The Impacts of the Multiscale Built Environment on Commuting Mode Choice: Spatial Heterogeneity, Moderating Effects, and Implications for Demand Estimation

Jian, Wenliang; Liu, Xiaoxiao; Liu, Hao; Hu, Yongshi; Gao, Liangpeng

doi:https://doi.org/10.1155/2023/9346631

Journal of Advanced Transportation

On this page

Abstract Introduction Literature Review Data Results and Analysis Conclusion Data Availability Conflicts of Interest Acknowledgments References Copyright Related Articles

Research Article | Open Access

Volume 2023 | Article ID 9346631 | https://doi.org/10.1155/2023/9346631

The Impacts of the Multiscale Built Environment on Commuting Mode Choice: Spatial Heterogeneity, Moderating Effects, and Implications for Demand Estimation

Wenliang Jian,^1,2Xiaoxiao Liu,¹Hao Liu,²Yongshi Hu,¹and Liangpeng Gao^1,3

Academic Editor: Luigi Dell’Olio

Received13 Jan 2023

Revised29 Mar 2023

Accepted19 Apr 2023

Published27 Apr 2023

Abstract

Promoting the usage of sustainable commuting modes requires in-depth understanding about residents’ commuting mode choice behavior. This study presents an empirical study to investigate the relationship between the built environment and commuting mode choice using CLDS 2016 cross-city questionnaire data. Several multilevel multinomial logit models including the null model, base model, and moderating effect model are developed to analyze the effects of built environments at both city and neighborhood levels on commuting mode choice. Estimation results of the null model reveal the significant spatial heterogeneities in commuting mode choice across different cities and different neighborhoods within a specific city. We then explore the potential built environment variables yielding the spatial heterogeneity via the base model. Results show that the built environment at the city level (including the urbanization rate, number of public transportation vehicles, metro operating mileage, GDP, city population density, and road area per capita) and neighborhood level (including neighborhood population density, air quality, neighborhood location, and land use diversity) could partially explain the spatial heterogeneities in commuting mode choice. In addition, the moderating effects of these built environments on the link between commuting time and commuting mode choice are examined. Results imply that the urbanization rate and neighborhood population density moderate the effect of commuting time on choosing nonmotorized modes, while neighborhood location moderates the effect of commuting time on choosing public transit. Also, the mode shares of nonmotorized mode and public transit under different levels of commuting time are estimated in different built environment contexts. The findings of this study are expected to provide serviceable support for urban planning and transportation policy making.

1. Introduction

Traffic congestion and environmental pollution are becoming increasingly severe in many Chinese cities, while the car remains to be an attractive commuting mode. To alleviate traffic congestion and protect the environment, the use of more sustainable commuting modes (e.g., nonmotorized mode and public transit) should be promoted. The built environment has been proven to play a significant role in residents’ commuting behavior. Thus, one possible approach to shift car users to sustainable commuting modes is to develop reasonable urban planning and create a suitably built environment for walking, cycling, and public transit [1–3]. To support this, it is necessary to explore how the built environment affects residents’ commuting mode choices.

The effect of the built environment on residents’ travel behavior (e.g., travel time choice, commuting mode choice, and car ownership) has attracted attention in previous works of the literature because of its derivative impacts on travel demand forecasting (or policy assessment). Many empirical studies have been conducted to determine the link between the built environment and travel behaviors. Ding et al. [4] and Tao et al. [5] pointed out that built environments are even more important than both socioeconomic characteristics and travel mode attributes in travel decision-making. Built environments at both city (or region) and neighborhood levels were found to have significant impacts on travel behavior in these studies. Nevertheless, the most existing studies dedicated to commuting mode choice focused on the impact of the built environment at the neighborhood level. Several studies investigated the effect of the city-level built environment on commuting mode choice [6–8]. However, it was rare that the effects of the built environment at both city and neighborhood levels on commuting mode choice were explored within the same framework. Besides, the relationship between individual attributes of residents (e.g., commuting time) and commuting mode choice is usually moderated by the built environment. As a result, residents with the same individual characteristics present preference heterogeneity for commuting mode choices in different built environment contexts [9]. Yet, previous studies paid more attention to the direct effect of the built environment on commuting mode choice, and the moderating effect between the built environment and individual attributes was rarely investigated. This limits the in-depth understanding about residents’ commuting mode choice behavior.

Therefore, this study aims to comprehensively investigate the effects of the built environment on residents’ commuting mode choice. In particular, we intend to reveal the link between the built environment and the usage of sustainable commuting modes (including nonmotorized mode and public transit) to provide support for shifting car users to a more sustainable mode. By building a multilevel multinomial logit (MML, including city level, neighborhood level, and individual level) model, built environment elements at the city level and neighborhood level are both considered in our research. Then, we add the interaction terms between built environment variables and critical variables (i.e., commuting time) of individual attributes into the model to explore how built environment moderates the impact of commuting time on commuting mode choice. Also, the mode shares of the nonmotorized mode (and public transit) under different levels of commuting time are estimated in different built environment contexts. The findings of this study are expected to provide serviceable support for urban planning and transportation policy making.

The remainder of this study is organized as follows: the next section briefly reviews the previous related research. Section 3 describes data and methodology; Section 4 presents the results of the model and analysis; and Section 5 summarizes the main findings of this study.

2. Literature Review

As one of the most important parts in travel demand models, travel mode choice modeling has received extensive attention over the past two decades. Many theoretical and empirical studies have been conducted to investigate mode choice behavior [10–12]. The discrete choice model has been wildly used to analyze travel mode choice behavior due to its both theoretical and empirical advantages over other models [13]. One of the critical issues in travel mode investigation was identifying the attributes that a travel decision-maker considered during the decision-making process. Research studies in earlier periods mainly focused on the travel mode attributes (including travel time, waiting time, travel time reliability, fare, in-vehicle crowding, and transfers) and demographic attributes (e.g., age, education, and income) [14–22]. In recent years, a number of studies have underlined the importance of spatial attributes (i.e., the spatial context in which individuals make travel mode decisions) as a determinant in travel mode choice [2–8, 23–36]. Nasri and Zhang [8] pointed out that improving the built environment and transit accessibility in regional areas would help encourage the use of public transit and decrease the reliance on driving around rail stations. Eldeeb et al. [26] found that improving the city built environment would increase the probability of choosing walking and biking while decreasing the probability of choosing other modes. Neves et al. [36] also found the positive effect of population density and diversity on walking mode choice for trips to school and work. Furthermore, spatial heterogeneity of travel mode choice behavior has been found in plenty of empirical studies [37]. Hong and Goodchild [38] indicated that ignoring the spatial heterogeneity in mode choice investigation could result in erroneous conclusions. To accommodate the spatial heterogeneity of travel mode choice behavior, the multilevel modeling framework has been employed to analyze the effect of the built environment on travel behavior [39–42]. However, the most previous literature built the LPLA (linear in the parameters and linear in the attributes) model specification to investigate travel mode choice. The LPLA specification assumed that built environments influenced preferences among travel modes but did not influence preferences for the travel mode attributes (e.g., travel time). This limited the research about the moderating effect of the built environment on the relationship between travel mode attributes and mode choice.

The measures of the built environment can be divided into two categories: (1) built environment measures at a neighborhood level and (2) built environment measures at a city (or region) level. Built environment measures at the neighborhood level have been wildly investigated in travel behavior research studies [2, 3, 22–33, 43, 44]. The scale of neighborhood is usually defined as traffic analysis zones, communities, or blocks. The “5Ds” principle, which includes density, diversity, design, destination accessibility, and distance to transit, is the most used method to measure the neighborhood-level built environment [45]. Density measure is always defined as the variable of interest per unit of area. The variable of interest can be population, dwelling units, employment, and building floor area. Diversity is usually defined as measures that relate to the number of different land uses in a given area. Many previous studies concluded that the increasing land use diversity could reduce the use of car significantly because more destinations are available within a short distance of an individual’s home [26, 31, 33, 46]. Design measures in the existing research studies include sidewalk coverage, average street widths, number of pedestrian crossings, and green coverage rate. Destination accessibility measures ease of access to trip destination, while distance to transit measures ease of access to public transit facilities (e.g., metro station and bus stop) [2, 23].

Due to the fact that most existing research studies analyzing the relationship between the built environment and travel behavior were conducted based on single-city samples, the city-level built environment is comparatively less frequently considered [47–50]. Nevertheless, several research studies considering the effect of the city-level built environment on commuting behavior are still found [6–8, 51–53]. In these studies, measures of the city-level built environment include city size, density, transport infrastructure, economic-related factors, and spatial structure. City size (usually measured by population or total built-up area) and density (usually measured by population density) are the most widely applied measures of the city-level built environment in empirical studies. For example, Guerra et al. [52] found that the city-level built environment, including population density, spatial compactness, roadway density, and kilometers of mass rapid transit per capita, impacted the probability of car use; and elasticity analysis results showed that the probability of car use is most sensitive to population density. However, most previous studies analyzed the importance of the city-level built environment for commuting behavior in Western countries’ cities, and the empirical evidence for developing countries (including China) is scarce [25]. Engelfriet and Koomen [53] indicated that the urban development processes of China were different from those of Western countries in at least three respects, including the higher urban density levels, the faster pace of urban expansion, and the stronger urban planning. Therefore, the relationships between the city-level built environment and commuting behavior found in Western countries cannot hold for China. A complete assessment on the importance of the built environment for commuting mode choice in Chinese cities should be conducted to support transportation policy making.

Table 1 summarizes the built environment variables used in the recent literature. In summary, we identify some research gaps and develop our contribution accordingly, as follows. First, using the nationwide labor-force survey data of China, we comprehensively analyze the effects of both city level and neighborhood-level built environment on commuting mode choice within a framework. Second, by adding the interaction terms between built environment variables and commuting time, a nonlinear specification is built to explore how the built environment moderates the impact of commuting time on commuting mode choice. Moreover, we conduct a mode share simulation to investigate the effects of spatial heterogeneity on demand forecasting. This is also new in the literature.

3. Data and Methodology

3.1. Data Source

The data used in this study are primarily derived from the China Labor-Force Dynamics Survey (CLDS), which was conducted by the Centre for Social Science Survey of Sun Yat-sen University in 2016. Residents in 29 provinces (excluding Hainan and Tibet) of mainland China participated in the survey. The CLDS 2016 collected the individual information, household information, work information, commuting trip information, and the information about the neighborhood built environment of residence via a three-level questionnaire survey (including individual level, household level, and neighborhood level). The variables of the neighborhood-level built environment, individual attributes, and commuting mode attributes in this study are directly drawn from the survey dataset. It needs to be mentioned that the neighborhood in this study is defined as the neighborhood committee (juweihui), which is the smallest administrative unit in Chinese cities. Besides, we obtain the data of the city-level built environment from China Urban Construction Statistical Yearbook 2016. Since we focus on the commuting mode choice, respondents who are unemployed or work at home are excluded. Finally, 4964 individual samples from 176 neighborhoods in 79 cities are used for this research. The spatial distribution of the valid samples is shown in Figure 1.

3.2. Variables’ Description

A multilevel multinomial logit model is built to analyze the impact of the built environment on commuting mode choice. The dependent variable of the model is residents’ commuting mode choice. Since we intend to reveal the link between the built environment and usage of sustainable commuting modes, residents’ commuting mode choice is divided into three categories: nonmotorized mode (including walking and cycling), public transit mode (including bus and metro), and car mode (including private car and taxi). The car mode is defined as the base kind of the commuting mode. In our data set, samples that choose the nonmotorized mode or public transit mode account for 61.7%. This is in line with the reported value from Traffic Analysis Report of Major Cities in China 2021 (60%–65%).

The independent variables are measured at the city level, neighborhood level, and individual level, respectively. At the city level, six variables including the gross domestic product (GDP), urbanization rate, city population density, number of public transportation vehicles, metro operating mileage, and road area per capita are selected to denote the city characteristics. The GDP and urbanization rate reflect the level of the city’s economic development. According to the sample statistics, we find that the proportion of commuting by car and public transit in developed cities is significantly higher than that in developing cities. This was also proved by Yin and Sun [51]. At the neighborhood level, the neighborhood built environment used for this analysis includes neighborhood population density, diversity, green coverage, air quality, distance to district centre, and neighborhood location. Diversity which is measured by the sum of six categories of facilities in the neighborhood (including bank, hospital, sports venue, library, senior centre, and square) reflects the spatial balance among diversified destinations. The neighborhood location is divided into two categories: neighborhood in the urban area and neighborhood in the suburban area. This variable is selected because commuting characteristics (e.g., commuting time and commuting distance) are significantly different between the urban area and the suburban area [32, 55]. The respondents’ socioeconomic characteristics and commuting characteristics are considered at the individual level. Respondents’ socioeconomic characteristics include gender, age, hukou type, household income, and car ownership. Commuting characteristics in this study include separation of workplace, residence, and commuting time. The descriptive statistics of the variables are presented in Table 2.

3.3. Model Framework

This study aims to investigate the impact of the multiscale built environment on residents’ commuting mode choice in China. Due to the hierarchical structure of the influential factors, in which the residents are nested within neighborhoods and neighborhoods are clustered within cities, the MML model is employed to address spatial heterogeneity. Five model specifications, including three null models (NMs), one base model (BM), and one moderating effects model (MEM), are utilized to achieve three research objectives: (1) to check whether the spatial heterogeneity in residents’ commuting mode choice exists and to reveal that the built environment at both city and neighborhood levels should be considered in commuting mode choice analysis; (2) to examine the direct effects of the built environment on commuting mode choice at different geographical scales so that potential sources leading to spatial heterogeneity of commuting mode choice can be determined; and (3) to explain how the built environment moderates individual behaviors towards commuting mode choice. The theoretical framework of this study is illustrated in Figure 2.

3.4. Model Specification

In the first step, we develop three NMs (defined as NM 1, NM 2, and NM 3, respectively) to check whether the spatial heterogeneity in residents’ commuting mode choice exist and to test whether built environment effects at both city and neighborhood levels should be considered. The null model, which only contains constant terms and random terms, is widely used to determine the level of spatial heterogeneity [54, 56]. NM 1 is a three-level model, of which the specification includes the city effect random term, neighborhood effect random term, residual error term, and constant term, while NM 2 and NM 3 are both two-level models. NM 2 is defined without the city effect random term, and NM 3 is defined without the neighborhood effect random term. All the specifications of NM 1–NM 3 are shown as follows:where i, j, k, and m represent the resident, neighborhood, city, and commuting mode, respectively. As mentioned in variables’ description, commuting modes include the nonmotorized mode, public transit mode, and car mode. In addition, the car mode is defined as the reference. P_ijkm is the probability of resident i of neighborhood j within city k choosing commuting mode m. β₀ is the mean response across all cities. u_k is the effect of city k, and is the effect of neighborhood j within city k. e is the residual error term. Parameter estimation results of NM 1–3 (shown in next section) show that the spatial heterogeneity in residents’ commuting mode choice exists, and built environment effects at both city and neighborhood levels should be considered in commuting mode choice investigation.

Therefore, in the second step, we build a three-level BM to investigate the direct effects of the built environment at both city and neighborhood levels on commuting mode choice. In BM, we add the independent variables at the city level, neighborhood level, and individual level (listed in Table 3) into the model specification. The specification of BM is shown as follows:where BEC, BEN, and SE are vectors of variables at the city level, neighborhood level, and individual level, respectively. SE includes individuals’ socioeconomic characteristics and commuting characteristics. β_BEC,β_BEN,and β_SE are vectors of parameters associated with different variables.

In the third step, we develop MEM to test how the built environment moderates the effects of commuting attributes (commuting time) on commuting mode choice by using interaction terms. The specification of MEM is shown as follows:where CT is the commuting time and β_m is the vector of parameters for interaction terms. The estimation results of these models are derived by using the maximum likelihood estimation method. Gllamm program in software Stata is applied for parameter estimation.

4. Results and Analysis

4.1. Spatial Heterogeneity

First, we estimate the parameters for NM 1–NM 3 to investigate the spatial heterogeneity in residents’ commuting mode choice. The level of spatial heterogeneity is measured by the intraclass correlation coefficient (ICC) index, of which the value ranges between 0 and 1. ICC is usually used to estimate the correlation among individuals’ choice within the cluster or nested structure. This is a measure of the proportion of variation in the outcome variable that occurs between the groups versus the total variation present. A higher value of ICC indicates that a greater share of the total variation in the outcome measure is associated with cluster membership, i.e., more significant spatial heterogeneity and the multilevel modeling approach should be applied. According to the classic method of Cohen [57], ICC < 0.059, 0.059 <ICC <0.138, and ICC > 0.138 represent a low, moderate, and high spatial heterogeneity between the groups. The formulas of ICC indexes at the neighborhood level and city level are as follows:where ICC_n and ICC_c represent ICC indexes at the neighborhood level and city level, respectively. σ_u and are standard deviations of u_k and , respectively.

Table 3 presents the calculation results of statistical indicators for NM 1–NM 3. As shown in Table 3, ICC_n and ICC_c in NM 1 (equal to 0.27 and 0.14, respectively) are both larger than 0.138. This indicates that there are significant spatial heterogeneities in commuting mode choice across different cities and different neighborhoods within a specific city. Besides, we compare the performance across NM 1, NM 2, and NM 3. Both of the comparison results for log likelihood (LL) and Akaike information criterion (AIC) measures (LL of NM 1 is significantly larger than that of NM 2 and NM 3, and AIC of NM 1 is significantly smaller than that of NM 2 and NM 3) indicate that a three-level model (NM 1) fits the sample data better than the two-level model (NM 2 and NM 3). Moreover, the two-level models will overestimate the variation across different cities (ICC_c in NM 1 and NM 3 equal to 0.14 and 0.19, respectively) while underestimating the variation across different neighborhoods within a specific city (ICC_n in NM 1 and NM 2 equal to 0.27 and 0.25, respectively). Therefore, to reduce the estimated bias, a three-level model considering impact factors at both city and neighborhood levels should be used for commuting mode choice analysis.

4.2. Direct Effects of the Built Environment

In this section, we estimate the parameters for BM to investigate the direct effects of the built environment on commuting mode choice. Table 4 presents the estimation results of the BM model. The explanatory variables remained by using the likelihood ratio test and the variance comparison of the multilevel model when the model was constructed. Even if some of the built environment variables only have a significant impact on a specific commuting mode (e.g., public transit), they have still remained. Remaining these factors can avoid missing important factors and make the results more accurate and reliable [2, 23, 44].

Among the variables at the city level, the GDP, urbanization rate, city population density, number of public transportation vehicles, metro operating mileage, and road area per capita have a significant impact on commuting mode choice. More specifically, GDP, city population density, and road area per capita have a negative impact on choosing public transit (parameters = −0.364, −2.087, and −0.033, respectively). This indicates respondents living in cities with larger GDP, higher city population density, and more road area per capita are less likely to commute by public transit than those living in cities with a smaller GDP, lower city population density, and less road area per capita. Besides, road area per capita also has a negative impact on choosing nonmotorized modes (parameter = −0.043). Respondents living in cities with more road area per capita are less likely to commute by nonmotorized modes than those living in cities with a less road area per capita. However, the urbanization rate, number of public transportation vehicles, and metro operating mileage have positive effects on choosing public transit (parameters = 0.926, 0.087, 1.163, and 2.268, respectively). This implies respondents living in cities with a medium urbanization rate (50%–75%), large number of public transportation vehicles per 10,000 persons, and long metro operating mileage are more likely to commute by public transit than those living in cities with a low urbanization rate (<50%), small number of public transportation vehicles per 10,000 persons, and short metro operating mileage. It needs to be mentioned that the odds ratio (OR) value of metro operating mileage >100 km (OR = 9.66) is larger than that of metro operating mileage <100 km (OR = 3.12), indicating that when metro operating mileage reaches a high level (≥100 km), the positive effect of metro operation mileage on choosing public transit is enhanced.

Second, among neighborhood-level variables, neighborhood population density has a positive effect on choosing public transit; green coverage, air quality, and neighborhood location have positive effects on choosing both public transit and nonmotorized modes. These results are in line with studies conducted by Antipova et al. [41] and Ding et al. [58]. Land-use diversity has a negative impact on choosing both public transit and nonmotorized modes. This is out of our expectations. Nonetheless, several previous research studies still obtained the same results, such as [2, 59, 60]. What these studies have in common is that research objects for most of these studies are developing countries (e.g., Chile, Iran, and Mexico). The reason may be that land-use diversity planning in developing countries pays more attention to the construction of supporting facilities for car travel. In addition, there are two possible explanations for this result: (1) neighborhoods with diversified land use usually have more facilities (e.g., squares), which can provide enough parking lots. This promotes the use of private cars [61]. (2) The mixed-use land could reduce the travel cost, which stimulates the car travel demand [62]. Finally, among individual-level variables, all of individual socioeconomic attributes listed in Table 3 have a significant impact on commuting mode choice. Respondents who are elder, female, with nonlocal hukou, have lower household income, and without a private car, they prefer to commute by nonmotorized modes and public transit. Besides that, variables of commuting characteristic including separation of workplace and residence and commuting time are negatively correlated with choosing nonmotorized modes but are positively correlated with choosing public transit.

4.3. Moderating Effects of the Built Environment

Since commuting time plays a dominant role in commuting mode choice, we add the interaction terms between built environment variables and commuting time into the BM to examine the moderating effect of the built environment (i.e., MEM). The MEM is important for understanding how the built environment moderates the impact of commuting time on commuting mode choice. Table 5 presents the estimation results of MEM. Both of the comparison results for LL and AIC measures between BM and MEM indicate that MEM fits the sample data better than BM. Therefore, the moderating effect of the built environment should be considered in commuting mode choice analysis.

As shown in Table 5, among built environment variables at both the neighborhood level and city level, the urbanization rate and neighborhood population density moderate the effect of commuting time on choosing nonmotorized modes significantly, while neighborhood location moderates the effect of commuting time on choosing public transit significantly. Specifically, the urbanization rate and neighborhood population density increase the negative effect of commuting time on choosing nonmotorized modes, indicating respondents living in cities with a low urbanization rate (<50%), or neighborhoods with low population density (<5000 persons/km²) have higher tolerance for commuting time of nonmotorized modes. This may be due to there being a better walking (and cycling) environment (e.g., better air quality, narrower roads, and fewer vehicles) in cities with a low urbanization rate and neighborhoods with low population density. Besides, neighborhood location decreases the positive effect of commuting time on choosing public transit. This implies that respondents living in the urban area have lower tolerance for commuting time of public transit than those living in the suburban area.

4.4. Implications for Demand Estimation

The estimation results of MEM indicate that respondents living in different built environment contexts do have preference heterogeneity for commuting time when they choose the commuting mode. If the moderating effects of the built environment are neglected, it will cause bias in the results of demand forecasting. Therefore, in this section, we calculate the mode shares of nonmotorized modes (and public transit) under different levels of commuting time in different built environment contexts using MEM.

Based on the built environment variables (including the urbanization rate, neighborhood population density, and neighborhood location) which moderate the impact of commuting time on commuting mode choice, nine categories of built environment contexts with heterogeneous preference for commuting time of nonmotorized modes can be obtained and the two categories of built environment contexts with heterogeneous preference for commuting time of public transit can be obtained (listed in Table 6).

Figure 3 shows the calculation results. It can be seen that significantly different mode shares among these built environment contexts are derived. For nonmotorized modes (shown in Figure 3(a)), a maximum of 10.1% difference in mode shares change is observed. When commuting time decreases, mode shares of nonmotorized modes for categories 6 and 9, in which built environments with high neighborhood population density (≥10000 persons/km²) increase much faster than that for other categories. For public transit (shown in Figure 3(b)), a maximum of 3.5% difference in mode shares change is observed. Compared with nonmotorized modes, the differences in the changes in the mode share of public transit with the decrease in the commuting time are much smaller. This is due to the fact that mode share of public transit accounts for only 15.4% in our sample data (shown in Table 2), which is a relatively low level. Additionally, the results of the mode share calculation indicate that decrease in commuting time can significantly increase the mode share of nonmotorized modes significantly while decreasing the mode share of public transit relatively insignificantly. This implies that adopting reasonable land use planning to reduce commuting time is effective for promoting the usage of sustainable commuting modes.

(a)

(b)

5. Conclusion

This study investigated the relationship between multiscale built environments and residents’ commuting mode choice using CLDS cross-city questionnaire data. In particular, we focused on the link between the built environment and the usage of sustainable commuting modes (including nonmotorized modes and public transit). First, by building a three-level MML model, we revealed the significant spatial heterogeneities in commuting mode choice across different cities and different neighborhoods within a specific city. Then, we explored the potential built environment variables that contribute to the spatial heterogeneity. The direct effects of these built environments on commuting mode choice and the moderating effects of these built environments on commuting time were examined. In addition, the mode shares of the nonmotorized mode (and public transit) under different levels of commuting time were estimated in different built environment contexts. The conclusions of this study were summarized as follows:(a)Built environments at both city and neighborhood levels influence the commuting mode choice behavior significantly, and they should be considered in mode choice modeling.(b)Among the built environments at the city level, the increasing urbanization rate, number of public transportation vehicles, and metro operating mileage are expected to increase the possibility of commuting by public transit; but the increasing GDP, city population density, and road area per capita are likely to decrease the possibility of commuting by public transit; increasing road area per capita can also decrease the possibility of commuting by the nonmotorized mode.(c)Among the built environments at the neighborhood level, high neighborhood population density has a positive effect on choosing public transit; large green coverage, good air quality, and neighborhood located in the urban area have a positive effect on choosing both public transit and nonmotorized modes, while land-use diversity has a negative impact on choosing both public transit and nonmotorized modes.(d)The effect of commuting time on choosing nonmotorized modes is moderated by the urbanization rate and neighborhood population density. Generally, the increasing urbanization rate and neighborhood population density would strengthen the negative effect of commuting time on choosing nonmotorized modes.(e)The effect of commuting time on choosing public transit is moderated by neighborhood location. Specifically, residents living in the urban area have lower tolerance for commuting time of public transit than those living in the suburban area.(f)Decreasing commuting time would increase the mode share of nonmotorized modes, especially in neighborhoods with high population density (≥10000 persons/km²). However, decreasing commuting time would decrease the mode share of public transit, especially for neighborhoods located in the urban area.

The recommendation for future work is to expand the analysis to other trip purposes so as to provide comprehensive suggestions for shifting car users to sustainable travel modes. Moreover, the multilevel model used in this study can only identify the statistical relationship between the built environment and commuting mode choice. In future research, the machine learning method (such as the gradient boosting decision tree) can be used to measure the relative importance of independent variables and visualize the nonlinear effects of the built environment on commuting mode choice.

Data Availability

The data used in this study are from the China Labor-Force Dynamics Survey (CLDS) by the Centre for Social Science Survey at Sun Yat-sen University in Guangzhou, China. Please refer to https://css.sysu.edu.cn for more information about the CLDS data.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The authors gratefully acknowledge financial support from the Natural Science Foundation of Fujian Province (item nos.: 2020J05194, 2021J05226, and 2022J01938) and Innovation and Entrepreneurship Training Program for College Students (item no.: s202210388089).

References

P. Zhu, S. N. Ho, Y. Jiang, and X. Tan, “Built environment, commuting behaviour and job accessibility in a rail-based dense urban context,” Transportation Research Part D: Transport and Environment, vol. 87, Article ID 102438, 2020.
View at: Publisher Site | Google Scholar
D. A. Bautista-Hernandez, “Mode choice in commuting and the built environment in México City. Is there a chance for non-motorized travel?” Journal of Transport Geography, vol. 92, Article ID 103024, 2021.
View at: Publisher Site | Google Scholar
T. Munshi, “Built environment and mode choice relationship for commute travel in the city of Rajkot, India,” Transportation Research Part D: Transport and Environment, vol. 44, pp. 239–253, 2016.
View at: Publisher Site | Google Scholar
C. Ding, X. Cao, and Y. Wang, “Synergistic effects of the built environment and commuting programs on commute mode choice,” Transportation Research Part A: Policy and Practice, vol. 118, pp. 104–118, 2018.
View at: Publisher Site | Google Scholar
T. Tao, J. Wang, and X. Cao, “Exploring the non-linear associations between spatial attributes and walking distance to transit,” Journal of Transport Geography, vol. 82, Article ID 102560, 2020.
View at: Publisher Site | Google Scholar
J. L. Renne, S. Hamidi, and R. Ewing, “Transit commuting, the network accessibility effect, and the built environment in station areas across the United States,” Research in Transportation Economics, vol. 60, no. 12, pp. 35–43, 2016.
View at: Publisher Site | Google Scholar
Z. Zhu, Z. Li, Y. Liu, H. Chen, and J. Zeng, “The impact of urban characteristics and residents’ income on commuting in China,” Transportation Research Part D: Transport and Environment, vol. 57, no. 12, pp. 474–483, 2017.
View at: Publisher Site | Google Scholar
A. Nasri and L. Zhang, “Multi-level urban form and commuting mode share in rail station areas across the United States; a seemingly unrelated regression approach,” Transport Policy, vol. 81, pp. 311–319, 2019.
View at: Publisher Site | Google Scholar
Y. Guo and S. Y. He, “The role of objective and perceived built environments in affecting dockless bike-sharing as a feeder mode choice of metro commuting,” Transportation Research Part A: Policy and Practice, vol. 149, pp. 377–396, 2021.
View at: Publisher Site | Google Scholar
Z. Li and D. A. Hensher, “Crowding and public transport: a review of willingness to pay evidence and its relevance in project appraisal,” Transport Policy, vol. 18, no. 6, pp. 880–887, 2011.
View at: Publisher Site | Google Scholar
C. Carrion and D. Levinson, “Value of travel time reliability: a review of current evidence,” Transportation Research Part A: Policy and Practice, vol. 46, no. 4, pp. 720–741, 2012.
View at: Publisher Site | Google Scholar
L. Redman, M. Friman, T. Gärling, and T. Hartig, “Quality attributes of public transport that attract car users: a research review,” Transport Policy, vol. 25, pp. 119–127, 2013.
View at: Publisher Site | Google Scholar
D. A. Hensher, J. M. Rose, and W. H. Greene, Applied Choice Analysis, Cambridge University Press, Cambridge, UK, 2 edition, 2015.
K. Gao, L. Sun, H. Tu, and H. Li, “Heterogeneity in valuation of travel time reliability and in-vehicle crowding for mode choices in multimodal networks,” Journal of Transportation Engineering, Part A: Systems, vol. 144, no. 10, Article ID 4018061, 2018.
View at: Publisher Site | Google Scholar
C. R. Bhat and R. Sardesai, “The impact of stop-making and travel time reliability on commute mode choice,” Transportation Research Part B: Methodological, vol. 40, no. 9, pp. 709–730, 2006.
View at: Publisher Site | Google Scholar
Z. Mao, D. Ettema, and M. Dijst, “Analysis of travel time and mode choice shift for non-work stops in commuting: case study of Beijing, China,” Transportation, vol. 45, no. 3, pp. 751–766, 2018.
View at: Publisher Site | Google Scholar
D. Nam, D. Park, and A. Khamkongkhun, “Estimation of value of travel time reliability,” Journal of Advanced Transportation, vol. 39, no. 1, pp. 39–61, 2005.
View at: Publisher Site | Google Scholar
J. Ha, S. Lee, and J. Ko, “Unraveling the impact of travel time, cost, and transit burdens on commute mode choice for different income and age groups,” Transportation Research Part A: Policy and Practice, vol. 141, pp. 147–166, 2020.
View at: Publisher Site | Google Scholar
D. Ton, S. Bekhor, O. Cats, D. C. Duives, S. Hoogendoorn-Lanser, and S. P. Hoogendoorn, “The experienced mode choice set and its determinants: commuting trips in The Netherlands,” Transportation Research Part A: Policy and Practice, vol. 132, pp. 744–758, 2020.
View at: Publisher Site | Google Scholar
P. Arbués, J. F. Baños, M. Mayor, and P. Suárez, “Determinants of ground transport modal choice in long-distance trips in Spain,” Transportation Research Part A: Policy and Practice, vol. 84, pp. 131–143, 2016.
View at: Publisher Site | Google Scholar
S. Ghader, A. Darzi, and L. Zhang, “Modeling effects of travel time reliability on mode choice using cumulative prospect theory,” Transportation Research Part C: Emerging Technologies, vol. 108, pp. 245–254, 2019.
View at: Publisher Site | Google Scholar
D. Ton, D. C. Duives, O. Cats, S. Hoogendoorn-Lanser, and S. P. Hoogendoorn, “Cycling or walking? Determinants of mode choice in The Netherlands,” Transportation Research Part A: Policy and Practice, vol. 123, pp. 7–23, 2019.
View at: Publisher Site | Google Scholar
Y. Ao, Y. Zhang, Y. Wang, Y. Chen, and L. Yang, “Influences of rural built environment on travel mode choice of rural residents: the case of rural Sichuan,” Journal of Transport Geography, vol. 85, Article ID 102708, 2020.
View at: Publisher Site | Google Scholar
B. Sun, A. Ermagun, and B. Dan, “Built environmental impacts on commuting mode choice and distance: evidence from Shanghai,” Transportation Research Part D: Transport and Environment, vol. 52, pp. 441–453, 2017.
View at: Publisher Site | Google Scholar
X. Wang, C. Shao, C. Yin, and C. Dong, “Exploring the effects of the built environment on commuting mode choice in neighborhoods near public transit stations: evidence from China,” Transportation Planning and Technology, vol. 44, no. 1, pp. 111–127, 2020.
View at: Publisher Site | Google Scholar
G. Eldeeb, M. Mohamed, and A. Páez, “Built for active travel? Investigating the contextual effects of the built environment on transportation mode choice,” Journal of Transport Geography, vol. 96, Article ID 103158, 2021.
View at: Publisher Site | Google Scholar
J. De Vos, L. Cheng, M. Kamruzzaman, and F. Witlox, “The indirect effect of the built environment on travel mode choice: a focus on recent movers,” Journal of Transport Geography, vol. 91, Article ID 102983, 2021.
View at: Publisher Site | Google Scholar
Y. Hu, A. Sobhani, and D. Ettema, “Exploring commute mode choice in dual-earner households in a small Chinese city,” Transportation Research Part D: Transport and Environment, vol. 102, Article ID 103148, 2022.
View at: Publisher Site | Google Scholar
B. Mo, Y. Shen, and J. Zhao, “Impact of built environment on first- and last-mile travel mode choice,” Transportation Research Record Journal of the Transportation Research Board, vol. 2672, no. 6, pp. 40–51, 2018.
View at: Publisher Site | Google Scholar
Z. Wang, Y. Yang, E. Yao, Y. Zhu, and Z. Rao, “An analysis of commute mode choice behavior considering the impacts of built environment in Beijing,” Transportation Letters, vol. 14, pp. 733–739, 2021.
View at: Publisher Site | Google Scholar
L. Yang, C. Ding, Y. Ju, and B. Yu, “Driving as a commuting travel mode choice of car owners in urban China: roles of the built environment,” Cities, vol. 112, Article ID 103114, 2021.
View at: Publisher Site | Google Scholar
J. Cheng, R. Yan, and Y. Gao, “Exploring spatial heterogeneity in accessibility and transit mode choice,” Transportation Research Part D: Transport and Environment, vol. 87, Article ID 102521, 2020.
View at: Publisher Site | Google Scholar
C. Ding, D. Wang, C. Liu, Y. Zhang, and J. Yang, “Exploring the influence of built environment on travel mode choice considering the mediating effects of car ownership and travel distance,” Transportation Research Part A: Policy and Practice, vol. 100, pp. 65–80, 2017.
View at: Publisher Site | Google Scholar
B. Martín and A. Páez, “Individual and geographic variations in the propensity to travel by active modes in Vitoria-Gasteiz, Spain,” Journal of Transport Geography, vol. 76, pp. 103–113, 2019.
View at: Publisher Site | Google Scholar
C. Mulley and L. Ma, “How the longer term success of a social marketing program is influenced by socio-demographics and the built environment,” Transportation, vol. 45, no. 2, pp. 291–309, 2017.
View at: Publisher Site | Google Scholar
C. E. T. Neves, A. R. Da Silva, and F. S. de Arruda, “Exploring the link between built environment and walking choice in São Paulo city, Brazil,” Journal of Transport Geography, vol. 93, Article ID 103064, 2021.
View at: Publisher Site | Google Scholar
D. Lin, A. Allan, and J. Cui, “The impacts of urban spatial structure and socio-economic factors on patterns of commuting: a review,” International Journal on the Unity of the Sciences, vol. 19, no. 2, pp. 238–255, 2015.
View at: Publisher Site | Google Scholar
J. Hong and A. Goodchild, “Land use policies and transport emissions: modeling the impact of trip speed, vehicle characteristics and residential location,” Transportation Research Part D: Transport and Environment, vol. 26, pp. 47–51, 2014.
View at: Publisher Site | Google Scholar
C. Ding, X. Cao, B. Yu, and Y. Ju, “Non-linear associations between zonal built environment attributes and transit commuting mode choice accounting for spatial heterogeneity,” Transportation Research Part A: Policy and Practice, vol. 148, pp. 22–35, 2021.
View at: Publisher Site | Google Scholar
W. Wu and J. Hong, “Does public transit improvement affect commuting behavior in Beijing, China? A spatial multilevel approach,” Transportation Research Part D: Transport and Environment, vol. 52, pp. 471–479, 2017.
View at: Publisher Site | Google Scholar
A. Antipova, F. Wang, and C. Wilmot, “Urban land uses, socio-demographic attributes and commuting: a multilevel modeling approach,” Applied Geography, vol. 31, no. 3, pp. 1010–1018, 2011.
View at: Publisher Site | Google Scholar
J. Huo, H. Yang, C. Li, R. Zheng, L. Yang, and Y. Wen, “Influence of the built environment on E-scooter sharing ridership: a tale of five cities,” Journal of Transport Geography, vol. 93, Article ID 103084, 2021.
View at: Publisher Site | Google Scholar
J. Xianyu, “Exploring the impacts of built environment on commute mode choice: evidence from China,” Mathematical Problems in Engineering, vol. 2022, Article ID 7939950, 11 pages, 2022.
View at: Publisher Site | Google Scholar
P. Nakshi and A. K. Debnath, “Impact of built environment on mode choice to major destinations in dhaka,” Transportation Research Record, vol. 2675, no. 4, pp. 281–296, 2021.
View at: Publisher Site | Google Scholar
R. Ewing and R. Cervero, “Travel and the built environment: a meta-analysis,” Journal of the American Planning Association, vol. 76, no. 3, pp. 265–294, 2010.
View at: Publisher Site | Google Scholar
S. Melia, G. Parkhurst, and H. Barton, “The paradox of intensification,” Transport Policy, vol. 18, no. 1, pp. 46–52, 2011.
View at: Publisher Site | Google Scholar
J. Yang, S. French, J. Holt, and X. Zhang, “Measuring the structure of U.S. Metropolitan areas, 1970–2000,” Journal of the American Planning Association, vol. 78, no. 2, pp. 197–209, 2012.
View at: Publisher Site | Google Scholar
E. Guerra, H. Zhang, L. Hassall, J. Wang, and A. Cheyette, “Who cycles to work and where? A comparative multilevel analysis of urban commuters in the US and Mexico,” Transportation Research Part D: Transport and Environment, vol. 87, Article ID 102554, 2020.
View at: Publisher Site | Google Scholar
X. Delclòs-Alió, C. Kanai, L. Soriano et al., “Cars in Latin America: an exploration of the urban landscape and street network correlates of motorization in 300 cities,” Travel Behaviour and Society, vol. 30, pp. 192–201, 2023.
View at: Publisher Site | Google Scholar
C. Yin, C. Shao, C. Dong, and X. Wang, “Happiness in urbanizing China: the role of commuting and multi-scale built environment across urban regions,” Transportation Research Part D: Transport and Environment, vol. 74, pp. 306–317, 2019.
View at: Publisher Site | Google Scholar
C. Yin and B. Sun, “Disentangling the effects of the built environment on car ownership: a multi-level analysis of Chinese cities,” Cities, vol. 74, no. 5, pp. 188–195, 2018.
View at: Publisher Site | Google Scholar
E. Guerra, C. Caudillo, P. Monkkonen, and J. Montejano, “Urban form, transit supply, and travel behavior in Latin America: evidence from Mexico's 100 largest urban areas,” Transport Policy, vol. 69, pp. 98–105, 2018.
View at: Publisher Site | Google Scholar
L. Engelfriet and E. Koomen, “The impact of urban form on commuting in large Chinese cities,” Transportation, vol. 45, no. 5, pp. 1269–1295, 2017.
View at: Publisher Site | Google Scholar
C. Ding, Y. Wang, T. Tang, S. Mishra, and C. Liu, “Joint analysis of the spatial impacts of built environment on car ownership and travel mode choice,” Transportation Research Part D: Transport and Environment, vol. 60, pp. 28–40, 2018.
View at: Publisher Site | Google Scholar
K. Mouratidis, D. Ettema, and P. Næss, “Urban form, travel behavior, and travel satisfaction,” Transportation Research Part A: Policy and Practice, vol. 129, pp. 306–320, 2019.
View at: Publisher Site | Google Scholar
T. A. Snijders and R. J. Bosker, Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling, SAGE, California, UK, 2011.
J. Cohen, Statistical Power Analysis for the Behavioral Sciences, Academic Press, Hillsdale, MI, USA, 2 edition, 1988.
C. Ding, X. Cao, and C. Liu, “How does the station-area built environment influence Metrorail ridership? Using gradient boosting decision trees to identify non-linear thresholds,” Journal of Transport Geography, vol. 77, pp. 70–78, 2019.
View at: Publisher Site | Google Scholar
A. Soltani, “Social and urban form determinants of vehicle ownership; evidence from a developing country,” Transportation Research Part A: Policy and Practice, vol. 96, pp. 90–100, 2017.
View at: Publisher Site | Google Scholar
C. Zegras, “The Built Environment and Motor Vehicle Ownership and Use: evidence from Santiago de Chile,” Urban Studies, vol. 47, no. 8, pp. 1793–1817, 2010.
View at: Publisher Site | Google Scholar
Z. Guo, “Does residential parking supply affect household car ownership? The case of New York City,” Journal of Transport Geography, vol. 26, pp. 18–28, 2013.
View at: Publisher Site | Google Scholar
M. G. Boarnet and R. Crane, Travel by Design: The Influence of Urban Form on Travel, Oxford University Press on Demand, Oxford, UK, 2001.

Copyright

Copyright © 2023 Wenliang Jian et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies