A study on a ranking method for the provincial road traffic safety index in vietnam

6 Figure 2.2: Twenty provinces and cities having the highest and the lowest total number of road traffic accidents, total number of deaths and injuries due to RTAs in 2016 .... 7 Figure

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

VU THI HUONG KHUE

A STUDY ON A RANKING METHOD FOR THE PROVINCIAL ROAD TRAFFIC SAFETY INDEX IN VIETNAM

MASTER’S THESIS

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

VU THI HUONG KHUE

A STUDY ON A RANKING METHOD FOR THE PROVINCIAL ROAD TRAFFIC SAFETY INDEX IN VIETNAM

MAJOR: INFRASTRUCTURE ENGINEERING

CODE: 8900201.04QTD

RESEARCH SUPERVISOR:

Assoc Prof VU HOAI NAM

Hanoi, 2021

I would first like to thank my thesis supervisor Assoc Prof Vu Hoai Nam of the Department of Highway and Traffic engineering at National University of Civil Engineering The door to Assoc Prof Vu Hoai Nam office was always open whenever

I ran into a trouble spot or had a question about my research or writing He consistently allowed this paper to be my own work, but steered me in the right the direction whenever he thought I needed it

I would also like to acknowledge the experts as the second readers of this thesis: Prof

Dr Sci Nguyen Dinh Duc – Director of the Infrastructure Engineering program at Vietnam Japan University; Prof Hironori Kato – Co-Director of the Infrastructure Engineering program at Vietnam Japan University; Dr Phan Le Binh – Lecturer of the Infrastructure Engineering program at Vietnam Japan University; Prof Shinichi Takeda – Lecturer of the Infrastructure Engineering program at Vietnam Japan University; and other lecturers of the Infrastructure Engineering program at Vietnam Japan University Without their valuable comments, this research could not have been successfully conducted

Finally, I must express my very profound gratitude to my parents and to my boyfriend for providing me with unfailing support and continuous encouragement throughout my years of study and through the process of researching and writing this thesis This accomplishment would not have been possible without them

Thank you

Author

Vu Thi Huong Khue

CONTENTS

LIST OF TABLES

LIST OF FIGURES

LIST OF ABBREVIATIONS

CHAPTER 1 INTRODUCTION 1

1.1 Overview 1

1.2 Background 1

1.3 Problem statement 3

1.4 Goals and objectives 4

1.5 Study methodology 4

1.6 Thesis organization 4

1.7 Findings and Research contributions 5

CHAPTER 2 LITERATURE REVIEW 6

2.1 Vietnam road safety in the recent years 6

2.1.1 Road safety statistics 6

2.1.2 Current road safety ranking by Vietnamese Government 12

2.2 Literature Review on Road safety ranking methods 18

2.2.1 Analytic hierarchy process (AHP) method 18

2.2.2 Data Envelopment Analysis (DEA) method 19

2.2.3 Statistical method 20

2.2.4 Delphi method 21

2.2.5 Weighting method 22

2.2.6 Aggregation method 23

2.2.7 Several studies by Vietnamese researchers 24

2.3 Summary 25

CHAPTER 3 METHOD DEVELOPMENT 26

3.1 The requirements of the new method in the context of Vietnam 26

3.2 The selection of new road safety performance indicators for the model 27

3.2.1 The requirements of the proposal indicators 27

3.2.2 Type of indicators 27

3.2.3 The proposal indicators 29

3.3 Ranking Model Development 30

3.3.1 Research’s hypothesis 30

3.3.2 Ranking principles 31

3.4 The method procedure 31

CHAPTER 4 MODEL PERFORMANCE 34

4.1 Data collection 34

4.1.1 Traffic accident data source 34

4.1.2 Population data source 34

4.2 Data description 34

4.3 Data treatment 37

4.4 Model performance 39

4.5 Results score ranking 44

CHAPTER 5 MODEL EVALUATION 48

5.1 The DEA as the reference method 48

5.1.1 Reasons of choosing DEA model 48

5.1.2 DEA Methodology 48

5.1.3 Results DEA score ranking 49

5.2 Comparison and Discussion 52

CHAPTER 6 CONCLUSIONS AND RECOMMENDATIONS 55

6.1 Conclusions 55

6.2 Limitations 55

6.3 Recommendations and further studies 55

REFERENCES 56

APPENDIXES 59

LIST OF TABLES

Page Table 2.1: The percentage changing of road traffic accidents, deaths and injuries due to

RTAs from 2016 to 2018 10

Table 2.2: List of criticized provinces by the Deputy Prime Minister 12

Table 3.1: Road traffic accidents statistics characteristics 29

Table 3.2 Risk indicators characteristics 30

Table 4.1: Data category 34

Table 4.2: List of provinces stayed in low position in increasing RSIs or tend to reduce RSIs but still being in the Government's criticism list (2016-2018) 36

Table 4.3: List of provinces and cities having sudden increase/decrease in statistics from 2016 to 2018 37

Table 4.4: z-value transforming of 63 Vietnamese provinces from 2016 to 2018 (R1) 40 Table 4.5: z-value transforming of 63 Vietnamese provinces from 2016 to 2018 (R2) 41 Table 5.1: Input and output data for DEA method 48

Table 0.1: The denominators Nij of the data set in 2016 77

Table 0.2: The denominators Nij of the data set in 2017 79

Table 0.3: The denominators Nij of the data set in 2018 81

ii

LIST OF FIGURES

Pages Figure 2.1: Mean and standard deviation of Total number of road traffic accidents,

total number of deaths and injuries due to RTAs from 2016 to 2018 6

Figure 2.2: Twenty provinces and cities having the highest and the lowest total number of road traffic accidents, total number of deaths and injuries due to RTAs in 2016 7

Figure 2.3: Twenty provinces and cities having the highest and the lowest total number of road traffic accidents, total number of deaths and injuries due to RTAs in 2017 8

Figure 2.4: Twenty provinces and cities having the highest and the lowest total number of road traffic accidents, total number of deaths and injuries due to RTAs in 2018 9

Figure 2.5: Road safety factors trend in 63 provinces nationwide from 2016 to 2018 13 Figure 2.6: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents and population of 63 provinces and cities nationwide in 2018 14

Figure 2.7: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents and car registered of 63 provinces and cities nationwide in 2018 15

Figure 2.8: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents of 63 provinces 2016-2018 17

Figure 3.1: Normal distribution color and scale for road safety ranking principles 31

Figure 3.2: Research methodology flowchart 33

Figure 4.1: Distribution chart of RTAs, F and J from the smallest to the largest in 63 provinces nationwide (2016-2018) 35

Figure 4.2: Distribution chart of R1 in 63 provinces and cities nationwide (2016-2018) 38

Figure 4.3: Distribution chart of R2 in 63 provinces and cities nationwide (2016-2018) 39

Figure 4.4: z-values distribution of R1 on control chart 43

Figure 4.5: z-values distribution of R2 on control chart 43

Figure 4.6: z-score and rank of 63 provinces from 2016 to 2018 by using R1 44

Figure 4.7: z-score and rank of 63 provinces from 2016 to 2018 by using R2 45

Figure 4.8: Road safety distribution map in 63 provinces from 2016 to 2018 by ranking R1 46

Figure 4.9: Road safety distribution map in 63 provinces from 2016 to 2018 by ranking R2 47

Figure 5.1: DEA score and rank of 63 provinces from 2016 to 2018 50

Figure 5.2: Road safety distribution map in 63 provinces from 2016 to 2018 by using DEA method 51

LIST OF ABBREVIATIONS

WHO World Health Organization

M.O.T Ministry of Transport

M.P.S Ministry of Public Security

N.T.S.C National Traffic Safety Committee

AHP Analytic Hierarchy Process

DEA Data Envelopment Analysis

RTAs Road traffic accidents

RSIs Road safety indicators/indices

RS Road safety

F Fatalities/Deaths

J Injuries/Injured people

S.D Standard deviation

On the one hand, to assess the effort of each nation as well as each area in ensuring road safety, countries use road safety indicators as a ranking scale As the result, experts and specialists study and develop policies, indices to improve and ensure road safety [1]

Vietnam has been applying simple indicators to indicate the road safety performance

of the provinces and the cities However, these indicators have shown the evidence of drawbacks and need to be adjusted or replaced to achieve better evaluations on the national and regional road safety level

1.2 Background

Safety can be simply illustrated as the nonappearance of a hazard or vulnerability In the field of transportation, road safety can be defined by the human capacity in moving freely without injury or fatality An admirably harmless traffic network would not occur collisions within differing travelers Although non-accident is an excellent situation in theory and various transportation offices set a goal of zero fatalities on routes, people endure getting wounded or died on paths and roadways along the country The objection offer in the road safety course is to reduce the density of road collisions and the following fatalities and wounds by fully applying presently possible instruments, expertise, and technology [1]

In additions, road safety experts usually assess safety by the amount and proportion of traffic collisions and by the severity of those accidents, which represented by three parameters: crash frequency, crash rates and crash outcomes In the simplest terms,

crash frequency is the total road accidents appearing per year or another time unit Meanwhile, crash rates are the road collisions standardized by a specific group of people or metric of exposure For example, crashes per 100,000 people; crashes per miles traveled, etc Crash outcomes can be determined by the kinds of injuries suffered

to the crashes’ victims, ordinarily classified by the casualties and fatalities [1]

There are three factors influence the road traffic accidents and safety: exposure, accident rate and injury severity (Nilsson, 2002)

Exposure stands for the amount of movements in which accidents may take place The traffic volume frequently attributes the travelling turns, calling the number of people per kilometer of driving achieved Therefore, any individual activity is displayed the hazard of collision, in particular the road traffic accidents [2]

It is evident that people have many ways to circulate: walking, by bike, by car, by public transport, etc However, these ways absolutely not affect the similar accidents stage Therefore, the chosen exposure is a condition that affect to the number of deaths and injured people in road crashes Moreover, the danger threat to the road users is possibly depended on different means of traffic transport For instance, the lower the hazard of accident to walkers, the higher the rate of pedestrians in traffic participating [2]

Accident rate is the threat of accident per unit of exposure and is a statistical factor of traffic collisions Despite the fact that an accident rate is not similar to a probability estimation, it is an effective parameter because the accident probability can be considered to be proportional to the accident rate, in the theoretical meaning In fact, the greater the accident percentage, the greater the probability of an accident on a trip Occasionally, the definitions of “risk of accident”, “level of risk”, or “accident risk” will be synonymous with accident rate Moreover, the accident rate is not independent

on the exposure although it is defined per unit of exposure Admirably, an exposure must be characterized when it is separated with the accident rate Unfortunately, this argument is still in theory [2]

Additionally, the probability of traffic collision is influenced by a huge amount of risk

3

control devices, vehicles and road users An accident risk indicator is any factor raising the probability of traffic collisions In another word, although risk indicators are demographically linked to the probability of accidents, but not all of them can be observed as road collisions causes [2]

Injury severity hand over the accident’s consequence in terms of injured people or property damage In theory, the harshness of an accident outcome is a continuous variable extending from the tiniest obstacle to accidents with numerous deaths In reality, elementary scales, assuming just several disconnected values, are frequently created to illustrate collision or harm severity For example, administrator road collision data in almost nations categorize road traffic crash by severity in consonance with the simple scale below: fatal accident, accident resulting in serious injury, accident resulting in slight injury and property damage accidents However, this raw classification is not commensurate between nations and regions [2]

Vietnam is the country having a high rate of 23.4 fatalities per 100,000 capitals due to traffic accidents (WHO, 2018) In the recent years, Vietnam puts a lot of effort to struggling with the deduction of human loss in road safety The common approaches that the Vietnamese government has been applying are the 5Es namely Engineering, Enforcement, Education, Economics and Enablement One of the issues in term of Engineering is the accident statistics and the road safety ranking to serve for the monitoring the national road safety trend as well as the measurement of local authority

in road safety

Although road safety ranking is very important, but there are a number of issues that should be sold in Vietnam These are:

 The inconsistency of the current ranking method that has been used by the country

 The nạve road safety indicators

 The lacking of sound scientific and engineering bases to pointing out the blackspot road accident at the provincial level

Because of these reasons, the research on a new ranking method is necessarily, and this is also the aim of my study

1.4 Goals and objectives

The general objective of this research is to propose a new method capable of enhancing the accuracy of road safety ranking in Vietnam The specific objectives are:

 To propose new road safety indicators to replace the ones that have been using by the M.O.T

 To propose a new method that would help to classify the best and the worst provinces in term of road safety performance and achievements

1.5 Study methodology

The research particularly proposes a ranking method basing on statistical process control theory The road safety database in the period from 2016 to 2018 was collected and then processed and used as inputs for the developed model New principles of ranking based on the new road safety indicators have been applied to classify the road safety of each province of the country

1.6 Thesis organization

The dissertation includes four chapters excluding introduction and conclusion parts

 Chapter 1 analyses the road safety statistics and the current situation of road safety ranking in Vietnam The road safety ranking methods applied worldwide are also mentioned in the literature review

 Chapter 2 presents the model development including the section of new road safety indicator, the hypothesis, the ranking methodology, and the procedure to develop the model

 Chapter 3 shows the data collection and data treatments

5

 Chapter 4 proves the model’s ranking performance in comparison with another model which is considered to be a good ranking one Some discussions on the proposal model are also in this chapter

1.7 Findings and Research contributions

 Proven the limitations of the current ranking method and pointed out the needs of a new alternative method

 The proposed road safety indicators in this research partially take into account the local socio-economic characteristics

 The proposed method has capability of ranking the provincial road safety with a statistically significant confidence level

 The proposed method is simple, easy to apply and understand for the current statistics level among local officials

CHAPTER 2 LITERATURE REVIEW

2.1 Vietnam road safety in the recent years

2.1.1 Road safety statistics

Figure 2.1 presents road safety of Vietnam from 2016 to 2018 As shown in the figure, the number of accidents in the nation are slightly decreasing year by year

Figure 2.1: Mean and standard deviation of Total number of road traffic accidents,

total number of deaths and injuries due to RTAs from 2016 to 2018

(Source: Modified data from Traffic Police Department – Ministry of Public Security)

To have further pictures of the road safety in Vietnam in general and its provinces in particular, we processed the raw data and we show here 10 top highest and lowest provinces and cities in term of road traffic accidents As shown in the figures, HCM city, Ha Noi, Binh Duong were constantly the top 3 most dangerous cities in term of human losses Four other provinces (e.g Vung Tau, Dak Lak, Gia Lai, Binh Thuan) were also in this list three years consecutively In the lowest side, most of the names in the list are the remote regions and low-income provinces

7

Figure 2.2: Twenty provinces and cities having the highest and the lowest total number

of road traffic accidents, total number of deaths and injuries due to RTAs in 2016 (Source: Modified data from Traffic Police Department – Ministry of Public Security)

Figure 2.3: Twenty provinces and cities having the highest and the lowest total number

of road traffic accidents, total number of deaths and injuries due to RTAs in 2017

9

Figure 2.4: Twenty provinces and cities having the highest and the lowest total number

of road traffic accidents, total number of deaths and injuries due to RTAs in 2018 (Source: Modified data from Traffic Police Department – Ministry of Public Security)

In a closer look, Table 2.1 presents the provinces which had the number of fatalities (F), injuries (J) and road traffic accidents (RTAs) simultaneously increased during the time from 2016 to 2018

Table 2.1: The percentage changing of road traffic accidents, deaths and injuries due to

RTAs from 2016 to 2018

Provinces RTAs Deaths Injuries RTAs Deaths Injuries

An Giang 3.03% 8.51% -5.00% -25.32% -23.68% -46.34% Binh Đinh -16.33% -16.05% -15.98% -15.83% -8.00% -20.99% Binh Duong -26.48% -0.93% -10.62% -1.14% -4.18% -1.07% Binh Phuoc -1.68% -0.59% -13.58% -2.77% -1.80% -9.05% Binh Thuan -12.85% -6.28% -22.44% -18.57% 6.69% -31.05%

Ba Ria-V.Tau -8.24% -1.16% -12.40% -10.23% -9.32% -28.57% Bac Giang -6.86% -9.88% -7.34% 57.76% 67.60% 51.90% Bac Kan -33.33% -61.90% -12.00% 19.40% 27.59% 11.76% Bac Ninh -5.26% -2.30% -13.95% -4.59% -2.35% -10.26% Bac Lieu -23.08% -19.44% -44.74% -26.83% -33.33% -26.67% Ben Tre -5.86% 2.54% -26.32% -8.33% -1.03% -27.73%

Ca Mau -72.63% -19.23% -80.00% -38.76% -8.33% -55.91% Cao Bang -67.03% -20.93% -76.53% 2.15% 10.42% 2.00% Can Tho 30.98% 16.13% 41.18% -6.36% 16.96% -33.86% Dak Lak 6.10% 10.07% -3.62% -22.08% -5.51% -23.21%

Da Nang -19.64% -20.00% -25.00% -20.43% -20.69% -20.63% Dac Nong -10.40% -1.61% -7.41% -14.68% -6.90% -5.88% Dong Nai -12.30% -13.52% -10.97% -3.68% -8.91% -4.41% Dong Thap -20.24% -7.91% -25.22% -12.00% -6.92% -59.72% Dien Bien -2.17% -12.90% 0.00% -6.98% -19.23% -5.26% Gia Lai -6.34% -0.82% -8.46% -0.99% -0.41% -9.49%

Ha Giang -51.43% -22.73% -158.82% -45.83% -46.67% -30.77%

Ha Nam -6.02% -3.90% -7.06% -2.31% -2.67% -7.59%

Ha Noi -6.95% 0.00% -15.84% -6.75% -5.64% -22.84%

Ha Tinh -27.08% -14.93% -28.95% -2.13% -2.29% -5.56% Hai Duong -4.03% -15.34% 5.03% -4.64% 13.76% -19.55% Hai Phong -1.01% 3.33% -2.04% -2.06% -8.43% -22.50% Hau Giang 31.62% 26.76% 24.59% -7.34% 21.98% -12.96% Hung Yen -7.84% 1.75% -5.61% -2.00% -2.70% 0.00% Hoa Binh -0.93% -1.20% -2.35% -0.93% 2.35% -3.66% Khanh Hoa -6.33% 1.39% -80.77% -12.86% -1.41% -44.44% Kien Giang -15.49% -24.30% -44.59% -7.04% 12.30% -10.45% Kon Tum -2.78% -1.32% 0.00% -4.35% -22.58% 3.13%

11

Provinces RTAs Deaths Injuries RTAs Deaths Injuries Lao Cai -5.93% -14.04% -1.28% -53.41% -23.91% -60.82% Lang Son -16.22% -8.62% -82.93% -23.33% -9.43% 0.00% Lai Chau 20.27% 15.00% 31.52% -15.63% 18.37% -27.78% Long An -22.18% -7.58% -30.00% -22.77% -16.81% -24.26% Nam Dinh -6.52% 1.59% 0.00% -0.73% -5.00% -3.20% Nghe An -1.37% -0.53% -5.45% -1.04% -5.59% -2.33% Ninh Binh -1.66% -4.00% -2.84% -7.10% -13.64% -9.30% Ninh Thuan -29.62% -24.07% -28.18% -42.08% 6.90% -48.19% Phu Tho -4.17% -1.56% -4.04% -15.38% -3.23% -59.68% Phu Yen -13.90% -4.96% -33.70% -31.18% -34.44% -20.67% Quang Binh -6.49% -1.94% -21.99% -7.94% 0.96% -19.38% Quang Nam -3.31% -2.63% -4.29% -12.04% 2.56% -22.63% Quang Ngai -3.52% -3.50% -9.47% -51.65% -1.42% -65.32% Quang Ninh -17.09% -30.65% -19.42% 0.00% 6.06% -0.98% Quang Tri -7.88% 10.43% -24.58% -7.98% 5.74% -24.31% Son La -10.37% -24.66% 3.01% -8.87% -1.39% -27.88% Soc Trang -9.78% 10.28% -13.15% 0.00% -0.94% -10.36% Tay Ninh -1.80% 1.59% -3.82% 30.13% 34.38% 16.93% Thai Binh -14.37% -1.52% -12.68% -27.94% -1.54% -36.54% Thai Nguyen -4.57% -8.64% -4.00% -1.74% -5.19% -2.74% Thanh Hoa -4.57% -13.29% -2.63% -11.63% -6.04% -14.57% Thua Th.Hue -8.51% -5.47% -15.07% 2.29% 11.72% -5.04% Tien Giang -5.69% 7.81% -31.25% -15.76% -14.96% -17.07% H.C.M city -0.39% -13.19% -8.17% -6.73% -0.43% -17.64% Tra Vinh -23.85% -11.43% -38.32% -18.48% -2.94% -52.86% Tuyen Quang 35.92% 18.31% 50.00% -21.37% -36.54% -37.61% Vinh Long -15.77% -2.19% -41.37% -17.72% -7.87% -22.06% Vinh Phuc -2.27% -2.78% -48.89% -7.32% -9.09% -25.00% Yen Bai -3.83% 0.00% -2.03% -7.73% -6.38% -7.89%

(Source: Traffic Police Department – M.P.S) Every year, the Government criticizes the provinces for having bad record in road safety The criteria for the criticism are the high number of fatalities and the increasing

of total road accidents in comparison with that number of the last year For example, Table 2.2 shows the list of such the provinces from 2016 to 2018

Table 2.2: List of criticized provinces by the Deputy Prime Minister

No 2016-6/2017 2017-9/2018

(Increase > 15%) (Increase > 30%) (Increase > 20%)

1 Yen Bai An Giang Dak Nong

2 Hai Phong Quang Tri Hau Giang

3 Khanh Hoa Can Tho Hai Duong

4 Hau Giang Lai Chau Quang Nam

5 Gia Lai Lai Chau

6 Ha Nam Cao Bang

However, this assessment does not consider other relevant factors such as geographic, socioeconomic, or transportation characteristics of each province Therefore, this creates the following inadequacies:

Firstly, on the one hand, the provinces and cities having an enormous population, a high traffic transport density and a complex road network, are often documented in the list of provinces with poor road safety level For example, Hanoi and HCM city (Fig 2.6 and 2.7) have a large population and vehicles, so the number of road traffic accidents, deaths and injuries is evidently the highest among the 63 provinces nationwide, in addition

13 Figure 2.5: Road safety factors trend in 63 provinces nationwide from 2016 to 2018

(Source: Graphics are created by the author, Data from Traffic Police Department – Ministry of Public Security)

14

Figure 2.6: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents and population of 63 provinces

and cities nationwide in 2018 (Source: Graphics are created by the author, Data from Traffic Police Department – Ministry of Public Security)

15

Figure 2.7: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents and car registered of 63

provinces and cities nationwide in 2018 (Source: Graphics are created by the author, Data from Traffic Police Department – Ministry of Public Security)

On the other hand, when comparing Hanoi with some low populated regions whose population 4.3 times lower than Hanoi, but representing road traffic accidents per capita 1.3 times higher than that of Hanoi, in 2018 Therefore, is it really fair to rank Hanoi as a more “dangerous” city than Tien Giang in terms of road safety? Moreover, once this ranking process is continuously maintained, big and densely populated areas such as Hanoi, HCM city will always be in the “bad” group regardless of their efforts in ensuring local road safety In addition, this ranking method also gives a

“misunderstanding” picture of the road safety situation in small provinces with small population and creates the subjectivity in ensuring local road safety

Secondly, the current assessment indicators in Vietnam lack consistency in ranking Observing Figure 2.8 and comparing the road safety situation of two provinces: Bac Kan and Bac Ninh It emerges that Bac Kan is less safe than Bac Ninh because it shown higher number of road traffic accidents and deaths in 2018 In contrast, Bac Ninh is a more dangerous city as it shows higher RTAs

It should be noted that the simple using of three indicators F, J and RTAs in road safety assessment occurs these above inadequacies due to the indicators’ nạve characteristics and inappropriate application

Last but not least, the current method has not pointed out which province is a “black spot” in terms of road safety, as well as there is no robustness scientific bases to assess and rank road safety in Vietnam Therefore, the research questions are “Whether a region basing on single criteria such as F, J, or RTAs, is actually more “dangerous” at

a statistical significance level than that having lower statistics?” and “Which area is considered as "black spots" in road safety at the provincial level?

17

Figure 2.8: The distribution chart of road traffic accidents, deaths and injuries due to road traffic accidents of 63 provinces 2016-2018

(Source: Graphics are created by the author, Data from Traffic Police Department – Ministry of Public Security)

Provinces

2.2 Literature Review on Road safety ranking methods

2.2.1 Analytic hierarchy process (AHP) method

Analytic hierarchy process (AHP) is a mathematics and psychology multi-criteria decision-making method, developing by Thomas L Saaty (1975) In transportation field, this method is regularly applied to traffic safety assessment in several countries Cheng, C et al (2011) determined the weights of the elements provoking road traffic accidents (e.g drivers, vehicles, roads, and environment) in China by building an AHP model of 15 leading factors such as speeding, drunk and fatigue driving, poor performance vehicles, linear road and slope, traffic control, weather condition, etc They demonstrated that driver factor – Speeding explained a huge percentage (25.23%) causing RTAs, and precautionary projects could be established to avoid and minimize RTAs, according to the weight index of these 15 affecting factors [4]

Agarwal, P et al (2013) applied AHP method for ranking road safety hazardous locations (straight section, curve section and intersection) by adding the weight index

of safety factors, i.e surface, geometrical, road furniture, etc and their condition rating decimal grade from 0 to 1 The authors noted a ranking list of road safety hazardous locations, in which, the greater safety hazardous index, the more hazardous conditions Furthermore, the efficiency of the methodology was checked by comparing determined results from various road sections between Jaipur and Kiashangarh (India) [5]

Sordyl, J (2017) studied the significance of three elements influencing road safety (drives, vehicles, and environment) in Poland, by calculating their weight index through the Analytic hierarchy process method The researcher concluded that the most affecting factor on RS was all conducted in the drivers group (82% of priorities value), the environment factors group has moderate importance, and the last position was belonging to the vehicles elements [6]

Fernandez, J et al (2020) interviewed 535 drivers about their understanding of traffic signs in Manila (Philippines) and used AHP method to rank the priority of six road accident causes (lack of knowledge of traffic signs, bad driving behaviors of drivers,

19

physical and emotional state of drivers, lack of proper driving training, and distracted drivers) relating to the drivers traffic signs comprehensives They found that the majority of interviewers has above average awareness of traffic signs (37%) and the most considering cause of road accidents is the bad driving behaviors of drivers, in contrast [7]

Although AHP method is very popular in scientific researches, particularly in road safety, this procedure still has some drawbacks and limitation in studies In point of fact, there is not always a solution to the linear equations, and the calculated demand is overwhelming for a tiny question In addition, AHP method just approves triangular fuzzy figures to be accepted The rank switch phenomenon should be, moreover, examined attentively when implementing It absolutely determines the order replacement of the appraisal opportunities when a different appraisal opportunity is adjoined into the researched problem In another hand, this mathematic method depends on the probability and possibility standards and admits a subjective model which cannot ensure the accuracy of the methodology Last but not least, the bigger the model levels exist, the more complicate the model calculation act [8]

2.2.2 Data Envelopment Analysis (DEA) method

Data envelopment analysis (DEA) is a non-parametric method established by Charles

et al (1978) and enhanced by Banker et al (1984) This mathematical technique assesses the efficiency of Decision Making Units (DMUs)

Rosić, M et al (2017) used DEA and TOSIS method for composite road safety index selection by PROMETHEE-RS model with three applied parameters (average correlation, average age range variation and average cluster variation) in 27 Serbian police departments [9]

Nikolaou, P and Dimitriou, L (2018) assessed road safety policies in 23 European countries by applying DEA method and DEA-Cross efficiency model, with three groups data of economic, social and demographic from a decade [10]

Tešić, M et al (2018) also utilized DEA method to analyze the most meaningful indicators (alcohol level, speed, protective system, vehicles, roads and, trauma

management) of the total road safety performance index (RSPI) on an area (21 European countries) and presented the country ranking list based on RSPI [11]

Fancello, G et al (2020) studied road safety in the Italian civil road system by comparing the results from CCR and BCC models, with output data (social cost of accidents) and two input data (average number of conflict points at intersections and traffic flow), to determine hazardous road locations, which depended on safety status [12]

Antić, B et al (2020) administered DEA method for road safety performance benchmarking on 21 communities in Montenegro (decision making units – DMUs) in context with the scenario of the Rosić, M et al ‘s research in 2017 The output data (number of road traffic fatalities per 100,000 inhabitants and number of injuries per 100,000 inhabitants) and input data (number of traffic violations according to driving under influence of drug and alcohol, number of traffic violations according driving speed above the maximum speed limit, number of traffic violations according to not wearing the seatbelt, number of registered vehicles and, road density) were analyzed to rank the efficiency of the DMUs [13]

These researches applying DEA method contain some limitation inside The method success depends on the data quality Moreover, the study scope and the number of input data is still small and need to verify and update the timeline Furthermore, the DEA method neglects the impact of external variables on the procedure and neglects the numerical errors At last, this method does not concern the efficiency improvement process and is still lack of index testing after calculating [14]

2.2.3 Statistical method

Statistics is a section of applied mathematics including collection, description, analysis, etc This technique is used to assess the robustness of research data In the transportation field, the statistical method is also studied and applied in various countries

Leur, P de and Sayed, T (2002) raised a road safety risk index (RSRI) in North America for urban and rural road, based on overall length of the passing zone, sight

21

Shbeeb, L I and Awad, W H (2016) applied a questionnaire with two participants’ groups (general public and road specialists), then statistical procedure to analysis the road safety behavior in Jordan [16]

Smirnovs, J and Lāma, A (2019) argued the advancement tendency of road safety in Latvia, Russia, standing on number of vehicles, number of inhabitants, information about road network and traffic volume [17]

Bąk, I et al (2019) studied statistical method (the zero normalization method) for assessing road traffic in Polish voivode ship cities by analyzing the number of traffic crime, number of deaths due to road traffic accidents and the rate of crime determination to ranking the cities in Poland [18]

Ibrayev, K et al (2019) examined and determined the results from number of road traffic accidents quantity per 100 km, number of fatality per 100 km and number of injured people per 100 km, for studying road safety in Kazakhstan [19]

Wang, D et al (2019) researched the harshness of road traffic accident based on the census, which was total up and calculated by human damage (HD) and case fatality rate (CFR) from big provinces and cities in the People's Republic country [20]

Ayodeji, O.A et al (2020) used the Autoregressive Distributed Lag (ADL) to analyze the data of GDP, GDP per capital, population, nominal exchange rate and general level

of education in the country for examining the relationship between economic growth, motorization and road safety in Nigeria from 1970 to 2016 [21]

It can be seen that the statistical researches have a subjective nature in the studied process The experts and researchers cannot ensure the accuracy of their problems, and some studies ignore the human factor during researching

2.2.4 Delphi method

Delphi method is an estimating method first developed in the Cold war to forecast the effect of weapon machinery and extended by Project RAND and Olaf, H et al (1959) This process is built based on the specialists’ opinions in questionnaires This is an example of the Delphi method implementation in road safety field

Li, Y et al (2018) supported the map matching and the space-time division of road traffic and GPS data to assess the driving attitude safety for cabman in Shanghai The authors derived the driving behavior and collected the safety indicators (speed, number

of brakes, and the vehicles drift angle) by standardization, then applied the Delphi method to find out the indexes weight (4 level from low to high: A, B, C, and D) They concluded that their driving behavior safety assessment model could help cabman to observe the drivers’ safety index and the driving attitude safety map [22]

In this research, the authors do not discuss the limitations of their study However, Delphi method also has limitations in progress This method is fairly time-spending and assiduous for analysts and attendants, being a constraint of DEA According to Donohoe and Needham (2009), participants may abandon the progress because of the prolonged assurance or failure in the method Furthermore, the Delphi method has bias into the results, and the researchers’ ideas may be less ownership due to the participants’ incognito [23]

2.2.5 Weighting method

Weighting method is an alteration procedure that is accepted by survey analyst This technique assigns statistical correction that are made to collected survey input to advance the accuracy of the survey assessments The weighting method is also applied

in the road safety field

Pešić, D et al (2013) supposed an advanced method – Benchmarked traffic safety level (BTSL) for traffic safety assessment in a Serbian area This method selects various indicators (annual number of fatalities per 108 vehicles kilometers, per 104

registered vehicles and, per 100,000 inhabitants and; percentage of drivers non-driving under alcohol consummation and, non-speeding; percentage of drivers and passengers using seatbelts and; traffic safety indicators relating to traffic safety level rate of an area), converts the indicators’ values, then appoint their weights and sum up the traffic safety level (TSL) [24]

Yang, C et al (2016) enhanced the Grey correlation combination the weighting method to evaluate the index system of traffic safety for highways (road alignment,

It is evident that the weighting method is based on the data set and demand the equivalence between the results values This method is also fairly hard to administer due to the accuracy of the input indicators [27]

Castro-Nuño, M and Arévalo-Quijada, M T (2018) applied the Promethee-GAIA method, description statistic and weighting method for determining civil road safety depending on economics, demographics and sustainable civic transportation criteria, and ranking the 50 Spanish territory ‘s road safety to decrease traffic accidents and deaths in the city [29]

Fancello, G et al (2019) compared the results between threshold choice (Electre III) and ranking method (Concordance analysis) to establish road crossing in Villacidro, Italy for enhancing road safety by Vikor and Topsis methods, based on 8 benchmarks (sight distance, road signs and markings, intersection lighting, road surface

maintenance, density of traffic conflict points, number of vehicles in the intersections, percentage of vehicles weight, and pedestrian flow) [30]

Chen, F., Lyu, J and Wang, T (2020) assessed road safety improvement in the 36 Organization for Economic Co-operation and Development (OECD) countries based

on six safety performance indicators (SPIs), including output (personal risk, traffic risk, and change trend) and input dimension (vehicle safety, road situation, road user behavior, socioeconomic situation, and enforcement) from 2009 to 2018 by using inverse variance weighting, VIKOR method, Fisher’s natural breaks classification (FNBC), and IV-VIKOR with FNBC method [31]

2.2.7 Several studies by Vietnamese researchers

Viet Nam has researched the way that worldwide nations assure transportation control safety and security, and the developing country has studied some researches by applying international references and method, such as questionnaire, standardized method, and hierarchical analysis (AHP) method

Ngo, A.D et al (2012), based on statistics on people, seasons and causes of death due

to traffic accidents from the survey questionnaire, implemented the standardized method to determine the death rate by age because of traffic accidents in 192 communes, 16 provinces in Vietnam [32]

Trung Bui, H et al (2020) conducted an investigation into road crash risks and traffic violations in Vietnam, by using questionnaires of motorcycle rider behavior questionnaire (MRBQ), then highlighted some important differences between car drivers from Vietnam and other countries [33]

Ngoc, A and My Thanh, T (2020), through collecting and analyzing the views of road users in Vietnam on road safety infrastructure, traffic landscape, traffic safety management system and traffic safety education, proposed a hierarchical analysis (AHP) method to assign weights to strategic solutions that experts give appropriate to each opinion in the study [34]

25

As it can be seen that there are not many scientific studies on road safety in Vietnam and these researches have not been applied in practice to overcome and solve the current traffic problems in Vietnam

2.3 Summary

In these studies, analysis about road safety tended to consider the indicators and factors that influence on the traffic safety or cause the traffic accidents by applying numerous mathematic methods In general, these methods are although used to rank road safety, they are mainly employed to assess and rank specific effects for a particular type of accident Moreover, these methods have not been applied to the road safety evaluation and classification as the research objectives because they are still heavily influenced by the limitations of budgets, time and level of implementation in Vietnam

CHAPTER 3 METHOD DEVELOPMENT

This chapter describes the indicators selection, the statistical bases and the procedure

of the method development

3.1 The requirements of the new method in the context of Vietnam

In addition to the lack of scientific research on road safety in Vietnam as described in Chapter 2, the statistics and analysis of data related to road safety in Vietnam still have many limitations and shortcomings

 First of all, the collected data on the road safety situation of regions, ministries and branches are not really close to the reality There is even an opinion that in some areas, the road traffic accident data are “hiding” for fear of being criticized and reprimanded in provinces and cities;

 Secondly, road traffic accident statistics from the Ministry of Public Security (M.P.S) often differ quite far from those from other ministries and branches (e.g Ministry of Health (M.O.H), insurance agencies, etc.) For example, the nationwide statistics of road traffic accidents in 2018 from the M.P.S reached more than 8,000 deaths, meanwhile, the M.O.H gave a larger number of 15,000 cases In addition, due to many reasons such as regulations of each ministry or sector; statistical methods; post-accident procedures; etc., that lead to differences and inconsistencies

in statistics on road traffic accidents in Vietnam and;

 Third, the current method for road safety assessment exists a drawback of the inconsistency in the ranking procedure as mentioned in chapter 2 Specifically, if ranking performance by F, province A is worse than province B, but if ranking by J

or (J+F), it gives an opposite result So, the questions are “How do we rank?” and

“Which province will get a “higher” safety rating?” Therefore, selecting new road safety indicators is a solution to ensure the consistency in road safety ranking

On the other hand, assessing road safety by simple indicators (F, J or RTAs) and without considering economic, social, geographical, or traffic factors, etc., will result

in incorrect ratings For instance, as discussed in chapter 2, it cannot be concluded that the big provinces and cities having a lot of road traffic accidents and related damages (which are proportional with the density of vehicles and population), are weak on road

27

safety So, the second research question is “The simple use of the deaths (F) or injuries (J) number per year to compare and evaluate road safety in Vietnam, without considering relevant factors (e.g., population, number of road kilometers x number of vehicles, or local socio-economic characteristics), is reasonable or not?”

Last but not least, it has not been pointed out that “Which method "surely" has the statistical significance of localities that are actually "black spots" at the provincial level?” in Vietnam at the present

Therefore, this study aims to propose a method in national conditions to overcome the aforementioned disadvantages

3.2 The selection of new road safety performance indicators for the model

3.2.1 The requirements of the proposal indicators

The new indicators should show at least one of the three information groups as follows:

 Accident shape provides the number of accidents and the type of accidents that occur based on the accident’s frequency and intensity [35]

 Accident affiliation provides the number of people and vehicles involved in the accident, usually based on the proportion of population and the type of vehicle [35]

 Degree of accident allows to obtain the severity of the accident, mainly the extent of human damage such as death rate, injury rate or property damage [35]

3.2.2 Type of indicators

There are a number of road safety statistics indices that should be used as the indicators for the road safety performance ranking at the provincial and national levels They are:

 Accident frequency is the number of accidents recorded at a spot, a road section or

an area in a time unit (cases/year) [35]

 Accident rate determines the number of traffic accidents on the total number of vehicles passing a point on a road segment (vehicle.km), or the total number of vehicles passing an intersection in a time unit (case/vehicle.km or case/vehicle) [35]

 Severity of accident ratio is traffic accident statistics based on the ratio of deaths, injuries, and accident affiliation per a unit of comparison The unit of comparison can be 10,000 people, 10,000 registered vehicles, over 10,000 km of roads or over

100 million vehicles.km [35]

 Severity index is determined by the number of deaths by total number of accidents occurring in a time unit [35]

 Property damage equivalence index (PDE)

This is the type of statistic commonly used in traffic accident statistics In which, traffic accidents are classified into three groups: Group F (fatalities) with at least one death; group I (injuries) divided into three subgroups (Class A: serious injury having long-term treatment in the hospital, class B: milder injuries having few days of hospitalization (less than 20 days), and class C: local treatment); and group P (property) only has property damage PDE is often applied in rating the level of danger

or safety in the traffic safety assessment and determined by the formula: [35]

where, F = the number of fatalities,

A = the number of serious injuries,

B = the number of moderate injuries,

C = the number of small injuries,

P = the damaged property,

K1, K2, K3 are weights Some areas take K1 = 5; K2 = 3; and K3 = 1 However, this is the reference values and these weights depend on the research in each country

T = a unit index which can be calculated by vehicle.km; by the total number of accidents; or by the unit of area, population, or registered vehicle

29

Table 3.1: Road traffic accidents statistics characteristics

No RTAs statistics Advantages Disadvantages

1 Accident frequency  The simplest form of

statistics for comparing and ranking the location’s traffic hazard level

 Only show the accident shape

 Do not reflect the traffic safety problems

2 Accident ratio  Better than accident

frequency

 Do not reflect the traffic safety problems

3 PDE index  Be often applied in rating

the level of danger or safety in the traffic safety assessment

 The weight K depend

on the research in each country

3.2.3 The proposal indicators

In this research, the author suggests the use of R1, R2, R3, S1 or S2 as follows

1

F JR

vehicle km



 (3.3) where, R1, R2, R3, S1, S2 are risk indicators;

RTAs is number of road traffic accidents;

F is number of deaths due to RTAs;

J is number of injuries due to RTAs;

PDE is property damage equivalence index

The reasons why this study select R1 and R2 are:

 R1 and R2 take into account one of the specific features of a region (i.e population)

 Existing data set can be used to calculate R1 and R2 and no extra cost is required

Table 3.2 Risk indicators characteristics Indicators Advantages Disadvantages

 Different road network distribution in each province, which not guaranteed in terms

of traffic volume and population

 Complicated calculation if separate road network

 It is difficult to find data to calculate PDE in Vietnam 3.3 Ranking Model Development

3.3.1 Research’s hypothesis

Traffic collisions and its consequences at a location are widely to be shaped as a Normal distribution The characteristics of normal curves are well-known and usually applied in transportation engineering In addition, the normal distribution is a bell-shaped curve with a mean of crashes and a standard deviation equal to the square root

of variance [36]

Any distribution is applied to represent either a spot should be considered as prone; a comprehension of the accurate application of statistical significance is essential for properly clarifying analytical outcomes The regular statistical method believed that the number of road accidents at an area followed a normal distribution [36]

collision-In this research, the following hypotheses are made:

 R1 and R2 to be considered as a normal distribution with the sample size N = 63 (e.g the number of provinces of Vietnam)

 If a province has R1 or R2 greater than twice of S.D from the mean expectation, it can be considered as “black spot” province at 95% statistical significant level

 Provinces having zR1, R2 located in the interval equal and greater than (μ + 2σ), are considered as “black spots” with serious traffic accident problems and weak on road safety

 Provinces having zR1, R2 stayed in the interval of [μ + 1σ, μ + 2σ] (high alert) or that

of [μ, μ + 1σ] (moderate alert), are subject to caution on road safety

 Provinces having zR1, R2 located in the interval less than μ, are considered as an area ensuring road safety

Figure 3.1: Normal distribution color and scale for road safety ranking principles 3.4 The method procedure

The study uses the statistical method combining with the normal distribution method and the standard normal distribution to analyze the data [37], following the steps as follows:

 Step 1: