' '-.A'.- . IWImnmmmmmmm L 1 00870 8921 This is to certify that the dissertation entitled DRIVER CHARACTERISTICS AND THEIR RELATION TO TRAFFIC CONFLICT OCCURRENCE presented by Khaled M. Abdulghani has been accepted towards fulfillment of the requirements for Doctor of Philosophy degree". Civil Engineering Major professor/ Date May 17, 1982 MS U i: an Affirmative Action/Equal Opportunity Institution 0-1277 1 .15“ £425 ' w » . , ‘ v _,' ' . 4 . ~ *7 I'Clfl‘“, MSU LIBRARIES -_—. RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. ' .I i'.. -.. a Q». . J‘g‘, n“ ’, )tiva .1 '35 ’,.fi DRIVER CHARACTERISTICS AND THEIR RELATION TO TRAFFIC CONFLICT OCCURRENCE by Khaled M. Abdulghani A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Civil and Sanitary Engineering 1982 l 1‘) . . . n;? "| . LM - ;~ e, ‘ IF- '.1 ' ,1. .I . ..‘ ‘ D‘- ‘5 ‘u I— '\ ." 3 : . ~“y ' 0 it. .‘ n ,‘K‘, c“ ‘ u F—n.» J A 4w 4w w-v—ngffl '3 i l ' u .1 .1 »- ABSTRACT DRIVER CHARACTERISTICS AND THEIR RELATION TO TRAFFIC CONFLICT OCCURRENCE by Khaled M. Abdulghani A traffic conflict is defined as a hazardous situation in which an accident is avoided by evasive action by the driver of one of the vehicles. Research on the use of conflicts as a traffic engineering tool has been focused on attempts to correlate conflicts and accidents so that conflicts, which occur more frequently than accidents, can be used as a basis for predicting future accidents. Several stratifica- ‘ticnas and/or refinements to conflict measures have been introduced to acfliieve better correlation, but the use of driver characteristics has ruot been used in these research studies. The objectives of this study were to determine the driver charac- teri stics that are overrepresented in conflict involved drivers when cxmnpared to all Saudi Arabian drivers; the relationship of these char- acteristics to types of conflict; and the interrelationships between types of conflict and traffic control at the intersection. A final objective was to determine the possibility of predicting future con- flicts by knowing the mix of driver characteristics at an urban intersection. Conflict involved drivers observed at six different locations were found to possess similar characteristics, indicating that there is a pattern which defines hazardous drivers. Age was the single vari- able which explained the greatest variation in conflict involvement, but nationality, occupation, past accident experience and driver training “>- ...I u- were also significantly correlated with conflicts. Thus, predicting future conflict by knowing the mix of driver characteristics may be possible, but larger samples and more research is needed to achieve better correlation of variables in the multiple regression model. The relationships between driver characteristics and conflicts found in this study were similar to relationships known to exist between driver characteristics and accidents. Thus, it appears that conflicts may be used to identify hazardous intersections when past accident records are not available. Another conclusion reached in this study was that there are sig- nificant differences in the conflict rate depending on the training received by the driver. It appears that more strict driver licensing and enforcement of traffic regulations would decrease the conflict rate, and presumably the accident rate, in Saudi Arabia. Dedicated to my father, Rashad, and my mother, Faigah, who sacri- ficed their life for the education of myself and my ten brothers and sisters. My father has been continuously concerned about our education and provided the impossible to overcome the obstacles in our paths. His courage in going against the wish of the elders in sending my sister Roshdia abroad to study medicine and make her dream come true left me with great admiration. My mother's closeness to me and my brothers and sisters, and her continuous prayers that God assist us and be with us were replied. It is for her outpouring feelings and upbringing that gave us recognition among friends. On behalf of my brothers and sisters, may I express our deep grati- tude for the great price you have paid which we can never repay as long as we live. iii 11:2”!- "H u ’ . an I. .. F‘w- .. or ‘. u . :'_. "o' u‘; ~ “Q m, I ‘ i ' Q ~ ‘n6 . A‘ «I_ .' -. . ‘ \ ‘I F i. a . §.. ‘1‘ ‘~q .' a A I. N ‘ ACKNOWLEDGMENTS I wish to express my deep appreciation to my major professor, Dr. William C. Taylor, for his continuous guidance and assistance from the initial process of proposal development to the completion of this research, and throughout my entire doctoral program. I am deeply indebted to him for sharing his knowledge and experience with myself and to him and his wife Norma for being the best of friends in our home away from home. I wish to thank Dr. James D. Brogan for his invaluable guidance during my Master's program, and in the conduct of this research. I also want to express my gratitude to Dr. Tapan K. Datta whose exper- ience with this area of research provided constructive suggestions and guidance. I am grateful to Dr. Gail C. Blomquist who was readily available to make contributions on research problems. Dr. Koroos Mehjoob-Behrooz deserves recognition for his guidance in the statisti- cal design and analysis. The author wishes to extend his appreciation to Dr. Fahad H. Al-Dakhil, Vice Rector of UPM, for his assistance in eliminating all obstacles during the data collection phase. Recognition must also be given to Captain Abdulqadir Talha, Head of Dammam's Traffic Authority, vvho participated in providing manpower for this research. Special appreciation is extended to Ms. Vicki Brannan who helped in the typing of this thesis. My beloved wife Amera has shown unlimited devotion and understanding iii in during the field studies and the entire period of this research. She was very devoted in her support of me, and in the upbringing of our son Yasser and our daughter Nasreen whose closeness generated a new vigour in me. My brothers Sami and Essam deserve my special gratitude for their encouragement and support throughout my doctoral program. I would like to thank all the unnamed friends who contributed morally in the development of this research. iv . in. A. 41-0 . "‘ a. L;) .0. I.) 't‘ . u '. l . ‘ 'o . ‘ D . -. s .. . u s .‘ .2“- Ink 4. LI ) v an .E I . o '1 .._‘ i p '5 1'1'1 "T v. List of List of CHAPTER 1.1 1.2 1.3 CHAPTER 2.1 2.2 2.3 2.4 CHAPTER 3.1 3.2 3.3 3.4 3.5 CHAPTER 4.1 4.2 4.3 4.4 TABLE OF CONTENTS Tables Figures 1 Introduction The Problem Objectives of This Study 2: LITERATURE REVIEW Exposure Measures Accident Patterns Driver Characteristics Traffic Conflicts 3: METHODOLOGY Location Choice of Intersections Observer Training Experimental Design Schedule of Activities 4: DATA ANALYSIS Observer Reliability Representation of Driver Characteristics Characteristics of Hazardous Drivers Test for Driver Characteristics That Explain Conflict Involvement Page vii ix 12 13 18 23 23 23 24 26 28 28 29 32 35 '; 'A-«r. 1i 1‘ ‘ ‘- -s a o p" I “ 5 . . A \ [AF Ho 11v 5 ' 1 my .’ \ .0. fl 4 ""l - V A»; .a“ . i ' 'Ov-v Page 4.5 Evidence of Patterns Among Conflict Involved Drivers 47 4.6 Conflicts by Type 47 4.7 The Model 52 CHAPTER 5: CONCLUSIONS AND RECOMMENDATIONS 58 5.1 Conclusions 58 5.2 Discussion and Recommendations 59 BIBLIOGRAPHY 61 APPENDICES 65 vi .‘ll .\1 4‘ Table MN .33 10. 11. 12. 13. 14. 15. 16. 17. LIST OF TABLES Two motor vehicle accidents at urban intersections. Percentage accident rate by type of collision. Percentage of faulty driving techniques. Observed conflict counts for observer reliability. Conflict observation, reliability test. Conflict observation, reliability test. Chi-square test for population assumption. Sample sizes for studied intersections. Chi-square test for driver characteristics in sample 1 and sample 2 using all intersection data. Chi-square test for driver characteristics in sample 1 and sample 2 using signalized intersection data. Chi-square test for driver characteristics in sample 1 and sample 2 using unsignalized intersection data. Chi-square test for driver characteristics in sample 1 and sample 2 using unsignalized intersection data for drivers with two conflicts. Percentage increase in representation of conflict involved drivers over intersection driving public. Chi-square test for conflict involved drivers at signalized intersections. Chi-square test for conflict involved drivers at all intersections. Frequencies of conflicts by type at unsignalized intersections. Frequencies of conflicts by type at signalized intersections. vii 44 44 45 45 46 48 48 50 3.! U Table 18. Characteristics with similar distribution to conflict involvement by type. viii IV. Figure LIST OF FIGURES Percentage of driver age in samples 1 and 2. Percentages of driver age in sample 1 and turning conflict sample. . Percentage of driver age in sample 1 and in right angle conflict sample. Percentage of driver age in sample 1_and in side swipe conflict sample. Percentage of driver age in sample 1 and in rear end conflict sample. Age of conflict involved drivers with two or more accidents in past two years. Relationships between conflict index and age of conflict involved drivers. Relationships between conflict index and age of conflict involved drivers with and without accident history. ix Page 36 37 38 39 40 41 56 57 ‘ (In oc'oq. . '. . ,3 r o .Di . u. -\ . a 4 ‘2‘ “n4 ., ".1-é‘ .1;. . . 0 i a; ‘ 2‘ s r‘ "(u-i: ‘v fl '. ‘ h. \ ’I ‘ u 4 I - \ ~‘ . ‘ . v ' h‘ 7 “-5 “s CHAPTER 1 1.1 Introduction The prediction of intersection accidents has long been a goal of transportation research. Much of this research has been focused on attempts to develop techniques for predicting accidents as a basis for implementing safety standards to prevent or minimize these accidents. Traditionally, accident records have been used as a measure of the hazardousness of intersections. The fact that an accident may be the result of a driver's error, a defective vehicle, an inappropriate road- way condition, inadequate or improper traffic control devices, inade- quate design, or a combination of these factors, however, makes it dif- ficult to accurately predict future accidents based on past accidents. Also, the fact that accident records may be distorted, incorrect, not readily available or not statistically reliable, adds to the difficulty of developing adequate models from these records. Researchers have acknowledged the need for new methods to predict the number of accidents expected at any given intersection. Most of the research work in this area has attempted to use some measure of traffic exposure as a predictive variable. The most common exposure measures have been the volume, or the volume to capacity ratio. How- ever, neither of these measures capture the influence of geometric features on accident rates. One of the newest methods proposed as a basis for accident predic- tion is the Traffic Conflict Technique. A traffic conflict is defined as a hazardous situation in which an accident is avoided by evasive ‘ ' ‘ a a. ' .'. All. in V... 0.,“ J I! "‘7 up“ '5 K i - § 1 . . . F- ‘.\! (‘7. 2 action by one or both drivers to avoid a collision. A conflict is identi- fied by the occurrence of evasive actions (such as braking or weaving to avoid an impending accident) or a traffic violation. A traffic violation is defined in accordance with the uniform traffic code, and is considered a traffic conflict even if no other vehicle is in close proximity to the violator. Twelve conflict types have traditionally been used at intersections and are defined as follows (53): 1. Left turn, same direction: A left turn, same direction con- flict occurs when an instigating vehicle slows to make a left turn, thus placing a following, conflicted vehicle in jeopardy of a rear-end collision. 2. Right turn, same direction: A right turn, same direction con- flict situation occurs when an instigating vehicle slows to make a right turn, thus placing a following, conflicted vehicle in jeopardy of a rear-end collision. 3. Slow vehicle, same direction: A slow vehicle, same direction conflict situation occurs when an instigating vehicle slows while ap- proaching or passing through an intersection, thus placing a following vehicle in jeopardy of a rear-end collision. 4. Lane change: A lane change conflict situation occurs when an instigating vehicle changes from one lane to another, thus placing a following, conflicted vehicle in the new lane in jeopardy of a rear- end or side-swipe collision. 5. Opposing left turn: An opposing left turn conflict situation occurs when an oncoming vehicle makes a left turn, thus placing the conflicted vehicle in jeopardy of a head-on or broadside collision. 6. Right turn cross traffic, from right: A right turn cross 1.. “ . . 1‘ ‘flo "- 4“, h. .. , ‘ :r I". | \‘D -," d! u traffic from right situation occurs when an instigating vehicle approach- ing from the right makes a right turn, thus placing the conflicted vehicle in jeopardy of a broadside or rear-end collision. 7. Left turn cross traffic, from left: A left turn cross traffic from left conflict situation occurs when an instigating vehicle approach- ing from the left makes a left turn, thus placing a conflicted vehicle in jeopardy of a broadside or rear-end collision. 8. Left turn cross traffic, from right: A left turn cross traf- fic from right conflict situation occurs when an instigating vehicle approaching from the right makes a left turn, thus placing the con- flicted vehicle in jeopardy of a broadside collision. 9. Thru cross traffic, from left: A thru cross traffic from left conflict situation occurs when an instigating vehicle approaching from the left crosses in front of a conflicted vehicle, thus placing it in jeopardy of a broadside collision. IO. Thru cross traffic, from right: A thru cross traffic from right conflict situation occurs when an instigating vehicle approaching from the right crosses in front of the conflicted vehicle, thus placing it in jeopardy of a broadside collision. 11. U-Turns: A U—Turn conflict situation occurs when an instigat- ing vehicle makes a U-Turn in the vicinity of an intersection in front of a conflicted vehicle, thus placing it in jeopardy of a collision. 12. Disobeying red light or stop sign: This traffic violation occurs when an instigating vehicle violates the uniform traffic code. Whenever a vehicle is involved in more than one observed conflict, each type of conflict is recorded for that vehicle. For analysis purposes, the conflicts are often grOuped into four . a.» “una- . t' v _"’3 ”I.“ - 0 l .I p 7 .0 fi ( 1‘. ’I g 4“- .-\ .‘H‘ “.:‘ . I. 7“. .v‘ 2‘- .1 H, gr. ‘5. l 5.;- \ ‘ . ' v P1? 'v. a... 'n. . "a‘ “ categories: 1. Right angle conflicts: the sum of conflict counts for: a. Thru cross traffic, from left. b. Thru cross traffic, from right. c. Disobeying red light or stop sign. 2. Rear end conflicts: the sum of conflict counts for: a. Left turn, same direction. b. Right turn, same direction. c. Slow vehicle. 3. Turning conflicts: the sum of conflict counts for: a. Opposing left turn. b. Right turn cross traffic, from right. c. Left turn cross traffic, from left. d. Left turn cross traffic, from right. e. U-Turns. 4. Side swipe conflicts: the sum of conflicts counts for lane change conflicts. Research on the use of this technique has attempted to correlate conflicts and accidents so conflicts, which occur more frequently than accidents, can be used as a basis for predicting future accidents. If this correlation can be verified, conflicts can be used to identify potential hazards and operational deficiencies, and to evaluate traffic improvements. Although the traffic conflict technique has been studied for more than ten years, there are still many components that have not been thoroughly investigated. For example, insufficient research has been conducted on characteristics of conflict involved drivers (those who contribute heavily to conflicts). This research is directed toward A. F ow ‘ u ':' «VJ '.¢J«p _ ..,." i .0 .. ~- |‘_ . p . .- u A-po‘ liv . . v Q 1 I: Q ""2 r. W 5:}; 'Ia u ‘r o ‘ ‘J .- . a 3' ’s ‘ . 5 ‘3 .‘I 1 i . . ‘6. ' ‘ .- . u \ “ \ o «Q F ~ I .Q i 5 the study of the characteristics of conflict involved drivers, and the relationships of these characteristics to the number and type of conflicts. Although the scope of this research does not include the testing of conflict-accident relationships, an accident is simply a conflict where the evasive action was too little or too late, and accidents and conflicts should be statistically related. Glauz and Miglez (53) argued that most attempts to find strong correlations between conflicts and accidents have been unsuccessful for one or more of the following reasons: 1. Not all intersection accidents are reported, and the reported accidents may have happened at different times than when con- flicts were measured. 2. Not all intersection accidents are the types representative of the conflicts measured. 3. Not all accident data used by previous researches was statis- tically repeatable. 4. Conflict definitions have not been precise enough to ensure interobserver reliability, and there have been various inter- pretations of these definitions. 1.2 The Problem The objective of all highway safety programs is to reduce the num- ber of accidents and fatalities which occur each year. Previous acci- dent studies examined such factors as the characteristics of the roadway, the vehicle, and the environment and how each of these contribute to the occurrence of accidents. However, insufficient attention has been given to one of the basic elements of traffic accidents, the driver. 5.. l I Hun .‘ ,. . I" vf\ I '6 It ‘nt;p| “:I'l\ . a ‘co‘v- l n.« a" A L...” 3 d “we: IOIv. "in”. l P ml ‘ n u. 1 Ii... ’ .Y" 'n'rb . :...r‘f -.,‘ = l . ‘ l on. .~ ‘ .T l I I 00 n I s. q i v S b s >“ ‘ O z I" ‘ v - .“‘ 6 Recent studies have shown that there exists a 'pattern' among drivers involved in traffic accidents, and that certain driver charac- teristics are overrepresented in persons involved in accidents. Al- though several authors have looked at the possibility of predicting future accidents using individual characteristics, the literature con- tained little information on the potential for using characteristic 'patterns' as a basis for predicting future accidents. In the literature review, no studies were found of the character- istics of conflict involved drivers or attempts to identify those char- acteristics that are overrepresented in conflicts. 1.3 Objectives of This Study This study was designed to achieve the following objectives: 1. To determine driver characteristics that are overrepresented in conflict involved drivers when compared to all drivers. 2. To determine the relationship between drivers' characteris- tics and types of conflicts. 3. To determine the contribution of each conflict type to total conflicts at urban intersections. 4. To examine the possibility of predicting the number of con- flicts and their types by knowing the mix of driver characteristics. The importance of this research lies in the potential use of the results to establish a new basis for predicting future conflicts, and ultimately, future accidents. ’;nr bkfi I 'l 'ql. I p z m A. 'r-‘c wt :‘l':bb U Q tun-‘- j.-: 1cu: : fl. ' . Jr‘ ‘0- ' U'Ub. I h I ‘~ , b . h:- " ‘n- ‘ c '. ‘GPr; r‘ y~ ‘9-6. p .: at Q ‘. ' § \" V. . . - ya. I'. h.) \ V ,f‘ 6.; ...‘. . ~' u..‘* a . I: ‘V p- \ .- .'d1= In I x , N‘ Q n. .3, a‘ v: ‘- ~‘o‘ ‘ \ ‘U _. ‘ a . ‘ A "‘ "~ 5 -. .é: ‘1‘ . CHAPTER 2 LITERATURE REVIEW The distribution of traffic accidents in time and space, factors which contribute to or are correlated with accidents, and the potential value of traffic control devices in reducing accidents have all been the subject of extensive research. A large number of accident prediction models using various surrogate measures (mostly exposure related) have been used to explain these variations in traffic accidents. In contrast, a relatively small number of published articles have discussed the use of driver characteristics in accident prediction models even though these variables generally are strongly correlated with accident frequencies. The literature review was conducted to determine whether sufficient evidence of a relationship between accidents and conflicts, and acci- dents and driver characteristics,exist. The relationships discussed by other authors was then used to develop the hypothesis to be tested in this study. 2.1 Exposure Measures Intersection accidents account for more than 50 percent of the total accidents in many nations, which gives the 'intersection' a world wide priority in finding accident prediction procedures and remedial measures (3). Traffic accidents are the most direct measure of safety deficien- cies at intersections. However, postponing potential safety improve- ments until an accident history has evolved delays the implementation of accident reduction measures. The lack of reliable accident records, due to incomplete records or errors in coding either important accident 7 -r V! ‘ ll;( 1" 0 ~- .‘ M: vh' . u‘bvli 'l‘ u, . ,— .s . \ a... U H... II .F‘: "'5 1. IA, an I o». "“2 r! I"...‘ D‘fi‘l I -.‘.A .rl‘ ,- "~ '-.. . . A. ' HP‘A 1. in ‘ u. ‘- s- q s \ A ‘. '3» h a \A u- ”s I. ‘ ‘. I"... . lu‘ \ V .. lyi- \ '9 . .. .. ‘ rt‘ 1 8 facts or the precise accident location, contributes to the insufficien- cies in using accidents alone as a basis for safety improvements. Several individuals and agencies have tried to find a solution to this problem by introducing new methods to estimate the expected number of accidents at intersections. Some of the early studies like Jorgen- sen (4) and Hall (5) calculated the expected number of accidents on the basis of the number of vehicle-kilometers within the intersection. How- ever, the risk per kilometer travelled at intersections was not a satis- factory measure, since the number of accidents at intersections is dependent on more factors than volume and the width of the intersection. This lack of success led to more sophisticated methods of predict- ing accidents at intersections based on a combination of measures of exposure. These studies used mathematical functions based on the stream flows passing through an intersection as the measure of exposure, and compared accident occurrences with this calculated exposure. Accidents and flows were compared for many locations and regression equations derived which correlated with the observed accident occurrences. Thorson (8), for example, defined exposure as the sum of all vehicles entering the intersection. He theorized that the number of accidents that occurred at each intersection divided by this sum would provide a basis for comparing the hazardousness of intersections of different types. However, this too was proved to be unacceptable as it does not consider all the factors that effect accidents. Several researchers defined exposure as a probability distribution based on the number of opportunities of being involved in an accident at the intersection. Mathewson and Brenner (6) and Breuning and Bone (7) were among the first to point out the difference between exposure on a road situation, which is related to distance travelled, and - III'U b 97" In I}: an 1 un'd 32"» "vll 9 exposure at intersections, which is related to the maneuvers through the intersection. They suggest that an exposure index at intersections should take into account the flows on each crossing, merging, or diverg- ing path. These studies led to several measures of exposure being used to predict collisions at intersections. The following paragraphs summarize briefly what has been accomplished with these models; the traffic con- flict technique will be discussed in more detail later in this chapter. Early research on exposure was reported by Feuchtinger (13), Raff (14), Thorson (8), and Breuning and Bone (7). Each found that the number of accidents per vehicle was lower for locations with a high measure of exposure than for locations with a low measure of exposure. In each of these studies, however, measures of exposure were found by observing traffic at different locations rather than by taking observations at one location over different time periods. Thus, this phenomenon may be explained by the fact that locations with higher flows generally are built to higher design levels than those with lower flows. Thus none of these studies can claim to have shown conclusively the effect of flow on the ratio of accidents per vehicle. McDonald (15) and Thorson (8) found that accident risks do not vary much between heavily travelled intersections, even though the flows varied widely. For the intersections they studied, there does not seem to be a strong case for the use of the sum of entering traffic as a measure of exposure. Grossman (10) calculated the number of locations where one traffic stream crossed another traffic stream in an intersection, added the flows for each stream, and summed this over all locations and used it as the measure of exposure. This measure was used to compare different .- . at I I‘ - u \ 'fl .4 0 :6 I 'I 9 u r. .' MI. -. “‘2ro; 'v-., ' 'Iaqn - '-d l p A h o .1 '4 a. ’v . “ n _. s .11 4‘ O“ .I F ~4- ... I 10 intersection layouts. However, these models have not calibrated well with accident experience, indicating that not all traffic stream cross- ings are equally hazardous. A second group of researchers used the product of conflicting flows as a measure of exposure. This measure takes into account the number of times vehicles from different directions wish to occupy the same area of road space simultaneously. Surti (11) proposed using this pro- duct from a number of points (depending on type of intersection) where potential collisions may take place as a basis for an accident exposure indes. The essential parameters for this model are: number of con- flict points, ADT, and time. Time is a measure of the period of expo- sure to conflict and can be arbitrary since it only has a relative value. The purpose of Surti's work was to determine an "exposure index" for each of the conflict points to develop a general equation for the entire intersection. Surti (16) followed his initial work by using the same points con- cept and adding merging maneuvers to test the model. He considered, however, the same likelihood of accidents for different maneuvers. Chapman (17) agreed with the use of the product rather than the sum of flows as an exposure measure, but added that the square root of the product may be a better predictor of accident occurrences. McDonald (15) was among the leading researchers to use the square root of the product of flows as a measure of exposure. He studied intersections between a divided highway and a cross street and obtained a relation between the number of accidents per year 'AV', the average daily traffic entering from the divided highway 'Vd', and the cross street traffic ll. Vc . ’( v D» ‘r e‘_‘ 1r an. 0r: fiff‘. v.p.. h :p; ... ‘v .p. n ,‘n.. ‘1 ‘C. A : A. ‘.‘ 11 A = .OOO783 vd'455 vc°633 Thrope (21),in an investigation of accident rates at signalized intersections,related the number of accidents to the square root of the product of entering traffic, but did not justify the use of such a measure for signalized intersections, where the two traffic streams are separated by time. Webb (20) developed graphs (for urban, semi-urban, and rural sites) for signalized intersections, showing the average expected number of accidents per year according to a non-linear function of the average daily main road and minor road traffic. This study, like those of Raff (14) and others, was based on single data points from many locations rather than many data points for a single location. Thus the cause- effect relationship cannot be determined. Gwynn (12) conducted one of the most comprehensive studies of the relationship between accident rates (or involvement) and hourly volumes. Although this was a highway (not an urban intersection) study, it demon- strated that accident rates and traffic volumes do not parallel each other. This may be related to the fact that there is a ten- dency for a disproportionately high accident rate to occur during the hours with the lowest volumes (early morning hours). This is followed by a decrease in accident rate with an intermediate traffic flow, an in- crease in accidents with an increase in volume up to a point of conges- tion, and then the accident rate again falls off, even though the volume of traffic continues to increase. This study illustrated the fact that a stable relationship between exposure and accidents probably does not exist unless other conditions are eliminated from the analysis . .np :‘ :8} ‘ ‘v d \ , \uvu .‘g . A; .' all .0. HI — 1 r1 ; v. :"e 3,. :.... r. 'Ouuu‘n' .,‘ :V“ i 1.C Q... I.e “‘9. , . 12 (such as a high number of drivers under the influence of alcohol between midnight and 5:00 a.m., congested traffic). Thus, it is not surprising that attempts to develop accident predictive models based on exposure measures have not been successful. 2.2 Accident Patterns Intersection accident patterns have proven to be very stable over time and in many different traffic situations. Intersection accidents account for a high and relatively constant percentage of total accidents in many nations. Smeed (3) reported that in Great Britain over 50% of all injury accidents in 1970 occurred in urban intersections. In France, intersection accidents accounted for more than 45% of total injury acci- dents in 1972; in Denmark 38% in 1973; in Ireland 44% in 1973; in Holland 44% in 1971; and in the United States an average of 41% of all reported accidents between 1971 and 1979 occurred at intersections. Accident statistics published by the National Safety Council show little variation in the number of accidents when classified by type of collision over a period of ten years (see Table 1). The statistics also show a constant pattern in the type of accidents, with angle accidents having the highest frequency, followed by rear-end accidents and left turn accidents. The effect of intersection geometry on accidents has also been studied extensively. In one two-year study to find relationships be- tween motor vehicle accidents and the geometric and traffic features of highway intersections, Stanford Research Institute analyzed 558 inter- sections (57). The results were similar to the National Safety Council statistics as they found that angle accident rates are the highest for all types of intersections followed by rear-end accidents and weaving 13 and turning accidents. Table 2 presents the percentages of collisions by type for different geometrics and traffic control at intersections. Thus, if a particular intersection is experiencing an abnormal accident history, some other factors must be contributing to these accidents. 2.3 Driver Characteristics There have been fewer studies reported on the relationship between accidents and the driver. Yet the characteristics of drivers, their capabilities and their driving skills are important factors that affect accidents. The National Safety Council concluded that an average of 88 percent of all accidents at urban intersections were a result of im- proper driving (58). Table 3 presents the percentages of different faulty driving techniques that caused accidents over a ten year period compiled by the National Safety Council. In an attempt to find an easily measurable, yet reliable, surrogate measure of an accident prone driver, Spicer (26) examined the difference in car following behavior (as measured by the time gap between vehicles) between drivers with accident records and the general driving public. This study concluded that the smaller the following gap, the higher the probability of prior involvement in accidents. In the same study, Spicer also analyzed the relationship between intersection accidents and drivers age. He concluded that a crossing maneuver involving a young driver on the major road and an older driver on the minor road is more likely to result in accidents than any other combination of age groups. Siebrecht, Schumacher, and Lauer (22) made a comparison of two samples of drivers. The first was taken in 1950 and the second three years later from the vehicle license files in Iowa. A follow-up mail mnvey was sent to all individuals in the samples. The survey included I. \‘.Ill.|z lI Ill 1|. 1“. III, 1|!" l' I: ll" I‘ll II .N. CJEQT...) 3?...LC>< M.\mu— m~\fi.— h\$~ C\3— n\3~ Qkfimu N\3~ -N\3~ -KQ~ ~ CN 3. m L 1.0% fires: UO_\ KO C§XK no ..I..~.cc-u- ooooooooo u.. ..a..‘ u ..l\ ,u. .~.\v .m on mpasmm as» mmuzomc cows: cowumpau_mu mucmwcm> new mmmcw>m ms“ :_ cmu:_u:w mw mmczmww mmmgp to mac apco .cmmmwn m=_mn we zuwpwncnoca one mmoauocucw gown: mcowumwcm> o: sogm mmofi .flmmH .ouoa :_ =m>wm mmczm_m on» a 14 mucoamm mmpuwcosu=< uwmcmch mumum :0 women Fwoczou xamwmm Pacowumz .muomm acmuwou< "mucaom E. m; 3 o.~ : E Z 3 ms to no to 29:0 .2 mm 8. Z 3 j 3 Z Z 3 E 3 .3 3 2323 25 £3 25 mm 8. mm. to so m... no 2 a; .3 3 i: 3 2323 9:8 58 mm mm. o; .2 m; o; no as 9o 3 3. Two no 233 :< mm 3 m8 ~.m m8 2 3 2 3 E 3 .3 3 888m 25 mm a. ma ~.m ma 3 3 2 ma 2 3 1: m.m 2323 25 .EE 95 mm. mm.) M: S 2 on 3 3 Z 3 9m 9m o.m .2323 8.8 58 w H.N m.o~ a.mfi ~.- m.- m.- ~.m~ ~.c~ m.mH a.mfi a.- ¢.~fi apmc< 8a m=_cau=u 3 n: 8.8 “.2 I; 53 98 8mm 9: 98 98 gm mcwwfiwmwfiafimcncakes” a 85:2, 82a: 22 22 E: 23 £2 :2 23 a2: :2: 35 as; 28.82 to 8.: mcomuummcmucm sang: um macmuvoo< mpovsm> Lopez 03H .H m4m
ogo
m.e m.¢ m.¢ ~.o m.m m.m m.m m.m o.¢ o.¢ Pmcmwm umugmmmcmwo
m.cm c.0w a.mm a.mw H.Hm a.mm m.om m.mw a.mm w.mm mcw>wco cmaocasm quoh
mm ms mm mm mm em mu us an an mcw>Pgo cmaosaEH mo vcwx
mmzawczomh mcw>wgo zupzmu mo mmmwucougmq
.m ”Sm:
I‘-’
(-1
' v
.T" ‘A
n
‘1.
9"»..1.
63””‘r
"' ‘i.
U
. :q‘
I, g. ‘
lg. 9
‘ ‘ 3
"-aw,‘
‘5 Q.
‘1].
“'5?
'1
‘2. ..
“ a):
l
hr“.
1
17
data on the methods of learning to drive, mileage driven yearly, educa-
tional level, and other information. The authors concluded:
1. Violations and accidents are closely correlated and records
of violations give a fair index of potential accidents.
2. About one person in four has an accident record and the same
ratio holds true for violations.
3. A higher percentage of accidents involve men under 30 than
other age groups.
4. The hypothesis that accidents are distributed equally among
age groups is rejected.
Lynn (23) indicated that there is evidence to support the fact that
accidents do not occur completely by chance but rather according to a
demographic pattern. For example, in 1970, males between the age of
20 and 24 comprised 10.1% of all males in Virginia and 5% of the total
population of the state. However, this demographic group comprised
20.7% of all male accident fatalities during the same year. Lynn gave
other examples and concluded that certain demographic variables can
be used to predict accident patterns.
Waller and Koch (27) evaluated accident data and types of driving
exposure in terms of sociodemographic variables. They found that more
highly educated individuals drive more than those with less education
but there was no significant difference in the accident rate.
Plez and Schuman (28) found that among young drivers, those who
were alienated from the educational system (dropouts, poor'grades) had
more accidents and incurred more violations.
Baker (29) found that more males and more young drivers were at
fault in fatal accidents and that lower economic class drivers were also
an
olf‘
in i
I)
V
Turf:
l
u ul-
s-‘
uu'u ,
..
.123 F
.
t...
. .1-
-
‘vv ‘0‘
1w q
n. J
Ca
.1 .:
u... n.
a...
..V‘:[
U I. ‘
.1"
.‘
18
more often at fault.
McGuire (59) compiled data on 2,799 drivers in an attempt to iden-
tify personal characteristics of drivers that possessed a causal rela-
tionship to accident-producing behavior. The study reported a multiple
correlation between characteristics of drivers and accidents, with the
highest observed simple R values corresponding to age (.38), occupation
(.38), and income (.30), among 29 other tested variables.
All of these studies indicate that there are dirver characteris-
tics related to accidents which if known in advance, could be used in
accident prediction models. However, many of these characteristics
can only be determined through interviews with drivers involved in acci-
dents or stopped at a survey station. Unless the distribution of these
characteristics can be determined, it is not possible to determine the
relative hazardness of intersections.
2.4 Traffic Conflicts
Traffic conflict analysis has been proposed as a method of addres-
sing these concerns. Because traffic conflicts are measured on-site,
variations due to different driver characteristics or changing volumes
can be captured in the data. The adoption of this technique, however,
has been limited by the lack of research indicating that these conflicts
are strongly related to accidents.
The traffic conflict technique was developed by Perkins and Harris
(30) to be used as a dynamic tool in evaluating accident potential and
operational deficiencies of intersections. Using this technique, it
was.hypothesized that intersections could be evaluated with no accident
data. Instead, the safety evaluation would be based on the observed
corlflict frequencies at the intersection..
l‘)
a
a...
.-
V'Iv‘
III
a O
O-A'.
I1. “
19
Since this initial paper on the traffic conflict technique, re-
searchers have directed their studies in two main areas:
1. Studies concerned with finding relationships between accidents
and conflicts.
2. Studies using the technique to identify potential hazards and
operational deficiencies at intersections, and to substantiate
the value of operational traffic improvements by measuring a
reduction in conflict frequency or rate.
The Federal Highway Administration (FHWA), in cooperation with the
State Highway Departments of Washington, Ohio and Virginia, evaluated
the traffic conflict technique in an effort to find statistical rela-
tionships between accidents and conflicts. Their objectives were also
to field test the technique to determine if conflict data proVided in-
formation advantageous to pinpointing the need for safety improvements
at intersections. The studies concluded the following (31):
1. The hypothesis that conflicts and accidents are associated is
accepted.
2. Safety deficiencies at intersections can be described more
quickly and reliably by using the traffic conflict technique
than by conventional methods.
3. The traffic conflict technique costs less than conventional
methods for finding intersection deficiencies and operational
problems, and is also useful in analyzing spot improvements
(before/after studies).
Paddock and Spence (32) developed prediction equations to
estimate the number of accidents expected over a two year period using
ccwrflicts (and additional parameters) as components of these equations.
Equaltions (1) and (2) were developed from their data for signalized and
Fllv
Nd-
20
non-signalized intersections respectively:
AP2Y = 1.1653 + 11.6345 (ADT) - .0503 (CPT)
- .0321 (RROPP) + .0387 (OCP02) (1)
+ 0.285 (TTOPP) - .02255 (OPOPP)
AP2Y = 0.36 + ADT (22.3568 + 17.773 (SPLIT)
- 36.7045 (ADT)% - 1.6785 (SPLIT)2 (2)
+ 18.2544 (ADT) - .0264 (OPOPP) + .8385 (OPCON)
where
AP2Y = Accidents predicted in 2 years
ADT = Average Daily Traffic in 10,000's
SPLIT = Ratio of the sum of the cross volume to the approach volume
CPT = Total of conflicts/10 opportunities
OPOPP = Opposing conflict opportunities (opportunity is a term used
to denote traffic volume)
RROPP = Rear-end conflict opportunities
OPCON = Opposing conflicts
TTOPP = Total conflict opportunities
OCP02 = Square of opposing conflicts/10 opportunities
The regression equations had a prediction error for all points (611
locations, 220 signalized, and 391 nonsignalized intersections) of i 4.2
accidents/year at a 95% confidence level.
Glennon (33) pointed out that these equations are based more on
traffic volume counts (ADT, SPLIT, OPOPP, RROPP, AND TTOPP) than the
conflict counts OPCON, OCP02 and CPT). In fact, in equation (1), the
rnunber of accidents is negatively correlated with conflicts (CPT and
OCP02). Thus, the use of these regression equations adds little to the
expcnsure based models discussed earlier.
21
All studies of conflict rates did not reach the same conclusion.
Cooper (37) and Glennon (33) found that conflicts are very much volume
dependent and could not account for differences in accidents when cor-
rected for exposure. They argued that studies which concluded strong
association between accidents and conflicts are either brief and not
statistically reliable, or have disregarded important parameters that
may explain the association (such as site parameters).
Not all conflicts are, in fact, near accidents. Drivers vary in
their response to other vehicles, some braking when it would not be
necessary to avoid a collision, for example. In an attempt to gain
reliability, Spicer (34,35,36) classified conflicts by severity (routine,
moderate, and severe) and attempted to find relationships between severe
conflicts and intersection accidents. He concluded that using only
severe conflicts improves the prediction equation. However, accidents
at only six intersections were included in this analysis.
Hayward (53) used Time Measured to Collision (TMTC) as another
method of identifying conflicts. TMTC is defined as "the time required
for two vehicle to collide if they continue at their present speeds on
the same path." This too improves the reliability of prediction equa-
tions be reducing the variation due to a driver related response in
the equation.
A second type of research used the Traffic Conflict Technique to
identify potential hazards and operational deficiencies of highways
and intersections. Findings from these studies were used to justify
arui substantiate thevvalue of many types of operational traffic im-
prxivements. This use seems to have had more success with the Traffic
Conf‘l ict Technique.
Among the reports indicating successful use of the Traffic Conflict
up!“ ‘
.~ ‘ "'
n
O
(\J
-,
”.1.
Qa-
t-a -
‘I
n
“A...
5'-‘
22
Technique in practice are:
‘1. Before and after improvement studies (38,40,41,42),
2. Construction zones (43),
3. Freeway acceleration lanes (44),
Freeway lane drops (45),
Freeway weave areas (46),
01014:-
Intersection improvements (42,41,39,40,47,48),
7. Interchanges (49),
Oversize loads (50),
9. Pedestrians (51),
10. Signing and traffic signals (38,39,48,43,40,52).
The conclusion of all these studies was that traffic conflict fre-
quencies or rates are sensitive to changes in geometry or traffic control
measures. Thus, traffic conflicts appear to possess one of the proper-
ties necessary for a valid surrogate measure - that they respond to
changes in the independent variable being tested. The remaining ques-
tion is whether the conflict response maps reliably into an accident
response, and thus can be used to make an accurate prediction of future
accidents.
- nag...
c¢“bu‘ u
.1
O:{§3l.
"4' H.»
:II( ”All!
0,... ..
:7f’;:1
a. |
- .H~4
1"1‘
'~—l
[\J
(.k.’
CHAPTER 3
METHODOLOGY
As the literature review indicated, many past studies have
suggested a statistical relationship between a change in the
number of conflicts and accidents. In addition, research (59) (22)
has identified the existence of an overrepresentation of certain char-
acteristics among drivers involved in accidents. Therefore, it is a
reasonable hypothesis that there also exists a relationship between
conflict involvement and driver characteristics. This hypothesis was
tested by studying the characteristics of conflict involved drivers-
and comparing these with the general driving public to see if they are
significantly different.
3.1 Location
The experiment was conducted in the Eastern province of Saudi
Arabia in the city of Dammam. Dammam is about 18 kilometers from Dhah-
ran, the home of the University of Petroleum and Minerals.
This location was chosen based on the following considerations:
1. The University's interest in this type of research and their
willingness to fund the data collection phase.
2. The cooperative attitude of the Dammam police agency.
3. The availability of facilities and students to assist in the
data collection activities.
3.2 Choice of Intersections
Due to the extensive construction and development program in Saudi
Arat>ia, many of Dammam's streets were under construction. The task of
23
M" 'he I
t
“-‘
. t I\
.
I
o
. ‘5-
‘. try.“
"“V Kr
::'biy,
Q
in
:.r :.
a,
. 5 A
. '. \-
‘.‘ H
-
u
. 4
Q.‘e “
.' .
' u
. .
.‘\ .-
. V.‘
.
3"}
..\\
1“.‘
.~
“A ‘
.~ I
.
..-!‘
"'51
‘.a 1
.'_~
i ~ ‘
an! c
S
75:.
._“
24
locating suitable sites to undertake the experiment was a difficult one.
One day was spent touring the city, and nineteen intersections suitable
for the experiment were identified. Five of those were signalized and
fourteen were controlled by stop signs.
The city was then divided into three radial sections extending from
the central business district. Two intersections were chosen randomly
from each of the three sections, to serve as data collection sites. One
of those was disqualified after the second visit and a similar inter-
section in the same sector was substituted for that intersections.
The intersections chosen had no obstructions to the traffic stream
for at least 300 feet from the center. They all were properly signed,
paved, and painted. The signalized intersections were lighted at night,
and there were no observed driver distractions.
The Average Daily Traffic (ADT) at these intersections ranged from
2000 to 26000. The two signalized intersections had ADT's of 21000 and
26000, with the unsignalized intersections carrying lower volumes.
3.3 Observer Training
Special attention was given to the training of conflict observers
for the data collection phase. Fifteen candidates were selected as pos-
sible conflict observers, and all were required to attend four training
sessions designed to provide them with an overall understanding of traf-
fic law, traffic safety, and traffic conflict techniques. The sessions
consisted of movies, slides, video tapes and lectures. At the end of
the last session, the candidates were taken to different intersections
arm! each was asked to conduct volume counting, interviewing, and conflict
observations.
Three of the fifteen participants were chosen as conflict observers,
01h
25
with the others assigned to volume counts or interviews.
The choice of conflict observers was based on:
1. Their background in traffic engineering. (Two were seniors,
one a graduate student, all majoring in Transportation.)
2. Their attendance and participation in all of the training
sessions.
3. Their understanding of the definition of traffic conflicts,
and their ability to distinguish conflicts in the field.
3.4 Experimental Design
Each intersection was observed for three consecutive working days
during which two different samples were collected.
1. Intersection p0pulati0n sample (subsequently referred to as
sample 1).
The first day of observation at each intersection was used to
collect data to identify the characteristics of the general driver
population using the intersection. Three interviewers were sta-
tioned about 200 feet upstream from intersections on each approach,
and for a period of twenty minutes, every tenth vehicle was stopped
and the driver interviewed. The interviewers then relocated on a
second approach, and the process was repeated until samples from
all intersection approaches were collected.
2. Conflict involved drivers sample (subsequently referred to as
sample 2).
The second and third days were used to collect data on con-
flict involved drivers. Conflict observers identified those
vehicles that caused a particular conflict, and then signalled con-
flict type and make and color of that vehicle to interviewers who
(JD
. a .
.~ \"
.1: vs
—
.1
'r
up
n.
‘9
1'.
. -~
.
~0-
I
D'-
I "
'7.
.
.
‘2
u
, .
. 1 i
'b
‘A
V O
- h C
‘ l
26
stopped and interviewed the driver. Table 8 gives the number of
samples collected at each intersection.
3.5 Schedule of Activities
The data collection phase of this project took three weeks. The
first week was devoted to training participants and selecting adequate
sites, and the following two weeks were devoted to data collection.
A total of six hours of three consecutive working days were re-
quired to collect both the data for sample 1 and sample 2 at each inter-
section. The first two hours of the first day were used to collect the
general driving population sample of each intersection with a twenty
minute period designated to collect data for each approach. The same
two hour period of the following two days was used to collect data on
conflict involved drivers.
A minimum of seven participants for each observation period were
needed to collect the required data. The participants included one
conflict observer, one volume observer, three interviewers, and the
two policemen.
Data collection on conflict involved drivers was designed so that
data on each traffic movement is collected one time (for twenty minutes)
during the total four-hour period of two consecutive days. This was
accomplished by first stationing the conflict observer on the south
approach and the interviewer on the north approach. The conflict ob-
server would record all conflicts on the south approach and identify for
the interviewer the subset of those drivers traveling north. After
twenty minutes, the interviewer was relocated on the west approach,
iwher'e the subset of conflict involved drivers turning left from the
Stnrtliapproach were interviewed. Finally the interviewer was moved to
l
a
o ‘O
a
,, au'
.
.-.a E
10 u
u v
o
f
5.;
p
“ UV
A "I
l. J)
27
the east approach to obtain the last data set. In the second hour of
observation, the conflict observer relocated to the north approach and
the process was repeated. The east and west approach were then used in
the data collection phase on the second day of data collection for
sample 2.
This accounted for all traffic movements with the use of only two
police officers. Simultaneous collection of all movements was not pos-
sible because the use of communications devices by individuals other
than policemen is banned in Saudi Arabia.
l' A .
. A
to vb-
.—
It“
uU
Q
=4: r:
o . 5|
' 4.
WV..- 4‘
4‘}
{Stung
.. I. .
"its .4
' U
CHAPTER 4
DATA ANALYSIS
4.1 Observer Reliability
Observer reliability is essential to this experiment, since the samples
are relatively small, and some analyses required the aggregation of data
from different observers. To determine whether data from the various ob-
servers could be combined, a test was designed to determine if observa-
tions made by different observers were significantly different.
Two observers simultaneously recorded conflicts on eleven different
occasions. In seven cases observers one and two recorded simultaneous
data, and in four cases the observers were numbers one and three (see
Table 4).
The Pearson's xztest for goodness of fit was used to determine
whether the resulting conflict counts were observer dependent. This
test is used to study data which can be grouped into multiple categories
across at least two dimensions. The categorization can be either scaler
or vector level (contingency tables or cross tabulations), and is used
to determine whether the categories are independent or whether certain
levels of categories tend to be associated.
An r x c contingency table with columns E A, rows E B, and
cell frequency n is used to test the independance of A andlB by calcu-
lating the value of x2 (60, 61):
X2 = Z ("13' 7 £13)
all Eij
rc cells
where:
28
V .
out
kit
7
‘1‘.
.xv
N»
29
"ij = frequency of observations in cell AiBj
n. x n .
Eij = expected frequency of A, Bj = -19—N——91-if the charac-
teristics are independent
ith row total, or frequency of Aj
3
ll
i0
"oj = jth column total, frequency of Bj
n : Z nio = Z noj
The null hypothesis of independance is rejected at a specified
level of significance (0), if the calculated x: is greater than the upper
tail of a x? distribution with degrees of freedom = (r-1)(c-1).
The null hypothesis for observer reliability was; Ho: there is
:10 difference in conflict counts by type of conflict between observers.
'The null hypothesis was accepted at the a = .1 level of significance,
611d thus data from different observers could be combined for further
analysis (Tables 5 and 6).
4. 2 Representation of Driver Characteristics
Since data on the characteristics of conflict involved drivers
(sample 2) and the data on characteristics of the general driving public
(sample 1) were collected on different days, an analysis was conducted
two 'test the assumption that "drivers passing through an intersection at
a. Stoecific hour during working days possess similar characteristics to
'd?‘i\/ers passing through that intersection at the same hour of other
WCJr‘king days."
Intersection populations were sampled for each signalized inter-
SeCtion on the first and third working day during the specified hours
0f data collection. The null hypothesis that there is no difference
b‘i‘tWeen characteristics of the population during these days was tested.
h
\-\ol
1
5...)
Qovu.
o‘.‘
\ \‘§o\\
3O
8N NN NN NN HN 3N NN NN NN NN NH NN NN NN NN NN NH HH NH NH N oH HNHoN essHoN
N N a o N H a a N N N a N N N N H H H H H H NNHN Lo HNNNHN NNHNNNONHN
N N a N N N N N N N H H N a N N o o o o o o =N:N-N
o N o H o o H H o o H H o o o o o o o o H o HNNHN EOLN UHNNNNH NNoaN
N N N H o N N N o N o o H H o o o o o o H H HNNH socc UHNNNLH NNOLN
o o H o H o o o o o o o o o o H o o o o o o HNNH acct caaH HNNN
N o N N o o o o N o N N H H N N H H H o o o HNNHL ENLN ccsH HNNNN
H o o o o o o o o H o o o o o o o o o H H N =L=H HNNH NNHNoaao
N N N N N N N N N N N N N N a N N a N N N a NNNNNU NNNN
H N N N N N N N N N N H N N N N N - N N N N N «HUHNms onN
N N o o N N N N N N H o N N N N o o N o o o NOHHUNLHN aeNN .caaH HNNHN
H H N N N N N N H o H N o o N N N H N H o o coHHuaLHN NENN .=L=H HNNH
H N H N H N H N N H N H N H N H N H N H N H .62 LN>LNNNN aaNH
NHzoN Hmaz NHLoz HNNN HNmz NHLoz HNNN NNNON NHNON HNNN Hmaz NaNoaaa< HaHHNNON
Hza N-Nv NEHH Nza N-NV asHH N28 N-NV aeHH
Na: NHN .mHasNN NNN
N :oNHUNNNmucu
Nae NHN .NHaeNN NNN
H :owuuomcmuca
Noe vcm .NFNENN ecu
H coNHUmNcmucH
.NHNNNN HNNHN=NN NNNNNNNN .4 NHNNH
TABLE 5.
31
Conflict observation, reliability test for observers 1 and 2.
Total Observer Counts
(Seven simultaneous counts)
Total
Conflict Type 1 2 rows
Right Angle 21 (20.36) (19.64) 40
Rear End (49.87 (48.13) 98
Turning O (30.02) 9 (28.96) 59
Side Swipe (42.75) (41.25) 84
Total Columns 143 138 281
Hypothesis:
0 type of conflict.
x2 calculated:
at 3 d.f.:accept H0.C
TABLE 6.
T
H : The frequencies counted by observers 1 and 2 are similar for each
1.16 < x2 distribution
7.81
Conflict observation, reliability test for observers 1 and 3.
Total Observer Counts
(Four simultaneous counts)
Total
Conflict Type 1 3 rows
fight Angle 6(26. 20) (25.80) 52
:ear End 4.(55 42) 56 (54.58) 110
.1:ir711n93(22'17) 21 (21.83) 44
Side Swipe 30 (29. 22) 28 (28.78) 58
Totai Columns 133 131 264
———HH___
Hypothesis:
*1C,: The frequencies counted by observers 1 and 3 are similar for each
type of conflict.
x: calculated =
.181 < 7.81 at 3 d.fJ accept H0.
"H. '
1'
»
u
‘\
.7
.
TIM
32
A xz-test was run to test for differences between population char-
acteristics on the first day, and population characteristics on the
third day of data collected for the same intersection. The sample sizes
were (n; = 99, n; = 27) for population samples on the first and third
day respectively. Results of this test are summarized in Table 8, which
indicates that there is no significant difference between drivers of the
first and third day when age, nationality, trip purpose, occupation, and
income are tested. The null hypothesis for these characteristics is
not rejected at a = .1.
When the driver characteristics "miles driven per year" and "number
of accidents in the last two years" were tested, they were found to be
significantly different at the a = .1 level. About 58% of the drivers
()n the first day drove < 13000 miles per year compared to 30% of the
cirivers on the third day, and three times as many drivers have had two
(or: more accidents in the past two years on the third day than on the
fLirst. This implies that there is more variance in these characteris-
‘ti<:s than in the others tested, or that a different group of drivers use
the road on different days.
4.N3 Characteristics of Hazardous Drivers
Nineteen characteristics of both conflict involved drivers, and
the general driving public were collected in the field (see Appendix A).
The average driver is 29 years old, with an income of about $1100 U.S.
a. "nonth. Almost sixteen percent of the drivers did not possess a valid
dr‘i ver's license. One of every three drivers had been involved in at
1east one accident in the last two years, and one of every two drivers
rui<1 been involved in at least one accident since he started driving.
37 Xty percent of the drivers had completed a driving training course.
33
TABLE 7. Chi-square test of population assumption.
x
Characteristics of drivers a =T.1 xc
Age 2.706 .122
Nationality 2.706 2.015
Trip purpose 4.605 4.278
Occupation 4.606 1.717
Miles driven/year 2.706 5.562*
Number of accidents in past 2 years 2.605 9;§49_
Income 4.605 2.271
* H0 is rejected for underlined characteristic.
HO: There is no difference between characteristics of drivers on
different days of observation.
TVXBLE 8. Sample sizes for studied intersection, number of drivers
interviewed in each sample for each intersection.
Sample 1 Sample 2 Double
(Intersection (Conflict conflict
population involved involved
Intersection sample) driver sample) drivers
1 Signalized 56 66 --
2 Signalized 70 64 --
3 Unsignalized 19 41 34
4 Unsignalized 20 _ 23 8
5 Unsignalized 24 51 28
6 Unsignalized 23 51 10
(Total 7
l\ll intersections 212 296 --
S i gnal i zed 126 130 --
) lJtisignalized 86 166 80
\
'L‘vJ'
..Q‘
1“ pi
n‘u I
rar-
.u-u
V
8
n
‘r' or
o H,-
a
'L’JF‘
- v' ‘
- .
:q'tr
"d'
1
l
u
I
I
C
u .
A.
‘- .“
"Ubu
I.
«We
5-
‘U1
34
About 40% of all drivers were professionals, followed by students (36%)
and all other occupations (24%). Two of every three drivers were Saudi
Arabian nationals.
Determination of characteristics of conflict involved drivers
In order to identify the characteristics of conflict involved
drivers, it was hypothesized that conflict involved drivers possess dif-
ferent characteristics from that of the general driving public. The
testing of this hypothesis was done in two steps:
1. By comparing the intersection population sample to that of
conflict involved drivers.
2. Statistically testing the significance of the characteristics
that appeared to be different from step 1.
'The following paragraphs will discuss the findings of step 1, while
section 4.4 will discuss the findings of step 2.
The characteristics of conflict involved drivers (sample 2) were
ccnnpared to those of the general driving public (sample 1) at both sig-
ruilized and non-signalized intersections (Table 14), to determine whether
haxzardous drivers can be identified by these characteristics.
Driver characteristics overrepresented in the conflict sample at
bcrth signalized and non-signalized intersections include the age groups
< 25, the age group > 46, Middle Eastern (but non-Saudi) nationality,
Occupation listed as a salesman, craftsman, student, or professional
dr‘T‘ver, has completed 9 years of schooling, drives an average of 11,000
mi Tes per year, has had two or more accidents in the past two years, and
has attended a local driver training school.
On the other hand, safe drivers (those underrepresented in the con-
Fl'T<:t samples) are of the age group (26-45), of Saudi or Western
35
nationality, have a professional occupation, college education, have had
no accidents in the past two years, an income of $1800 U.S. a month, and
had driving training outside the Kingdom.
Several plots (1-5) were used to show the representations of dif-
ferent age groups in conflict involvement, in order to determine whether
a certain age group contributes more to conflicts than other age groups.
The graphs show that drivers of age group (5_25 years of age)
are overrepresented in conflict involvement, and that drivers of age
group (> 45 years of age) were slightly overrepresented in conflict
involvement.
These observations support the hypothesis that there is a positive
relationship between conflicts and accidents, as young drivers and
drivers with an accident history were significantly more likely to be:
rwapresented in the conflict sample than in the sample of the general
toublic (see Fig. 6).
4 .4- Test for Driver Characteristics that Explain Conflict Involvement
Since most characteristics of the general driving public are not
significantly different across test sites and samples, these data were
Dcnoled, and x2 tests conducted to determine whether the differences in
characteristics identified in the previous section are statistically
Significant.
The frequency with which characteristics occur in the general popu-
1ation were compared to the frequency with which these same characteris-
tics occur in drivers involved in conflicts. This comparison was con-
duCted for data from all intersections, data from signalized intersec-
tiOns only, and data from unsignalized intersections only.
The sample size for each intersection is given in Table 9, and the
36
‘
6 4 9... Nu mermv 4 m2
1 1 1 1
umuzmmmcamccm>o Ncm>wco to pcmucma
1‘
no N» no 8 02
1 1
umucmmmcamccmucs Ncm>wco to pcwucma
om A
omnmv.
mvnfic
ovuom
mmuHm
omuom
mm v
Age
Representation of driver age in conflict involvement.
Fig. 1.
37
6 A.» 0 more 4.4 Md
1 1 1
12‘
cmuemmmcamccm>o Ncm>wco No ucmocma
4
9... A... no 8 02
1 1
umpcmchamccmuc: NLN>NLQ mo ucmocma
cm A
mvnflv
oeuom
mmuflm
omnmm
mm v
Age
Representation of driver age in turning conflict
sample.
F519. 2.
38
1!
6 A... 2. mu 80 4 2
1 1.. 1.. 1
cmpcmmmcamccmso Ncm>wco No Hcmucma
A
2 4 no mo 0 2
1 1
emucmmmcamccmvcz Ncm>wco No ucmucma
om A
omaoe
mequ
oesom
mmuHm
omuom
mN v
Age
Representation of driver age in right angle conflict
sample.
Fig. 3.
39
‘ ll ‘1
No A» mu no ,0 n4 ma
.1 1 1..
12+
umucmmwcamcgm>o Ncm>Ngo mo ucwucma
4
no 4 6 mo mum/H
1 1
umpcmmmcamctmccz Ncm>Hco to ucmucma
cm A
omuov
mvafiv
oenom
mmufim
omamm
mm v
Age
Representation of driver age in side swipe conflict
sample.
Fig. 4.
4O
1‘
6 4 W. W 80 4 2
1 1. 1. l.
umucmmmcamccm>o Ncm>wco No ucmucma
‘ 11
9. m. .b no my A2
11
umpcmmmcamccmvca NLm>HLo to pcmucma
cm A
menav
oqnom
mmuHm
omnom
mm v
Age
Representation of driver age in rear end conflict
sample.
. 5.
1'9
F
41
Population Sample
Conflict involved drivers
with 2 or more accidents
in last 2 years
DI
'U
C)
>
I— U,
OQ’L
>Lt‘U
:20.)
.— >5
1323“ 40
u— H
F050
2&5
,—
o 30
US:
44'!-
“-‘r-
0319 20
«Hm:
588
u)..— 10
L-t-U
GLO
QD<
L“ 3 3 N
1—4
N I I I <1'
(‘4 to '43
V N 0") A
Age
Fig. 6. Age of conflict involved drivers with 2 or more
accidents.
42
test results are shown in Tables 9, 10 and 11.
These results indicate that there are significant differences in
the characteristics of conflict involved drivers and the general driving
public.
When all intersection data was combined, the null hypothesis of no
difference was rejected (at a = .1) for only nationality and driving
schools (Table 9). When using only signalized intersection data, the
null hypothesis was rejected (at a = .1) for age, nationality, occupa-
tion, driving schools, miles driven per year and income (Table 10).
These results suggest that there are differences between the general
driving public and conflict involved drivers at signalized intersections.
Some of these characteristics may be useful in predicting accidents at
signalized intersections.
When using only unsignalized intersection data, the null hypothesis
of no difference was rejected (at a = .1) only for driving schools
(Table 11).
The fact that the null hypothesis is not rejected for nearly all
tested characteristics at unsignalized intersection is due (in part) to
the fact that nearly all drivers are conflict involved drivers. This is
true because a violation of a traffic control device is considered a
conflict, and nearly all drivers failed to observe the stop signs at
these intersections. (An average of 80% of all conflicts observed at
unsignalized intersections were right angle conflicts, which include
control device violations.)
In an effort to more clearly identify conflict involvement at un-
signalized intersections, the test was repeated with Sample 2 including
only those drivers involved in at least two conflicts. Using these
data, the null hypothesis was rejected (at a = .1) for age, occupation,
43
educational level, and number of accidents in the past two years (Table
12).
These results indicate that including violations as a conflict
simply tends to mask real differences between hazardous drivers and the
general driving public. At signalized intersections, like unsignalized
intersections, those drivers involved in conflicts have significantly
different characteristics than the general driving public.
The age of drivers was found to be significantly different at
a = .1, at both types of intersections. Thus, age appears to be the
variable with the greatest explanatory power among the nineteen charac-
teristics tested.
This result is consistent with findings by Lynn (23) in which he
concluded that age is the variable which best explains differences in
accident involvement, and McGuire (59) who reported that age had the
highest simple R value (.38) among all characteristics correlated with
accident involvement.
Drivers of age groups (< 18, 19-25, and > 46) were overrepresented
in conflict involvement at both intersection types. The highest in-
crease in representation found was for age group < 18 among all age
groups. The age group (26-45) was underrepresented in conflict involve-
ment at both types of intersections. Drivers with a middle eastern
nationality (but non-Saudi), those with two or more accidents in the
past two years, students with 9 years of education or less, and drivers
with an occupation as salesman, craftsman, or professional driver were
also overrepresented at both intersection types.
On the other hand, drivers with a professional occupation, those
from a high income group, and drivers with no previous accident history
were underrepresented at both intersection types.
44
TABLE 9. Chi-square test for driver characteristics in Sample 1 vs.
Sample 2 using all intersections.
X2
T 2
Characteristics of Drivers at a = .1 X
Age 6.251 4.605
Nationality 7.779 12.013
Occupation 10.645 8.200
Educational level 7.779 6.076
Driving schools 4.605 11.938
Miles driven/year 4.605 1.993
Number of accidents in past 2 years 6.251 1.112
Income 4.605 3.217
Drivers license availability 4.605 .649
0
combined.
TABLE 10.
Sample 2 (signalized intersections only).
H : There is no difference between characteristics of population
drivers and conflict involved drivers at all intersections
Chi-square test for driver characteristics in Sample 1 vs.
2
XT X2
Characteristics of Drivers at a = .1
Age 6.251 7.802
Nationality 6.251 12.463
Occupation 9.236 18.109
Educational level 7.779 3.490
Driving schools 4.605 6.174
Miles driven/year 4.605 5.568
Number of accidents in past 2 years 4.605 .523
Income 4.605 5.450
Drivers license availability 2.706 2.203
H :
There is no difference between characteristics of intersection
population drivers and conflict involved drivers at signalized
intersections.
TABLE 11.
45
Chi-square test for driver characteristics in Sample 1 vs.
Sample 2 (unsignalized intersections only).
X2
T 2
Characteristics of Drivers at a = .1 Xc
Age 6.251 .750
Nationality 6.251 .384
Educational level 7.779 1.405
Driving schools 4.605 12.636
Miles driven/year 4.605 1.590
Number of accidents in past 2 years 4.605 1.619
Income 4.605 1.997
H :
0
TABLE 12.
There is no difference between characteristics of intersection
population drivers and conflict involved drivers at unsignalized
intersections.
Chi-square test for driver characteristics in Sample 1 vs.
Sample 2 (unsignalized intersections, drivers with two
conflicts only).
2
XT X2
Characteristics of Drivers at a = .1 c
Age 4.605 6.212
Nationality 4.605 1.562
Occupation 6.251 6.088
Educational level 6.251 18.359
Miles driven/year 4.605 .264
Number of accidents in past 2 years 4.605 12.530
Income 4.605 2.207
H :
0
There is no difference between characteristics of intersection
population drivers and drivers involved in two conflicts at
unsignalized intersections.
46
TABLE 13. Percentage increase or decrease in representations of con-
flict involved drivers over intersection population sample.
percent over or
under in representation
Unsignalized
Signalized (drivers with 2
Characteristic Intersection conflicts only)
Age:
< 18 + 23% + 95%
19-25 + 28% + 1%
26-45 - 26% - 21%
> 46 +150% N T *
Nationality:
Saudi + 1% - 6%
Middle Eastern + 70% + 39%
Far Eastern + 38% - 18%
Others — 73% N T
Occupation:
Salesman + 53% + 43%
Teacher - 17% + 43%
Professional - 49%
Craftsman + 40% + 46%
Student + 94% + 50%
Professional Driver + 82% N T
Educational level:
None + 18% - 47%
Preparatory + 17% - 47%
Intermediate + 23% + 28%
High School - 9% + 74%
College - 34% - 61%
Driving Schools:
No - 5%
Dallah + 34% N.T.
Others - 51%
Miles Driven/Year:
< 9000 miles - 33% + 2%
9000-13000 miles + 55% + 12%
> 13,000 miles + 6% - 7%
Number of Accidents Past 2 Years:
None - 3% - 23%
One + 4% - 7%
__Two orgmore + 31% +150%
Monthly Income (S):
< 900 + 45% — 20%
900-1800 - 10% + 26%
> 1800 - 29% - 13%
* N.T. Not tested because of low frequencies of occurrences.
47
4.5 Evidence of Patterns Among_Conflict Involved Drivers
. Data from conflict involved drivers were separated by intersection
for each of the signalized intersections, and the distribution of these
characteristics were compared with the characteristics found when data
from all unsignalized intersections for drivers with two conflicts were
combined.
Pearson's goodness of fit test was used to test the null hypothesis
of no difference among conflict involved driver's characteristics. The
results are given in Table 14 and 15, which indicate that conflict in-
volved drivers possess similar characteristics at all intersections.
It may be concluded that a 'pattern' does exist among conflict
involved drivers, and that this pattern may be useful in predicting
future involvement in conflicts, and ultimately in accidents.
It is suspected that the reason certain characteristics (occupa-
tion, educational level, and income) were accepted in the first test
(Table 14) and rejected in the second (Table 15) was due to the large
number of young conflict involved drivers (high school students) with-
out driver's license traveling through some of the 'back streets' away
from police detection.
4.6 Conflicts by Type
The literature (32,34,35) contains several articles that suggest
a better correlation exists between accidents and specific types of
conflicts than between accidents and total conflicts. Thus, an analy-
sis was performed to determine whether specific conflict types are
prevalent among conflict involved drivers. The null hypothesis of no
difference between characteristics of drivers involved in a specific
conflict type and the general driver population was tested. The result
48
TABLE 14. Chi-square test for conflict involved
1 vs. Intersection 2 signalized).
drivers (Intersection
X2
T X2
Characteristic a = .1 c
Age 4.605 2.453
Nationality 4.605 5.161
Occupation 7.779 4.959
Educational level 7.779 .766
Driving schools 4.605 2.039
Miles driven/year 4.605 1.810
Number of accidents in last 2 years 4.605 4.107
Income 4.605 3.286
Driver's license availability 2.706 1.431
H : There is no difference between conflict involved driver
0 characteristics.
TABLE 15. Chi-square test for conflict involved drivers (Intersection
1 vs. Intersection 2 vs. all unsignalized intersections with
drivers of 2 conflicts).
X2
T X2
Characteristics 0 = .1 c
Age 7.779 4.586
Nationality 7.779 7.192
Occupation 13.362 24.656
Educational level 13.362 23.807
Driving schools 7.779 2.381
Miles driven/year 7.779 7.733
Number of accidents in past 2 years 7.779 13.290
Income 7.779 5.710
Driver's license availability 4.605 32.599
H : There is no difference between conflicted involved driver
0 characteristics.
‘P
In Or
on a
4
Ir‘
(d
.er
I
0
"'nl
, .
'lu.v
3..
r
5..
‘
"zr‘
’
A
'i’li
49
of this analysis is provided in Table 18.
Since no single characteristic is significantly overrepresented in
drivers involved in all types of conflicts, predicting type of conflict
using driver characteristics did not appear feasible. In contrast to
the literature, total conflicts were more highly related to characteris-
tics than were conflict types.
One possible explanation for these results would be that the
variance between days or locations in the dependent variable (conflict
type) was large, and thus combining these data masked any true relation-
ship. To determine whether this is true, a series of chi-square tests
were conducted to see if the variance by day or location was signifi-
cant. The specific tests conducted were: ‘
1. HO: conflict type frequencies are similar at each unsigna-
lized intersection.
2. H0: conflict type frequencies are similar for each day of
conflict observation at unsignalized intersections.
3. H0: conflict type frequencies are similar at each signalized
intersection.
4. HO: conflict type frequencies are similar for each day of
conflict observation at signalized intersections.
In tests 1 and 2 frequencies of conflict type were found to be
similar on all days at all unsignalized intersections. Thus prediction
of conflict type at these intersections cannot be improved by analyzing
each data set separately. In tests 3 and 4 frequencies of conflict
types at signalized intersections were found to differ significantly
with the observation period and the intersection, indicating that
separating the data may be helpful in determining the relationship
between driver characteristics and conflict types.
50
TABLE 16. Frequencies of conflicts by type for unsignalized intersections
(drivers involved in two conflicts only).
Conflict Type Frequency Percent of Total
Right angle 48 60
Turning 11 13.7
Rear End 9 11.2
Side Swipe 12 15.0
ES 136'"
TABLE 17. Frequencies of conflicts by type for signalized intersections
Conflict Type Frequency Percent of Total
Right angle 14 11.2
Turning 23 18.4
Rear End 40 32.0
Side Swipe 48 38.4
125 I 100
51
TABLE 18. Characteristics of conflict involved drivers that explain
types of conflict.
Characteristics Found Significant at a = .1
Con§$gct Signalized Unsignalized All Intersections
Right angle Nationality Income Nationality, number
of accidents
last 2 years
Turning Age None None
Rear end Occupation, None Educational level
educational
level
Side Swipe Occupation, None Number of
miles accidents last 2
years, miles driven
per eyar
52
A test to determine whether subsets of driver characteristics are
related to the distribution of conflict involvement by type for each of
the signalized intersections was conducted. Each characteristic that
explained conflict involvement (as shown in Table 10) was tested using
a multiplier to maintain the proportionality of the test.
_ 11.
n .
n.
OJ J
n = total conflict frequency observed
"j = total column frequency of characteristic subset
noj = new cell frequency of that subset
Several subsets of characteristics were found to have a significant
relationship to type of conflict. The significant variables are drivers
of age group < 20, and with a Saudi nationality. As in previous tests,
age and nationality proved to be important variables in explaining the
variation in conflicts.
4.7 The Model
One of the principal objectives of this study was to determine
whether models could be developed to predict the conflict rate at inter-
sections, based on driver characteristics. The fact that a 'pattern'
of conflict involved driver characteristics, defined in (4.5), exists
indicates that some characteristics of conflict involved drivers are
consistently related to the conflict rate. It is, therefore, hypothe-
sized that the overrepresentation of these characteristics can be used
in the model to predict conflicts.
Multiple linear regression was used in this study to relate
53
dependent variables to one or more independent variables by means of a
linear relationship. One of the methods used in multiple regression is
step wise regression where an independent variable is added or removed
from the equation to achieve the highest coefficient of determination
with the fewest number of independent variables. The model tested was
of the form:
Y X X
k = a000 + a11k 11k + °°° T a14k 14k
where:
% of conflict involvement (pattern) for characteristic
X = i subset j intersection k
ijk % ofigeneral driving public for characteristic i
subset j intersection k
Total conflict count
Y = conflict rate/ = ( for each approach ) (1000)
k 1000 vehicles Corresponding VOTUme (20 min.)
n
n = number of observations
corresponding volume = 20 minute volume for that approach
The driver characteristic most clearly related to conflicts, age,
was used as the first independent variable in the model tested.
The multiple linear regression models developed as a result are
Y = 508.9 - 242.9 X
k 12k (1)
X13 corresponds to age group (19-25)
Y = 925.2 - 95.8 X - 349.5 X
k - 149.3 X
13 (2)
11 12
X11 corresponds to age group 5.18 and > 46 years of age
54
X13 corresponds to age group 26-45
In all cases, the 't' test indicated that the coefficients were
not significantly different from zero.
The coefficient of determination R2 was .32 (32% of variations ex-
plained by regression) with an F significance of .244 for the first
model, and R2 = .58 for an F significance of .558 for the second model.
All coefficients of the independent variables were negative with
the lowest coefficient (95.8) corresponding to age group (< 18 and
> 46) and the highest coefficient corresponds to age group (26-45).
Because the sum of the percentages of all age groups amounts to 100%,
the lowest negative coefficient represent that subset of the age that
makes the greatest contribution to the conflict rate.
Thus, a well defined and statistically reliable linear relationship
between conflict rates and age does not exist in this data even though
the x2 test indicated that there was a relationship between these
variables. The number of data points was too small to test non-linear
regression analysis, or to stratify the data into multiple regions for
separate linear analysis.
To illustrate the empirical relationship between these variables, a
plot of the conflict index, which is defined as the ratio of the per-
centage of conflict involved drivers of a certain characteristic to the
percentage of the same characteristic of the general public, and driver's
age is shown in Fig. 7. The graph suggests two linear relationships,
with the conflict index decreasing with age for ages 16-40 and then
increasing with age after age 40. More data would be required to test
the statistical reliability of each of these relationships, but this
explains why the x2 test was significant and the linear regression not
55
strong. 4
The conflict index plotted against age for both drivers who had
been involved in at least one accident in the past two years, and
drivers with no accident in this time period is shown in Fig. 8.
Drivers with no accident history exhibited the same general pattern
as all drivers. However, drivers with one or more accidents in the
past two years exhibited a similar relationship for the under 40 age
group, but the conflict index did not increase with age beyond this
point.
56
S
ml
e
v
.1
P:
D
.T.
C
3|
1|.
..T
n
O
C
1'
1|
A
1098765432109876543
.
221111111111
xaucH HUHHNNON
om A
omume
mvnflv
ovumm
mmnam
omnom
mNuHN
omumfi
Conflict index by age of all conflict involved drivers.
Fig. 7.
4.0
3.8
3.6
3.4
3.2
3.0
2.8
g 2.6
E 2.4
.3 2.2
E 2.0
3; 1.8
.3 1.6
1.4
1.2
1.0
.8
.6
.4
.2
Fig. 8.
57
—--Conflict Involved Drivers with No Accident History
“‘ Conflict Involved Drivers with One or More Accidents
,in the Past Two Years
k“-)S-oo.X- - X
L!) 0 L0
C? Q' d'
I I
H go H
m 0") <2-
16-20
21-25
26-30
46-50
0
L0
A
Conflict index by age of conflict involved drivers with and
without an accident history.
CHAPTER 5
CONCLUSIONS AND RECOMMENDATIONS
5.1 Conclusions
Based on the data collected and analyzed in this study, the follow-
ing conclusions were reached: .
1. The most hazardous driver is in the age group 25, has not
completed more than 9 years of schooling, has been involved in two or
more accidents in the past two years, works as a salesman, craftsman,
professional driver, or a student. His income is of a lower income
group, and he drives an average of 11,000 miles per year.
2. Drivers with prior involvement in accidents were involved in
”more conflicts than drivers with no accident history.
3. Age was found to be the single most explanatory variable of
conflict involvement. Age and occupation form the best multiple re-
gression model among all tested characteristics.
4. Uniformity of conflict involved driver characteristics was
found to improve when intersection type is considered. There was no
significant difference between the characteristics of the general
driving public and conflict involved driver for combined data from all
intersections. However, differences were identified when the data was
segregated by type of intersection control.
5. Characteristics of conflict involved drivers were found to be
similar at all intersections, and the potential of using this pattern
is promising in predicting future conflict occurrence and ultimately
future accidents.
6. No single driver characteristic has a distribution matching
t:he distribution of type of conflict occurrence. However, combinations
58
59
of individual characteristics (young Saudi drivers) correlates well with
the occurrence of conflict types at signalized intersections.
7. Drivers licensed by local driving schools were involved in
more conflicts than drivers without driving training or with training
from schools outside the Kingdom.
8. The multiple regression model results indicate that a simple
linear relationship between conflict rate and driver's age does not
exist. However, when the relationship was presented vectorially two
different relationships were observed; one for the age group 16-40 and
a second one for the age group 40 and over. More data is needed to
construct and test these relationships in a model.
5.2 Discussion and Recommendations
The traffic conflict technique was determined to be a promising
technique for identifying characteristics of hazardous drivers at urban
intersections in Saudi Arabia. The relationship between the charac-
teristics of hazardous drivers and conflicts was found to be analogous
to that between driver characteristics and accidents. This strengthens
the arguments for using conflicts as a surrogate safety measure when-
ever accident history is not developed or not available, such as in
the case of Saudi Arabia.
The similarities in the observed characteristics of conflict in—
volved drivers at different locations provides the basis for the iden-
tification of a driver profile (or pattern) necessary for predicting
future conflict rates. *Based on the data collected in this study, it
appears that the prediction process should be carried out separately
fflor signalized and unsignalized intersections.
Young drivers were found to be overrepresented in conflict
6D
involvement as well as drivers with one or more accidents in the past
two years. One of every two drivers interviewed 18 years of age or
younger, was involved in one or more accidents in the past two years.
This group of drivers is considered a high risk group, and issuance of
special driving permits, which is allowed for drivers below 18 years
of age (the legal driving age), should be reviewed. These young
drivers tended to traverse back streets away from police detection.
It is recommended that local police introduce more patrol surveillance
to cover remote roads, and better enforcement action be promptly and
strictly applied to traffic violators to reduce the risk of accidents.
Local driving schools have not been effective in minimizing the
risk of involvement in conflicts. These schools should be evaluated
and upgraded to achieve greater effectiveness in driver training.
The establishment of an accident location system should be
initiated in Saudi Arabia to provide a data base for research, improve-
ment, and development programs.
BIBLIOGRAPHY
10.
11.
12.
l3.
14.
15.
BIBLIOGRAPHY
Russan, Sabey 8., "Accident and Traffic Conflicts at Junctions,:
TRRL Report LR 514.
Waller, P. F., and G. G. Koch, "Characteristics of North Carolina
Drivers," Highway Safety Research Center, Univeristy of North
Carolina, Chapel Hill, N.C. (December 1971).
Smead, R. J., "The Statistics of Road Accidents," 3rd Int. Sympo-
sium on Urban Traffic Safety, October, Budapest, 1972.
Jorgenson, N. O., "The Statistical Detection of Accident Black
Spots," 11th Int. Study Weak in Traffic Engineering and Safety,
Brussels, 1972.
Hall, P. A., "The Identification of High Accident Locations," 11th
Int. Study Week in Traffic Engineering and Safety, Brussels, 1972.
Mattewson, J., and Brenner, R., Indexes of Motor Vehicle Accident
Liklihood, Highway Research Board Bulletin 161, 1957.
Breunning, S. M., and Bone, A. J., Interchange Accident Exposure,
Highway Research Board, Bulletin 240, 44-52, 1959.
Thorson, 0., "Traffic Accidents and Road Layout," The Technical
University of Denmark, Copenhagen, Denmark, 1967.
Smith, w. L., Probability Study of Accident Locations in Kansas
City, Missouri, Traffic Engineering 40 (7), 42-49, 1970.
Gorssman, L., Accident Exposure Index, Highway Research Board Proc.
33! 129’ 19, 54.
Sutri, V. H., "Accident Exposure for At-Grade Intersections,"
Traffic Engineering, December 1965.
Gwynn, D. H., "Relationship of Accident Rates and Accident Involve-
ments with Hourly Volumes," Traffic Quarterly, July 1967.
Feuchtinger (1949) Road Junctions and Trans. Die Bautechnik (Germany).
Raff, M. S. (1953) Interstate Highway Accident Study, HRB Bulletin
74, 18-45.
McDonald, J. w., Relation Between Number of Accidents and Traffic
Volumes at Divided Highway Intersections, HRB Bulletin 74, 7-17, 1953.
61
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
x29.
30.
/31.
1,32.
62
Surti, V. H., Accident Exposure and Intersection Safety for At-
Grade, Unsignalized Intersections, HRB 286, 81-95.
Chapman, R. A., Traffic Collision Exposure, New Zealand Roading
Symp., 184-198, 1967.
Tanner, J. C., 1958, Accidents at Rural Three-Way Junctions, J. of
Instn. Highway Engineers, 2(11), 56-67.
Colgate, M. and Tanner, J. C. (1967), Accidents at Rural Three-Way
Junctions, Road Research Laboratory Report, LR-87, Crowthorne,
England.
Webb,(i.M., 1955, The Relation Between Accidents and Traffic Volumes
at Signalized Intersections, Proc. Inst. of Traffic Engr., 149-160.
Thrope, J. D., 1964, Calculating Relative Involvement Rates in
Accidents Without Determining Exposure ARRB Aust. Road Res. 2(1),
26-36.
Seibrecht, E. B., Schumacher, C. F., Lauer, A. R., "Accident Char-
acteristics of Drivers at Various Age Levels," Iowa State College,
Traffic Safety, December 1959.
Lynn, C. W., "Psychological and Sociodemographic Characteristics
of Accident Involved Drivers," Highway Safety, Virginia.
Betz, M. H., and Bauman, R. 0., "Driver Characteristics at Inter-
sections," HRR, 19S.
Faulkner, C. R., 1968, "Accident Debris and Reported Accidents at
Roundabouts," TRRL Report LR 202, Crowthorne, England.
Spicer, B. R., "A Traffic Conflict Study at an Intersection on the
Andoversford By-Pass," Transport and Road Research Laboratory,
TRRL Report LR 520, 1972.
Waller, P. F., Koch, G. 6., "Characteristics of North Carolina
Drivers," Highway Safety Research Center, University of North
Carolina, Chapel Hill, N.C., December 1971.
Pelz, D., Schuman, 5., "Drinking, Hostility, and Alienation in the
Driving of Young Men," A paper delivered to the Third Annual Alco-
holism Conference, NIAAA, Washington, June 1973.
Baker, W. T., "Evaluating the Traffic Conflict Technique," FHWA.
Perkins, S. R., Harris, J. 1., "Traffic Conflict Characteristics -
Accident Potential at Intersections," GMR, December 1967.
Baker, W. T., "Evaluating the Traffic Conflict Technique," FHWA.
Paddock, R. D., Spence, D. E., "The Conflict Technique: An Acci-
dent Prediction Method," Ohio DOT, August 1973.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
‘49.
63
Glennon, J. C., "Critique of the Traffic Conflicts Technique,"
TRR #630. 1977.
Spicer, B. R., "A Traffic Conflict Study at an Intersection on the
Andoversford By-Pass,“ Transport and Road Research Laboratory,
TRRL Report LR 520, 1972.
Spicer, B. R., "A Study of Traffic Conflict at the Intersection
Transport and Road Research Laboratory," TRRL, Department of the
Environment, Crowthorne, Berkshire, 1973.
Spicer, B. R., "A Pilot Study of TC at a Dual Carriage-Way Inter-
section," Road Research Laboratory, TRRL Report LR 410, 1971.
Cooper, P. J., "Predicting Intersection Accidents," Ministry of
Transport, Canada Road and Motor Vehicle Traffic Safety Branch,
September, 1973.
Pugh, David, "Relevance of Conflict Studies to Safety Improvements
at Intersections," Washington DOT, 1977.
Pugh, D. E., Haplin, T. J., "Traffic Conflict in Washington State,"
Washington State DOT, April 18, 1975.
Al-Ashini, Nasa, "Alternative Methods of Examining Correlation of
Conflicts with Accidents," Michigan DOT Traffic Engineering,
October, 1976.
Parker, M. R., Jr., "Right-Turn-on-red," Virginia Highway at
Transportation Research Council, September, 1975.
Blunden, W. R., Munro, R. D., "Report on the Study of Traffic Con-
flicts and Accident Exposure," Australian DOT, June 1976.
Seymour, W. M., Deen, R. C., Havens, J. H., "Traffic Control for
Maintenance of High-Speed Highways," Research Report 380, Kentucky
DOT, December 1973.
Cima, B. T., "An Evaluation of Freeway Merging Safety as Influenced
by Ramp Metering Control," TRR 30, 1977.
Pigman, J. 6., Agent, K. R., "Raised Pavement Markings as a Traffic
Control Measure at Lane Drop," Public Roads, Vol. 39, No. 1, June
1975.
Pigman, J. G., Seymour, W. M., Agent, K. R., Cornell, D. L., "An
Operational Analysis of the I-69, I-65 and I-71 Route Junction at
Louisville," Kentucky DOT, April 1972.
Zegeer, C. V., "Effectiveness of Green-Extension of Systems at High
Speed Intersections," Kentucky DOT, Divisions of Highways, May 1977.
Clyde,WL N., Paper presented at Traffic Conflicts Seminar, March
20-21, 1974, Michigan DOT.
SO.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
64
Parker, M. R., Jr., "Use of TCT to Assess the Hazards of Trans-
porting Oversize Loads," Virginia Highway and Transportation
Research Council, December 1977.
Guttinger, V. A., "Conflict Observation Technique of the Work
Done and On-Going Research in Holland," Holland 1977.
Paddock, R. D., "The Conflict Technique Procedures Manual,” Ohio
DOT, July 1975.
Blanz, W. D., and Migletz, D. J., "Application and Traffic Con-
flict Analysis at Intersections," Midwest Research Institute,
October 1979.
Clayton, M. E., Deen, R. C., "Evaluation of Urban Intersections
Using TC Measures," Kentuck DOT, August 1971.
Hayward, J. C., "New-Miss Determination Throng Use of a Scale of
Danger," Pennsylvania Transportation and Traffic Safety Center,
HRB Record #384.
Lightburn, A., and Howarth, C., A Study of Observer Variability
and Reliability in the Detection and Grading of Traffic Conflict,
2nd International Traffic Conflict Workshop.
"Motor Vehicle Accidents in Relation to Geometric and Traffic
Features of Highway Intersections," Stanford Research Institute.
National Safety Council, "Accident Facts," 1970-1979.
McGuire, F. L., "The Understanding and Prediction of Accident-
Producing Behavior," North Carolina Symposium on Highway Safety,
Chapel Hill, N.C., 1969.
Nie, Hull, Jenkins, Steinbrenner, and Bent, "Statistical Package
for Social Sciences," McGraw-Hill, 1970.
APPENDIX A
Driver Interview Form for Both Sample 1 and 2
A-l 55
mt" w)‘ 4” r-
.l l l' . | . “4".H4h4, Counmmtlu '"I""“""" 44444441
”and! . u‘ only 4444' 80444444445: Hadron
‘by
We“!
he on on you?
“4’5 I I JJ’Q- 1 I Jay-1p l a...
Doc 4. your 44.44.44.444” “444“ Ion-6.444“ (Spocuyl
"b,oblg‘J' 4 i 4.44 4 ”.44 4 I 3.....44 4 1.1354.-
1 do)
“to on you got”? 44.4444 Moo. Shopping 0444:"
42444:.le
4 “aw“ b an- lg'k I Solemn/Inn:
I
“I. to m melon 05‘ ’91- I l Tanner/Proton”:
4,4» I 4» I .4».- 1 l Prolonlooal/‘hchnlcol/Mmrut
g4— J-b ’ *9 Mb 1 l Craftmn/nochantc/Poctory lotto:
A .35 lfifi- I ,u I l cur4c44/ucncu4u/oulc- 4444444.:
4'35 4 1 "who:
.4... 4 1 I444 Iron!
4.1—— 4 ) omor
'Jchuarde-Hb 1 I‘ll... 4 Din-,1- I IJ'O C 4.3-Io! 1.3;.
In: 4- pl Mind in" m tum—flu- 4444444 34:44:31 mu.)- m
tyuohdh» 4 I... I 4 1
II. you can “to an no ‘ 440 Mm Uh“
25.5.4.4 mas-1_—
4 4,544.3“...5. 3.4. 4 4;. 4-4 4 1 .45. 4-4 4 4a,..- 3,44
In: my ”an did you an”? A 1-2 gun )-5 you-- nor. 4
' “*J‘JHVL ‘ “fibh 4 A Jb‘aH- ‘ 4m! A 4405‘ 3.45-
me tn- “ on you «In? Pun-90¢ a: mu I44. 0444."
'Mowuowbfi
be my ”on 440'. you 11'. 1. 8M. om?
'oanlquI-Jlfi C I '0 I I-‘alga I “85‘ I My“ I dun-I
love. you has n «4444.. about I. cum "one. on." «mum
~- I was! mfl'JeI-MMAV ' '
.0 any in...“ you «tn 1. a day. can. at you?
4wwumdhoowp; 4 4.3.4. 4 3.5 I Lab. 4 Dana-i“
In. .0447 munc- a“ you no. «c444. I044. 044. 1540 Mon :44 . cu
’m "" mv...u.n,.. 434.4,“, 4 4,... 4 4-4.4 4.4.44 4 a.m.,;
an my mime. “at“ In you: an:
«4444-. «9031044440. I944. 044. M More than in
4 and.» 4 H 4 4..- 4 4...: 4 41'
m 10! actual? . In You In In
4 44.4.1.4.» Jo
no you have children
IIIUJJEUH o
mmmvtun‘clzi‘y? ‘ '0‘...” ‘ ’fim ‘ .zytium
4qb¢y¢na 1 HmJ-n 4 Ha— ! lab—b
I. you have m“ beau? Vary an All. r4944: Duh 0
mum-m» 4 4404-"...444
an to you linen, no.
"ooo-’ooo ‘ ,‘."I'fi.fi'
Lou than “.3000 It)”. - «on lion than 60“
“final—5,4...» 4 I... 4 H 4 45449.11
in you In. 0 I.“ Antiv- «no; no..." In In La“:
44. v 444
I no.
“JV-”a-
APPENDIX B
Chi-Square Test of Population Assumption
66
Na u w
.toommmm mo mwmwuo ~ Ivan ohmmn
ozocumau ac umxwuo n rpm: ¢m~mu
5.99H nogm
sud — cm
H Nona
— mono
mono — moc—
hh ” mu
— @oh
H muhn
flown — no¢m
hm ~ w
~-"'-"'
mom
4<>oh >
:cm «omuzcmz
arccpucoonaocuo
«Jazabmmcau ...
. Nm\wnxmo u uhqo zanh¢u¢uv
zo-<>zummo oz~wm~r we «mean:
. n u¢H¢o Hz«:c
Hno
81
mm H monH<>¢ummo qummHz ma «mean:
u
.zocuumm no muwmouc :HH1 «mmH~.n u um<3cw Hzo 34x
:.o¢H c.cm c... 4
H Hggou
I I I I I I I I I I I.I I
I.I
AU>H4 onp