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 beIow. SAFETY IMPACTS OF VEHICLE DESIGN AND HIGHWAY GEOMETRY 3? Koji Kuroda 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 1984 / (KI/I I)! ABSTRACT SAFETY IMPACTS OF VEHICLE DESIGN AND HIGHWAY GEOMETRY Bv Koji Kuroda The purpose of this study is to provide insight on the relationships among automobile size. highway geometry, and traffic accidents. The curb weight was selected as the parameter to define automobile size for this study, and automobiles are classified into seven groups by their curb weight. To examine other possible factors in accident involvement, the model-year of the automobile is also considered. Data for this study are! based on police reported accidents that occurred in Michigan in 1982. Only accidents which occurred on the Interstate system or on the Michigan Trunkline system are analyzed. A total of 51,740 accidents were included in the data base. This study uses a new exposure approach based on two hypotheses: 1. The likelihood of an automobile being an object (the second vehicle) of an accident is proportional to its exposure. 2. The likelihood of an automobile being an object of an accident is common to any design if the exposure is the same. The validation of these hypotheses are consistently supported by data. The new exposure approach is found to be a useful tool for the quantification of exposure. Small automobiles are found to have a unique risk of accident involvement. and are found to be more likely to be involved in an accident in the following conditions: - single vehicle accidents - overturned vehicle accidents - on icy or snowy highway surfaces - at midblocks - in rural areas On the other hand. large automobiles are found to be more likely to be involved in an accident in the following conditions: - accidents with pedestrians - accidents with parked vehicles - accidents with other vehicles - at intersections - in urban areas Results for several geometric features are obtained by examining the accident data for rural and urban areas separately. In rural areas, small automobiles are found to be more likely to be involved in an accident at the following geometric features. - midblocks - 2 lane-2 way highways - no passing zones In urban areas. large automobiles are found to be more likely to be involved in an accident at the following geometric features: — intersections — 6 lane-divided highways — low posted speed limits. Drivers in small automobiles are found to have a greater risk of being injured regardless of whether they are in the automobiles identified to be responsible for the accident or in the second automobile. Results of analyses for vehicle classifications using wheelbase as a measure of automobile size provides very similar results, however the curb weight categories show more consistent results. The model year is also found to be an important factor in determining accident involvement. Newer model-year automobiles are found to be less likely to be involved in an accident than earlier model-year automobiles. ACKNOWLEDGEMENTS I wish to extend my sincere appreciation to the many officials of the Michigan Department of Transportation who assisted me in the research on this project. Especially, the assistance of Jack Boneck was invaluable and this paper could not have been produced without his help. I also wish to thank Dr. Thomas L. Maleck, Dr. William C. Taylor, Dr. John Kreer and Dr. Snehamy Khasnabis for their individual assistance. Thanks are due to William Sproule who helped with the revision of sentences. The help of Vicki Switzer in typing this dissertation is gratefully acknowledged. The Highway LoSs Data Institute's provision of VINDTCATOR 83 was invaluable in decoding the Vehicle Identification Numbers. ii TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES 1.0 INTRODUCTION 2.0 LITERATURE REVIEW 3.0 PROBLEM STATEMENT 4.0 METHODOLOGY 5.0 DATA BASE 6.0 VALIDATION OF THE NEW EXPOSURE APPROACH 7.0 RESULTS 7.1 Statewide Results 7.1.1 Findings for Locations 7.1.2 Findings for Geometric Features 7.2 Relationships for Urban-Rural Subsets 7.2.1 Findings for Rural Areas 7.3 Findings for Midblock Locations in Rural Areas 7.4 Findings for Urban Areas 7.5 Results for Driver Injuries 7.6 Comparison of Results by Curb Weight, Wheelbase, and Model Year 8.0 CONCLUSIONS BIBLIOGRAPHY iii PAGE iv vi 24 26 47 50 60 63 74 83 89 94 104 115 119 134 169 171 LIST OF TABLES Percentage Distribution of Compact Vehicle Accidents versus All Other Accidents - 1963 Percentage Distribution of Compact Vehicle Accidents versus All Other Accidents by Road Location - 1962 Summary of Literature Review (Data Elements) Summary of Literature Review (Contents of Studies) Summary of Literature Review (Results of Studies: Small Cars were found to have the following characteristics) VINDICATOR Data Available Distribution of Automobiles Regisered in Ingham County (1981) Distribution of VEH #1 and VEH #2 Accidents in Ingham County (1982) Statistical Test for Population Distributions of VEH #2 Accidents by Weight for Accident Type The Best Fitting Curves for Geometric Features Statistical Test for Population Distributions of VEH #2 Accidents for Rural and Urban Areas Best Fitting Curves for Geometric Features in Rural Areas Best Fitting Curves for Geometric Features at Midblock Locations in Rural Areas Best Fitting Curves for Geometric Features in Urban Areas iv PAGE 20 21 22 28 52 52 59 93 106 116 121 TABLE Number of Driver Injuries in VEH #1 and VEH #2 by Curb Weight (1982) Results of a Least Square Fit of Its Linear Transform for Driver Injury in VEH #1 Results of a Least Square Fit of Its Linear Transform for Driver Injury in VEH #2 Crosstabulation 1 Crosstabulation 2 Crosstabulation 3 Number of Drivers Injured VEH #1 (Per Two Passenger Car Accidents) Number of Drivers Injured VEH #2 (Per Two Passenger Car Accidents) Best Fitting Curves for Results by Wheelbase Categories PAGE 122 127 128 129 130 132 133 133 154 FIGURE 4. 4O 1 2 LIST OF FIGURES Sample Output from VINDICATOR 83 Accident Report Form used by Michigan Police Agencies Accident File from Michigan Department of Transportation Michigan Department of Transportation Master Accident File Relationship between Vehicle Weight and Wheelbase Relationship between Model Year and Vehicle Weight Michigan Highway Districts Control Section Map for Kalkaska Co. Michigan MIDAS Segment Record Accidents in Calhoun Co. by Sex of Driver of VEH #2 Accidents in Calhoun Co. by Sex of Driver of VEH #1 Difference between Exposure Measures Exposure Methods Population Distribution #1 Population Distribution #2 Research Procedure Total Accidents by Curb Weight Accidents by Sex of Driver vi PAGE 29 30 31 32 35 37 41 42 44 51 51 53 55 57 58 61 66 68 FIGURE PAGE 7.4 Drinking/Drugs 69 7.5 Single Vehicle Accidents 71 7.6 Overturned Accidents 72 7.7 Other Vehicles I 73 7.8 Parked Vehicles 75 7.9 Pedestrian/Fixed Object 76 7.10 Animals/Bicycles 77 7.11 Surface Condition 78 7.12 Area Type 80 7.13 Route Class 81 7.14 District 82 7.15 Number of Lanes 84 7.16 No Passing 85 7.17 Degree of Curvature 86 7.2.1 Distribution #3 91 7.2.2 Distribution #4 92 7.2.3 Total Accidents 95 7.2.4 Area Type 97 7.2.5 Number of Lanes 98 7.2.6 Lane Width 99 7.2.7 Shoulder Width 101 7.2.8 Degree of Curvature 103 7.2.9 No Passing 105 7.3.1 Total Accidents 108 7.3.2 Midblock Accidents 109 7.3.3 Lane Width 110 7.3.4 Shoulder Width 112 vii FIGURE 7.6.10 7.6.11 7.6.12 7.6.13 7.6.14 7.6.15 7.6.16 7.6.17 7.6.18 7.6.19 No Passing Overturned Total Accidents Area Type Number of Lanes Driver Injury VEH #1 Driver Injury VEH #2 Scattergram 1 Scattergram 2 Model Year Model Year by Curb Weight Model Year by Wheelbase Scattergram 3 Total Accidents Driver Injury VEH #1 Driver Injury VEH #2 Sex of Driver Age of Driver Single Vehicle Overturned Vehicle Other Vehicles Pedestrians/Fixed Objects Area Type Surface Type Total Accidents Driver Injury VEH #1 viii PAGE 113 114 117 118 120 123 125 135 137 138 139 140 141 143 144 145 146 147 148 149 150 151 152 153 155 156 FIGURE 7.6.20 7.6.21 7.6.22 7.6.23 7.6.24 7.6.25 7.6.26 7.6.27 7.6.28 7.6.29 Driver Injury VEH #2 Sex of Driver Age of Driver Single Vehicle Overturned Vehicle Other Vehicles Pedestrian/Fixed Object Area Type Surface Condition Drinking/Drugs ix PAGE 157 158 159 160 162 163 164 165 166 167 1.0 INTRODUCTION It is widely known that small automobiles are increasing in population in America. Approximately 30 million automobiles whose weight does not exceed 3000 lbs. are registered in a population of about 100 million passenger cars. The median curbweight of passenger cars is 3500 lbs. and the median wheelbase length falls in the 110-115 inch group (1). This downsizing of passenger automobiles has an effect on such vehicle design parameters as vehicle length and width, eye height, center of gravity. braking ability, horsepower and engine size. The relationship between vehicle design and accident involvement or severity have been investigated intensively by many researchers, and a concensus has developed that occupants of small automobiles have a greater risk of injury when involved in an accident. However, it is not clear that small automobiles have a higher accident involvement rate or even a higher injury accident involvement rate than large automobiles. There has not yet been a thorough examination of the effect of downsizing of vehicles on accident rates at various geometric features. Certainly the fact that small automobiles represent about 32 percent of registered vehicles (1) would indicate an urgency to study issues 1 2 concerned not only with accident risk and vehicle design but also with accident risk and highway design. This paper attempts to provide insight on the relationship among automobile size, highway geometry, and traffic accidents. 2.0 LITERATURE REVIEW As early as 1963, the Division of Research and Development, New York State Department of Motor Vehicles (2) studied the extent of accident involvement of compact vehicles as compared to the overall vehicle population. The definition of compact vehicles used in this study was furnished by the U.S. Bureau of Public Roads. Any vehicle whose points totaled less than 1.0 on the following scale was considered to be a compact vehicle. 1. Wheel base, more than 112 inches - .5 points 2. Empty weight, more than 3,000 pounds - .5 points 3. Gross brake horsepower, more than 130 - 1.0 points 4. Overall length, more than 200 inches - 1.0 points They found that in 1962, compacts accounted for 11.4 percent of all passenger vehicle registrations but represented only 10.0 percent of all passenger cars involved in accidents, 10.4 percent of all passenger cars involved in fatal accidents, 9.8 percent of all passenger cars involved in injury-producing accidents, and 10.2 percent of all passenger cars involved in property damage only accidents.. The study also included a comparison of type and location of accidents for compacts and other automobiles. The results showed a relatively smaller involvement of compact automobiles in pedestrian accidents, a higher percentage participation in collision accidents with other 3 vehicles, a higher frequency of overturning vehicle accidents and a larger percentage of compact automobile accidents during poor weather and poor road conditions (Table 2.1). Little comparative significance in regard to road locations was found. Nevertheless, a greater percentage of compact automobile accidents occurred on curves, both level and on grade. No statistical test was implemented to determine if this difference was statistically significant (Table 2.2). In 1964 Solomon (3) studied accidents on main rural highways related to speed, driver and vehicle characteristics. This study was not designed specifically for evaluating small and large automobiles. However, he compared the accident involvement rate and injury rate with regard to size, age, and horsepower of passenger cars. The data base for this study consisted of 10,000 accidents which occurred on main rural highways. The vehicle size classification used in this study was as follows: 1. Small - all foreign automobiles, Crosley, Henry J. and Willys. 2. Low priced - Ford, Chevrolet, Plymouth, Studebaker, etc. 3. Medium priced - Pontiac, Buick, Oldsmobile, Edsel, Mercury, etc. 4. High priced - Cadillac, Lincoln Continental, Chrysler Imperial, and Packard. Comparison made by grouping passenger automobiles according to size showed that the involvement rate tended Table 2.1. Percentage Distribution of Compact Vehicle Accidents versus All Other Accidents - 1962 Total Accidents Type of Accident Compact Other Collision with: Pedestrian 2.5 7.8 Other Motor Vehicle 87.2 77.7 Ped. - Other M.V. 0.1 0.3 Railroad Train 0.0 0.1 Animal 0.4 0.5 Fixed Object 6.0 9.0 Bicycle 0.7 1.5 Motorcycle 0.1 0.1 Non-collision: Overturned on Road 1.0 0.5 Ran Off Road 1.8 1.6 Other 0.2 0.9 TOTAL 100.0 100.0 Weather Conditions Clear ' 71.6 73 a Cloudy -- -- Raining 16.8 14.6 Snowing 7.0 6.2 Sleeting 0.3 0.3 Fog 0.7 0.7 Not Stated 3.6 4.9 TOTAL 100.0 100.0 Road Condition Dry 60.5 63.3 Wet 23.0 20.0 Muddy -- -- Snowy 12.5 11.7 Icy 0.4 0.1 Not Stated 3 6 4.9 TOTAL 100.0 100.0 Table 2.2. Percentage Distribution Of Compact Vehicle Accidents versus All Other Accidents by Road Location - 1962 ion Road Locat Not at Intersection At Intersection Driveway Underpass Railroad Crossing Bridge Ove rpass cter Road Chara Straight, Curve, Straight, Curve, Curve, Not Stated level level grade grade Straight, hillcrest hillcrest TOTAL TOTAL Total Accidents Compact 100. '40th090 H O O #0! 0000000 ~10taa-m13 ONWNbMfl O .O Other #0! OOOQNU NIHHUNJH H O O IO 0’ IDOMUQUIID mchhnmtno-fi H O O .0 7 to decrease as the price of the automobiles increased and that there was also a tendency for the injury rate and the number of persons injured per 100 involvements to decrease as the price of the automobile increased. Passenger automobiles were also divided into three horsepower groups, 110 or less, 111-170, and 171 or more. A comparison of accident involvement showed that drivers of automobiles having 110 horsepower or less had the highest accident involvement rate. This is the reverse of the results found in the New York study, but as in that study, no test was conducted to determine statistical significance. Also in 1964, Jaakko K. Kihlberg, Eugene A. Narragon and B. J. Campbell (4) investigated accident injury data from 12,835 injury-producing rural motor-vehicle accidents. Passenger automobiles were grouped in three vehicle weight categories for this study: 1. Small automobiles: 771 vehicles, under 2000 pounds. 2. Compact automobiles: 1085 vehicles, 2000-2999 pounds. 3. Standard automobiles: 10,979 vehicles, 3000 pounds and over. This study was only concerned with injury accidents and with a comparison of the percentage of fatalities and severe injury accidents for light and heavy automobiles. The number of injury-producing vehicles was used as the exposure index. Three major findings of this study were: 1. For all occupants, frequency of injury (to any degree) in small automobiles was about ten percent higher than in standard automobiles. 2. The frequency of dangerous and fatal injuries in small automobiles was about twenty percent higher than in standard automobiles. 3. The percentage of fatalities among small automobile occupants was about fifty percent higher than among standard automobile occupants (7.0 percent of occupants were killed in small automobiles, as compared with 6.0 and 4.6 percent in compact and standard automobiles, respectively). It was not clear from the report if these results were statistically significant. This study also included a review of eight previously published automobile size studies. The authors characterized the findings in the reviewed literature as follows. 1. No evidence was found to suggest that small automobiles have a different accident involvement rate than that of other automobile classes. 2; Once involved in an accident, the risk of injury appears to be higher for small automobile occupants than for large automobile occupants. In 1968, John W. Garrett and Arthur Stern (5) expanded the earlier Automotive Crash Injury Research study by Kihlberg, Narragon and Campbell, comparing the performance of the Volkswagen two door sedan "beetle" models with that of American and other imported automobiles involved in rural injury-producing accidents in 30 states. The accident sample was divided into seven groups, with Volkswagen comprising one of these groups. The seven summary groups and the number of cases were as follows: 1. Volkswagen 879 2. Renault 325 3. Foreign Sedan 391 4. Corvair 674 5. Light U.S. 1582 6. Intermediate U.S. 22568 7. Heavy U.S. 1135 TOTAL 27552 This study compared the percentage of dangerous and fatal injuries, the percentage of rollovers, collisions and the percentage of ejection of occupants from the vehicle. using injury-producing accidents as the exposure. Dangerous and fatal injuries were more frequent among the occupants of Volkswagen and other small automobiles than among the occupants of larger automobiles. This appeared to be more closely related to the door opening and/or the door latch design than to vehicle size. Ejection accounted for 40 percent of the Volkswagen fatalities and 32 percent of those injured to a dangerous degree. Rollover, with its high incidence of lethal ejection, occurred more frequently for Volkswagen, Renault and other small foreign sedans, than for American automobiles involved in injury accidents. These results were statistically significant. In 1970, B. J. Campbell (6) dealt with the variation in injury to unbelted drivers involved in crashes while driving various automobile makes and models. Accident reports from 270,000 vehicles involved in crashes in North Carolina in 1966 and 1968 were analyzed. The aggregated driver injury rate of all vehicles was calculated and compared to driver injury in each automobile 10 make on the basis of a grouping of accident circumstances, speed, impact site and accident type. VIN data were utilized to identify automobile years and makes. Definitions of vehicle size were not given. However, each vehicle make was classified into six groups which were: 1. "The Big 3" (standard Chevrolet, Ford and Plymouth). 2. The largest automobiles (Buick Electra, Dodge Monaco, Oldsmobile 98, Pontiac Bonneville, etc.). 3. Standard size Buick, Dodge, Mercury, Oldsmobile, Pontiac. 4. Cars just smaller than standard (Buick Special, Chevrolet Chevelle, Dodge Dart, etc.). 5. Domestic compact automobiles (Chevrolet Chevy II. Chevrolet Corvair, Ford Falcon, Plymouth Valiant). 6. Other automobiles; foreign, American specialty automobiles, and a regrouping of American compact automobiles. The two major results of this study were: 1) driver injuries were less frequent and less' severe under comparable crash circumstances in the later model years than in earlier years; 2) driver injuries tended to be more severe in smaller automobiles than in large automobiles under comparable conditions. These results were statistically significant. In 1973, James O'Day, Daniel H. Golomb, and Peter Cooley (7) made an accident study to assess the risk of injury to occupants of small and large automobiles. Accident data for Washtenaw County, Michigan during the period 1968-1970, which contained 16,360 passenger automobiles was used for this study. Small automobiles 11 were defined as those with a licensing weight of 3,100 pounds or less, and large automobiles as those with a licensing weight of 3,300 pounds or greater. The 200 pound range in the middle of the distribution was not used. The percentage of injury involvement in small and large automobiles for single and two automobile collisions was compared, using the numbers of small and large automobiles involved in accidents as the exposure. They found that, in single vehicle accidents, small automobile occupants were more likely to incur an injury, and in small-large collisions, small automobile occupants sustain more injuries. No statistical test was conducted. The number of injuries per accident involvement as a function of vehicle weight was expressed in equation form as follows. N = 10’8 w2 - 1.32 x 10“ w + .642 where N = the number of injuries per accident involvement W a the weight of the involved vehicle in pounds. In 1974. Theodore E. Anderson (8) examined the influence of various accident parameters on the probability of passenger compartment intrusion. Data pertaining to one and two automobile collisions involving American made automobiles manufactured between 1969 and 1973, involved in accidents within an eight-county area surrounding Buffalo, New York were used. Only those accidents which had hospital treated injury cases were used (approximately 360 12 cases annually). Automobile sizes were defined as small (0-2999 pounds), medium (3000-3900 pounds), and large (4000 pounds and over). The small automobiles had a 55.5% intrusion incidence, the medium automobiles had 57.0%, and the large automobiles 44.4%. Thus, the large vehicles were found to be less likely to experience intrusion than small or medium vehicles. This result showed a statistical significance. In 1974, Donald W. Reinfurt and B. J. Campbell (9) presented crash rates per million vehicle miles for several makes, models, and years of automobiles. Accidents occurring in North Carolina in 1966 and 1968 were used as accident data for this study. Estimates of annual mileage for a given make-model—year combination were derived from the annual vehicle inspection program in North Carolina. *VIN data were used to identify make and year for both accident and inspection data. Definition of vehicle size was not clearly given. However, automobiles were classified into big, standard and small. Overall and single vehicle accident rates per million vehicle miles by model year, from 1960 to 1966, were calculated and compared for several makes. The results indicated that the accident rate within a given make tended to be lower for each model year than the rates for the preceding model year. The average overall accident rate of 30 makes was 8.47 per million vehicle miles for 1960 models, and 3.52 for 1966 models. On the other hand, there were no consistent differences in the overall accident rate 13 among the various makes within a model year nor were there any clear differences according to size of vehicle. No statistical test was conducted. In 1975, James O'Day and Richard Kaplan (10) compared fatal injury rates by vehicle size and by driver age. An equation P(F) = P(A) x P(F/A) was used to describe the probability of a driver being killed by an accident in a year. P(A) was gained from a survey conducted in 1970 (1341 accidents among 4221 individuals), and P(F/A) was obtained from the 1972 Texas accident file (1061 driver fatalities among 13,932 accidents). The definition of vehicle size was not given, but was dichotomized by designating standard model autos as "large," and compact and sports models as "small." Intermediate body styles were excluded. The results showed that the probability of receiving a fatal injury was greater in small automobiles (P(F) 8 .000388) than in large automobiles (P(F) s .000183), and that the difference increased with age (.000605 for ages 55+ against .000183 for all ages). The results showed statistical significance for the difference in the fatal injury rate between large and small automobiles. Also in 1975, P. L. Yu, C. Wrather and G. Kozmetsky (11) studied the relationship between vehicle weight and safety, using the data from the State of Texas for the year 1973. Passenger automobiles were classified into four classes. 1. Small automobiles 0-2999 pounds l4 2. Intermediate automobiles 3000—3999 pounds 3. Large automobiles 4000—5000 pounds 4. Super large automobiles > 5000 pounds A sample of 1,204 accidents and 148 serious injury/or fatal accidents were broken down into weight classes along with the percentage distribution of the automobiles registered in that weight class in Texas for the year 1973. The major findings were: 1. Larger automobiles had a much higher frequency of accident involvement. 2. Once an accident occurred, smaller automobiles had a slightly higher frequency of getting into a serious or fatal accident. 3. Overall, larger automobiles had a much higher frequency of getting into a serious injury, or fatal accident. The results of 1 and 3 were found to be statistically significant. Also in 1975, Leon S. Robertson and Susan P. Baker (12) examined motor vehicle related fatalities by vehicle size. Accident records of all reported fatalities in Maryland involving a motor vehicle during 1970 and 1971 were used. Automobiles were classified by automobile-wheelbase in inches into five groups (105 or less, 106-110, 111-115, 116-120, and 121 or more). Fatal crash involvement rates per 100,000 registered vehicles were calculated by dividing the number of involved vehicles by the number registered for those five groups. In single vehicle crashes, the smallest vehicles (105 15 inches or less) were involved at a rate almost three times the rate of the largest vehicles (121 inches or more), 12.0 compared to 4.1 per 100,000 registered vehicles. In multiple vehicle crashes, the smallest vehicles were involved at a rate 12.6 compared to 7.2 for the largest. However, these results were not examined statistically. In 1977, Amitabh K. Dutt, Donald W. Reinfurt, and Jance C. Stutts (13) investigated accident involvement and crash injury rate per million miles of vehicle travel by make, model and year of automobile. The accident and injury information was obtained from the North Carolina accident file over the period October 1973 through October 1974. The exposure data, derived from paired odometer readings recorded on a statewide sample of motor vehicle inspection receipts (approximately 300,000), were also obtained over the same period. Passenger automobiles were classified into three major groups by make-model. 1. Full size automobiles (luxury, medium, and standard) Buick, Cadillac, Standard Ford, etc. 2 Middle-sized automobiles (intermediate, compact) Chevrolet Chevelle, Intermediate Ford, Ford Mustang, etc. 3. Small-sized automobiles (subcompact) Ford Pinto, Datsun, Toyota, VW Beetle, etc. 16 Calculations were given by: 6 2 5‘ 0 '1 (I {J p. ll number of involvements of group i. 3 ll estimated annual mileage for group i. Ni = estimated registration frequency for group i. The overall or total accident rate showed a steady decline with newer model years (6.24 for 1960 models against 3.56 accidents per million vehicle miles for 1974 models for full-sized automobiles). Injury rates also showed a steady decline with newer model years (0.92 against 0.22). Full-sized automobiles showed far better rates than middle or small-sized automobiles for accident involvement and injury. This was particularly true for the newer model year (injury rate: 0.22 for full, 0.32 for middle and 0.57 for small for 1974 models)- These results proved to be statistically significant. In 1979, G. Grime and T. P. Hutchinson (14) studied relationships between fatal rates, injury severities and mass ratio of the vehicles involved in collisions. Accidents involving injury occurring in England during the years 1969 to 1972 were analyzed. Vehicle weights were decoded into ten classes. Severities of injury were judged by comparing the percentages of fatal, serious and slight injury and of uninjured drivers. Analyses were separated for head-on and intersection, and for rural (speed limit more than 40 17 miles) and urban (elsewhere) areas. They found that the mass ratio had the greatest effect on deaths. For example, the percentage of deaths in light vehicles was seven times that in heavier vehicles when there was a collision involving two vehicles with a mass ratio of 2. 0n the other hand, little or no effect of vehicle weight was found in single—vehicle accidents. No statistical test was conducted in this study. In 1980, J. Richard Steward and Carol Lederhaus Carroll (15) updated and extended the previous study by A. K. Dutt and D. W. Reinfurt (1977). As in the previous study, this study estimated average annual mileage by automobile make and model year based on paired odometer readings obtained from motor vehicle inspection receipts. 1979 data obtained in North Carolina were studied and compared with previous results. Automobiles were classified into four automobile size categories — full sized, intermediate, compact, and subcompact. Comparison with the previous study showed a much higher average annual mileage for relatively new subcompacts (under five years old) in 1979 than in 1974 or in 1975. Comparisons of accident rates averaged over the entire span of fourteen model years showed full sized automobiles to have significantly lower rates than intermediates and compacts. Rates for subcompacts were also significantly lower than those for intermediate and compacts. However, when averaged over the newer nine model years, full sized automobiles still had a lower rate than 18 the other three classes, but lthose three did not differ significantly from one another. These results were examined statistically. In 1982, Leonard Evans (16) estimated the relationship between vehicle mass and the likelihood of an occupany fatality. Data from the Fatal Accident Reporting System for 1978 (28687 occupants) were used to study both non-two automobile and two automobile accidents. Another set of data were prepared to determine the number of registered vehicles of the same mass having owners of the same age, using 1980 registration data from the State of Michigan and R. L. Polk. Vehicle mass ranges were 500-900 kg, 900-1100 kg, 1100-1300 kg, 1300-1500 kg, 1500-1800 kg, and 1300-2400 kg. For non-two automobile accidents. numbers of fatalities by driver age and vehicle mass were divided by the estimated percentage of vehicles registered to obtain the likelihoods. To get the likelihood of an occupant fatality of two car accidents, the number of occupant fatalities in automobiles 0f mass ”1 in collisions with vehicles of mass M2 was divided by assumed exposures which were derived from the results of non-two car accident data. This study showed that the likelihood of a fatality in small vehicles (900 kg) was 1.5 times (non-two automobile accidents) or 3.4 times (two-automobile accidents) greater than that in large vehicles (1800 kg). The statistical significance was not examined. In 1983, Leonard Evans (17) also studied the 19 likelihood of involvement in a potentially fatal accident using a new method to account for exposure. He hypothesized that the likelihood of a pedestrian or a motorcyclist being killed in an accident does not depend on the mass of the automobile, and ’if a vehicle has more exposure, it has a greater probability of hitting a pedestrian or a motorcyclist. Thus, the number of pedestrians or motorcyclists killed in accidentes with vehicles of mass m can be used as the measure of exposure. Data from the Fatal Accident Reporting System for 1975 through 1980 combined were used to account for those fatalities. The likelihood of a driver fatality for small automobiles (900 kg) was 2.60 times that for large automobiles (1800 kg) with 99 percent confidence. The aforementioned studies represent 20 years of attempting to advance the understanding of the relationship between vehicle size and accidents. Table 2.3 shows a summary of data elements from these studies. Seven out of 16 studies used vehicle weight as the definition of size. Another 7 studies used the make and model of automobiles while the remainder used the wheelbase and point system. Table 2.4 summarizes the contents of the studies. Related to these tables, Table 2.5 shows a summary of the results of 16 studies. Nine reports are (in varying degree) related to the problem of injury and death resulting from an accident. These reports, without exception, conclude that the injury and death rate or severity of injury are higher in small 20 0:650: sensor 000% .Hmooz .oxmz phases umm> .Hmuoz .mxmz mmmnaomsz pause: Hmooz .033: new» .Hmvoz .mxmz 0:630: 0:650: Home: .mxmz Honor .oxmz ..3.> unmade umo> .00aum .mxmz Emum>m vanom nuaw 000 no CofiufiCnuwn Haumm awash .o.z UGmHucu .o.z .02 Haven .xua Haydn .xmh anvmh .o.z .s.z .ncH .50”: .o.z Hausa .ncn amusm .ncH Hausa .>.z mama UUN>~0G¢ AmuCOEOHN mummy owlmhmn mhmfi mum” Nblmomn fibthmfi HNIONGH mm whmfl thu .wmmH Gran Ohlmmmn om .momH molwmmn mmlonmn hnlmmmd «mad new» mama .mmoauam may :« vmou>oum uoz fl www.mu N M M own new Hoo.H 0 can omn.mn OO0.0NN Non.hu 0mm.NH OO0.0H M deem unmemm mcm>u mcm>m mausoum define uufia conunmnom 5» >mo.o unamcuom nonsmoca >mo.o saunasoo vacuumc unobseflx coeoHom >29 .>.z peruse >Bsum 30a>om undumnuuaq no >umeesm Am mama «man Oman mum“ hhmu ohm” mhmu ohm“ vhma vwmn mum” Ohm” mwma vomn *wmn mom" new» .n.N OHDGB 21 Table 2.4. Summary of Literature Review (Contents of Studies) P(F/A) or P(F) Author Exposure Used P(A) P(I/A) or P(I) Geometry Percentage N.Y. DMV Registered No No‘ No (2) Solomon Veh-miles Yes No Yes No Kihlberg Inj. acc No Yes No No Garrett Inj. acc No Yes No No Campbell Acc No Yes No No O'Day Acc No Yes No No Anderson Acc No Yes No No Reinfurt Veh-miles Yes No Yes No O'Day Drivers No No Yes No Yu Registration Yes Yes Yes No Robertson Registration No No Yes No Dutt Veh-miles Yes No Yes No Grime Inj. acc No Yes ‘ No (3) Stewart Veh-miles Yes No No No Evans Registration No Yes No No Evans (1) No Yes No No P(A) - Accident Involvement P(F/A), P(I/A) - Fatality, Injury Rate by Accident P(F), P(I) - Fatal, Injury Accident Involvement (1) Pedestrian and motorcyclist killed. (2) Percentages of accidents involving compact cars for various geometric design conditions compared to percentage of compact car registration. (3) Urban, rural only. 22 00» 00% 00> 300 00> 00> 00w 0300000cw0m >00300030000m 00> 00» 00» 00? oz 00» 00% 00» 00» 00% 00% 00% 00% A000 0o A<\Hva Amvm 0o A05ficu “easy 00:»«: 0000 >0=fi00 ueuew 00£wuz 00w 0000800 .000000800ucu 00:0 0000 \oz 00000000 0030a 0000 game .uuefim 000000000 0Hw:«a 00w uceouuucwua 00: .50an:0 Heueu 0002 00% 00>~0>cu 00000000 >0afi00 000“ 0:» 0030: 0:0 00>~0>cq 0:000000 0005 .503n0u 0002 0000 a05n00 “even 00:&«: 02 0000 H0000 00 >0sncH 005w“; 000E0>~0>00 06000000 00304 0000 a05fi:« aeuom 005w“: .w.z 0000 £0000 0030a 05» ~000§ 0030: 0:0 .uceuuuucwqo uoz 0000:00c0 000: m0w0zficu 0005 .06000000 p0>~0>00 00:0 00>0m 000A 0:0 H0008 0030: 0:0 .h05fi00 000>0m :0w03mx~0> w:0&0 00030000 e0uu¢~mueh 00003fic0 0008 .00000000 00>~0>00 00:0 000 0000 £0000 0030A 050 0030: 0:0 .00>~0>:« 00000000 0008 .a0amca 0002 .w.z unmouugcwum uoz uC0E0>~0>cu 0000000< A0£ 00 0::0m 0003 e000 naofim "m0quum mo muuam0¢v 300>0¢ 005000000; H0 >00555m mcm>m oce>m 00030um 05000 0030 00000090“ aw zoo.o 00smca0m 00000vc< 036.6 00033530 UUOHHQO w0030z00 COanom >29 .w.z .m.N 0HDQP 23 automobiles once involved in an accident. The other reports are mainly relevant to the question of how often small automobiles are involved in an accident, or in an injury or fatal accident. These studies tended to disagree with each other. For accident involvement, two studies found small automobiles have a higher rate than larger automobiles, while one study found the opposite result, and two studies did not find any significant difference. For the injury or fatal accident involvement, four studies found small automobiles have a higher rate, while one study found the opposite result. Four studies also examined the accident involvement rate for various model years. These studies found a consistant result, which is that older model year automobiles have the higher accident involvement rate. The reviewer feels that the question of accident involvement requires a measure of exposure to the risk of an accident or injury accident, and because various measures have been used to quantify exposure, the results have been inconsistant. Little research was found which would contribute to an understanding of the effects of geometric design on vehicle size and accidents. The studies reviewed did not address the full range of geometric features. 3.0 PROBLEM STATEMENT Previous analyses about the relationship between automobile design and accidents were associated mainly with crashworthiness during a collision. These studies genereally agree on the inferiority of small automobiles in relation to driver injury or death. However, little information was given about the likelihood of small vehicles being involved in an accident. There is some argument about whether higher maneuverability and lower aggressivity of small vehicle drivers could offset the inferiority of their crashworthiness. To determine the impact on highway safety caused by the increasing share of small automobiles, it must be determined if they have a unique risk of accident involvement. This problem necessitates a quantification of exposure. Various measurements have been used, including estimated vehicle miles, number of registered vehicles, and number of drivers. Nevertheless, none of these is available or appropriate when trying to examine accident rates under specific circumstances, such as on rainy days or by day of week; or at specific locations such as at intersections or in parking lots; or on roads with certain geometric design characteristics. Traditional definitions of small automobiles are given by several criteria, although curbweight and wheelbase length are the two most commonly used. It has not yet been 24 25 determined which is the most suitable indicator of characteristics of small automobiles. Although a study to identify the relationship between vehicle design and accidents is important it is of equal importance to understand the relationship between highway geometry, vehicle design and accidents. In the past, little has been done in regards to highway design, partially because this relationship may be subtle and partially because appropriate measures of exposure were not available. This study provides not only quantitative information on the impact of the changing population of small automobiles on highway safety, but also insight on the impact of small automobiles on accidents at selected highway geometric design features. 4.0 METHODOLOGY The central theme of this study is the analysis of the relationship between highway geometric design and related automobile parameters with accidents being a measure of performance. The methodology consists of two parts. The first part of the methodology involves an investigation of the variation in accident involvement and in injury accident involvement as related to these parameters. The second part extends the investigation to an examination of the relationship between automobile design, highway geometric design and accidents. The methodology consists of the following steps. 1. Definition of vehicle parameters to be studied. The curb weight and the wheelbase were selected as the two parameters to define automobile size. To examine other possible factors in accident involvement, the model-year is also used. 2. Igentification of :informgtion on automobile design. In the United States, each new vehicle sold is required to automobilery a Vehicle Identification Number (VIN). VIN's consist of unique alphanumeric strings of up to 17 characters assigned by manufacturers. VINs from the accident file of the Michigan Department of State Police (MSP Accident File) are utilized to obtain information on vehicle make, series, model, curb weight, wheelbase, and year of production. The VINDICATOR 83 program (18) is used 26 27 to decode VINs. Table 4.1 lists the make and model year span the VINDICATOR 83 handles. Examples of the output of VINDICATOR 83 are given in Figure 4.1. 3. Development of g new accigpnt filg (File 1). A standard accident report form has been used for many . years to report accidents that occur in Michigan. Figure 4.2 is an example of the form that has been used since 1979 by all state and local agencies in Michigan. With this report, information on accident location, conditions. vehicles and drivers are put on record. The VIN numbers for the vehicles are also indicated. Vehicle number one is identified as the vehicle designated as responsible for initiating the accident and vehicle number two is the second vehicle in the accident. When a copy of the accident report is sent to the Michigan Department of State Police (MSP), each report is given a unique number, called the Accident Report Number (AR). This number is attached to any accident file of the Department of State Police or the Department of Transportation to identify the accident. As mentioned in step 2, VIN is extracted from the MSP Accident File. Figure 4.3 shows a description of this file. The necessary accident information is extracted from the Highway Accident Master Data of the Department of Transportation. Figure 4.4 shows the data file description of this file. These two pieces of information are combined by matching the AR and a new file is created (File 1) that has information for both accident data and vehicle design. Table 4.1. MAKE VALUES iDCDNIOIOIIFQNl-I 28 VINDICATOR Data Available MAKE NAME Chevrolet Ford Pontiac Buick Plymouth Oldsmobile Dodge Volkswagen Mercury Cadillac American Chrysler Lincoln Opel/Isuzu Datsun/Nissan Toyota Capri Mazda Fiat Volvo Audi Dodge/Mitsubishi Honda Porsche MG Subaru Plymouth/Mitsubishi GM of Canada Chevrolet Truck GMC Truck Ford Truck Dodge Truck Plymouth Truck Jeep International Mercedes Benz BMW Renault Saab Peugeot Triumph Ferrari Lancia Jaguar Alfa Romeo Rolls Royce Bentley Aston Martin Delorean Avanti Mitsubishi Lotus MODEL YEAR SPAN 1967-1983 1967-1983 1967-1983 -1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1983 1967-1979 1967-1983 1967-1983 1972-1977 1967-1983 1972-1983 1972-1983 1972-1983 1972-1983 1973-1983 1972-1983 1970-1980 1972-1983 1976-1983 1968-1983 1973-1983 1972-1983 1973-1983 1973-1983 1975-1983 1973-1983 1975-1983 1981-1983 1981-1983 1981-1983 1981-1983 1981-1983 1981 1981-1983 1981-1982 1981-1983 1981-1983 1981-1983 1981-1983 1981-1983 1981-1982 1981-1983 1983 1981-1983 I 1981-1983 29 mm moe 8000 030000 oHQEmm .H.v 003600 .....v.oo...oo.oo..o.....44....oesooooo...44.4.0.9...o....aooeeooeoeoeeee.9000000000.eeeeoeoeeooeeeOeeeeeoeee00000000000 m0 mm-mm 0n “02-063 mu0mn¢o .mZ—OZu Ov 60-4.aw0 Umu_awv. Ow.-Cm. UdI en. nmm..I3 Zcowm 300040 .>oom wa—mDZmODO ”axe: n.n- NpIO—wa. <~oc “JwOO: 05m.unhm- hu maoucw OZ .xudiwc vcwn mmwm Ow. Om. 9n. n.nt C n. 0 cm 0 v. new. new. @ O O 0 02.: axoom Juno: mu_uum scam» au axe: )m04< 3mmza new. .0w—Jdaam u2— CZ. @hOn.m2nImen N. n Omhun nu H0302— ....4....ooeeooeo.oo.oo.oooo.ooooooo..ooo.eooeoeoeeo90¢.o9000000000000.00¢9000000000..0.00000009000000000000000000000. m. 0.-qn.o. H01:03 <0v\n-mmv\0mn-m nwz_02u av mm-<0.wo Umw_umm 0mm-0mm .3: en. UmmDIJ ape: a0004v H>000 wm_mo:m050 Durex emOe h2.0—w: <.oom awooz mw_awm 4u ua wxez 3m04< 3mmz< may. .ow_nad:m a2_ 02. mnmnnnzomnvmn n. . 0m>.n oh .0302— ......4.......o.o..ooo..o..ooviooeooo..o.,.v...9.0.9..coo...coo.00000.0eeeeeo0000000000coo000000009400090000000000.o9.00.. a. mm-m..nn .a:\003 . <0.-0mn-w Hw2_02w . be <0<2oom cued “uxas mm.n U2.0:: <0m.-m-m. Hm maoaau oz “xaazwa a.n- w.nn hm 0m 0.. ma.n C m C 00 0 .n ohm. 050. N O O 0 02.: axoom Juno: mw_cmm scam» a¢ wxcz zmpme 3mmz< cau— .ow_ada:m u2_ oz. wnnsnnnnmzm h n mes.n m- .0302— .....o..o.ooooooooov9.4...ooaoo....o.oo.oo..o.ooooeeo.900000.000.00.000000.eGeeeeeeeeeeeeeeeeeeeeeeeee99.000000000900000 ...mn-...mn .d:\003 <0~-0mn-m .wz_02w an wmaw>w10 .mu_uww mu.-mw. La: n.. “moat: coma muv 0001 coco n H>000 0w40c>m10 “3.4: mean ”pzc_m3 <0 mcocau oz .xcazwu ..mn ..mn me. me. n.. seen 0 cm 0 an 0 m. mum. mum. . 0 0 o c:_: axoom awoo: mu_awm anew» acaw> wxdz rmbne amaze unu_ .om_naa:m uz. oz. oanOv.nI>no. c . mes.n ms “502. coo.oooeoo90.090.000.000...09.00.0000...o000600000009000OOOOOOOOOOOOOOO.oOOOOOOOOOOOOOOOOOOOOOOOO0000.00.00.00909.0000 30 an- d I... OFFICIAL TRAFFIC ACCIDENT REPORT 1 i It a ON u u s 8 L5“? “”9935 W85 '30-. alt-um 00v mine-v.00 mks-Du. mm min lion- 90'me Cute-00009 I vac-M. D I Van-annua- love-Mum Dyna-Dena .00.“. 0 .mm and) Donald! Pinon-In ”hummmnm Dan-II- whee-«Over'mvm Dun-Oca— "I. m ”ammo. a! bun uni-n. we.” e! 40-. e! I!” am Figure 4.2. Accident Report Form used by Michigan Police Agencies 31 ' ,*\\ PACE_5__OF8_ «p I.) DATA FILE/RECORD DESCRIPTION one Ffie'fj‘ ”no “/79! _ . -9:.5.'.8.0 "LE "WE use - ACCIDENT MASTER RECORD ".“E ”2" “MT .— 'LE '0 0200221 “Em" 5'“ .a. o .0. 156 mm El 0,... g“ :1; g 5;: [j D»...~ RECORD nee 3 o, l. TAPE TYPE 17"“! ” KEY/SEQUENCE ‘ 0mm IBPII DO 3:0 83°: 8:5; 5“ C] ’°° AR I (Positions 150-155) :3? go'“ C] E a (TNRZ) 0st FILE ORGANIZATION a y ' FILE TYPE [1) Ema [j Veoiebla' (’1 0»...- APPROX. no. or RECORDS BLOCKSIZE _.-3129 (“1004. mama... "EéFATEfii"""" --. -Yariabcho 360.000-...— EmeDE [)0 EDCDIC [j] DCI. [’ )ASCII (j )0»...- coquNTs 1. "2" Format is identified by a 2 in Position 156. 2. Each "2" record contains traffic unit information for one vehicle only beginning with 1980 accident year. Prior years contained data for two vehicles. FIELD DESCRIPTION FORHAT NUMBRIC LO AGE OF DRIVER HIT G YEAR VEH. MPCD. YER. TYPE SUBSCRIPT VEHICLE HARE VEHICLE TYPE DRIVER INTENT CONTRIBUTING (VIN) ALPHA/ NUMERIC OBJECT HIT VEHICLE IDENTIFICATION NUMBER TRAFFIC UNIT NUMBER DRIVER AGE SUBSCRIPT FILLER ) OR OPERATOR NUMBER HTS UNDER Figure 4.3. Accident File from Michigan Department of Transportation PACLLOF; DATA FILE/RECORD DESCRIPTION . . DATE Reyised was II sm 7 1-26-79 7-25-80 "LE ““5 Hwy. Accident Master HRECMD ““5 Accident ‘ '° 01.302121 "”5“" "2‘ .... 0 .. 252 ‘n‘o D nigh 8 Tape D Coed E: Ofiu' “RECORD TYPE 1 e‘ 1 7‘" "'5 D 7"" D ”m2 E ”2 KEY/SEQUENCE °°""" ""’ 5 “°° D '°° D 5“ D 2'” leage within Control SectiOn within District L'“ E] 94 C] "L D N” 5.4 mt FILE ORGANIZATION Pom, E1 Odd D Even UNRZ) Sequential I-ILE TYPE a In.“ C] Van»: [:1 0M: APPROX. no. or RECORDS sLoCIIsIZE _22£9__Q sou. E] Cheeses." “an” or 130190 Exraooe E escouc [j ICL D ASCII D o»...- Q/430/02 coaaeu'rs File is created from Program 0/630/02 which strips data from MSP Accident File "0200221". Format effective for 1978 accident year. Traffic Control Hwy District Control Special Ta; Section Number me Control Section Hilepoint Number of Vehicles Filler Distance from Crossroad Numeric Direction to/from Intersectin; Street Name (Thru Position 82) Alpha Numeric Figure 4.4. Michigan Department of Transportation Master Accident File III 09/?“ Figure 4.4. 33 “Not on all files -continusd- No. of Injuries Vehicle ll-Subscript Vehicle fl-Hake Driver ll-Age '0! - Driver Fl-Incenc Violation ll Circumstance ll Vehicle Fl-Obj act [it Vehicle Fl-b‘pe Filler Vehicle 02-Subecript Vehicle FZ-Hska Driver lZ-ue Driver lz-lntent Violation 02 Circumstance 02 Vehicle FZ-Obj ect lit Vehicle IZ-me (continued) use 2 or)... Filler Vehicle 0.1-Subscript Vehicle ll-Haks Driver FS-Ags Driver III-Int ant Violation '3 Circumstance '3 Vehicle III-Object lit Vehicle Oil-W *Average Daily Traffic (T.V.H.) Number of Fatalities Number of Injuries Number 34 MSP MDOT Accident IVINDICATOR Accident Master 83 Master Vehicle Design with AR . , Merge —<-————-—- File 1 Data This study considers vehicle number one (VEH #1) and vehicle number two (VEH #2) only. 4. Classificggion of passenger automobiles by design. The curb weight, wheelbase, and model year of passenger automobiles involved in accidents in 1982, were first examined using the accident file for Calhoun County. Fig. 4.5 shows the relationship of the curb weight and the wheelbase for these automobiles. Passenger automobiles are classified into seven classes by their curb weights or by the wheelbase. CURB WEIGHT (1) less than 2000 lbs. (2) 2000-2499 lbs. (3) 2500-2999 lbs. (4) 3000-3499 lbs. (5) 3500-3999 lbs. (6) 4000-4499 lbs. (7) 4500 lbs. or more 35 Ononamwnz poo 0:000: 0H00£0> 0003000 chmcoflumawm .m.v 005000 8.80. 00.8mm 8.80m 8.00am 00.08Q 8.809 8.980 8.00am 8.08« 8.8.! 8.000. sane-QIIIIOIIII¢IIII¢IIII9IIII¢IIIIo-III«IIIIQIIIIQIIIIOIIIIsaooueooIIeIIIIeIIIIOIIIIeIIIIeIIIIeIIIoo. 00.0. e e a 8.00 n a a a u — n — 00.no e e 00.00 — n u e u ~ ~ ~ — 00.00 0 4 00.00 ~ — u n a — u _ 00.0. e 0 saw ~ — a _ ~ u n a 00.00. 4 c 00,00. — u u u ~ ~ — — 00.n0. o o 00 no. u — u — u on unnuewee — u see u 8.0: e Dune-n e o 00.0: a e e e e u a be en fin u n a u e e he «an e — 8.0.. e e e e 8.nu. — e s can. as n e — ~ _ n e e u u «a e a 00.0“. e e es e a 8.0“. u s e e e. as e u u e e e n u e e — ~ nwfl — 8.0“. e 0 00,6“— ~ — u e e e e Qe ~ — n u a 00.00. s e a e e 8.0a. .oIIIIeIIIIe IIIIIII Le IIIIIIII e IIIIIIII (IIIIeIIIIe IIIIIIIIIIII eIIIIQIIIIeIIIIeIIIIeIoIIe IIIIIIII o. e e o e e _ 0 8.00pm 8.8ua 8.8hv 8.8«e 8.0099 8.8«0 8.89“ 8.8mm 8.8a. 8.0»: 93> >0 «20> been: can «run; .022.qu .3) >0 893) and. am: 8‘0 «lg .238. no lelunwwueuw 36 WHEELBASE (8) less than 95 inches (9) 95-100 inches (10) 101-106 inches (11) 107-112 inches (12) 113-118 inches (13) 119-124 inches (14) 125 inches or more MODEL YEAR (15) 1972 or older (16) 1973 (17) 1974 (18) 1975 (19) 1976 (20) 1977 (21) 1978 (22) 1979 (23) 1980 (24) 1981 (25) 1982 The model year is used to classify vehicles into 11 groups (1972-1982). Those automobiles older than 1972 are classified as 1972. Fig. 4.6 shows the relationship of the model year and the curb weight of automobiles involved in accidents in Calhoun County. 5. Acguisition of the number of accidents and injury accidents associated with automobile design. Only accidents which involved passenger automobiles are used in the analyses. The passenger automobile data is then broken down by curb weight, wheelbase, and model year. The accidents are cross tabulated by location, weather condition, age, sex of driver, road surface condition, and accident type for each classification category. To eliminate problems due to a difference in the number of occupants by different automobile designs, only 37 050003 0H00£0> 0:0 H000 H0002 0003000 mwzmcofiumH0m .m.v 0Hsmflm 00 «00. 00. .00. 00. 000. 00. 050. 00. 050. 00. pho. 00. :0... 00. 0h0.¢ 00. Qh0. 00. :0... 00.«h0. .on-s|4:oon¢nn:n¢nnua9¢tno¢otun¢ uuuuuuuu QuuuoOUIouo uuuuuuuuuuuuuuuu o IIIIIIIIIIIIIIIIIIIIIIII 9. 00.000. o 9 00.000. _ u n u . n n . 00.000. 0 9 00.000. — 0 u o u u o o o o «— — « o o o o o 0 00.0000 fl). N n b 0 00.0000 — o v o « ()Md .0 « a e u u o o n u o n n « 0 u O0.00n« 0b b D N « e «9 00 000R . n o v n « o 00 u n u v o « o 0 — o o o u . . « «. 8800 0 N 0 fl 0 o e 00 8.3 u « N o v 0 on u o « a 0 0 Q. . o « o o 0- ~ 0 0 a - oooonn F-) n b h n n ho 8.039 _ n o 0 n o «u . o « n « « o a. . o o o o 0. ~ « o « o o 0 0008' o n « « o e 00 8.§Q — o o o o (\Hu — o « an . o o o o 0 u — o o o. 8.00: o o o no 000001 — o 0 « m u a. . e o 0 n 0 00.0000 9 o o .9 00.0000 . o o— . o 0 u . — 0 00.009. . 9 00.0000 . . ~ 0 _ . _ . oooooa . . o 8.80. .ou-uno.-nuo uuuuuuuu o oooooooo o oooooooo o uuuuuuuuuuuuuuuuuuuuuuuu Olnnoouuooo oooooooo oounoon0ooo. On. .00. On. 000. 00. Ono. 00. 0b.. 00. rho. 00. 0.0. 00. 0b.. 00. vh0. 00. 090. 00. «>0. / 2.) >0 «10> ¢<0> 0000! «0¢0>> .000004. 2.) >0 «10> >10000 040 «bu00> .2300. 00 800000>>¢00 38 the drivers are considered when recording injuries. Fatal, incapacitating, non-incapacitating, and possible injury of drivers are considered as the injury classifications. 6. Determination of the exposure associated with automobile design. In the present study a new exposure approach is used. Hypotheses are made as follows: 1. The likelihood of being an object (the second vehicle) of an accident is proportional to the exposure . 2. The likelihood of being an object of an accident is common to any design if the exposure is the same. The basis of the first hypothesis is if an automobile is driven for a longer time or a longer distance, it will have a larger risk of being involved as the VEH #2. For example, any automobile, regardless of its design characteristics, could be hit by another vehicle while stopping on a red signal at an intersection. The number of times a vehicle must stop on a red signal is proportional to the number of miles driven (and on what roads the driving takes place) in that vehicle. Thus, the number of times a vehicle is involved in an accident as the V28 #2 is a direct measure of exposure. The basis of the second hypothesis is since this study is concerned with passenger automobiles only, we infer that there would be no difference in the likelihood of being hit regardless of whether the automobile is small or large. We assume, for instance, when a vehicle is involved in a rear-end accident with an automobile which has stopped on 39 red, the risk of the automobile being hit would not be related to its size. These two hypotheses imply that the number of VEH #2 accidents of automobiles with some design (D1) should be proportional to the exposure of these automobiles. All accidents in which a passenger automobile was involved as VEH #2 are considered for the exposures. The likelihood of a VEH #1 accident for an automobile with design (D1) is equal to the total accidents multiplied by VEH #2 class D1 relative to all VEH #2 accidents: EIVEH #1, Class 01] I Total Accidents x VB" ”2: Class DI (1) Total VEH #2 Dividing the number of VEH #1, Class D1 accidents by both sides of equation 1, we have the following. VEH 01. Class 01 ‘ Total van #2 K van #1, Class n1 EIVEH 01, Class 01] Total Accidents VEH #2, Class 01 (2) Equation 2 is the ratio of the actual number of accidents to the expected number of accidents for automobiles of class D1. This ratio will be referred to as the A/E ratio in this paper. Likewise, the expected number of injury accidents is equal to the total injury accidents multiplied by VEH #2 Class D1 relative to all VEH #2 accidents. EIInjury Acc. Class Dl] I Total Injury Acc. VEH fig) Class 01 Total VEH 02 (3) 40 Dividing the number of injury accidents of Class D1 by both sides of the equation 3, the ratio of the actual number of injury accidents to the expected number of injury accidents for automobiles of Class D1 is expressed as follows. Injury Acc. Class 01 - Total VEH #2 Ellnjury Acc. Class 01] Total Injury Acc. Injury Acc. Class 01 (4) V88 #2, Class D1 X The accident file used in this study includes the information for both VEH #1 and VEH #2. This permits the analysis of accident rates and injury accident rates by various automobile designs for various kinds of accidents. This same analysis can be used to compare the actual and expected number of accidents for any given geometric characteristic, vehicle characteristic or location, by defining Class D1 to be the parameter to be studied. 7. Development of accident File 2. The Michigan Department of Transportation has divided 83 counties into nine districts. Figure 4.7 is a map of Michigan illustrating the location of each county and the nine districts. Each state trunkline in each county is assigned a unique number and mileage point (distance in miles along the route from the point of origin of the control section to the point in question). Figure 4.8 is an example of a control section map. This control section is further divided into segments with respect to the geometric ppserruvr fl?BflflfilflXflfl383¥§3flffl333§393333$. AWL...— “CIR—n ALLIOL AWNL—u— ANTI“...— M000;- ML...— IAIN? “Y “02“....— "OMEN...— 000%.. CALM-ON.— 008.....— CHAILIVO'X. CHEW“. WWII“ “MI...— ”CINC— 00. YIAVII‘I. OIA‘I’IOT... NI LL80A LI. uwou'rou- HURON...— WM... ML... 00800...." “00...... ME LLA. “ORION... (AWN. “Luau. KIN? K "Ilfllflm 00:70.07 c.0000-00000000000000000000000000000000000 Figure 4.7. 41 DISTRICT LAKE—a... 3 M II“...— LIIWU... LINAMI—u. LIVW nth—m “KW“... ”mt-“TI MANUEL... “IOUITTI... W. “Wk...- “WONCL. MIDLAND—u... MK I‘... “0°C....- WIN.— ”WIN”. ”noon... “WAY“...- CAK LAND J!" x! ANA... OMWW—m "MOON. mom... “OWL—u. WOO...- wflmgflfi new: I‘LL sacs-mow- “MMM_~ “NI LAG...— WLCRAFY. “Imus“... 87.0““...‘(7'0 “at"..- 7 WOW..- I VAN ”RI“... 7 fiA‘NTINAW. I WAYNE-“METRO NIX FORD... 3 andao-na-uOuuo-uo 3fi?833§3¢RfifiiiflfiflfiIfififl933$§3lfifiRFS§F§Ffi$b ouooaauaouaaooau and COUNTY NUMBERS 61L .. nu. I41 .3 70 m 3 B :zjz=.f {34:19 ——--O Michigan Highway Districts 42 530:0... .00 00.0933. How. am: :ofluo0m Honucoo .m... 0090.0 Cznow...m.w_..mk5---C>.LEFT-TURN .— , —. REAR-END 4T \ é i g i '5 ' ! i i i ' ' 1 Under 20 25 30 35 40 - 45 Over X100 lbs Figure 6.5. CURB WEIGHT Population Distribution #1 PERCENTAGE (%) 20 10 58 I T £&——1Q.ANGLE O- - - O LEFT-TURN \ r . _. REAR-END Under 95 Figure 6.6. 101 107 113 119 125 Over : WHEBLBASE Population Distribution #2 59 Table 6.3. Statistical Test for Population Distributions of VEH #2 Accidents by Weight for Accident Type. Weight Class 1 2 3 4 5 6 7 Total ObserVEd 5.1 15.1 17.3 24.1 19.9 13.2 5.3 100% (Angle) (3:253:23) 6.4 17.0 18.3 24.1 18.2 11.7 4.3 100% 191%13 .26 .21 .05 0 .16 .19 .23 1.10 x2 = 1.10 Wheelbase Class Whgeizzse 1 2 3 4 5 6 7 Total ?::;§:fd 8.9 17.1 8.4 26.7 20.7 14.6 3.5 100% (Eggjfzig) 10.5 19.1 9.6 25.3 20.0 12.8 2.7 100% 19%;)? .24 .21 .15 .08 .02 .25 .24 1.19 2 7.0 RESULTS The ratios of the actual to the expected number of accidents for seven groups of vehicle curb weights are obtained in this study using the new exposure approach. As described in the methodology, this study uses the number of VEH #2 accidents as a measure of exposure. This measure enables one to obtain relative exposures for any size of automobile, location, and duration. For example, a relative exposure of the“ smallest size of automobiles on Sundays is obtained by identifying accidents in which VEH #2 is of the smallest vehicle class on Sundays out of the total VEH #2 accidents on Sundays. This procedure is utilized throughout this study except for.those accidents which do nOt include VEH #2 accidents. VEH #2 accidents on a state-wide basis for each category of automObile size are used to obtain the relative exposure for those accidents which do not include a VEH #2, such as single-vehicle-accidents, turnover vehicle accidents, accidents with a pedestrian, or driver injury accidents. The three stages used in analyzing the accident data are shown in Figure 7.1. The total accident data set was initially used to determine statewide A/E ratios. For example, all accidents occurring at intersections were first analyzed to study intersection accident involvements for the seven groups of automobile size. Similarly, the A/E ratio for the following conditions were determined 60 61 ALL ACCIDENTS Figure 7.1. Resear 51,740 1) General H w r-1 0 .H Location m Geometry N m H -:--l I 1 “l" RURAL URBAN 10,283 12,022 Geometry Geometry MIDBLOCK 6,907 Geometry Total Accidents Number of Lanes Lane Width Shoulder Width Posted Speed Limit Degree of Curve No Passing Zones ch Procedure Single Variable ple Variable Multi «(346 62 using the state—wide accident data base: General - Total accidents - Driver characteristics (age, sex, alcohol related) - Accident type — Time of day, day of week — Roadway conditions Location - Rural or urban - Michigan Department of Transportation districts Geometry - Number of lanes — Lane width - Shoulder width - Posted speed limit - Degree of curve - No passing zones The data set used for general and location analyses consisted of 51,740 accidents that occurred on the Interstate system and the Michigan Trunkline system in 1982. The data set used for analyzing the geometric features consisted of 38,828 accidents that occurred only on the Michigan Trunkline system. The accident data was then divided into two subsets, rural and urban, and another analysis of the geometric features was conducted. The rural data set consists of 10,283 accidents, and the urban data set consists of 12,022 63 accidents. The 16,518 accidents which occurred in strip-fringe areas were excluded from these analyses. Lastly, 6,907 accidents which occurred at midblock locations in rural areas were used to study geometric conditions, and the relative accident involvement for the seven automobile size groups at rural midblock locations were determined. For example, accidents occurring at midblock locations in rural areas where no passing restrictions apply were analyzed to compare the accident involvements for the seven groups of automobiles. The results obtained by the second and third stage analyses are discussed later. 7.1 Statewide Results Results obtained by examining A/E ratios with a single independent factor were that small automobiles have a higher ratio in the following conditions: - single vehicle accidents - overturned vehicle accidents - late at night - on weekends - in darkness — on icy or snowy highway surfaces - at midblocks — in rural areas - on 2 lane-2 way highways - in no-passing zones 64 On the other hand, large automobiles were found to have a higher ratio in the following conditions: - accidents with pedestrians - accidents with parked vehicles - accidents with other vehicles - at intersections - in urban areas The statistical procedure used in this study consists of the —square test utilizing the observed number of VEH #1 accidents and the expected number of VEH #1 accidents. The expected number of VEH #1 accidents is obtained by equation 3 (pageiw). The relationship is determined to be statistically significant if the level of confidence of the test is greater than 0.95. Appendix A includes the number of VEH #1 accidents, the number of VEH #2 accidents, and the ratio of actual to expected number of accidents by curb weight for each study parameter. Appendix 8 includes the number of VEH #1 accidents, the number of VEH #2 accidents, and the ratio of actual to expected number of accidents by wheelbase for each study parameter. Appendix 0 includes the number of VEH #1 accidents, the number of VEH #2 accidents, and the ratio of actual to expected number of accidents by mode1:year for each study parameter. Using curb weight categories, the equations for the best fitting curves are obtained by testing seven types of 65 curves: linear, exponential, power, logarithmic functions and three kinds of hyperbolic functions. A curve type is selected by comparing the coefficient of determination R2, The ratios of the actual to the expected number of accidents are also obtained for wheelbase groups and by model-year. The curb weight and the wheelbase categories provide very similar results, however the curb weight categories show more consistent results, and are discussed in this study. The automobile model-year is used to examine other possible factors in accident involvement. It was found that newer models have lower accident involvement rate than earlier model automobiles. Newer model-year automobiles have lower A/E ratios regardless of the age of driver, the sex of driver, and the type of accident, except for certain types of accidents for which small automobiles are found to be overrepresented. The average curb weight of the newer model-year automobiles is much smaller than that of the older model-years. Because of this fact, the effect of model-year is sometimes offset by the effect of automobile size. The A/E ratio for total accidents shows a slightly higher value for large automobiles, with the ratio varying from 0.96 to 1.04 (Figure 7.2). Only the two smallest groups (less than 2000 pounds and 2000-2500 pounds) are found to have an A/E ratio lower than one. This is not unexpected because this analysis includes all accidents at 66 1.1 _JEML_. swmun 1.0 0.9 TOTAL ACCIDENTS /\/ AH f Under 20 25 30 35 CURB WEIGHT 40 45 Over x1001b Figure 7.2. Total Accidents by Curb Weight 67 all locations on the Interstate and Trunkline system. The data points are based on relatively large numbers of accidents, with the actual and expected accidents for the seven vehicle classifications ranging from 2362 for the largest size vehicle class to 12,617 for the 3000-3499 pound vehicle class. The same general trend is true regardless of the sex of the driver, as shown in Figure 7.3. Contrary to some reports, there is no evidence that females are "safer" drivers than males. These results indicate that any difference in the number of accidents involving males or females is explained by their relative exposure more than their sex. The results of an analysis of accidents in which the driver was suspected to be under the influence of alcohol or drugs were interesting. There was a clear trend toward a higher percentage of drivers of large vehicles being under the influence of alcohol or drugs. This may be explained by the fact that younger drivers tend to drive older automobiles which are generally larger than new vehicles. Since there was no reliable data on the number of VEH #2 drivers that might have been driving while under the influence of alcohol or drugs, it was not possible to obtain a value for the expected number of involvements by VEH #1 drivers. Thus, Figure 7.4 simply portrays the involvement as a percent of all VEH #1 drivers, and not the A/E ratio. When the type of accident was analyzed, some much 68 1.1 [firt Male Female _ysiti. 8W1) 1.0 sex or DRIWBR 0.9 l 1' Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.3. Accidents by Sex of Driver 69 19 18 DRINKING OR USE F DRU 17 16___ 15 E g .14 Id 0 a: K. 2: / 13 k// 12 L\\\figg ‘ 11 _ 10 I -1 1 Under 20 25 30 35 45 Over x1001b CURB WEIGHT Figure 7.4. Drinking/Drugs 70 stronger relationships were found between accident involvement and automobile size. Small cars are much more likely to be involved in single car accidents than their exposure would indicate, with an A/E ratio of 1.11 for those automobiles which weigh less than 2000 pounds. The ratio decreases consistently with vehicle size with each of the two large vehicle classes having an A/E ratio of approximately 0.9 as shown in Figure 7.5. Due to the analytic procedure being used, a result above the 1.0 axis for a small curb weight has a nearly equal and opposite result for a large curb weight. This dependency is prevalent. Since different types of accidents tend to occur in urban and rural areas, it is not clear whether this observation is related only to vehicle size or also to location. As might be expected, the same phenomenon is exhibited when overturned vehicle accidents are analyzed. In this case, however, the ratio varies even more as a function of vehicle size, with the smallest automobiles having an A/E ratio of 2.4, while vehicles over 3000 pounds have a ratio of less than 0.5 (Figure 7.6). This indicates that the probability of overturning in a small vehicle is 4.8 times as likely for the same level of exposure. This may indicate a need to carefully review highway design standards as the vehicle fleet continues to get smaller. The probability of being involved in a two car accident appears to be independent of vehicle size (Figure 7.7). When the A/E ratio is plotted against vehicle size, 71 1.1 ‘k\ ._MEU_. Ennaui) 1.0 0.9 SING VEHICLE ACCIDENT T Under 20 25 3o 35 40 45 Over CURB WEIGHT x1001b Figure 7.5. Single Vehicle Accidents 72 OW ERTURNI ED ACC. IDENTS K4 H Under 20 25 3O 35 CURB WEIGHT 40 45 Over x1001b Figure 7.6. Overturned Accidents 73 1.5 VEHfl «4 Efifiitfl ACCIDELTS WITH OTHER VEHhCLES Under 20 25 3o 35 CURB WEIGHT 4O 45 Over x1001b Figure 7.7. Other Vehicles 74 the result is nearly a horizontal line at an A/E ratio of 1.0. Large vehicles exhibit an A/E ratio higher than one for accidents with parked vehicles, bicycles and pedestrians (Figures 7.8, 7.9 and 7.10). It is not known whether this is related to the size of the automobile or the location in which most of these accidents occur (urban areas). We will gain some additional information on this when the urban and rural areas are separated for further study. When the' accident data was stratified by road condition, two interesting trends were noted. There is a relatively large difference between the A/E ratio for small and large cars on icy roads, with the large cars having the lower ratio as shown in Figure 7.11. The reverse is true on both dry and wet roads, where small cars have the lower ratio. The differences in both cases are roughly 20 percent. The findings discussed above are all statistically significant. 7.1.1 Findings for Locations The relationship between accidents and highway area types (intersection or midblock), road classes (Interstate, U.S. routes, and Michigan routes), and state highway districts were also investigated. Small automobiles are found to have an A/E ratio greater than 1.0 at midblock locations and lower than 1.0 at intersections on all 75 VEH 1 RIVER! 15 1 ACCIdENTs warn PARKED V#HICLEE Under 20 25 30 35 40 45 Over x1001b CURB WEIGHT Figure 7.8. Parked Vehicles 76 E VEfitl Htl £y_£5nmm rams H FIXED OBJEC‘ i Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.9. Pedestrian/Fixed Object 77 1.5 vain E #1 A ‘\ \\\ \ A 1.0 ‘\ 4” ‘ x\ A" 4 . \A \\ A 0.5 AA ACCIiENTS ITH ANIMALS H ACCI ENTS ITH PitDALCYiELBS Under 20 25 3o 35 4o 45 Over xlOOlb CURB WEIGHT Figure 7.10. Animals/Bicycles 78 CURB WEIGHT J 1.1 _ML. Eflnifll) 1.0 0.9 SURFACE CO ITIOJ ER\ Zk'iSDRy \ w... \ ICY. FNOWY ' J 4' Under 20 25 30 35 40 45 Over x1001b Figure 7.11. Surface Condition 79 classes of road. The ratio for the group of smallest automobiles is 21 percent higher than that for the group of largest automobiles at midblock locations for all roadway types combined. At intersections, the largest automobiles have about a 25 percent higher ratio than the group of smallest automobiles. These results are illustrated in Figure 7.12. These results are consistent with the observations for single and multiple vehicle accidents. Small automobiles appear to be involved in a disproportionate number of single vehicle accidents at midblock locations, while large automobiles are involved in a disproportionate number of accidents at intersections. This may explain the urban-rural differences noted previously, since intersection accidents tend to dominate the urban accident experience. The A/E ratios for U.S. routes and Michigan routes follow the same trend when plotted against vehice size as shown in Figure 7.13. Thus, data from both groups were combined for the analysis. Figure 7.14 shows the results for three of the 9 state highway districts. These results are consistent with previous findings. For example, in District 4, large automobiles have a lower ratio than small automobiles, while District 8 and District 9 have the reverse trend, with the higher ratio associated with large automobiles. This is probably due to the differences in the traffic environment, as District 9 (metro area) and District 8 are 80 i A ’\ ll \ 1.1 ; ‘x o I, \ _vsm_ ' , \ 3W” , A I / / I A / I], 1.0 _ T I / I , I / I I l I 009 43 / / / ,’ 816mm: AREA TYPE J A A‘ .A INTERir-zc'rro HIDBLQLCK 1. T Under 20 25 30 35 40 45 Over x1001b 'CURE WEIGHT Figure 7.12. Area Type 81 1.1 flflfll E[VEH#1) 2 \ ;§\\ ll, /\ I L0 i/ ‘1 a; 2 \ ,/ \ [4 \\\ A}? a [17’ 0.9 ROUTE Lr-A [INC] t CLAS? US R0 MICHI TE RT 20 25 30 35 CURB WEIGHT Under 40 45 Over x1001b Figure 7.13. Route Class 82 fi DD DISTRICT 4 H DIST ICT 8 H DIST ICT 9 1.1 _ML Emu) 1.0 \F{ 0.9 4" 1’ Under 20 25 30 35 40 45 Over x1001b CURB WEIGHT Figure 7.14. District 83 primarily urban, and District 4 is basically a rural district. 7.1.2 Findings for Geometric Features File 2 (the file consisting of information on geometric features and accidents) was used to determine the relationship between vehicle size, accidents and selected geometric features. This data file consists of the 38,828 accidents which occurred on the Michigan Trunkline system. The A/E at any location is calculated by Equation 6 (page 46). The geometric features investigated in this phase of the study were: - number of lanes ~ lane width - shoulder width — posted speed limit - roadside development - no-passing zones — degree of curve — number of intersection legs (intersection, midblock) Figures 7.15 through 7.17 show the results obtained for some of the features listed above. The common denominator for the results of these analyses is that small automobiles have a ratio greater than one for features associated with rural areas, and a ratio less than one for features common to urban areas. Small automobiles are found to be over-represented in accidents at the following 84 VEH! l 21mml1 f/Eik 1.1 4 \ r71 [/KL—S A I : m9 ‘~. ’ i 3 0.8 ' H 21ane1»2way A'fl 4lane<72way D'D Slane-vaay ONO 41anevdevidld 0. 7 I l T Under 20 25 30 35 4o 45 Over x1001b CURB WEIGHT Figure 7.15. Number of Lanes 85 ‘mfill E‘flflll 0.9 0.7 [> PASS NO E [NG AL: ASSING LOWED i .r Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.16. No Passing 86 CURB WEIGHT LT— wmn E‘flflll IJ §\ \ \ \ i . LO . m9 }“ m8 A—«A 0-2 DEGREES D—--CJ 3-4 5- OJ 1L 1’ Under 30 35 40 45 Over x1001b Figure 7.17. Degree of Curvature 87 features: - 2 lane-2 way roads 55 mph posted speed limit - no-passing zones midblocks These features are characteristic of rural areas, and the results confirm that small automobiles are more likely to be involved in an accident in rural areas. This could be because small automobiles are less stable when driven at the high speeds than are large automobiles. Consistent with the urban-rural dichotomy, small automobiles are found to have a ratio of less than one for the following features: multiple lanes - cross sections with curbs - low posted speed limits - curves with a degree of curve of 5 degrees or more - intersections zones in which passing is allowed These features are characteristic of urban areas. Small automobiles may be able to maneuver better than large automobiles in urban areas, and there are more vehicles, pedestrians and bicycles in urban areas. As described earlier, curves were fit to the data points shown in the preceding figures, and an equation of the line of best fit determined for each relationship. Table 7.1 contains the equations for the relationships between the A/E ratio and vehicle size using the entire 88 Table 7.1. The Best Fitting Curves Condition 2 lane-2 way Curb 25 MPH 55 MPH Rural Urban Passing Allowed No Passing 5—Degree Intersection Eggation Y: K 0< < K: '< < K: *< < II where Y = The ratio of the of accidents x = Curb weight .875 + X/(907 X/(727 1.16 - 1.25 - X/(807 X/(429 .932 + X/(85O X/(766 actual for Geometric Features R2 387 /X .759 + .708 X) .879 + .786 X) .536 .0000545 X .768 .0000780 X .954 + .742 X) .877 + .865 X) .738 230 /X .524 + .728 X) .796 + .752 X) .940 to the expected number R2 = Coefficient of determination 89 accident data base. It is reasonably clear from the preceding analyses that one of the major variables in explaining the over- or under-representation of different size vehicles in accidents is the urban-rural split. In almost all cases, those locations and geometric features related to urban areas exhibit a small automobile A/E ratio lower than 1.0, and those features related to rural areas have an A/E ratio greater than 1.0. Because there is also a significant difference between accident involvement ratio at intersections and non—intersections, it is probable that there is some interaction between this factor and the urban-rural factor. To separate these two factors, the accident data set was divided into urban, fringe area and rural subsets. The accidents occuring in fringe areas were then set aside, and separate analyses conducted on the urban and rural subsets. The data sets are still relatively large, with the urban subset containing 12,022 accidents and the rural subset containing 10,283 accidents. 7.2 Relationships for Urban-Rural Subsets The A/B ratios were obtained for several location factors and geometric features in rural and urban areas separately. The relationships among automobile design, geometric features and accidents were expected to become more clear when the accident data were separated, since the relative value of these ratios seemed to be opposite each 90 other for urban and rural areas. Once again, the distribution of accidents by type were plotted for both rural and urban areas to verify the use of VEH #2 accidents as an unbiased estimator of exposure. Figure 7.2.1 shows the distribution of VEH #2 accidents by weight class for '"angle," "left-turn," and "rear-end" accidents in rural areas. Figure 7.2.2 shows population distributions of VEH #2 accidents by weight class for "angle," "left-turn," and "rear-end" accidents in urban areas. Statistical tests were conducted and the null hypothesis "population distributions of VEH #2 are the same" is accepted at a = .975 for both cases (Table 7.2.1). Thus, the use of the new exposure approach is supported for both rural and urban areas. The accident file consisting of _information on geometric features and accidents (File 2) is used in the following studies. The data base consists of 10,283 accidents for rural areas, and 12,022 accidents for urban areas. The accidents occurred on the Michigan Trunkline system. The expected number of accidents is calculated by Equation 6 on page 46 for the following categories: - total accidents (rural, urban) - highway area type (intersection or midblock) - number of lanes lane width PERCENTAGE (5) 91 RURAL ANGLE “.0 LEFT- RN .__. REAR- ND 20 1. .. §.\ i f e I i J , Under 20 25 30 35 40 45 Over x1001bs Figure 7.2.1. Distribution #3 PERCENTAGE (%) N O 10 92 URBiN ANGLE LEFT- RN REAR- ND \ y A 4 \ - \ \ \ W A Under 20 25 7 3o 35 4o 45 Over x1001bs Figure 7.2.2. Distribution #4 93 Table 7.2.1. Statistical Test for Population Distributions of VEH 02 Accidents for Rural and Urban Areas. RURAL Weight Class 1 2 3 4 5 6 7 Total Ob'°’ved 6.39 15.28 18.32 25.35 19.50 11.05 4.01 100.00 (Angle) Expected ' 6.14 17.53 18.34 23.30 17.75 13.39 3.55 100.00 (Rear-End) 2 -£QE§1- .01 .26 .oo .18 .17 .41 .06 1.09 x2 - 1.09 URBAN weight 1 2 3 4 5 6 7 Total Class Observed 4.17 13.97 17.34 25.10 20.70 13.31 5.41 100.00 (Angle) Expected 7 7 (Rear_end) 6.03 15.31 1 .24 25. 1 18.59 12.06 5.06 100.00 O-E 2 E .57 .12 .oo .01 .24 .13 .02 1.09 x2 - 1.09 94 shoulder width posted speed limit degree of curve no passing zones As expected, the results are more consistent than the results obtained when all accident data were combined. There were 8,559 accidents on rural 2 lane—2 way highways and 878 accidents on urban 2 lane-2 way highways. When the accident data were separated, the coefficient of determination (R2) for total accidents on rural 2 lane-2 way highways is .993. This value is considerably greater than .759 which was obtained for 2 lane-2 way highways for all highways combined. 7.2.1 Findings for Rural Areas The relationship between accidents and geometric features in rural areas shows remarkable consistency for these geometric features typically found in rural areas; The following geometric features were analyzed: - total accidents - midblock accidents - 2 lane-2 way highways — no-passing zones — 10 ft., 12 ft. lane width - highways with shoulders — 0-2 degrees of curve Figure 7.2.3 shows the results for total accidents on rural Michigan trunklines. Small automobiles are found to 95 RURNL VEHII Bhflfilli 1.1 1.0 0.9 TOTAL ACCID NTS 0.7 ,L -f Under 20 25 30 3S ‘CURB WEIGHT 40 45 Over x1001b Figure 7.2.3. Total Accidents 96 have a higher A/E ratio than large automobiles with the difference being about 26 percent. Figure 7.2.4 shows the results for highway area types, with the accident data stratified into intersection and midblock locations. The number of accidents at midblock locataions is dominant in rural areas, with 3,194 accidents at intersections and 7,094 accidents at midblock locations. The group of the smallest automobiles is found to have an A/E ratio 48 percent higher than the group of the largest automobiles at mid-block locations. This stratification helps explain the accident phenomena, as the intersection accidents A/E ratio is not related to vehicle size on rural highways. Figure 7.2.5 shows the results by number of lanes. There were 8,559 accidents on 2 lane-2 way highways and 874 accidents on 4 lane—divided highways. The sample size was inadequate for any other lane configuration in rural areas. 0n 2 lane-2 way highways, small automobiles are found to have a high A/E ratio. The ratio for the group of the smallest automobiles is about 36 percent higher than the group of the largest automobiles. No trend is apparent for the 4-lane divided highways, as the accident data appears to be randomly distributed over automobile size. The results for lane width in rural areas are shown in Figure 7.2.6. There were 1,730 accidents on 10 foot lanes, and 5,494 accidents on 12 foot lanes. Small automobiles are found to have a higher ratio than large automobiles regardless of lane width. However, small automobiles seem 97 RURAL CURB WEIGHT L0 03 o 8 BIG BY AREA TYRE ‘ AHA MIDBILOCK 5 % INTERSECTIJyN 37 ,L —r Under 20 25 30 35 4O 45 Over x1001b Figure 7.2.4. Area Type 98 RURAL 1 ,2 /\ VEH“ ‘\ ‘ 8 [mm x ‘1 0.8 LANEAGE A'A ZLANlF-ZWAY H 4LAN11-DIVI ED 0.7 ’ ‘f \ Under 20 25 30 35 40 45 Over x1001b CURB WEIGHT Figure 7.2.5. Number of Lanes 99 mnmh L9 VEHII Ewmn lJ L0 ‘ \ \ \ \ \ \ \ 39 X \ \ \ \ \ LAWS WIDTH \ I \ 38 A--A 10 PT H12 PT OJ ,9 —r Under 20 25 30 35 CURB WEIGHT 40 45 Over x1001b Figure 7.2.6. Lane Width 100 to be more sensitive to narrow lane width than are large automobiles. The trend toward a lower A/E ratio as vehicle size increases on 12 foot lanes is less than that for total accidents, and probably reflects this general trend rather than a relationship to lane width. The variation in the A/E ratio on 10 foot lanes is more erratic than for the wider lanes, and the relationship appears to be more pronounced. With the exception of the smallest automobile groups and the largest automobile groups (which are based on an expected value of 36 and 24 respectively) the trend is fairly consistent. More data will be required to determine whether the extreme points are as shown in the figure, or whether these variations are due to small samples. The results for the shoulder width in rural areas are shown in Figure 7.2.7. There were 3,461 accidents on highways with 4-8 foot shoulder width, 5,797 accidents on highways with 8-10 foot shoulder width, and 432 accidents on highways with curbs. Small automobiles are found to have a higher A/E ratio on highways with shoulders and a lower ratio on highways with curbs. The best fitting curves show that the ratio for the group of the smallest automobiles is about 32 percent higher on highways with 4-8 foot shoulder, and about 24 percent higher than the group of the largest automobiles where the shoulder width is 8-10 feet. These results are difficult to interpret, and may be complicated by other factors. The curbed sections are 101 RU 1., 1.1 1.0 0.9 0.8 ‘4- : SHOULDER IDTHV I A-A cm Arab. 0“)” P) 8-10;T—___ ~ Jh‘ OJ I II I L :15: ;. $.454 Under 20 25 30 35 40 45 Over x1001b CURB WEIGHT Shoulder Width Figure 7.2.7. 102 probably in more developed areas with a higher density of intersections, and thus may simply be reflecting accident type. The sections with shoulders appear to reflect the general trend in rural accidents, and thus the A/E ratio may be independent of the shoulder width. An explanation of the relationship between shoulder width and accidents is probably only possible if the data are further stratified by accident type, and single vehicle run-off-road type accidents are analyzed. The results for the degree of curve in rural areas are shown in Figure 7.2.8. There were 9,693 accidents (94 percent of the accidents in rural areas) on highways with 0-2 degrees of curve and 323 (6 percent) accidents on highways with 5 or greater degrees of curve. Small automobiles are found to have a higher A/E ratio than large automobiles regardless of the degree of curve. However, Small automobiles seem to be more sensitive to the degree of curve. The best fitting curves show that the group of the smallest automobiles has about a 35 percent higher ratio than the group of the largest automObiles for 5 or greater degrees of curve and about a 25 percent higher ratio for 0-2 degrees of curve. In this analysis, the expected number of accidents was based on all rural accidents, not just those on curves. This is a different concept than is used for the other figures, and thus the results should be interpreted accordingly. The results for no-passing zones in rural areas are 103 RURAJL 1.2 VEHI1 EWEHM) A 1.1 45- A \ 1.0 0.9 0.8 DEGREE OF CURVE; lyfl 0-2 GREE H 5- CREE: Under 20 25 30 35 CURB WEIGHT 40 45 Over x1001b Figure 7.2.8. Degree of Curvature 104 shown in Figure 7.2.9. There were 7,372 accidents on highways where passing is allowed and 2,911 accidents on highways where no passing is allowed. All VEH #2 accidents in rural areas were used for the exposure in this analysis as well. Small automobiles have a higher ratio regardless of whether passing is allowed. However, the ratio for the group of the smallest automobiles is about 37 percent higher than the group of the largest automobiles in no passing zones, while it is only about 22 percent higher where passing is allowed. This result may be related to the geometry of no passing zones, the difference in eye height of drivers in small automobiles, or some combination of these factors. Additional stratification will be required to make this determination. Table 7.2.2 shows the best fitting curves for the geometric features shown in the preceding figures. 7.3 Findings for Midblock Locations in Rural Areas In an attempt to better understand the relationship among automobile design, geometric features, and accidents, all intersection accidents occuring in rural areas were removed from the file and only those accidents occurring at midblock locations were analyzed. Many of the geometric features being studied are more likely to be related with safety at midblock locations than at intersections. The relationships obtained by examining this data set are better defined than when using combined intersection and non-intersection accident data. Small automobiles are 105 CURB WEIGHT 1mmu L7 wmn flwmnj IA A--. 2 L0 ‘ ”-25 \ as X 38 Au-APASSI 6 AL WED . . NO PA SING 11.1.0178: 37 ,L f Under 20 25 30 35 40 45 Over x1001b Figure 7.2.9. No Passing 106 Table 7.2.2. Best Fitting Curves for Geometric Features in Rural Areas Geometric Features Equation 3? Total Accidents Y = 1.25 — .0000780 X .954 Midblock Y = 1.36 - .000112 .990 2 Lane-2 Way Y = X/(-910 + 1.31 X) .993 10 Ft. Lane Width Y = 1/(.478 + .000168 X) .650 12 Ft. Lane Width Y = 1.21 - .0000708 x .750 4—8 Ft. Shoulder Y = .713 - .871 /X .814 8-10 Ft. Shoulder Y 8 1.24 - .0000735 x .652 0-2 Degrees Y B 1.23 - .0000758 X .928 5-Degrees Y = .683 + 958 /X .620 Passing Allowed Y = 1.20 - .0000670 X .836 No Passing Y = 3.70 - .335 log X .951 where Y = The ratio of the actual to the expected number of accidents X = Curb weight R2 = Coefficient of determination 107 found to have a higher A/E ratio at midblock locations in rural areas for all geometric features studied. In particular, the relationship between accidents and lane width and shoulder width are more pronounced than in the previous analysis. All VEH #2 accident in rural areas are used in determining the exposures. The ' A/E ratio for total accidents occuring in mid-block 'locations in rural areas is shown in Figure 7.3.1. The elimination of intersection accidents helps to illustrate the relationship between this type of accident and vehicle size. The difference in the A/E ratio between the smallest and largest automobiles is about 1.2 versus 0.8. There were 6,224 mid-block accidents on 2 lane-2 way highways. The number of accidents for other laneage are too small to obtain results which are statistically Significant. The results shown in Figure 7.3.2 show less of an effect than the results obtained for all rural accidents. This would suggest that the laneage is related with accidents not only at midblock locations but also at intersections. There were 1,265 accidents on 10 foot lanes and 3,523 accidents on 12 foot lanes. As shown in Figure 7.3.3, small automobiles seem to be more sensitive to narrow lane width than are large automobiles. The best fitting curves show that the A/E ratio for the group of the smallest automobiles is about 59 percent higher on 10 foot lanes, and about 33 percent higher on 12 foot lanes at midblock 108 RURALTMIDBLJDCK VEH! 1 Z ElVEHIl i \ 1.0 {-7 0.9 38 TOTAL ACCIDENTS 1 er Under 20 25 30 35 40 45 Over x1001b CURB WEIGHT Figure 7.3.1. Total Accidents 109 CURB WEIGHT RURAL+MIDELUCR 1.2 VEH11 E(VEHIIJ A 1.1 §\\ 1.0 0.9 >\\\\‘. A 0.8 2LANEv2WAY HIGHWAYS 0.7 L 1’ Under 20 25 30 35 40 45 Over xIOOlb Figure 7.3.2. Midblock Accidents llO RURAL-MIDBLDCK A I, \ I’ \\ 1.2 A", X vaull \ Ewmufl \ lJ L0 03 AMA \ m8 IAMEWHHH OJ "[3 IOFT lk-ik 12?? ,L ’f Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.3.3. Lane Width 111 locations. This compares to 35 and 25 percent respectively when all rural accidents were used in the analysis. There were 2,373 accidents at midblock locations with 4—8 foot shoulder width, 4,006 accidents where there are 8-10 foot shoulders, and 287 accidents where there are 10-12 foot shoulders. The results shown in Figure 7.3.4 are consistent with the results obtained for total accidents in rural areas. Small automobiles have a higher A/E ratio than large automobiles regardless of the shoulder width. The results for 4-8 foot and 8-10 foot shoulder widths are statistically significant, but the sample size for 10-12 foot shoulder width is too small to obtain statistical significnce. There were 4,896 accidents at midblock locations where passing is allowed and 2,011 accidents at midblock locations where no passing is allowed.‘ Once again, the greatest difference in the A/E ratio as a function of automobile size is in the areas with no passing allowed (Figure 7.3.5). The best fitting curve for no passing zones shows that the ratio for the group of the smallest automobiles is about 56 percent higher than that for large automobile ratio compared to about 33 percent higher for midblock locations where passing is allowed. Figure 7.3.6 shows the results for overturned accidents at midblock locations in rural areas. The results show a very similar tendency to the results obtained for overturned accidents for statewide data. Small automobiles are found to be more involved in 1112 CURB WEI GHT */;.35fi \_ RURALPMIDBLOCK 1.2 \\k 4‘“ ‘G vsall AA £[vautlj ‘~"’< 1.1 . Z§\, \\‘\ \\ E \ 7 \\ 1.0 0.9 SHO?LDER WIDTH 0.8 zfiy-ifil4-8 FT 0.0 8-10 1="r ‘ u. 12 FT ‘ O C) 0.7 ‘\2 \o t 4’ Under 20 25 30 35 40 45 Over x1001b Figure 7.3.4. Shoulder Width 211.3 VEHII £[VEHIlj RURALuMIDBL4 )CK 0.9 0.8 passxnc ALLOWED A"A H NO pussmc immwai: L ”f Under 20 25 3O 35 CURB WEIGHT 40 45 Over x1001b Figure 7.3.5. No Passing 114 VEHtl EWEHQ 1] RURALWMIDBLOCK OVERTURNED ACCIDENTJ Under 20 25 3O 35 40 45 Over CURB WEIGHT x1001b Figure 7.3.6. Overturned 115 overturned accidents than large automobiles at midblock locations in rural areas, with the ratio being about 4:1 for the extremes. Table 7.3.1 shows the equation of the best fitting curves for geometric features at midblock locations in rural areas. 7.4 Findings for Urban Areas Small automobiles are found to have a lower A/E ratio than large automobiles for all geometric features in urban areas. The results for the following geometric features in urban areas are discussed: - total accidents - intersections - multi—lane facilities The results for total accidents in urban areas are shown in Figure 7.4.1. Small automobiles have a lower A/E ratio than large automobiles, with the group of the smallest automobiles about 26 percent lower than the group of the largest automobiles. The results for intersections and midblock locations in urban areas are shown in Figure 7.4.2. There were 9,715 accidents at intersections and 2,307 accidents at midblocks in the sample. Intersection accidents appear to have a consistent relationship with vehicle size, with the small automobiles having the lower A/E ratio. The group of the smallest automobiles is about 29 percent lower than the group of the largest automobiles. Conversely, the 116 Table 7.3.1. Best Fitting Curves for Geometric Features at Midblock Locations in Rural Areas Geometric Features Total Accidents 2 Lane-2 Way 10 Ft. Lane Width 12 Ft. Lane Width 4-8 Ft. Shoulder 8—10 Ft. Shoulder 12 Ft. Shoulder Passing Allowed No Passing Eguation Y = 1.36 - .000112 Y = 1.43 exp (-.000113 X) Y = l/(.501 + .000163 X) Y = 1.28 - .0000929 X Y = 3.54 - .316 log x Y = 1.40 exp(-.000108 X) Y a 6.89 - .733 log K Y = 1.31 - .0000961 X Y = l/(.522 + .000157 X) 3.2 .990 .984 .934 .817 .956 .945 .909 .961 .940 where Y = The ratio of the actual to the expected number of accidents X R2 Curb weight Coefficient of determination 117 lflflll W 0.9 TOTAL ACCIDENTS 0.8 b ( \ Under 20 25 30 35 40 45 Over XIOOID CURB WEIGHT Figure 7.4.1. Total Accidents 118 wmn EWBUI URsAfi 1., IA \}§ II I, 1.0 I! 009 0‘8 HIG AY A ryp‘ Ck' tuba K INTE ECTION 0.7 i ( Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.4.2. Area Type 119 mid-block accidents appear to be randomly distributed over vehicle size. The relationship between accidents and the number of lanes in urban areas are shown in Figure 7.4.3. There were 878 accidents on 2 lane-2 way highways, 1,820 accidents on 5 lane-2 way highways, and 1,481 accidents on 6 lane—divided highways. All of the laneage combinations show the same trend as all urban accidents. It does not appear that there is a significant difference in results as a function of laneage. The group of the smallest automobiles is found to have about a 50 percent lower ratio than the group of the largest automobiles for all lane combinations. The equations for the relationships discussed above are given in Table 7.4.1. 7.5 Results for Driver Injuries To eliminate any bias due to a difference in the number of occupants by different automobile designs, only injuries sustained by the drivers are used in the analyses. Fatal (F-injury), incapacitating (A-injury), non-incapicitating (B-injury), and possible injury (C-injury) of drivers are considered as the injury classifications. Driver injuries in VEH #1 and driver injuries in VEH #2 are analyzed separately. The number of driver injuries in VEH #1 and VEH #2 by curb weight are displayed in Table 7.5.1. 120 mm E (van: 1 OJ [firigznmm-mnx C}%jsnmmmflmy Hams-1 DIVIDiD $ Under 20 25 30 35 40 45 Over CURB WEIGHT x1001b Figure 7.4.3. Number of Lanes Table 7.4.1. 121 in Urban Areas Geometric Features Total Accidents Eguation Y = X/(807 + .742 X) Intersection Y = X/(918 + .706 X) 6 Lane-Divided Y = .511 + .000151 X 10 Ft. Lane Width Y = X/(1190 + .626 X) 12 Ft. Lane Width Y = X/(905 + .711 X) where R Best Fitting Curves for Geometric Features 33 .877 .905 .972 .847 .843 Y = The ratio of actual to the expected number of accidents X = Curb weight 2 '= Coefficient of determination 122 Table 7.5.1. Number of Driver Injuries in VEH #1 and VEH #2 by Curb Weight (1982) Less 2000- 2500- 3000- 3500- 4000- 4500 Than 2499 2999 3499 3999 4499 lbs or 2000 lbs lbs lbs lbs -lbs lbs more Total (VEH #1) Injury 571 1761 1889 2116 1547 923 318 9225 F + A 128 278 348 350 243 142 48 1530 (VEH #2) Injury 554 1517 1469 1762 1352 818 288 7760 F + A 67 194 176 195 112 60 23 827 The A/E ratio for automobiles of any class (D1) is calculated using Equation 4 (page 40). Drivers in small automobiles were found to have a greater risk of being injured. Driver injuries in VEH #1 and driver injuries in VEH #2 are analyzed separately, with the drivers of vehicle number 1 in the group of smallest automobiles found to have a 72 percent higher risk of being injured, and a 108 percent higher risk of being seriously injured than drivers in the group of largest automobiles for the same exposure. Similarly, drivers of vehicle number 2 in the group of smallest automobiles have a 56 percent higher risk of being injured, and a 178 percent higher risk of being seriously injured than drivers of the largest automobiles. The results for VEH #1 are shown in Figure 7.5.1. This shows that drivers of small automobiles are more likely to be injured (and more likely to sustain a serious injury) given that an accident occurs, than are drivers of larger automobiles. The method of least squares indicates 123 AuA INJURj . ‘ ”FII+I II L5 In2.Acc. IBInj.Aoc Lo 1 . \\\ 5| i I P I 0.5 7 DRFVER Ipauny IN vsnpi Under 20 25 30 35 40 45 Over xlOOIb CURB WEIGHT Figure 7.5.1. Driver Injury VEH #1 124 that a hyperbolic function Y = 1/(.422 + .000189 X), is the curve of best fit. In this equation: Y is the ratio of the actual to the expected number of injury accidents, and X is the curb weight. The coefficient of determination R2 = .971. h The best fitting‘ curve for the F + A injuries is a hyperbolic function: Y = 1/(.222 - .000258 X) with R2 = .949. Using these curves the drivers in automobiles weighing less than 2000 lbs. would be estimated to have a 72 percent higher_ risk of being injured and a 108 percent higher risk of being seriously injured than drivers in automobiles weighing 4,500 lbs. or ‘more for the same exposure. The results for VEH #2 are shown in Figure 7.5.2. The results are similar to those for drivers of VEH #1. The best fitting curve is a power function: Y = 1.39 x X-o'413 with R2 = .963 for all injuries, and an exponential function: Y = 2.88 exp (.000348 X) with R2 = .932 for serious injuries. These curves provide estimates that the drivers in automobiles weighing less. than 2000 lbs. have a 56 percent higher risk of being an injured driver of VEH #2, and a 178 percent higher risk of being a seriously injured driver of VEH #2 than drivers in automobiles weighing 4,500 lbs. or more for the same exposure. This suggests that the likelihood of being seriously injured in VEH #2 would be greatly affected by the size of the vehicle. The results of a least square fit of the linear transform of these equations are shown for 125 0.5 .,V."”_.#_._ir._unnm__1 DRIVER INJURY IW vantz zfi-thmmafi . . wa+n1Fo 1 Under 20 25 30 35 40 45 Over xlOOlb CURB WEIGHT Figure 7.5.2. Driver Injury VEH #2 126 VEH #1 and VEH #2 in Tables 7.5.1 and 7.5.2 respectively. Although small automobiles have a lower A/E ratio for certain geometric and locational parameters, once a small automobile i§_ involved in an accident the driver has a higher risk of being injured than drivers of larger automobiles. The likelihood of being injured when a driver is involved in an accident with automobiles of various sizes was also examined. The number of injured drivers in automobiles of design 01 in collision with automobiles of design DZ were obtained for the seven groups of curb weight. Table 7.5.4 shows the number of injured drivers in VEH #1 as the result of accidents involving two passenger automobiles. The table entries are recorded as the driver in an automobile of design Di, listed in the row, in collision with an automobile of design D2, listed in the column. For example, only four drivers in automobiles weighing 4,500 lbs. or more were injured when their automobiles hit automobiles weighing less than 2000 lbs. Table 7.5.5 shows the number of injured drivers in VEH #2 as the result of accidents involving two passenger automobiles. For this table, the entries are for drivers in an automobile of design D1, listed in the row, in collision within an automobile of design 02, listed in the column, sustaining an injury. For example, nineteen drivers in the group of smallest automobile were injured when their automobiles were hit by an automobile from the group of largest automobiles. This corresponds to the four 127 Table 7.5.2. Results of a Least Square Fit of Its Linear Transform for Driver Injury in VEH #1 * Injury Y = 1/(.422 + .000189 X), R2 = .971 X-ACTUAL Y-ACTUAL Y-CALC PCT DIFFER 1867 1.239 1.28901 -3.8 2245 1.183 1.18017 .2 2701 1.159 1.07107 8.2 3223 .939 .968573 -3 3729 .884 .886352 - .2 4215 .811 .819533 -1 4802 .755 .751139 .5 * F+A Y = 1/(.222 - .000258 X), R2 = .949 X-ACTUAL Y-ACTUAL Y-CALC PCT DIFFER 1867 1.425 1.4214 .2 2245 1.126 1.24831 -9.7 2701 1.261 1.08842 15.8 3223 .936 .949236 -1.3 3729 .837 .844549 - .8 4215 .752 .763657 -1.5 4802 .687 .684473 .3 X = Curb weight (lbs.) Y = The ratio of the actual to the expected number of accidents Table 7.5.3. * Injury Y X-ACTUAL * F+A Y = 2.88 exp(.000348 X), R2 = .932 X-ACTUAL X 1867 2245 2701 3223 3729 4215 4802 1867 2245 2701 3223 3729 4215 4802 128 Results of a Least Square Fit of Its Linear Transform for Driver Injury in VEH #2 = 1.39 x x"°‘13. R2 = .963 Y-ACTUAL 1.216 1.211 1.071 .929 .918 .854 .812 Y-ACTUAL 1.38 1.454 1.204 .965 .714 .588 .609 = Curb weight (lbs.) Y-CALC 1.25558 1.15171 1.05616 .972279 .908092 .857454 .806661 Y-CALC 1.50506 1.31959 1.12601 .939007 .787435 .66494 .542112 PCT DIFFER -3.1 PCT DIFFER Y = The ratio of the actual to the expected number of accidents 129 O 2~> >m NIw> O . N.Ov 0mm 0.5— .m03 m.nm hum ..ON 0.5 h.m. Oww ache» 30a 0 O O O Q ammo? u m20_»<>awmmo ozumm~1 no ameDZ N.w O.m¢ w..N N.VN '“wp 0.N¢ 0.0 thOh m.n Onm hwh 0mm mwm flmv mm. 223400 — 11111111 — 11111111 — 11111111 u 11111111 u 11111111 u 11111111 ~ 11111111 —I ~ _ ~ mac: do mm; OOmv _ w ¢N F“ on ON n. AMWV u .h a 1111111111111111111111111111111111111111111111 u 11111111 n: ~ — mm; mmvv1000v ~ mm vm hm mm 0m. #0 N. u .w n 11111111 u 111111111111111111111111111111111111111111111111 ~n — ~ mm; mmmnuoomn u mm mm fin. wV— vm nm .N u .m H 111111111111111111111111 ~ uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu ~1 ~ ~ mm; meDIOOOn ~ Om —N. .m— wmv NV. mm mm n .v u uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu nu — ~ mma mmmNuOOmN ~ mv mow pfiw v@— 0" #0 ND — .fl ~ 111111111111111111111111111111111111111111111111 ~ 11111111 ~1 _ _ mm; mmvn-ooo~ u mm '0— mm, wh¢ mm 00 RN ~ .N ~ uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu ~ 11111111 u: n — 4 000” 241» mmwa n as NO mv vm mv Q” 0. n .w ~ 111111111111111111111111 m 1111111111111111 u 111111 tau 111111 :1—111 11111 uh~m3> ~.h .0 .m .v .n .N .— p deZ do mm; 9 mm; m wma m mmJ 0 mm; m J OOON Zn mm; OOmv QVVIOOOV mmncoomn mvnnOOOn mmN1OOMN QVNuOOON (Ih mmw4~ ~ #2300 Nh~w3> 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 bIO~w3 ado n»—w3> >m 2~> >0 .Iw) FIG—w) ado .hnw3> 4 4 4 4 u 0 2 D — h < J D m < h m m C a U 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 H coflumaonmummouu .v.m.h want 5055 a m20~h<>awmmo 02—mm_2 no uwszZ 130 0.00. 0 n «.0. 4.4. n.un «.0. 4.4. o 0 34.0» 0m44 mm. 4m4 so. .40. 4mm who .0n 220300 3 ........ 3 ........ 3 ........ 3--------3--------3 ........ 3 ........ 3- 0.m 3 3 3 3 3 3 3 3 340: no mm3 00m4 043 3 m 3 mm 3 E 3 mm 3 on 3 n4 3 e 3 .4 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3-- -----3- 4.4. 3 3 3 3 3 3 3 3 mm3 mm44-0004 000 3 ma 3 00 3 ma. 3 n4. 3 no. 3 4.. 3 m4 3 0 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3- 0.- 3 3 3 3 3 3 3 3 mm3 mmmn-00mn 400. 3 4n 3 .0. 3 44. 3 o.n 3 .08 3 004 3 0» 3 m 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3- 4.mn 3 3 3 3 3 3 3 3 mm3 444n-0oon 04.. 3 4n 3 m0. 3 00m 3 .0“ 3 man 3 nnu 3 00 3 4 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3- 0.0. 3 3 3 3 3 3 3 3 mm3 mmmn-00mn 0.. 3 mm 3 .4 3 .3. 3 n0. 3 4n. 3 .4. 3 3m 3 n 3 ........................................................ 3- 0... 3 3 mm3 mm4w-000n m.m 3 m. an m. 4a. 40 40. An 3 .n 3 ............. r--3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3- m n 3 3 3 3 3 3 3 3 3 000a 24:4 mm33 0.. 3 4 3 .u 3 an 3 on 3 mm 3 n4 3 n. 3 .. 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ....... -3 ........ .333:> 3 4 3 0 3 m 3.4 3.n 3 a 3 . 3 3430. 340: 40 mm3 m mm3 m mm3 m mm3 m 003 m 3 0008 23 304 mm3 00m4 444-0004 mmn-00mn 044-000n mm~-00mn m4n-000n 4:3 mm333 3 .2300 Nb—w3> 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 2—> >m NIw> FIG—NB U >m Zn) >0 —Iw> h10~w3 adv wh~w3> 4 4 4 4 4 4 4 4 4 4 4 u C 2 O — h d .— 3 m < .P m m 0 a U 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 m 003343004344030 .m.m.n 03349 131 drivers of the large automobiles who were injured when they hit the small automobiles. Table 7.5.6 shows the number of two passenger automobile accidents with VEH #1, listed in the row, and VEH #2, listed in the column. For example, there were 58 accidents that involved the group of largest automobiles as VEH #1 and the group of smallest automobiles as VEH #2. The numbers from Table 7.5.4 and Table 7.5.5 are divided by the numbers from Table 7.5.6, and the ratios are shown in Table 7.5.7 and Table 7.5.8 respectively. The numbers in these tables are the average number of injured drivers in VEH #1 and VEH #2 per accident. For example, there were .069 injured drivers of the group of largest automobiles per accident in which the group of largest automobiles hit the group of smallest automobiles. 0n the other hand, there were .286 injured drivers of the group of smallest automobiles per accident in which the group of smallest automobiles hit one of the largest automobiles. The ratio (.286/.069 = 4.1) provides a relative likelihood of the drivers of VEH #1 being injured. The numbers shown in Table 7.5.8 illustrate a relative likelihood of being injured for the drivers of VEH #2. For example, the ratio (.328/.255 = 1.3) shows the likelihood of being injured is 30 percent greater for the drivers in automobiles weighing less than 2000 lbs. being hit by an automobile weighing 4,500 lbs. or more than when hit by an automobile of their same weight group. The ratio (.214/.095 = 2.3) shows the likelihood of being injured is 132 w—h-n 4 w20~h<>awmmo UZ—mm—I no ameDZ 0.00. 0.4 0.0. 4.0. 0.40 0.0. 0.0. 4.0 3430» 4000. 000 0000 0000 0004 0040 00.0 000. 220300 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3- 0 4 3 3 3 3 3 3 3 3 000: 00 003 0004 000 3 04 3 00. 3 4.0 3 040 3 04. 3 .0. 3 0 3 .3. 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3 ........ 3-- --3- 0.0. 3 3 3 3 3 3 3 3 003 0044-0004 0.00 3 00. 3 000 3 000 3 .00 3 004 3 004 3 04. 3 .0 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3- 0.0. 3 3 3 3 3 3 3 3 003 0000-0000 4000 3 00. 3 004 3 000 3 000 3 000 3 000 3 4.0 3 .0 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3- 0.00 3 3 3 3 3 3 3 3 003 0040-0000 .000 3 0.0 3 000 3 000 3 000. 3 000 3 000 3 000 3 .4 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3- 4.0. 3 3 3 3 3 3 3 3 003 0000-0000 0040 3 00. 3 004 3 000 3 400 3 0.0 3 000 3 00. 3 .0 3 -------- 3 -------- 3 -------- 3 ----- --3 -------- 3 --------- 3 -------- 3- 0.4. 3 3 3 3 3 3 3 3 003 0040-0000 4400 3 00. 3 000 3 000 3 000 3 004 3 004 3 00. 3 .0 3 -------- 3 ........ 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3- 0.0 3 3 3 3 3 3 3 3 3 0000 24:. 0003 .00. 3 04 3 40. 3 00. 3 000 3 40. 3 00. 3 .0 3 .. 3 ........ 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- 3 -------- .3303> 3.0 3.0 3.0 3.4 3 0 3.0 3.. 3 34303 000: 00 003 0 003 0 003 0 003 0 003 0 3 0000 23 300 003 0004 044-0004 000-0000 040-0000 000-0000 040-0000 4:3 00033 3 32:00 03303> O O O O O O O O O O O O O O C C O O C O O O C O O O O O O O O O O O I C O O O O O 9 O O O O 0 O 2~> >0 NIw> FIG—w) ad Nh_w3> >m Zn) >m .Iw) bIO~w3 adv vh—m3> 44444444444 0.0 ZD~h0 .Iw> h20—0) 8‘0 00.00nv 00.00h0 .h.w)> .000801. 00.00am 00.00.“ 00.00«« 2—> >0 .1») 00‘04001) CCU 8.02.. 8.03. .31; .2303 50 218000..100 136 108.7 inches. the relationship between curb weight and wheelbase can be expressed by a simple regression equation: Y = 71.9 + .0116 X, where: Y = wheelbase, X = curb weight. The correlation coefficient is R2 = .903. The relationship between the curb weight and model—year of the automobiles involved in accidents as VEH #1 is shown in Figure 7.6.2. The variance of curb weight for new model—year automobiles is much smaller than that of older model-years. The average curb weight of new model-year automobiles is also much smaller than that of older model-years. Figure 7.6.3 shows the average curb weight of VEH #1 and VEH #2 for each model-year. The average weight has decreased by about 1000 lbs. from 1973 (3,673 lbs.) to 1982 (2,674 lbs.). The description of the curb weight and wheelbase by model-year are shown in Figure 7.6.4 and in Figure 7.6.5. Figure 7.6.6 shows the relationship between the wheelbase and model-year of automobiles involved in accidents as VEH #1. The same tendency as in the curb weight can be seen. The average wheelbase is also getting shorter. The curb weight, wheelbase, and model-year were all found to be important measures of accident potential. The curb weight and wheelbase categories show similar results. Newer model—year automobiles are found to have a lower A/E ratio than earlier model-year automobiles. The conditions examined by wheelbase and model-year are as follows: 137 N Emnmuwuumom .~.m.5 madman 00.n00. 00..00. 00.000. 00.050. 00.050. 00.550. 00.050. 00.050. 00.V50. 00.050. 00.050. .IIII‘III-‘-'--"-'-.'-'-"'II.""‘|'II.-'""-'I‘-‘l-.---'.-l‘l.-|-"----.'"""--.'--I."I'.--l" { hu—o-n—0 0-00-0-0—9 Q—oo—u—n— 0". N O — .0000 Q—o—n—t—lu—o‘ 0.0—Innin— EDNLD‘DF Ch ”005190101 ammo-00:0 $010101 90,0!an '000'000000 0000 N'Oflo “000 00010 000000.00 000 0 n 19 n O Q If N GOOO'OMO'OO'OQ ...:... u) fifimDQGNQHQFDQmOOOOFQO 0 ”(V OFVQQOOFOU’OO C (N «n OWNONN'GOO 00090100 IDQOOOIDCDOODCDQ 030lnra0lnrinr-v!55-0r-vv~0 0 tit 00' we 0 00 05000000000000000040 000.0 0 0!I0¢no 0o 0(0000f50twdbocflr-0ev0tw 00 HID in. 0 r O NNN'O 00 000001 f" 0 _ (1000010 WOtDOQ'WWCFQ .0000 Q——~—¢————Q—p———¢_——~ 0 O ——0~—¢0—~——9 oo.ooom --. out-«Ont:no-cuusucaaonlalolItIOIIIl0tooIOIIOIQII:IQIIII#OI000|OIIQII01¢!IllfiavluOIIOIQOIIuOOIIu§. 0m..mm. om.owm. 00.050. 00.050. 00.550. 00.050. 00.050. 00.050. 00.050. 00.N5m. .u¢u>> .wmocuq. 2_> >m .:w> »:o_w) can .p.u)> _zroo. no aaaouuppqom Q 138 4000 ‘ CURB WEIGHT (LBS) 3000 2000 H Zk-{Bfimm EA "EM N \— Y 6 7) 7 7 9 8 8‘ or older MODEL YEAR Figure 7.6.3. Model Year unmflmz page an Ham» ammo: .e.m.5 musmflm 9 3 l .50; 5.m . mwm40 oz_mm_: u mmmcu .3505 .nm. . .555.mvvo~e m5.v.mvm anon..uma oooo.wmo5nv .nom. .a> .mmmn . «www.mm5nnn .nv5.55m 5mnw.n5mu oooo.m0m.mm5 .uom. .a«w>> Am05e . 50am.nwm5mn umm5.5mm cm5m.ov5n oooo.Ommnnmn. ..om. .a4m>> Avwvm . .vmm.on.mm~ 0505.5nm mmoo.mown oooo.O5o05mv. .Oam. .c«m>> .mo.m . canv.0mvwov oowv.5mw 5.5m.omm« 0000.”..o0no. .m5m. .u«m>> Ae5mm v n.nm.v00emv m..m.no5 wOOm.m..n oooo.omucnmo. .o5m. .u> Amv5m . mm5~.v5m0vv mmmo.vww m5no.ewvn oooo.momm.mm. .55m. .u> Amw5v v mvmm.nwvmwm w5mn.nm5 umv...nmn oooo.m0mwnow. .w5m. .u> Amoco 0 m.n..oum~nm cumu.mm5 5nnw.0omn oooo.u.oumnu. .m5m. .u> .mmwn . m..v.5vmm5w mw.m..uo omvw.uovn oooo..mnowcu. .v5m. .a> .uovn 0 .uno.vwmoow m5mn.Om5 m.n..n5mm oooo.mvmmm5u. .n5m. .a> Avnnm . .mwo.no.m0m mmo5.o.5 mmev.5.vn oooo.wm5«mo5. .«5m. .a«u>> A0v5.m 0 55e5.mom5om m5mw.m55 mmon.mm.n oooo.o5nmnmmw. zo.»<4:aon wu~»2u «an 2 muz‘_u<> >uo ohm 24m: 32m Jum<4 m34<> woou udmc~ag> .a«m>> >m zzoo Zuxoum z_> >m .zw> pzo.wa a wam«_m«> zo_au»~uu ----------------- mzo_»<4:aoam:m no zo.»a_uumuo ----------------- mmmnawmsz an ummw Hwooz .m.m.5 musmflm 0 na 1 .50¢ w.m no omen . mumco oz_mm~: wmocm . mwmqo 445a» .nm. . uvm5 mm .mww.5 mm5o.vo. oooo.wvow. .nmm. .a> ..mm~ 0 «man cm wmmn.5 muOv.uo. oooo.5oum0n .uom. .a> Am.5v 0 5m5m.mm m.mv.5 mewv.no. oooo.0um5oc ..am. .c> Awmvm 0 wv.u.5m 0vwm.5 ..mm.no. oooo.m5mvmm .owm. .u> Acmow . u5m5 w5 ..m5.m mnm~.mo_ oooo.voo5vu .m5m. .a> Anmmm v 5mm~.~o v.5o.m mo5o.oo. oooo.5nvvvm .m5m. .u> .5mom V nnmw.w5 .mm5.m n.n~.... oooo.mv.ovm .55m. .u«m>> Amm5v . wmmm.oo. vvmo.o. ammo.... oooo.n5mmnm .w5m. .u«w>> .vnvm . ammo.uo. one..o. .w55..._ oooo.mnonon .m5m. .a> Ammon . 50mo.vo. mmm..0. .mm..... oooo.mnoo.v .v5m. .c4m>> Amman . 5.n..nm 55...m 5mmn.v.. oooo.ovwomn .n5m. .a> Amnnm 0 mmn5.m5 .ow5.m mnmm.n.. 0000.50vmmm .u5m. .a‘w>> 309m 0 00no.vm vmmwd 03560. 00805300.. 20.55.30,“. waCZm no... 2 ~024~a<> >wo o5m 2.“: 22m 4mm¢4 ~24<> wooo m4m«_a<> .¢> >m zroo zuxoum z_> >m .zw> um‘mammxz can .m4:3> m4m<~u¢> zo_uw5_au - - - - - - - - - - - - - - - - - m z o _ 5 < A a a o a m a m a o z o _ p a . a o m u o - - - - - - - - - - - - - - - - - 141 m Emumumuumom .m.o.5 musmflm 00.000. 00..00. 00.000. 00.050. 00.050. 00.550. 00.050. 00.050. 00.050. 00.050. 00.050. . 90IIUOIIll‘l-ll9|ll|§lIIIQIIII§IIII§IIIIQ|OICOIIIIOIIIIQIIII’IIII‘Ol-00!IIIOIIIIOIOOI0IIII0IIIIOIIIl0. 8 .00 n ~~—Q~——-¢~———¢ N '5 0 H N 0. uuuonuu—Q-uuuo .~ 00 ~ 0) NO DU «00 ~Q~ ~~ O V OUOOOOQDDU «0005009. tun-n DQNOHGOHOD. 0 o- _ 000000000. aauooouoc5n 'ONF'F’ no « «nonaun 0055 ---------—9 O 017* o o N C! 0 no C no No 0 n O a O .- .06 E Q Ci 0 6 8.00. ~~ .— 00 0 I050 00 0 0 0 0 .0.. —O——n~— _- 75 Q0! 0 00 0000000. O 00 0000 000 000 000 O OGFOO O Q 19 000 DOB ONO'F 0 — n n 0 N00 00 000000000000 0 NOOI’N'ONOO ”0N0 G 00 NO CODE 00 ~- 0 «00000 0 000 000 0 000000 ' Quaint-tho u—umq—un— ' 0 H I) — u N o 0 ¢ 8.00. .o----¢---n+-uu-o----+u---¢c-n-¢n-u-+--usouucn‘c--:ouu--o---uou-uuou---4--c-¢---u4-u-uousau4u--u+-uuno. 0m..00. 00.000. om.OhG. 00.00.. 00.000. 00.000. 00.00.. Ofl.QhQ. 00.000. OO.NFO. .B(U>> AWWOCUC~ 2—’ >0 v1w> UW¢DJUUI’ CCU .0JZ’) ~ZIOOv no "flaflwbhduw 142 e total accidents - driver injury (VEH #1, VEH #2) - sex of driver - age of driver — single vehicle accidents - overturned vehicle accidents - accident with other vehicles - accidents with pedestrians - accidents with fixed objects - highway area type — highway surface condition The results by wheelbase are shown in Figures 7.6.7 through 7.6.17. However. using wheelbase as a measure of vehicle size does not show as consistent results as do the curb weight categories. Table 7.6.1 shows equations and coefficients of determination for the best fitting curves. Only overturned vehicle accidents and serious injury in VEH #2 produce larger coefficients of determination than the curb weight categories. The results by model-year are shown in Figures 7.6.18 through 7.6.23. It is apparent that newer model-year automobiles are less likely to be involved in an accident than earlier model-year automobiles (for the same exposure), regardless of the sex or age of the driver. While there seems to be a major effect of model-year on highway safety. the model-year is not the only significant factor in explaining relative accident involvement. As shown in the early part of this chapter, the size of automobile is also an important factor. For example, small 143 1.1 TOTAL ACCID NT \Eflil £05331; Ijk\\\x F 1 o /‘ < 1/ 0.9 ( Under 95 101 107 113 119 125 Over inches WHEELBAS F. Figure 7.6.7. Total Accidents 144 --4 WHEELBASE 1.5 na. Acc E:Inj.lkxfl 1.0 0.5 e i +—- L ”ususflfii-iwnn INJURQ IN vrnsi [8'157INJU ‘H 'P"+" " i I 3 Under 95 101 107 113 119 125 Over inches Figure 7.6.8. Driver Injury VEH #1 1455 0.5 —~~— - ~J-*----—1----*r--———---— INJURY IN 8:2 A'fls INJU 1 ”F04.” 0 Under ’ 95 101 107 113 119 125 Over WHEELBASE inches Figure 7.6.9. Driver Injury VEH #2 ' 146 1.1 VEHIl m 1.0 0.9 i —( Under 95 101 107 113 119 125 Over inches WHEELBASE Figure 7.6.10. Sex of Driver ' 147 L231 1.1 VEI-Ifl Eifiilj 1.0 0.9 ,1. “f , Under 95 101 107 113 119 125 Over inches WHEELBASE Figure 7.6.11. Age of Driver ' 2148 1.1 vanti Efiiitij 1.0 0.9 SINfLE v3+1cns ACCIDETT A. T Under 95 101 107 113 119 125 Over inches WHEELBASE Figure 7.6.12. Single Vehicle v 149 ovaamvawzo VEHICLE Accxnnfir VEH 1 _-.--_-+ .- -_fi E vsati \ \. Under 95 101 107 113 119 125 Over WHEELBASE inches Figure 7.6.13. Overturned Vehicle ' 150 1.5 m— i I L 3W1) ACCID NT WITH 0mm VEHICLES A/AN/ ‘ Under 95 101 107 113 119 125 Over inches WHEELBASE Figure 7.6.14. Other Vehicles 151 1.5 _yImI. swam] A45- . “;7:\\é ACCID BNT WITH PED .ACCIDfNT WITH FIX} STRIA£S ocmm ECTS Under 95 101 107 113 WHEELBASE 119 125 Over inches Figure 7.6.15. Pedestrians/Fixed Objects .152 1.1 VEHOl Efiiiilj AREA TYPE Iursnsscrxob / mnmmx A ~15. H MIDBIJOCK Under 95 101 107 113 119 125 Over inches WHEBLBASE Area Type Figure 7.6.16. 1523 1.1 \nfifll Efiifitlj 1.0 0.9 [Tl -A ID u ;_‘ T pa CE DRY WET ICY OR 8N0“ Under 95 101 107 113 119 125 Over inches WHEELBASE Figure 7.6.17. Surface Type 154 Table 7.6.1. Best Fitting Curves for Results by Wheelbase Categories Conditions Eguations g? Inj. VEH #1 Y = X/(-17O + 2.62 X) .868 Serious Inj. VEH #1 Y = -1.03 + 2.19 /X .852 Inj. VEH #2 Y = 1/(-282 + .0119 X) .955 Serious Inj. VEH #2 Y = -1.86 + 310 /X .969 Single Vehicle Y = X/(-70.4 + 1.66 X) .855 Overturned Y = 837 exp(-.0637 X) .968 Parked Vehicles. Y = 1/(3.91 - .0261 X) .866 Pedestrians Y = 1/(2.21 — .0111 X) .551 where Y = Ratio of the actual to the expected number of accidents X = Wheelbase R2 = Coefficient of determination 155 TOTAL AcCIDtNT L2 mm“. A ‘ E L1 R LO N as k. \5 m8 m7 I J 72 73 74 75 75 77 78 79 80 81 82 or older umm.nmn Figure 7.6.18. Total Accidents 156 g y Y . A‘" lINJURY ‘ ”FN*Ial D IVER INJUURY IN Vfitfll 1.5 In'. Acc. E 1173. Aoc 15‘ I, \ A 4. \ A 1.0 K\ 0.5 72 73 74 75 76 77 78 79 80 81 82 .or older MDELYEAR Figure 7.6.19. Driver Injury VEH #1 157 I L5 In'. Acc. E In). Acc 1.0 67 as DRIVER INJURY 1N vdaiz [\w-A INJURY h ' 'F'4”A' 72 73. 74 75 76 77 78 79 80 81 82 or older DUDE]. YEAR Figure 7.6.20. Driver Injury VEH #2 158 1.2 A \EMl \ E \ \ \\ ,’ \\ \ I \ I L1 2K 2% \ \ \ \ \ 1.0 0.9 SEX OF DRIVER 0.8 Z§"¢§wh£ ‘ffiflxflbmqa 0.7 ! I T I . 72 73 74 75 76 77 78 79 80 81 82 orolder uma;umk Figure 7.6.21. Sex of Driver 159 VEHfl 1.2 1.1 1.0 /: .,53 \ I‘I‘1“ 0.9 ” “ . \Ith \ 45 0.8 AGE or URIva MQs-422><».24 .25—34 0.7 - J~_“ p|f~~ 4'-54 KDr' ss— . JV 7 ' ‘ T ! =. ' - l l 1 1 72 73 74 7s 76 77 78 79 80 81 82 orolder MODEL YEAR Figure 7.6.22. Age of Driver 160 VEI'HI 1.2 1.1 ”K / 5 89 e. ‘ I SINGLfl VEHICLfi AC¢Iosfir ms 87 ' 3 2- ? : L3 L A L l 72 73 74 7s 76 77 78 79 80 81 82 orolder nommvmm Figure 7.6.23. Single Vehicle 161 automobiles are found to have a high A/E ratio in conditions of driver injury, single vehicle accident, and overturned vehicle accidents. It was also shown that there is a clear tendency that newer model-year automobiles are smaller than earlier model-years.. Therefore, the results obtained by using model-year categories cannot be explained only by model-year itself. Two effects are involved in the results, and they tend to offset the effects of each other in most rural conditions, and they tend to compound the effects in urban conditions. Examples of where the two factors may offset each other would include single vehicle accidents, overturned vehicle accidents, accidents at midblocks, and accidents on icy or snowy surface conditions (Figures 7.6.24 through 7.6.28). In addition to the size difference, newer automobiles usually have new tires, and thus might be able to maneuver better on icy or snowy surface conditions, even though small automobiles are found to be more hazardous during these conditions. An example of the compounding effect can be seen in results of accidents at intersections. One factor which explains the overrepresentation of large automobiles is the fact that the accidents involving earlier model-year automobiles “are more related to drinking drivers or drivers using drugs than accidents involving newer model-year automobiles. Figure 7.6.29 shows the percentage of drivers involved in accidents as VEH #1 who were drinking or using drugs. The results obtained above are statistically 162 VEHIl Iii-GE!” 1' H OLERTWRNEk VB ICL] 5 ACC :IDfiflT I 1 ‘72 73 or older 74 7 S 76 MHDIII, 77 78 YEAR 79 80 81 82 Figure 7.6.24. Overturned Vehicle 163 VEH l EWEHO 1i “MN c -1— \ . N‘ i 9 a i ' i j l I ACC Dani wan Crank VEHiCLE . as «L» T g i i L i . : Z 1 1 ‘ i 1 i 3 72 73 74 7s 6 77 78 79 80 81 82 or older MODEL YEAR Figure 7.6.25. Other Vehicles 164 VEH l E Eflfl LO \ \ \\ A1 \ \ ‘z 0.5 [fir-£3 ACCI NT WITH pa ST ansk ti? ACCI NT WITH FI D JEG s 72 73 74 7s 76 77 78 79 80 81 82 or older MODEL YEAR Figure 7.6.26. Pedestrian/Fixed Object 165 1‘ A5 \ [x \ ll \ \ I \ L2 1 , L .\ I, f‘ V5151 ‘ I \ \ \ 1.1 x \ \ 2:. ' \ 1.0 ' x A“ I i ‘zx, as ;7 . X i x \ \\ [4% 2x’ \ HIGHwai TYPE ‘ \ . m8 Aqm ik—d Ann-A INTARSECHIOW I Air—Ilnn>uxw m7 I ! J- I u T | t . 1 72 73 74 75 76 77 or 01w 78 79 80 81 32 Wm Figure 7.6.27. Area TYpe .166 1.2 10 E fifl/ \EQ . V I \ *1 1 , \ . 4‘ 08 j 45‘ .49 l X HILHWAT SURFACE cosDITion 0.8 Zl- --A DRY ‘k--1.LWET [}-~-{j ICY on snow 0.7 J, . T l i ; 72 73 74 7s 76 77 78 79 80 81 82 or older mm Figure 7.6.28. Surface Condition 167 19 DRINKING 0R U53 OF DRflGS 17 16 15 PERCENT (3) 14 L§\\z 13 12 A 11 H. 11. /A 10 a; 4b ' t 72 73 74 75 76 77 78 79 80 81 82 or older MODEL YEAR Figure 7.6.29. Drinking/Drugs 168 significant except for the results of single vehicle accidents. 8.0 CONCLUSIONS There is consistent evidence that small automobiles have a unique risk of accident involvement. Small automobiles were found to be involved in more accidents than would be expected for their exposure in each of the following cases: single vehicle accidents - overturned vehicle accidents - on icy or snowy highway surface - at midblocks — in rural areas On the other hand, large automobiles were found to be over represented in the following conditions: - accidents with other vehicles - at intersections - in urban areas In addition to these general findings, results for geometric features were obtained by examining the accident data for rural and urban areas separately. The results show remarkable consistency. In rural areas, small automobiles were found to have a high A/E ratio in the following geometric features: - midblocks - 2 lane-2 way highways - no passing zones In urban areas, large automobiles exhibited a high A/E 169 170 ratio at intersections. Drivers of small automobiles were found to have a greater risk of being injured for the same exposure. Drivers of small automobiles are exposed to the risk of being injured regardless of whether they are in VEH #1 (identified as being responsible for the accident) or in VEH #2 (the second automobile). Drivers of automobiles weighing less than 2,000 lbs. have a 72 percent higher risk of being injured in VEH #1, and have a 56 percent higher risk of being injured in VEH #2 than the drivers of automobiles weighing 4,500 lbs. or more for the same exposure. The curb weight, wheelbase, and model-year were all found to be important measures of accidents. The curb weight and wheelbase categories show similar results, but the curb weight categories provide more consistent results than do the wheelbase categories. It was found that newer model-year automobiles have lower A/E ratios than earlier model—years. The new exposure approach used in the present study is found to be a useful tool for quantification of exposure. The validation of two assumptions made for the new exposure approach are consistently supported by the data. The results obtained in the study show remarkable consistency and suggest validation of the assumptions. BIBLIOGRAPHY BIBLIOGRAPHY RFP/Contract Continuation Sheet, "Small Car Safety Study," Statement of Work Section B 4.1, DTNH22-83-R-O7287, National Highway Traffic Safety Administration, U.S. Department of Transportation, 1983. "Study of Compact Vehicles Registered in New York State 1962," Division of Research and Development, New York State Department of Motor Vehicles, 1963. David Solomon, "Accidents on Main Rural Highways Related to Speed, Driver, and Vehicle," U.S. Department of Commerce, Bureau of Public Roads, Office of Research and Development, 1964. Jaakko K. Kihlberg, Eugene A. Narragon and B. J. Campbell, "Automobile Crash Injury in Relation to Car Size," Cornell Aeronautical Laboratory, Inc. of Cornell University, CAL Report No. VJ-1823-R11, November 1964. John W. Garrett and Arthur Stern, "A Study of Volkswagen Accidents in the United States," Cornell Aeronautical Laboratory, Inc. of Cornell University, CAL Report No. VJ-1823-R32, November 1968. B. J. Capmbell, "Driver Injury in Automobile Accidents Involving Certain Car Models," University of North Carolina, Highway Safety Research Center, Journal of Safety Research, Vol. 2, No. 4, pp. 207-228, December 1970. James O'Day, Daniel H. Golomb and Peter Cooley, "A Statistical Description of Large and Small Car Involvement in Accidents," Highway Research Institute, University of Michigan, HIT Lab Reports, Vol. 3, No. 9. May 1973. Theodore B. Anderson, "Passenger Compartment Intrusion in Automobile Accidents," Calspan Corporation, Buffalo, N.Y., Report No. ZQ-5276—V—3R, October 1974. Donald W. Reinfurt, B. J. Campbell, "Mileage Crash Rates for Certain Car Make and Model Year Combination: A Preliminary Study," University of North Carolina, Highway Safety Research Center, December 1974. 171 10. 11. 12. 13. 14. 15. 16. 17. 18. 172 James O'Day and Richard Kaplan, "How Much Safer Are You in a Large Car?", Highway Safety Research Institute, University of Michigan, HIT Lab Reports, Vol. 5, No. 9, May 1975. P. L. Yu, C. Wrather and G. Kozmetsky, "Auto Weight and Public Safety, A Statistical Study of Transportation Hazards," Texas University at Austin, June 1975. Leon S. Robertson and Susan P. Baker, "Motor Vehicle Sizes in 1440 Fatal Crashes," Washington, D.C., Insurance Institute for Highway Safety, July 1975. Amitabh K. Dutt, Donald W. Reinfurt and Jane C. Stutts, "Accident Involvement and Crash Injury Rates: An Investigation by Make, Model and Year of Car," University of North Carolina. Highway Safety Research Center, Accident Analysis and Prevention, Vol. 9, No. 4, December 1977. G. Grime and T. P. Hutchinson, "Some Implications of Vehicle-Weight for the Risk of Injury to Drivers," London, Great Britain, University College London, Transport Studies Group, Research Report ISSN 0142-6052, February 1979. J. Richard Stewart and Carbl Lederhaus Carroll, "Annual Mileage Comparisons and Accident and Injury Rates by Make, Model," University of North Carolina, Highway Safety Research Center, October 1980. Leonard Evans, "Car Mass and Likelihood of Occupant Fatality," Research Laboratories, General Motors _ Corp., SAE Technical Paper Series 820807, June 1982. Leonard Evans, "Driver Fatalities versus Car Mass Using a New Exposure Approach," Transportation Research Department, General Motors Research Laboratories, GMR-4241, January 1983. J. Richard Stewart and Carol Lederhaus Carroll, "Annual Mileage Comparisons and Accident and Injury Rates by Make, Model," University of North Carolina, Highway Safety Research Center, October 1980.