STUDY ON RISKY BEHAVIOUR, RISK PERCEPTION AND POSITIVE AFFECT OF
MOTORCYCLISTS IN MALAYSIA
RICHARD CHEAH JUN XIAN
SCHOOL OF CIVIL ENGINEERING
UNIVERSITI SAINS MALAYSIA
STUDY ON RISKY BEHAVIOUR, RISK PERCEPTION AND POSITIVE AFFECT OF MOTORCYCLIST IN MALAYSIA
By
RICHARD CHEAH JUN XIAN
This dissertation is submitted to UNIVERSITI SAINS MALAYSIA
As partial fulfilment of requirement for the degree of
BACHELOR OF ENGINEERING (HONS) (CIVIL ENGINEERING)
School of Civil Engineering Universiti Sains Malaysia
June 2019
ACKNOWLEDGEMENT
I would like to express my sincere gratitude to one and all, who directly or indirectly, helped me during my journey of this research study. First and foremost, I would like to express my sincere gratitude to my supervisor, Assoc. Prof. Ir. Dr.
Leong Lee Vien for her guidance, continuous encouragement, invaluable support and advises throughout this study.
I would also like to express my gratitude to Prof. Ahmad Shukri Yahya for his guidance in some of the statistical analysis work. I would also like to thank my fellow course mates for their assistance in the work of data survey.
Last but not least, I would like to express my utmost gratitude to my family for their unconditionally supports, patience and encouragements as none of this could have happened without them.
ABSTRAK
Kajian ini dijalankan untuk menentukan kepentingan persepsi risiko dan keminatan terhadap tingkah laku menunggang motosikal yang berisiko dan hubungan antara ketiga-tiga komponen ini. Survei soal selidik telah dijalankan terhadap penunggang motosikal untuk mengumpul maklumat yang berkenaan seperti umur, jantina, pengalaman memandu dan tanggapan risiko, keminatan dan tingkah laku menunggang berisiko. Sistem skala 5-titik telah digunakan dalam soal selidik tersebut. Kajian ini mendapati bahawa keminatan penunggang motosikal mempengaruhi lebih banyak daripada persepsi risiko terhadap tingkah laku menunggang berisiko berdasarkan keputusan yang diperolehi daripada analisis Model Persamaan Struktur (SEM). Ini menunjukkan bahawa proses intuitif adalah proses utama bagi penunggang motosikal di Malaysia semasa mereka menunggang di jalan raya. Di samping itu, umur penunggang motosikal memberi kesan ketara terhadap tingkah laku berisiko dengan anggaran pekali -0.037. Ini menunjukkan bahawa penunggang motosikal yang lebih tua jarang melakukan tingkah laku berisiko semasa menunggang di jalan raya. Selain itu, pengalaman dalam memandu (tahun mendapat lesen) didapati ada hubungan positif dengan persepsi risiko pengawal motif dengan anggaran pekali 0.012 dan hubungan negatif dengan kesan positif dengan anggaran pekali -0.032. Ini menunjukkan bahawa penunggang motosikal yang baru sahaja mendapat lesen penunggang motosikal lebih cenderung untuk berfikir bahawa kelakuan risiko yang dinyatakan tersebut adalah tidak begitu risiko dan mereka lebih suka untuk bertindak tingkah laku berisiko tersebut ketika menunggang motosikal.
ABSTRACT
This study was done to determine the significance of risk perception and positive affect towards motorcyclists’ risky riding behaviour and the relationship among these three components. A questionnaire survey had been conducted to collect the related information such as age, gender, driving experience and self-reported risk perception, positive affect, risky riding behaviour from the motorcyclists. A 5-point scale system was used for respondents to answer the self-reported questions. The study found that positive affect influences more than risk perception on risky riding behaviour based on the results obtained from Structural Equation Modeling (SEM) analysis. This shows that intuitive processes are the major process for the motorcyclists in Malaysia rather than rational processes while they are riding on the road. Moreover, the age of motorcyclists significantly affects the risky behaviour with an estimate coefficient of -0.037 which showed that elder motorcyclists seldom perform risky behaviour while riding on the road. Besides, the driving experience (years of obtaining license) obtained positive relationship on the motorcyclists’ risk perception with an estimate coeffiicient of 0.012 and negative relationship with positive affect with estimate coefficient of -0.032. This showed that motorcyclists that just newly obtained the motorcyclist license would tend to think that those stated behaviour were lower risk in overall and more likely to behave the particular riding risky behaviour.
TABLE OF CONTENTS
ACKNOWLEDGEMENT………….………II ABSTRAK………...……….………III ABSTRACT……….………...……….…IV TABLE OF CONTENTS.………..………….V LIST OF FIGURES…….………..……….VII LIST OF TABLES……..……….……..…....VIII CHAPTER 1 INTRODUCTION...1- 4
1.1 Background... 1
1.2 Problem Statement... 2
1.3 Objectives...3
1.4 Scope of Study...3
CHAPTER 2 LITERATURE REVIEW...5-16 2.1 Overview... 5
2.2 Dual-process models of decisions making... 5
2.3 Risk Perception...6
2.4 Positive Affect...7
2.5 Risky Behaviour...8
2.6 Age... 10
2.7 Driving Experience...11
2.8 Cronbach’s Alpha...12
2.9 Principal Component Analysis (PCA)... 13
2.10 Structural Equation Modeling (SEM)... 13
2.11 Summary... 16 CHAPTER 3 METHODOLOGY...17-25
3.1 Overview... 17
3.2 Questionnaire………19
3.3 Data Analysis………..21
3.3.1 Reliability Test (Cronbach’s Alpha)………...22
3.3.2 Principal Component Analysis (PCA)………...…22
3.3.3 Structural Equation Modeling (SEM)………...…….23
3.4 Summary………...………25
CHAPTER 4 RESULT AND DISCUSSION...26-48 4.1 Introduction... 26
4.2 Descriptive Statistics………..………..27
4.3 Preliminary Study………...………..…29
4.4 Reliability Test (Cronbach’s Alpha)……….………..………37
4.5 Principal Component Analysis (PCA)…….……..………37
4.6 Structural Equation Modeling (SEM)………..………..41
4.7 Summary………..….………48
CHAPTER 5 CONCLUSION AND RECOMMENDATION...50-52 5.1 Conclusion...50
5.2 Recommendations………..………51 REFERENCES...53-57 APPENDIX A
LIST OF FIGURES
Figure 3.1 Flowchart of the methodology of the study………18
Figure 3.2 Relationship Model of the Study……….…….…24
Figure 4.1 Percentage of Age Group of Respondent for Male and Female…………..27
Figure 4.2 Percentage of Driving Experience of Respondents (Number of Years of Obtaining License)………28
Figure 4.3 Percentage of Type of Motorcycles Obtained by the Respondents..……28
Figure 4.4 Self-reported Risky Perception of Motorcyclists in Malaysia...30
Figure 4.5 Self-reported Positive Affect of Motorcyclists in Malaysia... 31
Figure 4.6 Self-reported Risky Behaviours of Motorcyclists in Malaysia...32
Figure 4.7 Mean of Self-reported Risky Perception...33
Figure 4.8 Mean of Self-reported Positive Affect...34
Figure 4.9 Mean of Self-reported Risky Behaviour...35
Figure 4.10 Structural Equation Modeling (SEM)……...………...42
LIST OF TABLES
Table 2.1 Structural Equation Modelling Evaluations: Rules of Thumb...15
Table 2.2. Goodness-of-fit for structural equation model...15
Table 3.1 Fifteen Behaviour Measures...20
Table 4.1 Mean and Standard Deviation of Each Observed Variable...36
Table 4.2 Reliability Test (Cronbach's Alpha) of the Self-reported Variables...37
Table 4.3 Variables and Reliable Analysis of Latent and Observed Variable...39
Table 4.4 Variables and Reliable Analysis of Latent and Observed Variable (cont.)...40
Table 4.5 Goodness-of-Fit measures of the developed model...43
Table 4.6 Standardised Coeffiecient Weight of the Motorcylists Model...45
CHAPTER 1 INTRODUCTION
1.1 Background
It is a well known fact that the road safety is currently a public health issue as traffic accidents are the top eight leading factors of death worldwide, (Loo et al., 2005).
In South East Asia countries, most of the countries have very high volume of motorcycles on the road which are facing high motorcyclist fatalities and this includes Malaysia.
From a road safety perspective, motorcyclist is the category of people that are importantly concerned (Elliott, 2010). The number of traffic fatality involving motorcyclists are still very high although few efforts had been taken to decrease the number of accidents occurrence (Karim et al., 2003). In most of the developing and developed countries, road accident that involve motorcycle is still a problem that yet to be solved or reduced. A previous studies found that driving a passenger car is 17 times safer than riding a motorcycle (Radin Umar et al.,1995). That is mainly because huge number of vehicles on road may lead to the high rates of accidents that involve motorcyclists.
In Malaysia, motorcycles accidents is one of the highest mode of transportation that involve in road fatality. More than 50% of road accident involve motorcyclist (Marizwan and Andras 2012). Traffic safety of motorcyclists is always an important and inevitable issue. In most of the urban areas, mixed traffic (traffic that consists of motorised and non-motorised vehicles) is very common to be seen. The traffic flow of motorcycle is strongly affected by the driver characteristics, external environment as
affected by driver characteristics and therefore, it is very important to examine the impacts of motorcyclists’ personal riding characteristics.
Moreover, those accidents may be due to the risk behaviours of motorcyclists.
In the previous study conducted in Denspasar, which had also used this model in the study of risky riding behaviour of the motorcyclists. The risky behaviour of motorcyclists were measured by referring to positive affect and risky perception and the relationship among the three variables were being discussed (Wedegama 2011).
In the past, dual-process models had been conducted to health risk behaviour such as smoking and consumption of alcohol (Moss and Albery, 2009), but this dual-process model is seldom used in traffic-related studies. Therefore, it is interesting to understand the relationship or influence of risk perception and positive affect upon risky riding behaviour of the motorcyclists in Malaysia.
The objectives of this study is to find out the relationship of the positive affect and risk perception of the motorcyclists towards the risky riding behaviour and also the significance of the impact of each observed variable from risk perception and positive affect towards the behaviour. Questionnaire survey is used to collect the data of the motorcyclists. The collected data is then analysed by using Principal Component Analysis (PCA) and the further interpretation and prediction are showed by developing a model by using Structural Equation Modeling (SEM).
1.2 Problem Statement
Motorcycle is one of the highest selling modes of transportation among the road users in Malaysia. However, most of the road accident fatalities involved motorcyclists. Malaysia has the highest road fatality risk (per 100,000 population)
among the ASEAN countries and more than 50% of the road accident fatalities involve motorcyclist (Marizwan and Andras 2012).
According to a study by Zahid Sultan, et al. (2016), the highest factor that contributes to the motorcycle crashes in Malaysia is the human behavior factor. This previous research showed that the change of road user behaviour is crucial to reduce the road accidents.
Most of the previous studies were focusing on the traffic violation and motorcyclists characteristic while motorcyclists’ perspective and their personal preference upon risky riding behaviour were seldom being discussed. Therefore, a better or full comprehension on motorcyclists risky behaviour, positive affect and their risk perception upon riding behaviour is needed for a better improvement in road safety.
1.3 Objectives
The objectives of this study are:
To identify the general pattern of risky behaviour, risk perception and positive affect of motorcyclists.
To determine the significant observed variables of risky behaviour, risk perception and positive affect of motorcyclists.
To determine the relationship of risk perception and positive affect towards the risky riding behaviour of motorcyclists
1.4 Scope of Study
The study is carried out through questionnaire survey via Face-to-Face Method and On-line Method. A number of respondents who obtained B2 Driving License are targeted. In this study, every respondent is required to answer 3 sections of questions
where 15 questions each. In the first section, respondents were required to indicate on a 5-point scale, to tell how frequently they engaged in the risky behaviour when riding, from “Never” to “Always. Next, the respondents indicated their perception of how risky the stated behaviors was to engage on a 5-point scale from “Not Risky at All” to
“Extremely Risky”. Lastly, the respondents indicated their positive affect towards the stated behaviors on a 5-point scale from “Extremely Dislike” to “Like It Very Much”.
Significance of the factors of each latent variable, namely risky perception, positive affect and risky riding behaviour are analysed by using Principal Component Analysis (PCA) via the statistical software of IBM SPSS.
A path model, Structural Equation Modeling (SEM) is constructed and analysed to identify the relationship among the risky behaviour, risky perception and positive affect based on the estimated values obtained.
CHAPTER 2
LITERATURE REVIEW
2.1 Overview
Many studies have found out that drivers’ risk behavior would influence the possibility of motorcycle crash accidents. Researches on the risk behaviors have been extensively conducted.
In this chapter, literature reviews on previous that are study similar to this study are carry out to understand more about the objectives of this study. In Section 2.2, a dual-process models of decision making are reviewed. Based on the research by Rhodes and Pivik (2011), the decision making of motorists on the road can be considered as a dual process which consists of rational and intuitive processes. In Section 2.3 to 2.5, risk perception, positive affect and risky riding behaviour of motorcyclists are discussed.
Recent literature on the influences of age and driving experience are review in Section 2.6 and 2.7. Type of analysis such as Cronbach’s Alpha, Principal Component Analysis (PCA) and Structural Equation Modeling (SEM) are reviewed in Section 2.8 to 2.10 as well.
2.2 Dual-process models of decisions making
Dual-process models of decisions making was mainly the implication of behavioral decision making work (Chaiken and Trope, 1999). There are two information processing systems in this particular model. One of the systems is referring to the logical process in our mind to analyse the information. This is. prone to rational and analytic which the decisions made are generally based on logical analysis of the information and this could be described as rational process. On the other hands, the
another system is intuitive and experiential. This particular process of decision making is based on affect or personal sensation. Both process influence the risky behaviour of a motorcyclist while riding on the road. Study on behaviour such as alcohol taking had been applied with the dual-process model (Moss and Albery, 2009). Moreover, this particular model had also been commonly applied in the field of traffic safety.
2.3 Risk Perception
Theoretically, risk perception described as the perspective upon the overall risk of a particular situation. Theoretical models that have been commonly applied to decision making about health risk behaviour, such as the health belief model (Rosenstock, 1974) and the theories of reasoned action and planned behaviour which assume that decision making about risk behaviour occurs through rational processes such as evaluating the risk and benefits of a given action (Ajzen, 1991). If the decision making was purely a rational process, interventions which inform the motorcyclists of the risk inherent in riding, such as driver’s education level, should be sufficient to encourage safer motorcycle riding.
Risks and hazards perception are getting important in many studies as the researchers are realising that motorcyclists are more likely to be involved in road fatalities than other road users. The motorcyclists tend to concern more on their perspective and perception from experience of their routine and less focused on the psychological attribute (Weissenfeld et al., 2014).
Moreover, research has failed to find a significant improvement in crash rates as a function of traditional driver’s education program (Senserrick et al., 2009).
Therefore, it is very important to move beyond models of rational decision making in understanding the risky of riding motorcycle on the roads.
By referring to the dual-process of decision making model, risk perception can be considered in rational process category system which the good or bad of a certain action or information is concerned in the evaluation. Understanding the driving risks and law enforcement policies are important to reduce the risk behaviour of motorcyclist such as speeding (Wedagama 2017).
2.4 Positive Affect
Positive affect of certain riding behaviors could be described as the enjoyment of doing or loving of risk riding such as the higher tendency of running a red light while there is no vehicles crossing the intersection (Rhodes and Pivik 2011). Driving had already been characterised by emotions together with the pleasures of doing (Sheller, 2004). Those emotions usually influenced by the automobiles marketing advertisements which portray the joy of driving.
Previous study had found out that positive emotional functions of riding a motorcycle such as having fun with friends or passengers are related to risky driving (Møller and Gregersen, 2008). However, there are many investigations of dual-process models which focus negative affect, especially like anxiety and fear. As riding had already a common transportation of human being in the society today, a focus on the importance of positive affect would come out a complete understanding of risky riding behaviour.
Positive affect is also important in gender differences in driving behaviour. A previous study shows that young men are usually more interested in riding and enjoy riding than young women (Harre et al.,1996). For the fast-paced decision making situation such as driving or riding a vehicle, experiential or intuitive decision making
seems to be the dominant, and therefore positive affect would play an important role in the outcome of riding decisions (riding behaviour).
Thus, investigating the influence of positive affect on risk driving behaviour may provide a full comprehend view on this. Therefore, the purpose of this study is to determine the relationship between the risky perception and positive affect towards the risky riding behaviour and how both systems influence the risky behaviour of motorcyclists. Hence, in the study of Rhodes and Pivik (2011), a list of hypothesis were required to test in the study.
Hypothesis 1. - Male drivers report greater positive affect for risky driving behavior than female drivers.
Hypothesis 2. - Positive affect and perceived risk mediate the relationship between gender and risky driving behavior.
Hypothesis 3. - Teen drivers report greater positive affect for risky driving behavior than adult drivers.
Hypothesis 4. - Positive affect and perceived risk mediate the relationship between age and risky driving behavior.
than risky perception
2.5 Risky Behaviour
In the research conducted by Allen and Brown (2008), they stated that the motorised vehicle is more than just a mode of transportation for teenager. As driving was previously unavailable to them as a child or adolescents, it provides a sense of social status achievement which is driving license, and also a status of freedom from parental supervision. Peer influences is usually a lesser awareness related factor that adults would concern about on teen driving. Allen and Brown (2008) agreed that
adolescents’ driving behaviour are depending on who is the passenger in the vehicle with them. For instance, teenagers tends to drive faster and perform high-risk driving or riding behaviour especially when the peers that they are carrying are young men. On the other hand, teenagers would tend to drive or ride safely when they are carrying adults as passengers.
The effectiveness of media coverage the severity of road accidents in deterring risk taking was being investigated by Job, (1990) which suggested that media coverage may instill the sense of overconfidence in young drivers, who would tend to believe that they have a superb riding skill because they may have not been involved in an vehicle accident. This may affect the young motorcyclist to take greater risk while riding on road which might lead to injuries and possible fatality.
Driving behaviour is considered as the connecting chain that links the human to different consequences. A few of the behaviors might just be considered as disrespectful behaviour such as not to comply with parking disciplines (Saba and Vafa, 2014).
However, dangerous driving or risky driving behaviour are significantly endangering or at least having the possibility to put the driver or other road users in danger situation (Dula and Geller 2004). To be exact, some patterns of driving behaviour like improper passing and lane usage, speeding, tailgating, illegal turns, right-of-way violations and control signal violations that place drivers at risk for morbidity and mortality are considered as risky driving behaviour (Shams and Rahimi-Movaghar, 2009). These behaviors can be either intentional or non-deliberate.
Human behaviour had been found to be one of the significant factors that could influence the motorcyclist accidents. These factors are generally indicated in motorcyclists behaviour such as traffic violation (Joewono and Susilo, 2017).
Disobeying the traffic rules and regulations are considered as intentional actions. This is
resulted from social factors and would increase the risk of traffic accidents (Steg and Brussel, 2009).
Moreover, against traffic rules or speeding were counted as one of the typical motorcyclists’ risky behaviour (Zamani-Alavijeh et al., 2009). Actually, motorcyclists’
actual riding behaviour can be estimated by analysing their perception and intentions (Tunnicliff, 2006).
In Thailand, the impact of alcohol is proven to be the most outstanding factor affecting the severity of motorcycles accidents (Kasantikul et al., 2015). In Denpasar, drunk riding is the most risky perceived, the least frequent behavior conducted by motorcyclists. (Wedagama, 2015). In fact, riders that have alcohol before riding that caused single-vehicle crashes are more likely to result fatal or severe injury (Shaheed and Gkritza, 2014).
2.6 Age
From the previous study, it is known that adolescents are more likely than adults to perform risky behaviour while driving or riding on the road (Mohd et al., 2012). For instance, adolescents are more likely than adults to drive recklessly, to drive while intoxicated, to use illegal substances, to have unprotected sex, and to engage in both minor and major anti-social behaviours (Arnett, 1992). The tendency for younger age drivers to do risky riding activities has been well stated in the previous studies as well (Evans and Wasielewski, 1983; Jonah, 1986, 1990). For an instance, high-risk driving behaviour in younger drivers have always been identified as a major factor in this age group’s basic motivation in not using seat-belts, which is one of the reasons that their fatal crash rates are higher than those of older age groups’ (Chliaoutakis et al., 2000;
Jonah, 1986; Mayhew and Simpson, 1999; Williams and Shabanova, 2002).
There are many evidences which have proven that risky riding behaviour is the most important major reason that causes the high vehicle accident rates among the teenagers (Finn and Bragg, 1986; Jonah, 1986). However, teen crashes are not only because of their lower driving experience (Senserrick, 2006), risk-taking behaviour are also the factors that causes road accidents (Rhodes et al., 2005). In fact, gender had also been determined as one of the factors that influences the risky behaviour (Oltedal and Rundmo, 2006). Speeding is more likely to be the male fatal accident factor by comparing to female rider of the same age (Harre et al., 1996).
Previous study has found out that most of the motorcyclists were younger in age, less than 3 years of driving experience and male (Pang et al., 2000). Young male rider that involve in traffic accident may probably due to their risky riding behaviour whereas young female riders may due to lack of riding experience (Chang and Yeh, 2007). While exceeding traffic speed limit were found to be the most usual violation risky riding behaviour of motorcyclists (Steg and Brussel, 2009). This might explain why majority of the road accidents involved young motorcyclists.
2.7 Driving Experience
Matthew and Simpson (1990) had reevaluated the related researches more than a decade ago and figured out a few studies were intending to determine the relationship of driver age and experience. Some of the studies mentioned that the driving experience which was referring to the number of years that the driver had obtained the particular vehicles license was more important whereas the age-related aspects were related more significantly to the involvement of road accident. In their study was mentioned that both age and driving experience have independent relationship with the participation in road accident and the influences in driving experience were significantly greater among the
older drivers than the younger drivers. In fact, Anne et al.,(2009) also mentioned that driving experience and drivers’ age have significant and independent influences on crash risk.
Drivers that are below 20 years old are having eight times more likely to be participated in road fatality or injuries when they were just obtained the drivers’ license in the first three months compared to the previous three months on a learner's driving license (Pnina et al., 2018). In fact, vehicle crash rates were highest during the first few months after the newly obtained full-privilege licensed teenage drivers started to drive without any guidance or supervision of other licensed drivers (McCartt et al., 2003).
However, in the study of Waller et al., (2001), the types of road accident were assumed to be less affected by increasing experience as the intentionally risk-taking behaviors are rather a factor of age and personal characteristics than the driving experience. While for the unlicensed drivers, they are having higher risk for involving in a car crash injury after taking other crash-related risk factors into account (Stephanie et al., 2005).
2.8 Cronbach’s Alpha
It is common to see the reliability of instruments used in published science education studies framed in terms of a statistic known as Cronbach’s alpha. Cronbach’s alpha has been described as ‘one of the most important and pervasive statistics in research involving test construction and use’ to the extent that its use in research with multiple-item measurements is considered routine (Schmitt, 1996). Cronbach’s Alpha is commonly reported for the development of scales intended to measure attitudes and other affective constructs. However, the literature also includes reports of the
development of tests of student knowledge and understanding that cite Cronbach’s alpha as an indicator of instrument quality.
2.9 Principal Component Analysis (PCA)
Principal Component Analysis (PCA) was invented in 1901 by Karl Pearson, as an analogue of the principal axis theorem in mechanics. It was later independently developed and named by Harold Hotelling in the 1930s. Depending on the field of application, it is also named the discrete Karhunen–Loève transform (KLT) in signal processing, the Hotelling transform in multivariate quality control, proper orthogonal decomposition (POD) in mechanical engineering, singular value decomposition (SVD) of X (Golub and Van Loan, 1983), eigenvalue decomposition (EVD) of XTX in linear algebra, factor analysis (Jolliffe, 2002), Eckart–Young theorem (Harman, 1960), or empirical orthogonal functions (EOF) in meteorological science, empirical eigenfunction decomposition (Sirovich and Kirby, 1987), empirical component analysis (Lorenz, 1956), spectral decomposition in noise and vibration, and empirical modal analysis in structural dynamics.
The results of a PCA are usually discussed in terms of component scores, sometimes called factor scores (the transformed variable values corresponding to a particular data point), and loadings (the weight by which each standardized original variable should be multiplied to get the component score).
2.10 Structural Equation Modeling (SEM)
Structural equation modeling (SEM) is a form of causal modeling that includes a diverse set of mathematical models and computer algorithms. It is a statistical methods that fit networks of constructs to variety of data. SEM includes confirmatory
factor analysis, confirmatory composite analysis, path analysis, partial least squares path modeling, and latent growth modeling. SEM is used in this study to simplify the complex relationships among the variables. Matrix co-variance is basically used in the analysis which could give a more accurate results than the linear regression method (Hair et al., 2014). This kind of modelling can be used to determine the dependence variables simultaneously. Therefore, it is very suitable and helpful in interpreting the human behavioral factors.
SEM can be classified into covariance based SEM and component based on SEM. The first approach has been developed by Karl Joreskog and second approach by Herman Wold under the name of Partial Least Squares. Covariance based SEM is usually used with an objective of model validation and requires a large sample.
Component based SEM is mainly used for score computation and can be carried out on very small samples (Tenenhaus, 2008). SEM is a technique used to specify, estimate, and evaluate models of linear models among a set of observed variables in terms of an often smaller number of unobserved variables. SEM may be used to build or test a theory. When constructing the SEM, care should be taken to consider a theory’s stage of development. Exploratory techniques are well suited for establishing and whether it explains a meaningful amount of variance in an endogenous construct. Because of the components-based approach to estimating relationships, exploratory techniques such as PLS are less prone to Type I error and better suited for small, non-normal data-sets which were often collected for initial tests of relationships. Regardless of whether the SEM technique is exploratory or confirmatory,it possesses the ability to integrate measurement and structural models (Roberts et al., 2010).
Goodness-of-fits and null models are used to verify the created model (Hooper et al., 2008). Chi-Squared test, Root Mean Square Error of Approximation (RMSEA),
Adjusted Goodness-of-Fit Index (AGFI) and Goodness of-Fit Index (GFI) are also taken into account as the measures of the model.
In Table 2.1, the acceptable fit and good fit of each fit measures is shown whereas the Goodness-of-fit for structural equation model is shown in Table 2.2.
Table 2.1. Structural Equation Modelling Evaluations: Rules of Thumb
Fit Measures Good Fit Acceptable Fit
Root Mean Square Error of Approcimation (RMSEA)
0 < RMSEA< 0.05 0.05 ≤ RMSEA ≤ 0.10
Normed Fit Index (NFI) 0.95 ≤ NFI ≤ 1 0.90 ≤ NFI ≤ 0.95 Comparative Fit Index (CFI) 0.97 ≤ CFI ≤ 1 0.95 ≤ CFI ≤ 0.97 Goodness-of-Fit Index (GFI) 0.95 ≤ GFI ≤ 1 0.90 ≤ GFI ≤ 0.95 Adjusted Goodness-of-Fit Index 0.90 ≤ AGFI ≤ 1 0.85 ≤ AGFI ≤ 0.90
Table 2.2. Goodness-of-fit for structural equation model
Test Statistics Fit Indices Indicator Value
Absolute fit test χ2/df <5
p-value <0.05
RMSEA <0.08
Incremental fit test NFI Between 0 and 1, close to
1 is better
CFI Between 0 and 1, close to
1 is better
Parsimonious fit test PNFI >0.5
2.11 Summary
The purpose of this study is to understand the risk perception, positive affect and risky riding behaviour of motorcyclists as well as the relationship among these variables. These three variables are commonly discussed in the risky riding behaviour study. Cronbach Alpha is the most common method to conduct reliability test. While Principal Component Analysis (PCA) is found to be more suitable in determining the significance observed variables. Moreover, Structural Equation Modeling (SEM) is the best method to understand the causal relationship among the variables.
CHAPTER 3 METHODOLOGY
3.1 Overview
In this chapter, the methodology undertaken for the study on motorcyclists’
risky behaviour and their risk perception are being discussed. Hypothesis are being created based on the preliminary study of the motorcyclists’ characteristics. The hypothesis were then tested in the study. A simple questionnaire was constructed for collecting the data. The questions were grouped into three sections, to determine how frequently they were to behave the particular behaviour, how risky they think the particular behaviours are and how much they like to do it respectively. In fact, basic demographic information and personal characteristic of the respondents were collected, such as age, gender, driving experience, type of motorcycles and how many distance they normally used to travel per trip. Factor analysis was then being used to analyse the data collected. A model was developed to assess the relationship among the variables based on Structural Equation Modeling (SEM).
Figure 3.1 shows the flow of the steps taken in this study. Factors of motorcyclists’ risky riding behaviour are determined before the questionnaire survey is designed and conducted. Data collection is done in face-to-face and online methods.
The data collected are then analysed through three methods, namely preliminary analysis, Principal Component Analysis (PCA) and model estimations and causal relationship analysis via Structural Equation Modeling (SEM).
Figure 3.1 Flowchart of the methodology of the study Determine the factors of motorcyclists’ risky riding
behaviour that could lead to accident
Questionnaire:
Demographics of the respondents
Ranking the risky riding behaviour, risk perception and positive affect
Data Collection:
Online and Face-to-Face Method
Data Analysis
Results and Conclusion
Preliminary Analysis Principal Component Analysis (PCA)
Model Estimations and Causal Relationship Analysis Structural Equation Modelling (SEM)
3.2 Questionnaire
The questionnaire survey is an important instrument used to obtain a variety of information on the properties of any particular aspect, behaviour, beliefs and reasons for any action (Bulmer, 2004). There are lots of studies which discussed about motorcyclists safety via questionnaires in order to determine the mental properties of motorcyclists that lead to risky riding behaviour and road fatalities (Scott-Parker et al., 2009).
The survey target is only for respondents that currently hold the B, B1 and B2 driving license. Questionnaire is a survey tool that assess the motorcyclists on their frequency of behaving the risky act while riding, perception towards risky riding behaviour and preference upon those behaviour. Additional questions are added to collect basic information of the respondents, such as age, gender, driving experience, type of motorcycle and how far they used to travel per trip.
Based on the literature review, there are total of 15 questions are created in this questionnaire. For every section, a 5-point scale system is used for respondents to answer the questions. In section 1, every respondent was asked to rate the questions on how frequently were they behave like these 15 behaviours with ranging from 1 (Never) to 5 (Always). While in section 2, the respondents were asked about how risky do they think of these 15 behaviours (1-Not Risky At All; 5- Extremely Risky), whereas respondents were need to rate the questions upon how much did they like to behave like these 15 behaviours in section 3 (1-Extremely Dislike ; 5- Like it Very Much). The questionnaire is shown in the Appendix A. In summary, all the question were given a variable code to represent themselves which is as shown in Table 3.1.
Table 3.1 Fifteen Behaviour Measures
No. Behaviour Measures Variable Code
Risky
Behaviour Risk
Perception Positive Affect 1 Frequently changing lane to
overtake the vehicle in front. X1 Y1 Z1
2 Speeding up and suddenly braking. X2 Y2 Z2
3 Exceeding speed limit even feeling
unsafe. X3 Y3 Z3
4 Riding fast on the curve. X4 Y4 Z4
5 Continue riding although feeling
sleepy. X5 Y5 Z5
6 Taking alcohol before riding. X6 Y6 Z6
7 Run a red light. X7 Y7 Z7
8 Racing with other vehicles. X8 Y8 Z8
9 Riding during peak hour. X9 Y9 Z9
10 Fail to keep a proper distance with
other vehicles. X10 Y10 Z10
11 Making overtake/turn without
using signal lights. X11 Y11 Z11
12 Riding without wearing crash
helmet. X12 Y12 Z12
13 Crossing a stop-junction without
fully stopping. X13 Y13 Z13
14 Using phone while riding X14 Y14 Z14
15 Not switching on the headlights
during daytime. X15 Y15 Z15
For simplicity, each behaviour measure are represented by different variable code such as X, Y and Z in order to easily state in the following chapters.
A total of 349 respondents from different states in Malaysia such as Pulau Pinang, Johor, Kedah, Kelantan, Melaka, Negeri Sembilan, Pahang, Perak, Selangor, Sarawak, Terengganu and Wilayah Persekutuan were participated in this survey in February-April 2019. The participants were told about the purpose of the survey and asked for their permission and willingness to participate before the survey commence.
However, only 204 of samples were completed and used.
All the personal data and characteristics as well as the self-reported of risk perception, positive affect and risky behaviour were collected anonymously from the motorcyclists. Self-reported method is chosen as with such variety of risky behaviour, it is very convenient for the respondents to answer (Rhodes and Pivik, 2011).
3.3 Data Analysis
The data collected were first analysed and a simple direct percentage chart is created to have a simple insight of the results by showing the motorcyclists’ answers based on percentage. In this study, the self-reported risk perception, positive affect and risky behaviour are latent variables while there are fifteen risky behaviour measures items are observed variables.
Reliability test was conducted by determine the Cronbach’s Alpha value prior to any other analysis in the study. To obtain the significance of each item’s factor loading, the results were then analysed by using Principal Component Analysis (PCA) and a causal modeling is constructed to show the simplify version of complex relationship among the variables of the data. Lastly, a model was created via Structural Equation Modelling (SEM) to present the relationship between the latent
3.3.1 Reliability Test (Cronbach’s Alpha)
In statistics and psychometrics, reliability is usually used as a measure to determine the overall consistency. Reliability is the consistency of results when the experiment is replicated under the same conditions. Reliability test is conducted in this study to ensure the reliability of the study is good enough. Study results must be reliable else the research questions would not meet the aim of the study which in turn fail to generalise any susceptible and useless findings. Generally, the coefficients of reliability would be in the range of 0 to 1 where “0” means too many errors and “1”
defines no error. This value is to indicate the amount of error in the study results (Neil et al., 2009).
Cronbach’s Alpha was used to conducted the reliability analysis in this study.
In assessing an unidirectional latent construct, Cronbach’s Alpha is playing a role as a scale reliability of measurement to quantify the goodness of items or variables (Hassan and Abdel-aty, 2013). Alpha value that higher than 0.8 indicates the items within a study or test is high in internal consistency (Loo et al., 2015). While for Cronbach’s Alpha coefficient that is more than 0.7, it is generally be considered as minimum acceptable value.
3.3.2 Principal Component Analysis (PCA)
In order to differentiate the correlated and uncorrelated linear combinations of the covariates to avoid multicollinearity, Principal Component Analysis (PCA) is chosen for this study to analyse the significance of the self-reported perception and preference towards the particular questions.
PCA is usually used to verify that the linear combinations are having maximum variance. While the results obtained from the PCA are generally discussed
scores are the transformed variable values correlated with a particular data point.
While for loadings, it is refering to the weight where every standardised orginal variable should be multiplied and obtain the component scores (Shaw, 2003).
In this study, Principal Component Analysis (PCA) was performed under the condition where Eigenvalue is more than to 1. Number of factors in an analysis of a particular variable can be usefully determined by Eigenvalue (Fyhri and Backer Grondahl, 2012). Other than that, reliability value that higher than 0.7, 20% or higher variance explained and rotation is conducted in Principal Component Analysis (PCA) to obtain factor loading are also the important conditions met for running the Principal Component Analysis (PCA) (Hooper et al., 2008).
Factor analysis is a technique that is used to reduce a large number of variables into fewer numbers of factors. This technique extracts maximum common variance from all variables and puts them into a common score. As an index of all variables, this score can be used for further analysis.
3.3.3 Structural Equation Modeling (SEM)
A model estimations and a causal relationship analysis was conducted. A statistical model was firstly constructed based on the latent variables’ correlations, correlation between the latent variables and observed variables and also the hypothesis model. Figure 3.2 shows the entire relationship among the latent variables and observed variables in this study.
*RP= Risk Perception; PA= Positive Affect; RB= Risky Behaviour; PC= Personal Characteristic; EX= Driving Experience; CC= Type of Motorcycles
Figure 3.2 Relationship Model of the Study
Hence, a hypothesised model is initially involved to investigate the relationship among the latent measures shown in Figure 3.3 as follows:
H1- Risky perception would have a negatively influence on risky behaviour H2- Positive affect would have a positively influence on risky behaviour
CC
H3 - Positive affect are expected to have greater influence on risky riding behaviour than risky perception
3.4 Summary
In overall, three major analysis, namely Reliability Test, Principal Component Analysis (PCA) and Structural Equation Modeling (SEM) were being conducted to understand the relationship among the variables. Reliability test was conducted by using Cronbach’s Alpha at first to test the reliability of the study.
Principal Component Analysis (PCA) was then used to determine the significance of the observed variables. A Structural Equation Modeling (SEM) was constructed to determine the relationship among the variables.
CHAPTER 4
RESULTS AND DISCUSSION
4.1 Introduction
In this chapter, the study of the relationship between three latent variables namely risky behaviour, risk perception and positive affect of the motorcyclists is being discussed based on the results obtained from questionnaire survey and a variety of analysis. A simple and direct preliminary study analysis was conducted to know the perception and preference of the motorcyclists towards the risky riding behaviour as well as the frequency of doing the particular behaviour in terms of percentage.
Subsequently, Principal Component Analysis (PCA) and reliability test were conducted to know the significance of each observed variables and also the reliability of the survey itself. The significance factors were sorted out based on the the factor loading of the observed items and also the percentage of variance explained.
Moreover, a statistical path modeling was constructed to present and identify the estimates and the causal relationship among the three latent variables. An estimate value were obtained from the Structural Equation Modeling (SEM) where the relationship among the observed variables and latent variables were identified.
4.2 Description Statistics
Figure 4.1 : Percentage of Age Group of Respondent for Male and Female
As an insight of the data, all respondents aged between 17 and 63 years old.
In total, there are 71.1% of respondents was male and 28.9% was female. Most of the respondents are from the age group of 21-25 years old. For female, 83.1% of female respondents are from this age group whereas 73.1% of male respondents are within the age of 21 to 25 years old. In fact, the mean and standard deviation of age of the respondents were 23.54 years old and 4.02 years respectively.
Figure 4.2 : Percentage of Driving Experience of Respondents (Number of Years of Obtaining License)
Besides that, the mean of the driving experience (number of years that had already obtained the motorcycle license) of the respondents is 6.20 years. Majority of the respondents have 5 and 6 years of driving experience which are 20% and 15% of total respondents respectively.
Figure 4.3 : Percentage of Type of Motorcycles Obtained by the Respondents
No. Of Years
Among the respondents, most of the respondents are using motorcycle with 100cc- 250cc (68%). Secondly is motorcycle that is less than 100cc and next is 250cc- 500cc which consists of 26% and 12% of total respondents respectively. While, only 0.005% of respondents obtain motorcycles with 750cc-1000cc.
4.3 Preliminary Data Analysis
Preliminary data analysis was conducted prior to the detailed analysis. Figure 4.4 to 4.6 shows the survey results of the motorcyclists risky behaviour measures in Malaysia. The first figure, Figure 4.4, shows that more than 60% of respondents think that taking alcohol before riding is extremely risky behaviour. Although it is stated in the law of traffic safety, more than 20% of motorcyclists still perceived that “Not Switching on the Headlight During Daytime” is not risky at all. While basically, less than 10% of motorcyclists think that all measures are not risky at all except not switching on the headlight during daytime. In a nutshell, most of the motorcyclists believe that all the stated measures can lead to the occurrence of road accidents.
Figure 4.4 Self-reported Risky Perception of Motorcyclists in Malaysia
Figure 4.5 shows that the how much the motorcyclists favour (Positive Affect) to behave like the questions stated. The second figure indicates more than 75%
extremely dislike in drunk riding. More than 55% of motorcyclists are potentially like to frequently changing the lane to overtake the vehicle in front. However, just less than 10% of motorcyclist actually like to do those listed risky behaviour measures while riding. This shows that most of the motorcyclists are fully aware of the importance of traffic safety.
Figure 4.5 Self-reported Positive Affect of Motorcyclists in Malaysia
In Figure 4.6, more than 85% of motorcyclists never takes alcohol before riding. In fact, more than 85% of motorcyclists would ride during peak hour although it is unsafe. Moreover, more than 40% of motorcyclists have raced with other vehicles on the road and surprisingly, almost 13% of motorcyclists have never change lane to overtake the vehicles in front while riding.
Figure 4.6 Self-reported Risky Behaviours of Motorcyclists in Malaysia
In Figure 4.7 to 4.9, the mean of self-reported risk perception, positive affect and risky behaviour is discussed. This shows the average rating of each observed variables which indicate the trend of the respondents’ riding behaviour, risk perception and positive affect.
The mean of each self-reported component (latent variable) namely risky perception, positive affect or personal preferences and risky behaviour are presented in following graphs.
From Figure 4.7, it shows that most of the motorcyclist perceived that riding without switching on the headlights during daytime (Y15) is the least risky behaviour compared to the other statements. In fact, the majority of the motorcyclists think that
taking alcohol before riding motorcycle (Y6) is an extremely risky action which would cause severe traffic accident.
Figure 4.7 Mean of Self-reported Risky Perception
As in Figure 4.8, most of the motorcyclists tend to frequently changing lane to overtake the vehicles in front (Z1). This is actually very common in Malaysia and this is most probably because of the the high volume of vehicles on road and lack of motorcycle lanes in Malaysia. Thus, the motorcyclists need to share lane with the other vehicles and in turn the motorcyclists like to keep changing lane to overtake the other vehicles especially when the traffic is congested.
On the other hand, most of the motorcyclists do not like to take alcohol before riding the motorcycles (Z6).
Figure 4.8 Mean of Self-reported Positive Affect
The mean of the self-reported risky riding behaviour is shown in Figure 4.9.
From the Figure 4.9, taking alcohol before riding (X6) can be considered as the least favourable riding behaviour among the motorcyclist. In fact, most of the motorcyclist tend to ride during peak hour (X9). As described in the Figure 4.4, most of the motorcyclists perceived that riding during peak hour is the least risky behaviour among the other, thus this is also the most frequently risky riding behaviour.
Figure 4.9 Mean of Self-reported Risky Behaviour
In the following Table 4.1, the mean and standard deviation of each self-reported 15 observed variables were listed. The mean value indicates the rating of each of the observed variable averagely whereas the standard deviation of each observed variables showshow much the each observed variable differ from the mean value for the latent variables.
Table 4.1 Mean and Standard Deviation of Each Observed Variable
Observed Variables Mean SD. Mean SD. Mean SD.
Risky Behaviour Risk Perception Positive Affect Frequently changing lane to
overtake the vehicle in front. 2.941 1.214 3.618 1.136 2.642 1.238 Speeding up and suddenly
braking. 2.142 0.928 3.936 1.166 1.941 1.035
Exceeding speed limit even
feeling unsafe. 2.074 1.131 3.873 1.241 2.069 1.134
Riding fast on the curve. 1.995 0.970 4.059 1.202 2.069 1.181 Continue riding although
feeling sleepy. 2.049 1.122 4.019 1.267 1.706 0.826
Taking alcohol before riding. 1.211 0.587 4.201 1.314 1.348 0.750
Run a red light. 2.020 1.096 3.897 1.253 1.921 1.024
Racing with other vehicles. 1.500 0.874 4.054 1.245 1.735 1.036 Riding during peak hour 3.216 1.245 3.015 1.129 2.265 1.140 Fail to keep a proper distance
with other vehicles 2.206 1.016 3.603 1.151 2.010 0.915 Making overtake/turn without
using signal lights. 1.887 0.968 3.775 1.207 1.814 0.965 Riding without wearing crash
helmet. 1.608 0.933 4.123 1.236 1.730 1.046
Crossing a stop-junction
without fully stopping. 2.191 1.109 3.632 1.156 2.138 1.128 Using phone while riding 1.549 0.872 4.088 1.233 1.721 1.029 Not switching on the
headlights during daytime. 1.931 1.345 2.936 1.368 2.240 1.230
*SD = Standard Deviation
4.3 Reliability Test
Table 4.2 shows that the values of Cronbach’s Alpha obtained for the three latent variables, risky behaviour, risky perception and positive affect are 0.833, 0.945 and 0.889. These values indicate that three variables have have level of internal consistency and in turn showed that the study is reliable.
Table 4.2 Reliability Test (Cronbach's Alpha) of the Self-reported Variables.
Latent Variables Cronbach’s Alpha
Risky Behaviours (X) 0.833
Risky Perception (Y) 0.945
Positive Affect (Z) 0.889
4.4 Principal Component Analysis (PCA)
From the results obtained, 31.493%, 60.134% and 41.404% of variances describe three components, risky behaviour, risk perception and positive affect respectively.
As the results shown, the items that have factor loadings lower than 0.5 were considered as insignificant. Those insignificant items consists of X2, X6, X7, X13, Y10, Y13, Z5 and Z15. Moreover, there are 15 observed variables are categorised into a group where the variance explained is less then 20%. Observations that explained lower thatn 20% of variance are also considered as insignificant. These 15 observed variables are X5, X9, X10, X11, X12, X15, Y9, Y15, Z1, Z2, Z3, Z4, Z8, Z9 and Z10.
Table 4.3 shows that motorcyclists in Malaysia were significantly connected with the risky behaviours namely “Frequently changing lanes to overtake (X1)”,
“Exceeding speed limit even feeling unsafe (X3)”, “Riding fast on curve (X4)”,
factor loadings 0.580, 0.653, 0.669, 0.651 and 0.556 respectively. This is compatible with the previous studies (Steg and Brussel, 2009) saying that manner or behaviour of a motorcyclists is one of the issues that would lead to vehicle crash.
In fact, motorcyclists were also associated with their perception on risky riding behaviours which consisting the items of variables Y1, Y2, Y3, Y4, Y5, Y6, Y7, Y8, Y11, Y12 and Y14 with loading factor of 0.577, 0.790, 0.829, 0.864, 0.869, 0.900, 0.744, 0.884, 0.738, 0.832 and 0.832 respectively. This showed that risky perception of motorcyclists in Malaysia would significantly affect their riding behaviour. Risk perception would precisely affect the one’s risky behaviour and take protective action to ensure their safeness (Noel et al., 2004). Statement like “How Risky Do You Think Taking alcohol Before Riding (Y6)” scored the highest factor loading among the rest of the risky perception statements.
Moreover, the motorcyclists were also related to their preference on riding behaviour such as “Taking alcohol before riding (Z6)”, “Run a red light (Z7)”,
“Making overtake/turn without using signal lights (Z11)”, “Riding without wearing crash helmet (Z12)”, “Crossing a stop-junction without fully stopping (Z13)” and
“Using phone while riding (Z14)”. The factor loading were 0.574, 0.509, 0.698, 0.781 and 0.693 respectively. Sensation seeking of motorcyclists would affect the riding behaviour negatively (Ulleberg and Rundmo, 2003). The results was consistent with the previous studies which also stated that emotional are closely related to the risky riding behaviour (Britt, 2003; Ferguson et al., 2003).
Table 4.3 shows that the summary of factor loadings of the variables, percentage of variance of each latent variables that had explained and also the Cronbach’s Alpha of each latent variables.
Table 4.3 Variables and Reliable Analysis of Latent and Observed Variable Latent
Variables
Observed Variables
Variables Code
Loading Factors
Variance Explained
Cronbach’s Alpha
Risk Perception
Frequently changing lane to
overtake the vehicle in front.
Y1 0.577
60.134% 0.945 Speeding up and
suddenly
braking. Y2 0.790
Exceeding speed limit even
feeling unsafe. Y3 0.829 Riding fast on
the curve. Y4 0.864
Continue riding although feeling
sleepy. Y5 0.869
Taking alcohol
before riding. Y6 0.900 Run a red light. Y7 0.774
Racing with
other vehicles. Y8 0.844 Making
overtake/turn without using signal lights.
Y11 0.738
Riding without wearing crash
helmet. Y12 0.832
Using phone
while riding Y14 0.832
Table 4.4 Variables and Reliable Analysis of Latent and Observed Variable (cont.) Latent
Variables
Observed Variables
Variables Code
Loading Factors
Variance Explained
Cronbach’s Alpha
Positive Affect
Taking alcohol
before riding. Z6 0.574
41.404% 0.889 Run a red light. Z7 0.509
Making overtake/turn without using signal lights.
Z11 0.698
Riding without wearing crash
helmet. Z12 0.781
Crossing a stop-junction without fully
stopping.
Z13 0.693
Using phone
while riding. Z14 0.654
Risk Behaviour
Frequently changing lane to
overtake the vehicle in front.
X1 0.580
31.493% 0.833 Exceeding speed
limit even feeling
unsafe. X3 0.653
Riding fast on the
curve. X4 0.669
Racing with other
vehicles. X8 0.651
Using phone
while riding X14 0.556
4.5 Structural Equation Modeling (SEM)
In order to present a model estimations and a causal relationship analysis, a statistical model was constructed based on the latent variables’ correlations, correlation between the latent variables and observed variables and also the hypothesis model. A statistical software IBM SPSS AMOS (Statistical Package for the Social Sciences- AMOS) is used to test hypotheses on complex variable relationship and construct a model to present the causal relationship among the variables. The correlations among the measures may include three components as the risky perception and positive affect of motorcyclists are the measures of factors for their risky behaviour. In accordance to the completed Principal Component Analysis, Reliability test and also the hypothesized models, the relationship among the three components are drawn as shown in Figure 4.10.
*RP= Risk Perception; PA= Positive Affect; RB= Risky Behaviour; PC= Personal Characteristic; EX= Driving Experience
Figure 4.10 Relationship of Latent Variables Components via Structural Equation Modeling (SEM)
A validity criteria from Schermelleh-Engle and Moosbrugger (2003) and Lai (2011) was used in this study to obtain the goodness of fit of the constructed model.
Consequently, indicators such as NFI and CFI values are less than the cut-off point.
However, value such as χ2/df, RMSEA and PNFI are statistically acceptable.
Table 4.5 Goodness-of-Fit measures of the developed model.
Fit Measure Developed Model Acceptable Fit
χ2 941.2 -
df 229 -
P-value 0.000 <0.05
χ2/df 4.11 <5
RMSEA 0.074 <0.08
NFI 0.595 0 < NFI < 1,
close to 1 is better
CFI 0.646 0 < CFI < 1,
close to 1 is better
GFI 0.926 0.90 ≤ GFI ≤ 0.95
AGFI 0.852 0.85 ≤ GFI ≤ 0.90
PNFI 0.59 >0.5
Figure 4.10 shows a influential path model from two latent variables, motorcyclists’ perception and positive affect to another latent variables, risky behaviour. Latent variable of risk perception is determined by 11 observed variables.
These included Y1 (Frequently changing lane to overtake the vehicle in front.), Y2 (Speeding up and suddenly braking), Y3 (Exceeding speed limit even feeling unsafe), Y4 (Riding fast on the curve), Y5 (Continue riding although feeling sleepy), Y6 (Taking alcohol before riding), Y7 (Run a red light), Y8 (Racing with other vehicles), Y11 (Making overtake/turn without using signal lights), Y12 (Riding without
wearing crash helmet) and Y14 (Using phone while riding). Moreover, six observed variables were also determined on latent variable positive affect. The six observed variables consists of Z6 (Taking alcohol before riding), Z7 (Run a red light), Z11 (Making overtake/turn without using signal lights), Z12 (Riding without wearing crash helmet), Z13 (Crossing a stop-junction without fully stopping) and Z14 (Using phone while riding).
In fact, latent variable risky behaviour was determined by five observed variables, namely X1 (Frequently changing lane to overtake the vehicle in front), X3 (Exceeding speed limit even feeling unsafe), X4 (Riding fast on the curve), X8 (Racing with other vehicles) and X14 (Using phone while riding).
Table 4.6 shows the standardised estimate coefficient of the path analysis model which indicate the relationship of the observed variables with the latent variables
Table 4.6 Standardised Coeffiecient Weight of the Motorcylists Model
Path Estimate
Risky Behaviour <== Age -0.037
Risk Perception <== Driving Experience 0.012 Positive Affect <== Driving Experience -0.032 Risky Behaviour <== Risky Perception 0.036 Risky Behaviour <== Positive Affect 1.003
Y1 <== Risky Perception 0.729
Y2 <== Risky Perception 0.894
Y3 <== Risky Perception 1.009
Y4 <== Risky Perception 1.000
Y5 <== Risky Perception 1.062
Y6 <== Risky Perception 1.112
Y7 <== Risky Perception 0.979
Y8 <== Risky Perception 1.024
Y11 <== Risky Perception 0.922
Y12<== Risky Perception 0.985
Y14 <== Risky Perception 0.990
Z12 <== Positive Affect 0.624
Z11 <== Positive Affect 0.634
Z13 <== Positive Affect 0.719
Z14 <== Positive Affect 0.628
Z6 <== Positive Affect 0.345
Z7 <== Positive Affect 0.581
X14 <== Risky Behaviour 0.298
X4 <== Risky Behaviour 0.410
X8 <== Risky Behaviour 0.350
X3 <== Risky Behaviour 0.437
X1 <== Risky Behaviour 0.459
The model constructed is based on the age which consists of younger (under 23 years old) and elder communities (above and equal 23 years old) of motorcyclists. Moreover, the model is constructed with the variety of respondents’
characteristics, such as male and female motorcyclists, different driving experiences etc. However, the model results are not significant. This is most probably due to the insufficient sample size for each of the age group and gender. Thus, if there is further studies upon this, sample size is expected to be concerned in the modelling specially
This study also taking the age and driving experience of motorcyclists into consideration of the path analysis. From Table 4.6, the age of motorcyclists has significant affect on the risky behaviour (estimate= -0.037). This indicates that young motorcyclists were more frequently to perform risky riding behaviour while riding on the road. Besides, the findings also show that driving experience has significant influence on motorcyclists’ risk perception and positive affect which the estimate scores 0.012 and -0.032 respectively. This shows that motorcyclists that already obtained the motorcyclist license for a longer time would perceive safer risky behaviour and dislike to perform those risky riding behaviour.
Positive affect (loading factor = 1.003) is known as more influences than the risk perception measures on risky behaviour. Thus, this shows that intuitive processes are the major process for the motorcyclists in Malaysia rather than rational processes while they are riding on the road. In a previous study, the findings concluded that positive emotions in driving pleasures (Positive affect) strongly affect the risk behaviour for the drivers (Rhodes and Pivik, 2011). Therefore, this study suggested that the positive affect should be highly concerned in the further study with the expectation to figure out some suggestions to minimise the motorcyclists risky riding behaviour.
Besides, relationship path between risk perception and risky behaviour were found to be weak positive (estimated value= 0.036). This category of measures consists of 11 components, namely “frequently changing lane to overtake the vehicle in front”, “speeding up and suddenly braking”, “exceeding speed limit even feeling unsafe”, “riding fast on the curve”, “continue riding although feeling sleepy”, “taking alcohol before riding”, “run a red light”, “racing with other vehicles”, “making overtake/turn without using signal lights”, “riding without wearing crash helmet and
using phone while riding”. In a nutshell, these 11 risk perception measures were weakly directly related to riding behaviour of motorcyclists. The highest positive influencing measure on risky riding behaviour is “Taking Alcohol Before Riding (Y6)” with loading factor of 1.112. This is statistically known as the riskiest riding behaviour. Not only that, more than 60% of motorcyclists think that it is extremely risky and most of the motorcyclist (around 95% of them) have never or rarely doing that while riding on the road.
On the other hand, positive affect or motorcyclists preference of riding behaviour is significantly associated with risky behaviour with estimated value 1.003.
This can be said that positive affect has more impact on risky behaviour than risk perception. These included taking alcohol before riding, run a red light, making overtake/turn without using signal lights, riding without wearing crash helmet, crossing a stop-junction without fully stopping and using phone while riding. The strongest measure of positive affect is crossing a stop-junction without fully stopping with estimate value 0.719. Surprisingly, less than 10% of motorcyclists think that it is not risky at all. In fact, 5% of motorcyclists like it very much, and around 10% of motorcyclists are often and always doing that.
Positive affect was one of the primary influence the risky behavior intention of the motorcyclists (Watso et al, 2007). Generally, the high scoring of motorcyclists on positive affect are defined to be more probably to against the traffic rule in future (Wedagama, 2017). This measure towards risky riding behaviour shows that motorcyclists in Malaysia are intending to take part in a particular dangerous behaviour while riding on the roads. In Table 4.6, it suggests that when the positive affect or personal preference towards risky riding behaviour by one unit, the measures of risky behaviour would increase by 1.004 unit. In the previous research
by Zamani-Alavijeh et al. (2009), it was also found that positive affect or personal preference was significantly associated towards the intentions to against the traffic rules.
For the most significant measure for risky behaviour, “frequently changing lane to overtake the vehicle in front” with estimate coefficient value 0.459, can be described as positively enjoy and slightly positive perceived by the motorcyclists.
More than 90% of motorcyclists have been doing that while riding on road while less than 5% of motorcyclists perceived it extremely risky and less than 30% of motorcyclists extremely dislike it. Motorcyclists may seriously injure or encounter minor crash with the vehicles on the road because of the unstable and unpredictable movement of the vehicle which increase the probability of accidents. This also indicated that frequently changing lanes to overtake the vehicle in front of motorcyclists can be considered for both rational and intuitive processes of motorcyclists’ decision making while riding.
4.6 Summary
In this study, the use of self-reported perceptions, behaviour and positive affect of the motorcyclists while riding on the road is applicable. It may exactly bring back their involvement or events when they were against the traffic rules and regulations in the past. Anonymous surveys that have in-depth and detailed information is able to be conducted by using self-reported technique (Steg and Brussel, 2009).
From the study results obtained, increasing awareness of traffic rules and traffic safety that could increase the practices to road safety could be provided to the schools and universities, households and workplaces. This is also particularly
involves the civilised values of the society (Chakrabarty et al, 2013). Educational awareness initiatives could basically reduce the motorcyclists personal riding behaviour such as frequently changing lane to overtake the vehicle in front (estimate value= 0.459 ), exceeding speed limit even feeling unsafe (estimate value= 0.437) and riding fast on curve (estimate value= 0.410)
In fact, enforcement on traffic rules and regulations are suggested to reduce the traffic violations by the motorcyclists. Traffic rules and regulations are suggested to be comply with the educational initiatives to reduce the traffic violations effectively (Wedagama 2017). Thus, this could be effectively countering the significant risky riding behaviour of the motorcyclists such as racing with other vehicles (estimate value= 0.350) and using phone while riding (estimate value=
0.298).
