Is the decline of active travel to school unavoidable by-products of economic growth and urbanization in developing countries?

Abstract

Background:

Economic growth and urbanization may contribute to the decline of active travel to school (ATS). We aim to explain the change of ATS in China between 1997 and 2011 and to predict the prevalence of ATS in China within the next 30 years using various scenario.

Methods:

We developed a system dynamics model to study ATS and the model assumes the prevalence of ATS is determined by the dynamic interaction of four exogenous and eight endogenous variables.

Results:

The simulated prevalence of ATS is roughly consistent with empirical data. Economic development and urban sprawl are more influential than urban design and crime in terms of ATS. Under a relatively reasonable scenario, the prevalence of ATS is projected to decrease from 73% in 2011 to 65% in 2014, and the prevalence of childhood overweight & obesity is projected to increase from 24% in 2011 to 34% in 2041. With the maintaining of economic development grow, to control urban sprawl is the most effective measure to promote ATS and decrease childhood obesity.

Conclusions:

Overall, the model enabled us to conduct experiments to test the possible effects of changing one or more factors taking into account their dynamic interrelationship, and our study may provide implications for policy intervention.

Introduction

Active travel to school (ATS), defining as walking or bicycling to and from school, is associated with multiple benefits among children including higher overall physical activity [1–8], less likely to be obese and to develop chronic disease risk factors [9–12]. Also, ATS reduces children’s dependence on parents, improves mental health and social interactions, and promotes healthy lifestyles which may be maintained into adulthood [13, 14]. The significance of ATS is not limited to children’s health. ATS is an important part of sustainable transportation, with environmental benefits such as reductions in traffic congestion and air pollution, social benefits such as increase of social cohesion in the neighborhood, improvements in safety from traffic and crime, and economic benefits such as the saving of time and cost for parents.

However, a global decline of ATS has been observed, although the magnitude of decline and the underlying drivers may vary by the specific country. In the U.S., the prevalence of ATS among children aged 5–14 decreased from 47.7% in 1969 to 12.7% in 2009, and half of the ATS decline was due to the increase of distance to school [15]. Similarly, In Sydney, Australia, the prevalence of ATS among children aged 5–14 decreased from 44.2–57.7% in 1971 to 21.1–25.5% during 1999–2003 [16]. In Switzerland, ATS prevalence among children aged 6–14 decreased slightly from 78.4% in 1994 to 71.4% in 2005, may be due to the decreased bikes availability and increased number of cars per household [17]. In China, ATS among children aged 6–18 has decreased from 96.4% in 1997 to 85.9% in 2006, and the decline may be contributed by the increased distance to school and the rapid increase in private vehicle ownerships [18]. A declined pattern of ATS was also found in Brazil between 1997 and 2007 [19]. In most of the above countries, the ATS decline has been concurrent with economic growth that increases the prevalence of private vehicles among the population and urban growth that unfortunately, often results in urban sprawl, that is, stretched form of the city with low density at the periphery [20]. Generally, the literature supports the notion that the economic growth and urbanization process may contribute to the decline of physical activity because of the change of life style and living conditions (e.g. less physical activity required from occupation, passive travel, and more sedentary activities) [21–23]. A more specific question: is the decline of ATS an unavoidable byproduct of economic growth and urbanization, particularly in developing countries?

Evidence shows that ATS is associated with multiple factors, especially distance from home to school, ownership of private vehicles, traffic, crime, family time budget, social norm towards ATS, and parents’ attitude toward ATS [24–26]. It is plausible that the decline of ATS may be driven by a complex interaction of above factors including the changes of built environment, transportation, policy, social norm and perception, safety concern, and change at the household level, such as car ownership rates and the number of parents that are working [16]. For example, dynamic interactions exist between travel behavior, other behaviors, and the built and social environment [27]. Urban sprawl, as a global phenomenon, has expanded cities into rural areas and the resulted low-density and segregated land use make ATS very challenging. Although a number of factors were found to be associated with ATS, some factors may influence ATS directly, while some factors may influence ATS through the proxies of the other factors. For example, economic development may indirectly influence ATS through its direct influence on private vehicle ownership, family time budget, social norm towards ATS, childhood obesity, and etc. Similarly, urban sprawl may indirectly influence ATS through its direct influence on the distance from home to school, private vehicle ownership, traffic, and etc.

Over the past several decades, China has undergone a significant economic development and urbanization. During the same period, the prevalence of overweight and obesity among children and adolescents in China has increased rapidly [28, 29] and the ATS prevalence has been dropped sharply [18]. Understanding the basic mechanisms of ATS change, as an example, may contribute to our knowledge on how economic growth and urbanization may influence our society from both positive and negative sides. The aims of this study were to develop a system dynamics model to study the change of ATS in China and to apply the model to the following two questions. First, to explain the change of ATS in China between 1997 and 2011. Second, to predict the prevalence of ATS in China within the next 30 years using various scenario simulation. More specifically, (1) to what extent will the factors of urban sprawl, urban design (refers to how the urban design supports active travel, consisting of features such as sidewalk, connectivity, proximity, parks and public spaces, pedestrian amenities, traffic safety, and etc), economic development, and crime influence ATS, both separately and jointly. (2) to predict ATS under the combination of a various trend for major factors, and (3) to explore the best control measure of improving ATS under the condition that the economic development will continue.

The use of systems thinking and complex systems have been promoted in the study of sustainable city [30]. Over the past several decades, system dynamics (SD) model, as one method of complex systems, has been extensively applied to obtain a better understanding of the system structure and associated behaviors by capturing delays (material and information delays) and feedbacks embedded in the structure. SD models have been applied in public health [31–33], urban planning [34], and transportation [35, 36], but never been used to in the study of ATS. The aim of this research is to gain in-depth understanding of how multilevel factors simultaneously affect the health behaviors of children and consequently provide constructive inputs for building high leverage policy or interventions to curb the negative externalities of economic development to the public health.

Methods

Model description

We developed a system dynamics model using Vensim Professional, Version 6.3D (Ventana Systems, Inc, 2015). As Figure 1 shows, we assume the prevalence of ATS is determined by the dynamic interactions of 12 variables. In this model, all variables are within a range between 0 and 1, with 0 indicating the lowest level and 1 indicating the highest level in general. The configuration of 0–1 value is an effort to normalize the associated factors given the difference in units, because it provides a base for the across-factor comparison with regard to the relative magnitude of the effects on ATS, which is critical to examine the dynamic complex relationships in the system.

Figure 1.

Key variables and the associated dynamic processes in our System Dynamics model. The outcome is the prevalence of ATS (at top-right corner). The four variables on the left (highlighted with underline) are exogenous variables and all other variables are endogenous variables.

Table 1 shows the parameters configuration for all 12 variables. Considering the current knowledge, the configuration may be an empirically feasible choice to make the model operational and serve our study aims. Eight variables are endogenous, that is, their initial values in 1997 were set according to the best available empirical studies or estimations. For example, according to the China Health and Nutrition Survey [37], ATS, childhood overweight & obesity, and private vehicle in 1997 were set to be 0.96, 0.24, and 0.17 because in 1997 the prevalence of ATS was 96% [38], the prevalence of childhood overweight & obesity was 24% [28], and the percent of people owning private vehicle was 17% [38]. Other endogenous variables that were challenging to be obtained were estimated (e.g., 0.10 for distance to school) or set to be extreme status (e.g., 0.001 for traffic, and 1 for parents’ attitude towards ATS). For the eight endogenous variables, their values over the following years are determined by the dynamic interaction of exogenous variables or other endogenous variables. The simulated patterns of these variables (especial major outcomes such as ATS and childhood overweight & obesity) over 1997 and 2011 will be compared with the empirical data to validate our model.

Table 1.

Parameters configuration for the system dynamics model

Endogenous parameters
Parameters	Initial values in 1997	Simulated values in 2011	1 Update mechanisms
ATS	0.96	0.73	Distance (N, medium), Vehicle (N, medium), Traffic (N, medium), Urban design (P, medium) Crime (N, short), Time (P, short), Attitude (P, short)
Childhood overweight & obesity	0.15	0.24	Econ (P, long), ATS (N, short), Urban design (N, medium)
Distance to school	0.10	0.20	Urban sprawl (N, medium)
Private vehicle	0.17	0.29	Econ (P, short), Distance (P, short), Urban sprawl (P, short)
Traffic	0.01	0.22	Vehicle (P, short), Urban sprawl (P, long)
Family time budget	1	0.74	Econ (N, short)
Parents’ attitude towards ATS	1	0.53	Childhood overweight & obesity (P, short), Norm (P, short), ATS (P, short)
Social norm towards ATS	1	0.63	Econ (P, medium), ATS (P, short)
Exogenous parameters
Parameters	Initial values in 1997	Value in 2011
Economic development	0.001	0.4
Urban sprawl	0.001	0.4
Urban design	1	0.7
Crime	0.001	0.2

¹Update mechanisms: relationship: N for negative, and P for positive; and effecting durations: short term (5 years), median term (10 years), and long term (20 years).

Four variables are exogenous, that is, their values are feed in from outside and not affected and determined in the system. For all four variables, both values in 1997 and 2011 were assigned, then we assume linear change patterns from 1997 to 2011. Evidence confirmed that overall, the period between 1997 and 2011 observed a growth of economic development [39], an increase of urban sprawl [40, 41], a decline of urban design in terms of the support of active travel [42–44], and an increase of crime rate [45]. We run a sensitivity analysis to explore how ATS is influenced by ranges of each exogenous variables.

In the model, each step is one year. If factor B has a positive influence on factor A, then at t year, the value of A_t is updated using the Equation 1.

$$A_t=A_{t-1}+\frac{B_{t-1}-A_{t-1}}{N_{BA}}$$ E.1

If factor C has a negative influence on A, then at t year, the value of A_t is updated using the Equa. 2.

$$A_t=A_{t-1}+\frac{\left(1-{\text{C}}_{t-1}\right)-A_{t-1}}{N_{CA}}$$ E.2

If both factors B and C have influence on A, the value of A_t is updated using the Equa. 3.

$$A_t=A_{t-1}+\frac{B_{t-1}-A_{t-1}}{N_{BA}}+\frac{\left(1-C_{t-1}\right)-A_{t-1}}{N_{CA}}$$ E.3

For all equations, N_BA is a parameter to decide how fast factor B could influence factor A. The smaller value of N_BA, the quicker for factor A to be similar to factor B (for positive relation) or be similar to 1-B (if negative relation). As Table 1 shows, we assume three optional values for N, as short (5 years), medium (10 years), and long terms (20 years).

Here, we illustrate the update mechanism using the variable of childhood overweight & obesity (O).

$$O_t=O_{t-1}+\frac{Eco{n}_{t-1}-O_{t-1}}{20}+\frac{\left(1-{\text{ATS}}_{t-1}\right)-O_{t-1}}{10}+\frac{\left(1-{\text{UrbanDesign}}_{t-1}\right)-O_{t-1}}{10}$$ E.4

Model assessment

The model will be assessed by the consistence of ATS prevalence between our simulated results and empirical data during 1997–2011. To further validate the model, for the endogenous variables, we will compare their simulated change patterns with the reality in China during 1997–2011.

Scenario analysis: to predict ATS from 2011 to 2041

We run three series of scenario analysis to predict the change of ATS over the next 30 years (2011–2041), depending on the combination of various hypothetical changes of four exogenous variables including economic development, urban sprawl, urban design, and crime. First, we explore how will the ATS be sensitive to the change of each of the four exogenous variables with the assumption that the other three variables will be stable over 2011–2041. Second, we run four typical scenarios including (1) Baseline (the four exogenous variables will keep going with the current trend); (2) Stable (the four exogenous variables will keep constant); (3) Best for ATS (the four exogenous variables will reverse the current trend); and (4) Reasonable for ATS (the economic developing will keep growing, and the other three variables will reverse the current trend). Third, with the assumption that economic developing will keep growing, we explore the combination of other three variables. For the second and the third series, the setting of parameters is shown in Table 2.

Table 2.

Parameters setting for the period between 2011 and 2041 and the simulated ATS in 2041

	Change of exogenous parameters from 2011 to 2041				Outcomes in 2041
Scenarios	Economic Development	Urban Sprawl	Urban Design	Crime	Prevalence of ATS, with 95% range	Prevalence of childhood overweight & obesity, with 95% range
Baseline	Increase to [0.7–0.9]	Increase to [0.7–0.9]	Decrease to [0.2–0.4]	Increase to [0.5–0.7]	37 [33,40]	64 [60,68]
Stable	Stable to [0.3–0.5]	Stable to [0.3–0.5]	Stable to [0.6–0.8]	Stable to [0.1–0.3]	63 [60, 67]	35 [31,39]
Best	Decrease to [0.1–0.3]	Decrease to [0–0.2]	Increase to [0.8–1.0]	Decrease to [0–0.2]	77 [73,80]	21 [17, 25]
Reasonable	Increase to [0.7–0.9]	Decrease to [0–0.2]	Increase to [0.8–1.0]	Decrease to [0–0.2]	65 [61,68]	34 [31,38]
Control Urban Sprawl	Increase to [0.7–0.9]	Decrease to [0–0.2]	Decrease to [0.2–0.4]	Increase to [0.5–0.7]	49 [46,52]	58 [54,62]
Control urban design		Increase to [0.7–0.9]	Increase to [0.8–1.0]	Increase to [0.5–0.7]	42 [39,45]	46 [42,50]
Control crime		Increase to [0.7–0.9]	Decrease to [0.2–0.4]	Decrease to [0–0.2]	47 [44,50]	59 [55,63]
Control urban sprawl and design		Decrease to [0–0.2]	Increase to [0.8–1.0]	Increase to [0.5–0.7]	54 [51,58]	40 [36, 44]
Control urban sprawl and crime		Decrease to [0–0.2]	Decrease to [0.2–0.4]	Decrease to [0–0.2]	59 [56,63]	53 [49,57]
Control urban design and crime		Increase to [0.7–0.9]	Increase to [0.8–1.0]	Decrease to [0–0.2]	52 [49,56]	41 [37,44]

Results

For model assessment

Using the parameters configuration in Table 1, the simulated change of ATS is shown in Figure 2. First, the simulated pattern of ATS is roughly consistent with empirical data in China. From 1997 and 2011, CHNS data has six waves and the ATS prevalence was kept dropping from 96% in 1997 to 73% in 2011 [38]. Correspondingly, our simulated ATS percent was 73% in 2011. As Table 2 shows, the prevalence of childhood overweight & obesity was set to be 15% in 1997 and was simulated to be 24% in 2011. This is consistent with the empirical data in the same period [28]. The simulated change patterns for most endogenous variables were consistent with the reality in China as well, for example, the rapid increase of childhood obesity [46] and private vehicle ownership [47].

Figure 2.

The change of the ATS between 1997 and 2011, with economic development ranged in [0.3–0.5], urban sprawl ranged in [0.3–0.5], urban design ranged in [0.6–0.8], and crime ranged in [0.1–0.3].

Scenario analysis 1

For the four exogenous variables, we found a linear relationship between the variable and the prevalence of ATS and childhood overweight & obesity. On average, each 0.1 increase in economic development results in 2.0 percent decrease of ATS and 2.3 percent increase in childhood overweight & obesity. Each 0.1 increase of urban sprawl results in 1.8 percent decrease of ATS and 0.9 percent increase in childhood overweight & obesity. Each 0.1 increase of crime results in 2.1 percent decrease of ATS and 1.1 percent increase in childhood overweight & obesity. Each 0.1 increase of urban design results in 0.9 percent increase of ATS and 3.0 percent decrease in childhood overweight & obesity.

Scenario analysis 2

Figure 3 shows the predicted change of ATS and childhood overweight & obesity between 2011 and 2041 of four common scenarios (the exact values of ATS in 2041 are shown in Table 2). Beginning from 73 percent of ATS in 2011, if the four exogenous variables keep the current trend (i.e., Baseline scenario, increase for economic development, urban sprawl, and crime, and decrease for urban design), the ATS will keep dropping to 37 percent and the childhood overweight & obesity will increase to 64 percent in 2041. Even if all the four variables keep stable with their values in 2011 (i.e., Stable scenario), the ATS will see a mild drop to 63 percent and the childhood overweight & obesity will increase to 25 percent in 2041. In the best for ATS scenario, both the ATS and childhood overweight & obesity will keep nearly constant. Although this scenario may be unrealistic because of the assumption of the economic development dropdown, it may provide a theoretical maximum threshold. In a scenario which may be more reasonable (with an increase of economic development and urban design, and a decrease of urban sprawl and crime), the ATS will decrease to 65 percent and childhood overweight & obesity will increase to 34 percent in 2041.

Figure 3.

Predicted change of ATS and childhood overweight & obesity between 2011 and 2041 of four common scenarios

Scenario analysis 3

As shown in Table 2, with the maintain of economic development grow, if only one variable could be controlled, then to control urban sprawl is more effective than the other two regarding the improvement of ATS. If two variables could be controlled, combination urban sprawl with a crime will be the most effective to the increase of ATS. For childhood overweight & obesity, to control crime is the most effective than the other two measures.

Discussions

Multilevel factors associated with ATS have been examined by a growing body of literature using traditional regression-based methods, e.g. on age, gender, ethnicity at the individual level, family SES, parents’ perception, built environment, perceived safety, aesthetics of the environment etc. However, the interrelationships among these factors, complex nonlinear effects and possible mechanisms that may give rise to the declining ATS worldwide have not been well studied and understood. Our model was designed to capture the key features in a dynamic system to reveal the leveraging points and mechanisms in the system. We cautiously avoided adding unnecessary complexity into the model while also ensuring the model’s capacity to reflect and mimic the real world phenomena relevant to the study. We evaluated the model performance by comparing the model-generated ATS prevalence and the observed empirical data. The simulated results and empirical data exhibited high agreement, suggesting valid model design and good model performance.

As our simulated results indicated, economic development and urban sprawl are more influential than urban design and crime in terms of ATS. This is consistent with empirical studies of ATS in both China [18, 48] and other countries [49, 50] that the long distance to school is a primary barrier that prevents children from ATS, as well as our hypothesis at the beginning that the decline of ATS, may be a by-product of urbanization. If the trend of economic growth keeps stable in China, it will be extremely challenging to reverse the decline pattern of ATS. Although there is no data on ATS in China at the national level available to compare, the growing trend and the latest data of children overweight & obesity is consistent with our estimation [51].

Urban sprawl has been criticized for its adverse effects on health including more reliance on private vehicles and a decrease of active travel [20, 52]. Although urban sprawl is a common feature of the built environment in the United States and other industrialized nations, China may still has the opportunity to control sprawling expansion, for example, through smart growth that is characterized by high population density, walkable and bikeable neighborhoods, preserved green spaces, mixed-use development, available mass transit, and limited road construction [52–54]. A recent study has confirmed the continuing trend of urban sprawl among most Chinese cities between 1995 and 2015. However, at the same time, it found that cities that have controlled urban sprawl more effectively have larger GDPs [55].

Our study has contributed to the study of ATS and sustainable transportation. As the first study using simulation-based systems models to examine ATS, this study revealed many important aspects of ATS in a fast changing context like China, which also has substantial public health implications for other developing and developed countries. Our simulated results may deepen our understanding of the impact of each factor on ATS and how factors work together to determine ATS. Our model illustrates the utilities of systems dynamics model, mainly, the emergency of dynamic complexities by the interactions of a number of factors. This model may help researchers to test their hypotheses and let decision makers explore and examine policies or intervention program to promote ATS. Specifically, our model enabled us to conduct experiments to test the possible effects of changing one or more factors taking into account their dynamic interrelationship. Our experiments suggested that controlling the urban sprawl would be more effective than controlling crime rate and urban design to promote ATS when economic development remained steady. Such experiments and finding are especially helpful and informative in policy design and decision-making. Resources are always scarce. Budget limit is a common constraint for any policy and intervention strategy. Given a limited budget, how to maximize policy effect is of great importance.

The study has limitations. Primarily, due to the lack of existing studies examining the complex mechanisms influencing ATS in children, we could not use empirically estimated parameters and related functional relationships to develop our model. One example is that we separate variables associated with ATS as exogenous and endogenous variables. This categorization may be valid for simplification benefit, however, it is quite arbitrary. Another example is that our model parameters by their nature are ordinal. They reflect the effect order on a scale 0 to 1, instead of empirically quantifying the magnitude. Additionally, the durations of short, median and long terms for the update mechanism are reasonably estimated by the authors rather than based on empirical evidence. Thus, the above aspects may limit our capability to meaningfully model and quantify intervention effects. This study is more leaned to exploratory than confirmatory. Our results and findings provide the possible explanations and solutions to the declining ATS, while not excluding the possibilities of other possible mechanisms. To improve the model in these aspects will require the development of empirical research in related fields. When adequate empirical evidence is accumulated, we may be able to express these causal relationships in a more precise manner. It should be noted that our model and the result were specific to China and may be applicable to other countries featured by rapid economic growth and urbanization at most. However, the causal mechanisms underlying the decline of ATS may differ in other countries. For example, in high population density regions such as the Netherlands, Sweden, Switzerland, and the U.K., features such as urban form and traffic safety may be more influential to children’s ATS than urban sprawl [17, 56–58]. The model will have to be tailored to any specific country which it was targeted to. For example, a model for the ATS in the Netherland may consider the Netherlands’ cycling culture and deal with walking and cycling separately.

Due to accelerated development in economy and overemphasizing on pursuit of quantity over quality, economic growth in China in the past three decades created so many negative externalities to the society including but not limited to pollution (haze and smog), imbalanced resources in education, fast urban sprawl, more traffic accidents, distrust among people, and insufficient law enforcement powers. They are all major factors contributing to the decrease of ATS. Having recognized these issues, fighting on the pollution, enhancement in the law enforcement forces and urban design such as building more trails will encourage an increase in ATS. Furthermore, with designing more balanced allocating mechanism of education resources, students could attend schools that are within walking distance. At the same time, encouraging “walk pool” of students living in surrounding communities will assure parents the safety concern. Promoting education programs on the importance of ATS to preventing childhood obesity to city governors, school administration, parents, and children would also be conducive to increase ATS. The decline of active travel was a general pattern in China including other groups such as adults [59]. Specific policies was advocated to reverse the trend including slowdown massive roadway investment, expanding and improving public transport, cycling, and walking facilities, and restrict motor vehicle use in congested areas [60]. A sustainable transportation planning should embrace the contributions of smart growth, sustainable growth, and inclusive growth [61].