rush hours, the possibility of danger in the process of commuting is further exacerbated. In view of this, the government and families should strengthen the monitoring of primary school students who commute alone. Specifically, the government can consider cooperating with technology companies to develop a path monitoring system for primary school students, and install it in the daily items carried by primary school students, such as watches and shoes. When the primary school students’ commute track or speed is abnormal, the guardian shall be notified in time to pay attention. This measure can provide security for students who commute alone and avoid accidents. Additionally, the government also needs to introduce policies to supervise such systems to avoid personal privacy disclosure.

6.3 Expand the coverage of school public transport in peripheral areas

The research results confirm that schools in remote areas increase the possibility of primary school students choosing motorized commuting modes because of the low coverage of subway and bus in the surrounding areas. Shaaban and Abdur-Rouf (2021) confirmed that improving the infrastructure around the school can greatly enhance the enthusiasm of more children to walk and ride bicycles to school. Therefore, we suggest that the government should take measures such as adding bus stops and subway stations around schools in remote areas, expanding the coverage of shared bicycles, and supplementing the area of bicycle parking areas to improve the level of transportation infrastructure in schools in remote areas, so as to provide more convenient conditions for primary school students to choose low-carbon commuting modes.

6.4 Strengthen the traffic flow control of housing around the school

The results of this study show that some families choose to rent or borrow school district houses in order to reduce their children’s fatigue in commuting. These houses are close to schools, so primary school students choose to commute by walking or cycling (Lin, 2018). The road traffic flow around the school is always busy and congested during commuting hours, which also poses a threat to the travel of primary school students. Therefore, it is suggested that the traffic management department should take measures such as vehicle speed limit, setting no parking areas, and increasing the punishment for illegal acts to strengthen the traffic flow management around the school during commuting hours. In addition, the government should also strengthen publicity to promote the whole society to form a sense of comity to primary school students, which can provide long-term safety guarantee for various travel behaviors of primary school students, including commuting.

6.5 Establish school gate management regulations and prolong the security time of school gates

In 2019, Beijing issued Regulations of Beijing Municipality on Safety Management of Primary and Middle School Kindergarten (Beijing Municipal Education Commission, 2022). It is clearly pointed out that there are security guards and police working together to maintain order at the school gate. The research results confirm that the commuting time of primary school students in some regions is widely distributed and flexible. Therefore, it is proposed that the government cooperate with the campus to establish complete school gate management regulations, especially during commuting hours. Among them, the security work time should be appropriately lengthened on the basis of the current time, giving full consideration to every primary school student.

6.6 Formulate preferential policies for family commuting on public transport

The study found that primary school students tend to choose motorized commuting modes such as private cars and motorcycles when commuting with their parents. On the one hand, private cars offer more convenience with more freedom in travel time. On the other hand, the travel attraction of public transport needs to be improved. Based on this problem, we suggest the government formulate preferential policies in public transport for family commuting. During the commuting period, parents and students can enjoy preferential policies such as ticket discount and low-carbon travel points when taking public transportation such as buses and subways. This policy can reduce the commuting cost of families and promote the low-carbon structure of commuting modes.

7 Discussion and conclusions

7.1 Classify and assign commuting modes based on sample size

The classification and assignment of the dependent variable determine the accuracy and effectiveness of the zonal heterogeneity analysis. This section will consider the classification and assignment of commuting modes based on the sample size from a low-carbon perspective. Commuting modes were eventually classified into four categories: strong low carbon, slightly strong low carbon, slightly weak low carbon, and weak low carbon. Four categories were assigned values from 1 to 4. The classification results are shown in Table 5.
Table 5 The division and assignment of commuting modes based on the sample size
Commuting mode Category Assignment
Walk Strong low carbon 1
Bicycle Slightly strong low carbon 2
Two-wheel electric bicycle
Three-wheel electric bicycle Slightly weak low carbon 3
Bus
Subway
School bus
Motorcycle
Shuttle bus
Taxi
Car Weak low carbon 4
Based on the classification result, this paper takes the commuting escort type as an example to compare two classification methods and discuss their correlation and differences. The comparison results are shown in Figure 11. It can be found that compared to the classification method based on the sample size, the classification method based on the sample type enhance the overall correlation between variables, while the negative correlation in the upper-middle region weakens compared to other regions. However, it is also worth noting that the overall characteristics of the correlation between variables have been ensured in the classification method based on the sample type, with a gradually increasing negative correlation from southeast to northwest. Additionally, the relative correlations of all regions except the upper-middle region remain consistent across the study area, which ensures the effectiveness of zonal heterogeneity analysis.
Figure 11 Spatial distribution of average estimation coefficients of commuting escort types in classification methods based on the sample type (a) and the sample size (b)

7.2 General conclusions and future directions of research

Based on the data from Beijing School Commute Survey, combined with various characteristic variables, this paper discussed the influencing factors of primary school students’ commuting modes. The GWR and OLS models were built to fit the research data and compared the fitting results. From three aspects of personal characteristics, family factors and school factors, this paper studied the detailed information about the zonal heterogeneity of explanatory variables’ impacts on the commute-mode choices of primary school students from a low- carbon perspective. Results show that the grade, commuting time, commuting distance, family housing category, commuting escort types, and the distance from the city center have a significant impact on the choice of commuting modes. The important conclusions about the influencing factors on primary school students’ commute-mode choices are summarized as follows:
(1) Personal characteristics
The grade of primary school students is negatively correlated with the degree of motorized commuting modes. The older primary school students are, the more they prefer low-carbon commuting modes, which is more obvious in the western region. On the contrary, the increase of commuting distance will raise the possibility for primary school students to choose motorized commuting modes. Affected by the regional economic conditions, the significance of the eastern and western regions is stronger than that of the central region, and the significance of the northern region is stronger than that of the southern region. The lagging degree of primary school students’ commuting time is in direct proportion to the low- carbon degree of commuting modes. This phenomenon is more common in the central and southern regions.
(2) Family factors
Family factors mainly include commuting escort types and housing categories. The correlation coefficient between commuting escort types and motorized commuting modes of primary school students is negative, and the absolute value gradually decreases from west to east. Primary school students who commute alone or accompanied by elders are more inclined to choose low-carbon commuting modes. The housing category is negatively related to primary school students’ willingness to choose motorized commuting modes, and the degree is gradually deepening from southeast to northwest. Compared with primary school students whose housing category is tenancy and lodge, primary school students with their own houses would like to choose motorized commuting modes such as private cars.
(3) School factors
The distance from the city center is positively related to the motorized degree of primary school students’ commuting modes, especially in the southeastern region. The conclusion shows that the more remote the school is, the more likely the primary school students are to choose non low-carbon commuting modes.
This study has several limitations. First, limited by the survey data, this paper only takes the central urban area of Beijing as the study area. However, regions vary in climate, economic development level and landform, and primary school students also differ in commuting modes. Second, this paper studies the travel behavior of primary school students based on the time background of commuting days, namely working days. It lacks the comparison of primary school students’ travel behavior at different times. Futhermore, this paper only considers some of the family factors and school factors, which needs to be supplemented. Future research could expand the study area to Shanghai and other developed cities in China using jiedao as the study unit. Attention also needs to be given to other family factors (e.g., household income, vehicle usage and parents’ work locations) and school factors (e.g., class schedule and the surrounding environment). On this basis, the geographically and temporally weighted regression (GTWR) model can be used to analyze the changes in the commuting mode choice of primary school students from the spatiotemporal perspective. Finally, the commuting modes and explanatory variables need to be further refined. For example, the ride-hailing service can be taken into account in commuting modes, and the gender and commuting frequency of primary school students can be taken into account in personal characteristics. At the same time, social aspects (such as building environment) and government aspects (such as traffic restriction policies) also merit consideration, which will help better understand the impact of various factors on the choice of commuting modes of primary school students.

Appendix

Table A1 lists the verification results of the multicollinearity of explanatory variables of the OLS model. In addition to the VIF value, it also includes the coefficient value and P-value. Table A2 shows the test results of spatial autocorrelation of explanatory variables, involving Moran’s I, Z-score, and P-value. Table A3 contains the legend names and descriptions of figures.
Table A1 VIF values of explanatory variables of OLS model
Variable Coef. P-value VIF
Grade -0.057 0.047 1.021
Distance from the city center 0.034 0.000 1.024
Commuting time -1.311 0.000 1.135
Commuting escort type -1.542 0.000 1.039
Commuting distance 0.317 0.000 1.159
Housing category -1.181 0.000 1.037
Table A2 Results of spatial autocorrelation test on explanatory variables
Variable Moran’s I Z-score P-value
Grade 0.094 103.306 0.000
Distance from the city center 0.651 2520.812 0.000
Commuting time 0.152 170.284 0.000
Commuting escort type 0.044 49.300 0.000
Commuting distance 0.177 197.302 0.000
Housing category 0.287 320.557 0.000
Table A3 The legend names and descriptions of figures
Figure Legend name Description
Figure 1 Map of the central urban area
of Beijing and TAZs		 Location of the central urban area of Beijing and the division results of TAZs
Figure 2 The sample size The sample size contained at each dot
Figure 3 The commuting mode coefficients The commuting mode coefficients of primary school students in different regions
Figure 7 The standard residuals The standard residuals for the GWR model
Figure 8 The coefficients for the model The coefficients of the grade and commuting distance for the model
Figure 9 The coefficients for the model The coefficients of the distance from the city center and commuting time for the model
Figure 10 The coefficients for the model The coefficients of commuting escort types and housing categories for the model
Figure 11 The coefficients for the model The coefficients of commuting escort types for the model

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Beck L F, Nguyen D D, 2017. School transportation mode, by distance between home and school, United States, ConsumerStyles 2012. Journal of Safety Research, 62: 245-251.

DOI PMID

[2]
Beijing Municipal Education Commission, 2022. Regulations of Beijing on Safety Management of Primary and Middle School Kindergarten for trial implementation. http://jw.beijing.gov.cn/jyzx/ztzl/bjjyxx/jyxxzcfg/202207/t20220706_2764772.html.

[3]
Beijing Xiaoshengchu Network BXN, 2022. Basic information on the number of primary school graduates and the number of students in all grades in Beijing in the 2021-2022 academic year. https://www.xschu.com/xiaoshengchu/17/44263.html.

[4]
Beijing Youshengxiao Network BYN, 2022. Number of primary schools and classes in each grade in Beijing in the 2021-2022 school year. https://www.ysxiao.cn/c/202203/51422.html.

[5]
Booth M L, Okely A D, Denney-Wilson E et al., 2007. Characteristics of travel to and from school among adolescents in NSW, Australia. Journal of Paediatrics and Child Health, 43(11): 755-761.

PMID

[6]
Broberg A, Sarjala S, 2015. School travel mode choice and the characteristics of the urban built environment: The case of Helsinki, Finland. Transport Policy, 37: 1-10.

DOI

[7]
Brunsdon C, Fotheringham A S, Charlton M E, 1996. Geographically weighted regression: A method for exploring spatial nonstationarity. Geographical Analysis, 28(4): 281-298.

DOI

[8]
Buli B G, Tillander A, Fell T et al., 2022. Active commuting and healthy behavior among adolescents in neighborhoods with varying socioeconomic status: The NESLA study. International Journal of Environmental Research and Public Health, 19(7): 3784.

DOI

[9]
Carlson J A, Sallis J F, Kerr J et al., 2014. Built environment characteristics and parent active transportation are associated with active travel to school in youth age 12-15. British Journal of Sports Medicine, 48(22): 1634-1676.

DOI PMID

[10]
Chen B L, Bai X X, 2015. Family socioeconomic status, peer group pressure of parents and tutoring of urban primary school students: Based on a survey of primary schools in Haidian district, Beijing. Tsinghua University Education Research, 36(5): 102-109. (in Chinese)

[11]
Chen Y G, 2021. An analytical process of spatial autocorrelation functions based on Moran’s index. Plos One, 16(4): Article e0249589.

[12]
Chen Y L, Tolfrey K, Pearson N et al., 2021. Stand out in class: Investigating the potential impact of a sit-stand desk intervention on children’s sitting and physical activity during class time and after school. International Journal of Environmental Research and Public Health, 18(9): 14: Article 4759.

DOI

[13]
Chillon P, Herrador-Colmenero M, Migueles J H et al., 2017. Convergent validation of a questionnaire to assess the mode and frequency of commuting to and from school. Scandinavian Journal of Public Health, 45(6): 612-620.

DOI PMID

[14]
Chillon P, Ortega F B, Ruiz J R et al., 2009. Socio-economic factors and active commuting to school in urban Spanish adolescents: The AVENA study. European Journal of Public Health, 19(5): 470-476.

DOI PMID

[15]
Chui C H K, Jordan L, 2018. Child protection in China: Threats and opportunities. Asia Pacific Journal of Social Work and Development, 28(2): 84-96.

DOI

[16]
Cima E G, Uribe-Opazo M A, Johann J A et al., 2018. Analysis of spatial autocorrelation of grain production and agricultural storage in Parana. Engenharia Agricola, 38(3): 395-402.

DOI

[17]
Comber A, Brunsdon C, Charlton M et al., 2022. A route map for successful applications of geographically weighted regression. Geographical Analysis: 55(1): 155-178.

[18]
Corral-Abos A, Aibar A, Estrada-Tenorio S et al., 2021. Implications of school type for active commuting to school in primary education students. Travel Behaviour and Society, 24: 143-151.

DOI

[19]
Dai T Q, Wang L, Zhang Y C et al., 2019. The cost of school consolidation policy: Implications from decomposing school commuting distances in Yanqing, Beijing. Applied Spatial Analysis and Policy, 12(2): 191-204.

DOI

[20]
Darand M, Dostkamyan M, Rehmanic M I A, 2017. Spatial autocorrelation analysis of extreme precipitation in Iran. Russian Meteorology and Hydrology, 42(6): 415-424.

DOI

[21]
De Angelis M, Prati G, Tusl M et al., 2021. Mobility behaviors of Italian university students and staff: Exploring the moderating role of commuting distances. International Journal of Sustainable Transportation, 15(8): 581-591.

DOI

[22]
De Bruijn G J, Gardner B, 2011. Active commuting and habit strength: An interactive and discriminant analyses approach. American Journal of Health Promotion: AJHP, 25(3): e27-36.

DOI

[23]
Dray S, 2011. A new perspective about Moran’s coefficient: Spatial autocorrelation as a linear regression problem. Geographical Analysis, 43(2): 127-141.

DOI

[24]
Du M Y, Cheng L, Li X F et al., 2022. Spatial variation of ridesplitting adoption rate in Chicago. Transportation Research Part A: Policy and Practice, 164: 13-37.

[25]
Elias W, Katoshevski-Cavari R, 2014. The role of socio-economic and environmental characteristics in school-commuting behavior: A comparative study of Jewish and Arab children in Israel. Transport Policy, 32: 79-87.

DOI

[26]
Ermagun A, Samimi A, 2016. How are children accompanied to school? Journal of Urban Planning and Development, 142(3): 04016002.

DOI

[27]
Fast I, 2020. Unequal traveling: How school district and family characteristics shape the duration of students’ commute to school. Travel Behaviour and Society, 20: 165-173.

DOI

[28]
Ferri-Garcia R, Fernandez-Luna J M, Rodriguez-Lopez C et al., 2020. Data mining techniques to analyze the factors influencing active commuting to school. International Journal of Sustainable Transportation, 14(4): 308-323.

DOI

[29]
Fitch D T, Rhemtulla M, Handy S L, 2019. The relation of the road environment and bicycling attitudes to usual travel mode to school in teenagers. Transportation Research Part A: Policy and Practice, 123: 35-53.

[30]
Galvez-Fernandez P, Chillon P, Timperio A et al., 2022. The patterns of active commuting to school change along the school-life in Spanish youth. International Journal of Sustainable Transportation, 1-8.

[31]
Gao Y E, Chen X H, Shan X N et al., 2018. Active commuting among junior high school students in a Chinese medium-sized city: Application of the theory of planned behavior. Transportation Research Part F: Traffic Psychology and Behaviour, 56: 46-53.

DOI

[32]
Ghorbani H, 2020. Ill-conditioning in linear regression models and its diagnostics. Journal of the Korean Society of Mathematical Education Series B: Pure and Applied Mathematics, 27(2): 71-81.

[33]
Hamylton S M, Barnes R S K, 2018. The effect of sampling effort on spatial autocorrelation in macrobenthic intertidal invertebrates. Hydrobiologia, 811(1): 239-250.

DOI

[34]
He X D, Mai X M, Shen G Q, 2021. Poverty and physical geographic factors: An empirical analysis of Sichuan province using the GWR model. Sustainability, 13(1): Article 100.

[35]
Herrador-Colmenero M, Escabias M, Ortega F B et al., 2019. Mode of commuting to and from school: A similar or different pattern? Sustainability, 11(4): Article 1026.

[36]
Hurvich C M, Simonoff J S, Tsai C L, 1998. Smoothing parameter selection in nonparametric regression using an improved Akaike information criterion. Journal of the Royal Statistical Society Series B: Statistical Methodology, 60: 271-293.

DOI

[37]
Ikeda E, Stewart T, Garrett N et al., 2018. Built environment associates of active school travel in New Zealand children and youth: A systematic meta-analysis using individual participant data. Journal of Transport and Health, 9: 117-131.

DOI

[38]
Janacek j rybacek v, 2020. Commuting time and life satisfaction of high school students in the Czech Republic, Italy and Slovenia. Prague Economic Papers, 29(5): 561-574.

DOI

[39]
Jiang H X, Liu Q F, Li Y L et al., 2022. The relationship between school districts and parental commuting behavior: Analysis of gender differences in the Chinese context. Frontiers in Environmental Science, 10: Article 1019753.

[40]
Jin H T, Jin F J, Wang J E et al., 2019. Competition and cooperation between shared bicycles and public transit: A case study of Beijing. Sustainability, 11(5): Article 1323.

[41]
Jurak G, Soric M, Ocvirk T et al., 2021. Barriers and determinants of active commuting to school in Slovenia. Sustainability, 13(24): Article 13808.

[42]
Kelly J A, Fu M, 2014. Sustainable school commuting: Understanding choices and identifying opportunities: A case study in Dublin, Ireland. Journal of Transport Geography, 34: 221-230.

DOI

[43]
Kerr J, Rosenberg D, Sallis J F et al., 2006. Active commuting to school: Associations with environment and parental concerns. Medicine and Science in Sports and Exercise, 38(4): 787-794.

DOI

[44]
Khan M, Choudhury C F, Wang J, 2011. Modeling preference for school bus service in Dhaka, Bangladesh stated preference approach. Transportation Research Record, 2239: 64-73.

DOI

[45]
Kobus M B W, Van Ommeren J N, Rietveld P, 2015. Student commute time, university presence and academic achievement. Regional Science and Urban Economics, 52: 129-140.

DOI

[46]
Kouda K, Ohara K, Fujita Y et al., 2019. Relationships between serum leptin levels and bone mineral parameters in school-aged children: A 3-year follow-up study. Journal of Bone and Mineral Metabolism, 37(1): 152-160.

DOI PMID

[47]
Larsen K, Gilliland J, Hess P et al., 2009. The influence of the physical environment and sociodemographic characteristics on children’s mode of travel to and from school. American Journal of Public Health, 99(3): 520-526.

DOI

[48]
Lavery M R, Acharya P, Sivo S A et al., 2019. Number of predictors and multicollinearity: What are their effects on error and bias in regression? Communications in Statistics: Simulation and Computation, 48(1): 27-38.

[49]
Leslie E, Kremer P, Toumbourou J W et al., 2010. Gender differences in personal, social and environmental influences on active travel to and from school for Australian adolescents. Journal of Science and Medicine in Sport, 13(6): 597-601.

DOI PMID

[50]
Li A Y, Zhao P X, Huang Y Z et al., 2020. An empirical analysis of dockless bike-sharing utilization and its explanatory factors: Case study from Shanghai, China. Journal of Transport Geography, 88: Article 102828.

[51]
Li J, Lo K, Guo M, 2018. Do socio-economic characteristics affect travel behavior? A comparative study of low-carbon and non-low-carbon shopping travel in Shenyang city, China. International Journal of Environmental Research and Public Health, 15(7): Article 1346.

[52]
Liao C, Dai T Q, 2022. Is “Attending Nearby School” near? An analysis of travel-to-school distances of primary students in Beijing using smart card data. Sustainability, 14(7): Article 4344.

[53]
Liang L, Guo Z, Li J et al., 2019. Analysis of the premium of second-hand housing prices in surrounding communities in middle school districts: Taking Shenzhen as an example. Acta Scientiarum Naturalium Universitatis Pekinensis, 55(3): 537-542. (in Chinese)

[54]
Lin C, 2018. The impact of school district on housing prices: Based on an evolutionary game model. In:Proceedings of the Eighth International Conference on Management, Education and Information, 980-985.

[55]
Liu W T, Ma C M, Liu I J et al., 2015. Effects of commuting mode on air pollution exposure and cardiovascular health among young adults in Taipei, Taiwan. International Journal of Hygiene and Environmental Health, 218(3): 319-323.

DOI

[56]
Lu B B, Brunsdon C, Charlton M et al., 2017. Geographically weighted regression with parameter-specific distance metrics. International Journal of Geographical Information Science, 31(5): 982-998.

DOI

[57]
Luo X B, Peng Y D, 2016. Scale effects of the relationships between urban heat islands and impact factors based on a geographically-weighted regression model. Remote Sensing, 8(9): Article 760.

[58]
Ma L, Xiong H, Wang Z et al., 2019. Impact of weather conditions on middle school students’ commute mode choices: Empirical findings from Beijing, China. Transportation Research Part D: Transport and Environment, 68: 39-51.

DOI

[59]
Ma X W, Ji Y J, Yuan Y F et al., 2020. A comparison in travel patterns and determinants of user demand between docked and dockless bike-sharing systems using multi-sourced data. Transportation Research Part A: Policy and Practice, 139: 148-173.

[60]
Mandic S, de la Barra S L, Bengoechea E G et al., 2015. Personal, social and environmental correlates of active transport to school among adolescents in Otago, New Zealand. Journal of Science and Medicine in Sport, 18(4): 432-437.

DOI PMID

[61]
Marshall J D, Wilson R D, Meyer K L et al., 2010. Vehicle emissions during children’s school commuting: Impacts of education policy. Environmental Science and Technology, 44(5): 1537-1543.

DOI PMID

[62]
Masoumi H E, Zanoli G, Papageorgiou A et al., 2017. Patterns of children’s travel to school, their body weight, spatial factors, and perceptions: A survey on nine European cities. Geoscape, 11(2): 52-75.

DOI

[63]
Matthews S A, Yang T C, 2012. Mapping the results of local statistics: Using geographically weighted regression. Demographic Research, 26: 151-166.

PMID

[64]
McMillan T E, 2005. Urban form and a child’s trip to school: The current literature and a framework for future research. Journal of Planning Literature. 19(4): 440-456.

DOI

[65]
Mei D W, Xiu C L, Feng X H et al., 2019. Study of the school-residence spatial relationship and the characteristics of travel-to-school distance in Shenyang. Sustainability, 11(16): 15

DOI

[66]
Miao Q, Bouchard M, Chen D M et al., 2015. Commuting behaviors and exposure to air pollution in Montreal, Canada. Science of the Total Environment, 508: 193-198.

DOI

[67]
Mickells G E, Figueroa J, West K W et al., 2021. Adherence to masking requirement during the COVID-19 pandemic by early elementary school children. Journal of School Health, 91(7): 555-561.

DOI

[68]
Molina J A, Gimenez-Nadal J I, Velilla J, 2020. Sustainable commuting: Results from a social approach and international evidence on carpooling. Sustainability, 12(22): Article 9587.

[69]
Molina-Garcia J, Queralt A, 2017. Neighborhood built environment and socioeconomic status in relation to active commuting to school in children. Journal of Physical Activity and Health, 14(10): 761-765.

DOI

[70]
Moran P A P, 1950. Notes on continuous stochastic phenomena. Biometrika, 37: 17-23.

PMID

[71]
Nakamura K, Hayashi Y, 2013. Strategies and instruments for low-carbon urban transport: An international review on trends and effects. Transport Policy, 29: 264-274.

DOI

[72]
Netease, 2022a. Haidian builds the urban slow traffic system, and the first bicycle dedicated road will be developed in the south. https://www.163.com/dy/article/HB7KKBQS0552UKSD.html.

[73]
Netease, 2022b. Ranking of top 100 cities in China’s GDP in 2021. https://www.163.com/dy/article/ H1QK7JUM0552NMG9.html.

[74]
Nguyen H V, 2020. Association of Canada’s provincial bans on electronic cigarette sales to minors with electronic cigarette use among youths. Jama Pediatrics, 174(1): Article e193912.

[75]
Nirupama N, Hafezi H, 2014. A short communication on school bus accidents: A review and analysis. Natural Hazards, 74(3): 2305-2310.

DOI

[76]
Ogura L M, 2010. Effects of urban growth controls on intercity commuting. Urban Studies, 47(10): 2173-2193.

PMID

[77]
Peng J D, Qi J J, Cui C W et al., 2021. Research on the impact of the built environment on the characteristics of metropolis rail transit school commuting: Take Wuhan as an example. International Journal of Environmental Research and Public Health, 18(18): Article 9885.

[78]
People’s Government of Beijing Municipality PGBM, 2022. Main data of the seventh national population census in Beijing. http://www.beijing.gov.cn/gongkai/shuju/sjjd/202105/t20210519_2392877.html.

[79]
Pocock T, Moore A, Keall M et al., 2019. Physical and spatial assessment of school neighbourhood built environments for active transport to school in adolescents from Dunedin New Zealand. Health and Place, 55: 1-8.

DOI

[80]
Rayens M K, McGeeney T, Wiggins A T et al., 2022. Smoke-free ordinances and youth tobacco use in Kentucky. American Journal of Health Promotion, 36(4): 673-677.

DOI

[81]
Raza A, Pulakka A, Hanson L L M et al., 2021. Commuting distance and behavior-related health: A longitudinal study. Preventive Medicine, 150: Article 106665.

[82]
Rodriguez-Lopez C, Villa-Gonzalez E, Perez-Lopez I J et al., 2013. Family factors influence active commuting to school in Spanish children. Nutricion Hospitalaria, 28(3): 756-763.

[83]
Rodriguez-Rodriguez F, Jara O P, Kuthe N M et al., 2019. Influence of distance, area, and cultural context in active commuting: Continental and insular children. Plos One, 14(3): Article e0213159.

[84]
Rong P J, Kwan M P, Qin Y C et al., 2022. A review of research on low-carbon school trips and their implications for human-environment relationship. Journal of Transport Geography, 99: Article 103306.

[85]
Schonbach D M I, Altenburg T M, Chinapaw M J M et al., 2019. Strategies and effects of promising school-based interventions to promote active school transportation by bicycle among children and adolescents: Protocol for a systematic review. Systematic Reviews, 8(1): 6.

DOI

[86]
Selby B, Kockelman K M, 2013. Spatial prediction of traffic levels in unmeasured locations: Applications of universal kriging and geographically weighted regression. Journal of Transport Geography, 29: 24-32.

DOI

[87]
Sener I N, Lee R J, Sidharthan R, 2019. An examination of children’s school travel: A focus on active travel and parental effects. Transportation Research Part A: Policy and Practice, 123: 24-34.

[88]
Shaaban K, Abdur-Rouf K, 2020. Assessing walking and cycling around schools. Sustainability, 12(24): Article 10607.

[89]
Shafahi A, Aliari S, Haghani A, 2018. Balanced scheduling of school bus trips using a perfect matching heuristic. Transportation Research Record, 2672(48): 1-11.

[90]
Sirard J R, McDonald K, Mustain P et al., 2015. Effect of a school choice policy change on active commuting to elementary school. American Journal of Health Promotion, 30(1): 28-35.

DOI PMID

[91]
Soltani A, Allan A, Nguyen H A et al., 2019. Students’ commuting pattern from the viewpoint of environmentalism: Comparing Australia with China. International Journal of Sustainability in Higher Education, 20(1): 91-114.

DOI

[92]
Song Y, Zhang L, Zhang M, 2022. Research on the impact of public climate policy cognition on low-carbon travel based on SOR theory: Evidence from China. Energy, 261: Article 125192.

[93]
Tamura R, Kobayashi K, Takano Y et al., 2019. Mixed integer quadratic optimization formulations for eliminating multicollinearity based on variance inflation factor. Journal of Global Optimization, 73(2): 431-446.

DOI

[94]
Tencent, 2022. What are the high-quality schools in Chaoyang District—Elementary School. https://new. qq.com/rain/a/20220107A05Q5L00.

[95]
Tian F, Qiu G Y, Yang Y H et al., 2012. Studies on the relationships between land surface temperature and environmental factors in an inland river catchment based on geographically weighted regression and MODIS data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 5(3): 687-698.

DOI

[96]
Tobler W, 2004. On the first law of geography: A reply. Annals of the Association of American Geographers, 94(2): 304-310.

DOI

[97]
Tudor-Locke C, Ainsworth B E, Popkin B M, 2001. Active commuting to school: An overlooked source of childrens’ physical activity? Sports Medicine, 31(50): 309-313.

DOI

[98]
Van Goeverden C D, de Boer E, 2013. School travel behaviour in the Netherlands and Flanders. Transport Policy, 26: 73-84.

DOI

[99]
Vorosmarty G, Dobos I, 2020. Green purchasing frameworks considering firm size: A multicollinearity analysis using variance inflation factor. Supply Chain Forum, 21(4): 290-301.

[100]
Wang W J, Zhu L Y, 2021. Research on the high-quality development of Beijing’s economy. Journal of Beijing Industrial and Commercial University Social Science Edition, 36(3): 102-111. (in Chinese)

[101]
Wati K, Tranter P J, 2015. Spatial and socio-demographic determinants of South East Queensland students' school cycling. Journal of Transport Geography, 47: 23-36.

DOI

[102]
Wen L M, Fry D, Rissel C et al., 2008. Factors associated with children being driven to school: Implications for walk to school programs. Health Education Research, 23(2): 325-334.

PMID

[103]
Whalen K E, Páez A, Carrasco J A, 2013. Mode choice of university students commuting to school and the role of active travel. Journal of Transport Geography, 31: 132-142.

DOI

[104]
Wikipedia, 2022. Beijing Haidian Experimental Primary School. https://zh.wikipedia.org/wiki/%E5%8C%97% E4%BA%AC%E5%B8%82%E6%B5%B7%E6%B7%80%E5%8C%BA%E5%AE%9E%E9%AA%8C%E5%B0%8F%E5%AD%A6.

[105]
Xiao Z P, Lin T Y, Liao J Y et al., 2021. School travel inequity between students from public and private schools in the city of Shenzhen, China. Journal of Advanced Transportation, 2021: Article 5032726.

[106]
Yacim J A, Boshoff D G B, 2019. A comparison of bandwidth and kernel function selection in geographically weighted regression for house valuation. International Journal of Technology, 10(1): 58-68.

DOI

[107]
Yang X D, Abbasi Q H, Alomainy A et al., 2011. Spatial correlation analysis of on-body radio channels considering statistical significance. IEEE Antennas and Wireless Propagation Letters, 10: 780-783.

DOI

[108]
Yin S, Hu X, Ma Z et al., 2019. Spatial and temporal effects of housing price in urban school districts based on education equity: A case study of public primary school in the main urban area of Nanjing. Economic Geography, 39(9): 82-93. (in Chinese)

[109]
YNET, 2022. Dongcheng district of Beijing promotes the reconstruction of the old city: Old hutongs usher in a new life. https://t.ynet.cn/baijia/32753188.html.

[110]
Zhang H, Guo Y P, Su L X et al., 2021. Analysis on the coordinated development of human settlements and economy in the urban-rural fringe: Taking Chaoyang district of Beijing as an example. Human Geography, 36(6): 108-116. DOI: 10.13959/j.issn.1003-2398.2021.06.012.2398.2021.06.012. (in Chinese)

[111]
Zhang H, Wan J, 2010. Research on the development mode of slow traffic system in cities. Proceedings of the Second International Conference of Urbanization and Land Resource Utilization, 452-457.

[112]
Zhang W, Zhang Y J, 2014. On the current development of technology finance industry in Haidian district. Proceedings of the International Conference on Social, Education and Management Engineering: 391-394.

[113]
Zhang Y H, 2018. Research on spatial differentiation and influencing factors of second-hand housing prices in Beijing based on GWR model[D]. Beijing: China University of Geosciences Beijing. (in Chinese)

[114]
Zhang Z, Fujii H, Managi S, 2014. How does commuting behavior change due to incentives? An empirical study of the Beijing Subway System. Transportation Research Part F: Traffic Psychology and Behaviour, 24: 17-26.

DOI

[115]
Zhao P, Xu Y, Liu X et al., 2020. Space-time dynamics of cab drivers’ stay behaviors and their relationships with built environment characteristics. Cities, 101: 102689.

DOI

[116]
Zhao P J, Zhang Y X, 2018. Travel behaviour and life course: Examining changes in car use after residential relocation in Beijing. Journal of Transport Geography, 73: 41-53.

DOI

Options
Outlines

/

〈 〉