Travel behavior analysis focusing on private vehicle usage and switch to public transport in ho chi minh city doctor of philosophy major engineering

Travel Behavior Analysis Focusing on Private Vehicle Usage and Switch to Public Transport in Ho Chi Minh City NGUYEN Ngoc Thi TRAVEL BEHAVIOR ANALYSIS FOCUSING ON PRIVATE VEHICLE USAGE AND SWITCH TO P[.]

Travel Behavior Analysis Focusing on Private Vehicle Usage and Switch

to Public Transport in Ho Chi Minh City

NGUYEN Ngoc Thi

TRAVEL BEHAVIOR ANALYSIS FOCUSING ON PRIVATE VEHICLE USAGE AND SWITCH TO PUBLIC TRANSPORT

IN HO CHI MINH CITY

by

NGUYEN Ngoc Thi

Submitted to the Graduate School of Environmental Studies

In Partial Fulfillment of the Requirements for the Degree of

Doctor of Engineering

at Nagoya University

July 2018

Academic Advisers:

Professor Morikawa Takayuki Professor Tanikawa Hiroki Associate Professor Miwa Tomio

3

Abstract

A public transport (PT) system that is a low carbon transport option is vital for sustainable urban development However, introducing this type of system in developing cities may be a challenge given residents’ common practice of using private vehicles, especially motorcycles Take Ho Chi Minh City as an example, the usage preference of private vehicle users for their own vehicles, the current bus system, and the future PT system were investigated This research aims to develop analyses based on using practical data about vehicle usage and applying feasible methods and/or improving them for understanding the situation and for finding solutions boosting sustainable travel behaviors in the city Methodologies of demand modeling, discrete choice models were used to explore individual behaviors on the private vehicle usage and switch to PT The advantage of revealed preference data, stated preference data, and experiment design were taken in specific analyses

The preference of private vehicle users on their own vehicles is analyzed using revealed data

A joint discrete-continuous model based on the copula approach is used to overcome selectivity bias in the data and to address the relationship between vehicle type choice (a discrete outcome) and usage (a continuous outcome) by specifying a joint distribution The analysis found significant roles of socioeconomic attributes on the individual choice, a trend

of modal shift from motorcycles to cars, and CO2 emissions from this trend

The motivations so that the private vehicle users switch to PT were analyzed using a revealed-stated preference data The analyses followed a two-stage approach consisting of a multiple-indicator–multiple-cause model for capturing psychological determinants and a bivariate ordered probit model for explaining the decisions of each user group on usage frequencies of the current buses and the new PT system The new PT usage was found to be correlated with bus usage The significant roles were explored in factors of access/egress time, fare/cost, congestion/comfort, social interaction, agreement to the PT projects, dissatisfaction with PT, distance from home to workplace, motorcycle ownership, occupation, and age In addition, by adding a component of latent class assignment, the heterogeneity among motorcycle users was detected in choice models in the latent classes The “collectivistic” and “individualistic” tendencies in the two latent classes were found to make the individuals behave in different ways

Lastly, by adding a dimension of in-vehicle occupancy into the traditional social interaction that reflects individual’s behavior and other people’s behavior, an equilibrium calculation

on both positive and negative mass effects was able to conduct based on loop procedures The results give envision on travel demand for commuting trips by PT in the future situation

Thesis Supervisor: Morikawa Takayuki

Title: Professor, Graduate School of Environmental Studies

4

Acknowledgments

This work was partly supported by the Japanese Government (Monbukagakusho: MEXT) Scholarship awarded through the Forefront Studies Program on Civil and Environmental Engineering for Sustainable Co-Development of Nagoya University It was also partly supported by the Nagoya University Transportation and Environment Dynamics (NU TREND) Laboratory

I wish to express my appreciation to those who have helped in carrying out this thesis

I am particularly indebted to Professor Morikawa Takayuki who gave me the opportunity to study in Nagoya University As my thesis supervisor, he provided me with valuable suggestions and insightful guidance during the planning and development of research works Special thanks should be also given to Associate Professor Miwa Tomio, who worked as my academic advisor and gave me precious advice throughout my research works I am grateful for his willingness, patience, and enthusiastic encouragement Professor Tanikawa Hiroki has served as a committee member He gave me very useful comments and suggestions from his extensive experiences that significantly improved the clarity of this thesis

My gratitude are also expressed to Lecturer Sato Hitomi for her great help in collecting data Professor Yamamoto Toshiyuki and Associate Professor Kanamori Ryo gave me helpful comments Lecturer Tashiro Mutsumi, Ms Kikata Chiharu, and Ms Tsuda Junko provided

me with their valuable assistances through my stay in Nagoya University

I would like to acknowledge the support of Associate Professor To Hien T., Ms Nguyen Nguyen T., and Mr Tran Vu of VNUHCM-University of Science in conducting the pilot survey

I would like to thank my lab mates, Chu Dung T., Tosa Cristian, Gong Lei, Phyu Phyu Thwe, Mothafer Ghasak, Li Yan Yan, Liu Zhiquang, Hao Mingyang, Ye Lanhang, Ban Takumi, Tang Routian, and Sangeetha Ann I learned both academic and non-academic things from them Discussions with Chu and Tosa are extremely useful in conducting empirical analyses Finally, my sincere thanks are extended to my parents and my husband for their understanding and encouragement throughout my study

5

Table of contents

Abstract 3

Acknowledgments 4

Table of contents 5

List of Figures 8

List of Tables 9

List of Abbreviations 10

Introduction 11

1.1 Background 11

1.1.1 Overloaded Traffic Infrastructure in Ho Chi Minh City 11

1.1.2 Select a “Green” Mode: the Planned Public Transport System 12

1.1.3 Travel behavior analysis 13

1.2 Problem Statements and Objective of this Research 14

1.3 Outline of the Thesis 14

Literature Reviews 16

2.1 Local Context and Public Transport Development 16

2.2 Role of Travel Behavior Analysis in Transport Planning 17

2.3 Habitual Travel Choice 17

2.4 Barriers and Motivations of Sustainable Travel Behavior 19

2.5 Summary 19

Data Collection 20

3.1 A Dataset of Vehicle Type Choice and Usage 20

3.2 First Revealed Preference-Stated Preference Survey on Public Transport 21

3.3 Second Revealed Preference-Stated Preference Survey on Public Transport 23

3.4 Auxiliary Data 25

Vehicle Type Choice, Usage, and CO 2 Emission 27

4.1 Methodology for Discrete-Continuous Choices 28

6

4.2 Modeling Framework 29

4.2.1 Multinomial Logit Component of Vehicle Type Choice 30

4.2.2 Regression Component of Usage 30

4.2.3 Joint Model 31

4.3 Data Preparation 33

4.4 Results and Discussion 38

4.4.1 Model Estimation Results 38

4.4.2 Model Application 42

4.5 Summary 44

Motivations on Switching to Public Transport Modes for Commuting Trips 46

5.1 State of the Analysis 46

5.1.1 Role of Psychological Determinants in Travel Behavior Analysis 46

5.1.2 Effects of Social Interaction 47

5.1.3 Combination of Revealed Preferences and Stated Preferences 48

5.2 Modeling Framework 49

5.2.1 MIMIC Model 49

5.2.2 Bivariate Ordered Probit Model 50

5.3 Data Preparation 52

5.4 Results and Discussion 56

5.4.1 Results of the MIMIC Model 56

5.4.2 Results of the Bivariate Ordered Probit Model 58

5.5 Summary 62

Latent Classes in Response to the Planned Public Transport System 64

6.1 Latent Class Assignment: A brief review 64

6.2 Methodological Framework 65

6.3 Data Preparation 69

6.4 Results and Discussion 71

6.4.1 Latent Classes 71

6.4.2 Utility Function of PT Usage 73

7

6.4.3 Implications for Policy Making 75

6.5 Summary 76

Positive and Negative Mass Effects: Equilibrium in Public Transport Usage 78

7.1 Methodological Framework 79

7.1.1 Modeling 79

7.1.2 Equilibrium Analysis 81

7.2 Data Preparation 83

7.3 Results and Discussion 86

7.3.1 Psychological Determinants 86

7.3.2 Utility of PT Usage 88

7.3.3 Mass Effects and Equilibrium 91

7.3.4 Implications for Policy Making 93

7.4 Summary 94

Conclusions and Future works 95

References 99

Appendix A 109

Appendix B 121

8

List of Figures

Figure 1.1 Travel demand is growing 12

Figure 2.1 The process of script-based choice (Gärling et al., 2001) 18

Figure 3.1 Example of a case in the first experiment 22

Figure 3.2 Example of a case in the second experiment 24

Figure 4.1 CO2 emissions in various scenarios 44

Figure 5.1 Distribution of occupation among motorcycle users and car users 52

Figure 5.2 Distribution of decisions on usage frequencies of PT 55

Figure 5.3 Reasons not to use the new PT system more frequently 55

Figure 5.4 Relationships among latent variables 58

Figure 6.1 Methodological framework 66

Figure 6.2 Improvement in modeling framework 67

Figure 6.3 Outcome probabilities 74

Figure 6.4 Sensitivity analysis for Class 1 “Collectivistic” 76

Figure 6.5 Sensitivity analysis for Class 2 “Individualistic” 76

Figure 7.1 Modeling framework 80

Figure 7.2 Flow chart of the algorithm for calculating equilibrium points 82

Figure 7.3 Distribution of choice in frequencies of PT usage for commuting trips 86

Figure 7.4 Structural relationships for latent variables 88

Figure 7.5 Equilibrium for the two-dimensional social interaction 92 Figure 7.6 Output of the equilibrium for the three-dimensional social interaction analysis 92

9

List of Tables

Table 3.1 Description of factors in the first experiment 22

Table 3.2 Description of factors in the second experiment 24

Table 4.1 Distribution of vehicle type choices and usage 35

Table 4.2 Descriptions of the explanatory variables 36

Table 4.3 Estimation result of the ordered probit model for income 37

Table 4.4 Estimation result of the ordered probit model for expenditure 37

Table 4.5 Estimated parameters of the joint model 40

Table 4.6 Validation results 41

Table 4.7 Changes in the 10th year compared with the base case 43

Table 5.1 Intervals specific for categories 51

Table 5.2 Summary statistics by private vehicle users 53

Table 5.3 Descriptive statistics of indicators and socioeconomic variables for psychological determinants 54

Table 5.4 Relationship between latent variables and indicators 57

Table 5.5 Relationship between causal variables and latent variables and DIF effects 57

Table 5.6 Estimated parameters of the bivariate ordered probit models 60

Table 6.1 Descriptive statistics (N=591) 70

Table 6.2 Latent class choice model for the usage frequency of PT of motorcycle users 72

Table 7.1 Demographic Statistics (N=1030) 84

Table 7.2 Descriptive Statistics of Indicators for Psychological Determinants (N=1030) 85 Table 7.3 Relationship between Latent Variables/Psychological Determinants and Indicators 87

Table 7.4 Testing DIF 87

Table 7.5 The utility model of PT usage 90

Table 7.6 Sensitivity 94

10

List of Abbreviations

AIC Akaike’s information criterion

APV Awareness of problems relating to private vehicles APP Agreement to public transport projects

CNG Compressed Natural Gas

DB Dissatisfaction with bus

DIF Differential item functioning

DPT Dissatisfaction with the new public transport system

MIMIC Multiple-indicator–multiple-cause

RUM Random Utility Maximization

VND Vietnamese Dong, the currency of Vietnam

11

Introduction

1.1 Background

The motivation of this study is from the transportation situation of Ho Chi Minh City (HCMC) in that the current traffic infrastructure is overloaded by a huge amount of private vehicles and the effectiveness of the planned public transport (PT) system is questioned Along with the rapid economic growth, population growth, and urbanization, the transportation system would be worse if the private vehicle dependency is not reduced and the new PT system is not preferred

1.1.1 Overloaded Traffic Infrastructure in Ho Chi Minh City

HCMC is the biggest city Vietnam with an area of 2096 km2 and a population of 8.4 million people Because most of the population (80%) live in urban areas, the density in these areas has reached about 14000 people/km2 (HCMC Statistical Office, 2017) Regarding the transportation system, based on roadway, both the road density and land area for transportation are now substandard The road density is 1.9 km/km2 compared with the standard of 10 km/km2 The land area for transportation is 8.2% compared with the standard

of 24% (Vietnam Government, 2011) While the infrastructure is overloaded, travel demand

is growing year by year The data of five recent years shows a quick increase in the number

of motorcycles and cars along with an increase of population (see Figure 1.1) Notably, the

PT system with only buses witnesses a steady decrease in usage in the same period Emberger (2016) reported a modal split of around 3% pedestrians, 1% cycling, 6% PT, 9% cars, and 81% motorcycles The trend of private vehicle dependency may continue to increase because the quality of the current bus system is poor, and people’s wealth may encourage the use of private vehicles (Dargay and Gately, 1999; Dalkmann and Huizenga, 2010)

Consequently, the transport system faces challenges in terms of capacity constraints and environmental problems The urban space has been pressured by traffic congestion, not only during rush hours This causes a certain loss of time and difficulties in accessing services Also, the unsafety becomes serious There were over 700 people died because of traffic accidents in 2017 (Ministry of Public Security, 2018) The ambient air is polluted by noise and traffic emissions that are mainly caused by motorcycles (Ho and Clappier, 2011)

12

Figure 1.1 Travel demand is growing

1.1.2 Select a “Green” Mode: the Planned Public Transport System

The environmental problem of transportation is related to the energy used by traditional modes of transportation When considering urban sustainability and efficiencies in transport and energy use, both private and PT have been explored In general, PT, which is a mass transit system, has been suggested as more effective than private vehicles because of its high capacity, resulting in less congestion on roadways and less energy consumption per passenger Kii and Hanaoka (2003) argue that the effectiveness of PT depends on the urban structure If population density is low, demand will be small, resulting in higher energy consumption per passenger with PT than with private vehicles Moreover, technological progress towards cleaner vehicles makes private vehicles more attractive when considering environmental benefits

Even in a dense city, transit-oriented development has the unavoidable challenge of public adoption Kenworthy (2008) found in many cities that a successful PT system depends

on its extent and quality In Los Angeles, Chester et al (2013) found that the environmental goal is only achieved with a modal shift of between 20 and 30%

A combination of high capacity, high population density, and a modal shift contribute to the cleanness of PT systems In HCMC, there are several reasons PT should be

a good selection First, PT helps reduce the pressure created by a huge number of private vehicles (7.4 million) on urban spaces with a population density of about 4,000 persons/km2 (HCMC Statistical Office, 2017) Second, most of the private vehicles currently operated in HCMC use traditional fuels, which pollute the environment A mass replacement of these vehicles with cleaner private vehicles is not easy to accomplish (see Egbue et al (2017) for the case of electric vehicles) Third, the new PT system includes cleaner technologies The